From: Pierre-Marc Fournier Date: Mon, 30 Mar 2009 22:27:41 +0000 (-0400) Subject: merge libmarkers, libtracing and libtracectl in a single libust library X-Git-Tag: v0.1~242 X-Git-Url: https://git.lttng.org./?a=commitdiff_plain;h=b4512257eb71d0432554047acf6278dc42a15a75;p=ust.git merge libmarkers, libtracing and libtracectl in a single libust library --- diff --git a/Makefile.am b/Makefile.am index f8620df..996aedd 100644 --- a/Makefile.am +++ b/Makefile.am @@ -1 +1 @@ -SUBDIRS = libmarkers libtracectl libtracing hello libmallocwrap +SUBDIRS = libust hello libmallocwrap diff --git a/bootstrap b/bootstrap index 1199ae2..0374e38 100755 --- a/bootstrap +++ b/bootstrap @@ -1,7 +1,7 @@ #! /bin/sh set -x -aclocal -I config +aclocal libtoolize --force --copy autoheader automake --add-missing --copy diff --git a/configure.ac b/configure.ac index 454faac..38d6b8b 100644 --- a/configure.ac +++ b/configure.ac @@ -6,7 +6,7 @@ AC_INIT([ust], [0.0], [pierre-marc dot fournier at polymtl dot ca]) AC_CONFIG_AUX_DIR(make_scripts) AM_INIT_AUTOMAKE([0.0 foreign]) AC_CONFIG_SRCDIR([ust/localerr.h]) -#AC_CONFIG_HEADERS([config.h]) +AC_CONFIG_HEADERS([config.h]) # Checks for programs. AC_PROG_CC @@ -63,5 +63,5 @@ AC_SUBST(KCOMPAT_PATH) # ust/Makefile # ustd/Makefile]) -AC_CONFIG_FILES([Makefile libmarkers/Makefile libtracectl/Makefile libtracing/Makefile hello/Makefile libmallocwrap/Makefile]) +AC_CONFIG_FILES([Makefile libust/Makefile hello/Makefile libmallocwrap/Makefile]) AC_OUTPUT diff --git a/hello/Makefile.am b/hello/Makefile.am index 81b4604..2a4e6b7 100644 --- a/hello/Makefile.am +++ b/hello/Makefile.am @@ -1,11 +1,8 @@ bin_PROGRAMS = hello -hello_SOURCES = hello.c tp.c tp.h $(top_builddir)/share/kref.c -hello_LDADD = $(top_builddir)/libmarkers/libmarkers.la $(top_builddir)/libtracectl/libtracectl.la $(top_builddir)/libtracing/libtracing.la @URCU_PATH@/liburcu.so -##hello_LDFLAGS = -L$(top_builddir)/libmarkers +hello_SOURCES = hello.c tp.c tp.h +hello_LDADD = $(top_builddir)/libust/libust.la @URCU_PATH@/liburcu.so INCLUDES = -I$(top_builddir)/share INCLUDES += -I@URCU_PATH@ INCLUDES += -I@KCOMPAT_PATH@ -INCLUDES += -I$(top_builddir)/libtracing -INCLUDES += -I$(top_builddir)/libmarkers -INCLUDES += -I$(top_builddir)/libtracectl +INCLUDES += -I$(top_builddir)/libust diff --git a/libmallocwrap/Makefile.am b/libmallocwrap/Makefile.am new file mode 100644 index 0000000..c9192b0 --- /dev/null +++ b/libmallocwrap/Makefile.am @@ -0,0 +1,8 @@ +lib_LTLIBRARIES = libmallocwrap.la +libmallocwrap_la_SOURCES = mallocwrap.c +noinst_SCRIPTS = run + +INCLUDES = -I$(top_builddir)/share +INCLUDES += -I@URCU_PATH@ +INCLUDES += -I@KCOMPAT_PATH@ +INCLUDES += -I$(top_builddir)/libust diff --git a/libmarkers/Makefile.am b/libmarkers/Makefile.am deleted file mode 100644 index 88643f3..0000000 --- a/libmarkers/Makefile.am +++ /dev/null @@ -1,8 +0,0 @@ -lib_LTLIBRARIES = libmarkers.la -libmarkers_la_SOURCES = marker.c tracepoint.c immediate.h jhash.h rcupdate.h marker.h tracepoint.h -libmarkers_la_LIBADD = @URCU_PATH@/liburcu.so - -INCLUDES = -I$(top_builddir)/share -INCLUDES += -I@URCU_PATH@ -INCLUDES += -I@KCOMPAT_PATH@ -INCLUDES += -I$(top_builddir)/libtracing diff --git a/libmarkers/immediate.h b/libmarkers/immediate.h deleted file mode 100644 index ee94ecf..0000000 --- a/libmarkers/immediate.h +++ /dev/null @@ -1,86 +0,0 @@ -#ifndef _LINUX_IMMEDIATE_H -#define _LINUX_IMMEDIATE_H - -/* - * Immediate values, can be updated at runtime and save cache lines. - * - * (C) Copyright 2007 Mathieu Desnoyers - * - * This file is released under the GPLv2. - * See the file COPYING for more details. - */ - -#ifdef USE_IMMEDIATE - -#include - -/** - * imv_set - set immediate variable (with locking) - * @name: immediate value name - * @i: required value - * - * Sets the value of @name, taking the module_mutex if required by - * the architecture. - */ -#define imv_set(name, i) \ - do { \ - name##__imv = (i); \ - core_imv_update(); \ - module_imv_update(); \ - } while (0) - -/* - * Internal update functions. - */ -extern void core_imv_update(void); -extern void imv_update_range(const struct __imv *begin, - const struct __imv *end); -extern void imv_unref_core_init(void); -extern void imv_unref(struct __imv *begin, struct __imv *end, void *start, - unsigned long size); - -#else - -/* - * Generic immediate values: a simple, standard, memory load. - */ - -/** - * imv_read - read immediate variable - * @name: immediate value name - * - * Reads the value of @name. - */ -#define imv_read(name) _imv_read(name) - -/** - * imv_set - set immediate variable (with locking) - * @name: immediate value name - * @i: required value - * - * Sets the value of @name, taking the module_mutex if required by - * the architecture. - */ -#define imv_set(name, i) (name##__imv = (i)) - -static inline void core_imv_update(void) { } -static inline void imv_unref_core_init(void) { } - -#endif - -#define DECLARE_IMV(type, name) extern __typeof__(type) name##__imv -#define DEFINE_IMV(type, name) __typeof__(type) name##__imv - -#define EXPORT_IMV_SYMBOL(name) EXPORT_SYMBOL(name##__imv) -#define EXPORT_IMV_SYMBOL_GPL(name) EXPORT_SYMBOL_GPL(name##__imv) - -/** - * _imv_read - Read immediate value with standard memory load. - * @name: immediate value name - * - * Force a data read of the immediate value instead of the immediate value - * based mechanism. Useful for __init and __exit section data read. - */ -#define _imv_read(name) (name##__imv) - -#endif diff --git a/libmarkers/marker.c b/libmarkers/marker.c deleted file mode 100644 index ab0fc77..0000000 --- a/libmarkers/marker.c +++ /dev/null @@ -1,1532 +0,0 @@ -/* - * Copyright (C) 2007 Mathieu Desnoyers - * - * This program is free software; you can redistribute it and/or modify - * it under the terms of the GNU General Public License as published by - * the Free Software Foundation; either version 2 of the License, or - * (at your option) any later version. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * GNU General Public License for more details. - * - * You should have received a copy of the GNU General Public License - * along with this program; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. - */ -//ust// #include -//ust// #include -//ust// #include -#include "jhash.h" -#include "list.h" -#include "rcupdate.h" -//ust// #include -#include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include - -#include "marker.h" -#include "kernelcompat.h" -#include "usterr.h" -#include "channels.h" -#include "tracercore.h" -#include "tracer.h" - -extern struct marker __start___markers[] __attribute__((visibility("hidden"))); -extern struct marker __stop___markers[] __attribute__((visibility("hidden"))); - -/* Set to 1 to enable marker debug output */ -static const int marker_debug; - -/* - * markers_mutex nests inside module_mutex. Markers mutex protects the builtin - * and module markers and the hash table. - */ -static DEFINE_MUTEX(markers_mutex); - -void lock_markers(void) -{ - mutex_lock(&markers_mutex); -} - -void unlock_markers(void) -{ - mutex_unlock(&markers_mutex); -} - -/* - * Marker hash table, containing the active markers. - * Protected by module_mutex. - */ -#define MARKER_HASH_BITS 6 -#define MARKER_TABLE_SIZE (1 << MARKER_HASH_BITS) -static struct hlist_head marker_table[MARKER_TABLE_SIZE]; - -/* - * Note about RCU : - * It is used to make sure every handler has finished using its private data - * between two consecutive operation (add or remove) on a given marker. It is - * also used to delay the free of multiple probes array until a quiescent state - * is reached. - * marker entries modifications are protected by the markers_mutex. - */ -struct marker_entry { - struct hlist_node hlist; - char *format; - char *name; - /* Probe wrapper */ - void (*call)(const struct marker *mdata, void *call_private, ...); - struct marker_probe_closure single; - struct marker_probe_closure *multi; - int refcount; /* Number of times armed. 0 if disarmed. */ - struct rcu_head rcu; - void *oldptr; - int rcu_pending; - u16 channel_id; - u16 event_id; - unsigned char ptype:1; - unsigned char format_allocated:1; - char channel[0]; /* Contains channel'\0'name'\0'format'\0' */ -}; - -#ifdef CONFIG_MARKERS_USERSPACE -static void marker_update_processes(void); -#else -static void marker_update_processes(void) -{ -} -#endif - -/** - * __mark_empty_function - Empty probe callback - * @mdata: marker data - * @probe_private: probe private data - * @call_private: call site private data - * @fmt: format string - * @...: variable argument list - * - * Empty callback provided as a probe to the markers. By providing this to a - * disabled marker, we make sure the execution flow is always valid even - * though the function pointer change and the marker enabling are two distinct - * operations that modifies the execution flow of preemptible code. - */ -notrace void __mark_empty_function(const struct marker *mdata, - void *probe_private, void *call_private, const char *fmt, va_list *args) -{ -} -//ust// EXPORT_SYMBOL_GPL(__mark_empty_function); - -/* - * marker_probe_cb Callback that prepares the variable argument list for probes. - * @mdata: pointer of type struct marker - * @call_private: caller site private data - * @...: Variable argument list. - * - * Since we do not use "typical" pointer based RCU in the 1 argument case, we - * need to put a full smp_rmb() in this branch. This is why we do not use - * rcu_dereference() for the pointer read. - */ -notrace void marker_probe_cb(const struct marker *mdata, - void *call_private, ...) -{ - va_list args; - char ptype; - - /* - * rcu_read_lock_sched does two things : disabling preemption to make - * sure the teardown of the callbacks can be done correctly when they - * are in modules and they insure RCU read coherency. - */ -//ust// rcu_read_lock_sched_notrace(); - ptype = mdata->ptype; - if (likely(!ptype)) { - marker_probe_func *func; - /* Must read the ptype before ptr. They are not data dependant, - * so we put an explicit smp_rmb() here. */ - smp_rmb(); - func = mdata->single.func; - /* Must read the ptr before private data. They are not data - * dependant, so we put an explicit smp_rmb() here. */ - smp_rmb(); - va_start(args, call_private); - func(mdata, mdata->single.probe_private, call_private, - mdata->format, &args); - va_end(args); - } else { - struct marker_probe_closure *multi; - int i; - /* - * Read mdata->ptype before mdata->multi. - */ - smp_rmb(); - multi = mdata->multi; - /* - * multi points to an array, therefore accessing the array - * depends on reading multi. However, even in this case, - * we must insure that the pointer is read _before_ the array - * data. Same as rcu_dereference, but we need a full smp_rmb() - * in the fast path, so put the explicit barrier here. - */ - smp_read_barrier_depends(); - for (i = 0; multi[i].func; i++) { - va_start(args, call_private); - multi[i].func(mdata, multi[i].probe_private, - call_private, mdata->format, &args); - va_end(args); - } - } -//ust// rcu_read_unlock_sched_notrace(); -} -//ust// EXPORT_SYMBOL_GPL(marker_probe_cb); - -/* - * marker_probe_cb Callback that does not prepare the variable argument list. - * @mdata: pointer of type struct marker - * @call_private: caller site private data - * @...: Variable argument list. - * - * Should be connected to markers "MARK_NOARGS". - */ -static notrace void marker_probe_cb_noarg(const struct marker *mdata, - void *call_private, ...) -{ - va_list args; /* not initialized */ - char ptype; - -//ust// rcu_read_lock_sched_notrace(); - ptype = mdata->ptype; - if (likely(!ptype)) { - marker_probe_func *func; - /* Must read the ptype before ptr. They are not data dependant, - * so we put an explicit smp_rmb() here. */ - smp_rmb(); - func = mdata->single.func; - /* Must read the ptr before private data. They are not data - * dependant, so we put an explicit smp_rmb() here. */ - smp_rmb(); - func(mdata, mdata->single.probe_private, call_private, - mdata->format, &args); - } else { - struct marker_probe_closure *multi; - int i; - /* - * Read mdata->ptype before mdata->multi. - */ - smp_rmb(); - multi = mdata->multi; - /* - * multi points to an array, therefore accessing the array - * depends on reading multi. However, even in this case, - * we must insure that the pointer is read _before_ the array - * data. Same as rcu_dereference, but we need a full smp_rmb() - * in the fast path, so put the explicit barrier here. - */ - smp_read_barrier_depends(); - for (i = 0; multi[i].func; i++) - multi[i].func(mdata, multi[i].probe_private, - call_private, mdata->format, &args); - } -//ust// rcu_read_unlock_sched_notrace(); -} - -static void free_old_closure(struct rcu_head *head) -{ - struct marker_entry *entry = container_of(head, - struct marker_entry, rcu); - kfree(entry->oldptr); - /* Make sure we free the data before setting the pending flag to 0 */ - smp_wmb(); - entry->rcu_pending = 0; -} - -static void debug_print_probes(struct marker_entry *entry) -{ - int i; - - if (!marker_debug) - return; - - if (!entry->ptype) { - printk(KERN_DEBUG "Single probe : %p %p\n", - entry->single.func, - entry->single.probe_private); - } else { - for (i = 0; entry->multi[i].func; i++) - printk(KERN_DEBUG "Multi probe %d : %p %p\n", i, - entry->multi[i].func, - entry->multi[i].probe_private); - } -} - -static struct marker_probe_closure * -marker_entry_add_probe(struct marker_entry *entry, - marker_probe_func *probe, void *probe_private) -{ - int nr_probes = 0; - struct marker_probe_closure *old, *new; - - WARN_ON(!probe); - - debug_print_probes(entry); - old = entry->multi; - if (!entry->ptype) { - if (entry->single.func == probe && - entry->single.probe_private == probe_private) - return ERR_PTR(-EBUSY); - if (entry->single.func == __mark_empty_function) { - /* 0 -> 1 probes */ - entry->single.func = probe; - entry->single.probe_private = probe_private; - entry->refcount = 1; - entry->ptype = 0; - debug_print_probes(entry); - return NULL; - } else { - /* 1 -> 2 probes */ - nr_probes = 1; - old = NULL; - } - } else { - /* (N -> N+1), (N != 0, 1) probes */ - for (nr_probes = 0; old[nr_probes].func; nr_probes++) - if (old[nr_probes].func == probe - && old[nr_probes].probe_private - == probe_private) - return ERR_PTR(-EBUSY); - } - /* + 2 : one for new probe, one for NULL func */ - new = kzalloc((nr_probes + 2) * sizeof(struct marker_probe_closure), - GFP_KERNEL); - if (new == NULL) - return ERR_PTR(-ENOMEM); - if (!old) - new[0] = entry->single; - else - memcpy(new, old, - nr_probes * sizeof(struct marker_probe_closure)); - new[nr_probes].func = probe; - new[nr_probes].probe_private = probe_private; - entry->refcount = nr_probes + 1; - entry->multi = new; - entry->ptype = 1; - debug_print_probes(entry); - return old; -} - -static struct marker_probe_closure * -marker_entry_remove_probe(struct marker_entry *entry, - marker_probe_func *probe, void *probe_private) -{ - int nr_probes = 0, nr_del = 0, i; - struct marker_probe_closure *old, *new; - - old = entry->multi; - - debug_print_probes(entry); - if (!entry->ptype) { - /* 0 -> N is an error */ - WARN_ON(entry->single.func == __mark_empty_function); - /* 1 -> 0 probes */ - WARN_ON(probe && entry->single.func != probe); - WARN_ON(entry->single.probe_private != probe_private); - entry->single.func = __mark_empty_function; - entry->refcount = 0; - entry->ptype = 0; - debug_print_probes(entry); - return NULL; - } else { - /* (N -> M), (N > 1, M >= 0) probes */ - for (nr_probes = 0; old[nr_probes].func; nr_probes++) { - if ((!probe || old[nr_probes].func == probe) - && old[nr_probes].probe_private - == probe_private) - nr_del++; - } - } - - if (nr_probes - nr_del == 0) { - /* N -> 0, (N > 1) */ - entry->single.func = __mark_empty_function; - entry->refcount = 0; - entry->ptype = 0; - } else if (nr_probes - nr_del == 1) { - /* N -> 1, (N > 1) */ - for (i = 0; old[i].func; i++) - if ((probe && old[i].func != probe) || - old[i].probe_private != probe_private) - entry->single = old[i]; - entry->refcount = 1; - entry->ptype = 0; - } else { - int j = 0; - /* N -> M, (N > 1, M > 1) */ - /* + 1 for NULL */ - new = kzalloc((nr_probes - nr_del + 1) - * sizeof(struct marker_probe_closure), GFP_KERNEL); - if (new == NULL) - return ERR_PTR(-ENOMEM); - for (i = 0; old[i].func; i++) - if ((probe && old[i].func != probe) || - old[i].probe_private != probe_private) - new[j++] = old[i]; - entry->refcount = nr_probes - nr_del; - entry->ptype = 1; - entry->multi = new; - } - debug_print_probes(entry); - return old; -} - -/* - * Get marker if the marker is present in the marker hash table. - * Must be called with markers_mutex held. - * Returns NULL if not present. - */ -static struct marker_entry *get_marker(const char *channel, const char *name) -{ - struct hlist_head *head; - struct hlist_node *node; - struct marker_entry *e; - size_t channel_len = strlen(channel) + 1; - size_t name_len = strlen(name) + 1; - u32 hash; - - hash = jhash(channel, channel_len-1, 0) ^ jhash(name, name_len-1, 0); - head = &marker_table[hash & ((1 << MARKER_HASH_BITS)-1)]; - hlist_for_each_entry(e, node, head, hlist) { - if (!strcmp(channel, e->channel) && !strcmp(name, e->name)) - return e; - } - return NULL; -} - -/* - * Add the marker to the marker hash table. Must be called with markers_mutex - * held. - */ -static struct marker_entry *add_marker(const char *channel, const char *name, - const char *format) -{ - struct hlist_head *head; - struct hlist_node *node; - struct marker_entry *e; - size_t channel_len = strlen(channel) + 1; - size_t name_len = strlen(name) + 1; - size_t format_len = 0; - u32 hash; - - hash = jhash(channel, channel_len-1, 0) ^ jhash(name, name_len-1, 0); - if (format) - format_len = strlen(format) + 1; - head = &marker_table[hash & ((1 << MARKER_HASH_BITS)-1)]; - hlist_for_each_entry(e, node, head, hlist) { - if (!strcmp(channel, e->channel) && !strcmp(name, e->name)) { - printk(KERN_NOTICE - "Marker %s.%s busy\n", channel, name); - return ERR_PTR(-EBUSY); /* Already there */ - } - } - /* - * Using kmalloc here to allocate a variable length element. Could - * cause some memory fragmentation if overused. - */ - e = kmalloc(sizeof(struct marker_entry) - + channel_len + name_len + format_len, - GFP_KERNEL); - if (!e) - return ERR_PTR(-ENOMEM); - memcpy(e->channel, channel, channel_len); - e->name = &e->channel[channel_len]; - memcpy(e->name, name, name_len); - if (format) { - e->format = &e->name[channel_len + name_len]; - memcpy(e->format, format, format_len); - if (strcmp(e->format, MARK_NOARGS) == 0) - e->call = marker_probe_cb_noarg; - else - e->call = marker_probe_cb; - trace_mark(metadata, core_marker_format, - "channel %s name %s format %s", - e->channel, e->name, e->format); - } else { - e->format = NULL; - e->call = marker_probe_cb; - } - e->single.func = __mark_empty_function; - e->single.probe_private = NULL; - e->multi = NULL; - e->ptype = 0; - e->format_allocated = 0; - e->refcount = 0; - e->rcu_pending = 0; - hlist_add_head(&e->hlist, head); - return e; -} - -/* - * Remove the marker from the marker hash table. Must be called with mutex_lock - * held. - */ -static int remove_marker(const char *channel, const char *name) -{ - struct hlist_head *head; - struct hlist_node *node; - struct marker_entry *e; - int found = 0; - size_t channel_len = strlen(channel) + 1; - size_t name_len = strlen(name) + 1; - u32 hash; - int ret; - - hash = jhash(channel, channel_len-1, 0) ^ jhash(name, name_len-1, 0); - head = &marker_table[hash & ((1 << MARKER_HASH_BITS)-1)]; - hlist_for_each_entry(e, node, head, hlist) { - if (!strcmp(channel, e->channel) && !strcmp(name, e->name)) { - found = 1; - break; - } - } - if (!found) - return -ENOENT; - if (e->single.func != __mark_empty_function) - return -EBUSY; - hlist_del(&e->hlist); - if (e->format_allocated) - kfree(e->format); - ret = ltt_channels_unregister(e->channel); - WARN_ON(ret); - /* Make sure the call_rcu has been executed */ - if (e->rcu_pending) - rcu_barrier_sched(); - kfree(e); - return 0; -} - -/* - * Set the mark_entry format to the format found in the element. - */ -static int marker_set_format(struct marker_entry *entry, const char *format) -{ - entry->format = kstrdup(format, GFP_KERNEL); - if (!entry->format) - return -ENOMEM; - entry->format_allocated = 1; - - trace_mark(metadata, core_marker_format, - "channel %s name %s format %s", - entry->channel, entry->name, entry->format); - return 0; -} - -/* - * Sets the probe callback corresponding to one marker. - */ -static int set_marker(struct marker_entry *entry, struct marker *elem, - int active) -{ - int ret = 0; - WARN_ON(strcmp(entry->name, elem->name) != 0); - - if (entry->format) { - if (strcmp(entry->format, elem->format) != 0) { - printk(KERN_NOTICE - "Format mismatch for probe %s " - "(%s), marker (%s)\n", - entry->name, - entry->format, - elem->format); - return -EPERM; - } - } else { - ret = marker_set_format(entry, elem->format); - if (ret) - return ret; - } - - /* - * probe_cb setup (statically known) is done here. It is - * asynchronous with the rest of execution, therefore we only - * pass from a "safe" callback (with argument) to an "unsafe" - * callback (does not set arguments). - */ - elem->call = entry->call; - elem->channel_id = entry->channel_id; - elem->event_id = entry->event_id; - /* - * Sanity check : - * We only update the single probe private data when the ptr is - * set to a _non_ single probe! (0 -> 1 and N -> 1, N != 1) - */ - WARN_ON(elem->single.func != __mark_empty_function - && elem->single.probe_private != entry->single.probe_private - && !elem->ptype); - elem->single.probe_private = entry->single.probe_private; - /* - * Make sure the private data is valid when we update the - * single probe ptr. - */ - smp_wmb(); - elem->single.func = entry->single.func; - /* - * We also make sure that the new probe callbacks array is consistent - * before setting a pointer to it. - */ - rcu_assign_pointer(elem->multi, entry->multi); - /* - * Update the function or multi probe array pointer before setting the - * ptype. - */ - smp_wmb(); - elem->ptype = entry->ptype; - -//ust// if (elem->tp_name && (active ^ _imv_read(elem->state))) { -//ust// WARN_ON(!elem->tp_cb); -//ust// /* -//ust// * It is ok to directly call the probe registration because type -//ust// * checking has been done in the __trace_mark_tp() macro. -//ust// */ -//ust// -//ust// if (active) { -//ust// /* -//ust// * try_module_get should always succeed because we hold -//ust// * markers_mutex to get the tp_cb address. -//ust// */ -//ust// ret = try_module_get(__module_text_address( -//ust// (unsigned long)elem->tp_cb)); -//ust// BUG_ON(!ret); -//ust// ret = tracepoint_probe_register_noupdate( -//ust// elem->tp_name, -//ust// elem->tp_cb); -//ust// } else { -//ust// ret = tracepoint_probe_unregister_noupdate( -//ust// elem->tp_name, -//ust// elem->tp_cb); -//ust// /* -//ust// * tracepoint_probe_update_all() must be called -//ust// * before the module containing tp_cb is unloaded. -//ust// */ -//ust// module_put(__module_text_address( -//ust// (unsigned long)elem->tp_cb)); -//ust// } -//ust// } - elem->state__imv = active; - - return ret; -} - -/* - * Disable a marker and its probe callback. - * Note: only waiting an RCU period after setting elem->call to the empty - * function insures that the original callback is not used anymore. This insured - * by rcu_read_lock_sched around the call site. - */ -static void disable_marker(struct marker *elem) -{ - int ret; - - /* leave "call" as is. It is known statically. */ -//ust// if (elem->tp_name && _imv_read(elem->state)) { -//ust// WARN_ON(!elem->tp_cb); -//ust// /* -//ust// * It is ok to directly call the probe registration because type -//ust// * checking has been done in the __trace_mark_tp() macro. -//ust// */ -//ust// ret = tracepoint_probe_unregister_noupdate(elem->tp_name, -//ust// elem->tp_cb); -//ust// WARN_ON(ret); -//ust// /* -//ust// * tracepoint_probe_update_all() must be called -//ust// * before the module containing tp_cb is unloaded. -//ust// */ -//ust// module_put(__module_text_address((unsigned long)elem->tp_cb)); -//ust// } - elem->state__imv = 0; - elem->single.func = __mark_empty_function; - /* Update the function before setting the ptype */ - smp_wmb(); - elem->ptype = 0; /* single probe */ - /* - * Leave the private data and channel_id/event_id there, because removal - * is racy and should be done only after an RCU period. These are never - * used until the next initialization anyway. - */ -} - -/** - * marker_update_probe_range - Update a probe range - * @begin: beginning of the range - * @end: end of the range - * - * Updates the probe callback corresponding to a range of markers. - */ -void marker_update_probe_range(struct marker *begin, - struct marker *end) -{ - struct marker *iter; - struct marker_entry *mark_entry; - - mutex_lock(&markers_mutex); - for (iter = begin; iter < end; iter++) { - mark_entry = get_marker(iter->channel, iter->name); - if (mark_entry) { - set_marker(mark_entry, iter, !!mark_entry->refcount); - /* - * ignore error, continue - */ - - /* This is added for UST. We emit a core_marker_id event - * for markers that are already registered to a probe - * upon library load. Otherwise, no core_marker_id will - * be generated for these markers. Is this the right thing - * to do? - */ - trace_mark(metadata, core_marker_id, - "channel %s name %s event_id %hu " - "int #1u%zu long #1u%zu pointer #1u%zu " - "size_t #1u%zu alignment #1u%u", - iter->channel, iter->name, mark_entry->event_id, - sizeof(int), sizeof(long), sizeof(void *), - sizeof(size_t), ltt_get_alignment()); - } else { - disable_marker(iter); - } - } - mutex_unlock(&markers_mutex); -} - -/* - * Update probes, removing the faulty probes. - * - * Internal callback only changed before the first probe is connected to it. - * Single probe private data can only be changed on 0 -> 1 and 2 -> 1 - * transitions. All other transitions will leave the old private data valid. - * This makes the non-atomicity of the callback/private data updates valid. - * - * "special case" updates : - * 0 -> 1 callback - * 1 -> 0 callback - * 1 -> 2 callbacks - * 2 -> 1 callbacks - * Other updates all behave the same, just like the 2 -> 3 or 3 -> 2 updates. - * Site effect : marker_set_format may delete the marker entry (creating a - * replacement). - */ -static void marker_update_probes(void) -{ - /* Core kernel markers */ -//ust// marker_update_probe_range(__start___markers, __stop___markers); - /* Markers in modules. */ -//ust// module_update_markers(); - lib_update_markers(); -//ust// tracepoint_probe_update_all(); - /* Update immediate values */ - core_imv_update(); -//ust// module_imv_update(); /* FIXME: need to port for libs? */ - marker_update_processes(); -} - -/** - * marker_probe_register - Connect a probe to a marker - * @channel: marker channel - * @name: marker name - * @format: format string - * @probe: probe handler - * @probe_private: probe private data - * - * private data must be a valid allocated memory address, or NULL. - * Returns 0 if ok, error value on error. - * The probe address must at least be aligned on the architecture pointer size. - */ -int marker_probe_register(const char *channel, const char *name, - const char *format, marker_probe_func *probe, - void *probe_private) -{ - struct marker_entry *entry; - int ret = 0, ret_err; - struct marker_probe_closure *old; - int first_probe = 0; - - mutex_lock(&markers_mutex); - entry = get_marker(channel, name); - if (!entry) { - first_probe = 1; - entry = add_marker(channel, name, format); - if (IS_ERR(entry)) - ret = PTR_ERR(entry); - if (ret) - goto end; - ret = ltt_channels_register(channel); - if (ret) - goto error_remove_marker; - ret = ltt_channels_get_index_from_name(channel); - if (ret < 0) - goto error_unregister_channel; - entry->channel_id = ret; - ret = ltt_channels_get_event_id(channel, name); - if (ret < 0) - goto error_unregister_channel; - entry->event_id = ret; - ret = 0; - trace_mark(metadata, core_marker_id, - "channel %s name %s event_id %hu " - "int #1u%zu long #1u%zu pointer #1u%zu " - "size_t #1u%zu alignment #1u%u", - channel, name, entry->event_id, - sizeof(int), sizeof(long), sizeof(void *), - sizeof(size_t), ltt_get_alignment()); - } else if (format) { - if (!entry->format) - ret = marker_set_format(entry, format); - else if (strcmp(entry->format, format)) - ret = -EPERM; - if (ret) - goto end; - } - - /* - * If we detect that a call_rcu is pending for this marker, - * make sure it's executed now. - */ - if (entry->rcu_pending) - rcu_barrier_sched(); - old = marker_entry_add_probe(entry, probe, probe_private); - if (IS_ERR(old)) { - ret = PTR_ERR(old); - if (first_probe) - goto error_unregister_channel; - else - goto end; - } - mutex_unlock(&markers_mutex); - - marker_update_probes(); - - mutex_lock(&markers_mutex); - entry = get_marker(channel, name); - if (!entry) - goto end; - if (entry->rcu_pending) - rcu_barrier_sched(); - entry->oldptr = old; - entry->rcu_pending = 1; - /* write rcu_pending before calling the RCU callback */ - smp_wmb(); - call_rcu_sched(&entry->rcu, free_old_closure); - /*synchronize_rcu(); free_old_closure();*/ - goto end; - -error_unregister_channel: - ret_err = ltt_channels_unregister(channel); - WARN_ON(ret_err); -error_remove_marker: - ret_err = remove_marker(channel, name); - WARN_ON(ret_err); -end: - mutex_unlock(&markers_mutex); - return ret; -} -//ust// EXPORT_SYMBOL_GPL(marker_probe_register); - -/** - * marker_probe_unregister - Disconnect a probe from a marker - * @channel: marker channel - * @name: marker name - * @probe: probe function pointer - * @probe_private: probe private data - * - * Returns the private data given to marker_probe_register, or an ERR_PTR(). - * We do not need to call a synchronize_sched to make sure the probes have - * finished running before doing a module unload, because the module unload - * itself uses stop_machine(), which insures that every preempt disabled section - * have finished. - */ -int marker_probe_unregister(const char *channel, const char *name, - marker_probe_func *probe, void *probe_private) -{ - struct marker_entry *entry; - struct marker_probe_closure *old; - int ret = -ENOENT; - - mutex_lock(&markers_mutex); - entry = get_marker(channel, name); - if (!entry) - goto end; - if (entry->rcu_pending) - rcu_barrier_sched(); - old = marker_entry_remove_probe(entry, probe, probe_private); - mutex_unlock(&markers_mutex); - - marker_update_probes(); - - mutex_lock(&markers_mutex); - entry = get_marker(channel, name); - if (!entry) - goto end; - if (entry->rcu_pending) - rcu_barrier_sched(); - entry->oldptr = old; - entry->rcu_pending = 1; - /* write rcu_pending before calling the RCU callback */ - smp_wmb(); - call_rcu_sched(&entry->rcu, free_old_closure); - remove_marker(channel, name); /* Ignore busy error message */ - ret = 0; -end: - mutex_unlock(&markers_mutex); - return ret; -} -//ust// EXPORT_SYMBOL_GPL(marker_probe_unregister); - -static struct marker_entry * -get_marker_from_private_data(marker_probe_func *probe, void *probe_private) -{ - struct marker_entry *entry; - unsigned int i; - struct hlist_head *head; - struct hlist_node *node; - - for (i = 0; i < MARKER_TABLE_SIZE; i++) { - head = &marker_table[i]; - hlist_for_each_entry(entry, node, head, hlist) { - if (!entry->ptype) { - if (entry->single.func == probe - && entry->single.probe_private - == probe_private) - return entry; - } else { - struct marker_probe_closure *closure; - closure = entry->multi; - for (i = 0; closure[i].func; i++) { - if (closure[i].func == probe && - closure[i].probe_private - == probe_private) - return entry; - } - } - } - } - return NULL; -} - -/** - * marker_probe_unregister_private_data - Disconnect a probe from a marker - * @probe: probe function - * @probe_private: probe private data - * - * Unregister a probe by providing the registered private data. - * Only removes the first marker found in hash table. - * Return 0 on success or error value. - * We do not need to call a synchronize_sched to make sure the probes have - * finished running before doing a module unload, because the module unload - * itself uses stop_machine(), which insures that every preempt disabled section - * have finished. - */ -int marker_probe_unregister_private_data(marker_probe_func *probe, - void *probe_private) -{ - struct marker_entry *entry; - int ret = 0; - struct marker_probe_closure *old; - const char *channel = NULL, *name = NULL; - - mutex_lock(&markers_mutex); - entry = get_marker_from_private_data(probe, probe_private); - if (!entry) { - ret = -ENOENT; - goto end; - } - if (entry->rcu_pending) - rcu_barrier_sched(); - old = marker_entry_remove_probe(entry, NULL, probe_private); - channel = kstrdup(entry->channel, GFP_KERNEL); - name = kstrdup(entry->name, GFP_KERNEL); - mutex_unlock(&markers_mutex); - - marker_update_probes(); - - mutex_lock(&markers_mutex); - entry = get_marker(channel, name); - if (!entry) - goto end; - if (entry->rcu_pending) - rcu_barrier_sched(); - entry->oldptr = old; - entry->rcu_pending = 1; - /* write rcu_pending before calling the RCU callback */ - smp_wmb(); - call_rcu_sched(&entry->rcu, free_old_closure); - /* Ignore busy error message */ - remove_marker(channel, name); -end: - mutex_unlock(&markers_mutex); - kfree(channel); - kfree(name); - return ret; -} -//ust// EXPORT_SYMBOL_GPL(marker_probe_unregister_private_data); - -/** - * marker_get_private_data - Get a marker's probe private data - * @channel: marker channel - * @name: marker name - * @probe: probe to match - * @num: get the nth matching probe's private data - * - * Returns the nth private data pointer (starting from 0) matching, or an - * ERR_PTR. - * Returns the private data pointer, or an ERR_PTR. - * The private data pointer should _only_ be dereferenced if the caller is the - * owner of the data, or its content could vanish. This is mostly used to - * confirm that a caller is the owner of a registered probe. - */ -void *marker_get_private_data(const char *channel, const char *name, - marker_probe_func *probe, int num) -{ - struct hlist_head *head; - struct hlist_node *node; - struct marker_entry *e; - size_t channel_len = strlen(channel) + 1; - size_t name_len = strlen(name) + 1; - int i; - u32 hash; - - hash = jhash(channel, channel_len-1, 0) ^ jhash(name, name_len-1, 0); - head = &marker_table[hash & ((1 << MARKER_HASH_BITS)-1)]; - hlist_for_each_entry(e, node, head, hlist) { - if (!strcmp(channel, e->channel) && !strcmp(name, e->name)) { - if (!e->ptype) { - if (num == 0 && e->single.func == probe) - return e->single.probe_private; - } else { - struct marker_probe_closure *closure; - int match = 0; - closure = e->multi; - for (i = 0; closure[i].func; i++) { - if (closure[i].func != probe) - continue; - if (match++ == num) - return closure[i].probe_private; - } - } - break; - } - } - return ERR_PTR(-ENOENT); -} -//ust// EXPORT_SYMBOL_GPL(marker_get_private_data); - -/** - * markers_compact_event_ids - Compact markers event IDs and reassign channels - * - * Called when no channel users are active by the channel infrastructure. - * Called with lock_markers() and channel mutex held. - */ -//ust// void markers_compact_event_ids(void) -//ust// { -//ust// struct marker_entry *entry; -//ust// unsigned int i; -//ust// struct hlist_head *head; -//ust// struct hlist_node *node; -//ust// int ret; -//ust// -//ust// for (i = 0; i < MARKER_TABLE_SIZE; i++) { -//ust// head = &marker_table[i]; -//ust// hlist_for_each_entry(entry, node, head, hlist) { -//ust// ret = ltt_channels_get_index_from_name(entry->channel); -//ust// WARN_ON(ret < 0); -//ust// entry->channel_id = ret; -//ust// ret = _ltt_channels_get_event_id(entry->channel, -//ust// entry->name); -//ust// WARN_ON(ret < 0); -//ust// entry->event_id = ret; -//ust// } -//ust// } -//ust// } - -//ust//#ifdef CONFIG_MODULES - -/** - * marker_get_iter_range - Get a next marker iterator given a range. - * @marker: current markers (in), next marker (out) - * @begin: beginning of the range - * @end: end of the range - * - * Returns whether a next marker has been found (1) or not (0). - * Will return the first marker in the range if the input marker is NULL. - */ -int marker_get_iter_range(struct marker **marker, struct marker *begin, - struct marker *end) -{ - if (!*marker && begin != end) { - *marker = begin; - return 1; - } - if (*marker >= begin && *marker < end) - return 1; - return 0; -} -//ust// EXPORT_SYMBOL_GPL(marker_get_iter_range); - -static void marker_get_iter(struct marker_iter *iter) -{ - int found = 0; - - /* Core kernel markers */ - if (!iter->lib) { - /* ust FIXME: how come we cannot disable the following line? we shouldn't need core stuff */ - found = marker_get_iter_range(&iter->marker, - __start___markers, __stop___markers); - if (found) - goto end; - } - /* Markers in modules. */ - found = lib_get_iter_markers(iter); -end: - if (!found) - marker_iter_reset(iter); -} - -void marker_iter_start(struct marker_iter *iter) -{ - marker_get_iter(iter); -} -//ust// EXPORT_SYMBOL_GPL(marker_iter_start); - -void marker_iter_next(struct marker_iter *iter) -{ - iter->marker++; - /* - * iter->marker may be invalid because we blindly incremented it. - * Make sure it is valid by marshalling on the markers, getting the - * markers from following modules if necessary. - */ - marker_get_iter(iter); -} -//ust// EXPORT_SYMBOL_GPL(marker_iter_next); - -void marker_iter_stop(struct marker_iter *iter) -{ -} -//ust// EXPORT_SYMBOL_GPL(marker_iter_stop); - -void marker_iter_reset(struct marker_iter *iter) -{ - iter->lib = NULL; - iter->marker = NULL; -} -//ust// EXPORT_SYMBOL_GPL(marker_iter_reset); - -#ifdef CONFIG_MARKERS_USERSPACE -/* - * must be called with current->user_markers_mutex held - */ -static void free_user_marker(char __user *state, struct hlist_head *head) -{ - struct user_marker *umark; - struct hlist_node *pos, *n; - - hlist_for_each_entry_safe(umark, pos, n, head, hlist) { - if (umark->state == state) { - hlist_del(&umark->hlist); - kfree(umark); - } - } -} - -//ust// asmlinkage long sys_marker(char __user *name, char __user *format, -//ust// char __user *state, int reg) -//ust// { -//ust// struct user_marker *umark; -//ust// long len; -//ust// struct marker_entry *entry; -//ust// int ret = 0; -//ust// -//ust// printk(KERN_DEBUG "Program %s %s marker [%p, %p]\n", -//ust// current->comm, reg ? "registers" : "unregisters", -//ust// name, state); -//ust// if (reg) { -//ust// umark = kmalloc(sizeof(struct user_marker), GFP_KERNEL); -//ust// umark->name[MAX_USER_MARKER_NAME_LEN - 1] = '\0'; -//ust// umark->format[MAX_USER_MARKER_FORMAT_LEN - 1] = '\0'; -//ust// umark->state = state; -//ust// len = strncpy_from_user(umark->name, name, -//ust// MAX_USER_MARKER_NAME_LEN - 1); -//ust// if (len < 0) { -//ust// ret = -EFAULT; -//ust// goto error; -//ust// } -//ust// len = strncpy_from_user(umark->format, format, -//ust// MAX_USER_MARKER_FORMAT_LEN - 1); -//ust// if (len < 0) { -//ust// ret = -EFAULT; -//ust// goto error; -//ust// } -//ust// printk(KERN_DEBUG "Marker name : %s, format : %s", umark->name, -//ust// umark->format); -//ust// mutex_lock(&markers_mutex); -//ust// entry = get_marker("userspace", umark->name); -//ust// if (entry) { -//ust// if (entry->format && -//ust// strcmp(entry->format, umark->format) != 0) { -//ust// printk(" error, wrong format in process %s", -//ust// current->comm); -//ust// ret = -EPERM; -//ust// goto error_unlock; -//ust// } -//ust// printk(" %s", !!entry->refcount -//ust// ? "enabled" : "disabled"); -//ust// if (put_user(!!entry->refcount, state)) { -//ust// ret = -EFAULT; -//ust// goto error_unlock; -//ust// } -//ust// printk("\n"); -//ust// } else { -//ust// printk(" disabled\n"); -//ust// if (put_user(0, umark->state)) { -//ust// printk(KERN_WARNING -//ust// "Marker in %s caused a fault\n", -//ust// current->comm); -//ust// goto error_unlock; -//ust// } -//ust// } -//ust// mutex_lock(¤t->group_leader->user_markers_mutex); -//ust// hlist_add_head(&umark->hlist, -//ust// ¤t->group_leader->user_markers); -//ust// current->group_leader->user_markers_sequence++; -//ust// mutex_unlock(¤t->group_leader->user_markers_mutex); -//ust// mutex_unlock(&markers_mutex); -//ust// } else { -//ust// mutex_lock(¤t->group_leader->user_markers_mutex); -//ust// free_user_marker(state, -//ust// ¤t->group_leader->user_markers); -//ust// current->group_leader->user_markers_sequence++; -//ust// mutex_unlock(¤t->group_leader->user_markers_mutex); -//ust// } -//ust// goto end; -//ust// error_unlock: -//ust// mutex_unlock(&markers_mutex); -//ust// error: -//ust// kfree(umark); -//ust// end: -//ust// return ret; -//ust// } -//ust// -//ust// /* -//ust// * Types : -//ust// * string : 0 -//ust// */ -//ust// asmlinkage long sys_trace(int type, uint16_t id, -//ust// char __user *ubuf) -//ust// { -//ust// long ret = -EPERM; -//ust// char *page; -//ust// int len; -//ust// -//ust// switch (type) { -//ust// case 0: /* String */ -//ust// ret = -ENOMEM; -//ust// page = (char *)__get_free_page(GFP_TEMPORARY); -//ust// if (!page) -//ust// goto string_out; -//ust// len = strncpy_from_user(page, ubuf, PAGE_SIZE); -//ust// if (len < 0) { -//ust// ret = -EFAULT; -//ust// goto string_err; -//ust// } -//ust// trace_mark(userspace, string, "string %s", page); -//ust// string_err: -//ust// free_page((unsigned long) page); -//ust// string_out: -//ust// break; -//ust// default: -//ust// break; -//ust// } -//ust// return ret; -//ust// } - -//ust// static void marker_update_processes(void) -//ust// { -//ust// struct task_struct *g, *t; -//ust// -//ust// /* -//ust// * markers_mutex is taken to protect the p->user_markers read. -//ust// */ -//ust// mutex_lock(&markers_mutex); -//ust// read_lock(&tasklist_lock); -//ust// for_each_process(g) { -//ust// WARN_ON(!thread_group_leader(g)); -//ust// if (hlist_empty(&g->user_markers)) -//ust// continue; -//ust// if (strcmp(g->comm, "testprog") == 0) -//ust// printk(KERN_DEBUG "set update pending for testprog\n"); -//ust// t = g; -//ust// do { -//ust// /* TODO : implement this thread flag in each arch. */ -//ust// set_tsk_thread_flag(t, TIF_MARKER_PENDING); -//ust// } while ((t = next_thread(t)) != g); -//ust// } -//ust// read_unlock(&tasklist_lock); -//ust// mutex_unlock(&markers_mutex); -//ust// } - -/* - * Update current process. - * Note that we have to wait a whole scheduler period before we are sure that - * every running userspace threads have their markers updated. - * (synchronize_sched() can be used to insure this). - */ -void marker_update_process(void) -{ - struct user_marker *umark; - struct hlist_node *pos; - struct marker_entry *entry; - - mutex_lock(&markers_mutex); - mutex_lock(¤t->group_leader->user_markers_mutex); - if (strcmp(current->comm, "testprog") == 0) - printk(KERN_DEBUG "do update pending for testprog\n"); - hlist_for_each_entry(umark, pos, - ¤t->group_leader->user_markers, hlist) { - printk(KERN_DEBUG "Updating marker %s in %s\n", - umark->name, current->comm); - entry = get_marker("userspace", umark->name); - if (entry) { - if (entry->format && - strcmp(entry->format, umark->format) != 0) { - printk(KERN_WARNING - " error, wrong format in process %s\n", - current->comm); - break; - } - if (put_user(!!entry->refcount, umark->state)) { - printk(KERN_WARNING - "Marker in %s caused a fault\n", - current->comm); - break; - } - } else { - if (put_user(0, umark->state)) { - printk(KERN_WARNING - "Marker in %s caused a fault\n", - current->comm); - break; - } - } - } - clear_thread_flag(TIF_MARKER_PENDING); - mutex_unlock(¤t->group_leader->user_markers_mutex); - mutex_unlock(&markers_mutex); -} - -/* - * Called at process exit and upon do_execve(). - * We assume that when the leader exits, no more references can be done to the - * leader structure by the other threads. - */ -void exit_user_markers(struct task_struct *p) -{ - struct user_marker *umark; - struct hlist_node *pos, *n; - - if (thread_group_leader(p)) { - mutex_lock(&markers_mutex); - mutex_lock(&p->user_markers_mutex); - hlist_for_each_entry_safe(umark, pos, n, &p->user_markers, - hlist) - kfree(umark); - INIT_HLIST_HEAD(&p->user_markers); - p->user_markers_sequence++; - mutex_unlock(&p->user_markers_mutex); - mutex_unlock(&markers_mutex); - } -} - -int is_marker_enabled(const char *channel, const char *name) -{ - struct marker_entry *entry; - - mutex_lock(&markers_mutex); - entry = get_marker(channel, name); - mutex_unlock(&markers_mutex); - - return entry && !!entry->refcount; -} -//ust// #endif - -int marker_module_notify(struct notifier_block *self, - unsigned long val, void *data) -{ - struct module *mod = data; - - switch (val) { - case MODULE_STATE_COMING: - marker_update_probe_range(mod->markers, - mod->markers + mod->num_markers); - break; - case MODULE_STATE_GOING: - marker_update_probe_range(mod->markers, - mod->markers + mod->num_markers); - break; - } - return 0; -} - -struct notifier_block marker_module_nb = { - .notifier_call = marker_module_notify, - .priority = 0, -}; - -//ust// static int init_markers(void) -//ust// { -//ust// return register_module_notifier(&marker_module_nb); -//ust// } -//ust// __initcall(init_markers); -/* TODO: call marker_module_nb() when a library is linked at runtime (dlopen)? */ - -#endif /* CONFIG_MODULES */ - -void ltt_dump_marker_state(struct ltt_trace_struct *trace) -{ - struct marker_iter iter; - struct ltt_probe_private_data call_data; - const char *channel; - - call_data.trace = trace; - call_data.serializer = NULL; - - marker_iter_reset(&iter); - marker_iter_start(&iter); - for (; iter.marker != NULL; marker_iter_next(&iter)) { - if (!_imv_read(iter.marker->state)) - continue; - channel = ltt_channels_get_name_from_index( - iter.marker->channel_id); - __trace_mark(0, metadata, core_marker_id, - &call_data, - "channel %s name %s event_id %hu " - "int #1u%zu long #1u%zu pointer #1u%zu " - "size_t #1u%zu alignment #1u%u", - channel, - iter.marker->name, - iter.marker->event_id, - sizeof(int), sizeof(long), - sizeof(void *), sizeof(size_t), - ltt_get_alignment()); - if (iter.marker->format) - __trace_mark(0, metadata, - core_marker_format, - &call_data, - "channel %s name %s format %s", - channel, - iter.marker->name, - iter.marker->format); - } - marker_iter_stop(&iter); -} -//ust// EXPORT_SYMBOL_GPL(ltt_dump_marker_state); - - -static LIST_HEAD(libs); - -/* - * Returns 0 if current not found. - * Returns 1 if current found. - */ -int lib_get_iter_markers(struct marker_iter *iter) -{ - struct lib *iter_lib; - int found = 0; - -//ust// mutex_lock(&module_mutex); - list_for_each_entry(iter_lib, &libs, list) { - if (iter_lib < iter->lib) - continue; - else if (iter_lib > iter->lib) - iter->marker = NULL; - found = marker_get_iter_range(&iter->marker, - iter_lib->markers_start, - iter_lib->markers_start + iter_lib->markers_count); - if (found) { - iter->lib = iter_lib; - break; - } - } -//ust// mutex_unlock(&module_mutex); - return found; -} - -void lib_update_markers(void) -{ - struct lib *lib; - -//ust// mutex_lock(&module_mutex); - list_for_each_entry(lib, &libs, list) - marker_update_probe_range(lib->markers_start, - lib->markers_start + lib->markers_count); -//ust// mutex_unlock(&module_mutex); -} - -static void (*new_marker_cb)(struct marker *) = NULL; - -void marker_set_new_marker_cb(void (*cb)(struct marker *)) -{ - new_marker_cb = cb; -} - -static void new_markers(struct marker *start, struct marker *end) -{ - if(new_marker_cb) { - struct marker *m; - for(m=start; m < end; m++) { - new_marker_cb(m); - } - } -} - -int marker_register_lib(struct marker *markers_start, int markers_count) -{ - struct lib *pl; - - pl = (struct lib *) malloc(sizeof(struct lib)); - - pl->markers_start = markers_start; - pl->markers_count = markers_count; - - /* FIXME: maybe protect this with its own mutex? */ - lock_markers(); - list_add(&pl->list, &libs); - unlock_markers(); - - new_markers(markers_start, markers_start + markers_count); - - /* FIXME: update just the loaded lib */ - lib_update_markers(); - - DBG("just registered a markers section from %p and having %d markers", markers_start, markers_count); - - return 0; -} - -int marker_unregister_lib(struct marker *markers_start, int markers_count) -{ - /*FIXME: implement; but before implementing, marker_register_lib must - have appropriate locking. */ - - return 0; -} - -static int initialized = 0; - -void __attribute__((constructor)) init_markers(void) -{ - if(!initialized) { - marker_register_lib(__start___markers, (((long)__stop___markers)-((long)__start___markers))/sizeof(struct marker)); - printf("markers_start: %p, markers_stop: %p\n", __start___markers, __stop___markers); - initialized = 1; - } -} diff --git a/libmarkers/marker.h b/libmarkers/marker.h deleted file mode 100644 index 7c3d8c5..0000000 --- a/libmarkers/marker.h +++ /dev/null @@ -1,298 +0,0 @@ -/* - * Code markup for dynamic and static tracing. - * - * See Documentation/marker.txt. - * - * (C) Copyright 2006 Mathieu Desnoyers - * (C) Copyright 2009 Pierre-Marc Fournier - * - * This file is released under the GPLv2. - * See the file COPYING for more details. - */ - -#ifndef _LINUX_MARKER_H -#define _LINUX_MARKER_H - -#include -//ust// #include -#include "immediate.h" -//ust// #include -#include "kernelcompat.h" -#include "compiler.h" -#include "list.h" - -//ust// struct module; -//ust// struct task_struct; -struct marker; - -/** - * marker_probe_func - Type of a marker probe function - * @mdata: marker data - * @probe_private: probe private data - * @call_private: call site private data - * @fmt: format string - * @args: variable argument list pointer. Use a pointer to overcome C's - * inability to pass this around as a pointer in a portable manner in - * the callee otherwise. - * - * Type of marker probe functions. They receive the mdata and need to parse the - * format string to recover the variable argument list. - */ -typedef void marker_probe_func(const struct marker *mdata, - void *probe_private, void *call_private, - const char *fmt, va_list *args); - -struct marker_probe_closure { - marker_probe_func *func; /* Callback */ - void *probe_private; /* Private probe data */ -}; - -struct marker { - const char *channel; /* Name of channel where to send data */ - const char *name; /* Marker name */ - const char *format; /* Marker format string, describing the - * variable argument list. - */ - DEFINE_IMV(char, state);/* Immediate value state. */ - char ptype; /* probe type : 0 : single, 1 : multi */ - /* Probe wrapper */ - u16 channel_id; /* Numeric channel identifier, dynamic */ - u16 event_id; /* Numeric event identifier, dynamic */ - void (*call)(const struct marker *mdata, void *call_private, ...); - struct marker_probe_closure single; - struct marker_probe_closure *multi; - const char *tp_name; /* Optional tracepoint name */ - void *tp_cb; /* Optional tracepoint callback */ -} __attribute__((aligned(8))); - -//ust// #ifdef CONFIG_MARKERS - -#define _DEFINE_MARKER(channel, name, tp_name_str, tp_cb, format) \ - static const char __mstrtab_##channel##_##name[] \ - __attribute__((section("__markers_strings"))) \ - = #channel "\0" #name "\0" format; \ - static struct marker __mark_##channel##_##name \ - __attribute__((section("__markers"), aligned(8))) = \ - { __mstrtab_##channel##_##name, \ - &__mstrtab_##channel##_##name[sizeof(#channel)], \ - &__mstrtab_##channel##_##name[sizeof(#channel) + \ - sizeof(#name)], \ - 0, 0, 0, 0, marker_probe_cb, \ - { __mark_empty_function, NULL}, \ - NULL, tp_name_str, tp_cb } - -#define DEFINE_MARKER(channel, name, format) \ - _DEFINE_MARKER(channel, name, NULL, NULL, format) - -#define DEFINE_MARKER_TP(channel, name, tp_name, tp_cb, format) \ - _DEFINE_MARKER(channel, name, #tp_name, tp_cb, format) - -/* - * Make sure the alignment of the structure in the __markers section will - * not add unwanted padding between the beginning of the section and the - * structure. Force alignment to the same alignment as the section start. - * - * The "generic" argument controls which marker enabling mechanism must be used. - * If generic is true, a variable read is used. - * If generic is false, immediate values are used. - */ -#define __trace_mark(generic, channel, name, call_private, format, args...) \ - do { \ - DEFINE_MARKER(channel, name, format); \ - __mark_check_format(format, ## args); \ - if (!generic) { \ - if (unlikely(imv_read( \ - __mark_##channel##_##name.state))) \ - (*__mark_##channel##_##name.call) \ - (&__mark_##channel##_##name, \ - call_private, ## args); \ - } else { \ - if (unlikely(_imv_read( \ - __mark_##channel##_##name.state))) \ - (*__mark_##channel##_##name.call) \ - (&__mark_##channel##_##name, \ - call_private, ## args); \ - } \ - } while (0) - -#define __trace_mark_tp(channel, name, call_private, tp_name, tp_cb, \ - format, args...) \ - do { \ - void __check_tp_type(void) \ - { \ - register_trace_##tp_name(tp_cb); \ - } \ - DEFINE_MARKER_TP(channel, name, tp_name, tp_cb, format);\ - __mark_check_format(format, ## args); \ - (*__mark_##channel##_##name.call)(&__mark_##channel##_##name, \ - call_private, ## args); \ - } while (0) - -extern void marker_update_probe_range(struct marker *begin, - struct marker *end); - -#define GET_MARKER(channel, name) (__mark_##channel##_##name) - -//ust// #else /* !CONFIG_MARKERS */ -//ust// #define DEFINE_MARKER(channel, name, tp_name, tp_cb, format) -//ust// #define __trace_mark(generic, channel, name, call_private, format, args...) \ -//ust// __mark_check_format(format, ## args) -//ust// #define __trace_mark_tp(channel, name, call_private, tp_name, tp_cb, \ -//ust// format, args...) \ -//ust// do { \ -//ust// void __check_tp_type(void) \ -//ust// { \ -//ust// register_trace_##tp_name(tp_cb); \ -//ust// } \ -//ust// __mark_check_format(format, ## args); \ -//ust// } while (0) -//ust// static inline void marker_update_probe_range(struct marker *begin, -//ust// struct marker *end) -//ust// { } -//ust// #define GET_MARKER(channel, name) -//ust// #endif /* CONFIG_MARKERS */ - -/** - * trace_mark - Marker using code patching - * @channel: marker channel (where to send the data), not quoted. - * @name: marker name, not quoted. - * @format: format string - * @args...: variable argument list - * - * Places a marker using optimized code patching technique (imv_read()) - * to be enabled when immediate values are present. - */ -#define trace_mark(channel, name, format, args...) \ - __trace_mark(0, channel, name, NULL, format, ## args) - -/** - * _trace_mark - Marker using variable read - * @channel: marker channel (where to send the data), not quoted. - * @name: marker name, not quoted. - * @format: format string - * @args...: variable argument list - * - * Places a marker using a standard memory read (_imv_read()) to be - * enabled. Should be used for markers in code paths where instruction - * modification based enabling is not welcome. (__init and __exit functions, - * lockdep, some traps, printk). - */ -#define _trace_mark(channel, name, format, args...) \ - __trace_mark(1, channel, name, NULL, format, ## args) - -/** - * trace_mark_tp - Marker in a tracepoint callback - * @channel: marker channel (where to send the data), not quoted. - * @name: marker name, not quoted. - * @tp_name: tracepoint name, not quoted. - * @tp_cb: tracepoint callback. Should have an associated global symbol so it - * is not optimized away by the compiler (should not be static). - * @format: format string - * @args...: variable argument list - * - * Places a marker in a tracepoint callback. - */ -#define trace_mark_tp(channel, name, tp_name, tp_cb, format, args...) \ - __trace_mark_tp(channel, name, NULL, tp_name, tp_cb, format, ## args) - -/** - * MARK_NOARGS - Format string for a marker with no argument. - */ -#define MARK_NOARGS " " - -extern void lock_markers(void); -extern void unlock_markers(void); - -extern void markers_compact_event_ids(void); - -/* To be used for string format validity checking with gcc */ -static inline void __printf(1, 2) ___mark_check_format(const char *fmt, ...) -{ -} - -#define __mark_check_format(format, args...) \ - do { \ - if (0) \ - ___mark_check_format(format, ## args); \ - } while (0) - -extern marker_probe_func __mark_empty_function; - -extern void marker_probe_cb(const struct marker *mdata, - void *call_private, ...); - -/* - * Connect a probe to a marker. - * private data pointer must be a valid allocated memory address, or NULL. - */ -extern int marker_probe_register(const char *channel, const char *name, - const char *format, marker_probe_func *probe, void *probe_private); - -/* - * Returns the private data given to marker_probe_register. - */ -extern int marker_probe_unregister(const char *channel, const char *name, - marker_probe_func *probe, void *probe_private); -/* - * Unregister a marker by providing the registered private data. - */ -extern int marker_probe_unregister_private_data(marker_probe_func *probe, - void *probe_private); - -extern void *marker_get_private_data(const char *channel, const char *name, - marker_probe_func *probe, int num); - -/* - * marker_synchronize_unregister must be called between the last marker probe - * unregistration and the first one of - * - the end of module exit function - * - the free of any resource used by the probes - * to ensure the code and data are valid for any possibly running probes. - */ -#define marker_synchronize_unregister() synchronize_sched() - -struct marker_iter { -//ust// struct module *module; - struct lib *lib; - struct marker *marker; -}; - -extern void marker_iter_start(struct marker_iter *iter); -extern void marker_iter_next(struct marker_iter *iter); -extern void marker_iter_stop(struct marker_iter *iter); -extern void marker_iter_reset(struct marker_iter *iter); -extern int marker_get_iter_range(struct marker **marker, struct marker *begin, - struct marker *end); - -extern void marker_update_process(void); -extern int is_marker_enabled(const char *channel, const char *name); - -//ust// #ifdef CONFIG_MARKERS_USERSPACE -//ust// extern void exit_user_markers(struct task_struct *p); -//ust// #else -//ust// static inline void exit_user_markers(struct task_struct *p) -//ust// { -//ust// } -//ust// #endif - - -struct lib { - struct marker *markers_start; - int markers_count; - struct list_head list; -}; - -int marker_register_lib(struct marker *markers_start, int markers_count); - -#define MARKER_LIB \ -extern struct marker __start___markers[] __attribute__((visibility("hidden"))); \ -extern struct marker __stop___markers[] __attribute__((visibility("hidden"))); \ - \ -static void __attribute__((constructor)) __markers__init(void) \ -{ \ - marker_register_lib(__start___markers, (((long)__stop___markers)-((long)__start___markers))/sizeof(struct marker));\ -} - -void marker_set_new_marker_cb(void (*cb)(struct marker *)); - -#endif diff --git a/libmarkers/tracepoint.c b/libmarkers/tracepoint.c deleted file mode 100644 index ba4509b..0000000 --- a/libmarkers/tracepoint.c +++ /dev/null @@ -1,685 +0,0 @@ -/* - * Copyright (C) 2008 Mathieu Desnoyers - * Copyright (C) 2009 Pierre-Marc Fournier - * - * This program is free software; you can redistribute it and/or - * modify it under the terms of the GNU General Public License - * as published by the Free Software Foundation; either version 2 - * of the License, or (at your option) any later version. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * GNU General Public License for more details. - * - * You should have received a copy of the GNU General Public License - * along with this program; if not, write to the Free Software - * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. - * - * Ported to userspace by Pierre-Marc Fournier. - */ - -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include - -#include - -#include "kernelcompat.h" -#include "tracepoint.h" -#include "usterr.h" -#include "list.h" - -//extern struct tracepoint __start___tracepoints[] __attribute__((visibility("hidden"))); -//extern struct tracepoint __stop___tracepoints[] __attribute__((visibility("hidden"))); - -/* Set to 1 to enable tracepoint debug output */ -static const int tracepoint_debug; - -/* libraries that contain tracepoints (struct tracepoint_lib) */ -static LIST_HEAD(libs); - -/* - * tracepoints_mutex nests inside module_mutex. Tracepoints mutex protects the - * builtin and module tracepoints and the hash table. - */ -static DEFINE_MUTEX(tracepoints_mutex); - -/* - * Tracepoint hash table, containing the active tracepoints. - * Protected by tracepoints_mutex. - */ -#define TRACEPOINT_HASH_BITS 6 -#define TRACEPOINT_TABLE_SIZE (1 << TRACEPOINT_HASH_BITS) -static struct hlist_head tracepoint_table[TRACEPOINT_TABLE_SIZE]; - -/* - * Note about RCU : - * It is used to to delay the free of multiple probes array until a quiescent - * state is reached. - * Tracepoint entries modifications are protected by the tracepoints_mutex. - */ -struct tracepoint_entry { - struct hlist_node hlist; - void **funcs; - int refcount; /* Number of times armed. 0 if disarmed. */ - char name[0]; -}; - -struct tp_probes { - union { -//ust// struct rcu_head rcu; - struct list_head list; - } u; - void *probes[0]; -}; - -static inline void *allocate_probes(int count) -{ - struct tp_probes *p = kmalloc(count * sizeof(void *) - + sizeof(struct tp_probes), GFP_KERNEL); - return p == NULL ? NULL : p->probes; -} - -//ust// static void rcu_free_old_probes(struct rcu_head *head) -//ust// { -//ust// kfree(container_of(head, struct tp_probes, u.rcu)); -//ust// } - -static inline void release_probes(void *old) -{ - if (old) { - struct tp_probes *tp_probes = container_of(old, - struct tp_probes, probes[0]); -//ust// call_rcu_sched(&tp_probes->u.rcu, rcu_free_old_probes); - synchronize_rcu(); - kfree(tp_probes); - } -} - -static void debug_print_probes(struct tracepoint_entry *entry) -{ - int i; - - if (!tracepoint_debug || !entry->funcs) - return; - - for (i = 0; entry->funcs[i]; i++) - printk(KERN_DEBUG "Probe %d : %p\n", i, entry->funcs[i]); -} - -static void * -tracepoint_entry_add_probe(struct tracepoint_entry *entry, void *probe) -{ - int nr_probes = 0; - void **old, **new; - - WARN_ON(!probe); - - debug_print_probes(entry); - old = entry->funcs; - if (old) { - /* (N -> N+1), (N != 0, 1) probes */ - for (nr_probes = 0; old[nr_probes]; nr_probes++) - if (old[nr_probes] == probe) - return ERR_PTR(-EEXIST); - } - /* + 2 : one for new probe, one for NULL func */ - new = allocate_probes(nr_probes + 2); - if (new == NULL) - return ERR_PTR(-ENOMEM); - if (old) - memcpy(new, old, nr_probes * sizeof(void *)); - new[nr_probes] = probe; - new[nr_probes + 1] = NULL; - entry->refcount = nr_probes + 1; - entry->funcs = new; - debug_print_probes(entry); - return old; -} - -static void * -tracepoint_entry_remove_probe(struct tracepoint_entry *entry, void *probe) -{ - int nr_probes = 0, nr_del = 0, i; - void **old, **new; - - old = entry->funcs; - - if (!old) - return ERR_PTR(-ENOENT); - - debug_print_probes(entry); - /* (N -> M), (N > 1, M >= 0) probes */ - for (nr_probes = 0; old[nr_probes]; nr_probes++) { - if ((!probe || old[nr_probes] == probe)) - nr_del++; - } - - if (nr_probes - nr_del == 0) { - /* N -> 0, (N > 1) */ - entry->funcs = NULL; - entry->refcount = 0; - debug_print_probes(entry); - return old; - } else { - int j = 0; - /* N -> M, (N > 1, M > 0) */ - /* + 1 for NULL */ - new = allocate_probes(nr_probes - nr_del + 1); - if (new == NULL) - return ERR_PTR(-ENOMEM); - for (i = 0; old[i]; i++) - if ((probe && old[i] != probe)) - new[j++] = old[i]; - new[nr_probes - nr_del] = NULL; - entry->refcount = nr_probes - nr_del; - entry->funcs = new; - } - debug_print_probes(entry); - return old; -} - -/* - * Get tracepoint if the tracepoint is present in the tracepoint hash table. - * Must be called with tracepoints_mutex held. - * Returns NULL if not present. - */ -static struct tracepoint_entry *get_tracepoint(const char *name) -{ - struct hlist_head *head; - struct hlist_node *node; - struct tracepoint_entry *e; - u32 hash = jhash(name, strlen(name), 0); - - head = &tracepoint_table[hash & (TRACEPOINT_TABLE_SIZE - 1)]; - hlist_for_each_entry(e, node, head, hlist) { - if (!strcmp(name, e->name)) - return e; - } - return NULL; -} - -/* - * Add the tracepoint to the tracepoint hash table. Must be called with - * tracepoints_mutex held. - */ -static struct tracepoint_entry *add_tracepoint(const char *name) -{ - struct hlist_head *head; - struct hlist_node *node; - struct tracepoint_entry *e; - size_t name_len = strlen(name) + 1; - u32 hash = jhash(name, name_len-1, 0); - - head = &tracepoint_table[hash & (TRACEPOINT_TABLE_SIZE - 1)]; - hlist_for_each_entry(e, node, head, hlist) { - if (!strcmp(name, e->name)) { - printk(KERN_NOTICE - "tracepoint %s busy\n", name); - return ERR_PTR(-EEXIST); /* Already there */ - } - } - /* - * Using kmalloc here to allocate a variable length element. Could - * cause some memory fragmentation if overused. - */ - e = kmalloc(sizeof(struct tracepoint_entry) + name_len, GFP_KERNEL); - if (!e) - return ERR_PTR(-ENOMEM); - memcpy(&e->name[0], name, name_len); - e->funcs = NULL; - e->refcount = 0; - hlist_add_head(&e->hlist, head); - return e; -} - -/* - * Remove the tracepoint from the tracepoint hash table. Must be called with - * mutex_lock held. - */ -static inline void remove_tracepoint(struct tracepoint_entry *e) -{ - hlist_del(&e->hlist); - kfree(e); -} - -/* - * Sets the probe callback corresponding to one tracepoint. - */ -static void set_tracepoint(struct tracepoint_entry **entry, - struct tracepoint *elem, int active) -{ - WARN_ON(strcmp((*entry)->name, elem->name) != 0); - - /* - * rcu_assign_pointer has a smp_wmb() which makes sure that the new - * probe callbacks array is consistent before setting a pointer to it. - * This array is referenced by __DO_TRACE from - * include/linux/tracepoints.h. A matching smp_read_barrier_depends() - * is used. - */ - rcu_assign_pointer(elem->funcs, (*entry)->funcs); - elem->state__imv = active; -} - -/* - * Disable a tracepoint and its probe callback. - * Note: only waiting an RCU period after setting elem->call to the empty - * function insures that the original callback is not used anymore. This insured - * by preempt_disable around the call site. - */ -static void disable_tracepoint(struct tracepoint *elem) -{ - elem->state__imv = 0; - rcu_assign_pointer(elem->funcs, NULL); -} - -/** - * tracepoint_update_probe_range - Update a probe range - * @begin: beginning of the range - * @end: end of the range - * - * Updates the probe callback corresponding to a range of tracepoints. - */ -void tracepoint_update_probe_range(struct tracepoint *begin, - struct tracepoint *end) -{ - struct tracepoint *iter; - struct tracepoint_entry *mark_entry; - - mutex_lock(&tracepoints_mutex); - for (iter = begin; iter < end; iter++) { - mark_entry = get_tracepoint(iter->name); - if (mark_entry) { - set_tracepoint(&mark_entry, iter, - !!mark_entry->refcount); - } else { - disable_tracepoint(iter); - } - } - mutex_unlock(&tracepoints_mutex); -} - -/* - * Update probes, removing the faulty probes. - */ -static void tracepoint_update_probes(void) -{ - /* Core kernel tracepoints */ -//ust// tracepoint_update_probe_range(__start___tracepoints, -//ust// __stop___tracepoints); - /* tracepoints in modules. */ - lib_update_tracepoints(); - /* Update immediate values */ - core_imv_update(); -//ust// module_imv_update(); -} - -static void *tracepoint_add_probe(const char *name, void *probe) -{ - struct tracepoint_entry *entry; - void *old; - - entry = get_tracepoint(name); - if (!entry) { - entry = add_tracepoint(name); - if (IS_ERR(entry)) - return entry; - } - old = tracepoint_entry_add_probe(entry, probe); - if (IS_ERR(old) && !entry->refcount) - remove_tracepoint(entry); - return old; -} - -/** - * tracepoint_probe_register - Connect a probe to a tracepoint - * @name: tracepoint name - * @probe: probe handler - * - * Returns 0 if ok, error value on error. - * The probe address must at least be aligned on the architecture pointer size. - */ -int tracepoint_probe_register(const char *name, void *probe) -{ - void *old; - - mutex_lock(&tracepoints_mutex); - old = tracepoint_add_probe(name, probe); - mutex_unlock(&tracepoints_mutex); - if (IS_ERR(old)) - return PTR_ERR(old); - - tracepoint_update_probes(); /* may update entry */ - release_probes(old); - return 0; -} -//ust// EXPORT_SYMBOL_GPL(tracepoint_probe_register); - -static void *tracepoint_remove_probe(const char *name, void *probe) -{ - struct tracepoint_entry *entry; - void *old; - - entry = get_tracepoint(name); - if (!entry) - return ERR_PTR(-ENOENT); - old = tracepoint_entry_remove_probe(entry, probe); - if (IS_ERR(old)) - return old; - if (!entry->refcount) - remove_tracepoint(entry); - return old; -} - -/** - * tracepoint_probe_unregister - Disconnect a probe from a tracepoint - * @name: tracepoint name - * @probe: probe function pointer - * - * We do not need to call a synchronize_sched to make sure the probes have - * finished running before doing a module unload, because the module unload - * itself uses stop_machine(), which insures that every preempt disabled section - * have finished. - */ -int tracepoint_probe_unregister(const char *name, void *probe) -{ - void *old; - - mutex_lock(&tracepoints_mutex); - old = tracepoint_remove_probe(name, probe); - mutex_unlock(&tracepoints_mutex); - if (IS_ERR(old)) - return PTR_ERR(old); - - tracepoint_update_probes(); /* may update entry */ - release_probes(old); - return 0; -} -//ust// EXPORT_SYMBOL_GPL(tracepoint_probe_unregister); - -static LIST_HEAD(old_probes); -static int need_update; - -static void tracepoint_add_old_probes(void *old) -{ - need_update = 1; - if (old) { - struct tp_probes *tp_probes = container_of(old, - struct tp_probes, probes[0]); - list_add(&tp_probes->u.list, &old_probes); - } -} - -/** - * tracepoint_probe_register_noupdate - register a probe but not connect - * @name: tracepoint name - * @probe: probe handler - * - * caller must call tracepoint_probe_update_all() - */ -int tracepoint_probe_register_noupdate(const char *name, void *probe) -{ - void *old; - - mutex_lock(&tracepoints_mutex); - old = tracepoint_add_probe(name, probe); - if (IS_ERR(old)) { - mutex_unlock(&tracepoints_mutex); - return PTR_ERR(old); - } - tracepoint_add_old_probes(old); - mutex_unlock(&tracepoints_mutex); - return 0; -} -//ust// EXPORT_SYMBOL_GPL(tracepoint_probe_register_noupdate); - -/** - * tracepoint_probe_unregister_noupdate - remove a probe but not disconnect - * @name: tracepoint name - * @probe: probe function pointer - * - * caller must call tracepoint_probe_update_all() - */ -int tracepoint_probe_unregister_noupdate(const char *name, void *probe) -{ - void *old; - - mutex_lock(&tracepoints_mutex); - old = tracepoint_remove_probe(name, probe); - if (IS_ERR(old)) { - mutex_unlock(&tracepoints_mutex); - return PTR_ERR(old); - } - tracepoint_add_old_probes(old); - mutex_unlock(&tracepoints_mutex); - return 0; -} -//ust// EXPORT_SYMBOL_GPL(tracepoint_probe_unregister_noupdate); - -/** - * tracepoint_probe_update_all - update tracepoints - */ -void tracepoint_probe_update_all(void) -{ - LIST_HEAD(release_probes); - struct tp_probes *pos, *next; - - mutex_lock(&tracepoints_mutex); - if (!need_update) { - mutex_unlock(&tracepoints_mutex); - return; - } - if (!list_empty(&old_probes)) - list_replace_init(&old_probes, &release_probes); - need_update = 0; - mutex_unlock(&tracepoints_mutex); - - tracepoint_update_probes(); - list_for_each_entry_safe(pos, next, &release_probes, u.list) { - list_del(&pos->u.list); -//ust// call_rcu_sched(&pos->u.rcu, rcu_free_old_probes); - synchronize_rcu(); - kfree(pos); - } -} -//ust// EXPORT_SYMBOL_GPL(tracepoint_probe_update_all); - -/** - * tracepoint_get_iter_range - Get a next tracepoint iterator given a range. - * @tracepoint: current tracepoints (in), next tracepoint (out) - * @begin: beginning of the range - * @end: end of the range - * - * Returns whether a next tracepoint has been found (1) or not (0). - * Will return the first tracepoint in the range if the input tracepoint is - * NULL. - */ -int tracepoint_get_iter_range(struct tracepoint **tracepoint, - struct tracepoint *begin, struct tracepoint *end) -{ - if (!*tracepoint && begin != end) { - *tracepoint = begin; - return 1; - } - if (*tracepoint >= begin && *tracepoint < end) - return 1; - return 0; -} -//ust// EXPORT_SYMBOL_GPL(tracepoint_get_iter_range); - -static void tracepoint_get_iter(struct tracepoint_iter *iter) -{ - int found = 0; - -//ust// /* Core kernel tracepoints */ -//ust// if (!iter->module) { -//ust// found = tracepoint_get_iter_range(&iter->tracepoint, -//ust// __start___tracepoints, __stop___tracepoints); -//ust// if (found) -//ust// goto end; -//ust// } - /* tracepoints in libs. */ - found = lib_get_iter_tracepoints(iter); -end: - if (!found) - tracepoint_iter_reset(iter); -} - -void tracepoint_iter_start(struct tracepoint_iter *iter) -{ - tracepoint_get_iter(iter); -} -//ust// EXPORT_SYMBOL_GPL(tracepoint_iter_start); - -void tracepoint_iter_next(struct tracepoint_iter *iter) -{ - iter->tracepoint++; - /* - * iter->tracepoint may be invalid because we blindly incremented it. - * Make sure it is valid by marshalling on the tracepoints, getting the - * tracepoints from following modules if necessary. - */ - tracepoint_get_iter(iter); -} -//ust// EXPORT_SYMBOL_GPL(tracepoint_iter_next); - -void tracepoint_iter_stop(struct tracepoint_iter *iter) -{ -} -//ust// EXPORT_SYMBOL_GPL(tracepoint_iter_stop); - -void tracepoint_iter_reset(struct tracepoint_iter *iter) -{ -//ust// iter->module = NULL; - iter->tracepoint = NULL; -} -//ust// EXPORT_SYMBOL_GPL(tracepoint_iter_reset); - -//ust// #ifdef CONFIG_MODULES - -//ust// int tracepoint_module_notify(struct notifier_block *self, -//ust// unsigned long val, void *data) -//ust// { -//ust// struct module *mod = data; -//ust// -//ust// switch (val) { -//ust// case MODULE_STATE_COMING: -//ust// tracepoint_update_probe_range(mod->tracepoints, -//ust// mod->tracepoints + mod->num_tracepoints); -//ust// break; -//ust// case MODULE_STATE_GOING: -//ust// tracepoint_update_probe_range(mod->tracepoints, -//ust// mod->tracepoints + mod->num_tracepoints); -//ust// break; -//ust// } -//ust// return 0; -//ust// } - -//ust// struct notifier_block tracepoint_module_nb = { -//ust// .notifier_call = tracepoint_module_notify, -//ust// .priority = 0, -//ust// }; - -//ust// static int init_tracepoints(void) -//ust// { -//ust// return register_module_notifier(&tracepoint_module_nb); -//ust// } -//ust// __initcall(init_tracepoints); - -//ust// #endif /* CONFIG_MODULES */ - -/* - * Returns 0 if current not found. - * Returns 1 if current found. - */ -int lib_get_iter_tracepoints(struct tracepoint_iter *iter) -{ - struct tracepoint_lib *iter_lib; - int found = 0; - -//ust// mutex_lock(&module_mutex); - list_for_each_entry(iter_lib, &libs, list) { - if (iter_lib < iter->lib) - continue; - else if (iter_lib > iter->lib) - iter->tracepoint = NULL; - found = marker_get_iter_range(&iter->tracepoint, - iter_lib->tracepoints_start, - iter_lib->tracepoints_start + iter_lib->tracepoints_count); - if (found) { - iter->lib = iter_lib; - break; - } - } -//ust// mutex_unlock(&module_mutex); - return found; -} - -void lib_update_tracepoints(void) -{ - struct tracepoint_lib *lib; - -//ust// mutex_lock(&module_mutex); - list_for_each_entry(lib, &libs, list) - tracepoint_update_probe_range(lib->tracepoints_start, - lib->tracepoints_start + lib->tracepoints_count); -//ust// mutex_unlock(&module_mutex); -} - -static void (*new_tracepoint_cb)(struct tracepoint *) = NULL; - -void tracepoint_set_new_tracepoint_cb(void (*cb)(struct tracepoint *)) -{ - new_tracepoint_cb = cb; -} - -static void new_tracepoints(struct tracepoint *start, struct tracepoint *end) -{ - if(new_tracepoint_cb) { - struct tracepoint *t; - for(t=start; t < end; t++) { - new_tracepoint_cb(t); - } - } -} - -int tracepoint_register_lib(struct tracepoint *tracepoints_start, int tracepoints_count) -{ - struct tracepoint_lib *pl; - - pl = (struct tracepoint_lib *) malloc(sizeof(struct tracepoint_lib)); - - pl->tracepoints_start = tracepoints_start; - pl->tracepoints_count = tracepoints_count; - - /* FIXME: maybe protect this with its own mutex? */ - mutex_lock(&tracepoints_mutex); - list_add(&pl->list, &libs); - mutex_unlock(&tracepoints_mutex); - - new_tracepoints(tracepoints_start, tracepoints_start + tracepoints_count); - - /* FIXME: update just the loaded lib */ - lib_update_tracepoints(); - - DBG("just registered a tracepoints section from %p and having %d tracepoints", tracepoints_start, tracepoints_count); - - return 0; -} - -int tracepoint_unregister_lib(struct tracepoint *tracepoints_start, int tracepoints_count) -{ - /*FIXME: implement; but before implementing, tracepoint_register_lib must - have appropriate locking. */ - - return 0; -} diff --git a/libmarkers/tracepoint.h b/libmarkers/tracepoint.h deleted file mode 100644 index 6f14ea0..0000000 --- a/libmarkers/tracepoint.h +++ /dev/null @@ -1,205 +0,0 @@ -#ifndef _LINUX_TRACEPOINT_H -#define _LINUX_TRACEPOINT_H - -/* - * Copyright (C) 2008 Mathieu Desnoyers - * Copyright (C) 2009 Pierre-Marc Fournier - * - * This program is free software; you can redistribute it and/or - * modify it under the terms of the GNU General Public License - * as published by the Free Software Foundation; either version 2 - * of the License, or (at your option) any later version. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * GNU General Public License for more details. - * - * You should have received a copy of the GNU General Public License - * along with this program; if not, write to the Free Software - * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. - * - * Heavily inspired from the Linux Kernel Markers. - * - * Ported to userspace by Pierre-Marc Fournier. - */ - -//#include -//#include -//#include - -#include "immediate.h" -#include "kernelcompat.h" - -struct module; -struct tracepoint; - -struct tracepoint { - const char *name; /* Tracepoint name */ - DEFINE_IMV(char, state); /* State. */ - void **funcs; -} __attribute__((aligned(32))); /* - * Aligned on 32 bytes because it is - * globally visible and gcc happily - * align these on the structure size. - * Keep in sync with vmlinux.lds.h. - */ - -#define TPPROTO(args...) args -#define TPARGS(args...) args - -//ust// #ifdef CONFIG_TRACEPOINTS - -/* - * it_func[0] is never NULL because there is at least one element in the array - * when the array itself is non NULL. - */ -#define __DO_TRACE(tp, proto, args) \ - do { \ - void **it_func; \ - \ - rcu_read_lock_sched_notrace(); \ - it_func = rcu_dereference((tp)->funcs); \ - if (it_func) { \ - do { \ - ((void(*)(proto))(*it_func))(args); \ - } while (*(++it_func)); \ - } \ - rcu_read_unlock_sched_notrace(); \ - } while (0) - -#define __CHECK_TRACE(name, generic, proto, args) \ - do { \ - if (!generic) { \ - if (unlikely(imv_read(__tracepoint_##name.state))) \ - __DO_TRACE(&__tracepoint_##name, \ - TPPROTO(proto), TPARGS(args)); \ - } else { \ - if (unlikely(_imv_read(__tracepoint_##name.state))) \ - __DO_TRACE(&__tracepoint_##name, \ - TPPROTO(proto), TPARGS(args)); \ - } \ - } while (0) - -/* - * Make sure the alignment of the structure in the __tracepoints section will - * not add unwanted padding between the beginning of the section and the - * structure. Force alignment to the same alignment as the section start. - * - * The "generic" argument, passed to the declared __trace_##name inline - * function controls which tracepoint enabling mechanism must be used. - * If generic is true, a variable read is used. - * If generic is false, immediate values are used. - */ -#define DECLARE_TRACE(name, proto, args) \ - extern struct tracepoint __tracepoint_##name; \ - static inline void trace_##name(proto) \ - { \ - __CHECK_TRACE(name, 0, TPPROTO(proto), TPARGS(args)); \ - } \ - static inline void _trace_##name(proto) \ - { \ - __CHECK_TRACE(name, 1, TPPROTO(proto), TPARGS(args)); \ - } \ - static inline int register_trace_##name(void (*probe)(proto)) \ - { \ - return tracepoint_probe_register(#name, (void *)probe); \ - } \ - static inline int unregister_trace_##name(void (*probe)(proto)) \ - { \ - return tracepoint_probe_unregister(#name, (void *)probe);\ - } - -#define DEFINE_TRACE(name) \ - static const char __tpstrtab_##name[] \ - __attribute__((section("__tracepoints_strings"))) = #name; \ - struct tracepoint __tracepoint_##name \ - __attribute__((section("__tracepoints"), aligned(32))) = \ - { __tpstrtab_##name, 0, NULL } - -#define EXPORT_TRACEPOINT_SYMBOL_GPL(name) \ - EXPORT_SYMBOL_GPL(__tracepoint_##name) -#define EXPORT_TRACEPOINT_SYMBOL(name) \ - EXPORT_SYMBOL(__tracepoint_##name) - -extern void tracepoint_update_probe_range(struct tracepoint *begin, - struct tracepoint *end); - -//ust// #else /* !CONFIG_TRACEPOINTS */ -//ust// #define DECLARE_TRACE(name, proto, args) \ -//ust// static inline void trace_##name(proto) \ -//ust// { } \ -//ust// static inline void _trace_##name(proto) \ -//ust// { } \ -//ust// static inline int register_trace_##name(void (*probe)(proto)) \ -//ust// { \ -//ust// return -ENOSYS; \ -//ust// } \ -//ust// static inline int unregister_trace_##name(void (*probe)(proto)) \ -//ust// { \ -//ust// return -ENOSYS; \ -//ust// } -//ust// -//ust// #define DEFINE_TRACE(name) -//ust// #define EXPORT_TRACEPOINT_SYMBOL_GPL(name) -//ust// #define EXPORT_TRACEPOINT_SYMBOL(name) -//ust// -//ust// static inline void tracepoint_update_probe_range(struct tracepoint *begin, -//ust// struct tracepoint *end) -//ust// { } -//ust// #endif /* CONFIG_TRACEPOINTS */ - -/* - * Connect a probe to a tracepoint. - * Internal API, should not be used directly. - */ -extern int tracepoint_probe_register(const char *name, void *probe); - -/* - * Disconnect a probe from a tracepoint. - * Internal API, should not be used directly. - */ -extern int tracepoint_probe_unregister(const char *name, void *probe); - -extern int tracepoint_probe_register_noupdate(const char *name, void *probe); -extern int tracepoint_probe_unregister_noupdate(const char *name, void *probe); -extern void tracepoint_probe_update_all(void); - -struct tracepoint_iter { -//ust// struct module *module; - struct tracepoint_lib *lib; - struct tracepoint *tracepoint; -}; - -extern void tracepoint_iter_start(struct tracepoint_iter *iter); -extern void tracepoint_iter_next(struct tracepoint_iter *iter); -extern void tracepoint_iter_stop(struct tracepoint_iter *iter); -extern void tracepoint_iter_reset(struct tracepoint_iter *iter); -extern int tracepoint_get_iter_range(struct tracepoint **tracepoint, - struct tracepoint *begin, struct tracepoint *end); - -/* - * tracepoint_synchronize_unregister must be called between the last tracepoint - * probe unregistration and the end of module exit to make sure there is no - * caller executing a probe when it is freed. - */ -static inline void tracepoint_synchronize_unregister(void) -{ - synchronize_sched(); -} - -struct tracepoint_lib { - struct tracepoint *tracepoints_start; - int tracepoints_count; - struct list_head list; -}; - -#define TRACEPOINT_LIB \ -extern struct tracepoint __start___tracepoints[] __attribute__((visibility("hidden"))); \ -extern struct tracepoint __stop___tracepoints[] __attribute__((visibility("hidden"))); \ - \ -static void __attribute__((constructor)) __tracepoints__init(void) \ -{ \ - tracepoint_register_lib(__start___tracepoints, (((long)__stop___tracepoints)-((long)__start___tracepoints))/sizeof(struct tracepoint));\ -} -#endif diff --git a/libtracectl/Makefile.am b/libtracectl/Makefile.am deleted file mode 100644 index e13ebe5..0000000 --- a/libtracectl/Makefile.am +++ /dev/null @@ -1,9 +0,0 @@ -lib_LTLIBRARIES = libtracectl.la -libtracectl_la_SOURCES = marker-control.c localerr.h tracectl.c ../libustcomm/ustcomm.c ../libustcomm/ustcomm.h - -INCLUDES = -I$(top_builddir)/share -INCLUDES += -I@URCU_PATH@ -INCLUDES += -I@KCOMPAT_PATH@ -INCLUDES += -I$(top_builddir)/libtracing -INCLUDES += -I$(top_builddir)/libmarkers -INCLUDES += -I$(top_builddir)/libustcomm diff --git a/libtracectl/localerr.h b/libtracectl/localerr.h deleted file mode 100644 index eef0d4f..0000000 --- a/libtracectl/localerr.h +++ /dev/null @@ -1 +0,0 @@ -#include "usterr.h" diff --git a/libtracectl/marker-control.c b/libtracectl/marker-control.c deleted file mode 100644 index 29c43b0..0000000 --- a/libtracectl/marker-control.c +++ /dev/null @@ -1,459 +0,0 @@ -/* - * Copyright (C) 2007 Mathieu Desnoyers - * - * This program is free software; you can redistribute it and/or modify - * it under the terms of the GNU General Public License as published by - * the Free Software Foundation; either version 2 of the License, or - * (at your option) any later version. - * - * This program is distributed in the hope that it will be useful, - * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * GNU General Public License for more details. - * - * You should have received a copy of the GNU General Public License - * along with this program; if not, write to the Free Software - * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. - * - * LTT marker control module over /proc - */ - -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -#include "kernelcompat.h" -#include "list.h" -#include "tracer.h" -#include "localerr.h" - -#define DEFAULT_CHANNEL "cpu" -#define DEFAULT_PROBE "default" - -LIST_HEAD(probes_list); - -/* - * Mutex protecting the probe slab cache. - * Nests inside the traces mutex. - */ -DEFINE_MUTEX(probes_mutex); - -struct ltt_available_probe default_probe = { - .name = "default", - .format = NULL, - .probe_func = ltt_vtrace, - .callbacks[0] = ltt_serialize_data, -}; - -//ust//static struct kmem_cache *markers_loaded_cachep; -static LIST_HEAD(markers_loaded_list); -/* - * List sorted by name strcmp order. - */ -static LIST_HEAD(probes_registered_list); - -//ust// static struct proc_dir_entry *pentry; - -//ust// static struct file_operations ltt_fops; - -static struct ltt_available_probe *get_probe_from_name(const char *pname) -{ - struct ltt_available_probe *iter; - int comparison, found = 0; - - if (!pname) - pname = DEFAULT_PROBE; - list_for_each_entry(iter, &probes_registered_list, node) { - comparison = strcmp(pname, iter->name); - if (!comparison) - found = 1; - if (comparison <= 0) - break; - } - if (found) - return iter; - else - return NULL; -} - -static char *skip_spaces(char *buf) -{ - while (*buf != '\0' && isspace(*buf)) - buf++; - return buf; -} - -static char *skip_nonspaces(char *buf) -{ - while (*buf != '\0' && !isspace(*buf)) - buf++; - return buf; -} - -static void get_marker_string(char *buf, char **start, - char **end) -{ - *start = skip_spaces(buf); - *end = skip_nonspaces(*start); - **end = '\0'; -} - -int ltt_probe_register(struct ltt_available_probe *pdata) -{ - int ret = 0; - int comparison; - struct ltt_available_probe *iter; - - mutex_lock(&probes_mutex); - list_for_each_entry_reverse(iter, &probes_registered_list, node) { - comparison = strcmp(pdata->name, iter->name); - if (!comparison) { - ret = -EBUSY; - goto end; - } else if (comparison > 0) { - /* We belong to the location right after iter. */ - list_add(&pdata->node, &iter->node); - goto end; - } - } - /* Should be added at the head of the list */ - list_add(&pdata->node, &probes_registered_list); -end: - mutex_unlock(&probes_mutex); - return ret; -} -EXPORT_SYMBOL_GPL(ltt_probe_register); - -/* - * Called when a probe does not want to be called anymore. - */ -int ltt_probe_unregister(struct ltt_available_probe *pdata) -{ - int ret = 0; - struct ltt_active_marker *amark, *tmp; - - mutex_lock(&probes_mutex); - list_for_each_entry_safe(amark, tmp, &markers_loaded_list, node) { - if (amark->probe == pdata) { - ret = marker_probe_unregister_private_data( - pdata->probe_func, amark); - if (ret) - goto end; - list_del(&amark->node); - free(amark); - } - } - list_del(&pdata->node); -end: - mutex_unlock(&probes_mutex); - return ret; -} -EXPORT_SYMBOL_GPL(ltt_probe_unregister); - -/* - * Connect marker "mname" to probe "pname". - * Only allow _only_ probe instance to be connected to a marker. - */ -int ltt_marker_connect(const char *channel, const char *mname, - const char *pname) - -{ - int ret; - struct ltt_active_marker *pdata; - struct ltt_available_probe *probe; - - ltt_lock_traces(); - mutex_lock(&probes_mutex); - probe = get_probe_from_name(pname); - if (!probe) { - ret = -ENOENT; - goto end; - } - pdata = marker_get_private_data(channel, mname, probe->probe_func, 0); - if (pdata && !IS_ERR(pdata)) { - ret = -EEXIST; - goto end; - } - pdata = zmalloc(sizeof(struct ltt_active_marker)); - if (!pdata) { - ret = -ENOMEM; - goto end; - } - pdata->probe = probe; - /* - * ID has priority over channel in case of conflict. - */ - ret = marker_probe_register(channel, mname, NULL, - probe->probe_func, pdata); - if (ret) - free(pdata); - else - list_add(&pdata->node, &markers_loaded_list); -end: - mutex_unlock(&probes_mutex); - ltt_unlock_traces(); - return ret; -} -EXPORT_SYMBOL_GPL(ltt_marker_connect); - -/* - * Disconnect marker "mname", probe "pname". - */ -int ltt_marker_disconnect(const char *channel, const char *mname, - const char *pname) -{ - struct ltt_active_marker *pdata; - struct ltt_available_probe *probe; - int ret = 0; - - mutex_lock(&probes_mutex); - probe = get_probe_from_name(pname); - if (!probe) { - ret = -ENOENT; - goto end; - } - pdata = marker_get_private_data(channel, mname, probe->probe_func, 0); - if (IS_ERR(pdata)) { - ret = PTR_ERR(pdata); - goto end; - } else if (!pdata) { - /* - * Not registered by us. - */ - ret = -EPERM; - goto end; - } - ret = marker_probe_unregister(channel, mname, probe->probe_func, pdata); - if (ret) - goto end; - else { - list_del(&pdata->node); - free(pdata); - } -end: - mutex_unlock(&probes_mutex); - return ret; -} -EXPORT_SYMBOL_GPL(ltt_marker_disconnect); - -/* - * function handling proc entry write. - * - * connect []] - * disconnect [] - */ -//ust// static ssize_t ltt_write(struct file *file, const char __user *buffer, -//ust// size_t count, loff_t *offset) -//ust// { -//ust// char *kbuf; -//ust// char *iter, *marker_action, *arg[4]; -//ust// ssize_t ret; -//ust// int i; -//ust// -//ust// if (!count) -//ust// return -EINVAL; -//ust// -//ust// kbuf = vmalloc(count + 1); -//ust// kbuf[count] = '\0'; /* Transform into a string */ -//ust// ret = copy_from_user(kbuf, buffer, count); -//ust// if (ret) { -//ust// ret = -EINVAL; -//ust// goto end; -//ust// } -//ust// get_marker_string(kbuf, &marker_action, &iter); -//ust// if (!marker_action || marker_action == iter) { -//ust// ret = -EINVAL; -//ust// goto end; -//ust// } -//ust// for (i = 0; i < 4; i++) { -//ust// arg[i] = NULL; -//ust// if (iter < kbuf + count) { -//ust// iter++; /* skip the added '\0' */ -//ust// get_marker_string(iter, &arg[i], &iter); -//ust// if (arg[i] == iter) -//ust// arg[i] = NULL; -//ust// } -//ust// } -//ust// -//ust// if (!arg[0] || !arg[1]) { -//ust// ret = -EINVAL; -//ust// goto end; -//ust// } -//ust// -//ust// if (!strcmp(marker_action, "connect")) { -//ust// ret = ltt_marker_connect(arg[0], arg[1], arg[2]); -//ust// if (ret) -//ust// goto end; -//ust// } else if (!strcmp(marker_action, "disconnect")) { -//ust// ret = ltt_marker_disconnect(arg[0], arg[1], arg[2]); -//ust// if (ret) -//ust// goto end; -//ust// } -//ust// ret = count; -//ust// end: -//ust// vfree(kbuf); -//ust// return ret; -//ust// } -//ust// -//ust// static void *s_next(struct seq_file *m, void *p, loff_t *pos) -//ust// { -//ust// struct marker_iter *iter = m->private; -//ust// -//ust// marker_iter_next(iter); -//ust// if (!iter->marker) { -//ust// /* -//ust// * Setting the iter module to -1UL will make sure -//ust// * that no module can possibly hold the current marker. -//ust// */ -//ust// iter->module = (void *)-1UL; -//ust// return NULL; -//ust// } -//ust// return iter->marker; -//ust// } -//ust// -//ust// static void *s_start(struct seq_file *m, loff_t *pos) -//ust// { -//ust// struct marker_iter *iter = m->private; -//ust// -//ust// if (!*pos) -//ust// marker_iter_reset(iter); -//ust// marker_iter_start(iter); -//ust// if (!iter->marker) { -//ust// /* -//ust// * Setting the iter module to -1UL will make sure -//ust// * that no module can possibly hold the current marker. -//ust// */ -//ust// iter->module = (void *)-1UL; -//ust// return NULL; -//ust// } -//ust// return iter->marker; -//ust// } -//ust// -//ust// static void s_stop(struct seq_file *m, void *p) -//ust// { -//ust// marker_iter_stop(m->private); -//ust// } -//ust// -//ust// static int s_show(struct seq_file *m, void *p) -//ust// { -//ust// struct marker_iter *iter = m->private; -//ust// -//ust// seq_printf(m, "channel: %s marker: %s format: \"%s\" state: %d " -//ust// "event_id: %hu call: 0x%p probe %s : 0x%p\n", -//ust// iter->marker->channel, -//ust// iter->marker->name, iter->marker->format, -//ust// _imv_read(iter->marker->state), -//ust// iter->marker->event_id, -//ust// iter->marker->call, -//ust// iter->marker->ptype ? "multi" : "single", -//ust// iter->marker->ptype ? -//ust// (void*)iter->marker->multi : (void*)iter->marker->single.func); -//ust// return 0; -//ust// } -//ust// -//ust// static const struct seq_operations ltt_seq_op = { -//ust// .start = s_start, -//ust// .next = s_next, -//ust// .stop = s_stop, -//ust// .show = s_show, -//ust// }; -//ust// -//ust// static int ltt_open(struct inode *inode, struct file *file) -//ust// { -//ust// /* -//ust// * Iterator kept in m->private. -//ust// * Restart iteration on all modules between reads because we do not lock -//ust// * the module mutex between those. -//ust// */ -//ust// int ret; -//ust// struct marker_iter *iter; -//ust// -//ust// iter = kzalloc(sizeof(*iter), GFP_KERNEL); -//ust// if (!iter) -//ust// return -ENOMEM; -//ust// -//ust// ret = seq_open(file, <t_seq_op); -//ust// if (ret == 0) -//ust// ((struct seq_file *)file->private_data)->private = iter; -//ust// else -//ust// kfree(iter); -//ust// return ret; -//ust// } -//ust// -//ust// static struct file_operations ltt_fops = { -//ust// .write = ltt_write, -//ust// .open = ltt_open, -//ust// .read = seq_read, -//ust// .llseek = seq_lseek, -//ust// .release = seq_release_private, -//ust// }; - -static void disconnect_all_markers(void) -{ - struct ltt_active_marker *pdata, *tmp; - - list_for_each_entry_safe(pdata, tmp, &markers_loaded_list, node) { - marker_probe_unregister_private_data(pdata->probe->probe_func, - pdata); - list_del(&pdata->node); - free(pdata); - } -} - -static char initialized = 0; - -void __attribute__((constructor)) init_marker_control(void) -{ - if(!initialized) { - int ret; - -//ust// pentry = create_proc_entry("ltt", S_IRUSR|S_IWUSR, NULL); -//ust// if (!pentry) -//ust// return -EBUSY; -//ust// markers_loaded_cachep = KMEM_CACHE(ltt_active_marker, 0); - - ret = ltt_probe_register(&default_probe); - BUG_ON(ret); - ret = ltt_marker_connect("metadata", "core_marker_format", - DEFAULT_PROBE); - BUG_ON(ret); - ret = ltt_marker_connect("metadata", "core_marker_id", DEFAULT_PROBE); - BUG_ON(ret); -//ust// pentry->proc_fops = <t_fops; - - initialized = 1; - } -} -//ust// module_init(marker_control_init); - -static void __exit marker_control_exit(void) -{ - int ret; - -//ust// remove_proc_entry("ltt", NULL); - ret = ltt_marker_disconnect("metadata", "core_marker_format", - DEFAULT_PROBE); - BUG_ON(ret); - ret = ltt_marker_disconnect("metadata", "core_marker_id", - DEFAULT_PROBE); - BUG_ON(ret); - ret = ltt_probe_unregister(&default_probe); - BUG_ON(ret); - disconnect_all_markers(); -//ust// kmem_cache_destroy(markers_loaded_cachep); -//ust// marker_synchronize_unregister(); -} -//ust// module_exit(marker_control_exit); - -//ust// MODULE_LICENSE("GPL"); -//ust// MODULE_AUTHOR("Mathieu Desnoyers"); -//ust// MODULE_DESCRIPTION("Linux Trace Toolkit Marker Control"); diff --git a/libtracectl/marker-control.h b/libtracectl/marker-control.h deleted file mode 100644 index 0b602db..0000000 --- a/libtracectl/marker-control.h +++ /dev/null @@ -1,7 +0,0 @@ -#ifndef MARKER_CONTROL_H -#define MARKER_CONTROL_H - -int marker_control_init(void); -int ltt_probe_register(struct ltt_available_probe *pdata); - -#endif /* MARKER_CONTROL_H */ diff --git a/libtracectl/tracectl.c b/libtracectl/tracectl.c deleted file mode 100644 index 2101d5c..0000000 --- a/libtracectl/tracectl.c +++ /dev/null @@ -1,820 +0,0 @@ -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#include "marker.h" -#include "tracer.h" -#include "localerr.h" -#include "ustcomm.h" -#include "relay.h" /* FIXME: remove */ - -//#define USE_CLONE - -#define USTSIGNAL SIGIO - -#define MAX_MSG_SIZE (100) -#define MSG_NOTIF 1 -#define MSG_REGISTER_NOTIF 2 - -char consumer_stack[10000]; - -struct list_head blocked_consumers = LIST_HEAD_INIT(blocked_consumers); - -static struct ustcomm_app ustcomm_app; - -struct tracecmd { /* no padding */ - uint32_t size; - uint16_t command; -}; - -//struct listener_arg { -// int pipe_fd; -//}; - -struct trctl_msg { - /* size: the size of all the fields except size itself */ - uint32_t size; - uint16_t type; - /* Only the necessary part of the payload is transferred. It - * may even be none of it. - */ - char payload[94]; -}; - -struct consumer_channel { - int fd; - struct ltt_channel_struct *chan; -}; - -struct blocked_consumer { - int fd_consumer; - int fd_producer; - int tmp_poll_idx; - - /* args to ustcomm_send_reply */ - struct ustcomm_server server; - struct ustcomm_source src; - - /* args to ltt_do_get_subbuf */ - struct rchan_buf *rbuf; - struct ltt_channel_buf_struct *lttbuf; - - struct list_head list; -}; - -static void print_markers(void) -{ - struct marker_iter iter; - - lock_markers(); - marker_iter_reset(&iter); - marker_iter_start(&iter); - - while(iter.marker) { - fprintf(stderr, "marker: %s_%s \"%s\"\n", iter.marker->channel, iter.marker->name, iter.marker->format); - marker_iter_next(&iter); - } - unlock_markers(); -} - -void do_command(struct tracecmd *cmd) -{ -} - -void receive_commands() -{ -} - -int fd_notif = -1; -void notif_cb(void) -{ - int result; - struct trctl_msg msg; - - /* FIXME: fd_notif should probably be protected by a spinlock */ - - if(fd_notif == -1) - return; - - msg.type = MSG_NOTIF; - msg.size = sizeof(msg.type); - - /* FIXME: don't block here */ - result = write(fd_notif, &msg, msg.size+sizeof(msg.size)); - if(result == -1) { - PERROR("write"); - return; - } -} - -static int inform_consumer_daemon(void) -{ - ustcomm_request_consumer(getpid(), "metadata"); - ustcomm_request_consumer(getpid(), "ust"); -} - -void process_blocked_consumers(void) -{ - int n_fds = 0; - struct pollfd *fds; - struct blocked_consumer *bc; - int idx = 0; - char inbuf; - int result; - - list_for_each_entry(bc, &blocked_consumers, list) { - n_fds++; - } - - fds = (struct pollfd *) malloc(n_fds * sizeof(struct pollfd)); - if(fds == NULL) { - ERR("malloc returned NULL"); - return; - } - - list_for_each_entry(bc, &blocked_consumers, list) { - fds[idx].fd = bc->fd_producer; - fds[idx].events = POLLIN; - bc->tmp_poll_idx = idx; - idx++; - } - - result = poll(fds, n_fds, 0); - if(result == -1) { - PERROR("poll"); - return -1; - } - - list_for_each_entry(bc, &blocked_consumers, list) { - if(fds[bc->tmp_poll_idx].revents) { - long consumed_old = 0; - char *reply; - - result = read(bc->fd_producer, &inbuf, 1); - if(result == -1) { - PERROR("read"); - continue; - } - if(result == 0) { - DBG("PRODUCER END"); - - close(bc->fd_producer); - - __list_del(bc->list.prev, bc->list.next); - - result = ustcomm_send_reply(&bc->server, "END", &bc->src); - if(result < 0) { - ERR("ustcomm_send_reply failed"); - continue; - } - - continue; - } - - result = ltt_do_get_subbuf(bc->rbuf, bc->lttbuf, &consumed_old); - if(result == -EAGAIN) { - WARN("missed buffer?"); - continue; - } - else if(result < 0) { - DBG("ltt_do_get_subbuf: error: %s", strerror(-result)); - } - asprintf(&reply, "%s %ld", "OK", consumed_old); - result = ustcomm_send_reply(&bc->server, reply, &bc->src); - if(result < 0) { - ERR("ustcomm_send_reply failed"); - free(reply); - continue; - } - free(reply); - - __list_del(bc->list.prev, bc->list.next); - } - } - -} - -int listener_main(void *p) -{ - int result; - - DBG("LISTENER"); - - for(;;) { - uint32_t size; - struct sockaddr_un addr; - socklen_t addrlen = sizeof(addr); - char trace_name[] = "auto"; - char trace_type[] = "ustrelay"; - char *recvbuf; - int len; - struct ustcomm_source src; - - process_blocked_consumers(); - - result = ustcomm_app_recv_message(&ustcomm_app, &recvbuf, &src, 5); - if(result < 0) { - WARN("error in ustcomm_app_recv_message"); - continue; - } - else if(result == 0) { - /* no message */ - continue; - } - - DBG("received a message! it's: %s\n", recvbuf); - len = strlen(recvbuf); - - if(!strcmp(recvbuf, "print_markers")) { - print_markers(); - } - else if(!strcmp(recvbuf, "trace_setup")) { - DBG("trace setup"); - - result = ltt_trace_setup(trace_name); - if(result < 0) { - ERR("ltt_trace_setup failed"); - return; - } - - result = ltt_trace_set_type(trace_name, trace_type); - if(result < 0) { - ERR("ltt_trace_set_type failed"); - return; - } - } - else if(!strcmp(recvbuf, "trace_alloc")) { - DBG("trace alloc"); - - result = ltt_trace_alloc(trace_name); - if(result < 0) { - ERR("ltt_trace_alloc failed"); - return; - } - } - else if(!strcmp(recvbuf, "trace_start")) { - DBG("trace start"); - - result = ltt_trace_start(trace_name); - if(result < 0) { - ERR("ltt_trace_start failed"); - continue; - } - } - else if(!strcmp(recvbuf, "trace_stop")) { - DBG("trace stop"); - - result = ltt_trace_stop(trace_name); - if(result < 0) { - ERR("ltt_trace_stop failed"); - return; - } - } - else if(!strcmp(recvbuf, "trace_destroy")) { - - DBG("trace destroy"); - - result = ltt_trace_destroy(trace_name); - if(result < 0) { - ERR("ltt_trace_destroy failed"); - return; - } - } - else if(nth_token_is(recvbuf, "get_shmid", 0) == 1) { - struct ltt_trace_struct *trace; - char trace_name[] = "auto"; - int i; - char *channel_name; - - DBG("get_shmid"); - - channel_name = nth_token(recvbuf, 1); - if(channel_name == NULL) { - ERR("get_shmid: cannot parse channel"); - goto next_cmd; - } - - ltt_lock_traces(); - trace = _ltt_trace_find(trace_name); - ltt_unlock_traces(); - - if(trace == NULL) { - CPRINTF("cannot find trace!"); - return 1; - } - - for(i=0; inr_channels; i++) { - struct rchan *rchan = trace->channels[i].trans_channel_data; - struct rchan_buf *rbuf = rchan->buf; - struct ltt_channel_struct *ltt_channel = (struct ltt_channel_struct *)rchan->private_data; - struct ltt_channel_buf_struct *ltt_buf = ltt_channel->buf; - - if(!strcmp(trace->channels[i].channel_name, channel_name)) { - char *reply; - - DBG("the shmid for the requested channel is %d", rbuf->shmid); - DBG("the shmid for its buffer structure is %d", ltt_channel->buf_shmid); - asprintf(&reply, "%d %d", rbuf->shmid, ltt_channel->buf_shmid); - - result = ustcomm_send_reply(&ustcomm_app.server, reply, &src); - if(result) { - ERR("listener: get_shmid: ustcomm_send_reply failed"); - goto next_cmd; - } - - free(reply); - - break; - } - } - } - else if(nth_token_is(recvbuf, "get_n_subbufs", 0) == 1) { - struct ltt_trace_struct *trace; - char trace_name[] = "auto"; - int i; - char *channel_name; - - DBG("get_n_subbufs"); - - channel_name = nth_token(recvbuf, 1); - if(channel_name == NULL) { - ERR("get_n_subbufs: cannot parse channel"); - goto next_cmd; - } - - ltt_lock_traces(); - trace = _ltt_trace_find(trace_name); - ltt_unlock_traces(); - - if(trace == NULL) { - CPRINTF("cannot find trace!"); - return 1; - } - - for(i=0; inr_channels; i++) { - struct rchan *rchan = trace->channels[i].trans_channel_data; - - if(!strcmp(trace->channels[i].channel_name, channel_name)) { - char *reply; - - DBG("the n_subbufs for the requested channel is %d", rchan->n_subbufs); - asprintf(&reply, "%d", rchan->n_subbufs); - - result = ustcomm_send_reply(&ustcomm_app.server, reply, &src); - if(result) { - ERR("listener: get_n_subbufs: ustcomm_send_reply failed"); - goto next_cmd; - } - - free(reply); - - break; - } - } - } - else if(nth_token_is(recvbuf, "get_subbuf_size", 0) == 1) { - struct ltt_trace_struct *trace; - char trace_name[] = "auto"; - int i; - char *channel_name; - - DBG("get_subbuf_size"); - - channel_name = nth_token(recvbuf, 1); - if(channel_name == NULL) { - ERR("get_subbuf_size: cannot parse channel"); - goto next_cmd; - } - - ltt_lock_traces(); - trace = _ltt_trace_find(trace_name); - ltt_unlock_traces(); - - if(trace == NULL) { - CPRINTF("cannot find trace!"); - return 1; - } - - for(i=0; inr_channels; i++) { - struct rchan *rchan = trace->channels[i].trans_channel_data; - - if(!strcmp(trace->channels[i].channel_name, channel_name)) { - char *reply; - - DBG("the subbuf_size for the requested channel is %d", rchan->subbuf_size); - asprintf(&reply, "%d", rchan->subbuf_size); - - result = ustcomm_send_reply(&ustcomm_app.server, reply, &src); - if(result) { - ERR("listener: get_subbuf_size: ustcomm_send_reply failed"); - goto next_cmd; - } - - free(reply); - - break; - } - } - } - else if(nth_token_is(recvbuf, "load_probe_lib", 0) == 1) { - char *libfile; - - libfile = nth_token(recvbuf, 1); - - DBG("load_probe_lib loading %s", libfile); - } - else if(nth_token_is(recvbuf, "get_subbuffer", 0) == 1) { - struct ltt_trace_struct *trace; - char trace_name[] = "auto"; - int i; - char *channel_name; - - DBG("get_subbuf"); - - channel_name = nth_token(recvbuf, 1); - if(channel_name == NULL) { - ERR("get_subbuf: cannot parse channel"); - goto next_cmd; - } - - ltt_lock_traces(); - trace = _ltt_trace_find(trace_name); - ltt_unlock_traces(); - - if(trace == NULL) { - CPRINTF("cannot find trace!"); - return 1; - } - - for(i=0; inr_channels; i++) { - struct rchan *rchan = trace->channels[i].trans_channel_data; - - if(!strcmp(trace->channels[i].channel_name, channel_name)) { - struct rchan_buf *rbuf = rchan->buf; - struct ltt_channel_buf_struct *lttbuf = trace->channels[i].buf; - char *reply; - long consumed_old=0; - int fd; - struct blocked_consumer *bc; - - bc = (struct blocked_consumer *) malloc(sizeof(struct blocked_consumer)); - if(bc == NULL) { - ERR("malloc returned NULL"); - goto next_cmd; - } - bc->fd_consumer = src.fd; - bc->fd_producer = lttbuf->data_ready_fd_read; - bc->rbuf = rbuf; - bc->lttbuf = lttbuf; - bc->src = src; - bc->server = ustcomm_app.server; - - list_add(&bc->list, &blocked_consumers); - - break; - } - } - } - else if(nth_token_is(recvbuf, "put_subbuffer", 0) == 1) { - struct ltt_trace_struct *trace; - char trace_name[] = "auto"; - int i; - char *channel_name; - long consumed_old; - char *consumed_old_str; - char *endptr; - - DBG("put_subbuf"); - - channel_name = strdup_malloc(nth_token(recvbuf, 1)); - if(channel_name == NULL) { - ERR("put_subbuf_size: cannot parse channel"); - goto next_cmd; - } - - consumed_old_str = strdup_malloc(nth_token(recvbuf, 2)); - if(consumed_old_str == NULL) { - ERR("put_subbuf: cannot parse consumed_old"); - goto next_cmd; - } - consumed_old = strtol(consumed_old_str, &endptr, 10); - if(*endptr != '\0') { - ERR("put_subbuf: invalid value for consumed_old"); - goto next_cmd; - } - - ltt_lock_traces(); - trace = _ltt_trace_find(trace_name); - ltt_unlock_traces(); - - if(trace == NULL) { - CPRINTF("cannot find trace!"); - return 1; - } - - for(i=0; inr_channels; i++) { - struct rchan *rchan = trace->channels[i].trans_channel_data; - - if(!strcmp(trace->channels[i].channel_name, channel_name)) { - struct rchan_buf *rbuf = rchan->buf; - struct ltt_channel_buf_struct *lttbuf = trace->channels[i].buf; - char *reply; - long consumed_old=0; - - result = ltt_do_put_subbuf(rbuf, lttbuf, consumed_old); - if(result < 0) { - WARN("ltt_do_put_subbuf: error"); - } - else { - DBG("ltt_do_put_subbuf: success"); - } - asprintf(&reply, "%s", "OK", consumed_old); - - result = ustcomm_send_reply(&ustcomm_app.server, reply, &src); - if(result) { - ERR("listener: put_subbuf: ustcomm_send_reply failed"); - goto next_cmd; - } - - free(reply); - - break; - } - } - - free(channel_name); - free(consumed_old_str); - } -// else if(nth_token_is(recvbuf, "get_notifications", 0) == 1) { -// struct ltt_trace_struct *trace; -// char trace_name[] = "auto"; -// int i; -// char *channel_name; -// -// DBG("get_notifications"); -// -// channel_name = strdup_malloc(nth_token(recvbuf, 1)); -// if(channel_name == NULL) { -// ERR("put_subbuf_size: cannot parse channel"); -// goto next_cmd; -// } -// -// ltt_lock_traces(); -// trace = _ltt_trace_find(trace_name); -// ltt_unlock_traces(); -// -// if(trace == NULL) { -// CPRINTF("cannot find trace!"); -// return 1; -// } -// -// for(i=0; inr_channels; i++) { -// struct rchan *rchan = trace->channels[i].trans_channel_data; -// int fd; -// -// if(!strcmp(trace->channels[i].channel_name, channel_name)) { -// struct rchan_buf *rbuf = rchan->buf; -// struct ltt_channel_buf_struct *lttbuf = trace->channels[i].buf; -// -// result = fd = ustcomm_app_detach_client(&ustcomm_app, &src); -// if(result == -1) { -// ERR("ustcomm_app_detach_client failed"); -// goto next_cmd; -// } -// -// lttbuf->wake_consumer_arg = (void *) fd; -// -// smp_wmb(); -// -// lttbuf->call_wake_consumer = 1; -// -// break; -// } -// } -// -// free(channel_name); -// } - else { - ERR("unable to parse message: %s", recvbuf); - } - - next_cmd: - free(recvbuf); - } -} - -static char listener_stack[16384]; - -void create_listener(void) -{ - int result; - static char listener_stack[16384]; - //char *listener_stack = malloc(16384); - -#ifdef USE_CLONE - result = clone(listener_main, listener_stack+sizeof(listener_stack)-1, CLONE_FS | CLONE_FILES | CLONE_VM | CLONE_SIGHAND | CLONE_THREAD, NULL); - if(result == -1) { - perror("clone"); - } -#else - pthread_t thread; - - pthread_create(&thread, NULL, listener_main, NULL); -#endif -} - -/* The signal handler itself. Signals must be setup so there cannot be - nested signals. */ - -void sighandler(int sig) -{ - static char have_listener = 0; - DBG("sighandler"); - - if(!have_listener) { - create_listener(); - have_listener = 1; - } -} - -/* Called by the app signal handler to chain it to us. */ - -void chain_signal(void) -{ - sighandler(USTSIGNAL); -} - -static int init_socket(void) -{ - return ustcomm_init_app(getpid(), &ustcomm_app); -} - -static void destroy_socket(void) -{ -// int result; -// -// if(mysocketfile[0] == '\0') -// return; -// -// result = unlink(mysocketfile); -// if(result == -1) { -// PERROR("unlink"); -// } -} - -static int init_signal_handler(void) -{ - /* Attempt to handler SIGIO. If the main program wants to - * handle it, fine, it'll override us. They it'll have to - * use the chaining function. - */ - - int result; - struct sigaction act; - - result = sigemptyset(&act.sa_mask); - if(result == -1) { - PERROR("sigemptyset"); - return -1; - } - - act.sa_handler = sighandler; - act.sa_flags = SA_RESTART; - - /* Only defer ourselves. Also, try to restart interrupted - * syscalls to disturb the traced program as little as possible. - */ - result = sigaction(SIGIO, &act, NULL); - if(result == -1) { - PERROR("sigaction"); - return -1; - } - - return 0; -} - -static void auto_probe_connect(struct marker *m) -{ - int result; - - result = ltt_marker_connect(m->channel, m->name, "default"); - if(result) - ERR("ltt_marker_connect"); - - DBG("just auto connected marker %s %s to probe default", m->channel, m->name); -} - -static void __attribute__((constructor(101))) init0() -{ - DBG("UST_AUTOPROBE constructor"); - if(getenv("UST_AUTOPROBE")) { - marker_set_new_marker_cb(auto_probe_connect); - } -} - -static void fini(void); - -static void __attribute__((constructor(1000))) init() -{ - int result; - - DBG("UST_TRACE constructor"); - - /* Must create socket before signal handler to prevent races. - */ - result = init_socket(); - if(result == -1) { - ERR("init_socket error"); - return; - } - result = init_signal_handler(); - if(result == -1) { - ERR("init_signal_handler error"); - return; - } - - if(getenv("UST_TRACE")) { - char trace_name[] = "auto"; - char trace_type[] = "ustrelay"; - - DBG("starting early tracing"); - - /* Ensure marker control is initialized */ - init_marker_control(); - - /* Ensure relay is initialized */ - init_ustrelay_transport(); - - /* Ensure markers are initialized */ - init_markers(); - - /* In case. */ - ltt_channels_register("ust"); - - result = ltt_trace_setup(trace_name); - if(result < 0) { - ERR("ltt_trace_setup failed"); - return; - } - - result = ltt_trace_set_type(trace_name, trace_type); - if(result < 0) { - ERR("ltt_trace_set_type failed"); - return; - } - - result = ltt_trace_alloc(trace_name); - if(result < 0) { - ERR("ltt_trace_alloc failed"); - return; - } - - result = ltt_trace_start(trace_name); - if(result < 0) { - ERR("ltt_trace_start failed"); - return; - } - //start_consumer(); - inform_consumer_daemon(); - } - - - return; - - /* should decrementally destroy stuff if error */ - -} - -/* This is only called if we terminate normally, not with an unhandled signal, - * so we cannot rely on it. */ - -static void __attribute__((destructor)) fini() -{ - int result; - - /* if trace running, finish it */ - - DBG("destructor stopping traces"); - - result = ltt_trace_stop("auto"); - if(result == -1) { - ERR("ltt_trace_stop error"); - } - - result = ltt_trace_destroy("auto"); - if(result == -1) { - ERR("ltt_trace_destroy error"); - } - - /* FIXME: wait for the consumer to be done */ - //DBG("waiting 5 sec for consume"); - //sleep(5); - - destroy_socket(); -} diff --git a/libtracing/Makefile.am b/libtracing/Makefile.am deleted file mode 100644 index 8b1c36c..0000000 --- a/libtracing/Makefile.am +++ /dev/null @@ -1,7 +0,0 @@ -lib_LTLIBRARIES = libtracing.la -libtracing_la_SOURCES = channels.c channels.h relay.c relay.h serialize.c tracer.c tracer.h tracercore.c tracercore.h - -INCLUDES = -I$(top_builddir)/share -INCLUDES += -I@URCU_PATH@ -INCLUDES += -I@KCOMPAT_PATH@ -INCLUDES += -I$(top_builddir)/libmarkers diff --git a/libtracing/channels.c b/libtracing/channels.c deleted file mode 100644 index a8df3c4..0000000 --- a/libtracing/channels.c +++ /dev/null @@ -1,348 +0,0 @@ -/* - * ltt/ltt-channels.c - * - * (C) Copyright 2008 - Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) - * - * LTTng channel management. - * - * Author: - * Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) - */ - -//ust// #include -//ust// #include -//ust// #include -//ust// #include - -#include "kernelcompat.h" -#include "channels.h" -#include "usterr.h" - -/* - * ltt_channel_mutex may be nested inside the LTT trace mutex. - * ltt_channel_mutex mutex may be nested inside markers mutex. - */ -static DEFINE_MUTEX(ltt_channel_mutex); -static LIST_HEAD(ltt_channels); -/* - * Index of next channel in array. Makes sure that as long as a trace channel is - * allocated, no array index will be re-used when a channel is freed and then - * another channel is allocated. This index is cleared and the array indexeds - * get reassigned when the index_kref goes back to 0, which indicates that no - * more trace channels are allocated. - */ -static unsigned int free_index; -static struct kref index_kref; /* Keeps track of allocated trace channels */ - -static struct ltt_channel_setting *lookup_channel(const char *name) -{ - struct ltt_channel_setting *iter; - - list_for_each_entry(iter, <t_channels, list) - if (strcmp(name, iter->name) == 0) - return iter; - return NULL; -} - -/* - * Must be called when channel refcount falls to 0 _and_ also when the last - * trace is freed. This function is responsible for compacting the channel and - * event IDs when no users are active. - * - * Called with lock_markers() and channels mutex held. - */ -static void release_channel_setting(struct kref *kref) -{ - struct ltt_channel_setting *setting = container_of(kref, - struct ltt_channel_setting, kref); - struct ltt_channel_setting *iter; - - if (atomic_read(&index_kref.refcount) == 0 - && atomic_read(&setting->kref.refcount) == 0) { - list_del(&setting->list); - kfree(setting); - - free_index = 0; - list_for_each_entry(iter, <t_channels, list) { - iter->index = free_index++; - iter->free_event_id = 0; - } -//ust// markers_compact_event_ids(); - } -} - -/* - * Perform channel index compaction when the last trace channel is freed. - * - * Called with lock_markers() and channels mutex held. - */ -static void release_trace_channel(struct kref *kref) -{ - struct ltt_channel_setting *iter, *n; - - list_for_each_entry_safe(iter, n, <t_channels, list) - release_channel_setting(&iter->kref); -} - -/** - * ltt_channels_register - Register a trace channel. - * @name: channel name - * - * Uses refcounting. - */ -int ltt_channels_register(const char *name) -{ - struct ltt_channel_setting *setting; - int ret = 0; - - mutex_lock(<t_channel_mutex); - setting = lookup_channel(name); - if (setting) { - if (atomic_read(&setting->kref.refcount) == 0) - goto init_kref; - else { - kref_get(&setting->kref); - goto end; - } - } - setting = kzalloc(sizeof(*setting), GFP_KERNEL); - if (!setting) { - ret = -ENOMEM; - goto end; - } - list_add(&setting->list, <t_channels); - strncpy(setting->name, name, PATH_MAX-1); - setting->index = free_index++; -init_kref: - kref_init(&setting->kref); -end: - mutex_unlock(<t_channel_mutex); - return ret; -} -//ust// EXPORT_SYMBOL_GPL(ltt_channels_register); - -/** - * ltt_channels_unregister - Unregister a trace channel. - * @name: channel name - * - * Must be called with markers mutex held. - */ -int ltt_channels_unregister(const char *name) -{ - struct ltt_channel_setting *setting; - int ret = 0; - - mutex_lock(<t_channel_mutex); - setting = lookup_channel(name); - if (!setting || atomic_read(&setting->kref.refcount) == 0) { - ret = -ENOENT; - goto end; - } - kref_put(&setting->kref, release_channel_setting); -end: - mutex_unlock(<t_channel_mutex); - return ret; -} -//ust// EXPORT_SYMBOL_GPL(ltt_channels_unregister); - -/** - * ltt_channels_set_default - Set channel default behavior. - * @name: default channel name - * @subbuf_size: size of the subbuffers - * @subbuf_cnt: number of subbuffers - */ -int ltt_channels_set_default(const char *name, - unsigned int subbuf_size, - unsigned int subbuf_cnt) -{ - struct ltt_channel_setting *setting; - int ret = 0; - - mutex_lock(<t_channel_mutex); - setting = lookup_channel(name); - if (!setting || atomic_read(&setting->kref.refcount) == 0) { - ret = -ENOENT; - goto end; - } - setting->subbuf_size = subbuf_size; - setting->subbuf_cnt = subbuf_cnt; -end: - mutex_unlock(<t_channel_mutex); - return ret; -} -//ust// EXPORT_SYMBOL_GPL(ltt_channels_set_default); - -/** - * ltt_channels_get_name_from_index - get channel name from channel index - * @index: channel index - * - * Allows to lookup the channel name given its index. Done to keep the name - * information outside of each trace channel instance. - */ -const char *ltt_channels_get_name_from_index(unsigned int index) -{ - struct ltt_channel_setting *iter; - - list_for_each_entry(iter, <t_channels, list) - if (iter->index == index && atomic_read(&iter->kref.refcount)) - return iter->name; - return NULL; -} -//ust// EXPORT_SYMBOL_GPL(ltt_channels_get_name_from_index); - -static struct ltt_channel_setting * -ltt_channels_get_setting_from_name(const char *name) -{ - struct ltt_channel_setting *iter; - - list_for_each_entry(iter, <t_channels, list) - if (!strcmp(iter->name, name) - && atomic_read(&iter->kref.refcount)) - return iter; - return NULL; -} - -/** - * ltt_channels_get_index_from_name - get channel index from channel name - * @name: channel name - * - * Allows to lookup the channel index given its name. Done to keep the name - * information outside of each trace channel instance. - * Returns -1 if not found. - */ -int ltt_channels_get_index_from_name(const char *name) -{ - struct ltt_channel_setting *setting; - - setting = ltt_channels_get_setting_from_name(name); - if (setting) - return setting->index; - else - return -1; -} -//ust// EXPORT_SYMBOL_GPL(ltt_channels_get_index_from_name); - -/** - * ltt_channels_trace_alloc - Allocate channel structures for a trace - * @subbuf_size: subbuffer size. 0 uses default. - * @subbuf_cnt: number of subbuffers per per-cpu buffers. 0 uses default. - * @flags: Default channel flags - * - * Use the current channel list to allocate the channels for a trace. - * Called with trace lock held. Does not perform the trace buffer allocation, - * because we must let the user overwrite specific channel sizes. - */ -struct ltt_channel_struct *ltt_channels_trace_alloc(unsigned int *nr_channels, - int overwrite, - int active) -{ - struct ltt_channel_struct *channel = NULL; - struct ltt_channel_setting *iter; - - mutex_lock(<t_channel_mutex); - if (!free_index) { - WARN("ltt_channels_trace_alloc: no free_index; are there any probes connected?"); - goto end; - } - if (!atomic_read(&index_kref.refcount)) - kref_init(&index_kref); - else - kref_get(&index_kref); - *nr_channels = free_index; - channel = kzalloc(sizeof(struct ltt_channel_struct) * free_index, - GFP_KERNEL); - if (!channel) { - WARN("ltt_channel_struct: channel null after alloc"); - goto end; - } - list_for_each_entry(iter, <t_channels, list) { - if (!atomic_read(&iter->kref.refcount)) - continue; - channel[iter->index].subbuf_size = iter->subbuf_size; - channel[iter->index].subbuf_cnt = iter->subbuf_cnt; - channel[iter->index].overwrite = overwrite; - channel[iter->index].active = active; - channel[iter->index].channel_name = iter->name; - } -end: - mutex_unlock(<t_channel_mutex); - return channel; -} -//ust// EXPORT_SYMBOL_GPL(ltt_channels_trace_alloc); - -/** - * ltt_channels_trace_free - Free one trace's channels - * @channels: channels to free - * - * Called with trace lock held. The actual channel buffers must be freed before - * this function is called. - */ -void ltt_channels_trace_free(struct ltt_channel_struct *channels) -{ - lock_markers(); - mutex_lock(<t_channel_mutex); - kfree(channels); - kref_put(&index_kref, release_trace_channel); - mutex_unlock(<t_channel_mutex); - unlock_markers(); -} -//ust// EXPORT_SYMBOL_GPL(ltt_channels_trace_free); - -/** - * _ltt_channels_get_event_id - get next event ID for a marker - * @channel: channel name - * @name: event name - * - * Returns a unique event ID (for this channel) or < 0 on error. - * Must be called with channels mutex held. - */ -int _ltt_channels_get_event_id(const char *channel, const char *name) -{ - struct ltt_channel_setting *setting; - int ret; - - setting = ltt_channels_get_setting_from_name(channel); - if (!setting) { - ret = -ENOENT; - goto end; - } - if (strcmp(channel, "metadata") == 0) { - if (strcmp(name, "core_marker_id") == 0) - ret = 0; - else if (strcmp(name, "core_marker_format") == 0) - ret = 1; - else if (strcmp(name, "testev") == 0) - ret = 2; - else - ret = -ENOENT; - goto end; - } - if (setting->free_event_id == EVENTS_PER_CHANNEL - 1) { - ret = -ENOSPC; - goto end; - } - ret = setting->free_event_id++; -end: - return ret; -} - -/** - * ltt_channels_get_event_id - get next event ID for a marker - * @channel: channel name - * @name: event name - * - * Returns a unique event ID (for this channel) or < 0 on error. - */ -int ltt_channels_get_event_id(const char *channel, const char *name) -{ - int ret; - - mutex_lock(<t_channel_mutex); - ret = _ltt_channels_get_event_id(channel, name); - mutex_unlock(<t_channel_mutex); - return ret; -} - -//ust// MODULE_LICENSE("GPL"); -//ust// MODULE_AUTHOR("Mathieu Desnoyers"); -//ust// MODULE_DESCRIPTION("Linux Trace Toolkit Next Generation Channel Management"); diff --git a/libtracing/channels.h b/libtracing/channels.h deleted file mode 100644 index dedc1f8..0000000 --- a/libtracing/channels.h +++ /dev/null @@ -1,84 +0,0 @@ -#ifndef _LTT_CHANNELS_H -#define _LTT_CHANNELS_H - -/* - * Copyright (C) 2008 Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) - * - * Dynamic tracer channel allocation. - */ - -#include -//ust// #include -//ust// #include -#include - -#include "kernelcompat.h" -#include "kref.h" -#include "list.h" - -#define EVENTS_PER_CHANNEL 65536 - -struct ltt_trace_struct; -struct rchan_buf; - -struct ltt_channel_struct { - /* First 32 bytes cache-hot cacheline */ - struct ltt_trace_struct *trace; - void *buf; - void *trans_channel_data; - int overwrite:1; - int active:1; - unsigned int n_subbufs_order; - unsigned long commit_count_mask; /* - * Commit count mask, removing - * the MSBs corresponding to - * bits used to represent the - * subbuffer index. - */ - /* End of first 32 bytes cacheline */ - - /* - * buffer_begin - called on buffer-switch to a new sub-buffer - * @buf: the channel buffer containing the new sub-buffer - */ - void (*buffer_begin) (struct rchan_buf *buf, - u64 tsc, unsigned int subbuf_idx); - /* - * buffer_end - called on buffer-switch to a new sub-buffer - * @buf: the channel buffer containing the previous sub-buffer - */ - void (*buffer_end) (struct rchan_buf *buf, - u64 tsc, unsigned int offset, unsigned int subbuf_idx); - struct kref kref; /* Channel transport reference count */ - unsigned int subbuf_size; - unsigned int subbuf_cnt; - const char *channel_name; - - int buf_shmid; -} ____cacheline_aligned; - -struct ltt_channel_setting { - unsigned int subbuf_size; - unsigned int subbuf_cnt; - struct kref kref; /* Number of references to structure content */ - struct list_head list; - unsigned int index; /* index of channel in trace channel array */ - u16 free_event_id; /* Next event ID to allocate */ - char name[PATH_MAX]; -}; - -int ltt_channels_register(const char *name); -int ltt_channels_unregister(const char *name); -int ltt_channels_set_default(const char *name, - unsigned int subbuf_size, - unsigned int subbuf_cnt); -const char *ltt_channels_get_name_from_index(unsigned int index); -int ltt_channels_get_index_from_name(const char *name); -struct ltt_channel_struct *ltt_channels_trace_alloc(unsigned int *nr_channels, - int overwrite, - int active); -void ltt_channels_trace_free(struct ltt_channel_struct *channels); -int _ltt_channels_get_event_id(const char *channel, const char *name); -int ltt_channels_get_event_id(const char *channel, const char *name); - -#endif /* _LTT_CHANNELS_H */ diff --git a/libtracing/relay.c b/libtracing/relay.c deleted file mode 100644 index 6bb5a35..0000000 --- a/libtracing/relay.c +++ /dev/null @@ -1,2490 +0,0 @@ -/* - * Public API and common code for kernel->userspace relay file support. - * - * Copyright (C) 2002-2005 - Tom Zanussi (zanussi@us.ibm.com), IBM Corp - * Copyright (C) 1999-2005 - Karim Yaghmour (karim@opersys.com) - * Copyright (C) 2008 - Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) - * - * Moved to kernel/relay.c by Paul Mundt, 2006. - * November 2006 - CPU hotplug support by Mathieu Desnoyers - * (mathieu.desnoyers@polymtl.ca) - * - * This file is released under the GPL. - */ -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -#include "kernelcompat.h" -#include -#include -#include -#include "list.h" -#include "relay.h" -#include "channels.h" -#include "kref.h" -#include "tracer.h" -#include "tracercore.h" -#include "usterr.h" - -/* list of open channels, for cpu hotplug */ -static DEFINE_MUTEX(relay_channels_mutex); -static LIST_HEAD(relay_channels); - - -static struct dentry *ltt_create_buf_file_callback(struct rchan_buf *buf); - -/** - * relay_alloc_buf - allocate a channel buffer - * @buf: the buffer struct - * @size: total size of the buffer - */ -//ust// static int relay_alloc_buf(struct rchan_buf *buf, size_t *size) -//ust//{ -//ust// unsigned int i, n_pages; -//ust// struct buf_page *buf_page, *n; -//ust// -//ust// *size = PAGE_ALIGN(*size); -//ust// n_pages = *size >> PAGE_SHIFT; -//ust// -//ust// INIT_LIST_HEAD(&buf->pages); -//ust// -//ust// for (i = 0; i < n_pages; i++) { -//ust// buf_page = kmalloc_node(sizeof(*buf_page), GFP_KERNEL, -//ust// cpu_to_node(buf->cpu)); -//ust// if (unlikely(!buf_page)) -//ust// goto depopulate; -//ust// buf_page->page = alloc_pages_node(cpu_to_node(buf->cpu), -//ust// GFP_KERNEL | __GFP_ZERO, 0); -//ust// if (unlikely(!buf_page->page)) { -//ust// kfree(buf_page); -//ust// goto depopulate; -//ust// } -//ust// list_add_tail(&buf_page->list, &buf->pages); -//ust// buf_page->offset = (size_t)i << PAGE_SHIFT; -//ust// buf_page->buf = buf; -//ust// set_page_private(buf_page->page, (unsigned long)buf_page); -//ust// if (i == 0) { -//ust// buf->wpage = buf_page; -//ust// buf->hpage[0] = buf_page; -//ust// buf->hpage[1] = buf_page; -//ust// buf->rpage = buf_page; -//ust// } -//ust// } -//ust// buf->page_count = n_pages; -//ust// return 0; -//ust// -//ust//depopulate: -//ust// list_for_each_entry_safe(buf_page, n, &buf->pages, list) { -//ust// list_del_init(&buf_page->list); -//ust// __free_page(buf_page->page); -//ust// kfree(buf_page); -//ust// } -//ust// return -ENOMEM; -//ust//} - -static int relay_alloc_buf(struct rchan_buf *buf, size_t *size) -{ -//ust// unsigned int n_pages; -//ust// struct buf_page *buf_page, *n; - - void *ptr; - int result; - - *size = PAGE_ALIGN(*size); - - result = buf->shmid = shmget(getpid(), *size, IPC_CREAT | IPC_EXCL | 0700); - if(buf->shmid == -1) { - PERROR("shmget"); - return -1; - } - - ptr = shmat(buf->shmid, NULL, 0); - if(ptr == (void *) -1) { - perror("shmat"); - goto destroy_shmem; - } - - /* Already mark the shared memory for destruction. This will occur only - * when all users have detached. - */ - result = shmctl(buf->shmid, IPC_RMID, NULL); - if(result == -1) { - perror("shmctl"); - return -1; - } - - buf->buf_data = ptr; - buf->buf_size = *size; - - return 0; - - destroy_shmem: - result = shmctl(buf->shmid, IPC_RMID, NULL); - if(result == -1) { - perror("shmctl"); - } - - return -1; -} - -/** - * relay_create_buf - allocate and initialize a channel buffer - * @chan: the relay channel - * @cpu: cpu the buffer belongs to - * - * Returns channel buffer if successful, %NULL otherwise. - */ -static struct rchan_buf *relay_create_buf(struct rchan *chan) -{ - int ret; - struct rchan_buf *buf = kzalloc(sizeof(struct rchan_buf), GFP_KERNEL); - if (!buf) - return NULL; - -// buf->cpu = cpu; - ret = relay_alloc_buf(buf, &chan->alloc_size); - if (ret) - goto free_buf; - - buf->chan = chan; - kref_get(&buf->chan->kref); - return buf; - -free_buf: - kfree(buf); - return NULL; -} - -/** - * relay_destroy_channel - free the channel struct - * @kref: target kernel reference that contains the relay channel - * - * Should only be called from kref_put(). - */ -static void relay_destroy_channel(struct kref *kref) -{ - struct rchan *chan = container_of(kref, struct rchan, kref); - kfree(chan); -} - -/** - * relay_destroy_buf - destroy an rchan_buf struct and associated buffer - * @buf: the buffer struct - */ -static void relay_destroy_buf(struct rchan_buf *buf) -{ - struct rchan *chan = buf->chan; - struct buf_page *buf_page, *n; - int result; - - result = munmap(buf->buf_data, buf->buf_size); - if(result == -1) { - PERROR("munmap"); - } - -//ust// chan->buf[buf->cpu] = NULL; - kfree(buf); - kref_put(&chan->kref, relay_destroy_channel); -} - -/** - * relay_remove_buf - remove a channel buffer - * @kref: target kernel reference that contains the relay buffer - * - * Removes the file from the fileystem, which also frees the - * rchan_buf_struct and the channel buffer. Should only be called from - * kref_put(). - */ -static void relay_remove_buf(struct kref *kref) -{ - struct rchan_buf *buf = container_of(kref, struct rchan_buf, kref); -//ust// buf->chan->cb->remove_buf_file(buf); - relay_destroy_buf(buf); -} - -/* - * High-level relay kernel API and associated functions. - */ - -/* - * rchan_callback implementations defining default channel behavior. Used - * in place of corresponding NULL values in client callback struct. - */ - -/* - * create_buf_file_create() default callback. Does nothing. - */ -static struct dentry *create_buf_file_default_callback(const char *filename, - struct dentry *parent, - int mode, - struct rchan_buf *buf) -{ - return NULL; -} - -/* - * remove_buf_file() default callback. Does nothing. - */ -static int remove_buf_file_default_callback(struct dentry *dentry) -{ - return -EINVAL; -} - -/** - * wakeup_readers - wake up readers waiting on a channel - * @data: contains the channel buffer - * - * This is the timer function used to defer reader waking. - */ -//ust// static void wakeup_readers(unsigned long data) -//ust// { -//ust// struct rchan_buf *buf = (struct rchan_buf *)data; -//ust// wake_up_interruptible(&buf->read_wait); -//ust// } - -/** - * __relay_reset - reset a channel buffer - * @buf: the channel buffer - * @init: 1 if this is a first-time initialization - * - * See relay_reset() for description of effect. - */ -static void __relay_reset(struct rchan_buf *buf, unsigned int init) -{ - if (init) { -//ust// init_waitqueue_head(&buf->read_wait); - kref_init(&buf->kref); -//ust// setup_timer(&buf->timer, wakeup_readers, (unsigned long)buf); - } else -//ust// del_timer_sync(&buf->timer); - - buf->finalized = 0; -} - -/* - * relay_open_buf - create a new relay channel buffer - * - * used by relay_open() and CPU hotplug. - */ -static struct rchan_buf *relay_open_buf(struct rchan *chan) -{ - struct rchan_buf *buf = NULL; - struct dentry *dentry; -//ust// char *tmpname; - -//ust// tmpname = kzalloc(NAME_MAX + 1, GFP_KERNEL); -//ust// if (!tmpname) -//ust// goto end; -//ust// snprintf(tmpname, NAME_MAX, "%s%d", chan->base_filename, cpu); - - buf = relay_create_buf(chan); - if (!buf) - goto free_name; - - __relay_reset(buf, 1); - - /* Create file in fs */ -//ust// dentry = chan->cb->create_buf_file(tmpname, chan->parent, S_IRUSR, -//ust// buf); - - ltt_create_buf_file_callback(buf); // ust // - -//ust// if (!dentry) -//ust// goto free_buf; -//ust// -//ust// buf->dentry = dentry; - - goto free_name; - -free_buf: - relay_destroy_buf(buf); - buf = NULL; -free_name: -//ust// kfree(tmpname); -end: - return buf; -} - -/** - * relay_close_buf - close a channel buffer - * @buf: channel buffer - * - * Marks the buffer finalized and restores the default callbacks. - * The channel buffer and channel buffer data structure are then freed - * automatically when the last reference is given up. - */ -static void relay_close_buf(struct rchan_buf *buf) -{ -//ust// del_timer_sync(&buf->timer); - kref_put(&buf->kref, relay_remove_buf); -} - -//ust// static void setup_callbacks(struct rchan *chan, -//ust// struct rchan_callbacks *cb) -//ust// { -//ust// if (!cb) { -//ust// chan->cb = &default_channel_callbacks; -//ust// return; -//ust// } -//ust// -//ust// if (!cb->create_buf_file) -//ust// cb->create_buf_file = create_buf_file_default_callback; -//ust// if (!cb->remove_buf_file) -//ust// cb->remove_buf_file = remove_buf_file_default_callback; -//ust// chan->cb = cb; -//ust// } - -/** - * relay_hotcpu_callback - CPU hotplug callback - * @nb: notifier block - * @action: hotplug action to take - * @hcpu: CPU number - * - * Returns the success/failure of the operation. (%NOTIFY_OK, %NOTIFY_BAD) - */ -//ust// static int __cpuinit relay_hotcpu_callback(struct notifier_block *nb, -//ust// unsigned long action, -//ust// void *hcpu) -//ust// { -//ust// unsigned int hotcpu = (unsigned long)hcpu; -//ust// struct rchan *chan; -//ust// -//ust// switch (action) { -//ust// case CPU_UP_PREPARE: -//ust// case CPU_UP_PREPARE_FROZEN: -//ust// mutex_lock(&relay_channels_mutex); -//ust// list_for_each_entry(chan, &relay_channels, list) { -//ust// if (chan->buf[hotcpu]) -//ust// continue; -//ust// chan->buf[hotcpu] = relay_open_buf(chan, hotcpu); -//ust// if (!chan->buf[hotcpu]) { -//ust// printk(KERN_ERR -//ust// "relay_hotcpu_callback: cpu %d buffer " -//ust// "creation failed\n", hotcpu); -//ust// mutex_unlock(&relay_channels_mutex); -//ust// return NOTIFY_BAD; -//ust// } -//ust// } -//ust// mutex_unlock(&relay_channels_mutex); -//ust// break; -//ust// case CPU_DEAD: -//ust// case CPU_DEAD_FROZEN: -//ust// /* No need to flush the cpu : will be flushed upon -//ust// * final relay_flush() call. */ -//ust// break; -//ust// } -//ust// return NOTIFY_OK; -//ust// } - -/** - * ltt_relay_open - create a new relay channel - * @base_filename: base name of files to create - * @parent: dentry of parent directory, %NULL for root directory - * @subbuf_size: size of sub-buffers - * @n_subbufs: number of sub-buffers - * @cb: client callback functions - * @private_data: user-defined data - * - * Returns channel pointer if successful, %NULL otherwise. - * - * Creates a channel buffer for each cpu using the sizes and - * attributes specified. The created channel buffer files - * will be named base_filename0...base_filenameN-1. File - * permissions will be %S_IRUSR. - */ -struct rchan *ltt_relay_open(const char *base_filename, - struct dentry *parent, - size_t subbuf_size, - size_t n_subbufs, - void *private_data) -{ - unsigned int i; - struct rchan *chan; -//ust// if (!base_filename) -//ust// return NULL; - - if (!(subbuf_size && n_subbufs)) - return NULL; - - chan = kzalloc(sizeof(struct rchan), GFP_KERNEL); - if (!chan) - return NULL; - - chan->version = LTT_RELAY_CHANNEL_VERSION; - chan->n_subbufs = n_subbufs; - chan->subbuf_size = subbuf_size; - chan->subbuf_size_order = get_count_order(subbuf_size); - chan->alloc_size = FIX_SIZE(subbuf_size * n_subbufs); - chan->parent = parent; - chan->private_data = private_data; -//ust// strlcpy(chan->base_filename, base_filename, NAME_MAX); -//ust// setup_callbacks(chan, cb); - kref_init(&chan->kref); - - mutex_lock(&relay_channels_mutex); -//ust// for_each_online_cpu(i) { - chan->buf = relay_open_buf(chan); - if (!chan->buf) - goto error; -//ust// } - list_add(&chan->list, &relay_channels); - mutex_unlock(&relay_channels_mutex); - - return chan; - -//ust//free_bufs: -//ust// for_each_possible_cpu(i) { -//ust// if (!chan->buf[i]) -//ust// break; -//ust// relay_close_buf(chan->buf[i]); -//ust// } - - error: - kref_put(&chan->kref, relay_destroy_channel); - mutex_unlock(&relay_channels_mutex); - return NULL; -} -//ust// EXPORT_SYMBOL_GPL(ltt_relay_open); - -/** - * ltt_relay_close - close the channel - * @chan: the channel - * - * Closes all channel buffers and frees the channel. - */ -void ltt_relay_close(struct rchan *chan) -{ - unsigned int i; - - if (!chan) - return; - - mutex_lock(&relay_channels_mutex); -//ust// for_each_possible_cpu(i) - if (chan->buf) - relay_close_buf(chan->buf); - - list_del(&chan->list); - kref_put(&chan->kref, relay_destroy_channel); - mutex_unlock(&relay_channels_mutex); -} -//ust// EXPORT_SYMBOL_GPL(ltt_relay_close); - -/* - * Start iteration at the previous element. Skip the real list head. - */ -//ust// struct buf_page *ltt_relay_find_prev_page(struct rchan_buf *buf, -//ust// struct buf_page *page, size_t offset, ssize_t diff_offset) -//ust// { -//ust// struct buf_page *iter; -//ust// size_t orig_iter_off; -//ust// unsigned int i = 0; -//ust// -//ust// orig_iter_off = page->offset; -//ust// list_for_each_entry_reverse(iter, &page->list, list) { -//ust// /* -//ust// * Skip the real list head. -//ust// */ -//ust// if (&iter->list == &buf->pages) -//ust// continue; -//ust// i++; -//ust// if (offset >= iter->offset -//ust// && offset < iter->offset + PAGE_SIZE) { -//ust// #ifdef CONFIG_LTT_RELAY_CHECK_RANDOM_ACCESS -//ust// if (i > 1) { -//ust// printk(KERN_WARNING -//ust// "Backward random access detected in " -//ust// "ltt_relay. Iterations %u, " -//ust// "offset %zu, orig iter->off %zu, " -//ust// "iter->off %zu diff_offset %zd.\n", i, -//ust// offset, orig_iter_off, iter->offset, -//ust// diff_offset); -//ust// WARN_ON(1); -//ust// } -//ust// #endif -//ust// return iter; -//ust// } -//ust// } -//ust// WARN_ON(1); -//ust// return NULL; -//ust// } -//ust// EXPORT_SYMBOL_GPL(ltt_relay_find_prev_page); - -/* - * Start iteration at the next element. Skip the real list head. - */ -//ust// struct buf_page *ltt_relay_find_next_page(struct rchan_buf *buf, -//ust// struct buf_page *page, size_t offset, ssize_t diff_offset) -//ust// { -//ust// struct buf_page *iter; -//ust// unsigned int i = 0; -//ust// size_t orig_iter_off; -//ust// -//ust// orig_iter_off = page->offset; -//ust// list_for_each_entry(iter, &page->list, list) { -//ust// /* -//ust// * Skip the real list head. -//ust// */ -//ust// if (&iter->list == &buf->pages) -//ust// continue; -//ust// i++; -//ust// if (offset >= iter->offset -//ust// && offset < iter->offset + PAGE_SIZE) { -//ust// #ifdef CONFIG_LTT_RELAY_CHECK_RANDOM_ACCESS -//ust// if (i > 1) { -//ust// printk(KERN_WARNING -//ust// "Forward random access detected in " -//ust// "ltt_relay. Iterations %u, " -//ust// "offset %zu, orig iter->off %zu, " -//ust// "iter->off %zu diff_offset %zd.\n", i, -//ust// offset, orig_iter_off, iter->offset, -//ust// diff_offset); -//ust// WARN_ON(1); -//ust// } -//ust// #endif -//ust// return iter; -//ust// } -//ust// } -//ust// WARN_ON(1); -//ust// return NULL; -//ust// } -//ust// EXPORT_SYMBOL_GPL(ltt_relay_find_next_page); - -/** - * ltt_relay_write - write data to a ltt_relay buffer. - * @buf : buffer - * @offset : offset within the buffer - * @src : source address - * @len : length to write - * @page : cached buffer page - * @pagecpy : page size copied so far - */ -void _ltt_relay_write(struct rchan_buf *buf, size_t offset, - const void *src, size_t len, ssize_t cpy) -{ - do { - len -= cpy; - src += cpy; - offset += cpy; - /* - * Underlying layer should never ask for writes across - * subbuffers. - */ - WARN_ON(offset >= buf->buf_size); - - cpy = min_t(size_t, len, buf->buf_size - offset); - ltt_relay_do_copy(buf->buf_data + offset, src, cpy); - } while (unlikely(len != cpy)); -} -//ust// EXPORT_SYMBOL_GPL(_ltt_relay_write); - -/** - * ltt_relay_read - read data from ltt_relay_buffer. - * @buf : buffer - * @offset : offset within the buffer - * @dest : destination address - * @len : length to write - */ -//ust// int ltt_relay_read(struct rchan_buf *buf, size_t offset, -//ust// void *dest, size_t len) -//ust// { -//ust// struct buf_page *page; -//ust// ssize_t pagecpy, orig_len; -//ust// -//ust// orig_len = len; -//ust// offset &= buf->chan->alloc_size - 1; -//ust// page = buf->rpage; -//ust// if (unlikely(!len)) -//ust// return 0; -//ust// for (;;) { -//ust// page = ltt_relay_cache_page(buf, &buf->rpage, page, offset); -//ust// pagecpy = min_t(size_t, len, PAGE_SIZE - (offset & ~PAGE_MASK)); -//ust// memcpy(dest, page_address(page->page) + (offset & ~PAGE_MASK), -//ust// pagecpy); -//ust// len -= pagecpy; -//ust// if (likely(!len)) -//ust// break; -//ust// dest += pagecpy; -//ust// offset += pagecpy; -//ust// /* -//ust// * Underlying layer should never ask for reads across -//ust// * subbuffers. -//ust// */ -//ust// WARN_ON(offset >= buf->chan->alloc_size); -//ust// } -//ust// return orig_len; -//ust// } -//ust// EXPORT_SYMBOL_GPL(ltt_relay_read); - -/** - * ltt_relay_read_get_page - Get a whole page to read from - * @buf : buffer - * @offset : offset within the buffer - */ -//ust// struct buf_page *ltt_relay_read_get_page(struct rchan_buf *buf, size_t offset) -//ust// { -//ust// struct buf_page *page; - -//ust// offset &= buf->chan->alloc_size - 1; -//ust// page = buf->rpage; -//ust// page = ltt_relay_cache_page(buf, &buf->rpage, page, offset); -//ust// return page; -//ust// } -//ust// EXPORT_SYMBOL_GPL(ltt_relay_read_get_page); - -/** - * ltt_relay_offset_address - get address of a location within the buffer - * @buf : buffer - * @offset : offset within the buffer. - * - * Return the address where a given offset is located. - * Should be used to get the current subbuffer header pointer. Given we know - * it's never on a page boundary, it's safe to write directly to this address, - * as long as the write is never bigger than a page size. - */ -void *ltt_relay_offset_address(struct rchan_buf *buf, size_t offset) -{ -//ust// struct buf_page *page; -//ust// unsigned int odd; -//ust// -//ust// offset &= buf->chan->alloc_size - 1; -//ust// odd = !!(offset & buf->chan->subbuf_size); -//ust// page = buf->hpage[odd]; -//ust// if (offset < page->offset || offset >= page->offset + PAGE_SIZE) -//ust// buf->hpage[odd] = page = buf->wpage; -//ust// page = ltt_relay_cache_page(buf, &buf->hpage[odd], page, offset); -//ust// return page_address(page->page) + (offset & ~PAGE_MASK); - return ((char *)buf->buf_data)+offset; - return NULL; -} -//ust// EXPORT_SYMBOL_GPL(ltt_relay_offset_address); - -/** - * relay_file_open - open file op for relay files - * @inode: the inode - * @filp: the file - * - * Increments the channel buffer refcount. - */ -//ust// static int relay_file_open(struct inode *inode, struct file *filp) -//ust// { -//ust// struct rchan_buf *buf = inode->i_private; -//ust// kref_get(&buf->kref); -//ust// filp->private_data = buf; -//ust// -//ust// return nonseekable_open(inode, filp); -//ust// } - -/** - * relay_file_release - release file op for relay files - * @inode: the inode - * @filp: the file - * - * Decrements the channel refcount, as the filesystem is - * no longer using it. - */ -//ust// static int relay_file_release(struct inode *inode, struct file *filp) -//ust// { -//ust// struct rchan_buf *buf = filp->private_data; -//ust// kref_put(&buf->kref, relay_remove_buf); -//ust// -//ust// return 0; -//ust// } - -//ust// const struct file_operations ltt_relay_file_operations = { -//ust// .open = relay_file_open, -//ust// .release = relay_file_release, -//ust// }; -//ust// EXPORT_SYMBOL_GPL(ltt_relay_file_operations); - -//ust// static __init int relay_init(void) -//ust// { -//ust// hotcpu_notifier(relay_hotcpu_callback, 5); -//ust// return 0; -//ust// } - -//ust// module_init(relay_init); -/* - * ltt/ltt-relay.c - * - * (C) Copyright 2005-2008 - Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) - * - * LTTng lockless buffer space management (reader/writer). - * - * Author: - * Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) - * - * Inspired from LTT : - * Karim Yaghmour (karim@opersys.com) - * Tom Zanussi (zanussi@us.ibm.com) - * Bob Wisniewski (bob@watson.ibm.com) - * And from K42 : - * Bob Wisniewski (bob@watson.ibm.com) - * - * Changelog: - * 08/10/08, Cleanup. - * 19/10/05, Complete lockless mechanism. - * 27/05/05, Modular redesign and rewrite. - * - * Userspace reader semantic : - * while (poll fd != POLLHUP) { - * - ioctl RELAY_GET_SUBBUF_SIZE - * while (1) { - * - ioctl GET_SUBBUF - * - splice 1 subbuffer worth of data to a pipe - * - splice the data from pipe to disk/network - * - ioctl PUT_SUBBUF, check error value - * if err val < 0, previous subbuffer was corrupted. - * } - * } - */ - -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include - -#if 0 -#define printk_dbg(fmt, args...) printk(fmt, args) -#else -#define printk_dbg(fmt, args...) -#endif - -/* - * Last TSC comparison functions. Check if the current TSC overflows - * LTT_TSC_BITS bits from the last TSC read. Reads and writes last_tsc - * atomically. - */ - -#if (BITS_PER_LONG == 32) -static inline void save_last_tsc(struct ltt_channel_buf_struct *ltt_buf, - u64 tsc) -{ - ltt_buf->last_tsc = (unsigned long)(tsc >> LTT_TSC_BITS); -} - -static inline int last_tsc_overflow(struct ltt_channel_buf_struct *ltt_buf, - u64 tsc) -{ - unsigned long tsc_shifted = (unsigned long)(tsc >> LTT_TSC_BITS); - - if (unlikely((tsc_shifted - ltt_buf->last_tsc))) - return 1; - else - return 0; -} -#else -static inline void save_last_tsc(struct ltt_channel_buf_struct *ltt_buf, - u64 tsc) -{ - ltt_buf->last_tsc = (unsigned long)tsc; -} - -static inline int last_tsc_overflow(struct ltt_channel_buf_struct *ltt_buf, - u64 tsc) -{ - if (unlikely((tsc - ltt_buf->last_tsc) >> LTT_TSC_BITS)) - return 1; - else - return 0; -} -#endif - -//ust// static struct file_operations ltt_file_operations; - -/* - * A switch is done during tracing or as a final flush after tracing (so it - * won't write in the new sub-buffer). - */ -enum force_switch_mode { FORCE_ACTIVE, FORCE_FLUSH }; - -static int ltt_relay_create_buffer(struct ltt_trace_struct *trace, - struct ltt_channel_struct *ltt_chan, - struct rchan_buf *buf, - unsigned int n_subbufs); - -static void ltt_relay_destroy_buffer(struct ltt_channel_struct *ltt_chan); - -static void ltt_force_switch(struct rchan_buf *buf, - enum force_switch_mode mode); - -/* - * Trace callbacks - */ -static void ltt_buffer_begin_callback(struct rchan_buf *buf, - u64 tsc, unsigned int subbuf_idx) -{ - struct ltt_channel_struct *channel = - (struct ltt_channel_struct *)buf->chan->private_data; - struct ltt_subbuffer_header *header = - (struct ltt_subbuffer_header *) - ltt_relay_offset_address(buf, - subbuf_idx * buf->chan->subbuf_size); - - header->cycle_count_begin = tsc; - header->lost_size = 0xFFFFFFFF; /* for debugging */ - header->buf_size = buf->chan->subbuf_size; - ltt_write_trace_header(channel->trace, header); -} - -/* - * offset is assumed to never be 0 here : never deliver a completely empty - * subbuffer. The lost size is between 0 and subbuf_size-1. - */ -static notrace void ltt_buffer_end_callback(struct rchan_buf *buf, - u64 tsc, unsigned int offset, unsigned int subbuf_idx) -{ - struct ltt_channel_struct *channel = - (struct ltt_channel_struct *)buf->chan->private_data; - struct ltt_channel_buf_struct *ltt_buf = channel->buf; - struct ltt_subbuffer_header *header = - (struct ltt_subbuffer_header *) - ltt_relay_offset_address(buf, - subbuf_idx * buf->chan->subbuf_size); - - header->lost_size = SUBBUF_OFFSET((buf->chan->subbuf_size - offset), - buf->chan); - header->cycle_count_end = tsc; - header->events_lost = local_read(<t_buf->events_lost); - header->subbuf_corrupt = local_read(<t_buf->corrupted_subbuffers); - -} - -void (*wake_consumer)(void *, int) = NULL; - -void relay_set_wake_consumer(void (*wake)(void *, int)) -{ - wake_consumer = wake; -} - -void relay_wake_consumer(void *arg, int finished) -{ - if(wake_consumer) - wake_consumer(arg, finished); -} - -static notrace void ltt_deliver(struct rchan_buf *buf, unsigned int subbuf_idx, - void *subbuf) -{ - struct ltt_channel_struct *channel = - (struct ltt_channel_struct *)buf->chan->private_data; - struct ltt_channel_buf_struct *ltt_buf = channel->buf; - int result; - - result = write(ltt_buf->data_ready_fd_write, "1", 1); - if(result == -1) { - PERROR("write (in ltt_relay_buffer_flush)"); - ERR("this should never happen!"); - } -//ust// atomic_set(<t_buf->wakeup_readers, 1); -} - -static struct dentry *ltt_create_buf_file_callback(struct rchan_buf *buf) -{ - struct ltt_channel_struct *ltt_chan; - int err; -//ust// struct dentry *dentry; - - ltt_chan = buf->chan->private_data; - err = ltt_relay_create_buffer(ltt_chan->trace, ltt_chan, buf, buf->chan->n_subbufs); - if (err) - return ERR_PTR(err); - -//ust// dentry = debugfs_create_file(filename, mode, parent, buf, -//ust// <t_file_operations); -//ust// if (!dentry) -//ust// goto error; -//ust// return dentry; - return NULL; //ust// -//ust//error: - ltt_relay_destroy_buffer(ltt_chan); - return NULL; -} - -static int ltt_remove_buf_file_callback(struct rchan_buf *buf) -{ -//ust// struct rchan_buf *buf = dentry->d_inode->i_private; - struct ltt_channel_struct *ltt_chan = buf->chan->private_data; - -//ust// debugfs_remove(dentry); - ltt_relay_destroy_buffer(ltt_chan); - - return 0; -} - -/* - * Wake writers : - * - * This must be done after the trace is removed from the RCU list so that there - * are no stalled writers. - */ -//ust// static void ltt_relay_wake_writers(struct ltt_channel_buf_struct *ltt_buf) -//ust// { -//ust// -//ust// if (waitqueue_active(<t_buf->write_wait)) -//ust// wake_up_interruptible(<t_buf->write_wait); -//ust// } - -/* - * This function should not be called from NMI interrupt context - */ -static notrace void ltt_buf_unfull(struct rchan_buf *buf, - unsigned int subbuf_idx, - long offset) -{ -//ust// struct ltt_channel_struct *ltt_channel = -//ust// (struct ltt_channel_struct *)buf->chan->private_data; -//ust// struct ltt_channel_buf_struct *ltt_buf = ltt_channel->buf; -//ust// -//ust// ltt_relay_wake_writers(ltt_buf); -} - -/** - * ltt_open - open file op for ltt files - * @inode: opened inode - * @file: opened file - * - * Open implementation. Makes sure only one open instance of a buffer is - * done at a given moment. - */ -//ust// static int ltt_open(struct inode *inode, struct file *file) -//ust// { -//ust// struct rchan_buf *buf = inode->i_private; -//ust// struct ltt_channel_struct *ltt_channel = -//ust// (struct ltt_channel_struct *)buf->chan->private_data; -//ust// struct ltt_channel_buf_struct *ltt_buf = -//ust// percpu_ptr(ltt_channel->buf, buf->cpu); -//ust// -//ust// if (!atomic_long_add_unless(<t_buf->active_readers, 1, 1)) -//ust// return -EBUSY; -//ust// return ltt_relay_file_operations.open(inode, file); -//ust// } - -/** - * ltt_release - release file op for ltt files - * @inode: opened inode - * @file: opened file - * - * Release implementation. - */ -//ust// static int ltt_release(struct inode *inode, struct file *file) -//ust// { -//ust// struct rchan_buf *buf = inode->i_private; -//ust// struct ltt_channel_struct *ltt_channel = -//ust// (struct ltt_channel_struct *)buf->chan->private_data; -//ust// struct ltt_channel_buf_struct *ltt_buf = -//ust// percpu_ptr(ltt_channel->buf, buf->cpu); -//ust// int ret; -//ust// -//ust// WARN_ON(atomic_long_read(<t_buf->active_readers) != 1); -//ust// atomic_long_dec(<t_buf->active_readers); -//ust// ret = ltt_relay_file_operations.release(inode, file); -//ust// WARN_ON(ret); -//ust// return ret; -//ust// } - -/** - * ltt_poll - file op for ltt files - * @filp: the file - * @wait: poll table - * - * Poll implementation. - */ -//ust// static unsigned int ltt_poll(struct file *filp, poll_table *wait) -//ust// { -//ust// unsigned int mask = 0; -//ust// struct inode *inode = filp->f_dentry->d_inode; -//ust// struct rchan_buf *buf = inode->i_private; -//ust// struct ltt_channel_struct *ltt_channel = -//ust// (struct ltt_channel_struct *)buf->chan->private_data; -//ust// struct ltt_channel_buf_struct *ltt_buf = -//ust// percpu_ptr(ltt_channel->buf, buf->cpu); -//ust// -//ust// if (filp->f_mode & FMODE_READ) { -//ust// poll_wait_set_exclusive(wait); -//ust// poll_wait(filp, &buf->read_wait, wait); -//ust// -//ust// WARN_ON(atomic_long_read(<t_buf->active_readers) != 1); -//ust// if (SUBBUF_TRUNC(local_read(<t_buf->offset), -//ust// buf->chan) -//ust// - SUBBUF_TRUNC(atomic_long_read(<t_buf->consumed), -//ust// buf->chan) -//ust// == 0) { -//ust// if (buf->finalized) -//ust// return POLLHUP; -//ust// else -//ust// return 0; -//ust// } else { -//ust// struct rchan *rchan = -//ust// ltt_channel->trans_channel_data; -//ust// if (SUBBUF_TRUNC(local_read(<t_buf->offset), -//ust// buf->chan) -//ust// - SUBBUF_TRUNC(atomic_long_read( -//ust// <t_buf->consumed), -//ust// buf->chan) -//ust// >= rchan->alloc_size) -//ust// return POLLPRI | POLLRDBAND; -//ust// else -//ust// return POLLIN | POLLRDNORM; -//ust// } -//ust// } -//ust// return mask; -//ust// } - -int ltt_do_get_subbuf(struct rchan_buf *buf, struct ltt_channel_buf_struct *ltt_buf, long *pconsumed_old) -{ - struct ltt_channel_struct *ltt_channel = (struct ltt_channel_struct *)buf->chan->private_data; - long consumed_old, consumed_idx, commit_count, write_offset; - consumed_old = atomic_long_read(<t_buf->consumed); - consumed_idx = SUBBUF_INDEX(consumed_old, buf->chan); - commit_count = local_read(<t_buf->commit_count[consumed_idx]); - /* - * Make sure we read the commit count before reading the buffer - * data and the write offset. Correct consumed offset ordering - * wrt commit count is insured by the use of cmpxchg to update - * the consumed offset. - */ - smp_rmb(); - write_offset = local_read(<t_buf->offset); - /* - * Check that the subbuffer we are trying to consume has been - * already fully committed. - */ - if (((commit_count - buf->chan->subbuf_size) - & ltt_channel->commit_count_mask) - - (BUFFER_TRUNC(consumed_old, buf->chan) - >> ltt_channel->n_subbufs_order) - != 0) { - return -EAGAIN; - } - /* - * Check that we are not about to read the same subbuffer in - * which the writer head is. - */ - if ((SUBBUF_TRUNC(write_offset, buf->chan) - - SUBBUF_TRUNC(consumed_old, buf->chan)) - == 0) { - return -EAGAIN; - } - - *pconsumed_old = consumed_old; - return 0; -} - -int ltt_do_put_subbuf(struct rchan_buf *buf, struct ltt_channel_buf_struct *ltt_buf, u32 uconsumed_old) -{ - long consumed_new, consumed_old; - - consumed_old = atomic_long_read(<t_buf->consumed); - consumed_old = consumed_old & (~0xFFFFFFFFL); - consumed_old = consumed_old | uconsumed_old; - consumed_new = SUBBUF_ALIGN(consumed_old, buf->chan); - -//ust// spin_lock(<t_buf->full_lock); - if (atomic_long_cmpxchg(<t_buf->consumed, consumed_old, - consumed_new) - != consumed_old) { - /* We have been pushed by the writer : the last - * buffer read _is_ corrupted! It can also - * happen if this is a buffer we never got. */ -//ust// spin_unlock(<t_buf->full_lock); - return -EIO; - } else { - /* tell the client that buffer is now unfull */ - int index; - long data; - index = SUBBUF_INDEX(consumed_old, buf->chan); - data = BUFFER_OFFSET(consumed_old, buf->chan); - ltt_buf_unfull(buf, index, data); -//ust// spin_unlock(<t_buf->full_lock); - } - return 0; -} - -/** - * ltt_ioctl - control on the debugfs file - * - * @inode: the inode - * @filp: the file - * @cmd: the command - * @arg: command arg - * - * This ioctl implements three commands necessary for a minimal - * producer/consumer implementation : - * RELAY_GET_SUBBUF - * Get the next sub buffer that can be read. It never blocks. - * RELAY_PUT_SUBBUF - * Release the currently read sub-buffer. Parameter is the last - * put subbuffer (returned by GET_SUBBUF). - * RELAY_GET_N_BUBBUFS - * returns the number of sub buffers in the per cpu channel. - * RELAY_GET_SUBBUF_SIZE - * returns the size of the sub buffers. - */ -//ust// static int ltt_ioctl(struct inode *inode, struct file *filp, -//ust// unsigned int cmd, unsigned long arg) -//ust// { -//ust// struct rchan_buf *buf = inode->i_private; -//ust// struct ltt_channel_struct *ltt_channel = -//ust// (struct ltt_channel_struct *)buf->chan->private_data; -//ust// struct ltt_channel_buf_struct *ltt_buf = -//ust// percpu_ptr(ltt_channel->buf, buf->cpu); -//ust// u32 __user *argp = (u32 __user *)arg; -//ust// -//ust// WARN_ON(atomic_long_read(<t_buf->active_readers) != 1); -//ust// switch (cmd) { -//ust// case RELAY_GET_SUBBUF: -//ust// { -//ust// int ret; -//ust// ret = ltt_do_get_subbuf(buf, ltt_buf, &consumed_old); -//ust// if(ret < 0) -//ust// return ret; -//ust// return put_user((u32)consumed_old, argp); -//ust// } -//ust// case RELAY_PUT_SUBBUF: -//ust// { -//ust// int ret; -//ust// u32 uconsumed_old; -//ust// ret = get_user(uconsumed_old, argp); -//ust// if (ret) -//ust// return ret; /* will return -EFAULT */ -//ust// return ltt_do_put_subbuf(buf, ltt_buf, uconsumed_old); -//ust// } -//ust// case RELAY_GET_N_SUBBUFS: -//ust// return put_user((u32)buf->chan->n_subbufs, argp); -//ust// break; -//ust// case RELAY_GET_SUBBUF_SIZE: -//ust// return put_user((u32)buf->chan->subbuf_size, argp); -//ust// break; -//ust// default: -//ust// return -ENOIOCTLCMD; -//ust// } -//ust// return 0; -//ust// } - -//ust// #ifdef CONFIG_COMPAT -//ust// static long ltt_compat_ioctl(struct file *file, unsigned int cmd, -//ust// unsigned long arg) -//ust// { -//ust// long ret = -ENOIOCTLCMD; -//ust// -//ust// lock_kernel(); -//ust// ret = ltt_ioctl(file->f_dentry->d_inode, file, cmd, arg); -//ust// unlock_kernel(); -//ust// -//ust// return ret; -//ust// } -//ust// #endif - -//ust// static void ltt_relay_pipe_buf_release(struct pipe_inode_info *pipe, -//ust// struct pipe_buffer *pbuf) -//ust// { -//ust// } -//ust// -//ust// static struct pipe_buf_operations ltt_relay_pipe_buf_ops = { -//ust// .can_merge = 0, -//ust// .map = generic_pipe_buf_map, -//ust// .unmap = generic_pipe_buf_unmap, -//ust// .confirm = generic_pipe_buf_confirm, -//ust// .release = ltt_relay_pipe_buf_release, -//ust// .steal = generic_pipe_buf_steal, -//ust// .get = generic_pipe_buf_get, -//ust// }; - -//ust// static void ltt_relay_page_release(struct splice_pipe_desc *spd, unsigned int i) -//ust// { -//ust// } - -/* - * subbuf_splice_actor - splice up to one subbuf's worth of data - */ -//ust// static int subbuf_splice_actor(struct file *in, -//ust// loff_t *ppos, -//ust// struct pipe_inode_info *pipe, -//ust// size_t len, -//ust// unsigned int flags) -//ust// { -//ust// struct rchan_buf *buf = in->private_data; -//ust// struct ltt_channel_struct *ltt_channel = -//ust// (struct ltt_channel_struct *)buf->chan->private_data; -//ust// struct ltt_channel_buf_struct *ltt_buf = -//ust// percpu_ptr(ltt_channel->buf, buf->cpu); -//ust// unsigned int poff, subbuf_pages, nr_pages; -//ust// struct page *pages[PIPE_BUFFERS]; -//ust// struct partial_page partial[PIPE_BUFFERS]; -//ust// struct splice_pipe_desc spd = { -//ust// .pages = pages, -//ust// .nr_pages = 0, -//ust// .partial = partial, -//ust// .flags = flags, -//ust// .ops = <t_relay_pipe_buf_ops, -//ust// .spd_release = ltt_relay_page_release, -//ust// }; -//ust// long consumed_old, consumed_idx, roffset; -//ust// unsigned long bytes_avail; -//ust// -//ust// /* -//ust// * Check that a GET_SUBBUF ioctl has been done before. -//ust// */ -//ust// WARN_ON(atomic_long_read(<t_buf->active_readers) != 1); -//ust// consumed_old = atomic_long_read(<t_buf->consumed); -//ust// consumed_old += *ppos; -//ust// consumed_idx = SUBBUF_INDEX(consumed_old, buf->chan); -//ust// -//ust// /* -//ust// * Adjust read len, if longer than what is available -//ust// */ -//ust// bytes_avail = SUBBUF_TRUNC(local_read(<t_buf->offset), buf->chan) -//ust// - consumed_old; -//ust// WARN_ON(bytes_avail > buf->chan->alloc_size); -//ust// len = min_t(size_t, len, bytes_avail); -//ust// subbuf_pages = bytes_avail >> PAGE_SHIFT; -//ust// nr_pages = min_t(unsigned int, subbuf_pages, PIPE_BUFFERS); -//ust// roffset = consumed_old & PAGE_MASK; -//ust// poff = consumed_old & ~PAGE_MASK; -//ust// printk_dbg(KERN_DEBUG "SPLICE actor len %zu pos %zd write_pos %ld\n", -//ust// len, (ssize_t)*ppos, local_read(<t_buf->offset)); -//ust// -//ust// for (; spd.nr_pages < nr_pages; spd.nr_pages++) { -//ust// unsigned int this_len; -//ust// struct buf_page *page; -//ust// -//ust// if (!len) -//ust// break; -//ust// printk_dbg(KERN_DEBUG "SPLICE actor loop len %zu roffset %ld\n", -//ust// len, roffset); -//ust// -//ust// this_len = PAGE_SIZE - poff; -//ust// page = ltt_relay_read_get_page(buf, roffset); -//ust// spd.pages[spd.nr_pages] = page->page; -//ust// spd.partial[spd.nr_pages].offset = poff; -//ust// spd.partial[spd.nr_pages].len = this_len; -//ust// -//ust// poff = 0; -//ust// roffset += PAGE_SIZE; -//ust// len -= this_len; -//ust// } -//ust// -//ust// if (!spd.nr_pages) -//ust// return 0; -//ust// -//ust// return splice_to_pipe(pipe, &spd); -//ust// } - -//ust// static ssize_t ltt_relay_file_splice_read(struct file *in, -//ust// loff_t *ppos, -//ust// struct pipe_inode_info *pipe, -//ust// size_t len, -//ust// unsigned int flags) -//ust// { -//ust// ssize_t spliced; -//ust// int ret; -//ust// -//ust// ret = 0; -//ust// spliced = 0; -//ust// -//ust// printk_dbg(KERN_DEBUG "SPLICE read len %zu pos %zd\n", -//ust// len, (ssize_t)*ppos); -//ust// while (len && !spliced) { -//ust// ret = subbuf_splice_actor(in, ppos, pipe, len, flags); -//ust// printk_dbg(KERN_DEBUG "SPLICE read loop ret %d\n", ret); -//ust// if (ret < 0) -//ust// break; -//ust// else if (!ret) { -//ust// if (flags & SPLICE_F_NONBLOCK) -//ust// ret = -EAGAIN; -//ust// break; -//ust// } -//ust// -//ust// *ppos += ret; -//ust// if (ret > len) -//ust// len = 0; -//ust// else -//ust// len -= ret; -//ust// spliced += ret; -//ust// } -//ust// -//ust// if (spliced) -//ust// return spliced; -//ust// -//ust// return ret; -//ust// } - -static void ltt_relay_print_subbuffer_errors( - struct ltt_channel_struct *ltt_chan, - long cons_off) -{ - struct rchan *rchan = ltt_chan->trans_channel_data; - struct ltt_channel_buf_struct *ltt_buf = ltt_chan->buf; - long cons_idx, commit_count, write_offset; - - cons_idx = SUBBUF_INDEX(cons_off, rchan); - commit_count = local_read(<t_buf->commit_count[cons_idx]); - /* - * No need to order commit_count and write_offset reads because we - * execute after trace is stopped when there are no readers left. - */ - write_offset = local_read(<t_buf->offset); - printk(KERN_WARNING - "LTT : unread channel %s offset is %ld " - "and cons_off : %ld\n", - ltt_chan->channel_name, write_offset, cons_off); - /* Check each sub-buffer for non filled commit count */ - if (((commit_count - rchan->subbuf_size) & ltt_chan->commit_count_mask) - - (BUFFER_TRUNC(cons_off, rchan) >> ltt_chan->n_subbufs_order) - != 0) - printk(KERN_ALERT - "LTT : %s : subbuffer %lu has non filled " - "commit count %lu.\n", - ltt_chan->channel_name, cons_idx, commit_count); - printk(KERN_ALERT "LTT : %s : commit count : %lu, subbuf size %zd\n", - ltt_chan->channel_name, commit_count, - rchan->subbuf_size); -} - -static void ltt_relay_print_errors(struct ltt_trace_struct *trace, - struct ltt_channel_struct *ltt_chan) -{ - struct rchan *rchan = ltt_chan->trans_channel_data; - struct ltt_channel_buf_struct *ltt_buf = ltt_chan->buf; - long cons_off; - - for (cons_off = atomic_long_read(<t_buf->consumed); - (SUBBUF_TRUNC(local_read(<t_buf->offset), - rchan) - - cons_off) > 0; - cons_off = SUBBUF_ALIGN(cons_off, rchan)) - ltt_relay_print_subbuffer_errors(ltt_chan, cons_off); -} - -static void ltt_relay_print_buffer_errors(struct ltt_channel_struct *ltt_chan) -{ - struct ltt_trace_struct *trace = ltt_chan->trace; - struct ltt_channel_buf_struct *ltt_buf = ltt_chan->buf; - - if (local_read(<t_buf->events_lost)) - printk(KERN_ALERT - "LTT : %s : %ld events lost " - "in %s channel.\n", - ltt_chan->channel_name, - local_read(<t_buf->events_lost), - ltt_chan->channel_name); - if (local_read(<t_buf->corrupted_subbuffers)) - printk(KERN_ALERT - "LTT : %s : %ld corrupted subbuffers " - "in %s channel.\n", - ltt_chan->channel_name, - local_read(<t_buf->corrupted_subbuffers), - ltt_chan->channel_name); - - ltt_relay_print_errors(trace, ltt_chan); -} - -static void ltt_relay_remove_dirs(struct ltt_trace_struct *trace) -{ -//ust// debugfs_remove(trace->dentry.trace_root); -} - -static void ltt_relay_release_channel(struct kref *kref) -{ - struct ltt_channel_struct *ltt_chan = container_of(kref, - struct ltt_channel_struct, kref); - free(ltt_chan->buf); -} - -/* - * Create ltt buffer. - */ -//ust// static int ltt_relay_create_buffer(struct ltt_trace_struct *trace, -//ust// struct ltt_channel_struct *ltt_chan, struct rchan_buf *buf, -//ust// unsigned int cpu, unsigned int n_subbufs) -//ust// { -//ust// struct ltt_channel_buf_struct *ltt_buf = -//ust// percpu_ptr(ltt_chan->buf, cpu); -//ust// unsigned int j; -//ust// -//ust// ltt_buf->commit_count = -//ust// kzalloc_node(sizeof(ltt_buf->commit_count) * n_subbufs, -//ust// GFP_KERNEL, cpu_to_node(cpu)); -//ust// if (!ltt_buf->commit_count) -//ust// return -ENOMEM; -//ust// kref_get(&trace->kref); -//ust// kref_get(&trace->ltt_transport_kref); -//ust// kref_get(<t_chan->kref); -//ust// local_set(<t_buf->offset, ltt_subbuffer_header_size()); -//ust// atomic_long_set(<t_buf->consumed, 0); -//ust// atomic_long_set(<t_buf->active_readers, 0); -//ust// for (j = 0; j < n_subbufs; j++) -//ust// local_set(<t_buf->commit_count[j], 0); -//ust// init_waitqueue_head(<t_buf->write_wait); -//ust// atomic_set(<t_buf->wakeup_readers, 0); -//ust// spin_lock_init(<t_buf->full_lock); -//ust// -//ust// ltt_buffer_begin_callback(buf, trace->start_tsc, 0); -//ust// /* atomic_add made on local variable on data that belongs to -//ust// * various CPUs : ok because tracing not started (for this cpu). */ -//ust// local_add(ltt_subbuffer_header_size(), <t_buf->commit_count[0]); -//ust// -//ust// local_set(<t_buf->events_lost, 0); -//ust// local_set(<t_buf->corrupted_subbuffers, 0); -//ust// -//ust// return 0; -//ust// } - -static int ltt_relay_create_buffer(struct ltt_trace_struct *trace, - struct ltt_channel_struct *ltt_chan, struct rchan_buf *buf, - unsigned int n_subbufs) -{ - struct ltt_channel_buf_struct *ltt_buf = ltt_chan->buf; - unsigned int j; - int fds[2]; - int result; - -//ust// ltt_buf->commit_count = -//ust// zmalloc(sizeof(ltt_buf->commit_count) * n_subbufs); -//ust// if (!ltt_buf->commit_count) -//ust// return -ENOMEM; - kref_get(&trace->kref); - kref_get(&trace->ltt_transport_kref); - kref_get(<t_chan->kref); - local_set(<t_buf->offset, ltt_subbuffer_header_size()); - atomic_long_set(<t_buf->consumed, 0); - atomic_long_set(<t_buf->active_readers, 0); - for (j = 0; j < n_subbufs; j++) - local_set(<t_buf->commit_count[j], 0); -//ust// init_waitqueue_head(<t_buf->write_wait); -//ust// atomic_set(<t_buf->wakeup_readers, 0); -//ust// spin_lock_init(<t_buf->full_lock); - - ltt_buffer_begin_callback(buf, trace->start_tsc, 0); - - local_add(ltt_subbuffer_header_size(), <t_buf->commit_count[0]); - - local_set(<t_buf->events_lost, 0); - local_set(<t_buf->corrupted_subbuffers, 0); - - result = pipe(fds); - if(result == -1) { - PERROR("pipe"); - return -1; - } - ltt_buf->data_ready_fd_read = fds[0]; - ltt_buf->data_ready_fd_write = fds[1]; - - return 0; -} - -static void ltt_relay_destroy_buffer(struct ltt_channel_struct *ltt_chan) -{ - struct ltt_trace_struct *trace = ltt_chan->trace; - struct ltt_channel_buf_struct *ltt_buf = ltt_chan->buf; - - kref_put(<t_chan->trace->ltt_transport_kref, - ltt_release_transport); - ltt_relay_print_buffer_errors(ltt_chan); -//ust// kfree(ltt_buf->commit_count); -//ust// ltt_buf->commit_count = NULL; - kref_put(<t_chan->kref, ltt_relay_release_channel); - kref_put(&trace->kref, ltt_release_trace); -//ust// wake_up_interruptible(&trace->kref_wq); -} - -static void ltt_chan_alloc_ltt_buf(struct ltt_channel_struct *ltt_chan) -{ - void *ptr; - int result; - - /* Get one page */ - /* FIXME: increase size if we have a commit_count array that overflows the page */ - size_t size = PAGE_ALIGN(1); - - result = ltt_chan->buf_shmid = shmget(getpid(), size, IPC_CREAT | IPC_EXCL | 0700); - if(ltt_chan->buf_shmid == -1) { - PERROR("shmget"); - return -1; - } - - ptr = shmat(ltt_chan->buf_shmid, NULL, 0); - if(ptr == (void *) -1) { - perror("shmat"); - goto destroy_shmem; - } - - /* Already mark the shared memory for destruction. This will occur only - * when all users have detached. - */ - result = shmctl(ltt_chan->buf_shmid, IPC_RMID, NULL); - if(result == -1) { - perror("shmctl"); - return -1; - } - - ltt_chan->buf = ptr; - - return 0; - - destroy_shmem: - result = shmctl(ltt_chan->buf_shmid, IPC_RMID, NULL); - if(result == -1) { - perror("shmctl"); - } - - return -1; -} - -/* - * Create channel. - */ -static int ltt_relay_create_channel(const char *trace_name, - struct ltt_trace_struct *trace, struct dentry *dir, - const char *channel_name, struct ltt_channel_struct *ltt_chan, - unsigned int subbuf_size, unsigned int n_subbufs, - int overwrite) -{ - char *tmpname; - unsigned int tmpname_len; - int err = 0; - - tmpname = kmalloc(PATH_MAX, GFP_KERNEL); - if (!tmpname) - return EPERM; - if (overwrite) { - strncpy(tmpname, LTT_FLIGHT_PREFIX, PATH_MAX-1); - strncat(tmpname, channel_name, - PATH_MAX-1-sizeof(LTT_FLIGHT_PREFIX)); - } else { - strncpy(tmpname, channel_name, PATH_MAX-1); - } - strncat(tmpname, "_", PATH_MAX-1-strlen(tmpname)); - - kref_init(<t_chan->kref); - - ltt_chan->trace = trace; - ltt_chan->buffer_begin = ltt_buffer_begin_callback; - ltt_chan->buffer_end = ltt_buffer_end_callback; - ltt_chan->overwrite = overwrite; - ltt_chan->n_subbufs_order = get_count_order(n_subbufs); - ltt_chan->commit_count_mask = (~0UL >> ltt_chan->n_subbufs_order); -//ust// ltt_chan->buf = percpu_alloc_mask(sizeof(struct ltt_channel_buf_struct), GFP_KERNEL, cpu_possible_map); - - ltt_chan_alloc_ltt_buf(ltt_chan); - -//ust// ltt_chan->buf = malloc(sizeof(struct ltt_channel_buf_struct)); - if (!ltt_chan->buf) - goto alloc_error; - ltt_chan->trans_channel_data = ltt_relay_open(tmpname, - dir, - subbuf_size, - n_subbufs, - ltt_chan); - tmpname_len = strlen(tmpname); - if (tmpname_len > 0) { - /* Remove final _ for pretty printing */ - tmpname[tmpname_len-1] = '\0'; - } - if (ltt_chan->trans_channel_data == NULL) { - printk(KERN_ERR "LTT : Can't open %s channel for trace %s\n", - tmpname, trace_name); - goto relay_open_error; - } - - err = 0; - goto end; - -relay_open_error: -//ust// percpu_free(ltt_chan->buf); -alloc_error: - err = EPERM; -end: - kfree(tmpname); - return err; -} - -static int ltt_relay_create_dirs(struct ltt_trace_struct *new_trace) -{ -//ust// new_trace->dentry.trace_root = debugfs_create_dir(new_trace->trace_name, -//ust// get_ltt_root()); -//ust// if (new_trace->dentry.trace_root == NULL) { -//ust// printk(KERN_ERR "LTT : Trace directory name %s already taken\n", -//ust// new_trace->trace_name); -//ust// return EEXIST; -//ust// } - -//ust// new_trace->callbacks.create_buf_file = ltt_create_buf_file_callback; -//ust// new_trace->callbacks.remove_buf_file = ltt_remove_buf_file_callback; - - return 0; -} - -/* - * LTTng channel flush function. - * - * Must be called when no tracing is active in the channel, because of - * accesses across CPUs. - */ -static notrace void ltt_relay_buffer_flush(struct rchan_buf *buf) -{ - struct ltt_channel_struct *channel = - (struct ltt_channel_struct *)buf->chan->private_data; - struct ltt_channel_buf_struct *ltt_buf = channel->buf; - int result; - - buf->finalized = 1; - ltt_force_switch(buf, FORCE_FLUSH); - - result = write(ltt_buf->data_ready_fd_write, "1", 1); - if(result == -1) { - PERROR("write (in ltt_relay_buffer_flush)"); - ERR("this should never happen!"); - } -} - -static void ltt_relay_async_wakeup_chan(struct ltt_channel_struct *ltt_channel) -{ -//ust// unsigned int i; -//ust// struct rchan *rchan = ltt_channel->trans_channel_data; -//ust// -//ust// for_each_possible_cpu(i) { -//ust// struct ltt_channel_buf_struct *ltt_buf = -//ust// percpu_ptr(ltt_channel->buf, i); -//ust// -//ust// if (atomic_read(<t_buf->wakeup_readers) == 1) { -//ust// atomic_set(<t_buf->wakeup_readers, 0); -//ust// wake_up_interruptible(&rchan->buf[i]->read_wait); -//ust// } -//ust// } -} - -static void ltt_relay_finish_buffer(struct ltt_channel_struct *ltt_channel) -{ - struct rchan *rchan = ltt_channel->trans_channel_data; - int result; - - if (rchan->buf) { - struct ltt_channel_buf_struct *ltt_buf = ltt_channel->buf; - ltt_relay_buffer_flush(rchan->buf); -//ust// ltt_relay_wake_writers(ltt_buf); - /* closing the pipe tells the consumer the buffer is finished */ - - //result = write(ltt_buf->data_ready_fd_write, "D", 1); - //if(result == -1) { - // PERROR("write (in ltt_relay_finish_buffer)"); - // ERR("this should never happen!"); - //} - close(ltt_buf->data_ready_fd_write); - } -} - - -static void ltt_relay_finish_channel(struct ltt_channel_struct *ltt_channel) -{ - unsigned int i; - -//ust// for_each_possible_cpu(i) - ltt_relay_finish_buffer(ltt_channel); -} - -static void ltt_relay_remove_channel(struct ltt_channel_struct *channel) -{ - struct rchan *rchan = channel->trans_channel_data; - - ltt_relay_close(rchan); - kref_put(&channel->kref, ltt_relay_release_channel); -} - -struct ltt_reserve_switch_offsets { - long begin, end, old; - long begin_switch, end_switch_current, end_switch_old; - long commit_count, reserve_commit_diff; - size_t before_hdr_pad, size; -}; - -/* - * Returns : - * 0 if ok - * !0 if execution must be aborted. - */ -static inline int ltt_relay_try_reserve( - struct ltt_channel_struct *ltt_channel, - struct ltt_channel_buf_struct *ltt_buf, struct rchan *rchan, - struct rchan_buf *buf, - struct ltt_reserve_switch_offsets *offsets, size_t data_size, - u64 *tsc, unsigned int *rflags, int largest_align) -{ - offsets->begin = local_read(<t_buf->offset); - offsets->old = offsets->begin; - offsets->begin_switch = 0; - offsets->end_switch_current = 0; - offsets->end_switch_old = 0; - - *tsc = trace_clock_read64(); - if (last_tsc_overflow(ltt_buf, *tsc)) - *rflags = LTT_RFLAG_ID_SIZE_TSC; - - if (SUBBUF_OFFSET(offsets->begin, buf->chan) == 0) { - offsets->begin_switch = 1; /* For offsets->begin */ - } else { - offsets->size = ltt_get_header_size(ltt_channel, - offsets->begin, data_size, - &offsets->before_hdr_pad, *rflags); - offsets->size += ltt_align(offsets->begin + offsets->size, - largest_align) - + data_size; - if ((SUBBUF_OFFSET(offsets->begin, buf->chan) + offsets->size) - > buf->chan->subbuf_size) { - offsets->end_switch_old = 1; /* For offsets->old */ - offsets->begin_switch = 1; /* For offsets->begin */ - } - } - if (offsets->begin_switch) { - long subbuf_index; - - if (offsets->end_switch_old) - offsets->begin = SUBBUF_ALIGN(offsets->begin, - buf->chan); - offsets->begin = offsets->begin + ltt_subbuffer_header_size(); - /* Test new buffer integrity */ - subbuf_index = SUBBUF_INDEX(offsets->begin, buf->chan); - offsets->reserve_commit_diff = - (BUFFER_TRUNC(offsets->begin, buf->chan) - >> ltt_channel->n_subbufs_order) - - (local_read(<t_buf->commit_count[subbuf_index]) - & ltt_channel->commit_count_mask); - if (offsets->reserve_commit_diff == 0) { - /* Next buffer not corrupted. */ - if (!ltt_channel->overwrite && - (SUBBUF_TRUNC(offsets->begin, buf->chan) - - SUBBUF_TRUNC(atomic_long_read( - <t_buf->consumed), - buf->chan)) - >= rchan->alloc_size) { - /* - * We do not overwrite non consumed buffers - * and we are full : event is lost. - */ - local_inc(<t_buf->events_lost); - return -1; - } else { - /* - * next buffer not corrupted, we are either in - * overwrite mode or the buffer is not full. - * It's safe to write in this new subbuffer. - */ - } - } else { - /* - * Next subbuffer corrupted. Force pushing reader even - * in normal mode. It's safe to write in this new - * subbuffer. - */ - } - offsets->size = ltt_get_header_size(ltt_channel, - offsets->begin, data_size, - &offsets->before_hdr_pad, *rflags); - offsets->size += ltt_align(offsets->begin + offsets->size, - largest_align) - + data_size; - if ((SUBBUF_OFFSET(offsets->begin, buf->chan) + offsets->size) - > buf->chan->subbuf_size) { - /* - * Event too big for subbuffers, report error, don't - * complete the sub-buffer switch. - */ - local_inc(<t_buf->events_lost); - return -1; - } else { - /* - * We just made a successful buffer switch and the event - * fits in the new subbuffer. Let's write. - */ - } - } else { - /* - * Event fits in the current buffer and we are not on a switch - * boundary. It's safe to write. - */ - } - offsets->end = offsets->begin + offsets->size; - - if ((SUBBUF_OFFSET(offsets->end, buf->chan)) == 0) { - /* - * The offset_end will fall at the very beginning of the next - * subbuffer. - */ - offsets->end_switch_current = 1; /* For offsets->begin */ - } - return 0; -} - -/* - * Returns : - * 0 if ok - * !0 if execution must be aborted. - */ -static inline int ltt_relay_try_switch( - enum force_switch_mode mode, - struct ltt_channel_struct *ltt_channel, - struct ltt_channel_buf_struct *ltt_buf, struct rchan *rchan, - struct rchan_buf *buf, - struct ltt_reserve_switch_offsets *offsets, - u64 *tsc) -{ - long subbuf_index; - - offsets->begin = local_read(<t_buf->offset); - offsets->old = offsets->begin; - offsets->begin_switch = 0; - offsets->end_switch_old = 0; - - *tsc = trace_clock_read64(); - - if (SUBBUF_OFFSET(offsets->begin, buf->chan) != 0) { - offsets->begin = SUBBUF_ALIGN(offsets->begin, buf->chan); - offsets->end_switch_old = 1; - } else { - /* we do not have to switch : buffer is empty */ - return -1; - } - if (mode == FORCE_ACTIVE) - offsets->begin += ltt_subbuffer_header_size(); - /* - * Always begin_switch in FORCE_ACTIVE mode. - * Test new buffer integrity - */ - subbuf_index = SUBBUF_INDEX(offsets->begin, buf->chan); - offsets->reserve_commit_diff = - (BUFFER_TRUNC(offsets->begin, buf->chan) - >> ltt_channel->n_subbufs_order) - - (local_read(<t_buf->commit_count[subbuf_index]) - & ltt_channel->commit_count_mask); - if (offsets->reserve_commit_diff == 0) { - /* Next buffer not corrupted. */ - if (mode == FORCE_ACTIVE - && !ltt_channel->overwrite - && offsets->begin - atomic_long_read(<t_buf->consumed) - >= rchan->alloc_size) { - /* - * We do not overwrite non consumed buffers and we are - * full : ignore switch while tracing is active. - */ - return -1; - } - } else { - /* - * Next subbuffer corrupted. Force pushing reader even in normal - * mode - */ - } - offsets->end = offsets->begin; - return 0; -} - -static inline void ltt_reserve_push_reader( - struct ltt_channel_struct *ltt_channel, - struct ltt_channel_buf_struct *ltt_buf, - struct rchan *rchan, - struct rchan_buf *buf, - struct ltt_reserve_switch_offsets *offsets) -{ - long consumed_old, consumed_new; - - do { - consumed_old = atomic_long_read(<t_buf->consumed); - /* - * If buffer is in overwrite mode, push the reader consumed - * count if the write position has reached it and we are not - * at the first iteration (don't push the reader farther than - * the writer). This operation can be done concurrently by many - * writers in the same buffer, the writer being at the farthest - * write position sub-buffer index in the buffer being the one - * which will win this loop. - * If the buffer is not in overwrite mode, pushing the reader - * only happens if a sub-buffer is corrupted. - */ - if ((SUBBUF_TRUNC(offsets->end-1, buf->chan) - - SUBBUF_TRUNC(consumed_old, buf->chan)) - >= rchan->alloc_size) - consumed_new = SUBBUF_ALIGN(consumed_old, buf->chan); - else { - consumed_new = consumed_old; - break; - } - } while (atomic_long_cmpxchg(<t_buf->consumed, consumed_old, - consumed_new) != consumed_old); - - if (consumed_old != consumed_new) { - /* - * Reader pushed : we are the winner of the push, we can - * therefore reequilibrate reserve and commit. Atomic increment - * of the commit count permits other writers to play around - * with this variable before us. We keep track of - * corrupted_subbuffers even in overwrite mode : - * we never want to write over a non completely committed - * sub-buffer : possible causes : the buffer size is too low - * compared to the unordered data input, or there is a writer - * that died between the reserve and the commit. - */ - if (offsets->reserve_commit_diff) { - /* - * We have to alter the sub-buffer commit count. - * We do not deliver the previous subbuffer, given it - * was either corrupted or not consumed (overwrite - * mode). - */ - local_add(offsets->reserve_commit_diff, - <t_buf->commit_count[ - SUBBUF_INDEX(offsets->begin, - buf->chan)]); - if (!ltt_channel->overwrite - || offsets->reserve_commit_diff - != rchan->subbuf_size) { - /* - * The reserve commit diff was not subbuf_size : - * it means the subbuffer was partly written to - * and is therefore corrupted. If it is multiple - * of subbuffer size and we are in flight - * recorder mode, we are skipping over a whole - * subbuffer. - */ - local_inc(<t_buf->corrupted_subbuffers); - } - } - } -} - - -/* - * ltt_reserve_switch_old_subbuf: switch old subbuffer - * - * Concurrency safe because we are the last and only thread to alter this - * sub-buffer. As long as it is not delivered and read, no other thread can - * alter the offset, alter the reserve_count or call the - * client_buffer_end_callback on this sub-buffer. - * - * The only remaining threads could be the ones with pending commits. They will - * have to do the deliver themselves. Not concurrency safe in overwrite mode. - * We detect corrupted subbuffers with commit and reserve counts. We keep a - * corrupted sub-buffers count and push the readers across these sub-buffers. - * - * Not concurrency safe if a writer is stalled in a subbuffer and another writer - * switches in, finding out it's corrupted. The result will be than the old - * (uncommited) subbuffer will be declared corrupted, and that the new subbuffer - * will be declared corrupted too because of the commit count adjustment. - * - * Note : offset_old should never be 0 here. - */ -static inline void ltt_reserve_switch_old_subbuf( - struct ltt_channel_struct *ltt_channel, - struct ltt_channel_buf_struct *ltt_buf, struct rchan *rchan, - struct rchan_buf *buf, - struct ltt_reserve_switch_offsets *offsets, u64 *tsc) -{ - long oldidx = SUBBUF_INDEX(offsets->old - 1, rchan); - - ltt_channel->buffer_end(buf, *tsc, offsets->old, oldidx); - /* Must write buffer end before incrementing commit count */ - smp_wmb(); - offsets->commit_count = - local_add_return(rchan->subbuf_size - - (SUBBUF_OFFSET(offsets->old - 1, rchan) - + 1), - <t_buf->commit_count[oldidx]); - if ((BUFFER_TRUNC(offsets->old - 1, rchan) - >> ltt_channel->n_subbufs_order) - - ((offsets->commit_count - rchan->subbuf_size) - & ltt_channel->commit_count_mask) == 0) - ltt_deliver(buf, oldidx, NULL); -} - -/* - * ltt_reserve_switch_new_subbuf: Populate new subbuffer. - * - * This code can be executed unordered : writers may already have written to the - * sub-buffer before this code gets executed, caution. The commit makes sure - * that this code is executed before the deliver of this sub-buffer. - */ -static /*inline*/ void ltt_reserve_switch_new_subbuf( - struct ltt_channel_struct *ltt_channel, - struct ltt_channel_buf_struct *ltt_buf, struct rchan *rchan, - struct rchan_buf *buf, - struct ltt_reserve_switch_offsets *offsets, u64 *tsc) -{ - long beginidx = SUBBUF_INDEX(offsets->begin, rchan); - - ltt_channel->buffer_begin(buf, *tsc, beginidx); - /* Must write buffer end before incrementing commit count */ - smp_wmb(); - offsets->commit_count = local_add_return(ltt_subbuffer_header_size(), - <t_buf->commit_count[beginidx]); - /* Check if the written buffer has to be delivered */ - if ((BUFFER_TRUNC(offsets->begin, rchan) - >> ltt_channel->n_subbufs_order) - - ((offsets->commit_count - rchan->subbuf_size) - & ltt_channel->commit_count_mask) == 0) - ltt_deliver(buf, beginidx, NULL); -} - - -/* - * ltt_reserve_end_switch_current: finish switching current subbuffer - * - * Concurrency safe because we are the last and only thread to alter this - * sub-buffer. As long as it is not delivered and read, no other thread can - * alter the offset, alter the reserve_count or call the - * client_buffer_end_callback on this sub-buffer. - * - * The only remaining threads could be the ones with pending commits. They will - * have to do the deliver themselves. Not concurrency safe in overwrite mode. - * We detect corrupted subbuffers with commit and reserve counts. We keep a - * corrupted sub-buffers count and push the readers across these sub-buffers. - * - * Not concurrency safe if a writer is stalled in a subbuffer and another writer - * switches in, finding out it's corrupted. The result will be than the old - * (uncommited) subbuffer will be declared corrupted, and that the new subbuffer - * will be declared corrupted too because of the commit count adjustment. - */ -static inline void ltt_reserve_end_switch_current( - struct ltt_channel_struct *ltt_channel, - struct ltt_channel_buf_struct *ltt_buf, struct rchan *rchan, - struct rchan_buf *buf, - struct ltt_reserve_switch_offsets *offsets, u64 *tsc) -{ - long endidx = SUBBUF_INDEX(offsets->end - 1, rchan); - - ltt_channel->buffer_end(buf, *tsc, offsets->end, endidx); - /* Must write buffer begin before incrementing commit count */ - smp_wmb(); - offsets->commit_count = - local_add_return(rchan->subbuf_size - - (SUBBUF_OFFSET(offsets->end - 1, rchan) - + 1), - <t_buf->commit_count[endidx]); - if ((BUFFER_TRUNC(offsets->end - 1, rchan) - >> ltt_channel->n_subbufs_order) - - ((offsets->commit_count - rchan->subbuf_size) - & ltt_channel->commit_count_mask) == 0) - ltt_deliver(buf, endidx, NULL); -} - -/** - * ltt_relay_reserve_slot - Atomic slot reservation in a LTTng buffer. - * @trace: the trace structure to log to. - * @ltt_channel: channel structure - * @transport_data: data structure specific to ltt relay - * @data_size: size of the variable length data to log. - * @slot_size: pointer to total size of the slot (out) - * @buf_offset : pointer to reserved buffer offset (out) - * @tsc: pointer to the tsc at the slot reservation (out) - * @cpu: cpuid - * - * Return : -ENOSPC if not enough space, else returns 0. - * It will take care of sub-buffer switching. - */ -static notrace int ltt_relay_reserve_slot(struct ltt_trace_struct *trace, - struct ltt_channel_struct *ltt_channel, void **transport_data, - size_t data_size, size_t *slot_size, long *buf_offset, u64 *tsc, - unsigned int *rflags, int largest_align) -{ - struct rchan *rchan = ltt_channel->trans_channel_data; - struct rchan_buf *buf = *transport_data = rchan->buf; - struct ltt_channel_buf_struct *ltt_buf = ltt_channel->buf; - struct ltt_reserve_switch_offsets offsets; - - offsets.reserve_commit_diff = 0; - offsets.size = 0; - - /* - * Perform retryable operations. - */ - if (ltt_nesting > 4) { - local_inc(<t_buf->events_lost); - return -EPERM; - } - do { - if (ltt_relay_try_reserve(ltt_channel, ltt_buf, - rchan, buf, &offsets, data_size, tsc, rflags, - largest_align)) - return -ENOSPC; - } while (local_cmpxchg(<t_buf->offset, offsets.old, - offsets.end) != offsets.old); - - /* - * Atomically update last_tsc. This update races against concurrent - * atomic updates, but the race will always cause supplementary full TSC - * events, never the opposite (missing a full TSC event when it would be - * needed). - */ - save_last_tsc(ltt_buf, *tsc); - - /* - * Push the reader if necessary - */ - ltt_reserve_push_reader(ltt_channel, ltt_buf, rchan, buf, &offsets); - - /* - * Switch old subbuffer if needed. - */ - if (offsets.end_switch_old) - ltt_reserve_switch_old_subbuf(ltt_channel, ltt_buf, rchan, buf, - &offsets, tsc); - - /* - * Populate new subbuffer. - */ - if (offsets.begin_switch) - ltt_reserve_switch_new_subbuf(ltt_channel, ltt_buf, rchan, - buf, &offsets, tsc); - - if (offsets.end_switch_current) - ltt_reserve_end_switch_current(ltt_channel, ltt_buf, rchan, - buf, &offsets, tsc); - - *slot_size = offsets.size; - *buf_offset = offsets.begin + offsets.before_hdr_pad; - return 0; -} - -/* - * Force a sub-buffer switch for a per-cpu buffer. This operation is - * completely reentrant : can be called while tracing is active with - * absolutely no lock held. - * - * Note, however, that as a local_cmpxchg is used for some atomic - * operations, this function must be called from the CPU which owns the buffer - * for a ACTIVE flush. - */ -static notrace void ltt_force_switch(struct rchan_buf *buf, - enum force_switch_mode mode) -{ - struct ltt_channel_struct *ltt_channel = - (struct ltt_channel_struct *)buf->chan->private_data; - struct ltt_channel_buf_struct *ltt_buf = ltt_channel->buf; - struct rchan *rchan = ltt_channel->trans_channel_data; - struct ltt_reserve_switch_offsets offsets; - u64 tsc; - - offsets.reserve_commit_diff = 0; - offsets.size = 0; - - /* - * Perform retryable operations. - */ - do { - if (ltt_relay_try_switch(mode, ltt_channel, ltt_buf, - rchan, buf, &offsets, &tsc)) - return; - } while (local_cmpxchg(<t_buf->offset, offsets.old, - offsets.end) != offsets.old); - - /* - * Atomically update last_tsc. This update races against concurrent - * atomic updates, but the race will always cause supplementary full TSC - * events, never the opposite (missing a full TSC event when it would be - * needed). - */ - save_last_tsc(ltt_buf, tsc); - - /* - * Push the reader if necessary - */ - if (mode == FORCE_ACTIVE) - ltt_reserve_push_reader(ltt_channel, ltt_buf, rchan, - buf, &offsets); - - /* - * Switch old subbuffer if needed. - */ - if (offsets.end_switch_old) - ltt_reserve_switch_old_subbuf(ltt_channel, ltt_buf, rchan, buf, - &offsets, &tsc); - - /* - * Populate new subbuffer. - */ - if (mode == FORCE_ACTIVE) - ltt_reserve_switch_new_subbuf(ltt_channel, - ltt_buf, rchan, buf, &offsets, &tsc); -} - -/* - * for flight recording. must be called after relay_commit. - * This function decrements de subbuffer's lost_size each time the commit count - * reaches back the reserve offset (module subbuffer size). It is useful for - * crash dump. - * We use slot_size - 1 to make sure we deal correctly with the case where we - * fill the subbuffer completely (so the subbuf index stays in the previous - * subbuffer). - */ -//ust// #ifdef CONFIG_LTT_VMCORE -static /*inline*/ void ltt_write_commit_counter(struct rchan_buf *buf, - long buf_offset, size_t slot_size) -{ - struct ltt_channel_struct *ltt_channel = - (struct ltt_channel_struct *)buf->chan->private_data; - struct ltt_channel_buf_struct *ltt_buf = ltt_channel->buf; - struct ltt_subbuffer_header *header; - long offset, subbuf_idx, commit_count; - uint32_t lost_old, lost_new; - - subbuf_idx = SUBBUF_INDEX(buf_offset - 1, buf->chan); - offset = buf_offset + slot_size; - header = (struct ltt_subbuffer_header *) - ltt_relay_offset_address(buf, - subbuf_idx * buf->chan->subbuf_size); - for (;;) { - lost_old = header->lost_size; - commit_count = - local_read(<t_buf->commit_count[subbuf_idx]); - /* SUBBUF_OFFSET includes commit_count_mask */ - if (!SUBBUF_OFFSET(offset - commit_count, buf->chan)) { - lost_new = (uint32_t)buf->chan->subbuf_size - - SUBBUF_OFFSET(commit_count, buf->chan); - lost_old = cmpxchg_local(&header->lost_size, lost_old, - lost_new); - if (lost_old <= lost_new) - break; - } else { - break; - } - } -} -//ust// #else -//ust// static inline void ltt_write_commit_counter(struct rchan_buf *buf, -//ust// long buf_offset, size_t slot_size) -//ust// { -//ust// } -//ust// #endif - -/* - * Atomic unordered slot commit. Increments the commit count in the - * specified sub-buffer, and delivers it if necessary. - * - * Parameters: - * - * @ltt_channel : channel structure - * @transport_data: transport-specific data - * @buf_offset : offset following the event header. - * @slot_size : size of the reserved slot. - */ -static notrace void ltt_relay_commit_slot( - struct ltt_channel_struct *ltt_channel, - void **transport_data, long buf_offset, size_t slot_size) -{ - struct rchan_buf *buf = *transport_data; - struct ltt_channel_buf_struct *ltt_buf = ltt_channel->buf; - struct rchan *rchan = buf->chan; - long offset_end = buf_offset; - long endidx = SUBBUF_INDEX(offset_end - 1, rchan); - long commit_count; - - /* Must write slot data before incrementing commit count */ - smp_wmb(); - commit_count = local_add_return(slot_size, - <t_buf->commit_count[endidx]); - /* Check if all commits have been done */ - if ((BUFFER_TRUNC(offset_end - 1, rchan) - >> ltt_channel->n_subbufs_order) - - ((commit_count - rchan->subbuf_size) - & ltt_channel->commit_count_mask) == 0) - ltt_deliver(buf, endidx, NULL); - /* - * Update lost_size for each commit. It's needed only for extracting - * ltt buffers from vmcore, after crash. - */ - ltt_write_commit_counter(buf, buf_offset, slot_size); - - DBG("commited slot. now commit count is %ld", commit_count); -} - -/* - * This is called with preemption disabled when user space has requested - * blocking mode. If one of the active traces has free space below a - * specific threshold value, we reenable preemption and block. - */ -static int ltt_relay_user_blocking(struct ltt_trace_struct *trace, - unsigned int chan_index, size_t data_size, - struct user_dbg_data *dbg) -{ -//ust// struct rchan *rchan; -//ust// struct ltt_channel_buf_struct *ltt_buf; -//ust// struct ltt_channel_struct *channel; -//ust// struct rchan_buf *relay_buf; -//ust// int cpu; -//ust// DECLARE_WAITQUEUE(wait, current); -//ust// -//ust// channel = &trace->channels[chan_index]; -//ust// rchan = channel->trans_channel_data; -//ust// cpu = smp_processor_id(); -//ust// relay_buf = rchan->buf[cpu]; -//ust// ltt_buf = percpu_ptr(channel->buf, cpu); -//ust// -//ust// /* -//ust// * Check if data is too big for the channel : do not -//ust// * block for it. -//ust// */ -//ust// if (LTT_RESERVE_CRITICAL + data_size > relay_buf->chan->subbuf_size) -//ust// return 0; -//ust// -//ust// /* -//ust// * If free space too low, we block. We restart from the -//ust// * beginning after we resume (cpu id may have changed -//ust// * while preemption is active). -//ust// */ -//ust// spin_lock(<t_buf->full_lock); -//ust// if (!channel->overwrite) { -//ust// dbg->write = local_read(<t_buf->offset); -//ust// dbg->read = atomic_long_read(<t_buf->consumed); -//ust// dbg->avail_size = dbg->write + LTT_RESERVE_CRITICAL + data_size -//ust// - SUBBUF_TRUNC(dbg->read, -//ust// relay_buf->chan); -//ust// if (dbg->avail_size > rchan->alloc_size) { -//ust// __set_current_state(TASK_INTERRUPTIBLE); -//ust// add_wait_queue(<t_buf->write_wait, &wait); -//ust// spin_unlock(<t_buf->full_lock); -//ust// preempt_enable(); -//ust// schedule(); -//ust// __set_current_state(TASK_RUNNING); -//ust// remove_wait_queue(<t_buf->write_wait, &wait); -//ust// if (signal_pending(current)) -//ust// return -ERESTARTSYS; -//ust// preempt_disable(); -//ust// return 1; -//ust// } -//ust// } -//ust// spin_unlock(<t_buf->full_lock); - return 0; -} - -static void ltt_relay_print_user_errors(struct ltt_trace_struct *trace, - unsigned int chan_index, size_t data_size, - struct user_dbg_data *dbg) -{ - struct rchan *rchan; - struct ltt_channel_buf_struct *ltt_buf; - struct ltt_channel_struct *channel; - struct rchan_buf *relay_buf; - - channel = &trace->channels[chan_index]; - rchan = channel->trans_channel_data; - relay_buf = rchan->buf; - ltt_buf = channel->buf; - - printk(KERN_ERR "Error in LTT usertrace : " - "buffer full : event lost in blocking " - "mode. Increase LTT_RESERVE_CRITICAL.\n"); - printk(KERN_ERR "LTT nesting level is %u.\n", ltt_nesting); - printk(KERN_ERR "LTT avail size %lu.\n", - dbg->avail_size); - printk(KERN_ERR "avai write : %lu, read : %lu\n", - dbg->write, dbg->read); - - dbg->write = local_read(<t_buf->offset); - dbg->read = atomic_long_read(<t_buf->consumed); - - printk(KERN_ERR "LTT cur size %lu.\n", - dbg->write + LTT_RESERVE_CRITICAL + data_size - - SUBBUF_TRUNC(dbg->read, relay_buf->chan)); - printk(KERN_ERR "cur write : %lu, read : %lu\n", - dbg->write, dbg->read); -} - -//ust// static struct ltt_transport ltt_relay_transport = { -//ust// .name = "relay", -//ust// .owner = THIS_MODULE, -//ust// .ops = { -//ust// .create_dirs = ltt_relay_create_dirs, -//ust// .remove_dirs = ltt_relay_remove_dirs, -//ust// .create_channel = ltt_relay_create_channel, -//ust// .finish_channel = ltt_relay_finish_channel, -//ust// .remove_channel = ltt_relay_remove_channel, -//ust// .wakeup_channel = ltt_relay_async_wakeup_chan, -//ust// .commit_slot = ltt_relay_commit_slot, -//ust// .reserve_slot = ltt_relay_reserve_slot, -//ust// .user_blocking = ltt_relay_user_blocking, -//ust// .user_errors = ltt_relay_print_user_errors, -//ust// }, -//ust// }; - -static struct ltt_transport ust_relay_transport = { - .name = "ustrelay", -//ust// .owner = THIS_MODULE, - .ops = { - .create_dirs = ltt_relay_create_dirs, - .remove_dirs = ltt_relay_remove_dirs, - .create_channel = ltt_relay_create_channel, - .finish_channel = ltt_relay_finish_channel, - .remove_channel = ltt_relay_remove_channel, - .wakeup_channel = ltt_relay_async_wakeup_chan, - .commit_slot = ltt_relay_commit_slot, - .reserve_slot = ltt_relay_reserve_slot, - .user_blocking = ltt_relay_user_blocking, - .user_errors = ltt_relay_print_user_errors, - }, -}; - -//ust// static int __init ltt_relay_init(void) -//ust// { -//ust// printk(KERN_INFO "LTT : ltt-relay init\n"); -//ust// -//ust// ltt_file_operations = ltt_relay_file_operations; -//ust// ltt_file_operations.owner = THIS_MODULE; -//ust// ltt_file_operations.open = ltt_open; -//ust// ltt_file_operations.release = ltt_release; -//ust// ltt_file_operations.poll = ltt_poll; -//ust// ltt_file_operations.splice_read = ltt_relay_file_splice_read, -//ust// ltt_file_operations.ioctl = ltt_ioctl; -//ust//#ifdef CONFIG_COMPAT -//ust// ltt_file_operations.compat_ioctl = ltt_compat_ioctl; -//ust//#endif -//ust// -//ust// ltt_transport_register(<t_relay_transport); -//ust// -//ust// return 0; -//ust// } - -static char initialized = 0; - -void __attribute__((constructor)) init_ustrelay_transport(void) -{ - if(!initialized) { - ltt_transport_register(&ust_relay_transport); - initialized = 1; - } -} - -static void __exit ltt_relay_exit(void) -{ -//ust// printk(KERN_INFO "LTT : ltt-relay exit\n"); - - ltt_transport_unregister(&ust_relay_transport); -} - -//ust// module_init(ltt_relay_init); -//ust// module_exit(ltt_relay_exit); -//ust// -//ust// MODULE_LICENSE("GPL"); -//ust// MODULE_AUTHOR("Mathieu Desnoyers"); -//ust// MODULE_DESCRIPTION("Linux Trace Toolkit Next Generation Lockless Relay"); diff --git a/libtracing/relay.h b/libtracing/relay.h deleted file mode 100644 index 4b27b5b..0000000 --- a/libtracing/relay.h +++ /dev/null @@ -1,367 +0,0 @@ -/* - * linux/include/linux/ltt-relay.h - * - * Copyright (C) 2002, 2003 - Tom Zanussi (zanussi@us.ibm.com), IBM Corp - * Copyright (C) 1999, 2000, 2001, 2002 - Karim Yaghmour (karim@opersys.com) - * Copyright (C) 2008 - Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) - * - * CONFIG_RELAY definitions and declarations - */ - -#ifndef _LINUX_LTT_RELAY_H -#define _LINUX_LTT_RELAY_H - -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -#include "kref.h" -#include "list.h" - -/* Needs a _much_ better name... */ -#define FIX_SIZE(x) ((((x) - 1) & PAGE_MASK) + PAGE_SIZE) - -/* - * Tracks changes to rchan/rchan_buf structs - */ -#define LTT_RELAY_CHANNEL_VERSION 8 - -struct rchan_buf; - -struct buf_page { - struct page *page; - struct rchan_buf *buf; /* buffer the page belongs to */ - size_t offset; /* page offset in the buffer */ - struct list_head list; /* buffer linked list */ -}; - -/* - * Per-cpu relay channel buffer - */ -struct rchan_buf { - struct rchan *chan; /* associated channel */ -//ust// wait_queue_head_t read_wait; /* reader wait queue */ -//ust// struct timer_list timer; /* reader wake-up timer */ -//ust// struct dentry *dentry; /* channel file dentry */ - struct kref kref; /* channel buffer refcount */ -//ust// struct list_head pages; /* list of buffer pages */ - void *buf_data; //ust// - size_t buf_size; -//ust// struct buf_page *wpage; /* current write page (cache) */ -//ust// struct buf_page *hpage[2]; /* current subbuf header page (cache) */ -//ust// struct buf_page *rpage; /* current subbuf read page (cache) */ -//ust// unsigned int page_count; /* number of current buffer pages */ - unsigned int finalized; /* buffer has been finalized */ -//ust// unsigned int cpu; /* this buf's cpu */ - int shmid; /* the shmid of the buffer data pages */ -} ____cacheline_aligned; - -/* - * Relay channel data structure - */ -struct rchan { - u32 version; /* the version of this struct */ - size_t subbuf_size; /* sub-buffer size */ - size_t n_subbufs; /* number of sub-buffers per buffer */ - size_t alloc_size; /* total buffer size allocated */ - struct rchan_callbacks *cb; /* client callbacks */ - struct kref kref; /* channel refcount */ - void *private_data; /* for user-defined data */ -//ust// struct rchan_buf *buf[NR_CPUS]; /* per-cpu channel buffers */ - struct rchan_buf *buf; - struct list_head list; /* for channel list */ - struct dentry *parent; /* parent dentry passed to open */ - int subbuf_size_order; /* order of sub-buffer size */ -//ust// char base_filename[NAME_MAX]; /* saved base filename */ -}; - -/* - * Relay channel client callbacks - */ -struct rchan_callbacks { - /* - * subbuf_start - called on buffer-switch to a new sub-buffer - * @buf: the channel buffer containing the new sub-buffer - * @subbuf: the start of the new sub-buffer - * @prev_subbuf: the start of the previous sub-buffer - * @prev_padding: unused space at the end of previous sub-buffer - * - * The client should return 1 to continue logging, 0 to stop - * logging. - * - * NOTE: subbuf_start will also be invoked when the buffer is - * created, so that the first sub-buffer can be initialized - * if necessary. In this case, prev_subbuf will be NULL. - * - * NOTE: the client can reserve bytes at the beginning of the new - * sub-buffer by calling subbuf_start_reserve() in this callback. - */ - int (*subbuf_start) (struct rchan_buf *buf, - void *subbuf, - void *prev_subbuf, - size_t prev_padding); - - /* - * create_buf_file - create file to represent a relay channel buffer - * @filename: the name of the file to create - * @parent: the parent of the file to create - * @mode: the mode of the file to create - * @buf: the channel buffer - * - * Called during relay_open(), once for each per-cpu buffer, - * to allow the client to create a file to be used to - * represent the corresponding channel buffer. If the file is - * created outside of relay, the parent must also exist in - * that filesystem. - * - * The callback should return the dentry of the file created - * to represent the relay buffer. - * - * Setting the is_global outparam to a non-zero value will - * cause relay_open() to create a single global buffer rather - * than the default set of per-cpu buffers. - * - * See Documentation/filesystems/relayfs.txt for more info. - */ - struct dentry *(*create_buf_file)(const char *filename, - struct dentry *parent, - int mode, - struct rchan_buf *buf); - - /* - * remove_buf_file - remove file representing a relay channel buffer - * @dentry: the dentry of the file to remove - * - * Called during relay_close(), once for each per-cpu buffer, - * to allow the client to remove a file used to represent a - * channel buffer. - * - * The callback should return 0 if successful, negative if not. - */ -//ust// int (*remove_buf_file)(struct rchan_buf *buf); -}; - -extern struct buf_page *ltt_relay_find_prev_page(struct rchan_buf *buf, - struct buf_page *page, size_t offset, ssize_t diff_offset); - -extern struct buf_page *ltt_relay_find_next_page(struct rchan_buf *buf, - struct buf_page *page, size_t offset, ssize_t diff_offset); - -extern void _ltt_relay_write(struct rchan_buf *buf, size_t offset, - const void *src, size_t len, ssize_t cpy); - -extern int ltt_relay_read(struct rchan_buf *buf, size_t offset, - void *dest, size_t len); - -extern struct buf_page *ltt_relay_read_get_page(struct rchan_buf *buf, - size_t offset); - -/* - * Return the address where a given offset is located. - * Should be used to get the current subbuffer header pointer. Given we know - * it's never on a page boundary, it's safe to write directly to this address, - * as long as the write is never bigger than a page size. - */ -extern void *ltt_relay_offset_address(struct rchan_buf *buf, - size_t offset); - -/* - * Find the page containing "offset". Cache it if it is after the currently - * cached page. - */ -static inline struct buf_page *ltt_relay_cache_page(struct rchan_buf *buf, - struct buf_page **page_cache, - struct buf_page *page, size_t offset) -{ - ssize_t diff_offset; - ssize_t half_buf_size = buf->chan->alloc_size >> 1; - - /* - * Make sure this is the page we want to write into. The current - * page is changed concurrently by other writers. [wrh]page are - * used as a cache remembering the last page written - * to/read/looked up for header address. No synchronization; - * could have to find the previous page is a nested write - * occured. Finding the right page is done by comparing the - * dest_offset with the buf_page offsets. - * When at the exact opposite of the buffer, bias towards forward search - * because it will be cached. - */ - - diff_offset = (ssize_t)offset - (ssize_t)page->offset; - if (diff_offset <= -(ssize_t)half_buf_size) - diff_offset += buf->chan->alloc_size; - else if (diff_offset > half_buf_size) - diff_offset -= buf->chan->alloc_size; - - if (unlikely(diff_offset >= (ssize_t)PAGE_SIZE)) { - page = ltt_relay_find_next_page(buf, page, offset, diff_offset); - *page_cache = page; - } else if (unlikely(diff_offset < 0)) { - page = ltt_relay_find_prev_page(buf, page, offset, diff_offset); - } - return page; -} - -//ust// #ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS - static inline void ltt_relay_do_copy(void *dest, const void *src, size_t len) -{ - switch (len) { - case 0: break; - case 1: *(u8 *)dest = *(const u8 *)src; - break; - case 2: *(u16 *)dest = *(const u16 *)src; - break; - case 4: *(u32 *)dest = *(const u32 *)src; - break; -//ust// #if (BITS_PER_LONG == 64) - case 8: *(u64 *)dest = *(const u64 *)src; - break; -//ust// #endif - default: - memcpy(dest, src, len); - } -} -//ust// #else -//ust// /* -//ust// * Returns whether the dest and src addresses are aligned on -//ust// * min(sizeof(void *), len). Call this with statically known len for efficiency. -//ust// */ -//ust// static inline int addr_aligned(const void *dest, const void *src, size_t len) -//ust// { -//ust// if (ltt_align((size_t)dest, len)) -//ust// return 0; -//ust// if (ltt_align((size_t)src, len)) -//ust// return 0; -//ust// return 1; -//ust// } -//ust// -//ust// static inline void ltt_relay_do_copy(void *dest, const void *src, size_t len) -//ust// { -//ust// switch (len) { -//ust// case 0: break; -//ust// case 1: *(u8 *)dest = *(const u8 *)src; -//ust// break; -//ust// case 2: if (unlikely(!addr_aligned(dest, src, 2))) -//ust// goto memcpy_fallback; -//ust// *(u16 *)dest = *(const u16 *)src; -//ust// break; -//ust// case 4: if (unlikely(!addr_aligned(dest, src, 4))) -//ust// goto memcpy_fallback; -//ust// *(u32 *)dest = *(const u32 *)src; -//ust// break; -//ust// #if (BITS_PER_LONG == 64) -//ust// case 8: if (unlikely(!addr_aligned(dest, src, 8))) -//ust// goto memcpy_fallback; -//ust// *(u64 *)dest = *(const u64 *)src; -//ust// break; -//ust// #endif -//ust// default: -//ust// goto memcpy_fallback; -//ust// } -//ust// return; -//ust// memcpy_fallback: -//ust// memcpy(dest, src, len); -//ust// } -//ust// #endif - -static inline int ltt_relay_write(struct rchan_buf *buf, size_t offset, - const void *src, size_t len) -{ -//ust// struct buf_page *page; -//ust// ssize_t pagecpy; -//ust// -//ust// offset &= buf->chan->alloc_size - 1; -//ust// page = buf->wpage; -//ust// -//ust// page = ltt_relay_cache_page(buf, &buf->wpage, page, offset); -//ust// pagecpy = min_t(size_t, len, PAGE_SIZE - (offset & ~PAGE_MASK)); -//ust// ltt_relay_do_copy(page_address(page->page) -//ust// + (offset & ~PAGE_MASK), src, pagecpy); -//ust// -//ust// if (unlikely(len != pagecpy)) -//ust// _ltt_relay_write(buf, offset, src, len, page, pagecpy); -//ust// return len; - - - size_t cpy; - cpy = min_t(size_t, len, buf->buf_size - offset); - ltt_relay_do_copy(buf->buf_data + offset, src, cpy); - - if (unlikely(len != cpy)) - _ltt_relay_write(buf, offset, src, len, cpy); - return len; -} - -/* - * CONFIG_LTT_RELAY kernel API, ltt/ltt-relay-alloc.c - */ - -struct rchan *ltt_relay_open(const char *base_filename, - struct dentry *parent, - size_t subbuf_size, - size_t n_subbufs, - void *private_data); -extern void ltt_relay_close(struct rchan *chan); - -/* - * exported ltt_relay file operations, ltt/ltt-relay-alloc.c - */ -extern const struct file_operations ltt_relay_file_operations; - - -/* LTTng lockless logging buffer info */ -struct ltt_channel_buf_struct { - /* First 32 bytes cache-hot cacheline */ - local_t offset; /* Current offset in the buffer */ -//ust// local_t *commit_count; /* Commit count per sub-buffer */ - atomic_long_t consumed; /* - * Current offset in the buffer - * standard atomic access (shared) - */ - unsigned long last_tsc; /* - * Last timestamp written in the buffer. - */ - /* End of first 32 bytes cacheline */ - atomic_long_t active_readers; /* - * Active readers count - * standard atomic access (shared) - */ - local_t events_lost; - local_t corrupted_subbuffers; -//ust// spinlock_t full_lock; /* -//ust// * buffer full condition spinlock, only -//ust// * for userspace tracing blocking mode -//ust// * synchronization with reader. -//ust// */ -//ust// wait_queue_head_t write_wait; /* -//ust// * Wait queue for blocking user space -//ust// * writers -//ust// */ -//ust// atomic_t wakeup_readers; /* Boolean : wakeup readers waiting ? */ - /* one byte is written to this pipe when data is available, in order - to wake the consumer */ - /* portability: Single byte writes must be as quick as possible. The kernel-side - buffer must be large enough so the writer doesn't block. From the pipe(7) - man page: Since linux 2.6.11, the pipe capacity is 65536 bytes. */ - int data_ready_fd_write; - /* the reading end of the pipe */ - int data_ready_fd_read; - - /* commit count per subbuffer; must be at end of struct */ - local_t commit_count[0] ____cacheline_aligned; -} ____cacheline_aligned; - -int ltt_do_get_subbuf(struct rchan_buf *buf, struct ltt_channel_buf_struct *ltt_buf, long *pconsumed_old); - -int ltt_do_put_subbuf(struct rchan_buf *buf, struct ltt_channel_buf_struct *ltt_buf, u32 uconsumed_old); - - -#endif /* _LINUX_LTT_RELAY_H */ - diff --git a/libtracing/serialize.c b/libtracing/serialize.c deleted file mode 100644 index bef0178..0000000 --- a/libtracing/serialize.c +++ /dev/null @@ -1,696 +0,0 @@ -/* - * LTTng serializing code. - * - * Copyright Mathieu Desnoyers, March 2007. - * - * Licensed under the GPLv2. - * - * See this discussion about weirdness about passing va_list and then va_list to - * functions. (related to array argument passing). va_list seems to be - * implemented as an array on x86_64, but not on i386... This is why we pass a - * va_list * to ltt_vtrace. - */ - -#include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -#include -#include -#include "kernelcompat.h" -#include "relay.h" -#include "tracer.h" -#include "list.h" -#include "usterr.h" - -enum ltt_type { - LTT_TYPE_SIGNED_INT, - LTT_TYPE_UNSIGNED_INT, - LTT_TYPE_STRING, - LTT_TYPE_NONE, -}; - -#define LTT_ATTRIBUTE_NETWORK_BYTE_ORDER (1<<1) - -/* - * Inspired from vsnprintf - * - * The serialization format string supports the basic printf format strings. - * In addition, it defines new formats that can be used to serialize more - * complex/non portable data structures. - * - * Typical use: - * - * field_name %ctype - * field_name #tracetype %ctype - * field_name #tracetype %ctype1 %ctype2 ... - * - * A conversion is performed between format string types supported by GCC and - * the trace type requested. GCC type is used to perform type checking on format - * strings. Trace type is used to specify the exact binary representation - * in the trace. A mapping is done between one or more GCC types to one trace - * type. Sign extension, if required by the conversion, is performed following - * the trace type. - * - * If a gcc format is not declared with a trace format, the gcc format is - * also used as binary representation in the trace. - * - * Strings are supported with %s. - * A single tracetype (sequence) can take multiple c types as parameter. - * - * c types: - * - * see printf(3). - * - * Note: to write a uint32_t in a trace, the following expression is recommended - * si it can be portable: - * - * ("#4u%lu", (unsigned long)var) - * - * trace types: - * - * Serialization specific formats : - * - * Fixed size integers - * #1u writes uint8_t - * #2u writes uint16_t - * #4u writes uint32_t - * #8u writes uint64_t - * #1d writes int8_t - * #2d writes int16_t - * #4d writes int32_t - * #8d writes int64_t - * i.e.: - * #1u%lu #2u%lu #4d%lu #8d%lu #llu%hu #d%lu - * - * * Attributes: - * - * n: (for network byte order) - * #ntracetype%ctype - * is written in the trace in network byte order. - * - * i.e.: #bn4u%lu, #n%lu, #b%u - * - * TODO (eventually) - * Variable length sequence - * #a #tracetype1 #tracetype2 %array_ptr %elem_size %num_elems - * In the trace: - * #a specifies that this is a sequence - * #tracetype1 is the type of elements in the sequence - * #tracetype2 is the type of the element count - * GCC input: - * array_ptr is a pointer to an array that contains members of size - * elem_size. - * num_elems is the number of elements in the array. - * i.e.: #a #lu #lu %p %lu %u - * - * Callback - * #k callback (taken from the probe data) - * The following % arguments are exepected by the callback - * - * i.e.: #a #lu #lu #k %p - * - * Note: No conversion is done from floats to integers, nor from integers to - * floats between c types and trace types. float conversion from double to float - * or from float to double is also not supported. - * - * REMOVE - * %*b expects sizeof(data), data - * where sizeof(data) is 1, 2, 4 or 8 - * - * Fixed length struct, union or array. - * FIXME: unable to extract those sizes statically. - * %*r expects sizeof(*ptr), ptr - * %*.*r expects sizeof(*ptr), __alignof__(*ptr), ptr - * struct and unions removed. - * Fixed length array: - * [%p]#a[len #tracetype] - * i.e.: [%p]#a[12 #lu] - * - * Variable length sequence - * %*.*:*v expects sizeof(*ptr), __alignof__(*ptr), elem_num, ptr - * where elem_num is the number of elements in the sequence - */ -static inline const char *parse_trace_type(const char *fmt, - char *trace_size, enum ltt_type *trace_type, - unsigned long *attributes) -{ - int qualifier; /* 'h', 'l', or 'L' for integer fields */ - /* 'z' support added 23/7/1999 S.H. */ - /* 'z' changed to 'Z' --davidm 1/25/99 */ - /* 't' added for ptrdiff_t */ - - /* parse attributes. */ -repeat: - switch (*fmt) { - case 'n': - *attributes |= LTT_ATTRIBUTE_NETWORK_BYTE_ORDER; - ++fmt; - goto repeat; - } - - /* get the conversion qualifier */ - qualifier = -1; - if (*fmt == 'h' || *fmt == 'l' || *fmt == 'L' || - *fmt == 'Z' || *fmt == 'z' || *fmt == 't' || - *fmt == 'S' || *fmt == '1' || *fmt == '2' || - *fmt == '4' || *fmt == 8) { - qualifier = *fmt; - ++fmt; - if (qualifier == 'l' && *fmt == 'l') { - qualifier = 'L'; - ++fmt; - } - } - - switch (*fmt) { - case 'c': - *trace_type = LTT_TYPE_UNSIGNED_INT; - *trace_size = sizeof(unsigned char); - goto parse_end; - case 's': - *trace_type = LTT_TYPE_STRING; - goto parse_end; - case 'p': - *trace_type = LTT_TYPE_UNSIGNED_INT; - *trace_size = sizeof(void *); - goto parse_end; - case 'd': - case 'i': - *trace_type = LTT_TYPE_SIGNED_INT; - break; - case 'o': - case 'u': - case 'x': - case 'X': - *trace_type = LTT_TYPE_UNSIGNED_INT; - break; - default: - if (!*fmt) - --fmt; - goto parse_end; - } - switch (qualifier) { - case 'L': - *trace_size = sizeof(long long); - break; - case 'l': - *trace_size = sizeof(long); - break; - case 'Z': - case 'z': - *trace_size = sizeof(size_t); - break; -//ust// case 't': -//ust// *trace_size = sizeof(ptrdiff_t); -//ust// break; - case 'h': - *trace_size = sizeof(short); - break; - case '1': - *trace_size = sizeof(uint8_t); - break; - case '2': - *trace_size = sizeof(uint16_t); - break; - case '4': - *trace_size = sizeof(uint32_t); - break; - case '8': - *trace_size = sizeof(uint64_t); - break; - default: - *trace_size = sizeof(int); - } - -parse_end: - return fmt; -} - -/* - * Restrictions: - * Field width and precision are *not* supported. - * %n not supported. - */ -static inline const char *parse_c_type(const char *fmt, - char *c_size, enum ltt_type *c_type) -{ - int qualifier; /* 'h', 'l', or 'L' for integer fields */ - /* 'z' support added 23/7/1999 S.H. */ - /* 'z' changed to 'Z' --davidm 1/25/99 */ - /* 't' added for ptrdiff_t */ - - /* process flags : ignore standard print formats for now. */ -repeat: - switch (*fmt) { - case '-': - case '+': - case ' ': - case '#': - case '0': - ++fmt; - goto repeat; - } - - /* get the conversion qualifier */ - qualifier = -1; - if (*fmt == 'h' || *fmt == 'l' || *fmt == 'L' || - *fmt == 'Z' || *fmt == 'z' || *fmt == 't' || - *fmt == 'S') { - qualifier = *fmt; - ++fmt; - if (qualifier == 'l' && *fmt == 'l') { - qualifier = 'L'; - ++fmt; - } - } - - switch (*fmt) { - case 'c': - *c_type = LTT_TYPE_UNSIGNED_INT; - *c_size = sizeof(unsigned char); - goto parse_end; - case 's': - *c_type = LTT_TYPE_STRING; - goto parse_end; - case 'p': - *c_type = LTT_TYPE_UNSIGNED_INT; - *c_size = sizeof(void *); - goto parse_end; - case 'd': - case 'i': - *c_type = LTT_TYPE_SIGNED_INT; - break; - case 'o': - case 'u': - case 'x': - case 'X': - *c_type = LTT_TYPE_UNSIGNED_INT; - break; - default: - if (!*fmt) - --fmt; - goto parse_end; - } - switch (qualifier) { - case 'L': - *c_size = sizeof(long long); - break; - case 'l': - *c_size = sizeof(long); - break; - case 'Z': - case 'z': - *c_size = sizeof(size_t); - break; -//ust// case 't': -//ust// *c_size = sizeof(ptrdiff_t); -//ust// break; - case 'h': - *c_size = sizeof(short); - break; - default: - *c_size = sizeof(int); - } - -parse_end: - return fmt; -} - -static inline size_t serialize_trace_data(struct rchan_buf *buf, - size_t buf_offset, - char trace_size, enum ltt_type trace_type, - char c_size, enum ltt_type c_type, - int *largest_align, va_list *args) -{ - union { - unsigned long v_ulong; - uint64_t v_uint64; - struct { - const char *s; - size_t len; - } v_string; - } tmp; - - /* - * Be careful about sign extension here. - * Sign extension is done with the destination (trace) type. - */ - switch (trace_type) { - case LTT_TYPE_SIGNED_INT: - switch (c_size) { - case 1: - tmp.v_ulong = (long)(int8_t)va_arg(*args, int); - break; - case 2: - tmp.v_ulong = (long)(int16_t)va_arg(*args, int); - break; - case 4: - tmp.v_ulong = (long)(int32_t)va_arg(*args, int); - break; - case 8: - tmp.v_uint64 = va_arg(*args, int64_t); - break; - default: - BUG(); - } - break; - case LTT_TYPE_UNSIGNED_INT: - switch (c_size) { - case 1: - tmp.v_ulong = (unsigned long)(uint8_t) - va_arg(*args, unsigned int); - break; - case 2: - tmp.v_ulong = (unsigned long)(uint16_t) - va_arg(*args, unsigned int); - break; - case 4: - tmp.v_ulong = (unsigned long)(uint32_t) - va_arg(*args, unsigned int); - break; - case 8: - tmp.v_uint64 = va_arg(*args, uint64_t); - break; - default: - BUG(); - } - break; - case LTT_TYPE_STRING: - tmp.v_string.s = va_arg(*args, const char *); - if ((unsigned long)tmp.v_string.s < PAGE_SIZE) - tmp.v_string.s = ""; - tmp.v_string.len = strlen(tmp.v_string.s)+1; - if (buf) - ltt_relay_write(buf, buf_offset, tmp.v_string.s, - tmp.v_string.len); - buf_offset += tmp.v_string.len; - goto copydone; - default: - BUG(); - } - - /* - * If trace_size is lower or equal to 4 bytes, there is no sign - * extension to do because we are already encoded in a long. Therefore, - * we can combine signed and unsigned ops. 4 bytes float also works - * with this, because we do a simple copy of 4 bytes into 4 bytes - * without manipulation (and we do not support conversion from integers - * to floats). - * It is also the case if c_size is 8 bytes, which is the largest - * possible integer. - */ - if (ltt_get_alignment()) { - buf_offset += ltt_align(buf_offset, trace_size); - if (largest_align) - *largest_align = max_t(int, *largest_align, trace_size); - } - if (trace_size <= 4 || c_size == 8) { - if (buf) { - switch (trace_size) { - case 1: - if (c_size == 8) - ltt_relay_write(buf, buf_offset, - (uint8_t[]){ (uint8_t)tmp.v_uint64 }, - sizeof(uint8_t)); - else - ltt_relay_write(buf, buf_offset, - (uint8_t[]){ (uint8_t)tmp.v_ulong }, - sizeof(uint8_t)); - break; - case 2: - if (c_size == 8) - ltt_relay_write(buf, buf_offset, - (uint16_t[]){ (uint16_t)tmp.v_uint64 }, - sizeof(uint16_t)); - else - ltt_relay_write(buf, buf_offset, - (uint16_t[]){ (uint16_t)tmp.v_ulong }, - sizeof(uint16_t)); - break; - case 4: - if (c_size == 8) - ltt_relay_write(buf, buf_offset, - (uint32_t[]){ (uint32_t)tmp.v_uint64 }, - sizeof(uint32_t)); - else - ltt_relay_write(buf, buf_offset, - (uint32_t[]){ (uint32_t)tmp.v_ulong }, - sizeof(uint32_t)); - break; - case 8: - /* - * c_size cannot be other than 8 here because - * trace_size > 4. - */ - ltt_relay_write(buf, buf_offset, - (uint64_t[]){ (uint64_t)tmp.v_uint64 }, - sizeof(uint64_t)); - break; - default: - BUG(); - } - } - buf_offset += trace_size; - goto copydone; - } else { - /* - * Perform sign extension. - */ - if (buf) { - switch (trace_type) { - case LTT_TYPE_SIGNED_INT: - ltt_relay_write(buf, buf_offset, - (int64_t[]){ (int64_t)tmp.v_ulong }, - sizeof(int64_t)); - break; - case LTT_TYPE_UNSIGNED_INT: - ltt_relay_write(buf, buf_offset, - (uint64_t[]){ (uint64_t)tmp.v_ulong }, - sizeof(uint64_t)); - break; - default: - BUG(); - } - } - buf_offset += trace_size; - goto copydone; - } - -copydone: - return buf_offset; -} - -notrace size_t ltt_serialize_data(struct rchan_buf *buf, size_t buf_offset, - struct ltt_serialize_closure *closure, - void *serialize_private, int *largest_align, - const char *fmt, va_list *args) -{ - char trace_size = 0, c_size = 0; /* - * 0 (unset), 1, 2, 4, 8 bytes. - */ - enum ltt_type trace_type = LTT_TYPE_NONE, c_type = LTT_TYPE_NONE; - unsigned long attributes = 0; - - for (; *fmt ; ++fmt) { - switch (*fmt) { - case '#': - /* tracetypes (#) */ - ++fmt; /* skip first '#' */ - if (*fmt == '#') /* Escaped ## */ - break; - attributes = 0; - fmt = parse_trace_type(fmt, &trace_size, &trace_type, - &attributes); - break; - case '%': - /* c types (%) */ - ++fmt; /* skip first '%' */ - if (*fmt == '%') /* Escaped %% */ - break; - fmt = parse_c_type(fmt, &c_size, &c_type); - /* - * Output c types if no trace types has been - * specified. - */ - if (!trace_size) - trace_size = c_size; - if (trace_type == LTT_TYPE_NONE) - trace_type = c_type; - if (c_type == LTT_TYPE_STRING) - trace_type = LTT_TYPE_STRING; - /* perform trace write */ - buf_offset = serialize_trace_data(buf, - buf_offset, trace_size, - trace_type, c_size, c_type, - largest_align, args); - trace_size = 0; - c_size = 0; - trace_type = LTT_TYPE_NONE; - c_size = LTT_TYPE_NONE; - attributes = 0; - break; - /* default is to skip the text, doing nothing */ - } - } - return buf_offset; -} -EXPORT_SYMBOL_GPL(ltt_serialize_data); - -/* - * Calculate data size - * Assume that the padding for alignment starts at a sizeof(void *) address. - */ -static notrace size_t ltt_get_data_size(struct ltt_serialize_closure *closure, - void *serialize_private, int *largest_align, - const char *fmt, va_list *args) -{ - ltt_serialize_cb cb = closure->callbacks[0]; - closure->cb_idx = 0; - return (size_t)cb(NULL, 0, closure, serialize_private, - largest_align, fmt, args); -} - -static notrace -void ltt_write_event_data(struct rchan_buf *buf, size_t buf_offset, - struct ltt_serialize_closure *closure, - void *serialize_private, int largest_align, - const char *fmt, va_list *args) -{ - ltt_serialize_cb cb = closure->callbacks[0]; - closure->cb_idx = 0; - buf_offset += ltt_align(buf_offset, largest_align); - cb(buf, buf_offset, closure, serialize_private, NULL, fmt, args); -} - - -notrace void ltt_vtrace(const struct marker *mdata, void *probe_data, - void *call_data, const char *fmt, va_list *args) -{ - int largest_align, ret; - struct ltt_active_marker *pdata; - uint16_t eID; - size_t data_size, slot_size; - unsigned int chan_index; - struct ltt_channel_struct *channel; - struct ltt_trace_struct *trace, *dest_trace = NULL; - struct rchan_buf *buf; - void *transport_data; - u64 tsc; - long buf_offset; - va_list args_copy; - struct ltt_serialize_closure closure; - struct ltt_probe_private_data *private_data = call_data; - void *serialize_private = NULL; - int cpu; - unsigned int rflags; - - /* - * This test is useful for quickly exiting static tracing when no trace - * is active. We expect to have an active trace when we get here. - */ - if (unlikely(ltt_traces.num_active_traces == 0)) - return; - - rcu_read_lock_sched_notrace(); - cpu = smp_processor_id(); -//ust// __get_cpu_var(ltt_nesting)++; - ltt_nesting++; - - pdata = (struct ltt_active_marker *)probe_data; - eID = mdata->event_id; - chan_index = mdata->channel_id; - closure.callbacks = pdata->probe->callbacks; - - if (unlikely(private_data)) { - dest_trace = private_data->trace; - if (private_data->serializer) - closure.callbacks = &private_data->serializer; - serialize_private = private_data->serialize_private; - } - - va_copy(args_copy, *args); - /* - * Assumes event payload to start on largest_align alignment. - */ - largest_align = 1; /* must be non-zero for ltt_align */ - data_size = ltt_get_data_size(&closure, serialize_private, - &largest_align, fmt, &args_copy); - largest_align = min_t(int, largest_align, sizeof(void *)); - va_end(args_copy); - - /* Iterate on each trace */ - list_for_each_entry_rcu(trace, <t_traces.head, list) { - /* - * Expect the filter to filter out events. If we get here, - * we went through tracepoint activation as a first step. - */ - if (unlikely(dest_trace && trace != dest_trace)) - continue; - if (unlikely(!trace->active)) - continue; - if (unlikely(!ltt_run_filter(trace, eID))) - continue; -#ifdef CONFIG_LTT_DEBUG_EVENT_SIZE - rflags = LTT_RFLAG_ID_SIZE; -#else - if (unlikely(eID >= LTT_FREE_EVENTS)) - rflags = LTT_RFLAG_ID; - else - rflags = 0; -#endif - /* - * Skip channels added after trace creation. - */ - if (unlikely(chan_index >= trace->nr_channels)) - continue; - channel = &trace->channels[chan_index]; - if (!channel->active) - continue; - - /* reserve space : header and data */ - ret = ltt_reserve_slot(trace, channel, &transport_data, - data_size, &slot_size, &buf_offset, - &tsc, &rflags, - largest_align); - if (unlikely(ret < 0)) - continue; /* buffer full */ - - va_copy(args_copy, *args); - /* FIXME : could probably encapsulate transport better. */ -//ust// buf = ((struct rchan *)channel->trans_channel_data)->buf[cpu]; - buf = ((struct rchan *)channel->trans_channel_data)->buf; - /* Out-of-order write : header and data */ - buf_offset = ltt_write_event_header(trace, - channel, buf, buf_offset, - eID, data_size, tsc, rflags); - ltt_write_event_data(buf, buf_offset, &closure, - serialize_private, - largest_align, fmt, &args_copy); - va_end(args_copy); - /* Out-of-order commit */ - ltt_commit_slot(channel, &transport_data, buf_offset, - slot_size); - printf("just commited event at offset %d and size %d\n", buf_offset, slot_size); - } -//ust// __get_cpu_var(ltt_nesting)--; - ltt_nesting--; - rcu_read_unlock_sched_notrace(); -} -EXPORT_SYMBOL_GPL(ltt_vtrace); - -notrace void ltt_trace(const struct marker *mdata, void *probe_data, - void *call_data, const char *fmt, ...) -{ - va_list args; - - va_start(args, fmt); - ltt_vtrace(mdata, probe_data, call_data, fmt, &args); - va_end(args); -} -EXPORT_SYMBOL_GPL(ltt_trace); - -//ust// MODULE_LICENSE("GPL"); -//ust// MODULE_AUTHOR("Mathieu Desnoyers"); -//ust// MODULE_DESCRIPTION("Linux Trace Toolkit Next Generation Serializer"); diff --git a/libtracing/tracer.c b/libtracing/tracer.c deleted file mode 100644 index b807a50..0000000 --- a/libtracing/tracer.c +++ /dev/null @@ -1,1148 +0,0 @@ -/* - * ltt/ltt-tracer.c - * - * (C) Copyright 2005-2008 - - * Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) - * - * Tracing management internal kernel API. Trace buffer allocation/free, tracing - * start/stop. - * - * Author: - * Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) - * - * Inspired from LTT : - * Karim Yaghmour (karim@opersys.com) - * Tom Zanussi (zanussi@us.ibm.com) - * Bob Wisniewski (bob@watson.ibm.com) - * And from K42 : - * Bob Wisniewski (bob@watson.ibm.com) - * - * Changelog: - * 22/09/06, Move to the marker/probes mechanism. - * 19/10/05, Complete lockless mechanism. - * 27/05/05, Modular redesign and rewrite. - */ - -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -//ust// #include -#include "kernelcompat.h" -#include "tracercore.h" -#include "tracer.h" -#include "usterr.h" - -//ust// static void async_wakeup(unsigned long data); -//ust// -//ust// static DEFINE_TIMER(ltt_async_wakeup_timer, async_wakeup, 0, 0); - -/* Default callbacks for modules */ -notrace int ltt_filter_control_default(enum ltt_filter_control_msg msg, - struct ltt_trace_struct *trace) -{ - return 0; -} - -int ltt_statedump_default(struct ltt_trace_struct *trace) -{ - return 0; -} - -/* Callbacks for registered modules */ - -int (*ltt_filter_control_functor) - (enum ltt_filter_control_msg msg, struct ltt_trace_struct *trace) = - ltt_filter_control_default; -struct module *ltt_filter_control_owner; - -/* These function pointers are protected by a trace activation check */ -struct module *ltt_run_filter_owner; -int (*ltt_statedump_functor)(struct ltt_trace_struct *trace) = - ltt_statedump_default; -struct module *ltt_statedump_owner; - -struct chan_info_struct { - const char *name; - unsigned int def_subbufsize; - unsigned int def_subbufcount; -} chan_infos[] = { - [LTT_CHANNEL_METADATA] = { - LTT_METADATA_CHANNEL, - LTT_DEFAULT_SUBBUF_SIZE_LOW, - LTT_DEFAULT_N_SUBBUFS_LOW, - }, - [LTT_CHANNEL_UST] = { - LTT_UST_CHANNEL, - LTT_DEFAULT_SUBBUF_SIZE_HIGH, - LTT_DEFAULT_N_SUBBUFS_HIGH, - }, -}; - -static enum ltt_channels get_channel_type_from_name(const char *name) -{ - int i; - - if (!name) - return LTT_CHANNEL_UST; - - for (i = 0; i < ARRAY_SIZE(chan_infos); i++) - if (chan_infos[i].name && !strcmp(name, chan_infos[i].name)) - return (enum ltt_channels)i; - - return LTT_CHANNEL_UST; -} - -/** - * ltt_module_register - LTT module registration - * @name: module type - * @function: callback to register - * @owner: module which owns the callback - * - * The module calling this registration function must ensure that no - * trap-inducing code will be executed by "function". E.g. vmalloc_sync_all() - * must be called between a vmalloc and the moment the memory is made visible to - * "function". This registration acts as a vmalloc_sync_all. Therefore, only if - * the module allocates virtual memory after its registration must it - * synchronize the TLBs. - */ -//ust// int ltt_module_register(enum ltt_module_function name, void *function, -//ust// struct module *owner) -//ust// { -//ust// int ret = 0; -//ust// -//ust// /* -//ust// * Make sure no page fault can be triggered by the module about to be -//ust// * registered. We deal with this here so we don't have to call -//ust// * vmalloc_sync_all() in each module's init. -//ust// */ -//ust// vmalloc_sync_all(); -//ust// -//ust// switch (name) { -//ust// case LTT_FUNCTION_RUN_FILTER: -//ust// if (ltt_run_filter_owner != NULL) { -//ust// ret = -EEXIST; -//ust// goto end; -//ust// } -//ust// ltt_filter_register((ltt_run_filter_functor)function); -//ust// ltt_run_filter_owner = owner; -//ust// break; -//ust// case LTT_FUNCTION_FILTER_CONTROL: -//ust// if (ltt_filter_control_owner != NULL) { -//ust// ret = -EEXIST; -//ust// goto end; -//ust// } -//ust// ltt_filter_control_functor = -//ust// (int (*)(enum ltt_filter_control_msg, -//ust// struct ltt_trace_struct *))function; -//ust// ltt_filter_control_owner = owner; -//ust// break; -//ust// case LTT_FUNCTION_STATEDUMP: -//ust// if (ltt_statedump_owner != NULL) { -//ust// ret = -EEXIST; -//ust// goto end; -//ust// } -//ust// ltt_statedump_functor = -//ust// (int (*)(struct ltt_trace_struct *))function; -//ust// ltt_statedump_owner = owner; -//ust// break; -//ust// } -//ust// -//ust// end: -//ust// -//ust// return ret; -//ust// } -//ust// EXPORT_SYMBOL_GPL(ltt_module_register); - -/** - * ltt_module_unregister - LTT module unregistration - * @name: module type - */ -//ust// void ltt_module_unregister(enum ltt_module_function name) -//ust// { -//ust// switch (name) { -//ust// case LTT_FUNCTION_RUN_FILTER: -//ust// ltt_filter_unregister(); -//ust// ltt_run_filter_owner = NULL; -//ust// /* Wait for preempt sections to finish */ -//ust// synchronize_sched(); -//ust// break; -//ust// case LTT_FUNCTION_FILTER_CONTROL: -//ust// ltt_filter_control_functor = ltt_filter_control_default; -//ust// ltt_filter_control_owner = NULL; -//ust// break; -//ust// case LTT_FUNCTION_STATEDUMP: -//ust// ltt_statedump_functor = ltt_statedump_default; -//ust// ltt_statedump_owner = NULL; -//ust// break; -//ust// } -//ust// -//ust// } -//ust// EXPORT_SYMBOL_GPL(ltt_module_unregister); - -static LIST_HEAD(ltt_transport_list); - -/** - * ltt_transport_register - LTT transport registration - * @transport: transport structure - * - * Registers a transport which can be used as output to extract the data out of - * LTTng. The module calling this registration function must ensure that no - * trap-inducing code will be executed by the transport functions. E.g. - * vmalloc_sync_all() must be called between a vmalloc and the moment the memory - * is made visible to the transport function. This registration acts as a - * vmalloc_sync_all. Therefore, only if the module allocates virtual memory - * after its registration must it synchronize the TLBs. - */ -void ltt_transport_register(struct ltt_transport *transport) -{ - /* - * Make sure no page fault can be triggered by the module about to be - * registered. We deal with this here so we don't have to call - * vmalloc_sync_all() in each module's init. - */ -//ust// vmalloc_sync_all(); - - ltt_lock_traces(); - list_add_tail(&transport->node, <t_transport_list); - ltt_unlock_traces(); -} -//ust// EXPORT_SYMBOL_GPL(ltt_transport_register); - -/** - * ltt_transport_unregister - LTT transport unregistration - * @transport: transport structure - */ -void ltt_transport_unregister(struct ltt_transport *transport) -{ - ltt_lock_traces(); - list_del(&transport->node); - ltt_unlock_traces(); -} -//ust// EXPORT_SYMBOL_GPL(ltt_transport_unregister); - -static inline int is_channel_overwrite(enum ltt_channels chan, - enum trace_mode mode) -{ - switch (mode) { - case LTT_TRACE_NORMAL: - return 0; - case LTT_TRACE_FLIGHT: - switch (chan) { - case LTT_CHANNEL_METADATA: - return 0; - default: - return 1; - } - case LTT_TRACE_HYBRID: - switch (chan) { - case LTT_CHANNEL_METADATA: - return 0; - default: - return 1; - } - default: - return 0; - } -} - -/** - * ltt_write_trace_header - Write trace header - * @trace: Trace information - * @header: Memory address where the information must be written to - */ -void notrace ltt_write_trace_header(struct ltt_trace_struct *trace, - struct ltt_subbuffer_header *header) -{ - header->magic_number = LTT_TRACER_MAGIC_NUMBER; - header->major_version = LTT_TRACER_VERSION_MAJOR; - header->minor_version = LTT_TRACER_VERSION_MINOR; - header->arch_size = sizeof(void *); - header->alignment = ltt_get_alignment(); - header->start_time_sec = trace->start_time.tv_sec; - header->start_time_usec = trace->start_time.tv_usec; - header->start_freq = trace->start_freq; - header->freq_scale = trace->freq_scale; -} -//ust// EXPORT_SYMBOL_GPL(ltt_write_trace_header); - -static void trace_async_wakeup(struct ltt_trace_struct *trace) -{ - int i; - struct ltt_channel_struct *chan; - - /* Must check each channel for pending read wakeup */ - for (i = 0; i < trace->nr_channels; i++) { - chan = &trace->channels[i]; - if (chan->active) - trace->ops->wakeup_channel(chan); - } -} - -//ust// /* Timer to send async wakeups to the readers */ -//ust// static void async_wakeup(unsigned long data) -//ust// { -//ust// struct ltt_trace_struct *trace; -//ust// -//ust// /* -//ust// * PREEMPT_RT does not allow spinlocks to be taken within preempt -//ust// * disable sections (spinlock taken in wake_up). However, mainline won't -//ust// * allow mutex to be taken in interrupt context. Ugly. -//ust// * A proper way to do this would be to turn the timer into a -//ust// * periodically woken up thread, but it adds to the footprint. -//ust// */ -//ust// #ifndef CONFIG_PREEMPT_RT -//ust// rcu_read_lock_sched(); -//ust// #else -//ust// ltt_lock_traces(); -//ust// #endif -//ust// list_for_each_entry_rcu(trace, <t_traces.head, list) { -//ust// trace_async_wakeup(trace); -//ust// } -//ust// #ifndef CONFIG_PREEMPT_RT -//ust// rcu_read_unlock_sched(); -//ust// #else -//ust// ltt_unlock_traces(); -//ust// #endif -//ust// -//ust// mod_timer(<t_async_wakeup_timer, jiffies + LTT_PERCPU_TIMER_INTERVAL); -//ust// } - -/** - * _ltt_trace_find - find a trace by given name. - * trace_name: trace name - * - * Returns a pointer to the trace structure, NULL if not found. - */ -struct ltt_trace_struct *_ltt_trace_find(const char *trace_name) -{ - struct ltt_trace_struct *trace; - - list_for_each_entry(trace, <t_traces.head, list) - if (!strncmp(trace->trace_name, trace_name, NAME_MAX)) - return trace; - - return NULL; -} - -/* _ltt_trace_find_setup : - * find a trace in setup list by given name. - * - * Returns a pointer to the trace structure, NULL if not found. - */ -struct ltt_trace_struct *_ltt_trace_find_setup(const char *trace_name) -{ - struct ltt_trace_struct *trace; - - list_for_each_entry(trace, <t_traces.setup_head, list) - if (!strncmp(trace->trace_name, trace_name, NAME_MAX)) - return trace; - - return NULL; -} -//ust// EXPORT_SYMBOL_GPL(_ltt_trace_find_setup); - -/** - * ltt_release_transport - Release an LTT transport - * @kref : reference count on the transport - */ -void ltt_release_transport(struct kref *kref) -{ - struct ltt_trace_struct *trace = container_of(kref, - struct ltt_trace_struct, ltt_transport_kref); -//ust// trace->ops->remove_dirs(trace); -} -//ust// EXPORT_SYMBOL_GPL(ltt_release_transport); - -/** - * ltt_release_trace - Release a LTT trace - * @kref : reference count on the trace - */ -void ltt_release_trace(struct kref *kref) -{ - struct ltt_trace_struct *trace = container_of(kref, - struct ltt_trace_struct, kref); - ltt_channels_trace_free(trace->channels); - kfree(trace); -} -//ust// EXPORT_SYMBOL_GPL(ltt_release_trace); - -static inline void prepare_chan_size_num(unsigned int *subbuf_size, - unsigned int *n_subbufs) -{ - *subbuf_size = 1 << get_count_order(*subbuf_size); - *n_subbufs = 1 << get_count_order(*n_subbufs); - - /* Subbuf size and number must both be power of two */ - WARN_ON(hweight32(*subbuf_size) != 1); - WARN_ON(hweight32(*n_subbufs) != 1); -} - -int _ltt_trace_setup(const char *trace_name) -{ - int err = 0; - struct ltt_trace_struct *new_trace = NULL; - int metadata_index; - unsigned int chan; - enum ltt_channels chantype; - - if (_ltt_trace_find_setup(trace_name)) { - printk(KERN_ERR "LTT : Trace name %s already used.\n", - trace_name); - err = -EEXIST; - goto traces_error; - } - - if (_ltt_trace_find(trace_name)) { - printk(KERN_ERR "LTT : Trace name %s already used.\n", - trace_name); - err = -EEXIST; - goto traces_error; - } - - new_trace = kzalloc(sizeof(struct ltt_trace_struct), GFP_KERNEL); - if (!new_trace) { - printk(KERN_ERR - "LTT : Unable to allocate memory for trace %s\n", - trace_name); - err = -ENOMEM; - goto traces_error; - } - strncpy(new_trace->trace_name, trace_name, NAME_MAX); - new_trace->channels = ltt_channels_trace_alloc(&new_trace->nr_channels, - 0, 1); - if (!new_trace->channels) { - printk(KERN_ERR - "LTT : Unable to allocate memory for chaninfo %s\n", - trace_name); - err = -ENOMEM; - goto trace_free; - } - - /* - * Force metadata channel to active, no overwrite. - */ - metadata_index = ltt_channels_get_index_from_name("metadata"); - WARN_ON(metadata_index < 0); - new_trace->channels[metadata_index].overwrite = 0; - new_trace->channels[metadata_index].active = 1; - - /* - * Set hardcoded tracer defaults for some channels - */ - for (chan = 0; chan < new_trace->nr_channels; chan++) { - if (!(new_trace->channels[chan].active)) - continue; - - chantype = get_channel_type_from_name( - ltt_channels_get_name_from_index(chan)); - new_trace->channels[chan].subbuf_size = - chan_infos[chantype].def_subbufsize; - new_trace->channels[chan].subbuf_cnt = - chan_infos[chantype].def_subbufcount; - } - - list_add(&new_trace->list, <t_traces.setup_head); - return 0; - -trace_free: - kfree(new_trace); -traces_error: - return err; -} -//ust// EXPORT_SYMBOL_GPL(_ltt_trace_setup); - - -int ltt_trace_setup(const char *trace_name) -{ - int ret; - ltt_lock_traces(); - ret = _ltt_trace_setup(trace_name); - ltt_unlock_traces(); - return ret; -} -//ust// EXPORT_SYMBOL_GPL(ltt_trace_setup); - -/* must be called from within a traces lock. */ -static void _ltt_trace_free(struct ltt_trace_struct *trace) -{ - list_del(&trace->list); - kfree(trace); -} - -int ltt_trace_set_type(const char *trace_name, const char *trace_type) -{ - int err = 0; - struct ltt_trace_struct *trace; - struct ltt_transport *tran_iter, *transport = NULL; - - ltt_lock_traces(); - - trace = _ltt_trace_find_setup(trace_name); - if (!trace) { - printk(KERN_ERR "LTT : Trace not found %s\n", trace_name); - err = -ENOENT; - goto traces_error; - } - - list_for_each_entry(tran_iter, <t_transport_list, node) { - if (!strcmp(tran_iter->name, trace_type)) { - transport = tran_iter; - break; - } - } - if (!transport) { - printk(KERN_ERR "LTT : Transport %s is not present.\n", - trace_type); - err = -EINVAL; - goto traces_error; - } - - trace->transport = transport; - -traces_error: - ltt_unlock_traces(); - return err; -} -//ust// EXPORT_SYMBOL_GPL(ltt_trace_set_type); - -int ltt_trace_set_channel_subbufsize(const char *trace_name, - const char *channel_name, unsigned int size) -{ - int err = 0; - struct ltt_trace_struct *trace; - int index; - - ltt_lock_traces(); - - trace = _ltt_trace_find_setup(trace_name); - if (!trace) { - printk(KERN_ERR "LTT : Trace not found %s\n", trace_name); - err = -ENOENT; - goto traces_error; - } - - index = ltt_channels_get_index_from_name(channel_name); - if (index < 0) { - printk(KERN_ERR "LTT : Channel %s not found\n", channel_name); - err = -ENOENT; - goto traces_error; - } - trace->channels[index].subbuf_size = size; - -traces_error: - ltt_unlock_traces(); - return err; -} -//ust// EXPORT_SYMBOL_GPL(ltt_trace_set_channel_subbufsize); - -int ltt_trace_set_channel_subbufcount(const char *trace_name, - const char *channel_name, unsigned int cnt) -{ - int err = 0; - struct ltt_trace_struct *trace; - int index; - - ltt_lock_traces(); - - trace = _ltt_trace_find_setup(trace_name); - if (!trace) { - printk(KERN_ERR "LTT : Trace not found %s\n", trace_name); - err = -ENOENT; - goto traces_error; - } - - index = ltt_channels_get_index_from_name(channel_name); - if (index < 0) { - printk(KERN_ERR "LTT : Channel %s not found\n", channel_name); - err = -ENOENT; - goto traces_error; - } - trace->channels[index].subbuf_cnt = cnt; - -traces_error: - ltt_unlock_traces(); - return err; -} -//ust// EXPORT_SYMBOL_GPL(ltt_trace_set_channel_subbufcount); - -int ltt_trace_set_channel_enable(const char *trace_name, - const char *channel_name, unsigned int enable) -{ - int err = 0; - struct ltt_trace_struct *trace; - int index; - - ltt_lock_traces(); - - trace = _ltt_trace_find_setup(trace_name); - if (!trace) { - printk(KERN_ERR "LTT : Trace not found %s\n", trace_name); - err = -ENOENT; - goto traces_error; - } - - /* - * Datas in metadata channel(marker info) is necessary to be able to - * read the trace, we always enable this channel. - */ - if (!enable && !strcmp(channel_name, "metadata")) { - printk(KERN_ERR "LTT : Trying to disable metadata channel\n"); - err = -EINVAL; - goto traces_error; - } - - index = ltt_channels_get_index_from_name(channel_name); - if (index < 0) { - printk(KERN_ERR "LTT : Channel %s not found\n", channel_name); - err = -ENOENT; - goto traces_error; - } - - trace->channels[index].active = enable; - -traces_error: - ltt_unlock_traces(); - return err; -} -//ust// EXPORT_SYMBOL_GPL(ltt_trace_set_channel_enable); - -int ltt_trace_set_channel_overwrite(const char *trace_name, - const char *channel_name, unsigned int overwrite) -{ - int err = 0; - struct ltt_trace_struct *trace; - int index; - - ltt_lock_traces(); - - trace = _ltt_trace_find_setup(trace_name); - if (!trace) { - printk(KERN_ERR "LTT : Trace not found %s\n", trace_name); - err = -ENOENT; - goto traces_error; - } - - /* - * Always put the metadata channel in non-overwrite mode : - * This is a very low traffic channel and it can't afford to have its - * data overwritten : this data (marker info) is necessary to be - * able to read the trace. - */ - if (overwrite && !strcmp(channel_name, "metadata")) { - printk(KERN_ERR "LTT : Trying to set metadata channel to " - "overwrite mode\n"); - err = -EINVAL; - goto traces_error; - } - - index = ltt_channels_get_index_from_name(channel_name); - if (index < 0) { - printk(KERN_ERR "LTT : Channel %s not found\n", channel_name); - err = -ENOENT; - goto traces_error; - } - - trace->channels[index].overwrite = overwrite; - -traces_error: - ltt_unlock_traces(); - return err; -} -//ust// EXPORT_SYMBOL_GPL(ltt_trace_set_channel_overwrite); - -int ltt_trace_alloc(const char *trace_name) -{ - int err = 0; - struct ltt_trace_struct *trace; - int subbuf_size, subbuf_cnt; - unsigned long flags; - int chan; - const char *channel_name; - - ltt_lock_traces(); - - trace = _ltt_trace_find_setup(trace_name); - if (!trace) { - printk(KERN_ERR "LTT : Trace not found %s\n", trace_name); - err = -ENOENT; - goto traces_error; - } - - kref_init(&trace->kref); - kref_init(&trace->ltt_transport_kref); -//ust// init_waitqueue_head(&trace->kref_wq); - trace->active = 0; -//ust// get_trace_clock(); - trace->freq_scale = trace_clock_freq_scale(); - - if (!trace->transport) { - printk(KERN_ERR "LTT : Transport is not set.\n"); - err = -EINVAL; - goto transport_error; - } -//ust// if (!try_module_get(trace->transport->owner)) { -//ust// printk(KERN_ERR "LTT : Can't lock transport module.\n"); -//ust// err = -ENODEV; -//ust// goto transport_error; -//ust// } - trace->ops = &trace->transport->ops; - -//ust// err = trace->ops->create_dirs(trace); -//ust// if (err) { -//ust// printk(KERN_ERR "LTT : Can't create dir for trace %s.\n", -//ust// trace_name); -//ust// goto dirs_error; -//ust// } - -//ust// local_irq_save(flags); - trace->start_freq = trace_clock_frequency(); - trace->start_tsc = trace_clock_read64(); - gettimeofday(&trace->start_time, NULL); //ust// changed -//ust// local_irq_restore(flags); - - for (chan = 0; chan < trace->nr_channels; chan++) { - if (!(trace->channels[chan].active)) - continue; - - channel_name = ltt_channels_get_name_from_index(chan); - WARN_ON(!channel_name); - subbuf_size = trace->channels[chan].subbuf_size; - subbuf_cnt = trace->channels[chan].subbuf_cnt; - prepare_chan_size_num(&subbuf_size, &subbuf_cnt); - err = trace->ops->create_channel(trace_name, trace, - trace->dentry.trace_root, - channel_name, - &trace->channels[chan], - subbuf_size, - subbuf_cnt, - trace->channels[chan].overwrite); - if (err != 0) { - printk(KERN_ERR "LTT : Can't create channel %s.\n", - channel_name); - goto create_channel_error; - } - } - - list_del(&trace->list); -//ust// if (list_empty(<t_traces.head)) { -//ust// mod_timer(<t_async_wakeup_timer, -//ust// jiffies + LTT_PERCPU_TIMER_INTERVAL); -//ust// set_kernel_trace_flag_all_tasks(); -//ust// } - list_add_rcu(&trace->list, <t_traces.head); -//ust// synchronize_sched(); - - ltt_unlock_traces(); - - return 0; - -create_channel_error: - for (chan--; chan >= 0; chan--) - if (trace->channels[chan].active) - trace->ops->remove_channel(&trace->channels[chan]); - -dirs_error: -//ust// module_put(trace->transport->owner); -transport_error: -//ust// put_trace_clock(); -traces_error: - ltt_unlock_traces(); - return err; -} -//ust// EXPORT_SYMBOL_GPL(ltt_trace_alloc); - -/* - * It is worked as a wrapper for current version of ltt_control.ko. - * We will make a new ltt_control based on debugfs, and control each channel's - * buffer. - */ -static int ltt_trace_create(const char *trace_name, const char *trace_type, - enum trace_mode mode, - unsigned int subbuf_size_low, unsigned int n_subbufs_low, - unsigned int subbuf_size_med, unsigned int n_subbufs_med, - unsigned int subbuf_size_high, unsigned int n_subbufs_high) -{ - int err = 0; - - err = ltt_trace_setup(trace_name); - if (IS_ERR_VALUE(err)) - return err; - - err = ltt_trace_set_type(trace_name, trace_type); - if (IS_ERR_VALUE(err)) - return err; - - err = ltt_trace_alloc(trace_name); - if (IS_ERR_VALUE(err)) - return err; - - return err; -} - -/* Must be called while sure that trace is in the list. */ -static int _ltt_trace_destroy(struct ltt_trace_struct *trace) -{ - int err = -EPERM; - - if (trace == NULL) { - err = -ENOENT; - goto traces_error; - } - if (trace->active) { - printk(KERN_ERR - "LTT : Can't destroy trace %s : tracer is active\n", - trace->trace_name); - err = -EBUSY; - goto active_error; - } - /* Everything went fine */ -//ust// list_del_rcu(&trace->list); -//ust// synchronize_sched(); - if (list_empty(<t_traces.head)) { -//ust// clear_kernel_trace_flag_all_tasks(); - /* - * We stop the asynchronous delivery of reader wakeup, but - * we must make one last check for reader wakeups pending - * later in __ltt_trace_destroy. - */ -//ust// del_timer_sync(<t_async_wakeup_timer); - } - return 0; - - /* error handling */ -active_error: -traces_error: - return err; -} - -/* Sleepable part of the destroy */ -static void __ltt_trace_destroy(struct ltt_trace_struct *trace) -{ - int i; - struct ltt_channel_struct *chan; - - for (i = 0; i < trace->nr_channels; i++) { - chan = &trace->channels[i]; - if (chan->active) - trace->ops->finish_channel(chan); - } - - return; /* FIXME: temporary for ust */ -//ust// flush_scheduled_work(); - - /* - * The currently destroyed trace is not in the trace list anymore, - * so it's safe to call the async wakeup ourself. It will deliver - * the last subbuffers. - */ - trace_async_wakeup(trace); - - for (i = 0; i < trace->nr_channels; i++) { - chan = &trace->channels[i]; - if (chan->active) - trace->ops->remove_channel(chan); - } - - kref_put(&trace->ltt_transport_kref, ltt_release_transport); - -//ust// module_put(trace->transport->owner); - - /* - * Wait for lttd readers to release the files, therefore making sure - * the last subbuffers have been read. - */ -//ust// if (atomic_read(&trace->kref.refcount) > 1) { -//ust// int ret = 0; -//ust// __wait_event_interruptible(trace->kref_wq, -//ust// (atomic_read(&trace->kref.refcount) == 1), ret); -//ust// } - kref_put(&trace->kref, ltt_release_trace); -} - -int ltt_trace_destroy(const char *trace_name) -{ - int err = 0; - struct ltt_trace_struct *trace; - - ltt_lock_traces(); - - trace = _ltt_trace_find(trace_name); - if (trace) { - err = _ltt_trace_destroy(trace); - if (err) - goto error; - - ltt_unlock_traces(); - - __ltt_trace_destroy(trace); -//ust// put_trace_clock(); - - return 0; - } - - trace = _ltt_trace_find_setup(trace_name); - if (trace) { - _ltt_trace_free(trace); - ltt_unlock_traces(); - return 0; - } - - err = -ENOENT; - - /* Error handling */ -error: - ltt_unlock_traces(); - return err; -} -//ust// EXPORT_SYMBOL_GPL(ltt_trace_destroy); - -/* must be called from within a traces lock. */ -static int _ltt_trace_start(struct ltt_trace_struct *trace) -{ - int err = 0; - - if (trace == NULL) { - err = -ENOENT; - goto traces_error; - } - if (trace->active) - printk(KERN_INFO "LTT : Tracing already active for trace %s\n", - trace->trace_name); -//ust// if (!try_module_get(ltt_run_filter_owner)) { -//ust// err = -ENODEV; -//ust// printk(KERN_ERR "LTT : Can't lock filter module.\n"); -//ust// goto get_ltt_run_filter_error; -//ust// } - trace->active = 1; - /* Read by trace points without protection : be careful */ - ltt_traces.num_active_traces++; - return err; - - /* error handling */ -get_ltt_run_filter_error: -traces_error: - return err; -} - -int ltt_trace_start(const char *trace_name) -{ - int err = 0; - struct ltt_trace_struct *trace; - - ltt_lock_traces(); - - trace = _ltt_trace_find(trace_name); - err = _ltt_trace_start(trace); - if (err) - goto no_trace; - - ltt_unlock_traces(); - - /* - * Call the kernel state dump. - * Events will be mixed with real kernel events, it's ok. - * Notice that there is no protection on the trace : that's exactly - * why we iterate on the list and check for trace equality instead of - * directly using this trace handle inside the logging function. - */ - - ltt_dump_marker_state(trace); - -//ust// if (!try_module_get(ltt_statedump_owner)) { -//ust// err = -ENODEV; -//ust// printk(KERN_ERR -//ust// "LTT : Can't lock state dump module.\n"); -//ust// } else { - ltt_statedump_functor(trace); -//ust// module_put(ltt_statedump_owner); -//ust// } - - return err; - - /* Error handling */ -no_trace: - ltt_unlock_traces(); - return err; -} -//ust// EXPORT_SYMBOL_GPL(ltt_trace_start); - -/* must be called from within traces lock */ -static int _ltt_trace_stop(struct ltt_trace_struct *trace) -{ - int err = -EPERM; - - if (trace == NULL) { - err = -ENOENT; - goto traces_error; - } - if (!trace->active) - printk(KERN_INFO "LTT : Tracing not active for trace %s\n", - trace->trace_name); - if (trace->active) { - trace->active = 0; - ltt_traces.num_active_traces--; -//ust// synchronize_sched(); /* Wait for each tracing to be finished */ - } -//ust// module_put(ltt_run_filter_owner); - /* Everything went fine */ - return 0; - - /* Error handling */ -traces_error: - return err; -} - -int ltt_trace_stop(const char *trace_name) -{ - int err = 0; - struct ltt_trace_struct *trace; - - ltt_lock_traces(); - trace = _ltt_trace_find(trace_name); - err = _ltt_trace_stop(trace); - ltt_unlock_traces(); - return err; -} -//ust// EXPORT_SYMBOL_GPL(ltt_trace_stop); - -/** - * ltt_control - Trace control in-kernel API - * @msg: Action to perform - * @trace_name: Trace on which the action must be done - * @trace_type: Type of trace (normal, flight, hybrid) - * @args: Arguments specific to the action - */ -//ust// int ltt_control(enum ltt_control_msg msg, const char *trace_name, -//ust// const char *trace_type, union ltt_control_args args) -//ust// { -//ust// int err = -EPERM; -//ust// -//ust// printk(KERN_ALERT "ltt_control : trace %s\n", trace_name); -//ust// switch (msg) { -//ust// case LTT_CONTROL_START: -//ust// printk(KERN_DEBUG "Start tracing %s\n", trace_name); -//ust// err = ltt_trace_start(trace_name); -//ust// break; -//ust// case LTT_CONTROL_STOP: -//ust// printk(KERN_DEBUG "Stop tracing %s\n", trace_name); -//ust// err = ltt_trace_stop(trace_name); -//ust// break; -//ust// case LTT_CONTROL_CREATE_TRACE: -//ust// printk(KERN_DEBUG "Creating trace %s\n", trace_name); -//ust// err = ltt_trace_create(trace_name, trace_type, -//ust// args.new_trace.mode, -//ust// args.new_trace.subbuf_size_low, -//ust// args.new_trace.n_subbufs_low, -//ust// args.new_trace.subbuf_size_med, -//ust// args.new_trace.n_subbufs_med, -//ust// args.new_trace.subbuf_size_high, -//ust// args.new_trace.n_subbufs_high); -//ust// break; -//ust// case LTT_CONTROL_DESTROY_TRACE: -//ust// printk(KERN_DEBUG "Destroying trace %s\n", trace_name); -//ust// err = ltt_trace_destroy(trace_name); -//ust// break; -//ust// } -//ust// return err; -//ust// } -//ust// EXPORT_SYMBOL_GPL(ltt_control); - -/** - * ltt_filter_control - Trace filter control in-kernel API - * @msg: Action to perform on the filter - * @trace_name: Trace on which the action must be done - */ -int ltt_filter_control(enum ltt_filter_control_msg msg, const char *trace_name) -{ - int err; - struct ltt_trace_struct *trace; - - printk(KERN_DEBUG "ltt_filter_control : trace %s\n", trace_name); - ltt_lock_traces(); - trace = _ltt_trace_find(trace_name); - if (trace == NULL) { - printk(KERN_ALERT - "Trace does not exist. Cannot proxy control request\n"); - err = -ENOENT; - goto trace_error; - } -//ust// if (!try_module_get(ltt_filter_control_owner)) { -//ust// err = -ENODEV; -//ust// goto get_module_error; -//ust// } - switch (msg) { - case LTT_FILTER_DEFAULT_ACCEPT: - printk(KERN_DEBUG - "Proxy filter default accept %s\n", trace_name); - err = (*ltt_filter_control_functor)(msg, trace); - break; - case LTT_FILTER_DEFAULT_REJECT: - printk(KERN_DEBUG - "Proxy filter default reject %s\n", trace_name); - err = (*ltt_filter_control_functor)(msg, trace); - break; - default: - err = -EPERM; - } -//ust// module_put(ltt_filter_control_owner); - -get_module_error: -trace_error: - ltt_unlock_traces(); - return err; -} -//ust// EXPORT_SYMBOL_GPL(ltt_filter_control); - -//ust// int __init ltt_init(void) -//ust// { -//ust// /* Make sure no page fault can be triggered by this module */ -//ust// vmalloc_sync_all(); -//ust// return 0; -//ust// } - -//ust// module_init(ltt_init) - -//ust// static void __exit ltt_exit(void) -//ust// { -//ust// struct ltt_trace_struct *trace; -//ust// struct list_head *pos, *n; -//ust// -//ust// ltt_lock_traces(); -//ust// /* Stop each trace, currently being read by RCU read-side */ -//ust// list_for_each_entry_rcu(trace, <t_traces.head, list) -//ust// _ltt_trace_stop(trace); -//ust// /* Wait for quiescent state. Readers have preemption disabled. */ -//ust// synchronize_sched(); -//ust// /* Safe iteration is now permitted. It does not have to be RCU-safe -//ust// * because no readers are left. */ -//ust// list_for_each_safe(pos, n, <t_traces.head) { -//ust// trace = container_of(pos, struct ltt_trace_struct, list); -//ust// /* _ltt_trace_destroy does a synchronize_sched() */ -//ust// _ltt_trace_destroy(trace); -//ust// __ltt_trace_destroy(trace); -//ust// } -//ust// /* free traces in pre-alloc status */ -//ust// list_for_each_safe(pos, n, <t_traces.setup_head) { -//ust// trace = container_of(pos, struct ltt_trace_struct, list); -//ust// _ltt_trace_free(trace); -//ust// } -//ust// -//ust// ltt_unlock_traces(); -//ust// } - -//ust// module_exit(ltt_exit) - -//ust// MODULE_LICENSE("GPL"); -//ust// MODULE_AUTHOR("Mathieu Desnoyers"); -//ust// MODULE_DESCRIPTION("Linux Trace Toolkit Next Generation Tracer Kernel API"); diff --git a/libtracing/tracer.h b/libtracing/tracer.h deleted file mode 100644 index cc86d54..0000000 --- a/libtracing/tracer.h +++ /dev/null @@ -1,624 +0,0 @@ -/* - * Copyright (C) 2005,2006,2008 Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) - * Copyright (C) 2009 Pierre-Marc Fournier - * - * This contains the definitions for the Linux Trace Toolkit tracer. - * - * Ported to userspace by Pierre-Marc Fournier. - * - * This file is released under the GPLv2. - */ - -#ifndef _LTT_TRACER_H -#define _LTT_TRACER_H - -#include -#include -#include "relay.h" -#include "list.h" -#include "kernelcompat.h" -#include "channels.h" -#include "tracercore.h" -#include "marker.h" - -/* Number of bytes to log with a read/write event */ -#define LTT_LOG_RW_SIZE 32L - -/* Interval (in jiffies) at which the LTT per-CPU timer fires */ -#define LTT_PERCPU_TIMER_INTERVAL 1 - -#ifndef LTT_ARCH_TYPE -#define LTT_ARCH_TYPE LTT_ARCH_TYPE_UNDEFINED -#endif - -#ifndef LTT_ARCH_VARIANT -#define LTT_ARCH_VARIANT LTT_ARCH_VARIANT_NONE -#endif - -struct ltt_active_marker; - -/* Maximum number of callbacks per marker */ -#define LTT_NR_CALLBACKS 10 - -struct ltt_serialize_closure; -struct ltt_probe_private_data; - -/* Serialization callback '%k' */ -typedef size_t (*ltt_serialize_cb)(struct rchan_buf *buf, size_t buf_offset, - struct ltt_serialize_closure *closure, - void *serialize_private, int *largest_align, - const char *fmt, va_list *args); - -struct ltt_serialize_closure { - ltt_serialize_cb *callbacks; - long cb_args[LTT_NR_CALLBACKS]; - unsigned int cb_idx; -}; - -size_t ltt_serialize_data(struct rchan_buf *buf, size_t buf_offset, - struct ltt_serialize_closure *closure, - void *serialize_private, - int *largest_align, const char *fmt, va_list *args); - -struct ltt_available_probe { - const char *name; /* probe name */ - const char *format; - marker_probe_func *probe_func; - ltt_serialize_cb callbacks[LTT_NR_CALLBACKS]; - struct list_head node; /* registered probes list */ -}; - -struct ltt_probe_private_data { - struct ltt_trace_struct *trace; /* - * Target trace, for metadata - * or statedump. - */ - ltt_serialize_cb serializer; /* - * Serialization function override. - */ - void *serialize_private; /* - * Private data for serialization - * functions. - */ -}; - -enum ltt_channels { - LTT_CHANNEL_METADATA, - LTT_CHANNEL_UST, -}; - -struct ltt_active_marker { - struct list_head node; /* active markers list */ - const char *channel; - const char *name; - const char *format; - struct ltt_available_probe *probe; -}; - -struct marker; //ust// -extern void ltt_vtrace(const struct marker *mdata, void *probe_data, - void *call_data, const char *fmt, va_list *args); -extern void ltt_trace(const struct marker *mdata, void *probe_data, - void *call_data, const char *fmt, ...); - -/* - * Unique ID assigned to each registered probe. - */ -enum marker_id { - MARKER_ID_SET_MARKER_ID = 0, /* Static IDs available (range 0-7) */ - MARKER_ID_SET_MARKER_FORMAT, - MARKER_ID_COMPACT, /* Compact IDs (range: 8-127) */ - MARKER_ID_DYNAMIC, /* Dynamic IDs (range: 128-65535) */ -}; - -/* static ids 0-1 reserved for internal use. */ -#define MARKER_CORE_IDS 2 -static inline enum marker_id marker_id_type(uint16_t id) -{ - if (id < MARKER_CORE_IDS) - return (enum marker_id)id; - else - return MARKER_ID_DYNAMIC; -} - -struct user_dbg_data { - unsigned long avail_size; - unsigned long write; - unsigned long read; -}; - -struct ltt_trace_ops { - /* First 32 bytes cache-hot cacheline */ - int (*reserve_slot) (struct ltt_trace_struct *trace, - struct ltt_channel_struct *channel, - void **transport_data, size_t data_size, - size_t *slot_size, long *buf_offset, u64 *tsc, - unsigned int *rflags, - int largest_align); - void (*commit_slot) (struct ltt_channel_struct *channel, - void **transport_data, long buf_offset, - size_t slot_size); - void (*wakeup_channel) (struct ltt_channel_struct *ltt_channel); - int (*user_blocking) (struct ltt_trace_struct *trace, - unsigned int index, size_t data_size, - struct user_dbg_data *dbg); - /* End of first 32 bytes cacheline */ - int (*create_dirs) (struct ltt_trace_struct *new_trace); - void (*remove_dirs) (struct ltt_trace_struct *new_trace); - int (*create_channel) (const char *trace_name, - struct ltt_trace_struct *trace, - struct dentry *dir, const char *channel_name, - struct ltt_channel_struct *ltt_chan, - unsigned int subbuf_size, - unsigned int n_subbufs, int overwrite); - void (*finish_channel) (struct ltt_channel_struct *channel); - void (*remove_channel) (struct ltt_channel_struct *channel); - void (*user_errors) (struct ltt_trace_struct *trace, - unsigned int index, size_t data_size, - struct user_dbg_data *dbg); -} ____cacheline_aligned; - -struct ltt_transport { - char *name; - struct module *owner; - struct list_head node; - struct ltt_trace_ops ops; -}; - -enum trace_mode { LTT_TRACE_NORMAL, LTT_TRACE_FLIGHT, LTT_TRACE_HYBRID }; - -#define CHANNEL_FLAG_ENABLE (1U<<0) -#define CHANNEL_FLAG_OVERWRITE (1U<<1) - -/* Per-trace information - each trace/flight recorder represented by one */ -struct ltt_trace_struct { - /* First 32 bytes cache-hot cacheline */ - struct list_head list; - struct ltt_trace_ops *ops; - int active; - /* Second 32 bytes cache-hot cacheline */ - struct ltt_channel_struct *channels; - unsigned int nr_channels; - u32 freq_scale; - u64 start_freq; - u64 start_tsc; - unsigned long long start_monotonic; - struct timeval start_time; - struct ltt_channel_setting *settings; - struct { - struct dentry *trace_root; - } dentry; - struct kref kref; /* Each channel has a kref of the trace struct */ - struct ltt_transport *transport; - struct kref ltt_transport_kref; - char trace_name[NAME_MAX]; -} ____cacheline_aligned; - -/* Hardcoded event headers - * - * event header for a trace with active heartbeat : 27 bits timestamps - * - * headers are 32-bits aligned. In order to insure such alignment, a dynamic per - * trace alignment value must be done. - * - * Remember that the C compiler does align each member on the boundary - * equivalent to their own size. - * - * As relay subbuffers are aligned on pages, we are sure that they are 4 and 8 - * bytes aligned, so the buffer header and trace header are aligned. - * - * Event headers are aligned depending on the trace alignment option. - * - * Note using C structure bitfields for cross-endianness and portability - * concerns. - */ - -#define LTT_RESERVED_EVENTS 3 -#define LTT_EVENT_BITS 5 -#define LTT_FREE_EVENTS ((1 << LTT_EVENT_BITS) - LTT_RESERVED_EVENTS) -#define LTT_TSC_BITS 27 -#define LTT_TSC_MASK ((1 << LTT_TSC_BITS) - 1) - -struct ltt_event_header { - u32 id_time; /* 5 bits event id (MSB); 27 bits time (LSB) */ -}; - -/* Reservation flags */ -#define LTT_RFLAG_ID (1 << 0) -#define LTT_RFLAG_ID_SIZE (1 << 1) -#define LTT_RFLAG_ID_SIZE_TSC (1 << 2) - -/* - * We use asm/timex.h : cpu_khz/HZ variable in here : we might have to deal - * specifically with CPU frequency scaling someday, so using an interpolation - * between the start and end of buffer values is not flexible enough. Using an - * immediate frequency value permits to calculate directly the times for parts - * of a buffer that would be before a frequency change. - * - * Keep the natural field alignment for _each field_ within this structure if - * you ever add/remove a field from this header. Packed attribute is not used - * because gcc generates poor code on at least powerpc and mips. Don't ever - * let gcc add padding between the structure elements. - */ -struct ltt_subbuffer_header { - uint64_t cycle_count_begin; /* Cycle count at subbuffer start */ - uint64_t cycle_count_end; /* Cycle count at subbuffer end */ - uint32_t magic_number; /* - * Trace magic number. - * contains endianness information. - */ - uint8_t major_version; - uint8_t minor_version; - uint8_t arch_size; /* Architecture pointer size */ - uint8_t alignment; /* LTT data alignment */ - uint64_t start_time_sec; /* NTP-corrected start time */ - uint64_t start_time_usec; - uint64_t start_freq; /* - * Frequency at trace start, - * used all along the trace. - */ - uint32_t freq_scale; /* Frequency scaling (divisor) */ - uint32_t lost_size; /* Size unused at end of subbuffer */ - uint32_t buf_size; /* Size of this subbuffer */ - uint32_t events_lost; /* - * Events lost in this subbuffer since - * the beginning of the trace. - * (may overflow) - */ - uint32_t subbuf_corrupt; /* - * Corrupted (lost) subbuffers since - * the begginig of the trace. - * (may overflow) - */ - uint8_t header_end[0]; /* End of header */ -}; - -/** - * ltt_subbuffer_header_size - called on buffer-switch to a new sub-buffer - * - * Return header size without padding after the structure. Don't use packed - * structure because gcc generates inefficient code on some architectures - * (powerpc, mips..) - */ -static inline size_t ltt_subbuffer_header_size(void) -{ - return offsetof(struct ltt_subbuffer_header, header_end); -} - -/* - * ltt_get_header_size - * - * Calculate alignment offset to 32-bits. This is the alignment offset of the - * event header. - * - * Important note : - * The event header must be 32-bits. The total offset calculated here : - * - * Alignment of header struct on 32 bits (min arch size, header size) - * + sizeof(header struct) (32-bits) - * + (opt) u16 (ext. event id) - * + (opt) u16 (event_size) (if event_size == 0xFFFFUL, has ext. event size) - * + (opt) u32 (ext. event size) - * + (opt) u64 full TSC (aligned on min(64-bits, arch size)) - * - * The payload must itself determine its own alignment from the biggest type it - * contains. - * */ -static inline unsigned char ltt_get_header_size( - struct ltt_channel_struct *channel, - size_t offset, - size_t data_size, - size_t *before_hdr_pad, - unsigned int rflags) -{ - size_t orig_offset = offset; - size_t padding; - - padding = ltt_align(offset, sizeof(struct ltt_event_header)); - offset += padding; - offset += sizeof(struct ltt_event_header); - - switch (rflags) { - case LTT_RFLAG_ID_SIZE_TSC: - offset += sizeof(u16) + sizeof(u16); - if (data_size >= 0xFFFFU) - offset += sizeof(u32); - offset += ltt_align(offset, sizeof(u64)); - offset += sizeof(u64); - break; - case LTT_RFLAG_ID_SIZE: - offset += sizeof(u16) + sizeof(u16); - if (data_size >= 0xFFFFU) - offset += sizeof(u32); - break; - case LTT_RFLAG_ID: - offset += sizeof(u16); - break; - } - - *before_hdr_pad = padding; - return offset - orig_offset; -} - -/* - * ltt_write_event_header - * - * Writes the event header to the offset (already aligned on 32-bits). - * - * @trace : trace to write to. - * @channel : pointer to the channel structure.. - * @buf : buffer to write to. - * @buf_offset : buffer offset to write to (aligned on 32 bits). - * @eID : event ID - * @event_size : size of the event, excluding the event header. - * @tsc : time stamp counter. - * @rflags : reservation flags. - * - * returns : offset where the event data must be written. - */ -static inline size_t ltt_write_event_header(struct ltt_trace_struct *trace, - struct ltt_channel_struct *channel, - struct rchan_buf *buf, long buf_offset, - u16 eID, size_t event_size, - u64 tsc, unsigned int rflags) -{ - struct ltt_event_header header; - size_t small_size; - - switch (rflags) { - case LTT_RFLAG_ID_SIZE_TSC: - header.id_time = 29 << LTT_TSC_BITS; - break; - case LTT_RFLAG_ID_SIZE: - header.id_time = 30 << LTT_TSC_BITS; - break; - case LTT_RFLAG_ID: - header.id_time = 31 << LTT_TSC_BITS; - break; - default: - header.id_time = eID << LTT_TSC_BITS; - break; - } - header.id_time |= (u32)tsc & LTT_TSC_MASK; - ltt_relay_write(buf, buf_offset, &header, sizeof(header)); - buf_offset += sizeof(header); - - switch (rflags) { - case LTT_RFLAG_ID_SIZE_TSC: - small_size = min_t(size_t, event_size, 0xFFFFU); - ltt_relay_write(buf, buf_offset, - (u16[]){ (u16)eID }, sizeof(u16)); - buf_offset += sizeof(u16); - ltt_relay_write(buf, buf_offset, - (u16[]){ (u16)small_size }, sizeof(u16)); - buf_offset += sizeof(u16); - if (small_size == 0xFFFFU) { - ltt_relay_write(buf, buf_offset, - (u32[]){ (u32)event_size }, sizeof(u32)); - buf_offset += sizeof(u32); - } - buf_offset += ltt_align(buf_offset, sizeof(u64)); - ltt_relay_write(buf, buf_offset, - (u64[]){ (u64)tsc }, sizeof(u64)); - buf_offset += sizeof(u64); - break; - case LTT_RFLAG_ID_SIZE: - small_size = min_t(size_t, event_size, 0xFFFFU); - ltt_relay_write(buf, buf_offset, - (u16[]){ (u16)eID }, sizeof(u16)); - buf_offset += sizeof(u16); - ltt_relay_write(buf, buf_offset, - (u16[]){ (u16)small_size }, sizeof(u16)); - buf_offset += sizeof(u16); - if (small_size == 0xFFFFU) { - ltt_relay_write(buf, buf_offset, - (u32[]){ (u32)event_size }, sizeof(u32)); - buf_offset += sizeof(u32); - } - break; - case LTT_RFLAG_ID: - ltt_relay_write(buf, buf_offset, - (u16[]){ (u16)eID }, sizeof(u16)); - buf_offset += sizeof(u16); - break; - default: - break; - } - - return buf_offset; -} - -/* Lockless LTTng */ - -/* Buffer offset macros */ - -/* - * BUFFER_TRUNC zeroes the subbuffer offset and the subbuffer number parts of - * the offset, which leaves only the buffer number. - */ -#define BUFFER_TRUNC(offset, chan) \ - ((offset) & (~((chan)->alloc_size-1))) -#define BUFFER_OFFSET(offset, chan) ((offset) & ((chan)->alloc_size - 1)) -#define SUBBUF_OFFSET(offset, chan) ((offset) & ((chan)->subbuf_size - 1)) -#define SUBBUF_ALIGN(offset, chan) \ - (((offset) + (chan)->subbuf_size) & (~((chan)->subbuf_size - 1))) -#define SUBBUF_TRUNC(offset, chan) \ - ((offset) & (~((chan)->subbuf_size - 1))) -#define SUBBUF_INDEX(offset, chan) \ - (BUFFER_OFFSET((offset), chan) >> (chan)->subbuf_size_order) - -/* - * ltt_reserve_slot - * - * Atomic slot reservation in a LTTng buffer. It will take care of - * sub-buffer switching. - * - * Parameters: - * - * @trace : the trace structure to log to. - * @channel : the chanel to reserve space into. - * @transport_data : specific transport data. - * @data_size : size of the variable length data to log. - * @slot_size : pointer to total size of the slot (out) - * @buf_offset : pointer to reserve offset (out) - * @tsc : pointer to the tsc at the slot reservation (out) - * @rflags : reservation flags (header specificity) - * @cpu : cpu id - * - * Return : -ENOSPC if not enough space, else 0. - */ -static inline int ltt_reserve_slot( - struct ltt_trace_struct *trace, - struct ltt_channel_struct *channel, - void **transport_data, - size_t data_size, - size_t *slot_size, - long *buf_offset, - u64 *tsc, - unsigned int *rflags, - int largest_align) -{ - return trace->ops->reserve_slot(trace, channel, transport_data, - data_size, slot_size, buf_offset, tsc, rflags, - largest_align); -} - - -/* - * ltt_commit_slot - * - * Atomic unordered slot commit. Increments the commit count in the - * specified sub-buffer, and delivers it if necessary. - * - * Parameters: - * - * @channel : the chanel to reserve space into. - * @transport_data : specific transport data. - * @buf_offset : offset of beginning of reserved slot - * @slot_size : size of the reserved slot. - */ -static inline void ltt_commit_slot( - struct ltt_channel_struct *channel, - void **transport_data, - long buf_offset, - size_t slot_size) -{ - struct ltt_trace_struct *trace = channel->trace; - - trace->ops->commit_slot(channel, transport_data, buf_offset, slot_size); -} - -/* - * Control channels : - * control/metadata - * control/interrupts - * control/... - * - * cpu channel : - * cpu - */ - -#define LTT_METADATA_CHANNEL "metadata_state" -#define LTT_UST_CHANNEL "ust" - -#define LTT_FLIGHT_PREFIX "flight-" - -/* Tracer properties */ -//#define LTT_DEFAULT_SUBBUF_SIZE_LOW 65536 -#define LTT_DEFAULT_SUBBUF_SIZE_LOW 4096 -#define LTT_DEFAULT_N_SUBBUFS_LOW 2 -//#define LTT_DEFAULT_SUBBUF_SIZE_MED 262144 -#define LTT_DEFAULT_SUBBUF_SIZE_MED 4096 -#define LTT_DEFAULT_N_SUBBUFS_MED 2 -//#define LTT_DEFAULT_SUBBUF_SIZE_HIGH 1048576 -#define LTT_DEFAULT_SUBBUF_SIZE_HIGH 4096 -#define LTT_DEFAULT_N_SUBBUFS_HIGH 2 -#define LTT_TRACER_MAGIC_NUMBER 0x00D6B7ED -#define LTT_TRACER_VERSION_MAJOR 2 -#define LTT_TRACER_VERSION_MINOR 3 - -/* - * Size reserved for high priority events (interrupts, NMI, BH) at the end of a - * nearly full buffer. User space won't use this last amount of space when in - * blocking mode. This space also includes the event header that would be - * written by this user space event. - */ -#define LTT_RESERVE_CRITICAL 4096 - -/* Register and unregister function pointers */ - -enum ltt_module_function { - LTT_FUNCTION_RUN_FILTER, - LTT_FUNCTION_FILTER_CONTROL, - LTT_FUNCTION_STATEDUMP -}; - -void ltt_transport_register(struct ltt_transport *transport); -void ltt_transport_unregister(struct ltt_transport *transport); - -/* Exported control function */ - -union ltt_control_args { - struct { - enum trace_mode mode; - unsigned int subbuf_size_low; - unsigned int n_subbufs_low; - unsigned int subbuf_size_med; - unsigned int n_subbufs_med; - unsigned int subbuf_size_high; - unsigned int n_subbufs_high; - } new_trace; -}; - -int _ltt_trace_setup(const char *trace_name); -int ltt_trace_setup(const char *trace_name); -struct ltt_trace_struct *_ltt_trace_find_setup(const char *trace_name); -int ltt_trace_set_type(const char *trace_name, const char *trace_type); -int ltt_trace_set_channel_subbufsize(const char *trace_name, - const char *channel_name, unsigned int size); -int ltt_trace_set_channel_subbufcount(const char *trace_name, - const char *channel_name, unsigned int cnt); -int ltt_trace_set_channel_enable(const char *trace_name, - const char *channel_name, unsigned int enable); -int ltt_trace_set_channel_overwrite(const char *trace_name, - const char *channel_name, unsigned int overwrite); -int ltt_trace_alloc(const char *trace_name); -int ltt_trace_destroy(const char *trace_name); -int ltt_trace_start(const char *trace_name); -int ltt_trace_stop(const char *trace_name); - -enum ltt_filter_control_msg { - LTT_FILTER_DEFAULT_ACCEPT, - LTT_FILTER_DEFAULT_REJECT -}; - -extern int ltt_filter_control(enum ltt_filter_control_msg msg, - const char *trace_name); - -extern struct dentry *get_filter_root(void); - -void ltt_write_trace_header(struct ltt_trace_struct *trace, - struct ltt_subbuffer_header *header); -extern void ltt_buffer_destroy(struct ltt_channel_struct *ltt_chan); - -void ltt_core_register(int (*function)(u8, void *)); - -void ltt_core_unregister(void); - -void ltt_release_trace(struct kref *kref); -void ltt_release_transport(struct kref *kref); - -extern int ltt_probe_register(struct ltt_available_probe *pdata); -extern int ltt_probe_unregister(struct ltt_available_probe *pdata); -extern int ltt_marker_connect(const char *channel, const char *mname, - const char *pname); -extern int ltt_marker_disconnect(const char *channel, const char *mname, - const char *pname); -extern void ltt_dump_marker_state(struct ltt_trace_struct *trace); - -void ltt_lock_traces(void); -void ltt_unlock_traces(void); - -struct ltt_trace_struct *_ltt_trace_find(const char *trace_name); - -#endif /* _LTT_TRACER_H */ diff --git a/libtracing/tracercore.c b/libtracing/tracercore.c deleted file mode 100644 index 3b571f1..0000000 --- a/libtracing/tracercore.c +++ /dev/null @@ -1,74 +0,0 @@ -/* - * LTT core in-kernel infrastructure. - * - * Copyright 2006 - Mathieu Desnoyers mathieu.desnoyers@polymtl.ca - * - * Distributed under the GPL license - */ - -//ust// #include -//ust// #include -//ust// #include -//ust// #include -#include "kernelcompat.h" -#include "tracercore.h" - -/* Traces structures */ -struct ltt_traces ltt_traces = { - .setup_head = LIST_HEAD_INIT(ltt_traces.setup_head), - .head = LIST_HEAD_INIT(ltt_traces.head), -}; -//ust// EXPORT_SYMBOL(ltt_traces); - -/* Traces list writer locking */ -static DEFINE_MUTEX(ltt_traces_mutex); - -/* dentry of ltt's root dir */ -//ust// static struct dentry *ltt_root_dentry; -//ust// struct dentry *get_ltt_root(void) -//ust// { -//ust// if (!ltt_root_dentry) { -//ust// ltt_root_dentry = debugfs_create_dir(LTT_ROOT, NULL); -//ust// if (!ltt_root_dentry) -//ust// printk(KERN_ERR "LTT : create ltt root dir failed\n"); -//ust// } -//ust// return ltt_root_dentry; -//ust// } -//ust// EXPORT_SYMBOL_GPL(get_ltt_root); - -void ltt_lock_traces(void) -{ - mutex_lock(<t_traces_mutex); -} -//ust// EXPORT_SYMBOL_GPL(ltt_lock_traces); - -void ltt_unlock_traces(void) -{ - mutex_unlock(<t_traces_mutex); -} -//ust// EXPORT_SYMBOL_GPL(ltt_unlock_traces); - -//ust// DEFINE_PER_CPU(unsigned int, ltt_nesting); -//ust// EXPORT_PER_CPU_SYMBOL(ltt_nesting); -unsigned int ltt_nesting; - -int ltt_run_filter_default(void *trace, uint16_t eID) -{ - return 1; -} - -/* This function pointer is protected by a trace activation check */ -ltt_run_filter_functor ltt_run_filter = ltt_run_filter_default; -//ust// EXPORT_SYMBOL_GPL(ltt_run_filter); - -void ltt_filter_register(ltt_run_filter_functor func) -{ - ltt_run_filter = func; -} -//ust// EXPORT_SYMBOL_GPL(ltt_filter_register); - -void ltt_filter_unregister(void) -{ - ltt_run_filter = ltt_run_filter_default; -} -//ust// EXPORT_SYMBOL_GPL(ltt_filter_unregister); diff --git a/libtracing/tracercore.h b/libtracing/tracercore.h deleted file mode 100644 index e5ff00f..0000000 --- a/libtracing/tracercore.h +++ /dev/null @@ -1,83 +0,0 @@ -/* - * Copyright (C) 2005,2006 Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) - * - * This contains the core definitions for the Linux Trace Toolkit. - */ - -#ifndef LTT_CORE_H -#define LTT_CORE_H - -#include "list.h" -#include "kernelcompat.h" -//ust// #include - -/* ltt's root dir in debugfs */ -#define LTT_ROOT "ltt" - -/* - * All modifications of ltt_traces must be done by ltt-tracer.c, while holding - * the semaphore. Only reading of this information can be done elsewhere, with - * the RCU mechanism : the preemption must be disabled while reading the - * list. - */ -struct ltt_traces { - struct list_head setup_head; /* Pre-allocated traces list */ - struct list_head head; /* Allocated Traces list */ - unsigned int num_active_traces; /* Number of active traces */ -} ____cacheline_aligned; - -extern struct ltt_traces ltt_traces; - -/* - * get dentry of ltt's root dir - */ -struct dentry *get_ltt_root(void); - -/* Keep track of trap nesting inside LTT */ -//ust// DECLARE_PER_CPU(unsigned int, ltt_nesting); -extern unsigned int ltt_nesting; - -typedef int (*ltt_run_filter_functor)(void *trace, uint16_t eID); -//typedef int (*ltt_run_filter_functor)(void *, __u16); - -extern ltt_run_filter_functor ltt_run_filter; - -extern void ltt_filter_register(ltt_run_filter_functor func); -extern void ltt_filter_unregister(void); - -#if defined(CONFIG_LTT) && defined(CONFIG_LTT_ALIGNMENT) - -/* - * Calculate the offset needed to align the type. - * size_of_type must be non-zero. - */ -static inline unsigned int ltt_align(size_t align_drift, size_t size_of_type) -{ - size_t alignment = min(sizeof(void *), size_of_type); - return (alignment - align_drift) & (alignment - 1); -} -/* Default arch alignment */ -#define LTT_ALIGN - -static inline int ltt_get_alignment(void) -{ - return sizeof(void *); -} - -#else - -static inline unsigned int ltt_align(size_t align_drift, - size_t size_of_type) -{ - return 0; -} - -#define LTT_ALIGN __attribute__((packed)) - -static inline int ltt_get_alignment(void) -{ - return 0; -} -#endif /* defined(CONFIG_LTT) && defined(CONFIG_LTT_ALIGNMENT) */ - -#endif /* LTT_CORE_H */ diff --git a/libust/Makefile.am b/libust/Makefile.am new file mode 100644 index 0000000..d2a8187 --- /dev/null +++ b/libust/Makefile.am @@ -0,0 +1,9 @@ +lib_LTLIBRARIES = libust.la +libust_la_SOURCES = marker.c marker.h tracepoint.c tracepoint.h immediate.h channels.c channels.h marker-control.c marker-control.h relay.c relay.h tracer.c tracer.h tracercore.c tracercore.h serialize.c tracectl.c $(top_builddir)/libustcomm/ustcomm.c $(top_builddir)/share/kref.c $(top_builddir)/share/usterr.c +libust_la_LIBADD = @URCU_PATH@/liburcu.so +libust_la_LDFLAGS = -lpthread + +INCLUDES = -I$(top_builddir)/share +INCLUDES += -I$(top_builddir)/libustcomm +INCLUDES += -I@URCU_PATH@ +INCLUDES += -I@KCOMPAT_PATH@ diff --git a/libust/channels.c b/libust/channels.c new file mode 100644 index 0000000..a8df3c4 --- /dev/null +++ b/libust/channels.c @@ -0,0 +1,348 @@ +/* + * ltt/ltt-channels.c + * + * (C) Copyright 2008 - Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) + * + * LTTng channel management. + * + * Author: + * Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) + */ + +//ust// #include +//ust// #include +//ust// #include +//ust// #include + +#include "kernelcompat.h" +#include "channels.h" +#include "usterr.h" + +/* + * ltt_channel_mutex may be nested inside the LTT trace mutex. + * ltt_channel_mutex mutex may be nested inside markers mutex. + */ +static DEFINE_MUTEX(ltt_channel_mutex); +static LIST_HEAD(ltt_channels); +/* + * Index of next channel in array. Makes sure that as long as a trace channel is + * allocated, no array index will be re-used when a channel is freed and then + * another channel is allocated. This index is cleared and the array indexeds + * get reassigned when the index_kref goes back to 0, which indicates that no + * more trace channels are allocated. + */ +static unsigned int free_index; +static struct kref index_kref; /* Keeps track of allocated trace channels */ + +static struct ltt_channel_setting *lookup_channel(const char *name) +{ + struct ltt_channel_setting *iter; + + list_for_each_entry(iter, <t_channels, list) + if (strcmp(name, iter->name) == 0) + return iter; + return NULL; +} + +/* + * Must be called when channel refcount falls to 0 _and_ also when the last + * trace is freed. This function is responsible for compacting the channel and + * event IDs when no users are active. + * + * Called with lock_markers() and channels mutex held. + */ +static void release_channel_setting(struct kref *kref) +{ + struct ltt_channel_setting *setting = container_of(kref, + struct ltt_channel_setting, kref); + struct ltt_channel_setting *iter; + + if (atomic_read(&index_kref.refcount) == 0 + && atomic_read(&setting->kref.refcount) == 0) { + list_del(&setting->list); + kfree(setting); + + free_index = 0; + list_for_each_entry(iter, <t_channels, list) { + iter->index = free_index++; + iter->free_event_id = 0; + } +//ust// markers_compact_event_ids(); + } +} + +/* + * Perform channel index compaction when the last trace channel is freed. + * + * Called with lock_markers() and channels mutex held. + */ +static void release_trace_channel(struct kref *kref) +{ + struct ltt_channel_setting *iter, *n; + + list_for_each_entry_safe(iter, n, <t_channels, list) + release_channel_setting(&iter->kref); +} + +/** + * ltt_channels_register - Register a trace channel. + * @name: channel name + * + * Uses refcounting. + */ +int ltt_channels_register(const char *name) +{ + struct ltt_channel_setting *setting; + int ret = 0; + + mutex_lock(<t_channel_mutex); + setting = lookup_channel(name); + if (setting) { + if (atomic_read(&setting->kref.refcount) == 0) + goto init_kref; + else { + kref_get(&setting->kref); + goto end; + } + } + setting = kzalloc(sizeof(*setting), GFP_KERNEL); + if (!setting) { + ret = -ENOMEM; + goto end; + } + list_add(&setting->list, <t_channels); + strncpy(setting->name, name, PATH_MAX-1); + setting->index = free_index++; +init_kref: + kref_init(&setting->kref); +end: + mutex_unlock(<t_channel_mutex); + return ret; +} +//ust// EXPORT_SYMBOL_GPL(ltt_channels_register); + +/** + * ltt_channels_unregister - Unregister a trace channel. + * @name: channel name + * + * Must be called with markers mutex held. + */ +int ltt_channels_unregister(const char *name) +{ + struct ltt_channel_setting *setting; + int ret = 0; + + mutex_lock(<t_channel_mutex); + setting = lookup_channel(name); + if (!setting || atomic_read(&setting->kref.refcount) == 0) { + ret = -ENOENT; + goto end; + } + kref_put(&setting->kref, release_channel_setting); +end: + mutex_unlock(<t_channel_mutex); + return ret; +} +//ust// EXPORT_SYMBOL_GPL(ltt_channels_unregister); + +/** + * ltt_channels_set_default - Set channel default behavior. + * @name: default channel name + * @subbuf_size: size of the subbuffers + * @subbuf_cnt: number of subbuffers + */ +int ltt_channels_set_default(const char *name, + unsigned int subbuf_size, + unsigned int subbuf_cnt) +{ + struct ltt_channel_setting *setting; + int ret = 0; + + mutex_lock(<t_channel_mutex); + setting = lookup_channel(name); + if (!setting || atomic_read(&setting->kref.refcount) == 0) { + ret = -ENOENT; + goto end; + } + setting->subbuf_size = subbuf_size; + setting->subbuf_cnt = subbuf_cnt; +end: + mutex_unlock(<t_channel_mutex); + return ret; +} +//ust// EXPORT_SYMBOL_GPL(ltt_channels_set_default); + +/** + * ltt_channels_get_name_from_index - get channel name from channel index + * @index: channel index + * + * Allows to lookup the channel name given its index. Done to keep the name + * information outside of each trace channel instance. + */ +const char *ltt_channels_get_name_from_index(unsigned int index) +{ + struct ltt_channel_setting *iter; + + list_for_each_entry(iter, <t_channels, list) + if (iter->index == index && atomic_read(&iter->kref.refcount)) + return iter->name; + return NULL; +} +//ust// EXPORT_SYMBOL_GPL(ltt_channels_get_name_from_index); + +static struct ltt_channel_setting * +ltt_channels_get_setting_from_name(const char *name) +{ + struct ltt_channel_setting *iter; + + list_for_each_entry(iter, <t_channels, list) + if (!strcmp(iter->name, name) + && atomic_read(&iter->kref.refcount)) + return iter; + return NULL; +} + +/** + * ltt_channels_get_index_from_name - get channel index from channel name + * @name: channel name + * + * Allows to lookup the channel index given its name. Done to keep the name + * information outside of each trace channel instance. + * Returns -1 if not found. + */ +int ltt_channels_get_index_from_name(const char *name) +{ + struct ltt_channel_setting *setting; + + setting = ltt_channels_get_setting_from_name(name); + if (setting) + return setting->index; + else + return -1; +} +//ust// EXPORT_SYMBOL_GPL(ltt_channels_get_index_from_name); + +/** + * ltt_channels_trace_alloc - Allocate channel structures for a trace + * @subbuf_size: subbuffer size. 0 uses default. + * @subbuf_cnt: number of subbuffers per per-cpu buffers. 0 uses default. + * @flags: Default channel flags + * + * Use the current channel list to allocate the channels for a trace. + * Called with trace lock held. Does not perform the trace buffer allocation, + * because we must let the user overwrite specific channel sizes. + */ +struct ltt_channel_struct *ltt_channels_trace_alloc(unsigned int *nr_channels, + int overwrite, + int active) +{ + struct ltt_channel_struct *channel = NULL; + struct ltt_channel_setting *iter; + + mutex_lock(<t_channel_mutex); + if (!free_index) { + WARN("ltt_channels_trace_alloc: no free_index; are there any probes connected?"); + goto end; + } + if (!atomic_read(&index_kref.refcount)) + kref_init(&index_kref); + else + kref_get(&index_kref); + *nr_channels = free_index; + channel = kzalloc(sizeof(struct ltt_channel_struct) * free_index, + GFP_KERNEL); + if (!channel) { + WARN("ltt_channel_struct: channel null after alloc"); + goto end; + } + list_for_each_entry(iter, <t_channels, list) { + if (!atomic_read(&iter->kref.refcount)) + continue; + channel[iter->index].subbuf_size = iter->subbuf_size; + channel[iter->index].subbuf_cnt = iter->subbuf_cnt; + channel[iter->index].overwrite = overwrite; + channel[iter->index].active = active; + channel[iter->index].channel_name = iter->name; + } +end: + mutex_unlock(<t_channel_mutex); + return channel; +} +//ust// EXPORT_SYMBOL_GPL(ltt_channels_trace_alloc); + +/** + * ltt_channels_trace_free - Free one trace's channels + * @channels: channels to free + * + * Called with trace lock held. The actual channel buffers must be freed before + * this function is called. + */ +void ltt_channels_trace_free(struct ltt_channel_struct *channels) +{ + lock_markers(); + mutex_lock(<t_channel_mutex); + kfree(channels); + kref_put(&index_kref, release_trace_channel); + mutex_unlock(<t_channel_mutex); + unlock_markers(); +} +//ust// EXPORT_SYMBOL_GPL(ltt_channels_trace_free); + +/** + * _ltt_channels_get_event_id - get next event ID for a marker + * @channel: channel name + * @name: event name + * + * Returns a unique event ID (for this channel) or < 0 on error. + * Must be called with channels mutex held. + */ +int _ltt_channels_get_event_id(const char *channel, const char *name) +{ + struct ltt_channel_setting *setting; + int ret; + + setting = ltt_channels_get_setting_from_name(channel); + if (!setting) { + ret = -ENOENT; + goto end; + } + if (strcmp(channel, "metadata") == 0) { + if (strcmp(name, "core_marker_id") == 0) + ret = 0; + else if (strcmp(name, "core_marker_format") == 0) + ret = 1; + else if (strcmp(name, "testev") == 0) + ret = 2; + else + ret = -ENOENT; + goto end; + } + if (setting->free_event_id == EVENTS_PER_CHANNEL - 1) { + ret = -ENOSPC; + goto end; + } + ret = setting->free_event_id++; +end: + return ret; +} + +/** + * ltt_channels_get_event_id - get next event ID for a marker + * @channel: channel name + * @name: event name + * + * Returns a unique event ID (for this channel) or < 0 on error. + */ +int ltt_channels_get_event_id(const char *channel, const char *name) +{ + int ret; + + mutex_lock(<t_channel_mutex); + ret = _ltt_channels_get_event_id(channel, name); + mutex_unlock(<t_channel_mutex); + return ret; +} + +//ust// MODULE_LICENSE("GPL"); +//ust// MODULE_AUTHOR("Mathieu Desnoyers"); +//ust// MODULE_DESCRIPTION("Linux Trace Toolkit Next Generation Channel Management"); diff --git a/libust/channels.h b/libust/channels.h new file mode 100644 index 0000000..dedc1f8 --- /dev/null +++ b/libust/channels.h @@ -0,0 +1,84 @@ +#ifndef _LTT_CHANNELS_H +#define _LTT_CHANNELS_H + +/* + * Copyright (C) 2008 Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) + * + * Dynamic tracer channel allocation. + */ + +#include +//ust// #include +//ust// #include +#include + +#include "kernelcompat.h" +#include "kref.h" +#include "list.h" + +#define EVENTS_PER_CHANNEL 65536 + +struct ltt_trace_struct; +struct rchan_buf; + +struct ltt_channel_struct { + /* First 32 bytes cache-hot cacheline */ + struct ltt_trace_struct *trace; + void *buf; + void *trans_channel_data; + int overwrite:1; + int active:1; + unsigned int n_subbufs_order; + unsigned long commit_count_mask; /* + * Commit count mask, removing + * the MSBs corresponding to + * bits used to represent the + * subbuffer index. + */ + /* End of first 32 bytes cacheline */ + + /* + * buffer_begin - called on buffer-switch to a new sub-buffer + * @buf: the channel buffer containing the new sub-buffer + */ + void (*buffer_begin) (struct rchan_buf *buf, + u64 tsc, unsigned int subbuf_idx); + /* + * buffer_end - called on buffer-switch to a new sub-buffer + * @buf: the channel buffer containing the previous sub-buffer + */ + void (*buffer_end) (struct rchan_buf *buf, + u64 tsc, unsigned int offset, unsigned int subbuf_idx); + struct kref kref; /* Channel transport reference count */ + unsigned int subbuf_size; + unsigned int subbuf_cnt; + const char *channel_name; + + int buf_shmid; +} ____cacheline_aligned; + +struct ltt_channel_setting { + unsigned int subbuf_size; + unsigned int subbuf_cnt; + struct kref kref; /* Number of references to structure content */ + struct list_head list; + unsigned int index; /* index of channel in trace channel array */ + u16 free_event_id; /* Next event ID to allocate */ + char name[PATH_MAX]; +}; + +int ltt_channels_register(const char *name); +int ltt_channels_unregister(const char *name); +int ltt_channels_set_default(const char *name, + unsigned int subbuf_size, + unsigned int subbuf_cnt); +const char *ltt_channels_get_name_from_index(unsigned int index); +int ltt_channels_get_index_from_name(const char *name); +struct ltt_channel_struct *ltt_channels_trace_alloc(unsigned int *nr_channels, + int overwrite, + int active); +void ltt_channels_trace_free(struct ltt_channel_struct *channels); +int _ltt_channels_get_event_id(const char *channel, const char *name); +int ltt_channels_get_event_id(const char *channel, const char *name); + +#endif /* _LTT_CHANNELS_H */ diff --git a/libust/immediate.h b/libust/immediate.h new file mode 100644 index 0000000..ee94ecf --- /dev/null +++ b/libust/immediate.h @@ -0,0 +1,86 @@ +#ifndef _LINUX_IMMEDIATE_H +#define _LINUX_IMMEDIATE_H + +/* + * Immediate values, can be updated at runtime and save cache lines. + * + * (C) Copyright 2007 Mathieu Desnoyers + * + * This file is released under the GPLv2. + * See the file COPYING for more details. + */ + +#ifdef USE_IMMEDIATE + +#include + +/** + * imv_set - set immediate variable (with locking) + * @name: immediate value name + * @i: required value + * + * Sets the value of @name, taking the module_mutex if required by + * the architecture. + */ +#define imv_set(name, i) \ + do { \ + name##__imv = (i); \ + core_imv_update(); \ + module_imv_update(); \ + } while (0) + +/* + * Internal update functions. + */ +extern void core_imv_update(void); +extern void imv_update_range(const struct __imv *begin, + const struct __imv *end); +extern void imv_unref_core_init(void); +extern void imv_unref(struct __imv *begin, struct __imv *end, void *start, + unsigned long size); + +#else + +/* + * Generic immediate values: a simple, standard, memory load. + */ + +/** + * imv_read - read immediate variable + * @name: immediate value name + * + * Reads the value of @name. + */ +#define imv_read(name) _imv_read(name) + +/** + * imv_set - set immediate variable (with locking) + * @name: immediate value name + * @i: required value + * + * Sets the value of @name, taking the module_mutex if required by + * the architecture. + */ +#define imv_set(name, i) (name##__imv = (i)) + +static inline void core_imv_update(void) { } +static inline void imv_unref_core_init(void) { } + +#endif + +#define DECLARE_IMV(type, name) extern __typeof__(type) name##__imv +#define DEFINE_IMV(type, name) __typeof__(type) name##__imv + +#define EXPORT_IMV_SYMBOL(name) EXPORT_SYMBOL(name##__imv) +#define EXPORT_IMV_SYMBOL_GPL(name) EXPORT_SYMBOL_GPL(name##__imv) + +/** + * _imv_read - Read immediate value with standard memory load. + * @name: immediate value name + * + * Force a data read of the immediate value instead of the immediate value + * based mechanism. Useful for __init and __exit section data read. + */ +#define _imv_read(name) (name##__imv) + +#endif diff --git a/libust/localerr.h b/libust/localerr.h new file mode 100644 index 0000000..eef0d4f --- /dev/null +++ b/libust/localerr.h @@ -0,0 +1 @@ +#include "usterr.h" diff --git a/libust/marker-control.c b/libust/marker-control.c new file mode 100644 index 0000000..29c43b0 --- /dev/null +++ b/libust/marker-control.c @@ -0,0 +1,459 @@ +/* + * Copyright (C) 2007 Mathieu Desnoyers + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. + * + * LTT marker control module over /proc + */ + +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +#include "kernelcompat.h" +#include "list.h" +#include "tracer.h" +#include "localerr.h" + +#define DEFAULT_CHANNEL "cpu" +#define DEFAULT_PROBE "default" + +LIST_HEAD(probes_list); + +/* + * Mutex protecting the probe slab cache. + * Nests inside the traces mutex. + */ +DEFINE_MUTEX(probes_mutex); + +struct ltt_available_probe default_probe = { + .name = "default", + .format = NULL, + .probe_func = ltt_vtrace, + .callbacks[0] = ltt_serialize_data, +}; + +//ust//static struct kmem_cache *markers_loaded_cachep; +static LIST_HEAD(markers_loaded_list); +/* + * List sorted by name strcmp order. + */ +static LIST_HEAD(probes_registered_list); + +//ust// static struct proc_dir_entry *pentry; + +//ust// static struct file_operations ltt_fops; + +static struct ltt_available_probe *get_probe_from_name(const char *pname) +{ + struct ltt_available_probe *iter; + int comparison, found = 0; + + if (!pname) + pname = DEFAULT_PROBE; + list_for_each_entry(iter, &probes_registered_list, node) { + comparison = strcmp(pname, iter->name); + if (!comparison) + found = 1; + if (comparison <= 0) + break; + } + if (found) + return iter; + else + return NULL; +} + +static char *skip_spaces(char *buf) +{ + while (*buf != '\0' && isspace(*buf)) + buf++; + return buf; +} + +static char *skip_nonspaces(char *buf) +{ + while (*buf != '\0' && !isspace(*buf)) + buf++; + return buf; +} + +static void get_marker_string(char *buf, char **start, + char **end) +{ + *start = skip_spaces(buf); + *end = skip_nonspaces(*start); + **end = '\0'; +} + +int ltt_probe_register(struct ltt_available_probe *pdata) +{ + int ret = 0; + int comparison; + struct ltt_available_probe *iter; + + mutex_lock(&probes_mutex); + list_for_each_entry_reverse(iter, &probes_registered_list, node) { + comparison = strcmp(pdata->name, iter->name); + if (!comparison) { + ret = -EBUSY; + goto end; + } else if (comparison > 0) { + /* We belong to the location right after iter. */ + list_add(&pdata->node, &iter->node); + goto end; + } + } + /* Should be added at the head of the list */ + list_add(&pdata->node, &probes_registered_list); +end: + mutex_unlock(&probes_mutex); + return ret; +} +EXPORT_SYMBOL_GPL(ltt_probe_register); + +/* + * Called when a probe does not want to be called anymore. + */ +int ltt_probe_unregister(struct ltt_available_probe *pdata) +{ + int ret = 0; + struct ltt_active_marker *amark, *tmp; + + mutex_lock(&probes_mutex); + list_for_each_entry_safe(amark, tmp, &markers_loaded_list, node) { + if (amark->probe == pdata) { + ret = marker_probe_unregister_private_data( + pdata->probe_func, amark); + if (ret) + goto end; + list_del(&amark->node); + free(amark); + } + } + list_del(&pdata->node); +end: + mutex_unlock(&probes_mutex); + return ret; +} +EXPORT_SYMBOL_GPL(ltt_probe_unregister); + +/* + * Connect marker "mname" to probe "pname". + * Only allow _only_ probe instance to be connected to a marker. + */ +int ltt_marker_connect(const char *channel, const char *mname, + const char *pname) + +{ + int ret; + struct ltt_active_marker *pdata; + struct ltt_available_probe *probe; + + ltt_lock_traces(); + mutex_lock(&probes_mutex); + probe = get_probe_from_name(pname); + if (!probe) { + ret = -ENOENT; + goto end; + } + pdata = marker_get_private_data(channel, mname, probe->probe_func, 0); + if (pdata && !IS_ERR(pdata)) { + ret = -EEXIST; + goto end; + } + pdata = zmalloc(sizeof(struct ltt_active_marker)); + if (!pdata) { + ret = -ENOMEM; + goto end; + } + pdata->probe = probe; + /* + * ID has priority over channel in case of conflict. + */ + ret = marker_probe_register(channel, mname, NULL, + probe->probe_func, pdata); + if (ret) + free(pdata); + else + list_add(&pdata->node, &markers_loaded_list); +end: + mutex_unlock(&probes_mutex); + ltt_unlock_traces(); + return ret; +} +EXPORT_SYMBOL_GPL(ltt_marker_connect); + +/* + * Disconnect marker "mname", probe "pname". + */ +int ltt_marker_disconnect(const char *channel, const char *mname, + const char *pname) +{ + struct ltt_active_marker *pdata; + struct ltt_available_probe *probe; + int ret = 0; + + mutex_lock(&probes_mutex); + probe = get_probe_from_name(pname); + if (!probe) { + ret = -ENOENT; + goto end; + } + pdata = marker_get_private_data(channel, mname, probe->probe_func, 0); + if (IS_ERR(pdata)) { + ret = PTR_ERR(pdata); + goto end; + } else if (!pdata) { + /* + * Not registered by us. + */ + ret = -EPERM; + goto end; + } + ret = marker_probe_unregister(channel, mname, probe->probe_func, pdata); + if (ret) + goto end; + else { + list_del(&pdata->node); + free(pdata); + } +end: + mutex_unlock(&probes_mutex); + return ret; +} +EXPORT_SYMBOL_GPL(ltt_marker_disconnect); + +/* + * function handling proc entry write. + * + * connect []] + * disconnect [] + */ +//ust// static ssize_t ltt_write(struct file *file, const char __user *buffer, +//ust// size_t count, loff_t *offset) +//ust// { +//ust// char *kbuf; +//ust// char *iter, *marker_action, *arg[4]; +//ust// ssize_t ret; +//ust// int i; +//ust// +//ust// if (!count) +//ust// return -EINVAL; +//ust// +//ust// kbuf = vmalloc(count + 1); +//ust// kbuf[count] = '\0'; /* Transform into a string */ +//ust// ret = copy_from_user(kbuf, buffer, count); +//ust// if (ret) { +//ust// ret = -EINVAL; +//ust// goto end; +//ust// } +//ust// get_marker_string(kbuf, &marker_action, &iter); +//ust// if (!marker_action || marker_action == iter) { +//ust// ret = -EINVAL; +//ust// goto end; +//ust// } +//ust// for (i = 0; i < 4; i++) { +//ust// arg[i] = NULL; +//ust// if (iter < kbuf + count) { +//ust// iter++; /* skip the added '\0' */ +//ust// get_marker_string(iter, &arg[i], &iter); +//ust// if (arg[i] == iter) +//ust// arg[i] = NULL; +//ust// } +//ust// } +//ust// +//ust// if (!arg[0] || !arg[1]) { +//ust// ret = -EINVAL; +//ust// goto end; +//ust// } +//ust// +//ust// if (!strcmp(marker_action, "connect")) { +//ust// ret = ltt_marker_connect(arg[0], arg[1], arg[2]); +//ust// if (ret) +//ust// goto end; +//ust// } else if (!strcmp(marker_action, "disconnect")) { +//ust// ret = ltt_marker_disconnect(arg[0], arg[1], arg[2]); +//ust// if (ret) +//ust// goto end; +//ust// } +//ust// ret = count; +//ust// end: +//ust// vfree(kbuf); +//ust// return ret; +//ust// } +//ust// +//ust// static void *s_next(struct seq_file *m, void *p, loff_t *pos) +//ust// { +//ust// struct marker_iter *iter = m->private; +//ust// +//ust// marker_iter_next(iter); +//ust// if (!iter->marker) { +//ust// /* +//ust// * Setting the iter module to -1UL will make sure +//ust// * that no module can possibly hold the current marker. +//ust// */ +//ust// iter->module = (void *)-1UL; +//ust// return NULL; +//ust// } +//ust// return iter->marker; +//ust// } +//ust// +//ust// static void *s_start(struct seq_file *m, loff_t *pos) +//ust// { +//ust// struct marker_iter *iter = m->private; +//ust// +//ust// if (!*pos) +//ust// marker_iter_reset(iter); +//ust// marker_iter_start(iter); +//ust// if (!iter->marker) { +//ust// /* +//ust// * Setting the iter module to -1UL will make sure +//ust// * that no module can possibly hold the current marker. +//ust// */ +//ust// iter->module = (void *)-1UL; +//ust// return NULL; +//ust// } +//ust// return iter->marker; +//ust// } +//ust// +//ust// static void s_stop(struct seq_file *m, void *p) +//ust// { +//ust// marker_iter_stop(m->private); +//ust// } +//ust// +//ust// static int s_show(struct seq_file *m, void *p) +//ust// { +//ust// struct marker_iter *iter = m->private; +//ust// +//ust// seq_printf(m, "channel: %s marker: %s format: \"%s\" state: %d " +//ust// "event_id: %hu call: 0x%p probe %s : 0x%p\n", +//ust// iter->marker->channel, +//ust// iter->marker->name, iter->marker->format, +//ust// _imv_read(iter->marker->state), +//ust// iter->marker->event_id, +//ust// iter->marker->call, +//ust// iter->marker->ptype ? "multi" : "single", +//ust// iter->marker->ptype ? +//ust// (void*)iter->marker->multi : (void*)iter->marker->single.func); +//ust// return 0; +//ust// } +//ust// +//ust// static const struct seq_operations ltt_seq_op = { +//ust// .start = s_start, +//ust// .next = s_next, +//ust// .stop = s_stop, +//ust// .show = s_show, +//ust// }; +//ust// +//ust// static int ltt_open(struct inode *inode, struct file *file) +//ust// { +//ust// /* +//ust// * Iterator kept in m->private. +//ust// * Restart iteration on all modules between reads because we do not lock +//ust// * the module mutex between those. +//ust// */ +//ust// int ret; +//ust// struct marker_iter *iter; +//ust// +//ust// iter = kzalloc(sizeof(*iter), GFP_KERNEL); +//ust// if (!iter) +//ust// return -ENOMEM; +//ust// +//ust// ret = seq_open(file, <t_seq_op); +//ust// if (ret == 0) +//ust// ((struct seq_file *)file->private_data)->private = iter; +//ust// else +//ust// kfree(iter); +//ust// return ret; +//ust// } +//ust// +//ust// static struct file_operations ltt_fops = { +//ust// .write = ltt_write, +//ust// .open = ltt_open, +//ust// .read = seq_read, +//ust// .llseek = seq_lseek, +//ust// .release = seq_release_private, +//ust// }; + +static void disconnect_all_markers(void) +{ + struct ltt_active_marker *pdata, *tmp; + + list_for_each_entry_safe(pdata, tmp, &markers_loaded_list, node) { + marker_probe_unregister_private_data(pdata->probe->probe_func, + pdata); + list_del(&pdata->node); + free(pdata); + } +} + +static char initialized = 0; + +void __attribute__((constructor)) init_marker_control(void) +{ + if(!initialized) { + int ret; + +//ust// pentry = create_proc_entry("ltt", S_IRUSR|S_IWUSR, NULL); +//ust// if (!pentry) +//ust// return -EBUSY; +//ust// markers_loaded_cachep = KMEM_CACHE(ltt_active_marker, 0); + + ret = ltt_probe_register(&default_probe); + BUG_ON(ret); + ret = ltt_marker_connect("metadata", "core_marker_format", + DEFAULT_PROBE); + BUG_ON(ret); + ret = ltt_marker_connect("metadata", "core_marker_id", DEFAULT_PROBE); + BUG_ON(ret); +//ust// pentry->proc_fops = <t_fops; + + initialized = 1; + } +} +//ust// module_init(marker_control_init); + +static void __exit marker_control_exit(void) +{ + int ret; + +//ust// remove_proc_entry("ltt", NULL); + ret = ltt_marker_disconnect("metadata", "core_marker_format", + DEFAULT_PROBE); + BUG_ON(ret); + ret = ltt_marker_disconnect("metadata", "core_marker_id", + DEFAULT_PROBE); + BUG_ON(ret); + ret = ltt_probe_unregister(&default_probe); + BUG_ON(ret); + disconnect_all_markers(); +//ust// kmem_cache_destroy(markers_loaded_cachep); +//ust// marker_synchronize_unregister(); +} +//ust// module_exit(marker_control_exit); + +//ust// MODULE_LICENSE("GPL"); +//ust// MODULE_AUTHOR("Mathieu Desnoyers"); +//ust// MODULE_DESCRIPTION("Linux Trace Toolkit Marker Control"); diff --git a/libust/marker-control.h b/libust/marker-control.h new file mode 100644 index 0000000..0b602db --- /dev/null +++ b/libust/marker-control.h @@ -0,0 +1,7 @@ +#ifndef MARKER_CONTROL_H +#define MARKER_CONTROL_H + +int marker_control_init(void); +int ltt_probe_register(struct ltt_available_probe *pdata); + +#endif /* MARKER_CONTROL_H */ diff --git a/libust/marker.c b/libust/marker.c new file mode 100644 index 0000000..e417b62 --- /dev/null +++ b/libust/marker.c @@ -0,0 +1,1532 @@ +/* + * Copyright (C) 2007 Mathieu Desnoyers + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. + */ +//ust// #include +//ust// #include +//ust// #include +//#include "jhash.h" +//#include "list.h" +//#include "rcupdate.h" +//ust// #include +#include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include + +#include "marker.h" +#include "kernelcompat.h" +#include "usterr.h" +#include "channels.h" +#include "tracercore.h" +#include "tracer.h" + +extern struct marker __start___markers[] __attribute__((visibility("hidden"))); +extern struct marker __stop___markers[] __attribute__((visibility("hidden"))); + +/* Set to 1 to enable marker debug output */ +static const int marker_debug; + +/* + * markers_mutex nests inside module_mutex. Markers mutex protects the builtin + * and module markers and the hash table. + */ +static DEFINE_MUTEX(markers_mutex); + +void lock_markers(void) +{ + mutex_lock(&markers_mutex); +} + +void unlock_markers(void) +{ + mutex_unlock(&markers_mutex); +} + +/* + * Marker hash table, containing the active markers. + * Protected by module_mutex. + */ +#define MARKER_HASH_BITS 6 +#define MARKER_TABLE_SIZE (1 << MARKER_HASH_BITS) +static struct hlist_head marker_table[MARKER_TABLE_SIZE]; + +/* + * Note about RCU : + * It is used to make sure every handler has finished using its private data + * between two consecutive operation (add or remove) on a given marker. It is + * also used to delay the free of multiple probes array until a quiescent state + * is reached. + * marker entries modifications are protected by the markers_mutex. + */ +struct marker_entry { + struct hlist_node hlist; + char *format; + char *name; + /* Probe wrapper */ + void (*call)(const struct marker *mdata, void *call_private, ...); + struct marker_probe_closure single; + struct marker_probe_closure *multi; + int refcount; /* Number of times armed. 0 if disarmed. */ + struct rcu_head rcu; + void *oldptr; + int rcu_pending; + u16 channel_id; + u16 event_id; + unsigned char ptype:1; + unsigned char format_allocated:1; + char channel[0]; /* Contains channel'\0'name'\0'format'\0' */ +}; + +#ifdef CONFIG_MARKERS_USERSPACE +static void marker_update_processes(void); +#else +static void marker_update_processes(void) +{ +} +#endif + +/** + * __mark_empty_function - Empty probe callback + * @mdata: marker data + * @probe_private: probe private data + * @call_private: call site private data + * @fmt: format string + * @...: variable argument list + * + * Empty callback provided as a probe to the markers. By providing this to a + * disabled marker, we make sure the execution flow is always valid even + * though the function pointer change and the marker enabling are two distinct + * operations that modifies the execution flow of preemptible code. + */ +notrace void __mark_empty_function(const struct marker *mdata, + void *probe_private, void *call_private, const char *fmt, va_list *args) +{ +} +//ust// EXPORT_SYMBOL_GPL(__mark_empty_function); + +/* + * marker_probe_cb Callback that prepares the variable argument list for probes. + * @mdata: pointer of type struct marker + * @call_private: caller site private data + * @...: Variable argument list. + * + * Since we do not use "typical" pointer based RCU in the 1 argument case, we + * need to put a full smp_rmb() in this branch. This is why we do not use + * rcu_dereference() for the pointer read. + */ +notrace void marker_probe_cb(const struct marker *mdata, + void *call_private, ...) +{ + va_list args; + char ptype; + + /* + * rcu_read_lock_sched does two things : disabling preemption to make + * sure the teardown of the callbacks can be done correctly when they + * are in modules and they insure RCU read coherency. + */ +//ust// rcu_read_lock_sched_notrace(); + ptype = mdata->ptype; + if (likely(!ptype)) { + marker_probe_func *func; + /* Must read the ptype before ptr. They are not data dependant, + * so we put an explicit smp_rmb() here. */ + smp_rmb(); + func = mdata->single.func; + /* Must read the ptr before private data. They are not data + * dependant, so we put an explicit smp_rmb() here. */ + smp_rmb(); + va_start(args, call_private); + func(mdata, mdata->single.probe_private, call_private, + mdata->format, &args); + va_end(args); + } else { + struct marker_probe_closure *multi; + int i; + /* + * Read mdata->ptype before mdata->multi. + */ + smp_rmb(); + multi = mdata->multi; + /* + * multi points to an array, therefore accessing the array + * depends on reading multi. However, even in this case, + * we must insure that the pointer is read _before_ the array + * data. Same as rcu_dereference, but we need a full smp_rmb() + * in the fast path, so put the explicit barrier here. + */ + smp_read_barrier_depends(); + for (i = 0; multi[i].func; i++) { + va_start(args, call_private); + multi[i].func(mdata, multi[i].probe_private, + call_private, mdata->format, &args); + va_end(args); + } + } +//ust// rcu_read_unlock_sched_notrace(); +} +//ust// EXPORT_SYMBOL_GPL(marker_probe_cb); + +/* + * marker_probe_cb Callback that does not prepare the variable argument list. + * @mdata: pointer of type struct marker + * @call_private: caller site private data + * @...: Variable argument list. + * + * Should be connected to markers "MARK_NOARGS". + */ +static notrace void marker_probe_cb_noarg(const struct marker *mdata, + void *call_private, ...) +{ + va_list args; /* not initialized */ + char ptype; + +//ust// rcu_read_lock_sched_notrace(); + ptype = mdata->ptype; + if (likely(!ptype)) { + marker_probe_func *func; + /* Must read the ptype before ptr. They are not data dependant, + * so we put an explicit smp_rmb() here. */ + smp_rmb(); + func = mdata->single.func; + /* Must read the ptr before private data. They are not data + * dependant, so we put an explicit smp_rmb() here. */ + smp_rmb(); + func(mdata, mdata->single.probe_private, call_private, + mdata->format, &args); + } else { + struct marker_probe_closure *multi; + int i; + /* + * Read mdata->ptype before mdata->multi. + */ + smp_rmb(); + multi = mdata->multi; + /* + * multi points to an array, therefore accessing the array + * depends on reading multi. However, even in this case, + * we must insure that the pointer is read _before_ the array + * data. Same as rcu_dereference, but we need a full smp_rmb() + * in the fast path, so put the explicit barrier here. + */ + smp_read_barrier_depends(); + for (i = 0; multi[i].func; i++) + multi[i].func(mdata, multi[i].probe_private, + call_private, mdata->format, &args); + } +//ust// rcu_read_unlock_sched_notrace(); +} + +static void free_old_closure(struct rcu_head *head) +{ + struct marker_entry *entry = container_of(head, + struct marker_entry, rcu); + kfree(entry->oldptr); + /* Make sure we free the data before setting the pending flag to 0 */ + smp_wmb(); + entry->rcu_pending = 0; +} + +static void debug_print_probes(struct marker_entry *entry) +{ + int i; + + if (!marker_debug) + return; + + if (!entry->ptype) { + printk(KERN_DEBUG "Single probe : %p %p\n", + entry->single.func, + entry->single.probe_private); + } else { + for (i = 0; entry->multi[i].func; i++) + printk(KERN_DEBUG "Multi probe %d : %p %p\n", i, + entry->multi[i].func, + entry->multi[i].probe_private); + } +} + +static struct marker_probe_closure * +marker_entry_add_probe(struct marker_entry *entry, + marker_probe_func *probe, void *probe_private) +{ + int nr_probes = 0; + struct marker_probe_closure *old, *new; + + WARN_ON(!probe); + + debug_print_probes(entry); + old = entry->multi; + if (!entry->ptype) { + if (entry->single.func == probe && + entry->single.probe_private == probe_private) + return ERR_PTR(-EBUSY); + if (entry->single.func == __mark_empty_function) { + /* 0 -> 1 probes */ + entry->single.func = probe; + entry->single.probe_private = probe_private; + entry->refcount = 1; + entry->ptype = 0; + debug_print_probes(entry); + return NULL; + } else { + /* 1 -> 2 probes */ + nr_probes = 1; + old = NULL; + } + } else { + /* (N -> N+1), (N != 0, 1) probes */ + for (nr_probes = 0; old[nr_probes].func; nr_probes++) + if (old[nr_probes].func == probe + && old[nr_probes].probe_private + == probe_private) + return ERR_PTR(-EBUSY); + } + /* + 2 : one for new probe, one for NULL func */ + new = kzalloc((nr_probes + 2) * sizeof(struct marker_probe_closure), + GFP_KERNEL); + if (new == NULL) + return ERR_PTR(-ENOMEM); + if (!old) + new[0] = entry->single; + else + memcpy(new, old, + nr_probes * sizeof(struct marker_probe_closure)); + new[nr_probes].func = probe; + new[nr_probes].probe_private = probe_private; + entry->refcount = nr_probes + 1; + entry->multi = new; + entry->ptype = 1; + debug_print_probes(entry); + return old; +} + +static struct marker_probe_closure * +marker_entry_remove_probe(struct marker_entry *entry, + marker_probe_func *probe, void *probe_private) +{ + int nr_probes = 0, nr_del = 0, i; + struct marker_probe_closure *old, *new; + + old = entry->multi; + + debug_print_probes(entry); + if (!entry->ptype) { + /* 0 -> N is an error */ + WARN_ON(entry->single.func == __mark_empty_function); + /* 1 -> 0 probes */ + WARN_ON(probe && entry->single.func != probe); + WARN_ON(entry->single.probe_private != probe_private); + entry->single.func = __mark_empty_function; + entry->refcount = 0; + entry->ptype = 0; + debug_print_probes(entry); + return NULL; + } else { + /* (N -> M), (N > 1, M >= 0) probes */ + for (nr_probes = 0; old[nr_probes].func; nr_probes++) { + if ((!probe || old[nr_probes].func == probe) + && old[nr_probes].probe_private + == probe_private) + nr_del++; + } + } + + if (nr_probes - nr_del == 0) { + /* N -> 0, (N > 1) */ + entry->single.func = __mark_empty_function; + entry->refcount = 0; + entry->ptype = 0; + } else if (nr_probes - nr_del == 1) { + /* N -> 1, (N > 1) */ + for (i = 0; old[i].func; i++) + if ((probe && old[i].func != probe) || + old[i].probe_private != probe_private) + entry->single = old[i]; + entry->refcount = 1; + entry->ptype = 0; + } else { + int j = 0; + /* N -> M, (N > 1, M > 1) */ + /* + 1 for NULL */ + new = kzalloc((nr_probes - nr_del + 1) + * sizeof(struct marker_probe_closure), GFP_KERNEL); + if (new == NULL) + return ERR_PTR(-ENOMEM); + for (i = 0; old[i].func; i++) + if ((probe && old[i].func != probe) || + old[i].probe_private != probe_private) + new[j++] = old[i]; + entry->refcount = nr_probes - nr_del; + entry->ptype = 1; + entry->multi = new; + } + debug_print_probes(entry); + return old; +} + +/* + * Get marker if the marker is present in the marker hash table. + * Must be called with markers_mutex held. + * Returns NULL if not present. + */ +static struct marker_entry *get_marker(const char *channel, const char *name) +{ + struct hlist_head *head; + struct hlist_node *node; + struct marker_entry *e; + size_t channel_len = strlen(channel) + 1; + size_t name_len = strlen(name) + 1; + u32 hash; + + hash = jhash(channel, channel_len-1, 0) ^ jhash(name, name_len-1, 0); + head = &marker_table[hash & ((1 << MARKER_HASH_BITS)-1)]; + hlist_for_each_entry(e, node, head, hlist) { + if (!strcmp(channel, e->channel) && !strcmp(name, e->name)) + return e; + } + return NULL; +} + +/* + * Add the marker to the marker hash table. Must be called with markers_mutex + * held. + */ +static struct marker_entry *add_marker(const char *channel, const char *name, + const char *format) +{ + struct hlist_head *head; + struct hlist_node *node; + struct marker_entry *e; + size_t channel_len = strlen(channel) + 1; + size_t name_len = strlen(name) + 1; + size_t format_len = 0; + u32 hash; + + hash = jhash(channel, channel_len-1, 0) ^ jhash(name, name_len-1, 0); + if (format) + format_len = strlen(format) + 1; + head = &marker_table[hash & ((1 << MARKER_HASH_BITS)-1)]; + hlist_for_each_entry(e, node, head, hlist) { + if (!strcmp(channel, e->channel) && !strcmp(name, e->name)) { + printk(KERN_NOTICE + "Marker %s.%s busy\n", channel, name); + return ERR_PTR(-EBUSY); /* Already there */ + } + } + /* + * Using kmalloc here to allocate a variable length element. Could + * cause some memory fragmentation if overused. + */ + e = kmalloc(sizeof(struct marker_entry) + + channel_len + name_len + format_len, + GFP_KERNEL); + if (!e) + return ERR_PTR(-ENOMEM); + memcpy(e->channel, channel, channel_len); + e->name = &e->channel[channel_len]; + memcpy(e->name, name, name_len); + if (format) { + e->format = &e->name[channel_len + name_len]; + memcpy(e->format, format, format_len); + if (strcmp(e->format, MARK_NOARGS) == 0) + e->call = marker_probe_cb_noarg; + else + e->call = marker_probe_cb; + trace_mark(metadata, core_marker_format, + "channel %s name %s format %s", + e->channel, e->name, e->format); + } else { + e->format = NULL; + e->call = marker_probe_cb; + } + e->single.func = __mark_empty_function; + e->single.probe_private = NULL; + e->multi = NULL; + e->ptype = 0; + e->format_allocated = 0; + e->refcount = 0; + e->rcu_pending = 0; + hlist_add_head(&e->hlist, head); + return e; +} + +/* + * Remove the marker from the marker hash table. Must be called with mutex_lock + * held. + */ +static int remove_marker(const char *channel, const char *name) +{ + struct hlist_head *head; + struct hlist_node *node; + struct marker_entry *e; + int found = 0; + size_t channel_len = strlen(channel) + 1; + size_t name_len = strlen(name) + 1; + u32 hash; + int ret; + + hash = jhash(channel, channel_len-1, 0) ^ jhash(name, name_len-1, 0); + head = &marker_table[hash & ((1 << MARKER_HASH_BITS)-1)]; + hlist_for_each_entry(e, node, head, hlist) { + if (!strcmp(channel, e->channel) && !strcmp(name, e->name)) { + found = 1; + break; + } + } + if (!found) + return -ENOENT; + if (e->single.func != __mark_empty_function) + return -EBUSY; + hlist_del(&e->hlist); + if (e->format_allocated) + kfree(e->format); + ret = ltt_channels_unregister(e->channel); + WARN_ON(ret); + /* Make sure the call_rcu has been executed */ + if (e->rcu_pending) + rcu_barrier_sched(); + kfree(e); + return 0; +} + +/* + * Set the mark_entry format to the format found in the element. + */ +static int marker_set_format(struct marker_entry *entry, const char *format) +{ + entry->format = kstrdup(format, GFP_KERNEL); + if (!entry->format) + return -ENOMEM; + entry->format_allocated = 1; + + trace_mark(metadata, core_marker_format, + "channel %s name %s format %s", + entry->channel, entry->name, entry->format); + return 0; +} + +/* + * Sets the probe callback corresponding to one marker. + */ +static int set_marker(struct marker_entry *entry, struct marker *elem, + int active) +{ + int ret = 0; + WARN_ON(strcmp(entry->name, elem->name) != 0); + + if (entry->format) { + if (strcmp(entry->format, elem->format) != 0) { + printk(KERN_NOTICE + "Format mismatch for probe %s " + "(%s), marker (%s)\n", + entry->name, + entry->format, + elem->format); + return -EPERM; + } + } else { + ret = marker_set_format(entry, elem->format); + if (ret) + return ret; + } + + /* + * probe_cb setup (statically known) is done here. It is + * asynchronous with the rest of execution, therefore we only + * pass from a "safe" callback (with argument) to an "unsafe" + * callback (does not set arguments). + */ + elem->call = entry->call; + elem->channel_id = entry->channel_id; + elem->event_id = entry->event_id; + /* + * Sanity check : + * We only update the single probe private data when the ptr is + * set to a _non_ single probe! (0 -> 1 and N -> 1, N != 1) + */ + WARN_ON(elem->single.func != __mark_empty_function + && elem->single.probe_private != entry->single.probe_private + && !elem->ptype); + elem->single.probe_private = entry->single.probe_private; + /* + * Make sure the private data is valid when we update the + * single probe ptr. + */ + smp_wmb(); + elem->single.func = entry->single.func; + /* + * We also make sure that the new probe callbacks array is consistent + * before setting a pointer to it. + */ + rcu_assign_pointer(elem->multi, entry->multi); + /* + * Update the function or multi probe array pointer before setting the + * ptype. + */ + smp_wmb(); + elem->ptype = entry->ptype; + +//ust// if (elem->tp_name && (active ^ _imv_read(elem->state))) { +//ust// WARN_ON(!elem->tp_cb); +//ust// /* +//ust// * It is ok to directly call the probe registration because type +//ust// * checking has been done in the __trace_mark_tp() macro. +//ust// */ +//ust// +//ust// if (active) { +//ust// /* +//ust// * try_module_get should always succeed because we hold +//ust// * markers_mutex to get the tp_cb address. +//ust// */ +//ust// ret = try_module_get(__module_text_address( +//ust// (unsigned long)elem->tp_cb)); +//ust// BUG_ON(!ret); +//ust// ret = tracepoint_probe_register_noupdate( +//ust// elem->tp_name, +//ust// elem->tp_cb); +//ust// } else { +//ust// ret = tracepoint_probe_unregister_noupdate( +//ust// elem->tp_name, +//ust// elem->tp_cb); +//ust// /* +//ust// * tracepoint_probe_update_all() must be called +//ust// * before the module containing tp_cb is unloaded. +//ust// */ +//ust// module_put(__module_text_address( +//ust// (unsigned long)elem->tp_cb)); +//ust// } +//ust// } + elem->state__imv = active; + + return ret; +} + +/* + * Disable a marker and its probe callback. + * Note: only waiting an RCU period after setting elem->call to the empty + * function insures that the original callback is not used anymore. This insured + * by rcu_read_lock_sched around the call site. + */ +static void disable_marker(struct marker *elem) +{ + int ret; + + /* leave "call" as is. It is known statically. */ +//ust// if (elem->tp_name && _imv_read(elem->state)) { +//ust// WARN_ON(!elem->tp_cb); +//ust// /* +//ust// * It is ok to directly call the probe registration because type +//ust// * checking has been done in the __trace_mark_tp() macro. +//ust// */ +//ust// ret = tracepoint_probe_unregister_noupdate(elem->tp_name, +//ust// elem->tp_cb); +//ust// WARN_ON(ret); +//ust// /* +//ust// * tracepoint_probe_update_all() must be called +//ust// * before the module containing tp_cb is unloaded. +//ust// */ +//ust// module_put(__module_text_address((unsigned long)elem->tp_cb)); +//ust// } + elem->state__imv = 0; + elem->single.func = __mark_empty_function; + /* Update the function before setting the ptype */ + smp_wmb(); + elem->ptype = 0; /* single probe */ + /* + * Leave the private data and channel_id/event_id there, because removal + * is racy and should be done only after an RCU period. These are never + * used until the next initialization anyway. + */ +} + +/** + * marker_update_probe_range - Update a probe range + * @begin: beginning of the range + * @end: end of the range + * + * Updates the probe callback corresponding to a range of markers. + */ +void marker_update_probe_range(struct marker *begin, + struct marker *end) +{ + struct marker *iter; + struct marker_entry *mark_entry; + + mutex_lock(&markers_mutex); + for (iter = begin; iter < end; iter++) { + mark_entry = get_marker(iter->channel, iter->name); + if (mark_entry) { + set_marker(mark_entry, iter, !!mark_entry->refcount); + /* + * ignore error, continue + */ + + /* This is added for UST. We emit a core_marker_id event + * for markers that are already registered to a probe + * upon library load. Otherwise, no core_marker_id will + * be generated for these markers. Is this the right thing + * to do? + */ + trace_mark(metadata, core_marker_id, + "channel %s name %s event_id %hu " + "int #1u%zu long #1u%zu pointer #1u%zu " + "size_t #1u%zu alignment #1u%u", + iter->channel, iter->name, mark_entry->event_id, + sizeof(int), sizeof(long), sizeof(void *), + sizeof(size_t), ltt_get_alignment()); + } else { + disable_marker(iter); + } + } + mutex_unlock(&markers_mutex); +} + +/* + * Update probes, removing the faulty probes. + * + * Internal callback only changed before the first probe is connected to it. + * Single probe private data can only be changed on 0 -> 1 and 2 -> 1 + * transitions. All other transitions will leave the old private data valid. + * This makes the non-atomicity of the callback/private data updates valid. + * + * "special case" updates : + * 0 -> 1 callback + * 1 -> 0 callback + * 1 -> 2 callbacks + * 2 -> 1 callbacks + * Other updates all behave the same, just like the 2 -> 3 or 3 -> 2 updates. + * Site effect : marker_set_format may delete the marker entry (creating a + * replacement). + */ +static void marker_update_probes(void) +{ + /* Core kernel markers */ +//ust// marker_update_probe_range(__start___markers, __stop___markers); + /* Markers in modules. */ +//ust// module_update_markers(); + lib_update_markers(); +//ust// tracepoint_probe_update_all(); + /* Update immediate values */ + core_imv_update(); +//ust// module_imv_update(); /* FIXME: need to port for libs? */ + marker_update_processes(); +} + +/** + * marker_probe_register - Connect a probe to a marker + * @channel: marker channel + * @name: marker name + * @format: format string + * @probe: probe handler + * @probe_private: probe private data + * + * private data must be a valid allocated memory address, or NULL. + * Returns 0 if ok, error value on error. + * The probe address must at least be aligned on the architecture pointer size. + */ +int marker_probe_register(const char *channel, const char *name, + const char *format, marker_probe_func *probe, + void *probe_private) +{ + struct marker_entry *entry; + int ret = 0, ret_err; + struct marker_probe_closure *old; + int first_probe = 0; + + mutex_lock(&markers_mutex); + entry = get_marker(channel, name); + if (!entry) { + first_probe = 1; + entry = add_marker(channel, name, format); + if (IS_ERR(entry)) + ret = PTR_ERR(entry); + if (ret) + goto end; + ret = ltt_channels_register(channel); + if (ret) + goto error_remove_marker; + ret = ltt_channels_get_index_from_name(channel); + if (ret < 0) + goto error_unregister_channel; + entry->channel_id = ret; + ret = ltt_channels_get_event_id(channel, name); + if (ret < 0) + goto error_unregister_channel; + entry->event_id = ret; + ret = 0; + trace_mark(metadata, core_marker_id, + "channel %s name %s event_id %hu " + "int #1u%zu long #1u%zu pointer #1u%zu " + "size_t #1u%zu alignment #1u%u", + channel, name, entry->event_id, + sizeof(int), sizeof(long), sizeof(void *), + sizeof(size_t), ltt_get_alignment()); + } else if (format) { + if (!entry->format) + ret = marker_set_format(entry, format); + else if (strcmp(entry->format, format)) + ret = -EPERM; + if (ret) + goto end; + } + + /* + * If we detect that a call_rcu is pending for this marker, + * make sure it's executed now. + */ + if (entry->rcu_pending) + rcu_barrier_sched(); + old = marker_entry_add_probe(entry, probe, probe_private); + if (IS_ERR(old)) { + ret = PTR_ERR(old); + if (first_probe) + goto error_unregister_channel; + else + goto end; + } + mutex_unlock(&markers_mutex); + + marker_update_probes(); + + mutex_lock(&markers_mutex); + entry = get_marker(channel, name); + if (!entry) + goto end; + if (entry->rcu_pending) + rcu_barrier_sched(); + entry->oldptr = old; + entry->rcu_pending = 1; + /* write rcu_pending before calling the RCU callback */ + smp_wmb(); + call_rcu_sched(&entry->rcu, free_old_closure); + /*synchronize_rcu(); free_old_closure();*/ + goto end; + +error_unregister_channel: + ret_err = ltt_channels_unregister(channel); + WARN_ON(ret_err); +error_remove_marker: + ret_err = remove_marker(channel, name); + WARN_ON(ret_err); +end: + mutex_unlock(&markers_mutex); + return ret; +} +//ust// EXPORT_SYMBOL_GPL(marker_probe_register); + +/** + * marker_probe_unregister - Disconnect a probe from a marker + * @channel: marker channel + * @name: marker name + * @probe: probe function pointer + * @probe_private: probe private data + * + * Returns the private data given to marker_probe_register, or an ERR_PTR(). + * We do not need to call a synchronize_sched to make sure the probes have + * finished running before doing a module unload, because the module unload + * itself uses stop_machine(), which insures that every preempt disabled section + * have finished. + */ +int marker_probe_unregister(const char *channel, const char *name, + marker_probe_func *probe, void *probe_private) +{ + struct marker_entry *entry; + struct marker_probe_closure *old; + int ret = -ENOENT; + + mutex_lock(&markers_mutex); + entry = get_marker(channel, name); + if (!entry) + goto end; + if (entry->rcu_pending) + rcu_barrier_sched(); + old = marker_entry_remove_probe(entry, probe, probe_private); + mutex_unlock(&markers_mutex); + + marker_update_probes(); + + mutex_lock(&markers_mutex); + entry = get_marker(channel, name); + if (!entry) + goto end; + if (entry->rcu_pending) + rcu_barrier_sched(); + entry->oldptr = old; + entry->rcu_pending = 1; + /* write rcu_pending before calling the RCU callback */ + smp_wmb(); + call_rcu_sched(&entry->rcu, free_old_closure); + remove_marker(channel, name); /* Ignore busy error message */ + ret = 0; +end: + mutex_unlock(&markers_mutex); + return ret; +} +//ust// EXPORT_SYMBOL_GPL(marker_probe_unregister); + +static struct marker_entry * +get_marker_from_private_data(marker_probe_func *probe, void *probe_private) +{ + struct marker_entry *entry; + unsigned int i; + struct hlist_head *head; + struct hlist_node *node; + + for (i = 0; i < MARKER_TABLE_SIZE; i++) { + head = &marker_table[i]; + hlist_for_each_entry(entry, node, head, hlist) { + if (!entry->ptype) { + if (entry->single.func == probe + && entry->single.probe_private + == probe_private) + return entry; + } else { + struct marker_probe_closure *closure; + closure = entry->multi; + for (i = 0; closure[i].func; i++) { + if (closure[i].func == probe && + closure[i].probe_private + == probe_private) + return entry; + } + } + } + } + return NULL; +} + +/** + * marker_probe_unregister_private_data - Disconnect a probe from a marker + * @probe: probe function + * @probe_private: probe private data + * + * Unregister a probe by providing the registered private data. + * Only removes the first marker found in hash table. + * Return 0 on success or error value. + * We do not need to call a synchronize_sched to make sure the probes have + * finished running before doing a module unload, because the module unload + * itself uses stop_machine(), which insures that every preempt disabled section + * have finished. + */ +int marker_probe_unregister_private_data(marker_probe_func *probe, + void *probe_private) +{ + struct marker_entry *entry; + int ret = 0; + struct marker_probe_closure *old; + const char *channel = NULL, *name = NULL; + + mutex_lock(&markers_mutex); + entry = get_marker_from_private_data(probe, probe_private); + if (!entry) { + ret = -ENOENT; + goto end; + } + if (entry->rcu_pending) + rcu_barrier_sched(); + old = marker_entry_remove_probe(entry, NULL, probe_private); + channel = kstrdup(entry->channel, GFP_KERNEL); + name = kstrdup(entry->name, GFP_KERNEL); + mutex_unlock(&markers_mutex); + + marker_update_probes(); + + mutex_lock(&markers_mutex); + entry = get_marker(channel, name); + if (!entry) + goto end; + if (entry->rcu_pending) + rcu_barrier_sched(); + entry->oldptr = old; + entry->rcu_pending = 1; + /* write rcu_pending before calling the RCU callback */ + smp_wmb(); + call_rcu_sched(&entry->rcu, free_old_closure); + /* Ignore busy error message */ + remove_marker(channel, name); +end: + mutex_unlock(&markers_mutex); + kfree(channel); + kfree(name); + return ret; +} +//ust// EXPORT_SYMBOL_GPL(marker_probe_unregister_private_data); + +/** + * marker_get_private_data - Get a marker's probe private data + * @channel: marker channel + * @name: marker name + * @probe: probe to match + * @num: get the nth matching probe's private data + * + * Returns the nth private data pointer (starting from 0) matching, or an + * ERR_PTR. + * Returns the private data pointer, or an ERR_PTR. + * The private data pointer should _only_ be dereferenced if the caller is the + * owner of the data, or its content could vanish. This is mostly used to + * confirm that a caller is the owner of a registered probe. + */ +void *marker_get_private_data(const char *channel, const char *name, + marker_probe_func *probe, int num) +{ + struct hlist_head *head; + struct hlist_node *node; + struct marker_entry *e; + size_t channel_len = strlen(channel) + 1; + size_t name_len = strlen(name) + 1; + int i; + u32 hash; + + hash = jhash(channel, channel_len-1, 0) ^ jhash(name, name_len-1, 0); + head = &marker_table[hash & ((1 << MARKER_HASH_BITS)-1)]; + hlist_for_each_entry(e, node, head, hlist) { + if (!strcmp(channel, e->channel) && !strcmp(name, e->name)) { + if (!e->ptype) { + if (num == 0 && e->single.func == probe) + return e->single.probe_private; + } else { + struct marker_probe_closure *closure; + int match = 0; + closure = e->multi; + for (i = 0; closure[i].func; i++) { + if (closure[i].func != probe) + continue; + if (match++ == num) + return closure[i].probe_private; + } + } + break; + } + } + return ERR_PTR(-ENOENT); +} +//ust// EXPORT_SYMBOL_GPL(marker_get_private_data); + +/** + * markers_compact_event_ids - Compact markers event IDs and reassign channels + * + * Called when no channel users are active by the channel infrastructure. + * Called with lock_markers() and channel mutex held. + */ +//ust// void markers_compact_event_ids(void) +//ust// { +//ust// struct marker_entry *entry; +//ust// unsigned int i; +//ust// struct hlist_head *head; +//ust// struct hlist_node *node; +//ust// int ret; +//ust// +//ust// for (i = 0; i < MARKER_TABLE_SIZE; i++) { +//ust// head = &marker_table[i]; +//ust// hlist_for_each_entry(entry, node, head, hlist) { +//ust// ret = ltt_channels_get_index_from_name(entry->channel); +//ust// WARN_ON(ret < 0); +//ust// entry->channel_id = ret; +//ust// ret = _ltt_channels_get_event_id(entry->channel, +//ust// entry->name); +//ust// WARN_ON(ret < 0); +//ust// entry->event_id = ret; +//ust// } +//ust// } +//ust// } + +//ust//#ifdef CONFIG_MODULES + +/** + * marker_get_iter_range - Get a next marker iterator given a range. + * @marker: current markers (in), next marker (out) + * @begin: beginning of the range + * @end: end of the range + * + * Returns whether a next marker has been found (1) or not (0). + * Will return the first marker in the range if the input marker is NULL. + */ +int marker_get_iter_range(struct marker **marker, struct marker *begin, + struct marker *end) +{ + if (!*marker && begin != end) { + *marker = begin; + return 1; + } + if (*marker >= begin && *marker < end) + return 1; + return 0; +} +//ust// EXPORT_SYMBOL_GPL(marker_get_iter_range); + +static void marker_get_iter(struct marker_iter *iter) +{ + int found = 0; + + /* Core kernel markers */ + if (!iter->lib) { + /* ust FIXME: how come we cannot disable the following line? we shouldn't need core stuff */ + found = marker_get_iter_range(&iter->marker, + __start___markers, __stop___markers); + if (found) + goto end; + } + /* Markers in modules. */ + found = lib_get_iter_markers(iter); +end: + if (!found) + marker_iter_reset(iter); +} + +void marker_iter_start(struct marker_iter *iter) +{ + marker_get_iter(iter); +} +//ust// EXPORT_SYMBOL_GPL(marker_iter_start); + +void marker_iter_next(struct marker_iter *iter) +{ + iter->marker++; + /* + * iter->marker may be invalid because we blindly incremented it. + * Make sure it is valid by marshalling on the markers, getting the + * markers from following modules if necessary. + */ + marker_get_iter(iter); +} +//ust// EXPORT_SYMBOL_GPL(marker_iter_next); + +void marker_iter_stop(struct marker_iter *iter) +{ +} +//ust// EXPORT_SYMBOL_GPL(marker_iter_stop); + +void marker_iter_reset(struct marker_iter *iter) +{ + iter->lib = NULL; + iter->marker = NULL; +} +//ust// EXPORT_SYMBOL_GPL(marker_iter_reset); + +#ifdef CONFIG_MARKERS_USERSPACE +/* + * must be called with current->user_markers_mutex held + */ +static void free_user_marker(char __user *state, struct hlist_head *head) +{ + struct user_marker *umark; + struct hlist_node *pos, *n; + + hlist_for_each_entry_safe(umark, pos, n, head, hlist) { + if (umark->state == state) { + hlist_del(&umark->hlist); + kfree(umark); + } + } +} + +//ust// asmlinkage long sys_marker(char __user *name, char __user *format, +//ust// char __user *state, int reg) +//ust// { +//ust// struct user_marker *umark; +//ust// long len; +//ust// struct marker_entry *entry; +//ust// int ret = 0; +//ust// +//ust// printk(KERN_DEBUG "Program %s %s marker [%p, %p]\n", +//ust// current->comm, reg ? "registers" : "unregisters", +//ust// name, state); +//ust// if (reg) { +//ust// umark = kmalloc(sizeof(struct user_marker), GFP_KERNEL); +//ust// umark->name[MAX_USER_MARKER_NAME_LEN - 1] = '\0'; +//ust// umark->format[MAX_USER_MARKER_FORMAT_LEN - 1] = '\0'; +//ust// umark->state = state; +//ust// len = strncpy_from_user(umark->name, name, +//ust// MAX_USER_MARKER_NAME_LEN - 1); +//ust// if (len < 0) { +//ust// ret = -EFAULT; +//ust// goto error; +//ust// } +//ust// len = strncpy_from_user(umark->format, format, +//ust// MAX_USER_MARKER_FORMAT_LEN - 1); +//ust// if (len < 0) { +//ust// ret = -EFAULT; +//ust// goto error; +//ust// } +//ust// printk(KERN_DEBUG "Marker name : %s, format : %s", umark->name, +//ust// umark->format); +//ust// mutex_lock(&markers_mutex); +//ust// entry = get_marker("userspace", umark->name); +//ust// if (entry) { +//ust// if (entry->format && +//ust// strcmp(entry->format, umark->format) != 0) { +//ust// printk(" error, wrong format in process %s", +//ust// current->comm); +//ust// ret = -EPERM; +//ust// goto error_unlock; +//ust// } +//ust// printk(" %s", !!entry->refcount +//ust// ? "enabled" : "disabled"); +//ust// if (put_user(!!entry->refcount, state)) { +//ust// ret = -EFAULT; +//ust// goto error_unlock; +//ust// } +//ust// printk("\n"); +//ust// } else { +//ust// printk(" disabled\n"); +//ust// if (put_user(0, umark->state)) { +//ust// printk(KERN_WARNING +//ust// "Marker in %s caused a fault\n", +//ust// current->comm); +//ust// goto error_unlock; +//ust// } +//ust// } +//ust// mutex_lock(¤t->group_leader->user_markers_mutex); +//ust// hlist_add_head(&umark->hlist, +//ust// ¤t->group_leader->user_markers); +//ust// current->group_leader->user_markers_sequence++; +//ust// mutex_unlock(¤t->group_leader->user_markers_mutex); +//ust// mutex_unlock(&markers_mutex); +//ust// } else { +//ust// mutex_lock(¤t->group_leader->user_markers_mutex); +//ust// free_user_marker(state, +//ust// ¤t->group_leader->user_markers); +//ust// current->group_leader->user_markers_sequence++; +//ust// mutex_unlock(¤t->group_leader->user_markers_mutex); +//ust// } +//ust// goto end; +//ust// error_unlock: +//ust// mutex_unlock(&markers_mutex); +//ust// error: +//ust// kfree(umark); +//ust// end: +//ust// return ret; +//ust// } +//ust// +//ust// /* +//ust// * Types : +//ust// * string : 0 +//ust// */ +//ust// asmlinkage long sys_trace(int type, uint16_t id, +//ust// char __user *ubuf) +//ust// { +//ust// long ret = -EPERM; +//ust// char *page; +//ust// int len; +//ust// +//ust// switch (type) { +//ust// case 0: /* String */ +//ust// ret = -ENOMEM; +//ust// page = (char *)__get_free_page(GFP_TEMPORARY); +//ust// if (!page) +//ust// goto string_out; +//ust// len = strncpy_from_user(page, ubuf, PAGE_SIZE); +//ust// if (len < 0) { +//ust// ret = -EFAULT; +//ust// goto string_err; +//ust// } +//ust// trace_mark(userspace, string, "string %s", page); +//ust// string_err: +//ust// free_page((unsigned long) page); +//ust// string_out: +//ust// break; +//ust// default: +//ust// break; +//ust// } +//ust// return ret; +//ust// } + +//ust// static void marker_update_processes(void) +//ust// { +//ust// struct task_struct *g, *t; +//ust// +//ust// /* +//ust// * markers_mutex is taken to protect the p->user_markers read. +//ust// */ +//ust// mutex_lock(&markers_mutex); +//ust// read_lock(&tasklist_lock); +//ust// for_each_process(g) { +//ust// WARN_ON(!thread_group_leader(g)); +//ust// if (hlist_empty(&g->user_markers)) +//ust// continue; +//ust// if (strcmp(g->comm, "testprog") == 0) +//ust// printk(KERN_DEBUG "set update pending for testprog\n"); +//ust// t = g; +//ust// do { +//ust// /* TODO : implement this thread flag in each arch. */ +//ust// set_tsk_thread_flag(t, TIF_MARKER_PENDING); +//ust// } while ((t = next_thread(t)) != g); +//ust// } +//ust// read_unlock(&tasklist_lock); +//ust// mutex_unlock(&markers_mutex); +//ust// } + +/* + * Update current process. + * Note that we have to wait a whole scheduler period before we are sure that + * every running userspace threads have their markers updated. + * (synchronize_sched() can be used to insure this). + */ +void marker_update_process(void) +{ + struct user_marker *umark; + struct hlist_node *pos; + struct marker_entry *entry; + + mutex_lock(&markers_mutex); + mutex_lock(¤t->group_leader->user_markers_mutex); + if (strcmp(current->comm, "testprog") == 0) + printk(KERN_DEBUG "do update pending for testprog\n"); + hlist_for_each_entry(umark, pos, + ¤t->group_leader->user_markers, hlist) { + printk(KERN_DEBUG "Updating marker %s in %s\n", + umark->name, current->comm); + entry = get_marker("userspace", umark->name); + if (entry) { + if (entry->format && + strcmp(entry->format, umark->format) != 0) { + printk(KERN_WARNING + " error, wrong format in process %s\n", + current->comm); + break; + } + if (put_user(!!entry->refcount, umark->state)) { + printk(KERN_WARNING + "Marker in %s caused a fault\n", + current->comm); + break; + } + } else { + if (put_user(0, umark->state)) { + printk(KERN_WARNING + "Marker in %s caused a fault\n", + current->comm); + break; + } + } + } + clear_thread_flag(TIF_MARKER_PENDING); + mutex_unlock(¤t->group_leader->user_markers_mutex); + mutex_unlock(&markers_mutex); +} + +/* + * Called at process exit and upon do_execve(). + * We assume that when the leader exits, no more references can be done to the + * leader structure by the other threads. + */ +void exit_user_markers(struct task_struct *p) +{ + struct user_marker *umark; + struct hlist_node *pos, *n; + + if (thread_group_leader(p)) { + mutex_lock(&markers_mutex); + mutex_lock(&p->user_markers_mutex); + hlist_for_each_entry_safe(umark, pos, n, &p->user_markers, + hlist) + kfree(umark); + INIT_HLIST_HEAD(&p->user_markers); + p->user_markers_sequence++; + mutex_unlock(&p->user_markers_mutex); + mutex_unlock(&markers_mutex); + } +} + +int is_marker_enabled(const char *channel, const char *name) +{ + struct marker_entry *entry; + + mutex_lock(&markers_mutex); + entry = get_marker(channel, name); + mutex_unlock(&markers_mutex); + + return entry && !!entry->refcount; +} +//ust// #endif + +int marker_module_notify(struct notifier_block *self, + unsigned long val, void *data) +{ + struct module *mod = data; + + switch (val) { + case MODULE_STATE_COMING: + marker_update_probe_range(mod->markers, + mod->markers + mod->num_markers); + break; + case MODULE_STATE_GOING: + marker_update_probe_range(mod->markers, + mod->markers + mod->num_markers); + break; + } + return 0; +} + +struct notifier_block marker_module_nb = { + .notifier_call = marker_module_notify, + .priority = 0, +}; + +//ust// static int init_markers(void) +//ust// { +//ust// return register_module_notifier(&marker_module_nb); +//ust// } +//ust// __initcall(init_markers); +/* TODO: call marker_module_nb() when a library is linked at runtime (dlopen)? */ + +#endif /* CONFIG_MODULES */ + +void ltt_dump_marker_state(struct ltt_trace_struct *trace) +{ + struct marker_iter iter; + struct ltt_probe_private_data call_data; + const char *channel; + + call_data.trace = trace; + call_data.serializer = NULL; + + marker_iter_reset(&iter); + marker_iter_start(&iter); + for (; iter.marker != NULL; marker_iter_next(&iter)) { + if (!_imv_read(iter.marker->state)) + continue; + channel = ltt_channels_get_name_from_index( + iter.marker->channel_id); + __trace_mark(0, metadata, core_marker_id, + &call_data, + "channel %s name %s event_id %hu " + "int #1u%zu long #1u%zu pointer #1u%zu " + "size_t #1u%zu alignment #1u%u", + channel, + iter.marker->name, + iter.marker->event_id, + sizeof(int), sizeof(long), + sizeof(void *), sizeof(size_t), + ltt_get_alignment()); + if (iter.marker->format) + __trace_mark(0, metadata, + core_marker_format, + &call_data, + "channel %s name %s format %s", + channel, + iter.marker->name, + iter.marker->format); + } + marker_iter_stop(&iter); +} +//ust// EXPORT_SYMBOL_GPL(ltt_dump_marker_state); + + +static LIST_HEAD(libs); + +/* + * Returns 0 if current not found. + * Returns 1 if current found. + */ +int lib_get_iter_markers(struct marker_iter *iter) +{ + struct lib *iter_lib; + int found = 0; + +//ust// mutex_lock(&module_mutex); + list_for_each_entry(iter_lib, &libs, list) { + if (iter_lib < iter->lib) + continue; + else if (iter_lib > iter->lib) + iter->marker = NULL; + found = marker_get_iter_range(&iter->marker, + iter_lib->markers_start, + iter_lib->markers_start + iter_lib->markers_count); + if (found) { + iter->lib = iter_lib; + break; + } + } +//ust// mutex_unlock(&module_mutex); + return found; +} + +void lib_update_markers(void) +{ + struct lib *lib; + +//ust// mutex_lock(&module_mutex); + list_for_each_entry(lib, &libs, list) + marker_update_probe_range(lib->markers_start, + lib->markers_start + lib->markers_count); +//ust// mutex_unlock(&module_mutex); +} + +static void (*new_marker_cb)(struct marker *) = NULL; + +void marker_set_new_marker_cb(void (*cb)(struct marker *)) +{ + new_marker_cb = cb; +} + +static void new_markers(struct marker *start, struct marker *end) +{ + if(new_marker_cb) { + struct marker *m; + for(m=start; m < end; m++) { + new_marker_cb(m); + } + } +} + +int marker_register_lib(struct marker *markers_start, int markers_count) +{ + struct lib *pl; + + pl = (struct lib *) malloc(sizeof(struct lib)); + + pl->markers_start = markers_start; + pl->markers_count = markers_count; + + /* FIXME: maybe protect this with its own mutex? */ + lock_markers(); + list_add(&pl->list, &libs); + unlock_markers(); + + new_markers(markers_start, markers_start + markers_count); + + /* FIXME: update just the loaded lib */ + lib_update_markers(); + + DBG("just registered a markers section from %p and having %d markers", markers_start, markers_count); + + return 0; +} + +int marker_unregister_lib(struct marker *markers_start, int markers_count) +{ + /*FIXME: implement; but before implementing, marker_register_lib must + have appropriate locking. */ + + return 0; +} + +static int initialized = 0; + +void __attribute__((constructor)) init_markers(void) +{ + if(!initialized) { + marker_register_lib(__start___markers, (((long)__stop___markers)-((long)__start___markers))/sizeof(struct marker)); + printf("markers_start: %p, markers_stop: %p\n", __start___markers, __stop___markers); + initialized = 1; + } +} diff --git a/libust/marker.h b/libust/marker.h new file mode 100644 index 0000000..7c3d8c5 --- /dev/null +++ b/libust/marker.h @@ -0,0 +1,298 @@ +/* + * Code markup for dynamic and static tracing. + * + * See Documentation/marker.txt. + * + * (C) Copyright 2006 Mathieu Desnoyers + * (C) Copyright 2009 Pierre-Marc Fournier + * + * This file is released under the GPLv2. + * See the file COPYING for more details. + */ + +#ifndef _LINUX_MARKER_H +#define _LINUX_MARKER_H + +#include +//ust// #include +#include "immediate.h" +//ust// #include +#include "kernelcompat.h" +#include "compiler.h" +#include "list.h" + +//ust// struct module; +//ust// struct task_struct; +struct marker; + +/** + * marker_probe_func - Type of a marker probe function + * @mdata: marker data + * @probe_private: probe private data + * @call_private: call site private data + * @fmt: format string + * @args: variable argument list pointer. Use a pointer to overcome C's + * inability to pass this around as a pointer in a portable manner in + * the callee otherwise. + * + * Type of marker probe functions. They receive the mdata and need to parse the + * format string to recover the variable argument list. + */ +typedef void marker_probe_func(const struct marker *mdata, + void *probe_private, void *call_private, + const char *fmt, va_list *args); + +struct marker_probe_closure { + marker_probe_func *func; /* Callback */ + void *probe_private; /* Private probe data */ +}; + +struct marker { + const char *channel; /* Name of channel where to send data */ + const char *name; /* Marker name */ + const char *format; /* Marker format string, describing the + * variable argument list. + */ + DEFINE_IMV(char, state);/* Immediate value state. */ + char ptype; /* probe type : 0 : single, 1 : multi */ + /* Probe wrapper */ + u16 channel_id; /* Numeric channel identifier, dynamic */ + u16 event_id; /* Numeric event identifier, dynamic */ + void (*call)(const struct marker *mdata, void *call_private, ...); + struct marker_probe_closure single; + struct marker_probe_closure *multi; + const char *tp_name; /* Optional tracepoint name */ + void *tp_cb; /* Optional tracepoint callback */ +} __attribute__((aligned(8))); + +//ust// #ifdef CONFIG_MARKERS + +#define _DEFINE_MARKER(channel, name, tp_name_str, tp_cb, format) \ + static const char __mstrtab_##channel##_##name[] \ + __attribute__((section("__markers_strings"))) \ + = #channel "\0" #name "\0" format; \ + static struct marker __mark_##channel##_##name \ + __attribute__((section("__markers"), aligned(8))) = \ + { __mstrtab_##channel##_##name, \ + &__mstrtab_##channel##_##name[sizeof(#channel)], \ + &__mstrtab_##channel##_##name[sizeof(#channel) + \ + sizeof(#name)], \ + 0, 0, 0, 0, marker_probe_cb, \ + { __mark_empty_function, NULL}, \ + NULL, tp_name_str, tp_cb } + +#define DEFINE_MARKER(channel, name, format) \ + _DEFINE_MARKER(channel, name, NULL, NULL, format) + +#define DEFINE_MARKER_TP(channel, name, tp_name, tp_cb, format) \ + _DEFINE_MARKER(channel, name, #tp_name, tp_cb, format) + +/* + * Make sure the alignment of the structure in the __markers section will + * not add unwanted padding between the beginning of the section and the + * structure. Force alignment to the same alignment as the section start. + * + * The "generic" argument controls which marker enabling mechanism must be used. + * If generic is true, a variable read is used. + * If generic is false, immediate values are used. + */ +#define __trace_mark(generic, channel, name, call_private, format, args...) \ + do { \ + DEFINE_MARKER(channel, name, format); \ + __mark_check_format(format, ## args); \ + if (!generic) { \ + if (unlikely(imv_read( \ + __mark_##channel##_##name.state))) \ + (*__mark_##channel##_##name.call) \ + (&__mark_##channel##_##name, \ + call_private, ## args); \ + } else { \ + if (unlikely(_imv_read( \ + __mark_##channel##_##name.state))) \ + (*__mark_##channel##_##name.call) \ + (&__mark_##channel##_##name, \ + call_private, ## args); \ + } \ + } while (0) + +#define __trace_mark_tp(channel, name, call_private, tp_name, tp_cb, \ + format, args...) \ + do { \ + void __check_tp_type(void) \ + { \ + register_trace_##tp_name(tp_cb); \ + } \ + DEFINE_MARKER_TP(channel, name, tp_name, tp_cb, format);\ + __mark_check_format(format, ## args); \ + (*__mark_##channel##_##name.call)(&__mark_##channel##_##name, \ + call_private, ## args); \ + } while (0) + +extern void marker_update_probe_range(struct marker *begin, + struct marker *end); + +#define GET_MARKER(channel, name) (__mark_##channel##_##name) + +//ust// #else /* !CONFIG_MARKERS */ +//ust// #define DEFINE_MARKER(channel, name, tp_name, tp_cb, format) +//ust// #define __trace_mark(generic, channel, name, call_private, format, args...) \ +//ust// __mark_check_format(format, ## args) +//ust// #define __trace_mark_tp(channel, name, call_private, tp_name, tp_cb, \ +//ust// format, args...) \ +//ust// do { \ +//ust// void __check_tp_type(void) \ +//ust// { \ +//ust// register_trace_##tp_name(tp_cb); \ +//ust// } \ +//ust// __mark_check_format(format, ## args); \ +//ust// } while (0) +//ust// static inline void marker_update_probe_range(struct marker *begin, +//ust// struct marker *end) +//ust// { } +//ust// #define GET_MARKER(channel, name) +//ust// #endif /* CONFIG_MARKERS */ + +/** + * trace_mark - Marker using code patching + * @channel: marker channel (where to send the data), not quoted. + * @name: marker name, not quoted. + * @format: format string + * @args...: variable argument list + * + * Places a marker using optimized code patching technique (imv_read()) + * to be enabled when immediate values are present. + */ +#define trace_mark(channel, name, format, args...) \ + __trace_mark(0, channel, name, NULL, format, ## args) + +/** + * _trace_mark - Marker using variable read + * @channel: marker channel (where to send the data), not quoted. + * @name: marker name, not quoted. + * @format: format string + * @args...: variable argument list + * + * Places a marker using a standard memory read (_imv_read()) to be + * enabled. Should be used for markers in code paths where instruction + * modification based enabling is not welcome. (__init and __exit functions, + * lockdep, some traps, printk). + */ +#define _trace_mark(channel, name, format, args...) \ + __trace_mark(1, channel, name, NULL, format, ## args) + +/** + * trace_mark_tp - Marker in a tracepoint callback + * @channel: marker channel (where to send the data), not quoted. + * @name: marker name, not quoted. + * @tp_name: tracepoint name, not quoted. + * @tp_cb: tracepoint callback. Should have an associated global symbol so it + * is not optimized away by the compiler (should not be static). + * @format: format string + * @args...: variable argument list + * + * Places a marker in a tracepoint callback. + */ +#define trace_mark_tp(channel, name, tp_name, tp_cb, format, args...) \ + __trace_mark_tp(channel, name, NULL, tp_name, tp_cb, format, ## args) + +/** + * MARK_NOARGS - Format string for a marker with no argument. + */ +#define MARK_NOARGS " " + +extern void lock_markers(void); +extern void unlock_markers(void); + +extern void markers_compact_event_ids(void); + +/* To be used for string format validity checking with gcc */ +static inline void __printf(1, 2) ___mark_check_format(const char *fmt, ...) +{ +} + +#define __mark_check_format(format, args...) \ + do { \ + if (0) \ + ___mark_check_format(format, ## args); \ + } while (0) + +extern marker_probe_func __mark_empty_function; + +extern void marker_probe_cb(const struct marker *mdata, + void *call_private, ...); + +/* + * Connect a probe to a marker. + * private data pointer must be a valid allocated memory address, or NULL. + */ +extern int marker_probe_register(const char *channel, const char *name, + const char *format, marker_probe_func *probe, void *probe_private); + +/* + * Returns the private data given to marker_probe_register. + */ +extern int marker_probe_unregister(const char *channel, const char *name, + marker_probe_func *probe, void *probe_private); +/* + * Unregister a marker by providing the registered private data. + */ +extern int marker_probe_unregister_private_data(marker_probe_func *probe, + void *probe_private); + +extern void *marker_get_private_data(const char *channel, const char *name, + marker_probe_func *probe, int num); + +/* + * marker_synchronize_unregister must be called between the last marker probe + * unregistration and the first one of + * - the end of module exit function + * - the free of any resource used by the probes + * to ensure the code and data are valid for any possibly running probes. + */ +#define marker_synchronize_unregister() synchronize_sched() + +struct marker_iter { +//ust// struct module *module; + struct lib *lib; + struct marker *marker; +}; + +extern void marker_iter_start(struct marker_iter *iter); +extern void marker_iter_next(struct marker_iter *iter); +extern void marker_iter_stop(struct marker_iter *iter); +extern void marker_iter_reset(struct marker_iter *iter); +extern int marker_get_iter_range(struct marker **marker, struct marker *begin, + struct marker *end); + +extern void marker_update_process(void); +extern int is_marker_enabled(const char *channel, const char *name); + +//ust// #ifdef CONFIG_MARKERS_USERSPACE +//ust// extern void exit_user_markers(struct task_struct *p); +//ust// #else +//ust// static inline void exit_user_markers(struct task_struct *p) +//ust// { +//ust// } +//ust// #endif + + +struct lib { + struct marker *markers_start; + int markers_count; + struct list_head list; +}; + +int marker_register_lib(struct marker *markers_start, int markers_count); + +#define MARKER_LIB \ +extern struct marker __start___markers[] __attribute__((visibility("hidden"))); \ +extern struct marker __stop___markers[] __attribute__((visibility("hidden"))); \ + \ +static void __attribute__((constructor)) __markers__init(void) \ +{ \ + marker_register_lib(__start___markers, (((long)__stop___markers)-((long)__start___markers))/sizeof(struct marker));\ +} + +void marker_set_new_marker_cb(void (*cb)(struct marker *)); + +#endif diff --git a/libust/relay.c b/libust/relay.c new file mode 100644 index 0000000..6bb5a35 --- /dev/null +++ b/libust/relay.c @@ -0,0 +1,2490 @@ +/* + * Public API and common code for kernel->userspace relay file support. + * + * Copyright (C) 2002-2005 - Tom Zanussi (zanussi@us.ibm.com), IBM Corp + * Copyright (C) 1999-2005 - Karim Yaghmour (karim@opersys.com) + * Copyright (C) 2008 - Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) + * + * Moved to kernel/relay.c by Paul Mundt, 2006. + * November 2006 - CPU hotplug support by Mathieu Desnoyers + * (mathieu.desnoyers@polymtl.ca) + * + * This file is released under the GPL. + */ +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +#include "kernelcompat.h" +#include +#include +#include +#include "list.h" +#include "relay.h" +#include "channels.h" +#include "kref.h" +#include "tracer.h" +#include "tracercore.h" +#include "usterr.h" + +/* list of open channels, for cpu hotplug */ +static DEFINE_MUTEX(relay_channels_mutex); +static LIST_HEAD(relay_channels); + + +static struct dentry *ltt_create_buf_file_callback(struct rchan_buf *buf); + +/** + * relay_alloc_buf - allocate a channel buffer + * @buf: the buffer struct + * @size: total size of the buffer + */ +//ust// static int relay_alloc_buf(struct rchan_buf *buf, size_t *size) +//ust//{ +//ust// unsigned int i, n_pages; +//ust// struct buf_page *buf_page, *n; +//ust// +//ust// *size = PAGE_ALIGN(*size); +//ust// n_pages = *size >> PAGE_SHIFT; +//ust// +//ust// INIT_LIST_HEAD(&buf->pages); +//ust// +//ust// for (i = 0; i < n_pages; i++) { +//ust// buf_page = kmalloc_node(sizeof(*buf_page), GFP_KERNEL, +//ust// cpu_to_node(buf->cpu)); +//ust// if (unlikely(!buf_page)) +//ust// goto depopulate; +//ust// buf_page->page = alloc_pages_node(cpu_to_node(buf->cpu), +//ust// GFP_KERNEL | __GFP_ZERO, 0); +//ust// if (unlikely(!buf_page->page)) { +//ust// kfree(buf_page); +//ust// goto depopulate; +//ust// } +//ust// list_add_tail(&buf_page->list, &buf->pages); +//ust// buf_page->offset = (size_t)i << PAGE_SHIFT; +//ust// buf_page->buf = buf; +//ust// set_page_private(buf_page->page, (unsigned long)buf_page); +//ust// if (i == 0) { +//ust// buf->wpage = buf_page; +//ust// buf->hpage[0] = buf_page; +//ust// buf->hpage[1] = buf_page; +//ust// buf->rpage = buf_page; +//ust// } +//ust// } +//ust// buf->page_count = n_pages; +//ust// return 0; +//ust// +//ust//depopulate: +//ust// list_for_each_entry_safe(buf_page, n, &buf->pages, list) { +//ust// list_del_init(&buf_page->list); +//ust// __free_page(buf_page->page); +//ust// kfree(buf_page); +//ust// } +//ust// return -ENOMEM; +//ust//} + +static int relay_alloc_buf(struct rchan_buf *buf, size_t *size) +{ +//ust// unsigned int n_pages; +//ust// struct buf_page *buf_page, *n; + + void *ptr; + int result; + + *size = PAGE_ALIGN(*size); + + result = buf->shmid = shmget(getpid(), *size, IPC_CREAT | IPC_EXCL | 0700); + if(buf->shmid == -1) { + PERROR("shmget"); + return -1; + } + + ptr = shmat(buf->shmid, NULL, 0); + if(ptr == (void *) -1) { + perror("shmat"); + goto destroy_shmem; + } + + /* Already mark the shared memory for destruction. This will occur only + * when all users have detached. + */ + result = shmctl(buf->shmid, IPC_RMID, NULL); + if(result == -1) { + perror("shmctl"); + return -1; + } + + buf->buf_data = ptr; + buf->buf_size = *size; + + return 0; + + destroy_shmem: + result = shmctl(buf->shmid, IPC_RMID, NULL); + if(result == -1) { + perror("shmctl"); + } + + return -1; +} + +/** + * relay_create_buf - allocate and initialize a channel buffer + * @chan: the relay channel + * @cpu: cpu the buffer belongs to + * + * Returns channel buffer if successful, %NULL otherwise. + */ +static struct rchan_buf *relay_create_buf(struct rchan *chan) +{ + int ret; + struct rchan_buf *buf = kzalloc(sizeof(struct rchan_buf), GFP_KERNEL); + if (!buf) + return NULL; + +// buf->cpu = cpu; + ret = relay_alloc_buf(buf, &chan->alloc_size); + if (ret) + goto free_buf; + + buf->chan = chan; + kref_get(&buf->chan->kref); + return buf; + +free_buf: + kfree(buf); + return NULL; +} + +/** + * relay_destroy_channel - free the channel struct + * @kref: target kernel reference that contains the relay channel + * + * Should only be called from kref_put(). + */ +static void relay_destroy_channel(struct kref *kref) +{ + struct rchan *chan = container_of(kref, struct rchan, kref); + kfree(chan); +} + +/** + * relay_destroy_buf - destroy an rchan_buf struct and associated buffer + * @buf: the buffer struct + */ +static void relay_destroy_buf(struct rchan_buf *buf) +{ + struct rchan *chan = buf->chan; + struct buf_page *buf_page, *n; + int result; + + result = munmap(buf->buf_data, buf->buf_size); + if(result == -1) { + PERROR("munmap"); + } + +//ust// chan->buf[buf->cpu] = NULL; + kfree(buf); + kref_put(&chan->kref, relay_destroy_channel); +} + +/** + * relay_remove_buf - remove a channel buffer + * @kref: target kernel reference that contains the relay buffer + * + * Removes the file from the fileystem, which also frees the + * rchan_buf_struct and the channel buffer. Should only be called from + * kref_put(). + */ +static void relay_remove_buf(struct kref *kref) +{ + struct rchan_buf *buf = container_of(kref, struct rchan_buf, kref); +//ust// buf->chan->cb->remove_buf_file(buf); + relay_destroy_buf(buf); +} + +/* + * High-level relay kernel API and associated functions. + */ + +/* + * rchan_callback implementations defining default channel behavior. Used + * in place of corresponding NULL values in client callback struct. + */ + +/* + * create_buf_file_create() default callback. Does nothing. + */ +static struct dentry *create_buf_file_default_callback(const char *filename, + struct dentry *parent, + int mode, + struct rchan_buf *buf) +{ + return NULL; +} + +/* + * remove_buf_file() default callback. Does nothing. + */ +static int remove_buf_file_default_callback(struct dentry *dentry) +{ + return -EINVAL; +} + +/** + * wakeup_readers - wake up readers waiting on a channel + * @data: contains the channel buffer + * + * This is the timer function used to defer reader waking. + */ +//ust// static void wakeup_readers(unsigned long data) +//ust// { +//ust// struct rchan_buf *buf = (struct rchan_buf *)data; +//ust// wake_up_interruptible(&buf->read_wait); +//ust// } + +/** + * __relay_reset - reset a channel buffer + * @buf: the channel buffer + * @init: 1 if this is a first-time initialization + * + * See relay_reset() for description of effect. + */ +static void __relay_reset(struct rchan_buf *buf, unsigned int init) +{ + if (init) { +//ust// init_waitqueue_head(&buf->read_wait); + kref_init(&buf->kref); +//ust// setup_timer(&buf->timer, wakeup_readers, (unsigned long)buf); + } else +//ust// del_timer_sync(&buf->timer); + + buf->finalized = 0; +} + +/* + * relay_open_buf - create a new relay channel buffer + * + * used by relay_open() and CPU hotplug. + */ +static struct rchan_buf *relay_open_buf(struct rchan *chan) +{ + struct rchan_buf *buf = NULL; + struct dentry *dentry; +//ust// char *tmpname; + +//ust// tmpname = kzalloc(NAME_MAX + 1, GFP_KERNEL); +//ust// if (!tmpname) +//ust// goto end; +//ust// snprintf(tmpname, NAME_MAX, "%s%d", chan->base_filename, cpu); + + buf = relay_create_buf(chan); + if (!buf) + goto free_name; + + __relay_reset(buf, 1); + + /* Create file in fs */ +//ust// dentry = chan->cb->create_buf_file(tmpname, chan->parent, S_IRUSR, +//ust// buf); + + ltt_create_buf_file_callback(buf); // ust // + +//ust// if (!dentry) +//ust// goto free_buf; +//ust// +//ust// buf->dentry = dentry; + + goto free_name; + +free_buf: + relay_destroy_buf(buf); + buf = NULL; +free_name: +//ust// kfree(tmpname); +end: + return buf; +} + +/** + * relay_close_buf - close a channel buffer + * @buf: channel buffer + * + * Marks the buffer finalized and restores the default callbacks. + * The channel buffer and channel buffer data structure are then freed + * automatically when the last reference is given up. + */ +static void relay_close_buf(struct rchan_buf *buf) +{ +//ust// del_timer_sync(&buf->timer); + kref_put(&buf->kref, relay_remove_buf); +} + +//ust// static void setup_callbacks(struct rchan *chan, +//ust// struct rchan_callbacks *cb) +//ust// { +//ust// if (!cb) { +//ust// chan->cb = &default_channel_callbacks; +//ust// return; +//ust// } +//ust// +//ust// if (!cb->create_buf_file) +//ust// cb->create_buf_file = create_buf_file_default_callback; +//ust// if (!cb->remove_buf_file) +//ust// cb->remove_buf_file = remove_buf_file_default_callback; +//ust// chan->cb = cb; +//ust// } + +/** + * relay_hotcpu_callback - CPU hotplug callback + * @nb: notifier block + * @action: hotplug action to take + * @hcpu: CPU number + * + * Returns the success/failure of the operation. (%NOTIFY_OK, %NOTIFY_BAD) + */ +//ust// static int __cpuinit relay_hotcpu_callback(struct notifier_block *nb, +//ust// unsigned long action, +//ust// void *hcpu) +//ust// { +//ust// unsigned int hotcpu = (unsigned long)hcpu; +//ust// struct rchan *chan; +//ust// +//ust// switch (action) { +//ust// case CPU_UP_PREPARE: +//ust// case CPU_UP_PREPARE_FROZEN: +//ust// mutex_lock(&relay_channels_mutex); +//ust// list_for_each_entry(chan, &relay_channels, list) { +//ust// if (chan->buf[hotcpu]) +//ust// continue; +//ust// chan->buf[hotcpu] = relay_open_buf(chan, hotcpu); +//ust// if (!chan->buf[hotcpu]) { +//ust// printk(KERN_ERR +//ust// "relay_hotcpu_callback: cpu %d buffer " +//ust// "creation failed\n", hotcpu); +//ust// mutex_unlock(&relay_channels_mutex); +//ust// return NOTIFY_BAD; +//ust// } +//ust// } +//ust// mutex_unlock(&relay_channels_mutex); +//ust// break; +//ust// case CPU_DEAD: +//ust// case CPU_DEAD_FROZEN: +//ust// /* No need to flush the cpu : will be flushed upon +//ust// * final relay_flush() call. */ +//ust// break; +//ust// } +//ust// return NOTIFY_OK; +//ust// } + +/** + * ltt_relay_open - create a new relay channel + * @base_filename: base name of files to create + * @parent: dentry of parent directory, %NULL for root directory + * @subbuf_size: size of sub-buffers + * @n_subbufs: number of sub-buffers + * @cb: client callback functions + * @private_data: user-defined data + * + * Returns channel pointer if successful, %NULL otherwise. + * + * Creates a channel buffer for each cpu using the sizes and + * attributes specified. The created channel buffer files + * will be named base_filename0...base_filenameN-1. File + * permissions will be %S_IRUSR. + */ +struct rchan *ltt_relay_open(const char *base_filename, + struct dentry *parent, + size_t subbuf_size, + size_t n_subbufs, + void *private_data) +{ + unsigned int i; + struct rchan *chan; +//ust// if (!base_filename) +//ust// return NULL; + + if (!(subbuf_size && n_subbufs)) + return NULL; + + chan = kzalloc(sizeof(struct rchan), GFP_KERNEL); + if (!chan) + return NULL; + + chan->version = LTT_RELAY_CHANNEL_VERSION; + chan->n_subbufs = n_subbufs; + chan->subbuf_size = subbuf_size; + chan->subbuf_size_order = get_count_order(subbuf_size); + chan->alloc_size = FIX_SIZE(subbuf_size * n_subbufs); + chan->parent = parent; + chan->private_data = private_data; +//ust// strlcpy(chan->base_filename, base_filename, NAME_MAX); +//ust// setup_callbacks(chan, cb); + kref_init(&chan->kref); + + mutex_lock(&relay_channels_mutex); +//ust// for_each_online_cpu(i) { + chan->buf = relay_open_buf(chan); + if (!chan->buf) + goto error; +//ust// } + list_add(&chan->list, &relay_channels); + mutex_unlock(&relay_channels_mutex); + + return chan; + +//ust//free_bufs: +//ust// for_each_possible_cpu(i) { +//ust// if (!chan->buf[i]) +//ust// break; +//ust// relay_close_buf(chan->buf[i]); +//ust// } + + error: + kref_put(&chan->kref, relay_destroy_channel); + mutex_unlock(&relay_channels_mutex); + return NULL; +} +//ust// EXPORT_SYMBOL_GPL(ltt_relay_open); + +/** + * ltt_relay_close - close the channel + * @chan: the channel + * + * Closes all channel buffers and frees the channel. + */ +void ltt_relay_close(struct rchan *chan) +{ + unsigned int i; + + if (!chan) + return; + + mutex_lock(&relay_channels_mutex); +//ust// for_each_possible_cpu(i) + if (chan->buf) + relay_close_buf(chan->buf); + + list_del(&chan->list); + kref_put(&chan->kref, relay_destroy_channel); + mutex_unlock(&relay_channels_mutex); +} +//ust// EXPORT_SYMBOL_GPL(ltt_relay_close); + +/* + * Start iteration at the previous element. Skip the real list head. + */ +//ust// struct buf_page *ltt_relay_find_prev_page(struct rchan_buf *buf, +//ust// struct buf_page *page, size_t offset, ssize_t diff_offset) +//ust// { +//ust// struct buf_page *iter; +//ust// size_t orig_iter_off; +//ust// unsigned int i = 0; +//ust// +//ust// orig_iter_off = page->offset; +//ust// list_for_each_entry_reverse(iter, &page->list, list) { +//ust// /* +//ust// * Skip the real list head. +//ust// */ +//ust// if (&iter->list == &buf->pages) +//ust// continue; +//ust// i++; +//ust// if (offset >= iter->offset +//ust// && offset < iter->offset + PAGE_SIZE) { +//ust// #ifdef CONFIG_LTT_RELAY_CHECK_RANDOM_ACCESS +//ust// if (i > 1) { +//ust// printk(KERN_WARNING +//ust// "Backward random access detected in " +//ust// "ltt_relay. Iterations %u, " +//ust// "offset %zu, orig iter->off %zu, " +//ust// "iter->off %zu diff_offset %zd.\n", i, +//ust// offset, orig_iter_off, iter->offset, +//ust// diff_offset); +//ust// WARN_ON(1); +//ust// } +//ust// #endif +//ust// return iter; +//ust// } +//ust// } +//ust// WARN_ON(1); +//ust// return NULL; +//ust// } +//ust// EXPORT_SYMBOL_GPL(ltt_relay_find_prev_page); + +/* + * Start iteration at the next element. Skip the real list head. + */ +//ust// struct buf_page *ltt_relay_find_next_page(struct rchan_buf *buf, +//ust// struct buf_page *page, size_t offset, ssize_t diff_offset) +//ust// { +//ust// struct buf_page *iter; +//ust// unsigned int i = 0; +//ust// size_t orig_iter_off; +//ust// +//ust// orig_iter_off = page->offset; +//ust// list_for_each_entry(iter, &page->list, list) { +//ust// /* +//ust// * Skip the real list head. +//ust// */ +//ust// if (&iter->list == &buf->pages) +//ust// continue; +//ust// i++; +//ust// if (offset >= iter->offset +//ust// && offset < iter->offset + PAGE_SIZE) { +//ust// #ifdef CONFIG_LTT_RELAY_CHECK_RANDOM_ACCESS +//ust// if (i > 1) { +//ust// printk(KERN_WARNING +//ust// "Forward random access detected in " +//ust// "ltt_relay. Iterations %u, " +//ust// "offset %zu, orig iter->off %zu, " +//ust// "iter->off %zu diff_offset %zd.\n", i, +//ust// offset, orig_iter_off, iter->offset, +//ust// diff_offset); +//ust// WARN_ON(1); +//ust// } +//ust// #endif +//ust// return iter; +//ust// } +//ust// } +//ust// WARN_ON(1); +//ust// return NULL; +//ust// } +//ust// EXPORT_SYMBOL_GPL(ltt_relay_find_next_page); + +/** + * ltt_relay_write - write data to a ltt_relay buffer. + * @buf : buffer + * @offset : offset within the buffer + * @src : source address + * @len : length to write + * @page : cached buffer page + * @pagecpy : page size copied so far + */ +void _ltt_relay_write(struct rchan_buf *buf, size_t offset, + const void *src, size_t len, ssize_t cpy) +{ + do { + len -= cpy; + src += cpy; + offset += cpy; + /* + * Underlying layer should never ask for writes across + * subbuffers. + */ + WARN_ON(offset >= buf->buf_size); + + cpy = min_t(size_t, len, buf->buf_size - offset); + ltt_relay_do_copy(buf->buf_data + offset, src, cpy); + } while (unlikely(len != cpy)); +} +//ust// EXPORT_SYMBOL_GPL(_ltt_relay_write); + +/** + * ltt_relay_read - read data from ltt_relay_buffer. + * @buf : buffer + * @offset : offset within the buffer + * @dest : destination address + * @len : length to write + */ +//ust// int ltt_relay_read(struct rchan_buf *buf, size_t offset, +//ust// void *dest, size_t len) +//ust// { +//ust// struct buf_page *page; +//ust// ssize_t pagecpy, orig_len; +//ust// +//ust// orig_len = len; +//ust// offset &= buf->chan->alloc_size - 1; +//ust// page = buf->rpage; +//ust// if (unlikely(!len)) +//ust// return 0; +//ust// for (;;) { +//ust// page = ltt_relay_cache_page(buf, &buf->rpage, page, offset); +//ust// pagecpy = min_t(size_t, len, PAGE_SIZE - (offset & ~PAGE_MASK)); +//ust// memcpy(dest, page_address(page->page) + (offset & ~PAGE_MASK), +//ust// pagecpy); +//ust// len -= pagecpy; +//ust// if (likely(!len)) +//ust// break; +//ust// dest += pagecpy; +//ust// offset += pagecpy; +//ust// /* +//ust// * Underlying layer should never ask for reads across +//ust// * subbuffers. +//ust// */ +//ust// WARN_ON(offset >= buf->chan->alloc_size); +//ust// } +//ust// return orig_len; +//ust// } +//ust// EXPORT_SYMBOL_GPL(ltt_relay_read); + +/** + * ltt_relay_read_get_page - Get a whole page to read from + * @buf : buffer + * @offset : offset within the buffer + */ +//ust// struct buf_page *ltt_relay_read_get_page(struct rchan_buf *buf, size_t offset) +//ust// { +//ust// struct buf_page *page; + +//ust// offset &= buf->chan->alloc_size - 1; +//ust// page = buf->rpage; +//ust// page = ltt_relay_cache_page(buf, &buf->rpage, page, offset); +//ust// return page; +//ust// } +//ust// EXPORT_SYMBOL_GPL(ltt_relay_read_get_page); + +/** + * ltt_relay_offset_address - get address of a location within the buffer + * @buf : buffer + * @offset : offset within the buffer. + * + * Return the address where a given offset is located. + * Should be used to get the current subbuffer header pointer. Given we know + * it's never on a page boundary, it's safe to write directly to this address, + * as long as the write is never bigger than a page size. + */ +void *ltt_relay_offset_address(struct rchan_buf *buf, size_t offset) +{ +//ust// struct buf_page *page; +//ust// unsigned int odd; +//ust// +//ust// offset &= buf->chan->alloc_size - 1; +//ust// odd = !!(offset & buf->chan->subbuf_size); +//ust// page = buf->hpage[odd]; +//ust// if (offset < page->offset || offset >= page->offset + PAGE_SIZE) +//ust// buf->hpage[odd] = page = buf->wpage; +//ust// page = ltt_relay_cache_page(buf, &buf->hpage[odd], page, offset); +//ust// return page_address(page->page) + (offset & ~PAGE_MASK); + return ((char *)buf->buf_data)+offset; + return NULL; +} +//ust// EXPORT_SYMBOL_GPL(ltt_relay_offset_address); + +/** + * relay_file_open - open file op for relay files + * @inode: the inode + * @filp: the file + * + * Increments the channel buffer refcount. + */ +//ust// static int relay_file_open(struct inode *inode, struct file *filp) +//ust// { +//ust// struct rchan_buf *buf = inode->i_private; +//ust// kref_get(&buf->kref); +//ust// filp->private_data = buf; +//ust// +//ust// return nonseekable_open(inode, filp); +//ust// } + +/** + * relay_file_release - release file op for relay files + * @inode: the inode + * @filp: the file + * + * Decrements the channel refcount, as the filesystem is + * no longer using it. + */ +//ust// static int relay_file_release(struct inode *inode, struct file *filp) +//ust// { +//ust// struct rchan_buf *buf = filp->private_data; +//ust// kref_put(&buf->kref, relay_remove_buf); +//ust// +//ust// return 0; +//ust// } + +//ust// const struct file_operations ltt_relay_file_operations = { +//ust// .open = relay_file_open, +//ust// .release = relay_file_release, +//ust// }; +//ust// EXPORT_SYMBOL_GPL(ltt_relay_file_operations); + +//ust// static __init int relay_init(void) +//ust// { +//ust// hotcpu_notifier(relay_hotcpu_callback, 5); +//ust// return 0; +//ust// } + +//ust// module_init(relay_init); +/* + * ltt/ltt-relay.c + * + * (C) Copyright 2005-2008 - Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) + * + * LTTng lockless buffer space management (reader/writer). + * + * Author: + * Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) + * + * Inspired from LTT : + * Karim Yaghmour (karim@opersys.com) + * Tom Zanussi (zanussi@us.ibm.com) + * Bob Wisniewski (bob@watson.ibm.com) + * And from K42 : + * Bob Wisniewski (bob@watson.ibm.com) + * + * Changelog: + * 08/10/08, Cleanup. + * 19/10/05, Complete lockless mechanism. + * 27/05/05, Modular redesign and rewrite. + * + * Userspace reader semantic : + * while (poll fd != POLLHUP) { + * - ioctl RELAY_GET_SUBBUF_SIZE + * while (1) { + * - ioctl GET_SUBBUF + * - splice 1 subbuffer worth of data to a pipe + * - splice the data from pipe to disk/network + * - ioctl PUT_SUBBUF, check error value + * if err val < 0, previous subbuffer was corrupted. + * } + * } + */ + +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include + +#if 0 +#define printk_dbg(fmt, args...) printk(fmt, args) +#else +#define printk_dbg(fmt, args...) +#endif + +/* + * Last TSC comparison functions. Check if the current TSC overflows + * LTT_TSC_BITS bits from the last TSC read. Reads and writes last_tsc + * atomically. + */ + +#if (BITS_PER_LONG == 32) +static inline void save_last_tsc(struct ltt_channel_buf_struct *ltt_buf, + u64 tsc) +{ + ltt_buf->last_tsc = (unsigned long)(tsc >> LTT_TSC_BITS); +} + +static inline int last_tsc_overflow(struct ltt_channel_buf_struct *ltt_buf, + u64 tsc) +{ + unsigned long tsc_shifted = (unsigned long)(tsc >> LTT_TSC_BITS); + + if (unlikely((tsc_shifted - ltt_buf->last_tsc))) + return 1; + else + return 0; +} +#else +static inline void save_last_tsc(struct ltt_channel_buf_struct *ltt_buf, + u64 tsc) +{ + ltt_buf->last_tsc = (unsigned long)tsc; +} + +static inline int last_tsc_overflow(struct ltt_channel_buf_struct *ltt_buf, + u64 tsc) +{ + if (unlikely((tsc - ltt_buf->last_tsc) >> LTT_TSC_BITS)) + return 1; + else + return 0; +} +#endif + +//ust// static struct file_operations ltt_file_operations; + +/* + * A switch is done during tracing or as a final flush after tracing (so it + * won't write in the new sub-buffer). + */ +enum force_switch_mode { FORCE_ACTIVE, FORCE_FLUSH }; + +static int ltt_relay_create_buffer(struct ltt_trace_struct *trace, + struct ltt_channel_struct *ltt_chan, + struct rchan_buf *buf, + unsigned int n_subbufs); + +static void ltt_relay_destroy_buffer(struct ltt_channel_struct *ltt_chan); + +static void ltt_force_switch(struct rchan_buf *buf, + enum force_switch_mode mode); + +/* + * Trace callbacks + */ +static void ltt_buffer_begin_callback(struct rchan_buf *buf, + u64 tsc, unsigned int subbuf_idx) +{ + struct ltt_channel_struct *channel = + (struct ltt_channel_struct *)buf->chan->private_data; + struct ltt_subbuffer_header *header = + (struct ltt_subbuffer_header *) + ltt_relay_offset_address(buf, + subbuf_idx * buf->chan->subbuf_size); + + header->cycle_count_begin = tsc; + header->lost_size = 0xFFFFFFFF; /* for debugging */ + header->buf_size = buf->chan->subbuf_size; + ltt_write_trace_header(channel->trace, header); +} + +/* + * offset is assumed to never be 0 here : never deliver a completely empty + * subbuffer. The lost size is between 0 and subbuf_size-1. + */ +static notrace void ltt_buffer_end_callback(struct rchan_buf *buf, + u64 tsc, unsigned int offset, unsigned int subbuf_idx) +{ + struct ltt_channel_struct *channel = + (struct ltt_channel_struct *)buf->chan->private_data; + struct ltt_channel_buf_struct *ltt_buf = channel->buf; + struct ltt_subbuffer_header *header = + (struct ltt_subbuffer_header *) + ltt_relay_offset_address(buf, + subbuf_idx * buf->chan->subbuf_size); + + header->lost_size = SUBBUF_OFFSET((buf->chan->subbuf_size - offset), + buf->chan); + header->cycle_count_end = tsc; + header->events_lost = local_read(<t_buf->events_lost); + header->subbuf_corrupt = local_read(<t_buf->corrupted_subbuffers); + +} + +void (*wake_consumer)(void *, int) = NULL; + +void relay_set_wake_consumer(void (*wake)(void *, int)) +{ + wake_consumer = wake; +} + +void relay_wake_consumer(void *arg, int finished) +{ + if(wake_consumer) + wake_consumer(arg, finished); +} + +static notrace void ltt_deliver(struct rchan_buf *buf, unsigned int subbuf_idx, + void *subbuf) +{ + struct ltt_channel_struct *channel = + (struct ltt_channel_struct *)buf->chan->private_data; + struct ltt_channel_buf_struct *ltt_buf = channel->buf; + int result; + + result = write(ltt_buf->data_ready_fd_write, "1", 1); + if(result == -1) { + PERROR("write (in ltt_relay_buffer_flush)"); + ERR("this should never happen!"); + } +//ust// atomic_set(<t_buf->wakeup_readers, 1); +} + +static struct dentry *ltt_create_buf_file_callback(struct rchan_buf *buf) +{ + struct ltt_channel_struct *ltt_chan; + int err; +//ust// struct dentry *dentry; + + ltt_chan = buf->chan->private_data; + err = ltt_relay_create_buffer(ltt_chan->trace, ltt_chan, buf, buf->chan->n_subbufs); + if (err) + return ERR_PTR(err); + +//ust// dentry = debugfs_create_file(filename, mode, parent, buf, +//ust// <t_file_operations); +//ust// if (!dentry) +//ust// goto error; +//ust// return dentry; + return NULL; //ust// +//ust//error: + ltt_relay_destroy_buffer(ltt_chan); + return NULL; +} + +static int ltt_remove_buf_file_callback(struct rchan_buf *buf) +{ +//ust// struct rchan_buf *buf = dentry->d_inode->i_private; + struct ltt_channel_struct *ltt_chan = buf->chan->private_data; + +//ust// debugfs_remove(dentry); + ltt_relay_destroy_buffer(ltt_chan); + + return 0; +} + +/* + * Wake writers : + * + * This must be done after the trace is removed from the RCU list so that there + * are no stalled writers. + */ +//ust// static void ltt_relay_wake_writers(struct ltt_channel_buf_struct *ltt_buf) +//ust// { +//ust// +//ust// if (waitqueue_active(<t_buf->write_wait)) +//ust// wake_up_interruptible(<t_buf->write_wait); +//ust// } + +/* + * This function should not be called from NMI interrupt context + */ +static notrace void ltt_buf_unfull(struct rchan_buf *buf, + unsigned int subbuf_idx, + long offset) +{ +//ust// struct ltt_channel_struct *ltt_channel = +//ust// (struct ltt_channel_struct *)buf->chan->private_data; +//ust// struct ltt_channel_buf_struct *ltt_buf = ltt_channel->buf; +//ust// +//ust// ltt_relay_wake_writers(ltt_buf); +} + +/** + * ltt_open - open file op for ltt files + * @inode: opened inode + * @file: opened file + * + * Open implementation. Makes sure only one open instance of a buffer is + * done at a given moment. + */ +//ust// static int ltt_open(struct inode *inode, struct file *file) +//ust// { +//ust// struct rchan_buf *buf = inode->i_private; +//ust// struct ltt_channel_struct *ltt_channel = +//ust// (struct ltt_channel_struct *)buf->chan->private_data; +//ust// struct ltt_channel_buf_struct *ltt_buf = +//ust// percpu_ptr(ltt_channel->buf, buf->cpu); +//ust// +//ust// if (!atomic_long_add_unless(<t_buf->active_readers, 1, 1)) +//ust// return -EBUSY; +//ust// return ltt_relay_file_operations.open(inode, file); +//ust// } + +/** + * ltt_release - release file op for ltt files + * @inode: opened inode + * @file: opened file + * + * Release implementation. + */ +//ust// static int ltt_release(struct inode *inode, struct file *file) +//ust// { +//ust// struct rchan_buf *buf = inode->i_private; +//ust// struct ltt_channel_struct *ltt_channel = +//ust// (struct ltt_channel_struct *)buf->chan->private_data; +//ust// struct ltt_channel_buf_struct *ltt_buf = +//ust// percpu_ptr(ltt_channel->buf, buf->cpu); +//ust// int ret; +//ust// +//ust// WARN_ON(atomic_long_read(<t_buf->active_readers) != 1); +//ust// atomic_long_dec(<t_buf->active_readers); +//ust// ret = ltt_relay_file_operations.release(inode, file); +//ust// WARN_ON(ret); +//ust// return ret; +//ust// } + +/** + * ltt_poll - file op for ltt files + * @filp: the file + * @wait: poll table + * + * Poll implementation. + */ +//ust// static unsigned int ltt_poll(struct file *filp, poll_table *wait) +//ust// { +//ust// unsigned int mask = 0; +//ust// struct inode *inode = filp->f_dentry->d_inode; +//ust// struct rchan_buf *buf = inode->i_private; +//ust// struct ltt_channel_struct *ltt_channel = +//ust// (struct ltt_channel_struct *)buf->chan->private_data; +//ust// struct ltt_channel_buf_struct *ltt_buf = +//ust// percpu_ptr(ltt_channel->buf, buf->cpu); +//ust// +//ust// if (filp->f_mode & FMODE_READ) { +//ust// poll_wait_set_exclusive(wait); +//ust// poll_wait(filp, &buf->read_wait, wait); +//ust// +//ust// WARN_ON(atomic_long_read(<t_buf->active_readers) != 1); +//ust// if (SUBBUF_TRUNC(local_read(<t_buf->offset), +//ust// buf->chan) +//ust// - SUBBUF_TRUNC(atomic_long_read(<t_buf->consumed), +//ust// buf->chan) +//ust// == 0) { +//ust// if (buf->finalized) +//ust// return POLLHUP; +//ust// else +//ust// return 0; +//ust// } else { +//ust// struct rchan *rchan = +//ust// ltt_channel->trans_channel_data; +//ust// if (SUBBUF_TRUNC(local_read(<t_buf->offset), +//ust// buf->chan) +//ust// - SUBBUF_TRUNC(atomic_long_read( +//ust// <t_buf->consumed), +//ust// buf->chan) +//ust// >= rchan->alloc_size) +//ust// return POLLPRI | POLLRDBAND; +//ust// else +//ust// return POLLIN | POLLRDNORM; +//ust// } +//ust// } +//ust// return mask; +//ust// } + +int ltt_do_get_subbuf(struct rchan_buf *buf, struct ltt_channel_buf_struct *ltt_buf, long *pconsumed_old) +{ + struct ltt_channel_struct *ltt_channel = (struct ltt_channel_struct *)buf->chan->private_data; + long consumed_old, consumed_idx, commit_count, write_offset; + consumed_old = atomic_long_read(<t_buf->consumed); + consumed_idx = SUBBUF_INDEX(consumed_old, buf->chan); + commit_count = local_read(<t_buf->commit_count[consumed_idx]); + /* + * Make sure we read the commit count before reading the buffer + * data and the write offset. Correct consumed offset ordering + * wrt commit count is insured by the use of cmpxchg to update + * the consumed offset. + */ + smp_rmb(); + write_offset = local_read(<t_buf->offset); + /* + * Check that the subbuffer we are trying to consume has been + * already fully committed. + */ + if (((commit_count - buf->chan->subbuf_size) + & ltt_channel->commit_count_mask) + - (BUFFER_TRUNC(consumed_old, buf->chan) + >> ltt_channel->n_subbufs_order) + != 0) { + return -EAGAIN; + } + /* + * Check that we are not about to read the same subbuffer in + * which the writer head is. + */ + if ((SUBBUF_TRUNC(write_offset, buf->chan) + - SUBBUF_TRUNC(consumed_old, buf->chan)) + == 0) { + return -EAGAIN; + } + + *pconsumed_old = consumed_old; + return 0; +} + +int ltt_do_put_subbuf(struct rchan_buf *buf, struct ltt_channel_buf_struct *ltt_buf, u32 uconsumed_old) +{ + long consumed_new, consumed_old; + + consumed_old = atomic_long_read(<t_buf->consumed); + consumed_old = consumed_old & (~0xFFFFFFFFL); + consumed_old = consumed_old | uconsumed_old; + consumed_new = SUBBUF_ALIGN(consumed_old, buf->chan); + +//ust// spin_lock(<t_buf->full_lock); + if (atomic_long_cmpxchg(<t_buf->consumed, consumed_old, + consumed_new) + != consumed_old) { + /* We have been pushed by the writer : the last + * buffer read _is_ corrupted! It can also + * happen if this is a buffer we never got. */ +//ust// spin_unlock(<t_buf->full_lock); + return -EIO; + } else { + /* tell the client that buffer is now unfull */ + int index; + long data; + index = SUBBUF_INDEX(consumed_old, buf->chan); + data = BUFFER_OFFSET(consumed_old, buf->chan); + ltt_buf_unfull(buf, index, data); +//ust// spin_unlock(<t_buf->full_lock); + } + return 0; +} + +/** + * ltt_ioctl - control on the debugfs file + * + * @inode: the inode + * @filp: the file + * @cmd: the command + * @arg: command arg + * + * This ioctl implements three commands necessary for a minimal + * producer/consumer implementation : + * RELAY_GET_SUBBUF + * Get the next sub buffer that can be read. It never blocks. + * RELAY_PUT_SUBBUF + * Release the currently read sub-buffer. Parameter is the last + * put subbuffer (returned by GET_SUBBUF). + * RELAY_GET_N_BUBBUFS + * returns the number of sub buffers in the per cpu channel. + * RELAY_GET_SUBBUF_SIZE + * returns the size of the sub buffers. + */ +//ust// static int ltt_ioctl(struct inode *inode, struct file *filp, +//ust// unsigned int cmd, unsigned long arg) +//ust// { +//ust// struct rchan_buf *buf = inode->i_private; +//ust// struct ltt_channel_struct *ltt_channel = +//ust// (struct ltt_channel_struct *)buf->chan->private_data; +//ust// struct ltt_channel_buf_struct *ltt_buf = +//ust// percpu_ptr(ltt_channel->buf, buf->cpu); +//ust// u32 __user *argp = (u32 __user *)arg; +//ust// +//ust// WARN_ON(atomic_long_read(<t_buf->active_readers) != 1); +//ust// switch (cmd) { +//ust// case RELAY_GET_SUBBUF: +//ust// { +//ust// int ret; +//ust// ret = ltt_do_get_subbuf(buf, ltt_buf, &consumed_old); +//ust// if(ret < 0) +//ust// return ret; +//ust// return put_user((u32)consumed_old, argp); +//ust// } +//ust// case RELAY_PUT_SUBBUF: +//ust// { +//ust// int ret; +//ust// u32 uconsumed_old; +//ust// ret = get_user(uconsumed_old, argp); +//ust// if (ret) +//ust// return ret; /* will return -EFAULT */ +//ust// return ltt_do_put_subbuf(buf, ltt_buf, uconsumed_old); +//ust// } +//ust// case RELAY_GET_N_SUBBUFS: +//ust// return put_user((u32)buf->chan->n_subbufs, argp); +//ust// break; +//ust// case RELAY_GET_SUBBUF_SIZE: +//ust// return put_user((u32)buf->chan->subbuf_size, argp); +//ust// break; +//ust// default: +//ust// return -ENOIOCTLCMD; +//ust// } +//ust// return 0; +//ust// } + +//ust// #ifdef CONFIG_COMPAT +//ust// static long ltt_compat_ioctl(struct file *file, unsigned int cmd, +//ust// unsigned long arg) +//ust// { +//ust// long ret = -ENOIOCTLCMD; +//ust// +//ust// lock_kernel(); +//ust// ret = ltt_ioctl(file->f_dentry->d_inode, file, cmd, arg); +//ust// unlock_kernel(); +//ust// +//ust// return ret; +//ust// } +//ust// #endif + +//ust// static void ltt_relay_pipe_buf_release(struct pipe_inode_info *pipe, +//ust// struct pipe_buffer *pbuf) +//ust// { +//ust// } +//ust// +//ust// static struct pipe_buf_operations ltt_relay_pipe_buf_ops = { +//ust// .can_merge = 0, +//ust// .map = generic_pipe_buf_map, +//ust// .unmap = generic_pipe_buf_unmap, +//ust// .confirm = generic_pipe_buf_confirm, +//ust// .release = ltt_relay_pipe_buf_release, +//ust// .steal = generic_pipe_buf_steal, +//ust// .get = generic_pipe_buf_get, +//ust// }; + +//ust// static void ltt_relay_page_release(struct splice_pipe_desc *spd, unsigned int i) +//ust// { +//ust// } + +/* + * subbuf_splice_actor - splice up to one subbuf's worth of data + */ +//ust// static int subbuf_splice_actor(struct file *in, +//ust// loff_t *ppos, +//ust// struct pipe_inode_info *pipe, +//ust// size_t len, +//ust// unsigned int flags) +//ust// { +//ust// struct rchan_buf *buf = in->private_data; +//ust// struct ltt_channel_struct *ltt_channel = +//ust// (struct ltt_channel_struct *)buf->chan->private_data; +//ust// struct ltt_channel_buf_struct *ltt_buf = +//ust// percpu_ptr(ltt_channel->buf, buf->cpu); +//ust// unsigned int poff, subbuf_pages, nr_pages; +//ust// struct page *pages[PIPE_BUFFERS]; +//ust// struct partial_page partial[PIPE_BUFFERS]; +//ust// struct splice_pipe_desc spd = { +//ust// .pages = pages, +//ust// .nr_pages = 0, +//ust// .partial = partial, +//ust// .flags = flags, +//ust// .ops = <t_relay_pipe_buf_ops, +//ust// .spd_release = ltt_relay_page_release, +//ust// }; +//ust// long consumed_old, consumed_idx, roffset; +//ust// unsigned long bytes_avail; +//ust// +//ust// /* +//ust// * Check that a GET_SUBBUF ioctl has been done before. +//ust// */ +//ust// WARN_ON(atomic_long_read(<t_buf->active_readers) != 1); +//ust// consumed_old = atomic_long_read(<t_buf->consumed); +//ust// consumed_old += *ppos; +//ust// consumed_idx = SUBBUF_INDEX(consumed_old, buf->chan); +//ust// +//ust// /* +//ust// * Adjust read len, if longer than what is available +//ust// */ +//ust// bytes_avail = SUBBUF_TRUNC(local_read(<t_buf->offset), buf->chan) +//ust// - consumed_old; +//ust// WARN_ON(bytes_avail > buf->chan->alloc_size); +//ust// len = min_t(size_t, len, bytes_avail); +//ust// subbuf_pages = bytes_avail >> PAGE_SHIFT; +//ust// nr_pages = min_t(unsigned int, subbuf_pages, PIPE_BUFFERS); +//ust// roffset = consumed_old & PAGE_MASK; +//ust// poff = consumed_old & ~PAGE_MASK; +//ust// printk_dbg(KERN_DEBUG "SPLICE actor len %zu pos %zd write_pos %ld\n", +//ust// len, (ssize_t)*ppos, local_read(<t_buf->offset)); +//ust// +//ust// for (; spd.nr_pages < nr_pages; spd.nr_pages++) { +//ust// unsigned int this_len; +//ust// struct buf_page *page; +//ust// +//ust// if (!len) +//ust// break; +//ust// printk_dbg(KERN_DEBUG "SPLICE actor loop len %zu roffset %ld\n", +//ust// len, roffset); +//ust// +//ust// this_len = PAGE_SIZE - poff; +//ust// page = ltt_relay_read_get_page(buf, roffset); +//ust// spd.pages[spd.nr_pages] = page->page; +//ust// spd.partial[spd.nr_pages].offset = poff; +//ust// spd.partial[spd.nr_pages].len = this_len; +//ust// +//ust// poff = 0; +//ust// roffset += PAGE_SIZE; +//ust// len -= this_len; +//ust// } +//ust// +//ust// if (!spd.nr_pages) +//ust// return 0; +//ust// +//ust// return splice_to_pipe(pipe, &spd); +//ust// } + +//ust// static ssize_t ltt_relay_file_splice_read(struct file *in, +//ust// loff_t *ppos, +//ust// struct pipe_inode_info *pipe, +//ust// size_t len, +//ust// unsigned int flags) +//ust// { +//ust// ssize_t spliced; +//ust// int ret; +//ust// +//ust// ret = 0; +//ust// spliced = 0; +//ust// +//ust// printk_dbg(KERN_DEBUG "SPLICE read len %zu pos %zd\n", +//ust// len, (ssize_t)*ppos); +//ust// while (len && !spliced) { +//ust// ret = subbuf_splice_actor(in, ppos, pipe, len, flags); +//ust// printk_dbg(KERN_DEBUG "SPLICE read loop ret %d\n", ret); +//ust// if (ret < 0) +//ust// break; +//ust// else if (!ret) { +//ust// if (flags & SPLICE_F_NONBLOCK) +//ust// ret = -EAGAIN; +//ust// break; +//ust// } +//ust// +//ust// *ppos += ret; +//ust// if (ret > len) +//ust// len = 0; +//ust// else +//ust// len -= ret; +//ust// spliced += ret; +//ust// } +//ust// +//ust// if (spliced) +//ust// return spliced; +//ust// +//ust// return ret; +//ust// } + +static void ltt_relay_print_subbuffer_errors( + struct ltt_channel_struct *ltt_chan, + long cons_off) +{ + struct rchan *rchan = ltt_chan->trans_channel_data; + struct ltt_channel_buf_struct *ltt_buf = ltt_chan->buf; + long cons_idx, commit_count, write_offset; + + cons_idx = SUBBUF_INDEX(cons_off, rchan); + commit_count = local_read(<t_buf->commit_count[cons_idx]); + /* + * No need to order commit_count and write_offset reads because we + * execute after trace is stopped when there are no readers left. + */ + write_offset = local_read(<t_buf->offset); + printk(KERN_WARNING + "LTT : unread channel %s offset is %ld " + "and cons_off : %ld\n", + ltt_chan->channel_name, write_offset, cons_off); + /* Check each sub-buffer for non filled commit count */ + if (((commit_count - rchan->subbuf_size) & ltt_chan->commit_count_mask) + - (BUFFER_TRUNC(cons_off, rchan) >> ltt_chan->n_subbufs_order) + != 0) + printk(KERN_ALERT + "LTT : %s : subbuffer %lu has non filled " + "commit count %lu.\n", + ltt_chan->channel_name, cons_idx, commit_count); + printk(KERN_ALERT "LTT : %s : commit count : %lu, subbuf size %zd\n", + ltt_chan->channel_name, commit_count, + rchan->subbuf_size); +} + +static void ltt_relay_print_errors(struct ltt_trace_struct *trace, + struct ltt_channel_struct *ltt_chan) +{ + struct rchan *rchan = ltt_chan->trans_channel_data; + struct ltt_channel_buf_struct *ltt_buf = ltt_chan->buf; + long cons_off; + + for (cons_off = atomic_long_read(<t_buf->consumed); + (SUBBUF_TRUNC(local_read(<t_buf->offset), + rchan) + - cons_off) > 0; + cons_off = SUBBUF_ALIGN(cons_off, rchan)) + ltt_relay_print_subbuffer_errors(ltt_chan, cons_off); +} + +static void ltt_relay_print_buffer_errors(struct ltt_channel_struct *ltt_chan) +{ + struct ltt_trace_struct *trace = ltt_chan->trace; + struct ltt_channel_buf_struct *ltt_buf = ltt_chan->buf; + + if (local_read(<t_buf->events_lost)) + printk(KERN_ALERT + "LTT : %s : %ld events lost " + "in %s channel.\n", + ltt_chan->channel_name, + local_read(<t_buf->events_lost), + ltt_chan->channel_name); + if (local_read(<t_buf->corrupted_subbuffers)) + printk(KERN_ALERT + "LTT : %s : %ld corrupted subbuffers " + "in %s channel.\n", + ltt_chan->channel_name, + local_read(<t_buf->corrupted_subbuffers), + ltt_chan->channel_name); + + ltt_relay_print_errors(trace, ltt_chan); +} + +static void ltt_relay_remove_dirs(struct ltt_trace_struct *trace) +{ +//ust// debugfs_remove(trace->dentry.trace_root); +} + +static void ltt_relay_release_channel(struct kref *kref) +{ + struct ltt_channel_struct *ltt_chan = container_of(kref, + struct ltt_channel_struct, kref); + free(ltt_chan->buf); +} + +/* + * Create ltt buffer. + */ +//ust// static int ltt_relay_create_buffer(struct ltt_trace_struct *trace, +//ust// struct ltt_channel_struct *ltt_chan, struct rchan_buf *buf, +//ust// unsigned int cpu, unsigned int n_subbufs) +//ust// { +//ust// struct ltt_channel_buf_struct *ltt_buf = +//ust// percpu_ptr(ltt_chan->buf, cpu); +//ust// unsigned int j; +//ust// +//ust// ltt_buf->commit_count = +//ust// kzalloc_node(sizeof(ltt_buf->commit_count) * n_subbufs, +//ust// GFP_KERNEL, cpu_to_node(cpu)); +//ust// if (!ltt_buf->commit_count) +//ust// return -ENOMEM; +//ust// kref_get(&trace->kref); +//ust// kref_get(&trace->ltt_transport_kref); +//ust// kref_get(<t_chan->kref); +//ust// local_set(<t_buf->offset, ltt_subbuffer_header_size()); +//ust// atomic_long_set(<t_buf->consumed, 0); +//ust// atomic_long_set(<t_buf->active_readers, 0); +//ust// for (j = 0; j < n_subbufs; j++) +//ust// local_set(<t_buf->commit_count[j], 0); +//ust// init_waitqueue_head(<t_buf->write_wait); +//ust// atomic_set(<t_buf->wakeup_readers, 0); +//ust// spin_lock_init(<t_buf->full_lock); +//ust// +//ust// ltt_buffer_begin_callback(buf, trace->start_tsc, 0); +//ust// /* atomic_add made on local variable on data that belongs to +//ust// * various CPUs : ok because tracing not started (for this cpu). */ +//ust// local_add(ltt_subbuffer_header_size(), <t_buf->commit_count[0]); +//ust// +//ust// local_set(<t_buf->events_lost, 0); +//ust// local_set(<t_buf->corrupted_subbuffers, 0); +//ust// +//ust// return 0; +//ust// } + +static int ltt_relay_create_buffer(struct ltt_trace_struct *trace, + struct ltt_channel_struct *ltt_chan, struct rchan_buf *buf, + unsigned int n_subbufs) +{ + struct ltt_channel_buf_struct *ltt_buf = ltt_chan->buf; + unsigned int j; + int fds[2]; + int result; + +//ust// ltt_buf->commit_count = +//ust// zmalloc(sizeof(ltt_buf->commit_count) * n_subbufs); +//ust// if (!ltt_buf->commit_count) +//ust// return -ENOMEM; + kref_get(&trace->kref); + kref_get(&trace->ltt_transport_kref); + kref_get(<t_chan->kref); + local_set(<t_buf->offset, ltt_subbuffer_header_size()); + atomic_long_set(<t_buf->consumed, 0); + atomic_long_set(<t_buf->active_readers, 0); + for (j = 0; j < n_subbufs; j++) + local_set(<t_buf->commit_count[j], 0); +//ust// init_waitqueue_head(<t_buf->write_wait); +//ust// atomic_set(<t_buf->wakeup_readers, 0); +//ust// spin_lock_init(<t_buf->full_lock); + + ltt_buffer_begin_callback(buf, trace->start_tsc, 0); + + local_add(ltt_subbuffer_header_size(), <t_buf->commit_count[0]); + + local_set(<t_buf->events_lost, 0); + local_set(<t_buf->corrupted_subbuffers, 0); + + result = pipe(fds); + if(result == -1) { + PERROR("pipe"); + return -1; + } + ltt_buf->data_ready_fd_read = fds[0]; + ltt_buf->data_ready_fd_write = fds[1]; + + return 0; +} + +static void ltt_relay_destroy_buffer(struct ltt_channel_struct *ltt_chan) +{ + struct ltt_trace_struct *trace = ltt_chan->trace; + struct ltt_channel_buf_struct *ltt_buf = ltt_chan->buf; + + kref_put(<t_chan->trace->ltt_transport_kref, + ltt_release_transport); + ltt_relay_print_buffer_errors(ltt_chan); +//ust// kfree(ltt_buf->commit_count); +//ust// ltt_buf->commit_count = NULL; + kref_put(<t_chan->kref, ltt_relay_release_channel); + kref_put(&trace->kref, ltt_release_trace); +//ust// wake_up_interruptible(&trace->kref_wq); +} + +static void ltt_chan_alloc_ltt_buf(struct ltt_channel_struct *ltt_chan) +{ + void *ptr; + int result; + + /* Get one page */ + /* FIXME: increase size if we have a commit_count array that overflows the page */ + size_t size = PAGE_ALIGN(1); + + result = ltt_chan->buf_shmid = shmget(getpid(), size, IPC_CREAT | IPC_EXCL | 0700); + if(ltt_chan->buf_shmid == -1) { + PERROR("shmget"); + return -1; + } + + ptr = shmat(ltt_chan->buf_shmid, NULL, 0); + if(ptr == (void *) -1) { + perror("shmat"); + goto destroy_shmem; + } + + /* Already mark the shared memory for destruction. This will occur only + * when all users have detached. + */ + result = shmctl(ltt_chan->buf_shmid, IPC_RMID, NULL); + if(result == -1) { + perror("shmctl"); + return -1; + } + + ltt_chan->buf = ptr; + + return 0; + + destroy_shmem: + result = shmctl(ltt_chan->buf_shmid, IPC_RMID, NULL); + if(result == -1) { + perror("shmctl"); + } + + return -1; +} + +/* + * Create channel. + */ +static int ltt_relay_create_channel(const char *trace_name, + struct ltt_trace_struct *trace, struct dentry *dir, + const char *channel_name, struct ltt_channel_struct *ltt_chan, + unsigned int subbuf_size, unsigned int n_subbufs, + int overwrite) +{ + char *tmpname; + unsigned int tmpname_len; + int err = 0; + + tmpname = kmalloc(PATH_MAX, GFP_KERNEL); + if (!tmpname) + return EPERM; + if (overwrite) { + strncpy(tmpname, LTT_FLIGHT_PREFIX, PATH_MAX-1); + strncat(tmpname, channel_name, + PATH_MAX-1-sizeof(LTT_FLIGHT_PREFIX)); + } else { + strncpy(tmpname, channel_name, PATH_MAX-1); + } + strncat(tmpname, "_", PATH_MAX-1-strlen(tmpname)); + + kref_init(<t_chan->kref); + + ltt_chan->trace = trace; + ltt_chan->buffer_begin = ltt_buffer_begin_callback; + ltt_chan->buffer_end = ltt_buffer_end_callback; + ltt_chan->overwrite = overwrite; + ltt_chan->n_subbufs_order = get_count_order(n_subbufs); + ltt_chan->commit_count_mask = (~0UL >> ltt_chan->n_subbufs_order); +//ust// ltt_chan->buf = percpu_alloc_mask(sizeof(struct ltt_channel_buf_struct), GFP_KERNEL, cpu_possible_map); + + ltt_chan_alloc_ltt_buf(ltt_chan); + +//ust// ltt_chan->buf = malloc(sizeof(struct ltt_channel_buf_struct)); + if (!ltt_chan->buf) + goto alloc_error; + ltt_chan->trans_channel_data = ltt_relay_open(tmpname, + dir, + subbuf_size, + n_subbufs, + ltt_chan); + tmpname_len = strlen(tmpname); + if (tmpname_len > 0) { + /* Remove final _ for pretty printing */ + tmpname[tmpname_len-1] = '\0'; + } + if (ltt_chan->trans_channel_data == NULL) { + printk(KERN_ERR "LTT : Can't open %s channel for trace %s\n", + tmpname, trace_name); + goto relay_open_error; + } + + err = 0; + goto end; + +relay_open_error: +//ust// percpu_free(ltt_chan->buf); +alloc_error: + err = EPERM; +end: + kfree(tmpname); + return err; +} + +static int ltt_relay_create_dirs(struct ltt_trace_struct *new_trace) +{ +//ust// new_trace->dentry.trace_root = debugfs_create_dir(new_trace->trace_name, +//ust// get_ltt_root()); +//ust// if (new_trace->dentry.trace_root == NULL) { +//ust// printk(KERN_ERR "LTT : Trace directory name %s already taken\n", +//ust// new_trace->trace_name); +//ust// return EEXIST; +//ust// } + +//ust// new_trace->callbacks.create_buf_file = ltt_create_buf_file_callback; +//ust// new_trace->callbacks.remove_buf_file = ltt_remove_buf_file_callback; + + return 0; +} + +/* + * LTTng channel flush function. + * + * Must be called when no tracing is active in the channel, because of + * accesses across CPUs. + */ +static notrace void ltt_relay_buffer_flush(struct rchan_buf *buf) +{ + struct ltt_channel_struct *channel = + (struct ltt_channel_struct *)buf->chan->private_data; + struct ltt_channel_buf_struct *ltt_buf = channel->buf; + int result; + + buf->finalized = 1; + ltt_force_switch(buf, FORCE_FLUSH); + + result = write(ltt_buf->data_ready_fd_write, "1", 1); + if(result == -1) { + PERROR("write (in ltt_relay_buffer_flush)"); + ERR("this should never happen!"); + } +} + +static void ltt_relay_async_wakeup_chan(struct ltt_channel_struct *ltt_channel) +{ +//ust// unsigned int i; +//ust// struct rchan *rchan = ltt_channel->trans_channel_data; +//ust// +//ust// for_each_possible_cpu(i) { +//ust// struct ltt_channel_buf_struct *ltt_buf = +//ust// percpu_ptr(ltt_channel->buf, i); +//ust// +//ust// if (atomic_read(<t_buf->wakeup_readers) == 1) { +//ust// atomic_set(<t_buf->wakeup_readers, 0); +//ust// wake_up_interruptible(&rchan->buf[i]->read_wait); +//ust// } +//ust// } +} + +static void ltt_relay_finish_buffer(struct ltt_channel_struct *ltt_channel) +{ + struct rchan *rchan = ltt_channel->trans_channel_data; + int result; + + if (rchan->buf) { + struct ltt_channel_buf_struct *ltt_buf = ltt_channel->buf; + ltt_relay_buffer_flush(rchan->buf); +//ust// ltt_relay_wake_writers(ltt_buf); + /* closing the pipe tells the consumer the buffer is finished */ + + //result = write(ltt_buf->data_ready_fd_write, "D", 1); + //if(result == -1) { + // PERROR("write (in ltt_relay_finish_buffer)"); + // ERR("this should never happen!"); + //} + close(ltt_buf->data_ready_fd_write); + } +} + + +static void ltt_relay_finish_channel(struct ltt_channel_struct *ltt_channel) +{ + unsigned int i; + +//ust// for_each_possible_cpu(i) + ltt_relay_finish_buffer(ltt_channel); +} + +static void ltt_relay_remove_channel(struct ltt_channel_struct *channel) +{ + struct rchan *rchan = channel->trans_channel_data; + + ltt_relay_close(rchan); + kref_put(&channel->kref, ltt_relay_release_channel); +} + +struct ltt_reserve_switch_offsets { + long begin, end, old; + long begin_switch, end_switch_current, end_switch_old; + long commit_count, reserve_commit_diff; + size_t before_hdr_pad, size; +}; + +/* + * Returns : + * 0 if ok + * !0 if execution must be aborted. + */ +static inline int ltt_relay_try_reserve( + struct ltt_channel_struct *ltt_channel, + struct ltt_channel_buf_struct *ltt_buf, struct rchan *rchan, + struct rchan_buf *buf, + struct ltt_reserve_switch_offsets *offsets, size_t data_size, + u64 *tsc, unsigned int *rflags, int largest_align) +{ + offsets->begin = local_read(<t_buf->offset); + offsets->old = offsets->begin; + offsets->begin_switch = 0; + offsets->end_switch_current = 0; + offsets->end_switch_old = 0; + + *tsc = trace_clock_read64(); + if (last_tsc_overflow(ltt_buf, *tsc)) + *rflags = LTT_RFLAG_ID_SIZE_TSC; + + if (SUBBUF_OFFSET(offsets->begin, buf->chan) == 0) { + offsets->begin_switch = 1; /* For offsets->begin */ + } else { + offsets->size = ltt_get_header_size(ltt_channel, + offsets->begin, data_size, + &offsets->before_hdr_pad, *rflags); + offsets->size += ltt_align(offsets->begin + offsets->size, + largest_align) + + data_size; + if ((SUBBUF_OFFSET(offsets->begin, buf->chan) + offsets->size) + > buf->chan->subbuf_size) { + offsets->end_switch_old = 1; /* For offsets->old */ + offsets->begin_switch = 1; /* For offsets->begin */ + } + } + if (offsets->begin_switch) { + long subbuf_index; + + if (offsets->end_switch_old) + offsets->begin = SUBBUF_ALIGN(offsets->begin, + buf->chan); + offsets->begin = offsets->begin + ltt_subbuffer_header_size(); + /* Test new buffer integrity */ + subbuf_index = SUBBUF_INDEX(offsets->begin, buf->chan); + offsets->reserve_commit_diff = + (BUFFER_TRUNC(offsets->begin, buf->chan) + >> ltt_channel->n_subbufs_order) + - (local_read(<t_buf->commit_count[subbuf_index]) + & ltt_channel->commit_count_mask); + if (offsets->reserve_commit_diff == 0) { + /* Next buffer not corrupted. */ + if (!ltt_channel->overwrite && + (SUBBUF_TRUNC(offsets->begin, buf->chan) + - SUBBUF_TRUNC(atomic_long_read( + <t_buf->consumed), + buf->chan)) + >= rchan->alloc_size) { + /* + * We do not overwrite non consumed buffers + * and we are full : event is lost. + */ + local_inc(<t_buf->events_lost); + return -1; + } else { + /* + * next buffer not corrupted, we are either in + * overwrite mode or the buffer is not full. + * It's safe to write in this new subbuffer. + */ + } + } else { + /* + * Next subbuffer corrupted. Force pushing reader even + * in normal mode. It's safe to write in this new + * subbuffer. + */ + } + offsets->size = ltt_get_header_size(ltt_channel, + offsets->begin, data_size, + &offsets->before_hdr_pad, *rflags); + offsets->size += ltt_align(offsets->begin + offsets->size, + largest_align) + + data_size; + if ((SUBBUF_OFFSET(offsets->begin, buf->chan) + offsets->size) + > buf->chan->subbuf_size) { + /* + * Event too big for subbuffers, report error, don't + * complete the sub-buffer switch. + */ + local_inc(<t_buf->events_lost); + return -1; + } else { + /* + * We just made a successful buffer switch and the event + * fits in the new subbuffer. Let's write. + */ + } + } else { + /* + * Event fits in the current buffer and we are not on a switch + * boundary. It's safe to write. + */ + } + offsets->end = offsets->begin + offsets->size; + + if ((SUBBUF_OFFSET(offsets->end, buf->chan)) == 0) { + /* + * The offset_end will fall at the very beginning of the next + * subbuffer. + */ + offsets->end_switch_current = 1; /* For offsets->begin */ + } + return 0; +} + +/* + * Returns : + * 0 if ok + * !0 if execution must be aborted. + */ +static inline int ltt_relay_try_switch( + enum force_switch_mode mode, + struct ltt_channel_struct *ltt_channel, + struct ltt_channel_buf_struct *ltt_buf, struct rchan *rchan, + struct rchan_buf *buf, + struct ltt_reserve_switch_offsets *offsets, + u64 *tsc) +{ + long subbuf_index; + + offsets->begin = local_read(<t_buf->offset); + offsets->old = offsets->begin; + offsets->begin_switch = 0; + offsets->end_switch_old = 0; + + *tsc = trace_clock_read64(); + + if (SUBBUF_OFFSET(offsets->begin, buf->chan) != 0) { + offsets->begin = SUBBUF_ALIGN(offsets->begin, buf->chan); + offsets->end_switch_old = 1; + } else { + /* we do not have to switch : buffer is empty */ + return -1; + } + if (mode == FORCE_ACTIVE) + offsets->begin += ltt_subbuffer_header_size(); + /* + * Always begin_switch in FORCE_ACTIVE mode. + * Test new buffer integrity + */ + subbuf_index = SUBBUF_INDEX(offsets->begin, buf->chan); + offsets->reserve_commit_diff = + (BUFFER_TRUNC(offsets->begin, buf->chan) + >> ltt_channel->n_subbufs_order) + - (local_read(<t_buf->commit_count[subbuf_index]) + & ltt_channel->commit_count_mask); + if (offsets->reserve_commit_diff == 0) { + /* Next buffer not corrupted. */ + if (mode == FORCE_ACTIVE + && !ltt_channel->overwrite + && offsets->begin - atomic_long_read(<t_buf->consumed) + >= rchan->alloc_size) { + /* + * We do not overwrite non consumed buffers and we are + * full : ignore switch while tracing is active. + */ + return -1; + } + } else { + /* + * Next subbuffer corrupted. Force pushing reader even in normal + * mode + */ + } + offsets->end = offsets->begin; + return 0; +} + +static inline void ltt_reserve_push_reader( + struct ltt_channel_struct *ltt_channel, + struct ltt_channel_buf_struct *ltt_buf, + struct rchan *rchan, + struct rchan_buf *buf, + struct ltt_reserve_switch_offsets *offsets) +{ + long consumed_old, consumed_new; + + do { + consumed_old = atomic_long_read(<t_buf->consumed); + /* + * If buffer is in overwrite mode, push the reader consumed + * count if the write position has reached it and we are not + * at the first iteration (don't push the reader farther than + * the writer). This operation can be done concurrently by many + * writers in the same buffer, the writer being at the farthest + * write position sub-buffer index in the buffer being the one + * which will win this loop. + * If the buffer is not in overwrite mode, pushing the reader + * only happens if a sub-buffer is corrupted. + */ + if ((SUBBUF_TRUNC(offsets->end-1, buf->chan) + - SUBBUF_TRUNC(consumed_old, buf->chan)) + >= rchan->alloc_size) + consumed_new = SUBBUF_ALIGN(consumed_old, buf->chan); + else { + consumed_new = consumed_old; + break; + } + } while (atomic_long_cmpxchg(<t_buf->consumed, consumed_old, + consumed_new) != consumed_old); + + if (consumed_old != consumed_new) { + /* + * Reader pushed : we are the winner of the push, we can + * therefore reequilibrate reserve and commit. Atomic increment + * of the commit count permits other writers to play around + * with this variable before us. We keep track of + * corrupted_subbuffers even in overwrite mode : + * we never want to write over a non completely committed + * sub-buffer : possible causes : the buffer size is too low + * compared to the unordered data input, or there is a writer + * that died between the reserve and the commit. + */ + if (offsets->reserve_commit_diff) { + /* + * We have to alter the sub-buffer commit count. + * We do not deliver the previous subbuffer, given it + * was either corrupted or not consumed (overwrite + * mode). + */ + local_add(offsets->reserve_commit_diff, + <t_buf->commit_count[ + SUBBUF_INDEX(offsets->begin, + buf->chan)]); + if (!ltt_channel->overwrite + || offsets->reserve_commit_diff + != rchan->subbuf_size) { + /* + * The reserve commit diff was not subbuf_size : + * it means the subbuffer was partly written to + * and is therefore corrupted. If it is multiple + * of subbuffer size and we are in flight + * recorder mode, we are skipping over a whole + * subbuffer. + */ + local_inc(<t_buf->corrupted_subbuffers); + } + } + } +} + + +/* + * ltt_reserve_switch_old_subbuf: switch old subbuffer + * + * Concurrency safe because we are the last and only thread to alter this + * sub-buffer. As long as it is not delivered and read, no other thread can + * alter the offset, alter the reserve_count or call the + * client_buffer_end_callback on this sub-buffer. + * + * The only remaining threads could be the ones with pending commits. They will + * have to do the deliver themselves. Not concurrency safe in overwrite mode. + * We detect corrupted subbuffers with commit and reserve counts. We keep a + * corrupted sub-buffers count and push the readers across these sub-buffers. + * + * Not concurrency safe if a writer is stalled in a subbuffer and another writer + * switches in, finding out it's corrupted. The result will be than the old + * (uncommited) subbuffer will be declared corrupted, and that the new subbuffer + * will be declared corrupted too because of the commit count adjustment. + * + * Note : offset_old should never be 0 here. + */ +static inline void ltt_reserve_switch_old_subbuf( + struct ltt_channel_struct *ltt_channel, + struct ltt_channel_buf_struct *ltt_buf, struct rchan *rchan, + struct rchan_buf *buf, + struct ltt_reserve_switch_offsets *offsets, u64 *tsc) +{ + long oldidx = SUBBUF_INDEX(offsets->old - 1, rchan); + + ltt_channel->buffer_end(buf, *tsc, offsets->old, oldidx); + /* Must write buffer end before incrementing commit count */ + smp_wmb(); + offsets->commit_count = + local_add_return(rchan->subbuf_size + - (SUBBUF_OFFSET(offsets->old - 1, rchan) + + 1), + <t_buf->commit_count[oldidx]); + if ((BUFFER_TRUNC(offsets->old - 1, rchan) + >> ltt_channel->n_subbufs_order) + - ((offsets->commit_count - rchan->subbuf_size) + & ltt_channel->commit_count_mask) == 0) + ltt_deliver(buf, oldidx, NULL); +} + +/* + * ltt_reserve_switch_new_subbuf: Populate new subbuffer. + * + * This code can be executed unordered : writers may already have written to the + * sub-buffer before this code gets executed, caution. The commit makes sure + * that this code is executed before the deliver of this sub-buffer. + */ +static /*inline*/ void ltt_reserve_switch_new_subbuf( + struct ltt_channel_struct *ltt_channel, + struct ltt_channel_buf_struct *ltt_buf, struct rchan *rchan, + struct rchan_buf *buf, + struct ltt_reserve_switch_offsets *offsets, u64 *tsc) +{ + long beginidx = SUBBUF_INDEX(offsets->begin, rchan); + + ltt_channel->buffer_begin(buf, *tsc, beginidx); + /* Must write buffer end before incrementing commit count */ + smp_wmb(); + offsets->commit_count = local_add_return(ltt_subbuffer_header_size(), + <t_buf->commit_count[beginidx]); + /* Check if the written buffer has to be delivered */ + if ((BUFFER_TRUNC(offsets->begin, rchan) + >> ltt_channel->n_subbufs_order) + - ((offsets->commit_count - rchan->subbuf_size) + & ltt_channel->commit_count_mask) == 0) + ltt_deliver(buf, beginidx, NULL); +} + + +/* + * ltt_reserve_end_switch_current: finish switching current subbuffer + * + * Concurrency safe because we are the last and only thread to alter this + * sub-buffer. As long as it is not delivered and read, no other thread can + * alter the offset, alter the reserve_count or call the + * client_buffer_end_callback on this sub-buffer. + * + * The only remaining threads could be the ones with pending commits. They will + * have to do the deliver themselves. Not concurrency safe in overwrite mode. + * We detect corrupted subbuffers with commit and reserve counts. We keep a + * corrupted sub-buffers count and push the readers across these sub-buffers. + * + * Not concurrency safe if a writer is stalled in a subbuffer and another writer + * switches in, finding out it's corrupted. The result will be than the old + * (uncommited) subbuffer will be declared corrupted, and that the new subbuffer + * will be declared corrupted too because of the commit count adjustment. + */ +static inline void ltt_reserve_end_switch_current( + struct ltt_channel_struct *ltt_channel, + struct ltt_channel_buf_struct *ltt_buf, struct rchan *rchan, + struct rchan_buf *buf, + struct ltt_reserve_switch_offsets *offsets, u64 *tsc) +{ + long endidx = SUBBUF_INDEX(offsets->end - 1, rchan); + + ltt_channel->buffer_end(buf, *tsc, offsets->end, endidx); + /* Must write buffer begin before incrementing commit count */ + smp_wmb(); + offsets->commit_count = + local_add_return(rchan->subbuf_size + - (SUBBUF_OFFSET(offsets->end - 1, rchan) + + 1), + <t_buf->commit_count[endidx]); + if ((BUFFER_TRUNC(offsets->end - 1, rchan) + >> ltt_channel->n_subbufs_order) + - ((offsets->commit_count - rchan->subbuf_size) + & ltt_channel->commit_count_mask) == 0) + ltt_deliver(buf, endidx, NULL); +} + +/** + * ltt_relay_reserve_slot - Atomic slot reservation in a LTTng buffer. + * @trace: the trace structure to log to. + * @ltt_channel: channel structure + * @transport_data: data structure specific to ltt relay + * @data_size: size of the variable length data to log. + * @slot_size: pointer to total size of the slot (out) + * @buf_offset : pointer to reserved buffer offset (out) + * @tsc: pointer to the tsc at the slot reservation (out) + * @cpu: cpuid + * + * Return : -ENOSPC if not enough space, else returns 0. + * It will take care of sub-buffer switching. + */ +static notrace int ltt_relay_reserve_slot(struct ltt_trace_struct *trace, + struct ltt_channel_struct *ltt_channel, void **transport_data, + size_t data_size, size_t *slot_size, long *buf_offset, u64 *tsc, + unsigned int *rflags, int largest_align) +{ + struct rchan *rchan = ltt_channel->trans_channel_data; + struct rchan_buf *buf = *transport_data = rchan->buf; + struct ltt_channel_buf_struct *ltt_buf = ltt_channel->buf; + struct ltt_reserve_switch_offsets offsets; + + offsets.reserve_commit_diff = 0; + offsets.size = 0; + + /* + * Perform retryable operations. + */ + if (ltt_nesting > 4) { + local_inc(<t_buf->events_lost); + return -EPERM; + } + do { + if (ltt_relay_try_reserve(ltt_channel, ltt_buf, + rchan, buf, &offsets, data_size, tsc, rflags, + largest_align)) + return -ENOSPC; + } while (local_cmpxchg(<t_buf->offset, offsets.old, + offsets.end) != offsets.old); + + /* + * Atomically update last_tsc. This update races against concurrent + * atomic updates, but the race will always cause supplementary full TSC + * events, never the opposite (missing a full TSC event when it would be + * needed). + */ + save_last_tsc(ltt_buf, *tsc); + + /* + * Push the reader if necessary + */ + ltt_reserve_push_reader(ltt_channel, ltt_buf, rchan, buf, &offsets); + + /* + * Switch old subbuffer if needed. + */ + if (offsets.end_switch_old) + ltt_reserve_switch_old_subbuf(ltt_channel, ltt_buf, rchan, buf, + &offsets, tsc); + + /* + * Populate new subbuffer. + */ + if (offsets.begin_switch) + ltt_reserve_switch_new_subbuf(ltt_channel, ltt_buf, rchan, + buf, &offsets, tsc); + + if (offsets.end_switch_current) + ltt_reserve_end_switch_current(ltt_channel, ltt_buf, rchan, + buf, &offsets, tsc); + + *slot_size = offsets.size; + *buf_offset = offsets.begin + offsets.before_hdr_pad; + return 0; +} + +/* + * Force a sub-buffer switch for a per-cpu buffer. This operation is + * completely reentrant : can be called while tracing is active with + * absolutely no lock held. + * + * Note, however, that as a local_cmpxchg is used for some atomic + * operations, this function must be called from the CPU which owns the buffer + * for a ACTIVE flush. + */ +static notrace void ltt_force_switch(struct rchan_buf *buf, + enum force_switch_mode mode) +{ + struct ltt_channel_struct *ltt_channel = + (struct ltt_channel_struct *)buf->chan->private_data; + struct ltt_channel_buf_struct *ltt_buf = ltt_channel->buf; + struct rchan *rchan = ltt_channel->trans_channel_data; + struct ltt_reserve_switch_offsets offsets; + u64 tsc; + + offsets.reserve_commit_diff = 0; + offsets.size = 0; + + /* + * Perform retryable operations. + */ + do { + if (ltt_relay_try_switch(mode, ltt_channel, ltt_buf, + rchan, buf, &offsets, &tsc)) + return; + } while (local_cmpxchg(<t_buf->offset, offsets.old, + offsets.end) != offsets.old); + + /* + * Atomically update last_tsc. This update races against concurrent + * atomic updates, but the race will always cause supplementary full TSC + * events, never the opposite (missing a full TSC event when it would be + * needed). + */ + save_last_tsc(ltt_buf, tsc); + + /* + * Push the reader if necessary + */ + if (mode == FORCE_ACTIVE) + ltt_reserve_push_reader(ltt_channel, ltt_buf, rchan, + buf, &offsets); + + /* + * Switch old subbuffer if needed. + */ + if (offsets.end_switch_old) + ltt_reserve_switch_old_subbuf(ltt_channel, ltt_buf, rchan, buf, + &offsets, &tsc); + + /* + * Populate new subbuffer. + */ + if (mode == FORCE_ACTIVE) + ltt_reserve_switch_new_subbuf(ltt_channel, + ltt_buf, rchan, buf, &offsets, &tsc); +} + +/* + * for flight recording. must be called after relay_commit. + * This function decrements de subbuffer's lost_size each time the commit count + * reaches back the reserve offset (module subbuffer size). It is useful for + * crash dump. + * We use slot_size - 1 to make sure we deal correctly with the case where we + * fill the subbuffer completely (so the subbuf index stays in the previous + * subbuffer). + */ +//ust// #ifdef CONFIG_LTT_VMCORE +static /*inline*/ void ltt_write_commit_counter(struct rchan_buf *buf, + long buf_offset, size_t slot_size) +{ + struct ltt_channel_struct *ltt_channel = + (struct ltt_channel_struct *)buf->chan->private_data; + struct ltt_channel_buf_struct *ltt_buf = ltt_channel->buf; + struct ltt_subbuffer_header *header; + long offset, subbuf_idx, commit_count; + uint32_t lost_old, lost_new; + + subbuf_idx = SUBBUF_INDEX(buf_offset - 1, buf->chan); + offset = buf_offset + slot_size; + header = (struct ltt_subbuffer_header *) + ltt_relay_offset_address(buf, + subbuf_idx * buf->chan->subbuf_size); + for (;;) { + lost_old = header->lost_size; + commit_count = + local_read(<t_buf->commit_count[subbuf_idx]); + /* SUBBUF_OFFSET includes commit_count_mask */ + if (!SUBBUF_OFFSET(offset - commit_count, buf->chan)) { + lost_new = (uint32_t)buf->chan->subbuf_size + - SUBBUF_OFFSET(commit_count, buf->chan); + lost_old = cmpxchg_local(&header->lost_size, lost_old, + lost_new); + if (lost_old <= lost_new) + break; + } else { + break; + } + } +} +//ust// #else +//ust// static inline void ltt_write_commit_counter(struct rchan_buf *buf, +//ust// long buf_offset, size_t slot_size) +//ust// { +//ust// } +//ust// #endif + +/* + * Atomic unordered slot commit. Increments the commit count in the + * specified sub-buffer, and delivers it if necessary. + * + * Parameters: + * + * @ltt_channel : channel structure + * @transport_data: transport-specific data + * @buf_offset : offset following the event header. + * @slot_size : size of the reserved slot. + */ +static notrace void ltt_relay_commit_slot( + struct ltt_channel_struct *ltt_channel, + void **transport_data, long buf_offset, size_t slot_size) +{ + struct rchan_buf *buf = *transport_data; + struct ltt_channel_buf_struct *ltt_buf = ltt_channel->buf; + struct rchan *rchan = buf->chan; + long offset_end = buf_offset; + long endidx = SUBBUF_INDEX(offset_end - 1, rchan); + long commit_count; + + /* Must write slot data before incrementing commit count */ + smp_wmb(); + commit_count = local_add_return(slot_size, + <t_buf->commit_count[endidx]); + /* Check if all commits have been done */ + if ((BUFFER_TRUNC(offset_end - 1, rchan) + >> ltt_channel->n_subbufs_order) + - ((commit_count - rchan->subbuf_size) + & ltt_channel->commit_count_mask) == 0) + ltt_deliver(buf, endidx, NULL); + /* + * Update lost_size for each commit. It's needed only for extracting + * ltt buffers from vmcore, after crash. + */ + ltt_write_commit_counter(buf, buf_offset, slot_size); + + DBG("commited slot. now commit count is %ld", commit_count); +} + +/* + * This is called with preemption disabled when user space has requested + * blocking mode. If one of the active traces has free space below a + * specific threshold value, we reenable preemption and block. + */ +static int ltt_relay_user_blocking(struct ltt_trace_struct *trace, + unsigned int chan_index, size_t data_size, + struct user_dbg_data *dbg) +{ +//ust// struct rchan *rchan; +//ust// struct ltt_channel_buf_struct *ltt_buf; +//ust// struct ltt_channel_struct *channel; +//ust// struct rchan_buf *relay_buf; +//ust// int cpu; +//ust// DECLARE_WAITQUEUE(wait, current); +//ust// +//ust// channel = &trace->channels[chan_index]; +//ust// rchan = channel->trans_channel_data; +//ust// cpu = smp_processor_id(); +//ust// relay_buf = rchan->buf[cpu]; +//ust// ltt_buf = percpu_ptr(channel->buf, cpu); +//ust// +//ust// /* +//ust// * Check if data is too big for the channel : do not +//ust// * block for it. +//ust// */ +//ust// if (LTT_RESERVE_CRITICAL + data_size > relay_buf->chan->subbuf_size) +//ust// return 0; +//ust// +//ust// /* +//ust// * If free space too low, we block. We restart from the +//ust// * beginning after we resume (cpu id may have changed +//ust// * while preemption is active). +//ust// */ +//ust// spin_lock(<t_buf->full_lock); +//ust// if (!channel->overwrite) { +//ust// dbg->write = local_read(<t_buf->offset); +//ust// dbg->read = atomic_long_read(<t_buf->consumed); +//ust// dbg->avail_size = dbg->write + LTT_RESERVE_CRITICAL + data_size +//ust// - SUBBUF_TRUNC(dbg->read, +//ust// relay_buf->chan); +//ust// if (dbg->avail_size > rchan->alloc_size) { +//ust// __set_current_state(TASK_INTERRUPTIBLE); +//ust// add_wait_queue(<t_buf->write_wait, &wait); +//ust// spin_unlock(<t_buf->full_lock); +//ust// preempt_enable(); +//ust// schedule(); +//ust// __set_current_state(TASK_RUNNING); +//ust// remove_wait_queue(<t_buf->write_wait, &wait); +//ust// if (signal_pending(current)) +//ust// return -ERESTARTSYS; +//ust// preempt_disable(); +//ust// return 1; +//ust// } +//ust// } +//ust// spin_unlock(<t_buf->full_lock); + return 0; +} + +static void ltt_relay_print_user_errors(struct ltt_trace_struct *trace, + unsigned int chan_index, size_t data_size, + struct user_dbg_data *dbg) +{ + struct rchan *rchan; + struct ltt_channel_buf_struct *ltt_buf; + struct ltt_channel_struct *channel; + struct rchan_buf *relay_buf; + + channel = &trace->channels[chan_index]; + rchan = channel->trans_channel_data; + relay_buf = rchan->buf; + ltt_buf = channel->buf; + + printk(KERN_ERR "Error in LTT usertrace : " + "buffer full : event lost in blocking " + "mode. Increase LTT_RESERVE_CRITICAL.\n"); + printk(KERN_ERR "LTT nesting level is %u.\n", ltt_nesting); + printk(KERN_ERR "LTT avail size %lu.\n", + dbg->avail_size); + printk(KERN_ERR "avai write : %lu, read : %lu\n", + dbg->write, dbg->read); + + dbg->write = local_read(<t_buf->offset); + dbg->read = atomic_long_read(<t_buf->consumed); + + printk(KERN_ERR "LTT cur size %lu.\n", + dbg->write + LTT_RESERVE_CRITICAL + data_size + - SUBBUF_TRUNC(dbg->read, relay_buf->chan)); + printk(KERN_ERR "cur write : %lu, read : %lu\n", + dbg->write, dbg->read); +} + +//ust// static struct ltt_transport ltt_relay_transport = { +//ust// .name = "relay", +//ust// .owner = THIS_MODULE, +//ust// .ops = { +//ust// .create_dirs = ltt_relay_create_dirs, +//ust// .remove_dirs = ltt_relay_remove_dirs, +//ust// .create_channel = ltt_relay_create_channel, +//ust// .finish_channel = ltt_relay_finish_channel, +//ust// .remove_channel = ltt_relay_remove_channel, +//ust// .wakeup_channel = ltt_relay_async_wakeup_chan, +//ust// .commit_slot = ltt_relay_commit_slot, +//ust// .reserve_slot = ltt_relay_reserve_slot, +//ust// .user_blocking = ltt_relay_user_blocking, +//ust// .user_errors = ltt_relay_print_user_errors, +//ust// }, +//ust// }; + +static struct ltt_transport ust_relay_transport = { + .name = "ustrelay", +//ust// .owner = THIS_MODULE, + .ops = { + .create_dirs = ltt_relay_create_dirs, + .remove_dirs = ltt_relay_remove_dirs, + .create_channel = ltt_relay_create_channel, + .finish_channel = ltt_relay_finish_channel, + .remove_channel = ltt_relay_remove_channel, + .wakeup_channel = ltt_relay_async_wakeup_chan, + .commit_slot = ltt_relay_commit_slot, + .reserve_slot = ltt_relay_reserve_slot, + .user_blocking = ltt_relay_user_blocking, + .user_errors = ltt_relay_print_user_errors, + }, +}; + +//ust// static int __init ltt_relay_init(void) +//ust// { +//ust// printk(KERN_INFO "LTT : ltt-relay init\n"); +//ust// +//ust// ltt_file_operations = ltt_relay_file_operations; +//ust// ltt_file_operations.owner = THIS_MODULE; +//ust// ltt_file_operations.open = ltt_open; +//ust// ltt_file_operations.release = ltt_release; +//ust// ltt_file_operations.poll = ltt_poll; +//ust// ltt_file_operations.splice_read = ltt_relay_file_splice_read, +//ust// ltt_file_operations.ioctl = ltt_ioctl; +//ust//#ifdef CONFIG_COMPAT +//ust// ltt_file_operations.compat_ioctl = ltt_compat_ioctl; +//ust//#endif +//ust// +//ust// ltt_transport_register(<t_relay_transport); +//ust// +//ust// return 0; +//ust// } + +static char initialized = 0; + +void __attribute__((constructor)) init_ustrelay_transport(void) +{ + if(!initialized) { + ltt_transport_register(&ust_relay_transport); + initialized = 1; + } +} + +static void __exit ltt_relay_exit(void) +{ +//ust// printk(KERN_INFO "LTT : ltt-relay exit\n"); + + ltt_transport_unregister(&ust_relay_transport); +} + +//ust// module_init(ltt_relay_init); +//ust// module_exit(ltt_relay_exit); +//ust// +//ust// MODULE_LICENSE("GPL"); +//ust// MODULE_AUTHOR("Mathieu Desnoyers"); +//ust// MODULE_DESCRIPTION("Linux Trace Toolkit Next Generation Lockless Relay"); diff --git a/libust/relay.h b/libust/relay.h new file mode 100644 index 0000000..4b27b5b --- /dev/null +++ b/libust/relay.h @@ -0,0 +1,367 @@ +/* + * linux/include/linux/ltt-relay.h + * + * Copyright (C) 2002, 2003 - Tom Zanussi (zanussi@us.ibm.com), IBM Corp + * Copyright (C) 1999, 2000, 2001, 2002 - Karim Yaghmour (karim@opersys.com) + * Copyright (C) 2008 - Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) + * + * CONFIG_RELAY definitions and declarations + */ + +#ifndef _LINUX_LTT_RELAY_H +#define _LINUX_LTT_RELAY_H + +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +#include "kref.h" +#include "list.h" + +/* Needs a _much_ better name... */ +#define FIX_SIZE(x) ((((x) - 1) & PAGE_MASK) + PAGE_SIZE) + +/* + * Tracks changes to rchan/rchan_buf structs + */ +#define LTT_RELAY_CHANNEL_VERSION 8 + +struct rchan_buf; + +struct buf_page { + struct page *page; + struct rchan_buf *buf; /* buffer the page belongs to */ + size_t offset; /* page offset in the buffer */ + struct list_head list; /* buffer linked list */ +}; + +/* + * Per-cpu relay channel buffer + */ +struct rchan_buf { + struct rchan *chan; /* associated channel */ +//ust// wait_queue_head_t read_wait; /* reader wait queue */ +//ust// struct timer_list timer; /* reader wake-up timer */ +//ust// struct dentry *dentry; /* channel file dentry */ + struct kref kref; /* channel buffer refcount */ +//ust// struct list_head pages; /* list of buffer pages */ + void *buf_data; //ust// + size_t buf_size; +//ust// struct buf_page *wpage; /* current write page (cache) */ +//ust// struct buf_page *hpage[2]; /* current subbuf header page (cache) */ +//ust// struct buf_page *rpage; /* current subbuf read page (cache) */ +//ust// unsigned int page_count; /* number of current buffer pages */ + unsigned int finalized; /* buffer has been finalized */ +//ust// unsigned int cpu; /* this buf's cpu */ + int shmid; /* the shmid of the buffer data pages */ +} ____cacheline_aligned; + +/* + * Relay channel data structure + */ +struct rchan { + u32 version; /* the version of this struct */ + size_t subbuf_size; /* sub-buffer size */ + size_t n_subbufs; /* number of sub-buffers per buffer */ + size_t alloc_size; /* total buffer size allocated */ + struct rchan_callbacks *cb; /* client callbacks */ + struct kref kref; /* channel refcount */ + void *private_data; /* for user-defined data */ +//ust// struct rchan_buf *buf[NR_CPUS]; /* per-cpu channel buffers */ + struct rchan_buf *buf; + struct list_head list; /* for channel list */ + struct dentry *parent; /* parent dentry passed to open */ + int subbuf_size_order; /* order of sub-buffer size */ +//ust// char base_filename[NAME_MAX]; /* saved base filename */ +}; + +/* + * Relay channel client callbacks + */ +struct rchan_callbacks { + /* + * subbuf_start - called on buffer-switch to a new sub-buffer + * @buf: the channel buffer containing the new sub-buffer + * @subbuf: the start of the new sub-buffer + * @prev_subbuf: the start of the previous sub-buffer + * @prev_padding: unused space at the end of previous sub-buffer + * + * The client should return 1 to continue logging, 0 to stop + * logging. + * + * NOTE: subbuf_start will also be invoked when the buffer is + * created, so that the first sub-buffer can be initialized + * if necessary. In this case, prev_subbuf will be NULL. + * + * NOTE: the client can reserve bytes at the beginning of the new + * sub-buffer by calling subbuf_start_reserve() in this callback. + */ + int (*subbuf_start) (struct rchan_buf *buf, + void *subbuf, + void *prev_subbuf, + size_t prev_padding); + + /* + * create_buf_file - create file to represent a relay channel buffer + * @filename: the name of the file to create + * @parent: the parent of the file to create + * @mode: the mode of the file to create + * @buf: the channel buffer + * + * Called during relay_open(), once for each per-cpu buffer, + * to allow the client to create a file to be used to + * represent the corresponding channel buffer. If the file is + * created outside of relay, the parent must also exist in + * that filesystem. + * + * The callback should return the dentry of the file created + * to represent the relay buffer. + * + * Setting the is_global outparam to a non-zero value will + * cause relay_open() to create a single global buffer rather + * than the default set of per-cpu buffers. + * + * See Documentation/filesystems/relayfs.txt for more info. + */ + struct dentry *(*create_buf_file)(const char *filename, + struct dentry *parent, + int mode, + struct rchan_buf *buf); + + /* + * remove_buf_file - remove file representing a relay channel buffer + * @dentry: the dentry of the file to remove + * + * Called during relay_close(), once for each per-cpu buffer, + * to allow the client to remove a file used to represent a + * channel buffer. + * + * The callback should return 0 if successful, negative if not. + */ +//ust// int (*remove_buf_file)(struct rchan_buf *buf); +}; + +extern struct buf_page *ltt_relay_find_prev_page(struct rchan_buf *buf, + struct buf_page *page, size_t offset, ssize_t diff_offset); + +extern struct buf_page *ltt_relay_find_next_page(struct rchan_buf *buf, + struct buf_page *page, size_t offset, ssize_t diff_offset); + +extern void _ltt_relay_write(struct rchan_buf *buf, size_t offset, + const void *src, size_t len, ssize_t cpy); + +extern int ltt_relay_read(struct rchan_buf *buf, size_t offset, + void *dest, size_t len); + +extern struct buf_page *ltt_relay_read_get_page(struct rchan_buf *buf, + size_t offset); + +/* + * Return the address where a given offset is located. + * Should be used to get the current subbuffer header pointer. Given we know + * it's never on a page boundary, it's safe to write directly to this address, + * as long as the write is never bigger than a page size. + */ +extern void *ltt_relay_offset_address(struct rchan_buf *buf, + size_t offset); + +/* + * Find the page containing "offset". Cache it if it is after the currently + * cached page. + */ +static inline struct buf_page *ltt_relay_cache_page(struct rchan_buf *buf, + struct buf_page **page_cache, + struct buf_page *page, size_t offset) +{ + ssize_t diff_offset; + ssize_t half_buf_size = buf->chan->alloc_size >> 1; + + /* + * Make sure this is the page we want to write into. The current + * page is changed concurrently by other writers. [wrh]page are + * used as a cache remembering the last page written + * to/read/looked up for header address. No synchronization; + * could have to find the previous page is a nested write + * occured. Finding the right page is done by comparing the + * dest_offset with the buf_page offsets. + * When at the exact opposite of the buffer, bias towards forward search + * because it will be cached. + */ + + diff_offset = (ssize_t)offset - (ssize_t)page->offset; + if (diff_offset <= -(ssize_t)half_buf_size) + diff_offset += buf->chan->alloc_size; + else if (diff_offset > half_buf_size) + diff_offset -= buf->chan->alloc_size; + + if (unlikely(diff_offset >= (ssize_t)PAGE_SIZE)) { + page = ltt_relay_find_next_page(buf, page, offset, diff_offset); + *page_cache = page; + } else if (unlikely(diff_offset < 0)) { + page = ltt_relay_find_prev_page(buf, page, offset, diff_offset); + } + return page; +} + +//ust// #ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS + static inline void ltt_relay_do_copy(void *dest, const void *src, size_t len) +{ + switch (len) { + case 0: break; + case 1: *(u8 *)dest = *(const u8 *)src; + break; + case 2: *(u16 *)dest = *(const u16 *)src; + break; + case 4: *(u32 *)dest = *(const u32 *)src; + break; +//ust// #if (BITS_PER_LONG == 64) + case 8: *(u64 *)dest = *(const u64 *)src; + break; +//ust// #endif + default: + memcpy(dest, src, len); + } +} +//ust// #else +//ust// /* +//ust// * Returns whether the dest and src addresses are aligned on +//ust// * min(sizeof(void *), len). Call this with statically known len for efficiency. +//ust// */ +//ust// static inline int addr_aligned(const void *dest, const void *src, size_t len) +//ust// { +//ust// if (ltt_align((size_t)dest, len)) +//ust// return 0; +//ust// if (ltt_align((size_t)src, len)) +//ust// return 0; +//ust// return 1; +//ust// } +//ust// +//ust// static inline void ltt_relay_do_copy(void *dest, const void *src, size_t len) +//ust// { +//ust// switch (len) { +//ust// case 0: break; +//ust// case 1: *(u8 *)dest = *(const u8 *)src; +//ust// break; +//ust// case 2: if (unlikely(!addr_aligned(dest, src, 2))) +//ust// goto memcpy_fallback; +//ust// *(u16 *)dest = *(const u16 *)src; +//ust// break; +//ust// case 4: if (unlikely(!addr_aligned(dest, src, 4))) +//ust// goto memcpy_fallback; +//ust// *(u32 *)dest = *(const u32 *)src; +//ust// break; +//ust// #if (BITS_PER_LONG == 64) +//ust// case 8: if (unlikely(!addr_aligned(dest, src, 8))) +//ust// goto memcpy_fallback; +//ust// *(u64 *)dest = *(const u64 *)src; +//ust// break; +//ust// #endif +//ust// default: +//ust// goto memcpy_fallback; +//ust// } +//ust// return; +//ust// memcpy_fallback: +//ust// memcpy(dest, src, len); +//ust// } +//ust// #endif + +static inline int ltt_relay_write(struct rchan_buf *buf, size_t offset, + const void *src, size_t len) +{ +//ust// struct buf_page *page; +//ust// ssize_t pagecpy; +//ust// +//ust// offset &= buf->chan->alloc_size - 1; +//ust// page = buf->wpage; +//ust// +//ust// page = ltt_relay_cache_page(buf, &buf->wpage, page, offset); +//ust// pagecpy = min_t(size_t, len, PAGE_SIZE - (offset & ~PAGE_MASK)); +//ust// ltt_relay_do_copy(page_address(page->page) +//ust// + (offset & ~PAGE_MASK), src, pagecpy); +//ust// +//ust// if (unlikely(len != pagecpy)) +//ust// _ltt_relay_write(buf, offset, src, len, page, pagecpy); +//ust// return len; + + + size_t cpy; + cpy = min_t(size_t, len, buf->buf_size - offset); + ltt_relay_do_copy(buf->buf_data + offset, src, cpy); + + if (unlikely(len != cpy)) + _ltt_relay_write(buf, offset, src, len, cpy); + return len; +} + +/* + * CONFIG_LTT_RELAY kernel API, ltt/ltt-relay-alloc.c + */ + +struct rchan *ltt_relay_open(const char *base_filename, + struct dentry *parent, + size_t subbuf_size, + size_t n_subbufs, + void *private_data); +extern void ltt_relay_close(struct rchan *chan); + +/* + * exported ltt_relay file operations, ltt/ltt-relay-alloc.c + */ +extern const struct file_operations ltt_relay_file_operations; + + +/* LTTng lockless logging buffer info */ +struct ltt_channel_buf_struct { + /* First 32 bytes cache-hot cacheline */ + local_t offset; /* Current offset in the buffer */ +//ust// local_t *commit_count; /* Commit count per sub-buffer */ + atomic_long_t consumed; /* + * Current offset in the buffer + * standard atomic access (shared) + */ + unsigned long last_tsc; /* + * Last timestamp written in the buffer. + */ + /* End of first 32 bytes cacheline */ + atomic_long_t active_readers; /* + * Active readers count + * standard atomic access (shared) + */ + local_t events_lost; + local_t corrupted_subbuffers; +//ust// spinlock_t full_lock; /* +//ust// * buffer full condition spinlock, only +//ust// * for userspace tracing blocking mode +//ust// * synchronization with reader. +//ust// */ +//ust// wait_queue_head_t write_wait; /* +//ust// * Wait queue for blocking user space +//ust// * writers +//ust// */ +//ust// atomic_t wakeup_readers; /* Boolean : wakeup readers waiting ? */ + /* one byte is written to this pipe when data is available, in order + to wake the consumer */ + /* portability: Single byte writes must be as quick as possible. The kernel-side + buffer must be large enough so the writer doesn't block. From the pipe(7) + man page: Since linux 2.6.11, the pipe capacity is 65536 bytes. */ + int data_ready_fd_write; + /* the reading end of the pipe */ + int data_ready_fd_read; + + /* commit count per subbuffer; must be at end of struct */ + local_t commit_count[0] ____cacheline_aligned; +} ____cacheline_aligned; + +int ltt_do_get_subbuf(struct rchan_buf *buf, struct ltt_channel_buf_struct *ltt_buf, long *pconsumed_old); + +int ltt_do_put_subbuf(struct rchan_buf *buf, struct ltt_channel_buf_struct *ltt_buf, u32 uconsumed_old); + + +#endif /* _LINUX_LTT_RELAY_H */ + diff --git a/libust/serialize.c b/libust/serialize.c new file mode 100644 index 0000000..bef0178 --- /dev/null +++ b/libust/serialize.c @@ -0,0 +1,696 @@ +/* + * LTTng serializing code. + * + * Copyright Mathieu Desnoyers, March 2007. + * + * Licensed under the GPLv2. + * + * See this discussion about weirdness about passing va_list and then va_list to + * functions. (related to array argument passing). va_list seems to be + * implemented as an array on x86_64, but not on i386... This is why we pass a + * va_list * to ltt_vtrace. + */ + +#include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +#include +#include +#include "kernelcompat.h" +#include "relay.h" +#include "tracer.h" +#include "list.h" +#include "usterr.h" + +enum ltt_type { + LTT_TYPE_SIGNED_INT, + LTT_TYPE_UNSIGNED_INT, + LTT_TYPE_STRING, + LTT_TYPE_NONE, +}; + +#define LTT_ATTRIBUTE_NETWORK_BYTE_ORDER (1<<1) + +/* + * Inspired from vsnprintf + * + * The serialization format string supports the basic printf format strings. + * In addition, it defines new formats that can be used to serialize more + * complex/non portable data structures. + * + * Typical use: + * + * field_name %ctype + * field_name #tracetype %ctype + * field_name #tracetype %ctype1 %ctype2 ... + * + * A conversion is performed between format string types supported by GCC and + * the trace type requested. GCC type is used to perform type checking on format + * strings. Trace type is used to specify the exact binary representation + * in the trace. A mapping is done between one or more GCC types to one trace + * type. Sign extension, if required by the conversion, is performed following + * the trace type. + * + * If a gcc format is not declared with a trace format, the gcc format is + * also used as binary representation in the trace. + * + * Strings are supported with %s. + * A single tracetype (sequence) can take multiple c types as parameter. + * + * c types: + * + * see printf(3). + * + * Note: to write a uint32_t in a trace, the following expression is recommended + * si it can be portable: + * + * ("#4u%lu", (unsigned long)var) + * + * trace types: + * + * Serialization specific formats : + * + * Fixed size integers + * #1u writes uint8_t + * #2u writes uint16_t + * #4u writes uint32_t + * #8u writes uint64_t + * #1d writes int8_t + * #2d writes int16_t + * #4d writes int32_t + * #8d writes int64_t + * i.e.: + * #1u%lu #2u%lu #4d%lu #8d%lu #llu%hu #d%lu + * + * * Attributes: + * + * n: (for network byte order) + * #ntracetype%ctype + * is written in the trace in network byte order. + * + * i.e.: #bn4u%lu, #n%lu, #b%u + * + * TODO (eventually) + * Variable length sequence + * #a #tracetype1 #tracetype2 %array_ptr %elem_size %num_elems + * In the trace: + * #a specifies that this is a sequence + * #tracetype1 is the type of elements in the sequence + * #tracetype2 is the type of the element count + * GCC input: + * array_ptr is a pointer to an array that contains members of size + * elem_size. + * num_elems is the number of elements in the array. + * i.e.: #a #lu #lu %p %lu %u + * + * Callback + * #k callback (taken from the probe data) + * The following % arguments are exepected by the callback + * + * i.e.: #a #lu #lu #k %p + * + * Note: No conversion is done from floats to integers, nor from integers to + * floats between c types and trace types. float conversion from double to float + * or from float to double is also not supported. + * + * REMOVE + * %*b expects sizeof(data), data + * where sizeof(data) is 1, 2, 4 or 8 + * + * Fixed length struct, union or array. + * FIXME: unable to extract those sizes statically. + * %*r expects sizeof(*ptr), ptr + * %*.*r expects sizeof(*ptr), __alignof__(*ptr), ptr + * struct and unions removed. + * Fixed length array: + * [%p]#a[len #tracetype] + * i.e.: [%p]#a[12 #lu] + * + * Variable length sequence + * %*.*:*v expects sizeof(*ptr), __alignof__(*ptr), elem_num, ptr + * where elem_num is the number of elements in the sequence + */ +static inline const char *parse_trace_type(const char *fmt, + char *trace_size, enum ltt_type *trace_type, + unsigned long *attributes) +{ + int qualifier; /* 'h', 'l', or 'L' for integer fields */ + /* 'z' support added 23/7/1999 S.H. */ + /* 'z' changed to 'Z' --davidm 1/25/99 */ + /* 't' added for ptrdiff_t */ + + /* parse attributes. */ +repeat: + switch (*fmt) { + case 'n': + *attributes |= LTT_ATTRIBUTE_NETWORK_BYTE_ORDER; + ++fmt; + goto repeat; + } + + /* get the conversion qualifier */ + qualifier = -1; + if (*fmt == 'h' || *fmt == 'l' || *fmt == 'L' || + *fmt == 'Z' || *fmt == 'z' || *fmt == 't' || + *fmt == 'S' || *fmt == '1' || *fmt == '2' || + *fmt == '4' || *fmt == 8) { + qualifier = *fmt; + ++fmt; + if (qualifier == 'l' && *fmt == 'l') { + qualifier = 'L'; + ++fmt; + } + } + + switch (*fmt) { + case 'c': + *trace_type = LTT_TYPE_UNSIGNED_INT; + *trace_size = sizeof(unsigned char); + goto parse_end; + case 's': + *trace_type = LTT_TYPE_STRING; + goto parse_end; + case 'p': + *trace_type = LTT_TYPE_UNSIGNED_INT; + *trace_size = sizeof(void *); + goto parse_end; + case 'd': + case 'i': + *trace_type = LTT_TYPE_SIGNED_INT; + break; + case 'o': + case 'u': + case 'x': + case 'X': + *trace_type = LTT_TYPE_UNSIGNED_INT; + break; + default: + if (!*fmt) + --fmt; + goto parse_end; + } + switch (qualifier) { + case 'L': + *trace_size = sizeof(long long); + break; + case 'l': + *trace_size = sizeof(long); + break; + case 'Z': + case 'z': + *trace_size = sizeof(size_t); + break; +//ust// case 't': +//ust// *trace_size = sizeof(ptrdiff_t); +//ust// break; + case 'h': + *trace_size = sizeof(short); + break; + case '1': + *trace_size = sizeof(uint8_t); + break; + case '2': + *trace_size = sizeof(uint16_t); + break; + case '4': + *trace_size = sizeof(uint32_t); + break; + case '8': + *trace_size = sizeof(uint64_t); + break; + default: + *trace_size = sizeof(int); + } + +parse_end: + return fmt; +} + +/* + * Restrictions: + * Field width and precision are *not* supported. + * %n not supported. + */ +static inline const char *parse_c_type(const char *fmt, + char *c_size, enum ltt_type *c_type) +{ + int qualifier; /* 'h', 'l', or 'L' for integer fields */ + /* 'z' support added 23/7/1999 S.H. */ + /* 'z' changed to 'Z' --davidm 1/25/99 */ + /* 't' added for ptrdiff_t */ + + /* process flags : ignore standard print formats for now. */ +repeat: + switch (*fmt) { + case '-': + case '+': + case ' ': + case '#': + case '0': + ++fmt; + goto repeat; + } + + /* get the conversion qualifier */ + qualifier = -1; + if (*fmt == 'h' || *fmt == 'l' || *fmt == 'L' || + *fmt == 'Z' || *fmt == 'z' || *fmt == 't' || + *fmt == 'S') { + qualifier = *fmt; + ++fmt; + if (qualifier == 'l' && *fmt == 'l') { + qualifier = 'L'; + ++fmt; + } + } + + switch (*fmt) { + case 'c': + *c_type = LTT_TYPE_UNSIGNED_INT; + *c_size = sizeof(unsigned char); + goto parse_end; + case 's': + *c_type = LTT_TYPE_STRING; + goto parse_end; + case 'p': + *c_type = LTT_TYPE_UNSIGNED_INT; + *c_size = sizeof(void *); + goto parse_end; + case 'd': + case 'i': + *c_type = LTT_TYPE_SIGNED_INT; + break; + case 'o': + case 'u': + case 'x': + case 'X': + *c_type = LTT_TYPE_UNSIGNED_INT; + break; + default: + if (!*fmt) + --fmt; + goto parse_end; + } + switch (qualifier) { + case 'L': + *c_size = sizeof(long long); + break; + case 'l': + *c_size = sizeof(long); + break; + case 'Z': + case 'z': + *c_size = sizeof(size_t); + break; +//ust// case 't': +//ust// *c_size = sizeof(ptrdiff_t); +//ust// break; + case 'h': + *c_size = sizeof(short); + break; + default: + *c_size = sizeof(int); + } + +parse_end: + return fmt; +} + +static inline size_t serialize_trace_data(struct rchan_buf *buf, + size_t buf_offset, + char trace_size, enum ltt_type trace_type, + char c_size, enum ltt_type c_type, + int *largest_align, va_list *args) +{ + union { + unsigned long v_ulong; + uint64_t v_uint64; + struct { + const char *s; + size_t len; + } v_string; + } tmp; + + /* + * Be careful about sign extension here. + * Sign extension is done with the destination (trace) type. + */ + switch (trace_type) { + case LTT_TYPE_SIGNED_INT: + switch (c_size) { + case 1: + tmp.v_ulong = (long)(int8_t)va_arg(*args, int); + break; + case 2: + tmp.v_ulong = (long)(int16_t)va_arg(*args, int); + break; + case 4: + tmp.v_ulong = (long)(int32_t)va_arg(*args, int); + break; + case 8: + tmp.v_uint64 = va_arg(*args, int64_t); + break; + default: + BUG(); + } + break; + case LTT_TYPE_UNSIGNED_INT: + switch (c_size) { + case 1: + tmp.v_ulong = (unsigned long)(uint8_t) + va_arg(*args, unsigned int); + break; + case 2: + tmp.v_ulong = (unsigned long)(uint16_t) + va_arg(*args, unsigned int); + break; + case 4: + tmp.v_ulong = (unsigned long)(uint32_t) + va_arg(*args, unsigned int); + break; + case 8: + tmp.v_uint64 = va_arg(*args, uint64_t); + break; + default: + BUG(); + } + break; + case LTT_TYPE_STRING: + tmp.v_string.s = va_arg(*args, const char *); + if ((unsigned long)tmp.v_string.s < PAGE_SIZE) + tmp.v_string.s = ""; + tmp.v_string.len = strlen(tmp.v_string.s)+1; + if (buf) + ltt_relay_write(buf, buf_offset, tmp.v_string.s, + tmp.v_string.len); + buf_offset += tmp.v_string.len; + goto copydone; + default: + BUG(); + } + + /* + * If trace_size is lower or equal to 4 bytes, there is no sign + * extension to do because we are already encoded in a long. Therefore, + * we can combine signed and unsigned ops. 4 bytes float also works + * with this, because we do a simple copy of 4 bytes into 4 bytes + * without manipulation (and we do not support conversion from integers + * to floats). + * It is also the case if c_size is 8 bytes, which is the largest + * possible integer. + */ + if (ltt_get_alignment()) { + buf_offset += ltt_align(buf_offset, trace_size); + if (largest_align) + *largest_align = max_t(int, *largest_align, trace_size); + } + if (trace_size <= 4 || c_size == 8) { + if (buf) { + switch (trace_size) { + case 1: + if (c_size == 8) + ltt_relay_write(buf, buf_offset, + (uint8_t[]){ (uint8_t)tmp.v_uint64 }, + sizeof(uint8_t)); + else + ltt_relay_write(buf, buf_offset, + (uint8_t[]){ (uint8_t)tmp.v_ulong }, + sizeof(uint8_t)); + break; + case 2: + if (c_size == 8) + ltt_relay_write(buf, buf_offset, + (uint16_t[]){ (uint16_t)tmp.v_uint64 }, + sizeof(uint16_t)); + else + ltt_relay_write(buf, buf_offset, + (uint16_t[]){ (uint16_t)tmp.v_ulong }, + sizeof(uint16_t)); + break; + case 4: + if (c_size == 8) + ltt_relay_write(buf, buf_offset, + (uint32_t[]){ (uint32_t)tmp.v_uint64 }, + sizeof(uint32_t)); + else + ltt_relay_write(buf, buf_offset, + (uint32_t[]){ (uint32_t)tmp.v_ulong }, + sizeof(uint32_t)); + break; + case 8: + /* + * c_size cannot be other than 8 here because + * trace_size > 4. + */ + ltt_relay_write(buf, buf_offset, + (uint64_t[]){ (uint64_t)tmp.v_uint64 }, + sizeof(uint64_t)); + break; + default: + BUG(); + } + } + buf_offset += trace_size; + goto copydone; + } else { + /* + * Perform sign extension. + */ + if (buf) { + switch (trace_type) { + case LTT_TYPE_SIGNED_INT: + ltt_relay_write(buf, buf_offset, + (int64_t[]){ (int64_t)tmp.v_ulong }, + sizeof(int64_t)); + break; + case LTT_TYPE_UNSIGNED_INT: + ltt_relay_write(buf, buf_offset, + (uint64_t[]){ (uint64_t)tmp.v_ulong }, + sizeof(uint64_t)); + break; + default: + BUG(); + } + } + buf_offset += trace_size; + goto copydone; + } + +copydone: + return buf_offset; +} + +notrace size_t ltt_serialize_data(struct rchan_buf *buf, size_t buf_offset, + struct ltt_serialize_closure *closure, + void *serialize_private, int *largest_align, + const char *fmt, va_list *args) +{ + char trace_size = 0, c_size = 0; /* + * 0 (unset), 1, 2, 4, 8 bytes. + */ + enum ltt_type trace_type = LTT_TYPE_NONE, c_type = LTT_TYPE_NONE; + unsigned long attributes = 0; + + for (; *fmt ; ++fmt) { + switch (*fmt) { + case '#': + /* tracetypes (#) */ + ++fmt; /* skip first '#' */ + if (*fmt == '#') /* Escaped ## */ + break; + attributes = 0; + fmt = parse_trace_type(fmt, &trace_size, &trace_type, + &attributes); + break; + case '%': + /* c types (%) */ + ++fmt; /* skip first '%' */ + if (*fmt == '%') /* Escaped %% */ + break; + fmt = parse_c_type(fmt, &c_size, &c_type); + /* + * Output c types if no trace types has been + * specified. + */ + if (!trace_size) + trace_size = c_size; + if (trace_type == LTT_TYPE_NONE) + trace_type = c_type; + if (c_type == LTT_TYPE_STRING) + trace_type = LTT_TYPE_STRING; + /* perform trace write */ + buf_offset = serialize_trace_data(buf, + buf_offset, trace_size, + trace_type, c_size, c_type, + largest_align, args); + trace_size = 0; + c_size = 0; + trace_type = LTT_TYPE_NONE; + c_size = LTT_TYPE_NONE; + attributes = 0; + break; + /* default is to skip the text, doing nothing */ + } + } + return buf_offset; +} +EXPORT_SYMBOL_GPL(ltt_serialize_data); + +/* + * Calculate data size + * Assume that the padding for alignment starts at a sizeof(void *) address. + */ +static notrace size_t ltt_get_data_size(struct ltt_serialize_closure *closure, + void *serialize_private, int *largest_align, + const char *fmt, va_list *args) +{ + ltt_serialize_cb cb = closure->callbacks[0]; + closure->cb_idx = 0; + return (size_t)cb(NULL, 0, closure, serialize_private, + largest_align, fmt, args); +} + +static notrace +void ltt_write_event_data(struct rchan_buf *buf, size_t buf_offset, + struct ltt_serialize_closure *closure, + void *serialize_private, int largest_align, + const char *fmt, va_list *args) +{ + ltt_serialize_cb cb = closure->callbacks[0]; + closure->cb_idx = 0; + buf_offset += ltt_align(buf_offset, largest_align); + cb(buf, buf_offset, closure, serialize_private, NULL, fmt, args); +} + + +notrace void ltt_vtrace(const struct marker *mdata, void *probe_data, + void *call_data, const char *fmt, va_list *args) +{ + int largest_align, ret; + struct ltt_active_marker *pdata; + uint16_t eID; + size_t data_size, slot_size; + unsigned int chan_index; + struct ltt_channel_struct *channel; + struct ltt_trace_struct *trace, *dest_trace = NULL; + struct rchan_buf *buf; + void *transport_data; + u64 tsc; + long buf_offset; + va_list args_copy; + struct ltt_serialize_closure closure; + struct ltt_probe_private_data *private_data = call_data; + void *serialize_private = NULL; + int cpu; + unsigned int rflags; + + /* + * This test is useful for quickly exiting static tracing when no trace + * is active. We expect to have an active trace when we get here. + */ + if (unlikely(ltt_traces.num_active_traces == 0)) + return; + + rcu_read_lock_sched_notrace(); + cpu = smp_processor_id(); +//ust// __get_cpu_var(ltt_nesting)++; + ltt_nesting++; + + pdata = (struct ltt_active_marker *)probe_data; + eID = mdata->event_id; + chan_index = mdata->channel_id; + closure.callbacks = pdata->probe->callbacks; + + if (unlikely(private_data)) { + dest_trace = private_data->trace; + if (private_data->serializer) + closure.callbacks = &private_data->serializer; + serialize_private = private_data->serialize_private; + } + + va_copy(args_copy, *args); + /* + * Assumes event payload to start on largest_align alignment. + */ + largest_align = 1; /* must be non-zero for ltt_align */ + data_size = ltt_get_data_size(&closure, serialize_private, + &largest_align, fmt, &args_copy); + largest_align = min_t(int, largest_align, sizeof(void *)); + va_end(args_copy); + + /* Iterate on each trace */ + list_for_each_entry_rcu(trace, <t_traces.head, list) { + /* + * Expect the filter to filter out events. If we get here, + * we went through tracepoint activation as a first step. + */ + if (unlikely(dest_trace && trace != dest_trace)) + continue; + if (unlikely(!trace->active)) + continue; + if (unlikely(!ltt_run_filter(trace, eID))) + continue; +#ifdef CONFIG_LTT_DEBUG_EVENT_SIZE + rflags = LTT_RFLAG_ID_SIZE; +#else + if (unlikely(eID >= LTT_FREE_EVENTS)) + rflags = LTT_RFLAG_ID; + else + rflags = 0; +#endif + /* + * Skip channels added after trace creation. + */ + if (unlikely(chan_index >= trace->nr_channels)) + continue; + channel = &trace->channels[chan_index]; + if (!channel->active) + continue; + + /* reserve space : header and data */ + ret = ltt_reserve_slot(trace, channel, &transport_data, + data_size, &slot_size, &buf_offset, + &tsc, &rflags, + largest_align); + if (unlikely(ret < 0)) + continue; /* buffer full */ + + va_copy(args_copy, *args); + /* FIXME : could probably encapsulate transport better. */ +//ust// buf = ((struct rchan *)channel->trans_channel_data)->buf[cpu]; + buf = ((struct rchan *)channel->trans_channel_data)->buf; + /* Out-of-order write : header and data */ + buf_offset = ltt_write_event_header(trace, + channel, buf, buf_offset, + eID, data_size, tsc, rflags); + ltt_write_event_data(buf, buf_offset, &closure, + serialize_private, + largest_align, fmt, &args_copy); + va_end(args_copy); + /* Out-of-order commit */ + ltt_commit_slot(channel, &transport_data, buf_offset, + slot_size); + printf("just commited event at offset %d and size %d\n", buf_offset, slot_size); + } +//ust// __get_cpu_var(ltt_nesting)--; + ltt_nesting--; + rcu_read_unlock_sched_notrace(); +} +EXPORT_SYMBOL_GPL(ltt_vtrace); + +notrace void ltt_trace(const struct marker *mdata, void *probe_data, + void *call_data, const char *fmt, ...) +{ + va_list args; + + va_start(args, fmt); + ltt_vtrace(mdata, probe_data, call_data, fmt, &args); + va_end(args); +} +EXPORT_SYMBOL_GPL(ltt_trace); + +//ust// MODULE_LICENSE("GPL"); +//ust// MODULE_AUTHOR("Mathieu Desnoyers"); +//ust// MODULE_DESCRIPTION("Linux Trace Toolkit Next Generation Serializer"); diff --git a/libust/tracectl.c b/libust/tracectl.c new file mode 100644 index 0000000..2101d5c --- /dev/null +++ b/libust/tracectl.c @@ -0,0 +1,820 @@ +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "marker.h" +#include "tracer.h" +#include "localerr.h" +#include "ustcomm.h" +#include "relay.h" /* FIXME: remove */ + +//#define USE_CLONE + +#define USTSIGNAL SIGIO + +#define MAX_MSG_SIZE (100) +#define MSG_NOTIF 1 +#define MSG_REGISTER_NOTIF 2 + +char consumer_stack[10000]; + +struct list_head blocked_consumers = LIST_HEAD_INIT(blocked_consumers); + +static struct ustcomm_app ustcomm_app; + +struct tracecmd { /* no padding */ + uint32_t size; + uint16_t command; +}; + +//struct listener_arg { +// int pipe_fd; +//}; + +struct trctl_msg { + /* size: the size of all the fields except size itself */ + uint32_t size; + uint16_t type; + /* Only the necessary part of the payload is transferred. It + * may even be none of it. + */ + char payload[94]; +}; + +struct consumer_channel { + int fd; + struct ltt_channel_struct *chan; +}; + +struct blocked_consumer { + int fd_consumer; + int fd_producer; + int tmp_poll_idx; + + /* args to ustcomm_send_reply */ + struct ustcomm_server server; + struct ustcomm_source src; + + /* args to ltt_do_get_subbuf */ + struct rchan_buf *rbuf; + struct ltt_channel_buf_struct *lttbuf; + + struct list_head list; +}; + +static void print_markers(void) +{ + struct marker_iter iter; + + lock_markers(); + marker_iter_reset(&iter); + marker_iter_start(&iter); + + while(iter.marker) { + fprintf(stderr, "marker: %s_%s \"%s\"\n", iter.marker->channel, iter.marker->name, iter.marker->format); + marker_iter_next(&iter); + } + unlock_markers(); +} + +void do_command(struct tracecmd *cmd) +{ +} + +void receive_commands() +{ +} + +int fd_notif = -1; +void notif_cb(void) +{ + int result; + struct trctl_msg msg; + + /* FIXME: fd_notif should probably be protected by a spinlock */ + + if(fd_notif == -1) + return; + + msg.type = MSG_NOTIF; + msg.size = sizeof(msg.type); + + /* FIXME: don't block here */ + result = write(fd_notif, &msg, msg.size+sizeof(msg.size)); + if(result == -1) { + PERROR("write"); + return; + } +} + +static int inform_consumer_daemon(void) +{ + ustcomm_request_consumer(getpid(), "metadata"); + ustcomm_request_consumer(getpid(), "ust"); +} + +void process_blocked_consumers(void) +{ + int n_fds = 0; + struct pollfd *fds; + struct blocked_consumer *bc; + int idx = 0; + char inbuf; + int result; + + list_for_each_entry(bc, &blocked_consumers, list) { + n_fds++; + } + + fds = (struct pollfd *) malloc(n_fds * sizeof(struct pollfd)); + if(fds == NULL) { + ERR("malloc returned NULL"); + return; + } + + list_for_each_entry(bc, &blocked_consumers, list) { + fds[idx].fd = bc->fd_producer; + fds[idx].events = POLLIN; + bc->tmp_poll_idx = idx; + idx++; + } + + result = poll(fds, n_fds, 0); + if(result == -1) { + PERROR("poll"); + return -1; + } + + list_for_each_entry(bc, &blocked_consumers, list) { + if(fds[bc->tmp_poll_idx].revents) { + long consumed_old = 0; + char *reply; + + result = read(bc->fd_producer, &inbuf, 1); + if(result == -1) { + PERROR("read"); + continue; + } + if(result == 0) { + DBG("PRODUCER END"); + + close(bc->fd_producer); + + __list_del(bc->list.prev, bc->list.next); + + result = ustcomm_send_reply(&bc->server, "END", &bc->src); + if(result < 0) { + ERR("ustcomm_send_reply failed"); + continue; + } + + continue; + } + + result = ltt_do_get_subbuf(bc->rbuf, bc->lttbuf, &consumed_old); + if(result == -EAGAIN) { + WARN("missed buffer?"); + continue; + } + else if(result < 0) { + DBG("ltt_do_get_subbuf: error: %s", strerror(-result)); + } + asprintf(&reply, "%s %ld", "OK", consumed_old); + result = ustcomm_send_reply(&bc->server, reply, &bc->src); + if(result < 0) { + ERR("ustcomm_send_reply failed"); + free(reply); + continue; + } + free(reply); + + __list_del(bc->list.prev, bc->list.next); + } + } + +} + +int listener_main(void *p) +{ + int result; + + DBG("LISTENER"); + + for(;;) { + uint32_t size; + struct sockaddr_un addr; + socklen_t addrlen = sizeof(addr); + char trace_name[] = "auto"; + char trace_type[] = "ustrelay"; + char *recvbuf; + int len; + struct ustcomm_source src; + + process_blocked_consumers(); + + result = ustcomm_app_recv_message(&ustcomm_app, &recvbuf, &src, 5); + if(result < 0) { + WARN("error in ustcomm_app_recv_message"); + continue; + } + else if(result == 0) { + /* no message */ + continue; + } + + DBG("received a message! it's: %s\n", recvbuf); + len = strlen(recvbuf); + + if(!strcmp(recvbuf, "print_markers")) { + print_markers(); + } + else if(!strcmp(recvbuf, "trace_setup")) { + DBG("trace setup"); + + result = ltt_trace_setup(trace_name); + if(result < 0) { + ERR("ltt_trace_setup failed"); + return; + } + + result = ltt_trace_set_type(trace_name, trace_type); + if(result < 0) { + ERR("ltt_trace_set_type failed"); + return; + } + } + else if(!strcmp(recvbuf, "trace_alloc")) { + DBG("trace alloc"); + + result = ltt_trace_alloc(trace_name); + if(result < 0) { + ERR("ltt_trace_alloc failed"); + return; + } + } + else if(!strcmp(recvbuf, "trace_start")) { + DBG("trace start"); + + result = ltt_trace_start(trace_name); + if(result < 0) { + ERR("ltt_trace_start failed"); + continue; + } + } + else if(!strcmp(recvbuf, "trace_stop")) { + DBG("trace stop"); + + result = ltt_trace_stop(trace_name); + if(result < 0) { + ERR("ltt_trace_stop failed"); + return; + } + } + else if(!strcmp(recvbuf, "trace_destroy")) { + + DBG("trace destroy"); + + result = ltt_trace_destroy(trace_name); + if(result < 0) { + ERR("ltt_trace_destroy failed"); + return; + } + } + else if(nth_token_is(recvbuf, "get_shmid", 0) == 1) { + struct ltt_trace_struct *trace; + char trace_name[] = "auto"; + int i; + char *channel_name; + + DBG("get_shmid"); + + channel_name = nth_token(recvbuf, 1); + if(channel_name == NULL) { + ERR("get_shmid: cannot parse channel"); + goto next_cmd; + } + + ltt_lock_traces(); + trace = _ltt_trace_find(trace_name); + ltt_unlock_traces(); + + if(trace == NULL) { + CPRINTF("cannot find trace!"); + return 1; + } + + for(i=0; inr_channels; i++) { + struct rchan *rchan = trace->channels[i].trans_channel_data; + struct rchan_buf *rbuf = rchan->buf; + struct ltt_channel_struct *ltt_channel = (struct ltt_channel_struct *)rchan->private_data; + struct ltt_channel_buf_struct *ltt_buf = ltt_channel->buf; + + if(!strcmp(trace->channels[i].channel_name, channel_name)) { + char *reply; + + DBG("the shmid for the requested channel is %d", rbuf->shmid); + DBG("the shmid for its buffer structure is %d", ltt_channel->buf_shmid); + asprintf(&reply, "%d %d", rbuf->shmid, ltt_channel->buf_shmid); + + result = ustcomm_send_reply(&ustcomm_app.server, reply, &src); + if(result) { + ERR("listener: get_shmid: ustcomm_send_reply failed"); + goto next_cmd; + } + + free(reply); + + break; + } + } + } + else if(nth_token_is(recvbuf, "get_n_subbufs", 0) == 1) { + struct ltt_trace_struct *trace; + char trace_name[] = "auto"; + int i; + char *channel_name; + + DBG("get_n_subbufs"); + + channel_name = nth_token(recvbuf, 1); + if(channel_name == NULL) { + ERR("get_n_subbufs: cannot parse channel"); + goto next_cmd; + } + + ltt_lock_traces(); + trace = _ltt_trace_find(trace_name); + ltt_unlock_traces(); + + if(trace == NULL) { + CPRINTF("cannot find trace!"); + return 1; + } + + for(i=0; inr_channels; i++) { + struct rchan *rchan = trace->channels[i].trans_channel_data; + + if(!strcmp(trace->channels[i].channel_name, channel_name)) { + char *reply; + + DBG("the n_subbufs for the requested channel is %d", rchan->n_subbufs); + asprintf(&reply, "%d", rchan->n_subbufs); + + result = ustcomm_send_reply(&ustcomm_app.server, reply, &src); + if(result) { + ERR("listener: get_n_subbufs: ustcomm_send_reply failed"); + goto next_cmd; + } + + free(reply); + + break; + } + } + } + else if(nth_token_is(recvbuf, "get_subbuf_size", 0) == 1) { + struct ltt_trace_struct *trace; + char trace_name[] = "auto"; + int i; + char *channel_name; + + DBG("get_subbuf_size"); + + channel_name = nth_token(recvbuf, 1); + if(channel_name == NULL) { + ERR("get_subbuf_size: cannot parse channel"); + goto next_cmd; + } + + ltt_lock_traces(); + trace = _ltt_trace_find(trace_name); + ltt_unlock_traces(); + + if(trace == NULL) { + CPRINTF("cannot find trace!"); + return 1; + } + + for(i=0; inr_channels; i++) { + struct rchan *rchan = trace->channels[i].trans_channel_data; + + if(!strcmp(trace->channels[i].channel_name, channel_name)) { + char *reply; + + DBG("the subbuf_size for the requested channel is %d", rchan->subbuf_size); + asprintf(&reply, "%d", rchan->subbuf_size); + + result = ustcomm_send_reply(&ustcomm_app.server, reply, &src); + if(result) { + ERR("listener: get_subbuf_size: ustcomm_send_reply failed"); + goto next_cmd; + } + + free(reply); + + break; + } + } + } + else if(nth_token_is(recvbuf, "load_probe_lib", 0) == 1) { + char *libfile; + + libfile = nth_token(recvbuf, 1); + + DBG("load_probe_lib loading %s", libfile); + } + else if(nth_token_is(recvbuf, "get_subbuffer", 0) == 1) { + struct ltt_trace_struct *trace; + char trace_name[] = "auto"; + int i; + char *channel_name; + + DBG("get_subbuf"); + + channel_name = nth_token(recvbuf, 1); + if(channel_name == NULL) { + ERR("get_subbuf: cannot parse channel"); + goto next_cmd; + } + + ltt_lock_traces(); + trace = _ltt_trace_find(trace_name); + ltt_unlock_traces(); + + if(trace == NULL) { + CPRINTF("cannot find trace!"); + return 1; + } + + for(i=0; inr_channels; i++) { + struct rchan *rchan = trace->channels[i].trans_channel_data; + + if(!strcmp(trace->channels[i].channel_name, channel_name)) { + struct rchan_buf *rbuf = rchan->buf; + struct ltt_channel_buf_struct *lttbuf = trace->channels[i].buf; + char *reply; + long consumed_old=0; + int fd; + struct blocked_consumer *bc; + + bc = (struct blocked_consumer *) malloc(sizeof(struct blocked_consumer)); + if(bc == NULL) { + ERR("malloc returned NULL"); + goto next_cmd; + } + bc->fd_consumer = src.fd; + bc->fd_producer = lttbuf->data_ready_fd_read; + bc->rbuf = rbuf; + bc->lttbuf = lttbuf; + bc->src = src; + bc->server = ustcomm_app.server; + + list_add(&bc->list, &blocked_consumers); + + break; + } + } + } + else if(nth_token_is(recvbuf, "put_subbuffer", 0) == 1) { + struct ltt_trace_struct *trace; + char trace_name[] = "auto"; + int i; + char *channel_name; + long consumed_old; + char *consumed_old_str; + char *endptr; + + DBG("put_subbuf"); + + channel_name = strdup_malloc(nth_token(recvbuf, 1)); + if(channel_name == NULL) { + ERR("put_subbuf_size: cannot parse channel"); + goto next_cmd; + } + + consumed_old_str = strdup_malloc(nth_token(recvbuf, 2)); + if(consumed_old_str == NULL) { + ERR("put_subbuf: cannot parse consumed_old"); + goto next_cmd; + } + consumed_old = strtol(consumed_old_str, &endptr, 10); + if(*endptr != '\0') { + ERR("put_subbuf: invalid value for consumed_old"); + goto next_cmd; + } + + ltt_lock_traces(); + trace = _ltt_trace_find(trace_name); + ltt_unlock_traces(); + + if(trace == NULL) { + CPRINTF("cannot find trace!"); + return 1; + } + + for(i=0; inr_channels; i++) { + struct rchan *rchan = trace->channels[i].trans_channel_data; + + if(!strcmp(trace->channels[i].channel_name, channel_name)) { + struct rchan_buf *rbuf = rchan->buf; + struct ltt_channel_buf_struct *lttbuf = trace->channels[i].buf; + char *reply; + long consumed_old=0; + + result = ltt_do_put_subbuf(rbuf, lttbuf, consumed_old); + if(result < 0) { + WARN("ltt_do_put_subbuf: error"); + } + else { + DBG("ltt_do_put_subbuf: success"); + } + asprintf(&reply, "%s", "OK", consumed_old); + + result = ustcomm_send_reply(&ustcomm_app.server, reply, &src); + if(result) { + ERR("listener: put_subbuf: ustcomm_send_reply failed"); + goto next_cmd; + } + + free(reply); + + break; + } + } + + free(channel_name); + free(consumed_old_str); + } +// else if(nth_token_is(recvbuf, "get_notifications", 0) == 1) { +// struct ltt_trace_struct *trace; +// char trace_name[] = "auto"; +// int i; +// char *channel_name; +// +// DBG("get_notifications"); +// +// channel_name = strdup_malloc(nth_token(recvbuf, 1)); +// if(channel_name == NULL) { +// ERR("put_subbuf_size: cannot parse channel"); +// goto next_cmd; +// } +// +// ltt_lock_traces(); +// trace = _ltt_trace_find(trace_name); +// ltt_unlock_traces(); +// +// if(trace == NULL) { +// CPRINTF("cannot find trace!"); +// return 1; +// } +// +// for(i=0; inr_channels; i++) { +// struct rchan *rchan = trace->channels[i].trans_channel_data; +// int fd; +// +// if(!strcmp(trace->channels[i].channel_name, channel_name)) { +// struct rchan_buf *rbuf = rchan->buf; +// struct ltt_channel_buf_struct *lttbuf = trace->channels[i].buf; +// +// result = fd = ustcomm_app_detach_client(&ustcomm_app, &src); +// if(result == -1) { +// ERR("ustcomm_app_detach_client failed"); +// goto next_cmd; +// } +// +// lttbuf->wake_consumer_arg = (void *) fd; +// +// smp_wmb(); +// +// lttbuf->call_wake_consumer = 1; +// +// break; +// } +// } +// +// free(channel_name); +// } + else { + ERR("unable to parse message: %s", recvbuf); + } + + next_cmd: + free(recvbuf); + } +} + +static char listener_stack[16384]; + +void create_listener(void) +{ + int result; + static char listener_stack[16384]; + //char *listener_stack = malloc(16384); + +#ifdef USE_CLONE + result = clone(listener_main, listener_stack+sizeof(listener_stack)-1, CLONE_FS | CLONE_FILES | CLONE_VM | CLONE_SIGHAND | CLONE_THREAD, NULL); + if(result == -1) { + perror("clone"); + } +#else + pthread_t thread; + + pthread_create(&thread, NULL, listener_main, NULL); +#endif +} + +/* The signal handler itself. Signals must be setup so there cannot be + nested signals. */ + +void sighandler(int sig) +{ + static char have_listener = 0; + DBG("sighandler"); + + if(!have_listener) { + create_listener(); + have_listener = 1; + } +} + +/* Called by the app signal handler to chain it to us. */ + +void chain_signal(void) +{ + sighandler(USTSIGNAL); +} + +static int init_socket(void) +{ + return ustcomm_init_app(getpid(), &ustcomm_app); +} + +static void destroy_socket(void) +{ +// int result; +// +// if(mysocketfile[0] == '\0') +// return; +// +// result = unlink(mysocketfile); +// if(result == -1) { +// PERROR("unlink"); +// } +} + +static int init_signal_handler(void) +{ + /* Attempt to handler SIGIO. If the main program wants to + * handle it, fine, it'll override us. They it'll have to + * use the chaining function. + */ + + int result; + struct sigaction act; + + result = sigemptyset(&act.sa_mask); + if(result == -1) { + PERROR("sigemptyset"); + return -1; + } + + act.sa_handler = sighandler; + act.sa_flags = SA_RESTART; + + /* Only defer ourselves. Also, try to restart interrupted + * syscalls to disturb the traced program as little as possible. + */ + result = sigaction(SIGIO, &act, NULL); + if(result == -1) { + PERROR("sigaction"); + return -1; + } + + return 0; +} + +static void auto_probe_connect(struct marker *m) +{ + int result; + + result = ltt_marker_connect(m->channel, m->name, "default"); + if(result) + ERR("ltt_marker_connect"); + + DBG("just auto connected marker %s %s to probe default", m->channel, m->name); +} + +static void __attribute__((constructor(101))) init0() +{ + DBG("UST_AUTOPROBE constructor"); + if(getenv("UST_AUTOPROBE")) { + marker_set_new_marker_cb(auto_probe_connect); + } +} + +static void fini(void); + +static void __attribute__((constructor(1000))) init() +{ + int result; + + DBG("UST_TRACE constructor"); + + /* Must create socket before signal handler to prevent races. + */ + result = init_socket(); + if(result == -1) { + ERR("init_socket error"); + return; + } + result = init_signal_handler(); + if(result == -1) { + ERR("init_signal_handler error"); + return; + } + + if(getenv("UST_TRACE")) { + char trace_name[] = "auto"; + char trace_type[] = "ustrelay"; + + DBG("starting early tracing"); + + /* Ensure marker control is initialized */ + init_marker_control(); + + /* Ensure relay is initialized */ + init_ustrelay_transport(); + + /* Ensure markers are initialized */ + init_markers(); + + /* In case. */ + ltt_channels_register("ust"); + + result = ltt_trace_setup(trace_name); + if(result < 0) { + ERR("ltt_trace_setup failed"); + return; + } + + result = ltt_trace_set_type(trace_name, trace_type); + if(result < 0) { + ERR("ltt_trace_set_type failed"); + return; + } + + result = ltt_trace_alloc(trace_name); + if(result < 0) { + ERR("ltt_trace_alloc failed"); + return; + } + + result = ltt_trace_start(trace_name); + if(result < 0) { + ERR("ltt_trace_start failed"); + return; + } + //start_consumer(); + inform_consumer_daemon(); + } + + + return; + + /* should decrementally destroy stuff if error */ + +} + +/* This is only called if we terminate normally, not with an unhandled signal, + * so we cannot rely on it. */ + +static void __attribute__((destructor)) fini() +{ + int result; + + /* if trace running, finish it */ + + DBG("destructor stopping traces"); + + result = ltt_trace_stop("auto"); + if(result == -1) { + ERR("ltt_trace_stop error"); + } + + result = ltt_trace_destroy("auto"); + if(result == -1) { + ERR("ltt_trace_destroy error"); + } + + /* FIXME: wait for the consumer to be done */ + //DBG("waiting 5 sec for consume"); + //sleep(5); + + destroy_socket(); +} diff --git a/libust/tracepoint.c b/libust/tracepoint.c new file mode 100644 index 0000000..ba4509b --- /dev/null +++ b/libust/tracepoint.c @@ -0,0 +1,685 @@ +/* + * Copyright (C) 2008 Mathieu Desnoyers + * Copyright (C) 2009 Pierre-Marc Fournier + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. + * + * Ported to userspace by Pierre-Marc Fournier. + */ + +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include + +#include + +#include "kernelcompat.h" +#include "tracepoint.h" +#include "usterr.h" +#include "list.h" + +//extern struct tracepoint __start___tracepoints[] __attribute__((visibility("hidden"))); +//extern struct tracepoint __stop___tracepoints[] __attribute__((visibility("hidden"))); + +/* Set to 1 to enable tracepoint debug output */ +static const int tracepoint_debug; + +/* libraries that contain tracepoints (struct tracepoint_lib) */ +static LIST_HEAD(libs); + +/* + * tracepoints_mutex nests inside module_mutex. Tracepoints mutex protects the + * builtin and module tracepoints and the hash table. + */ +static DEFINE_MUTEX(tracepoints_mutex); + +/* + * Tracepoint hash table, containing the active tracepoints. + * Protected by tracepoints_mutex. + */ +#define TRACEPOINT_HASH_BITS 6 +#define TRACEPOINT_TABLE_SIZE (1 << TRACEPOINT_HASH_BITS) +static struct hlist_head tracepoint_table[TRACEPOINT_TABLE_SIZE]; + +/* + * Note about RCU : + * It is used to to delay the free of multiple probes array until a quiescent + * state is reached. + * Tracepoint entries modifications are protected by the tracepoints_mutex. + */ +struct tracepoint_entry { + struct hlist_node hlist; + void **funcs; + int refcount; /* Number of times armed. 0 if disarmed. */ + char name[0]; +}; + +struct tp_probes { + union { +//ust// struct rcu_head rcu; + struct list_head list; + } u; + void *probes[0]; +}; + +static inline void *allocate_probes(int count) +{ + struct tp_probes *p = kmalloc(count * sizeof(void *) + + sizeof(struct tp_probes), GFP_KERNEL); + return p == NULL ? NULL : p->probes; +} + +//ust// static void rcu_free_old_probes(struct rcu_head *head) +//ust// { +//ust// kfree(container_of(head, struct tp_probes, u.rcu)); +//ust// } + +static inline void release_probes(void *old) +{ + if (old) { + struct tp_probes *tp_probes = container_of(old, + struct tp_probes, probes[0]); +//ust// call_rcu_sched(&tp_probes->u.rcu, rcu_free_old_probes); + synchronize_rcu(); + kfree(tp_probes); + } +} + +static void debug_print_probes(struct tracepoint_entry *entry) +{ + int i; + + if (!tracepoint_debug || !entry->funcs) + return; + + for (i = 0; entry->funcs[i]; i++) + printk(KERN_DEBUG "Probe %d : %p\n", i, entry->funcs[i]); +} + +static void * +tracepoint_entry_add_probe(struct tracepoint_entry *entry, void *probe) +{ + int nr_probes = 0; + void **old, **new; + + WARN_ON(!probe); + + debug_print_probes(entry); + old = entry->funcs; + if (old) { + /* (N -> N+1), (N != 0, 1) probes */ + for (nr_probes = 0; old[nr_probes]; nr_probes++) + if (old[nr_probes] == probe) + return ERR_PTR(-EEXIST); + } + /* + 2 : one for new probe, one for NULL func */ + new = allocate_probes(nr_probes + 2); + if (new == NULL) + return ERR_PTR(-ENOMEM); + if (old) + memcpy(new, old, nr_probes * sizeof(void *)); + new[nr_probes] = probe; + new[nr_probes + 1] = NULL; + entry->refcount = nr_probes + 1; + entry->funcs = new; + debug_print_probes(entry); + return old; +} + +static void * +tracepoint_entry_remove_probe(struct tracepoint_entry *entry, void *probe) +{ + int nr_probes = 0, nr_del = 0, i; + void **old, **new; + + old = entry->funcs; + + if (!old) + return ERR_PTR(-ENOENT); + + debug_print_probes(entry); + /* (N -> M), (N > 1, M >= 0) probes */ + for (nr_probes = 0; old[nr_probes]; nr_probes++) { + if ((!probe || old[nr_probes] == probe)) + nr_del++; + } + + if (nr_probes - nr_del == 0) { + /* N -> 0, (N > 1) */ + entry->funcs = NULL; + entry->refcount = 0; + debug_print_probes(entry); + return old; + } else { + int j = 0; + /* N -> M, (N > 1, M > 0) */ + /* + 1 for NULL */ + new = allocate_probes(nr_probes - nr_del + 1); + if (new == NULL) + return ERR_PTR(-ENOMEM); + for (i = 0; old[i]; i++) + if ((probe && old[i] != probe)) + new[j++] = old[i]; + new[nr_probes - nr_del] = NULL; + entry->refcount = nr_probes - nr_del; + entry->funcs = new; + } + debug_print_probes(entry); + return old; +} + +/* + * Get tracepoint if the tracepoint is present in the tracepoint hash table. + * Must be called with tracepoints_mutex held. + * Returns NULL if not present. + */ +static struct tracepoint_entry *get_tracepoint(const char *name) +{ + struct hlist_head *head; + struct hlist_node *node; + struct tracepoint_entry *e; + u32 hash = jhash(name, strlen(name), 0); + + head = &tracepoint_table[hash & (TRACEPOINT_TABLE_SIZE - 1)]; + hlist_for_each_entry(e, node, head, hlist) { + if (!strcmp(name, e->name)) + return e; + } + return NULL; +} + +/* + * Add the tracepoint to the tracepoint hash table. Must be called with + * tracepoints_mutex held. + */ +static struct tracepoint_entry *add_tracepoint(const char *name) +{ + struct hlist_head *head; + struct hlist_node *node; + struct tracepoint_entry *e; + size_t name_len = strlen(name) + 1; + u32 hash = jhash(name, name_len-1, 0); + + head = &tracepoint_table[hash & (TRACEPOINT_TABLE_SIZE - 1)]; + hlist_for_each_entry(e, node, head, hlist) { + if (!strcmp(name, e->name)) { + printk(KERN_NOTICE + "tracepoint %s busy\n", name); + return ERR_PTR(-EEXIST); /* Already there */ + } + } + /* + * Using kmalloc here to allocate a variable length element. Could + * cause some memory fragmentation if overused. + */ + e = kmalloc(sizeof(struct tracepoint_entry) + name_len, GFP_KERNEL); + if (!e) + return ERR_PTR(-ENOMEM); + memcpy(&e->name[0], name, name_len); + e->funcs = NULL; + e->refcount = 0; + hlist_add_head(&e->hlist, head); + return e; +} + +/* + * Remove the tracepoint from the tracepoint hash table. Must be called with + * mutex_lock held. + */ +static inline void remove_tracepoint(struct tracepoint_entry *e) +{ + hlist_del(&e->hlist); + kfree(e); +} + +/* + * Sets the probe callback corresponding to one tracepoint. + */ +static void set_tracepoint(struct tracepoint_entry **entry, + struct tracepoint *elem, int active) +{ + WARN_ON(strcmp((*entry)->name, elem->name) != 0); + + /* + * rcu_assign_pointer has a smp_wmb() which makes sure that the new + * probe callbacks array is consistent before setting a pointer to it. + * This array is referenced by __DO_TRACE from + * include/linux/tracepoints.h. A matching smp_read_barrier_depends() + * is used. + */ + rcu_assign_pointer(elem->funcs, (*entry)->funcs); + elem->state__imv = active; +} + +/* + * Disable a tracepoint and its probe callback. + * Note: only waiting an RCU period after setting elem->call to the empty + * function insures that the original callback is not used anymore. This insured + * by preempt_disable around the call site. + */ +static void disable_tracepoint(struct tracepoint *elem) +{ + elem->state__imv = 0; + rcu_assign_pointer(elem->funcs, NULL); +} + +/** + * tracepoint_update_probe_range - Update a probe range + * @begin: beginning of the range + * @end: end of the range + * + * Updates the probe callback corresponding to a range of tracepoints. + */ +void tracepoint_update_probe_range(struct tracepoint *begin, + struct tracepoint *end) +{ + struct tracepoint *iter; + struct tracepoint_entry *mark_entry; + + mutex_lock(&tracepoints_mutex); + for (iter = begin; iter < end; iter++) { + mark_entry = get_tracepoint(iter->name); + if (mark_entry) { + set_tracepoint(&mark_entry, iter, + !!mark_entry->refcount); + } else { + disable_tracepoint(iter); + } + } + mutex_unlock(&tracepoints_mutex); +} + +/* + * Update probes, removing the faulty probes. + */ +static void tracepoint_update_probes(void) +{ + /* Core kernel tracepoints */ +//ust// tracepoint_update_probe_range(__start___tracepoints, +//ust// __stop___tracepoints); + /* tracepoints in modules. */ + lib_update_tracepoints(); + /* Update immediate values */ + core_imv_update(); +//ust// module_imv_update(); +} + +static void *tracepoint_add_probe(const char *name, void *probe) +{ + struct tracepoint_entry *entry; + void *old; + + entry = get_tracepoint(name); + if (!entry) { + entry = add_tracepoint(name); + if (IS_ERR(entry)) + return entry; + } + old = tracepoint_entry_add_probe(entry, probe); + if (IS_ERR(old) && !entry->refcount) + remove_tracepoint(entry); + return old; +} + +/** + * tracepoint_probe_register - Connect a probe to a tracepoint + * @name: tracepoint name + * @probe: probe handler + * + * Returns 0 if ok, error value on error. + * The probe address must at least be aligned on the architecture pointer size. + */ +int tracepoint_probe_register(const char *name, void *probe) +{ + void *old; + + mutex_lock(&tracepoints_mutex); + old = tracepoint_add_probe(name, probe); + mutex_unlock(&tracepoints_mutex); + if (IS_ERR(old)) + return PTR_ERR(old); + + tracepoint_update_probes(); /* may update entry */ + release_probes(old); + return 0; +} +//ust// EXPORT_SYMBOL_GPL(tracepoint_probe_register); + +static void *tracepoint_remove_probe(const char *name, void *probe) +{ + struct tracepoint_entry *entry; + void *old; + + entry = get_tracepoint(name); + if (!entry) + return ERR_PTR(-ENOENT); + old = tracepoint_entry_remove_probe(entry, probe); + if (IS_ERR(old)) + return old; + if (!entry->refcount) + remove_tracepoint(entry); + return old; +} + +/** + * tracepoint_probe_unregister - Disconnect a probe from a tracepoint + * @name: tracepoint name + * @probe: probe function pointer + * + * We do not need to call a synchronize_sched to make sure the probes have + * finished running before doing a module unload, because the module unload + * itself uses stop_machine(), which insures that every preempt disabled section + * have finished. + */ +int tracepoint_probe_unregister(const char *name, void *probe) +{ + void *old; + + mutex_lock(&tracepoints_mutex); + old = tracepoint_remove_probe(name, probe); + mutex_unlock(&tracepoints_mutex); + if (IS_ERR(old)) + return PTR_ERR(old); + + tracepoint_update_probes(); /* may update entry */ + release_probes(old); + return 0; +} +//ust// EXPORT_SYMBOL_GPL(tracepoint_probe_unregister); + +static LIST_HEAD(old_probes); +static int need_update; + +static void tracepoint_add_old_probes(void *old) +{ + need_update = 1; + if (old) { + struct tp_probes *tp_probes = container_of(old, + struct tp_probes, probes[0]); + list_add(&tp_probes->u.list, &old_probes); + } +} + +/** + * tracepoint_probe_register_noupdate - register a probe but not connect + * @name: tracepoint name + * @probe: probe handler + * + * caller must call tracepoint_probe_update_all() + */ +int tracepoint_probe_register_noupdate(const char *name, void *probe) +{ + void *old; + + mutex_lock(&tracepoints_mutex); + old = tracepoint_add_probe(name, probe); + if (IS_ERR(old)) { + mutex_unlock(&tracepoints_mutex); + return PTR_ERR(old); + } + tracepoint_add_old_probes(old); + mutex_unlock(&tracepoints_mutex); + return 0; +} +//ust// EXPORT_SYMBOL_GPL(tracepoint_probe_register_noupdate); + +/** + * tracepoint_probe_unregister_noupdate - remove a probe but not disconnect + * @name: tracepoint name + * @probe: probe function pointer + * + * caller must call tracepoint_probe_update_all() + */ +int tracepoint_probe_unregister_noupdate(const char *name, void *probe) +{ + void *old; + + mutex_lock(&tracepoints_mutex); + old = tracepoint_remove_probe(name, probe); + if (IS_ERR(old)) { + mutex_unlock(&tracepoints_mutex); + return PTR_ERR(old); + } + tracepoint_add_old_probes(old); + mutex_unlock(&tracepoints_mutex); + return 0; +} +//ust// EXPORT_SYMBOL_GPL(tracepoint_probe_unregister_noupdate); + +/** + * tracepoint_probe_update_all - update tracepoints + */ +void tracepoint_probe_update_all(void) +{ + LIST_HEAD(release_probes); + struct tp_probes *pos, *next; + + mutex_lock(&tracepoints_mutex); + if (!need_update) { + mutex_unlock(&tracepoints_mutex); + return; + } + if (!list_empty(&old_probes)) + list_replace_init(&old_probes, &release_probes); + need_update = 0; + mutex_unlock(&tracepoints_mutex); + + tracepoint_update_probes(); + list_for_each_entry_safe(pos, next, &release_probes, u.list) { + list_del(&pos->u.list); +//ust// call_rcu_sched(&pos->u.rcu, rcu_free_old_probes); + synchronize_rcu(); + kfree(pos); + } +} +//ust// EXPORT_SYMBOL_GPL(tracepoint_probe_update_all); + +/** + * tracepoint_get_iter_range - Get a next tracepoint iterator given a range. + * @tracepoint: current tracepoints (in), next tracepoint (out) + * @begin: beginning of the range + * @end: end of the range + * + * Returns whether a next tracepoint has been found (1) or not (0). + * Will return the first tracepoint in the range if the input tracepoint is + * NULL. + */ +int tracepoint_get_iter_range(struct tracepoint **tracepoint, + struct tracepoint *begin, struct tracepoint *end) +{ + if (!*tracepoint && begin != end) { + *tracepoint = begin; + return 1; + } + if (*tracepoint >= begin && *tracepoint < end) + return 1; + return 0; +} +//ust// EXPORT_SYMBOL_GPL(tracepoint_get_iter_range); + +static void tracepoint_get_iter(struct tracepoint_iter *iter) +{ + int found = 0; + +//ust// /* Core kernel tracepoints */ +//ust// if (!iter->module) { +//ust// found = tracepoint_get_iter_range(&iter->tracepoint, +//ust// __start___tracepoints, __stop___tracepoints); +//ust// if (found) +//ust// goto end; +//ust// } + /* tracepoints in libs. */ + found = lib_get_iter_tracepoints(iter); +end: + if (!found) + tracepoint_iter_reset(iter); +} + +void tracepoint_iter_start(struct tracepoint_iter *iter) +{ + tracepoint_get_iter(iter); +} +//ust// EXPORT_SYMBOL_GPL(tracepoint_iter_start); + +void tracepoint_iter_next(struct tracepoint_iter *iter) +{ + iter->tracepoint++; + /* + * iter->tracepoint may be invalid because we blindly incremented it. + * Make sure it is valid by marshalling on the tracepoints, getting the + * tracepoints from following modules if necessary. + */ + tracepoint_get_iter(iter); +} +//ust// EXPORT_SYMBOL_GPL(tracepoint_iter_next); + +void tracepoint_iter_stop(struct tracepoint_iter *iter) +{ +} +//ust// EXPORT_SYMBOL_GPL(tracepoint_iter_stop); + +void tracepoint_iter_reset(struct tracepoint_iter *iter) +{ +//ust// iter->module = NULL; + iter->tracepoint = NULL; +} +//ust// EXPORT_SYMBOL_GPL(tracepoint_iter_reset); + +//ust// #ifdef CONFIG_MODULES + +//ust// int tracepoint_module_notify(struct notifier_block *self, +//ust// unsigned long val, void *data) +//ust// { +//ust// struct module *mod = data; +//ust// +//ust// switch (val) { +//ust// case MODULE_STATE_COMING: +//ust// tracepoint_update_probe_range(mod->tracepoints, +//ust// mod->tracepoints + mod->num_tracepoints); +//ust// break; +//ust// case MODULE_STATE_GOING: +//ust// tracepoint_update_probe_range(mod->tracepoints, +//ust// mod->tracepoints + mod->num_tracepoints); +//ust// break; +//ust// } +//ust// return 0; +//ust// } + +//ust// struct notifier_block tracepoint_module_nb = { +//ust// .notifier_call = tracepoint_module_notify, +//ust// .priority = 0, +//ust// }; + +//ust// static int init_tracepoints(void) +//ust// { +//ust// return register_module_notifier(&tracepoint_module_nb); +//ust// } +//ust// __initcall(init_tracepoints); + +//ust// #endif /* CONFIG_MODULES */ + +/* + * Returns 0 if current not found. + * Returns 1 if current found. + */ +int lib_get_iter_tracepoints(struct tracepoint_iter *iter) +{ + struct tracepoint_lib *iter_lib; + int found = 0; + +//ust// mutex_lock(&module_mutex); + list_for_each_entry(iter_lib, &libs, list) { + if (iter_lib < iter->lib) + continue; + else if (iter_lib > iter->lib) + iter->tracepoint = NULL; + found = marker_get_iter_range(&iter->tracepoint, + iter_lib->tracepoints_start, + iter_lib->tracepoints_start + iter_lib->tracepoints_count); + if (found) { + iter->lib = iter_lib; + break; + } + } +//ust// mutex_unlock(&module_mutex); + return found; +} + +void lib_update_tracepoints(void) +{ + struct tracepoint_lib *lib; + +//ust// mutex_lock(&module_mutex); + list_for_each_entry(lib, &libs, list) + tracepoint_update_probe_range(lib->tracepoints_start, + lib->tracepoints_start + lib->tracepoints_count); +//ust// mutex_unlock(&module_mutex); +} + +static void (*new_tracepoint_cb)(struct tracepoint *) = NULL; + +void tracepoint_set_new_tracepoint_cb(void (*cb)(struct tracepoint *)) +{ + new_tracepoint_cb = cb; +} + +static void new_tracepoints(struct tracepoint *start, struct tracepoint *end) +{ + if(new_tracepoint_cb) { + struct tracepoint *t; + for(t=start; t < end; t++) { + new_tracepoint_cb(t); + } + } +} + +int tracepoint_register_lib(struct tracepoint *tracepoints_start, int tracepoints_count) +{ + struct tracepoint_lib *pl; + + pl = (struct tracepoint_lib *) malloc(sizeof(struct tracepoint_lib)); + + pl->tracepoints_start = tracepoints_start; + pl->tracepoints_count = tracepoints_count; + + /* FIXME: maybe protect this with its own mutex? */ + mutex_lock(&tracepoints_mutex); + list_add(&pl->list, &libs); + mutex_unlock(&tracepoints_mutex); + + new_tracepoints(tracepoints_start, tracepoints_start + tracepoints_count); + + /* FIXME: update just the loaded lib */ + lib_update_tracepoints(); + + DBG("just registered a tracepoints section from %p and having %d tracepoints", tracepoints_start, tracepoints_count); + + return 0; +} + +int tracepoint_unregister_lib(struct tracepoint *tracepoints_start, int tracepoints_count) +{ + /*FIXME: implement; but before implementing, tracepoint_register_lib must + have appropriate locking. */ + + return 0; +} diff --git a/libust/tracepoint.h b/libust/tracepoint.h new file mode 100644 index 0000000..6f14ea0 --- /dev/null +++ b/libust/tracepoint.h @@ -0,0 +1,205 @@ +#ifndef _LINUX_TRACEPOINT_H +#define _LINUX_TRACEPOINT_H + +/* + * Copyright (C) 2008 Mathieu Desnoyers + * Copyright (C) 2009 Pierre-Marc Fournier + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version 2 + * of the License, or (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. + * + * Heavily inspired from the Linux Kernel Markers. + * + * Ported to userspace by Pierre-Marc Fournier. + */ + +//#include +//#include +//#include + +#include "immediate.h" +#include "kernelcompat.h" + +struct module; +struct tracepoint; + +struct tracepoint { + const char *name; /* Tracepoint name */ + DEFINE_IMV(char, state); /* State. */ + void **funcs; +} __attribute__((aligned(32))); /* + * Aligned on 32 bytes because it is + * globally visible and gcc happily + * align these on the structure size. + * Keep in sync with vmlinux.lds.h. + */ + +#define TPPROTO(args...) args +#define TPARGS(args...) args + +//ust// #ifdef CONFIG_TRACEPOINTS + +/* + * it_func[0] is never NULL because there is at least one element in the array + * when the array itself is non NULL. + */ +#define __DO_TRACE(tp, proto, args) \ + do { \ + void **it_func; \ + \ + rcu_read_lock_sched_notrace(); \ + it_func = rcu_dereference((tp)->funcs); \ + if (it_func) { \ + do { \ + ((void(*)(proto))(*it_func))(args); \ + } while (*(++it_func)); \ + } \ + rcu_read_unlock_sched_notrace(); \ + } while (0) + +#define __CHECK_TRACE(name, generic, proto, args) \ + do { \ + if (!generic) { \ + if (unlikely(imv_read(__tracepoint_##name.state))) \ + __DO_TRACE(&__tracepoint_##name, \ + TPPROTO(proto), TPARGS(args)); \ + } else { \ + if (unlikely(_imv_read(__tracepoint_##name.state))) \ + __DO_TRACE(&__tracepoint_##name, \ + TPPROTO(proto), TPARGS(args)); \ + } \ + } while (0) + +/* + * Make sure the alignment of the structure in the __tracepoints section will + * not add unwanted padding between the beginning of the section and the + * structure. Force alignment to the same alignment as the section start. + * + * The "generic" argument, passed to the declared __trace_##name inline + * function controls which tracepoint enabling mechanism must be used. + * If generic is true, a variable read is used. + * If generic is false, immediate values are used. + */ +#define DECLARE_TRACE(name, proto, args) \ + extern struct tracepoint __tracepoint_##name; \ + static inline void trace_##name(proto) \ + { \ + __CHECK_TRACE(name, 0, TPPROTO(proto), TPARGS(args)); \ + } \ + static inline void _trace_##name(proto) \ + { \ + __CHECK_TRACE(name, 1, TPPROTO(proto), TPARGS(args)); \ + } \ + static inline int register_trace_##name(void (*probe)(proto)) \ + { \ + return tracepoint_probe_register(#name, (void *)probe); \ + } \ + static inline int unregister_trace_##name(void (*probe)(proto)) \ + { \ + return tracepoint_probe_unregister(#name, (void *)probe);\ + } + +#define DEFINE_TRACE(name) \ + static const char __tpstrtab_##name[] \ + __attribute__((section("__tracepoints_strings"))) = #name; \ + struct tracepoint __tracepoint_##name \ + __attribute__((section("__tracepoints"), aligned(32))) = \ + { __tpstrtab_##name, 0, NULL } + +#define EXPORT_TRACEPOINT_SYMBOL_GPL(name) \ + EXPORT_SYMBOL_GPL(__tracepoint_##name) +#define EXPORT_TRACEPOINT_SYMBOL(name) \ + EXPORT_SYMBOL(__tracepoint_##name) + +extern void tracepoint_update_probe_range(struct tracepoint *begin, + struct tracepoint *end); + +//ust// #else /* !CONFIG_TRACEPOINTS */ +//ust// #define DECLARE_TRACE(name, proto, args) \ +//ust// static inline void trace_##name(proto) \ +//ust// { } \ +//ust// static inline void _trace_##name(proto) \ +//ust// { } \ +//ust// static inline int register_trace_##name(void (*probe)(proto)) \ +//ust// { \ +//ust// return -ENOSYS; \ +//ust// } \ +//ust// static inline int unregister_trace_##name(void (*probe)(proto)) \ +//ust// { \ +//ust// return -ENOSYS; \ +//ust// } +//ust// +//ust// #define DEFINE_TRACE(name) +//ust// #define EXPORT_TRACEPOINT_SYMBOL_GPL(name) +//ust// #define EXPORT_TRACEPOINT_SYMBOL(name) +//ust// +//ust// static inline void tracepoint_update_probe_range(struct tracepoint *begin, +//ust// struct tracepoint *end) +//ust// { } +//ust// #endif /* CONFIG_TRACEPOINTS */ + +/* + * Connect a probe to a tracepoint. + * Internal API, should not be used directly. + */ +extern int tracepoint_probe_register(const char *name, void *probe); + +/* + * Disconnect a probe from a tracepoint. + * Internal API, should not be used directly. + */ +extern int tracepoint_probe_unregister(const char *name, void *probe); + +extern int tracepoint_probe_register_noupdate(const char *name, void *probe); +extern int tracepoint_probe_unregister_noupdate(const char *name, void *probe); +extern void tracepoint_probe_update_all(void); + +struct tracepoint_iter { +//ust// struct module *module; + struct tracepoint_lib *lib; + struct tracepoint *tracepoint; +}; + +extern void tracepoint_iter_start(struct tracepoint_iter *iter); +extern void tracepoint_iter_next(struct tracepoint_iter *iter); +extern void tracepoint_iter_stop(struct tracepoint_iter *iter); +extern void tracepoint_iter_reset(struct tracepoint_iter *iter); +extern int tracepoint_get_iter_range(struct tracepoint **tracepoint, + struct tracepoint *begin, struct tracepoint *end); + +/* + * tracepoint_synchronize_unregister must be called between the last tracepoint + * probe unregistration and the end of module exit to make sure there is no + * caller executing a probe when it is freed. + */ +static inline void tracepoint_synchronize_unregister(void) +{ + synchronize_sched(); +} + +struct tracepoint_lib { + struct tracepoint *tracepoints_start; + int tracepoints_count; + struct list_head list; +}; + +#define TRACEPOINT_LIB \ +extern struct tracepoint __start___tracepoints[] __attribute__((visibility("hidden"))); \ +extern struct tracepoint __stop___tracepoints[] __attribute__((visibility("hidden"))); \ + \ +static void __attribute__((constructor)) __tracepoints__init(void) \ +{ \ + tracepoint_register_lib(__start___tracepoints, (((long)__stop___tracepoints)-((long)__start___tracepoints))/sizeof(struct tracepoint));\ +} +#endif diff --git a/libust/tracer.c b/libust/tracer.c new file mode 100644 index 0000000..b807a50 --- /dev/null +++ b/libust/tracer.c @@ -0,0 +1,1148 @@ +/* + * ltt/ltt-tracer.c + * + * (C) Copyright 2005-2008 - + * Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) + * + * Tracing management internal kernel API. Trace buffer allocation/free, tracing + * start/stop. + * + * Author: + * Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) + * + * Inspired from LTT : + * Karim Yaghmour (karim@opersys.com) + * Tom Zanussi (zanussi@us.ibm.com) + * Bob Wisniewski (bob@watson.ibm.com) + * And from K42 : + * Bob Wisniewski (bob@watson.ibm.com) + * + * Changelog: + * 22/09/06, Move to the marker/probes mechanism. + * 19/10/05, Complete lockless mechanism. + * 27/05/05, Modular redesign and rewrite. + */ + +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +//ust// #include +#include "kernelcompat.h" +#include "tracercore.h" +#include "tracer.h" +#include "usterr.h" + +//ust// static void async_wakeup(unsigned long data); +//ust// +//ust// static DEFINE_TIMER(ltt_async_wakeup_timer, async_wakeup, 0, 0); + +/* Default callbacks for modules */ +notrace int ltt_filter_control_default(enum ltt_filter_control_msg msg, + struct ltt_trace_struct *trace) +{ + return 0; +} + +int ltt_statedump_default(struct ltt_trace_struct *trace) +{ + return 0; +} + +/* Callbacks for registered modules */ + +int (*ltt_filter_control_functor) + (enum ltt_filter_control_msg msg, struct ltt_trace_struct *trace) = + ltt_filter_control_default; +struct module *ltt_filter_control_owner; + +/* These function pointers are protected by a trace activation check */ +struct module *ltt_run_filter_owner; +int (*ltt_statedump_functor)(struct ltt_trace_struct *trace) = + ltt_statedump_default; +struct module *ltt_statedump_owner; + +struct chan_info_struct { + const char *name; + unsigned int def_subbufsize; + unsigned int def_subbufcount; +} chan_infos[] = { + [LTT_CHANNEL_METADATA] = { + LTT_METADATA_CHANNEL, + LTT_DEFAULT_SUBBUF_SIZE_LOW, + LTT_DEFAULT_N_SUBBUFS_LOW, + }, + [LTT_CHANNEL_UST] = { + LTT_UST_CHANNEL, + LTT_DEFAULT_SUBBUF_SIZE_HIGH, + LTT_DEFAULT_N_SUBBUFS_HIGH, + }, +}; + +static enum ltt_channels get_channel_type_from_name(const char *name) +{ + int i; + + if (!name) + return LTT_CHANNEL_UST; + + for (i = 0; i < ARRAY_SIZE(chan_infos); i++) + if (chan_infos[i].name && !strcmp(name, chan_infos[i].name)) + return (enum ltt_channels)i; + + return LTT_CHANNEL_UST; +} + +/** + * ltt_module_register - LTT module registration + * @name: module type + * @function: callback to register + * @owner: module which owns the callback + * + * The module calling this registration function must ensure that no + * trap-inducing code will be executed by "function". E.g. vmalloc_sync_all() + * must be called between a vmalloc and the moment the memory is made visible to + * "function". This registration acts as a vmalloc_sync_all. Therefore, only if + * the module allocates virtual memory after its registration must it + * synchronize the TLBs. + */ +//ust// int ltt_module_register(enum ltt_module_function name, void *function, +//ust// struct module *owner) +//ust// { +//ust// int ret = 0; +//ust// +//ust// /* +//ust// * Make sure no page fault can be triggered by the module about to be +//ust// * registered. We deal with this here so we don't have to call +//ust// * vmalloc_sync_all() in each module's init. +//ust// */ +//ust// vmalloc_sync_all(); +//ust// +//ust// switch (name) { +//ust// case LTT_FUNCTION_RUN_FILTER: +//ust// if (ltt_run_filter_owner != NULL) { +//ust// ret = -EEXIST; +//ust// goto end; +//ust// } +//ust// ltt_filter_register((ltt_run_filter_functor)function); +//ust// ltt_run_filter_owner = owner; +//ust// break; +//ust// case LTT_FUNCTION_FILTER_CONTROL: +//ust// if (ltt_filter_control_owner != NULL) { +//ust// ret = -EEXIST; +//ust// goto end; +//ust// } +//ust// ltt_filter_control_functor = +//ust// (int (*)(enum ltt_filter_control_msg, +//ust// struct ltt_trace_struct *))function; +//ust// ltt_filter_control_owner = owner; +//ust// break; +//ust// case LTT_FUNCTION_STATEDUMP: +//ust// if (ltt_statedump_owner != NULL) { +//ust// ret = -EEXIST; +//ust// goto end; +//ust// } +//ust// ltt_statedump_functor = +//ust// (int (*)(struct ltt_trace_struct *))function; +//ust// ltt_statedump_owner = owner; +//ust// break; +//ust// } +//ust// +//ust// end: +//ust// +//ust// return ret; +//ust// } +//ust// EXPORT_SYMBOL_GPL(ltt_module_register); + +/** + * ltt_module_unregister - LTT module unregistration + * @name: module type + */ +//ust// void ltt_module_unregister(enum ltt_module_function name) +//ust// { +//ust// switch (name) { +//ust// case LTT_FUNCTION_RUN_FILTER: +//ust// ltt_filter_unregister(); +//ust// ltt_run_filter_owner = NULL; +//ust// /* Wait for preempt sections to finish */ +//ust// synchronize_sched(); +//ust// break; +//ust// case LTT_FUNCTION_FILTER_CONTROL: +//ust// ltt_filter_control_functor = ltt_filter_control_default; +//ust// ltt_filter_control_owner = NULL; +//ust// break; +//ust// case LTT_FUNCTION_STATEDUMP: +//ust// ltt_statedump_functor = ltt_statedump_default; +//ust// ltt_statedump_owner = NULL; +//ust// break; +//ust// } +//ust// +//ust// } +//ust// EXPORT_SYMBOL_GPL(ltt_module_unregister); + +static LIST_HEAD(ltt_transport_list); + +/** + * ltt_transport_register - LTT transport registration + * @transport: transport structure + * + * Registers a transport which can be used as output to extract the data out of + * LTTng. The module calling this registration function must ensure that no + * trap-inducing code will be executed by the transport functions. E.g. + * vmalloc_sync_all() must be called between a vmalloc and the moment the memory + * is made visible to the transport function. This registration acts as a + * vmalloc_sync_all. Therefore, only if the module allocates virtual memory + * after its registration must it synchronize the TLBs. + */ +void ltt_transport_register(struct ltt_transport *transport) +{ + /* + * Make sure no page fault can be triggered by the module about to be + * registered. We deal with this here so we don't have to call + * vmalloc_sync_all() in each module's init. + */ +//ust// vmalloc_sync_all(); + + ltt_lock_traces(); + list_add_tail(&transport->node, <t_transport_list); + ltt_unlock_traces(); +} +//ust// EXPORT_SYMBOL_GPL(ltt_transport_register); + +/** + * ltt_transport_unregister - LTT transport unregistration + * @transport: transport structure + */ +void ltt_transport_unregister(struct ltt_transport *transport) +{ + ltt_lock_traces(); + list_del(&transport->node); + ltt_unlock_traces(); +} +//ust// EXPORT_SYMBOL_GPL(ltt_transport_unregister); + +static inline int is_channel_overwrite(enum ltt_channels chan, + enum trace_mode mode) +{ + switch (mode) { + case LTT_TRACE_NORMAL: + return 0; + case LTT_TRACE_FLIGHT: + switch (chan) { + case LTT_CHANNEL_METADATA: + return 0; + default: + return 1; + } + case LTT_TRACE_HYBRID: + switch (chan) { + case LTT_CHANNEL_METADATA: + return 0; + default: + return 1; + } + default: + return 0; + } +} + +/** + * ltt_write_trace_header - Write trace header + * @trace: Trace information + * @header: Memory address where the information must be written to + */ +void notrace ltt_write_trace_header(struct ltt_trace_struct *trace, + struct ltt_subbuffer_header *header) +{ + header->magic_number = LTT_TRACER_MAGIC_NUMBER; + header->major_version = LTT_TRACER_VERSION_MAJOR; + header->minor_version = LTT_TRACER_VERSION_MINOR; + header->arch_size = sizeof(void *); + header->alignment = ltt_get_alignment(); + header->start_time_sec = trace->start_time.tv_sec; + header->start_time_usec = trace->start_time.tv_usec; + header->start_freq = trace->start_freq; + header->freq_scale = trace->freq_scale; +} +//ust// EXPORT_SYMBOL_GPL(ltt_write_trace_header); + +static void trace_async_wakeup(struct ltt_trace_struct *trace) +{ + int i; + struct ltt_channel_struct *chan; + + /* Must check each channel for pending read wakeup */ + for (i = 0; i < trace->nr_channels; i++) { + chan = &trace->channels[i]; + if (chan->active) + trace->ops->wakeup_channel(chan); + } +} + +//ust// /* Timer to send async wakeups to the readers */ +//ust// static void async_wakeup(unsigned long data) +//ust// { +//ust// struct ltt_trace_struct *trace; +//ust// +//ust// /* +//ust// * PREEMPT_RT does not allow spinlocks to be taken within preempt +//ust// * disable sections (spinlock taken in wake_up). However, mainline won't +//ust// * allow mutex to be taken in interrupt context. Ugly. +//ust// * A proper way to do this would be to turn the timer into a +//ust// * periodically woken up thread, but it adds to the footprint. +//ust// */ +//ust// #ifndef CONFIG_PREEMPT_RT +//ust// rcu_read_lock_sched(); +//ust// #else +//ust// ltt_lock_traces(); +//ust// #endif +//ust// list_for_each_entry_rcu(trace, <t_traces.head, list) { +//ust// trace_async_wakeup(trace); +//ust// } +//ust// #ifndef CONFIG_PREEMPT_RT +//ust// rcu_read_unlock_sched(); +//ust// #else +//ust// ltt_unlock_traces(); +//ust// #endif +//ust// +//ust// mod_timer(<t_async_wakeup_timer, jiffies + LTT_PERCPU_TIMER_INTERVAL); +//ust// } + +/** + * _ltt_trace_find - find a trace by given name. + * trace_name: trace name + * + * Returns a pointer to the trace structure, NULL if not found. + */ +struct ltt_trace_struct *_ltt_trace_find(const char *trace_name) +{ + struct ltt_trace_struct *trace; + + list_for_each_entry(trace, <t_traces.head, list) + if (!strncmp(trace->trace_name, trace_name, NAME_MAX)) + return trace; + + return NULL; +} + +/* _ltt_trace_find_setup : + * find a trace in setup list by given name. + * + * Returns a pointer to the trace structure, NULL if not found. + */ +struct ltt_trace_struct *_ltt_trace_find_setup(const char *trace_name) +{ + struct ltt_trace_struct *trace; + + list_for_each_entry(trace, <t_traces.setup_head, list) + if (!strncmp(trace->trace_name, trace_name, NAME_MAX)) + return trace; + + return NULL; +} +//ust// EXPORT_SYMBOL_GPL(_ltt_trace_find_setup); + +/** + * ltt_release_transport - Release an LTT transport + * @kref : reference count on the transport + */ +void ltt_release_transport(struct kref *kref) +{ + struct ltt_trace_struct *trace = container_of(kref, + struct ltt_trace_struct, ltt_transport_kref); +//ust// trace->ops->remove_dirs(trace); +} +//ust// EXPORT_SYMBOL_GPL(ltt_release_transport); + +/** + * ltt_release_trace - Release a LTT trace + * @kref : reference count on the trace + */ +void ltt_release_trace(struct kref *kref) +{ + struct ltt_trace_struct *trace = container_of(kref, + struct ltt_trace_struct, kref); + ltt_channels_trace_free(trace->channels); + kfree(trace); +} +//ust// EXPORT_SYMBOL_GPL(ltt_release_trace); + +static inline void prepare_chan_size_num(unsigned int *subbuf_size, + unsigned int *n_subbufs) +{ + *subbuf_size = 1 << get_count_order(*subbuf_size); + *n_subbufs = 1 << get_count_order(*n_subbufs); + + /* Subbuf size and number must both be power of two */ + WARN_ON(hweight32(*subbuf_size) != 1); + WARN_ON(hweight32(*n_subbufs) != 1); +} + +int _ltt_trace_setup(const char *trace_name) +{ + int err = 0; + struct ltt_trace_struct *new_trace = NULL; + int metadata_index; + unsigned int chan; + enum ltt_channels chantype; + + if (_ltt_trace_find_setup(trace_name)) { + printk(KERN_ERR "LTT : Trace name %s already used.\n", + trace_name); + err = -EEXIST; + goto traces_error; + } + + if (_ltt_trace_find(trace_name)) { + printk(KERN_ERR "LTT : Trace name %s already used.\n", + trace_name); + err = -EEXIST; + goto traces_error; + } + + new_trace = kzalloc(sizeof(struct ltt_trace_struct), GFP_KERNEL); + if (!new_trace) { + printk(KERN_ERR + "LTT : Unable to allocate memory for trace %s\n", + trace_name); + err = -ENOMEM; + goto traces_error; + } + strncpy(new_trace->trace_name, trace_name, NAME_MAX); + new_trace->channels = ltt_channels_trace_alloc(&new_trace->nr_channels, + 0, 1); + if (!new_trace->channels) { + printk(KERN_ERR + "LTT : Unable to allocate memory for chaninfo %s\n", + trace_name); + err = -ENOMEM; + goto trace_free; + } + + /* + * Force metadata channel to active, no overwrite. + */ + metadata_index = ltt_channels_get_index_from_name("metadata"); + WARN_ON(metadata_index < 0); + new_trace->channels[metadata_index].overwrite = 0; + new_trace->channels[metadata_index].active = 1; + + /* + * Set hardcoded tracer defaults for some channels + */ + for (chan = 0; chan < new_trace->nr_channels; chan++) { + if (!(new_trace->channels[chan].active)) + continue; + + chantype = get_channel_type_from_name( + ltt_channels_get_name_from_index(chan)); + new_trace->channels[chan].subbuf_size = + chan_infos[chantype].def_subbufsize; + new_trace->channels[chan].subbuf_cnt = + chan_infos[chantype].def_subbufcount; + } + + list_add(&new_trace->list, <t_traces.setup_head); + return 0; + +trace_free: + kfree(new_trace); +traces_error: + return err; +} +//ust// EXPORT_SYMBOL_GPL(_ltt_trace_setup); + + +int ltt_trace_setup(const char *trace_name) +{ + int ret; + ltt_lock_traces(); + ret = _ltt_trace_setup(trace_name); + ltt_unlock_traces(); + return ret; +} +//ust// EXPORT_SYMBOL_GPL(ltt_trace_setup); + +/* must be called from within a traces lock. */ +static void _ltt_trace_free(struct ltt_trace_struct *trace) +{ + list_del(&trace->list); + kfree(trace); +} + +int ltt_trace_set_type(const char *trace_name, const char *trace_type) +{ + int err = 0; + struct ltt_trace_struct *trace; + struct ltt_transport *tran_iter, *transport = NULL; + + ltt_lock_traces(); + + trace = _ltt_trace_find_setup(trace_name); + if (!trace) { + printk(KERN_ERR "LTT : Trace not found %s\n", trace_name); + err = -ENOENT; + goto traces_error; + } + + list_for_each_entry(tran_iter, <t_transport_list, node) { + if (!strcmp(tran_iter->name, trace_type)) { + transport = tran_iter; + break; + } + } + if (!transport) { + printk(KERN_ERR "LTT : Transport %s is not present.\n", + trace_type); + err = -EINVAL; + goto traces_error; + } + + trace->transport = transport; + +traces_error: + ltt_unlock_traces(); + return err; +} +//ust// EXPORT_SYMBOL_GPL(ltt_trace_set_type); + +int ltt_trace_set_channel_subbufsize(const char *trace_name, + const char *channel_name, unsigned int size) +{ + int err = 0; + struct ltt_trace_struct *trace; + int index; + + ltt_lock_traces(); + + trace = _ltt_trace_find_setup(trace_name); + if (!trace) { + printk(KERN_ERR "LTT : Trace not found %s\n", trace_name); + err = -ENOENT; + goto traces_error; + } + + index = ltt_channels_get_index_from_name(channel_name); + if (index < 0) { + printk(KERN_ERR "LTT : Channel %s not found\n", channel_name); + err = -ENOENT; + goto traces_error; + } + trace->channels[index].subbuf_size = size; + +traces_error: + ltt_unlock_traces(); + return err; +} +//ust// EXPORT_SYMBOL_GPL(ltt_trace_set_channel_subbufsize); + +int ltt_trace_set_channel_subbufcount(const char *trace_name, + const char *channel_name, unsigned int cnt) +{ + int err = 0; + struct ltt_trace_struct *trace; + int index; + + ltt_lock_traces(); + + trace = _ltt_trace_find_setup(trace_name); + if (!trace) { + printk(KERN_ERR "LTT : Trace not found %s\n", trace_name); + err = -ENOENT; + goto traces_error; + } + + index = ltt_channels_get_index_from_name(channel_name); + if (index < 0) { + printk(KERN_ERR "LTT : Channel %s not found\n", channel_name); + err = -ENOENT; + goto traces_error; + } + trace->channels[index].subbuf_cnt = cnt; + +traces_error: + ltt_unlock_traces(); + return err; +} +//ust// EXPORT_SYMBOL_GPL(ltt_trace_set_channel_subbufcount); + +int ltt_trace_set_channel_enable(const char *trace_name, + const char *channel_name, unsigned int enable) +{ + int err = 0; + struct ltt_trace_struct *trace; + int index; + + ltt_lock_traces(); + + trace = _ltt_trace_find_setup(trace_name); + if (!trace) { + printk(KERN_ERR "LTT : Trace not found %s\n", trace_name); + err = -ENOENT; + goto traces_error; + } + + /* + * Datas in metadata channel(marker info) is necessary to be able to + * read the trace, we always enable this channel. + */ + if (!enable && !strcmp(channel_name, "metadata")) { + printk(KERN_ERR "LTT : Trying to disable metadata channel\n"); + err = -EINVAL; + goto traces_error; + } + + index = ltt_channels_get_index_from_name(channel_name); + if (index < 0) { + printk(KERN_ERR "LTT : Channel %s not found\n", channel_name); + err = -ENOENT; + goto traces_error; + } + + trace->channels[index].active = enable; + +traces_error: + ltt_unlock_traces(); + return err; +} +//ust// EXPORT_SYMBOL_GPL(ltt_trace_set_channel_enable); + +int ltt_trace_set_channel_overwrite(const char *trace_name, + const char *channel_name, unsigned int overwrite) +{ + int err = 0; + struct ltt_trace_struct *trace; + int index; + + ltt_lock_traces(); + + trace = _ltt_trace_find_setup(trace_name); + if (!trace) { + printk(KERN_ERR "LTT : Trace not found %s\n", trace_name); + err = -ENOENT; + goto traces_error; + } + + /* + * Always put the metadata channel in non-overwrite mode : + * This is a very low traffic channel and it can't afford to have its + * data overwritten : this data (marker info) is necessary to be + * able to read the trace. + */ + if (overwrite && !strcmp(channel_name, "metadata")) { + printk(KERN_ERR "LTT : Trying to set metadata channel to " + "overwrite mode\n"); + err = -EINVAL; + goto traces_error; + } + + index = ltt_channels_get_index_from_name(channel_name); + if (index < 0) { + printk(KERN_ERR "LTT : Channel %s not found\n", channel_name); + err = -ENOENT; + goto traces_error; + } + + trace->channels[index].overwrite = overwrite; + +traces_error: + ltt_unlock_traces(); + return err; +} +//ust// EXPORT_SYMBOL_GPL(ltt_trace_set_channel_overwrite); + +int ltt_trace_alloc(const char *trace_name) +{ + int err = 0; + struct ltt_trace_struct *trace; + int subbuf_size, subbuf_cnt; + unsigned long flags; + int chan; + const char *channel_name; + + ltt_lock_traces(); + + trace = _ltt_trace_find_setup(trace_name); + if (!trace) { + printk(KERN_ERR "LTT : Trace not found %s\n", trace_name); + err = -ENOENT; + goto traces_error; + } + + kref_init(&trace->kref); + kref_init(&trace->ltt_transport_kref); +//ust// init_waitqueue_head(&trace->kref_wq); + trace->active = 0; +//ust// get_trace_clock(); + trace->freq_scale = trace_clock_freq_scale(); + + if (!trace->transport) { + printk(KERN_ERR "LTT : Transport is not set.\n"); + err = -EINVAL; + goto transport_error; + } +//ust// if (!try_module_get(trace->transport->owner)) { +//ust// printk(KERN_ERR "LTT : Can't lock transport module.\n"); +//ust// err = -ENODEV; +//ust// goto transport_error; +//ust// } + trace->ops = &trace->transport->ops; + +//ust// err = trace->ops->create_dirs(trace); +//ust// if (err) { +//ust// printk(KERN_ERR "LTT : Can't create dir for trace %s.\n", +//ust// trace_name); +//ust// goto dirs_error; +//ust// } + +//ust// local_irq_save(flags); + trace->start_freq = trace_clock_frequency(); + trace->start_tsc = trace_clock_read64(); + gettimeofday(&trace->start_time, NULL); //ust// changed +//ust// local_irq_restore(flags); + + for (chan = 0; chan < trace->nr_channels; chan++) { + if (!(trace->channels[chan].active)) + continue; + + channel_name = ltt_channels_get_name_from_index(chan); + WARN_ON(!channel_name); + subbuf_size = trace->channels[chan].subbuf_size; + subbuf_cnt = trace->channels[chan].subbuf_cnt; + prepare_chan_size_num(&subbuf_size, &subbuf_cnt); + err = trace->ops->create_channel(trace_name, trace, + trace->dentry.trace_root, + channel_name, + &trace->channels[chan], + subbuf_size, + subbuf_cnt, + trace->channels[chan].overwrite); + if (err != 0) { + printk(KERN_ERR "LTT : Can't create channel %s.\n", + channel_name); + goto create_channel_error; + } + } + + list_del(&trace->list); +//ust// if (list_empty(<t_traces.head)) { +//ust// mod_timer(<t_async_wakeup_timer, +//ust// jiffies + LTT_PERCPU_TIMER_INTERVAL); +//ust// set_kernel_trace_flag_all_tasks(); +//ust// } + list_add_rcu(&trace->list, <t_traces.head); +//ust// synchronize_sched(); + + ltt_unlock_traces(); + + return 0; + +create_channel_error: + for (chan--; chan >= 0; chan--) + if (trace->channels[chan].active) + trace->ops->remove_channel(&trace->channels[chan]); + +dirs_error: +//ust// module_put(trace->transport->owner); +transport_error: +//ust// put_trace_clock(); +traces_error: + ltt_unlock_traces(); + return err; +} +//ust// EXPORT_SYMBOL_GPL(ltt_trace_alloc); + +/* + * It is worked as a wrapper for current version of ltt_control.ko. + * We will make a new ltt_control based on debugfs, and control each channel's + * buffer. + */ +static int ltt_trace_create(const char *trace_name, const char *trace_type, + enum trace_mode mode, + unsigned int subbuf_size_low, unsigned int n_subbufs_low, + unsigned int subbuf_size_med, unsigned int n_subbufs_med, + unsigned int subbuf_size_high, unsigned int n_subbufs_high) +{ + int err = 0; + + err = ltt_trace_setup(trace_name); + if (IS_ERR_VALUE(err)) + return err; + + err = ltt_trace_set_type(trace_name, trace_type); + if (IS_ERR_VALUE(err)) + return err; + + err = ltt_trace_alloc(trace_name); + if (IS_ERR_VALUE(err)) + return err; + + return err; +} + +/* Must be called while sure that trace is in the list. */ +static int _ltt_trace_destroy(struct ltt_trace_struct *trace) +{ + int err = -EPERM; + + if (trace == NULL) { + err = -ENOENT; + goto traces_error; + } + if (trace->active) { + printk(KERN_ERR + "LTT : Can't destroy trace %s : tracer is active\n", + trace->trace_name); + err = -EBUSY; + goto active_error; + } + /* Everything went fine */ +//ust// list_del_rcu(&trace->list); +//ust// synchronize_sched(); + if (list_empty(<t_traces.head)) { +//ust// clear_kernel_trace_flag_all_tasks(); + /* + * We stop the asynchronous delivery of reader wakeup, but + * we must make one last check for reader wakeups pending + * later in __ltt_trace_destroy. + */ +//ust// del_timer_sync(<t_async_wakeup_timer); + } + return 0; + + /* error handling */ +active_error: +traces_error: + return err; +} + +/* Sleepable part of the destroy */ +static void __ltt_trace_destroy(struct ltt_trace_struct *trace) +{ + int i; + struct ltt_channel_struct *chan; + + for (i = 0; i < trace->nr_channels; i++) { + chan = &trace->channels[i]; + if (chan->active) + trace->ops->finish_channel(chan); + } + + return; /* FIXME: temporary for ust */ +//ust// flush_scheduled_work(); + + /* + * The currently destroyed trace is not in the trace list anymore, + * so it's safe to call the async wakeup ourself. It will deliver + * the last subbuffers. + */ + trace_async_wakeup(trace); + + for (i = 0; i < trace->nr_channels; i++) { + chan = &trace->channels[i]; + if (chan->active) + trace->ops->remove_channel(chan); + } + + kref_put(&trace->ltt_transport_kref, ltt_release_transport); + +//ust// module_put(trace->transport->owner); + + /* + * Wait for lttd readers to release the files, therefore making sure + * the last subbuffers have been read. + */ +//ust// if (atomic_read(&trace->kref.refcount) > 1) { +//ust// int ret = 0; +//ust// __wait_event_interruptible(trace->kref_wq, +//ust// (atomic_read(&trace->kref.refcount) == 1), ret); +//ust// } + kref_put(&trace->kref, ltt_release_trace); +} + +int ltt_trace_destroy(const char *trace_name) +{ + int err = 0; + struct ltt_trace_struct *trace; + + ltt_lock_traces(); + + trace = _ltt_trace_find(trace_name); + if (trace) { + err = _ltt_trace_destroy(trace); + if (err) + goto error; + + ltt_unlock_traces(); + + __ltt_trace_destroy(trace); +//ust// put_trace_clock(); + + return 0; + } + + trace = _ltt_trace_find_setup(trace_name); + if (trace) { + _ltt_trace_free(trace); + ltt_unlock_traces(); + return 0; + } + + err = -ENOENT; + + /* Error handling */ +error: + ltt_unlock_traces(); + return err; +} +//ust// EXPORT_SYMBOL_GPL(ltt_trace_destroy); + +/* must be called from within a traces lock. */ +static int _ltt_trace_start(struct ltt_trace_struct *trace) +{ + int err = 0; + + if (trace == NULL) { + err = -ENOENT; + goto traces_error; + } + if (trace->active) + printk(KERN_INFO "LTT : Tracing already active for trace %s\n", + trace->trace_name); +//ust// if (!try_module_get(ltt_run_filter_owner)) { +//ust// err = -ENODEV; +//ust// printk(KERN_ERR "LTT : Can't lock filter module.\n"); +//ust// goto get_ltt_run_filter_error; +//ust// } + trace->active = 1; + /* Read by trace points without protection : be careful */ + ltt_traces.num_active_traces++; + return err; + + /* error handling */ +get_ltt_run_filter_error: +traces_error: + return err; +} + +int ltt_trace_start(const char *trace_name) +{ + int err = 0; + struct ltt_trace_struct *trace; + + ltt_lock_traces(); + + trace = _ltt_trace_find(trace_name); + err = _ltt_trace_start(trace); + if (err) + goto no_trace; + + ltt_unlock_traces(); + + /* + * Call the kernel state dump. + * Events will be mixed with real kernel events, it's ok. + * Notice that there is no protection on the trace : that's exactly + * why we iterate on the list and check for trace equality instead of + * directly using this trace handle inside the logging function. + */ + + ltt_dump_marker_state(trace); + +//ust// if (!try_module_get(ltt_statedump_owner)) { +//ust// err = -ENODEV; +//ust// printk(KERN_ERR +//ust// "LTT : Can't lock state dump module.\n"); +//ust// } else { + ltt_statedump_functor(trace); +//ust// module_put(ltt_statedump_owner); +//ust// } + + return err; + + /* Error handling */ +no_trace: + ltt_unlock_traces(); + return err; +} +//ust// EXPORT_SYMBOL_GPL(ltt_trace_start); + +/* must be called from within traces lock */ +static int _ltt_trace_stop(struct ltt_trace_struct *trace) +{ + int err = -EPERM; + + if (trace == NULL) { + err = -ENOENT; + goto traces_error; + } + if (!trace->active) + printk(KERN_INFO "LTT : Tracing not active for trace %s\n", + trace->trace_name); + if (trace->active) { + trace->active = 0; + ltt_traces.num_active_traces--; +//ust// synchronize_sched(); /* Wait for each tracing to be finished */ + } +//ust// module_put(ltt_run_filter_owner); + /* Everything went fine */ + return 0; + + /* Error handling */ +traces_error: + return err; +} + +int ltt_trace_stop(const char *trace_name) +{ + int err = 0; + struct ltt_trace_struct *trace; + + ltt_lock_traces(); + trace = _ltt_trace_find(trace_name); + err = _ltt_trace_stop(trace); + ltt_unlock_traces(); + return err; +} +//ust// EXPORT_SYMBOL_GPL(ltt_trace_stop); + +/** + * ltt_control - Trace control in-kernel API + * @msg: Action to perform + * @trace_name: Trace on which the action must be done + * @trace_type: Type of trace (normal, flight, hybrid) + * @args: Arguments specific to the action + */ +//ust// int ltt_control(enum ltt_control_msg msg, const char *trace_name, +//ust// const char *trace_type, union ltt_control_args args) +//ust// { +//ust// int err = -EPERM; +//ust// +//ust// printk(KERN_ALERT "ltt_control : trace %s\n", trace_name); +//ust// switch (msg) { +//ust// case LTT_CONTROL_START: +//ust// printk(KERN_DEBUG "Start tracing %s\n", trace_name); +//ust// err = ltt_trace_start(trace_name); +//ust// break; +//ust// case LTT_CONTROL_STOP: +//ust// printk(KERN_DEBUG "Stop tracing %s\n", trace_name); +//ust// err = ltt_trace_stop(trace_name); +//ust// break; +//ust// case LTT_CONTROL_CREATE_TRACE: +//ust// printk(KERN_DEBUG "Creating trace %s\n", trace_name); +//ust// err = ltt_trace_create(trace_name, trace_type, +//ust// args.new_trace.mode, +//ust// args.new_trace.subbuf_size_low, +//ust// args.new_trace.n_subbufs_low, +//ust// args.new_trace.subbuf_size_med, +//ust// args.new_trace.n_subbufs_med, +//ust// args.new_trace.subbuf_size_high, +//ust// args.new_trace.n_subbufs_high); +//ust// break; +//ust// case LTT_CONTROL_DESTROY_TRACE: +//ust// printk(KERN_DEBUG "Destroying trace %s\n", trace_name); +//ust// err = ltt_trace_destroy(trace_name); +//ust// break; +//ust// } +//ust// return err; +//ust// } +//ust// EXPORT_SYMBOL_GPL(ltt_control); + +/** + * ltt_filter_control - Trace filter control in-kernel API + * @msg: Action to perform on the filter + * @trace_name: Trace on which the action must be done + */ +int ltt_filter_control(enum ltt_filter_control_msg msg, const char *trace_name) +{ + int err; + struct ltt_trace_struct *trace; + + printk(KERN_DEBUG "ltt_filter_control : trace %s\n", trace_name); + ltt_lock_traces(); + trace = _ltt_trace_find(trace_name); + if (trace == NULL) { + printk(KERN_ALERT + "Trace does not exist. Cannot proxy control request\n"); + err = -ENOENT; + goto trace_error; + } +//ust// if (!try_module_get(ltt_filter_control_owner)) { +//ust// err = -ENODEV; +//ust// goto get_module_error; +//ust// } + switch (msg) { + case LTT_FILTER_DEFAULT_ACCEPT: + printk(KERN_DEBUG + "Proxy filter default accept %s\n", trace_name); + err = (*ltt_filter_control_functor)(msg, trace); + break; + case LTT_FILTER_DEFAULT_REJECT: + printk(KERN_DEBUG + "Proxy filter default reject %s\n", trace_name); + err = (*ltt_filter_control_functor)(msg, trace); + break; + default: + err = -EPERM; + } +//ust// module_put(ltt_filter_control_owner); + +get_module_error: +trace_error: + ltt_unlock_traces(); + return err; +} +//ust// EXPORT_SYMBOL_GPL(ltt_filter_control); + +//ust// int __init ltt_init(void) +//ust// { +//ust// /* Make sure no page fault can be triggered by this module */ +//ust// vmalloc_sync_all(); +//ust// return 0; +//ust// } + +//ust// module_init(ltt_init) + +//ust// static void __exit ltt_exit(void) +//ust// { +//ust// struct ltt_trace_struct *trace; +//ust// struct list_head *pos, *n; +//ust// +//ust// ltt_lock_traces(); +//ust// /* Stop each trace, currently being read by RCU read-side */ +//ust// list_for_each_entry_rcu(trace, <t_traces.head, list) +//ust// _ltt_trace_stop(trace); +//ust// /* Wait for quiescent state. Readers have preemption disabled. */ +//ust// synchronize_sched(); +//ust// /* Safe iteration is now permitted. It does not have to be RCU-safe +//ust// * because no readers are left. */ +//ust// list_for_each_safe(pos, n, <t_traces.head) { +//ust// trace = container_of(pos, struct ltt_trace_struct, list); +//ust// /* _ltt_trace_destroy does a synchronize_sched() */ +//ust// _ltt_trace_destroy(trace); +//ust// __ltt_trace_destroy(trace); +//ust// } +//ust// /* free traces in pre-alloc status */ +//ust// list_for_each_safe(pos, n, <t_traces.setup_head) { +//ust// trace = container_of(pos, struct ltt_trace_struct, list); +//ust// _ltt_trace_free(trace); +//ust// } +//ust// +//ust// ltt_unlock_traces(); +//ust// } + +//ust// module_exit(ltt_exit) + +//ust// MODULE_LICENSE("GPL"); +//ust// MODULE_AUTHOR("Mathieu Desnoyers"); +//ust// MODULE_DESCRIPTION("Linux Trace Toolkit Next Generation Tracer Kernel API"); diff --git a/libust/tracer.h b/libust/tracer.h new file mode 100644 index 0000000..cc86d54 --- /dev/null +++ b/libust/tracer.h @@ -0,0 +1,624 @@ +/* + * Copyright (C) 2005,2006,2008 Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) + * Copyright (C) 2009 Pierre-Marc Fournier + * + * This contains the definitions for the Linux Trace Toolkit tracer. + * + * Ported to userspace by Pierre-Marc Fournier. + * + * This file is released under the GPLv2. + */ + +#ifndef _LTT_TRACER_H +#define _LTT_TRACER_H + +#include +#include +#include "relay.h" +#include "list.h" +#include "kernelcompat.h" +#include "channels.h" +#include "tracercore.h" +#include "marker.h" + +/* Number of bytes to log with a read/write event */ +#define LTT_LOG_RW_SIZE 32L + +/* Interval (in jiffies) at which the LTT per-CPU timer fires */ +#define LTT_PERCPU_TIMER_INTERVAL 1 + +#ifndef LTT_ARCH_TYPE +#define LTT_ARCH_TYPE LTT_ARCH_TYPE_UNDEFINED +#endif + +#ifndef LTT_ARCH_VARIANT +#define LTT_ARCH_VARIANT LTT_ARCH_VARIANT_NONE +#endif + +struct ltt_active_marker; + +/* Maximum number of callbacks per marker */ +#define LTT_NR_CALLBACKS 10 + +struct ltt_serialize_closure; +struct ltt_probe_private_data; + +/* Serialization callback '%k' */ +typedef size_t (*ltt_serialize_cb)(struct rchan_buf *buf, size_t buf_offset, + struct ltt_serialize_closure *closure, + void *serialize_private, int *largest_align, + const char *fmt, va_list *args); + +struct ltt_serialize_closure { + ltt_serialize_cb *callbacks; + long cb_args[LTT_NR_CALLBACKS]; + unsigned int cb_idx; +}; + +size_t ltt_serialize_data(struct rchan_buf *buf, size_t buf_offset, + struct ltt_serialize_closure *closure, + void *serialize_private, + int *largest_align, const char *fmt, va_list *args); + +struct ltt_available_probe { + const char *name; /* probe name */ + const char *format; + marker_probe_func *probe_func; + ltt_serialize_cb callbacks[LTT_NR_CALLBACKS]; + struct list_head node; /* registered probes list */ +}; + +struct ltt_probe_private_data { + struct ltt_trace_struct *trace; /* + * Target trace, for metadata + * or statedump. + */ + ltt_serialize_cb serializer; /* + * Serialization function override. + */ + void *serialize_private; /* + * Private data for serialization + * functions. + */ +}; + +enum ltt_channels { + LTT_CHANNEL_METADATA, + LTT_CHANNEL_UST, +}; + +struct ltt_active_marker { + struct list_head node; /* active markers list */ + const char *channel; + const char *name; + const char *format; + struct ltt_available_probe *probe; +}; + +struct marker; //ust// +extern void ltt_vtrace(const struct marker *mdata, void *probe_data, + void *call_data, const char *fmt, va_list *args); +extern void ltt_trace(const struct marker *mdata, void *probe_data, + void *call_data, const char *fmt, ...); + +/* + * Unique ID assigned to each registered probe. + */ +enum marker_id { + MARKER_ID_SET_MARKER_ID = 0, /* Static IDs available (range 0-7) */ + MARKER_ID_SET_MARKER_FORMAT, + MARKER_ID_COMPACT, /* Compact IDs (range: 8-127) */ + MARKER_ID_DYNAMIC, /* Dynamic IDs (range: 128-65535) */ +}; + +/* static ids 0-1 reserved for internal use. */ +#define MARKER_CORE_IDS 2 +static inline enum marker_id marker_id_type(uint16_t id) +{ + if (id < MARKER_CORE_IDS) + return (enum marker_id)id; + else + return MARKER_ID_DYNAMIC; +} + +struct user_dbg_data { + unsigned long avail_size; + unsigned long write; + unsigned long read; +}; + +struct ltt_trace_ops { + /* First 32 bytes cache-hot cacheline */ + int (*reserve_slot) (struct ltt_trace_struct *trace, + struct ltt_channel_struct *channel, + void **transport_data, size_t data_size, + size_t *slot_size, long *buf_offset, u64 *tsc, + unsigned int *rflags, + int largest_align); + void (*commit_slot) (struct ltt_channel_struct *channel, + void **transport_data, long buf_offset, + size_t slot_size); + void (*wakeup_channel) (struct ltt_channel_struct *ltt_channel); + int (*user_blocking) (struct ltt_trace_struct *trace, + unsigned int index, size_t data_size, + struct user_dbg_data *dbg); + /* End of first 32 bytes cacheline */ + int (*create_dirs) (struct ltt_trace_struct *new_trace); + void (*remove_dirs) (struct ltt_trace_struct *new_trace); + int (*create_channel) (const char *trace_name, + struct ltt_trace_struct *trace, + struct dentry *dir, const char *channel_name, + struct ltt_channel_struct *ltt_chan, + unsigned int subbuf_size, + unsigned int n_subbufs, int overwrite); + void (*finish_channel) (struct ltt_channel_struct *channel); + void (*remove_channel) (struct ltt_channel_struct *channel); + void (*user_errors) (struct ltt_trace_struct *trace, + unsigned int index, size_t data_size, + struct user_dbg_data *dbg); +} ____cacheline_aligned; + +struct ltt_transport { + char *name; + struct module *owner; + struct list_head node; + struct ltt_trace_ops ops; +}; + +enum trace_mode { LTT_TRACE_NORMAL, LTT_TRACE_FLIGHT, LTT_TRACE_HYBRID }; + +#define CHANNEL_FLAG_ENABLE (1U<<0) +#define CHANNEL_FLAG_OVERWRITE (1U<<1) + +/* Per-trace information - each trace/flight recorder represented by one */ +struct ltt_trace_struct { + /* First 32 bytes cache-hot cacheline */ + struct list_head list; + struct ltt_trace_ops *ops; + int active; + /* Second 32 bytes cache-hot cacheline */ + struct ltt_channel_struct *channels; + unsigned int nr_channels; + u32 freq_scale; + u64 start_freq; + u64 start_tsc; + unsigned long long start_monotonic; + struct timeval start_time; + struct ltt_channel_setting *settings; + struct { + struct dentry *trace_root; + } dentry; + struct kref kref; /* Each channel has a kref of the trace struct */ + struct ltt_transport *transport; + struct kref ltt_transport_kref; + char trace_name[NAME_MAX]; +} ____cacheline_aligned; + +/* Hardcoded event headers + * + * event header for a trace with active heartbeat : 27 bits timestamps + * + * headers are 32-bits aligned. In order to insure such alignment, a dynamic per + * trace alignment value must be done. + * + * Remember that the C compiler does align each member on the boundary + * equivalent to their own size. + * + * As relay subbuffers are aligned on pages, we are sure that they are 4 and 8 + * bytes aligned, so the buffer header and trace header are aligned. + * + * Event headers are aligned depending on the trace alignment option. + * + * Note using C structure bitfields for cross-endianness and portability + * concerns. + */ + +#define LTT_RESERVED_EVENTS 3 +#define LTT_EVENT_BITS 5 +#define LTT_FREE_EVENTS ((1 << LTT_EVENT_BITS) - LTT_RESERVED_EVENTS) +#define LTT_TSC_BITS 27 +#define LTT_TSC_MASK ((1 << LTT_TSC_BITS) - 1) + +struct ltt_event_header { + u32 id_time; /* 5 bits event id (MSB); 27 bits time (LSB) */ +}; + +/* Reservation flags */ +#define LTT_RFLAG_ID (1 << 0) +#define LTT_RFLAG_ID_SIZE (1 << 1) +#define LTT_RFLAG_ID_SIZE_TSC (1 << 2) + +/* + * We use asm/timex.h : cpu_khz/HZ variable in here : we might have to deal + * specifically with CPU frequency scaling someday, so using an interpolation + * between the start and end of buffer values is not flexible enough. Using an + * immediate frequency value permits to calculate directly the times for parts + * of a buffer that would be before a frequency change. + * + * Keep the natural field alignment for _each field_ within this structure if + * you ever add/remove a field from this header. Packed attribute is not used + * because gcc generates poor code on at least powerpc and mips. Don't ever + * let gcc add padding between the structure elements. + */ +struct ltt_subbuffer_header { + uint64_t cycle_count_begin; /* Cycle count at subbuffer start */ + uint64_t cycle_count_end; /* Cycle count at subbuffer end */ + uint32_t magic_number; /* + * Trace magic number. + * contains endianness information. + */ + uint8_t major_version; + uint8_t minor_version; + uint8_t arch_size; /* Architecture pointer size */ + uint8_t alignment; /* LTT data alignment */ + uint64_t start_time_sec; /* NTP-corrected start time */ + uint64_t start_time_usec; + uint64_t start_freq; /* + * Frequency at trace start, + * used all along the trace. + */ + uint32_t freq_scale; /* Frequency scaling (divisor) */ + uint32_t lost_size; /* Size unused at end of subbuffer */ + uint32_t buf_size; /* Size of this subbuffer */ + uint32_t events_lost; /* + * Events lost in this subbuffer since + * the beginning of the trace. + * (may overflow) + */ + uint32_t subbuf_corrupt; /* + * Corrupted (lost) subbuffers since + * the begginig of the trace. + * (may overflow) + */ + uint8_t header_end[0]; /* End of header */ +}; + +/** + * ltt_subbuffer_header_size - called on buffer-switch to a new sub-buffer + * + * Return header size without padding after the structure. Don't use packed + * structure because gcc generates inefficient code on some architectures + * (powerpc, mips..) + */ +static inline size_t ltt_subbuffer_header_size(void) +{ + return offsetof(struct ltt_subbuffer_header, header_end); +} + +/* + * ltt_get_header_size + * + * Calculate alignment offset to 32-bits. This is the alignment offset of the + * event header. + * + * Important note : + * The event header must be 32-bits. The total offset calculated here : + * + * Alignment of header struct on 32 bits (min arch size, header size) + * + sizeof(header struct) (32-bits) + * + (opt) u16 (ext. event id) + * + (opt) u16 (event_size) (if event_size == 0xFFFFUL, has ext. event size) + * + (opt) u32 (ext. event size) + * + (opt) u64 full TSC (aligned on min(64-bits, arch size)) + * + * The payload must itself determine its own alignment from the biggest type it + * contains. + * */ +static inline unsigned char ltt_get_header_size( + struct ltt_channel_struct *channel, + size_t offset, + size_t data_size, + size_t *before_hdr_pad, + unsigned int rflags) +{ + size_t orig_offset = offset; + size_t padding; + + padding = ltt_align(offset, sizeof(struct ltt_event_header)); + offset += padding; + offset += sizeof(struct ltt_event_header); + + switch (rflags) { + case LTT_RFLAG_ID_SIZE_TSC: + offset += sizeof(u16) + sizeof(u16); + if (data_size >= 0xFFFFU) + offset += sizeof(u32); + offset += ltt_align(offset, sizeof(u64)); + offset += sizeof(u64); + break; + case LTT_RFLAG_ID_SIZE: + offset += sizeof(u16) + sizeof(u16); + if (data_size >= 0xFFFFU) + offset += sizeof(u32); + break; + case LTT_RFLAG_ID: + offset += sizeof(u16); + break; + } + + *before_hdr_pad = padding; + return offset - orig_offset; +} + +/* + * ltt_write_event_header + * + * Writes the event header to the offset (already aligned on 32-bits). + * + * @trace : trace to write to. + * @channel : pointer to the channel structure.. + * @buf : buffer to write to. + * @buf_offset : buffer offset to write to (aligned on 32 bits). + * @eID : event ID + * @event_size : size of the event, excluding the event header. + * @tsc : time stamp counter. + * @rflags : reservation flags. + * + * returns : offset where the event data must be written. + */ +static inline size_t ltt_write_event_header(struct ltt_trace_struct *trace, + struct ltt_channel_struct *channel, + struct rchan_buf *buf, long buf_offset, + u16 eID, size_t event_size, + u64 tsc, unsigned int rflags) +{ + struct ltt_event_header header; + size_t small_size; + + switch (rflags) { + case LTT_RFLAG_ID_SIZE_TSC: + header.id_time = 29 << LTT_TSC_BITS; + break; + case LTT_RFLAG_ID_SIZE: + header.id_time = 30 << LTT_TSC_BITS; + break; + case LTT_RFLAG_ID: + header.id_time = 31 << LTT_TSC_BITS; + break; + default: + header.id_time = eID << LTT_TSC_BITS; + break; + } + header.id_time |= (u32)tsc & LTT_TSC_MASK; + ltt_relay_write(buf, buf_offset, &header, sizeof(header)); + buf_offset += sizeof(header); + + switch (rflags) { + case LTT_RFLAG_ID_SIZE_TSC: + small_size = min_t(size_t, event_size, 0xFFFFU); + ltt_relay_write(buf, buf_offset, + (u16[]){ (u16)eID }, sizeof(u16)); + buf_offset += sizeof(u16); + ltt_relay_write(buf, buf_offset, + (u16[]){ (u16)small_size }, sizeof(u16)); + buf_offset += sizeof(u16); + if (small_size == 0xFFFFU) { + ltt_relay_write(buf, buf_offset, + (u32[]){ (u32)event_size }, sizeof(u32)); + buf_offset += sizeof(u32); + } + buf_offset += ltt_align(buf_offset, sizeof(u64)); + ltt_relay_write(buf, buf_offset, + (u64[]){ (u64)tsc }, sizeof(u64)); + buf_offset += sizeof(u64); + break; + case LTT_RFLAG_ID_SIZE: + small_size = min_t(size_t, event_size, 0xFFFFU); + ltt_relay_write(buf, buf_offset, + (u16[]){ (u16)eID }, sizeof(u16)); + buf_offset += sizeof(u16); + ltt_relay_write(buf, buf_offset, + (u16[]){ (u16)small_size }, sizeof(u16)); + buf_offset += sizeof(u16); + if (small_size == 0xFFFFU) { + ltt_relay_write(buf, buf_offset, + (u32[]){ (u32)event_size }, sizeof(u32)); + buf_offset += sizeof(u32); + } + break; + case LTT_RFLAG_ID: + ltt_relay_write(buf, buf_offset, + (u16[]){ (u16)eID }, sizeof(u16)); + buf_offset += sizeof(u16); + break; + default: + break; + } + + return buf_offset; +} + +/* Lockless LTTng */ + +/* Buffer offset macros */ + +/* + * BUFFER_TRUNC zeroes the subbuffer offset and the subbuffer number parts of + * the offset, which leaves only the buffer number. + */ +#define BUFFER_TRUNC(offset, chan) \ + ((offset) & (~((chan)->alloc_size-1))) +#define BUFFER_OFFSET(offset, chan) ((offset) & ((chan)->alloc_size - 1)) +#define SUBBUF_OFFSET(offset, chan) ((offset) & ((chan)->subbuf_size - 1)) +#define SUBBUF_ALIGN(offset, chan) \ + (((offset) + (chan)->subbuf_size) & (~((chan)->subbuf_size - 1))) +#define SUBBUF_TRUNC(offset, chan) \ + ((offset) & (~((chan)->subbuf_size - 1))) +#define SUBBUF_INDEX(offset, chan) \ + (BUFFER_OFFSET((offset), chan) >> (chan)->subbuf_size_order) + +/* + * ltt_reserve_slot + * + * Atomic slot reservation in a LTTng buffer. It will take care of + * sub-buffer switching. + * + * Parameters: + * + * @trace : the trace structure to log to. + * @channel : the chanel to reserve space into. + * @transport_data : specific transport data. + * @data_size : size of the variable length data to log. + * @slot_size : pointer to total size of the slot (out) + * @buf_offset : pointer to reserve offset (out) + * @tsc : pointer to the tsc at the slot reservation (out) + * @rflags : reservation flags (header specificity) + * @cpu : cpu id + * + * Return : -ENOSPC if not enough space, else 0. + */ +static inline int ltt_reserve_slot( + struct ltt_trace_struct *trace, + struct ltt_channel_struct *channel, + void **transport_data, + size_t data_size, + size_t *slot_size, + long *buf_offset, + u64 *tsc, + unsigned int *rflags, + int largest_align) +{ + return trace->ops->reserve_slot(trace, channel, transport_data, + data_size, slot_size, buf_offset, tsc, rflags, + largest_align); +} + + +/* + * ltt_commit_slot + * + * Atomic unordered slot commit. Increments the commit count in the + * specified sub-buffer, and delivers it if necessary. + * + * Parameters: + * + * @channel : the chanel to reserve space into. + * @transport_data : specific transport data. + * @buf_offset : offset of beginning of reserved slot + * @slot_size : size of the reserved slot. + */ +static inline void ltt_commit_slot( + struct ltt_channel_struct *channel, + void **transport_data, + long buf_offset, + size_t slot_size) +{ + struct ltt_trace_struct *trace = channel->trace; + + trace->ops->commit_slot(channel, transport_data, buf_offset, slot_size); +} + +/* + * Control channels : + * control/metadata + * control/interrupts + * control/... + * + * cpu channel : + * cpu + */ + +#define LTT_METADATA_CHANNEL "metadata_state" +#define LTT_UST_CHANNEL "ust" + +#define LTT_FLIGHT_PREFIX "flight-" + +/* Tracer properties */ +//#define LTT_DEFAULT_SUBBUF_SIZE_LOW 65536 +#define LTT_DEFAULT_SUBBUF_SIZE_LOW 4096 +#define LTT_DEFAULT_N_SUBBUFS_LOW 2 +//#define LTT_DEFAULT_SUBBUF_SIZE_MED 262144 +#define LTT_DEFAULT_SUBBUF_SIZE_MED 4096 +#define LTT_DEFAULT_N_SUBBUFS_MED 2 +//#define LTT_DEFAULT_SUBBUF_SIZE_HIGH 1048576 +#define LTT_DEFAULT_SUBBUF_SIZE_HIGH 4096 +#define LTT_DEFAULT_N_SUBBUFS_HIGH 2 +#define LTT_TRACER_MAGIC_NUMBER 0x00D6B7ED +#define LTT_TRACER_VERSION_MAJOR 2 +#define LTT_TRACER_VERSION_MINOR 3 + +/* + * Size reserved for high priority events (interrupts, NMI, BH) at the end of a + * nearly full buffer. User space won't use this last amount of space when in + * blocking mode. This space also includes the event header that would be + * written by this user space event. + */ +#define LTT_RESERVE_CRITICAL 4096 + +/* Register and unregister function pointers */ + +enum ltt_module_function { + LTT_FUNCTION_RUN_FILTER, + LTT_FUNCTION_FILTER_CONTROL, + LTT_FUNCTION_STATEDUMP +}; + +void ltt_transport_register(struct ltt_transport *transport); +void ltt_transport_unregister(struct ltt_transport *transport); + +/* Exported control function */ + +union ltt_control_args { + struct { + enum trace_mode mode; + unsigned int subbuf_size_low; + unsigned int n_subbufs_low; + unsigned int subbuf_size_med; + unsigned int n_subbufs_med; + unsigned int subbuf_size_high; + unsigned int n_subbufs_high; + } new_trace; +}; + +int _ltt_trace_setup(const char *trace_name); +int ltt_trace_setup(const char *trace_name); +struct ltt_trace_struct *_ltt_trace_find_setup(const char *trace_name); +int ltt_trace_set_type(const char *trace_name, const char *trace_type); +int ltt_trace_set_channel_subbufsize(const char *trace_name, + const char *channel_name, unsigned int size); +int ltt_trace_set_channel_subbufcount(const char *trace_name, + const char *channel_name, unsigned int cnt); +int ltt_trace_set_channel_enable(const char *trace_name, + const char *channel_name, unsigned int enable); +int ltt_trace_set_channel_overwrite(const char *trace_name, + const char *channel_name, unsigned int overwrite); +int ltt_trace_alloc(const char *trace_name); +int ltt_trace_destroy(const char *trace_name); +int ltt_trace_start(const char *trace_name); +int ltt_trace_stop(const char *trace_name); + +enum ltt_filter_control_msg { + LTT_FILTER_DEFAULT_ACCEPT, + LTT_FILTER_DEFAULT_REJECT +}; + +extern int ltt_filter_control(enum ltt_filter_control_msg msg, + const char *trace_name); + +extern struct dentry *get_filter_root(void); + +void ltt_write_trace_header(struct ltt_trace_struct *trace, + struct ltt_subbuffer_header *header); +extern void ltt_buffer_destroy(struct ltt_channel_struct *ltt_chan); + +void ltt_core_register(int (*function)(u8, void *)); + +void ltt_core_unregister(void); + +void ltt_release_trace(struct kref *kref); +void ltt_release_transport(struct kref *kref); + +extern int ltt_probe_register(struct ltt_available_probe *pdata); +extern int ltt_probe_unregister(struct ltt_available_probe *pdata); +extern int ltt_marker_connect(const char *channel, const char *mname, + const char *pname); +extern int ltt_marker_disconnect(const char *channel, const char *mname, + const char *pname); +extern void ltt_dump_marker_state(struct ltt_trace_struct *trace); + +void ltt_lock_traces(void); +void ltt_unlock_traces(void); + +struct ltt_trace_struct *_ltt_trace_find(const char *trace_name); + +#endif /* _LTT_TRACER_H */ diff --git a/libust/tracercore.c b/libust/tracercore.c new file mode 100644 index 0000000..3b571f1 --- /dev/null +++ b/libust/tracercore.c @@ -0,0 +1,74 @@ +/* + * LTT core in-kernel infrastructure. + * + * Copyright 2006 - Mathieu Desnoyers mathieu.desnoyers@polymtl.ca + * + * Distributed under the GPL license + */ + +//ust// #include +//ust// #include +//ust// #include +//ust// #include +#include "kernelcompat.h" +#include "tracercore.h" + +/* Traces structures */ +struct ltt_traces ltt_traces = { + .setup_head = LIST_HEAD_INIT(ltt_traces.setup_head), + .head = LIST_HEAD_INIT(ltt_traces.head), +}; +//ust// EXPORT_SYMBOL(ltt_traces); + +/* Traces list writer locking */ +static DEFINE_MUTEX(ltt_traces_mutex); + +/* dentry of ltt's root dir */ +//ust// static struct dentry *ltt_root_dentry; +//ust// struct dentry *get_ltt_root(void) +//ust// { +//ust// if (!ltt_root_dentry) { +//ust// ltt_root_dentry = debugfs_create_dir(LTT_ROOT, NULL); +//ust// if (!ltt_root_dentry) +//ust// printk(KERN_ERR "LTT : create ltt root dir failed\n"); +//ust// } +//ust// return ltt_root_dentry; +//ust// } +//ust// EXPORT_SYMBOL_GPL(get_ltt_root); + +void ltt_lock_traces(void) +{ + mutex_lock(<t_traces_mutex); +} +//ust// EXPORT_SYMBOL_GPL(ltt_lock_traces); + +void ltt_unlock_traces(void) +{ + mutex_unlock(<t_traces_mutex); +} +//ust// EXPORT_SYMBOL_GPL(ltt_unlock_traces); + +//ust// DEFINE_PER_CPU(unsigned int, ltt_nesting); +//ust// EXPORT_PER_CPU_SYMBOL(ltt_nesting); +unsigned int ltt_nesting; + +int ltt_run_filter_default(void *trace, uint16_t eID) +{ + return 1; +} + +/* This function pointer is protected by a trace activation check */ +ltt_run_filter_functor ltt_run_filter = ltt_run_filter_default; +//ust// EXPORT_SYMBOL_GPL(ltt_run_filter); + +void ltt_filter_register(ltt_run_filter_functor func) +{ + ltt_run_filter = func; +} +//ust// EXPORT_SYMBOL_GPL(ltt_filter_register); + +void ltt_filter_unregister(void) +{ + ltt_run_filter = ltt_run_filter_default; +} +//ust// EXPORT_SYMBOL_GPL(ltt_filter_unregister); diff --git a/libust/tracercore.h b/libust/tracercore.h new file mode 100644 index 0000000..e5ff00f --- /dev/null +++ b/libust/tracercore.h @@ -0,0 +1,83 @@ +/* + * Copyright (C) 2005,2006 Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) + * + * This contains the core definitions for the Linux Trace Toolkit. + */ + +#ifndef LTT_CORE_H +#define LTT_CORE_H + +#include "list.h" +#include "kernelcompat.h" +//ust// #include + +/* ltt's root dir in debugfs */ +#define LTT_ROOT "ltt" + +/* + * All modifications of ltt_traces must be done by ltt-tracer.c, while holding + * the semaphore. Only reading of this information can be done elsewhere, with + * the RCU mechanism : the preemption must be disabled while reading the + * list. + */ +struct ltt_traces { + struct list_head setup_head; /* Pre-allocated traces list */ + struct list_head head; /* Allocated Traces list */ + unsigned int num_active_traces; /* Number of active traces */ +} ____cacheline_aligned; + +extern struct ltt_traces ltt_traces; + +/* + * get dentry of ltt's root dir + */ +struct dentry *get_ltt_root(void); + +/* Keep track of trap nesting inside LTT */ +//ust// DECLARE_PER_CPU(unsigned int, ltt_nesting); +extern unsigned int ltt_nesting; + +typedef int (*ltt_run_filter_functor)(void *trace, uint16_t eID); +//typedef int (*ltt_run_filter_functor)(void *, __u16); + +extern ltt_run_filter_functor ltt_run_filter; + +extern void ltt_filter_register(ltt_run_filter_functor func); +extern void ltt_filter_unregister(void); + +#if defined(CONFIG_LTT) && defined(CONFIG_LTT_ALIGNMENT) + +/* + * Calculate the offset needed to align the type. + * size_of_type must be non-zero. + */ +static inline unsigned int ltt_align(size_t align_drift, size_t size_of_type) +{ + size_t alignment = min(sizeof(void *), size_of_type); + return (alignment - align_drift) & (alignment - 1); +} +/* Default arch alignment */ +#define LTT_ALIGN + +static inline int ltt_get_alignment(void) +{ + return sizeof(void *); +} + +#else + +static inline unsigned int ltt_align(size_t align_drift, + size_t size_of_type) +{ + return 0; +} + +#define LTT_ALIGN __attribute__((packed)) + +static inline int ltt_get_alignment(void) +{ + return 0; +} +#endif /* defined(CONFIG_LTT) && defined(CONFIG_LTT_ALIGNMENT) */ + +#endif /* LTT_CORE_H */