+2011-12-23 lttng-tools 2.0-pre16
+ * Per-user lttng-sessiond now fully functionnal
+ * Fix nested double usage of hashtable iterators
+ * Support creds passing between lttng and sessiond
+ * Rename sessiond internal "uid" fields to "id"
+ * Create all trace directories and files with client user credentials
+ * Create output directory at session creation command
+ * Only allow the user or group owning the session (or root) to control a session
+ * Add disable kernel tracing option to sessiond
+
+2011-12-13 lttng-tools 2.0-pre15
+ * (MAJOR) User-space tracing support for global domain (lttng-ust 2.0)
+ * UST consumer 32/64 bit support
+ * Multiple bug fixes
+ * Multiple segfault, race and double lock fixes
+ * API change for lttng_list_domains, lttng_start/stop and lttng_destroy
+ * UST loglevel support
+ * ABIs now support 256 char event names
+ * Dependency on libpopt >= 1.13
+ * Dependency on URCU >= 0.6.7
+ * Rename kernel-ctl.c/.h to kernel.c/.h
+ * Change malloc to zmalloc in lttng-sessiond code
+ * RCU hashtable support for all UST data structure
+ * New libconsumer for UST, kernel and one generic
+ * Change "--all" option from lttng command line
+ * Remove function entry option to discourage its use
+
2011-09-30 lttng-tools 2.0-pre14
* Syscall tracing support
* Multiple bug fixes (nothing critical)
ACLOCAL_AMFLAGS = -I config
SUBDIRS = benchmark \
+ common \
liblttng-sessiond-comm \
libkernelctl \
+ liblttng-ht \
liblttng-kconsumer \
liblttng-ustconsumer \
liblttng-consumer \
--- /dev/null
+AM_CPPFLAGS = -I$(top_srcdir)/include
+
+AM_CFLAGS = -fno-strict-aliasing
+
+noinst_LTLIBRARIES = libcommon.la
+
+libcommon_la_SOURCES = runas.c
--- /dev/null
+/*
+ * Copyright (C) 2011 - David Goulet <david.goulet@polymtl.ca>
+ * Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ *
+ * 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; only version 2 of the License.
+ *
+ * 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.
+ */
+
+#define _GNU_SOURCE
+#include <errno.h>
+#include <limits.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/wait.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <sched.h>
+#include <sys/mman.h>
+
+#include <lttngerr.h>
+
+#define CHILD_STACK_SIZE 10485760
+
+struct run_as_data {
+ int (*cmd)(void *data);
+ void *data;
+ uid_t uid;
+ gid_t gid;
+ int retval_pipe;
+};
+
+struct mkdir_data {
+ const char *path;
+ mode_t mode;
+};
+
+struct open_data {
+ const char *path;
+ int flags;
+ mode_t mode;
+};
+
+/*
+ * Create recursively directory using the FULL path.
+ */
+static
+int _mkdir_recursive(void *_data)
+{
+ struct mkdir_data *data = _data;
+ const char *path;
+ char *p, tmp[PATH_MAX];
+ struct stat statbuf;
+ mode_t mode;
+ size_t len;
+ int ret;
+
+ path = data->path;
+ mode = data->mode;
+
+ ret = snprintf(tmp, sizeof(tmp), "%s", path);
+ if (ret < 0) {
+ PERROR("snprintf mkdir");
+ goto error;
+ }
+
+ len = ret;
+ if (tmp[len - 1] == '/') {
+ tmp[len - 1] = 0;
+ }
+
+ for (p = tmp + 1; *p; p++) {
+ if (*p == '/') {
+ *p = 0;
+ ret = stat(tmp, &statbuf);
+ if (ret < 0) {
+ ret = mkdir(tmp, mode);
+ if (ret < 0) {
+ if (!(errno == EEXIST)) {
+ PERROR("mkdir recursive");
+ ret = -errno;
+ goto error;
+ }
+ }
+ }
+ *p = '/';
+ }
+ }
+
+ ret = mkdir(tmp, mode);
+ if (ret < 0) {
+ if (!(errno == EEXIST)) {
+ PERROR("mkdir recursive last piece");
+ ret = -errno;
+ } else {
+ ret = 0;
+ }
+ }
+
+error:
+ return ret;
+}
+
+static
+int _mkdir(void *_data)
+{
+ struct mkdir_data *data = _data;
+ return mkdir(data->path, data->mode);
+}
+
+static
+int _open(void *_data)
+{
+ struct open_data *data = _data;
+ return open(data->path, data->flags, data->mode);
+}
+
+static
+int child_run_as(void *_data)
+{
+ struct run_as_data *data = _data;
+ size_t writelen, writeleft, index;
+ union {
+ int i;
+ char c[sizeof(int)];
+ } sendret;
+ int ret;
+
+ /*
+ * Child: it is safe to drop egid and euid while sharing the
+ * file descriptors with the parent process, since we do not
+ * drop "uid": therefore, the user we are dropping egid/euid to
+ * cannot attach to this process with, e.g. ptrace, nor map this
+ * process memory.
+ */
+ if (data->gid != getegid()) {
+ ret = setegid(data->gid);
+ if (ret < 0) {
+ perror("setegid");
+ return EXIT_FAILURE;
+ }
+ }
+ if (data->uid != geteuid()) {
+ ret = seteuid(data->uid);
+ if (ret < 0) {
+ perror("seteuid");
+ return EXIT_FAILURE;
+ }
+ }
+ /*
+ * Also set umask to 0 for mkdir executable bit.
+ */
+ umask(0);
+ sendret.i = (*data->cmd)(data->data);
+ /* send back return value */
+ writeleft = sizeof(sendret);
+ index = 0;
+ do {
+ writelen = write(data->retval_pipe, &sendret.c[index],
+ writeleft);
+ if (writelen < 0) {
+ perror("write");
+ return EXIT_FAILURE;
+ }
+ writeleft -= writelen;
+ index += writelen;
+ } while (writeleft > 0);
+ return EXIT_SUCCESS;
+}
+
+static
+int run_as(int (*cmd)(void *data), void *data, uid_t uid, gid_t gid)
+{
+ struct run_as_data run_as_data;
+ int ret = 0;
+ int status;
+ pid_t pid;
+ int retval_pipe[2];
+ ssize_t readlen, readleft, index;
+ void *child_stack;
+ union {
+ int i;
+ char c[sizeof(int)];
+ } retval;
+
+ /*
+ * If we are non-root, we can only deal with our own uid.
+ */
+ if (geteuid() != 0) {
+ if (uid != geteuid()) {
+ ERR("Client (%d)/Server (%d) UID mismatch (and sessiond is not root)",
+ uid, geteuid());
+ return -EPERM;
+ }
+ }
+
+ ret = pipe(retval_pipe);
+ if (ret < 0) {
+ perror("pipe");
+ goto end;
+ }
+ run_as_data.data = data;
+ run_as_data.cmd = cmd;
+ run_as_data.uid = uid;
+ run_as_data.gid = gid;
+ run_as_data.retval_pipe = retval_pipe[1]; /* write end */
+ child_stack = mmap(NULL, CHILD_STACK_SIZE,
+ PROT_WRITE | PROT_READ,
+ MAP_PRIVATE | MAP_GROWSDOWN | MAP_ANONYMOUS | MAP_STACK,
+ -1, 0);
+ if (child_stack == MAP_FAILED) {
+ perror("mmap");
+ ret = -ENOMEM;
+ goto close_pipe;
+ }
+ /*
+ * Pointing to the middle of the stack to support architectures
+ * where the stack grows up (HPPA).
+ */
+ pid = clone(child_run_as, child_stack + (CHILD_STACK_SIZE / 2),
+ CLONE_FILES | SIGCHLD | CLONE_VM,
+ &run_as_data, NULL);
+ if (pid < 0) {
+ perror("clone");
+ ret = pid;
+ goto unmap_stack;
+ }
+ /* receive return value */
+ readleft = sizeof(retval);
+ index = 0;
+ do {
+ readlen = read(retval_pipe[0], &retval.c[index], readleft);
+ if (readlen < 0) {
+ perror("read");
+ ret = -1;
+ break;
+ }
+ readleft -= readlen;
+ index += readlen;
+ } while (readleft > 0);
+
+ /*
+ * Parent: wait for child to return, in which case the
+ * shared memory map will have been created.
+ */
+ pid = waitpid(pid, &status, 0);
+ if (pid < 0 || !WIFEXITED(status) || WEXITSTATUS(status) != 0) {
+ perror("wait");
+ ret = -1;
+ }
+unmap_stack:
+ ret = munmap(child_stack, CHILD_STACK_SIZE);
+ if (ret < 0) {
+ perror("munmap");
+ }
+close_pipe:
+ close(retval_pipe[0]);
+ close(retval_pipe[1]);
+end:
+ return retval.i;
+}
+
+int mkdir_recursive_run_as(const char *path, mode_t mode, uid_t uid, gid_t gid)
+{
+ struct mkdir_data data;
+
+ DBG3("mkdir() recursive %s with mode %d for uid %d and gid %d",
+ path, mode, uid, gid);
+ data.path = path;
+ data.mode = mode;
+ return run_as(_mkdir_recursive, &data, uid, gid);
+}
+
+int mkdir_run_as(const char *path, mode_t mode, uid_t uid, gid_t gid)
+{
+ struct mkdir_data data;
+
+ DBG3("mkdir() %s with mode %d for uid %d and gid %d",
+ path, mode, uid, gid);
+ data.path = path;
+ data.mode = mode;
+ return run_as(_mkdir, &data, uid, gid);
+}
+
+/*
+ * Note: open_run_as is currently not working. We'd need to pass the fd
+ * opened in the child to the parent.
+ */
+int open_run_as(const char *path, int flags, mode_t mode, uid_t uid, gid_t gid)
+{
+ struct open_data data;
+
+ DBG3("open() %s with flags %X mode %d for uid %d and gid %d",
+ path, flags, mode, uid, gid);
+ data.path = path;
+ data.flags = flags;
+ data.mode = mode;
+ return run_as(_open, &data, uid, gid);
+}
-AC_INIT([lttng-tools],[2.0-pre14],[david.goulet@polymtl.ca],[],[http://lttng.org])
+AC_INIT([lttng-tools],[2.0-pre16],[david.goulet@polymtl.ca],[],[http://lttng.org])
AC_CONFIG_AUX_DIR([config])
AC_CANONICAL_TARGET
AC_CANONICAL_HOST
)
# URCU library version needed or newer
-liburcu_version=">= 0.6.6"
+liburcu_version=">= 0.6.7"
# Check liburcu needed function calls
AC_CHECK_DECL([cds_list_add], [],
Makefile
include/Makefile
benchmark/Makefile
+ common/Makefile
libkernelctl/Makefile
liblttng-consumer/Makefile
liblttng-kconsumer/Makefile
liblttng-ustconsumer/Makefile
liblttngctl/Makefile
liblttng-sessiond-comm/Makefile
+ liblttng-ht/Makefile
lttng-consumerd/Makefile
lttng-sessiond/Makefile
lttng/Makefile
+
+******** DEPRECATED **********
+* Kept for historic purposes *
+******** DEPRECATED **********
+
lttng-tools command line interface
(Note: as of June 8th, 2011, this document is at [RFC] stage.)
NOTES:
--------------
-2011-07-21 : User-space tracer is not released. Tracing in user-space at this
-point is not possible with lttng-tools.
+
+2011-12-12: For user-space tracing, only the global UST domain ("-u" alone) is
+supported meaning that if you enable a tracepoint for user-space it will be
+enabled for all applications for the current tracing session you are working
+on.
QUICKSTART
--------------
See the README file for installation procedure or use the various Linux
distribution packages.
-In order to trace the kernel, you'll need the lttng-modules >= 2.0 compiled and
+In order to trace the kernel, you'll need the lttng-modules 2.0 compiled and
installed. See http://lttng.org/lttng2.0 for more instructions for that part.
-For user-space tracing, you'll need an instrumented application, please see
-http://lttng.org/ust.
+For user-space tracing, you'll need an instrumented application with lttng-ust
+2.0.
lttng-tools provide a session daemon (lttng-sessiond) that acts as a tracing
registry. To trace any instrumented applications or the kernel, a registered
tracing session is needed beforehand. To interact with the session daemon and a
-tracing session, you should use the lttng command line UI (lttng).
+tracing session, you should use the lttng command line UI (lttng). It is also
+possible to use the liblttngctl library for tracing control (lttng.h).
Here is a list of some powerful features the LTTng 2.0 kernel tracer offers:
* Perf counter support
* Tracepoint support
-The next sections explain how to do tracing :)
+And for the LTTng UST 2.0 tracer:
+
+ * Applications registration
+ * Automatic tracepoints activation upon app. registration
+ * Context information support
+ * Safe buffers after application crash
+ * Per-user tracing (root access *not* mandatory)
+
+The next sections explains how to do tracing :)
Kernel Tracing
--------------
-You can start the session daemon by invoking the command "lttng-sessiond",
-or let the lttng command line tool do it for you. The session daemon
-loads the LTTng tracer modules for you if those modules can be found on
-your system. If they are not found, the kernel tracing feature will be
-unavailable.
+You can start the session daemon by invoking the command "lttng-sessiond", or
+let the lttng command line tool do it for you. The session daemon loads the
+LTTng tracer modules for you if those modules can be found on your system. If
+they are not found, the kernel tracing feature will be unavailable.
List available kernel events:
# lttng list -k
-1) Create a tracing session. A .lttngrc will be created in $HOME containing
-the session name (here 'mysession') you are working on.
+1) Create a tracing session. The .lttng directory will be created with .lttngrc
+file in $HOME containing the session name (here 'mysession') you are working
+on.
# lttng create mysession
# lttng add-context -k -e sched_switch -t pid
At this point, you will have to look at 'lttng add-context --help' for all
-possible context type which are integer values.
+possible context type.
You can on the same line activate multiple context:
Tracing is in progress at this point and traces will be written in
$HOME/lttng-traces/mysession-<date>-<time>
+NOTE: It will start tracing for *all* domain(s).
+
9) Stop tracing:
# lttng stop
-Note: At this point, you can restart the trace (lttng start), enable/disable
+NOTE: At this point, you can restart the trace (lttng start), enable/disable
events or just go take a break and come back 3 days later to start it again :).
+You can also read the trace since the buffers are flushed on stop command.
10) Destroy your session after you are done with tracing
User-space Tracing
--------------
-User-space tracer 2.0 not released at this point. For the 0.x versions,
-you need to use 'ustctl' to control user-space tracing.
+Like kernel tracing, you can start the session daemon by invoking the command
+"lttng-sessiond", or let the lttng command line tool do it for you.
+
+NOTE: You do *not* need root credentials in order to tracer user-space
+applications. However, if you run the session daemon under non-root user
+rights, only applications of that user will be traced.
+
+So, after instrumenting you applications with LTTng-ust 2.0
+(http://lttng.org/lttng2.0), upon startup, it will automatically register to
+the session daemon. If there is none running, it will simply wait on a seperate
+thread for a session daemon to appear and then register.
+
+Start your instrumented application at any time but at least before starting
+tracing :).
+
+List available registered applications:
+
+$ lttng list -u
+
+1) Create a tracing session. The .lttng directory will be created with a
+.lttngrc file in $HOME containing the session name (here 'mysession') you are
+working on.
+
+$ lttng create mysession
+
+If you have multiple sessions, you can change the current session by using:
+
+$ lttng set-session myothersession
+
+2) Enable all tracepoints for the global UST domain ("-u" alone).
+
+$ lttng enable-event -a -u
+
+or enable a single tracepoint event.
+
+$ lttng enable-event ust_tests_hello:tptest -u
+
+3) This is also a new feature which allows you to add context information to an
+event. For example, you can add the PID along with the event information:
+
+$ lttng add-context -t pid -e ust_tests_hello:tptest -u
+
+At this point, you will have to look at 'lttng add-context --help' for all
+possible context type.
+
+You can on the same line activate multiple context:
+
+$ lttng add-context -u -e ust_tests_hello:tptest -t pid -t nice -t tid
+
+4) Start tracing:
+
+$ lttng start
+
+Tracing is in progress at this point and traces will be written in
+$HOME/lttng-traces/mysession-<date>-<time>/ust/<procname>-<pid>-<date>-<time>
+
+NOTE: It will start tracing for *all* domain(s).
+
+5) Stop tracing:
+
+$ lttng stop
+
+NOTE: At this point, you can restart the trace (lttng start), enable/disable
+events or just go take a break and come back 3 days later to start it again :).
+You can also read the trace since the buffers are flushed on stop command.
+
+6) Destroy your session after you are done with tracing
+
+$ lttng destroy
+
+See "Reading a trace" section below to read you trace(s).
+
Reading a trace
--------------
The tool "Babeltrace" can be used to dump your binary trace into a
-human-readable text format. Please see
-http://www.efficios.com/babeltrace and git tree
-http://git.efficios.com/?p=babeltrace.git
+human-readable text format. Please see http://www.efficios.com/babeltrace and
+git tree http://git.efficios.com/?p=babeltrace.git
# babeltrace $HOME/lttng-traces/mysession-<date>-<time> | less
Voilà !
-Please report any bugs/comments on our mailing list
-(ltt-dev@lists.casi.polymtl.ca) or you can go on our IRC channel at
-irc.oftc.net, channel #lttng
+Please report any bugs/comments on our mailing list (lttng-dev@lists.lttng.org)
+or you can go on our IRC channel at irc.oftc.net, channel #lttng
+++ /dev/null
-/*
- * Copyright (C) - Bob Jenkins, May 2006, Public Domain.
- * Copyright (C) 2011 - David Goulet <david.goulet@polymtl.ca>
- * Copyright (C) 2011 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
- *
- * These are functions for producing 32-bit hashes for hash table lookup.
- * hashword(), hashlittle(), hashlittle2(), hashbig(), mix(), and final() are
- * externally useful functions. Routines to test the hash are included if
- * SELF_TEST is defined. You can use this free for any purpose. It's in the
- * public domain. It has no warranty.
- *
- * You probably want to use hashlittle(). hashlittle() and hashbig() hash byte
- * arrays. hashlittle() is is faster than hashbig() on little-endian machines.
- * Intel and AMD are little-endian machines. On second thought, you probably
- * want hashlittle2(), which is identical to hashlittle() except it returns two
- * 32-bit hashes for the price of one. You could implement hashbig2() if you
- * wanted but I haven't bothered here.
- *
- * If you want to find a hash of, say, exactly 7 integers, do
- * a = i1; b = i2; c = i3;
- * mix(a,b,c);
- * a += i4; b += i5; c += i6;
- * mix(a,b,c);
- * a += i7;
- * final(a,b,c);
- * then use c as the hash value. If you have a variable length array of
- * 4-byte integers to hash, use hashword(). If you have a byte array (like
- * a character string), use hashlittle(). If you have several byte arrays, or
- * a mix of things, see the comments above hashlittle().
- *
- * Why is this so big? I read 12 bytes at a time into 3 4-byte integers, then
- * mix those integers. This is fast (you can do a lot more thorough mixing
- * with 12*3 instructions on 3 integers than you can with 3 instructions on 1
- * byte), but shoehorning those bytes into integers efficiently is messy.
- */
-
-#include <stdio.h> /* defines printf for tests */
-#include <time.h> /* defines time_t for timings in the test */
-#include <stdint.h> /* defines uint32_t etc */
-#include <sys/param.h> /* attempt to define endianness */
-#include <endian.h> /* attempt to define endianness */
-#include <string.h>
-#include <assert.h>
-#include <urcu/compiler.h>
-
-/*
- * My best guess at if you are big-endian or little-endian. This may
- * need adjustment.
- */
-#if (defined(__BYTE_ORDER) && defined(__LITTLE_ENDIAN) && \
- __BYTE_ORDER == __LITTLE_ENDIAN) || \
- (defined(i386) || defined(__i386__) || defined(__i486__) || \
- defined(__i586__) || defined(__i686__) || defined(vax) || defined(MIPSEL))
-# define HASH_LITTLE_ENDIAN 1
-# define HASH_BIG_ENDIAN 0
-#elif (defined(__BYTE_ORDER) && defined(__BIG_ENDIAN) && \
- __BYTE_ORDER == __BIG_ENDIAN) || \
- (defined(sparc) || defined(POWERPC) || defined(mc68000) || defined(sel))
-# define HASH_LITTLE_ENDIAN 0
-# define HASH_BIG_ENDIAN 1
-#else
-# define HASH_LITTLE_ENDIAN 0
-# define HASH_BIG_ENDIAN 0
-#endif
-
-#define hashsize(n) ((uint32_t)1<<(n))
-#define hashmask(n) (hashsize(n)-1)
-#define rot(x,k) (((x)<<(k)) | ((x)>>(32-(k))))
-
-/*
- * mix -- mix 3 32-bit values reversibly.
- *
- * This is reversible, so any information in (a,b,c) before mix() is
- * still in (a,b,c) after mix().
- *
- * If four pairs of (a,b,c) inputs are run through mix(), or through
- * mix() in reverse, there are at least 32 bits of the output that
- * are sometimes the same for one pair and different for another pair.
- * This was tested for:
- * * pairs that differed by one bit, by two bits, in any combination
- * of top bits of (a,b,c), or in any combination of bottom bits of
- * (a,b,c).
- * * "differ" is defined as +, -, ^, or ~^. For + and -, I transformed
- * the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
- * is commonly produced by subtraction) look like a single 1-bit
- * difference.
- * * the base values were pseudorandom, all zero but one bit set, or
- * all zero plus a counter that starts at zero.
- *
- * Some k values for my "a-=c; a^=rot(c,k); c+=b;" arrangement that
- * satisfy this are
- * 4 6 8 16 19 4
- * 9 15 3 18 27 15
- * 14 9 3 7 17 3
- * Well, "9 15 3 18 27 15" didn't quite get 32 bits diffing
- * for "differ" defined as + with a one-bit base and a two-bit delta. I
- * used http://burtleburtle.net/bob/hash/avalanche.html to choose
- * the operations, constants, and arrangements of the variables.
- *
- * This does not achieve avalanche. There are input bits of (a,b,c)
- * that fail to affect some output bits of (a,b,c), especially of a. The
- * most thoroughly mixed value is c, but it doesn't really even achieve
- * avalanche in c.
- *
- * This allows some parallelism. Read-after-writes are good at doubling
- * the number of bits affected, so the goal of mixing pulls in the opposite
- * direction as the goal of parallelism. I did what I could. Rotates
- * seem to cost as much as shifts on every machine I could lay my hands
- * on, and rotates are much kinder to the top and bottom bits, so I used
- * rotates.
- */
-#define mix(a,b,c) \
-{ \
- a -= c; a ^= rot(c, 4); c += b; \
- b -= a; b ^= rot(a, 6); a += c; \
- c -= b; c ^= rot(b, 8); b += a; \
- a -= c; a ^= rot(c,16); c += b; \
- b -= a; b ^= rot(a,19); a += c; \
- c -= b; c ^= rot(b, 4); b += a; \
-}
-
-/*
- * final -- final mixing of 3 32-bit values (a,b,c) into c
- *
- * Pairs of (a,b,c) values differing in only a few bits will usually
- * produce values of c that look totally different. This was tested for
- * * pairs that differed by one bit, by two bits, in any combination
- * of top bits of (a,b,c), or in any combination of bottom bits of
- * (a,b,c).
- * * "differ" is defined as +, -, ^, or ~^. For + and -, I transformed
- * the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
- * is commonly produced by subtraction) look like a single 1-bit
- * difference.
- * * the base values were pseudorandom, all zero but one bit set, or
- * all zero plus a counter that starts at zero.
- *
- * These constants passed:
- * 14 11 25 16 4 14 24
- * 12 14 25 16 4 14 24
- * and these came close:
- * 4 8 15 26 3 22 24
- * 10 8 15 26 3 22 24
- * 11 8 15 26 3 22 24
- */
-#define final(a,b,c) \
-{ \
- c ^= b; c -= rot(b,14); \
- a ^= c; a -= rot(c,11); \
- b ^= a; b -= rot(a,25); \
- c ^= b; c -= rot(b,16); \
- a ^= c; a -= rot(c,4); \
- b ^= a; b -= rot(a,14); \
- c ^= b; c -= rot(b,24); \
-}
-
-static __attribute__((unused))
-uint32_t hashword(
- const uint32_t *k, /* the key, an array of uint32_t values */
- size_t length, /* the length of the key, in uint32_ts */
- uint32_t initval) /* the previous hash, or an arbitrary value */
-{
- uint32_t a, b, c;
-
- /* Set up the internal state */
- a = b = c = 0xdeadbeef + (((uint32_t) length) << 2) + initval;
-
- /*----------------------------------------- handle most of the key */
- while (length > 3) {
- a += k[0];
- b += k[1];
- c += k[2];
- mix(a, b, c);
- length -= 3;
- k += 3;
- }
-
- /*----------------------------------- handle the last 3 uint32_t's */
- switch (length) { /* all the case statements fall through */
- case 3: c += k[2];
- case 2: b += k[1];
- case 1: a += k[0];
- final(a, b, c);
- case 0: /* case 0: nothing left to add */
- break;
- }
- /*---------------------------------------------- report the result */
- return c;
-}
-
-
-/*
- * hashword2() -- same as hashword(), but take two seeds and return two 32-bit
- * values. pc and pb must both be nonnull, and *pc and *pb must both be
- * initialized with seeds. If you pass in (*pb)==0, the output (*pc) will be
- * the same as the return value from hashword().
- */
-static __attribute__((unused))
-void hashword2(const uint32_t *k, size_t length,
- uint32_t *pc, uint32_t *pb)
-{
- uint32_t a, b, c;
-
- /* Set up the internal state */
- a = b = c = 0xdeadbeef + ((uint32_t) (length << 2)) + *pc;
- c += *pb;
-
- while (length > 3) {
- a += k[0];
- b += k[1];
- c += k[2];
- mix(a, b, c);
- length -= 3;
- k += 3;
- }
-
- switch (length) {
- case 3 :
- c += k[2];
- case 2 :
- b += k[1];
- case 1 :
- a += k[0];
- final(a, b, c);
- case 0: /* case 0: nothing left to add */
- break;
- }
-
- *pc = c;
- *pb = b;
-}
-
-/*
- * hashlittle() -- hash a variable-length key into a 32-bit value
- * k : the key (the unaligned variable-length array of bytes)
- * length : the length of the key, counting by bytes
- * initval : can be any 4-byte value
- * Returns a 32-bit value. Every bit of the key affects every bit of
- * the return value. Two keys differing by one or two bits will have
- * totally different hash values.
- *
- * The best hash table sizes are powers of 2. There is no need to do
- * mod a prime (mod is sooo slow!). If you need less than 32 bits,
- * use a bitmask. For example, if you need only 10 bits, do
- * h = (h & hashmask(10));
- * In which case, the hash table should have hashsize(10) elements.
- *
- * If you are hashing n strings (uint8_t **)k, do it like this:
- * for (i=0, h=0; i<n; ++i) h = hashlittle( k[i], len[i], h);
- *
- * By Bob Jenkins, 2006. bob_jenkins@burtleburtle.net. You may use this
- * code any way you wish, private, educational, or commercial. It's free.
- *
- * Use for hash table lookup, or anything where one collision in 2^^32 is
- * acceptable. Do NOT use for cryptographic purposes.
- */
-
-static uint32_t hashlittle(const void *key, size_t length, uint32_t initval)
-{
- uint32_t a,b,c;
- union {
- const void *ptr;
- size_t i;
- } u; /* needed for Mac Powerbook G4 */
-
- /* Set up the internal state */
- a = b = c = 0xdeadbeef + ((uint32_t)length) + initval;
-
- u.ptr = key;
- if (HASH_LITTLE_ENDIAN && ((u.i & 0x3) == 0)) {
- const uint32_t *k = (const uint32_t *)key; /* read 32-bit chunks */
-
- /*------ all but last block: aligned reads and affect 32 bits of (a,b,c) */
- while (length > 12) {
- a += k[0];
- b += k[1];
- c += k[2];
- mix(a,b,c);
- length -= 12;
- k += 3;
- }
-
- /*
- * "k[2]&0xffffff" actually reads beyond the end of the string, but
- * then masks off the part it's not allowed to read. Because the
- * string is aligned, the masked-off tail is in the same word as the
- * rest of the string. Every machine with memory protection I've seen
- * does it on word boundaries, so is OK with this. But VALGRIND will
- * still catch it and complain. The masking trick does make the hash
- * noticably faster for short strings (like English words).
- */
-#ifndef VALGRIND
-
- switch (length) {
- case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
- case 11: c+=k[2]&0xffffff; b+=k[1]; a+=k[0]; break;
- case 10: c+=k[2]&0xffff; b+=k[1]; a+=k[0]; break;
- case 9 : c+=k[2]&0xff; b+=k[1]; a+=k[0]; break;
- case 8 : b+=k[1]; a+=k[0]; break;
- case 7 : b+=k[1]&0xffffff; a+=k[0]; break;
- case 6 : b+=k[1]&0xffff; a+=k[0]; break;
- case 5 : b+=k[1]&0xff; a+=k[0]; break;
- case 4 : a+=k[0]; break;
- case 3 : a+=k[0]&0xffffff; break;
- case 2 : a+=k[0]&0xffff; break;
- case 1 : a+=k[0]&0xff; break;
- case 0 : return c; /* zero length strings require no mixing */
- }
-#else /* make valgrind happy */
- const uint8_t *k8;
-
- k8 = (const uint8_t *)k;
- switch (length) {
- case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
- case 11: c+=((uint32_t)k8[10])<<16; /* fall through */
- case 10: c+=((uint32_t)k8[9])<<8; /* fall through */
- case 9 : c+=k8[8]; /* fall through */
- case 8 : b+=k[1]; a+=k[0]; break;
- case 7 : b+=((uint32_t)k8[6])<<16; /* fall through */
- case 6 : b+=((uint32_t)k8[5])<<8; /* fall through */
- case 5 : b+=k8[4]; /* fall through */
- case 4 : a+=k[0]; break;
- case 3 : a+=((uint32_t)k8[2])<<16; /* fall through */
- case 2 : a+=((uint32_t)k8[1])<<8; /* fall through */
- case 1 : a+=k8[0]; break;
- case 0 : return c;
- }
-#endif /* !valgrind */
- } else if (HASH_LITTLE_ENDIAN && ((u.i & 0x1) == 0)) {
- const uint16_t *k = (const uint16_t *)key; /* read 16-bit chunks */
- const uint8_t *k8;
-
- /*--------------- all but last block: aligned reads and different mixing */
- while (length > 12) {
- a += k[0] + (((uint32_t)k[1])<<16);
- b += k[2] + (((uint32_t)k[3])<<16);
- c += k[4] + (((uint32_t)k[5])<<16);
- mix(a,b,c);
- length -= 12;
- k += 6;
- }
-
- k8 = (const uint8_t *)k;
- switch (length) {
- case 12:
- c+=k[4]+(((uint32_t)k[5])<<16);
- b+=k[2]+(((uint32_t)k[3])<<16);
- a+=k[0]+(((uint32_t)k[1])<<16);
- break;
- case 11:
- c+=((uint32_t)k8[10])<<16; /* fall through */
- case 10:
- c+=k[4];
- b+=k[2]+(((uint32_t)k[3])<<16);
- a+=k[0]+(((uint32_t)k[1])<<16);
- break;
- case 9:
- c+=k8[8]; /* fall through */
- case 8:
- b+=k[2]+(((uint32_t)k[3])<<16);
- a+=k[0]+(((uint32_t)k[1])<<16);
- break;
- case 7:
- b+=((uint32_t)k8[6])<<16; /* fall through */
- case 6:
- b+=k[2];
- a+=k[0]+(((uint32_t)k[1])<<16);
- break;
- case 5:
- b+=k8[4]; /* fall through */
- case 4:
- a+=k[0]+(((uint32_t)k[1])<<16);
- break;
- case 3:
- a+=((uint32_t)k8[2])<<16; /* fall through */
- case 2:
- a+=k[0];
- break;
- case 1:
- a+=k8[0];
- break;
- case 0:
- return c; /* zero length requires no mixing */
- }
-
- } else { /* need to read the key one byte at a time */
- const uint8_t *k = (const uint8_t *)key;
-
- while (length > 12) {
- a += k[0];
- a += ((uint32_t)k[1])<<8;
- a += ((uint32_t)k[2])<<16;
- a += ((uint32_t)k[3])<<24;
- b += k[4];
- b += ((uint32_t)k[5])<<8;
- b += ((uint32_t)k[6])<<16;
- b += ((uint32_t)k[7])<<24;
- c += k[8];
- c += ((uint32_t)k[9])<<8;
- c += ((uint32_t)k[10])<<16;
- c += ((uint32_t)k[11])<<24;
- mix(a,b,c);
- length -= 12;
- k += 12;
- }
-
- switch(length) { /* all the case statements fall through */
- case 12: c+=((uint32_t)k[11])<<24;
- case 11: c+=((uint32_t)k[10])<<16;
- case 10: c+=((uint32_t)k[9])<<8;
- case 9: c+=k[8];
- case 8: b+=((uint32_t)k[7])<<24;
- case 7: b+=((uint32_t)k[6])<<16;
- case 6: b+=((uint32_t)k[5])<<8;
- case 5: b+=k[4];
- case 4: a+=((uint32_t)k[3])<<24;
- case 3: a+=((uint32_t)k[2])<<16;
- case 2: a+=((uint32_t)k[1])<<8;
- case 1:
- a+=k[0];
- break;
- case 0:
- return c;
- }
- }
-
- final(a,b,c);
- return c;
-}
-
-#if (CAA_BITS_PER_LONG == 64)
-/*
- * Hash function for number value.
- */
-unsigned long hash_key(void *_key, size_t length, unsigned long seed)
-{
- union {
- uint64_t v64;
- uint32_t v32[2];
- } v;
- union {
- uint64_t v64;
- uint32_t v32[2];
- } key;
-
- assert(length == sizeof(unsigned long));
- v.v64 = (uint64_t) seed;
- key.v64 = (uint64_t) _key;
- hashword2(key.v32, 2, &v.v32[0], &v.v32[1]);
- return v.v64;
-}
-#else
-/*
- * Hash function for number value.
- */
-unsigned long hash_key(void *_key, size_t length, unsigned long seed)
-{
- uint32_t key = (uint32_t) _key;
-
- assert(length == sizeof(uint32_t));
- return hashword(&key, 1, seed);
-}
-#endif
-
-/*
- * Hash function for string.
- */
-unsigned long hash_key_str(void *key, size_t length, unsigned long seed)
-{
- return hashlittle(key, length, seed);
-}
-
-/*
- * Hash function compare for number value.
- */
-unsigned long hash_compare_key(void *key1, size_t key1_len,
- void *key2, size_t key2_len)
-{
- if (key1_len != key2_len) {
- return -1;
- }
-
- if (key1 == key2) {
- return 0;
- }
-
- return 1;
-}
-
-/*
- * Hash compare function for string.
- */
-unsigned long hash_compare_key_str(void *key1, size_t key1_len,
- void *key2, size_t key2_len)
-{
- if (key1_len != key2_len) {
- return -1;
- }
-
- if (strncmp(key1, key2, key1_len) == 0) {
- return 0;
- }
-
- return 1;
-}
+++ /dev/null
-/*
- * Copyright (C) 2011 - David Goulet <david.goulet@polymtl.ca>
- *
- * 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; only version 2 of the License.
- *
- * 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.
- */
-
-#ifndef _LTT_HASH_H
-#define _LTT_HASH_H
-
-unsigned long hash_key(void *_key, size_t length, unsigned long seed);
-
-unsigned long hash_key_str(void *_key, size_t length, unsigned long seed);
-
-unsigned long hash_compare_key(void *key1, size_t key1_len,
- void *key2, size_t key2_len);
-
-unsigned long hash_compare_key_str(void *key1, size_t key1_len,
- void *key2, size_t key2_len);
-
-#endif /* _LTT_HASH_H */
+++ /dev/null
-/*
- * rculfhash.c
- *
- * Userspace RCU library - Lock-Free Resizable RCU Hash Table
- *
- * Copyright 2010-2011 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
- *
- * This library is free software; you can redistribute it and/or
- * modify it under the terms of the GNU Lesser General Public
- * License as published by the Free Software Foundation; either
- * version 2.1 of the License, or (at your option) any later version.
- *
- * This library 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
- * Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public
- * License along with this library; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- */
-
-/*
- * Based on the following articles:
- * - Ori Shalev and Nir Shavit. Split-ordered lists: Lock-free
- * extensible hash tables. J. ACM 53, 3 (May 2006), 379-405.
- * - Michael, M. M. High performance dynamic lock-free hash tables
- * and list-based sets. In Proceedings of the fourteenth annual ACM
- * symposium on Parallel algorithms and architectures, ACM Press,
- * (2002), 73-82.
- *
- * Some specificities of this Lock-Free Resizable RCU Hash Table
- * implementation:
- *
- * - RCU read-side critical section allows readers to perform hash
- * table lookups and use the returned objects safely by delaying
- * memory reclaim of a grace period.
- * - Add and remove operations are lock-free, and do not need to
- * allocate memory. They need to be executed within RCU read-side
- * critical section to ensure the objects they read are valid and to
- * deal with the cmpxchg ABA problem.
- * - add and add_unique operations are supported. add_unique checks if
- * the node key already exists in the hash table. It ensures no key
- * duplicata exists.
- * - The resize operation executes concurrently with add/remove/lookup.
- * - Hash table nodes are contained within a split-ordered list. This
- * list is ordered by incrementing reversed-bits-hash value.
- * - An index of dummy nodes is kept. These dummy nodes are the hash
- * table "buckets", and they are also chained together in the
- * split-ordered list, which allows recursive expansion.
- * - The resize operation for small tables only allows expanding the hash table.
- * It is triggered automatically by detecting long chains in the add
- * operation.
- * - The resize operation for larger tables (and available through an
- * API) allows both expanding and shrinking the hash table.
- * - Split-counters are used to keep track of the number of
- * nodes within the hash table for automatic resize triggering.
- * - Resize operation initiated by long chain detection is executed by a
- * call_rcu thread, which keeps lock-freedom of add and remove.
- * - Resize operations are protected by a mutex.
- * - The removal operation is split in two parts: first, a "removed"
- * flag is set in the next pointer within the node to remove. Then,
- * a "garbage collection" is performed in the bucket containing the
- * removed node (from the start of the bucket up to the removed node).
- * All encountered nodes with "removed" flag set in their next
- * pointers are removed from the linked-list. If the cmpxchg used for
- * removal fails (due to concurrent garbage-collection or concurrent
- * add), we retry from the beginning of the bucket. This ensures that
- * the node with "removed" flag set is removed from the hash table
- * (not visible to lookups anymore) before the RCU read-side critical
- * section held across removal ends. Furthermore, this ensures that
- * the node with "removed" flag set is removed from the linked-list
- * before its memory is reclaimed. Only the thread which removal
- * successfully set the "removed" flag (with a cmpxchg) into a node's
- * next pointer is considered to have succeeded its removal (and thus
- * owns the node to reclaim). Because we garbage-collect starting from
- * an invariant node (the start-of-bucket dummy node) up to the
- * "removed" node (or find a reverse-hash that is higher), we are sure
- * that a successful traversal of the chain leads to a chain that is
- * present in the linked-list (the start node is never removed) and
- * that is does not contain the "removed" node anymore, even if
- * concurrent delete/add operations are changing the structure of the
- * list concurrently.
- * - The add operation performs gargage collection of buckets if it
- * encounters nodes with removed flag set in the bucket where it wants
- * to add its new node. This ensures lock-freedom of add operation by
- * helping the remover unlink nodes from the list rather than to wait
- * for it do to so.
- * - A RCU "order table" indexed by log2(hash index) is copied and
- * expanded by the resize operation. This order table allows finding
- * the "dummy node" tables.
- * - There is one dummy node table per hash index order. The size of
- * each dummy node table is half the number of hashes contained in
- * this order (except for order 0).
- * - synchronzie_rcu is used to garbage-collect the old dummy node table.
- * - The per-order dummy node tables contain a compact version of the
- * hash table nodes. These tables are invariant after they are
- * populated into the hash table.
- *
- * Dummy node tables:
- *
- * hash table hash table the last all dummy node tables
- * order size dummy node 0 1 2 3 4 5 6(index)
- * table size
- * 0 1 1 1
- * 1 2 1 1 1
- * 2 4 2 1 1 2
- * 3 8 4 1 1 2 4
- * 4 16 8 1 1 2 4 8
- * 5 32 16 1 1 2 4 8 16
- * 6 64 32 1 1 2 4 8 16 32
- *
- * When growing/shrinking, we only focus on the last dummy node table
- * which size is (!order ? 1 : (1 << (order -1))).
- *
- * Example for growing/shrinking:
- * grow hash table from order 5 to 6: init the index=6 dummy node table
- * shrink hash table from order 6 to 5: fini the index=6 dummy node table
- *
- * A bit of ascii art explanation:
- *
- * Order index is the off-by-one compare to the actual power of 2 because
- * we use index 0 to deal with the 0 special-case.
- *
- * This shows the nodes for a small table ordered by reversed bits:
- *
- * bits reverse
- * 0 000 000
- * 4 100 001
- * 2 010 010
- * 6 110 011
- * 1 001 100
- * 5 101 101
- * 3 011 110
- * 7 111 111
- *
- * This shows the nodes in order of non-reversed bits, linked by
- * reversed-bit order.
- *
- * order bits reverse
- * 0 0 000 000
- * 1 | 1 001 100 <-
- * 2 | | 2 010 010 <- |
- * | | | 3 011 110 | <- |
- * 3 -> | | | 4 100 001 | |
- * -> | | 5 101 101 |
- * -> | 6 110 011
- * -> 7 111 111
- */
-
-#define _LGPL_SOURCE
-#include <stdlib.h>
-#include <errno.h>
-#include <assert.h>
-#include <stdio.h>
-#include <stdint.h>
-#include <string.h>
-
-#include "config.h"
-#include <urcu.h>
-#include <urcu-call-rcu.h>
-#include <urcu/arch.h>
-#include <urcu/uatomic.h>
-#include <urcu/compiler.h>
-#include <stdio.h>
-#include <pthread.h>
-
-#include "rculfhash.h"
-
-#ifdef DEBUG
-#define dbg_printf(fmt, args...) printf("[debug rculfhash] " fmt, ## args)
-#else
-#define dbg_printf(fmt, args...)
-#endif
-
-/*
- * Split-counters lazily update the global counter each 1024
- * addition/removal. It automatically keeps track of resize required.
- * We use the bucket length as indicator for need to expand for small
- * tables and machines lacking per-cpu data suppport.
- */
-#define COUNT_COMMIT_ORDER 10
-#define DEFAULT_SPLIT_COUNT_MASK 0xFUL
-#define CHAIN_LEN_TARGET 1
-#define CHAIN_LEN_RESIZE_THRESHOLD 3
-
-/*
- * Define the minimum table size.
- */
-#define MIN_TABLE_SIZE 1
-
-#if (CAA_BITS_PER_LONG == 32)
-#define MAX_TABLE_ORDER 32
-#else
-#define MAX_TABLE_ORDER 64
-#endif
-
-/*
- * Minimum number of dummy nodes to touch per thread to parallelize grow/shrink.
- */
-#define MIN_PARTITION_PER_THREAD_ORDER 12
-#define MIN_PARTITION_PER_THREAD (1UL << MIN_PARTITION_PER_THREAD_ORDER)
-
-#ifndef min
-#define min(a, b) ((a) < (b) ? (a) : (b))
-#endif
-
-#ifndef max
-#define max(a, b) ((a) > (b) ? (a) : (b))
-#endif
-
-/*
- * The removed flag needs to be updated atomically with the pointer.
- * It indicates that no node must attach to the node scheduled for
- * removal, and that node garbage collection must be performed.
- * The dummy flag does not require to be updated atomically with the
- * pointer, but it is added as a pointer low bit flag to save space.
- */
-#define REMOVED_FLAG (1UL << 0)
-#define DUMMY_FLAG (1UL << 1)
-#define FLAGS_MASK ((1UL << 2) - 1)
-
-/* Value of the end pointer. Should not interact with flags. */
-#define END_VALUE NULL
-
-/*
- * ht_items_count: Split-counters counting the number of node addition
- * and removal in the table. Only used if the CDS_LFHT_ACCOUNTING flag
- * is set at hash table creation.
- *
- * These are free-running counters, never reset to zero. They count the
- * number of add/remove, and trigger every (1 << COUNT_COMMIT_ORDER)
- * operations to update the global counter. We choose a power-of-2 value
- * for the trigger to deal with 32 or 64-bit overflow of the counter.
- */
-struct ht_items_count {
- unsigned long add, del;
-} __attribute__((aligned(CAA_CACHE_LINE_SIZE)));
-
-/*
- * rcu_level: Contains the per order-index-level dummy node table. The
- * size of each dummy node table is half the number of hashes contained
- * in this order (except for order 0). The minimum allocation size
- * parameter allows combining the dummy node arrays of the lowermost
- * levels to improve cache locality for small index orders.
- */
-struct rcu_level {
- /* Note: manually update allocation length when adding a field */
- struct _cds_lfht_node nodes[0];
-};
-
-/*
- * rcu_table: Contains the size and desired new size if a resize
- * operation is in progress, as well as the statically-sized array of
- * rcu_level pointers.
- */
-struct rcu_table {
- unsigned long size; /* always a power of 2, shared (RCU) */
- unsigned long resize_target;
- int resize_initiated;
- struct rcu_level *tbl[MAX_TABLE_ORDER];
-};
-
-/*
- * cds_lfht: Top-level data structure representing a lock-free hash
- * table. Defined in the implementation file to make it be an opaque
- * cookie to users.
- */
-struct cds_lfht {
- struct rcu_table t;
- cds_lfht_hash_fct hash_fct;
- cds_lfht_compare_fct compare_fct;
- unsigned long min_alloc_order;
- unsigned long min_alloc_size;
- unsigned long hash_seed;
- int flags;
- /*
- * We need to put the work threads offline (QSBR) when taking this
- * mutex, because we use synchronize_rcu within this mutex critical
- * section, which waits on read-side critical sections, and could
- * therefore cause grace-period deadlock if we hold off RCU G.P.
- * completion.
- */
- pthread_mutex_t resize_mutex; /* resize mutex: add/del mutex */
- unsigned int in_progress_resize, in_progress_destroy;
- void (*cds_lfht_call_rcu)(struct rcu_head *head,
- void (*func)(struct rcu_head *head));
- void (*cds_lfht_synchronize_rcu)(void);
- void (*cds_lfht_rcu_read_lock)(void);
- void (*cds_lfht_rcu_read_unlock)(void);
- void (*cds_lfht_rcu_thread_offline)(void);
- void (*cds_lfht_rcu_thread_online)(void);
- void (*cds_lfht_rcu_register_thread)(void);
- void (*cds_lfht_rcu_unregister_thread)(void);
- pthread_attr_t *resize_attr; /* Resize threads attributes */
- long count; /* global approximate item count */
- struct ht_items_count *split_count; /* split item count */
-};
-
-/*
- * rcu_resize_work: Contains arguments passed to RCU worker thread
- * responsible for performing lazy resize.
- */
-struct rcu_resize_work {
- struct rcu_head head;
- struct cds_lfht *ht;
-};
-
-/*
- * partition_resize_work: Contains arguments passed to worker threads
- * executing the hash table resize on partitions of the hash table
- * assigned to each processor's worker thread.
- */
-struct partition_resize_work {
- pthread_t thread_id;
- struct cds_lfht *ht;
- unsigned long i, start, len;
- void (*fct)(struct cds_lfht *ht, unsigned long i,
- unsigned long start, unsigned long len);
-};
-
-static
-void _cds_lfht_add(struct cds_lfht *ht,
- unsigned long size,
- struct cds_lfht_node *node,
- struct cds_lfht_iter *unique_ret,
- int dummy);
-
-/*
- * Algorithm to reverse bits in a word by lookup table, extended to
- * 64-bit words.
- * Source:
- * http://graphics.stanford.edu/~seander/bithacks.html#BitReverseTable
- * Originally from Public Domain.
- */
-
-static const uint8_t BitReverseTable256[256] =
-{
-#define R2(n) (n), (n) + 2*64, (n) + 1*64, (n) + 3*64
-#define R4(n) R2(n), R2((n) + 2*16), R2((n) + 1*16), R2((n) + 3*16)
-#define R6(n) R4(n), R4((n) + 2*4 ), R4((n) + 1*4 ), R4((n) + 3*4 )
- R6(0), R6(2), R6(1), R6(3)
-};
-#undef R2
-#undef R4
-#undef R6
-
-static
-uint8_t bit_reverse_u8(uint8_t v)
-{
- return BitReverseTable256[v];
-}
-
-static __attribute__((unused))
-uint32_t bit_reverse_u32(uint32_t v)
-{
- return ((uint32_t) bit_reverse_u8(v) << 24) |
- ((uint32_t) bit_reverse_u8(v >> 8) << 16) |
- ((uint32_t) bit_reverse_u8(v >> 16) << 8) |
- ((uint32_t) bit_reverse_u8(v >> 24));
-}
-
-static __attribute__((unused))
-uint64_t bit_reverse_u64(uint64_t v)
-{
- return ((uint64_t) bit_reverse_u8(v) << 56) |
- ((uint64_t) bit_reverse_u8(v >> 8) << 48) |
- ((uint64_t) bit_reverse_u8(v >> 16) << 40) |
- ((uint64_t) bit_reverse_u8(v >> 24) << 32) |
- ((uint64_t) bit_reverse_u8(v >> 32) << 24) |
- ((uint64_t) bit_reverse_u8(v >> 40) << 16) |
- ((uint64_t) bit_reverse_u8(v >> 48) << 8) |
- ((uint64_t) bit_reverse_u8(v >> 56));
-}
-
-static
-unsigned long bit_reverse_ulong(unsigned long v)
-{
-#if (CAA_BITS_PER_LONG == 32)
- return bit_reverse_u32(v);
-#else
- return bit_reverse_u64(v);
-#endif
-}
-
-/*
- * fls: returns the position of the most significant bit.
- * Returns 0 if no bit is set, else returns the position of the most
- * significant bit (from 1 to 32 on 32-bit, from 1 to 64 on 64-bit).
- */
-#if defined(__i386) || defined(__x86_64)
-static inline
-unsigned int fls_u32(uint32_t x)
-{
- int r;
-
- asm("bsrl %1,%0\n\t"
- "jnz 1f\n\t"
- "movl $-1,%0\n\t"
- "1:\n\t"
- : "=r" (r) : "rm" (x));
- return r + 1;
-}
-#define HAS_FLS_U32
-#endif
-
-#if defined(__x86_64)
-static inline
-unsigned int fls_u64(uint64_t x)
-{
- long r;
-
- asm("bsrq %1,%0\n\t"
- "jnz 1f\n\t"
- "movq $-1,%0\n\t"
- "1:\n\t"
- : "=r" (r) : "rm" (x));
- return r + 1;
-}
-#define HAS_FLS_U64
-#endif
-
-#ifndef HAS_FLS_U64
-static __attribute__((unused))
-unsigned int fls_u64(uint64_t x)
-{
- unsigned int r = 64;
-
- if (!x)
- return 0;
-
- if (!(x & 0xFFFFFFFF00000000ULL)) {
- x <<= 32;
- r -= 32;
- }
- if (!(x & 0xFFFF000000000000ULL)) {
- x <<= 16;
- r -= 16;
- }
- if (!(x & 0xFF00000000000000ULL)) {
- x <<= 8;
- r -= 8;
- }
- if (!(x & 0xF000000000000000ULL)) {
- x <<= 4;
- r -= 4;
- }
- if (!(x & 0xC000000000000000ULL)) {
- x <<= 2;
- r -= 2;
- }
- if (!(x & 0x8000000000000000ULL)) {
- x <<= 1;
- r -= 1;
- }
- return r;
-}
-#endif
-
-#ifndef HAS_FLS_U32
-static __attribute__((unused))
-unsigned int fls_u32(uint32_t x)
-{
- unsigned int r = 32;
-
- if (!x)
- return 0;
- if (!(x & 0xFFFF0000U)) {
- x <<= 16;
- r -= 16;
- }
- if (!(x & 0xFF000000U)) {
- x <<= 8;
- r -= 8;
- }
- if (!(x & 0xF0000000U)) {
- x <<= 4;
- r -= 4;
- }
- if (!(x & 0xC0000000U)) {
- x <<= 2;
- r -= 2;
- }
- if (!(x & 0x80000000U)) {
- x <<= 1;
- r -= 1;
- }
- return r;
-}
-#endif
-
-unsigned int fls_ulong(unsigned long x)
-{
-#if (CAA_BITS_PER_LONG == 32)
- return fls_u32(x);
-#else
- return fls_u64(x);
-#endif
-}
-
-/*
- * Return the minimum order for which x <= (1UL << order).
- * Return -1 if x is 0.
- */
-int get_count_order_u32(uint32_t x)
-{
- if (!x)
- return -1;
-
- return fls_u32(x - 1);
-}
-
-/*
- * Return the minimum order for which x <= (1UL << order).
- * Return -1 if x is 0.
- */
-int get_count_order_ulong(unsigned long x)
-{
- if (!x)
- return -1;
-
- return fls_ulong(x - 1);
-}
-
-#ifdef POISON_FREE
-#define poison_free(ptr) \
- do { \
- if (ptr) { \
- memset(ptr, 0x42, sizeof(*(ptr))); \
- free(ptr); \
- } \
- } while (0)
-#else
-#define poison_free(ptr) free(ptr)
-#endif
-
-static
-void cds_lfht_resize_lazy(struct cds_lfht *ht, unsigned long size, int growth);
-
-static
-void cds_lfht_resize_lazy_count(struct cds_lfht *ht, unsigned long size,
- unsigned long count);
-
-static long nr_cpus_mask = -1;
-static long split_count_mask = -1;
-
-#if defined(HAVE_SYSCONF)
-static void ht_init_nr_cpus_mask(void)
-{
- long maxcpus;
-
- maxcpus = sysconf(_SC_NPROCESSORS_CONF);
- if (maxcpus <= 0) {
- nr_cpus_mask = -2;
- return;
- }
- /*
- * round up number of CPUs to next power of two, so we
- * can use & for modulo.
- */
- maxcpus = 1UL << get_count_order_ulong(maxcpus);
- nr_cpus_mask = maxcpus - 1;
-}
-#else /* #if defined(HAVE_SYSCONF) */
-static void ht_init_nr_cpus_mask(void)
-{
- nr_cpus_mask = -2;
-}
-#endif /* #else #if defined(HAVE_SYSCONF) */
-
-static
-void alloc_split_items_count(struct cds_lfht *ht)
-{
- struct ht_items_count *count;
-
- if (nr_cpus_mask == -1) {
- ht_init_nr_cpus_mask();
- if (nr_cpus_mask < 0)
- split_count_mask = DEFAULT_SPLIT_COUNT_MASK;
- else
- split_count_mask = nr_cpus_mask;
- }
-
- assert(split_count_mask >= 0);
-
- if (ht->flags & CDS_LFHT_ACCOUNTING) {
- ht->split_count = calloc(split_count_mask + 1, sizeof(*count));
- assert(ht->split_count);
- } else {
- ht->split_count = NULL;
- }
-}
-
-static
-void free_split_items_count(struct cds_lfht *ht)
-{
- poison_free(ht->split_count);
-}
-
-#if defined(HAVE_SCHED_GETCPU)
-static
-int ht_get_split_count_index(unsigned long hash)
-{
- int cpu;
-
- assert(split_count_mask >= 0);
- cpu = sched_getcpu();
- if (caa_unlikely(cpu < 0))
- return hash & split_count_mask;
- else
- return cpu & split_count_mask;
-}
-#else /* #if defined(HAVE_SCHED_GETCPU) */
-static
-int ht_get_split_count_index(unsigned long hash)
-{
- return hash & split_count_mask;
-}
-#endif /* #else #if defined(HAVE_SCHED_GETCPU) */
-
-static
-void ht_count_add(struct cds_lfht *ht, unsigned long size, unsigned long hash)
-{
- unsigned long split_count;
- int index;
-
- if (caa_unlikely(!ht->split_count))
- return;
- index = ht_get_split_count_index(hash);
- split_count = uatomic_add_return(&ht->split_count[index].add, 1);
- if (caa_unlikely(!(split_count & ((1UL << COUNT_COMMIT_ORDER) - 1)))) {
- long count;
-
- dbg_printf("add split count %lu\n", split_count);
- count = uatomic_add_return(&ht->count,
- 1UL << COUNT_COMMIT_ORDER);
- /* If power of 2 */
- if (!(count & (count - 1))) {
- if ((count >> CHAIN_LEN_RESIZE_THRESHOLD) < size)
- return;
- dbg_printf("add set global %ld\n", count);
- cds_lfht_resize_lazy_count(ht, size,
- count >> (CHAIN_LEN_TARGET - 1));
- }
- }
-}
-
-static
-void ht_count_del(struct cds_lfht *ht, unsigned long size, unsigned long hash)
-{
- unsigned long split_count;
- int index;
-
- if (caa_unlikely(!ht->split_count))
- return;
- index = ht_get_split_count_index(hash);
- split_count = uatomic_add_return(&ht->split_count[index].del, 1);
- if (caa_unlikely(!(split_count & ((1UL << COUNT_COMMIT_ORDER) - 1)))) {
- long count;
-
- dbg_printf("del split count %lu\n", split_count);
- count = uatomic_add_return(&ht->count,
- -(1UL << COUNT_COMMIT_ORDER));
- /* If power of 2 */
- if (!(count & (count - 1))) {
- if ((count >> CHAIN_LEN_RESIZE_THRESHOLD) >= size)
- return;
- dbg_printf("del set global %ld\n", count);
- /*
- * Don't shrink table if the number of nodes is below a
- * certain threshold.
- */
- if (count < (1UL << COUNT_COMMIT_ORDER) * (split_count_mask + 1))
- return;
- cds_lfht_resize_lazy_count(ht, size,
- count >> (CHAIN_LEN_TARGET - 1));
- }
- }
-}
-
-static
-void check_resize(struct cds_lfht *ht, unsigned long size, uint32_t chain_len)
-{
- unsigned long count;
-
- if (!(ht->flags & CDS_LFHT_AUTO_RESIZE))
- return;
- count = uatomic_read(&ht->count);
- /*
- * Use bucket-local length for small table expand and for
- * environments lacking per-cpu data support.
- */
- if (count >= (1UL << COUNT_COMMIT_ORDER))
- return;
- if (chain_len > 100)
- dbg_printf("WARNING: large chain length: %u.\n",
- chain_len);
- if (chain_len >= CHAIN_LEN_RESIZE_THRESHOLD)
- cds_lfht_resize_lazy(ht, size,
- get_count_order_u32(chain_len - (CHAIN_LEN_TARGET - 1)));
-}
-
-static
-struct cds_lfht_node *clear_flag(struct cds_lfht_node *node)
-{
- return (struct cds_lfht_node *) (((unsigned long) node) & ~FLAGS_MASK);
-}
-
-static
-int is_removed(struct cds_lfht_node *node)
-{
- return ((unsigned long) node) & REMOVED_FLAG;
-}
-
-static
-struct cds_lfht_node *flag_removed(struct cds_lfht_node *node)
-{
- return (struct cds_lfht_node *) (((unsigned long) node) | REMOVED_FLAG);
-}
-
-static
-int is_dummy(struct cds_lfht_node *node)
-{
- return ((unsigned long) node) & DUMMY_FLAG;
-}
-
-static
-struct cds_lfht_node *flag_dummy(struct cds_lfht_node *node)
-{
- return (struct cds_lfht_node *) (((unsigned long) node) | DUMMY_FLAG);
-}
-
-static
-struct cds_lfht_node *get_end(void)
-{
- return (struct cds_lfht_node *) END_VALUE;
-}
-
-static
-int is_end(struct cds_lfht_node *node)
-{
- return clear_flag(node) == (struct cds_lfht_node *) END_VALUE;
-}
-
-static
-unsigned long _uatomic_max(unsigned long *ptr, unsigned long v)
-{
- unsigned long old1, old2;
-
- old1 = uatomic_read(ptr);
- do {
- old2 = old1;
- if (old2 >= v)
- return old2;
- } while ((old1 = uatomic_cmpxchg(ptr, old2, v)) != old2);
- return v;
-}
-
-static
-struct _cds_lfht_node *lookup_bucket(struct cds_lfht *ht, unsigned long size,
- unsigned long hash)
-{
- unsigned long index, order;
-
- assert(size > 0);
- index = hash & (size - 1);
-
- if (index < ht->min_alloc_size) {
- dbg_printf("lookup hash %lu index %lu order 0 aridx 0\n",
- hash, index);
- return &ht->t.tbl[0]->nodes[index];
- }
- /*
- * equivalent to get_count_order_ulong(index + 1), but optimizes
- * away the non-existing 0 special-case for
- * get_count_order_ulong.
- */
- order = fls_ulong(index);
- dbg_printf("lookup hash %lu index %lu order %lu aridx %lu\n",
- hash, index, order, index & ((1UL << (order - 1)) - 1));
- return &ht->t.tbl[order]->nodes[index & ((1UL << (order - 1)) - 1)];
-}
-
-/*
- * Remove all logically deleted nodes from a bucket up to a certain node key.
- */
-static
-void _cds_lfht_gc_bucket(struct cds_lfht_node *dummy, struct cds_lfht_node *node)
-{
- struct cds_lfht_node *iter_prev, *iter, *next, *new_next;
-
- assert(!is_dummy(dummy));
- assert(!is_removed(dummy));
- assert(!is_dummy(node));
- assert(!is_removed(node));
- for (;;) {
- iter_prev = dummy;
- /* We can always skip the dummy node initially */
- iter = rcu_dereference(iter_prev->p.next);
- assert(!is_removed(iter));
- assert(iter_prev->p.reverse_hash <= node->p.reverse_hash);
- /*
- * We should never be called with dummy (start of chain)
- * and logically removed node (end of path compression
- * marker) being the actual same node. This would be a
- * bug in the algorithm implementation.
- */
- assert(dummy != node);
- for (;;) {
- if (caa_unlikely(is_end(iter)))
- return;
- if (caa_likely(clear_flag(iter)->p.reverse_hash > node->p.reverse_hash))
- return;
- next = rcu_dereference(clear_flag(iter)->p.next);
- if (caa_likely(is_removed(next)))
- break;
- iter_prev = clear_flag(iter);
- iter = next;
- }
- assert(!is_removed(iter));
- if (is_dummy(iter))
- new_next = flag_dummy(clear_flag(next));
- else
- new_next = clear_flag(next);
- (void) uatomic_cmpxchg(&iter_prev->p.next, iter, new_next);
- }
- return;
-}
-
-static
-int _cds_lfht_replace(struct cds_lfht *ht, unsigned long size,
- struct cds_lfht_node *old_node,
- struct cds_lfht_node *old_next,
- struct cds_lfht_node *new_node)
-{
- struct cds_lfht_node *dummy, *ret_next;
- struct _cds_lfht_node *lookup;
-
- if (!old_node) /* Return -ENOENT if asked to replace NULL node */
- return -ENOENT;
-
- assert(!is_removed(old_node));
- assert(!is_dummy(old_node));
- assert(!is_removed(new_node));
- assert(!is_dummy(new_node));
- assert(new_node != old_node);
- for (;;) {
- /* Insert after node to be replaced */
- if (is_removed(old_next)) {
- /*
- * Too late, the old node has been removed under us
- * between lookup and replace. Fail.
- */
- return -ENOENT;
- }
- assert(!is_dummy(old_next));
- assert(new_node != clear_flag(old_next));
- new_node->p.next = clear_flag(old_next);
- /*
- * Here is the whole trick for lock-free replace: we add
- * the replacement node _after_ the node we want to
- * replace by atomically setting its next pointer at the
- * same time we set its removal flag. Given that
- * the lookups/get next use an iterator aware of the
- * next pointer, they will either skip the old node due
- * to the removal flag and see the new node, or use
- * the old node, but will not see the new one.
- */
- ret_next = uatomic_cmpxchg(&old_node->p.next,
- old_next, flag_removed(new_node));
- if (ret_next == old_next)
- break; /* We performed the replacement. */
- old_next = ret_next;
- }
-
- /*
- * Ensure that the old node is not visible to readers anymore:
- * lookup for the node, and remove it (along with any other
- * logically removed node) if found.
- */
- lookup = lookup_bucket(ht, size, bit_reverse_ulong(old_node->p.reverse_hash));
- dummy = (struct cds_lfht_node *) lookup;
- _cds_lfht_gc_bucket(dummy, new_node);
-
- assert(is_removed(rcu_dereference(old_node->p.next)));
- return 0;
-}
-
-/*
- * A non-NULL unique_ret pointer uses the "add unique" (or uniquify) add
- * mode. A NULL unique_ret allows creation of duplicate keys.
- */
-static
-void _cds_lfht_add(struct cds_lfht *ht,
- unsigned long size,
- struct cds_lfht_node *node,
- struct cds_lfht_iter *unique_ret,
- int dummy)
-{
- struct cds_lfht_node *iter_prev, *iter, *next, *new_node, *new_next,
- *return_node;
- struct _cds_lfht_node *lookup;
-
- assert(!is_dummy(node));
- assert(!is_removed(node));
- lookup = lookup_bucket(ht, size, bit_reverse_ulong(node->p.reverse_hash));
- for (;;) {
- uint32_t chain_len = 0;
-
- /*
- * iter_prev points to the non-removed node prior to the
- * insert location.
- */
- iter_prev = (struct cds_lfht_node *) lookup;
- /* We can always skip the dummy node initially */
- iter = rcu_dereference(iter_prev->p.next);
- assert(iter_prev->p.reverse_hash <= node->p.reverse_hash);
- for (;;) {
- if (caa_unlikely(is_end(iter)))
- goto insert;
- if (caa_likely(clear_flag(iter)->p.reverse_hash > node->p.reverse_hash))
- goto insert;
-
- /* dummy node is the first node of the identical-hash-value chain */
- if (dummy && clear_flag(iter)->p.reverse_hash == node->p.reverse_hash)
- goto insert;
-
- next = rcu_dereference(clear_flag(iter)->p.next);
- if (caa_unlikely(is_removed(next)))
- goto gc_node;
-
- /* uniquely add */
- if (unique_ret
- && !is_dummy(next)
- && clear_flag(iter)->p.reverse_hash == node->p.reverse_hash) {
- struct cds_lfht_iter d_iter = { .node = node, .next = iter, };
-
- /*
- * uniquely adding inserts the node as the first
- * node of the identical-hash-value node chain.
- *
- * This semantic ensures no duplicated keys
- * should ever be observable in the table
- * (including observe one node by one node
- * by forward iterations)
- */
- cds_lfht_next_duplicate(ht, &d_iter);
- if (!d_iter.node)
- goto insert;
-
- *unique_ret = d_iter;
- return;
- }
-
- /* Only account for identical reverse hash once */
- if (iter_prev->p.reverse_hash != clear_flag(iter)->p.reverse_hash
- && !is_dummy(next))
- check_resize(ht, size, ++chain_len);
- iter_prev = clear_flag(iter);
- iter = next;
- }
-
- insert:
- assert(node != clear_flag(iter));
- assert(!is_removed(iter_prev));
- assert(!is_removed(iter));
- assert(iter_prev != node);
- if (!dummy)
- node->p.next = clear_flag(iter);
- else
- node->p.next = flag_dummy(clear_flag(iter));
- if (is_dummy(iter))
- new_node = flag_dummy(node);
- else
- new_node = node;
- if (uatomic_cmpxchg(&iter_prev->p.next, iter,
- new_node) != iter) {
- continue; /* retry */
- } else {
- return_node = node;
- goto end;
- }
-
- gc_node:
- assert(!is_removed(iter));
- if (is_dummy(iter))
- new_next = flag_dummy(clear_flag(next));
- else
- new_next = clear_flag(next);
- (void) uatomic_cmpxchg(&iter_prev->p.next, iter, new_next);
- /* retry */
- }
-end:
- if (unique_ret) {
- unique_ret->node = return_node;
- /* unique_ret->next left unset, never used. */
- }
-}
-
-static
-int _cds_lfht_del(struct cds_lfht *ht, unsigned long size,
- struct cds_lfht_node *node,
- int dummy_removal)
-{
- struct cds_lfht_node *dummy, *next, *old;
- struct _cds_lfht_node *lookup;
-
- if (!node) /* Return -ENOENT if asked to delete NULL node */
- return -ENOENT;
-
- /* logically delete the node */
- assert(!is_dummy(node));
- assert(!is_removed(node));
- old = rcu_dereference(node->p.next);
- do {
- struct cds_lfht_node *new_next;
-
- next = old;
- if (caa_unlikely(is_removed(next)))
- return -ENOENT;
- if (dummy_removal)
- assert(is_dummy(next));
- else
- assert(!is_dummy(next));
- new_next = flag_removed(next);
- old = uatomic_cmpxchg(&node->p.next, next, new_next);
- } while (old != next);
- /* We performed the (logical) deletion. */
-
- /*
- * Ensure that the node is not visible to readers anymore: lookup for
- * the node, and remove it (along with any other logically removed node)
- * if found.
- */
- lookup = lookup_bucket(ht, size, bit_reverse_ulong(node->p.reverse_hash));
- dummy = (struct cds_lfht_node *) lookup;
- _cds_lfht_gc_bucket(dummy, node);
-
- assert(is_removed(rcu_dereference(node->p.next)));
- return 0;
-}
-
-static
-void *partition_resize_thread(void *arg)
-{
- struct partition_resize_work *work = arg;
-
- work->ht->cds_lfht_rcu_register_thread();
- work->fct(work->ht, work->i, work->start, work->len);
- work->ht->cds_lfht_rcu_unregister_thread();
- return NULL;
-}
-
-static
-void partition_resize_helper(struct cds_lfht *ht, unsigned long i,
- unsigned long len,
- void (*fct)(struct cds_lfht *ht, unsigned long i,
- unsigned long start, unsigned long len))
-{
- unsigned long partition_len;
- struct partition_resize_work *work;
- int thread, ret;
- unsigned long nr_threads;
-
- /*
- * Note: nr_cpus_mask + 1 is always power of 2.
- * We spawn just the number of threads we need to satisfy the minimum
- * partition size, up to the number of CPUs in the system.
- */
- if (nr_cpus_mask > 0) {
- nr_threads = min(nr_cpus_mask + 1,
- len >> MIN_PARTITION_PER_THREAD_ORDER);
- } else {
- nr_threads = 1;
- }
- partition_len = len >> get_count_order_ulong(nr_threads);
- work = calloc(nr_threads, sizeof(*work));
- assert(work);
- for (thread = 0; thread < nr_threads; thread++) {
- work[thread].ht = ht;
- work[thread].i = i;
- work[thread].len = partition_len;
- work[thread].start = thread * partition_len;
- work[thread].fct = fct;
- ret = pthread_create(&(work[thread].thread_id), ht->resize_attr,
- partition_resize_thread, &work[thread]);
- assert(!ret);
- }
- for (thread = 0; thread < nr_threads; thread++) {
- ret = pthread_join(work[thread].thread_id, NULL);
- assert(!ret);
- }
- free(work);
-}
-
-/*
- * Holding RCU read lock to protect _cds_lfht_add against memory
- * reclaim that could be performed by other call_rcu worker threads (ABA
- * problem).
- *
- * When we reach a certain length, we can split this population phase over
- * many worker threads, based on the number of CPUs available in the system.
- * This should therefore take care of not having the expand lagging behind too
- * many concurrent insertion threads by using the scheduler's ability to
- * schedule dummy node population fairly with insertions.
- */
-static
-void init_table_populate_partition(struct cds_lfht *ht, unsigned long i,
- unsigned long start, unsigned long len)
-{
- unsigned long j;
-
- assert(i > ht->min_alloc_order);
- ht->cds_lfht_rcu_read_lock();
- for (j = start; j < start + len; j++) {
- struct cds_lfht_node *new_node =
- (struct cds_lfht_node *) &ht->t.tbl[i]->nodes[j];
-
- dbg_printf("init populate: i %lu j %lu hash %lu\n",
- i, j, (1UL << (i - 1)) + j);
- new_node->p.reverse_hash =
- bit_reverse_ulong((1UL << (i - 1)) + j);
- _cds_lfht_add(ht, 1UL << (i - 1),
- new_node, NULL, 1);
- }
- ht->cds_lfht_rcu_read_unlock();
-}
-
-static
-void init_table_populate(struct cds_lfht *ht, unsigned long i,
- unsigned long len)
-{
- assert(nr_cpus_mask != -1);
- if (nr_cpus_mask < 0 || len < 2 * MIN_PARTITION_PER_THREAD) {
- ht->cds_lfht_rcu_thread_online();
- init_table_populate_partition(ht, i, 0, len);
- ht->cds_lfht_rcu_thread_offline();
- return;
- }
- partition_resize_helper(ht, i, len, init_table_populate_partition);
-}
-
-static
-void init_table(struct cds_lfht *ht,
- unsigned long first_order, unsigned long last_order)
-{
- unsigned long i;
-
- dbg_printf("init table: first_order %lu last_order %lu\n",
- first_order, last_order);
- assert(first_order > ht->min_alloc_order);
- for (i = first_order; i <= last_order; i++) {
- unsigned long len;
-
- len = 1UL << (i - 1);
- dbg_printf("init order %lu len: %lu\n", i, len);
-
- /* Stop expand if the resize target changes under us */
- if (CMM_LOAD_SHARED(ht->t.resize_target) < (1UL << i))
- break;
-
- ht->t.tbl[i] = calloc(1, len * sizeof(struct _cds_lfht_node));
- assert(ht->t.tbl[i]);
-
- /*
- * Set all dummy nodes reverse hash values for a level and
- * link all dummy nodes into the table.
- */
- init_table_populate(ht, i, len);
-
- /*
- * Update table size.
- */
- cmm_smp_wmb(); /* populate data before RCU size */
- CMM_STORE_SHARED(ht->t.size, 1UL << i);
-
- dbg_printf("init new size: %lu\n", 1UL << i);
- if (CMM_LOAD_SHARED(ht->in_progress_destroy))
- break;
- }
-}
-
-/*
- * Holding RCU read lock to protect _cds_lfht_remove against memory
- * reclaim that could be performed by other call_rcu worker threads (ABA
- * problem).
- * For a single level, we logically remove and garbage collect each node.
- *
- * As a design choice, we perform logical removal and garbage collection on a
- * node-per-node basis to simplify this algorithm. We also assume keeping good
- * cache locality of the operation would overweight possible performance gain
- * that could be achieved by batching garbage collection for multiple levels.
- * However, this would have to be justified by benchmarks.
- *
- * Concurrent removal and add operations are helping us perform garbage
- * collection of logically removed nodes. We guarantee that all logically
- * removed nodes have been garbage-collected (unlinked) before call_rcu is
- * invoked to free a hole level of dummy nodes (after a grace period).
- *
- * Logical removal and garbage collection can therefore be done in batch or on a
- * node-per-node basis, as long as the guarantee above holds.
- *
- * When we reach a certain length, we can split this removal over many worker
- * threads, based on the number of CPUs available in the system. This should
- * take care of not letting resize process lag behind too many concurrent
- * updater threads actively inserting into the hash table.
- */
-static
-void remove_table_partition(struct cds_lfht *ht, unsigned long i,
- unsigned long start, unsigned long len)
-{
- unsigned long j;
-
- assert(i > ht->min_alloc_order);
- ht->cds_lfht_rcu_read_lock();
- for (j = start; j < start + len; j++) {
- struct cds_lfht_node *fini_node =
- (struct cds_lfht_node *) &ht->t.tbl[i]->nodes[j];
-
- dbg_printf("remove entry: i %lu j %lu hash %lu\n",
- i, j, (1UL << (i - 1)) + j);
- fini_node->p.reverse_hash =
- bit_reverse_ulong((1UL << (i - 1)) + j);
- (void) _cds_lfht_del(ht, 1UL << (i - 1), fini_node, 1);
- }
- ht->cds_lfht_rcu_read_unlock();
-}
-
-static
-void remove_table(struct cds_lfht *ht, unsigned long i, unsigned long len)
-{
-
- assert(nr_cpus_mask != -1);
- if (nr_cpus_mask < 0 || len < 2 * MIN_PARTITION_PER_THREAD) {
- ht->cds_lfht_rcu_thread_online();
- remove_table_partition(ht, i, 0, len);
- ht->cds_lfht_rcu_thread_offline();
- return;
- }
- partition_resize_helper(ht, i, len, remove_table_partition);
-}
-
-static
-void fini_table(struct cds_lfht *ht,
- unsigned long first_order, unsigned long last_order)
-{
- long i;
- void *free_by_rcu = NULL;
-
- dbg_printf("fini table: first_order %lu last_order %lu\n",
- first_order, last_order);
- assert(first_order > ht->min_alloc_order);
- for (i = last_order; i >= first_order; i--) {
- unsigned long len;
-
- len = 1UL << (i - 1);
- dbg_printf("fini order %lu len: %lu\n", i, len);
-
- /* Stop shrink if the resize target changes under us */
- if (CMM_LOAD_SHARED(ht->t.resize_target) > (1UL << (i - 1)))
- break;
-
- cmm_smp_wmb(); /* populate data before RCU size */
- CMM_STORE_SHARED(ht->t.size, 1UL << (i - 1));
-
- /*
- * We need to wait for all add operations to reach Q.S. (and
- * thus use the new table for lookups) before we can start
- * releasing the old dummy nodes. Otherwise their lookup will
- * return a logically removed node as insert position.
- */
- ht->cds_lfht_synchronize_rcu();
- if (free_by_rcu)
- free(free_by_rcu);
-
- /*
- * Set "removed" flag in dummy nodes about to be removed.
- * Unlink all now-logically-removed dummy node pointers.
- * Concurrent add/remove operation are helping us doing
- * the gc.
- */
- remove_table(ht, i, len);
-
- free_by_rcu = ht->t.tbl[i];
-
- dbg_printf("fini new size: %lu\n", 1UL << i);
- if (CMM_LOAD_SHARED(ht->in_progress_destroy))
- break;
- }
-
- if (free_by_rcu) {
- ht->cds_lfht_synchronize_rcu();
- free(free_by_rcu);
- }
-}
-
-static
-void cds_lfht_create_dummy(struct cds_lfht *ht, unsigned long size)
-{
- struct _cds_lfht_node *prev, *node;
- unsigned long order, len, i, j;
-
- ht->t.tbl[0] = calloc(1, ht->min_alloc_size * sizeof(struct _cds_lfht_node));
- assert(ht->t.tbl[0]);
-
- dbg_printf("create dummy: order %lu index %lu hash %lu\n", 0, 0, 0);
- ht->t.tbl[0]->nodes[0].next = flag_dummy(get_end());
- ht->t.tbl[0]->nodes[0].reverse_hash = 0;
-
- for (order = 1; order < get_count_order_ulong(size) + 1; order++) {
- len = 1UL << (order - 1);
- if (order <= ht->min_alloc_order) {
- ht->t.tbl[order] = (struct rcu_level *) (ht->t.tbl[0]->nodes + len);
- } else {
- ht->t.tbl[order] = calloc(1, len * sizeof(struct _cds_lfht_node));
- assert(ht->t.tbl[order]);
- }
-
- i = 0;
- prev = ht->t.tbl[i]->nodes;
- for (j = 0; j < len; j++) {
- if (j & (j - 1)) { /* Between power of 2 */
- prev++;
- } else if (j) { /* At each power of 2 */
- i++;
- prev = ht->t.tbl[i]->nodes;
- }
-
- node = &ht->t.tbl[order]->nodes[j];
- dbg_printf("create dummy: order %lu index %lu hash %lu\n",
- order, j, j + len);
- node->next = prev->next;
- assert(is_dummy(node->next));
- node->reverse_hash = bit_reverse_ulong(j + len);
- prev->next = flag_dummy((struct cds_lfht_node *)node);
- }
- }
-}
-
-struct cds_lfht *_cds_lfht_new(cds_lfht_hash_fct hash_fct,
- cds_lfht_compare_fct compare_fct,
- unsigned long hash_seed,
- unsigned long init_size,
- unsigned long min_alloc_size,
- int flags,
- void (*cds_lfht_call_rcu)(struct rcu_head *head,
- void (*func)(struct rcu_head *head)),
- void (*cds_lfht_synchronize_rcu)(void),
- void (*cds_lfht_rcu_read_lock)(void),
- void (*cds_lfht_rcu_read_unlock)(void),
- void (*cds_lfht_rcu_thread_offline)(void),
- void (*cds_lfht_rcu_thread_online)(void),
- void (*cds_lfht_rcu_register_thread)(void),
- void (*cds_lfht_rcu_unregister_thread)(void),
- pthread_attr_t *attr)
-{
- struct cds_lfht *ht;
- unsigned long order;
-
- /* min_alloc_size must be power of two */
- if (!min_alloc_size || (min_alloc_size & (min_alloc_size - 1)))
- return NULL;
- /* init_size must be power of two */
- if (!init_size || (init_size & (init_size - 1)))
- return NULL;
- min_alloc_size = max(min_alloc_size, MIN_TABLE_SIZE);
- init_size = max(init_size, min_alloc_size);
- ht = calloc(1, sizeof(struct cds_lfht));
- assert(ht);
- ht->flags = flags;
- ht->hash_fct = hash_fct;
- ht->compare_fct = compare_fct;
- ht->hash_seed = hash_seed;
- ht->cds_lfht_call_rcu = cds_lfht_call_rcu;
- ht->cds_lfht_synchronize_rcu = cds_lfht_synchronize_rcu;
- ht->cds_lfht_rcu_read_lock = cds_lfht_rcu_read_lock;
- ht->cds_lfht_rcu_read_unlock = cds_lfht_rcu_read_unlock;
- ht->cds_lfht_rcu_thread_offline = cds_lfht_rcu_thread_offline;
- ht->cds_lfht_rcu_thread_online = cds_lfht_rcu_thread_online;
- ht->cds_lfht_rcu_register_thread = cds_lfht_rcu_register_thread;
- ht->cds_lfht_rcu_unregister_thread = cds_lfht_rcu_unregister_thread;
- ht->resize_attr = attr;
- alloc_split_items_count(ht);
- /* this mutex should not nest in read-side C.S. */
- pthread_mutex_init(&ht->resize_mutex, NULL);
- order = get_count_order_ulong(init_size);
- ht->t.resize_target = 1UL << order;
- ht->min_alloc_size = min_alloc_size;
- ht->min_alloc_order = get_count_order_ulong(min_alloc_size);
- cds_lfht_create_dummy(ht, 1UL << order);
- ht->t.size = 1UL << order;
- return ht;
-}
-
-void cds_lfht_lookup(struct cds_lfht *ht, void *key, size_t key_len,
- struct cds_lfht_iter *iter)
-{
- struct cds_lfht_node *node, *next, *dummy_node;
- struct _cds_lfht_node *lookup;
- unsigned long hash, reverse_hash, size;
-
- hash = ht->hash_fct(key, key_len, ht->hash_seed);
- reverse_hash = bit_reverse_ulong(hash);
-
- size = rcu_dereference(ht->t.size);
- lookup = lookup_bucket(ht, size, hash);
- dummy_node = (struct cds_lfht_node *) lookup;
- /* We can always skip the dummy node initially */
- node = rcu_dereference(dummy_node->p.next);
- node = clear_flag(node);
- for (;;) {
- if (caa_unlikely(is_end(node))) {
- node = next = NULL;
- break;
- }
- if (caa_unlikely(node->p.reverse_hash > reverse_hash)) {
- node = next = NULL;
- break;
- }
- next = rcu_dereference(node->p.next);
- assert(node == clear_flag(node));
- if (caa_likely(!is_removed(next))
- && !is_dummy(next)
- && node->p.reverse_hash == reverse_hash
- && caa_likely(!ht->compare_fct(node->key, node->key_len, key, key_len))) {
- break;
- }
- node = clear_flag(next);
- }
- assert(!node || !is_dummy(rcu_dereference(node->p.next)));
- iter->node = node;
- iter->next = next;
-}
-
-void cds_lfht_next_duplicate(struct cds_lfht *ht, struct cds_lfht_iter *iter)
-{
- struct cds_lfht_node *node, *next;
- unsigned long reverse_hash;
- void *key;
- size_t key_len;
-
- node = iter->node;
- reverse_hash = node->p.reverse_hash;
- key = node->key;
- key_len = node->key_len;
- next = iter->next;
- node = clear_flag(next);
-
- for (;;) {
- if (caa_unlikely(is_end(node))) {
- node = next = NULL;
- break;
- }
- if (caa_unlikely(node->p.reverse_hash > reverse_hash)) {
- node = next = NULL;
- break;
- }
- next = rcu_dereference(node->p.next);
- if (caa_likely(!is_removed(next))
- && !is_dummy(next)
- && caa_likely(!ht->compare_fct(node->key, node->key_len, key, key_len))) {
- break;
- }
- node = clear_flag(next);
- }
- assert(!node || !is_dummy(rcu_dereference(node->p.next)));
- iter->node = node;
- iter->next = next;
-}
-
-void cds_lfht_next(struct cds_lfht *ht, struct cds_lfht_iter *iter)
-{
- struct cds_lfht_node *node, *next;
-
- node = clear_flag(iter->next);
- for (;;) {
- if (caa_unlikely(is_end(node))) {
- node = next = NULL;
- break;
- }
- next = rcu_dereference(node->p.next);
- if (caa_likely(!is_removed(next))
- && !is_dummy(next)) {
- break;
- }
- node = clear_flag(next);
- }
- assert(!node || !is_dummy(rcu_dereference(node->p.next)));
- iter->node = node;
- iter->next = next;
-}
-
-void cds_lfht_first(struct cds_lfht *ht, struct cds_lfht_iter *iter)
-{
- struct _cds_lfht_node *lookup;
-
- /*
- * Get next after first dummy node. The first dummy node is the
- * first node of the linked list.
- */
- lookup = &ht->t.tbl[0]->nodes[0];
- iter->next = lookup->next;
- cds_lfht_next(ht, iter);
-}
-
-void cds_lfht_add(struct cds_lfht *ht, struct cds_lfht_node *node)
-{
- unsigned long hash, size;
-
- hash = ht->hash_fct(node->key, node->key_len, ht->hash_seed);
- node->p.reverse_hash = bit_reverse_ulong((unsigned long) hash);
-
- size = rcu_dereference(ht->t.size);
- _cds_lfht_add(ht, size, node, NULL, 0);
- ht_count_add(ht, size, hash);
-}
-
-struct cds_lfht_node *cds_lfht_add_unique(struct cds_lfht *ht,
- struct cds_lfht_node *node)
-{
- unsigned long hash, size;
- struct cds_lfht_iter iter;
-
- hash = ht->hash_fct(node->key, node->key_len, ht->hash_seed);
- node->p.reverse_hash = bit_reverse_ulong((unsigned long) hash);
-
- size = rcu_dereference(ht->t.size);
- _cds_lfht_add(ht, size, node, &iter, 0);
- if (iter.node == node)
- ht_count_add(ht, size, hash);
- return iter.node;
-}
-
-struct cds_lfht_node *cds_lfht_add_replace(struct cds_lfht *ht,
- struct cds_lfht_node *node)
-{
- unsigned long hash, size;
- struct cds_lfht_iter iter;
-
- hash = ht->hash_fct(node->key, node->key_len, ht->hash_seed);
- node->p.reverse_hash = bit_reverse_ulong((unsigned long) hash);
-
- size = rcu_dereference(ht->t.size);
- for (;;) {
- _cds_lfht_add(ht, size, node, &iter, 0);
- if (iter.node == node) {
- ht_count_add(ht, size, hash);
- return NULL;
- }
-
- if (!_cds_lfht_replace(ht, size, iter.node, iter.next, node))
- return iter.node;
- }
-}
-
-int cds_lfht_replace(struct cds_lfht *ht, struct cds_lfht_iter *old_iter,
- struct cds_lfht_node *new_node)
-{
- unsigned long size;
-
- size = rcu_dereference(ht->t.size);
- return _cds_lfht_replace(ht, size, old_iter->node, old_iter->next,
- new_node);
-}
-
-int cds_lfht_del(struct cds_lfht *ht, struct cds_lfht_iter *iter)
-{
- unsigned long size, hash;
- int ret;
-
- size = rcu_dereference(ht->t.size);
- ret = _cds_lfht_del(ht, size, iter->node, 0);
- if (!ret) {
- hash = bit_reverse_ulong(iter->node->p.reverse_hash);
- ht_count_del(ht, size, hash);
- }
- return ret;
-}
-
-static
-int cds_lfht_delete_dummy(struct cds_lfht *ht)
-{
- struct cds_lfht_node *node;
- struct _cds_lfht_node *lookup;
- unsigned long order, i, size;
-
- /* Check that the table is empty */
- lookup = &ht->t.tbl[0]->nodes[0];
- node = (struct cds_lfht_node *) lookup;
- do {
- node = clear_flag(node)->p.next;
- if (!is_dummy(node))
- return -EPERM;
- assert(!is_removed(node));
- } while (!is_end(node));
- /*
- * size accessed without rcu_dereference because hash table is
- * being destroyed.
- */
- size = ht->t.size;
- /* Internal sanity check: all nodes left should be dummy */
- for (order = 0; order < get_count_order_ulong(size) + 1; order++) {
- unsigned long len;
-
- len = !order ? 1 : 1UL << (order - 1);
- for (i = 0; i < len; i++) {
- dbg_printf("delete order %lu i %lu hash %lu\n",
- order, i,
- bit_reverse_ulong(ht->t.tbl[order]->nodes[i].reverse_hash));
- assert(is_dummy(ht->t.tbl[order]->nodes[i].next));
- }
-
- if (order == ht->min_alloc_order)
- poison_free(ht->t.tbl[0]);
- else if (order > ht->min_alloc_order)
- poison_free(ht->t.tbl[order]);
- /* Nothing to delete for order < ht->min_alloc_order */
- }
- return 0;
-}
-
-/*
- * Should only be called when no more concurrent readers nor writers can
- * possibly access the table.
- */
-int cds_lfht_destroy(struct cds_lfht *ht, pthread_attr_t **attr)
-{
- int ret;
-
- /* Wait for in-flight resize operations to complete */
- _CMM_STORE_SHARED(ht->in_progress_destroy, 1);
- cmm_smp_mb(); /* Store destroy before load resize */
- while (uatomic_read(&ht->in_progress_resize))
- poll(NULL, 0, 100); /* wait for 100ms */
- ret = cds_lfht_delete_dummy(ht);
- if (ret)
- return ret;
- free_split_items_count(ht);
- if (attr)
- *attr = ht->resize_attr;
- poison_free(ht);
- return ret;
-}
-
-void cds_lfht_count_nodes(struct cds_lfht *ht,
- long *approx_before,
- unsigned long *count,
- unsigned long *removed,
- long *approx_after)
-{
- struct cds_lfht_node *node, *next;
- struct _cds_lfht_node *lookup;
- unsigned long nr_dummy = 0;
-
- *approx_before = 0;
- if (ht->split_count) {
- int i;
-
- for (i = 0; i < split_count_mask + 1; i++) {
- *approx_before += uatomic_read(&ht->split_count[i].add);
- *approx_before -= uatomic_read(&ht->split_count[i].del);
- }
- }
-
- *count = 0;
- *removed = 0;
-
- /* Count non-dummy nodes in the table */
- lookup = &ht->t.tbl[0]->nodes[0];
- node = (struct cds_lfht_node *) lookup;
- do {
- next = rcu_dereference(node->p.next);
- if (is_removed(next)) {
- if (!is_dummy(next))
- (*removed)++;
- else
- (nr_dummy)++;
- } else if (!is_dummy(next))
- (*count)++;
- else
- (nr_dummy)++;
- node = clear_flag(next);
- } while (!is_end(node));
- dbg_printf("number of dummy nodes: %lu\n", nr_dummy);
- *approx_after = 0;
- if (ht->split_count) {
- int i;
-
- for (i = 0; i < split_count_mask + 1; i++) {
- *approx_after += uatomic_read(&ht->split_count[i].add);
- *approx_after -= uatomic_read(&ht->split_count[i].del);
- }
- }
-}
-
-/* called with resize mutex held */
-static
-void _do_cds_lfht_grow(struct cds_lfht *ht,
- unsigned long old_size, unsigned long new_size)
-{
- unsigned long old_order, new_order;
-
- old_order = get_count_order_ulong(old_size);
- new_order = get_count_order_ulong(new_size);
- dbg_printf("resize from %lu (order %lu) to %lu (order %lu) buckets\n",
- old_size, old_order, new_size, new_order);
- assert(new_size > old_size);
- init_table(ht, old_order + 1, new_order);
-}
-
-/* called with resize mutex held */
-static
-void _do_cds_lfht_shrink(struct cds_lfht *ht,
- unsigned long old_size, unsigned long new_size)
-{
- unsigned long old_order, new_order;
-
- new_size = max(new_size, ht->min_alloc_size);
- old_order = get_count_order_ulong(old_size);
- new_order = get_count_order_ulong(new_size);
- dbg_printf("resize from %lu (order %lu) to %lu (order %lu) buckets\n",
- old_size, old_order, new_size, new_order);
- assert(new_size < old_size);
-
- /* Remove and unlink all dummy nodes to remove. */
- fini_table(ht, new_order + 1, old_order);
-}
-
-
-/* called with resize mutex held */
-static
-void _do_cds_lfht_resize(struct cds_lfht *ht)
-{
- unsigned long new_size, old_size;
-
- /*
- * Resize table, re-do if the target size has changed under us.
- */
- do {
- assert(uatomic_read(&ht->in_progress_resize));
- if (CMM_LOAD_SHARED(ht->in_progress_destroy))
- break;
- ht->t.resize_initiated = 1;
- old_size = ht->t.size;
- new_size = CMM_LOAD_SHARED(ht->t.resize_target);
- if (old_size < new_size)
- _do_cds_lfht_grow(ht, old_size, new_size);
- else if (old_size > new_size)
- _do_cds_lfht_shrink(ht, old_size, new_size);
- ht->t.resize_initiated = 0;
- /* write resize_initiated before read resize_target */
- cmm_smp_mb();
- } while (ht->t.size != CMM_LOAD_SHARED(ht->t.resize_target));
-}
-
-static
-unsigned long resize_target_update(struct cds_lfht *ht, unsigned long size,
- int growth_order)
-{
- return _uatomic_max(&ht->t.resize_target,
- size << growth_order);
-}
-
-static
-void resize_target_update_count(struct cds_lfht *ht,
- unsigned long count)
-{
- count = max(count, ht->min_alloc_size);
- uatomic_set(&ht->t.resize_target, count);
-}
-
-void cds_lfht_resize(struct cds_lfht *ht, unsigned long new_size)
-{
- resize_target_update_count(ht, new_size);
- CMM_STORE_SHARED(ht->t.resize_initiated, 1);
- ht->cds_lfht_rcu_thread_offline();
- pthread_mutex_lock(&ht->resize_mutex);
- _do_cds_lfht_resize(ht);
- pthread_mutex_unlock(&ht->resize_mutex);
- ht->cds_lfht_rcu_thread_online();
-}
-
-static
-void do_resize_cb(struct rcu_head *head)
-{
- struct rcu_resize_work *work =
- caa_container_of(head, struct rcu_resize_work, head);
- struct cds_lfht *ht = work->ht;
-
- ht->cds_lfht_rcu_thread_offline();
- pthread_mutex_lock(&ht->resize_mutex);
- _do_cds_lfht_resize(ht);
- pthread_mutex_unlock(&ht->resize_mutex);
- ht->cds_lfht_rcu_thread_online();
- poison_free(work);
- cmm_smp_mb(); /* finish resize before decrement */
- uatomic_dec(&ht->in_progress_resize);
-}
-
-static
-void cds_lfht_resize_lazy(struct cds_lfht *ht, unsigned long size, int growth)
-{
- struct rcu_resize_work *work;
- unsigned long target_size;
-
- target_size = resize_target_update(ht, size, growth);
- /* Store resize_target before read resize_initiated */
- cmm_smp_mb();
- if (!CMM_LOAD_SHARED(ht->t.resize_initiated) && size < target_size) {
- uatomic_inc(&ht->in_progress_resize);
- cmm_smp_mb(); /* increment resize count before load destroy */
- if (CMM_LOAD_SHARED(ht->in_progress_destroy)) {
- uatomic_dec(&ht->in_progress_resize);
- return;
- }
- work = malloc(sizeof(*work));
- work->ht = ht;
- ht->cds_lfht_call_rcu(&work->head, do_resize_cb);
- CMM_STORE_SHARED(ht->t.resize_initiated, 1);
- }
-}
-
-static
-void cds_lfht_resize_lazy_count(struct cds_lfht *ht, unsigned long size,
- unsigned long count)
-{
- struct rcu_resize_work *work;
-
- if (!(ht->flags & CDS_LFHT_AUTO_RESIZE))
- return;
- resize_target_update_count(ht, count);
- /* Store resize_target before read resize_initiated */
- cmm_smp_mb();
- if (!CMM_LOAD_SHARED(ht->t.resize_initiated)) {
- uatomic_inc(&ht->in_progress_resize);
- cmm_smp_mb(); /* increment resize count before load destroy */
- if (CMM_LOAD_SHARED(ht->in_progress_destroy)) {
- uatomic_dec(&ht->in_progress_resize);
- return;
- }
- work = malloc(sizeof(*work));
- work->ht = ht;
- ht->cds_lfht_call_rcu(&work->head, do_resize_cb);
- CMM_STORE_SHARED(ht->t.resize_initiated, 1);
- }
-}
+++ /dev/null
-#ifndef _URCU_RCULFHASH_H
-#define _URCU_RCULFHASH_H
-
-/*
- * urcu/rculfhash.h
- *
- * Userspace RCU library - Lock-Free RCU Hash Table
- *
- * Copyright 2011 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
- *
- * This library is free software; you can redistribute it and/or
- * modify it under the terms of the GNU Lesser General Public
- * License as published by the Free Software Foundation; either
- * version 2.1 of the License, or (at your option) any later version.
- *
- * This library 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
- * Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public
- * License along with this library; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- *
- * Include this file _after_ including your URCU flavor.
- */
-
-#include <stdint.h>
-#include <urcu-call-rcu.h>
-
-#ifdef __cplusplus
-extern "C" {
-#endif
-
-/*
- * struct cds_lfht_node and struct _cds_lfht_node should be aligned on
- * 4-bytes boundaries because the two lower bits are used as flags.
- */
-
-/*
- * _cds_lfht_node: Contains the internal pointers and reverse-hash
- * value required for lookup and traversal of the hash table.
- */
-struct _cds_lfht_node {
- struct cds_lfht_node *next; /* ptr | DUMMY_FLAG | REMOVED_FLAG */
- unsigned long reverse_hash;
-} __attribute__((aligned(4)));
-
-/*
- * cds_lfht_node: Contains the full key and length required to check for
- * an actual match, and also contains an rcu_head structure that is used
- * by RCU to track a node through a given RCU grace period. There is an
- * instance of _cds_lfht_node enclosed as a field within each
- * _cds_lfht_node structure.
- *
- * struct cds_lfht_node can be embedded into a structure (as a field).
- * caa_container_of() can be used to get the structure from the struct
- * cds_lfht_node after a lookup.
- */
-struct cds_lfht_node {
- /* cache-hot for iteration */
- struct _cds_lfht_node p; /* needs to be first field */
- void *key;
- unsigned int key_len;
- /* cache-cold for iteration */
- struct rcu_head head;
-};
-
-/* cds_lfht_iter: Used to track state while traversing a hash chain. */
-struct cds_lfht_iter {
- struct cds_lfht_node *node, *next;
-};
-
-static inline
-struct cds_lfht_node *cds_lfht_iter_get_node(struct cds_lfht_iter *iter)
-{
- return iter->node;
-}
-
-struct cds_lfht;
-
-/*
- * Caution !
- * Ensure reader and writer threads are registered as urcu readers.
- */
-
-typedef unsigned long (*cds_lfht_hash_fct)(void *key, size_t length,
- unsigned long seed);
-typedef unsigned long (*cds_lfht_compare_fct)(void *key1, size_t key1_len,
- void *key2, size_t key2_len);
-
-/*
- * cds_lfht_node_init - initialize a hash table node
- */
-static inline
-void cds_lfht_node_init(struct cds_lfht_node *node, void *key,
- size_t key_len)
-{
- node->key = key;
- node->key_len = key_len;
-}
-
-/*
- * Hash table creation flags.
- */
-enum {
- CDS_LFHT_AUTO_RESIZE = (1U << 0),
- CDS_LFHT_ACCOUNTING = (1U << 1),
-};
-
-/*
- * _cds_lfht_new - API used by cds_lfht_new wrapper. Do not use directly.
- */
-struct cds_lfht *_cds_lfht_new(cds_lfht_hash_fct hash_fct,
- cds_lfht_compare_fct compare_fct,
- unsigned long hash_seed,
- unsigned long init_size,
- unsigned long min_alloc_size,
- int flags,
- void (*cds_lfht_call_rcu)(struct rcu_head *head,
- void (*func)(struct rcu_head *head)),
- void (*cds_lfht_synchronize_rcu)(void),
- void (*cds_lfht_rcu_read_lock)(void),
- void (*cds_lfht_rcu_read_unlock)(void),
- void (*cds_lfht_rcu_thread_offline)(void),
- void (*cds_lfht_rcu_thread_online)(void),
- void (*cds_lfht_rcu_register_thread)(void),
- void (*cds_lfht_rcu_unregister_thread)(void),
- pthread_attr_t *attr);
-
-/*
- * cds_lfht_new - allocate a hash table.
- * @hash_fct: the hashing function.
- * @compare_fct: the key comparison function.
- * @hash_seed: the seed for hash function.
- * @init_size: number of nodes to allocate initially. Must be power of two.
- * @min_alloc_size: the smallest allocation size to use. Must be power of two.
- * @flags: hash table creation flags (can be combined with bitwise or: '|').
- * 0: no flags.
- * CDS_LFHT_AUTO_RESIZE: automatically resize hash table.
- * @attr: optional resize worker thread attributes. NULL for default.
- *
- * Return NULL on error.
- * Note: the RCU flavor must be already included before the hash table header.
- *
- * The programmer is responsible for ensuring that resize operation has a
- * priority equal to hash table updater threads. It should be performed by
- * specifying the appropriate priority in the pthread "attr" argument, and,
- * for CDS_LFHT_AUTO_RESIZE, by ensuring that call_rcu worker threads also have
- * this priority level. Having lower priority for call_rcu and resize threads
- * does not pose any correctness issue, but the resize operations could be
- * starved by updates, thus leading to long hash table bucket chains.
- * Threads calling this API need to be registered RCU read-side threads.
- */
-static inline
-struct cds_lfht *cds_lfht_new(cds_lfht_hash_fct hash_fct,
- cds_lfht_compare_fct compare_fct,
- unsigned long hash_seed,
- unsigned long init_size,
- unsigned long min_alloc_size,
- int flags,
- pthread_attr_t *attr)
-{
- return _cds_lfht_new(hash_fct, compare_fct, hash_seed,
- init_size, min_alloc_size, flags,
- call_rcu, synchronize_rcu, rcu_read_lock,
- rcu_read_unlock, rcu_thread_offline,
- rcu_thread_online, rcu_register_thread,
- rcu_unregister_thread, attr);
-}
-
-/*
- * cds_lfht_destroy - destroy a hash table.
- * @ht: the hash table to destroy.
- * @attr: (output) resize worker thread attributes, as received by cds_lfht_new.
- * The caller will typically want to free this pointer if dynamically
- * allocated. The attr point can be NULL if the caller does not
- * need to be informed of the value passed to cds_lfht_new().
- *
- * Return 0 on success, negative error value on error.
- * Threads calling this API need to be registered RCU read-side threads.
- */
-int cds_lfht_destroy(struct cds_lfht *ht, pthread_attr_t **attr);
-
-/*
- * cds_lfht_count_nodes - count the number of nodes in the hash table.
- * @ht: the hash table.
- * @split_count_before: Sample the node count split-counter before traversal.
- * @count: Traverse the hash table, count the number of nodes observed.
- * @removed: Number of logically removed nodes observed during traversal.
- * @split_count_after: Sample the node count split-counter after traversal.
- * Call with rcu_read_lock held.
- * Threads calling this API need to be registered RCU read-side threads.
- */
-void cds_lfht_count_nodes(struct cds_lfht *ht,
- long *split_count_before,
- unsigned long *count,
- unsigned long *removed,
- long *split_count_after);
-
-/*
- * cds_lfht_lookup - lookup a node by key.
- *
- * Output in "*iter". *iter->node set to NULL if not found.
- * Call with rcu_read_lock held.
- * Threads calling this API need to be registered RCU read-side threads.
- */
-void cds_lfht_lookup(struct cds_lfht *ht, void *key, size_t key_len,
- struct cds_lfht_iter *iter);
-
-/*
- * cds_lfht_next_duplicate - get the next item with same key (after a lookup).
- *
- * Uses an iterator initialized by a lookup.
- * Sets *iter-node to the following node with same key.
- * Sets *iter->node to NULL if no following node exists with same key.
- * RCU read-side lock must be held across cds_lfht_lookup and
- * cds_lfht_next calls, and also between cds_lfht_next calls using the
- * node returned by a previous cds_lfht_next.
- * Call with rcu_read_lock held.
- * Threads calling this API need to be registered RCU read-side threads.
- */
-void cds_lfht_next_duplicate(struct cds_lfht *ht, struct cds_lfht_iter *iter);
-
-/*
- * cds_lfht_first - get the first node in the table.
- *
- * Output in "*iter". *iter->node set to NULL if table is empty.
- * Call with rcu_read_lock held.
- * Threads calling this API need to be registered RCU read-side threads.
- */
-void cds_lfht_first(struct cds_lfht *ht, struct cds_lfht_iter *iter);
-
-/*
- * cds_lfht_next - get the next node in the table.
- *
- * Input/Output in "*iter". *iter->node set to NULL if *iter was
- * pointing to the last table node.
- * Call with rcu_read_lock held.
- * Threads calling this API need to be registered RCU read-side threads.
- */
-void cds_lfht_next(struct cds_lfht *ht, struct cds_lfht_iter *iter);
-
-/*
- * cds_lfht_add - add a node to the hash table.
- *
- * This function supports adding redundant keys into the table.
- * Call with rcu_read_lock held.
- * Threads calling this API need to be registered RCU read-side threads.
- */
-void cds_lfht_add(struct cds_lfht *ht, struct cds_lfht_node *node);
-
-/*
- * cds_lfht_add_unique - add a node to hash table, if key is not present.
- *
- * Return the node added upon success.
- * Return the unique node already present upon failure. If
- * cds_lfht_add_unique fails, the node passed as parameter should be
- * freed by the caller.
- * Call with rcu_read_lock held.
- * Threads calling this API need to be registered RCU read-side threads.
- *
- * The semantic of this function is that if only this function is used
- * to add keys into the table, no duplicated keys should ever be
- * observable in the table. The same guarantee apply for combination of
- * add_unique and add_replace (see below).
- */
-struct cds_lfht_node *cds_lfht_add_unique(struct cds_lfht *ht,
- struct cds_lfht_node *node);
-
-/*
- * cds_lfht_add_replace - replace or add a node within hash table.
- *
- * Return the node replaced upon success. If no node matching the key
- * was present, return NULL, which also means the operation succeeded.
- * This replacement operation should never fail.
- * Call with rcu_read_lock held.
- * Threads calling this API need to be registered RCU read-side threads.
- * After successful replacement, a grace period must be waited for before
- * freeing the memory reserved for the returned node.
- *
- * The semantic of replacement vs lookups is the following: if lookups
- * are performed between a key unique insertion and its removal, we
- * guarantee that the lookups and get next will always find exactly one
- * instance of the key if it is replaced concurrently with the lookups.
- *
- * Providing this semantic allows us to ensure that replacement-only
- * schemes will never generate duplicated keys. It also allows us to
- * guarantee that a combination of add_replace and add_unique updates
- * will never generate duplicated keys.
- */
-struct cds_lfht_node *cds_lfht_add_replace(struct cds_lfht *ht,
- struct cds_lfht_node *node);
-
-/*
- * cds_lfht_replace - replace a node pointer to by iter within hash table.
- *
- * Return 0 if replacement is successful, negative value otherwise.
- * Replacing a NULL old node or an already removed node will fail with a
- * negative value.
- * Old node can be looked up with cds_lfht_lookup and cds_lfht_next.
- * RCU read-side lock must be held between lookup and replacement.
- * Call with rcu_read_lock held.
- * Threads calling this API need to be registered RCU read-side threads.
- * After successful replacement, a grace period must be waited for before
- * freeing the memory reserved for the old node (which can be accessed
- * with cds_lfht_iter_get_node).
- *
- * The semantic of replacement vs lookups is the following: if lookups
- * are performed between a key unique insertion and its removal, we
- * guarantee that the lookups and get next will always find exactly one
- * instance of the key if it is replaced concurrently with the lookups.
- *
- * Providing this semantic allows us to ensure that replacement-only
- * schemes will never generate duplicated keys. It also allows us to
- * guarantee that a combination of add_replace and add_unique updates
- * will never generate duplicated keys.
- */
-int cds_lfht_replace(struct cds_lfht *ht, struct cds_lfht_iter *old_iter,
- struct cds_lfht_node *new_node);
-
-/*
- * cds_lfht_del - remove node pointed to by iterator from hash table.
- *
- * Return 0 if the node is successfully removed, negative value
- * otherwise.
- * Replacing a NULL node or an already removed node will fail with a
- * negative value.
- * Node can be looked up with cds_lfht_lookup and cds_lfht_next.
- * cds_lfht_iter_get_node.
- * RCU read-side lock must be held between lookup and removal.
- * Call with rcu_read_lock held.
- * Threads calling this API need to be registered RCU read-side threads.
- * After successful removal, a grace period must be waited for before
- * freeing the memory reserved for old node (which can be accessed with
- * cds_lfht_iter_get_node).
- */
-int cds_lfht_del(struct cds_lfht *ht, struct cds_lfht_iter *iter);
-
-/*
- * cds_lfht_resize - Force a hash table resize
- * @new_size: update to this hash table size.
- *
- * Threads calling this API need to be registered RCU read-side threads.
- */
-void cds_lfht_resize(struct cds_lfht *ht, unsigned long new_size);
-
-/*
- * Note: cds_lfht_for_each are safe for element removal during
- * iteration.
- */
-#define cds_lfht_for_each(ht, iter, node) \
- for (cds_lfht_first(ht, iter), \
- node = cds_lfht_iter_get_node(iter); \
- node != NULL; \
- cds_lfht_next(ht, iter), \
- node = cds_lfht_iter_get_node(iter))
-
-#define cds_lfht_for_each_duplicate(ht, match, hash, key, iter, node) \
- for (cds_lfht_lookup(ht, match, hash, key, iter), \
- node = cds_lfht_iter_get_node(iter); \
- node != NULL; \
- cds_lfht_next_duplicate(ht, match, key, iter), \
- node = cds_lfht_iter_get_node(iter))
-
-#define cds_lfht_for_each_entry(ht, iter, pos, member) \
- for (cds_lfht_first(ht, iter), \
- pos = caa_container_of(cds_lfht_iter_get_node(iter), \
- typeof(*(pos)), member); \
- &(pos)->member != NULL; \
- cds_lfht_next(ht, iter), \
- pos = caa_container_of(cds_lfht_iter_get_node(iter), \
- typeof(*(pos)), member))
-
-#define cds_lfht_for_each_entry_duplicate(ht, match, hash, key, \
- iter, pos, member) \
-for (cds_lfht_lookup(ht, match, hash, key, iter), \
- pos = caa_container_of(cds_lfht_iter_get_node(iter), \
- typeof(*(pos)), member); \
- &(pos)->member != NULL; \
- cds_lfht_next_duplicate(ht, match, key, iter), \
- pos = caa_container_of(cds_lfht_iter_get_node(iter), \
- typeof(*(pos)), member))
-
-#ifdef __cplusplus
-}
-#endif
-
-#endif /* _URCU_RCULFHASH_H */
lttnginclude_HEADERS = lttng/lttng.h lttng/lttng-kconsumer.h \
- lttng/lttng-ustconsumer.h lttng/lttng-consumer.h
+ lttng/lttng-ustconsumer.h lttng/lttng-consumer.h
noinst_HEADERS = lttngerr.h lttng-kernel.h lttng-consumerd.h lttng-share.h \
- lttng-sessiond-comm.h lttng-kernel-ctl.h
+ lttng-sessiond-comm.h lttng-kernel-ctl.h lttng-ht.h runas.h
--- /dev/null
+/*
+ * Copyright (C) 2011 - David Goulet <david.goulet@polymtl.ca>
+ *
+ * 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; only version 2 of the License.
+ *
+ * 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.
+ */
+
+#ifndef _LTT_HT_H
+#define _LTT_HT_H
+
+#include <urcu.h>
+#include "../liblttng-ht/rculfhash.h"
+#include "../liblttng-ht/rculfhash-internal.h"
+
+typedef unsigned long (*hash_fct)(void *_key, unsigned long seed);
+typedef cds_lfht_match_fct hash_match_fct;
+
+enum lttng_ht_type {
+ LTTNG_HT_TYPE_STRING,
+ LTTNG_HT_TYPE_ULONG,
+};
+
+struct lttng_ht {
+ struct cds_lfht *ht;
+ cds_lfht_match_fct match_fct;
+ hash_fct hash_fct;
+};
+
+struct lttng_ht_iter {
+ struct cds_lfht_iter iter;
+};
+
+struct lttng_ht_node_str {
+ char *key;
+ struct cds_lfht_node node;
+ struct rcu_head head;
+};
+
+struct lttng_ht_node_ulong {
+ unsigned long key;
+ struct cds_lfht_node node;
+ struct rcu_head head;
+};
+
+/* Hashtable new and destroy */
+extern struct lttng_ht *lttng_ht_new(unsigned long size, int type);
+extern void lttng_ht_destroy(struct lttng_ht *ht);
+
+/* Specialized node init and free functions */
+extern void lttng_ht_node_init_str(struct lttng_ht_node_str *node, char *key);
+extern void lttng_ht_node_init_ulong(struct lttng_ht_node_ulong *node,
+ unsigned long key);
+extern void lttng_ht_node_free_str(struct lttng_ht_node_str *node);
+extern void lttng_ht_node_free_ulong(struct lttng_ht_node_ulong *node);
+
+extern void lttng_ht_lookup(struct lttng_ht *ht, void *key,
+ struct lttng_ht_iter *iter);
+
+/* Specialized add unique functions */
+extern void lttng_ht_add_unique_str(struct lttng_ht *ht,
+ struct lttng_ht_node_str *node);
+extern void lttng_ht_add_unique_ulong(struct lttng_ht *ht,
+ struct lttng_ht_node_ulong *node);
+
+extern int lttng_ht_del(struct lttng_ht *ht, struct lttng_ht_iter *iter);
+
+extern void lttng_ht_get_first(struct lttng_ht *ht,
+ struct lttng_ht_iter *iter);
+extern void lttng_ht_get_next(struct lttng_ht *ht, struct lttng_ht_iter *iter);
+
+extern unsigned long lttng_ht_get_count(struct lttng_ht *ht);
+
+extern struct lttng_ht_node_str *lttng_ht_iter_get_node_str(
+ struct lttng_ht_iter *iter);
+extern struct lttng_ht_node_ulong *lttng_ht_iter_get_node_ulong(
+ struct lttng_ht_iter *iter);
+
+#endif /* _LTT_HT_H */
* These declarations should NOT be considered stable API.
*/
+#define _GNU_SOURCE
#include <limits.h>
#include <lttng/lttng.h>
+#include <sys/socket.h>
#define LTTNG_RUNDIR "/var/run/lttng"
#define LTTNG_HOME_RUNDIR "%s/.lttng"
LTTCOMM_NO_EVENT, /* No event found */
LTTCOMM_CONNECT_FAIL, /* Unable to connect to unix socket */
LTTCOMM_APP_NOT_FOUND, /* App not found in traceable app list */
+ LTTCOMM_EPERM, /* Permission denied */
LTTCOMM_KERN_NA, /* Kernel tracer unavalable */
LTTCOMM_KERN_EVENT_EXIST, /* Kernel event already exists */
LTTCOMM_KERN_SESS_FAIL, /* Kernel create session failed */
uint32_t state; /* enum lttcomm_consumer_fd_state */
enum lttng_event_output output; /* use splice or mmap to consume this fd */
uint64_t mmap_len;
+ uid_t uid; /* User ID owning the session */
+ gid_t gid; /* Group ID owning the session */
char path_name[PATH_MAX];
} stream;
} u;
extern ssize_t lttcomm_recv_unix_sock(int sock, void *buf, size_t len);
extern ssize_t lttcomm_send_unix_sock(int sock, void *buf, size_t len);
+
+extern ssize_t lttcomm_send_creds_unix_sock(int sock, void *buf, size_t len);
+extern ssize_t lttcomm_recv_creds_unix_sock(int sock, void *buf, size_t len,
+ struct ucred *creds);
+
extern const char *lttcomm_get_readable_code(enum lttcomm_return_code code);
+extern int lttcomm_setsockopt_creds_unix_sock(int sock);
#endif /* _LTTNG_SESSIOND_COMM_H */
#include <lttng/lttng.h>
/* Default size of a hash table */
-#define DEFAULT_HT_SIZE 32
+#define DEFAULT_HT_SIZE 4
/* Default channel attributes */
#define DEFAULT_CHANNEL_NAME "channel0"
#include <limits.h>
#include <poll.h>
+#include <unistd.h>
#include <urcu/list.h>
#include <lttng/lttng.h>
struct lttng_ust_lib_ring_buffer *buf;
int cpu;
int hangup_flush_done;
+ /* UID/GID of the user owning the session to which stream belongs */
+ uid_t uid;
+ gid_t gid;
};
/*
enum lttng_consumer_stream_state state,
uint64_t mmap_len,
enum lttng_event_output output,
- const char *path_name);
+ const char *path_name,
+ uid_t uid,
+ gid_t gid);
extern int consumer_add_stream(struct lttng_consumer_stream *stream);
extern void consumer_del_stream(struct lttng_consumer_stream *stream);
extern void consumer_change_stream_state(int stream_key,
--- /dev/null
+#ifndef _RUNAS_H
+#define _RUNAS_H
+
+/*
+ * Copyright (C) 2011 - David Goulet <david.goulet@polymtl.ca>
+ * Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ *
+ * 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; only verion 2
+ * of the License.
+ *
+ * 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.
+ */
+
+#include <unistd.h>
+
+int mkdir_recursive_run_as(const char *path, mode_t mode, uid_t uid, gid_t gid);
+int mkdir_run_as(const char *path, mode_t mode, uid_t uid, gid_t gid);
+int open_run_as(const char *path, int flags, mode_t mode, uid_t uid, gid_t gid);
+
+#endif /* _RUNAS_H */
enum lttng_consumer_stream_state state,
uint64_t mmap_len,
enum lttng_event_output output,
- const char *path_name)
+ const char *path_name,
+ uid_t uid,
+ gid_t gid)
{
struct lttng_consumer_stream *stream;
int ret;
stream->mmap_len = mmap_len;
stream->mmap_base = NULL;
stream->output = output;
+ stream->uid = uid;
+ stream->gid = gid;
strncpy(stream->path_name, path_name, PATH_MAX - 1);
stream->path_name[PATH_MAX - 1] = '\0';
--- /dev/null
+AM_CPPFLAGS = -I$(top_srcdir)/include
+
+noinst_LTLIBRARIES = liblttng-ht.la
+
+liblttng_ht_la_SOURCES = lttng-ht.c \
+ utils.c utils.h \
+ rculfhash-internal.h urcu-flavor.h \
+ rculfhash.h rculfhash.c \
+ rculfhash-mm-chunk.c \
+ rculfhash-mm-mmap.c \
+ rculfhash-mm-order.c
+
+liblttng_ht_la_LIBADD = -lurcu-common -lurcu
--- /dev/null
+/*
+ * Copyright (C) 2011 - David Goulet <david.goulet@polymtl.ca>
+ *
+ * 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; only version 2 of the License.
+ *
+ * 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.
+ */
+
+#define _GNU_SOURCE
+#include <assert.h>
+#include <string.h>
+#include <urcu.h>
+#include <urcu/compiler.h>
+
+#include <lttng-ht.h>
+#include <lttng-share.h>
+#include <lttngerr.h>
+
+#include "utils.h"
+
+#define HASH_SEED 0x42UL /* The answer to life */
+
+static unsigned long min_hash_alloc_size = 1;
+static unsigned long max_hash_buckets_size = (1UL << 20);
+
+/*
+ * Match function for string node.
+ */
+static int match_str(struct cds_lfht_node *node, const void *key)
+{
+ struct lttng_ht_node_str *match_node =
+ caa_container_of(node, struct lttng_ht_node_str, node);
+
+ return hash_match_key_str(match_node->key, (void *) key);
+}
+
+/*
+ * Match function for ulong node.
+ */
+static int match_ulong(struct cds_lfht_node *node, const void *key)
+{
+ struct lttng_ht_node_ulong *match_node =
+ caa_container_of(node, struct lttng_ht_node_ulong, node);
+
+ return hash_match_key_ulong((void *) match_node->key, (void *) key);
+}
+
+/*
+ * Return an allocated lttng hashtable.
+ */
+struct lttng_ht *lttng_ht_new(unsigned long size, int type)
+{
+ struct lttng_ht *ht;
+
+ /* Test size */
+ size != 0 ? : (size = DEFAULT_HT_SIZE);
+
+ ht = zmalloc(sizeof(*ht));
+ if (ht == NULL) {
+ PERROR("zmalloc lttng_ht");
+ goto error;
+ }
+
+ ht->ht = cds_lfht_new(size, min_hash_alloc_size, max_hash_buckets_size,
+ CDS_LFHT_AUTO_RESIZE, NULL);
+ /*
+ * There is already an assert in the RCU hashtable code so if the ht is
+ * NULL here there is a *huge* problem.
+ */
+ assert(ht->ht);
+
+ switch (type) {
+ case LTTNG_HT_TYPE_STRING:
+ ht->match_fct = match_str;
+ ht->hash_fct = hash_key_str;
+ break;
+ case LTTNG_HT_TYPE_ULONG:
+ ht->match_fct = match_ulong;
+ ht->hash_fct = hash_key_ulong;
+ break;
+ default:
+ ERR("Unknown lttng hashtable type %d", type);
+ goto error;
+ }
+
+ DBG3("Created hashtable size %lu at %p of type %d", size, ht->ht, type);
+
+ return ht;
+
+error:
+ return NULL;
+}
+
+/*
+ * Free a lttng hashtable.
+ */
+void lttng_ht_destroy(struct lttng_ht *ht)
+{
+ int ret;
+
+ ret = cds_lfht_destroy(ht->ht, NULL);
+ assert(!ret);
+}
+
+/*
+ * Init lttng ht node string.
+ */
+void lttng_ht_node_init_str(struct lttng_ht_node_str *node, char *key)
+{
+ assert(node);
+
+ node->key = key;
+ cds_lfht_node_init(&node->node);
+}
+
+/*
+ * Init lttng ht node unsigned long.
+ */
+void lttng_ht_node_init_ulong(struct lttng_ht_node_ulong *node,
+ unsigned long key)
+{
+ assert(node);
+
+ node->key = key;
+ cds_lfht_node_init(&node->node);
+}
+
+/*
+ * Free lttng ht node string.
+ */
+void lttng_ht_node_free_str(struct lttng_ht_node_str *node)
+{
+ assert(node);
+ free(node);
+}
+
+/*
+ * Free lttng ht node unsigned long.
+ */
+void lttng_ht_node_free_ulong(struct lttng_ht_node_ulong *node)
+{
+ assert(node);
+ free(node);
+}
+
+/*
+ * Lookup function in hashtable.
+ */
+void lttng_ht_lookup(struct lttng_ht *ht, void *key,
+ struct lttng_ht_iter *iter)
+{
+ assert(ht);
+ assert(ht->ht);
+ assert(key);
+
+ cds_lfht_lookup(ht->ht, ht->hash_fct(key, HASH_SEED),
+ ht->match_fct, key, &iter->iter);
+}
+
+/*
+ * Add unique string node to hashtable.
+ */
+void lttng_ht_add_unique_str(struct lttng_ht *ht,
+ struct lttng_ht_node_str *node)
+{
+ struct cds_lfht_node *node_ptr;
+ assert(ht);
+ assert(ht->ht);
+ assert(node);
+
+ node_ptr = cds_lfht_add_unique(ht->ht, ht->hash_fct(node->key, HASH_SEED),
+ ht->match_fct, node->key, &node->node);
+ assert(node_ptr == &node->node);
+}
+
+/*
+ * Add unique unsigned long node to hashtable.
+ */
+void lttng_ht_add_unique_ulong(struct lttng_ht *ht,
+ struct lttng_ht_node_ulong *node)
+{
+ struct cds_lfht_node *node_ptr;
+ assert(ht);
+ assert(ht->ht);
+ assert(node);
+
+ node_ptr = cds_lfht_add_unique(ht->ht,
+ ht->hash_fct((void *) node->key, HASH_SEED), ht->match_fct,
+ (void *) node->key, &node->node);
+ assert(node_ptr == &node->node);
+}
+
+/*
+ * Delete node from hashtable.
+ */
+int lttng_ht_del(struct lttng_ht *ht, struct lttng_ht_iter *iter)
+{
+ assert(ht);
+ assert(ht->ht);
+ assert(iter);
+
+ return cds_lfht_del(ht->ht, iter->iter.node);
+}
+
+/*
+ * Get first node in the hashtable.
+ */
+void lttng_ht_get_first(struct lttng_ht *ht, struct lttng_ht_iter *iter)
+{
+ assert(ht);
+ assert(ht->ht);
+ assert(iter);
+
+ cds_lfht_first(ht->ht, &iter->iter);
+}
+
+/*
+ * Get next node in the hashtable.
+ */
+void lttng_ht_get_next(struct lttng_ht *ht, struct lttng_ht_iter *iter)
+{
+ assert(ht);
+ assert(ht->ht);
+ assert(iter);
+
+ cds_lfht_next(ht->ht, &iter->iter);
+}
+
+/*
+ * Return the number of nodes in the hashtable.
+ */
+unsigned long lttng_ht_get_count(struct lttng_ht *ht)
+{
+ long scb, sca;
+ unsigned long count;
+
+ assert(ht);
+ assert(ht->ht);
+
+ cds_lfht_count_nodes(ht->ht, &scb, &count, &sca);
+
+ return count;
+}
+
+/*
+ * Return lttng ht string node from iterator.
+ */
+struct lttng_ht_node_str *lttng_ht_iter_get_node_str(
+ struct lttng_ht_iter *iter)
+{
+ struct cds_lfht_node *node;
+
+ assert(iter);
+ node = cds_lfht_iter_get_node(&iter->iter);
+ if (!node) {
+ return NULL;
+ }
+ return caa_container_of(node, struct lttng_ht_node_str, node);
+}
+
+/*
+ * Return lttng ht unsigned long node from iterator.
+ */
+struct lttng_ht_node_ulong *lttng_ht_iter_get_node_ulong(
+ struct lttng_ht_iter *iter)
+{
+ struct cds_lfht_node *node;
+
+ assert(iter);
+ node = cds_lfht_iter_get_node(&iter->iter);
+ if (!node) {
+ return NULL;
+ }
+ return caa_container_of(node, struct lttng_ht_node_ulong, node);
+}
--- /dev/null
+#ifndef _URCU_RCULFHASH_INTERNAL_H
+#define _URCU_RCULFHASH_INTERNAL_H
+
+/*
+ * urcu/rculfhash-internal.h
+ *
+ * Internal header for Lock-Free RCU Hash Table
+ *
+ * Copyright 2011 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ * Copyright 2011 - Lai Jiangshan <laijs@cn.fujitsu.com>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library 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
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "rculfhash.h"
+
+#ifdef DEBUG
+#define dbg_printf(fmt, args...) printf("[debug rculfhash] " fmt, ## args)
+#else
+#define dbg_printf(fmt, args...)
+#endif
+
+#if (CAA_BITS_PER_LONG == 32)
+#define MAX_TABLE_ORDER 32
+#else
+#define MAX_TABLE_ORDER 64
+#endif
+
+#define MAX_CHUNK_TABLE (1UL << 10)
+
+#ifndef min
+#define min(a, b) ((a) < (b) ? (a) : (b))
+#endif
+
+#ifndef max
+#define max(a, b) ((a) > (b) ? (a) : (b))
+#endif
+
+struct ht_items_count;
+
+/*
+ * cds_lfht: Top-level data structure representing a lock-free hash
+ * table. Defined in the implementation file to make it be an opaque
+ * cookie to users.
+ *
+ * The fields used in fast-paths are placed near the end of the
+ * structure, because we need to have a variable-sized union to contain
+ * the mm plugin fields, which are used in the fast path.
+ */
+struct cds_lfht {
+ /* Initial configuration items */
+ unsigned long max_nr_buckets;
+ const struct cds_lfht_mm_type *mm; /* memory management plugin */
+ const struct rcu_flavor_struct *flavor; /* RCU flavor */
+
+ long count; /* global approximate item count */
+
+ /*
+ * We need to put the work threads offline (QSBR) when taking this
+ * mutex, because we use synchronize_rcu within this mutex critical
+ * section, which waits on read-side critical sections, and could
+ * therefore cause grace-period deadlock if we hold off RCU G.P.
+ * completion.
+ */
+ pthread_mutex_t resize_mutex; /* resize mutex: add/del mutex */
+ pthread_attr_t *resize_attr; /* Resize threads attributes */
+ unsigned int in_progress_resize, in_progress_destroy;
+ unsigned long resize_target;
+ int resize_initiated;
+
+ /*
+ * Variables needed for add and remove fast-paths.
+ */
+ int flags;
+ unsigned long min_alloc_buckets_order;
+ unsigned long min_nr_alloc_buckets;
+ struct ht_items_count *split_count; /* split item count */
+
+ /*
+ * Variables needed for the lookup, add and remove fast-paths.
+ */
+ unsigned long size; /* always a power of 2, shared (RCU) */
+ /*
+ * bucket_at pointer is kept here to skip the extra level of
+ * dereference needed to get to "mm" (this is a fast-path).
+ */
+ struct cds_lfht_node *(*bucket_at)(struct cds_lfht *ht,
+ unsigned long index);
+ /*
+ * Dynamic length "tbl_chunk" needs to be at the end of
+ * cds_lfht.
+ */
+ union {
+ /*
+ * Contains the per order-index-level bucket node table.
+ * The size of each bucket node table is half the number
+ * of hashes contained in this order (except for order 0).
+ * The minimum allocation buckets size parameter allows
+ * combining the bucket node arrays of the lowermost
+ * levels to improve cache locality for small index orders.
+ */
+ struct cds_lfht_node *tbl_order[MAX_TABLE_ORDER];
+
+ /*
+ * Contains the bucket node chunks. The size of each
+ * bucket node chunk is ->min_alloc_size (we avoid to
+ * allocate chunks with different size). Chunks improve
+ * cache locality for small index orders, and are more
+ * friendly with environments where allocation of large
+ * contiguous memory areas is challenging due to memory
+ * fragmentation concerns or inability to use virtual
+ * memory addressing.
+ */
+ struct cds_lfht_node *tbl_chunk[0];
+
+ /*
+ * Memory mapping with room for all possible buckets.
+ * Their memory is allocated when needed.
+ */
+ struct cds_lfht_node *tbl_mmap;
+ };
+ /*
+ * End of variables needed for the lookup, add and remove
+ * fast-paths.
+ */
+};
+
+extern unsigned int cds_lfht_fls_ulong(unsigned long x);
+extern int cds_lfht_get_count_order_ulong(unsigned long x);
+
+#ifdef POISON_FREE
+#define poison_free(ptr) \
+ do { \
+ if (ptr) { \
+ memset(ptr, 0x42, sizeof(*(ptr))); \
+ free(ptr); \
+ } \
+ } while (0)
+#else
+#define poison_free(ptr) free(ptr)
+#endif
+
+static inline
+struct cds_lfht *__default_alloc_cds_lfht(
+ const struct cds_lfht_mm_type *mm,
+ unsigned long cds_lfht_size,
+ unsigned long min_nr_alloc_buckets,
+ unsigned long max_nr_buckets)
+{
+ struct cds_lfht *ht;
+
+ ht = calloc(1, cds_lfht_size);
+ assert(ht);
+
+ ht->mm = mm;
+ ht->bucket_at = mm->bucket_at;
+ ht->min_nr_alloc_buckets = min_nr_alloc_buckets;
+ ht->min_alloc_buckets_order =
+ cds_lfht_get_count_order_ulong(min_nr_alloc_buckets);
+ ht->max_nr_buckets = max_nr_buckets;
+
+ return ht;
+}
+
+#endif /* _URCU_RCULFHASH_INTERNAL_H */
--- /dev/null
+/*
+ * rculfhash-mm-chunk.c
+ *
+ * Chunk based memory management for Lock-Free RCU Hash Table
+ *
+ * Copyright 2011 - Lai Jiangshan <laijs@cn.fujitsu.com>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library 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
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <stddef.h>
+#include "rculfhash-internal.h"
+
+static
+void cds_lfht_alloc_bucket_table(struct cds_lfht *ht, unsigned long order)
+{
+ if (order == 0) {
+ ht->tbl_chunk[0] = calloc(ht->min_nr_alloc_buckets,
+ sizeof(struct cds_lfht_node));
+ assert(ht->tbl_chunk[0]);
+ } else if (order > ht->min_alloc_buckets_order) {
+ unsigned long i, len = 1UL << (order - 1 - ht->min_alloc_buckets_order);
+
+ for (i = len; i < 2 * len; i++) {
+ ht->tbl_chunk[i] = calloc(ht->min_nr_alloc_buckets,
+ sizeof(struct cds_lfht_node));
+ assert(ht->tbl_chunk[i]);
+ }
+ }
+ /* Nothing to do for 0 < order && order <= ht->min_alloc_buckets_order */
+}
+
+/*
+ * cds_lfht_free_bucket_table() should be called with decreasing order.
+ * When cds_lfht_free_bucket_table(0) is called, it means the whole
+ * lfht is destroyed.
+ */
+static
+void cds_lfht_free_bucket_table(struct cds_lfht *ht, unsigned long order)
+{
+ if (order == 0)
+ poison_free(ht->tbl_chunk[0]);
+ else if (order > ht->min_alloc_buckets_order) {
+ unsigned long i, len = 1UL << (order - 1 - ht->min_alloc_buckets_order);
+
+ for (i = len; i < 2 * len; i++)
+ poison_free(ht->tbl_chunk[i]);
+ }
+ /* Nothing to do for 0 < order && order <= ht->min_alloc_buckets_order */
+}
+
+static
+struct cds_lfht_node *bucket_at(struct cds_lfht *ht, unsigned long index)
+{
+ unsigned long chunk, offset;
+
+ chunk = index >> ht->min_alloc_buckets_order;
+ offset = index & (ht->min_nr_alloc_buckets - 1);
+ return &ht->tbl_chunk[chunk][offset];
+}
+
+static
+struct cds_lfht *alloc_cds_lfht(unsigned long min_nr_alloc_buckets,
+ unsigned long max_nr_buckets)
+{
+ unsigned long nr_chunks, cds_lfht_size;
+
+ min_nr_alloc_buckets = max(min_nr_alloc_buckets,
+ max_nr_buckets / MAX_CHUNK_TABLE);
+ nr_chunks = max_nr_buckets / min_nr_alloc_buckets;
+ cds_lfht_size = offsetof(struct cds_lfht, tbl_chunk) +
+ sizeof(struct cds_lfht_node *) * nr_chunks;
+ cds_lfht_size = max(cds_lfht_size, sizeof(struct cds_lfht));
+
+ return __default_alloc_cds_lfht(
+ &cds_lfht_mm_chunk, cds_lfht_size,
+ min_nr_alloc_buckets, max_nr_buckets);
+}
+
+const struct cds_lfht_mm_type cds_lfht_mm_chunk = {
+ .alloc_cds_lfht = alloc_cds_lfht,
+ .alloc_bucket_table = cds_lfht_alloc_bucket_table,
+ .free_bucket_table = cds_lfht_free_bucket_table,
+ .bucket_at = bucket_at,
+};
--- /dev/null
+/*
+ * rculfhash-mm-mmap.c
+ *
+ * mmap/reservation based memory management for Lock-Free RCU Hash Table
+ *
+ * Copyright 2011 - Lai Jiangshan <laijs@cn.fujitsu.com>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library 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
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <unistd.h>
+#include <sys/mman.h>
+#include "rculfhash-internal.h"
+
+/* reserve inaccessible memory space without allocation any memory */
+static void *memory_map(size_t length)
+{
+ void *ret = mmap(NULL, length, PROT_NONE,
+ MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+
+ assert(ret != MAP_FAILED);
+ return ret;
+}
+
+static void memory_unmap(void *ptr, size_t length)
+{
+ int ret __attribute__((unused));
+
+ ret = munmap(ptr, length);
+
+ assert(ret == 0);
+}
+
+static void memory_populate(void *ptr, size_t length)
+{
+ void *ret __attribute__((unused));
+
+ ret = mmap(ptr, length, PROT_READ | PROT_WRITE,
+ MAP_FIXED | MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+
+ assert(ret == ptr);
+}
+
+/*
+ * Discard garbage memory and avoid system save it when try to swap it out.
+ * Make it still reserved, inaccessible.
+ */
+static void memory_discard(void *ptr, size_t length)
+{
+ void *ret __attribute__((unused));
+
+ ret = mmap(ptr, length, PROT_NONE,
+ MAP_FIXED | MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+
+ assert(ret == ptr);
+}
+
+static
+void cds_lfht_alloc_bucket_table(struct cds_lfht *ht, unsigned long order)
+{
+ if (order == 0) {
+ if (ht->min_nr_alloc_buckets == ht->max_nr_buckets) {
+ /* small table */
+ ht->tbl_mmap = calloc(ht->max_nr_buckets,
+ sizeof(*ht->tbl_mmap));
+ assert(ht->tbl_mmap);
+ return;
+ }
+ /* large table */
+ ht->tbl_mmap = memory_map(ht->max_nr_buckets
+ * sizeof(*ht->tbl_mmap));
+ memory_populate(ht->tbl_mmap,
+ ht->min_nr_alloc_buckets * sizeof(*ht->tbl_mmap));
+ } else if (order > ht->min_alloc_buckets_order) {
+ /* large table */
+ unsigned long len = 1UL << (order - 1);
+
+ assert(ht->min_nr_alloc_buckets < ht->max_nr_buckets);
+ memory_populate(ht->tbl_mmap + len,
+ len * sizeof(*ht->tbl_mmap));
+ }
+ /* Nothing to do for 0 < order && order <= ht->min_alloc_buckets_order */
+}
+
+/*
+ * cds_lfht_free_bucket_table() should be called with decreasing order.
+ * When cds_lfht_free_bucket_table(0) is called, it means the whole
+ * lfht is destroyed.
+ */
+static
+void cds_lfht_free_bucket_table(struct cds_lfht *ht, unsigned long order)
+{
+ if (order == 0) {
+ if (ht->min_nr_alloc_buckets == ht->max_nr_buckets) {
+ /* small table */
+ poison_free(ht->tbl_mmap);
+ return;
+ }
+ /* large table */
+ memory_unmap(ht->tbl_mmap,
+ ht->max_nr_buckets * sizeof(*ht->tbl_mmap));
+ } else if (order > ht->min_alloc_buckets_order) {
+ /* large table */
+ unsigned long len = 1UL << (order - 1);
+
+ assert(ht->min_nr_alloc_buckets < ht->max_nr_buckets);
+ memory_discard(ht->tbl_mmap + len, len * sizeof(*ht->tbl_mmap));
+ }
+ /* Nothing to do for 0 < order && order <= ht->min_alloc_buckets_order */
+}
+
+static
+struct cds_lfht_node *bucket_at(struct cds_lfht *ht, unsigned long index)
+{
+ return &ht->tbl_mmap[index];
+}
+
+static
+struct cds_lfht *alloc_cds_lfht(unsigned long min_nr_alloc_buckets,
+ unsigned long max_nr_buckets)
+{
+ unsigned long page_bucket_size;
+
+ page_bucket_size = getpagesize() / sizeof(struct cds_lfht_node);
+ if (max_nr_buckets <= page_bucket_size) {
+ /* small table */
+ min_nr_alloc_buckets = max_nr_buckets;
+ } else {
+ /* large table */
+ min_nr_alloc_buckets = max(min_nr_alloc_buckets,
+ page_bucket_size);
+ }
+
+ return __default_alloc_cds_lfht(
+ &cds_lfht_mm_mmap, sizeof(struct cds_lfht),
+ min_nr_alloc_buckets, max_nr_buckets);
+}
+
+const struct cds_lfht_mm_type cds_lfht_mm_mmap = {
+ .alloc_cds_lfht = alloc_cds_lfht,
+ .alloc_bucket_table = cds_lfht_alloc_bucket_table,
+ .free_bucket_table = cds_lfht_free_bucket_table,
+ .bucket_at = bucket_at,
+};
--- /dev/null
+/*
+ * rculfhash-mm-order.c
+ *
+ * Order based memory management for Lock-Free RCU Hash Table
+ *
+ * Copyright 2011 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ * Copyright 2011 - Lai Jiangshan <laijs@cn.fujitsu.com>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library 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
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "rculfhash-internal.h"
+
+static
+void cds_lfht_alloc_bucket_table(struct cds_lfht *ht, unsigned long order)
+{
+ if (order == 0) {
+ ht->tbl_order[0] = calloc(ht->min_nr_alloc_buckets,
+ sizeof(struct cds_lfht_node));
+ assert(ht->tbl_order[0]);
+ } else if (order > ht->min_alloc_buckets_order) {
+ ht->tbl_order[order] = calloc(1UL << (order -1),
+ sizeof(struct cds_lfht_node));
+ assert(ht->tbl_order[order]);
+ }
+ /* Nothing to do for 0 < order && order <= ht->min_alloc_buckets_order */
+}
+
+/*
+ * cds_lfht_free_bucket_table() should be called with decreasing order.
+ * When cds_lfht_free_bucket_table(0) is called, it means the whole
+ * lfht is destroyed.
+ */
+static
+void cds_lfht_free_bucket_table(struct cds_lfht *ht, unsigned long order)
+{
+ if (order == 0)
+ poison_free(ht->tbl_order[0]);
+ else if (order > ht->min_alloc_buckets_order)
+ poison_free(ht->tbl_order[order]);
+ /* Nothing to do for 0 < order && order <= ht->min_alloc_buckets_order */
+}
+
+static
+struct cds_lfht_node *bucket_at(struct cds_lfht *ht, unsigned long index)
+{
+ unsigned long order;
+
+ if (index < ht->min_nr_alloc_buckets) {
+ dbg_printf("bucket index %lu order 0 aridx 0\n", index);
+ return &ht->tbl_order[0][index];
+ }
+ /*
+ * equivalent to cds_lfht_get_count_order_ulong(index + 1), but
+ * optimizes away the non-existing 0 special-case for
+ * cds_lfht_get_count_order_ulong.
+ */
+ order = cds_lfht_fls_ulong(index);
+ dbg_printf("bucket index %lu order %lu aridx %lu\n",
+ index, order, index & ((1UL << (order - 1)) - 1));
+ return &ht->tbl_order[order][index & ((1UL << (order - 1)) - 1)];
+}
+
+static
+struct cds_lfht *alloc_cds_lfht(unsigned long min_nr_alloc_buckets,
+ unsigned long max_nr_buckets)
+{
+ return __default_alloc_cds_lfht(
+ &cds_lfht_mm_order, sizeof(struct cds_lfht),
+ min_nr_alloc_buckets, max_nr_buckets);
+}
+
+const struct cds_lfht_mm_type cds_lfht_mm_order = {
+ .alloc_cds_lfht = alloc_cds_lfht,
+ .alloc_bucket_table = cds_lfht_alloc_bucket_table,
+ .free_bucket_table = cds_lfht_free_bucket_table,
+ .bucket_at = bucket_at,
+};
--- /dev/null
+/*
+ * rculfhash.c
+ *
+ * Userspace RCU library - Lock-Free Resizable RCU Hash Table
+ *
+ * Copyright 2010-2011 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ * Copyright 2011 - Lai Jiangshan <laijs@cn.fujitsu.com>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library 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
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/*
+ * Based on the following articles:
+ * - Ori Shalev and Nir Shavit. Split-ordered lists: Lock-free
+ * extensible hash tables. J. ACM 53, 3 (May 2006), 379-405.
+ * - Michael, M. M. High performance dynamic lock-free hash tables
+ * and list-based sets. In Proceedings of the fourteenth annual ACM
+ * symposium on Parallel algorithms and architectures, ACM Press,
+ * (2002), 73-82.
+ *
+ * Some specificities of this Lock-Free Resizable RCU Hash Table
+ * implementation:
+ *
+ * - RCU read-side critical section allows readers to perform hash
+ * table lookups and use the returned objects safely by delaying
+ * memory reclaim of a grace period.
+ * - Add and remove operations are lock-free, and do not need to
+ * allocate memory. They need to be executed within RCU read-side
+ * critical section to ensure the objects they read are valid and to
+ * deal with the cmpxchg ABA problem.
+ * - add and add_unique operations are supported. add_unique checks if
+ * the node key already exists in the hash table. It ensures no key
+ * duplicata exists.
+ * - The resize operation executes concurrently with add/remove/lookup.
+ * - Hash table nodes are contained within a split-ordered list. This
+ * list is ordered by incrementing reversed-bits-hash value.
+ * - An index of bucket nodes is kept. These bucket nodes are the hash
+ * table "buckets", and they are also chained together in the
+ * split-ordered list, which allows recursive expansion.
+ * - The resize operation for small tables only allows expanding the hash table.
+ * It is triggered automatically by detecting long chains in the add
+ * operation.
+ * - The resize operation for larger tables (and available through an
+ * API) allows both expanding and shrinking the hash table.
+ * - Split-counters are used to keep track of the number of
+ * nodes within the hash table for automatic resize triggering.
+ * - Resize operation initiated by long chain detection is executed by a
+ * call_rcu thread, which keeps lock-freedom of add and remove.
+ * - Resize operations are protected by a mutex.
+ * - The removal operation is split in two parts: first, a "removed"
+ * flag is set in the next pointer within the node to remove. Then,
+ * a "garbage collection" is performed in the bucket containing the
+ * removed node (from the start of the bucket up to the removed node).
+ * All encountered nodes with "removed" flag set in their next
+ * pointers are removed from the linked-list. If the cmpxchg used for
+ * removal fails (due to concurrent garbage-collection or concurrent
+ * add), we retry from the beginning of the bucket. This ensures that
+ * the node with "removed" flag set is removed from the hash table
+ * (not visible to lookups anymore) before the RCU read-side critical
+ * section held across removal ends. Furthermore, this ensures that
+ * the node with "removed" flag set is removed from the linked-list
+ * before its memory is reclaimed. Only the thread which removal
+ * successfully set the "removed" flag (with a cmpxchg) into a node's
+ * next pointer is considered to have succeeded its removal (and thus
+ * owns the node to reclaim). Because we garbage-collect starting from
+ * an invariant node (the start-of-bucket bucket node) up to the
+ * "removed" node (or find a reverse-hash that is higher), we are sure
+ * that a successful traversal of the chain leads to a chain that is
+ * present in the linked-list (the start node is never removed) and
+ * that is does not contain the "removed" node anymore, even if
+ * concurrent delete/add operations are changing the structure of the
+ * list concurrently.
+ * - The add operation performs gargage collection of buckets if it
+ * encounters nodes with removed flag set in the bucket where it wants
+ * to add its new node. This ensures lock-freedom of add operation by
+ * helping the remover unlink nodes from the list rather than to wait
+ * for it do to so.
+ * - A RCU "order table" indexed by log2(hash index) is copied and
+ * expanded by the resize operation. This order table allows finding
+ * the "bucket node" tables.
+ * - There is one bucket node table per hash index order. The size of
+ * each bucket node table is half the number of hashes contained in
+ * this order (except for order 0).
+ * - synchronzie_rcu is used to garbage-collect the old bucket node table.
+ * - The per-order bucket node tables contain a compact version of the
+ * hash table nodes. These tables are invariant after they are
+ * populated into the hash table.
+ *
+ * Bucket node tables:
+ *
+ * hash table hash table the last all bucket node tables
+ * order size bucket node 0 1 2 3 4 5 6(index)
+ * table size
+ * 0 1 1 1
+ * 1 2 1 1 1
+ * 2 4 2 1 1 2
+ * 3 8 4 1 1 2 4
+ * 4 16 8 1 1 2 4 8
+ * 5 32 16 1 1 2 4 8 16
+ * 6 64 32 1 1 2 4 8 16 32
+ *
+ * When growing/shrinking, we only focus on the last bucket node table
+ * which size is (!order ? 1 : (1 << (order -1))).
+ *
+ * Example for growing/shrinking:
+ * grow hash table from order 5 to 6: init the index=6 bucket node table
+ * shrink hash table from order 6 to 5: fini the index=6 bucket node table
+ *
+ * A bit of ascii art explanation:
+ *
+ * Order index is the off-by-one compare to the actual power of 2 because
+ * we use index 0 to deal with the 0 special-case.
+ *
+ * This shows the nodes for a small table ordered by reversed bits:
+ *
+ * bits reverse
+ * 0 000 000
+ * 4 100 001
+ * 2 010 010
+ * 6 110 011
+ * 1 001 100
+ * 5 101 101
+ * 3 011 110
+ * 7 111 111
+ *
+ * This shows the nodes in order of non-reversed bits, linked by
+ * reversed-bit order.
+ *
+ * order bits reverse
+ * 0 0 000 000
+ * 1 | 1 001 100 <-
+ * 2 | | 2 010 010 <- |
+ * | | | 3 011 110 | <- |
+ * 3 -> | | | 4 100 001 | |
+ * -> | | 5 101 101 |
+ * -> | 6 110 011
+ * -> 7 111 111
+ */
+
+#define _LGPL_SOURCE
+#include <stdlib.h>
+#include <errno.h>
+#include <assert.h>
+#include <stdio.h>
+#include <stdint.h>
+#include <string.h>
+
+#include "config.h"
+#include <urcu.h>
+#include <urcu-call-rcu.h>
+#include <urcu/arch.h>
+#include <urcu/uatomic.h>
+#include <urcu/compiler.h>
+#include <stdio.h>
+#include <pthread.h>
+
+#include "rculfhash.h"
+#include "rculfhash-internal.h"
+#include "urcu-flavor.h"
+
+/*
+ * Split-counters lazily update the global counter each 1024
+ * addition/removal. It automatically keeps track of resize required.
+ * We use the bucket length as indicator for need to expand for small
+ * tables and machines lacking per-cpu data suppport.
+ */
+#define COUNT_COMMIT_ORDER 10
+#define DEFAULT_SPLIT_COUNT_MASK 0xFUL
+#define CHAIN_LEN_TARGET 1
+#define CHAIN_LEN_RESIZE_THRESHOLD 3
+
+/*
+ * Define the minimum table size.
+ */
+#define MIN_TABLE_ORDER 0
+#define MIN_TABLE_SIZE (1UL << MIN_TABLE_ORDER)
+
+/*
+ * Minimum number of bucket nodes to touch per thread to parallelize grow/shrink.
+ */
+#define MIN_PARTITION_PER_THREAD_ORDER 12
+#define MIN_PARTITION_PER_THREAD (1UL << MIN_PARTITION_PER_THREAD_ORDER)
+
+/*
+ * The removed flag needs to be updated atomically with the pointer.
+ * It indicates that no node must attach to the node scheduled for
+ * removal, and that node garbage collection must be performed.
+ * The bucket flag does not require to be updated atomically with the
+ * pointer, but it is added as a pointer low bit flag to save space.
+ */
+#define REMOVED_FLAG (1UL << 0)
+#define BUCKET_FLAG (1UL << 1)
+#define REMOVAL_OWNER_FLAG (1UL << 2)
+#define FLAGS_MASK ((1UL << 3) - 1)
+
+/* Value of the end pointer. Should not interact with flags. */
+#define END_VALUE NULL
+
+DEFINE_RCU_FLAVOR(rcu_flavor);
+
+/*
+ * ht_items_count: Split-counters counting the number of node addition
+ * and removal in the table. Only used if the CDS_LFHT_ACCOUNTING flag
+ * is set at hash table creation.
+ *
+ * These are free-running counters, never reset to zero. They count the
+ * number of add/remove, and trigger every (1 << COUNT_COMMIT_ORDER)
+ * operations to update the global counter. We choose a power-of-2 value
+ * for the trigger to deal with 32 or 64-bit overflow of the counter.
+ */
+struct ht_items_count {
+ unsigned long add, del;
+} __attribute__((aligned(CAA_CACHE_LINE_SIZE)));
+
+/*
+ * rcu_resize_work: Contains arguments passed to RCU worker thread
+ * responsible for performing lazy resize.
+ */
+struct rcu_resize_work {
+ struct rcu_head head;
+ struct cds_lfht *ht;
+};
+
+/*
+ * partition_resize_work: Contains arguments passed to worker threads
+ * executing the hash table resize on partitions of the hash table
+ * assigned to each processor's worker thread.
+ */
+struct partition_resize_work {
+ pthread_t thread_id;
+ struct cds_lfht *ht;
+ unsigned long i, start, len;
+ void (*fct)(struct cds_lfht *ht, unsigned long i,
+ unsigned long start, unsigned long len);
+};
+
+/*
+ * Algorithm to reverse bits in a word by lookup table, extended to
+ * 64-bit words.
+ * Source:
+ * http://graphics.stanford.edu/~seander/bithacks.html#BitReverseTable
+ * Originally from Public Domain.
+ */
+
+static const uint8_t BitReverseTable256[256] =
+{
+#define R2(n) (n), (n) + 2*64, (n) + 1*64, (n) + 3*64
+#define R4(n) R2(n), R2((n) + 2*16), R2((n) + 1*16), R2((n) + 3*16)
+#define R6(n) R4(n), R4((n) + 2*4 ), R4((n) + 1*4 ), R4((n) + 3*4 )
+ R6(0), R6(2), R6(1), R6(3)
+};
+#undef R2
+#undef R4
+#undef R6
+
+static
+uint8_t bit_reverse_u8(uint8_t v)
+{
+ return BitReverseTable256[v];
+}
+
+static __attribute__((unused))
+uint32_t bit_reverse_u32(uint32_t v)
+{
+ return ((uint32_t) bit_reverse_u8(v) << 24) |
+ ((uint32_t) bit_reverse_u8(v >> 8) << 16) |
+ ((uint32_t) bit_reverse_u8(v >> 16) << 8) |
+ ((uint32_t) bit_reverse_u8(v >> 24));
+}
+
+static __attribute__((unused))
+uint64_t bit_reverse_u64(uint64_t v)
+{
+ return ((uint64_t) bit_reverse_u8(v) << 56) |
+ ((uint64_t) bit_reverse_u8(v >> 8) << 48) |
+ ((uint64_t) bit_reverse_u8(v >> 16) << 40) |
+ ((uint64_t) bit_reverse_u8(v >> 24) << 32) |
+ ((uint64_t) bit_reverse_u8(v >> 32) << 24) |
+ ((uint64_t) bit_reverse_u8(v >> 40) << 16) |
+ ((uint64_t) bit_reverse_u8(v >> 48) << 8) |
+ ((uint64_t) bit_reverse_u8(v >> 56));
+}
+
+static
+unsigned long bit_reverse_ulong(unsigned long v)
+{
+#if (CAA_BITS_PER_LONG == 32)
+ return bit_reverse_u32(v);
+#else
+ return bit_reverse_u64(v);
+#endif
+}
+
+/*
+ * fls: returns the position of the most significant bit.
+ * Returns 0 if no bit is set, else returns the position of the most
+ * significant bit (from 1 to 32 on 32-bit, from 1 to 64 on 64-bit).
+ */
+#if defined(__i386) || defined(__x86_64)
+static inline
+unsigned int fls_u32(uint32_t x)
+{
+ int r;
+
+ asm("bsrl %1,%0\n\t"
+ "jnz 1f\n\t"
+ "movl $-1,%0\n\t"
+ "1:\n\t"
+ : "=r" (r) : "rm" (x));
+ return r + 1;
+}
+#define HAS_FLS_U32
+#endif
+
+#if defined(__x86_64)
+static inline
+unsigned int fls_u64(uint64_t x)
+{
+ long r;
+
+ asm("bsrq %1,%0\n\t"
+ "jnz 1f\n\t"
+ "movq $-1,%0\n\t"
+ "1:\n\t"
+ : "=r" (r) : "rm" (x));
+ return r + 1;
+}
+#define HAS_FLS_U64
+#endif
+
+#ifndef HAS_FLS_U64
+static __attribute__((unused))
+unsigned int fls_u64(uint64_t x)
+{
+ unsigned int r = 64;
+
+ if (!x)
+ return 0;
+
+ if (!(x & 0xFFFFFFFF00000000ULL)) {
+ x <<= 32;
+ r -= 32;
+ }
+ if (!(x & 0xFFFF000000000000ULL)) {
+ x <<= 16;
+ r -= 16;
+ }
+ if (!(x & 0xFF00000000000000ULL)) {
+ x <<= 8;
+ r -= 8;
+ }
+ if (!(x & 0xF000000000000000ULL)) {
+ x <<= 4;
+ r -= 4;
+ }
+ if (!(x & 0xC000000000000000ULL)) {
+ x <<= 2;
+ r -= 2;
+ }
+ if (!(x & 0x8000000000000000ULL)) {
+ x <<= 1;
+ r -= 1;
+ }
+ return r;
+}
+#endif
+
+#ifndef HAS_FLS_U32
+static __attribute__((unused))
+unsigned int fls_u32(uint32_t x)
+{
+ unsigned int r = 32;
+
+ if (!x)
+ return 0;
+ if (!(x & 0xFFFF0000U)) {
+ x <<= 16;
+ r -= 16;
+ }
+ if (!(x & 0xFF000000U)) {
+ x <<= 8;
+ r -= 8;
+ }
+ if (!(x & 0xF0000000U)) {
+ x <<= 4;
+ r -= 4;
+ }
+ if (!(x & 0xC0000000U)) {
+ x <<= 2;
+ r -= 2;
+ }
+ if (!(x & 0x80000000U)) {
+ x <<= 1;
+ r -= 1;
+ }
+ return r;
+}
+#endif
+
+unsigned int cds_lfht_fls_ulong(unsigned long x)
+{
+#if (CAA_BITS_PER_LONG == 32)
+ return fls_u32(x);
+#else
+ return fls_u64(x);
+#endif
+}
+
+/*
+ * Return the minimum order for which x <= (1UL << order).
+ * Return -1 if x is 0.
+ */
+int cds_lfht_get_count_order_u32(uint32_t x)
+{
+ if (!x)
+ return -1;
+
+ return fls_u32(x - 1);
+}
+
+/*
+ * Return the minimum order for which x <= (1UL << order).
+ * Return -1 if x is 0.
+ */
+int cds_lfht_get_count_order_ulong(unsigned long x)
+{
+ if (!x)
+ return -1;
+
+ return cds_lfht_fls_ulong(x - 1);
+}
+
+static
+void cds_lfht_resize_lazy_grow(struct cds_lfht *ht, unsigned long size, int growth);
+
+static
+void cds_lfht_resize_lazy_count(struct cds_lfht *ht, unsigned long size,
+ unsigned long count);
+
+static long nr_cpus_mask = -1;
+static long split_count_mask = -1;
+
+#if defined(HAVE_SYSCONF)
+static void ht_init_nr_cpus_mask(void)
+{
+ long maxcpus;
+
+ maxcpus = sysconf(_SC_NPROCESSORS_CONF);
+ if (maxcpus <= 0) {
+ nr_cpus_mask = -2;
+ return;
+ }
+ /*
+ * round up number of CPUs to next power of two, so we
+ * can use & for modulo.
+ */
+ maxcpus = 1UL << cds_lfht_get_count_order_ulong(maxcpus);
+ nr_cpus_mask = maxcpus - 1;
+}
+#else /* #if defined(HAVE_SYSCONF) */
+static void ht_init_nr_cpus_mask(void)
+{
+ nr_cpus_mask = -2;
+}
+#endif /* #else #if defined(HAVE_SYSCONF) */
+
+static
+void alloc_split_items_count(struct cds_lfht *ht)
+{
+ struct ht_items_count *count;
+
+ if (nr_cpus_mask == -1) {
+ ht_init_nr_cpus_mask();
+ if (nr_cpus_mask < 0)
+ split_count_mask = DEFAULT_SPLIT_COUNT_MASK;
+ else
+ split_count_mask = nr_cpus_mask;
+ }
+
+ assert(split_count_mask >= 0);
+
+ if (ht->flags & CDS_LFHT_ACCOUNTING) {
+ ht->split_count = calloc(split_count_mask + 1, sizeof(*count));
+ assert(ht->split_count);
+ } else {
+ ht->split_count = NULL;
+ }
+}
+
+static
+void free_split_items_count(struct cds_lfht *ht)
+{
+ poison_free(ht->split_count);
+}
+
+#if defined(HAVE_SCHED_GETCPU)
+static
+int ht_get_split_count_index(unsigned long hash)
+{
+ int cpu;
+
+ assert(split_count_mask >= 0);
+ cpu = sched_getcpu();
+ if (caa_unlikely(cpu < 0))
+ return hash & split_count_mask;
+ else
+ return cpu & split_count_mask;
+}
+#else /* #if defined(HAVE_SCHED_GETCPU) */
+static
+int ht_get_split_count_index(unsigned long hash)
+{
+ return hash & split_count_mask;
+}
+#endif /* #else #if defined(HAVE_SCHED_GETCPU) */
+
+static
+void ht_count_add(struct cds_lfht *ht, unsigned long size, unsigned long hash)
+{
+ unsigned long split_count;
+ int index;
+ long count;
+
+ if (caa_unlikely(!ht->split_count))
+ return;
+ index = ht_get_split_count_index(hash);
+ split_count = uatomic_add_return(&ht->split_count[index].add, 1);
+ if (caa_likely(split_count & ((1UL << COUNT_COMMIT_ORDER) - 1)))
+ return;
+ /* Only if number of add multiple of 1UL << COUNT_COMMIT_ORDER */
+
+ dbg_printf("add split count %lu\n", split_count);
+ count = uatomic_add_return(&ht->count,
+ 1UL << COUNT_COMMIT_ORDER);
+ if (caa_likely(count & (count - 1)))
+ return;
+ /* Only if global count is power of 2 */
+
+ if ((count >> CHAIN_LEN_RESIZE_THRESHOLD) < size)
+ return;
+ dbg_printf("add set global %ld\n", count);
+ cds_lfht_resize_lazy_count(ht, size,
+ count >> (CHAIN_LEN_TARGET - 1));
+}
+
+static
+void ht_count_del(struct cds_lfht *ht, unsigned long size, unsigned long hash)
+{
+ unsigned long split_count;
+ int index;
+ long count;
+
+ if (caa_unlikely(!ht->split_count))
+ return;
+ index = ht_get_split_count_index(hash);
+ split_count = uatomic_add_return(&ht->split_count[index].del, 1);
+ if (caa_likely(split_count & ((1UL << COUNT_COMMIT_ORDER) - 1)))
+ return;
+ /* Only if number of deletes multiple of 1UL << COUNT_COMMIT_ORDER */
+
+ dbg_printf("del split count %lu\n", split_count);
+ count = uatomic_add_return(&ht->count,
+ -(1UL << COUNT_COMMIT_ORDER));
+ if (caa_likely(count & (count - 1)))
+ return;
+ /* Only if global count is power of 2 */
+
+ if ((count >> CHAIN_LEN_RESIZE_THRESHOLD) >= size)
+ return;
+ dbg_printf("del set global %ld\n", count);
+ /*
+ * Don't shrink table if the number of nodes is below a
+ * certain threshold.
+ */
+ if (count < (1UL << COUNT_COMMIT_ORDER) * (split_count_mask + 1))
+ return;
+ cds_lfht_resize_lazy_count(ht, size,
+ count >> (CHAIN_LEN_TARGET - 1));
+}
+
+static
+void check_resize(struct cds_lfht *ht, unsigned long size, uint32_t chain_len)
+{
+ unsigned long count;
+
+ if (!(ht->flags & CDS_LFHT_AUTO_RESIZE))
+ return;
+ count = uatomic_read(&ht->count);
+ /*
+ * Use bucket-local length for small table expand and for
+ * environments lacking per-cpu data support.
+ */
+ if (count >= (1UL << COUNT_COMMIT_ORDER))
+ return;
+ if (chain_len > 100)
+ dbg_printf("WARNING: large chain length: %u.\n",
+ chain_len);
+ if (chain_len >= CHAIN_LEN_RESIZE_THRESHOLD)
+ cds_lfht_resize_lazy_grow(ht, size,
+ cds_lfht_get_count_order_u32(chain_len - (CHAIN_LEN_TARGET - 1)));
+}
+
+static
+struct cds_lfht_node *clear_flag(struct cds_lfht_node *node)
+{
+ return (struct cds_lfht_node *) (((unsigned long) node) & ~FLAGS_MASK);
+}
+
+static
+int is_removed(struct cds_lfht_node *node)
+{
+ return ((unsigned long) node) & REMOVED_FLAG;
+}
+
+static
+struct cds_lfht_node *flag_removed(struct cds_lfht_node *node)
+{
+ return (struct cds_lfht_node *) (((unsigned long) node) | REMOVED_FLAG);
+}
+
+static
+int is_bucket(struct cds_lfht_node *node)
+{
+ return ((unsigned long) node) & BUCKET_FLAG;
+}
+
+static
+struct cds_lfht_node *flag_bucket(struct cds_lfht_node *node)
+{
+ return (struct cds_lfht_node *) (((unsigned long) node) | BUCKET_FLAG);
+}
+
+static
+int is_removal_owner(struct cds_lfht_node *node)
+{
+ return ((unsigned long) node) & REMOVAL_OWNER_FLAG;
+}
+
+static
+struct cds_lfht_node *flag_removal_owner(struct cds_lfht_node *node)
+{
+ return (struct cds_lfht_node *) (((unsigned long) node) | REMOVAL_OWNER_FLAG);
+}
+
+static
+struct cds_lfht_node *get_end(void)
+{
+ return (struct cds_lfht_node *) END_VALUE;
+}
+
+static
+int is_end(struct cds_lfht_node *node)
+{
+ return clear_flag(node) == (struct cds_lfht_node *) END_VALUE;
+}
+
+static
+unsigned long _uatomic_xchg_monotonic_increase(unsigned long *ptr,
+ unsigned long v)
+{
+ unsigned long old1, old2;
+
+ old1 = uatomic_read(ptr);
+ do {
+ old2 = old1;
+ if (old2 >= v)
+ return old2;
+ } while ((old1 = uatomic_cmpxchg(ptr, old2, v)) != old2);
+ return old2;
+}
+
+static
+void cds_lfht_alloc_bucket_table(struct cds_lfht *ht, unsigned long order)
+{
+ return ht->mm->alloc_bucket_table(ht, order);
+}
+
+/*
+ * cds_lfht_free_bucket_table() should be called with decreasing order.
+ * When cds_lfht_free_bucket_table(0) is called, it means the whole
+ * lfht is destroyed.
+ */
+static
+void cds_lfht_free_bucket_table(struct cds_lfht *ht, unsigned long order)
+{
+ return ht->mm->free_bucket_table(ht, order);
+}
+
+static inline
+struct cds_lfht_node *bucket_at(struct cds_lfht *ht, unsigned long index)
+{
+ return ht->bucket_at(ht, index);
+}
+
+static inline
+struct cds_lfht_node *lookup_bucket(struct cds_lfht *ht, unsigned long size,
+ unsigned long hash)
+{
+ assert(size > 0);
+ return bucket_at(ht, hash & (size - 1));
+}
+
+/*
+ * Remove all logically deleted nodes from a bucket up to a certain node key.
+ */
+static
+void _cds_lfht_gc_bucket(struct cds_lfht_node *bucket, struct cds_lfht_node *node)
+{
+ struct cds_lfht_node *iter_prev, *iter, *next, *new_next;
+
+ assert(!is_bucket(bucket));
+ assert(!is_removed(bucket));
+ assert(!is_bucket(node));
+ assert(!is_removed(node));
+ for (;;) {
+ iter_prev = bucket;
+ /* We can always skip the bucket node initially */
+ iter = rcu_dereference(iter_prev->next);
+ assert(!is_removed(iter));
+ assert(iter_prev->reverse_hash <= node->reverse_hash);
+ /*
+ * We should never be called with bucket (start of chain)
+ * and logically removed node (end of path compression
+ * marker) being the actual same node. This would be a
+ * bug in the algorithm implementation.
+ */
+ assert(bucket != node);
+ for (;;) {
+ if (caa_unlikely(is_end(iter)))
+ return;
+ if (caa_likely(clear_flag(iter)->reverse_hash > node->reverse_hash))
+ return;
+ next = rcu_dereference(clear_flag(iter)->next);
+ if (caa_likely(is_removed(next)))
+ break;
+ iter_prev = clear_flag(iter);
+ iter = next;
+ }
+ assert(!is_removed(iter));
+ if (is_bucket(iter))
+ new_next = flag_bucket(clear_flag(next));
+ else
+ new_next = clear_flag(next);
+ (void) uatomic_cmpxchg(&iter_prev->next, iter, new_next);
+ }
+}
+
+static
+int _cds_lfht_replace(struct cds_lfht *ht, unsigned long size,
+ struct cds_lfht_node *old_node,
+ struct cds_lfht_node *old_next,
+ struct cds_lfht_node *new_node)
+{
+ struct cds_lfht_node *bucket, *ret_next;
+
+ if (!old_node) /* Return -ENOENT if asked to replace NULL node */
+ return -ENOENT;
+
+ assert(!is_removed(old_node));
+ assert(!is_bucket(old_node));
+ assert(!is_removed(new_node));
+ assert(!is_bucket(new_node));
+ assert(new_node != old_node);
+ for (;;) {
+ /* Insert after node to be replaced */
+ if (is_removed(old_next)) {
+ /*
+ * Too late, the old node has been removed under us
+ * between lookup and replace. Fail.
+ */
+ return -ENOENT;
+ }
+ assert(old_next == clear_flag(old_next));
+ assert(new_node != old_next);
+ new_node->next = old_next;
+ /*
+ * Here is the whole trick for lock-free replace: we add
+ * the replacement node _after_ the node we want to
+ * replace by atomically setting its next pointer at the
+ * same time we set its removal flag. Given that
+ * the lookups/get next use an iterator aware of the
+ * next pointer, they will either skip the old node due
+ * to the removal flag and see the new node, or use
+ * the old node, but will not see the new one.
+ * This is a replacement of a node with another node
+ * that has the same value: we are therefore not
+ * removing a value from the hash table.
+ */
+ ret_next = uatomic_cmpxchg(&old_node->next,
+ old_next, flag_removed(new_node));
+ if (ret_next == old_next)
+ break; /* We performed the replacement. */
+ old_next = ret_next;
+ }
+
+ /*
+ * Ensure that the old node is not visible to readers anymore:
+ * lookup for the node, and remove it (along with any other
+ * logically removed node) if found.
+ */
+ bucket = lookup_bucket(ht, size, bit_reverse_ulong(old_node->reverse_hash));
+ _cds_lfht_gc_bucket(bucket, new_node);
+
+ assert(is_removed(rcu_dereference(old_node->next)));
+ return 0;
+}
+
+/*
+ * A non-NULL unique_ret pointer uses the "add unique" (or uniquify) add
+ * mode. A NULL unique_ret allows creation of duplicate keys.
+ */
+static
+void _cds_lfht_add(struct cds_lfht *ht,
+ unsigned long hash,
+ cds_lfht_match_fct match,
+ const void *key,
+ unsigned long size,
+ struct cds_lfht_node *node,
+ struct cds_lfht_iter *unique_ret,
+ int bucket_flag)
+{
+ struct cds_lfht_node *iter_prev, *iter, *next, *new_node, *new_next,
+ *return_node;
+ struct cds_lfht_node *bucket;
+
+ assert(!is_bucket(node));
+ assert(!is_removed(node));
+ bucket = lookup_bucket(ht, size, hash);
+ for (;;) {
+ uint32_t chain_len = 0;
+
+ /*
+ * iter_prev points to the non-removed node prior to the
+ * insert location.
+ */
+ iter_prev = bucket;
+ /* We can always skip the bucket node initially */
+ iter = rcu_dereference(iter_prev->next);
+ assert(iter_prev->reverse_hash <= node->reverse_hash);
+ for (;;) {
+ if (caa_unlikely(is_end(iter)))
+ goto insert;
+ if (caa_likely(clear_flag(iter)->reverse_hash > node->reverse_hash))
+ goto insert;
+
+ /* bucket node is the first node of the identical-hash-value chain */
+ if (bucket_flag && clear_flag(iter)->reverse_hash == node->reverse_hash)
+ goto insert;
+
+ next = rcu_dereference(clear_flag(iter)->next);
+ if (caa_unlikely(is_removed(next)))
+ goto gc_node;
+
+ /* uniquely add */
+ if (unique_ret
+ && !is_bucket(next)
+ && clear_flag(iter)->reverse_hash == node->reverse_hash) {
+ struct cds_lfht_iter d_iter = { .node = node, .next = iter, };
+
+ /*
+ * uniquely adding inserts the node as the first
+ * node of the identical-hash-value node chain.
+ *
+ * This semantic ensures no duplicated keys
+ * should ever be observable in the table
+ * (including observe one node by one node
+ * by forward iterations)
+ */
+ cds_lfht_next_duplicate(ht, match, key, &d_iter);
+ if (!d_iter.node)
+ goto insert;
+
+ *unique_ret = d_iter;
+ return;
+ }
+
+ /* Only account for identical reverse hash once */
+ if (iter_prev->reverse_hash != clear_flag(iter)->reverse_hash
+ && !is_bucket(next))
+ check_resize(ht, size, ++chain_len);
+ iter_prev = clear_flag(iter);
+ iter = next;
+ }
+
+ insert:
+ assert(node != clear_flag(iter));
+ assert(!is_removed(iter_prev));
+ assert(!is_removed(iter));
+ assert(iter_prev != node);
+ if (!bucket_flag)
+ node->next = clear_flag(iter);
+ else
+ node->next = flag_bucket(clear_flag(iter));
+ if (is_bucket(iter))
+ new_node = flag_bucket(node);
+ else
+ new_node = node;
+ if (uatomic_cmpxchg(&iter_prev->next, iter,
+ new_node) != iter) {
+ continue; /* retry */
+ } else {
+ return_node = node;
+ goto end;
+ }
+
+ gc_node:
+ assert(!is_removed(iter));
+ if (is_bucket(iter))
+ new_next = flag_bucket(clear_flag(next));
+ else
+ new_next = clear_flag(next);
+ (void) uatomic_cmpxchg(&iter_prev->next, iter, new_next);
+ /* retry */
+ }
+end:
+ if (unique_ret) {
+ unique_ret->node = return_node;
+ /* unique_ret->next left unset, never used. */
+ }
+}
+
+static
+int _cds_lfht_del(struct cds_lfht *ht, unsigned long size,
+ struct cds_lfht_node *node)
+{
+ struct cds_lfht_node *bucket, *next;
+
+ if (!node) /* Return -ENOENT if asked to delete NULL node */
+ return -ENOENT;
+
+ /* logically delete the node */
+ assert(!is_bucket(node));
+ assert(!is_removed(node));
+ assert(!is_removal_owner(node));
+
+ /*
+ * We are first checking if the node had previously been
+ * logically removed (this check is not atomic with setting the
+ * logical removal flag). Return -ENOENT if the node had
+ * previously been removed.
+ */
+ next = rcu_dereference(node->next);
+ if (caa_unlikely(is_removed(next)))
+ return -ENOENT;
+ assert(!is_bucket(next));
+ /*
+ * We set the REMOVED_FLAG unconditionally. Note that there may
+ * be more than one concurrent thread setting this flag.
+ * Knowing which wins the race will be known after the garbage
+ * collection phase, stay tuned!
+ */
+ uatomic_or(&node->next, REMOVED_FLAG);
+ /* We performed the (logical) deletion. */
+
+ /*
+ * Ensure that the node is not visible to readers anymore: lookup for
+ * the node, and remove it (along with any other logically removed node)
+ * if found.
+ */
+ bucket = lookup_bucket(ht, size, bit_reverse_ulong(node->reverse_hash));
+ _cds_lfht_gc_bucket(bucket, node);
+
+ assert(is_removed(rcu_dereference(node->next)));
+ /*
+ * Last phase: atomically exchange node->next with a version
+ * having "REMOVAL_OWNER_FLAG" set. If the returned node->next
+ * pointer did _not_ have "REMOVAL_OWNER_FLAG" set, we now own
+ * the node and win the removal race.
+ * It is interesting to note that all "add" paths are forbidden
+ * to change the next pointer starting from the point where the
+ * REMOVED_FLAG is set, so here using a read, followed by a
+ * xchg() suffice to guarantee that the xchg() will ever only
+ * set the "REMOVAL_OWNER_FLAG" (or change nothing if the flag
+ * was already set).
+ */
+ if (!is_removal_owner(uatomic_xchg(&node->next,
+ flag_removal_owner(node->next))))
+ return 0;
+ else
+ return -ENOENT;
+}
+
+static
+void *partition_resize_thread(void *arg)
+{
+ struct partition_resize_work *work = arg;
+
+ work->ht->flavor->register_thread();
+ work->fct(work->ht, work->i, work->start, work->len);
+ work->ht->flavor->unregister_thread();
+ return NULL;
+}
+
+static
+void partition_resize_helper(struct cds_lfht *ht, unsigned long i,
+ unsigned long len,
+ void (*fct)(struct cds_lfht *ht, unsigned long i,
+ unsigned long start, unsigned long len))
+{
+ unsigned long partition_len;
+ struct partition_resize_work *work;
+ int thread, ret;
+ unsigned long nr_threads;
+
+ /*
+ * Note: nr_cpus_mask + 1 is always power of 2.
+ * We spawn just the number of threads we need to satisfy the minimum
+ * partition size, up to the number of CPUs in the system.
+ */
+ if (nr_cpus_mask > 0) {
+ nr_threads = min(nr_cpus_mask + 1,
+ len >> MIN_PARTITION_PER_THREAD_ORDER);
+ } else {
+ nr_threads = 1;
+ }
+ partition_len = len >> cds_lfht_get_count_order_ulong(nr_threads);
+ work = calloc(nr_threads, sizeof(*work));
+ assert(work);
+ for (thread = 0; thread < nr_threads; thread++) {
+ work[thread].ht = ht;
+ work[thread].i = i;
+ work[thread].len = partition_len;
+ work[thread].start = thread * partition_len;
+ work[thread].fct = fct;
+ ret = pthread_create(&(work[thread].thread_id), ht->resize_attr,
+ partition_resize_thread, &work[thread]);
+ assert(!ret);
+ }
+ for (thread = 0; thread < nr_threads; thread++) {
+ ret = pthread_join(work[thread].thread_id, NULL);
+ assert(!ret);
+ }
+ free(work);
+}
+
+/*
+ * Holding RCU read lock to protect _cds_lfht_add against memory
+ * reclaim that could be performed by other call_rcu worker threads (ABA
+ * problem).
+ *
+ * When we reach a certain length, we can split this population phase over
+ * many worker threads, based on the number of CPUs available in the system.
+ * This should therefore take care of not having the expand lagging behind too
+ * many concurrent insertion threads by using the scheduler's ability to
+ * schedule bucket node population fairly with insertions.
+ */
+static
+void init_table_populate_partition(struct cds_lfht *ht, unsigned long i,
+ unsigned long start, unsigned long len)
+{
+ unsigned long j, size = 1UL << (i - 1);
+
+ assert(i > MIN_TABLE_ORDER);
+ ht->flavor->read_lock();
+ for (j = size + start; j < size + start + len; j++) {
+ struct cds_lfht_node *new_node = bucket_at(ht, j);
+
+ assert(j >= size && j < (size << 1));
+ dbg_printf("init populate: order %lu index %lu hash %lu\n",
+ i, j, j);
+ new_node->reverse_hash = bit_reverse_ulong(j);
+ _cds_lfht_add(ht, j, NULL, NULL, size, new_node, NULL, 1);
+ }
+ ht->flavor->read_unlock();
+}
+
+static
+void init_table_populate(struct cds_lfht *ht, unsigned long i,
+ unsigned long len)
+{
+ assert(nr_cpus_mask != -1);
+ if (nr_cpus_mask < 0 || len < 2 * MIN_PARTITION_PER_THREAD) {
+ ht->flavor->thread_online();
+ init_table_populate_partition(ht, i, 0, len);
+ ht->flavor->thread_offline();
+ return;
+ }
+ partition_resize_helper(ht, i, len, init_table_populate_partition);
+}
+
+static
+void init_table(struct cds_lfht *ht,
+ unsigned long first_order, unsigned long last_order)
+{
+ unsigned long i;
+
+ dbg_printf("init table: first_order %lu last_order %lu\n",
+ first_order, last_order);
+ assert(first_order > MIN_TABLE_ORDER);
+ for (i = first_order; i <= last_order; i++) {
+ unsigned long len;
+
+ len = 1UL << (i - 1);
+ dbg_printf("init order %lu len: %lu\n", i, len);
+
+ /* Stop expand if the resize target changes under us */
+ if (CMM_LOAD_SHARED(ht->resize_target) < (1UL << i))
+ break;
+
+ cds_lfht_alloc_bucket_table(ht, i);
+
+ /*
+ * Set all bucket nodes reverse hash values for a level and
+ * link all bucket nodes into the table.
+ */
+ init_table_populate(ht, i, len);
+
+ /*
+ * Update table size.
+ */
+ cmm_smp_wmb(); /* populate data before RCU size */
+ CMM_STORE_SHARED(ht->size, 1UL << i);
+
+ dbg_printf("init new size: %lu\n", 1UL << i);
+ if (CMM_LOAD_SHARED(ht->in_progress_destroy))
+ break;
+ }
+}
+
+/*
+ * Holding RCU read lock to protect _cds_lfht_remove against memory
+ * reclaim that could be performed by other call_rcu worker threads (ABA
+ * problem).
+ * For a single level, we logically remove and garbage collect each node.
+ *
+ * As a design choice, we perform logical removal and garbage collection on a
+ * node-per-node basis to simplify this algorithm. We also assume keeping good
+ * cache locality of the operation would overweight possible performance gain
+ * that could be achieved by batching garbage collection for multiple levels.
+ * However, this would have to be justified by benchmarks.
+ *
+ * Concurrent removal and add operations are helping us perform garbage
+ * collection of logically removed nodes. We guarantee that all logically
+ * removed nodes have been garbage-collected (unlinked) before call_rcu is
+ * invoked to free a hole level of bucket nodes (after a grace period).
+ *
+ * Logical removal and garbage collection can therefore be done in batch or on a
+ * node-per-node basis, as long as the guarantee above holds.
+ *
+ * When we reach a certain length, we can split this removal over many worker
+ * threads, based on the number of CPUs available in the system. This should
+ * take care of not letting resize process lag behind too many concurrent
+ * updater threads actively inserting into the hash table.
+ */
+static
+void remove_table_partition(struct cds_lfht *ht, unsigned long i,
+ unsigned long start, unsigned long len)
+{
+ unsigned long j, size = 1UL << (i - 1);
+
+ assert(i > MIN_TABLE_ORDER);
+ ht->flavor->read_lock();
+ for (j = size + start; j < size + start + len; j++) {
+ struct cds_lfht_node *fini_bucket = bucket_at(ht, j);
+ struct cds_lfht_node *parent_bucket = bucket_at(ht, j - size);
+
+ assert(j >= size && j < (size << 1));
+ dbg_printf("remove entry: order %lu index %lu hash %lu\n",
+ i, j, j);
+ /* Set the REMOVED_FLAG to freeze the ->next for gc */
+ uatomic_or(&fini_bucket->next, REMOVED_FLAG);
+ _cds_lfht_gc_bucket(parent_bucket, fini_bucket);
+ }
+ ht->flavor->read_unlock();
+}
+
+static
+void remove_table(struct cds_lfht *ht, unsigned long i, unsigned long len)
+{
+
+ assert(nr_cpus_mask != -1);
+ if (nr_cpus_mask < 0 || len < 2 * MIN_PARTITION_PER_THREAD) {
+ ht->flavor->thread_online();
+ remove_table_partition(ht, i, 0, len);
+ ht->flavor->thread_offline();
+ return;
+ }
+ partition_resize_helper(ht, i, len, remove_table_partition);
+}
+
+/*
+ * fini_table() is never called for first_order == 0, which is why
+ * free_by_rcu_order == 0 can be used as criterion to know if free must
+ * be called.
+ */
+static
+void fini_table(struct cds_lfht *ht,
+ unsigned long first_order, unsigned long last_order)
+{
+ long i;
+ unsigned long free_by_rcu_order = 0;
+
+ dbg_printf("fini table: first_order %lu last_order %lu\n",
+ first_order, last_order);
+ assert(first_order > MIN_TABLE_ORDER);
+ for (i = last_order; i >= first_order; i--) {
+ unsigned long len;
+
+ len = 1UL << (i - 1);
+ dbg_printf("fini order %lu len: %lu\n", i, len);
+
+ /* Stop shrink if the resize target changes under us */
+ if (CMM_LOAD_SHARED(ht->resize_target) > (1UL << (i - 1)))
+ break;
+
+ cmm_smp_wmb(); /* populate data before RCU size */
+ CMM_STORE_SHARED(ht->size, 1UL << (i - 1));
+
+ /*
+ * We need to wait for all add operations to reach Q.S. (and
+ * thus use the new table for lookups) before we can start
+ * releasing the old bucket nodes. Otherwise their lookup will
+ * return a logically removed node as insert position.
+ */
+ ht->flavor->update_synchronize_rcu();
+ if (free_by_rcu_order)
+ cds_lfht_free_bucket_table(ht, free_by_rcu_order);
+
+ /*
+ * Set "removed" flag in bucket nodes about to be removed.
+ * Unlink all now-logically-removed bucket node pointers.
+ * Concurrent add/remove operation are helping us doing
+ * the gc.
+ */
+ remove_table(ht, i, len);
+
+ free_by_rcu_order = i;
+
+ dbg_printf("fini new size: %lu\n", 1UL << i);
+ if (CMM_LOAD_SHARED(ht->in_progress_destroy))
+ break;
+ }
+
+ if (free_by_rcu_order) {
+ ht->flavor->update_synchronize_rcu();
+ cds_lfht_free_bucket_table(ht, free_by_rcu_order);
+ }
+}
+
+static
+void cds_lfht_create_bucket(struct cds_lfht *ht, unsigned long size)
+{
+ struct cds_lfht_node *prev, *node;
+ unsigned long order, len, i;
+
+ cds_lfht_alloc_bucket_table(ht, 0);
+
+ dbg_printf("create bucket: order 0 index 0 hash 0\n");
+ node = bucket_at(ht, 0);
+ node->next = flag_bucket(get_end());
+ node->reverse_hash = 0;
+
+ for (order = 1; order < cds_lfht_get_count_order_ulong(size) + 1; order++) {
+ len = 1UL << (order - 1);
+ cds_lfht_alloc_bucket_table(ht, order);
+
+ for (i = 0; i < len; i++) {
+ /*
+ * Now, we are trying to init the node with the
+ * hash=(len+i) (which is also a bucket with the
+ * index=(len+i)) and insert it into the hash table,
+ * so this node has to be inserted after the bucket
+ * with the index=(len+i)&(len-1)=i. And because there
+ * is no other non-bucket node nor bucket node with
+ * larger index/hash inserted, so the bucket node
+ * being inserted should be inserted directly linked
+ * after the bucket node with index=i.
+ */
+ prev = bucket_at(ht, i);
+ node = bucket_at(ht, len + i);
+
+ dbg_printf("create bucket: order %lu index %lu hash %lu\n",
+ order, len + i, len + i);
+ node->reverse_hash = bit_reverse_ulong(len + i);
+
+ /* insert after prev */
+ assert(is_bucket(prev->next));
+ node->next = prev->next;
+ prev->next = flag_bucket(node);
+ }
+ }
+}
+
+struct cds_lfht *_cds_lfht_new(unsigned long init_size,
+ unsigned long min_nr_alloc_buckets,
+ unsigned long max_nr_buckets,
+ int flags,
+ const struct cds_lfht_mm_type *mm,
+ const struct rcu_flavor_struct *flavor,
+ pthread_attr_t *attr)
+{
+ struct cds_lfht *ht;
+ unsigned long order;
+
+ /* min_nr_alloc_buckets must be power of two */
+ if (!min_nr_alloc_buckets || (min_nr_alloc_buckets & (min_nr_alloc_buckets - 1)))
+ return NULL;
+
+ /* init_size must be power of two */
+ if (!init_size || (init_size & (init_size - 1)))
+ return NULL;
+
+ /*
+ * Memory management plugin default.
+ */
+ if (!mm) {
+ if (CAA_BITS_PER_LONG > 32
+ && max_nr_buckets
+ && max_nr_buckets <= (1ULL << 32)) {
+ /*
+ * For 64-bit architectures, with max number of
+ * buckets small enough not to use the entire
+ * 64-bit memory mapping space (and allowing a
+ * fair number of hash table instances), use the
+ * mmap allocator, which is faster than the
+ * order allocator.
+ */
+ mm = &cds_lfht_mm_mmap;
+ } else {
+ /*
+ * The fallback is to use the order allocator.
+ */
+ mm = &cds_lfht_mm_order;
+ }
+ }
+
+ /* max_nr_buckets == 0 for order based mm means infinite */
+ if (mm == &cds_lfht_mm_order && !max_nr_buckets)
+ max_nr_buckets = 1UL << (MAX_TABLE_ORDER - 1);
+
+ /* max_nr_buckets must be power of two */
+ if (!max_nr_buckets || (max_nr_buckets & (max_nr_buckets - 1)))
+ return NULL;
+
+ min_nr_alloc_buckets = max(min_nr_alloc_buckets, MIN_TABLE_SIZE);
+ init_size = max(init_size, MIN_TABLE_SIZE);
+ max_nr_buckets = max(max_nr_buckets, min_nr_alloc_buckets);
+ init_size = min(init_size, max_nr_buckets);
+
+ ht = mm->alloc_cds_lfht(min_nr_alloc_buckets, max_nr_buckets);
+ assert(ht);
+ assert(ht->mm == mm);
+ assert(ht->bucket_at == mm->bucket_at);
+
+ ht->flags = flags;
+ ht->flavor = flavor;
+ ht->resize_attr = attr;
+ alloc_split_items_count(ht);
+ /* this mutex should not nest in read-side C.S. */
+ pthread_mutex_init(&ht->resize_mutex, NULL);
+ order = cds_lfht_get_count_order_ulong(init_size);
+ ht->resize_target = 1UL << order;
+ cds_lfht_create_bucket(ht, 1UL << order);
+ ht->size = 1UL << order;
+ return ht;
+}
+
+void cds_lfht_lookup(struct cds_lfht *ht, unsigned long hash,
+ cds_lfht_match_fct match, const void *key,
+ struct cds_lfht_iter *iter)
+{
+ struct cds_lfht_node *node, *next, *bucket;
+ unsigned long reverse_hash, size;
+
+ reverse_hash = bit_reverse_ulong(hash);
+
+ size = rcu_dereference(ht->size);
+ bucket = lookup_bucket(ht, size, hash);
+ /* We can always skip the bucket node initially */
+ node = rcu_dereference(bucket->next);
+ node = clear_flag(node);
+ for (;;) {
+ if (caa_unlikely(is_end(node))) {
+ node = next = NULL;
+ break;
+ }
+ if (caa_unlikely(node->reverse_hash > reverse_hash)) {
+ node = next = NULL;
+ break;
+ }
+ next = rcu_dereference(node->next);
+ assert(node == clear_flag(node));
+ if (caa_likely(!is_removed(next))
+ && !is_bucket(next)
+ && node->reverse_hash == reverse_hash
+ && caa_likely(match(node, key))) {
+ break;
+ }
+ node = clear_flag(next);
+ }
+ assert(!node || !is_bucket(rcu_dereference(node->next)));
+ iter->node = node;
+ iter->next = next;
+}
+
+void cds_lfht_next_duplicate(struct cds_lfht *ht, cds_lfht_match_fct match,
+ const void *key, struct cds_lfht_iter *iter)
+{
+ struct cds_lfht_node *node, *next;
+ unsigned long reverse_hash;
+
+ node = iter->node;
+ reverse_hash = node->reverse_hash;
+ next = iter->next;
+ node = clear_flag(next);
+
+ for (;;) {
+ if (caa_unlikely(is_end(node))) {
+ node = next = NULL;
+ break;
+ }
+ if (caa_unlikely(node->reverse_hash > reverse_hash)) {
+ node = next = NULL;
+ break;
+ }
+ next = rcu_dereference(node->next);
+ if (caa_likely(!is_removed(next))
+ && !is_bucket(next)
+ && caa_likely(match(node, key))) {
+ break;
+ }
+ node = clear_flag(next);
+ }
+ assert(!node || !is_bucket(rcu_dereference(node->next)));
+ iter->node = node;
+ iter->next = next;
+}
+
+void cds_lfht_next(struct cds_lfht *ht, struct cds_lfht_iter *iter)
+{
+ struct cds_lfht_node *node, *next;
+
+ node = clear_flag(iter->next);
+ for (;;) {
+ if (caa_unlikely(is_end(node))) {
+ node = next = NULL;
+ break;
+ }
+ next = rcu_dereference(node->next);
+ if (caa_likely(!is_removed(next))
+ && !is_bucket(next)) {
+ break;
+ }
+ node = clear_flag(next);
+ }
+ assert(!node || !is_bucket(rcu_dereference(node->next)));
+ iter->node = node;
+ iter->next = next;
+}
+
+void cds_lfht_first(struct cds_lfht *ht, struct cds_lfht_iter *iter)
+{
+ /*
+ * Get next after first bucket node. The first bucket node is the
+ * first node of the linked list.
+ */
+ iter->next = bucket_at(ht, 0)->next;
+ cds_lfht_next(ht, iter);
+}
+
+void cds_lfht_add(struct cds_lfht *ht, unsigned long hash,
+ struct cds_lfht_node *node)
+{
+ unsigned long size;
+
+ node->reverse_hash = bit_reverse_ulong(hash);
+ size = rcu_dereference(ht->size);
+ _cds_lfht_add(ht, hash, NULL, NULL, size, node, NULL, 0);
+ ht_count_add(ht, size, hash);
+}
+
+struct cds_lfht_node *cds_lfht_add_unique(struct cds_lfht *ht,
+ unsigned long hash,
+ cds_lfht_match_fct match,
+ const void *key,
+ struct cds_lfht_node *node)
+{
+ unsigned long size;
+ struct cds_lfht_iter iter;
+
+ node->reverse_hash = bit_reverse_ulong(hash);
+ size = rcu_dereference(ht->size);
+ _cds_lfht_add(ht, hash, match, key, size, node, &iter, 0);
+ if (iter.node == node)
+ ht_count_add(ht, size, hash);
+ return iter.node;
+}
+
+struct cds_lfht_node *cds_lfht_add_replace(struct cds_lfht *ht,
+ unsigned long hash,
+ cds_lfht_match_fct match,
+ const void *key,
+ struct cds_lfht_node *node)
+{
+ unsigned long size;
+ struct cds_lfht_iter iter;
+
+ node->reverse_hash = bit_reverse_ulong(hash);
+ size = rcu_dereference(ht->size);
+ for (;;) {
+ _cds_lfht_add(ht, hash, match, key, size, node, &iter, 0);
+ if (iter.node == node) {
+ ht_count_add(ht, size, hash);
+ return NULL;
+ }
+
+ if (!_cds_lfht_replace(ht, size, iter.node, iter.next, node))
+ return iter.node;
+ }
+}
+
+int cds_lfht_replace(struct cds_lfht *ht,
+ struct cds_lfht_iter *old_iter,
+ unsigned long hash,
+ cds_lfht_match_fct match,
+ const void *key,
+ struct cds_lfht_node *new_node)
+{
+ unsigned long size;
+
+ new_node->reverse_hash = bit_reverse_ulong(hash);
+ if (!old_iter->node)
+ return -ENOENT;
+ if (caa_unlikely(old_iter->node->reverse_hash != new_node->reverse_hash))
+ return -EINVAL;
+ if (caa_unlikely(!match(old_iter->node, key)))
+ return -EINVAL;
+ size = rcu_dereference(ht->size);
+ return _cds_lfht_replace(ht, size, old_iter->node, old_iter->next,
+ new_node);
+}
+
+int cds_lfht_del(struct cds_lfht *ht, struct cds_lfht_node *node)
+{
+ unsigned long size, hash;
+ int ret;
+
+ size = rcu_dereference(ht->size);
+ ret = _cds_lfht_del(ht, size, node);
+ if (!ret) {
+ hash = bit_reverse_ulong(node->reverse_hash);
+ ht_count_del(ht, size, hash);
+ }
+ return ret;
+}
+
+static
+int cds_lfht_delete_bucket(struct cds_lfht *ht)
+{
+ struct cds_lfht_node *node;
+ unsigned long order, i, size;
+
+ /* Check that the table is empty */
+ node = bucket_at(ht, 0);
+ do {
+ node = clear_flag(node)->next;
+ if (!is_bucket(node))
+ return -EPERM;
+ assert(!is_removed(node));
+ } while (!is_end(node));
+ /*
+ * size accessed without rcu_dereference because hash table is
+ * being destroyed.
+ */
+ size = ht->size;
+ /* Internal sanity check: all nodes left should be bucket */
+ for (i = 0; i < size; i++) {
+ node = bucket_at(ht, i);
+ dbg_printf("delete bucket: index %lu expected hash %lu hash %lu\n",
+ i, i, bit_reverse_ulong(node->reverse_hash));
+ assert(is_bucket(node->next));
+ }
+
+ for (order = cds_lfht_get_count_order_ulong(size); (long)order >= 0; order--)
+ cds_lfht_free_bucket_table(ht, order);
+
+ return 0;
+}
+
+/*
+ * Should only be called when no more concurrent readers nor writers can
+ * possibly access the table.
+ */
+int cds_lfht_destroy(struct cds_lfht *ht, pthread_attr_t **attr)
+{
+ int ret;
+
+ /* Wait for in-flight resize operations to complete */
+ _CMM_STORE_SHARED(ht->in_progress_destroy, 1);
+ cmm_smp_mb(); /* Store destroy before load resize */
+ while (uatomic_read(&ht->in_progress_resize))
+ poll(NULL, 0, 100); /* wait for 100ms */
+ ret = cds_lfht_delete_bucket(ht);
+ if (ret)
+ return ret;
+ free_split_items_count(ht);
+ if (attr)
+ *attr = ht->resize_attr;
+ poison_free(ht);
+ return ret;
+}
+
+void cds_lfht_count_nodes(struct cds_lfht *ht,
+ long *approx_before,
+ unsigned long *count,
+ long *approx_after)
+{
+ struct cds_lfht_node *node, *next;
+ unsigned long nr_bucket = 0, nr_removed = 0;
+
+ *approx_before = 0;
+ if (ht->split_count) {
+ int i;
+
+ for (i = 0; i < split_count_mask + 1; i++) {
+ *approx_before += uatomic_read(&ht->split_count[i].add);
+ *approx_before -= uatomic_read(&ht->split_count[i].del);
+ }
+ }
+
+ *count = 0;
+
+ /* Count non-bucket nodes in the table */
+ node = bucket_at(ht, 0);
+ do {
+ next = rcu_dereference(node->next);
+ if (is_removed(next)) {
+ if (!is_bucket(next))
+ (nr_removed)++;
+ else
+ (nr_bucket)++;
+ } else if (!is_bucket(next))
+ (*count)++;
+ else
+ (nr_bucket)++;
+ node = clear_flag(next);
+ } while (!is_end(node));
+ dbg_printf("number of logically removed nodes: %lu\n", nr_removed);
+ dbg_printf("number of bucket nodes: %lu\n", nr_bucket);
+ *approx_after = 0;
+ if (ht->split_count) {
+ int i;
+
+ for (i = 0; i < split_count_mask + 1; i++) {
+ *approx_after += uatomic_read(&ht->split_count[i].add);
+ *approx_after -= uatomic_read(&ht->split_count[i].del);
+ }
+ }
+}
+
+/* called with resize mutex held */
+static
+void _do_cds_lfht_grow(struct cds_lfht *ht,
+ unsigned long old_size, unsigned long new_size)
+{
+ unsigned long old_order, new_order;
+
+ old_order = cds_lfht_get_count_order_ulong(old_size);
+ new_order = cds_lfht_get_count_order_ulong(new_size);
+ dbg_printf("resize from %lu (order %lu) to %lu (order %lu) buckets\n",
+ old_size, old_order, new_size, new_order);
+ assert(new_size > old_size);
+ init_table(ht, old_order + 1, new_order);
+}
+
+/* called with resize mutex held */
+static
+void _do_cds_lfht_shrink(struct cds_lfht *ht,
+ unsigned long old_size, unsigned long new_size)
+{
+ unsigned long old_order, new_order;
+
+ new_size = max(new_size, MIN_TABLE_SIZE);
+ old_order = cds_lfht_get_count_order_ulong(old_size);
+ new_order = cds_lfht_get_count_order_ulong(new_size);
+ dbg_printf("resize from %lu (order %lu) to %lu (order %lu) buckets\n",
+ old_size, old_order, new_size, new_order);
+ assert(new_size < old_size);
+
+ /* Remove and unlink all bucket nodes to remove. */
+ fini_table(ht, new_order + 1, old_order);
+}
+
+
+/* called with resize mutex held */
+static
+void _do_cds_lfht_resize(struct cds_lfht *ht)
+{
+ unsigned long new_size, old_size;
+
+ /*
+ * Resize table, re-do if the target size has changed under us.
+ */
+ do {
+ assert(uatomic_read(&ht->in_progress_resize));
+ if (CMM_LOAD_SHARED(ht->in_progress_destroy))
+ break;
+ ht->resize_initiated = 1;
+ old_size = ht->size;
+ new_size = CMM_LOAD_SHARED(ht->resize_target);
+ if (old_size < new_size)
+ _do_cds_lfht_grow(ht, old_size, new_size);
+ else if (old_size > new_size)
+ _do_cds_lfht_shrink(ht, old_size, new_size);
+ ht->resize_initiated = 0;
+ /* write resize_initiated before read resize_target */
+ cmm_smp_mb();
+ } while (ht->size != CMM_LOAD_SHARED(ht->resize_target));
+}
+
+static
+unsigned long resize_target_grow(struct cds_lfht *ht, unsigned long new_size)
+{
+ return _uatomic_xchg_monotonic_increase(&ht->resize_target, new_size);
+}
+
+static
+void resize_target_update_count(struct cds_lfht *ht,
+ unsigned long count)
+{
+ count = max(count, MIN_TABLE_SIZE);
+ count = min(count, ht->max_nr_buckets);
+ uatomic_set(&ht->resize_target, count);
+}
+
+void cds_lfht_resize(struct cds_lfht *ht, unsigned long new_size)
+{
+ resize_target_update_count(ht, new_size);
+ CMM_STORE_SHARED(ht->resize_initiated, 1);
+ ht->flavor->thread_offline();
+ pthread_mutex_lock(&ht->resize_mutex);
+ _do_cds_lfht_resize(ht);
+ pthread_mutex_unlock(&ht->resize_mutex);
+ ht->flavor->thread_online();
+}
+
+static
+void do_resize_cb(struct rcu_head *head)
+{
+ struct rcu_resize_work *work =
+ caa_container_of(head, struct rcu_resize_work, head);
+ struct cds_lfht *ht = work->ht;
+
+ ht->flavor->thread_offline();
+ pthread_mutex_lock(&ht->resize_mutex);
+ _do_cds_lfht_resize(ht);
+ pthread_mutex_unlock(&ht->resize_mutex);
+ ht->flavor->thread_online();
+ poison_free(work);
+ cmm_smp_mb(); /* finish resize before decrement */
+ uatomic_dec(&ht->in_progress_resize);
+}
+
+static
+void __cds_lfht_resize_lazy_launch(struct cds_lfht *ht)
+{
+ struct rcu_resize_work *work;
+
+ /* Store resize_target before read resize_initiated */
+ cmm_smp_mb();
+ if (!CMM_LOAD_SHARED(ht->resize_initiated)) {
+ uatomic_inc(&ht->in_progress_resize);
+ cmm_smp_mb(); /* increment resize count before load destroy */
+ if (CMM_LOAD_SHARED(ht->in_progress_destroy)) {
+ uatomic_dec(&ht->in_progress_resize);
+ return;
+ }
+ work = malloc(sizeof(*work));
+ work->ht = ht;
+ ht->flavor->update_call_rcu(&work->head, do_resize_cb);
+ CMM_STORE_SHARED(ht->resize_initiated, 1);
+ }
+}
+
+static
+void cds_lfht_resize_lazy_grow(struct cds_lfht *ht, unsigned long size, int growth)
+{
+ unsigned long target_size = size << growth;
+
+ target_size = min(target_size, ht->max_nr_buckets);
+ if (resize_target_grow(ht, target_size) >= target_size)
+ return;
+
+ __cds_lfht_resize_lazy_launch(ht);
+}
+
+/*
+ * We favor grow operations over shrink. A shrink operation never occurs
+ * if a grow operation is queued for lazy execution. A grow operation
+ * cancels any pending shrink lazy execution.
+ */
+static
+void cds_lfht_resize_lazy_count(struct cds_lfht *ht, unsigned long size,
+ unsigned long count)
+{
+ if (!(ht->flags & CDS_LFHT_AUTO_RESIZE))
+ return;
+ count = max(count, MIN_TABLE_SIZE);
+ count = min(count, ht->max_nr_buckets);
+ if (count == size)
+ return; /* Already the right size, no resize needed */
+ if (count > size) { /* lazy grow */
+ if (resize_target_grow(ht, count) >= count)
+ return;
+ } else { /* lazy shrink */
+ for (;;) {
+ unsigned long s;
+
+ s = uatomic_cmpxchg(&ht->resize_target, size, count);
+ if (s == size)
+ break; /* no resize needed */
+ if (s > size)
+ return; /* growing is/(was just) in progress */
+ if (s <= count)
+ return; /* some other thread do shrink */
+ size = s;
+ }
+ }
+ __cds_lfht_resize_lazy_launch(ht);
+}
--- /dev/null
+#ifndef _URCU_RCULFHASH_H
+#define _URCU_RCULFHASH_H
+
+/*
+ * urcu/rculfhash.h
+ *
+ * Userspace RCU library - Lock-Free RCU Hash Table
+ *
+ * Copyright 2011 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ * Copyright 2011 - Lai Jiangshan <laijs@cn.fujitsu.com>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library 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
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ *
+ * Include this file _after_ including your URCU flavor.
+ */
+
+#include <stdint.h>
+#include <urcu/compiler.h>
+#include <urcu-call-rcu.h>
+
+#include "urcu-flavor.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+ * cds_lfht_node: Contains the next pointers and reverse-hash
+ * value required for lookup and traversal of the hash table.
+ *
+ * struct cds_lfht_node should be aligned on 8-bytes boundaries because
+ * the three lower bits are used as flags. It is worth noting that the
+ * information contained within these three bits could be represented on
+ * two bits by re-using the same bit for REMOVAL_OWNER_FLAG and
+ * BUCKET_FLAG. This can be done if we ensure that no iterator nor
+ * updater check the BUCKET_FLAG after it detects that the REMOVED_FLAG
+ * is set. Given the minimum size of struct cds_lfht_node is 8 bytes on
+ * 32-bit architectures, we choose to go for simplicity and reserve
+ * three bits.
+ *
+ * struct cds_lfht_node can be embedded into a structure (as a field).
+ * caa_container_of() can be used to get the structure from the struct
+ * cds_lfht_node after a lookup.
+ *
+ * The structure which embeds it typically holds the key (or key-value
+ * pair) of the object. The caller code is responsible for calculation
+ * of the hash value for cds_lfht APIs.
+ */
+struct cds_lfht_node {
+ struct cds_lfht_node *next; /* ptr | REMOVAL_OWNER_FLAG | BUCKET_FLAG | REMOVED_FLAG */
+ unsigned long reverse_hash;
+} __attribute__((aligned(8)));
+
+/* cds_lfht_iter: Used to track state while traversing a hash chain. */
+struct cds_lfht_iter {
+ struct cds_lfht_node *node, *next;
+};
+
+static inline
+struct cds_lfht_node *cds_lfht_iter_get_node(struct cds_lfht_iter *iter)
+{
+ return iter->node;
+}
+
+struct cds_lfht;
+
+/*
+ * Caution !
+ * Ensure reader and writer threads are registered as urcu readers.
+ */
+
+typedef int (*cds_lfht_match_fct)(struct cds_lfht_node *node, const void *key);
+
+/*
+ * cds_lfht_node_init - initialize a hash table node
+ * @node: the node to initialize.
+ *
+ * This function is kept to be eventually used for debugging purposes
+ * (detection of memory corruption).
+ */
+static inline
+void cds_lfht_node_init(struct cds_lfht_node *node)
+{
+}
+
+/*
+ * Hash table creation flags.
+ */
+enum {
+ CDS_LFHT_AUTO_RESIZE = (1U << 0),
+ CDS_LFHT_ACCOUNTING = (1U << 1),
+};
+
+struct cds_lfht_mm_type {
+ struct cds_lfht *(*alloc_cds_lfht)(unsigned long min_nr_alloc_buckets,
+ unsigned long max_nr_buckets);
+ void (*alloc_bucket_table)(struct cds_lfht *ht, unsigned long order);
+ void (*free_bucket_table)(struct cds_lfht *ht, unsigned long order);
+ struct cds_lfht_node *(*bucket_at)(struct cds_lfht *ht,
+ unsigned long index);
+};
+
+extern const struct cds_lfht_mm_type cds_lfht_mm_order;
+extern const struct cds_lfht_mm_type cds_lfht_mm_chunk;
+extern const struct cds_lfht_mm_type cds_lfht_mm_mmap;
+
+/*
+ * _cds_lfht_new - API used by cds_lfht_new wrapper. Do not use directly.
+ */
+struct cds_lfht *_cds_lfht_new(unsigned long init_size,
+ unsigned long min_nr_alloc_buckets,
+ unsigned long max_nr_buckets,
+ int flags,
+ const struct cds_lfht_mm_type *mm,
+ const struct rcu_flavor_struct *flavor,
+ pthread_attr_t *attr);
+
+/*
+ * cds_lfht_new - allocate a hash table.
+ * @init_size: number of buckets to allocate initially. Must be power of two.
+ * @min_nr_alloc_buckets: the minimum number of allocated buckets.
+ * (must be power of two)
+ * @max_nr_buckets: the maximum number of hash table buckets allowed.
+ * (must be power of two)
+ * @flags: hash table creation flags (can be combined with bitwise or: '|').
+ * 0: no flags.
+ * CDS_LFHT_AUTO_RESIZE: automatically resize hash table.
+ * CDS_LFHT_ACCOUNTING: count the number of node addition
+ * and removal in the table
+ * @attr: optional resize worker thread attributes. NULL for default.
+ *
+ * Return NULL on error.
+ * Note: the RCU flavor must be already included before the hash table header.
+ *
+ * The programmer is responsible for ensuring that resize operation has a
+ * priority equal to hash table updater threads. It should be performed by
+ * specifying the appropriate priority in the pthread "attr" argument, and,
+ * for CDS_LFHT_AUTO_RESIZE, by ensuring that call_rcu worker threads also have
+ * this priority level. Having lower priority for call_rcu and resize threads
+ * does not pose any correctness issue, but the resize operations could be
+ * starved by updates, thus leading to long hash table bucket chains.
+ * Threads calling this API are NOT required to be registered RCU read-side
+ * threads. It can be called very early.(before rcu is initialized ...etc.)
+ */
+static inline
+struct cds_lfht *cds_lfht_new(unsigned long init_size,
+ unsigned long min_nr_alloc_buckets,
+ unsigned long max_nr_buckets,
+ int flags,
+ pthread_attr_t *attr)
+{
+ return _cds_lfht_new(init_size, min_nr_alloc_buckets, max_nr_buckets,
+ flags, NULL, &rcu_flavor, attr);
+}
+
+/*
+ * cds_lfht_destroy - destroy a hash table.
+ * @ht: the hash table to destroy.
+ * @attr: (output) resize worker thread attributes, as received by cds_lfht_new.
+ * The caller will typically want to free this pointer if dynamically
+ * allocated. The attr point can be NULL if the caller does not
+ * need to be informed of the value passed to cds_lfht_new().
+ *
+ * Return 0 on success, negative error value on error.
+ * Threads calling this API need to be registered RCU read-side threads.
+ */
+int cds_lfht_destroy(struct cds_lfht *ht, pthread_attr_t **attr);
+
+/*
+ * cds_lfht_count_nodes - count the number of nodes in the hash table.
+ * @ht: the hash table.
+ * @split_count_before: Sample the node count split-counter before traversal.
+ * @count: Traverse the hash table, count the number of nodes observed.
+ * @split_count_after: Sample the node count split-counter after traversal.
+ *
+ * Call with rcu_read_lock held.
+ * Threads calling this API need to be registered RCU read-side threads.
+ */
+void cds_lfht_count_nodes(struct cds_lfht *ht,
+ long *split_count_before,
+ unsigned long *count,
+ long *split_count_after);
+
+/*
+ * cds_lfht_lookup - lookup a node by key.
+ * @ht: the hash table.
+ * @hash: the key hash.
+ * @match: the key match function.
+ * @key: the current node key.
+ * @iter: Node, if found (output). *iter->node set to NULL if not found.
+ *
+ * Call with rcu_read_lock held.
+ * Threads calling this API need to be registered RCU read-side threads.
+ */
+void cds_lfht_lookup(struct cds_lfht *ht, unsigned long hash,
+ cds_lfht_match_fct match, const void *key,
+ struct cds_lfht_iter *iter);
+
+/*
+ * cds_lfht_next_duplicate - get the next item with same key (after a lookup).
+ * @ht: the hash table.
+ * @match: the key match function.
+ * @key: the current node key.
+ * @iter: Node, if found (output). *iter->node set to NULL if not found.
+ *
+ * Uses an iterator initialized by a lookup.
+ * Sets *iter-node to the following node with same key.
+ * Sets *iter->node to NULL if no following node exists with same key.
+ * RCU read-side lock must be held across cds_lfht_lookup and
+ * cds_lfht_next calls, and also between cds_lfht_next calls using the
+ * node returned by a previous cds_lfht_next.
+ * Call with rcu_read_lock held.
+ * Threads calling this API need to be registered RCU read-side threads.
+ */
+void cds_lfht_next_duplicate(struct cds_lfht *ht,
+ cds_lfht_match_fct match, const void *key,
+ struct cds_lfht_iter *iter);
+
+/*
+ * cds_lfht_first - get the first node in the table.
+ * @ht: the hash table.
+ * @iter: First node, if exists (output). *iter->node set to NULL if not found.
+ *
+ * Output in "*iter". *iter->node set to NULL if table is empty.
+ * Call with rcu_read_lock held.
+ * Threads calling this API need to be registered RCU read-side threads.
+ */
+void cds_lfht_first(struct cds_lfht *ht, struct cds_lfht_iter *iter);
+
+/*
+ * cds_lfht_next - get the next node in the table.
+ * @ht: the hash table.
+ * @iter: Next node, if exists (output). *iter->node set to NULL if not found.
+ *
+ * Input/Output in "*iter". *iter->node set to NULL if *iter was
+ * pointing to the last table node.
+ * Call with rcu_read_lock held.
+ * Threads calling this API need to be registered RCU read-side threads.
+ */
+void cds_lfht_next(struct cds_lfht *ht, struct cds_lfht_iter *iter);
+
+/*
+ * cds_lfht_add - add a node to the hash table.
+ * @ht: the hash table.
+ * @hash: the key hash.
+ * @node: the node to add.
+ *
+ * This function supports adding redundant keys into the table.
+ * Call with rcu_read_lock held.
+ * Threads calling this API need to be registered RCU read-side threads.
+ */
+void cds_lfht_add(struct cds_lfht *ht, unsigned long hash,
+ struct cds_lfht_node *node);
+
+/*
+ * cds_lfht_add_unique - add a node to hash table, if key is not present.
+ * @ht: the hash table.
+ * @hash: the node's hash.
+ * @match: the key match function.
+ * @key: the node's key.
+ * @node: the node to try adding.
+ *
+ * Return the node added upon success.
+ * Return the unique node already present upon failure. If
+ * cds_lfht_add_unique fails, the node passed as parameter should be
+ * freed by the caller.
+ * Call with rcu_read_lock held.
+ * Threads calling this API need to be registered RCU read-side threads.
+ *
+ * The semantic of this function is that if only this function is used
+ * to add keys into the table, no duplicated keys should ever be
+ * observable in the table. The same guarantee apply for combination of
+ * add_unique and add_replace (see below).
+ */
+struct cds_lfht_node *cds_lfht_add_unique(struct cds_lfht *ht,
+ unsigned long hash,
+ cds_lfht_match_fct match,
+ const void *key,
+ struct cds_lfht_node *node);
+
+/*
+ * cds_lfht_add_replace - replace or add a node within hash table.
+ * @ht: the hash table.
+ * @hash: the node's hash.
+ * @match: the key match function.
+ * @key: the node's key.
+ * @node: the node to add.
+ *
+ * Return the node replaced upon success. If no node matching the key
+ * was present, return NULL, which also means the operation succeeded.
+ * This replacement operation should never fail.
+ * Call with rcu_read_lock held.
+ * Threads calling this API need to be registered RCU read-side threads.
+ * After successful replacement, a grace period must be waited for before
+ * freeing the memory reserved for the returned node.
+ *
+ * The semantic of replacement vs lookups is the following: if lookups
+ * are performed between a key unique insertion and its removal, we
+ * guarantee that the lookups and get next will always find exactly one
+ * instance of the key if it is replaced concurrently with the lookups.
+ *
+ * Providing this semantic allows us to ensure that replacement-only
+ * schemes will never generate duplicated keys. It also allows us to
+ * guarantee that a combination of add_replace and add_unique updates
+ * will never generate duplicated keys.
+ */
+struct cds_lfht_node *cds_lfht_add_replace(struct cds_lfht *ht,
+ unsigned long hash,
+ cds_lfht_match_fct match,
+ const void *key,
+ struct cds_lfht_node *node);
+
+/*
+ * cds_lfht_replace - replace a node pointer to by iter within hash table.
+ * @ht: the hash table.
+ * @old_iter: the iterator position of the node to replace.
+ * @hash: the node's hash.
+ * @match: the key match function.
+ * @key: the node's key.
+ * @new_node: the new node to use as replacement.
+ *
+ * Return 0 if replacement is successful, negative value otherwise.
+ * Replacing a NULL old node or an already removed node will fail with
+ * -ENOENT.
+ * If the hash or value of the node to replace and the new node differ,
+ * this function returns -EINVAL without proceeding to the replacement.
+ * Old node can be looked up with cds_lfht_lookup and cds_lfht_next.
+ * RCU read-side lock must be held between lookup and replacement.
+ * Call with rcu_read_lock held.
+ * Threads calling this API need to be registered RCU read-side threads.
+ * After successful replacement, a grace period must be waited for before
+ * freeing the memory reserved for the old node (which can be accessed
+ * with cds_lfht_iter_get_node).
+ *
+ * The semantic of replacement vs lookups is the following: if lookups
+ * are performed between a key unique insertion and its removal, we
+ * guarantee that the lookups and get next will always find exactly one
+ * instance of the key if it is replaced concurrently with the lookups.
+ *
+ * Providing this semantic allows us to ensure that replacement-only
+ * schemes will never generate duplicated keys. It also allows us to
+ * guarantee that a combination of add_replace and add_unique updates
+ * will never generate duplicated keys.
+ */
+int cds_lfht_replace(struct cds_lfht *ht,
+ struct cds_lfht_iter *old_iter,
+ unsigned long hash,
+ cds_lfht_match_fct match,
+ const void *key,
+ struct cds_lfht_node *new_node);
+
+/*
+ * cds_lfht_del - remove node pointed to by iterator from hash table.
+ * @ht: the hash table.
+ * @node: the node to delete.
+ *
+ * Return 0 if the node is successfully removed, negative value
+ * otherwise.
+ * Deleting a NULL node or an already removed node will fail with a
+ * negative value.
+ * Node can be looked up with cds_lfht_lookup and cds_lfht_next,
+ * followed by use of cds_lfht_iter_get_node.
+ * RCU read-side lock must be held between lookup and removal.
+ * Call with rcu_read_lock held.
+ * Threads calling this API need to be registered RCU read-side threads.
+ * After successful removal, a grace period must be waited for before
+ * freeing the memory reserved for old node (which can be accessed with
+ * cds_lfht_iter_get_node).
+ */
+int cds_lfht_del(struct cds_lfht *ht, struct cds_lfht_node *node);
+
+/*
+ * cds_lfht_resize - Force a hash table resize
+ * @ht: the hash table.
+ * @new_size: update to this hash table size.
+ *
+ * Threads calling this API need to be registered RCU read-side threads.
+ */
+void cds_lfht_resize(struct cds_lfht *ht, unsigned long new_size);
+
+/*
+ * Note: cds_lfht_for_each are safe for element removal during
+ * iteration.
+ */
+#define cds_lfht_for_each(ht, iter, node) \
+ for (cds_lfht_first(ht, iter), \
+ node = cds_lfht_iter_get_node(iter); \
+ node != NULL; \
+ cds_lfht_next(ht, iter), \
+ node = cds_lfht_iter_get_node(iter))
+
+#define cds_lfht_for_each_duplicate(ht, hash, match, key, iter, node) \
+ for (cds_lfht_lookup(ht, hash, match, key, iter), \
+ node = cds_lfht_iter_get_node(iter); \
+ node != NULL; \
+ cds_lfht_next_duplicate(ht, match, key, iter), \
+ node = cds_lfht_iter_get_node(iter))
+
+#define cds_lfht_for_each_entry(ht, iter, pos, member) \
+ for (cds_lfht_first(ht, iter), \
+ pos = caa_container_of(cds_lfht_iter_get_node(iter), \
+ typeof(*(pos)), member); \
+ &(pos)->member != NULL; \
+ cds_lfht_next(ht, iter), \
+ pos = caa_container_of(cds_lfht_iter_get_node(iter), \
+ typeof(*(pos)), member))
+
+#define cds_lfht_for_each_entry_duplicate(ht, hash, match, key, \
+ iter, pos, member) \
+ for (cds_lfht_lookup(ht, hash, match, key, iter), \
+ pos = caa_container_of(cds_lfht_iter_get_node(iter), \
+ typeof(*(pos)), member); \
+ &(pos)->member != NULL; \
+ cds_lfht_next_duplicate(ht, match, key, iter), \
+ pos = caa_container_of(cds_lfht_iter_get_node(iter), \
+ typeof(*(pos)), member))
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _URCU_RCULFHASH_H */
--- /dev/null
+#ifndef _URCU_FLAVOR_H
+#define _URCU_FLAVOR_H
+
+/*
+ * urcu-flavor.h
+ *
+ * Userspace RCU header - rcu flavor declarations
+ *
+ * Copyright (c) 2011 Lai Jiangshan <laijs@cn.fujitsu.com>
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library 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
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+struct rcu_flavor_struct {
+ void (*read_lock)(void);
+ void (*read_unlock)(void);
+ void (*read_quiescent_state)(void);
+ void (*update_call_rcu)(struct rcu_head *head,
+ void (*func)(struct rcu_head *head));
+ void (*update_synchronize_rcu)(void);
+ void (*update_defer_rcu)(void (*fct)(void *p), void *p);
+
+ void (*thread_offline)(void);
+ void (*thread_online)(void);
+ void (*register_thread)(void);
+ void (*unregister_thread)(void);
+};
+
+#define DEFINE_RCU_FLAVOR(x) \
+const struct rcu_flavor_struct x = { \
+ .read_lock = rcu_read_lock, \
+ .read_unlock = rcu_read_unlock, \
+ .read_quiescent_state = rcu_quiescent_state, \
+ .update_call_rcu = call_rcu, \
+ .update_synchronize_rcu = synchronize_rcu, \
+ .update_defer_rcu = defer_rcu, \
+ .thread_offline = rcu_thread_offline, \
+ .thread_online = rcu_thread_online, \
+ .register_thread = rcu_register_thread, \
+ .unregister_thread = rcu_unregister_thread,\
+}
+
+extern const struct rcu_flavor_struct rcu_flavor;
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _URCU_FLAVOR_H */
--- /dev/null
+/*
+ * Copyright (C) - Bob Jenkins, May 2006, Public Domain.
+ * Copyright (C) 2011 - David Goulet <david.goulet@polymtl.ca>
+ * Copyright (C) 2011 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
+ *
+ * These are functions for producing 32-bit hashes for hash table lookup.
+ * hashword(), hashlittle(), hashlittle2(), hashbig(), mix(), and final() are
+ * externally useful functions. Routines to test the hash are included if
+ * SELF_TEST is defined. You can use this free for any purpose. It's in the
+ * public domain. It has no warranty.
+ *
+ * You probably want to use hashlittle(). hashlittle() and hashbig() hash byte
+ * arrays. hashlittle() is is faster than hashbig() on little-endian machines.
+ * Intel and AMD are little-endian machines. On second thought, you probably
+ * want hashlittle2(), which is identical to hashlittle() except it returns two
+ * 32-bit hashes for the price of one. You could implement hashbig2() if you
+ * wanted but I haven't bothered here.
+ *
+ * If you want to find a hash of, say, exactly 7 integers, do
+ * a = i1; b = i2; c = i3;
+ * mix(a,b,c);
+ * a += i4; b += i5; c += i6;
+ * mix(a,b,c);
+ * a += i7;
+ * final(a,b,c);
+ * then use c as the hash value. If you have a variable length array of
+ * 4-byte integers to hash, use hashword(). If you have a byte array (like
+ * a character string), use hashlittle(). If you have several byte arrays, or
+ * a mix of things, see the comments above hashlittle().
+ *
+ * Why is this so big? I read 12 bytes at a time into 3 4-byte integers, then
+ * mix those integers. This is fast (you can do a lot more thorough mixing
+ * with 12*3 instructions on 3 integers than you can with 3 instructions on 1
+ * byte), but shoehorning those bytes into integers efficiently is messy.
+ */
+
+#include <assert.h>
+#include <endian.h> /* attempt to define endianness */
+#include <stdint.h> /* defines uint32_t etc */
+#include <stdio.h> /* defines printf for tests */
+#include <string.h>
+#include <sys/param.h> /* attempt to define endianness */
+#include <time.h> /* defines time_t for timings in the test */
+#include <urcu/compiler.h>
+
+#include "utils.h"
+
+/*
+ * My best guess at if you are big-endian or little-endian. This may
+ * need adjustment.
+ */
+#if (defined(__BYTE_ORDER) && defined(__LITTLE_ENDIAN) && \
+ __BYTE_ORDER == __LITTLE_ENDIAN) || \
+ (defined(i386) || defined(__i386__) || defined(__i486__) || \
+ defined(__i586__) || defined(__i686__) || defined(vax) || defined(MIPSEL))
+# define HASH_LITTLE_ENDIAN 1
+# define HASH_BIG_ENDIAN 0
+#elif (defined(__BYTE_ORDER) && defined(__BIG_ENDIAN) && \
+ __BYTE_ORDER == __BIG_ENDIAN) || \
+ (defined(sparc) || defined(POWERPC) || defined(mc68000) || defined(sel))
+# define HASH_LITTLE_ENDIAN 0
+# define HASH_BIG_ENDIAN 1
+#else
+# define HASH_LITTLE_ENDIAN 0
+# define HASH_BIG_ENDIAN 0
+#endif
+
+#define hashsize(n) ((uint32_t)1<<(n))
+#define hashmask(n) (hashsize(n)-1)
+#define rot(x,k) (((x)<<(k)) | ((x)>>(32-(k))))
+
+/*
+ * mix -- mix 3 32-bit values reversibly.
+ *
+ * This is reversible, so any information in (a,b,c) before mix() is
+ * still in (a,b,c) after mix().
+ *
+ * If four pairs of (a,b,c) inputs are run through mix(), or through
+ * mix() in reverse, there are at least 32 bits of the output that
+ * are sometimes the same for one pair and different for another pair.
+ * This was tested for:
+ * * pairs that differed by one bit, by two bits, in any combination
+ * of top bits of (a,b,c), or in any combination of bottom bits of
+ * (a,b,c).
+ * * "differ" is defined as +, -, ^, or ~^. For + and -, I transformed
+ * the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
+ * is commonly produced by subtraction) look like a single 1-bit
+ * difference.
+ * * the base values were pseudorandom, all zero but one bit set, or
+ * all zero plus a counter that starts at zero.
+ *
+ * Some k values for my "a-=c; a^=rot(c,k); c+=b;" arrangement that
+ * satisfy this are
+ * 4 6 8 16 19 4
+ * 9 15 3 18 27 15
+ * 14 9 3 7 17 3
+ * Well, "9 15 3 18 27 15" didn't quite get 32 bits diffing
+ * for "differ" defined as + with a one-bit base and a two-bit delta. I
+ * used http://burtleburtle.net/bob/hash/avalanche.html to choose
+ * the operations, constants, and arrangements of the variables.
+ *
+ * This does not achieve avalanche. There are input bits of (a,b,c)
+ * that fail to affect some output bits of (a,b,c), especially of a. The
+ * most thoroughly mixed value is c, but it doesn't really even achieve
+ * avalanche in c.
+ *
+ * This allows some parallelism. Read-after-writes are good at doubling
+ * the number of bits affected, so the goal of mixing pulls in the opposite
+ * direction as the goal of parallelism. I did what I could. Rotates
+ * seem to cost as much as shifts on every machine I could lay my hands
+ * on, and rotates are much kinder to the top and bottom bits, so I used
+ * rotates.
+ */
+#define mix(a,b,c) \
+{ \
+ a -= c; a ^= rot(c, 4); c += b; \
+ b -= a; b ^= rot(a, 6); a += c; \
+ c -= b; c ^= rot(b, 8); b += a; \
+ a -= c; a ^= rot(c,16); c += b; \
+ b -= a; b ^= rot(a,19); a += c; \
+ c -= b; c ^= rot(b, 4); b += a; \
+}
+
+/*
+ * final -- final mixing of 3 32-bit values (a,b,c) into c
+ *
+ * Pairs of (a,b,c) values differing in only a few bits will usually
+ * produce values of c that look totally different. This was tested for
+ * * pairs that differed by one bit, by two bits, in any combination
+ * of top bits of (a,b,c), or in any combination of bottom bits of
+ * (a,b,c).
+ * * "differ" is defined as +, -, ^, or ~^. For + and -, I transformed
+ * the output delta to a Gray code (a^(a>>1)) so a string of 1's (as
+ * is commonly produced by subtraction) look like a single 1-bit
+ * difference.
+ * * the base values were pseudorandom, all zero but one bit set, or
+ * all zero plus a counter that starts at zero.
+ *
+ * These constants passed:
+ * 14 11 25 16 4 14 24
+ * 12 14 25 16 4 14 24
+ * and these came close:
+ * 4 8 15 26 3 22 24
+ * 10 8 15 26 3 22 24
+ * 11 8 15 26 3 22 24
+ */
+#define final(a,b,c) \
+{ \
+ c ^= b; c -= rot(b,14); \
+ a ^= c; a -= rot(c,11); \
+ b ^= a; b -= rot(a,25); \
+ c ^= b; c -= rot(b,16); \
+ a ^= c; a -= rot(c,4); \
+ b ^= a; b -= rot(a,14); \
+ c ^= b; c -= rot(b,24); \
+}
+
+/*
+ * k - the key, an array of uint32_t values
+ * length - the length of the key, in uint32_ts
+ * initval - the previous hash, or an arbitrary value
+ */
+static uint32_t __attribute__((unused)) hashword(const uint32_t *k,
+ size_t length, uint32_t initval)
+{
+ uint32_t a, b, c;
+
+ /* Set up the internal state */
+ a = b = c = 0xdeadbeef + (((uint32_t) length) << 2) + initval;
+
+ /*----------------------------------------- handle most of the key */
+ while (length > 3) {
+ a += k[0];
+ b += k[1];
+ c += k[2];
+ mix(a, b, c);
+ length -= 3;
+ k += 3;
+ }
+
+ /*----------------------------------- handle the last 3 uint32_t's */
+ switch (length) { /* all the case statements fall through */
+ case 3: c += k[2];
+ case 2: b += k[1];
+ case 1: a += k[0];
+ final(a, b, c);
+ case 0: /* case 0: nothing left to add */
+ break;
+ }
+ /*---------------------------------------------- report the result */
+ return c;
+}
+
+
+/*
+ * hashword2() -- same as hashword(), but take two seeds and return two 32-bit
+ * values. pc and pb must both be nonnull, and *pc and *pb must both be
+ * initialized with seeds. If you pass in (*pb)==0, the output (*pc) will be
+ * the same as the return value from hashword().
+ */
+static void __attribute__((unused)) hashword2(const uint32_t *k, size_t length,
+ uint32_t *pc, uint32_t *pb)
+{
+ uint32_t a, b, c;
+
+ /* Set up the internal state */
+ a = b = c = 0xdeadbeef + ((uint32_t) (length << 2)) + *pc;
+ c += *pb;
+
+ while (length > 3) {
+ a += k[0];
+ b += k[1];
+ c += k[2];
+ mix(a, b, c);
+ length -= 3;
+ k += 3;
+ }
+
+ switch (length) {
+ case 3 :
+ c += k[2];
+ case 2 :
+ b += k[1];
+ case 1 :
+ a += k[0];
+ final(a, b, c);
+ case 0: /* case 0: nothing left to add */
+ break;
+ }
+
+ *pc = c;
+ *pb = b;
+}
+
+/*
+ * hashlittle() -- hash a variable-length key into a 32-bit value
+ * k : the key (the unaligned variable-length array of bytes)
+ * length : the length of the key, counting by bytes
+ * initval : can be any 4-byte value
+ * Returns a 32-bit value. Every bit of the key affects every bit of
+ * the return value. Two keys differing by one or two bits will have
+ * totally different hash values.
+ *
+ * The best hash table sizes are powers of 2. There is no need to do
+ * mod a prime (mod is sooo slow!). If you need less than 32 bits,
+ * use a bitmask. For example, if you need only 10 bits, do
+ * h = (h & hashmask(10));
+ * In which case, the hash table should have hashsize(10) elements.
+ *
+ * If you are hashing n strings (uint8_t **)k, do it like this:
+ * for (i=0, h=0; i<n; ++i) h = hashlittle( k[i], len[i], h);
+ *
+ * By Bob Jenkins, 2006. bob_jenkins@burtleburtle.net. You may use this
+ * code any way you wish, private, educational, or commercial. It's free.
+ *
+ * Use for hash table lookup, or anything where one collision in 2^^32 is
+ * acceptable. Do NOT use for cryptographic purposes.
+ */
+static uint32_t __attribute__((unused)) hashlittle(const void *key,
+ size_t length, uint32_t initval)
+{
+ uint32_t a,b,c;
+ union {
+ const void *ptr;
+ size_t i;
+ } u; /* needed for Mac Powerbook G4 */
+
+ /* Set up the internal state */
+ a = b = c = 0xdeadbeef + ((uint32_t)length) + initval;
+
+ u.ptr = key;
+ if (HASH_LITTLE_ENDIAN && ((u.i & 0x3) == 0)) {
+ const uint32_t *k = (const uint32_t *)key; /* read 32-bit chunks */
+
+ /*------ all but last block: aligned reads and affect 32 bits of (a,b,c) */
+ while (length > 12) {
+ a += k[0];
+ b += k[1];
+ c += k[2];
+ mix(a,b,c);
+ length -= 12;
+ k += 3;
+ }
+
+ /*
+ * "k[2]&0xffffff" actually reads beyond the end of the string, but
+ * then masks off the part it's not allowed to read. Because the
+ * string is aligned, the masked-off tail is in the same word as the
+ * rest of the string. Every machine with memory protection I've seen
+ * does it on word boundaries, so is OK with this. But VALGRIND will
+ * still catch it and complain. The masking trick does make the hash
+ * noticably faster for short strings (like English words).
+ */
+#ifndef VALGRIND
+
+ switch (length) {
+ case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
+ case 11: c+=k[2]&0xffffff; b+=k[1]; a+=k[0]; break;
+ case 10: c+=k[2]&0xffff; b+=k[1]; a+=k[0]; break;
+ case 9 : c+=k[2]&0xff; b+=k[1]; a+=k[0]; break;
+ case 8 : b+=k[1]; a+=k[0]; break;
+ case 7 : b+=k[1]&0xffffff; a+=k[0]; break;
+ case 6 : b+=k[1]&0xffff; a+=k[0]; break;
+ case 5 : b+=k[1]&0xff; a+=k[0]; break;
+ case 4 : a+=k[0]; break;
+ case 3 : a+=k[0]&0xffffff; break;
+ case 2 : a+=k[0]&0xffff; break;
+ case 1 : a+=k[0]&0xff; break;
+ case 0 : return c; /* zero length strings require no mixing */
+ }
+#else /* make valgrind happy */
+ const uint8_t *k8;
+
+ k8 = (const uint8_t *)k;
+ switch (length) {
+ case 12: c+=k[2]; b+=k[1]; a+=k[0]; break;
+ case 11: c+=((uint32_t)k8[10])<<16; /* fall through */
+ case 10: c+=((uint32_t)k8[9])<<8; /* fall through */
+ case 9 : c+=k8[8]; /* fall through */
+ case 8 : b+=k[1]; a+=k[0]; break;
+ case 7 : b+=((uint32_t)k8[6])<<16; /* fall through */
+ case 6 : b+=((uint32_t)k8[5])<<8; /* fall through */
+ case 5 : b+=k8[4]; /* fall through */
+ case 4 : a+=k[0]; break;
+ case 3 : a+=((uint32_t)k8[2])<<16; /* fall through */
+ case 2 : a+=((uint32_t)k8[1])<<8; /* fall through */
+ case 1 : a+=k8[0]; break;
+ case 0 : return c;
+ }
+#endif /* !valgrind */
+ } else if (HASH_LITTLE_ENDIAN && ((u.i & 0x1) == 0)) {
+ const uint16_t *k = (const uint16_t *)key; /* read 16-bit chunks */
+ const uint8_t *k8;
+
+ /*--------------- all but last block: aligned reads and different mixing */
+ while (length > 12) {
+ a += k[0] + (((uint32_t)k[1])<<16);
+ b += k[2] + (((uint32_t)k[3])<<16);
+ c += k[4] + (((uint32_t)k[5])<<16);
+ mix(a,b,c);
+ length -= 12;
+ k += 6;
+ }
+
+ k8 = (const uint8_t *)k;
+ switch (length) {
+ case 12:
+ c+=k[4]+(((uint32_t)k[5])<<16);
+ b+=k[2]+(((uint32_t)k[3])<<16);
+ a+=k[0]+(((uint32_t)k[1])<<16);
+ break;
+ case 11:
+ c+=((uint32_t)k8[10])<<16; /* fall through */
+ case 10:
+ c+=k[4];
+ b+=k[2]+(((uint32_t)k[3])<<16);
+ a+=k[0]+(((uint32_t)k[1])<<16);
+ break;
+ case 9:
+ c+=k8[8]; /* fall through */
+ case 8:
+ b+=k[2]+(((uint32_t)k[3])<<16);
+ a+=k[0]+(((uint32_t)k[1])<<16);
+ break;
+ case 7:
+ b+=((uint32_t)k8[6])<<16; /* fall through */
+ case 6:
+ b+=k[2];
+ a+=k[0]+(((uint32_t)k[1])<<16);
+ break;
+ case 5:
+ b+=k8[4]; /* fall through */
+ case 4:
+ a+=k[0]+(((uint32_t)k[1])<<16);
+ break;
+ case 3:
+ a+=((uint32_t)k8[2])<<16; /* fall through */
+ case 2:
+ a+=k[0];
+ break;
+ case 1:
+ a+=k8[0];
+ break;
+ case 0:
+ return c; /* zero length requires no mixing */
+ }
+
+ } else { /* need to read the key one byte at a time */
+ const uint8_t *k = (const uint8_t *)key;
+
+ while (length > 12) {
+ a += k[0];
+ a += ((uint32_t)k[1])<<8;
+ a += ((uint32_t)k[2])<<16;
+ a += ((uint32_t)k[3])<<24;
+ b += k[4];
+ b += ((uint32_t)k[5])<<8;
+ b += ((uint32_t)k[6])<<16;
+ b += ((uint32_t)k[7])<<24;
+ c += k[8];
+ c += ((uint32_t)k[9])<<8;
+ c += ((uint32_t)k[10])<<16;
+ c += ((uint32_t)k[11])<<24;
+ mix(a,b,c);
+ length -= 12;
+ k += 12;
+ }
+
+ switch(length) { /* all the case statements fall through */
+ case 12: c+=((uint32_t)k[11])<<24;
+ case 11: c+=((uint32_t)k[10])<<16;
+ case 10: c+=((uint32_t)k[9])<<8;
+ case 9: c+=k[8];
+ case 8: b+=((uint32_t)k[7])<<24;
+ case 7: b+=((uint32_t)k[6])<<16;
+ case 6: b+=((uint32_t)k[5])<<8;
+ case 5: b+=k[4];
+ case 4: a+=((uint32_t)k[3])<<24;
+ case 3: a+=((uint32_t)k[2])<<16;
+ case 2: a+=((uint32_t)k[1])<<8;
+ case 1:
+ a+=k[0];
+ break;
+ case 0:
+ return c;
+ }
+ }
+
+ final(a,b,c);
+ return c;
+}
+
+#if (CAA_BITS_PER_LONG == 64)
+/*
+ * Hash function for number value.
+ */
+unsigned long hash_key_ulong(void *_key, unsigned long seed)
+{
+ union {
+ uint64_t v64;
+ uint32_t v32[2];
+ } v;
+ union {
+ uint64_t v64;
+ uint32_t v32[2];
+ } key;
+
+ v.v64 = (uint64_t) seed;
+ key.v64 = (uint64_t) _key;
+ hashword2(key.v32, 2, &v.v32[0], &v.v32[1]);
+ return v.v64;
+}
+#else
+/*
+ * Hash function for number value.
+ */
+unsigned long hash_key_ulong(void *_key, unsigned long seed)
+{
+ uint32_t key = (uint32_t) _key;
+
+ return hashword(&key, 1, seed);
+}
+#endif /* CAA_BITS_PER_LONG */
+
+/*
+ * Hash function for string.
+ */
+unsigned long hash_key_str(void *key, unsigned long seed)
+{
+ return hashlittle(key, strlen((char *) key), seed);
+}
+
+/*
+ * Hash function compare for number value.
+ */
+int hash_match_key_ulong(void *key1, void *key2)
+{
+ if (key1 == key2) {
+ return 1;
+ }
+
+ return 0;
+}
+
+/*
+ * Hash compare function for string.
+ */
+int hash_match_key_str(void *key1, void *key2)
+{
+ if (strcmp(key1, key2) == 0) {
+ return 1;
+ }
+
+ return 0;
+}
--- /dev/null
+/*
+ * Copyright (C) 2011 - David Goulet <david.goulet@polymtl.ca>
+ *
+ * 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; only version 2 of the License.
+ *
+ * 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.
+ */
+
+#ifndef _LTT_HT_UTILS_H
+#define _LTT_HT_UTILS_H
+
+#include <stdint.h>
+
+unsigned long hash_key_ulong(void *_key, unsigned long seed);
+unsigned long hash_key_str(void *key, unsigned long seed);
+int hash_match_key_ulong(void *key1, void *key2);
+int hash_match_key_str(void *key1, void *key2);
+
+#endif /* _LTT_HT_UTILS_H */
#include <lttng-sessiond-comm.h>
#include <lttng/lttng-kconsumer.h>
#include <lttngerr.h>
+#include <runas.h>
extern struct lttng_consumer_global_data consumer_data;
extern int consumer_poll_timeout;
msg.u.stream.state,
msg.u.stream.mmap_len,
msg.u.stream.output,
- msg.u.stream.path_name);
+ msg.u.stream.path_name,
+ msg.u.stream.uid,
+ msg.u.stream.gid);
if (new_stream == NULL) {
lttng_consumer_send_error(ctx, CONSUMERD_OUTFD_ERROR);
goto end;
/* Opening the tracefile in write mode */
if (stream->path_name != NULL) {
- ret = open(stream->path_name,
- O_WRONLY|O_CREAT|O_TRUNC, S_IRWXU|S_IRWXG|S_IRWXO);
+ ret = open_run_as(stream->path_name,
+ O_WRONLY|O_CREAT|O_TRUNC,
+ S_IRWXU|S_IRWXG|S_IRWXO,
+ stream->uid, stream->gid);
if (ret < 0) {
ERR("Opening %s", stream->path_name);
perror("open");
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
-#include <sys/socket.h>
#include <sys/stat.h>
#include <sys/types.h>
#include <sys/un.h>
[ LTTCOMM_ERR_INDEX(LTTCOMM_EXIST_SESS) ] = "Session name already exist",
[ LTTCOMM_ERR_INDEX(LTTCOMM_CONNECT_FAIL) ] = "Unable to connect to Unix socket",
[ LTTCOMM_ERR_INDEX(LTTCOMM_APP_NOT_FOUND) ] = "Application not found",
+ [ LTTCOMM_ERR_INDEX(LTTCOMM_EPERM) ] = "Permission denied",
[ LTTCOMM_ERR_INDEX(LTTCOMM_KERN_NA) ] = "Kernel tracer not available",
[ LTTCOMM_ERR_INDEX(LTTCOMM_KERN_EVENT_EXIST) ] = "Kernel event already exists",
[ LTTCOMM_ERR_INDEX(LTTCOMM_KERN_SESS_FAIL) ] = "Kernel create session failed",
new_fd = accept(sock, (struct sockaddr *) &sun, &len);
if (new_fd < 0) {
perror("accept");
- goto error;
}
return new_fd;
-
-error:
- return -1;
}
/*
end:
return ret;
}
+
+/*
+ * Send a message with credentials over a unix socket.
+ *
+ * Returns the size of data sent, or negative error value.
+ */
+ssize_t lttcomm_send_creds_unix_sock(int sock, void *buf, size_t len)
+{
+ struct msghdr msg = { 0 };
+ struct cmsghdr *cmptr;
+ struct iovec iov[1];
+ ssize_t ret = -1;
+ struct ucred *creds;
+ size_t sizeof_cred = sizeof(struct ucred);
+ char anc_buf[CMSG_SPACE(sizeof_cred)];
+
+ iov[0].iov_base = buf;
+ iov[0].iov_len = len;
+ msg.msg_iov = iov;
+ msg.msg_iovlen = 1;
+
+ msg.msg_control = (caddr_t) anc_buf;
+ msg.msg_controllen = CMSG_LEN(sizeof_cred);
+
+ cmptr = CMSG_FIRSTHDR(&msg);
+ cmptr->cmsg_level = SOL_SOCKET;
+ cmptr->cmsg_type = SCM_CREDENTIALS;
+ cmptr->cmsg_len = CMSG_LEN(sizeof_cred);
+
+ creds = (struct ucred *) CMSG_DATA(cmptr);
+
+ creds->uid = geteuid();
+ creds->gid = getegid();
+ creds->pid = getpid();
+
+ ret = sendmsg(sock, &msg, 0);
+ if (ret < 0) {
+ perror("sendmsg");
+ }
+
+ return ret;
+}
+
+/*
+ * Recv a message accompanied with credentials from a unix socket.
+ *
+ * Returns the size of received data, or negative error value.
+ */
+ssize_t lttcomm_recv_creds_unix_sock(int sock, void *buf, size_t len,
+ struct ucred *creds)
+{
+ struct msghdr msg = { 0 };
+ struct cmsghdr *cmptr;
+ struct iovec iov[1];
+ ssize_t ret;
+ size_t sizeof_cred = sizeof(struct ucred);
+ char anc_buf[CMSG_SPACE(sizeof_cred)];
+
+ /* Not allowed */
+ if (creds == NULL) {
+ ret = -1;
+ goto end;
+ }
+
+ /* Prepare to receive the structures */
+ iov[0].iov_base = buf;
+ iov[0].iov_len = len;
+ msg.msg_iov = iov;
+ msg.msg_iovlen = 1;
+
+ msg.msg_control = anc_buf;
+ msg.msg_controllen = sizeof(anc_buf);
+
+ ret = recvmsg(sock, &msg, 0);
+ if (ret < 0) {
+ perror("recvmsg fds");
+ goto end;
+ }
+
+ if (msg.msg_flags & MSG_CTRUNC) {
+ fprintf(stderr, "Error: Control message truncated.\n");
+ ret = -1;
+ goto end;
+ }
+
+ cmptr = CMSG_FIRSTHDR(&msg);
+ if (cmptr == NULL) {
+ fprintf(stderr, "Error: Invalid control message header\n");
+ ret = -1;
+ goto end;
+ }
+
+ if (cmptr->cmsg_level != SOL_SOCKET ||
+ cmptr->cmsg_type != SCM_CREDENTIALS) {
+ fprintf(stderr, "Didn't received any credentials\n");
+ ret = -1;
+ goto end;
+ }
+
+ if (cmptr->cmsg_len != CMSG_LEN(sizeof_cred)) {
+ fprintf(stderr, "Error: Received %zu bytes of ancillary data, expected %zu\n",
+ cmptr->cmsg_len, CMSG_LEN(sizeof_cred));
+ ret = -1;
+ goto end;
+ }
+
+ memcpy(creds, CMSG_DATA(cmptr), sizeof_cred);
+
+end:
+ return ret;
+}
+
+/*
+ * Set socket option to use credentials passing.
+ */
+int lttcomm_setsockopt_creds_unix_sock(int sock)
+{
+ int ret, on = 1;
+
+ /* Set socket for credentials retrieval */
+ ret = setsockopt(sock, SOL_SOCKET, SO_PASSCRED, &on, sizeof(on));
+ if (ret < 0) {
+ perror("setsockopt creds unix sock");
+ }
+
+ return ret;
+}
liblttng_ustconsumer_la_SOURCES = lttng-ustconsumer.c
liblttng_ustconsumer_la_LIBADD = -llttng-ust-ctl
+
endif
#include <lttng/lttng-ustconsumer.h>
#include <lttng/ust-ctl.h>
#include <lttngerr.h>
+#include <runas.h>
extern struct lttng_consumer_global_data consumer_data;
extern int consumer_poll_timeout;
msg.u.stream.state,
msg.u.stream.mmap_len,
msg.u.stream.output,
- msg.u.stream.path_name);
+ msg.u.stream.path_name,
+ msg.u.stream.uid,
+ msg.u.stream.gid);
if (new_stream == NULL) {
lttng_consumer_send_error(ctx, CONSUMERD_OUTFD_ERROR);
goto end;
/* Opening the tracefile in write mode */
if (stream->path_name != NULL) {
- ret = open(stream->path_name,
- O_WRONLY|O_CREAT|O_TRUNC, S_IRWXU|S_IRWXG|S_IRWXO);
+ ret = open_run_as(stream->path_name,
+ O_WRONLY|O_CREAT|O_TRUNC,
+ S_IRWXU|S_IRWXG|S_IRWXO,
+ stream->uid, stream->gid);
if (ret < 0) {
ERR("Opening %s", stream->path_name);
perror("open");
goto end;
}
- ret = lttcomm_send_unix_sock(sessiond_socket, lsm,
+ ret = lttcomm_send_creds_unix_sock(sessiond_socket, lsm,
sizeof(struct lttcomm_session_msg));
end:
lttng_consumerd_LDADD = \
$(top_builddir)/libkernelctl/libkernelctl.la \
$(top_builddir)/liblttng-consumer/liblttng-consumer.la \
- $(top_builddir)/liblttng-sessiond-comm/liblttng-sessiond-comm.la
+ $(top_builddir)/liblttng-sessiond-comm/liblttng-sessiond-comm.la \
+ $(top_builddir)/common/libcommon.la
if HAVE_LIBLTTNG_UST_CTL
lttng_consumerd_LDADD += -llttng-ust-ctl
endif
lttng_sessiond_SOURCES = utils.c utils.h \
- hashtable.c hashtable.h \
compat/poll.h $(COMPAT) \
trace-kernel.c trace-kernel.h \
kernel.c kernel.h \
futex.c futex.h \
shm.c shm.h \
session.c session.h \
- ../hashtable/rculfhash.c \
- ../hashtable/rculfhash.h \
- ../hashtable/hash.c ../hashtable/hash.h
+ lttng-ust-ctl.h lttng-ust-abi.h
if HAVE_LIBLTTNG_UST_CTL
lttng_sessiond_SOURCES += trace-ust.c ust-app.c ust-consumer.c ust-consumer.h
$(top_builddir)/liblttng-sessiond-comm/liblttng-sessiond-comm.la \
$(top_builddir)/libkernelctl/libkernelctl.la \
$(top_builddir)/liblttngctl/liblttngctl.la \
- $(top_builddir)/benchmark/liblttng-benchmark.la
+ $(top_builddir)/benchmark/liblttng-benchmark.la \
+ $(top_builddir)/common/libcommon.la \
+ $(top_builddir)/liblttng-ht/liblttng-ht.la
if HAVE_LIBLTTNG_UST_CTL
lttng_sessiond_LDADD += -llttng-ust-ctl
#include <unistd.h>
#include <lttng/lttng.h>
+#include <lttng-ht.h>
#include <lttng-sessiond-comm.h>
#include <lttngerr.h>
#include "channel.h"
-#include "hashtable.h"
#include "kernel.h"
#include "ust-ctl.h"
#include "utils.h"
struct lttng_channel *attr)
{
int ret = LTTCOMM_OK;
- struct cds_lfht *chan_ht;
+ struct lttng_ht *chan_ht;
struct ltt_ust_channel *uchan = NULL;
struct lttng_channel *defattr = NULL;
}
uchan->enabled = 1;
- hashtable_add_unique(chan_ht, &uchan->node);
+ lttng_ht_add_unique_str(chan_ht, &uchan->node);
DBG2("Channel %s created successfully", uchan->name);
free(defattr);
#include <unistd.h>
#include <urcu/list.h>
+#include <lttng-ht.h>
#include <lttng-sessiond-comm.h>
#include <lttngerr.h>
#include "context.h"
-#include "hashtable.h"
#include "kernel.h"
#include "ust-app.h"
#include "trace-ust.h"
}
/* Add ltt UST context node to ltt UST channel */
- hashtable_add_unique(uchan->ctx, &uctx->node);
+ lttng_ht_add_unique_ulong(uchan->ctx, &uctx->node);
return LTTCOMM_OK;
}
/* Add ltt UST context node to ltt UST event */
- hashtable_add_unique(uevent->ctx, &uctx->node);
+ lttng_ht_add_unique_ulong(uevent->ctx, &uctx->node);
return LTTCOMM_OK;
char *channel_name)
{
int ret = LTTCOMM_OK, have_event = 0;
- struct cds_lfht_iter iter, uiter;
- struct cds_lfht *chan_ht;
+ struct lttng_ht_iter iter, uiter;
+ struct lttng_ht *chan_ht;
struct ltt_ust_channel *uchan = NULL;
struct ltt_ust_event *uevent = NULL;
ret = add_uctx_to_channel(usess, domain, uchan, ctx);
} else if (!uchan && have_event) { /* Add ctx to event */
/* Add context to event without having the channel name */
- cds_lfht_for_each_entry(chan_ht, &iter, uchan, node) {
+ cds_lfht_for_each_entry(chan_ht->ht, &iter.iter, uchan, node.node) {
uevent = trace_ust_find_event_by_name(uchan->events, event_name);
if (uevent != NULL) {
ret = add_uctx_to_event(usess, domain, uchan, uevent, ctx);
goto error;
} else if (!uchan && !have_event) { /* Add ctx all events, all channels */
/* For all channels */
- cds_lfht_for_each_entry(chan_ht, &iter, uchan, node) {
+ cds_lfht_for_each_entry(chan_ht->ht, &iter.iter, uchan, node.node) {
ret = add_uctx_to_channel(usess, domain, uchan, ctx);
if (ret < 0) {
ERR("Context added to channel %s failed", uchan->name);
}
/* For all events in channel */
- cds_lfht_for_each_entry(uchan->events, &uiter, uevent, node) {
+ cds_lfht_for_each_entry(uchan->events->ht, &uiter.iter, uevent,
+ node.node) {
ret = add_uctx_to_event(usess, domain, uchan, uevent, ctx);
if (ret < 0) {
ERR("Context add to event %s in channel %s failed",
* Place - Suite 330, Boston, MA 02111-1307, USA.
*/
+#define _GNU_SOURCE
#include <errno.h>
#include <urcu/list.h>
#include <string.h>
#include <lttng/lttng.h>
+#include <lttng-ht.h>
#include <lttng-sessiond-comm.h>
#include <lttngerr.h>
#include "channel.h"
#include "event.h"
-#include "hashtable.h"
#include "kernel.h"
#include "ust-ctl.h"
#include "ust-app.h"
{
int ret, i;
size_t size;
- struct cds_lfht_iter iter;
+ struct lttng_ht_iter iter;
struct ltt_ust_event *uevent = NULL;
struct lttng_event *events = NULL;
case LTTNG_DOMAIN_UST:
{
/* Enable existing events */
- cds_lfht_for_each_entry(uchan->events, &iter, uevent, node) {
+ cds_lfht_for_each_entry(uchan->events->ht, &iter.iter, uevent,
+ node.node) {
if (uevent->enabled == 0) {
ret = ust_app_enable_event_glb(usess, uchan, uevent);
if (ret < 0) {
uevent->enabled = 1;
/* Add ltt ust event to channel */
rcu_read_lock();
- hashtable_add_unique(uchan->events, &uevent->node);
+ lttng_ht_add_unique_str(uchan->events, &uevent->node);
rcu_read_unlock();
}
/* Add ltt ust event to channel */
if (to_create) {
rcu_read_lock();
- hashtable_add_unique(uchan->events, &uevent->node);
+ lttng_ht_add_unique_str(uchan->events, &uevent->node);
rcu_read_unlock();
}
{
int ret, i;
size_t size;
- struct cds_lfht_iter iter;
+ struct lttng_ht_iter iter;
struct ltt_ust_event *uevent = NULL;
struct lttng_event *events = NULL;
case LTTNG_DOMAIN_UST:
{
/* Disabling existing events */
- cds_lfht_for_each_entry(uchan->events, &iter, uevent, node) {
+ cds_lfht_for_each_entry(uchan->events->ht, &iter.iter, uevent,
+ node.node) {
if (uevent->enabled == 1) {
ret = ust_app_disable_event_glb(usess, uchan, uevent);
if (ret < 0) {
+++ /dev/null
-/*
- * Copyright (C) 2011 - David Goulet <david.goulet@polymtl.ca>
- *
- * 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; only version 2 of the License.
- *
- * 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.
- */
-
-#include <urcu.h>
-
-#include <lttng-share.h>
-
-#include "hashtable.h"
-#include "../hashtable/rculfhash.h"
-#include "../hashtable/hash.h"
-
-struct cds_lfht *hashtable_new(unsigned long size)
-{
- if (size == 0) {
- size = DEFAULT_HT_SIZE;
- }
-
- return cds_lfht_new(hash_key, hash_compare_key, 0x42UL,
- size, size, CDS_LFHT_AUTO_RESIZE, NULL);
-}
-
-struct cds_lfht *hashtable_new_str(unsigned long size)
-{
- if (size == 0) {
- size = DEFAULT_HT_SIZE;
- }
-
- return cds_lfht_new(hash_key_str, hash_compare_key_str, 0x42UL,
- size, size, CDS_LFHT_AUTO_RESIZE, NULL);
-}
-
-struct cds_lfht_node *hashtable_iter_get_node(struct cds_lfht_iter *iter)
-{
- /* Safety net */
- if (iter == NULL) {
- return NULL;
- }
-
- return cds_lfht_iter_get_node(iter);
-}
-
-struct cds_lfht_node *hashtable_lookup(struct cds_lfht *ht, void *key,
- size_t key_len, struct cds_lfht_iter *iter)
-{
- /* Safety net */
- if (ht == NULL || iter == NULL || key == NULL) {
- return NULL;
- }
-
- cds_lfht_lookup(ht, key, key_len, iter);
-
- return hashtable_iter_get_node(iter);
-}
-
-void hashtable_get_first(struct cds_lfht *ht, struct cds_lfht_iter *iter)
-{
- cds_lfht_first(ht, iter);
-}
-
-void hashtable_get_next(struct cds_lfht *ht, struct cds_lfht_iter *iter)
-{
- cds_lfht_next(ht, iter);
-}
-
-void hashtable_add_unique(struct cds_lfht *ht, struct cds_lfht_node *node)
-{
- cds_lfht_add_unique(ht, node);
-}
-
-void hashtable_node_init(struct cds_lfht_node *node, void *key,
- size_t key_len)
-{
- cds_lfht_node_init(node, key, key_len);
-}
-
-int hashtable_del(struct cds_lfht *ht, struct cds_lfht_iter *iter)
-{
- /* Safety net */
- if (ht == NULL || iter == NULL) {
- return -1;
- }
-
- return cds_lfht_del(ht, iter);
-}
-
-unsigned long hashtable_get_count(struct cds_lfht *ht)
-{
- long ab, aa;
- unsigned long count, removed;
-
- cds_lfht_count_nodes(ht, &ab, &count, &removed, &aa);
-
- return count;
-}
-
-int hashtable_destroy(struct cds_lfht *ht)
-{
- return cds_lfht_destroy(ht, NULL);
-}
+++ /dev/null
-/*
- * Copyright (C) 2011 - David Goulet <david.goulet@polymtl.ca>
- *
- * 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; only version 2 of the License.
- *
- * 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.
- */
-
-#ifndef _LTT_HASHTABLE_H
-#define _LTT_HASHTABLE_H
-
-#include <urcu.h>
-#include "../hashtable/rculfhash.h"
-
-struct cds_lfht *hashtable_new(unsigned long size);
-struct cds_lfht *hashtable_new_str(unsigned long size);
-
-struct cds_lfht_node *hashtable_iter_get_node(struct cds_lfht_iter *iter);
-struct cds_lfht_node *hashtable_lookup(struct cds_lfht *ht, void *key,
- size_t key_len, struct cds_lfht_iter *iter);
-
-void hashtable_get_first(struct cds_lfht *ht, struct cds_lfht_iter *iter);
-void hashtable_get_next(struct cds_lfht *ht, struct cds_lfht_iter *iter);
-void hashtable_add_unique(struct cds_lfht *ht, struct cds_lfht_node *node);
-void hashtable_node_init(struct cds_lfht_node *node,
- void *key, size_t key_len);
-
-int hashtable_del(struct cds_lfht *ht, struct cds_lfht_iter *iter);
-unsigned long hashtable_get_count(struct cds_lfht *ht);
-int hashtable_destroy(struct cds_lfht *ht);
-
-#endif /* _LTT_HASHTABLE_H */
struct command_ctx {
int ust_sock;
unsigned int lttng_msg_size;
+ struct ucred creds;
struct ltt_session *session;
struct lttcomm_lttng_msg *llm;
struct lttcomm_session_msg *lsm;
/* Release object created by members of this API */
int ustctl_release_object(int sock, struct lttng_ust_object_data *data);
+/* Release handle returned by create session. */
+int ustctl_release_handle(int sock, int handle);
#endif /* _LTTNG_UST_CTL_H */
#include <config.h>
#include <lttng-consumerd.h>
+#include <lttng-ht.h>
#include <lttng-sessiond-comm.h>
#include <lttng/lttng-consumer.h>
#include <lttngerr.h>
+#include <runas.h>
#include "channel.h"
#include "compat/poll.h"
#include "context.h"
#include "event.h"
#include "futex.h"
-#include "hashtable.h"
#include "kernel.h"
#include "lttng-sessiond.h"
#include "shm.h"
const char *opt_tracing_group;
static int opt_sig_parent;
static int opt_daemon;
+static int opt_no_kernel;
static int is_root; /* Set to 1 if the daemon is running as root */
static pid_t ppid; /* Parent PID for --sig-parent option */
static char *rundir;
pthread_mutex_destroy(&kconsumer_data.pid_mutex);
- DBG("Closing kernel fd");
- close(kernel_tracer_fd);
-
- if (is_root) {
+ if (is_root && !opt_no_kernel) {
+ DBG2("Closing kernel fd");
+ close(kernel_tracer_fd);
DBG("Unloading kernel modules");
modprobe_remove_kernel_modules();
}
/* END BENCHMARK */
/* <fun> */
- MSG("%c[%d;%dm*** assert failed :-) *** ==> %c[%dm%c[%d;%dm"
+ DBG("%c[%d;%dm*** assert failed :-) *** ==> %c[%dm%c[%d;%dm"
"Matthew, BEET driven development works!%c[%dm",
27, 1, 31, 27, 0, 27, 1, 33, 27, 0);
/* </fun> */
* Send all stream fds of kernel channel to the consumer.
*/
static int send_kconsumer_channel_streams(struct consumer_data *consumer_data,
- int sock, struct ltt_kernel_channel *channel)
+ int sock, struct ltt_kernel_channel *channel,
+ uid_t uid, gid_t gid)
{
int ret;
struct ltt_kernel_stream *stream;
lkm.u.stream.state = stream->state;
lkm.u.stream.output = channel->channel->attr.output;
lkm.u.stream.mmap_len = 0; /* for kernel */
+ lkm.u.stream.uid = uid;
+ lkm.u.stream.gid = gid;
strncpy(lkm.u.stream.path_name, stream->pathname, PATH_MAX - 1);
lkm.u.stream.path_name[PATH_MAX - 1] = '\0';
DBG("Sending stream %d to consumer", lkm.u.stream.stream_key);
lkm.u.stream.state = LTTNG_CONSUMER_ACTIVE_STREAM;
lkm.u.stream.output = DEFAULT_KERNEL_CHANNEL_OUTPUT;
lkm.u.stream.mmap_len = 0; /* for kernel */
+ lkm.u.stream.uid = session->uid;
+ lkm.u.stream.gid = session->gid;
strncpy(lkm.u.stream.path_name, session->metadata->pathname, PATH_MAX - 1);
lkm.u.stream.path_name[PATH_MAX - 1] = '\0';
DBG("Sending metadata stream %d to consumer", lkm.u.stream.stream_key);
}
cds_list_for_each_entry(chan, &session->channel_list.head, list) {
- ret = send_kconsumer_channel_streams(consumer_data, sock, chan);
+ ret = send_kconsumer_channel_streams(consumer_data, sock, chan,
+ session->uid, session->gid);
if (ret < 0) {
goto error;
}
*/
if (session->kernel_session->consumer_fds_sent == 1) {
ret = send_kconsumer_channel_streams(consumer_data,
- session->kernel_session->consumer_fd, channel);
+ session->kernel_session->consumer_fd, channel,
+ session->uid, session->gid);
if (ret < 0) {
goto error;
}
*/
update_ust_app(ust_cmd.sock);
- ret = ustctl_register_done(ust_cmd.sock);
+ ret = ust_app_register_done(ust_cmd.sock);
if (ret < 0) {
/*
* If the registration is not possible, we simply
}
}
DBG("Using 64-bit UST consumer at: %s", consumerd64_bin);
- ret = execl(consumerd64_bin, verbosity, "-u",
+ ret = execl(consumerd64_bin, "lttng-consumerd", verbosity, "-u",
"--consumerd-cmd-sock", consumer_data->cmd_unix_sock_path,
"--consumerd-err-sock", consumer_data->err_unix_sock_path,
NULL);
}
}
DBG("Using 32-bit UST consumer at: %s", consumerd32_bin);
- ret = execl(consumerd32_bin, verbosity, "-u",
+ ret = execl(consumerd32_bin, "lttng-consumerd", verbosity, "-u",
"--consumerd-cmd-sock", consumer_data->cmd_unix_sock_path,
"--consumerd-err-sock", consumer_data->err_unix_sock_path,
NULL);
int ret;
char *type = "debugfs";
- ret = mkdir_recursive(path, S_IRWXU | S_IRWXG, geteuid(), getegid());
+ ret = mkdir_recursive_run_as(path, S_IRWXU | S_IRWXG, geteuid(), getegid());
if (ret < 0) {
PERROR("Cannot create debugfs path");
goto error;
static int create_ust_session(struct ltt_session *session,
struct lttng_domain *domain)
{
- int ret;
- unsigned int uid;
struct ltt_ust_session *lus = NULL;
+ int ret;
switch (domain->type) {
case LTTNG_DOMAIN_UST:
DBG("Creating UST session");
- uid = session->uid;
- lus = trace_ust_create_session(session->path, uid, domain);
+ lus = trace_ust_create_session(session->path, session->id, domain);
if (lus == NULL) {
ret = LTTCOMM_UST_SESS_FAIL;
goto error;
}
- ret = mkdir_recursive(lus->pathname, S_IRWXU | S_IRWXG,
- geteuid(), allowed_group());
+ ret = mkdir_recursive_run_as(lus->pathname, S_IRWXU | S_IRWXG,
+ session->uid, session->gid);
if (ret < 0) {
if (ret != -EEXIST) {
ERR("Trace directory creation error");
ERR("Unknown UST domain on create session %d", domain->type);
goto error;
}
+ lus->uid = session->uid;
+ lus->gid = session->gid;
session->ust_session = lus;
return LTTCOMM_OK;
session->kernel_session->consumer_fd = kconsumer_data.cmd_sock;
}
- ret = mkdir_recursive(session->kernel_session->trace_path,
- S_IRWXU | S_IRWXG, geteuid(), allowed_group());
+ ret = mkdir_recursive_run_as(session->kernel_session->trace_path,
+ S_IRWXU | S_IRWXG, session->uid, session->gid);
if (ret < 0) {
if (ret != -EEXIST) {
ERR("Trace directory creation error");
goto error;
}
}
+ session->kernel_session->uid = session->uid;
+ session->kernel_session->gid = session->gid;
error:
return ret;
}
+/*
+ * Check if the UID or GID match the session. Root user has access to
+ * all sessions.
+ */
+static int session_access_ok(struct ltt_session *session,
+ uid_t uid, gid_t gid)
+{
+ if (uid != session->uid && gid != session->gid
+ && uid != 0) {
+ return 0;
+ } else {
+ return 1;
+ }
+}
+
+static unsigned int lttng_sessions_count(uid_t uid, gid_t gid)
+{
+ unsigned int i = 0;
+ struct ltt_session *session;
+
+ DBG("Counting number of available session for UID %d GID %d",
+ uid, gid);
+ cds_list_for_each_entry(session, &session_list_ptr->head, list) {
+ /*
+ * Only list the sessions the user can control.
+ */
+ if (!session_access_ok(session, uid, gid)) {
+ continue;
+ }
+ i++;
+ }
+ return i;
+}
+
/*
* Using the session list, filled a lttng_session array to send back to the
* client for session listing.
* The session list lock MUST be acquired before calling this function. Use
* session_lock_list() and session_unlock_list().
*/
-static void list_lttng_sessions(struct lttng_session *sessions)
+static void list_lttng_sessions(struct lttng_session *sessions,
+ uid_t uid, gid_t gid)
{
- int i = 0;
+ unsigned int i = 0;
struct ltt_session *session;
- DBG("Getting all available session");
+ DBG("Getting all available session for UID %d GID %d",
+ uid, gid);
/*
* Iterate over session list and append data after the control struct in
* the buffer.
*/
cds_list_for_each_entry(session, &session_list_ptr->head, list) {
+ /*
+ * Only list the sessions the user can control.
+ */
+ if (!session_access_ok(session, uid, gid)) {
+ continue;
+ }
strncpy(sessions[i].path, session->path, PATH_MAX);
sessions[i].path[PATH_MAX - 1] = '\0';
strncpy(sessions[i].name, session->name, NAME_MAX);
break;
case LTTNG_DOMAIN_UST:
{
- struct cds_lfht_iter iter;
+ struct lttng_ht_iter iter;
struct ltt_ust_channel *uchan;
- cds_lfht_for_each_entry(session->ust_session->domain_global.channels,
- &iter, uchan, node) {
+ cds_lfht_for_each_entry(session->ust_session->domain_global.channels->ht,
+ &iter.iter, uchan, node.node) {
strncpy(channels[i].name, uchan->name, LTTNG_SYMBOL_NAME_LEN);
channels[i].attr.overwrite = uchan->attr.overwrite;
channels[i].attr.subbuf_size = uchan->attr.subbuf_size;
{
int i = 0, ret = 0;
unsigned int nb_event = 0;
- struct cds_lfht_iter iter;
- struct cds_lfht_node *node;
+ struct lttng_ht_iter iter;
+ struct lttng_ht_node_str *node;
struct ltt_ust_channel *uchan;
struct ltt_ust_event *uevent;
struct lttng_event *tmp;
rcu_read_lock();
- node = hashtable_lookup(ust_global->channels, (void *) channel_name,
- strlen(channel_name), &iter);
+ lttng_ht_lookup(ust_global->channels, (void *)channel_name, &iter);
+ node = lttng_ht_iter_get_node_str(&iter);
if (node == NULL) {
ret = -LTTCOMM_UST_CHAN_NOT_FOUND;
goto error;
}
- uchan = caa_container_of(node, struct ltt_ust_channel, node);
+ uchan = caa_container_of(&node->node, struct ltt_ust_channel, node.node);
- nb_event += hashtable_get_count(uchan->events);
+ nb_event += lttng_ht_get_count(uchan->events);
if (nb_event == 0) {
ret = nb_event;
goto error;
}
- cds_lfht_for_each_entry(uchan->events, &iter, uevent, node) {
+ cds_lfht_for_each_entry(uchan->events->ht, &iter.iter, uevent, node.node) {
strncpy(tmp[i].name, uevent->attr.name, LTTNG_SYMBOL_NAME_LEN);
tmp[i].name[LTTNG_SYMBOL_NAME_LEN - 1] = '\0';
tmp[i].enabled = uevent->enabled;
case LTTNG_DOMAIN_UST:
{
struct ltt_ust_channel *uchan;
- struct cds_lfht *chan_ht;
+ struct lttng_ht *chan_ht;
chan_ht = usess->domain_global.channels;
{
int ret;
struct ltt_ust_session *usess = session->ust_session;
- struct cds_lfht *chan_ht;
+ struct lttng_ht *chan_ht;
DBG("Enabling channel %s for session %s", attr->name, session->name);
}
/* At this point, the session and channel exist on the tracer */
-
ret = event_ust_enable_tracepoint(usess, domain, uchan, event);
if (ret != LTTCOMM_OK) {
goto error;
ksession = session->kernel_session;
usess = session->ust_session;
- if (session->enabled)
- return LTTCOMM_UST_START_FAIL;
+ if (session->enabled) {
+ ret = LTTCOMM_UST_START_FAIL;
+ goto error;
+ }
+
session->enabled = 1;
/* Kernel tracing */
ksession = session->kernel_session;
usess = session->ust_session;
- if (!session->enabled)
- return LTTCOMM_UST_START_FAIL;
+ if (!session->enabled) {
+ ret = LTTCOMM_UST_START_FAIL;
+ goto error;
+ }
+
session->enabled = 0;
/* Kernel tracer */
/*
* Command LTTNG_CREATE_SESSION processed by the client thread.
*/
-static int cmd_create_session(char *name, char *path)
+static int cmd_create_session(char *name, char *path, struct ucred *creds)
{
int ret;
- ret = session_create(name, path);
+ ret = session_create(name, path, creds->uid, creds->gid);
if (ret != LTTCOMM_OK) {
goto error;
}
switch (domain) {
case LTTNG_DOMAIN_KERNEL:
/* Can't register a consumer if there is already one */
- if (session->kernel_session->consumer_fd != 0) {
+ if (session->kernel_session->consumer_fds_sent != 0) {
ret = LTTCOMM_KERN_CONSUMER_FAIL;
goto error;
}
break;
case LTTNG_DOMAIN_UST:
if (session->ust_session != NULL) {
- nb_chan = hashtable_get_count(
+ nb_chan = lttng_ht_get_count(
session->ust_session->domain_global.channels);
}
DBG3("Number of UST global channels %zd", nb_chan);
DBG("Processing client command %d", cmd_ctx->lsm->cmd_type);
+ if (opt_no_kernel && cmd_ctx->lsm->domain.type == LTTNG_DOMAIN_KERNEL) {
+ ret = LTTCOMM_KERN_NA;
+ goto error;
+ }
+
/*
* Check for command that don't needs to allocate a returned payload. We do
* this here so we don't have to make the call for no payload at each
break;
}
+ /*
+ * Check that the UID or GID match that of the tracing session.
+ * The root user can interact with all sessions.
+ */
+ if (need_tracing_session) {
+ if (!session_access_ok(cmd_ctx->session,
+ cmd_ctx->creds.uid, cmd_ctx->creds.gid)) {
+ ret = LTTCOMM_EPERM;
+ goto error;
+ }
+ }
+
/* Process by command type */
switch (cmd_ctx->lsm->cmd_type) {
case LTTNG_ADD_CONTEXT:
{
tracepoint(create_session_start);
ret = cmd_create_session(cmd_ctx->lsm->session.name,
- cmd_ctx->lsm->session.path);
+ cmd_ctx->lsm->session.path, &cmd_ctx->creds);
tracepoint(create_session_end);
break;
}
}
case LTTNG_LIST_SESSIONS:
{
- session_lock_list();
+ unsigned int nr_sessions;
- if (session_list_ptr->count == 0) {
+ session_lock_list();
+ nr_sessions = lttng_sessions_count(cmd_ctx->creds.uid, cmd_ctx->creds.gid);
+ if (nr_sessions == 0) {
ret = LTTCOMM_NO_SESSION;
session_unlock_list();
goto error;
}
-
- ret = setup_lttng_msg(cmd_ctx, sizeof(struct lttng_session) *
- session_list_ptr->count);
+ ret = setup_lttng_msg(cmd_ctx, sizeof(struct lttng_session) * nr_sessions);
if (ret < 0) {
session_unlock_list();
goto setup_error;
}
/* Filled the session array */
- list_lttng_sessions((struct lttng_session *)(cmd_ctx->llm->payload));
+ list_lttng_sessions((struct lttng_session *)(cmd_ctx->llm->payload),
+ cmd_ctx->creds.uid, cmd_ctx->creds.gid);
session_unlock_list();
goto error;
}
+ /* Set socket option for credentials retrieval */
+ ret = lttcomm_setsockopt_creds_unix_sock(sock);
+ if (ret < 0) {
+ goto error;
+ }
+
/* Allocate context command to process the client request */
cmd_ctx = zmalloc(sizeof(struct command_ctx));
if (cmd_ctx == NULL) {
* the client.
*/
DBG("Receiving data from client ...");
- ret = lttcomm_recv_unix_sock(sock, cmd_ctx->lsm,
- sizeof(struct lttcomm_session_msg));
+ ret = lttcomm_recv_creds_unix_sock(sock, cmd_ctx->lsm,
+ sizeof(struct lttcomm_session_msg), &cmd_ctx->creds);
if (ret <= 0) {
DBG("Nothing recv() from client... continuing");
close(sock);
ERR("Failed to send data back to client");
}
- clean_command_ctx(&cmd_ctx);
-
/* End of transmission */
close(sock);
+
+ clean_command_ctx(&cmd_ctx);
}
error:
fprintf(stderr, " -q, --quiet No output at all.\n");
fprintf(stderr, " -v, --verbose Verbose mode. Activate DBG() macro.\n");
fprintf(stderr, " --verbose-consumer Verbose mode for consumer. Activate DBG() macro.\n");
+ fprintf(stderr, " --no-kernel Disable kernel tracer\n");
}
/*
{ "quiet", 0, 0, 'q' },
{ "verbose", 0, 0, 'v' },
{ "verbose-consumer", 0, 0, 'Z' },
+ { "no-kernel", 0, 0, 'N' },
{ NULL, 0, 0, 0 }
};
while (1) {
int option_index = 0;
- c = getopt_long(argc, argv, "dhqvVS" "a:c:g:s:C:E:D:F:Z:u:t",
+ c = getopt_long(argc, argv, "dhqvVSN" "a:c:g:s:C:E:D:F:Z:u:t",
long_options, &option_index);
if (c == -1) {
break;
case 'G':
snprintf(ustconsumer32_data.cmd_unix_sock_path, PATH_MAX, "%s", optarg);
break;
+ case 'N':
+ opt_no_kernel = 1;
+ break;
case 'q':
opt_quiet = 1;
break;
* Race window between mkdir and chown is OK because we are going from more
* permissive (root.root) to les permissive (root.tracing).
*/
-static int set_permissions(void)
+static int set_permissions(char *rundir)
{
int ret;
gid_t gid;
gid = allowed_group();
if (gid < 0) {
- if (is_root) {
- WARN("No tracing group detected");
- ret = 0;
- } else {
- ERR("Missing tracing group. Aborting execution.");
- ret = -1;
- }
+ WARN("No tracing group detected");
+ ret = 0;
goto end;
}
/* Set lttng run dir */
- ret = chown(LTTNG_RUNDIR, 0, gid);
+ ret = chown(rundir, 0, gid);
if (ret < 0) {
- ERR("Unable to set group on " LTTNG_RUNDIR);
+ ERR("Unable to set group on %s", rundir);
perror("chown");
}
}
/* Setup kernel tracer */
- init_kernel_tracer();
+ if (!opt_no_kernel) {
+ init_kernel_tracer();
+ }
/* Set ulimit for open files */
set_ulimit();
}
/* Set credentials to socket */
- if (is_root && ((ret = set_permissions()) < 0)) {
+ if (is_root && ((ret = set_permissions(rundir)) < 0)) {
goto exit;
}
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
+#include <sys/stat.h>
+#include <sys/types.h>
#include <urcu.h>
#include <lttng-sessiond-comm.h>
#include <lttngerr.h>
+#include <runas.h>
-#include "hashtable.h"
#include "session.h"
-#include "../hashtable/hash.h"
-
/*
* NOTES:
*
/*
* Create a brand new session and add it to the session list.
*/
-int session_create(char *name, char *path)
+int session_create(char *name, char *path, uid_t uid, gid_t gid)
{
int ret;
struct ltt_session *new_session;
/* Init lock */
pthread_mutex_init(&new_session->lock, NULL);
+ new_session->uid = uid;
+ new_session->gid = gid;
+
+ ret = mkdir_recursive_run_as(new_session->path, S_IRWXU | S_IRWXG,
+ new_session->uid, new_session->gid);
+ if (ret < 0) {
+ if (ret != -EEXIST) {
+ ERR("Trace directory creation error");
+ ret = LTTCOMM_CREATE_FAIL;
+ goto error;
+ }
+ }
+
/* Add new session to the session list */
session_lock_list();
- new_session->uid = add_session_list(new_session);
+ new_session->id = add_session_list(new_session);
session_unlock_list();
- DBG("Tracing session %s created in %s with UID %d", name, path, new_session->uid);
+ DBG("Tracing session %s created in %s with ID %d by UID %d GID %d",
+ name, path, new_session->id,
+ new_session->uid, new_session->gid);
return LTTCOMM_OK;
#define _LTT_SESSION_H
#include <pthread.h>
+#include <unistd.h>
#include <urcu/list.h>
#include "trace-kernel.h"
pthread_mutex_t lock;
struct cds_list_head list;
int enabled; /* enabled/started flag */
- int uid;
+ int id; /* session unique identifier */
+ /* UID/GID of the user owning the session */
+ uid_t uid;
+ gid_t gid;
};
/* Prototypes */
-int session_create(char *name, char *path);
+int session_create(char *name, char *path, uid_t uid, gid_t gid);
int session_destroy(struct ltt_session *session);
void session_lock(struct ltt_session *session);
char *trace_path;
struct ltt_kernel_metadata *metadata;
struct ltt_kernel_channel_list channel_list;
+ /* UID/GID of the user owning the session */
+ uid_t uid;
+ gid_t gid;
};
/*
#include <unistd.h>
#include <lttngerr.h>
+#include <lttng-ht.h>
#include <lttng-share.h>
-#include "hashtable.h"
#include "trace-ust.h"
/*
* Find the channel in the hashtable.
*/
-struct ltt_ust_channel *trace_ust_find_channel_by_name(struct cds_lfht *ht,
+struct ltt_ust_channel *trace_ust_find_channel_by_name(struct lttng_ht *ht,
char *name)
{
- struct cds_lfht_node *node;
- struct cds_lfht_iter iter;
+ struct lttng_ht_node_str *node;
+ struct lttng_ht_iter iter;
rcu_read_lock();
- node = hashtable_lookup(ht, (void *) name, strlen(name), &iter);
+ lttng_ht_lookup(ht, (void *)name, &iter);
+ node = lttng_ht_iter_get_node_str(&iter);
if (node == NULL) {
rcu_read_unlock();
goto error;
/*
* Find the event in the hashtable.
*/
-struct ltt_ust_event *trace_ust_find_event_by_name(struct cds_lfht *ht,
+struct ltt_ust_event *trace_ust_find_event_by_name(struct lttng_ht *ht,
char *name)
{
- struct cds_lfht_node *node;
- struct cds_lfht_iter iter;
+ struct lttng_ht_node_str *node;
+ struct lttng_ht_iter iter;
rcu_read_lock();
- node = hashtable_lookup(ht, (void *)name, strlen(name), &iter);
+ lttng_ht_lookup(ht, (void *) name, &iter);
+ node = lttng_ht_iter_get_node_str(&iter);
if (node == NULL) {
rcu_read_unlock();
goto error;
*
* Return pointer to structure or NULL.
*/
-struct ltt_ust_session *trace_ust_create_session(char *path, unsigned int uid,
+struct ltt_ust_session *trace_ust_create_session(char *path, int session_id,
struct lttng_domain *domain)
{
int ret;
}
/* Init data structure */
- lus->uid = uid;
+ lus->id = session_id;
lus->start_trace = 0;
/* Alloc UST domain hash tables */
- lus->domain_pid = hashtable_new(0);
- lus->domain_exec = hashtable_new_str(0);
+ lus->domain_pid = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG);
+ lus->domain_exec = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG);
/* Alloc UST global domain channels' HT */
- lus->domain_global.channels = hashtable_new_str(0);
+ lus->domain_global.channels = lttng_ht_new(0, LTTNG_HT_TYPE_STRING);
/* Set session path */
ret = snprintf(lus->pathname, PATH_MAX, "%s/ust", path);
luc->name[LTTNG_UST_SYM_NAME_LEN - 1] = '\0';
/* Init node */
- hashtable_node_init(&luc->node, (void *) luc->name, strlen(luc->name));
+ lttng_ht_node_init_str(&luc->node, luc->name);
/* Alloc hash tables */
- luc->events = hashtable_new_str(0);
- luc->ctx = hashtable_new(0);
+ luc->events = lttng_ht_new(0, LTTNG_HT_TYPE_STRING);
+ luc->ctx = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG);
/* Set trace output path */
ret = snprintf(luc->pathname, PATH_MAX, "%s", path);
lue->attr.name[LTTNG_UST_SYM_NAME_LEN - 1] = '\0';
/* Init node */
- hashtable_node_init(&lue->node, (void *) lue->attr.name,
- strlen(lue->attr.name));
+ lttng_ht_node_init_str(&lue->node, lue->attr.name);
/* Alloc context hash tables */
- lue->ctx = hashtable_new(0);
+ lue->ctx = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG);
DBG2("Trace UST event %s created", lue->attr.name);
}
uctx->ctx.ctx = ctx->ctx;
- hashtable_node_init(&uctx->node, (void *)((unsigned long) uctx->ctx.ctx),
- sizeof(void *));
+ lttng_ht_node_init_ulong(&uctx->node, (unsigned long) uctx->ctx.ctx);
return uctx;
*/
static void destroy_context_rcu(struct rcu_head *head)
{
- struct cds_lfht_node *node =
- caa_container_of(head, struct cds_lfht_node, head);
+ struct lttng_ht_node_ulong *node =
+ caa_container_of(head, struct lttng_ht_node_ulong, head);
struct ltt_ust_context *ctx =
caa_container_of(node, struct ltt_ust_context, node);
/*
* Cleanup UST context hash table.
*/
-static void destroy_context(struct cds_lfht *ht)
+static void destroy_context(struct lttng_ht *ht)
{
int ret;
- struct cds_lfht_node *node;
- struct cds_lfht_iter iter;
+ struct lttng_ht_node_ulong *node;
+ struct lttng_ht_iter iter;
- cds_lfht_for_each(ht, &iter, node) {
- ret = hashtable_del(ht, &iter);
+ cds_lfht_for_each_entry(ht->ht, &iter.iter, node, node) {
+ ret = lttng_ht_del(ht, &iter);
if (!ret) {
call_rcu(&node->head, destroy_context_rcu);
}
}
- hashtable_destroy(ht);
+ lttng_ht_destroy(ht);
}
/*
*/
static void destroy_event_rcu(struct rcu_head *head)
{
- struct cds_lfht_node *node =
- caa_container_of(head, struct cds_lfht_node, head);
+ struct lttng_ht_node_str *node =
+ caa_container_of(head, struct lttng_ht_node_str, head);
struct ltt_ust_event *event =
caa_container_of(node, struct ltt_ust_event, node);
/*
* Cleanup UST events hashtable.
*/
-static void destroy_event(struct cds_lfht *events)
+static void destroy_event(struct lttng_ht *events)
{
int ret;
- struct cds_lfht_node *node;
- struct cds_lfht_iter iter;
+ struct lttng_ht_node_str *node;
+ struct lttng_ht_iter iter;
- cds_lfht_for_each(events, &iter, node) {
- ret = hashtable_del(events, &iter);
+ cds_lfht_for_each_entry(events->ht, &iter.iter, node, node) {
+ ret = lttng_ht_del(events, &iter);
if (!ret) {
call_rcu(&node->head, destroy_event_rcu);
}
}
- hashtable_destroy(events);
+ lttng_ht_destroy(events);
}
/*
void trace_ust_destroy_channel(struct ltt_ust_channel *channel)
{
int ret;
- struct cds_lfht_node *node;
- struct cds_lfht_iter iter;
+ struct lttng_ht_node_str *node;
+ struct lttng_ht_iter iter;
DBG2("Trace destroy UST channel %s", channel->name);
rcu_read_lock();
- cds_lfht_for_each(channel->events, &iter, node) {
- ret = hashtable_del(channel->events, &iter);
+ cds_lfht_for_each_entry(channel->events->ht, &iter.iter, node, node) {
+ ret = lttng_ht_del(channel->events, &iter);
if (!ret) {
destroy_event(channel->events);
}
*/
static void destroy_channel_rcu(struct rcu_head *head)
{
- struct cds_lfht_node *node =
- caa_container_of(head, struct cds_lfht_node, head);
+ struct lttng_ht_node_str *node =
+ caa_container_of(head, struct lttng_ht_node_str, head);
struct ltt_ust_channel *channel =
caa_container_of(node, struct ltt_ust_channel, node);
/*
* Iterate over a hash table containing channels and cleanup safely.
*/
-static void destroy_channels(struct cds_lfht *channels)
+static void destroy_channels(struct lttng_ht *channels)
{
int ret;
- struct cds_lfht_node *node;
- struct cds_lfht_iter iter;
+ struct lttng_ht_node_str *node;
+ struct lttng_ht_iter iter;
- cds_lfht_for_each(channels, &iter, node) {
- ret = hashtable_del(channels, &iter);
+ cds_lfht_for_each_entry(channels->ht, &iter.iter, node, node) {
+ ret = lttng_ht_del(channels, &iter);
if (!ret) {
call_rcu(&node->head, destroy_channel_rcu);
}
}
- hashtable_destroy(channels);
+ lttng_ht_destroy(channels);
}
/*
* Cleanup UST pid domain.
*/
-static void destroy_domain_pid(struct cds_lfht *ht)
+static void destroy_domain_pid(struct lttng_ht *ht)
{
int ret;
- struct cds_lfht_iter iter;
+ struct lttng_ht_iter iter;
struct ltt_ust_domain_pid *dpid;
- cds_lfht_for_each_entry(ht, &iter, dpid, node) {
- ret = hashtable_del(ht , &iter);
+ cds_lfht_for_each_entry(ht->ht, &iter.iter, dpid, node.node) {
+ ret = lttng_ht_del(ht , &iter);
if (!ret) {
destroy_channels(dpid->channels);
}
}
- hashtable_destroy(ht);
+ lttng_ht_destroy(ht);
}
/*
* Cleanup UST exec name domain.
*/
-static void destroy_domain_exec(struct cds_lfht *ht)
+static void destroy_domain_exec(struct lttng_ht *ht)
{
int ret;
- struct cds_lfht_iter iter;
+ struct lttng_ht_iter iter;
struct ltt_ust_domain_exec *dexec;
- cds_lfht_for_each_entry(ht, &iter, dexec, node) {
- ret = hashtable_del(ht , &iter);
+ cds_lfht_for_each_entry(ht->ht, &iter.iter, dexec, node.node) {
+ ret = lttng_ht_del(ht , &iter);
if (!ret) {
destroy_channels(dexec->channels);
}
}
- hashtable_destroy(ht);
+ lttng_ht_destroy(ht);
}
/*
rcu_read_lock();
- DBG2("Trace UST destroy session %d", session->uid);
+ DBG2("Trace UST destroy session %d", session->id);
/* Cleaning up UST domain */
destroy_domain_global(&session->domain_global);
#include <limits.h>
#include <urcu.h>
#include <urcu/list.h>
+
#include <lttng/lttng.h>
+#include <lttng-ht.h>
#include "ust-ctl.h"
-#include "../hashtable/rculfhash.h"
-
/* UST Stream list */
struct ltt_ust_stream_list {
unsigned int count;
/* Context hash table nodes */
struct ltt_ust_context {
struct lttng_ust_context ctx;
- struct cds_lfht_node node;
+ struct lttng_ht_node_ulong node;
};
/* UST event */
struct ltt_ust_event {
unsigned int enabled;
struct lttng_ust_event attr;
- struct cds_lfht *ctx;
- struct cds_lfht_node node;
+ struct lttng_ht *ctx;
+ struct lttng_ht_node_str node;
};
/* UST stream */
char name[LTTNG_UST_SYM_NAME_LEN];
char pathname[PATH_MAX];
struct lttng_ust_channel attr;
- struct cds_lfht *ctx;
- struct cds_lfht *events;
- struct cds_lfht_node node;
+ struct lttng_ht *ctx;
+ struct lttng_ht *events;
+ struct lttng_ht_node_str node;
};
/* UST Metadata */
/* UST domain global (LTTNG_DOMAIN_UST) */
struct ltt_ust_domain_global {
- struct cds_lfht *channels;
+ struct lttng_ht *channels;
};
/* UST domain pid (LTTNG_DOMAIN_UST_PID) */
struct ltt_ust_domain_pid {
pid_t pid;
- struct cds_lfht *channels;
- struct cds_lfht_node node;
+ struct lttng_ht *channels;
+ struct lttng_ht_node_ulong node;
};
/* UST domain exec name (LTTNG_DOMAIN_UST_EXEC_NAME) */
struct ltt_ust_domain_exec {
char exec_name[LTTNG_UST_SYM_NAME_LEN];
- struct cds_lfht *channels;
- struct cds_lfht_node node;
+ struct lttng_ht *channels;
+ struct lttng_ht_node_str node;
};
/* UST session */
struct ltt_ust_session {
- int uid; /* Unique identifier of session */
+ int id; /* Unique identifier of session */
int start_trace;
char pathname[PATH_MAX];
struct ltt_ust_domain_global domain_global;
* contains a HT of channels. See ltt_ust_domain_exec and
* ltt_ust_domain_pid data structures.
*/
- struct cds_lfht *domain_pid;
- struct cds_lfht *domain_exec;
+ struct lttng_ht *domain_pid;
+ struct lttng_ht *domain_exec;
+ /* UID/GID of the user owning the session */
+ uid_t uid;
+ gid_t gid;
};
#ifdef HAVE_LIBLTTNG_UST_CTL
/*
* Lookup functions. NULL is returned if not found.
*/
-struct ltt_ust_event *trace_ust_find_event_by_name(struct cds_lfht *ht,
+struct ltt_ust_event *trace_ust_find_event_by_name(struct lttng_ht *ht,
char *name);
-struct ltt_ust_channel *trace_ust_find_channel_by_name(struct cds_lfht *ht,
+struct ltt_ust_channel *trace_ust_find_channel_by_name(struct lttng_ht *ht,
char *name);
/*
* Create functions malloc() the data structure.
*/
-struct ltt_ust_session *trace_ust_create_session(char *path, unsigned int uid,
+struct ltt_ust_session *trace_ust_create_session(char *path, int session_id,
struct lttng_domain *domain);
struct ltt_ust_channel *trace_ust_create_channel(struct lttng_channel *attr,
char *path);
#else /* HAVE_LIBLTTNG_UST_CTL */
static inline
-struct ltt_ust_event *trace_ust_find_event_by_name(struct cds_lfht *ht,
+struct ltt_ust_event *trace_ust_find_event_by_name(struct lttng_ht *ht,
char *name)
{
return NULL;
}
static inline
-struct ltt_ust_channel *trace_ust_find_channel_by_name(struct cds_lfht *ht,
+struct ltt_ust_channel *trace_ust_find_channel_by_name(struct lttng_ht *ht,
char *name)
{
return NULL;
#include <sys/stat.h>
#include <sys/types.h>
#include <unistd.h>
+#include <runas.h>
#include <urcu/compiler.h>
+
#include <lttngerr.h>
+#include <lttng-ht.h>
#include <lttng-share.h>
+#include <runas.h>
-#include "hashtable.h"
#include "ust-app.h"
#include "ust-consumer.h"
#include "ust-ctl.h"
void delete_ust_app_event(int sock, struct ust_app_event *ua_event)
{
int ret;
- struct cds_lfht_iter iter;
+ struct lttng_ht_iter iter;
struct ust_app_ctx *ua_ctx;
- cds_lfht_for_each_entry(ua_event->ctx, &iter, ua_ctx, node) {
- ret = hashtable_del(ua_event->ctx, &iter);
+ cds_lfht_for_each_entry(ua_event->ctx->ht, &iter.iter, ua_ctx,
+ node.node) {
+ ret = lttng_ht_del(ua_event->ctx, &iter);
assert(!ret);
delete_ust_app_ctx(sock, ua_ctx);
}
- ret = hashtable_destroy(ua_event->ctx);
- assert(!ret);
+ lttng_ht_destroy(ua_event->ctx);
if (ua_event->obj != NULL) {
ustctl_release_object(sock, ua_event->obj);
void delete_ust_app_channel(int sock, struct ust_app_channel *ua_chan)
{
int ret;
- struct cds_lfht_iter iter;
+ struct lttng_ht_iter iter;
struct ust_app_event *ua_event;
struct ust_app_ctx *ua_ctx;
struct ltt_ust_stream *stream, *stmp;
}
/* Wipe context */
- cds_lfht_for_each_entry(ua_chan->ctx, &iter, ua_ctx, node) {
- ret = hashtable_del(ua_chan->ctx, &iter);
+ cds_lfht_for_each_entry(ua_chan->ctx->ht, &iter.iter, ua_ctx, node.node) {
+ ret = lttng_ht_del(ua_chan->ctx, &iter);
assert(!ret);
delete_ust_app_ctx(sock, ua_ctx);
}
- ret = hashtable_destroy(ua_chan->ctx);
- assert(!ret);
+ lttng_ht_destroy(ua_chan->ctx);
/* Wipe events */
- cds_lfht_for_each_entry(ua_chan->events, &iter, ua_event, node) {
- ret = hashtable_del(ua_chan->events, &iter);
+ cds_lfht_for_each_entry(ua_chan->events->ht, &iter.iter, ua_event,
+ node.node) {
+ ret = lttng_ht_del(ua_chan->events, &iter);
assert(!ret);
delete_ust_app_event(sock, ua_event);
}
- ret = hashtable_destroy(ua_chan->events);
- assert(!ret);
+ lttng_ht_destroy(ua_chan->events);
if (ua_chan->obj != NULL) {
ustctl_release_object(sock, ua_chan->obj);
void delete_ust_app_session(int sock, struct ust_app_session *ua_sess)
{
int ret;
- struct cds_lfht_iter iter;
+ struct lttng_ht_iter iter;
struct ust_app_channel *ua_chan;
if (ua_sess->metadata) {
}
}
- cds_lfht_for_each_entry(ua_sess->channels, &iter, ua_chan, node) {
- ret = hashtable_del(ua_sess->channels, &iter);
+ cds_lfht_for_each_entry(ua_sess->channels->ht, &iter.iter, ua_chan,
+ node.node) {
+ ret = lttng_ht_del(ua_sess->channels, &iter);
assert(!ret);
delete_ust_app_channel(sock, ua_chan);
}
- ret = hashtable_destroy(ua_sess->channels);
- assert(!ret);
+ lttng_ht_destroy(ua_sess->channels);
+ if (ua_sess->handle != -1) {
+ ustctl_release_handle(sock, ua_sess->handle);
+ }
free(ua_sess);
}
void delete_ust_app(struct ust_app *app)
{
int ret, sock;
- struct cds_lfht_node *node;
- struct cds_lfht_iter iter;
+ struct lttng_ht_iter iter;
struct ust_app_session *ua_sess;
rcu_read_lock();
- /* Remove from key hash table */
- node = hashtable_lookup(ust_app_sock_key_map,
- (void *) ((unsigned long) app->key.sock), sizeof(void *), &iter);
- if (node == NULL) {
- /* Not suppose to happen */
- ERR("UST app key %d not found in key hash table", app->key.sock);
- goto end;
- }
-
- ret = hashtable_del(ust_app_sock_key_map, &iter);
- if (ret) {
- ERR("UST app unable to delete app sock %d from key hash table",
- app->key.sock);
- } else {
- DBG2("UST app pair sock %d key %d deleted",
- app->key.sock, app->key.pid);
- }
-
- /* Socket is already closed at this point */
-
/* Delete ust app sessions info */
sock = app->key.sock;
app->key.sock = -1;
/* Wipe sessions */
- cds_lfht_for_each_entry(app->sessions, &iter, ua_sess, node) {
- ret = hashtable_del(app->sessions, &iter);
+ cds_lfht_for_each_entry(app->sessions->ht, &iter.iter, ua_sess,
+ node.node) {
+ ret = lttng_ht_del(app->sessions, &iter);
assert(!ret);
delete_ust_app_session(app->key.sock, ua_sess);
}
- ret = hashtable_destroy(app->sessions);
- assert(!ret);
+ lttng_ht_destroy(app->sessions);
/*
- * Wait until we have removed the key from the sock hash table
- * before closing this socket, otherwise an application could
- * re-use the socket ID and race with the teardown, using the
- * same hash table entry.
+ * Wait until we have removed the key from the sock hash table before
+ * closing this socket, otherwise an application could re-use the socket ID
+ * and race with the teardown, using the same hash table entry.
*/
close(sock);
DBG2("UST app pid %d deleted", app->key.pid);
free(app);
-end:
+
rcu_read_unlock();
}
static
void delete_ust_app_rcu(struct rcu_head *head)
{
- struct cds_lfht_node *node =
- caa_container_of(head, struct cds_lfht_node, head);
+ struct lttng_ht_node_ulong *node =
+ caa_container_of(head, struct lttng_ht_node_ulong, head);
struct ust_app *app =
caa_container_of(node, struct ust_app, node);
+ DBG3("Call RCU deleting app PID %d", app->key.pid);
delete_ust_app(app);
}
}
ua_sess->handle = -1;
- ua_sess->channels = hashtable_new_str(0);
+ ua_sess->channels = lttng_ht_new(0, LTTNG_HT_TYPE_STRING);
return ua_sess;
ua_chan->enabled = 1;
ua_chan->handle = -1;
- ua_chan->ctx = hashtable_new(0);
- ua_chan->events = hashtable_new_str(0);
- hashtable_node_init(&ua_chan->node, (void *) ua_chan->name,
- strlen(ua_chan->name));
+ ua_chan->ctx = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG);
+ ua_chan->events = lttng_ht_new(0, LTTNG_HT_TYPE_STRING);
+ lttng_ht_node_init_str(&ua_chan->node, ua_chan->name);
CDS_INIT_LIST_HEAD(&ua_chan->streams.head);
ua_event->enabled = 1;
strncpy(ua_event->name, name, sizeof(ua_event->name));
ua_event->name[sizeof(ua_event->name) - 1] = '\0';
- ua_event->ctx = hashtable_new(0);
- hashtable_node_init(&ua_event->node, (void *) ua_event->name,
- strlen(ua_event->name));
+ ua_event->ctx = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG);
+ lttng_ht_node_init_str(&ua_event->node, ua_event->name);
/* Copy attributes */
if (attr) {
static
struct ust_app *find_app_by_sock(int sock)
{
- struct cds_lfht_node *node;
+ struct lttng_ht_node_ulong *node;
struct ust_app_key *key;
- struct cds_lfht_iter iter;
+ struct lttng_ht_iter iter;
- node = hashtable_lookup(ust_app_sock_key_map,
- (void *)((unsigned long) sock), sizeof(void *), &iter);
+ lttng_ht_lookup(ust_app_sock_key_map, (void *)((unsigned long) sock),
+ &iter);
+ node = lttng_ht_iter_get_node_ulong(&iter);
if (node == NULL) {
DBG2("UST app find by sock %d key not found", sock);
goto error;
}
-
key = caa_container_of(node, struct ust_app_key, node);
- node = hashtable_lookup(ust_app_ht,
- (void *)((unsigned long) key->pid), sizeof(void *), &iter);
+ lttng_ht_lookup(ust_app_ht, (void *)((unsigned long) key->pid), &iter);
+ node = lttng_ht_iter_get_node_ulong(&iter);
if (node == NULL) {
DBG2("UST app find by sock %d not found", sock);
goto error;
static void shadow_copy_event(struct ust_app_event *ua_event,
struct ltt_ust_event *uevent)
{
- struct cds_lfht_iter iter;
+ struct lttng_ht_iter iter;
struct ltt_ust_context *uctx;
struct ust_app_ctx *ua_ctx;
/* Copy event attributes */
memcpy(&ua_event->attr, &uevent->attr, sizeof(ua_event->attr));
- cds_lfht_for_each_entry(uevent->ctx, &iter, uctx, node) {
+ cds_lfht_for_each_entry(uevent->ctx->ht, &iter.iter, uctx, node.node) {
ua_ctx = alloc_ust_app_ctx(&uctx->ctx);
if (ua_ctx == NULL) {
continue;
}
- hashtable_node_init(&ua_ctx->node,
- (void *)((unsigned long) ua_ctx->ctx.ctx), sizeof(void *));
- hashtable_add_unique(ua_event->ctx, &ua_ctx->node);
+ lttng_ht_node_init_ulong(&ua_ctx->node,
+ (unsigned long) ua_ctx->ctx.ctx);
+ lttng_ht_add_unique_ulong(ua_event->ctx, &ua_ctx->node);
}
}
static void shadow_copy_channel(struct ust_app_channel *ua_chan,
struct ltt_ust_channel *uchan)
{
- struct cds_lfht_iter iter;
- struct cds_lfht_node *ua_event_node;
+ struct lttng_ht_iter iter;
+ struct lttng_ht_node_str *ua_event_node;
struct ltt_ust_event *uevent;
struct ltt_ust_context *uctx;
struct ust_app_event *ua_event;
/* Copy event attributes */
memcpy(&ua_chan->attr, &uchan->attr, sizeof(ua_chan->attr));
- cds_lfht_for_each_entry(uchan->ctx, &iter, uctx, node) {
+ cds_lfht_for_each_entry(uchan->ctx->ht, &iter.iter, uctx, node.node) {
ua_ctx = alloc_ust_app_ctx(&uctx->ctx);
if (ua_ctx == NULL) {
continue;
}
- hashtable_node_init(&ua_ctx->node,
- (void *)((unsigned long) ua_ctx->ctx.ctx), sizeof(void *));
- hashtable_add_unique(ua_chan->ctx, &ua_ctx->node);
+ lttng_ht_node_init_ulong(&ua_ctx->node,
+ (unsigned long) ua_ctx->ctx.ctx);
+ lttng_ht_add_unique_ulong(ua_chan->ctx, &ua_ctx->node);
}
/* Copy all events from ltt ust channel to ust app channel */
- cds_lfht_for_each_entry(uchan->events, &iter, uevent, node) {
- struct cds_lfht_iter uiter;
+ cds_lfht_for_each_entry(uchan->events->ht, &iter.iter, uevent, node.node) {
+ struct lttng_ht_iter uiter;
- ua_event_node = hashtable_lookup(ua_chan->events,
- (void *) uevent->attr.name, strlen(uevent->attr.name),
- &uiter);
+ lttng_ht_lookup(ua_chan->events, (void *) uevent->attr.name, &uiter);
+ ua_event_node = lttng_ht_iter_get_node_str(&uiter);
if (ua_event_node == NULL) {
DBG2("UST event %s not found on shadow copy channel",
uevent->attr.name);
continue;
}
shadow_copy_event(ua_event, uevent);
- hashtable_add_unique(ua_chan->events, &ua_event->node);
+ lttng_ht_add_unique_str(ua_chan->events, &ua_event->node);
}
}
* Copy data between a UST app session and a regular LTT session.
*/
static void shadow_copy_session(struct ust_app_session *ua_sess,
- struct ltt_ust_session *usess,
- struct ust_app *app)
+ struct ltt_ust_session *usess, struct ust_app *app)
{
- struct cds_lfht_node *ua_chan_node;
- struct cds_lfht_iter iter;
+ struct lttng_ht_node_str *ua_chan_node;
+ struct lttng_ht_iter iter;
struct ltt_ust_channel *uchan;
struct ust_app_channel *ua_chan;
time_t rawtime;
DBG2("Shadow copy of session handle %d", ua_sess->handle);
+ ua_sess->id = usess->id;
ua_sess->uid = usess->uid;
+ ua_sess->gid = usess->gid;
ret = snprintf(ua_sess->path, PATH_MAX,
"%s/%s-%d-%s",
/* TODO: support all UST domain */
/* Iterate over all channels in global domain. */
- cds_lfht_for_each_entry(usess->domain_global.channels, &iter,
- uchan, node) {
- struct cds_lfht_iter uiter;
+ cds_lfht_for_each_entry(usess->domain_global.channels->ht, &iter.iter,
+ uchan, node.node) {
+ struct lttng_ht_iter uiter;
- ua_chan_node = hashtable_lookup(ua_sess->channels,
- (void *)uchan->name, strlen(uchan->name),
- &uiter);
+ lttng_ht_lookup(ua_sess->channels, (void *)uchan->name, &uiter);
+ ua_chan_node = lttng_ht_iter_get_node_str(&uiter);
if (ua_chan_node != NULL) {
continue;
}
}
shadow_copy_channel(ua_chan, uchan);
- hashtable_add_unique(ua_sess->channels, &ua_chan->node);
+ lttng_ht_add_unique_str(ua_sess->channels, &ua_chan->node);
}
}
*/
static
void __lookup_session_by_app(struct ltt_ust_session *usess,
- struct ust_app *app, struct cds_lfht_iter *iter)
+ struct ust_app *app, struct lttng_ht_iter *iter)
{
/* Get right UST app session from app */
- (void) hashtable_lookup(app->sessions,
- (void *) ((unsigned long) usess->uid), sizeof(void *),
- iter);
+ lttng_ht_lookup(app->sessions, (void *)((unsigned long) usess->uid), iter);
}
/*
* Return ust app session from the app session hashtable using the UST session
- * uid.
+ * id.
*/
static struct ust_app_session *lookup_session_by_app(
struct ltt_ust_session *usess, struct ust_app *app)
{
- struct cds_lfht_iter iter;
- struct cds_lfht_node *node;
+ struct lttng_ht_iter iter;
+ struct lttng_ht_node_ulong *node;
__lookup_session_by_app(usess, app, &iter);
- node = hashtable_iter_get_node(&iter);
+ node = lttng_ht_iter_get_node_ulong(&iter);
if (node == NULL) {
goto error;
}
ua_sess = lookup_session_by_app(usess, app);
if (ua_sess == NULL) {
- DBG2("UST app pid: %d session uid %d not found, creating it",
- app->key.pid, usess->uid);
+ DBG2("UST app pid: %d session id %d not found, creating it",
+ app->key.pid, usess->id);
ua_sess = alloc_ust_app_session();
if (ua_sess == NULL) {
/* Only malloc can failed so something is really wrong */
ua_sess->handle = ret;
/* Add ust app session to app's HT */
- hashtable_node_init(&ua_sess->node,
- (void *)((unsigned long) ua_sess->uid), sizeof(void *));
- hashtable_add_unique(app->sessions, &ua_sess->node);
+ lttng_ht_node_init_ulong(&ua_sess->node, (unsigned long) ua_sess->uid);
+ lttng_ht_add_unique_ulong(app->sessions, &ua_sess->node);
DBG2("UST app session created successfully with handle %d", ret);
}
struct ust_app *app)
{
int ret = 0;
- struct cds_lfht_iter iter;
- struct cds_lfht_node *node;
+ struct lttng_ht_iter iter;
+ struct lttng_ht_node_ulong *node;
struct ust_app_ctx *ua_ctx;
DBG2("UST app adding context to channel %s", ua_chan->name);
- node = hashtable_lookup(ua_chan->ctx, (void *)((unsigned long)uctx->ctx),
- sizeof(void *), &iter);
+ lttng_ht_lookup(ua_chan->ctx, (void *)((unsigned long)uctx->ctx), &iter);
+ node = lttng_ht_iter_get_node_ulong(&iter);
if (node != NULL) {
ret = -EEXIST;
goto error;
goto error;
}
- hashtable_node_init(&ua_ctx->node,
- (void *)((unsigned long) ua_ctx->ctx.ctx), sizeof(void *));
- hashtable_add_unique(ua_chan->ctx, &ua_ctx->node);
+ lttng_ht_node_init_ulong(&ua_ctx->node, (unsigned long) ua_ctx->ctx.ctx);
+ lttng_ht_add_unique_ulong(ua_chan->ctx, &ua_ctx->node);
ret = create_ust_channel_context(ua_chan, ua_ctx, app);
if (ret < 0) {
struct ust_app *app)
{
int ret = 0;
- struct cds_lfht_iter iter;
- struct cds_lfht_node *node;
+ struct lttng_ht_iter iter;
+ struct lttng_ht_node_ulong *node;
struct ust_app_ctx *ua_ctx;
DBG2("UST app adding context to event %s", ua_event->name);
- node = hashtable_lookup(ua_event->ctx, (void *)((unsigned long)uctx->ctx),
- sizeof(void *), &iter);
+ lttng_ht_lookup(ua_event->ctx, (void *)((unsigned long)uctx->ctx), &iter);
+ node = lttng_ht_iter_get_node_ulong(&iter);
if (node != NULL) {
ret = -EEXIST;
goto error;
goto error;
}
- hashtable_node_init(&ua_ctx->node,
- (void *)((unsigned long) ua_ctx->ctx.ctx), sizeof(void *));
- hashtable_add_unique(ua_event->ctx, &ua_ctx->node);
+ lttng_ht_node_init_ulong(&ua_ctx->node, (unsigned long) ua_ctx->ctx.ctx);
+ lttng_ht_add_unique_ulong(ua_event->ctx, &ua_ctx->node);
ret = create_ust_event_context(ua_event, ua_ctx, app);
if (ret < 0) {
struct ltt_ust_channel *uchan, struct ust_app *app)
{
int ret = 0;
- struct cds_lfht_iter iter;
- struct cds_lfht_node *ua_chan_node;
+ struct lttng_ht_iter iter;
+ struct lttng_ht_node_str *ua_chan_node;
struct ust_app_channel *ua_chan;
- ua_chan_node = hashtable_lookup(ua_sess->channels,
- (void *)uchan->name, strlen(uchan->name), &iter);
+ lttng_ht_lookup(ua_sess->channels, (void *)uchan->name, &iter);
+ ua_chan_node = lttng_ht_iter_get_node_str(&iter);
if (ua_chan_node == NULL) {
- DBG2("Unable to find channel %s in ust session uid %u",
- uchan->name, ua_sess->uid);
+ DBG2("Unable to find channel %s in ust session id %u",
+ uchan->name, ua_sess->id);
goto error;
}
struct ust_app *app)
{
int ret = 0;
- struct cds_lfht_iter iter;
- struct cds_lfht_node *ua_chan_node;
+ struct lttng_ht_iter iter;
+ struct lttng_ht_node_str *ua_chan_node;
struct ust_app_channel *ua_chan;
/* Lookup channel in the ust app session */
- ua_chan_node = hashtable_lookup(ua_sess->channels,
- (void *)uchan->name, strlen(uchan->name), &iter);
+ lttng_ht_lookup(ua_sess->channels, (void *)uchan->name, &iter);
+ ua_chan_node = lttng_ht_iter_get_node_str(&iter);
if (ua_chan_node == NULL) {
- DBG2("Unable to find channel %s in ust session uid %u",
- uchan->name, ua_sess->uid);
+ DBG2("Unable to find channel %s in ust session id %u",
+ uchan->name, ua_sess->id);
ua_chan = alloc_ust_app_channel(uchan->name, &uchan->attr);
if (ua_chan == NULL) {
goto error;
}
shadow_copy_channel(ua_chan, uchan);
- hashtable_add_unique(ua_sess->channels, &ua_chan->node);
+ lttng_ht_add_unique_str(ua_sess->channels, &ua_chan->node);
} else {
ua_chan = caa_container_of(ua_chan_node, struct ust_app_channel, node);
}
struct ust_app *app)
{
int ret = 0;
- struct cds_lfht_iter iter;
- struct cds_lfht_node *ua_event_node;
+ struct lttng_ht_iter iter;
+ struct lttng_ht_node_str *ua_event_node;
struct ust_app_event *ua_event;
/* Get event node */
- ua_event_node = hashtable_lookup(ua_chan->events,
- (void *)uevent->attr.name, strlen(uevent->attr.name), &iter);
+ lttng_ht_lookup(ua_chan->events, (void *)uevent->attr.name, &iter);
+ ua_event_node = lttng_ht_iter_get_node_str(&iter);
if (ua_event_node != NULL) {
ERR("UST app event %s already exist. Stopping creation.",
uevent->attr.name);
ua_event->enabled = 1;
- hashtable_add_unique(ua_chan->events, &ua_event->node);
+ lttng_ht_add_unique_str(ua_chan->events, &ua_event->node);
DBG2("UST app create event %s for PID %d completed",
ua_event->name, app->key.pid);
char *pathname, struct ust_app *app)
{
int ret = 0;
- mode_t old_umask;
if (ua_sess->metadata == NULL) {
/* Allocate UST metadata */
goto error;
}
- old_umask = umask(0);
- ret = mkdir(ua_sess->path, S_IRWXU | S_IRWXG);
+ ret = mkdir_run_as(ua_sess->path, S_IRWXU | S_IRWXG,
+ ua_sess->uid, ua_sess->gid);
if (ret < 0) {
PERROR("mkdir UST metadata");
goto error;
}
- umask(old_umask);
ret = snprintf(ua_sess->metadata->pathname, PATH_MAX,
"%s/metadata", ua_sess->path);
/*
* Return pointer to traceable apps list.
*/
-struct cds_lfht *ust_app_get_ht(void)
+struct lttng_ht *ust_app_get_ht(void)
{
return ust_app_ht;
}
*/
struct ust_app *ust_app_find_by_pid(pid_t pid)
{
- struct cds_lfht_node *node;
- struct cds_lfht_iter iter;
+ struct lttng_ht_node_ulong *node;
+ struct lttng_ht_iter iter;
rcu_read_lock();
- node = hashtable_lookup(ust_app_ht,
- (void *)((unsigned long) pid), sizeof(void *), &iter);
+ lttng_ht_lookup(ust_app_ht, (void *)((unsigned long) pid), &iter);
+ node = lttng_ht_iter_get_node_ulong(&iter);
if (node == NULL) {
DBG2("UST app no found with pid %d", pid);
goto error;
lta->v_minor = msg->minor;
strncpy(lta->name, msg->name, sizeof(lta->name));
lta->name[16] = '\0';
- lta->sessions = hashtable_new(0);
+ lta->sessions = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG);
/* Set key map */
lta->key.pid = msg->pid;
- hashtable_node_init(<a->node, (void *)((unsigned long)lta->key.pid),
- sizeof(void *));
+ lttng_ht_node_init_ulong(<a->node, (unsigned long)lta->key.pid);
lta->key.sock = sock;
- hashtable_node_init(<a->key.node, (void *)((unsigned long)lta->key.sock),
- sizeof(void *));
+ lttng_ht_node_init_ulong(<a->key.node, (unsigned long)lta->key.sock);
rcu_read_lock();
- hashtable_add_unique(ust_app_sock_key_map, <a->key.node);
- hashtable_add_unique(ust_app_ht, <a->node);
+ lttng_ht_add_unique_ulong(ust_app_sock_key_map, <a->key.node);
+ lttng_ht_add_unique_ulong(ust_app_ht, <a->node);
rcu_read_unlock();
DBG("App registered with pid:%d ppid:%d uid:%d gid:%d sock:%d name:%s"
void ust_app_unregister(int sock)
{
struct ust_app *lta;
- struct cds_lfht_node *node;
- struct cds_lfht_iter iter;
+ struct lttng_ht_node_ulong *node;
+ struct lttng_ht_iter iter;
int ret;
rcu_read_lock();
DBG("PID %d unregistering with sock %d", lta->key.pid, sock);
/* Get the node reference for a call_rcu */
- node = hashtable_lookup(ust_app_ht,
- (void *)((unsigned long) lta->key.pid), sizeof(void *), &iter);
+ lttng_ht_lookup(ust_app_ht, (void *)((unsigned long) lta->key.pid), &iter);
+ node = lttng_ht_iter_get_node_ulong(&iter);
if (node == NULL) {
ERR("Unable to find app sock %d by pid %d", sock, lta->key.pid);
goto error;
}
- ret = hashtable_del(ust_app_ht, &iter);
+ ret = lttng_ht_del(ust_app_ht, &iter);
assert(!ret);
call_rcu(&node->head, delete_ust_app_rcu);
error:
unsigned long count;
rcu_read_lock();
- count = hashtable_get_count(ust_app_ht);
+ count = lttng_ht_get_count(ust_app_ht);
rcu_read_unlock();
return count;
{
int ret, handle;
size_t nbmem, count = 0;
- struct cds_lfht_iter iter;
+ struct lttng_ht_iter iter;
struct ust_app *app;
struct lttng_event *tmp;
rcu_read_lock();
- cds_lfht_for_each_entry(ust_app_ht, &iter, app, node) {
- struct lttng_ust_tracepoint_iter iter;
+ cds_lfht_for_each_entry(ust_app_ht->ht, &iter.iter, app, node.node) {
+ struct lttng_ust_tracepoint_iter uiter;
handle = ustctl_tracepoint_list(app->key.sock);
if (handle < 0) {
}
while ((ret = ustctl_tracepoint_list_get(app->key.sock, handle,
- &iter)) != -ENOENT) {
+ &uiter)) != -ENOENT) {
if (count >= nbmem) {
DBG2("Reallocating event list from %zu to %zu entries", nbmem,
2 * nbmem);
goto rcu_error;
}
}
- memcpy(tmp[count].name, iter.name, LTTNG_UST_SYM_NAME_LEN);
- memcpy(tmp[count].loglevel, iter.loglevel, LTTNG_UST_SYM_NAME_LEN);
- tmp[count].loglevel_value = iter.loglevel_value;
+ memcpy(tmp[count].name, uiter.name, LTTNG_UST_SYM_NAME_LEN);
+ memcpy(tmp[count].loglevel, uiter.loglevel, LTTNG_UST_SYM_NAME_LEN);
+ tmp[count].loglevel_value = uiter.loglevel_value;
tmp[count].type = LTTNG_UST_TRACEPOINT;
tmp[count].pid = app->key.pid;
tmp[count].enabled = -1;
void ust_app_clean_list(void)
{
int ret;
- struct cds_lfht_iter iter;
- struct ust_app *app;
+ struct lttng_ht_iter iter;
+ struct lttng_ht_node_ulong *node;
DBG2("UST app cleaning registered apps hash table");
rcu_read_lock();
- cds_lfht_for_each_entry(ust_app_ht, &iter, app, node) {
- ret = hashtable_del(ust_app_ht, &iter);
+ cds_lfht_for_each_entry(ust_app_ht->ht, &iter.iter, node, node) {
+ ret = lttng_ht_del(ust_app_ht, &iter);
assert(!ret);
- call_rcu(&iter.node->head, delete_ust_app_rcu);
+ call_rcu(&node->head, delete_ust_app_rcu);
}
+ /* Destroy is done only when the ht is empty */
+ lttng_ht_destroy(ust_app_ht);
- hashtable_destroy(ust_app_ht);
- hashtable_destroy(ust_app_sock_key_map);
+ cds_lfht_for_each_entry(ust_app_sock_key_map->ht, &iter.iter, node, node) {
+ ret = lttng_ht_del(ust_app_sock_key_map, &iter);
+ assert(!ret);
+ }
+ /* Destroy is done only when the ht is empty */
+ lttng_ht_destroy(ust_app_sock_key_map);
rcu_read_unlock();
}
*/
void ust_app_ht_alloc(void)
{
- ust_app_ht = hashtable_new(0);
- ust_app_sock_key_map = hashtable_new(0);
+ ust_app_ht = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG);
+ ust_app_sock_key_map = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG);
}
/*
struct ltt_ust_channel *uchan)
{
int ret = 0;
- struct cds_lfht_iter iter;
- struct cds_lfht_node *ua_chan_node;
+ struct lttng_ht_iter iter;
+ struct lttng_ht_node_str *ua_chan_node;
struct ust_app *app;
struct ust_app_session *ua_sess;
struct ust_app_channel *ua_chan;
goto error;
}
- DBG2("UST app disabling channel %s from global domain for session uid %d",
- uchan->name, usess->uid);
+ DBG2("UST app disabling channel %s from global domain for session id %d",
+ uchan->name, usess->id);
rcu_read_lock();
/* For every registered applications */
- cds_lfht_for_each_entry(ust_app_ht, &iter, app, node) {
+ cds_lfht_for_each_entry(ust_app_ht->ht, &iter.iter, app, node.node) {
+ struct lttng_ht_iter uiter;
ua_sess = lookup_session_by_app(usess, app);
if (ua_sess == NULL) {
continue;
}
/* Get channel */
- ua_chan_node = hashtable_lookup(ua_sess->channels,
- (void *)uchan->name, strlen(uchan->name), &iter);
+ lttng_ht_lookup(ua_sess->channels, (void *)uchan->name, &uiter);
+ ua_chan_node = lttng_ht_iter_get_node_str(&uiter);
/* If the session if found for the app, the channel must be there */
assert(ua_chan_node);
struct ltt_ust_channel *uchan)
{
int ret = 0;
- struct cds_lfht_iter iter;
+ struct lttng_ht_iter iter;
struct ust_app *app;
struct ust_app_session *ua_sess;
goto error;
}
- DBG2("UST app enabling channel %s to global domain for session uid %d",
- uchan->name, usess->uid);
+ DBG2("UST app enabling channel %s to global domain for session id %d",
+ uchan->name, usess->id);
rcu_read_lock();
/* For every registered applications */
- cds_lfht_for_each_entry(ust_app_ht, &iter, app, node) {
+ cds_lfht_for_each_entry(ust_app_ht->ht, &iter.iter, app, node.node) {
ua_sess = lookup_session_by_app(usess, app);
if (ua_sess == NULL) {
continue;
struct ltt_ust_channel *uchan, struct ltt_ust_event *uevent)
{
int ret = 0;
- struct cds_lfht_iter iter, uiter;
- struct cds_lfht_node *ua_chan_node, *ua_event_node;
+ struct lttng_ht_iter iter, uiter;
+ struct lttng_ht_node_str *ua_chan_node, *ua_event_node;
struct ust_app *app;
struct ust_app_session *ua_sess;
struct ust_app_channel *ua_chan;
struct ust_app_event *ua_event;
DBG("UST app disabling event %s for all apps in channel "
- "%s for session uid %d", uevent->attr.name, uchan->name, usess->uid);
+ "%s for session id %d", uevent->attr.name, uchan->name, usess->id);
rcu_read_lock();
/* For all registered applications */
- cds_lfht_for_each_entry(ust_app_ht, &iter, app, node) {
+ cds_lfht_for_each_entry(ust_app_ht->ht, &iter.iter, app, node.node) {
ua_sess = lookup_session_by_app(usess, app);
if (ua_sess == NULL) {
/* Next app */
}
/* Lookup channel in the ust app session */
- ua_chan_node = hashtable_lookup(ua_sess->channels,
- (void *)uchan->name, strlen(uchan->name), &uiter);
+ lttng_ht_lookup(ua_sess->channels, (void *)uchan->name, &uiter);
+ ua_chan_node = lttng_ht_iter_get_node_str(&uiter);
if (ua_chan_node == NULL) {
- DBG2("Channel %s not found in session uid %d for app pid %d."
- "Skipping", uchan->name, usess->uid, app->key.pid);
+ DBG2("Channel %s not found in session id %d for app pid %d."
+ "Skipping", uchan->name, usess->id, app->key.pid);
continue;
}
ua_chan = caa_container_of(ua_chan_node, struct ust_app_channel, node);
- ua_event_node = hashtable_lookup(ua_chan->events,
- (void *)uevent->attr.name, strlen(uevent->attr.name), &uiter);
+ lttng_ht_lookup(ua_chan->events, (void *)uevent->attr.name, &uiter);
+ ua_event_node = lttng_ht_iter_get_node_str(&uiter);
if (ua_event_node == NULL) {
DBG2("Event %s not found in channel %s for app pid %d."
"Skipping", uevent->attr.name, uchan->name, app->key.pid);
struct ltt_ust_channel *uchan)
{
int ret = 0;
- struct cds_lfht_iter iter, uiter;
- struct cds_lfht_node *ua_chan_node;
+ struct lttng_ht_iter iter, uiter;
+ struct lttng_ht_node_str *ua_chan_node;
struct ust_app *app;
struct ust_app_session *ua_sess;
struct ust_app_channel *ua_chan;
struct ust_app_event *ua_event;
DBG("UST app disabling all event for all apps in channel "
- "%s for session uid %d", uchan->name, usess->uid);
+ "%s for session id %d", uchan->name, usess->id);
rcu_read_lock();
/* For all registered applications */
- cds_lfht_for_each_entry(ust_app_ht, &iter, app, node) {
+ cds_lfht_for_each_entry(ust_app_ht->ht, &iter.iter, app, node.node) {
ua_sess = lookup_session_by_app(usess, app);
/* If ua_sess is NULL, there is a code flow error */
assert(ua_sess);
/* Lookup channel in the ust app session */
- ua_chan_node = hashtable_lookup(ua_sess->channels, (void *)uchan->name,
- strlen(uchan->name), &uiter);
+ lttng_ht_lookup(ua_sess->channels, (void *)uchan->name, &uiter);
+ ua_chan_node = lttng_ht_iter_get_node_str(&uiter);
/* If the channel is not found, there is a code flow error */
assert(ua_chan_node);
ua_chan = caa_container_of(ua_chan_node, struct ust_app_channel, node);
/* Disable each events of channel */
- cds_lfht_for_each_entry(ua_chan->events, &uiter, ua_event, node) {
+ cds_lfht_for_each_entry(ua_chan->events->ht, &uiter.iter, ua_event,
+ node.node) {
ret = disable_ust_app_event(ua_sess, ua_event, app);
if (ret < 0) {
/* XXX: Report error someday... */
struct ltt_ust_channel *uchan)
{
int ret = 0;
- struct cds_lfht_iter iter;
+ struct lttng_ht_iter iter;
struct ust_app *app;
struct ust_app_session *ua_sess;
struct ust_app_channel *ua_chan;
goto error;
}
- DBG2("UST app adding channel %s to global domain for session uid %d",
- uchan->name, usess->uid);
+ DBG2("UST app adding channel %s to global domain for session id %d",
+ uchan->name, usess->id);
rcu_read_lock();
/* For every registered applications */
- cds_lfht_for_each_entry(ust_app_ht, &iter, app, node) {
+ cds_lfht_for_each_entry(ust_app_ht->ht, &iter.iter, app, node.node) {
/*
* Create session on the tracer side and add it to app session HT. Note
* that if session exist, it will simply return a pointer to the ust
struct ltt_ust_channel *uchan, struct ltt_ust_event *uevent)
{
int ret = 0;
- struct cds_lfht_iter iter, uiter;
- struct cds_lfht_node *ua_chan_node, *ua_event_node;
+ struct lttng_ht_iter iter, uiter;
+ struct lttng_ht_node_str *ua_chan_node, *ua_event_node;
struct ust_app *app;
struct ust_app_session *ua_sess;
struct ust_app_channel *ua_chan;
struct ust_app_event *ua_event;
- DBG("UST app enabling event %s for all apps for session uid %d",
- uevent->attr.name, usess->uid);
+ DBG("UST app enabling event %s for all apps for session id %d",
+ uevent->attr.name, usess->id);
/*
* NOTE: At this point, this function is called only if the session and
rcu_read_lock();
/* For all registered applications */
- cds_lfht_for_each_entry(ust_app_ht, &iter, app, node) {
+ cds_lfht_for_each_entry(ust_app_ht->ht, &iter.iter, app, node.node) {
ua_sess = lookup_session_by_app(usess, app);
/* If ua_sess is NULL, there is a code flow error */
assert(ua_sess);
/* Lookup channel in the ust app session */
- ua_chan_node = hashtable_lookup(ua_sess->channels, (void *)uchan->name,
- strlen(uchan->name), &uiter);
+ lttng_ht_lookup(ua_sess->channels, (void *)uchan->name, &uiter);
+ ua_chan_node = lttng_ht_iter_get_node_str(&uiter);
/* If the channel is not found, there is a code flow error */
assert(ua_chan_node);
ua_chan = caa_container_of(ua_chan_node, struct ust_app_channel, node);
- ua_event_node = hashtable_lookup(ua_chan->events,
- (void*)uevent->attr.name, strlen(uevent->attr.name), &uiter);
+ lttng_ht_lookup(ua_chan->events, (void*)uevent->attr.name, &uiter);
+ ua_event_node = lttng_ht_iter_get_node_str(&uiter);
if (ua_event_node == NULL) {
DBG3("UST app enable event %s not found for app PID %d."
"Skipping app", uevent->attr.name, app->key.pid);
struct ltt_ust_channel *uchan, struct ltt_ust_event *uevent)
{
int ret = 0;
- struct cds_lfht_iter iter, uiter;
- struct cds_lfht_node *ua_chan_node;
+ struct lttng_ht_iter iter, uiter;
+ struct lttng_ht_node_str *ua_chan_node;
struct ust_app *app;
struct ust_app_session *ua_sess;
struct ust_app_channel *ua_chan;
- DBG("UST app creating event %s for all apps for session uid %d",
- uevent->attr.name, usess->uid);
+ DBG("UST app creating event %s for all apps for session id %d",
+ uevent->attr.name, usess->id);
/*
* NOTE: At this point, this function is called only if the session and
rcu_read_lock();
/* For all registered applications */
- cds_lfht_for_each_entry(ust_app_ht, &iter, app, node) {
+ cds_lfht_for_each_entry(ust_app_ht->ht, &iter.iter, app, node.node) {
ua_sess = lookup_session_by_app(usess, app);
/* If ua_sess is NULL, there is a code flow error */
assert(ua_sess);
/* Lookup channel in the ust app session */
- ua_chan_node = hashtable_lookup(ua_sess->channels, (void *)uchan->name,
- strlen(uchan->name), &uiter);
+ lttng_ht_lookup(ua_sess->channels, (void *)uchan->name, &uiter);
+ ua_chan_node = lttng_ht_iter_get_node_str(&uiter);
/* If the channel is not found, there is a code flow error */
assert(ua_chan_node);
int ust_app_start_trace(struct ltt_ust_session *usess, struct ust_app *app)
{
int ret = 0;
- struct cds_lfht_iter iter;
+ struct lttng_ht_iter iter;
struct ust_app_session *ua_sess;
struct ust_app_channel *ua_chan;
struct ltt_ust_stream *ustream;
}
/* For each channel */
- cds_lfht_for_each_entry(ua_sess->channels, &iter, ua_chan, node) {
+ cds_lfht_for_each_entry(ua_sess->channels->ht, &iter.iter, ua_chan,
+ node.node) {
/* Create all streams */
while (1) {
/* Create UST stream */
int ust_app_stop_trace(struct ltt_ust_session *usess, struct ust_app *app)
{
int ret = 0;
- struct cds_lfht_iter iter;
+ struct lttng_ht_iter iter;
struct ust_app_session *ua_sess;
struct ust_app_channel *ua_chan;
goto error_rcu_unlock;
}
- /* Flush all buffers before stopping */
- ret = ustctl_sock_flush_buffer(app->key.sock, ua_sess->metadata->obj);
+ /* This inhibits UST tracing */
+ ret = ustctl_stop_session(app->key.sock, ua_sess->handle);
if (ret < 0) {
- ERR("UST app PID %d metadata flush failed", app->key.pid);
- ERR("Ended with ret %d", ret);
+ ERR("Error stopping tracing for app pid: %d", app->key.pid);
+ goto error_rcu_unlock;
}
- cds_lfht_for_each_entry(ua_sess->channels, &iter, ua_chan, node) {
+ /* Quiescent wait after stopping trace */
+ ustctl_wait_quiescent(app->key.sock);
+
+ /* Flushing buffers */
+ cds_lfht_for_each_entry(ua_sess->channels->ht, &iter.iter, ua_chan,
+ node.node) {
ret = ustctl_sock_flush_buffer(app->key.sock, ua_chan->obj);
if (ret < 0) {
ERR("UST app PID %d channel %s flush failed",
}
}
- /* This inhibits UST tracing */
- ret = ustctl_stop_session(app->key.sock, ua_sess->handle);
+ /* Flush all buffers before stopping */
+ ret = ustctl_sock_flush_buffer(app->key.sock, ua_sess->metadata->obj);
if (ret < 0) {
- ERR("Error stopping tracing for app pid: %d", app->key.pid);
- goto error_rcu_unlock;
+ ERR("UST app PID %d metadata flush failed", app->key.pid);
+ ERR("Ended with ret %d", ret);
}
rcu_read_unlock();
- /* Quiescent wait after stopping trace */
- ustctl_wait_quiescent(app->key.sock);
-
return 0;
error_rcu_unlock:
{
struct ust_app_session *ua_sess;
struct lttng_ust_object_data obj;
- struct cds_lfht_iter iter;
- struct cds_lfht_node *node;
+ struct lttng_ht_iter iter;
+ struct lttng_ht_node_ulong *node;
int ret;
DBG("Destroy tracing for ust app pid %d", app->key.pid);
rcu_read_lock();
__lookup_session_by_app(usess, app, &iter);
- node = hashtable_iter_get_node(&iter);
+ node = lttng_ht_iter_get_node_ulong(&iter);
if (node == NULL) {
/* Only malloc can failed so something is really wrong */
goto error_rcu_unlock;
}
ua_sess = caa_container_of(node, struct ust_app_session, node);
- ret = hashtable_del(app->sessions, &iter);
+ ret = lttng_ht_del(app->sessions, &iter);
assert(!ret);
delete_ust_app_session(app->key.sock, ua_sess);
obj.handle = ua_sess->handle;
int ust_app_start_trace_all(struct ltt_ust_session *usess)
{
int ret = 0;
- struct cds_lfht_iter iter;
+ struct lttng_ht_iter iter;
struct ust_app *app;
DBG("Starting all UST traces");
rcu_read_lock();
- cds_lfht_for_each_entry(ust_app_ht, &iter, app, node) {
+ cds_lfht_for_each_entry(ust_app_ht->ht, &iter.iter, app, node.node) {
ret = ust_app_start_trace(usess, app);
if (ret < 0) {
/* Continue to next apps even on error */
int ust_app_stop_trace_all(struct ltt_ust_session *usess)
{
int ret = 0;
- struct cds_lfht_iter iter;
+ struct lttng_ht_iter iter;
struct ust_app *app;
DBG("Stopping all UST traces");
rcu_read_lock();
- cds_lfht_for_each_entry(ust_app_ht, &iter, app, node) {
+ cds_lfht_for_each_entry(ust_app_ht->ht, &iter.iter, app, node.node) {
ret = ust_app_stop_trace(usess, app);
if (ret < 0) {
/* Continue to next apps even on error */
int ust_app_destroy_trace_all(struct ltt_ust_session *usess)
{
int ret = 0;
- struct cds_lfht_iter iter;
+ struct lttng_ht_iter iter;
struct ust_app *app;
DBG("Destroy all UST traces");
rcu_read_lock();
- cds_lfht_for_each_entry(ust_app_ht, &iter, app, node) {
+ cds_lfht_for_each_entry(ust_app_ht->ht, &iter.iter, app, node.node) {
ret = ust_app_destroy_trace(usess, app);
if (ret < 0) {
/* Continue to next apps even on error */
void ust_app_global_update(struct ltt_ust_session *usess, int sock)
{
int ret = 0;
- struct cds_lfht_iter iter;
+ struct lttng_ht_iter iter, uiter;
struct ust_app *app;
struct ust_app_session *ua_sess;
struct ust_app_channel *ua_chan;
goto error;
}
- DBG2("UST app global update for app sock %d for session uid %d", sock,
- usess->uid);
+ DBG2("UST app global update for app sock %d for session id %d", sock,
+ usess->id);
rcu_read_lock();
* app session above made a shadow copy of the UST global domain from the
* ltt ust session.
*/
- cds_lfht_for_each_entry(ua_sess->channels, &iter, ua_chan, node) {
+ cds_lfht_for_each_entry(ua_sess->channels->ht, &iter.iter, ua_chan,
+ node.node) {
ret = create_ust_channel(app, ua_sess, ua_chan);
if (ret < 0) {
/* FIXME: Should we quit here or continue... */
}
/* For each events */
- cds_lfht_for_each_entry(ua_chan->events, &iter, ua_event, node) {
+ cds_lfht_for_each_entry(ua_chan->events->ht, &uiter.iter, ua_event,
+ node.node) {
ret = create_ust_event(app, ua_sess, ua_chan, ua_event);
if (ret < 0) {
/* FIXME: Should we quit here or continue... */
struct ltt_ust_channel *uchan, struct ltt_ust_context *uctx)
{
int ret = 0;
- struct cds_lfht_node *ua_chan_node;
- struct cds_lfht_iter iter, uiter;
+ struct lttng_ht_node_str *ua_chan_node;
+ struct lttng_ht_iter iter, uiter;
struct ust_app_channel *ua_chan = NULL;
struct ust_app_session *ua_sess;
struct ust_app *app;
rcu_read_lock();
- cds_lfht_for_each_entry(ust_app_ht, &iter, app, node) {
+ cds_lfht_for_each_entry(ust_app_ht->ht, &iter.iter, app, node.node) {
ua_sess = lookup_session_by_app(usess, app);
if (ua_sess == NULL) {
continue;
}
/* Lookup channel in the ust app session */
- ua_chan_node = hashtable_lookup(ua_sess->channels,
- (void *)uchan->name, strlen(uchan->name), &uiter);
+ lttng_ht_lookup(ua_sess->channels, (void *)uchan->name, &uiter);
+ ua_chan_node = lttng_ht_iter_get_node_str(&uiter);
if (ua_chan_node == NULL) {
continue;
}
struct ltt_ust_context *uctx)
{
int ret = 0;
- struct cds_lfht_node *ua_chan_node, *ua_event_node;
- struct cds_lfht_iter iter, uiter;
+ struct lttng_ht_node_str *ua_chan_node, *ua_event_node;
+ struct lttng_ht_iter iter, uiter;
struct ust_app_session *ua_sess;
struct ust_app_event *ua_event;
struct ust_app_channel *ua_chan = NULL;
rcu_read_lock();
- cds_lfht_for_each_entry(ust_app_ht, &iter, app, node) {
+ cds_lfht_for_each_entry(ust_app_ht->ht, &iter.iter, app, node.node) {
ua_sess = lookup_session_by_app(usess, app);
if (ua_sess == NULL) {
continue;
}
/* Lookup channel in the ust app session */
- ua_chan_node = hashtable_lookup(ua_sess->channels,
- (void *)uchan->name, strlen(uchan->name), &uiter);
+ lttng_ht_lookup(ua_sess->channels, (void *)uchan->name, &uiter);
+ ua_chan_node = lttng_ht_iter_get_node_str(&uiter);
if (ua_chan_node == NULL) {
continue;
}
ua_chan = caa_container_of(ua_chan_node, struct ust_app_channel,
node);
- ua_event_node = hashtable_lookup(ua_chan->events,
- (void *)uevent->attr.name, strlen(uevent->attr.name), &uiter);
+ lttng_ht_lookup(ua_chan->events, (void *)uevent->attr.name, &uiter);
+ ua_event_node = lttng_ht_iter_get_node_str(&uiter);
if (ua_event_node == NULL) {
continue;
}
struct ltt_ust_channel *uchan, struct ltt_ust_event *uevent, pid_t pid)
{
int ret = 0;
- struct cds_lfht_iter iter;
- struct cds_lfht_node *ua_chan_node, *ua_event_node;
+ struct lttng_ht_iter iter;
+ struct lttng_ht_node_str *ua_chan_node, *ua_event_node;
struct ust_app *app;
struct ust_app_session *ua_sess;
struct ust_app_channel *ua_chan;
assert(ua_sess);
/* Lookup channel in the ust app session */
- ua_chan_node = hashtable_lookup(ua_sess->channels, (void *)uchan->name,
- strlen(uchan->name), &iter);
+ lttng_ht_lookup(ua_sess->channels, (void *)uchan->name, &iter);
+ ua_chan_node = lttng_ht_iter_get_node_str(&iter);
/* If the channel is not found, there is a code flow error */
assert(ua_chan_node);
ua_chan = caa_container_of(ua_chan_node, struct ust_app_channel, node);
- ua_event_node = hashtable_lookup(ua_chan->events,
- (void*)uevent->attr.name, strlen(uevent->attr.name), &iter);
+ lttng_ht_lookup(ua_chan->events, (void *)uevent->attr.name, &iter);
+ ua_event_node = lttng_ht_iter_get_node_str(&iter);
if (ua_event_node == NULL) {
ret = create_ust_app_event(ua_sess, ua_chan, uevent, app);
if (ret < 0) {
struct ltt_ust_channel *uchan, struct ltt_ust_event *uevent, pid_t pid)
{
int ret = 0;
- struct cds_lfht_iter iter;
- struct cds_lfht_node *ua_chan_node, *ua_event_node;
+ struct lttng_ht_iter iter;
+ struct lttng_ht_node_str *ua_chan_node, *ua_event_node;
struct ust_app *app;
struct ust_app_session *ua_sess;
struct ust_app_channel *ua_chan;
assert(ua_sess);
/* Lookup channel in the ust app session */
- ua_chan_node = hashtable_lookup(ua_sess->channels, (void *)uchan->name,
- strlen(uchan->name), &iter);
+ lttng_ht_lookup(ua_sess->channels, (void *)uchan->name, &iter);
+ ua_chan_node = lttng_ht_iter_get_node_str(&iter);
if (ua_chan_node == NULL) {
/* Channel does not exist, skip disabling */
goto error;
}
ua_chan = caa_container_of(ua_chan_node, struct ust_app_channel, node);
- ua_event_node = hashtable_lookup(ua_chan->events,
- (void*)uevent->attr.name, strlen(uevent->attr.name), &iter);
+ lttng_ht_lookup(ua_chan->events, (void *)uevent->attr.name, &iter);
+ ua_event_node = lttng_ht_iter_get_node_str(&iter);
if (ua_event_node == NULL) {
/* Event does not exist, skip disabling */
goto error;
/*
* Global applications HT used by the session daemon.
*/
-struct cds_lfht *ust_app_ht;
+struct lttng_ht *ust_app_ht;
-struct cds_lfht *ust_app_sock_key_map;
+struct lttng_ht *ust_app_sock_key_map;
struct ust_app_key {
pid_t pid;
int sock;
- struct cds_lfht_node node;
+ struct lttng_ht_node_ulong node;
};
struct ust_app_ctx {
int handle;
struct lttng_ust_context ctx;
struct lttng_ust_object_data *obj;
- struct cds_lfht_node node;
+ struct lttng_ht_node_ulong node;
};
struct ust_app_event {
struct lttng_ust_object_data *obj;
struct lttng_ust_event attr;
char name[LTTNG_UST_SYM_NAME_LEN];
- struct cds_lfht *ctx;
- struct cds_lfht_node node;
+ struct lttng_ht *ctx;
+ struct lttng_ht_node_str node;
};
struct ust_app_channel {
struct lttng_ust_channel attr;
struct lttng_ust_object_data *obj;
struct ltt_ust_stream_list streams;
- struct cds_lfht *ctx;
- struct cds_lfht *events;
- struct cds_lfht_node node;
+ struct lttng_ht *ctx;
+ struct lttng_ht *events;
+ struct lttng_ht_node_str node;
};
struct ust_app_session {
int enabled;
/* started: has the session been in started state at any time ? */
int started; /* allows detection of start vs restart. */
- int handle; /* Used has unique identifier */
- unsigned int uid;
+ int handle; /* used has unique identifier for app session */
+ int id; /* session unique identifier */
struct ltt_ust_metadata *metadata;
- struct cds_lfht *channels; /* Registered channels */
- struct cds_lfht_node node;
+ struct lttng_ht *channels; /* Registered channels */
+ struct lttng_ht_node_ulong node;
char path[PATH_MAX];
+ /* UID/GID of the user owning the session */
+ uid_t uid;
+ gid_t gid;
};
/*
uint32_t v_major; /* Verion major number */
uint32_t v_minor; /* Verion minor number */
char name[17]; /* Process name (short) */
- struct cds_lfht *sessions;
- struct cds_lfht_node node;
+ struct lttng_ht *sessions;
+ struct lttng_ht_node_ulong node;
struct ust_app_key key;
};
#ifdef HAVE_LIBLTTNG_UST_CTL
int ust_app_register(struct ust_register_msg *msg, int sock);
+static inline
+int ust_app_register_done(int sock)
+{
+ return ustctl_register_done(sock);
+}
void ust_app_unregister(int sock);
unsigned long ust_app_list_count(void);
int ust_app_start_trace(struct ltt_ust_session *usess, struct ust_app *app);
void ust_app_clean_list(void);
void ust_app_ht_alloc(void);
-struct cds_lfht *ust_app_get_ht(void);
+struct lttng_ht *ust_app_get_ht(void);
struct ust_app *ust_app_find_by_pid(pid_t pid);
#else /* HAVE_LIBLTTNG_UST_CTL */
return -ENOSYS;
}
static inline
+int ust_app_register_done(int sock)
+{
+ return -ENOSYS;
+}
+static inline
void ust_app_unregister(int sock)
{
}
return NULL;
}
static inline
-struct cds_lfht *ust_app_get_ht(void)
+struct lttng_ht *ust_app_get_ht(void)
{
return NULL;
}
#include <unistd.h>
#include <lttngerr.h>
+#include <lttng-ht.h>
#include <lttng-share.h>
#include <lttng-sessiond-comm.h>
#include <lttng/lttng-consumer.h>
-#include "hashtable.h"
#include "ust-consumer.h"
/*
* Send all stream fds of UST channel to the consumer.
*/
static int send_channel_streams(int sock,
- struct ust_app_channel *uchan)
+ struct ust_app_channel *uchan,
+ uid_t uid, gid_t gid)
{
int ret, fd;
struct lttcomm_consumer_msg lum;
*/
lum.u.stream.output = DEFAULT_UST_CHANNEL_OUTPUT;
lum.u.stream.mmap_len = stream->obj->memory_map_size;
+ lum.u.stream.uid = uid;
+ lum.u.stream.gid = gid;
strncpy(lum.u.stream.path_name, stream->pathname, PATH_MAX - 1);
lum.u.stream.path_name[PATH_MAX - 1] = '\0';
DBG("Sending stream %d to consumer", lum.u.stream.stream_key);
{
int ret = 0;
int sock = consumer_fd;
- struct cds_lfht_iter iter;
- struct cds_lfht_node *node;
+ struct lttng_ht_iter iter;
struct lttcomm_consumer_msg lum;
- struct ust_app_channel *uchan;
+ struct ust_app_channel *ua_chan;
DBG("Sending metadata stream fd");
lum.u.stream.state = LTTNG_CONSUMER_ACTIVE_STREAM;
lum.u.stream.output = DEFAULT_UST_CHANNEL_OUTPUT;
lum.u.stream.mmap_len = usess->metadata->stream_obj->memory_map_size;
+ lum.u.stream.uid = usess->uid;
+ lum.u.stream.gid = usess->gid;
strncpy(lum.u.stream.path_name, usess->metadata->pathname, PATH_MAX - 1);
lum.u.stream.path_name[PATH_MAX - 1] = '\0';
DBG("Sending metadata stream %d to consumer", lum.u.stream.stream_key);
/* Send each channel fd streams of session */
rcu_read_lock();
- hashtable_get_first(usess->channels, &iter);
- while ((node = hashtable_iter_get_node(&iter)) != NULL) {
- uchan = caa_container_of(node, struct ust_app_channel, node);
-
- ret = send_channel_streams(sock, uchan);
+ cds_lfht_for_each_entry(usess->channels->ht, &iter.iter, ua_chan,
+ node.node) {
+ ret = send_channel_streams(sock, ua_chan, usess->uid, usess->gid);
if (ret < 0) {
rcu_read_unlock();
goto error;
}
- hashtable_get_next(usess->channels, &iter);
}
rcu_read_unlock();
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
-#include <sys/types.h>
-#include <sys/stat.h>
#include <unistd.h>
#include <lttngerr.h>
{
return ((const char *) getenv("HOME"));
}
-
-/*
- * Create recursively directory using the FULL path.
- */
-int mkdir_recursive(const char *path, mode_t mode, uid_t uid, gid_t gid)
-{
- int ret;
- char *p, tmp[PATH_MAX];
- size_t len;
- mode_t old_umask;
-
- ret = snprintf(tmp, sizeof(tmp), "%s", path);
- if (ret < 0) {
- PERROR("snprintf mkdir");
- goto error;
- }
-
- len = ret;
- if (tmp[len - 1] == '/') {
- tmp[len - 1] = 0;
- }
-
- old_umask = umask(0);
- for (p = tmp + 1; *p; p++) {
- if (*p == '/') {
- *p = 0;
- ret = mkdir(tmp, mode);
- if (ret < 0) {
- if (!(errno == EEXIST)) {
- PERROR("mkdir recursive");
- ret = -errno;
- goto umask_error;
- }
- } else if (ret == 0) {
- /*
- * We created the directory. Set its ownership to the
- * user/group specified.
- */
- ret = chown(tmp, uid, gid);
- if (ret < 0) {
- PERROR("chown in mkdir recursive");
- ret = -errno;
- goto umask_error;
- }
- }
- *p = '/';
- }
- }
-
- ret = mkdir(tmp, mode);
- if (ret < 0) {
- if (!(errno == EEXIST)) {
- PERROR("mkdir recursive last piece");
- ret = -errno;
- } else {
- ret = 0;
- }
- } else if (ret == 0) {
- /*
- * We created the directory. Set its ownership to the user/group
- * specified.
- */
- ret = chown(tmp, uid, gid);
- if (ret < 0) {
- PERROR("chown in mkdir recursive");
- ret = -errno;
- goto umask_error;
- }
- }
-
-umask_error:
- umask(old_umask);
-error:
- return ret;
-}
#ifndef _LTT_UTILS_H
#define _LTT_UTILS_H
-#include <unistd.h>
-
#ifndef __stringify
#define __stringify1(x) #x
#define __stringify(x) __stringify1(x)
#endif
-int mkdir_recursive(const char *path, mode_t mode, uid_t uid, gid_t gid);
const char *get_home_dir(void);
int notify_thread_pipe(int wpipe);
kernel_event_basic
UTILS=utils.h
-SESSIONS=$(top_srcdir)/lttng-sessiond/session.c $(top_srcdir)/lttng-sessiond/hashtable.c \
- $(top_srcdir)/hashtable/rculfhash.c $(top_srcdir)/hashtable/hash.c
+SESSIONS=$(top_srcdir)/lttng-sessiond/session.c
KERN_DATA_TRACE=$(top_srcdir)/lttng-sessiond/trace-kernel.c
LIBLTTNG=$(top_srcdir)/liblttngctl/lttngctl.c \
$(top_srcdir)/liblttng-sessiond-comm/lttng-sessiond-comm.c
test_sessions_SOURCES = test_sessions.c $(UTILS) $(SESSIONS)
+test_sessions_LDADD = $(top_builddir)/common/libcommon.la
test_kernel_data_trace_SOURCES = test_kernel_data_trace.c $(UTILS) $(KERN_DATA_TRACE)
#include <string.h>
#include <unistd.h>
#include <time.h>
+#include <sys/types.h>
#include <lttng-sessiond-comm.h>
{
int ret;
- ret = session_create(name, path);
+ ret = session_create(name, path, geteuid(), getegid());
if (ret == LTTCOMM_OK) {
/* Validate */
ret = find_session_name(name);
}
printf("Create 1 session %s: ", SESSION1);
+ fflush(stdout);
ret = create_one_session(SESSION1, PATH1);
if (ret < 0) {
return -1;
PRINT_OK();
printf("Validating created session %s: ", SESSION1);
+ fflush(stdout);
tmp = session_find_by_name(SESSION1);
if (tmp == NULL) {
return -1;
PRINT_OK();
printf("Destroy 1 session %s: ", SESSION1);
+ fflush(stdout);
ret = destroy_one_session(tmp);
if (ret < 0) {
return -1;
PRINT_OK();
printf("Two session with same name: ");
+ fflush(stdout);
ret = two_session_same_name();
if (ret < 0) {
return -1;
empty_session_list();
printf("Fuzzing create_session arguments: ");
+ fflush(stdout);
ret = fuzzing_create_args();
if (ret < 0) {
return -1;
PRINT_OK();
printf("Fuzzing destroy_session argument: ");
+ fflush(stdout);
ret = fuzzing_destroy_args();
if (ret < 0) {
return -1;
PRINT_OK();
printf("Creating %d sessions: ", MAX_SESSIONS);
+ fflush(stdout);
for (i = 0; i < MAX_SESSIONS; i++) {
tmp_name = get_random_string();
ret = create_one_session(tmp_name, PATH1);
PRINT_OK();
printf("Destroying %d sessions: ", MAX_SESSIONS);
+ fflush(stdout);
for (i = 0; i < MAX_SESSIONS; i++) {
cds_list_for_each_entry_safe(iter, tmp, &session_list->head, list) {
ret = destroy_one_session(iter);
projects / lttng-tools.git / commitdiff
