4 * Linux Trace Toolkit Next Generation Kernel State Dump
6 * Copyright 2005 Jean-Hugues Deschenes <jean-hugues.deschenes@polymtl.ca>
7 * Copyright 2006-2012 Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
9 * This library is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public
11 * License as published by the Free Software Foundation; only
12 * version 2.1 of the License.
14 * This library is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with this library; if not, write to the Free Software
21 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * Eric Clement: Add listing of network IP interface
25 * 2006, 2007 Mathieu Desnoyers Fix kernel threads
29 #include <linux/init.h>
30 #include <linux/module.h>
31 #include <linux/netlink.h>
32 #include <linux/inet.h>
34 #include <linux/kthread.h>
35 #include <linux/proc_fs.h>
36 #include <linux/file.h>
37 #include <linux/interrupt.h>
38 #include <linux/irqnr.h>
39 #include <linux/cpu.h>
40 #include <linux/netdevice.h>
41 #include <linux/inetdevice.h>
42 #include <linux/sched.h>
44 #include <linux/fdtable.h>
45 #include <linux/swap.h>
46 #include <linux/wait.h>
47 #include <linux/mutex.h>
48 #include <linux/device.h>
50 #include "lttng-events.h"
51 #include "lttng-tracer.h"
52 #include "wrapper/irqdesc.h"
53 #include "wrapper/spinlock.h"
54 #include "wrapper/fdtable.h"
55 #include "wrapper/nsproxy.h"
56 #include "wrapper/irq.h"
57 #include "wrapper/tracepoint.h"
58 #include "wrapper/genhd.h"
60 #ifdef CONFIG_LTTNG_HAS_LIST_IRQ
61 #include <linux/irq.h>
64 /* Define the tracepoints, but do not build the probes */
65 #define CREATE_TRACE_POINTS
66 #define TRACE_INCLUDE_PATH ../instrumentation/events/lttng-module
67 #define TRACE_INCLUDE_FILE lttng-statedump
68 #include "instrumentation/events/lttng-module/lttng-statedump.h"
70 DEFINE_TRACE(lttng_statedump_block_device
);
71 DEFINE_TRACE(lttng_statedump_end
);
72 DEFINE_TRACE(lttng_statedump_interrupt
);
73 DEFINE_TRACE(lttng_statedump_file_descriptor
);
74 DEFINE_TRACE(lttng_statedump_start
);
75 DEFINE_TRACE(lttng_statedump_process_state
);
76 DEFINE_TRACE(lttng_statedump_network_interface
);
80 struct lttng_session
*session
;
81 struct task_struct
*p
;
86 * Protected by the trace lock.
88 static struct delayed_work cpu_work
[NR_CPUS
];
89 static DECLARE_WAIT_QUEUE_HEAD(statedump_wq
);
90 static atomic_t kernel_threads_to_run
;
92 enum lttng_thread_type
{
93 LTTNG_USER_THREAD
= 0,
94 LTTNG_KERNEL_THREAD
= 1,
97 enum lttng_execution_mode
{
103 LTTNG_MODE_UNKNOWN
= 5,
106 enum lttng_execution_submode
{
111 enum lttng_process_status
{
123 int lttng_enumerate_block_devices(struct lttng_session
*session
)
125 struct class *ptr_block_class
;
126 struct device_type
*ptr_disk_type
;
127 struct class_dev_iter iter
;
130 ptr_block_class
= wrapper_get_block_class();
131 if (!ptr_block_class
)
133 ptr_disk_type
= wrapper_get_disk_type();
134 if (!ptr_disk_type
) {
137 class_dev_iter_init(&iter
, ptr_block_class
, NULL
, ptr_disk_type
);
138 while ((dev
= class_dev_iter_next(&iter
))) {
139 struct disk_part_iter piter
;
140 struct gendisk
*disk
= dev_to_disk(dev
);
141 struct hd_struct
*part
;
144 * Don't show empty devices or things that have been
147 if (get_capacity(disk
) == 0 ||
148 (disk
->flags
& GENHD_FL_SUPPRESS_PARTITION_INFO
))
151 disk_part_iter_init(&piter
, disk
, DISK_PITER_INCL_PART0
);
152 while ((part
= disk_part_iter_next(&piter
))) {
153 char name_buf
[BDEVNAME_SIZE
];
156 p
= wrapper_disk_name(disk
, part
->partno
, name_buf
);
158 disk_part_iter_exit(&piter
);
159 class_dev_iter_exit(&iter
);
162 trace_lttng_statedump_block_device(session
,
163 part_devt(part
), name_buf
);
165 disk_part_iter_exit(&piter
);
167 class_dev_iter_exit(&iter
);
174 void lttng_enumerate_device(struct lttng_session
*session
,
175 struct net_device
*dev
)
177 struct in_device
*in_dev
;
178 struct in_ifaddr
*ifa
;
180 if (dev
->flags
& IFF_UP
) {
181 in_dev
= in_dev_get(dev
);
183 for (ifa
= in_dev
->ifa_list
; ifa
!= NULL
;
184 ifa
= ifa
->ifa_next
) {
185 trace_lttng_statedump_network_interface(
191 trace_lttng_statedump_network_interface(
197 int lttng_enumerate_network_ip_interface(struct lttng_session
*session
)
199 struct net_device
*dev
;
201 read_lock(&dev_base_lock
);
202 for_each_netdev(&init_net
, dev
)
203 lttng_enumerate_device(session
, dev
);
204 read_unlock(&dev_base_lock
);
208 #else /* CONFIG_INET */
210 int lttng_enumerate_network_ip_interface(struct lttng_session
*session
)
214 #endif /* CONFIG_INET */
217 int lttng_dump_one_fd(const void *p
, struct file
*file
, unsigned int fd
)
219 const struct lttng_fd_ctx
*ctx
= p
;
220 const char *s
= d_path(&file
->f_path
, ctx
->page
, PAGE_SIZE
);
221 unsigned int flags
= file
->f_flags
;
224 * We don't expose kernel internal flags, only userspace-visible
227 flags
&= ~FMODE_NONOTIFY
;
228 if (test_bit(fd
, ctx
->fdt
->close_on_exec
))
231 struct dentry
*dentry
= file
->f_path
.dentry
;
233 /* Make sure we give at least some info */
234 spin_lock(&dentry
->d_lock
);
235 trace_lttng_statedump_file_descriptor(ctx
->session
, ctx
->p
, fd
,
236 dentry
->d_name
.name
, flags
, file
->f_mode
);
237 spin_unlock(&dentry
->d_lock
);
240 trace_lttng_statedump_file_descriptor(ctx
->session
, ctx
->p
, fd
, s
,
241 flags
, file
->f_mode
);
247 void lttng_enumerate_task_fd(struct lttng_session
*session
,
248 struct task_struct
*p
, char *tmp
)
250 struct lttng_fd_ctx ctx
= { .page
= tmp
, .session
= session
, .p
= p
};
253 ctx
.fdt
= files_fdtable(p
->files
);
254 lttng_iterate_fd(p
->files
, 0, lttng_dump_one_fd
, &ctx
);
259 int lttng_enumerate_file_descriptors(struct lttng_session
*session
)
261 struct task_struct
*p
;
262 char *tmp
= (char *) __get_free_page(GFP_KERNEL
);
264 /* Enumerate active file descriptors */
267 lttng_enumerate_task_fd(session
, p
, tmp
);
269 free_page((unsigned long) tmp
);
275 * FIXME: we cannot take a mmap_sem while in a RCU read-side critical section
276 * (scheduling in atomic). Normally, the tasklist lock protects this kind of
277 * iteration, but it is not exported to modules.
280 void lttng_enumerate_task_vm_maps(struct lttng_session
*session
,
281 struct task_struct
*p
)
283 struct mm_struct
*mm
;
284 struct vm_area_struct
*map
;
287 /* get_task_mm does a task_lock... */
294 down_read(&mm
->mmap_sem
);
297 ino
= map
->vm_file
->f_dentry
->d_inode
->i_ino
;
300 trace_lttng_statedump_vm_map(session
, p
, map
, ino
);
303 up_read(&mm
->mmap_sem
);
309 int lttng_enumerate_vm_maps(struct lttng_session
*session
)
311 struct task_struct
*p
;
315 lttng_enumerate_task_vm_maps(session
, p
);
321 #ifdef CONFIG_LTTNG_HAS_LIST_IRQ
323 #if (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,39))
324 #define irq_desc_get_chip(desc) get_irq_desc_chip(desc)
328 void lttng_list_interrupts(struct lttng_session
*session
)
331 unsigned long flags
= 0;
332 struct irq_desc
*desc
;
334 #define irq_to_desc wrapper_irq_to_desc
336 for_each_irq_desc(irq
, desc
) {
337 struct irqaction
*action
;
338 const char *irq_chip_name
=
339 irq_desc_get_chip(desc
)->name
? : "unnamed_irq_chip";
341 local_irq_save(flags
);
342 wrapper_desc_spin_lock(&desc
->lock
);
343 for (action
= desc
->action
; action
; action
= action
->next
) {
344 trace_lttng_statedump_interrupt(session
,
345 irq
, irq_chip_name
, action
);
347 wrapper_desc_spin_unlock(&desc
->lock
);
348 local_irq_restore(flags
);
354 void lttng_list_interrupts(struct lttng_session
*session
)
360 void lttng_statedump_process_ns(struct lttng_session
*session
,
361 struct task_struct
*p
,
362 enum lttng_thread_type type
,
363 enum lttng_execution_mode mode
,
364 enum lttng_execution_submode submode
,
365 enum lttng_process_status status
)
367 struct nsproxy
*proxy
;
368 struct pid_namespace
*pid_ns
;
371 proxy
= task_nsproxy(p
);
373 pid_ns
= lttng_get_proxy_pid_ns(proxy
);
375 trace_lttng_statedump_process_state(session
,
376 p
, type
, mode
, submode
, status
, pid_ns
);
377 pid_ns
= pid_ns
->parent
;
380 trace_lttng_statedump_process_state(session
,
381 p
, type
, mode
, submode
, status
, NULL
);
387 int lttng_enumerate_process_states(struct lttng_session
*session
)
389 struct task_struct
*g
, *p
;
392 for_each_process(g
) {
395 enum lttng_execution_mode mode
=
397 enum lttng_execution_submode submode
=
399 enum lttng_process_status status
;
400 enum lttng_thread_type type
;
403 if (p
->exit_state
== EXIT_ZOMBIE
)
404 status
= LTTNG_ZOMBIE
;
405 else if (p
->exit_state
== EXIT_DEAD
)
407 else if (p
->state
== TASK_RUNNING
) {
408 /* Is this a forked child that has not run yet? */
409 if (list_empty(&p
->rt
.run_list
))
410 status
= LTTNG_WAIT_FORK
;
413 * All tasks are considered as wait_cpu;
414 * the viewer will sort out if the task
415 * was really running at this time.
417 status
= LTTNG_WAIT_CPU
;
418 } else if (p
->state
&
419 (TASK_INTERRUPTIBLE
| TASK_UNINTERRUPTIBLE
)) {
420 /* Task is waiting for something to complete */
423 status
= LTTNG_UNNAMED
;
424 submode
= LTTNG_NONE
;
427 * Verification of t->mm is to filter out kernel
428 * threads; Viewer will further filter out if a
429 * user-space thread was in syscall mode or not.
432 type
= LTTNG_USER_THREAD
;
434 type
= LTTNG_KERNEL_THREAD
;
435 lttng_statedump_process_ns(session
,
436 p
, type
, mode
, submode
, status
);
438 } while_each_thread(g
, p
);
446 void lttng_statedump_work_func(struct work_struct
*work
)
448 if (atomic_dec_and_test(&kernel_threads_to_run
))
449 /* If we are the last thread, wake up do_lttng_statedump */
450 wake_up(&statedump_wq
);
454 int do_lttng_statedump(struct lttng_session
*session
)
458 trace_lttng_statedump_start(session
);
459 lttng_enumerate_process_states(session
);
460 lttng_enumerate_file_descriptors(session
);
461 /* FIXME lttng_enumerate_vm_maps(session); */
462 lttng_list_interrupts(session
);
463 lttng_enumerate_network_ip_interface(session
);
464 lttng_enumerate_block_devices(session
);
466 /* TODO lttng_dump_idt_table(session); */
467 /* TODO lttng_dump_softirq_vec(session); */
468 /* TODO lttng_list_modules(session); */
469 /* TODO lttng_dump_swap_files(session); */
472 * Fire off a work queue on each CPU. Their sole purpose in life
473 * is to guarantee that each CPU has been in a state where is was in
474 * syscall mode (i.e. not in a trap, an IRQ or a soft IRQ).
477 atomic_set(&kernel_threads_to_run
, num_online_cpus());
478 for_each_online_cpu(cpu
) {
479 INIT_DELAYED_WORK(&cpu_work
[cpu
], lttng_statedump_work_func
);
480 schedule_delayed_work_on(cpu
, &cpu_work
[cpu
], 0);
482 /* Wait for all threads to run */
483 __wait_event(statedump_wq
, (atomic_read(&kernel_threads_to_run
) == 0));
485 /* Our work is done */
486 trace_lttng_statedump_end(session
);
491 * Called with session mutex held.
493 int lttng_statedump_start(struct lttng_session
*session
)
495 return do_lttng_statedump(session
);
497 EXPORT_SYMBOL_GPL(lttng_statedump_start
);
500 int __init
lttng_statedump_init(void)
503 * Allow module to load even if the fixup cannot be done. This
504 * will allow seemless transition when the underlying issue fix
505 * is merged into the Linux kernel, and when tracepoint.c
506 * "tracepoint_module_notify" is turned into a static function.
508 (void) wrapper_lttng_fixup_sig(THIS_MODULE
);
512 module_init(lttng_statedump_init
);
515 void __exit
lttng_statedump_exit(void)
519 module_exit(lttng_statedump_exit
);
521 MODULE_LICENSE("GPL and additional rights");
522 MODULE_AUTHOR("Jean-Hugues Deschenes");
523 MODULE_DESCRIPTION("Linux Trace Toolkit Next Generation Statedump");
524 MODULE_VERSION(__stringify(LTTNG_MODULES_MAJOR_VERSION
) "."
525 __stringify(LTTNG_MODULES_MINOR_VERSION
) "."
526 __stringify(LTTNG_MODULES_PATCHLEVEL_VERSION
)
527 LTTNG_MODULES_EXTRAVERSION
);