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>
49 #include "lttng-events.h"
50 #include "wrapper/irqdesc.h"
52 #ifdef CONFIG_GENERIC_HARDIRQS
53 #include <linux/irq.h>
56 /* Define the tracepoints, but do not build the probes */
57 #define CREATE_TRACE_POINTS
58 #define TRACE_INCLUDE_PATH ../instrumentation/events/lttng-module
59 #define TRACE_INCLUDE_FILE lttng-statedump
60 #include "instrumentation/events/lttng-module/lttng-statedump.h"
63 * Protected by the trace lock.
65 static struct delayed_work cpu_work
[NR_CPUS
];
66 static DECLARE_WAIT_QUEUE_HEAD(statedump_wq
);
67 static atomic_t kernel_threads_to_run
;
69 enum lttng_thread_type
{
70 LTTNG_USER_THREAD
= 0,
71 LTTNG_KERNEL_THREAD
= 1,
74 enum lttng_execution_mode
{
80 LTTNG_MODE_UNKNOWN
= 5,
83 enum lttng_execution_submode
{
88 enum lttng_process_status
{
101 void lttng_enumerate_device(struct lttng_session
*session
,
102 struct net_device
*dev
)
104 struct in_device
*in_dev
;
105 struct in_ifaddr
*ifa
;
107 if (dev
->flags
& IFF_UP
) {
108 in_dev
= in_dev_get(dev
);
110 for (ifa
= in_dev
->ifa_list
; ifa
!= NULL
;
111 ifa
= ifa
->ifa_next
) {
112 trace_lttng_statedump_network_interface(
118 trace_lttng_statedump_network_interface(
124 int lttng_enumerate_network_ip_interface(struct lttng_session
*session
)
126 struct net_device
*dev
;
128 read_lock(&dev_base_lock
);
129 for_each_netdev(&init_net
, dev
)
130 lttng_enumerate_device(session
, dev
);
131 read_unlock(&dev_base_lock
);
135 #else /* CONFIG_INET */
137 int lttng_enumerate_network_ip_interface(struct lttng_session
*session
)
141 #endif /* CONFIG_INET */
145 void lttng_enumerate_task_fd(struct lttng_session
*session
,
146 struct task_struct
*p
, char *tmp
)
151 const unsigned char *path
;
156 spin_lock(&p
->files
->file_lock
);
157 fdt
= files_fdtable(p
->files
);
158 for (i
= 0; i
< fdt
->max_fds
; i
++) {
159 filp
= fcheck_files(p
->files
, i
);
162 path
= d_path(&filp
->f_path
, tmp
, PAGE_SIZE
);
163 /* Make sure we give at least some info */
164 trace_lttng_statedump_file_descriptor(session
, p
, i
,
166 filp
->f_dentry
->d_name
.name
:
169 spin_unlock(&p
->files
->file_lock
);
175 int lttng_enumerate_file_descriptors(struct lttng_session
*session
)
177 struct task_struct
*p
;
178 char *tmp
= (char *) __get_free_page(GFP_KERNEL
);
180 /* Enumerate active file descriptors */
183 lttng_enumerate_task_fd(session
, p
, tmp
);
185 free_page((unsigned long) tmp
);
191 * FIXME: we cannot take a mmap_sem while in a RCU read-side critical section
192 * (scheduling in atomic). Normally, the tasklist lock protects this kind of
193 * iteration, but it is not exported to modules.
196 void lttng_enumerate_task_vm_maps(struct lttng_session
*session
,
197 struct task_struct
*p
)
199 struct mm_struct
*mm
;
200 struct vm_area_struct
*map
;
203 /* get_task_mm does a task_lock... */
210 down_read(&mm
->mmap_sem
);
213 ino
= map
->vm_file
->f_dentry
->d_inode
->i_ino
;
216 trace_lttng_statedump_vm_map(session
, p
, map
, ino
);
219 up_read(&mm
->mmap_sem
);
225 int lttng_enumerate_vm_maps(struct lttng_session
*session
)
227 struct task_struct
*p
;
231 lttng_enumerate_task_vm_maps(session
, p
);
237 #ifdef CONFIG_GENERIC_HARDIRQS
239 #if (LINUX_VERSION_CODE < KERNEL_VERSION(2,6,39))
240 #define irq_desc_get_chip(desc) get_irq_desc_chip(desc)
244 void lttng_list_interrupts(struct lttng_session
*session
)
247 unsigned long flags
= 0;
248 struct irq_desc
*desc
;
250 #define irq_to_desc wrapper_irq_to_desc
252 for_each_irq_desc(irq
, desc
) {
253 struct irqaction
*action
;
254 const char *irq_chip_name
=
255 irq_desc_get_chip(desc
)->name
? : "unnamed_irq_chip";
257 local_irq_save(flags
);
258 raw_spin_lock(&desc
->lock
);
259 for (action
= desc
->action
; action
; action
= action
->next
) {
260 trace_lttng_statedump_interrupt(session
,
261 irq
, irq_chip_name
, action
);
263 raw_spin_unlock(&desc
->lock
);
264 local_irq_restore(flags
);
270 void list_interrupts(struct lttng_session
*session
)
276 void lttng_statedump_process_ns(struct lttng_session
*session
,
277 struct task_struct
*p
,
278 enum lttng_thread_type type
,
279 enum lttng_execution_mode mode
,
280 enum lttng_execution_submode submode
,
281 enum lttng_process_status status
)
283 struct nsproxy
*proxy
;
284 struct pid_namespace
*pid_ns
;
287 proxy
= task_nsproxy(p
);
289 pid_ns
= proxy
->pid_ns
;
291 trace_lttng_statedump_process_state(session
,
292 p
, type
, mode
, submode
, status
, pid_ns
);
295 trace_lttng_statedump_process_state(session
,
296 p
, type
, mode
, submode
, status
, NULL
);
302 int lttng_enumerate_process_states(struct lttng_session
*session
)
304 struct task_struct
*g
, *p
;
307 for_each_process(g
) {
310 enum lttng_execution_mode mode
=
312 enum lttng_execution_submode submode
=
314 enum lttng_process_status status
;
315 enum lttng_thread_type type
;
318 if (p
->exit_state
== EXIT_ZOMBIE
)
319 status
= LTTNG_ZOMBIE
;
320 else if (p
->exit_state
== EXIT_DEAD
)
322 else if (p
->state
== TASK_RUNNING
) {
323 /* Is this a forked child that has not run yet? */
324 if (list_empty(&p
->rt
.run_list
))
325 status
= LTTNG_WAIT_FORK
;
328 * All tasks are considered as wait_cpu;
329 * the viewer will sort out if the task
330 * was really running at this time.
332 status
= LTTNG_WAIT_CPU
;
333 } else if (p
->state
&
334 (TASK_INTERRUPTIBLE
| TASK_UNINTERRUPTIBLE
)) {
335 /* Task is waiting for something to complete */
338 status
= LTTNG_UNNAMED
;
339 submode
= LTTNG_NONE
;
342 * Verification of t->mm is to filter out kernel
343 * threads; Viewer will further filter out if a
344 * user-space thread was in syscall mode or not.
347 type
= LTTNG_USER_THREAD
;
349 type
= LTTNG_KERNEL_THREAD
;
350 lttng_statedump_process_ns(session
,
351 p
, type
, mode
, submode
, status
);
353 } while_each_thread(g
, p
);
361 void lttng_statedump_work_func(struct work_struct
*work
)
363 if (atomic_dec_and_test(&kernel_threads_to_run
))
364 /* If we are the last thread, wake up do_lttng_statedump */
365 wake_up(&statedump_wq
);
369 int do_lttng_statedump(struct lttng_session
*session
)
373 printk(KERN_DEBUG
"LTT state dump thread start\n");
374 trace_lttng_statedump_start(session
);
375 lttng_enumerate_process_states(session
);
376 lttng_enumerate_file_descriptors(session
);
377 /* FIXME lttng_enumerate_vm_maps(session); */
378 lttng_list_interrupts(session
);
379 lttng_enumerate_network_ip_interface(session
);
381 /* TODO lttng_dump_idt_table(session); */
382 /* TODO lttng_dump_softirq_vec(session); */
383 /* TODO lttng_list_modules(session); */
384 /* TODO lttng_dump_swap_files(session); */
387 * Fire off a work queue on each CPU. Their sole purpose in life
388 * is to guarantee that each CPU has been in a state where is was in
389 * syscall mode (i.e. not in a trap, an IRQ or a soft IRQ).
392 atomic_set(&kernel_threads_to_run
, num_online_cpus());
393 for_each_online_cpu(cpu
) {
394 INIT_DELAYED_WORK(&cpu_work
[cpu
], lttng_statedump_work_func
);
395 schedule_delayed_work_on(cpu
, &cpu_work
[cpu
], 0);
397 /* Wait for all threads to run */
398 __wait_event(statedump_wq
, (atomic_read(&kernel_threads_to_run
) != 0));
400 /* Our work is done */
401 printk(KERN_DEBUG
"LTT state dump end\n");
402 trace_lttng_statedump_end(session
);
407 * Called with session mutex held.
409 int lttng_statedump_start(struct lttng_session
*session
)
411 printk(KERN_DEBUG
"LTTng: state dump begin\n");
412 return do_lttng_statedump(session
);
414 EXPORT_SYMBOL_GPL(lttng_statedump_start
);
416 MODULE_LICENSE("GPL and additional rights");
417 MODULE_AUTHOR("Jean-Hugues Deschenes");
418 MODULE_DESCRIPTION("Linux Trace Toolkit Next Generation Statedump");