1 ftrace - Function Tracer
2 ========================
4 Copyright 2008 Red Hat Inc.
5 Author: Steven Rostedt <srostedt@redhat.com>
6 License: The GNU Free Documentation License, Version 1.2
7 (dual licensed under the GPL v2)
8 Reviewers: Elias Oltmanns, Randy Dunlap, Andrew Morton,
9 John Kacur, and David Teigland.
11 Written for: 2.6.28-rc2
16 Ftrace is an internal tracer designed to help out developers and
17 designers of systems to find what is going on inside the kernel.
18 It can be used for debugging or analyzing latencies and performance
19 issues that take place outside of user-space.
21 Although ftrace is the function tracer, it also includes an
22 infrastructure that allows for other types of tracing. Some of the
23 tracers that are currently in ftrace include a tracer to trace
24 context switches, the time it takes for a high priority task to
25 run after it was woken up, the time interrupts are disabled, and
26 more (ftrace allows for tracer plugins, which means that the list of
27 tracers can always grow).
33 Ftrace uses the debugfs file system to hold the control files as well
34 as the files to display output.
36 To mount the debugfs system:
39 # mount -t debugfs nodev /debug
41 (Note: it is more common to mount at /sys/kernel/debug, but for simplicity
42 this document will use /debug)
44 That's it! (assuming that you have ftrace configured into your kernel)
46 After mounting the debugfs, you can see a directory called
47 "tracing". This directory contains the control and output files
48 of ftrace. Here is a list of some of the key files:
51 Note: all time values are in microseconds.
53 current_tracer: This is used to set or display the current tracer
56 available_tracers: This holds the different types of tracers that
57 have been compiled into the kernel. The tracers
58 listed here can be configured by echoing their name
61 tracing_enabled: This sets or displays whether the current_tracer
62 is activated and tracing or not. Echo 0 into this
63 file to disable the tracer or 1 to enable it.
65 trace: This file holds the output of the trace in a human readable
66 format (described below).
68 latency_trace: This file shows the same trace but the information
69 is organized more to display possible latencies
70 in the system (described below).
72 trace_pipe: The output is the same as the "trace" file but this
73 file is meant to be streamed with live tracing.
74 Reads from this file will block until new data
75 is retrieved. Unlike the "trace" and "latency_trace"
76 files, this file is a consumer. This means reading
77 from this file causes sequential reads to display
78 more current data. Once data is read from this
79 file, it is consumed, and will not be read
80 again with a sequential read. The "trace" and
81 "latency_trace" files are static, and if the
82 tracer is not adding more data, they will display
83 the same information every time they are read.
85 trace_options: This file lets the user control the amount of data
86 that is displayed in one of the above output
89 trace_max_latency: Some of the tracers record the max latency.
90 For example, the time interrupts are disabled.
91 This time is saved in this file. The max trace
92 will also be stored, and displayed by either
93 "trace" or "latency_trace". A new max trace will
94 only be recorded if the latency is greater than
95 the value in this file. (in microseconds)
97 buffer_size_kb: This sets or displays the number of kilobytes each CPU
98 buffer can hold. The tracer buffers are the same size
99 for each CPU. The displayed number is the size of the
100 CPU buffer and not total size of all buffers. The
101 trace buffers are allocated in pages (blocks of memory
102 that the kernel uses for allocation, usually 4 KB in size).
103 If the last page allocated has room for more bytes
104 than requested, the rest of the page will be used,
105 making the actual allocation bigger than requested.
106 (Note, the size may not be a multiple of the page size due
107 to buffer managment overhead.)
109 This can only be updated when the current_tracer
112 tracing_cpumask: This is a mask that lets the user only trace
113 on specified CPUS. The format is a hex string
114 representing the CPUS.
116 set_ftrace_filter: When dynamic ftrace is configured in (see the
117 section below "dynamic ftrace"), the code is dynamically
118 modified (code text rewrite) to disable calling of the
119 function profiler (mcount). This lets tracing be configured
120 in with practically no overhead in performance. This also
121 has a side effect of enabling or disabling specific functions
122 to be traced. Echoing names of functions into this file
123 will limit the trace to only those functions.
125 set_ftrace_notrace: This has an effect opposite to that of
126 set_ftrace_filter. Any function that is added here will not
127 be traced. If a function exists in both set_ftrace_filter
128 and set_ftrace_notrace, the function will _not_ be traced.
130 available_filter_functions: This lists the functions that ftrace
131 has processed and can trace. These are the function
132 names that you can pass to "set_ftrace_filter" or
133 "set_ftrace_notrace". (See the section "dynamic ftrace"
134 below for more details.)
140 Here is the list of current tracers that may be configured.
142 function - function tracer that uses mcount to trace all functions.
144 sched_switch - traces the context switches between tasks.
146 irqsoff - traces the areas that disable interrupts and saves
147 the trace with the longest max latency.
148 See tracing_max_latency. When a new max is recorded,
149 it replaces the old trace. It is best to view this
150 trace via the latency_trace file.
152 preemptoff - Similar to irqsoff but traces and records the amount of
153 time for which preemption is disabled.
155 preemptirqsoff - Similar to irqsoff and preemptoff, but traces and
156 records the largest time for which irqs and/or preemption
159 wakeup - Traces and records the max latency that it takes for
160 the highest priority task to get scheduled after
161 it has been woken up.
163 nop - This is not a tracer. To remove all tracers from tracing
164 simply echo "nop" into current_tracer.
167 Examples of using the tracer
168 ----------------------------
170 Here are typical examples of using the tracers when controlling them only
171 with the debugfs interface (without using any user-land utilities).
176 Here is an example of the output format of the file "trace"
181 # TASK-PID CPU# TIMESTAMP FUNCTION
183 bash-4251 [01] 10152.583854: path_put <-path_walk
184 bash-4251 [01] 10152.583855: dput <-path_put
185 bash-4251 [01] 10152.583855: _atomic_dec_and_lock <-dput
188 A header is printed with the tracer name that is represented by the trace.
189 In this case the tracer is "function". Then a header showing the format. Task
190 name "bash", the task PID "4251", the CPU that it was running on
191 "01", the timestamp in <secs>.<usecs> format, the function name that was
192 traced "path_put" and the parent function that called this function
193 "path_walk". The timestamp is the time at which the function was
196 The sched_switch tracer also includes tracing of task wakeups and
199 ksoftirqd/1-7 [01] 1453.070013: 7:115:R + 2916:115:S
200 ksoftirqd/1-7 [01] 1453.070013: 7:115:R + 10:115:S
201 ksoftirqd/1-7 [01] 1453.070013: 7:115:R ==> 10:115:R
202 events/1-10 [01] 1453.070013: 10:115:S ==> 2916:115:R
203 kondemand/1-2916 [01] 1453.070013: 2916:115:S ==> 7:115:R
204 ksoftirqd/1-7 [01] 1453.070013: 7:115:S ==> 0:140:R
206 Wake ups are represented by a "+" and the context switches are shown as
207 "==>". The format is:
211 Previous task Next Task
213 <pid>:<prio>:<state> ==> <pid>:<prio>:<state>
217 Current task Task waking up
219 <pid>:<prio>:<state> + <pid>:<prio>:<state>
221 The prio is the internal kernel priority, which is the inverse of the
222 priority that is usually displayed by user-space tools. Zero represents
223 the highest priority (99). Prio 100 starts the "nice" priorities with
224 100 being equal to nice -20 and 139 being nice 19. The prio "140" is
225 reserved for the idle task which is the lowest priority thread (pid 0).
231 For traces that display latency times, the latency_trace file gives
232 somewhat more information to see why a latency happened. Here is a typical
237 irqsoff latency trace v1.1.5 on 2.6.26-rc8
238 --------------------------------------------------------------------
239 latency: 97 us, #3/3, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
241 | task: swapper-0 (uid:0 nice:0 policy:0 rt_prio:0)
243 => started at: apic_timer_interrupt
244 => ended at: do_softirq
247 # / _-----=> irqs-off
248 # | / _----=> need-resched
249 # || / _---=> hardirq/softirq
250 # ||| / _--=> preempt-depth
253 # cmd pid ||||| time | caller
255 <idle>-0 0d..1 0us+: trace_hardirqs_off_thunk (apic_timer_interrupt)
256 <idle>-0 0d.s. 97us : __do_softirq (do_softirq)
257 <idle>-0 0d.s1 98us : trace_hardirqs_on (do_softirq)
261 This shows that the current tracer is "irqsoff" tracing the time for which
262 interrupts were disabled. It gives the trace version and the version
263 of the kernel upon which this was executed on (2.6.26-rc8). Then it displays
264 the max latency in microsecs (97 us). The number of trace entries displayed
265 and the total number recorded (both are three: #3/3). The type of
266 preemption that was used (PREEMPT). VP, KP, SP, and HP are always zero
267 and are reserved for later use. #P is the number of online CPUS (#P:2).
269 The task is the process that was running when the latency occurred.
272 The start and stop (the functions in which the interrupts were disabled and
273 enabled respectively) that caused the latencies:
275 apic_timer_interrupt is where the interrupts were disabled.
276 do_softirq is where they were enabled again.
278 The next lines after the header are the trace itself. The header
279 explains which is which.
281 cmd: The name of the process in the trace.
283 pid: The PID of that process.
285 CPU#: The CPU which the process was running on.
287 irqs-off: 'd' interrupts are disabled. '.' otherwise.
288 Note: If the architecture does not support a way to
289 read the irq flags variable, an 'X' will always
292 need-resched: 'N' task need_resched is set, '.' otherwise.
295 'H' - hard irq occurred inside a softirq.
296 'h' - hard irq is running
297 's' - soft irq is running
298 '.' - normal context.
300 preempt-depth: The level of preempt_disabled
302 The above is mostly meaningful for kernel developers.
304 time: This differs from the trace file output. The trace file output
305 includes an absolute timestamp. The timestamp used by the
306 latency_trace file is relative to the start of the trace.
308 delay: This is just to help catch your eye a bit better. And
309 needs to be fixed to be only relative to the same CPU.
310 The marks are determined by the difference between this
311 current trace and the next trace.
312 '!' - greater than preempt_mark_thresh (default 100)
313 '+' - greater than 1 microsecond
314 ' ' - less than or equal to 1 microsecond.
316 The rest is the same as the 'trace' file.
322 The trace_options file is used to control what gets printed in the trace
323 output. To see what is available, simply cat the file:
325 cat /debug/tracing/trace_options
326 print-parent nosym-offset nosym-addr noverbose noraw nohex nobin \
327 noblock nostacktrace nosched-tree nouserstacktrace nosym-userobj
329 To disable one of the options, echo in the option prepended with "no".
331 echo noprint-parent > /debug/tracing/trace_options
333 To enable an option, leave off the "no".
335 echo sym-offset > /debug/tracing/trace_options
337 Here are the available options:
339 print-parent - On function traces, display the calling function
340 as well as the function being traced.
343 bash-4000 [01] 1477.606694: simple_strtoul <-strict_strtoul
346 bash-4000 [01] 1477.606694: simple_strtoul
349 sym-offset - Display not only the function name, but also the offset
350 in the function. For example, instead of seeing just
351 "ktime_get", you will see "ktime_get+0xb/0x20".
354 bash-4000 [01] 1477.606694: simple_strtoul+0x6/0xa0
356 sym-addr - this will also display the function address as well as
360 bash-4000 [01] 1477.606694: simple_strtoul <c0339346>
362 verbose - This deals with the latency_trace file.
364 bash 4000 1 0 00000000 00010a95 [58127d26] 1720.415ms \
365 (+0.000ms): simple_strtoul (strict_strtoul)
367 raw - This will display raw numbers. This option is best for use with
368 user applications that can translate the raw numbers better than
369 having it done in the kernel.
371 hex - Similar to raw, but the numbers will be in a hexadecimal format.
373 bin - This will print out the formats in raw binary.
375 block - TBD (needs update)
377 stacktrace - This is one of the options that changes the trace itself.
378 When a trace is recorded, so is the stack of functions.
379 This allows for back traces of trace sites.
381 userstacktrace - This option changes the trace.
382 It records a stacktrace of the current userspace thread.
384 sym-userobj - when user stacktrace are enabled, look up which object the
385 address belongs to, and print a relative address
386 This is especially useful when ASLR is on, otherwise you don't
387 get a chance to resolve the address to object/file/line after the app is no
390 The lookup is performed when you read trace,trace_pipe,latency_trace. Example:
392 a.out-1623 [000] 40874.465068: /root/a.out[+0x480] <-/root/a.out[+0
393 x494] <- /root/a.out[+0x4a8] <- /lib/libc-2.7.so[+0x1e1a6]
395 sched-tree - TBD (any users??)
401 This tracer simply records schedule switches. Here is an example
404 # echo sched_switch > /debug/tracing/current_tracer
405 # echo 1 > /debug/tracing/tracing_enabled
407 # echo 0 > /debug/tracing/tracing_enabled
408 # cat /debug/tracing/trace
410 # tracer: sched_switch
412 # TASK-PID CPU# TIMESTAMP FUNCTION
414 bash-3997 [01] 240.132281: 3997:120:R + 4055:120:R
415 bash-3997 [01] 240.132284: 3997:120:R ==> 4055:120:R
416 sleep-4055 [01] 240.132371: 4055:120:S ==> 3997:120:R
417 bash-3997 [01] 240.132454: 3997:120:R + 4055:120:S
418 bash-3997 [01] 240.132457: 3997:120:R ==> 4055:120:R
419 sleep-4055 [01] 240.132460: 4055:120:D ==> 3997:120:R
420 bash-3997 [01] 240.132463: 3997:120:R + 4055:120:D
421 bash-3997 [01] 240.132465: 3997:120:R ==> 4055:120:R
422 <idle>-0 [00] 240.132589: 0:140:R + 4:115:S
423 <idle>-0 [00] 240.132591: 0:140:R ==> 4:115:R
424 ksoftirqd/0-4 [00] 240.132595: 4:115:S ==> 0:140:R
425 <idle>-0 [00] 240.132598: 0:140:R + 4:115:S
426 <idle>-0 [00] 240.132599: 0:140:R ==> 4:115:R
427 ksoftirqd/0-4 [00] 240.132603: 4:115:S ==> 0:140:R
428 sleep-4055 [01] 240.133058: 4055:120:S ==> 3997:120:R
432 As we have discussed previously about this format, the header shows
433 the name of the trace and points to the options. The "FUNCTION"
434 is a misnomer since here it represents the wake ups and context
437 The sched_switch file only lists the wake ups (represented with '+')
438 and context switches ('==>') with the previous task or current task
439 first followed by the next task or task waking up. The format for both
440 of these is PID:KERNEL-PRIO:TASK-STATE. Remember that the KERNEL-PRIO
441 is the inverse of the actual priority with zero (0) being the highest
442 priority and the nice values starting at 100 (nice -20). Below is
443 a quick chart to map the kernel priority to user land priorities.
445 Kernel priority: 0 to 99 ==> user RT priority 99 to 0
446 Kernel priority: 100 to 139 ==> user nice -20 to 19
447 Kernel priority: 140 ==> idle task priority
451 R - running : wants to run, may not actually be running
452 S - sleep : process is waiting to be woken up (handles signals)
453 D - disk sleep (uninterruptible sleep) : process must be woken up
455 T - stopped : process suspended
456 t - traced : process is being traced (with something like gdb)
457 Z - zombie : process waiting to be cleaned up
464 The following tracers (listed below) give different output depending
465 on whether or not the sysctl ftrace_enabled is set. To set ftrace_enabled,
466 one can either use the sysctl function or set it via the proc
467 file system interface.
469 sysctl kernel.ftrace_enabled=1
473 echo 1 > /proc/sys/kernel/ftrace_enabled
475 To disable ftrace_enabled simply replace the '1' with '0' in
478 When ftrace_enabled is set the tracers will also record the functions
479 that are within the trace. The descriptions of the tracers
480 will also show an example with ftrace enabled.
486 When interrupts are disabled, the CPU can not react to any other
487 external event (besides NMIs and SMIs). This prevents the timer
488 interrupt from triggering or the mouse interrupt from letting the
489 kernel know of a new mouse event. The result is a latency with the
492 The irqsoff tracer tracks the time for which interrupts are disabled.
493 When a new maximum latency is hit, the tracer saves the trace leading up
494 to that latency point so that every time a new maximum is reached, the old
495 saved trace is discarded and the new trace is saved.
497 To reset the maximum, echo 0 into tracing_max_latency. Here is an
500 # echo irqsoff > /debug/tracing/current_tracer
501 # echo 0 > /debug/tracing/tracing_max_latency
502 # echo 1 > /debug/tracing/tracing_enabled
505 # echo 0 > /debug/tracing/tracing_enabled
506 # cat /debug/tracing/latency_trace
509 irqsoff latency trace v1.1.5 on 2.6.26
510 --------------------------------------------------------------------
511 latency: 12 us, #3/3, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
513 | task: bash-3730 (uid:0 nice:0 policy:0 rt_prio:0)
515 => started at: sys_setpgid
516 => ended at: sys_setpgid
519 # / _-----=> irqs-off
520 # | / _----=> need-resched
521 # || / _---=> hardirq/softirq
522 # ||| / _--=> preempt-depth
525 # cmd pid ||||| time | caller
527 bash-3730 1d... 0us : _write_lock_irq (sys_setpgid)
528 bash-3730 1d..1 1us+: _write_unlock_irq (sys_setpgid)
529 bash-3730 1d..2 14us : trace_hardirqs_on (sys_setpgid)
532 Here we see that that we had a latency of 12 microsecs (which is
533 very good). The _write_lock_irq in sys_setpgid disabled interrupts.
534 The difference between the 12 and the displayed timestamp 14us occurred
535 because the clock was incremented between the time of recording the max
536 latency and the time of recording the function that had that latency.
538 Note the above example had ftrace_enabled not set. If we set the
539 ftrace_enabled, we get a much larger output:
543 irqsoff latency trace v1.1.5 on 2.6.26-rc8
544 --------------------------------------------------------------------
545 latency: 50 us, #101/101, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
547 | task: ls-4339 (uid:0 nice:0 policy:0 rt_prio:0)
549 => started at: __alloc_pages_internal
550 => ended at: __alloc_pages_internal
553 # / _-----=> irqs-off
554 # | / _----=> need-resched
555 # || / _---=> hardirq/softirq
556 # ||| / _--=> preempt-depth
559 # cmd pid ||||| time | caller
561 ls-4339 0...1 0us+: get_page_from_freelist (__alloc_pages_internal)
562 ls-4339 0d..1 3us : rmqueue_bulk (get_page_from_freelist)
563 ls-4339 0d..1 3us : _spin_lock (rmqueue_bulk)
564 ls-4339 0d..1 4us : add_preempt_count (_spin_lock)
565 ls-4339 0d..2 4us : __rmqueue (rmqueue_bulk)
566 ls-4339 0d..2 5us : __rmqueue_smallest (__rmqueue)
567 ls-4339 0d..2 5us : __mod_zone_page_state (__rmqueue_smallest)
568 ls-4339 0d..2 6us : __rmqueue (rmqueue_bulk)
569 ls-4339 0d..2 6us : __rmqueue_smallest (__rmqueue)
570 ls-4339 0d..2 7us : __mod_zone_page_state (__rmqueue_smallest)
571 ls-4339 0d..2 7us : __rmqueue (rmqueue_bulk)
572 ls-4339 0d..2 8us : __rmqueue_smallest (__rmqueue)
574 ls-4339 0d..2 46us : __rmqueue_smallest (__rmqueue)
575 ls-4339 0d..2 47us : __mod_zone_page_state (__rmqueue_smallest)
576 ls-4339 0d..2 47us : __rmqueue (rmqueue_bulk)
577 ls-4339 0d..2 48us : __rmqueue_smallest (__rmqueue)
578 ls-4339 0d..2 48us : __mod_zone_page_state (__rmqueue_smallest)
579 ls-4339 0d..2 49us : _spin_unlock (rmqueue_bulk)
580 ls-4339 0d..2 49us : sub_preempt_count (_spin_unlock)
581 ls-4339 0d..1 50us : get_page_from_freelist (__alloc_pages_internal)
582 ls-4339 0d..2 51us : trace_hardirqs_on (__alloc_pages_internal)
586 Here we traced a 50 microsecond latency. But we also see all the
587 functions that were called during that time. Note that by enabling
588 function tracing, we incur an added overhead. This overhead may
589 extend the latency times. But nevertheless, this trace has provided
590 some very helpful debugging information.
596 When preemption is disabled, we may be able to receive interrupts but
597 the task cannot be preempted and a higher priority task must wait
598 for preemption to be enabled again before it can preempt a lower
601 The preemptoff tracer traces the places that disable preemption.
602 Like the irqsoff tracer, it records the maximum latency for which preemption
603 was disabled. The control of preemptoff tracer is much like the irqsoff
606 # echo preemptoff > /debug/tracing/current_tracer
607 # echo 0 > /debug/tracing/tracing_max_latency
608 # echo 1 > /debug/tracing/tracing_enabled
611 # echo 0 > /debug/tracing/tracing_enabled
612 # cat /debug/tracing/latency_trace
615 preemptoff latency trace v1.1.5 on 2.6.26-rc8
616 --------------------------------------------------------------------
617 latency: 29 us, #3/3, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
619 | task: sshd-4261 (uid:0 nice:0 policy:0 rt_prio:0)
621 => started at: do_IRQ
622 => ended at: __do_softirq
625 # / _-----=> irqs-off
626 # | / _----=> need-resched
627 # || / _---=> hardirq/softirq
628 # ||| / _--=> preempt-depth
631 # cmd pid ||||| time | caller
633 sshd-4261 0d.h. 0us+: irq_enter (do_IRQ)
634 sshd-4261 0d.s. 29us : _local_bh_enable (__do_softirq)
635 sshd-4261 0d.s1 30us : trace_preempt_on (__do_softirq)
638 This has some more changes. Preemption was disabled when an interrupt
639 came in (notice the 'h'), and was enabled while doing a softirq.
640 (notice the 's'). But we also see that interrupts have been disabled
641 when entering the preempt off section and leaving it (the 'd').
642 We do not know if interrupts were enabled in the mean time.
646 preemptoff latency trace v1.1.5 on 2.6.26-rc8
647 --------------------------------------------------------------------
648 latency: 63 us, #87/87, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
650 | task: sshd-4261 (uid:0 nice:0 policy:0 rt_prio:0)
652 => started at: remove_wait_queue
653 => ended at: __do_softirq
656 # / _-----=> irqs-off
657 # | / _----=> need-resched
658 # || / _---=> hardirq/softirq
659 # ||| / _--=> preempt-depth
662 # cmd pid ||||| time | caller
664 sshd-4261 0d..1 0us : _spin_lock_irqsave (remove_wait_queue)
665 sshd-4261 0d..1 1us : _spin_unlock_irqrestore (remove_wait_queue)
666 sshd-4261 0d..1 2us : do_IRQ (common_interrupt)
667 sshd-4261 0d..1 2us : irq_enter (do_IRQ)
668 sshd-4261 0d..1 2us : idle_cpu (irq_enter)
669 sshd-4261 0d..1 3us : add_preempt_count (irq_enter)
670 sshd-4261 0d.h1 3us : idle_cpu (irq_enter)
671 sshd-4261 0d.h. 4us : handle_fasteoi_irq (do_IRQ)
673 sshd-4261 0d.h. 12us : add_preempt_count (_spin_lock)
674 sshd-4261 0d.h1 12us : ack_ioapic_quirk_irq (handle_fasteoi_irq)
675 sshd-4261 0d.h1 13us : move_native_irq (ack_ioapic_quirk_irq)
676 sshd-4261 0d.h1 13us : _spin_unlock (handle_fasteoi_irq)
677 sshd-4261 0d.h1 14us : sub_preempt_count (_spin_unlock)
678 sshd-4261 0d.h1 14us : irq_exit (do_IRQ)
679 sshd-4261 0d.h1 15us : sub_preempt_count (irq_exit)
680 sshd-4261 0d..2 15us : do_softirq (irq_exit)
681 sshd-4261 0d... 15us : __do_softirq (do_softirq)
682 sshd-4261 0d... 16us : __local_bh_disable (__do_softirq)
683 sshd-4261 0d... 16us+: add_preempt_count (__local_bh_disable)
684 sshd-4261 0d.s4 20us : add_preempt_count (__local_bh_disable)
685 sshd-4261 0d.s4 21us : sub_preempt_count (local_bh_enable)
686 sshd-4261 0d.s5 21us : sub_preempt_count (local_bh_enable)
688 sshd-4261 0d.s6 41us : add_preempt_count (__local_bh_disable)
689 sshd-4261 0d.s6 42us : sub_preempt_count (local_bh_enable)
690 sshd-4261 0d.s7 42us : sub_preempt_count (local_bh_enable)
691 sshd-4261 0d.s5 43us : add_preempt_count (__local_bh_disable)
692 sshd-4261 0d.s5 43us : sub_preempt_count (local_bh_enable_ip)
693 sshd-4261 0d.s6 44us : sub_preempt_count (local_bh_enable_ip)
694 sshd-4261 0d.s5 44us : add_preempt_count (__local_bh_disable)
695 sshd-4261 0d.s5 45us : sub_preempt_count (local_bh_enable)
697 sshd-4261 0d.s. 63us : _local_bh_enable (__do_softirq)
698 sshd-4261 0d.s1 64us : trace_preempt_on (__do_softirq)
701 The above is an example of the preemptoff trace with ftrace_enabled
702 set. Here we see that interrupts were disabled the entire time.
703 The irq_enter code lets us know that we entered an interrupt 'h'.
704 Before that, the functions being traced still show that it is not
705 in an interrupt, but we can see from the functions themselves that
706 this is not the case.
708 Notice that __do_softirq when called does not have a preempt_count.
709 It may seem that we missed a preempt enabling. What really happened
710 is that the preempt count is held on the thread's stack and we
711 switched to the softirq stack (4K stacks in effect). The code
712 does not copy the preempt count, but because interrupts are disabled,
713 we do not need to worry about it. Having a tracer like this is good
714 for letting people know what really happens inside the kernel.
720 Knowing the locations that have interrupts disabled or preemption
721 disabled for the longest times is helpful. But sometimes we would
722 like to know when either preemption and/or interrupts are disabled.
724 Consider the following code:
727 call_function_with_irqs_off();
729 call_function_with_irqs_and_preemption_off();
731 call_function_with_preemption_off();
734 The irqsoff tracer will record the total length of
735 call_function_with_irqs_off() and
736 call_function_with_irqs_and_preemption_off().
738 The preemptoff tracer will record the total length of
739 call_function_with_irqs_and_preemption_off() and
740 call_function_with_preemption_off().
742 But neither will trace the time that interrupts and/or preemption
743 is disabled. This total time is the time that we can not schedule.
744 To record this time, use the preemptirqsoff tracer.
746 Again, using this trace is much like the irqsoff and preemptoff tracers.
748 # echo preemptirqsoff > /debug/tracing/current_tracer
749 # echo 0 > /debug/tracing/tracing_max_latency
750 # echo 1 > /debug/tracing/tracing_enabled
753 # echo 0 > /debug/tracing/tracing_enabled
754 # cat /debug/tracing/latency_trace
755 # tracer: preemptirqsoff
757 preemptirqsoff latency trace v1.1.5 on 2.6.26-rc8
758 --------------------------------------------------------------------
759 latency: 293 us, #3/3, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
761 | task: ls-4860 (uid:0 nice:0 policy:0 rt_prio:0)
763 => started at: apic_timer_interrupt
764 => ended at: __do_softirq
767 # / _-----=> irqs-off
768 # | / _----=> need-resched
769 # || / _---=> hardirq/softirq
770 # ||| / _--=> preempt-depth
773 # cmd pid ||||| time | caller
775 ls-4860 0d... 0us!: trace_hardirqs_off_thunk (apic_timer_interrupt)
776 ls-4860 0d.s. 294us : _local_bh_enable (__do_softirq)
777 ls-4860 0d.s1 294us : trace_preempt_on (__do_softirq)
781 The trace_hardirqs_off_thunk is called from assembly on x86 when
782 interrupts are disabled in the assembly code. Without the function
783 tracing, we do not know if interrupts were enabled within the preemption
784 points. We do see that it started with preemption enabled.
786 Here is a trace with ftrace_enabled set:
789 # tracer: preemptirqsoff
791 preemptirqsoff latency trace v1.1.5 on 2.6.26-rc8
792 --------------------------------------------------------------------
793 latency: 105 us, #183/183, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
795 | task: sshd-4261 (uid:0 nice:0 policy:0 rt_prio:0)
797 => started at: write_chan
798 => ended at: __do_softirq
801 # / _-----=> irqs-off
802 # | / _----=> need-resched
803 # || / _---=> hardirq/softirq
804 # ||| / _--=> preempt-depth
807 # cmd pid ||||| time | caller
809 ls-4473 0.N.. 0us : preempt_schedule (write_chan)
810 ls-4473 0dN.1 1us : _spin_lock (schedule)
811 ls-4473 0dN.1 2us : add_preempt_count (_spin_lock)
812 ls-4473 0d..2 2us : put_prev_task_fair (schedule)
814 ls-4473 0d..2 13us : set_normalized_timespec (ktime_get_ts)
815 ls-4473 0d..2 13us : __switch_to (schedule)
816 sshd-4261 0d..2 14us : finish_task_switch (schedule)
817 sshd-4261 0d..2 14us : _spin_unlock_irq (finish_task_switch)
818 sshd-4261 0d..1 15us : add_preempt_count (_spin_lock_irqsave)
819 sshd-4261 0d..2 16us : _spin_unlock_irqrestore (hrtick_set)
820 sshd-4261 0d..2 16us : do_IRQ (common_interrupt)
821 sshd-4261 0d..2 17us : irq_enter (do_IRQ)
822 sshd-4261 0d..2 17us : idle_cpu (irq_enter)
823 sshd-4261 0d..2 18us : add_preempt_count (irq_enter)
824 sshd-4261 0d.h2 18us : idle_cpu (irq_enter)
825 sshd-4261 0d.h. 18us : handle_fasteoi_irq (do_IRQ)
826 sshd-4261 0d.h. 19us : _spin_lock (handle_fasteoi_irq)
827 sshd-4261 0d.h. 19us : add_preempt_count (_spin_lock)
828 sshd-4261 0d.h1 20us : _spin_unlock (handle_fasteoi_irq)
829 sshd-4261 0d.h1 20us : sub_preempt_count (_spin_unlock)
831 sshd-4261 0d.h1 28us : _spin_unlock (handle_fasteoi_irq)
832 sshd-4261 0d.h1 29us : sub_preempt_count (_spin_unlock)
833 sshd-4261 0d.h2 29us : irq_exit (do_IRQ)
834 sshd-4261 0d.h2 29us : sub_preempt_count (irq_exit)
835 sshd-4261 0d..3 30us : do_softirq (irq_exit)
836 sshd-4261 0d... 30us : __do_softirq (do_softirq)
837 sshd-4261 0d... 31us : __local_bh_disable (__do_softirq)
838 sshd-4261 0d... 31us+: add_preempt_count (__local_bh_disable)
839 sshd-4261 0d.s4 34us : add_preempt_count (__local_bh_disable)
841 sshd-4261 0d.s3 43us : sub_preempt_count (local_bh_enable_ip)
842 sshd-4261 0d.s4 44us : sub_preempt_count (local_bh_enable_ip)
843 sshd-4261 0d.s3 44us : smp_apic_timer_interrupt (apic_timer_interrupt)
844 sshd-4261 0d.s3 45us : irq_enter (smp_apic_timer_interrupt)
845 sshd-4261 0d.s3 45us : idle_cpu (irq_enter)
846 sshd-4261 0d.s3 46us : add_preempt_count (irq_enter)
847 sshd-4261 0d.H3 46us : idle_cpu (irq_enter)
848 sshd-4261 0d.H3 47us : hrtimer_interrupt (smp_apic_timer_interrupt)
849 sshd-4261 0d.H3 47us : ktime_get (hrtimer_interrupt)
851 sshd-4261 0d.H3 81us : tick_program_event (hrtimer_interrupt)
852 sshd-4261 0d.H3 82us : ktime_get (tick_program_event)
853 sshd-4261 0d.H3 82us : ktime_get_ts (ktime_get)
854 sshd-4261 0d.H3 83us : getnstimeofday (ktime_get_ts)
855 sshd-4261 0d.H3 83us : set_normalized_timespec (ktime_get_ts)
856 sshd-4261 0d.H3 84us : clockevents_program_event (tick_program_event)
857 sshd-4261 0d.H3 84us : lapic_next_event (clockevents_program_event)
858 sshd-4261 0d.H3 85us : irq_exit (smp_apic_timer_interrupt)
859 sshd-4261 0d.H3 85us : sub_preempt_count (irq_exit)
860 sshd-4261 0d.s4 86us : sub_preempt_count (irq_exit)
861 sshd-4261 0d.s3 86us : add_preempt_count (__local_bh_disable)
863 sshd-4261 0d.s1 98us : sub_preempt_count (net_rx_action)
864 sshd-4261 0d.s. 99us : add_preempt_count (_spin_lock_irq)
865 sshd-4261 0d.s1 99us+: _spin_unlock_irq (run_timer_softirq)
866 sshd-4261 0d.s. 104us : _local_bh_enable (__do_softirq)
867 sshd-4261 0d.s. 104us : sub_preempt_count (_local_bh_enable)
868 sshd-4261 0d.s. 105us : _local_bh_enable (__do_softirq)
869 sshd-4261 0d.s1 105us : trace_preempt_on (__do_softirq)
872 This is a very interesting trace. It started with the preemption of
873 the ls task. We see that the task had the "need_resched" bit set
874 via the 'N' in the trace. Interrupts were disabled before the spin_lock
875 at the beginning of the trace. We see that a schedule took place to run
876 sshd. When the interrupts were enabled, we took an interrupt.
877 On return from the interrupt handler, the softirq ran. We took another
878 interrupt while running the softirq as we see from the capital 'H'.
884 In a Real-Time environment it is very important to know the wakeup
885 time it takes for the highest priority task that is woken up to the
886 time that it executes. This is also known as "schedule latency".
887 I stress the point that this is about RT tasks. It is also important
888 to know the scheduling latency of non-RT tasks, but the average
889 schedule latency is better for non-RT tasks. Tools like
890 LatencyTop are more appropriate for such measurements.
892 Real-Time environments are interested in the worst case latency.
893 That is the longest latency it takes for something to happen, and
894 not the average. We can have a very fast scheduler that may only
895 have a large latency once in a while, but that would not work well
896 with Real-Time tasks. The wakeup tracer was designed to record
897 the worst case wakeups of RT tasks. Non-RT tasks are not recorded
898 because the tracer only records one worst case and tracing non-RT
899 tasks that are unpredictable will overwrite the worst case latency
902 Since this tracer only deals with RT tasks, we will run this slightly
903 differently than we did with the previous tracers. Instead of performing
904 an 'ls', we will run 'sleep 1' under 'chrt' which changes the
905 priority of the task.
907 # echo wakeup > /debug/tracing/current_tracer
908 # echo 0 > /debug/tracing/tracing_max_latency
909 # echo 1 > /debug/tracing/tracing_enabled
911 # echo 0 > /debug/tracing/tracing_enabled
912 # cat /debug/tracing/latency_trace
915 wakeup latency trace v1.1.5 on 2.6.26-rc8
916 --------------------------------------------------------------------
917 latency: 4 us, #2/2, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
919 | task: sleep-4901 (uid:0 nice:0 policy:1 rt_prio:5)
923 # / _-----=> irqs-off
924 # | / _----=> need-resched
925 # || / _---=> hardirq/softirq
926 # ||| / _--=> preempt-depth
929 # cmd pid ||||| time | caller
931 <idle>-0 1d.h4 0us+: try_to_wake_up (wake_up_process)
932 <idle>-0 1d..4 4us : schedule (cpu_idle)
936 Running this on an idle system, we see that it only took 4 microseconds
937 to perform the task switch. Note, since the trace marker in the
938 schedule is before the actual "switch", we stop the tracing when
939 the recorded task is about to schedule in. This may change if
940 we add a new marker at the end of the scheduler.
942 Notice that the recorded task is 'sleep' with the PID of 4901 and it
943 has an rt_prio of 5. This priority is user-space priority and not
944 the internal kernel priority. The policy is 1 for SCHED_FIFO and 2
947 Doing the same with chrt -r 5 and ftrace_enabled set.
951 wakeup latency trace v1.1.5 on 2.6.26-rc8
952 --------------------------------------------------------------------
953 latency: 50 us, #60/60, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
955 | task: sleep-4068 (uid:0 nice:0 policy:2 rt_prio:5)
959 # / _-----=> irqs-off
960 # | / _----=> need-resched
961 # || / _---=> hardirq/softirq
962 # ||| / _--=> preempt-depth
965 # cmd pid ||||| time | caller
967 ksoftirq-7 1d.H3 0us : try_to_wake_up (wake_up_process)
968 ksoftirq-7 1d.H4 1us : sub_preempt_count (marker_probe_cb)
969 ksoftirq-7 1d.H3 2us : check_preempt_wakeup (try_to_wake_up)
970 ksoftirq-7 1d.H3 3us : update_curr (check_preempt_wakeup)
971 ksoftirq-7 1d.H3 4us : calc_delta_mine (update_curr)
972 ksoftirq-7 1d.H3 5us : __resched_task (check_preempt_wakeup)
973 ksoftirq-7 1d.H3 6us : task_wake_up_rt (try_to_wake_up)
974 ksoftirq-7 1d.H3 7us : _spin_unlock_irqrestore (try_to_wake_up)
976 ksoftirq-7 1d.H2 17us : irq_exit (smp_apic_timer_interrupt)
977 ksoftirq-7 1d.H2 18us : sub_preempt_count (irq_exit)
978 ksoftirq-7 1d.s3 19us : sub_preempt_count (irq_exit)
979 ksoftirq-7 1..s2 20us : rcu_process_callbacks (__do_softirq)
981 ksoftirq-7 1..s2 26us : __rcu_process_callbacks (rcu_process_callbacks)
982 ksoftirq-7 1d.s2 27us : _local_bh_enable (__do_softirq)
983 ksoftirq-7 1d.s2 28us : sub_preempt_count (_local_bh_enable)
984 ksoftirq-7 1.N.3 29us : sub_preempt_count (ksoftirqd)
985 ksoftirq-7 1.N.2 30us : _cond_resched (ksoftirqd)
986 ksoftirq-7 1.N.2 31us : __cond_resched (_cond_resched)
987 ksoftirq-7 1.N.2 32us : add_preempt_count (__cond_resched)
988 ksoftirq-7 1.N.2 33us : schedule (__cond_resched)
989 ksoftirq-7 1.N.2 33us : add_preempt_count (schedule)
990 ksoftirq-7 1.N.3 34us : hrtick_clear (schedule)
991 ksoftirq-7 1dN.3 35us : _spin_lock (schedule)
992 ksoftirq-7 1dN.3 36us : add_preempt_count (_spin_lock)
993 ksoftirq-7 1d..4 37us : put_prev_task_fair (schedule)
994 ksoftirq-7 1d..4 38us : update_curr (put_prev_task_fair)
996 ksoftirq-7 1d..5 47us : _spin_trylock (tracing_record_cmdline)
997 ksoftirq-7 1d..5 48us : add_preempt_count (_spin_trylock)
998 ksoftirq-7 1d..6 49us : _spin_unlock (tracing_record_cmdline)
999 ksoftirq-7 1d..6 49us : sub_preempt_count (_spin_unlock)
1000 ksoftirq-7 1d..4 50us : schedule (__cond_resched)
1002 The interrupt went off while running ksoftirqd. This task runs at
1003 SCHED_OTHER. Why did not we see the 'N' set early? This may be
1004 a harmless bug with x86_32 and 4K stacks. On x86_32 with 4K stacks
1005 configured, the interrupt and softirq run with their own stack.
1006 Some information is held on the top of the task's stack (need_resched
1007 and preempt_count are both stored there). The setting of the NEED_RESCHED
1008 bit is done directly to the task's stack, but the reading of the
1009 NEED_RESCHED is done by looking at the current stack, which in this case
1010 is the stack for the hard interrupt. This hides the fact that NEED_RESCHED
1011 has been set. We do not see the 'N' until we switch back to the task's
1017 This tracer is the function tracer. Enabling the function tracer
1018 can be done from the debug file system. Make sure the ftrace_enabled is
1019 set; otherwise this tracer is a nop.
1021 # sysctl kernel.ftrace_enabled=1
1022 # echo function > /debug/tracing/current_tracer
1023 # echo 1 > /debug/tracing/tracing_enabled
1025 # echo 0 > /debug/tracing/tracing_enabled
1026 # cat /debug/tracing/trace
1029 # TASK-PID CPU# TIMESTAMP FUNCTION
1031 bash-4003 [00] 123.638713: finish_task_switch <-schedule
1032 bash-4003 [00] 123.638714: _spin_unlock_irq <-finish_task_switch
1033 bash-4003 [00] 123.638714: sub_preempt_count <-_spin_unlock_irq
1034 bash-4003 [00] 123.638715: hrtick_set <-schedule
1035 bash-4003 [00] 123.638715: _spin_lock_irqsave <-hrtick_set
1036 bash-4003 [00] 123.638716: add_preempt_count <-_spin_lock_irqsave
1037 bash-4003 [00] 123.638716: _spin_unlock_irqrestore <-hrtick_set
1038 bash-4003 [00] 123.638717: sub_preempt_count <-_spin_unlock_irqrestore
1039 bash-4003 [00] 123.638717: hrtick_clear <-hrtick_set
1040 bash-4003 [00] 123.638718: sub_preempt_count <-schedule
1041 bash-4003 [00] 123.638718: sub_preempt_count <-preempt_schedule
1042 bash-4003 [00] 123.638719: wait_for_completion <-__stop_machine_run
1043 bash-4003 [00] 123.638719: wait_for_common <-wait_for_completion
1044 bash-4003 [00] 123.638720: _spin_lock_irq <-wait_for_common
1045 bash-4003 [00] 123.638720: add_preempt_count <-_spin_lock_irq
1049 Note: function tracer uses ring buffers to store the above entries.
1050 The newest data may overwrite the oldest data. Sometimes using echo to
1051 stop the trace is not sufficient because the tracing could have overwritten
1052 the data that you wanted to record. For this reason, it is sometimes better to
1053 disable tracing directly from a program. This allows you to stop the
1054 tracing at the point that you hit the part that you are interested in.
1055 To disable the tracing directly from a C program, something like following
1056 code snippet can be used:
1060 int main(int argc, char *argv[]) {
1062 trace_fd = open("/debug/tracing/tracing_enabled", O_WRONLY);
1064 if (condition_hit()) {
1065 write(trace_fd, "0", 1);
1070 Note: Here we hard coded the path name. The debugfs mount is not
1071 guaranteed to be at /debug (and is more commonly at /sys/kernel/debug).
1072 For simple one time traces, the above is sufficent. For anything else,
1073 a search through /proc/mounts may be needed to find where the debugfs
1074 file-system is mounted.
1079 If CONFIG_DYNAMIC_FTRACE is set, the system will run with
1080 virtually no overhead when function tracing is disabled. The way
1081 this works is the mcount function call (placed at the start of
1082 every kernel function, produced by the -pg switch in gcc), starts
1083 of pointing to a simple return. (Enabling FTRACE will include the
1084 -pg switch in the compiling of the kernel.)
1086 At compile time every C file object is run through the
1087 recordmcount.pl script (located in the scripts directory). This
1088 script will process the C object using objdump to find all the
1089 locations in the .text section that call mcount. (Note, only
1090 the .text section is processed, since processing other sections
1091 like .init.text may cause races due to those sections being freed).
1093 A new section called "__mcount_loc" is created that holds references
1094 to all the mcount call sites in the .text section. This section is
1095 compiled back into the original object. The final linker will add
1096 all these references into a single table.
1098 On boot up, before SMP is initialized, the dynamic ftrace code
1099 scans this table and updates all the locations into nops. It also
1100 records the locations, which are added to the available_filter_functions
1101 list. Modules are processed as they are loaded and before they are
1102 executed. When a module is unloaded, it also removes its functions from
1103 the ftrace function list. This is automatic in the module unload
1104 code, and the module author does not need to worry about it.
1106 When tracing is enabled, kstop_machine is called to prevent races
1107 with the CPUS executing code being modified (which can cause the
1108 CPU to do undesireable things), and the nops are patched back
1109 to calls. But this time, they do not call mcount (which is just
1110 a function stub). They now call into the ftrace infrastructure.
1112 One special side-effect to the recording of the functions being
1113 traced is that we can now selectively choose which functions we
1114 wish to trace and which ones we want the mcount calls to remain as
1117 Two files are used, one for enabling and one for disabling the tracing
1118 of specified functions. They are:
1126 A list of available functions that you can add to these files is listed
1129 available_filter_functions
1131 # cat /debug/tracing/available_filter_functions
1140 If I am only interested in sys_nanosleep and hrtimer_interrupt:
1142 # echo sys_nanosleep hrtimer_interrupt \
1143 > /debug/tracing/set_ftrace_filter
1144 # echo ftrace > /debug/tracing/current_tracer
1145 # echo 1 > /debug/tracing/tracing_enabled
1147 # echo 0 > /debug/tracing/tracing_enabled
1148 # cat /debug/tracing/trace
1151 # TASK-PID CPU# TIMESTAMP FUNCTION
1153 usleep-4134 [00] 1317.070017: hrtimer_interrupt <-smp_apic_timer_interrupt
1154 usleep-4134 [00] 1317.070111: sys_nanosleep <-syscall_call
1155 <idle>-0 [00] 1317.070115: hrtimer_interrupt <-smp_apic_timer_interrupt
1157 To see which functions are being traced, you can cat the file:
1159 # cat /debug/tracing/set_ftrace_filter
1164 Perhaps this is not enough. The filters also allow simple wild cards.
1165 Only the following are currently available
1167 <match>* - will match functions that begin with <match>
1168 *<match> - will match functions that end with <match>
1169 *<match>* - will match functions that have <match> in it
1171 These are the only wild cards which are supported.
1173 <match>*<match> will not work.
1175 # echo hrtimer_* > /debug/tracing/set_ftrace_filter
1181 # TASK-PID CPU# TIMESTAMP FUNCTION
1183 bash-4003 [00] 1480.611794: hrtimer_init <-copy_process
1184 bash-4003 [00] 1480.611941: hrtimer_start <-hrtick_set
1185 bash-4003 [00] 1480.611956: hrtimer_cancel <-hrtick_clear
1186 bash-4003 [00] 1480.611956: hrtimer_try_to_cancel <-hrtimer_cancel
1187 <idle>-0 [00] 1480.612019: hrtimer_get_next_event <-get_next_timer_interrupt
1188 <idle>-0 [00] 1480.612025: hrtimer_get_next_event <-get_next_timer_interrupt
1189 <idle>-0 [00] 1480.612032: hrtimer_get_next_event <-get_next_timer_interrupt
1190 <idle>-0 [00] 1480.612037: hrtimer_get_next_event <-get_next_timer_interrupt
1191 <idle>-0 [00] 1480.612382: hrtimer_get_next_event <-get_next_timer_interrupt
1194 Notice that we lost the sys_nanosleep.
1196 # cat /debug/tracing/set_ftrace_filter
1201 hrtimer_try_to_cancel
1205 hrtimer_force_reprogram
1206 hrtimer_get_next_event
1210 hrtimer_get_remaining
1212 hrtimer_init_sleeper
1215 This is because the '>' and '>>' act just like they do in bash.
1216 To rewrite the filters, use '>'
1217 To append to the filters, use '>>'
1219 To clear out a filter so that all functions will be recorded again:
1221 # echo > /debug/tracing/set_ftrace_filter
1222 # cat /debug/tracing/set_ftrace_filter
1225 Again, now we want to append.
1227 # echo sys_nanosleep > /debug/tracing/set_ftrace_filter
1228 # cat /debug/tracing/set_ftrace_filter
1230 # echo hrtimer_* >> /debug/tracing/set_ftrace_filter
1231 # cat /debug/tracing/set_ftrace_filter
1236 hrtimer_try_to_cancel
1240 hrtimer_force_reprogram
1241 hrtimer_get_next_event
1246 hrtimer_get_remaining
1248 hrtimer_init_sleeper
1251 The set_ftrace_notrace prevents those functions from being traced.
1253 # echo '*preempt*' '*lock*' > /debug/tracing/set_ftrace_notrace
1259 # TASK-PID CPU# TIMESTAMP FUNCTION
1261 bash-4043 [01] 115.281644: finish_task_switch <-schedule
1262 bash-4043 [01] 115.281645: hrtick_set <-schedule
1263 bash-4043 [01] 115.281645: hrtick_clear <-hrtick_set
1264 bash-4043 [01] 115.281646: wait_for_completion <-__stop_machine_run
1265 bash-4043 [01] 115.281647: wait_for_common <-wait_for_completion
1266 bash-4043 [01] 115.281647: kthread_stop <-stop_machine_run
1267 bash-4043 [01] 115.281648: init_waitqueue_head <-kthread_stop
1268 bash-4043 [01] 115.281648: wake_up_process <-kthread_stop
1269 bash-4043 [01] 115.281649: try_to_wake_up <-wake_up_process
1271 We can see that there's no more lock or preempt tracing.
1276 The trace_pipe outputs the same content as the trace file, but the effect
1277 on the tracing is different. Every read from trace_pipe is consumed.
1278 This means that subsequent reads will be different. The trace
1281 # echo function > /debug/tracing/current_tracer
1282 # cat /debug/tracing/trace_pipe > /tmp/trace.out &
1284 # echo 1 > /debug/tracing/tracing_enabled
1286 # echo 0 > /debug/tracing/tracing_enabled
1287 # cat /debug/tracing/trace
1290 # TASK-PID CPU# TIMESTAMP FUNCTION
1294 # cat /tmp/trace.out
1295 bash-4043 [00] 41.267106: finish_task_switch <-schedule
1296 bash-4043 [00] 41.267106: hrtick_set <-schedule
1297 bash-4043 [00] 41.267107: hrtick_clear <-hrtick_set
1298 bash-4043 [00] 41.267108: wait_for_completion <-__stop_machine_run
1299 bash-4043 [00] 41.267108: wait_for_common <-wait_for_completion
1300 bash-4043 [00] 41.267109: kthread_stop <-stop_machine_run
1301 bash-4043 [00] 41.267109: init_waitqueue_head <-kthread_stop
1302 bash-4043 [00] 41.267110: wake_up_process <-kthread_stop
1303 bash-4043 [00] 41.267110: try_to_wake_up <-wake_up_process
1304 bash-4043 [00] 41.267111: select_task_rq_rt <-try_to_wake_up
1307 Note, reading the trace_pipe file will block until more input is added.
1308 By changing the tracer, trace_pipe will issue an EOF. We needed
1309 to set the function tracer _before_ we "cat" the trace_pipe file.
1315 Having too much or not enough data can be troublesome in diagnosing
1316 an issue in the kernel. The file buffer_size_kb is used to modify
1317 the size of the internal trace buffers. The number listed
1318 is the number of entries that can be recorded per CPU. To know
1319 the full size, multiply the number of possible CPUS with the
1322 # cat /debug/tracing/buffer_size_kb
1323 1408 (units kilobytes)
1325 Note, to modify this, you must have tracing completely disabled. To do that,
1326 echo "nop" into the current_tracer. If the current_tracer is not set
1327 to "nop", an EINVAL error will be returned.
1329 # echo nop > /debug/tracing/current_tracer
1330 # echo 10000 > /debug/tracing/buffer_size_kb
1331 # cat /debug/tracing/buffer_size_kb
1332 10000 (units kilobytes)
1334 The number of pages which will be allocated is limited to a percentage
1335 of available memory. Allocating too much will produce an error.
1337 # echo 1000000000000 > /debug/tracing/buffer_size_kb
1338 -bash: echo: write error: Cannot allocate memory
1339 # cat /debug/tracing/buffer_size_kb