| blob | 272464bb7c60249fe93e84a4b81a65c85fe654ce |
1 ========================
2 ftrace - Function Tracer
3 ========================
5 Copyright 2008 Red Hat Inc.
7 :Author: Steven Rostedt <srostedt@redhat.com>
8 :License: The GNU Free Documentation License, Version 1.2
9 (dual licensed under the GPL v2)
10 :Original Reviewers: Elias Oltmanns, Randy Dunlap, Andrew Morton,
11 John Kacur, and David Teigland.
13 - Written for: 2.6.28-rc2
14 - Updated for: 3.10
15 - Updated for: 4.13 - Copyright 2017 VMware Inc. Steven Rostedt
16 - Converted to rst format - Changbin Du <changbin.du@intel.com>
18 Introduction
19 ------------
21 Ftrace is an internal tracer designed to help out developers and
22 designers of systems to find what is going on inside the kernel.
23 It can be used for debugging or analyzing latencies and
24 performance issues that take place outside of user-space.
26 Although ftrace is typically considered the function tracer, it
27 is really a framework of several assorted tracing utilities.
28 There's latency tracing to examine what occurs between interrupts
29 disabled and enabled, as well as for preemption and from a time
30 a task is woken to the task is actually scheduled in.
32 One of the most common uses of ftrace is the event tracing.
33 Throughout the kernel is hundreds of static event points that
34 can be enabled via the tracefs file system to see what is
35 going on in certain parts of the kernel.
37 See events.rst for more information.
40 Implementation Details
41 ----------------------
43 See Documentation/trace/ftrace-design.rst for details for arch porters and such.
46 The File System
47 ---------------
49 Ftrace uses the tracefs file system to hold the control files as
50 well as the files to display output.
52 When tracefs is configured into the kernel (which selecting any ftrace
53 option will do) the directory /sys/kernel/tracing will be created. To mount
54 this directory, you can add to your /etc/fstab file::
56 tracefs /sys/kernel/tracing tracefs defaults 0 0
58 Or you can mount it at run time with::
60 mount -t tracefs nodev /sys/kernel/tracing
62 For quicker access to that directory you may want to make a soft link to
63 it::
65 ln -s /sys/kernel/tracing /tracing
67 .. attention::
69 Before 4.1, all ftrace tracing control files were within the debugfs
70 file system, which is typically located at /sys/kernel/debug/tracing.
71 For backward compatibility, when mounting the debugfs file system,
72 the tracefs file system will be automatically mounted at:
74 /sys/kernel/debug/tracing
76 All files located in the tracefs file system will be located in that
77 debugfs file system directory as well.
79 .. attention::
81 Any selected ftrace option will also create the tracefs file system.
82 The rest of the document will assume that you are in the ftrace directory
83 (cd /sys/kernel/tracing) and will only concentrate on the files within that
84 directory and not distract from the content with the extended
85 "/sys/kernel/tracing" path name.
87 That's it! (assuming that you have ftrace configured into your kernel)
89 After mounting tracefs you will have access to the control and output files
90 of ftrace. Here is a list of some of the key files:
93 Note: all time values are in microseconds.
95 current_tracer:
97 This is used to set or display the current tracer
98 that is configured. Changing the current tracer clears
99 the ring buffer content as well as the "snapshot" buffer.
101 available_tracers:
103 This holds the different types of tracers that
104 have been compiled into the kernel. The
105 tracers listed here can be configured by
106 echoing their name into current_tracer.
108 tracing_on:
110 This sets or displays whether writing to the trace
111 ring buffer is enabled. Echo 0 into this file to disable
112 the tracer or 1 to enable it. Note, this only disables
113 writing to the ring buffer, the tracing overhead may
114 still be occurring.
116 The kernel function tracing_off() can be used within the
117 kernel to disable writing to the ring buffer, which will
118 set this file to "0". User space can re-enable tracing by
119 echoing "1" into the file.
121 Note, the function and event trigger "traceoff" will also
122 set this file to zero and stop tracing. Which can also
123 be re-enabled by user space using this file.
125 trace:
127 This file holds the output of the trace in a human
128 readable format (described below). Opening this file for
129 writing with the O_TRUNC flag clears the ring buffer content.
130 Note, this file is not a consumer. If tracing is off
131 (no tracer running, or tracing_on is zero), it will produce
132 the same output each time it is read. When tracing is on,
133 it may produce inconsistent results as it tries to read
134 the entire buffer without consuming it.
136 trace_pipe:
138 The output is the same as the "trace" file but this
139 file is meant to be streamed with live tracing.
140 Reads from this file will block until new data is
141 retrieved. Unlike the "trace" file, this file is a
142 consumer. This means reading from this file causes
143 sequential reads to display more current data. Once
144 data is read from this file, it is consumed, and
145 will not be read again with a sequential read. The
146 "trace" file is static, and if the tracer is not
147 adding more data, it will display the same
148 information every time it is read.
150 trace_options:
152 This file lets the user control the amount of data
153 that is displayed in one of the above output
154 files. Options also exist to modify how a tracer
155 or events work (stack traces, timestamps, etc).
157 options:
159 This is a directory that has a file for every available
160 trace option (also in trace_options). Options may also be set
161 or cleared by writing a "1" or "0" respectively into the
162 corresponding file with the option name.
164 tracing_max_latency:
166 Some of the tracers record the max latency.
167 For example, the maximum time that interrupts are disabled.
168 The maximum time is saved in this file. The max trace will also be
169 stored, and displayed by "trace". A new max trace will only be
170 recorded if the latency is greater than the value in this file
171 (in microseconds).
173 By echoing in a time into this file, no latency will be recorded
174 unless it is greater than the time in this file.
176 tracing_thresh:
178 Some latency tracers will record a trace whenever the
179 latency is greater than the number in this file.
180 Only active when the file contains a number greater than 0.
181 (in microseconds)
183 buffer_percent:
185 This is the watermark for how much the ring buffer needs to be filled
186 before a waiter is woken up. That is, if an application calls a
187 blocking read syscall on one of the per_cpu trace_pipe_raw files, it
188 will block until the given amount of data specified by buffer_percent
189 is in the ring buffer before it wakes the reader up. This also
190 controls how the splice system calls are blocked on this file::
192 0 - means to wake up as soon as there is any data in the ring buffer.
193 50 - means to wake up when roughly half of the ring buffer sub-buffers
194 are full.
195 100 - means to block until the ring buffer is totally full and is
196 about to start overwriting the older data.
198 buffer_size_kb:
200 This sets or displays the number of kilobytes each CPU
201 buffer holds. By default, the trace buffers are the same size
202 for each CPU. The displayed number is the size of the
203 CPU buffer and not total size of all buffers. The
204 trace buffers are allocated in pages (blocks of memory
205 that the kernel uses for allocation, usually 4 KB in size).
206 A few extra pages may be allocated to accommodate buffer management
207 meta-data. If the last page allocated has room for more bytes
208 than requested, the rest of the page will be used,
209 making the actual allocation bigger than requested or shown.
210 ( Note, the size may not be a multiple of the page size
211 due to buffer management meta-data. )
213 Buffer sizes for individual CPUs may vary
214 (see "per_cpu/cpu0/buffer_size_kb" below), and if they do
215 this file will show "X".
217 buffer_total_size_kb:
219 This displays the total combined size of all the trace buffers.
221 buffer_subbuf_size_kb:
223 This sets or displays the sub buffer size. The ring buffer is broken up
224 into several same size "sub buffers". An event can not be bigger than
225 the size of the sub buffer. Normally, the sub buffer is the size of the
226 architecture's page (4K on x86). The sub buffer also contains meta data
227 at the start which also limits the size of an event. That means when
228 the sub buffer is a page size, no event can be larger than the page
229 size minus the sub buffer meta data.
231 Note, the buffer_subbuf_size_kb is a way for the user to specify the
232 minimum size of the subbuffer. The kernel may make it bigger due to the
233 implementation details, or simply fail the operation if the kernel can
234 not handle the request.
236 Changing the sub buffer size allows for events to be larger than the
237 page size.
239 Note: When changing the sub-buffer size, tracing is stopped and any
240 data in the ring buffer and the snapshot buffer will be discarded.
242 free_buffer:
244 If a process is performing tracing, and the ring buffer should be
245 shrunk "freed" when the process is finished, even if it were to be
246 killed by a signal, this file can be used for that purpose. On close
247 of this file, the ring buffer will be resized to its minimum size.
248 Having a process that is tracing also open this file, when the process
249 exits its file descriptor for this file will be closed, and in doing so,
250 the ring buffer will be "freed".
252 It may also stop tracing if disable_on_free option is set.
254 tracing_cpumask:
256 This is a mask that lets the user only trace on specified CPUs.
257 The format is a hex string representing the CPUs.
259 set_ftrace_filter:
261 When dynamic ftrace is configured in (see the
262 section below "dynamic ftrace"), the code is dynamically
263 modified (code text rewrite) to disable calling of the
264 function profiler (mcount). This lets tracing be configured
265 in with practically no overhead in performance. This also
266 has a side effect of enabling or disabling specific functions
267 to be traced. Echoing names of functions into this file
268 will limit the trace to only those functions.
269 This influences the tracers "function" and "function_graph"
270 and thus also function profiling (see "function_profile_enabled").
272 The functions listed in "available_filter_functions" are what
273 can be written into this file.
275 This interface also allows for commands to be used. See the
276 "Filter commands" section for more details.
278 As a speed up, since processing strings can be quite expensive
279 and requires a check of all functions registered to tracing, instead
280 an index can be written into this file. A number (starting with "1")
281 written will instead select the same corresponding at the line position
282 of the "available_filter_functions" file.
284 set_ftrace_notrace:
286 This has an effect opposite to that of
287 set_ftrace_filter. Any function that is added here will not
288 be traced. If a function exists in both set_ftrace_filter
289 and set_ftrace_notrace, the function will _not_ be traced.
291 set_ftrace_pid:
293 Have the function tracer only trace the threads whose PID are
294 listed in this file.
296 If the "function-fork" option is set, then when a task whose
297 PID is listed in this file forks, the child's PID will
298 automatically be added to this file, and the child will be
299 traced by the function tracer as well. This option will also
300 cause PIDs of tasks that exit to be removed from the file.
302 set_ftrace_notrace_pid:
304 Have the function tracer ignore threads whose PID are listed in
305 this file.
307 If the "function-fork" option is set, then when a task whose
308 PID is listed in this file forks, the child's PID will
309 automatically be added to this file, and the child will not be
310 traced by the function tracer as well. This option will also
311 cause PIDs of tasks that exit to be removed from the file.
313 If a PID is in both this file and "set_ftrace_pid", then this
314 file takes precedence, and the thread will not be traced.
316 set_event_pid:
318 Have the events only trace a task with a PID listed in this file.
319 Note, sched_switch and sched_wake_up will also trace events
320 listed in this file.
322 To have the PIDs of children of tasks with their PID in this file
323 added on fork, enable the "event-fork" option. That option will also
324 cause the PIDs of tasks to be removed from this file when the task
325 exits.
327 set_event_notrace_pid:
329 Have the events not trace a task with a PID listed in this file.
330 Note, sched_switch and sched_wakeup will trace threads not listed
331 in this file, even if a thread's PID is in the file if the
332 sched_switch or sched_wakeup events also trace a thread that should
333 be traced.
335 To have the PIDs of children of tasks with their PID in this file
336 added on fork, enable the "event-fork" option. That option will also
337 cause the PIDs of tasks to be removed from this file when the task
338 exits.
340 set_graph_function:
342 Functions listed in this file will cause the function graph
343 tracer to only trace these functions and the functions that
344 they call. (See the section "dynamic ftrace" for more details).
345 Note, set_ftrace_filter and set_ftrace_notrace still affects
346 what functions are being traced.
348 set_graph_notrace:
350 Similar to set_graph_function, but will disable function graph
351 tracing when the function is hit until it exits the function.
352 This makes it possible to ignore tracing functions that are called
353 by a specific function.
355 available_filter_functions:
357 This lists the functions that ftrace has processed and can trace.
358 These are the function names that you can pass to
359 "set_ftrace_filter", "set_ftrace_notrace",
360 "set_graph_function", or "set_graph_notrace".
361 (See the section "dynamic ftrace" below for more details.)
363 available_filter_functions_addrs:
365 Similar to available_filter_functions, but with address displayed
366 for each function. The displayed address is the patch-site address
367 and can differ from /proc/kallsyms address.
369 dyn_ftrace_total_info:
371 This file is for debugging purposes. The number of functions that
372 have been converted to nops and are available to be traced.
374 enabled_functions:
376 This file is more for debugging ftrace, but can also be useful
377 in seeing if any function has a callback attached to it.
378 Not only does the trace infrastructure use ftrace function
379 trace utility, but other subsystems might too. This file
380 displays all functions that have a callback attached to them
381 as well as the number of callbacks that have been attached.
382 Note, a callback may also call multiple functions which will
383 not be listed in this count.
385 If the callback registered to be traced by a function with
386 the "save regs" attribute (thus even more overhead), a 'R'
387 will be displayed on the same line as the function that
388 is returning registers.
390 If the callback registered to be traced by a function with
391 the "ip modify" attribute (thus the regs->ip can be changed),
392 an 'I' will be displayed on the same line as the function that
393 can be overridden.
395 If a non ftrace trampoline is attached (BPF) a 'D' will be displayed.
396 Note, normal ftrace trampolines can also be attached, but only one
397 "direct" trampoline can be attached to a given function at a time.
399 Some architectures can not call direct trampolines, but instead have
400 the ftrace ops function located above the function entry point. In
401 such cases an 'O' will be displayed.
403 If a function had either the "ip modify" or a "direct" call attached to
404 it in the past, a 'M' will be shown. This flag is never cleared. It is
405 used to know if a function was every modified by the ftrace infrastructure,
406 and can be used for debugging.
408 If the architecture supports it, it will also show what callback
409 is being directly called by the function. If the count is greater
410 than 1 it most likely will be ftrace_ops_list_func().
412 If the callback of a function jumps to a trampoline that is
413 specific to the callback and which is not the standard trampoline,
414 its address will be printed as well as the function that the
415 trampoline calls.
417 touched_functions:
419 This file contains all the functions that ever had a function callback
420 to it via the ftrace infrastructure. It has the same format as
421 enabled_functions but shows all functions that have every been
422 traced.
424 To see any function that has every been modified by "ip modify" or a
425 direct trampoline, one can perform the following command:
427 grep ' M ' /sys/kernel/tracing/touched_functions
429 function_profile_enabled:
431 When set it will enable all functions with either the function
432 tracer, or if configured, the function graph tracer. It will
433 keep a histogram of the number of functions that were called
434 and if the function graph tracer was configured, it will also keep
435 track of the time spent in those functions. The histogram
436 content can be displayed in the files:
438 trace_stat/function<cpu> ( function0, function1, etc).
440 trace_stat:
442 A directory that holds different tracing stats.
444 kprobe_events:
446 Enable dynamic trace points. See kprobetrace.rst.
448 kprobe_profile:
450 Dynamic trace points stats. See kprobetrace.rst.
452 max_graph_depth:
454 Used with the function graph tracer. This is the max depth
455 it will trace into a function. Setting this to a value of
456 one will show only the first kernel function that is called
457 from user space.
459 printk_formats:
461 This is for tools that read the raw format files. If an event in
462 the ring buffer references a string, only a pointer to the string
463 is recorded into the buffer and not the string itself. This prevents
464 tools from knowing what that string was. This file displays the string
465 and address for the string allowing tools to map the pointers to what
466 the strings were.
468 saved_cmdlines:
470 Only the pid of the task is recorded in a trace event unless
471 the event specifically saves the task comm as well. Ftrace
472 makes a cache of pid mappings to comms to try to display
473 comms for events. If a pid for a comm is not listed, then
474 "<...>" is displayed in the output.
476 If the option "record-cmd" is set to "0", then comms of tasks
477 will not be saved during recording. By default, it is enabled.
479 saved_cmdlines_size:
481 By default, 128 comms are saved (see "saved_cmdlines" above). To
482 increase or decrease the amount of comms that are cached, echo
483 the number of comms to cache into this file.
485 saved_tgids:
487 If the option "record-tgid" is set, on each scheduling context switch
488 the Task Group ID of a task is saved in a table mapping the PID of
489 the thread to its TGID. By default, the "record-tgid" option is
490 disabled.
492 snapshot:
494 This displays the "snapshot" buffer and also lets the user
495 take a snapshot of the current running trace.
496 See the "Snapshot" section below for more details.
498 stack_max_size:
500 When the stack tracer is activated, this will display the
501 maximum stack size it has encountered.
502 See the "Stack Trace" section below.
504 stack_trace:
506 This displays the stack back trace of the largest stack
507 that was encountered when the stack tracer is activated.
508 See the "Stack Trace" section below.
510 stack_trace_filter:
512 This is similar to "set_ftrace_filter" but it limits what
513 functions the stack tracer will check.
515 trace_clock:
517 Whenever an event is recorded into the ring buffer, a
518 "timestamp" is added. This stamp comes from a specified
519 clock. By default, ftrace uses the "local" clock. This
520 clock is very fast and strictly per cpu, but on some
521 systems it may not be monotonic with respect to other
522 CPUs. In other words, the local clocks may not be in sync
523 with local clocks on other CPUs.
525 Usual clocks for tracing::
527 # cat trace_clock
528 [local] global counter x86-tsc
530 The clock with the square brackets around it is the one in effect.
532 local:
533 Default clock, but may not be in sync across CPUs
535 global:
536 This clock is in sync with all CPUs but may
537 be a bit slower than the local clock.
539 counter:
540 This is not a clock at all, but literally an atomic
541 counter. It counts up one by one, but is in sync
542 with all CPUs. This is useful when you need to
543 know exactly the order events occurred with respect to
544 each other on different CPUs.
546 uptime:
547 This uses the jiffies counter and the time stamp
548 is relative to the time since boot up.
550 perf:
551 This makes ftrace use the same clock that perf uses.
552 Eventually perf will be able to read ftrace buffers
553 and this will help out in interleaving the data.
555 x86-tsc:
556 Architectures may define their own clocks. For
557 example, x86 uses its own TSC cycle clock here.
559 ppc-tb:
560 This uses the powerpc timebase register value.
561 This is in sync across CPUs and can also be used
562 to correlate events across hypervisor/guest if
563 tb_offset is known.
565 mono:
566 This uses the fast monotonic clock (CLOCK_MONOTONIC)
567 which is monotonic and is subject to NTP rate adjustments.
569 mono_raw:
570 This is the raw monotonic clock (CLOCK_MONOTONIC_RAW)
571 which is monotonic but is not subject to any rate adjustments
572 and ticks at the same rate as the hardware clocksource.
574 boot:
575 This is the boot clock (CLOCK_BOOTTIME) and is based on the
576 fast monotonic clock, but also accounts for time spent in
577 suspend. Since the clock access is designed for use in
578 tracing in the suspend path, some side effects are possible
579 if clock is accessed after the suspend time is accounted before
580 the fast mono clock is updated. In this case, the clock update
581 appears to happen slightly sooner than it normally would have.
582 Also on 32-bit systems, it's possible that the 64-bit boot offset
583 sees a partial update. These effects are rare and post
584 processing should be able to handle them. See comments in the
585 ktime_get_boot_fast_ns() function for more information.
587 tai:
588 This is the tai clock (CLOCK_TAI) and is derived from the wall-
589 clock time. However, this clock does not experience
590 discontinuities and backwards jumps caused by NTP inserting leap
591 seconds. Since the clock access is designed for use in tracing,
592 side effects are possible. The clock access may yield wrong
593 readouts in case the internal TAI offset is updated e.g., caused
594 by setting the system time or using adjtimex() with an offset.
595 These effects are rare and post processing should be able to
596 handle them. See comments in the ktime_get_tai_fast_ns()
597 function for more information.
599 To set a clock, simply echo the clock name into this file::
601 # echo global > trace_clock
603 Setting a clock clears the ring buffer content as well as the
604 "snapshot" buffer.
606 trace_marker:
608 This is a very useful file for synchronizing user space
609 with events happening in the kernel. Writing strings into
610 this file will be written into the ftrace buffer.
612 It is useful in applications to open this file at the start
613 of the application and just reference the file descriptor
614 for the file::
616 void trace_write(const char *fmt, ...)
617 {
618 va_list ap;
619 char buf[256];
620 int n;
622 if (trace_fd < 0)
623 return;
625 va_start(ap, fmt);
626 n = vsnprintf(buf, 256, fmt, ap);
627 va_end(ap);
629 write(trace_fd, buf, n);
630 }
632 start::
634 trace_fd = open("trace_marker", O_WRONLY);
636 Note: Writing into the trace_marker file can also initiate triggers
637 that are written into /sys/kernel/tracing/events/ftrace/print/trigger
638 See "Event triggers" in Documentation/trace/events.rst and an
639 example in Documentation/trace/histogram.rst (Section 3.)
641 trace_marker_raw:
643 This is similar to trace_marker above, but is meant for binary data
644 to be written to it, where a tool can be used to parse the data
645 from trace_pipe_raw.
647 uprobe_events:
649 Add dynamic tracepoints in programs.
650 See uprobetracer.rst
652 uprobe_profile:
654 Uprobe statistics. See uprobetrace.txt
656 instances:
658 This is a way to make multiple trace buffers where different
659 events can be recorded in different buffers.
660 See "Instances" section below.
662 events:
664 This is the trace event directory. It holds event tracepoints
665 (also known as static tracepoints) that have been compiled
666 into the kernel. It shows what event tracepoints exist
667 and how they are grouped by system. There are "enable"
668 files at various levels that can enable the tracepoints
669 when a "1" is written to them.
671 See events.rst for more information.
673 set_event:
675 By echoing in the event into this file, will enable that event.
677 See events.rst for more information.
679 available_events:
681 A list of events that can be enabled in tracing.
683 See events.rst for more information.
685 timestamp_mode:
687 Certain tracers may change the timestamp mode used when
688 logging trace events into the event buffer. Events with
689 different modes can coexist within a buffer but the mode in
690 effect when an event is logged determines which timestamp mode
691 is used for that event. The default timestamp mode is
692 'delta'.
694 Usual timestamp modes for tracing:
696 # cat timestamp_mode
697 [delta] absolute
699 The timestamp mode with the square brackets around it is the
700 one in effect.
702 delta: Default timestamp mode - timestamp is a delta against
703 a per-buffer timestamp.
705 absolute: The timestamp is a full timestamp, not a delta
706 against some other value. As such it takes up more
707 space and is less efficient.
709 hwlat_detector:
711 Directory for the Hardware Latency Detector.
712 See "Hardware Latency Detector" section below.
714 per_cpu:
716 This is a directory that contains the trace per_cpu information.
718 per_cpu/cpu0/buffer_size_kb:
720 The ftrace buffer is defined per_cpu. That is, there's a separate
721 buffer for each CPU to allow writes to be done atomically,
722 and free from cache bouncing. These buffers may have different
723 size buffers. This file is similar to the buffer_size_kb
724 file, but it only displays or sets the buffer size for the
725 specific CPU. (here cpu0).
727 per_cpu/cpu0/trace:
729 This is similar to the "trace" file, but it will only display
730 the data specific for the CPU. If written to, it only clears
731 the specific CPU buffer.
733 per_cpu/cpu0/trace_pipe
735 This is similar to the "trace_pipe" file, and is a consuming
736 read, but it will only display (and consume) the data specific
737 for the CPU.
739 per_cpu/cpu0/trace_pipe_raw
741 For tools that can parse the ftrace ring buffer binary format,
742 the trace_pipe_raw file can be used to extract the data
743 from the ring buffer directly. With the use of the splice()
744 system call, the buffer data can be quickly transferred to
745 a file or to the network where a server is collecting the
746 data.
748 Like trace_pipe, this is a consuming reader, where multiple
749 reads will always produce different data.
751 per_cpu/cpu0/snapshot:
753 This is similar to the main "snapshot" file, but will only
754 snapshot the current CPU (if supported). It only displays
755 the content of the snapshot for a given CPU, and if
756 written to, only clears this CPU buffer.
758 per_cpu/cpu0/snapshot_raw:
760 Similar to the trace_pipe_raw, but will read the binary format
761 from the snapshot buffer for the given CPU.
763 per_cpu/cpu0/stats:
765 This displays certain stats about the ring buffer:
767 entries:
768 The number of events that are still in the buffer.
770 overrun:
771 The number of lost events due to overwriting when
772 the buffer was full.
774 commit overrun:
775 Should always be zero.
776 This gets set if so many events happened within a nested
777 event (ring buffer is re-entrant), that it fills the
778 buffer and starts dropping events.
780 bytes:
781 Bytes actually read (not overwritten).
783 oldest event ts:
784 The oldest timestamp in the buffer
786 now ts:
787 The current timestamp
789 dropped events:
790 Events lost due to overwrite option being off.
792 read events:
793 The number of events read.
795 The Tracers
796 -----------
798 Here is the list of current tracers that may be configured.
800 "function"
802 Function call tracer to trace all kernel functions.
804 "function_graph"
806 Similar to the function tracer except that the
807 function tracer probes the functions on their entry
808 whereas the function graph tracer traces on both entry
809 and exit of the functions. It then provides the ability
810 to draw a graph of function calls similar to C code
811 source.
813 "blk"
815 The block tracer. The tracer used by the blktrace user
816 application.
818 "hwlat"
820 The Hardware Latency tracer is used to detect if the hardware
821 produces any latency. See "Hardware Latency Detector" section
822 below.
824 "irqsoff"
826 Traces the areas that disable interrupts and saves
827 the trace with the longest max latency.
828 See tracing_max_latency. When a new max is recorded,
829 it replaces the old trace. It is best to view this
830 trace with the latency-format option enabled, which
831 happens automatically when the tracer is selected.
833 "preemptoff"
835 Similar to irqsoff but traces and records the amount of
836 time for which preemption is disabled.
838 "preemptirqsoff"
840 Similar to irqsoff and preemptoff, but traces and
841 records the largest time for which irqs and/or preemption
842 is disabled.
844 "wakeup"
846 Traces and records the max latency that it takes for
847 the highest priority task to get scheduled after
848 it has been woken up.
849 Traces all tasks as an average developer would expect.
851 "wakeup_rt"
853 Traces and records the max latency that it takes for just
854 RT tasks (as the current "wakeup" does). This is useful
855 for those interested in wake up timings of RT tasks.
857 "wakeup_dl"
859 Traces and records the max latency that it takes for
860 a SCHED_DEADLINE task to be woken (as the "wakeup" and
861 "wakeup_rt" does).
863 "mmiotrace"
865 A special tracer that is used to trace binary module.
866 It will trace all the calls that a module makes to the
867 hardware. Everything it writes and reads from the I/O
868 as well.
870 "branch"
872 This tracer can be configured when tracing likely/unlikely
873 calls within the kernel. It will trace when a likely and
874 unlikely branch is hit and if it was correct in its prediction
875 of being correct.
877 "nop"
879 This is the "trace nothing" tracer. To remove all
880 tracers from tracing simply echo "nop" into
881 current_tracer.
883 Error conditions
884 ----------------
886 For most ftrace commands, failure modes are obvious and communicated
887 using standard return codes.
889 For other more involved commands, extended error information may be
890 available via the tracing/error_log file. For the commands that
891 support it, reading the tracing/error_log file after an error will
892 display more detailed information about what went wrong, if
893 information is available. The tracing/error_log file is a circular
894 error log displaying a small number (currently, 8) of ftrace errors
895 for the last (8) failed commands.
897 The extended error information and usage takes the form shown in
898 this example::
900 # echo xxx > /sys/kernel/tracing/events/sched/sched_wakeup/trigger
901 echo: write error: Invalid argument
903 # cat /sys/kernel/tracing/error_log
904 [ 5348.887237] location: error: Couldn't yyy: zzz
905 Command: xxx
906 ^
907 [ 7517.023364] location: error: Bad rrr: sss
908 Command: ppp qqq
909 ^
911 To clear the error log, echo the empty string into it::
913 # echo > /sys/kernel/tracing/error_log
915 Examples of using the tracer
916 ----------------------------
918 Here are typical examples of using the tracers when controlling
919 them only with the tracefs interface (without using any
920 user-land utilities).
922 Output format:
923 --------------
925 Here is an example of the output format of the file "trace"::
927 # tracer: function
928 #
929 # entries-in-buffer/entries-written: 140080/250280 #P:4
930 #
931 # _-----=> irqs-off
932 # / _----=> need-resched
933 # | / _---=> hardirq/softirq
934 # || / _--=> preempt-depth
935 # ||| / delay
936 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
937 # | | | |||| | |
938 bash-1977 [000] .... 17284.993652: sys_close <-system_call_fastpath
939 bash-1977 [000] .... 17284.993653: __close_fd <-sys_close
940 bash-1977 [000] .... 17284.993653: _raw_spin_lock <-__close_fd
941 sshd-1974 [003] .... 17284.993653: __srcu_read_unlock <-fsnotify
942 bash-1977 [000] .... 17284.993654: add_preempt_count <-_raw_spin_lock
943 bash-1977 [000] ...1 17284.993655: _raw_spin_unlock <-__close_fd
944 bash-1977 [000] ...1 17284.993656: sub_preempt_count <-_raw_spin_unlock
945 bash-1977 [000] .... 17284.993657: filp_close <-__close_fd
946 bash-1977 [000] .... 17284.993657: dnotify_flush <-filp_close
947 sshd-1974 [003] .... 17284.993658: sys_select <-system_call_fastpath
948 ....
950 A header is printed with the tracer name that is represented by
951 the trace. In this case the tracer is "function". Then it shows the
952 number of events in the buffer as well as the total number of entries
953 that were written. The difference is the number of entries that were
954 lost due to the buffer filling up (250280 - 140080 = 110200 events
955 lost).
957 The header explains the content of the events. Task name "bash", the task
958 PID "1977", the CPU that it was running on "000", the latency format
959 (explained below), the timestamp in <secs>.<usecs> format, the
960 function name that was traced "sys_close" and the parent function that
961 called this function "system_call_fastpath". The timestamp is the time
962 at which the function was entered.
964 Latency trace format
965 --------------------
967 When the latency-format option is enabled or when one of the latency
968 tracers is set, the trace file gives somewhat more information to see
969 why a latency happened. Here is a typical trace::
971 # tracer: irqsoff
972 #
973 # irqsoff latency trace v1.1.5 on 3.8.0-test+
974 # --------------------------------------------------------------------
975 # latency: 259 us, #4/4, CPU#2 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
976 # -----------------
977 # | task: ps-6143 (uid:0 nice:0 policy:0 rt_prio:0)
978 # -----------------
979 # => started at: __lock_task_sighand
980 # => ended at: _raw_spin_unlock_irqrestore
981 #
982 #
983 # _------=> CPU#
984 # / _-----=> irqs-off
985 # | / _----=> need-resched
986 # || / _---=> hardirq/softirq
987 # ||| / _--=> preempt-depth
988 # |||| / delay
989 # cmd pid ||||| time | caller
990 # \ / ||||| \ | /
991 ps-6143 2d... 0us!: trace_hardirqs_off <-__lock_task_sighand
992 ps-6143 2d..1 259us+: trace_hardirqs_on <-_raw_spin_unlock_irqrestore
993 ps-6143 2d..1 263us+: time_hardirqs_on <-_raw_spin_unlock_irqrestore
994 ps-6143 2d..1 306us : <stack trace>
995 => trace_hardirqs_on_caller
996 => trace_hardirqs_on
997 => _raw_spin_unlock_irqrestore
998 => do_task_stat
999 => proc_tgid_stat
1000 => proc_single_show
1001 => seq_read
1002 => vfs_read
1003 => sys_read
1004 => system_call_fastpath
1007 This shows that the current tracer is "irqsoff" tracing the time
1008 for which interrupts were disabled. It gives the trace version (which
1009 never changes) and the version of the kernel upon which this was executed on
1010 (3.8). Then it displays the max latency in microseconds (259 us). The number
1011 of trace entries displayed and the total number (both are four: #4/4).
1012 VP, KP, SP, and HP are always zero and are reserved for later use.
1013 #P is the number of online CPUs (#P:4).
1015 The task is the process that was running when the latency
1016 occurred. (ps pid: 6143).
1018 The start and stop (the functions in which the interrupts were
1019 disabled and enabled respectively) that caused the latencies:
1021 - __lock_task_sighand is where the interrupts were disabled.
1022 - _raw_spin_unlock_irqrestore is where they were enabled again.
1024 The next lines after the header are the trace itself. The header
1025 explains which is which.
1027 cmd: The name of the process in the trace.
1029 pid: The PID of that process.
1031 CPU#: The CPU which the process was running on.
1033 irqs-off: 'd' interrupts are disabled. '.' otherwise.
1035 need-resched:
1036 - 'B' all, TIF_NEED_RESCHED, PREEMPT_NEED_RESCHED and TIF_RESCHED_LAZY is set,
1037 - 'N' both TIF_NEED_RESCHED and PREEMPT_NEED_RESCHED is set,
1038 - 'n' only TIF_NEED_RESCHED is set,
1039 - 'p' only PREEMPT_NEED_RESCHED is set,
1040 - 'L' both PREEMPT_NEED_RESCHED and TIF_RESCHED_LAZY is set,
1041 - 'b' both TIF_NEED_RESCHED and TIF_RESCHED_LAZY is set,
1042 - 'l' only TIF_RESCHED_LAZY is set
1043 - '.' otherwise.
1045 hardirq/softirq:
1046 - 'Z' - NMI occurred inside a hardirq
1047 - 'z' - NMI is running
1048 - 'H' - hard irq occurred inside a softirq.
1049 - 'h' - hard irq is running
1050 - 's' - soft irq is running
1051 - '.' - normal context.
1053 preempt-depth: The level of preempt_disabled
1055 The above is mostly meaningful for kernel developers.
1057 time:
1058 When the latency-format option is enabled, the trace file
1059 output includes a timestamp relative to the start of the
1060 trace. This differs from the output when latency-format
1061 is disabled, which includes an absolute timestamp.
1063 delay:
1064 This is just to help catch your eye a bit better. And
1065 needs to be fixed to be only relative to the same CPU.
1066 The marks are determined by the difference between this
1067 current trace and the next trace.
1069 - '$' - greater than 1 second
1070 - '@' - greater than 100 millisecond
1071 - '*' - greater than 10 millisecond
1072 - '#' - greater than 1000 microsecond
1073 - '!' - greater than 100 microsecond
1074 - '+' - greater than 10 microsecond
1075 - ' ' - less than or equal to 10 microsecond.
1077 The rest is the same as the 'trace' file.
1079 Note, the latency tracers will usually end with a back trace
1080 to easily find where the latency occurred.
1082 trace_options
1083 -------------
1085 The trace_options file (or the options directory) is used to control
1086 what gets printed in the trace output, or manipulate the tracers.
1087 To see what is available, simply cat the file::
1089 cat trace_options
1090 print-parent
1091 nosym-offset
1092 nosym-addr
1093 noverbose
1094 noraw
1095 nohex
1096 nobin
1097 noblock
1098 nofields
1099 trace_printk
1100 annotate
1101 nouserstacktrace
1102 nosym-userobj
1103 noprintk-msg-only
1104 context-info
1105 nolatency-format
1106 record-cmd
1107 norecord-tgid
1108 overwrite
1109 nodisable_on_free
1110 irq-info
1111 markers
1112 noevent-fork
1113 function-trace
1114 nofunction-fork
1115 nodisplay-graph
1116 nostacktrace
1117 nobranch
1119 To disable one of the options, echo in the option prepended with
1120 "no"::
1122 echo noprint-parent > trace_options
1124 To enable an option, leave off the "no"::
1126 echo sym-offset > trace_options
1128 Here are the available options:
1130 print-parent
1131 On function traces, display the calling (parent)
1132 function as well as the function being traced.
1133 ::
1135 print-parent:
1136 bash-4000 [01] 1477.606694: simple_strtoul <-kstrtoul
1138 noprint-parent:
1139 bash-4000 [01] 1477.606694: simple_strtoul
1142 sym-offset
1143 Display not only the function name, but also the
1144 offset in the function. For example, instead of
1145 seeing just "ktime_get", you will see
1146 "ktime_get+0xb/0x20".
1147 ::
1149 sym-offset:
1150 bash-4000 [01] 1477.606694: simple_strtoul+0x6/0xa0
1152 sym-addr
1153 This will also display the function address as well
1154 as the function name.
1155 ::
1157 sym-addr:
1158 bash-4000 [01] 1477.606694: simple_strtoul <c0339346>
1160 verbose
1161 This deals with the trace file when the
1162 latency-format option is enabled.
1163 ::
1165 bash 4000 1 0 00000000 00010a95 [58127d26] 1720.415ms \
1166 (+0.000ms): simple_strtoul (kstrtoul)
1168 raw
1169 This will display raw numbers. This option is best for
1170 use with user applications that can translate the raw
1171 numbers better than having it done in the kernel.
1173 hex
1174 Similar to raw, but the numbers will be in a hexadecimal format.
1176 bin
1177 This will print out the formats in raw binary.
1179 block
1180 When set, reading trace_pipe will not block when polled.
1182 fields
1183 Print the fields as described by their types. This is a better
1184 option than using hex, bin or raw, as it gives a better parsing
1185 of the content of the event.
1187 trace_printk
1188 Can disable trace_printk() from writing into the buffer.
1190 trace_printk_dest
1191 Set to have trace_printk() and similar internal tracing functions
1192 write into this instance. Note, only one trace instance can have
1193 this set. By setting this flag, it clears the trace_printk_dest flag
1194 of the instance that had it set previously. By default, the top
1195 level trace has this set, and will get it set again if another
1196 instance has it set then clears it.
1198 This flag cannot be cleared by the top level instance, as it is the
1199 default instance. The only way the top level instance has this flag
1200 cleared, is by it being set in another instance.
1202 annotate
1203 It is sometimes confusing when the CPU buffers are full
1204 and one CPU buffer had a lot of events recently, thus
1205 a shorter time frame, were another CPU may have only had
1206 a few events, which lets it have older events. When
1207 the trace is reported, it shows the oldest events first,
1208 and it may look like only one CPU ran (the one with the
1209 oldest events). When the annotate option is set, it will
1210 display when a new CPU buffer started::
1212 <idle>-0 [001] dNs4 21169.031481: wake_up_idle_cpu <-add_timer_on
1213 <idle>-0 [001] dNs4 21169.031482: _raw_spin_unlock_irqrestore <-add_timer_on
1214 <idle>-0 [001] .Ns4 21169.031484: sub_preempt_count <-_raw_spin_unlock_irqrestore
1215 ##### CPU 2 buffer started ####
1216 <idle>-0 [002] .N.1 21169.031484: rcu_idle_exit <-cpu_idle
1217 <idle>-0 [001] .Ns3 21169.031484: _raw_spin_unlock <-clocksource_watchdog
1218 <idle>-0 [001] .Ns3 21169.031485: sub_preempt_count <-_raw_spin_unlock
1220 userstacktrace
1221 This option changes the trace. It records a
1222 stacktrace of the current user space thread after
1223 each trace event.
1225 sym-userobj
1226 when user stacktrace are enabled, look up which
1227 object the address belongs to, and print a
1228 relative address. This is especially useful when
1229 ASLR is on, otherwise you don't get a chance to
1230 resolve the address to object/file/line after
1231 the app is no longer running
1233 The lookup is performed when you read
1234 trace,trace_pipe. Example::
1236 a.out-1623 [000] 40874.465068: /root/a.out[+0x480] <-/root/a.out[+0
1237 x494] <- /root/a.out[+0x4a8] <- /lib/libc-2.7.so[+0x1e1a6]
1240 printk-msg-only
1241 When set, trace_printk()s will only show the format
1242 and not their parameters (if trace_bprintk() or
1243 trace_bputs() was used to save the trace_printk()).
1245 context-info
1246 Show only the event data. Hides the comm, PID,
1247 timestamp, CPU, and other useful data.
1249 latency-format
1250 This option changes the trace output. When it is enabled,
1251 the trace displays additional information about the
1252 latency, as described in "Latency trace format".
1254 pause-on-trace
1255 When set, opening the trace file for read, will pause
1256 writing to the ring buffer (as if tracing_on was set to zero).
1257 This simulates the original behavior of the trace file.
1258 When the file is closed, tracing will be enabled again.
1260 hash-ptr
1261 When set, "%p" in the event printk format displays the
1262 hashed pointer value instead of real address.
1263 This will be useful if you want to find out which hashed
1264 value is corresponding to the real value in trace log.
1266 record-cmd
1267 When any event or tracer is enabled, a hook is enabled
1268 in the sched_switch trace point to fill comm cache
1269 with mapped pids and comms. But this may cause some
1270 overhead, and if you only care about pids, and not the
1271 name of the task, disabling this option can lower the
1272 impact of tracing. See "saved_cmdlines".
1274 record-tgid
1275 When any event or tracer is enabled, a hook is enabled
1276 in the sched_switch trace point to fill the cache of
1277 mapped Thread Group IDs (TGID) mapping to pids. See
1278 "saved_tgids".
1280 overwrite
1281 This controls what happens when the trace buffer is
1282 full. If "1" (default), the oldest events are
1283 discarded and overwritten. If "0", then the newest
1284 events are discarded.
1285 (see per_cpu/cpu0/stats for overrun and dropped)
1287 disable_on_free
1288 When the free_buffer is closed, tracing will
1289 stop (tracing_on set to 0).
1291 irq-info
1292 Shows the interrupt, preempt count, need resched data.
1293 When disabled, the trace looks like::
1295 # tracer: function
1296 #
1297 # entries-in-buffer/entries-written: 144405/9452052 #P:4
1298 #
1299 # TASK-PID CPU# TIMESTAMP FUNCTION
1300 # | | | | |
1301 <idle>-0 [002] 23636.756054: ttwu_do_activate.constprop.89 <-try_to_wake_up
1302 <idle>-0 [002] 23636.756054: activate_task <-ttwu_do_activate.constprop.89
1303 <idle>-0 [002] 23636.756055: enqueue_task <-activate_task
1306 markers
1307 When set, the trace_marker is writable (only by root).
1308 When disabled, the trace_marker will error with EINVAL
1309 on write.
1311 event-fork
1312 When set, tasks with PIDs listed in set_event_pid will have
1313 the PIDs of their children added to set_event_pid when those
1314 tasks fork. Also, when tasks with PIDs in set_event_pid exit,
1315 their PIDs will be removed from the file.
1317 This affects PIDs listed in set_event_notrace_pid as well.
1319 function-trace
1320 The latency tracers will enable function tracing
1321 if this option is enabled (default it is). When
1322 it is disabled, the latency tracers do not trace
1323 functions. This keeps the overhead of the tracer down
1324 when performing latency tests.
1326 function-fork
1327 When set, tasks with PIDs listed in set_ftrace_pid will
1328 have the PIDs of their children added to set_ftrace_pid
1329 when those tasks fork. Also, when tasks with PIDs in
1330 set_ftrace_pid exit, their PIDs will be removed from the
1331 file.
1333 This affects PIDs in set_ftrace_notrace_pid as well.
1335 display-graph
1336 When set, the latency tracers (irqsoff, wakeup, etc) will
1337 use function graph tracing instead of function tracing.
1339 stacktrace
1340 When set, a stack trace is recorded after any trace event
1341 is recorded.
1343 branch
1344 Enable branch tracing with the tracer. This enables branch
1345 tracer along with the currently set tracer. Enabling this
1346 with the "nop" tracer is the same as just enabling the
1347 "branch" tracer.
1349 .. tip:: Some tracers have their own options. They only appear in this
1350 file when the tracer is active. They always appear in the
1351 options directory.
1354 Here are the per tracer options:
1356 Options for function tracer:
1358 func_stack_trace
1359 When set, a stack trace is recorded after every
1360 function that is recorded. NOTE! Limit the functions
1361 that are recorded before enabling this, with
1362 "set_ftrace_filter" otherwise the system performance
1363 will be critically degraded. Remember to disable
1364 this option before clearing the function filter.
1366 Options for function_graph tracer:
1368 Since the function_graph tracer has a slightly different output
1369 it has its own options to control what is displayed.
1371 funcgraph-overrun
1372 When set, the "overrun" of the graph stack is
1373 displayed after each function traced. The
1374 overrun, is when the stack depth of the calls
1375 is greater than what is reserved for each task.
1376 Each task has a fixed array of functions to
1377 trace in the call graph. If the depth of the
1378 calls exceeds that, the function is not traced.
1379 The overrun is the number of functions missed
1380 due to exceeding this array.
1382 funcgraph-cpu
1383 When set, the CPU number of the CPU where the trace
1384 occurred is displayed.
1386 funcgraph-overhead
1387 When set, if the function takes longer than
1388 A certain amount, then a delay marker is
1389 displayed. See "delay" above, under the
1390 header description.
1392 funcgraph-proc
1393 Unlike other tracers, the process' command line
1394 is not displayed by default, but instead only
1395 when a task is traced in and out during a context
1396 switch. Enabling this options has the command
1397 of each process displayed at every line.
1399 funcgraph-duration
1400 At the end of each function (the return)
1401 the duration of the amount of time in the
1402 function is displayed in microseconds.
1404 funcgraph-abstime
1405 When set, the timestamp is displayed at each line.
1407 funcgraph-irqs
1408 When disabled, functions that happen inside an
1409 interrupt will not be traced.
1411 funcgraph-tail
1412 When set, the return event will include the function
1413 that it represents. By default this is off, and
1414 only a closing curly bracket "}" is displayed for
1415 the return of a function.
1417 funcgraph-retval
1418 When set, the return value of each traced function
1419 will be printed after an equal sign "=". By default
1420 this is off.
1422 funcgraph-retval-hex
1423 When set, the return value will always be printed
1424 in hexadecimal format. If the option is not set and
1425 the return value is an error code, it will be printed
1426 in signed decimal format; otherwise it will also be
1427 printed in hexadecimal format. By default, this option
1428 is off.
1430 sleep-time
1431 When running function graph tracer, to include
1432 the time a task schedules out in its function.
1433 When enabled, it will account time the task has been
1434 scheduled out as part of the function call.
1436 graph-time
1437 When running function profiler with function graph tracer,
1438 to include the time to call nested functions. When this is
1439 not set, the time reported for the function will only
1440 include the time the function itself executed for, not the
1441 time for functions that it called.
1443 Options for blk tracer:
1445 blk_classic
1446 Shows a more minimalistic output.
1449 irqsoff
1450 -------
1452 When interrupts are disabled, the CPU can not react to any other
1453 external event (besides NMIs and SMIs). This prevents the timer
1454 interrupt from triggering or the mouse interrupt from letting
1455 the kernel know of a new mouse event. The result is a latency
1456 with the reaction time.
1458 The irqsoff tracer tracks the time for which interrupts are
1459 disabled. When a new maximum latency is hit, the tracer saves
1460 the trace leading up to that latency point so that every time a
1461 new maximum is reached, the old saved trace is discarded and the
1462 new trace is saved.
1464 To reset the maximum, echo 0 into tracing_max_latency. Here is
1465 an example::
1467 # echo 0 > options/function-trace
1468 # echo irqsoff > current_tracer
1469 # echo 1 > tracing_on
1470 # echo 0 > tracing_max_latency
1471 # ls -ltr
1472 [...]
1473 # echo 0 > tracing_on
1474 # cat trace
1475 # tracer: irqsoff
1476 #
1477 # irqsoff latency trace v1.1.5 on 3.8.0-test+
1478 # --------------------------------------------------------------------
1479 # latency: 16 us, #4/4, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1480 # -----------------
1481 # | task: swapper/0-0 (uid:0 nice:0 policy:0 rt_prio:0)
1482 # -----------------
1483 # => started at: run_timer_softirq
1484 # => ended at: run_timer_softirq
1485 #
1486 #
1487 # _------=> CPU#
1488 # / _-----=> irqs-off
1489 # | / _----=> need-resched
1490 # || / _---=> hardirq/softirq
1491 # ||| / _--=> preempt-depth
1492 # |||| / delay
1493 # cmd pid ||||| time | caller
1494 # \ / ||||| \ | /
1495 <idle>-0 0d.s2 0us+: _raw_spin_lock_irq <-run_timer_softirq
1496 <idle>-0 0dNs3 17us : _raw_spin_unlock_irq <-run_timer_softirq
1497 <idle>-0 0dNs3 17us+: trace_hardirqs_on <-run_timer_softirq
1498 <idle>-0 0dNs3 25us : <stack trace>
1499 => _raw_spin_unlock_irq
1500 => run_timer_softirq
1501 => __do_softirq
1502 => call_softirq
1503 => do_softirq
1504 => irq_exit
1505 => smp_apic_timer_interrupt
1506 => apic_timer_interrupt
1507 => rcu_idle_exit
1508 => cpu_idle
1509 => rest_init
1510 => start_kernel
1511 => x86_64_start_reservations
1512 => x86_64_start_kernel
1514 Here we see that we had a latency of 16 microseconds (which is
1515 very good). The _raw_spin_lock_irq in run_timer_softirq disabled
1516 interrupts. The difference between the 16 and the displayed
1517 timestamp 25us occurred because the clock was incremented
1518 between the time of recording the max latency and the time of
1519 recording the function that had that latency.
1521 Note the above example had function-trace not set. If we set
1522 function-trace, we get a much larger output::
1524 with echo 1 > options/function-trace
1526 # tracer: irqsoff
1527 #
1528 # irqsoff latency trace v1.1.5 on 3.8.0-test+
1529 # --------------------------------------------------------------------
1530 # latency: 71 us, #168/168, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1531 # -----------------
1532 # | task: bash-2042 (uid:0 nice:0 policy:0 rt_prio:0)
1533 # -----------------
1534 # => started at: ata_scsi_queuecmd
1535 # => ended at: ata_scsi_queuecmd
1536 #
1537 #
1538 # _------=> CPU#
1539 # / _-----=> irqs-off
1540 # | / _----=> need-resched
1541 # || / _---=> hardirq/softirq
1542 # ||| / _--=> preempt-depth
1543 # |||| / delay
1544 # cmd pid ||||| time | caller
1545 # \ / ||||| \ | /
1546 bash-2042 3d... 0us : _raw_spin_lock_irqsave <-ata_scsi_queuecmd
1547 bash-2042 3d... 0us : add_preempt_count <-_raw_spin_lock_irqsave
1548 bash-2042 3d..1 1us : ata_scsi_find_dev <-ata_scsi_queuecmd
1549 bash-2042 3d..1 1us : __ata_scsi_find_dev <-ata_scsi_find_dev
1550 bash-2042 3d..1 2us : ata_find_dev.part.14 <-__ata_scsi_find_dev
1551 bash-2042 3d..1 2us : ata_qc_new_init <-__ata_scsi_queuecmd
1552 bash-2042 3d..1 3us : ata_sg_init <-__ata_scsi_queuecmd
1553 bash-2042 3d..1 4us : ata_scsi_rw_xlat <-__ata_scsi_queuecmd
1554 bash-2042 3d..1 4us : ata_build_rw_tf <-ata_scsi_rw_xlat
1555 [...]
1556 bash-2042 3d..1 67us : delay_tsc <-__delay
1557 bash-2042 3d..1 67us : add_preempt_count <-delay_tsc
1558 bash-2042 3d..2 67us : sub_preempt_count <-delay_tsc
1559 bash-2042 3d..1 67us : add_preempt_count <-delay_tsc
1560 bash-2042 3d..2 68us : sub_preempt_count <-delay_tsc
1561 bash-2042 3d..1 68us+: ata_bmdma_start <-ata_bmdma_qc_issue
1562 bash-2042 3d..1 71us : _raw_spin_unlock_irqrestore <-ata_scsi_queuecmd
1563 bash-2042 3d..1 71us : _raw_spin_unlock_irqrestore <-ata_scsi_queuecmd
1564 bash-2042 3d..1 72us+: trace_hardirqs_on <-ata_scsi_queuecmd
1565 bash-2042 3d..1 120us : <stack trace>
1566 => _raw_spin_unlock_irqrestore
1567 => ata_scsi_queuecmd
1568 => scsi_dispatch_cmd
1569 => scsi_request_fn
1570 => __blk_run_queue_uncond
1571 => __blk_run_queue
1572 => blk_queue_bio
1573 => submit_bio_noacct
1574 => submit_bio
1575 => submit_bh
1576 => __ext3_get_inode_loc
1577 => ext3_iget
1578 => ext3_lookup
1579 => lookup_real
1580 => __lookup_hash
1581 => walk_component
1582 => lookup_last
1583 => path_lookupat
1584 => filename_lookup
1585 => user_path_at_empty
1586 => user_path_at
1587 => vfs_fstatat
1588 => vfs_stat
1589 => sys_newstat
1590 => system_call_fastpath
1593 Here we traced a 71 microsecond latency. But we also see all the
1594 functions that were called during that time. Note that by
1595 enabling function tracing, we incur an added overhead. This
1596 overhead may extend the latency times. But nevertheless, this
1597 trace has provided some very helpful debugging information.
1599 If we prefer function graph output instead of function, we can set
1600 display-graph option::
1602 with echo 1 > options/display-graph
1604 # tracer: irqsoff
1605 #
1606 # irqsoff latency trace v1.1.5 on 4.20.0-rc6+
1607 # --------------------------------------------------------------------
1608 # latency: 3751 us, #274/274, CPU#0 | (M:desktop VP:0, KP:0, SP:0 HP:0 #P:4)
1609 # -----------------
1610 # | task: bash-1507 (uid:0 nice:0 policy:0 rt_prio:0)
1611 # -----------------
1612 # => started at: free_debug_processing
1613 # => ended at: return_to_handler
1614 #
1615 #
1616 # _-----=> irqs-off
1617 # / _----=> need-resched
1618 # | / _---=> hardirq/softirq
1619 # || / _--=> preempt-depth
1620 # ||| /
1621 # REL TIME CPU TASK/PID |||| DURATION FUNCTION CALLS
1622 # | | | | |||| | | | | | |
1623 0 us | 0) bash-1507 | d... | 0.000 us | _raw_spin_lock_irqsave();
1624 0 us | 0) bash-1507 | d..1 | 0.378 us | do_raw_spin_trylock();
1625 1 us | 0) bash-1507 | d..2 | | set_track() {
1626 2 us | 0) bash-1507 | d..2 | | save_stack_trace() {
1627 2 us | 0) bash-1507 | d..2 | | __save_stack_trace() {
1628 3 us | 0) bash-1507 | d..2 | | __unwind_start() {
1629 3 us | 0) bash-1507 | d..2 | | get_stack_info() {
1630 3 us | 0) bash-1507 | d..2 | 0.351 us | in_task_stack();
1631 4 us | 0) bash-1507 | d..2 | 1.107 us | }
1632 [...]
1633 3750 us | 0) bash-1507 | d..1 | 0.516 us | do_raw_spin_unlock();
1634 3750 us | 0) bash-1507 | d..1 | 0.000 us | _raw_spin_unlock_irqrestore();
1635 3764 us | 0) bash-1507 | d..1 | 0.000 us | tracer_hardirqs_on();
1636 bash-1507 0d..1 3792us : <stack trace>
1637 => free_debug_processing
1638 => __slab_free
1639 => kmem_cache_free
1640 => vm_area_free
1641 => remove_vma
1642 => exit_mmap
1643 => mmput
1644 => begin_new_exec
1645 => load_elf_binary
1646 => search_binary_handler
1647 => __do_execve_file.isra.32
1648 => __x64_sys_execve
1649 => do_syscall_64
1650 => entry_SYSCALL_64_after_hwframe
1652 preemptoff
1653 ----------
1655 When preemption is disabled, we may be able to receive
1656 interrupts but the task cannot be preempted and a higher
1657 priority task must wait for preemption to be enabled again
1658 before it can preempt a lower priority task.
1660 The preemptoff tracer traces the places that disable preemption.
1661 Like the irqsoff tracer, it records the maximum latency for
1662 which preemption was disabled. The control of preemptoff tracer
1663 is much like the irqsoff tracer.
1664 ::
1666 # echo 0 > options/function-trace
1667 # echo preemptoff > current_tracer
1668 # echo 1 > tracing_on
1669 # echo 0 > tracing_max_latency
1670 # ls -ltr
1671 [...]
1672 # echo 0 > tracing_on
1673 # cat trace
1674 # tracer: preemptoff
1675 #
1676 # preemptoff latency trace v1.1.5 on 3.8.0-test+
1677 # --------------------------------------------------------------------
1678 # latency: 46 us, #4/4, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1679 # -----------------
1680 # | task: sshd-1991 (uid:0 nice:0 policy:0 rt_prio:0)
1681 # -----------------
1682 # => started at: do_IRQ
1683 # => ended at: do_IRQ
1684 #
1685 #
1686 # _------=> CPU#
1687 # / _-----=> irqs-off
1688 # | / _----=> need-resched
1689 # || / _---=> hardirq/softirq
1690 # ||| / _--=> preempt-depth
1691 # |||| / delay
1692 # cmd pid ||||| time | caller
1693 # \ / ||||| \ | /
1694 sshd-1991 1d.h. 0us+: irq_enter <-do_IRQ
1695 sshd-1991 1d..1 46us : irq_exit <-do_IRQ
1696 sshd-1991 1d..1 47us+: trace_preempt_on <-do_IRQ
1697 sshd-1991 1d..1 52us : <stack trace>
1698 => sub_preempt_count
1699 => irq_exit
1700 => do_IRQ
1701 => ret_from_intr
1704 This has some more changes. Preemption was disabled when an
1705 interrupt came in (notice the 'h'), and was enabled on exit.
1706 But we also see that interrupts have been disabled when entering
1707 the preempt off section and leaving it (the 'd'). We do not know if
1708 interrupts were enabled in the mean time or shortly after this
1709 was over.
1710 ::
1712 # tracer: preemptoff
1713 #
1714 # preemptoff latency trace v1.1.5 on 3.8.0-test+
1715 # --------------------------------------------------------------------
1716 # latency: 83 us, #241/241, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1717 # -----------------
1718 # | task: bash-1994 (uid:0 nice:0 policy:0 rt_prio:0)
1719 # -----------------
1720 # => started at: wake_up_new_task
1721 # => ended at: task_rq_unlock
1722 #
1723 #
1724 # _------=> CPU#
1725 # / _-----=> irqs-off
1726 # | / _----=> need-resched
1727 # || / _---=> hardirq/softirq
1728 # ||| / _--=> preempt-depth
1729 # |||| / delay
1730 # cmd pid ||||| time | caller
1731 # \ / ||||| \ | /
1732 bash-1994 1d..1 0us : _raw_spin_lock_irqsave <-wake_up_new_task
1733 bash-1994 1d..1 0us : select_task_rq_fair <-select_task_rq
1734 bash-1994 1d..1 1us : __rcu_read_lock <-select_task_rq_fair
1735 bash-1994 1d..1 1us : source_load <-select_task_rq_fair
1736 bash-1994 1d..1 1us : source_load <-select_task_rq_fair
1737 [...]
1738 bash-1994 1d..1 12us : irq_enter <-smp_apic_timer_interrupt
1739 bash-1994 1d..1 12us : rcu_irq_enter <-irq_enter
1740 bash-1994 1d..1 13us : add_preempt_count <-irq_enter
1741 bash-1994 1d.h1 13us : exit_idle <-smp_apic_timer_interrupt
1742 bash-1994 1d.h1 13us : hrtimer_interrupt <-smp_apic_timer_interrupt
1743 bash-1994 1d.h1 13us : _raw_spin_lock <-hrtimer_interrupt
1744 bash-1994 1d.h1 14us : add_preempt_count <-_raw_spin_lock
1745 bash-1994 1d.h2 14us : ktime_get_update_offsets <-hrtimer_interrupt
1746 [...]
1747 bash-1994 1d.h1 35us : lapic_next_event <-clockevents_program_event
1748 bash-1994 1d.h1 35us : irq_exit <-smp_apic_timer_interrupt
1749 bash-1994 1d.h1 36us : sub_preempt_count <-irq_exit
1750 bash-1994 1d..2 36us : do_softirq <-irq_exit
1751 bash-1994 1d..2 36us : __do_softirq <-call_softirq
1752 bash-1994 1d..2 36us : __local_bh_disable <-__do_softirq
1753 bash-1994 1d.s2 37us : add_preempt_count <-_raw_spin_lock_irq
1754 bash-1994 1d.s3 38us : _raw_spin_unlock <-run_timer_softirq
1755 bash-1994 1d.s3 39us : sub_preempt_count <-_raw_spin_unlock
1756 bash-1994 1d.s2 39us : call_timer_fn <-run_timer_softirq
1757 [...]
1758 bash-1994 1dNs2 81us : cpu_needs_another_gp <-rcu_process_callbacks
1759 bash-1994 1dNs2 82us : __local_bh_enable <-__do_softirq
1760 bash-1994 1dNs2 82us : sub_preempt_count <-__local_bh_enable
1761 bash-1994 1dN.2 82us : idle_cpu <-irq_exit
1762 bash-1994 1dN.2 83us : rcu_irq_exit <-irq_exit
1763 bash-1994 1dN.2 83us : sub_preempt_count <-irq_exit
1764 bash-1994 1.N.1 84us : _raw_spin_unlock_irqrestore <-task_rq_unlock
1765 bash-1994 1.N.1 84us+: trace_preempt_on <-task_rq_unlock
1766 bash-1994 1.N.1 104us : <stack trace>
1767 => sub_preempt_count
1768 => _raw_spin_unlock_irqrestore
1769 => task_rq_unlock
1770 => wake_up_new_task
1771 => do_fork
1772 => sys_clone
1773 => stub_clone
1776 The above is an example of the preemptoff trace with
1777 function-trace set. Here we see that interrupts were not disabled
1778 the entire time. The irq_enter code lets us know that we entered
1779 an interrupt 'h'. Before that, the functions being traced still
1780 show that it is not in an interrupt, but we can see from the
1781 functions themselves that this is not the case.
1783 preemptirqsoff
1784 --------------
1786 Knowing the locations that have interrupts disabled or
1787 preemption disabled for the longest times is helpful. But
1788 sometimes we would like to know when either preemption and/or
1789 interrupts are disabled.
1791 Consider the following code::
1793 local_irq_disable();
1794 call_function_with_irqs_off();
1795 preempt_disable();
1796 call_function_with_irqs_and_preemption_off();
1797 local_irq_enable();
1798 call_function_with_preemption_off();
1799 preempt_enable();
1801 The irqsoff tracer will record the total length of
1802 call_function_with_irqs_off() and
1803 call_function_with_irqs_and_preemption_off().
1805 The preemptoff tracer will record the total length of
1806 call_function_with_irqs_and_preemption_off() and
1807 call_function_with_preemption_off().
1809 But neither will trace the time that interrupts and/or
1810 preemption is disabled. This total time is the time that we can
1811 not schedule. To record this time, use the preemptirqsoff
1812 tracer.
1814 Again, using this trace is much like the irqsoff and preemptoff
1815 tracers.
1816 ::
1818 # echo 0 > options/function-trace
1819 # echo preemptirqsoff > current_tracer
1820 # echo 1 > tracing_on
1821 # echo 0 > tracing_max_latency
1822 # ls -ltr
1823 [...]
1824 # echo 0 > tracing_on
1825 # cat trace
1826 # tracer: preemptirqsoff
1827 #
1828 # preemptirqsoff latency trace v1.1.5 on 3.8.0-test+
1829 # --------------------------------------------------------------------
1830 # latency: 100 us, #4/4, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1831 # -----------------
1832 # | task: ls-2230 (uid:0 nice:0 policy:0 rt_prio:0)
1833 # -----------------
1834 # => started at: ata_scsi_queuecmd
1835 # => ended at: ata_scsi_queuecmd
1836 #
1837 #
1838 # _------=> CPU#
1839 # / _-----=> irqs-off
1840 # | / _----=> need-resched
1841 # || / _---=> hardirq/softirq
1842 # ||| / _--=> preempt-depth
1843 # |||| / delay
1844 # cmd pid ||||| time | caller
1845 # \ / ||||| \ | /
1846 ls-2230 3d... 0us+: _raw_spin_lock_irqsave <-ata_scsi_queuecmd
1847 ls-2230 3...1 100us : _raw_spin_unlock_irqrestore <-ata_scsi_queuecmd
1848 ls-2230 3...1 101us+: trace_preempt_on <-ata_scsi_queuecmd
1849 ls-2230 3...1 111us : <stack trace>
1850 => sub_preempt_count
1851 => _raw_spin_unlock_irqrestore
1852 => ata_scsi_queuecmd
1853 => scsi_dispatch_cmd
1854 => scsi_request_fn
1855 => __blk_run_queue_uncond
1856 => __blk_run_queue
1857 => blk_queue_bio
1858 => submit_bio_noacct
1859 => submit_bio
1860 => submit_bh
1861 => ext3_bread
1862 => ext3_dir_bread
1863 => htree_dirblock_to_tree
1864 => ext3_htree_fill_tree
1865 => ext3_readdir
1866 => vfs_readdir
1867 => sys_getdents
1868 => system_call_fastpath
1871 The trace_hardirqs_off_thunk is called from assembly on x86 when
1872 interrupts are disabled in the assembly code. Without the
1873 function tracing, we do not know if interrupts were enabled
1874 within the preemption points. We do see that it started with
1875 preemption enabled.
1877 Here is a trace with function-trace set::
1879 # tracer: preemptirqsoff
1880 #
1881 # preemptirqsoff latency trace v1.1.5 on 3.8.0-test+
1882 # --------------------------------------------------------------------
1883 # latency: 161 us, #339/339, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1884 # -----------------
1885 # | task: ls-2269 (uid:0 nice:0 policy:0 rt_prio:0)
1886 # -----------------
1887 # => started at: schedule
1888 # => ended at: mutex_unlock
1889 #
1890 #
1891 # _------=> CPU#
1892 # / _-----=> irqs-off
1893 # | / _----=> need-resched
1894 # || / _---=> hardirq/softirq
1895 # ||| / _--=> preempt-depth
1896 # |||| / delay
1897 # cmd pid ||||| time | caller
1898 # \ / ||||| \ | /
1899 kworker/-59 3...1 0us : __schedule <-schedule
1900 kworker/-59 3d..1 0us : rcu_preempt_qs <-rcu_note_context_switch
1901 kworker/-59 3d..1 1us : add_preempt_count <-_raw_spin_lock_irq
1902 kworker/-59 3d..2 1us : deactivate_task <-__schedule
1903 kworker/-59 3d..2 1us : dequeue_task <-deactivate_task
1904 kworker/-59 3d..2 2us : update_rq_clock <-dequeue_task
1905 kworker/-59 3d..2 2us : dequeue_task_fair <-dequeue_task
1906 kworker/-59 3d..2 2us : update_curr <-dequeue_task_fair
1907 kworker/-59 3d..2 2us : update_min_vruntime <-update_curr
1908 kworker/-59 3d..2 3us : cpuacct_charge <-update_curr
1909 kworker/-59 3d..2 3us : __rcu_read_lock <-cpuacct_charge
1910 kworker/-59 3d..2 3us : __rcu_read_unlock <-cpuacct_charge
1911 kworker/-59 3d..2 3us : update_cfs_rq_blocked_load <-dequeue_task_fair
1912 kworker/-59 3d..2 4us : clear_buddies <-dequeue_task_fair
1913 kworker/-59 3d..2 4us : account_entity_dequeue <-dequeue_task_fair
1914 kworker/-59 3d..2 4us : update_min_vruntime <-dequeue_task_fair
1915 kworker/-59 3d..2 4us : update_cfs_shares <-dequeue_task_fair
1916 kworker/-59 3d..2 5us : hrtick_update <-dequeue_task_fair
1917 kworker/-59 3d..2 5us : wq_worker_sleeping <-__schedule
1918 kworker/-59 3d..2 5us : kthread_data <-wq_worker_sleeping
1919 kworker/-59 3d..2 5us : put_prev_task_fair <-__schedule
1920 kworker/-59 3d..2 6us : pick_next_task_fair <-pick_next_task
1921 kworker/-59 3d..2 6us : clear_buddies <-pick_next_task_fair
1922 kworker/-59 3d..2 6us : set_next_entity <-pick_next_task_fair
1923 kworker/-59 3d..2 6us : update_stats_wait_end <-set_next_entity
1924 ls-2269 3d..2 7us : finish_task_switch <-__schedule
1925 ls-2269 3d..2 7us : _raw_spin_unlock_irq <-finish_task_switch
1926 ls-2269 3d..2 8us : do_IRQ <-ret_from_intr
1927 ls-2269 3d..2 8us : irq_enter <-do_IRQ
1928 ls-2269 3d..2 8us : rcu_irq_enter <-irq_enter
1929 ls-2269 3d..2 9us : add_preempt_count <-irq_enter
1930 ls-2269 3d.h2 9us : exit_idle <-do_IRQ
1931 [...]
1932 ls-2269 3d.h3 20us : sub_preempt_count <-_raw_spin_unlock
1933 ls-2269 3d.h2 20us : irq_exit <-do_IRQ
1934 ls-2269 3d.h2 21us : sub_preempt_count <-irq_exit
1935 ls-2269 3d..3 21us : do_softirq <-irq_exit
1936 ls-2269 3d..3 21us : __do_softirq <-call_softirq
1937 ls-2269 3d..3 21us+: __local_bh_disable <-__do_softirq
1938 ls-2269 3d.s4 29us : sub_preempt_count <-_local_bh_enable_ip
1939 ls-2269 3d.s5 29us : sub_preempt_count <-_local_bh_enable_ip
1940 ls-2269 3d.s5 31us : do_IRQ <-ret_from_intr
1941 ls-2269 3d.s5 31us : irq_enter <-do_IRQ
1942 ls-2269 3d.s5 31us : rcu_irq_enter <-irq_enter
1943 [...]
1944 ls-2269 3d.s5 31us : rcu_irq_enter <-irq_enter
1945 ls-2269 3d.s5 32us : add_preempt_count <-irq_enter
1946 ls-2269 3d.H5 32us : exit_idle <-do_IRQ
1947 ls-2269 3d.H5 32us : handle_irq <-do_IRQ
1948 ls-2269 3d.H5 32us : irq_to_desc <-handle_irq
1949 ls-2269 3d.H5 33us : handle_fasteoi_irq <-handle_irq
1950 [...]
1951 ls-2269 3d.s5 158us : _raw_spin_unlock_irqrestore <-rtl8139_poll
1952 ls-2269 3d.s3 158us : net_rps_action_and_irq_enable.isra.65 <-net_rx_action
1953 ls-2269 3d.s3 159us : __local_bh_enable <-__do_softirq
1954 ls-2269 3d.s3 159us : sub_preempt_count <-__local_bh_enable
1955 ls-2269 3d..3 159us : idle_cpu <-irq_exit
1956 ls-2269 3d..3 159us : rcu_irq_exit <-irq_exit
1957 ls-2269 3d..3 160us : sub_preempt_count <-irq_exit
1958 ls-2269 3d... 161us : __mutex_unlock_slowpath <-mutex_unlock
1959 ls-2269 3d... 162us+: trace_hardirqs_on <-mutex_unlock
1960 ls-2269 3d... 186us : <stack trace>
1961 => __mutex_unlock_slowpath
1962 => mutex_unlock
1963 => process_output
1964 => n_tty_write
1965 => tty_write
1966 => vfs_write
1967 => sys_write
1968 => system_call_fastpath
1970 This is an interesting trace. It started with kworker running and
1971 scheduling out and ls taking over. But as soon as ls released the
1972 rq lock and enabled interrupts (but not preemption) an interrupt
1973 triggered. When the interrupt finished, it started running softirqs.
1974 But while the softirq was running, another interrupt triggered.
1975 When an interrupt is running inside a softirq, the annotation is 'H'.
1978 wakeup
1979 ------
1981 One common case that people are interested in tracing is the
1982 time it takes for a task that is woken to actually wake up.
1983 Now for non Real-Time tasks, this can be arbitrary. But tracing
1984 it nonetheless can be interesting.
1986 Without function tracing::
1988 # echo 0 > options/function-trace
1989 # echo wakeup > current_tracer
1990 # echo 1 > tracing_on
1991 # echo 0 > tracing_max_latency
1992 # chrt -f 5 sleep 1
1993 # echo 0 > tracing_on
1994 # cat trace
1995 # tracer: wakeup
1996 #
1997 # wakeup latency trace v1.1.5 on 3.8.0-test+
1998 # --------------------------------------------------------------------
1999 # latency: 15 us, #4/4, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
2000 # -----------------
2001 # | task: kworker/3:1H-312 (uid:0 nice:-20 policy:0 rt_prio:0)
2002 # -----------------
2003 #
2004 # _------=> CPU#
2005 # / _-----=> irqs-off
2006 # | / _----=> need-resched
2007 # || / _---=> hardirq/softirq
2008 # ||| / _--=> preempt-depth
2009 # |||| / delay
2010 # cmd pid ||||| time | caller
2011 # \ / ||||| \ | /
2012 <idle>-0 3dNs7 0us : 0:120:R + [003] 312:100:R kworker/3:1H
2013 <idle>-0 3dNs7 1us+: ttwu_do_activate.constprop.87 <-try_to_wake_up
2014 <idle>-0 3d..3 15us : __schedule <-schedule
2015 <idle>-0 3d..3 15us : 0:120:R ==> [003] 312:100:R kworker/3:1H
2017 The tracer only traces the highest priority task in the system
2018 to avoid tracing the normal circumstances. Here we see that
2019 the kworker with a nice priority of -20 (not very nice), took
2020 just 15 microseconds from the time it woke up, to the time it
2021 ran.
2023 Non Real-Time tasks are not that interesting. A more interesting
2024 trace is to concentrate only on Real-Time tasks.
2026 wakeup_rt
2027 ---------
2029 In a Real-Time environment it is very important to know the
2030 wakeup time it takes for the highest priority task that is woken
2031 up to the time that it executes. This is also known as "schedule
2032 latency". I stress the point that this is about RT tasks. It is
2033 also important to know the scheduling latency of non-RT tasks,
2034 but the average schedule latency is better for non-RT tasks.
2035 Tools like LatencyTop are more appropriate for such
2036 measurements.
2038 Real-Time environments are interested in the worst case latency.
2039 That is the longest latency it takes for something to happen,
2040 and not the average. We can have a very fast scheduler that may
2041 only have a large latency once in a while, but that would not
2042 work well with Real-Time tasks. The wakeup_rt tracer was designed
2043 to record the worst case wakeups of RT tasks. Non-RT tasks are
2044 not recorded because the tracer only records one worst case and
2045 tracing non-RT tasks that are unpredictable will overwrite the
2046 worst case latency of RT tasks (just run the normal wakeup
2047 tracer for a while to see that effect).
2049 Since this tracer only deals with RT tasks, we will run this
2050 slightly differently than we did with the previous tracers.
2051 Instead of performing an 'ls', we will run 'sleep 1' under
2052 'chrt' which changes the priority of the task.
2053 ::
2055 # echo 0 > options/function-trace
2056 # echo wakeup_rt > current_tracer
2057 # echo 1 > tracing_on
2058 # echo 0 > tracing_max_latency
2059 # chrt -f 5 sleep 1
2060 # echo 0 > tracing_on
2061 # cat trace
2062 # tracer: wakeup
2063 #
2064 # tracer: wakeup_rt
2065 #
2066 # wakeup_rt latency trace v1.1.5 on 3.8.0-test+
2067 # --------------------------------------------------------------------
2068 # latency: 5 us, #4/4, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
2069 # -----------------
2070 # | task: sleep-2389 (uid:0 nice:0 policy:1 rt_prio:5)
2071 # -----------------
2072 #
2073 # _------=> CPU#
2074 # / _-----=> irqs-off
2075 # | / _----=> need-resched
2076 # || / _---=> hardirq/softirq
2077 # ||| / _--=> preempt-depth
2078 # |||| / delay
2079 # cmd pid ||||| time | caller
2080 # \ / ||||| \ | /
2081 <idle>-0 3d.h4 0us : 0:120:R + [003] 2389: 94:R sleep
2082 <idle>-0 3d.h4 1us+: ttwu_do_activate.constprop.87 <-try_to_wake_up
2083 <idle>-0 3d..3 5us : __schedule <-schedule
2084 <idle>-0 3d..3 5us : 0:120:R ==> [003] 2389: 94:R sleep
2087 Running this on an idle system, we see that it only took 5 microseconds
2088 to perform the task switch. Note, since the trace point in the schedule
2089 is before the actual "switch", we stop the tracing when the recorded task
2090 is about to schedule in. This may change if we add a new marker at the
2091 end of the scheduler.
2093 Notice that the recorded task is 'sleep' with the PID of 2389
2094 and it has an rt_prio of 5. This priority is user-space priority
2095 and not the internal kernel priority. The policy is 1 for
2096 SCHED_FIFO and 2 for SCHED_RR.
2098 Note, that the trace data shows the internal priority (99 - rtprio).
2099 ::
2101 <idle>-0 3d..3 5us : 0:120:R ==> [003] 2389: 94:R sleep
2103 The 0:120:R means idle was running with a nice priority of 0 (120 - 120)
2104 and in the running state 'R'. The sleep task was scheduled in with
2105 2389: 94:R. That is the priority is the kernel rtprio (99 - 5 = 94)
2106 and it too is in the running state.
2108 Doing the same with chrt -r 5 and function-trace set.
2109 ::
2111 echo 1 > options/function-trace
2113 # tracer: wakeup_rt
2114 #
2115 # wakeup_rt latency trace v1.1.5 on 3.8.0-test+
2116 # --------------------------------------------------------------------
2117 # latency: 29 us, #85/85, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
2118 # -----------------
2119 # | task: sleep-2448 (uid:0 nice:0 policy:1 rt_prio:5)
2120 # -----------------
2121 #
2122 # _------=> CPU#
2123 # / _-----=> irqs-off
2124 # | / _----=> need-resched
2125 # || / _---=> hardirq/softirq
2126 # ||| / _--=> preempt-depth
2127 # |||| / delay
2128 # cmd pid ||||| time | caller
2129 # \ / ||||| \ | /
2130 <idle>-0 3d.h4 1us+: 0:120:R + [003] 2448: 94:R sleep
2131 <idle>-0 3d.h4 2us : ttwu_do_activate.constprop.87 <-try_to_wake_up
2132 <idle>-0 3d.h3 3us : check_preempt_curr <-ttwu_do_wakeup
2133 <idle>-0 3d.h3 3us : resched_curr <-check_preempt_curr
2134 <idle>-0 3dNh3 4us : task_woken_rt <-ttwu_do_wakeup
2135 <idle>-0 3dNh3 4us : _raw_spin_unlock <-try_to_wake_up
2136 <idle>-0 3dNh3 4us : sub_preempt_count <-_raw_spin_unlock
2137 <idle>-0 3dNh2 5us : ttwu_stat <-try_to_wake_up
2138 <idle>-0 3dNh2 5us : _raw_spin_unlock_irqrestore <-try_to_wake_up
2139 <idle>-0 3dNh2 6us : sub_preempt_count <-_raw_spin_unlock_irqrestore
2140 <idle>-0 3dNh1 6us : _raw_spin_lock <-__run_hrtimer
2141 <idle>-0 3dNh1 6us : add_preempt_count <-_raw_spin_lock
2142 <idle>-0 3dNh2 7us : _raw_spin_unlock <-hrtimer_interrupt
2143 <idle>-0 3dNh2 7us : sub_preempt_count <-_raw_spin_unlock
2144 <idle>-0 3dNh1 7us : tick_program_event <-hrtimer_interrupt
2145 <idle>-0 3dNh1 7us : clockevents_program_event <-tick_program_event
2146 <idle>-0 3dNh1 8us : ktime_get <-clockevents_program_event
2147 <idle>-0 3dNh1 8us : lapic_next_event <-clockevents_program_event
2148 <idle>-0 3dNh1 8us : irq_exit <-smp_apic_timer_interrupt
2149 <idle>-0 3dNh1 9us : sub_preempt_count <-irq_exit
2150 <idle>-0 3dN.2 9us : idle_cpu <-irq_exit
2151 <idle>-0 3dN.2 9us : rcu_irq_exit <-irq_exit
2152 <idle>-0 3dN.2 10us : rcu_eqs_enter_common.isra.45 <-rcu_irq_exit
2153 <idle>-0 3dN.2 10us : sub_preempt_count <-irq_exit
2154 <idle>-0 3.N.1 11us : rcu_idle_exit <-cpu_idle
2155 <idle>-0 3dN.1 11us : rcu_eqs_exit_common.isra.43 <-rcu_idle_exit
2156 <idle>-0 3.N.1 11us : tick_nohz_idle_exit <-cpu_idle
2157 <idle>-0 3dN.1 12us : menu_hrtimer_cancel <-tick_nohz_idle_exit
2158 <idle>-0 3dN.1 12us : ktime_get <-tick_nohz_idle_exit
2159 <idle>-0 3dN.1 12us : tick_do_update_jiffies64 <-tick_nohz_idle_exit
2160 <idle>-0 3dN.1 13us : cpu_load_update_nohz <-tick_nohz_idle_exit
2161 <idle>-0 3dN.1 13us : _raw_spin_lock <-cpu_load_update_nohz
2162 <idle>-0 3dN.1 13us : add_preempt_count <-_raw_spin_lock
2163 <idle>-0 3dN.2 13us : __cpu_load_update <-cpu_load_update_nohz
2164 <idle>-0 3dN.2 14us : sched_avg_update <-__cpu_load_update
2165 <idle>-0 3dN.2 14us : _raw_spin_unlock <-cpu_load_update_nohz
2166 <idle>-0 3dN.2 14us : sub_preempt_count <-_raw_spin_unlock
2167 <idle>-0 3dN.1 15us : calc_load_nohz_stop <-tick_nohz_idle_exit
2168 <idle>-0 3dN.1 15us : touch_softlockup_watchdog <-tick_nohz_idle_exit
2169 <idle>-0 3dN.1 15us : hrtimer_cancel <-tick_nohz_idle_exit
2170 <idle>-0 3dN.1 15us : hrtimer_try_to_cancel <-hrtimer_cancel
2171 <idle>-0 3dN.1 16us : lock_hrtimer_base.isra.18 <-hrtimer_try_to_cancel
2172 <idle>-0 3dN.1 16us : _raw_spin_lock_irqsave <-lock_hrtimer_base.isra.18
2173 <idle>-0 3dN.1 16us : add_preempt_count <-_raw_spin_lock_irqsave
2174 <idle>-0 3dN.2 17us : __remove_hrtimer <-remove_hrtimer.part.16
2175 <idle>-0 3dN.2 17us : hrtimer_force_reprogram <-__remove_hrtimer
2176 <idle>-0 3dN.2 17us : tick_program_event <-hrtimer_force_reprogram
2177 <idle>-0 3dN.2 18us : clockevents_program_event <-tick_program_event
2178 <idle>-0 3dN.2 18us : ktime_get <-clockevents_program_event
2179 <idle>-0 3dN.2 18us : lapic_next_event <-clockevents_program_event
2180 <idle>-0 3dN.2 19us : _raw_spin_unlock_irqrestore <-hrtimer_try_to_cancel
2181 <idle>-0 3dN.2 19us : sub_preempt_count <-_raw_spin_unlock_irqrestore
2182 <idle>-0 3dN.1 19us : hrtimer_forward <-tick_nohz_idle_exit
2183 <idle>-0 3dN.1 20us : ktime_add_safe <-hrtimer_forward
2184 <idle>-0 3dN.1 20us : ktime_add_safe <-hrtimer_forward
2185 <idle>-0 3dN.1 20us : hrtimer_start_range_ns <-hrtimer_start_expires.constprop.11
2186 <idle>-0 3dN.1 20us : __hrtimer_start_range_ns <-hrtimer_start_range_ns
2187 <idle>-0 3dN.1 21us : lock_hrtimer_base.isra.18 <-__hrtimer_start_range_ns
2188 <idle>-0 3dN.1 21us : _raw_spin_lock_irqsave <-lock_hrtimer_base.isra.18
2189 <idle>-0 3dN.1 21us : add_preempt_count <-_raw_spin_lock_irqsave
2190 <idle>-0 3dN.2 22us : ktime_add_safe <-__hrtimer_start_range_ns
2191 <idle>-0 3dN.2 22us : enqueue_hrtimer <-__hrtimer_start_range_ns
2192 <idle>-0 3dN.2 22us : tick_program_event <-__hrtimer_start_range_ns
2193 <idle>-0 3dN.2 23us : clockevents_program_event <-tick_program_event
2194 <idle>-0 3dN.2 23us : ktime_get <-clockevents_program_event
2195 <idle>-0 3dN.2 23us : lapic_next_event <-clockevents_program_event
2196 <idle>-0 3dN.2 24us : _raw_spin_unlock_irqrestore <-__hrtimer_start_range_ns
2197 <idle>-0 3dN.2 24us : sub_preempt_count <-_raw_spin_unlock_irqrestore
2198 <idle>-0 3dN.1 24us : account_idle_ticks <-tick_nohz_idle_exit
2199 <idle>-0 3dN.1 24us : account_idle_time <-account_idle_ticks
2200 <idle>-0 3.N.1 25us : sub_preempt_count <-cpu_idle
2201 <idle>-0 3.N.. 25us : schedule <-cpu_idle
2202 <idle>-0 3.N.. 25us : __schedule <-preempt_schedule
2203 <idle>-0 3.N.. 26us : add_preempt_count <-__schedule
2204 <idle>-0 3.N.1 26us : rcu_note_context_switch <-__schedule
2205 <idle>-0 3.N.1 26us : rcu_sched_qs <-rcu_note_context_switch
2206 <idle>-0 3dN.1 27us : rcu_preempt_qs <-rcu_note_context_switch
2207 <idle>-0 3.N.1 27us : _raw_spin_lock_irq <-__schedule
2208 <idle>-0 3dN.1 27us : add_preempt_count <-_raw_spin_lock_irq
2209 <idle>-0 3dN.2 28us : put_prev_task_idle <-__schedule
2210 <idle>-0 3dN.2 28us : pick_next_task_stop <-pick_next_task
2211 <idle>-0 3dN.2 28us : pick_next_task_rt <-pick_next_task
2212 <idle>-0 3dN.2 29us : dequeue_pushable_task <-pick_next_task_rt
2213 <idle>-0 3d..3 29us : __schedule <-preempt_schedule
2214 <idle>-0 3d..3 30us : 0:120:R ==> [003] 2448: 94:R sleep
2216 This isn't that big of a trace, even with function tracing enabled,
2217 so I included the entire trace.
2219 The interrupt went off while when the system was idle. Somewhere
2220 before task_woken_rt() was called, the NEED_RESCHED flag was set,
2221 this is indicated by the first occurrence of the 'N' flag.
2223 Latency tracing and events
2224 --------------------------
2225 As function tracing can induce a much larger latency, but without
2226 seeing what happens within the latency it is hard to know what
2227 caused it. There is a middle ground, and that is with enabling
2228 events.
2229 ::
2231 # echo 0 > options/function-trace
2232 # echo wakeup_rt > current_tracer
2233 # echo 1 > events/enable
2234 # echo 1 > tracing_on
2235 # echo 0 > tracing_max_latency
2236 # chrt -f 5 sleep 1
2237 # echo 0 > tracing_on
2238 # cat trace
2239 # tracer: wakeup_rt
2240 #
2241 # wakeup_rt latency trace v1.1.5 on 3.8.0-test+
2242 # --------------------------------------------------------------------
2243 # latency: 6 us, #12/12, CPU#2 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
2244 # -----------------
2245 # | task: sleep-5882 (uid:0 nice:0 policy:1 rt_prio:5)
2246 # -----------------
2247 #
2248 # _------=> CPU#
2249 # / _-----=> irqs-off
2250 # | / _----=> need-resched
2251 # || / _---=> hardirq/softirq
2252 # ||| / _--=> preempt-depth
2253 # |||| / delay
2254 # cmd pid ||||| time | caller
2255 # \ / ||||| \ | /
2256 <idle>-0 2d.h4 0us : 0:120:R + [002] 5882: 94:R sleep
2257 <idle>-0 2d.h4 0us : ttwu_do_activate.constprop.87 <-try_to_wake_up
2258 <idle>-0 2d.h4 1us : sched_wakeup: comm=sleep pid=5882 prio=94 success=1 target_cpu=002
2259 <idle>-0 2dNh2 1us : hrtimer_expire_exit: hrtimer=ffff88007796feb8
2260 <idle>-0 2.N.2 2us : power_end: cpu_id=2
2261 <idle>-0 2.N.2 3us : cpu_idle: state=4294967295 cpu_id=2
2262 <idle>-0 2dN.3 4us : hrtimer_cancel: hrtimer=ffff88007d50d5e0
2263 <idle>-0 2dN.3 4us : hrtimer_start: hrtimer=ffff88007d50d5e0 function=tick_sched_timer expires=34311211000000 softexpires=34311211000000
2264 <idle>-0 2.N.2 5us : rcu_utilization: Start context switch
2265 <idle>-0 2.N.2 5us : rcu_utilization: End context switch
2266 <idle>-0 2d..3 6us : __schedule <-schedule
2267 <idle>-0 2d..3 6us : 0:120:R ==> [002] 5882: 94:R sleep
2270 Hardware Latency Detector
2271 -------------------------
2273 The hardware latency detector is executed by enabling the "hwlat" tracer.
2275 NOTE, this tracer will affect the performance of the system as it will
2276 periodically make a CPU constantly busy with interrupts disabled.
2277 ::
2279 # echo hwlat > current_tracer
2280 # sleep 100
2281 # cat trace
2282 # tracer: hwlat
2283 #
2284 # entries-in-buffer/entries-written: 13/13 #P:8
2285 #
2286 # _-----=> irqs-off
2287 # / _----=> need-resched
2288 # | / _---=> hardirq/softirq
2289 # || / _--=> preempt-depth
2290 # ||| / delay
2291 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
2292 # | | | |||| | |
2293 <...>-1729 [001] d... 678.473449: #1 inner/outer(us): 11/12 ts:1581527483.343962693 count:6
2294 <...>-1729 [004] d... 689.556542: #2 inner/outer(us): 16/9 ts:1581527494.889008092 count:1
2295 <...>-1729 [005] d... 714.756290: #3 inner/outer(us): 16/16 ts:1581527519.678961629 count:5
2296 <...>-1729 [001] d... 718.788247: #4 inner/outer(us): 9/17 ts:1581527523.889012713 count:1
2297 <...>-1729 [002] d... 719.796341: #5 inner/outer(us): 13/9 ts:1581527524.912872606 count:1
2298 <...>-1729 [006] d... 844.787091: #6 inner/outer(us): 9/12 ts:1581527649.889048502 count:2
2299 <...>-1729 [003] d... 849.827033: #7 inner/outer(us): 18/9 ts:1581527654.889013793 count:1
2300 <...>-1729 [007] d... 853.859002: #8 inner/outer(us): 9/12 ts:1581527658.889065736 count:1
2301 <...>-1729 [001] d... 855.874978: #9 inner/outer(us): 9/11 ts:1581527660.861991877 count:1
2302 <...>-1729 [001] d... 863.938932: #10 inner/outer(us): 9/11 ts:1581527668.970010500 count:1 nmi-total:7 nmi-count:1
2303 <...>-1729 [007] d... 878.050780: #11 inner/outer(us): 9/12 ts:1581527683.385002600 count:1 nmi-total:5 nmi-count:1
2304 <...>-1729 [007] d... 886.114702: #12 inner/outer(us): 9/12 ts:1581527691.385001600 count:1
2307 The above output is somewhat the same in the header. All events will have
2308 interrupts disabled 'd'. Under the FUNCTION title there is:
2310 #1
2311 This is the count of events recorded that were greater than the
2312 tracing_threshold (See below).
2314 inner/outer(us): 11/11
2316 This shows two numbers as "inner latency" and "outer latency". The test
2317 runs in a loop checking a timestamp twice. The latency detected within
2318 the two timestamps is the "inner latency" and the latency detected
2319 after the previous timestamp and the next timestamp in the loop is
2320 the "outer latency".
2322 ts:1581527483.343962693
2324 The absolute timestamp that the first latency was recorded in the window.
2326 count:6
2328 The number of times a latency was detected during the window.
2330 nmi-total:7 nmi-count:1
2332 On architectures that support it, if an NMI comes in during the
2333 test, the time spent in NMI is reported in "nmi-total" (in
2334 microseconds).
2336 All architectures that have NMIs will show the "nmi-count" if an
2337 NMI comes in during the test.
2339 hwlat files:
2341 tracing_threshold
2342 This gets automatically set to "10" to represent 10
2343 microseconds. This is the threshold of latency that
2344 needs to be detected before the trace will be recorded.
2346 Note, when hwlat tracer is finished (another tracer is
2347 written into "current_tracer"), the original value for
2348 tracing_threshold is placed back into this file.
2350 hwlat_detector/width
2351 The length of time the test runs with interrupts disabled.
2353 hwlat_detector/window
2354 The length of time of the window which the test
2355 runs. That is, the test will run for "width"
2356 microseconds per "window" microseconds
2358 tracing_cpumask
2359 When the test is started. A kernel thread is created that
2360 runs the test. This thread will alternate between CPUs
2361 listed in the tracing_cpumask between each period
2362 (one "window"). To limit the test to specific CPUs
2363 set the mask in this file to only the CPUs that the test
2364 should run on.
2366 function
2367 --------
2369 This tracer is the function tracer. Enabling the function tracer
2370 can be done from the debug file system. Make sure the
2371 ftrace_enabled is set; otherwise this tracer is a nop.
2372 See the "ftrace_enabled" section below.
2373 ::
2375 # sysctl kernel.ftrace_enabled=1
2376 # echo function > current_tracer
2377 # echo 1 > tracing_on
2378 # usleep 1
2379 # echo 0 > tracing_on
2380 # cat trace
2381 # tracer: function
2382 #
2383 # entries-in-buffer/entries-written: 24799/24799 #P:4
2384 #
2385 # _-----=> irqs-off
2386 # / _----=> need-resched
2387 # | / _---=> hardirq/softirq
2388 # || / _--=> preempt-depth
2389 # ||| / delay
2390 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
2391 # | | | |||| | |
2392 bash-1994 [002] .... 3082.063030: mutex_unlock <-rb_simple_write
2393 bash-1994 [002] .... 3082.063031: __mutex_unlock_slowpath <-mutex_unlock
2394 bash-1994 [002] .... 3082.063031: __fsnotify_parent <-fsnotify_modify
2395 bash-1994 [002] .... 3082.063032: fsnotify <-fsnotify_modify
2396 bash-1994 [002] .... 3082.063032: __srcu_read_lock <-fsnotify
2397 bash-1994 [002] .... 3082.063032: add_preempt_count <-__srcu_read_lock
2398 bash-1994 [002] ...1 3082.063032: sub_preempt_count <-__srcu_read_lock
2399 bash-1994 [002] .... 3082.063033: __srcu_read_unlock <-fsnotify
2400 [...]
2403 Note: function tracer uses ring buffers to store the above
2404 entries. The newest data may overwrite the oldest data.
2405 Sometimes using echo to stop the trace is not sufficient because
2406 the tracing could have overwritten the data that you wanted to
2407 record. For this reason, it is sometimes better to disable
2408 tracing directly from a program. This allows you to stop the
2409 tracing at the point that you hit the part that you are
2410 interested in. To disable the tracing directly from a C program,
2411 something like following code snippet can be used::
2413 int trace_fd;
2414 [...]
2415 int main(int argc, char *argv[]) {
2416 [...]
2417 trace_fd = open(tracing_file("tracing_on"), O_WRONLY);
2418 [...]
2419 if (condition_hit()) {
2420 write(trace_fd, "0", 1);
2421 }
2422 [...]
2423 }
2426 Single thread tracing
2427 ---------------------
2429 By writing into set_ftrace_pid you can trace a
2430 single thread. For example::
2432 # cat set_ftrace_pid
2433 no pid
2434 # echo 3111 > set_ftrace_pid
2435 # cat set_ftrace_pid
2436 3111
2437 # echo function > current_tracer
2438 # cat trace | head
2439 # tracer: function
2440 #
2441 # TASK-PID CPU# TIMESTAMP FUNCTION
2442 # | | | | |
2443 yum-updatesd-3111 [003] 1637.254676: finish_task_switch <-thread_return
2444 yum-updatesd-3111 [003] 1637.254681: hrtimer_cancel <-schedule_hrtimeout_range
2445 yum-updatesd-3111 [003] 1637.254682: hrtimer_try_to_cancel <-hrtimer_cancel
2446 yum-updatesd-3111 [003] 1637.254683: lock_hrtimer_base <-hrtimer_try_to_cancel
2447 yum-updatesd-3111 [003] 1637.254685: fget_light <-do_sys_poll
2448 yum-updatesd-3111 [003] 1637.254686: pipe_poll <-do_sys_poll
2449 # echo > set_ftrace_pid
2450 # cat trace |head
2451 # tracer: function
2452 #
2453 # TASK-PID CPU# TIMESTAMP FUNCTION
2454 # | | | | |
2455 ##### CPU 3 buffer started ####
2456 yum-updatesd-3111 [003] 1701.957688: free_poll_entry <-poll_freewait
2457 yum-updatesd-3111 [003] 1701.957689: remove_wait_queue <-free_poll_entry
2458 yum-updatesd-3111 [003] 1701.957691: fput <-free_poll_entry
2459 yum-updatesd-3111 [003] 1701.957692: audit_syscall_exit <-sysret_audit
2460 yum-updatesd-3111 [003] 1701.957693: path_put <-audit_syscall_exit
2462 If you want to trace a function when executing, you could use
2463 something like this simple program.
2464 ::
2466 #include <stdio.h>
2467 #include <stdlib.h>
2468 #include <sys/types.h>
2469 #include <sys/stat.h>
2470 #include <fcntl.h>
2471 #include <unistd.h>
2472 #include <string.h>
2474 #define _STR(x) #x
2475 #define STR(x) _STR(x)
2476 #define MAX_PATH 256
2478 const char *find_tracefs(void)
2479 {
2480 static char tracefs[MAX_PATH+1];
2481 static int tracefs_found;
2482 char type[100];
2483 FILE *fp;
2485 if (tracefs_found)
2486 return tracefs;
2488 if ((fp = fopen("/proc/mounts","r")) == NULL) {
2489 perror("/proc/mounts");
2490 return NULL;
2491 }
2493 while (fscanf(fp, "%*s %"
2494 STR(MAX_PATH)
2495 "s %99s %*s %*d %*d\n",
2496 tracefs, type) == 2) {
2497 if (strcmp(type, "tracefs") == 0)
2498 break;
2499 }
2500 fclose(fp);
2502 if (strcmp(type, "tracefs") != 0) {
2503 fprintf(stderr, "tracefs not mounted");
2504 return NULL;
2505 }
2507 strcat(tracefs, "/tracing/");
2508 tracefs_found = 1;
2510 return tracefs;
2511 }
2513 const char *tracing_file(const char *file_name)
2514 {
2515 static char trace_file[MAX_PATH+1];
2516 snprintf(trace_file, MAX_PATH, "%s/%s", find_tracefs(), file_name);
2517 return trace_file;
2518 }
2520 int main (int argc, char **argv)
2521 {
2522 if (argc < 1)
2523 exit(-1);
2525 if (fork() > 0) {
2526 int fd, ffd;
2527 char line[64];
2528 int s;
2530 ffd = open(tracing_file("current_tracer"), O_WRONLY);
2531 if (ffd < 0)
2532 exit(-1);
2533 write(ffd, "nop", 3);
2535 fd = open(tracing_file("set_ftrace_pid"), O_WRONLY);
2536 s = sprintf(line, "%d\n", getpid());
2537 write(fd, line, s);
2539 write(ffd, "function", 8);
2541 close(fd);
2542 close(ffd);
2544 execvp(argv[1], argv+1);
2545 }
2547 return 0;
2548 }
2550 Or this simple script!
2551 ::
2553 #!/bin/bash
2555 tracefs=`sed -ne 's/^tracefs \(.*\) tracefs.*/\1/p' /proc/mounts`
2556 echo 0 > $tracefs/tracing_on
2557 echo $$ > $tracefs/set_ftrace_pid
2558 echo function > $tracefs/current_tracer
2559 echo 1 > $tracefs/tracing_on
2560 exec "$@"
2563 function graph tracer
2564 ---------------------------
2566 This tracer is similar to the function tracer except that it
2567 probes a function on its entry and its exit. This is done by
2568 using a dynamically allocated stack of return addresses in each
2569 task_struct. On function entry the tracer overwrites the return
2570 address of each function traced to set a custom probe. Thus the
2571 original return address is stored on the stack of return address
2572 in the task_struct.
2574 Probing on both ends of a function leads to special features
2575 such as:
2577 - measure of a function's time execution
2578 - having a reliable call stack to draw function calls graph
2580 This tracer is useful in several situations:
2582 - you want to find the reason of a strange kernel behavior and
2583 need to see what happens in detail on any areas (or specific
2584 ones).
2586 - you are experiencing weird latencies but it's difficult to
2587 find its origin.
2589 - you want to find quickly which path is taken by a specific
2590 function
2592 - you just want to peek inside a working kernel and want to see
2593 what happens there.
2595 ::
2597 # tracer: function_graph
2598 #
2599 # CPU DURATION FUNCTION CALLS
2600 # | | | | | | |
2602 0) | sys_open() {
2603 0) | do_sys_open() {
2604 0) | getname() {
2605 0) | kmem_cache_alloc() {
2606 0) 1.382 us | __might_sleep();
2607 0) 2.478 us | }
2608 0) | strncpy_from_user() {
2609 0) | might_fault() {
2610 0) 1.389 us | __might_sleep();
2611 0) 2.553 us | }
2612 0) 3.807 us | }
2613 0) 7.876 us | }
2614 0) | alloc_fd() {
2615 0) 0.668 us | _spin_lock();
2616 0) 0.570 us | expand_files();
2617 0) 0.586 us | _spin_unlock();
2620 There are several columns that can be dynamically
2621 enabled/disabled. You can use every combination of options you
2622 want, depending on your needs.
2624 - The cpu number on which the function executed is default
2625 enabled. It is sometimes better to only trace one cpu (see
2626 tracing_cpumask file) or you might sometimes see unordered
2627 function calls while cpu tracing switch.
2629 - hide: echo nofuncgraph-cpu > trace_options
2630 - show: echo funcgraph-cpu > trace_options
2632 - The duration (function's time of execution) is displayed on
2633 the closing bracket line of a function or on the same line
2634 than the current function in case of a leaf one. It is default
2635 enabled.
2637 - hide: echo nofuncgraph-duration > trace_options
2638 - show: echo funcgraph-duration > trace_options
2640 - The overhead field precedes the duration field in case of
2641 reached duration thresholds.
2643 - hide: echo nofuncgraph-overhead > trace_options
2644 - show: echo funcgraph-overhead > trace_options
2645 - depends on: funcgraph-duration
2647 ie::
2649 3) # 1837.709 us | } /* __switch_to */
2650 3) | finish_task_switch() {
2651 3) 0.313 us | _raw_spin_unlock_irq();
2652 3) 3.177 us | }
2653 3) # 1889.063 us | } /* __schedule */
2654 3) ! 140.417 us | } /* __schedule */
2655 3) # 2034.948 us | } /* schedule */
2656 3) * 33998.59 us | } /* schedule_preempt_disabled */
2658 [...]
2660 1) 0.260 us | msecs_to_jiffies();
2661 1) 0.313 us | __rcu_read_unlock();
2662 1) + 61.770 us | }
2663 1) + 64.479 us | }
2664 1) 0.313 us | rcu_bh_qs();
2665 1) 0.313 us | __local_bh_enable();
2666 1) ! 217.240 us | }
2667 1) 0.365 us | idle_cpu();
2668 1) | rcu_irq_exit() {
2669 1) 0.417 us | rcu_eqs_enter_common.isra.47();
2670 1) 3.125 us | }
2671 1) ! 227.812 us | }
2672 1) ! 457.395 us | }
2673 1) @ 119760.2 us | }
2675 [...]
2677 2) | handle_IPI() {
2678 1) 6.979 us | }
2679 2) 0.417 us | scheduler_ipi();
2680 1) 9.791 us | }
2681 1) + 12.917 us | }
2682 2) 3.490 us | }
2683 1) + 15.729 us | }
2684 1) + 18.542 us | }
2685 2) $ 3594274 us | }
2687 Flags::
2689 + means that the function exceeded 10 usecs.
2690 ! means that the function exceeded 100 usecs.
2691 # means that the function exceeded 1000 usecs.
2692 * means that the function exceeded 10 msecs.
2693 @ means that the function exceeded 100 msecs.
2694 $ means that the function exceeded 1 sec.
2697 - The task/pid field displays the thread cmdline and pid which
2698 executed the function. It is default disabled.
2700 - hide: echo nofuncgraph-proc > trace_options
2701 - show: echo funcgraph-proc > trace_options
2703 ie::
2705 # tracer: function_graph
2706 #
2707 # CPU TASK/PID DURATION FUNCTION CALLS
2708 # | | | | | | | | |
2709 0) sh-4802 | | d_free() {
2710 0) sh-4802 | | call_rcu() {
2711 0) sh-4802 | | __call_rcu() {
2712 0) sh-4802 | 0.616 us | rcu_process_gp_end();
2713 0) sh-4802 | 0.586 us | check_for_new_grace_period();
2714 0) sh-4802 | 2.899 us | }
2715 0) sh-4802 | 4.040 us | }
2716 0) sh-4802 | 5.151 us | }
2717 0) sh-4802 | + 49.370 us | }
2720 - The absolute time field is an absolute timestamp given by the
2721 system clock since it started. A snapshot of this time is
2722 given on each entry/exit of functions
2724 - hide: echo nofuncgraph-abstime > trace_options
2725 - show: echo funcgraph-abstime > trace_options
2727 ie::
2729 #
2730 # TIME CPU DURATION FUNCTION CALLS
2731 # | | | | | | | |
2732 360.774522 | 1) 0.541 us | }
2733 360.774522 | 1) 4.663 us | }
2734 360.774523 | 1) 0.541 us | __wake_up_bit();
2735 360.774524 | 1) 6.796 us | }
2736 360.774524 | 1) 7.952 us | }
2737 360.774525 | 1) 9.063 us | }
2738 360.774525 | 1) 0.615 us | journal_mark_dirty();
2739 360.774527 | 1) 0.578 us | __brelse();
2740 360.774528 | 1) | reiserfs_prepare_for_journal() {
2741 360.774528 | 1) | unlock_buffer() {
2742 360.774529 | 1) | wake_up_bit() {
2743 360.774529 | 1) | bit_waitqueue() {
2744 360.774530 | 1) 0.594 us | __phys_addr();
2747 The function name is always displayed after the closing bracket
2748 for a function if the start of that function is not in the
2749 trace buffer.
2751 Display of the function name after the closing bracket may be
2752 enabled for functions whose start is in the trace buffer,
2753 allowing easier searching with grep for function durations.
2754 It is default disabled.
2756 - hide: echo nofuncgraph-tail > trace_options
2757 - show: echo funcgraph-tail > trace_options
2759 Example with nofuncgraph-tail (default)::
2761 0) | putname() {
2762 0) | kmem_cache_free() {
2763 0) 0.518 us | __phys_addr();
2764 0) 1.757 us | }
2765 0) 2.861 us | }
2767 Example with funcgraph-tail::
2769 0) | putname() {
2770 0) | kmem_cache_free() {
2771 0) 0.518 us | __phys_addr();
2772 0) 1.757 us | } /* kmem_cache_free() */
2773 0) 2.861 us | } /* putname() */
2775 The return value of each traced function can be displayed after
2776 an equal sign "=". When encountering system call failures, it
2777 can be very helpful to quickly locate the function that first
2778 returns an error code.
2780 - hide: echo nofuncgraph-retval > trace_options
2781 - show: echo funcgraph-retval > trace_options
2783 Example with funcgraph-retval::
2785 1) | cgroup_migrate() {
2786 1) 0.651 us | cgroup_migrate_add_task(); /* = 0xffff93fcfd346c00 */
2787 1) | cgroup_migrate_execute() {
2788 1) | cpu_cgroup_can_attach() {
2789 1) | cgroup_taskset_first() {
2790 1) 0.732 us | cgroup_taskset_next(); /* = 0xffff93fc8fb20000 */
2791 1) 1.232 us | } /* cgroup_taskset_first = 0xffff93fc8fb20000 */
2792 1) 0.380 us | sched_rt_can_attach(); /* = 0x0 */
2793 1) 2.335 us | } /* cpu_cgroup_can_attach = -22 */
2794 1) 4.369 us | } /* cgroup_migrate_execute = -22 */
2795 1) 7.143 us | } /* cgroup_migrate = -22 */
2797 The above example shows that the function cpu_cgroup_can_attach
2798 returned the error code -22 firstly, then we can read the code
2799 of this function to get the root cause.
2801 When the option funcgraph-retval-hex is not set, the return value can
2802 be displayed in a smart way. Specifically, if it is an error code,
2803 it will be printed in signed decimal format, otherwise it will
2804 printed in hexadecimal format.
2806 - smart: echo nofuncgraph-retval-hex > trace_options
2807 - hexadecimal: echo funcgraph-retval-hex > trace_options
2809 Example with funcgraph-retval-hex::
2811 1) | cgroup_migrate() {
2812 1) 0.651 us | cgroup_migrate_add_task(); /* = 0xffff93fcfd346c00 */
2813 1) | cgroup_migrate_execute() {
2814 1) | cpu_cgroup_can_attach() {
2815 1) | cgroup_taskset_first() {
2816 1) 0.732 us | cgroup_taskset_next(); /* = 0xffff93fc8fb20000 */
2817 1) 1.232 us | } /* cgroup_taskset_first = 0xffff93fc8fb20000 */
2818 1) 0.380 us | sched_rt_can_attach(); /* = 0x0 */
2819 1) 2.335 us | } /* cpu_cgroup_can_attach = 0xffffffea */
2820 1) 4.369 us | } /* cgroup_migrate_execute = 0xffffffea */
2821 1) 7.143 us | } /* cgroup_migrate = 0xffffffea */
2823 At present, there are some limitations when using the funcgraph-retval
2824 option, and these limitations will be eliminated in the future:
2826 - Even if the function return type is void, a return value will still
2827 be printed, and you can just ignore it.
2829 - Even if return values are stored in multiple registers, only the
2830 value contained in the first register will be recorded and printed.
2831 To illustrate, in the x86 architecture, eax and edx are used to store
2832 a 64-bit return value, with the lower 32 bits saved in eax and the
2833 upper 32 bits saved in edx. However, only the value stored in eax
2834 will be recorded and printed.
2836 - In certain procedure call standards, such as arm64's AAPCS64, when a
2837 type is smaller than a GPR, it is the responsibility of the consumer
2838 to perform the narrowing, and the upper bits may contain UNKNOWN values.
2839 Therefore, it is advisable to check the code for such cases. For instance,
2840 when using a u8 in a 64-bit GPR, bits [63:8] may contain arbitrary values,
2841 especially when larger types are truncated, whether explicitly or implicitly.
2842 Here are some specific cases to illustrate this point:
2844 **Case One**:
2846 The function narrow_to_u8 is defined as follows::
2848 u8 narrow_to_u8(u64 val)
2849 {
2850 // implicitly truncated
2851 return val;
2852 }
2854 It may be compiled to::
2856 narrow_to_u8:
2857 < ... ftrace instrumentation ... >
2858 RET
2860 If you pass 0x123456789abcdef to this function and want to narrow it,
2861 it may be recorded as 0x123456789abcdef instead of 0xef.
2863 **Case Two**:
2865 The function error_if_not_4g_aligned is defined as follows::
2867 int error_if_not_4g_aligned(u64 val)
2868 {
2869 if (val & GENMASK(31, 0))
2870 return -EINVAL;
2872 return 0;
2873 }
2875 It could be compiled to::
2877 error_if_not_4g_aligned:
2878 CBNZ w0, .Lnot_aligned
2879 RET // bits [31:0] are zero, bits
2880 // [63:32] are UNKNOWN
2881 .Lnot_aligned:
2882 MOV x0, #-EINVAL
2883 RET
2885 When passing 0x2_0000_0000 to it, the return value may be recorded as
2886 0x2_0000_0000 instead of 0.
2888 You can put some comments on specific functions by using
2889 trace_printk() For example, if you want to put a comment inside
2890 the __might_sleep() function, you just have to include
2891 <linux/ftrace.h> and call trace_printk() inside __might_sleep()::
2893 trace_printk("I'm a comment!\n")
2895 will produce::
2897 1) | __might_sleep() {
2898 1) | /* I'm a comment! */
2899 1) 1.449 us | }
2902 You might find other useful features for this tracer in the
2903 following "dynamic ftrace" section such as tracing only specific
2904 functions or tasks.
2906 dynamic ftrace
2907 --------------
2909 If CONFIG_DYNAMIC_FTRACE is set, the system will run with
2910 virtually no overhead when function tracing is disabled. The way
2911 this works is the mcount function call (placed at the start of
2912 every kernel function, produced by the -pg switch in gcc),
2913 starts of pointing to a simple return. (Enabling FTRACE will
2914 include the -pg switch in the compiling of the kernel.)
2916 At compile time every C file object is run through the
2917 recordmcount program (located in the scripts directory). This
2918 program will parse the ELF headers in the C object to find all
2919 the locations in the .text section that call mcount. Starting
2920 with gcc version 4.6, the -mfentry has been added for x86, which
2921 calls "__fentry__" instead of "mcount". Which is called before
2922 the creation of the stack frame.
2924 Note, not all sections are traced. They may be prevented by either
2925 a notrace, or blocked another way and all inline functions are not
2926 traced. Check the "available_filter_functions" file to see what functions
2927 can be traced.
2929 A section called "__mcount_loc" is created that holds
2930 references to all the mcount/fentry call sites in the .text section.
2931 The recordmcount program re-links this section back into the
2932 original object. The final linking stage of the kernel will add all these
2933 references into a single table.
2935 On boot up, before SMP is initialized, the dynamic ftrace code
2936 scans this table and updates all the locations into nops. It
2937 also records the locations, which are added to the
2938 available_filter_functions list. Modules are processed as they
2939 are loaded and before they are executed. When a module is
2940 unloaded, it also removes its functions from the ftrace function
2941 list. This is automatic in the module unload code, and the
2942 module author does not need to worry about it.
2944 When tracing is enabled, the process of modifying the function
2945 tracepoints is dependent on architecture. The old method is to use
2946 kstop_machine to prevent races with the CPUs executing code being
2947 modified (which can cause the CPU to do undesirable things, especially
2948 if the modified code crosses cache (or page) boundaries), and the nops are
2949 patched back to calls. But this time, they do not call mcount
2950 (which is just a function stub). They now call into the ftrace
2951 infrastructure.
2953 The new method of modifying the function tracepoints is to place
2954 a breakpoint at the location to be modified, sync all CPUs, modify
2955 the rest of the instruction not covered by the breakpoint. Sync
2956 all CPUs again, and then remove the breakpoint with the finished
2957 version to the ftrace call site.
2959 Some archs do not even need to monkey around with the synchronization,
2960 and can just slap the new code on top of the old without any
2961 problems with other CPUs executing it at the same time.
2963 One special side-effect to the recording of the functions being
2964 traced is that we can now selectively choose which functions we
2965 wish to trace and which ones we want the mcount calls to remain
2966 as nops.
2968 Two files are used, one for enabling and one for disabling the
2969 tracing of specified functions. They are:
2971 set_ftrace_filter
2973 and
2975 set_ftrace_notrace
2977 A list of available functions that you can add to these files is
2978 listed in:
2980 available_filter_functions
2982 ::
2984 # cat available_filter_functions
2985 put_prev_task_idle
2986 kmem_cache_create
2987 pick_next_task_rt
2988 cpus_read_lock
2989 pick_next_task_fair
2990 mutex_lock
2991 [...]
2993 If I am only interested in sys_nanosleep and hrtimer_interrupt::
2995 # echo sys_nanosleep hrtimer_interrupt > set_ftrace_filter
2996 # echo function > current_tracer
2997 # echo 1 > tracing_on
2998 # usleep 1
2999 # echo 0 > tracing_on
3000 # cat trace
3001 # tracer: function
3002 #
3003 # entries-in-buffer/entries-written: 5/5 #P:4
3004 #
3005 # _-----=> irqs-off
3006 # / _----=> need-resched
3007 # | / _---=> hardirq/softirq
3008 # || / _--=> preempt-depth
3009 # ||| / delay
3010 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
3011 # | | | |||| | |
3012 usleep-2665 [001] .... 4186.475355: sys_nanosleep <-system_call_fastpath
3013 <idle>-0 [001] d.h1 4186.475409: hrtimer_interrupt <-smp_apic_timer_interrupt
3014 usleep-2665 [001] d.h1 4186.475426: hrtimer_interrupt <-smp_apic_timer_interrupt
3015 <idle>-0 [003] d.h1 4186.475426: hrtimer_interrupt <-smp_apic_timer_interrupt
3016 <idle>-0 [002] d.h1 4186.475427: hrtimer_interrupt <-smp_apic_timer_interrupt
3018 To see which functions are being traced, you can cat the file:
3019 ::
3021 # cat set_ftrace_filter
3022 hrtimer_interrupt
3023 sys_nanosleep
3026 Perhaps this is not enough. The filters also allow glob(7) matching.
3028 ``<match>*``
3029 will match functions that begin with <match>
3030 ``*<match>``
3031 will match functions that end with <match>
3032 ``*<match>*``
3033 will match functions that have <match> in it
3034 ``<match1>*<match2>``
3035 will match functions that begin with <match1> and end with <match2>
3037 .. note::
3038 It is better to use quotes to enclose the wild cards,
3039 otherwise the shell may expand the parameters into names
3040 of files in the local directory.
3042 ::
3044 # echo 'hrtimer_*' > set_ftrace_filter
3046 Produces::
3048 # tracer: function
3049 #
3050 # entries-in-buffer/entries-written: 897/897 #P:4
3051 #
3052 # _-----=> irqs-off
3053 # / _----=> need-resched
3054 # | / _---=> hardirq/softirq
3055 # || / _--=> preempt-depth
3056 # ||| / delay
3057 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
3058 # | | | |||| | |
3059 <idle>-0 [003] dN.1 4228.547803: hrtimer_cancel <-tick_nohz_idle_exit
3060 <idle>-0 [003] dN.1 4228.547804: hrtimer_try_to_cancel <-hrtimer_cancel
3061 <idle>-0 [003] dN.2 4228.547805: hrtimer_force_reprogram <-__remove_hrtimer
3062 <idle>-0 [003] dN.1 4228.547805: hrtimer_forward <-tick_nohz_idle_exit
3063 <idle>-0 [003] dN.1 4228.547805: hrtimer_start_range_ns <-hrtimer_start_expires.constprop.11
3064 <idle>-0 [003] d..1 4228.547858: hrtimer_get_next_event <-get_next_timer_interrupt
3065 <idle>-0 [003] d..1 4228.547859: hrtimer_start <-__tick_nohz_idle_enter
3066 <idle>-0 [003] d..2 4228.547860: hrtimer_force_reprogram <-__rem
3068 Notice that we lost the sys_nanosleep.
3069 ::
3071 # cat set_ftrace_filter
3072 hrtimer_run_queues
3073 hrtimer_run_pending
3074 hrtimer_init
3075 hrtimer_cancel
3076 hrtimer_try_to_cancel
3077 hrtimer_forward
3078 hrtimer_start
3079 hrtimer_reprogram
3080 hrtimer_force_reprogram
3081 hrtimer_get_next_event
3082 hrtimer_interrupt
3083 hrtimer_nanosleep
3084 hrtimer_wakeup
3085 hrtimer_get_remaining
3086 hrtimer_get_res
3087 hrtimer_init_sleeper
3090 This is because the '>' and '>>' act just like they do in bash.
3091 To rewrite the filters, use '>'
3092 To append to the filters, use '>>'
3094 To clear out a filter so that all functions will be recorded
3095 again::
3097 # echo > set_ftrace_filter
3098 # cat set_ftrace_filter
3099 #
3101 Again, now we want to append.
3103 ::
3105 # echo sys_nanosleep > set_ftrace_filter
3106 # cat set_ftrace_filter
3107 sys_nanosleep
3108 # echo 'hrtimer_*' >> set_ftrace_filter
3109 # cat set_ftrace_filter
3110 hrtimer_run_queues
3111 hrtimer_run_pending
3112 hrtimer_init
3113 hrtimer_cancel
3114 hrtimer_try_to_cancel
3115 hrtimer_forward
3116 hrtimer_start
3117 hrtimer_reprogram
3118 hrtimer_force_reprogram
3119 hrtimer_get_next_event
3120 hrtimer_interrupt
3121 sys_nanosleep
3122 hrtimer_nanosleep
3123 hrtimer_wakeup
3124 hrtimer_get_remaining
3125 hrtimer_get_res
3126 hrtimer_init_sleeper
3129 The set_ftrace_notrace prevents those functions from being
3130 traced.
3131 ::
3133 # echo '*preempt*' '*lock*' > set_ftrace_notrace
3135 Produces::
3137 # tracer: function
3138 #
3139 # entries-in-buffer/entries-written: 39608/39608 #P:4
3140 #
3141 # _-----=> irqs-off
3142 # / _----=> need-resched
3143 # | / _---=> hardirq/softirq
3144 # || / _--=> preempt-depth
3145 # ||| / delay
3146 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
3147 # | | | |||| | |
3148 bash-1994 [000] .... 4342.324896: file_ra_state_init <-do_dentry_open
3149 bash-1994 [000] .... 4342.324897: open_check_o_direct <-do_last
3150 bash-1994 [000] .... 4342.324897: ima_file_check <-do_last
3151 bash-1994 [000] .... 4342.324898: process_measurement <-ima_file_check
3152 bash-1994 [000] .... 4342.324898: ima_get_action <-process_measurement
3153 bash-1994 [000] .... 4342.324898: ima_match_policy <-ima_get_action
3154 bash-1994 [000] .... 4342.324899: do_truncate <-do_last
3155 bash-1994 [000] .... 4342.324899: setattr_should_drop_suidgid <-do_truncate
3156 bash-1994 [000] .... 4342.324899: notify_change <-do_truncate
3157 bash-1994 [000] .... 4342.324900: current_fs_time <-notify_change
3158 bash-1994 [000] .... 4342.324900: current_kernel_time <-current_fs_time
3159 bash-1994 [000] .... 4342.324900: timespec_trunc <-current_fs_time
3161 We can see that there's no more lock or preempt tracing.
3163 Selecting function filters via index
3164 ------------------------------------
3166 Because processing of strings is expensive (the address of the function
3167 needs to be looked up before comparing to the string being passed in),
3168 an index can be used as well to enable functions. This is useful in the
3169 case of setting thousands of specific functions at a time. By passing
3170 in a list of numbers, no string processing will occur. Instead, the function
3171 at the specific location in the internal array (which corresponds to the
3172 functions in the "available_filter_functions" file), is selected.
3174 ::
3176 # echo 1 > set_ftrace_filter
3178 Will select the first function listed in "available_filter_functions"
3180 ::
3182 # head -1 available_filter_functions
3183 trace_initcall_finish_cb
3185 # cat set_ftrace_filter
3186 trace_initcall_finish_cb
3188 # head -50 available_filter_functions | tail -1
3189 x86_pmu_commit_txn
3191 # echo 1 50 > set_ftrace_filter
3192 # cat set_ftrace_filter
3193 trace_initcall_finish_cb
3194 x86_pmu_commit_txn
3196 Dynamic ftrace with the function graph tracer
3197 ---------------------------------------------
3199 Although what has been explained above concerns both the
3200 function tracer and the function-graph-tracer, there are some
3201 special features only available in the function-graph tracer.
3203 If you want to trace only one function and all of its children,
3204 you just have to echo its name into set_graph_function::
3206 echo __do_fault > set_graph_function
3208 will produce the following "expanded" trace of the __do_fault()
3209 function::
3211 0) | __do_fault() {
3212 0) | filemap_fault() {
3213 0) | find_lock_page() {
3214 0) 0.804 us | find_get_page();
3215 0) | __might_sleep() {
3216 0) 1.329 us | }
3217 0) 3.904 us | }
3218 0) 4.979 us | }
3219 0) 0.653 us | _spin_lock();
3220 0) 0.578 us | page_add_file_rmap();
3221 0) 0.525 us | native_set_pte_at();
3222 0) 0.585 us | _spin_unlock();
3223 0) | unlock_page() {
3224 0) 0.541 us | page_waitqueue();
3225 0) 0.639 us | __wake_up_bit();
3226 0) 2.786 us | }
3227 0) + 14.237 us | }
3228 0) | __do_fault() {
3229 0) | filemap_fault() {
3230 0) | find_lock_page() {
3231 0) 0.698 us | find_get_page();
3232 0) | __might_sleep() {
3233 0) 1.412 us | }
3234 0) 3.950 us | }
3235 0) 5.098 us | }
3236 0) 0.631 us | _spin_lock();
3237 0) 0.571 us | page_add_file_rmap();
3238 0) 0.526 us | native_set_pte_at();
3239 0) 0.586 us | _spin_unlock();
3240 0) | unlock_page() {
3241 0) 0.533 us | page_waitqueue();
3242 0) 0.638 us | __wake_up_bit();
3243 0) 2.793 us | }
3244 0) + 14.012 us | }
3246 You can also expand several functions at once::
3248 echo sys_open > set_graph_function
3249 echo sys_close >> set_graph_function
3251 Now if you want to go back to trace all functions you can clear
3252 this special filter via::
3254 echo > set_graph_function
3257 ftrace_enabled
3258 --------------
3260 Note, the proc sysctl ftrace_enable is a big on/off switch for the
3261 function tracer. By default it is enabled (when function tracing is
3262 enabled in the kernel). If it is disabled, all function tracing is
3263 disabled. This includes not only the function tracers for ftrace, but
3264 also for any other uses (perf, kprobes, stack tracing, profiling, etc). It
3265 cannot be disabled if there is a callback with FTRACE_OPS_FL_PERMANENT set
3266 registered.
3268 Please disable this with care.
3270 This can be disable (and enabled) with::
3272 sysctl kernel.ftrace_enabled=0
3273 sysctl kernel.ftrace_enabled=1
3275 or
3277 echo 0 > /proc/sys/kernel/ftrace_enabled
3278 echo 1 > /proc/sys/kernel/ftrace_enabled
3281 Filter commands
3282 ---------------
3284 A few commands are supported by the set_ftrace_filter interface.
3285 Trace commands have the following format::
3287 <function>:<command>:<parameter>
3289 The following commands are supported:
3291 - mod:
3292 This command enables function filtering per module. The
3293 parameter defines the module. For example, if only the write*
3294 functions in the ext3 module are desired, run:
3296 echo 'write*:mod:ext3' > set_ftrace_filter
3298 This command interacts with the filter in the same way as
3299 filtering based on function names. Thus, adding more functions
3300 in a different module is accomplished by appending (>>) to the
3301 filter file. Remove specific module functions by prepending
3302 '!'::
3304 echo '!writeback*:mod:ext3' >> set_ftrace_filter
3306 Mod command supports module globbing. Disable tracing for all
3307 functions except a specific module::
3309 echo '!*:mod:!ext3' >> set_ftrace_filter
3311 Disable tracing for all modules, but still trace kernel::
3313 echo '!*:mod:*' >> set_ftrace_filter
3315 Enable filter only for kernel::
3317 echo '*write*:mod:!*' >> set_ftrace_filter
3319 Enable filter for module globbing::
3321 echo '*write*:mod:*snd*' >> set_ftrace_filter
3323 - traceon/traceoff:
3324 These commands turn tracing on and off when the specified
3325 functions are hit. The parameter determines how many times the
3326 tracing system is turned on and off. If unspecified, there is
3327 no limit. For example, to disable tracing when a schedule bug
3328 is hit the first 5 times, run::
3330 echo '__schedule_bug:traceoff:5' > set_ftrace_filter
3332 To always disable tracing when __schedule_bug is hit::
3334 echo '__schedule_bug:traceoff' > set_ftrace_filter
3336 These commands are cumulative whether or not they are appended
3337 to set_ftrace_filter. To remove a command, prepend it by '!'
3338 and drop the parameter::
3340 echo '!__schedule_bug:traceoff:0' > set_ftrace_filter
3342 The above removes the traceoff command for __schedule_bug
3343 that have a counter. To remove commands without counters::
3345 echo '!__schedule_bug:traceoff' > set_ftrace_filter
3347 - snapshot:
3348 Will cause a snapshot to be triggered when the function is hit.
3349 ::
3351 echo 'native_flush_tlb_others:snapshot' > set_ftrace_filter
3353 To only snapshot once:
3354 ::
3356 echo 'native_flush_tlb_others:snapshot:1' > set_ftrace_filter
3358 To remove the above commands::
3360 echo '!native_flush_tlb_others:snapshot' > set_ftrace_filter
3361 echo '!native_flush_tlb_others:snapshot:0' > set_ftrace_filter
3363 - enable_event/disable_event:
3364 These commands can enable or disable a trace event. Note, because
3365 function tracing callbacks are very sensitive, when these commands
3366 are registered, the trace point is activated, but disabled in
3367 a "soft" mode. That is, the tracepoint will be called, but
3368 just will not be traced. The event tracepoint stays in this mode
3369 as long as there's a command that triggers it.
3370 ::
3372 echo 'try_to_wake_up:enable_event:sched:sched_switch:2' > \
3373 set_ftrace_filter
3375 The format is::
3377 <function>:enable_event:<system>:<event>[:count]
3378 <function>:disable_event:<system>:<event>[:count]
3380 To remove the events commands::
3382 echo '!try_to_wake_up:enable_event:sched:sched_switch:0' > \
3383 set_ftrace_filter
3384 echo '!schedule:disable_event:sched:sched_switch' > \
3385 set_ftrace_filter
3387 - dump:
3388 When the function is hit, it will dump the contents of the ftrace
3389 ring buffer to the console. This is useful if you need to debug
3390 something, and want to dump the trace when a certain function
3391 is hit. Perhaps it's a function that is called before a triple
3392 fault happens and does not allow you to get a regular dump.
3394 - cpudump:
3395 When the function is hit, it will dump the contents of the ftrace
3396 ring buffer for the current CPU to the console. Unlike the "dump"
3397 command, it only prints out the contents of the ring buffer for the
3398 CPU that executed the function that triggered the dump.
3400 - stacktrace:
3401 When the function is hit, a stack trace is recorded.
3403 trace_pipe
3404 ----------
3406 The trace_pipe outputs the same content as the trace file, but
3407 the effect on the tracing is different. Every read from
3408 trace_pipe is consumed. This means that subsequent reads will be
3409 different. The trace is live.
3410 ::
3412 # echo function > current_tracer
3413 # cat trace_pipe > /tmp/trace.out &
3414 [1] 4153
3415 # echo 1 > tracing_on
3416 # usleep 1
3417 # echo 0 > tracing_on
3418 # cat trace
3419 # tracer: function
3420 #
3421 # entries-in-buffer/entries-written: 0/0 #P:4
3422 #
3423 # _-----=> irqs-off
3424 # / _----=> need-resched
3425 # | / _---=> hardirq/softirq
3426 # || / _--=> preempt-depth
3427 # ||| / delay
3428 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
3429 # | | | |||| | |
3431 #
3432 # cat /tmp/trace.out
3433 bash-1994 [000] .... 5281.568961: mutex_unlock <-rb_simple_write
3434 bash-1994 [000] .... 5281.568963: __mutex_unlock_slowpath <-mutex_unlock
3435 bash-1994 [000] .... 5281.568963: __fsnotify_parent <-fsnotify_modify
3436 bash-1994 [000] .... 5281.568964: fsnotify <-fsnotify_modify
3437 bash-1994 [000] .... 5281.568964: __srcu_read_lock <-fsnotify
3438 bash-1994 [000] .... 5281.568964: add_preempt_count <-__srcu_read_lock
3439 bash-1994 [000] ...1 5281.568965: sub_preempt_count <-__srcu_read_lock
3440 bash-1994 [000] .... 5281.568965: __srcu_read_unlock <-fsnotify
3441 bash-1994 [000] .... 5281.568967: sys_dup2 <-system_call_fastpath
3444 Note, reading the trace_pipe file will block until more input is
3445 added. This is contrary to the trace file. If any process opened
3446 the trace file for reading, it will actually disable tracing and
3447 prevent new entries from being added. The trace_pipe file does
3448 not have this limitation.
3450 trace entries
3451 -------------
3453 Having too much or not enough data can be troublesome in
3454 diagnosing an issue in the kernel. The file buffer_size_kb is
3455 used to modify the size of the internal trace buffers. The
3456 number listed is the number of entries that can be recorded per
3457 CPU. To know the full size, multiply the number of possible CPUs
3458 with the number of entries.
3459 ::
3461 # cat buffer_size_kb
3462 1408 (units kilobytes)
3464 Or simply read buffer_total_size_kb
3465 ::
3467 # cat buffer_total_size_kb
3468 5632
3470 To modify the buffer, simple echo in a number (in 1024 byte segments).
3471 ::
3473 # echo 10000 > buffer_size_kb
3474 # cat buffer_size_kb
3475 10000 (units kilobytes)
3477 It will try to allocate as much as possible. If you allocate too
3478 much, it can cause Out-Of-Memory to trigger.
3479 ::
3481 # echo 1000000000000 > buffer_size_kb
3482 -bash: echo: write error: Cannot allocate memory
3483 # cat buffer_size_kb
3484 85
3486 The per_cpu buffers can be changed individually as well:
3487 ::
3489 # echo 10000 > per_cpu/cpu0/buffer_size_kb
3490 # echo 100 > per_cpu/cpu1/buffer_size_kb
3492 When the per_cpu buffers are not the same, the buffer_size_kb
3493 at the top level will just show an X
3494 ::
3496 # cat buffer_size_kb
3497 X
3499 This is where the buffer_total_size_kb is useful:
3500 ::
3502 # cat buffer_total_size_kb
3503 12916
3505 Writing to the top level buffer_size_kb will reset all the buffers
3506 to be the same again.
3508 Snapshot
3509 --------
3510 CONFIG_TRACER_SNAPSHOT makes a generic snapshot feature
3511 available to all non latency tracers. (Latency tracers which
3512 record max latency, such as "irqsoff" or "wakeup", can't use
3513 this feature, since those are already using the snapshot
3514 mechanism internally.)
3516 Snapshot preserves a current trace buffer at a particular point
3517 in time without stopping tracing. Ftrace swaps the current
3518 buffer with a spare buffer, and tracing continues in the new
3519 current (=previous spare) buffer.
3521 The following tracefs files in "tracing" are related to this
3522 feature:
3524 snapshot:
3526 This is used to take a snapshot and to read the output
3527 of the snapshot. Echo 1 into this file to allocate a
3528 spare buffer and to take a snapshot (swap), then read
3529 the snapshot from this file in the same format as
3530 "trace" (described above in the section "The File
3531 System"). Both reads snapshot and tracing are executable
3532 in parallel. When the spare buffer is allocated, echoing
3533 0 frees it, and echoing else (positive) values clear the
3534 snapshot contents.
3535 More details are shown in the table below.
3537 +--------------+------------+------------+------------+
3538 |status\\input | 0 | 1 | else |
3539 +==============+============+============+============+
3540 |not allocated |(do nothing)| alloc+swap |(do nothing)|
3541 +--------------+------------+------------+------------+
3542 |allocated | free | swap | clear |
3543 +--------------+------------+------------+------------+
3545 Here is an example of using the snapshot feature.
3546 ::
3548 # echo 1 > events/sched/enable
3549 # echo 1 > snapshot
3550 # cat snapshot
3551 # tracer: nop
3552 #
3553 # entries-in-buffer/entries-written: 71/71 #P:8
3554 #
3555 # _-----=> irqs-off
3556 # / _----=> need-resched
3557 # | / _---=> hardirq/softirq
3558 # || / _--=> preempt-depth
3559 # ||| / delay
3560 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
3561 # | | | |||| | |
3562 <idle>-0 [005] d... 2440.603828: sched_switch: prev_comm=swapper/5 prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=snapshot-test-2 next_pid=2242 next_prio=120
3563 sleep-2242 [005] d... 2440.603846: sched_switch: prev_comm=snapshot-test-2 prev_pid=2242 prev_prio=120 prev_state=R ==> next_comm=kworker/5:1 next_pid=60 next_prio=120
3564 [...]
3565 <idle>-0 [002] d... 2440.707230: sched_switch: prev_comm=swapper/2 prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=snapshot-test-2 next_pid=2229 next_prio=120
3567 # cat trace
3568 # tracer: nop
3569 #
3570 # entries-in-buffer/entries-written: 77/77 #P:8
3571 #
3572 # _-----=> irqs-off
3573 # / _----=> need-resched
3574 # | / _---=> hardirq/softirq
3575 # || / _--=> preempt-depth
3576 # ||| / delay
3577 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
3578 # | | | |||| | |
3579 <idle>-0 [007] d... 2440.707395: sched_switch: prev_comm=swapper/7 prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=snapshot-test-2 next_pid=2243 next_prio=120
3580 snapshot-test-2-2229 [002] d... 2440.707438: sched_switch: prev_comm=snapshot-test-2 prev_pid=2229 prev_prio=120 prev_state=S ==> next_comm=swapper/2 next_pid=0 next_prio=120
3581 [...]
3584 If you try to use this snapshot feature when current tracer is
3585 one of the latency tracers, you will get the following results.
3586 ::
3588 # echo wakeup > current_tracer
3589 # echo 1 > snapshot
3590 bash: echo: write error: Device or resource busy
3591 # cat snapshot
3592 cat: snapshot: Device or resource busy
3595 Instances
3596 ---------
3597 In the tracefs tracing directory, there is a directory called "instances".
3598 This directory can have new directories created inside of it using
3599 mkdir, and removing directories with rmdir. The directory created
3600 with mkdir in this directory will already contain files and other
3601 directories after it is created.
3602 ::
3604 # mkdir instances/foo
3605 # ls instances/foo
3606 buffer_size_kb buffer_total_size_kb events free_buffer per_cpu
3607 set_event snapshot trace trace_clock trace_marker trace_options
3608 trace_pipe tracing_on
3610 As you can see, the new directory looks similar to the tracing directory
3611 itself. In fact, it is very similar, except that the buffer and
3612 events are agnostic from the main directory, or from any other
3613 instances that are created.
3615 The files in the new directory work just like the files with the
3616 same name in the tracing directory except the buffer that is used
3617 is a separate and new buffer. The files affect that buffer but do not
3618 affect the main buffer with the exception of trace_options. Currently,
3619 the trace_options affect all instances and the top level buffer
3620 the same, but this may change in future releases. That is, options
3621 may become specific to the instance they reside in.
3623 Notice that none of the function tracer files are there, nor is
3624 current_tracer and available_tracers. This is because the buffers
3625 can currently only have events enabled for them.
3626 ::
3628 # mkdir instances/foo
3629 # mkdir instances/bar
3630 # mkdir instances/zoot
3631 # echo 100000 > buffer_size_kb
3632 # echo 1000 > instances/foo/buffer_size_kb
3633 # echo 5000 > instances/bar/per_cpu/cpu1/buffer_size_kb
3634 # echo function > current_trace
3635 # echo 1 > instances/foo/events/sched/sched_wakeup/enable
3636 # echo 1 > instances/foo/events/sched/sched_wakeup_new/enable
3637 # echo 1 > instances/foo/events/sched/sched_switch/enable
3638 # echo 1 > instances/bar/events/irq/enable
3639 # echo 1 > instances/zoot/events/syscalls/enable
3640 # cat trace_pipe
3641 CPU:2 [LOST 11745 EVENTS]
3642 bash-2044 [002] .... 10594.481032: _raw_spin_lock_irqsave <-get_page_from_freelist
3643 bash-2044 [002] d... 10594.481032: add_preempt_count <-_raw_spin_lock_irqsave
3644 bash-2044 [002] d..1 10594.481032: __rmqueue <-get_page_from_freelist
3645 bash-2044 [002] d..1 10594.481033: _raw_spin_unlock <-get_page_from_freelist
3646 bash-2044 [002] d..1 10594.481033: sub_preempt_count <-_raw_spin_unlock
3647 bash-2044 [002] d... 10594.481033: get_pageblock_flags_group <-get_pageblock_migratetype
3648 bash-2044 [002] d... 10594.481034: __mod_zone_page_state <-get_page_from_freelist
3649 bash-2044 [002] d... 10594.481034: zone_statistics <-get_page_from_freelist
3650 bash-2044 [002] d... 10594.481034: __inc_zone_state <-zone_statistics
3651 bash-2044 [002] d... 10594.481034: __inc_zone_state <-zone_statistics
3652 bash-2044 [002] .... 10594.481035: arch_dup_task_struct <-copy_process
3653 [...]
3655 # cat instances/foo/trace_pipe
3656 bash-1998 [000] d..4 136.676759: sched_wakeup: comm=kworker/0:1 pid=59 prio=120 success=1 target_cpu=000
3657 bash-1998 [000] dN.4 136.676760: sched_wakeup: comm=bash pid=1998 prio=120 success=1 target_cpu=000
3658 <idle>-0 [003] d.h3 136.676906: sched_wakeup: comm=rcu_preempt pid=9 prio=120 success=1 target_cpu=003
3659 <idle>-0 [003] d..3 136.676909: sched_switch: prev_comm=swapper/3 prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=rcu_preempt next_pid=9 next_prio=120
3660 rcu_preempt-9 [003] d..3 136.676916: sched_switch: prev_comm=rcu_preempt prev_pid=9 prev_prio=120 prev_state=S ==> next_comm=swapper/3 next_pid=0 next_prio=120
3661 bash-1998 [000] d..4 136.677014: sched_wakeup: comm=kworker/0:1 pid=59 prio=120 success=1 target_cpu=000
3662 bash-1998 [000] dN.4 136.677016: sched_wakeup: comm=bash pid=1998 prio=120 success=1 target_cpu=000
3663 bash-1998 [000] d..3 136.677018: sched_switch: prev_comm=bash prev_pid=1998 prev_prio=120 prev_state=R+ ==> next_comm=kworker/0:1 next_pid=59 next_prio=120
3664 kworker/0:1-59 [000] d..4 136.677022: sched_wakeup: comm=sshd pid=1995 prio=120 success=1 target_cpu=001
3665 kworker/0:1-59 [000] d..3 136.677025: sched_switch: prev_comm=kworker/0:1 prev_pid=59 prev_prio=120 prev_state=S ==> next_comm=bash next_pid=1998 next_prio=120
3666 [...]
3668 # cat instances/bar/trace_pipe
3669 migration/1-14 [001] d.h3 138.732674: softirq_raise: vec=3 [action=NET_RX]
3670 <idle>-0 [001] dNh3 138.732725: softirq_raise: vec=3 [action=NET_RX]
3671 bash-1998 [000] d.h1 138.733101: softirq_raise: vec=1 [action=TIMER]
3672 bash-1998 [000] d.h1 138.733102: softirq_raise: vec=9 [action=RCU]
3673 bash-1998 [000] ..s2 138.733105: softirq_entry: vec=1 [action=TIMER]
3674 bash-1998 [000] ..s2 138.733106: softirq_exit: vec=1 [action=TIMER]
3675 bash-1998 [000] ..s2 138.733106: softirq_entry: vec=9 [action=RCU]
3676 bash-1998 [000] ..s2 138.733109: softirq_exit: vec=9 [action=RCU]
3677 sshd-1995 [001] d.h1 138.733278: irq_handler_entry: irq=21 name=uhci_hcd:usb4
3678 sshd-1995 [001] d.h1 138.733280: irq_handler_exit: irq=21 ret=unhandled
3679 sshd-1995 [001] d.h1 138.733281: irq_handler_entry: irq=21 name=eth0
3680 sshd-1995 [001] d.h1 138.733283: irq_handler_exit: irq=21 ret=handled
3681 [...]
3683 # cat instances/zoot/trace
3684 # tracer: nop
3685 #
3686 # entries-in-buffer/entries-written: 18996/18996 #P:4
3687 #
3688 # _-----=> irqs-off
3689 # / _----=> need-resched
3690 # | / _---=> hardirq/softirq
3691 # || / _--=> preempt-depth
3692 # ||| / delay
3693 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
3694 # | | | |||| | |
3695 bash-1998 [000] d... 140.733501: sys_write -> 0x2
3696 bash-1998 [000] d... 140.733504: sys_dup2(oldfd: a, newfd: 1)
3697 bash-1998 [000] d... 140.733506: sys_dup2 -> 0x1
3698 bash-1998 [000] d... 140.733508: sys_fcntl(fd: a, cmd: 1, arg: 0)
3699 bash-1998 [000] d... 140.733509: sys_fcntl -> 0x1
3700 bash-1998 [000] d... 140.733510: sys_close(fd: a)
3701 bash-1998 [000] d... 140.733510: sys_close -> 0x0
3702 bash-1998 [000] d... 140.733514: sys_rt_sigprocmask(how: 0, nset: 0, oset: 6e2768, sigsetsize: 8)
3703 bash-1998 [000] d... 140.733515: sys_rt_sigprocmask -> 0x0
3704 bash-1998 [000] d... 140.733516: sys_rt_sigaction(sig: 2, act: 7fff718846f0, oact: 7fff71884650, sigsetsize: 8)
3705 bash-1998 [000] d... 140.733516: sys_rt_sigaction -> 0x0
3707 You can see that the trace of the top most trace buffer shows only
3708 the function tracing. The foo instance displays wakeups and task
3709 switches.
3711 To remove the instances, simply delete their directories:
3712 ::
3714 # rmdir instances/foo
3715 # rmdir instances/bar
3716 # rmdir instances/zoot
3718 Note, if a process has a trace file open in one of the instance
3719 directories, the rmdir will fail with EBUSY.
3722 Stack trace
3723 -----------
3724 Since the kernel has a fixed sized stack, it is important not to
3725 waste it in functions. A kernel developer must be conscious of
3726 what they allocate on the stack. If they add too much, the system
3727 can be in danger of a stack overflow, and corruption will occur,
3728 usually leading to a system panic.
3730 There are some tools that check this, usually with interrupts
3731 periodically checking usage. But if you can perform a check
3732 at every function call that will become very useful. As ftrace provides
3733 a function tracer, it makes it convenient to check the stack size
3734 at every function call. This is enabled via the stack tracer.
3736 CONFIG_STACK_TRACER enables the ftrace stack tracing functionality.
3737 To enable it, write a '1' into /proc/sys/kernel/stack_tracer_enabled.
3738 ::
3740 # echo 1 > /proc/sys/kernel/stack_tracer_enabled
3742 You can also enable it from the kernel command line to trace
3743 the stack size of the kernel during boot up, by adding "stacktrace"
3744 to the kernel command line parameter.
3746 After running it for a few minutes, the output looks like:
3747 ::
3749 # cat stack_max_size
3750 2928
3752 # cat stack_trace
3753 Depth Size Location (18 entries)
3754 ----- ---- --------
3755 0) 2928 224 update_sd_lb_stats+0xbc/0x4ac
3756 1) 2704 160 find_busiest_group+0x31/0x1f1
3757 2) 2544 256 load_balance+0xd9/0x662
3758 3) 2288 80 idle_balance+0xbb/0x130
3759 4) 2208 128 __schedule+0x26e/0x5b9
3760 5) 2080 16 schedule+0x64/0x66
3761 6) 2064 128 schedule_timeout+0x34/0xe0
3762 7) 1936 112 wait_for_common+0x97/0xf1
3763 8) 1824 16 wait_for_completion+0x1d/0x1f
3764 9) 1808 128 flush_work+0xfe/0x119
3765 10) 1680 16 tty_flush_to_ldisc+0x1e/0x20
3766 11) 1664 48 input_available_p+0x1d/0x5c
3767 12) 1616 48 n_tty_poll+0x6d/0x134
3768 13) 1568 64 tty_poll+0x64/0x7f
3769 14) 1504 880 do_select+0x31e/0x511
3770 15) 624 400 core_sys_select+0x177/0x216
3771 16) 224 96 sys_select+0x91/0xb9
3772 17) 128 128 system_call_fastpath+0x16/0x1b
3774 Note, if -mfentry is being used by gcc, functions get traced before
3775 they set up the stack frame. This means that leaf level functions
3776 are not tested by the stack tracer when -mfentry is used.
3778 Currently, -mfentry is used by gcc 4.6.0 and above on x86 only.
3780 More
3781 ----
3782 More details can be found in the source code, in the `kernel/trace/*.c` files.