1 ftrace - Function Tracer
2 ========================
4 Copyright 2008 Red Hat Inc.
5 Author: Steven Rostedt <srostedt@redhat.com>
6 License: The GNU Free Documentation License, Version 1.2
7 (dual licensed under the GPL v2)
8 Reviewers: Elias Oltmanns, Randy Dunlap, Andrew Morton,
9 John Kacur, and David Teigland.
10 Written for: 2.6.28-rc2
16 Ftrace is an internal tracer designed to help out developers and
17 designers of systems to find what is going on inside the kernel.
18 It can be used for debugging or analyzing latencies and
19 performance issues that take place outside of user-space.
21 Although ftrace is typically considered the function tracer, it
22 is really a frame work of several assorted tracing utilities.
23 There's latency tracing to examine what occurs between interrupts
24 disabled and enabled, as well as for preemption and from a time
25 a task is woken to the task is actually scheduled in.
27 One of the most common uses of ftrace is the event tracing.
28 Through out the kernel is hundreds of static event points that
29 can be enabled via the debugfs file system to see what is
30 going on in certain parts of the kernel.
33 Implementation Details
34 ----------------------
36 See ftrace-design.txt for details for arch porters and such.
42 Ftrace uses the debugfs file system to hold the control files as
43 well as the files to display output.
45 When debugfs is configured into the kernel (which selecting any ftrace
46 option will do) the directory /sys/kernel/debug will be created. To mount
47 this directory, you can add to your /etc/fstab file:
49 debugfs /sys/kernel/debug debugfs defaults 0 0
51 Or you can mount it at run time with:
53 mount -t debugfs nodev /sys/kernel/debug
55 For quicker access to that directory you may want to make a soft link to
58 ln -s /sys/kernel/debug /debug
60 Any selected ftrace option will also create a directory called tracing
61 within the debugfs. The rest of the document will assume that you are in
62 the ftrace directory (cd /sys/kernel/debug/tracing) and will only concentrate
63 on the files within that directory and not distract from the content with
64 the extended "/sys/kernel/debug/tracing" path name.
66 That's it! (assuming that you have ftrace configured into your kernel)
68 After mounting debugfs, you can see a directory called
69 "tracing". This directory contains the control and output files
70 of ftrace. Here is a list of some of the key files:
73 Note: all time values are in microseconds.
77 This is used to set or display the current tracer
82 This holds the different types of tracers that
83 have been compiled into the kernel. The
84 tracers listed here can be configured by
85 echoing their name into current_tracer.
89 This sets or displays whether writing to the trace
90 ring buffer is enabled. Echo 0 into this file to disable
91 the tracer or 1 to enable it. Note, this only disables
92 writing to the ring buffer, the tracing overhead may
97 This file holds the output of the trace in a human
98 readable format (described below).
102 The output is the same as the "trace" file but this
103 file is meant to be streamed with live tracing.
104 Reads from this file will block until new data is
105 retrieved. Unlike the "trace" file, this file is a
106 consumer. This means reading from this file causes
107 sequential reads to display more current data. Once
108 data is read from this file, it is consumed, and
109 will not be read again with a sequential read. The
110 "trace" file is static, and if the tracer is not
111 adding more data,they will display the same
112 information every time they are read.
116 This file lets the user control the amount of data
117 that is displayed in one of the above output
118 files. Options also exist to modify how a tracer
119 or events work (stack traces, timestamps, etc).
123 This is a directory that has a file for every available
124 trace option (also in trace_options). Options may also be set
125 or cleared by writing a "1" or "0" respectively into the
126 corresponding file with the option name.
130 Some of the tracers record the max latency.
131 For example, the time interrupts are disabled.
132 This time is saved in this file. The max trace
133 will also be stored, and displayed by "trace".
134 A new max trace will only be recorded if the
135 latency is greater than the value in this
136 file. (in microseconds)
140 Some latency tracers will record a trace whenever the
141 latency is greater than the number in this file.
142 Only active when the file contains a number greater than 0.
147 This sets or displays the number of kilobytes each CPU
148 buffer holds. By default, the trace buffers are the same size
149 for each CPU. The displayed number is the size of the
150 CPU buffer and not total size of all buffers. The
151 trace buffers are allocated in pages (blocks of memory
152 that the kernel uses for allocation, usually 4 KB in size).
153 If the last page allocated has room for more bytes
154 than requested, the rest of the page will be used,
155 making the actual allocation bigger than requested.
156 ( Note, the size may not be a multiple of the page size
157 due to buffer management meta-data. )
159 buffer_total_size_kb:
161 This displays the total combined size of all the trace buffers.
165 If a process is performing the tracing, and the ring buffer
166 should be shrunk "freed" when the process is finished, even
167 if it were to be killed by a signal, this file can be used
168 for that purpose. On close of this file, the ring buffer will
169 be resized to its minimum size. Having a process that is tracing
170 also open this file, when the process exits its file descriptor
171 for this file will be closed, and in doing so, the ring buffer
174 It may also stop tracing if disable_on_free option is set.
178 This is a mask that lets the user only trace
179 on specified CPUs. The format is a hex string
180 representing the CPUs.
184 When dynamic ftrace is configured in (see the
185 section below "dynamic ftrace"), the code is dynamically
186 modified (code text rewrite) to disable calling of the
187 function profiler (mcount). This lets tracing be configured
188 in with practically no overhead in performance. This also
189 has a side effect of enabling or disabling specific functions
190 to be traced. Echoing names of functions into this file
191 will limit the trace to only those functions.
193 This interface also allows for commands to be used. See the
194 "Filter commands" section for more details.
198 This has an effect opposite to that of
199 set_ftrace_filter. Any function that is added here will not
200 be traced. If a function exists in both set_ftrace_filter
201 and set_ftrace_notrace, the function will _not_ be traced.
205 Have the function tracer only trace a single thread.
209 Set a "trigger" function where tracing should start
210 with the function graph tracer (See the section
211 "dynamic ftrace" for more details).
213 available_filter_functions:
215 This lists the functions that ftrace
216 has processed and can trace. These are the function
217 names that you can pass to "set_ftrace_filter" or
218 "set_ftrace_notrace". (See the section "dynamic ftrace"
219 below for more details.)
223 This file is more for debugging ftrace, but can also be useful
224 in seeing if any function has a callback attached to it.
225 Not only does the trace infrastructure use ftrace function
226 trace utility, but other subsystems might too. This file
227 displays all functions that have a callback attached to them
228 as well as the number of callbacks that have been attached.
229 Note, a callback may also call multiple functions which will
230 not be listed in this count.
232 If the callback registered to be traced by a function with
233 the "save regs" attribute (thus even more overhead), a 'R'
234 will be displayed on the same line as the function that
235 is returning registers.
237 If the callback registered to be traced by a function with
238 the "ip modify" attribute (thus the regs->ip can be changed),
239 an 'I' will be displayed on the same line as the function that
242 function_profile_enabled:
244 When set it will enable all functions with either the function
245 tracer, or if enabled, the function graph tracer. It will
246 keep a histogram of the number of functions that were called
247 and if run with the function graph tracer, it will also keep
248 track of the time spent in those functions. The histogram
249 content can be displayed in the files:
251 trace_stats/function<cpu> ( function0, function1, etc).
255 A directory that holds different tracing stats.
259 Enable dynamic trace points. See kprobetrace.txt.
263 Dynamic trace points stats. See kprobetrace.txt.
267 Used with the function graph tracer. This is the max depth
268 it will trace into a function. Setting this to a value of
269 one will show only the first kernel function that is called
274 This is for tools that read the raw format files. If an event in
275 the ring buffer references a string (currently only trace_printk()
276 does this), only a pointer to the string is recorded into the buffer
277 and not the string itself. This prevents tools from knowing what
278 that string was. This file displays the string and address for
279 the string allowing tools to map the pointers to what the
284 Only the pid of the task is recorded in a trace event unless
285 the event specifically saves the task comm as well. Ftrace
286 makes a cache of pid mappings to comms to try to display
287 comms for events. If a pid for a comm is not listed, then
288 "<...>" is displayed in the output.
292 This displays the "snapshot" buffer and also lets the user
293 take a snapshot of the current running trace.
294 See the "Snapshot" section below for more details.
298 When the stack tracer is activated, this will display the
299 maximum stack size it has encountered.
300 See the "Stack Trace" section below.
304 This displays the stack back trace of the largest stack
305 that was encountered when the stack tracer is activated.
306 See the "Stack Trace" section below.
310 This is similar to "set_ftrace_filter" but it limits what
311 functions the stack tracer will check.
315 Whenever an event is recorded into the ring buffer, a
316 "timestamp" is added. This stamp comes from a specified
317 clock. By default, ftrace uses the "local" clock. This
318 clock is very fast and strictly per cpu, but on some
319 systems it may not be monotonic with respect to other
320 CPUs. In other words, the local clocks may not be in sync
321 with local clocks on other CPUs.
323 Usual clocks for tracing:
326 [local] global counter x86-tsc
328 local: Default clock, but may not be in sync across CPUs
330 global: This clock is in sync with all CPUs but may
331 be a bit slower than the local clock.
333 counter: This is not a clock at all, but literally an atomic
334 counter. It counts up one by one, but is in sync
335 with all CPUs. This is useful when you need to
336 know exactly the order events occurred with respect to
337 each other on different CPUs.
339 uptime: This uses the jiffies counter and the time stamp
340 is relative to the time since boot up.
342 perf: This makes ftrace use the same clock that perf uses.
343 Eventually perf will be able to read ftrace buffers
344 and this will help out in interleaving the data.
346 x86-tsc: Architectures may define their own clocks. For
347 example, x86 uses its own TSC cycle clock here.
349 To set a clock, simply echo the clock name into this file.
351 echo global > trace_clock
355 This is a very useful file for synchronizing user space
356 with events happening in the kernel. Writing strings into
357 this file will be written into the ftrace buffer.
359 It is useful in applications to open this file at the start
360 of the application and just reference the file descriptor
363 void trace_write(const char *fmt, ...)
373 n = vsnprintf(buf, 256, fmt, ap);
376 write(trace_fd, buf, n);
381 trace_fd = open("trace_marker", WR_ONLY);
385 Add dynamic tracepoints in programs.
390 Uprobe statistics. See uprobetrace.txt
394 This is a way to make multiple trace buffers where different
395 events can be recorded in different buffers.
396 See "Instances" section below.
400 This is the trace event directory. It holds event tracepoints
401 (also known as static tracepoints) that have been compiled
402 into the kernel. It shows what event tracepoints exist
403 and how they are grouped by system. There are "enable"
404 files at various levels that can enable the tracepoints
405 when a "1" is written to them.
407 See events.txt for more information.
411 This is a directory that contains the trace per_cpu information.
413 per_cpu/cpu0/buffer_size_kb:
415 The ftrace buffer is defined per_cpu. That is, there's a separate
416 buffer for each CPU to allow writes to be done atomically,
417 and free from cache bouncing. These buffers may have different
418 size buffers. This file is similar to the buffer_size_kb
419 file, but it only displays or sets the buffer size for the
420 specific CPU. (here cpu0).
424 This is similar to the "trace" file, but it will only display
425 the data specific for the CPU. If written to, it only clears
426 the specific CPU buffer.
428 per_cpu/cpu0/trace_pipe
430 This is similar to the "trace_pipe" file, and is a consuming
431 read, but it will only display (and consume) the data specific
434 per_cpu/cpu0/trace_pipe_raw
436 For tools that can parse the ftrace ring buffer binary format,
437 the trace_pipe_raw file can be used to extract the data
438 from the ring buffer directly. With the use of the splice()
439 system call, the buffer data can be quickly transferred to
440 a file or to the network where a server is collecting the
443 Like trace_pipe, this is a consuming reader, where multiple
444 reads will always produce different data.
446 per_cpu/cpu0/snapshot:
448 This is similar to the main "snapshot" file, but will only
449 snapshot the current CPU (if supported). It only displays
450 the content of the snapshot for a given CPU, and if
451 written to, only clears this CPU buffer.
453 per_cpu/cpu0/snapshot_raw:
455 Similar to the trace_pipe_raw, but will read the binary format
456 from the snapshot buffer for the given CPU.
460 This displays certain stats about the ring buffer:
462 entries: The number of events that are still in the buffer.
464 overrun: The number of lost events due to overwriting when
467 commit overrun: Should always be zero.
468 This gets set if so many events happened within a nested
469 event (ring buffer is re-entrant), that it fills the
470 buffer and starts dropping events.
472 bytes: Bytes actually read (not overwritten).
474 oldest event ts: The oldest timestamp in the buffer
476 now ts: The current timestamp
478 dropped events: Events lost due to overwrite option being off.
480 read events: The number of events read.
485 Here is the list of current tracers that may be configured.
489 Function call tracer to trace all kernel functions.
493 Similar to the function tracer except that the
494 function tracer probes the functions on their entry
495 whereas the function graph tracer traces on both entry
496 and exit of the functions. It then provides the ability
497 to draw a graph of function calls similar to C code
502 Traces the areas that disable interrupts and saves
503 the trace with the longest max latency.
504 See tracing_max_latency. When a new max is recorded,
505 it replaces the old trace. It is best to view this
506 trace with the latency-format option enabled.
510 Similar to irqsoff but traces and records the amount of
511 time for which preemption is disabled.
515 Similar to irqsoff and preemptoff, but traces and
516 records the largest time for which irqs and/or preemption
521 Traces and records the max latency that it takes for
522 the highest priority task to get scheduled after
523 it has been woken up.
524 Traces all tasks as an average developer would expect.
528 Traces and records the max latency that it takes for just
529 RT tasks (as the current "wakeup" does). This is useful
530 for those interested in wake up timings of RT tasks.
534 This is the "trace nothing" tracer. To remove all
535 tracers from tracing simply echo "nop" into
539 Examples of using the tracer
540 ----------------------------
542 Here are typical examples of using the tracers when controlling
543 them only with the debugfs interface (without using any
544 user-land utilities).
549 Here is an example of the output format of the file "trace"
554 # entries-in-buffer/entries-written: 140080/250280 #P:4
557 # / _----=> need-resched
558 # | / _---=> hardirq/softirq
559 # || / _--=> preempt-depth
561 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
563 bash-1977 [000] .... 17284.993652: sys_close <-system_call_fastpath
564 bash-1977 [000] .... 17284.993653: __close_fd <-sys_close
565 bash-1977 [000] .... 17284.993653: _raw_spin_lock <-__close_fd
566 sshd-1974 [003] .... 17284.993653: __srcu_read_unlock <-fsnotify
567 bash-1977 [000] .... 17284.993654: add_preempt_count <-_raw_spin_lock
568 bash-1977 [000] ...1 17284.993655: _raw_spin_unlock <-__close_fd
569 bash-1977 [000] ...1 17284.993656: sub_preempt_count <-_raw_spin_unlock
570 bash-1977 [000] .... 17284.993657: filp_close <-__close_fd
571 bash-1977 [000] .... 17284.993657: dnotify_flush <-filp_close
572 sshd-1974 [003] .... 17284.993658: sys_select <-system_call_fastpath
575 A header is printed with the tracer name that is represented by
576 the trace. In this case the tracer is "function". Then it shows the
577 number of events in the buffer as well as the total number of entries
578 that were written. The difference is the number of entries that were
579 lost due to the buffer filling up (250280 - 140080 = 110200 events
582 The header explains the content of the events. Task name "bash", the task
583 PID "1977", the CPU that it was running on "000", the latency format
584 (explained below), the timestamp in <secs>.<usecs> format, the
585 function name that was traced "sys_close" and the parent function that
586 called this function "system_call_fastpath". The timestamp is the time
587 at which the function was entered.
592 When the latency-format option is enabled or when one of the latency
593 tracers is set, the trace file gives somewhat more information to see
594 why a latency happened. Here is a typical trace.
598 # irqsoff latency trace v1.1.5 on 3.8.0-test+
599 # --------------------------------------------------------------------
600 # latency: 259 us, #4/4, CPU#2 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
602 # | task: ps-6143 (uid:0 nice:0 policy:0 rt_prio:0)
604 # => started at: __lock_task_sighand
605 # => ended at: _raw_spin_unlock_irqrestore
609 # / _-----=> irqs-off
610 # | / _----=> need-resched
611 # || / _---=> hardirq/softirq
612 # ||| / _--=> preempt-depth
614 # cmd pid ||||| time | caller
616 ps-6143 2d... 0us!: trace_hardirqs_off <-__lock_task_sighand
617 ps-6143 2d..1 259us+: trace_hardirqs_on <-_raw_spin_unlock_irqrestore
618 ps-6143 2d..1 263us+: time_hardirqs_on <-_raw_spin_unlock_irqrestore
619 ps-6143 2d..1 306us : <stack trace>
620 => trace_hardirqs_on_caller
622 => _raw_spin_unlock_irqrestore
629 => system_call_fastpath
632 This shows that the current tracer is "irqsoff" tracing the time
633 for which interrupts were disabled. It gives the trace version (which
634 never changes) and the version of the kernel upon which this was executed on
635 (3.10). Then it displays the max latency in microseconds (259 us). The number
636 of trace entries displayed and the total number (both are four: #4/4).
637 VP, KP, SP, and HP are always zero and are reserved for later use.
638 #P is the number of online CPUs (#P:4).
640 The task is the process that was running when the latency
641 occurred. (ps pid: 6143).
643 The start and stop (the functions in which the interrupts were
644 disabled and enabled respectively) that caused the latencies:
646 __lock_task_sighand is where the interrupts were disabled.
647 _raw_spin_unlock_irqrestore is where they were enabled again.
649 The next lines after the header are the trace itself. The header
650 explains which is which.
652 cmd: The name of the process in the trace.
654 pid: The PID of that process.
656 CPU#: The CPU which the process was running on.
658 irqs-off: 'd' interrupts are disabled. '.' otherwise.
659 Note: If the architecture does not support a way to
660 read the irq flags variable, an 'X' will always
664 'N' both TIF_NEED_RESCHED and PREEMPT_NEED_RESCHED is set,
665 'n' only TIF_NEED_RESCHED is set,
666 'p' only PREEMPT_NEED_RESCHED is set,
670 'H' - hard irq occurred inside a softirq.
671 'h' - hard irq is running
672 's' - soft irq is running
673 '.' - normal context.
675 preempt-depth: The level of preempt_disabled
677 The above is mostly meaningful for kernel developers.
679 time: When the latency-format option is enabled, the trace file
680 output includes a timestamp relative to the start of the
681 trace. This differs from the output when latency-format
682 is disabled, which includes an absolute timestamp.
684 delay: This is just to help catch your eye a bit better. And
685 needs to be fixed to be only relative to the same CPU.
686 The marks are determined by the difference between this
687 current trace and the next trace.
688 '!' - greater than preempt_mark_thresh (default 100)
689 '+' - greater than 1 microsecond
690 ' ' - less than or equal to 1 microsecond.
692 The rest is the same as the 'trace' file.
694 Note, the latency tracers will usually end with a back trace
695 to easily find where the latency occurred.
700 The trace_options file (or the options directory) is used to control
701 what gets printed in the trace output, or manipulate the tracers.
702 To see what is available, simply cat the file:
732 To disable one of the options, echo in the option prepended with
735 echo noprint-parent > trace_options
737 To enable an option, leave off the "no".
739 echo sym-offset > trace_options
741 Here are the available options:
743 print-parent - On function traces, display the calling (parent)
744 function as well as the function being traced.
747 bash-4000 [01] 1477.606694: simple_strtoul <-kstrtoul
750 bash-4000 [01] 1477.606694: simple_strtoul
753 sym-offset - Display not only the function name, but also the
754 offset in the function. For example, instead of
755 seeing just "ktime_get", you will see
756 "ktime_get+0xb/0x20".
759 bash-4000 [01] 1477.606694: simple_strtoul+0x6/0xa0
761 sym-addr - this will also display the function address as well
762 as the function name.
765 bash-4000 [01] 1477.606694: simple_strtoul <c0339346>
767 verbose - This deals with the trace file when the
768 latency-format option is enabled.
770 bash 4000 1 0 00000000 00010a95 [58127d26] 1720.415ms \
771 (+0.000ms): simple_strtoul (kstrtoul)
773 raw - This will display raw numbers. This option is best for
774 use with user applications that can translate the raw
775 numbers better than having it done in the kernel.
777 hex - Similar to raw, but the numbers will be in a hexadecimal
780 bin - This will print out the formats in raw binary.
782 block - When set, reading trace_pipe will not block when polled.
784 stacktrace - This is one of the options that changes the trace
785 itself. When a trace is recorded, so is the stack
786 of functions. This allows for back traces of
789 trace_printk - Can disable trace_printk() from writing into the buffer.
791 branch - Enable branch tracing with the tracer.
793 annotate - It is sometimes confusing when the CPU buffers are full
794 and one CPU buffer had a lot of events recently, thus
795 a shorter time frame, were another CPU may have only had
796 a few events, which lets it have older events. When
797 the trace is reported, it shows the oldest events first,
798 and it may look like only one CPU ran (the one with the
799 oldest events). When the annotate option is set, it will
800 display when a new CPU buffer started:
802 <idle>-0 [001] dNs4 21169.031481: wake_up_idle_cpu <-add_timer_on
803 <idle>-0 [001] dNs4 21169.031482: _raw_spin_unlock_irqrestore <-add_timer_on
804 <idle>-0 [001] .Ns4 21169.031484: sub_preempt_count <-_raw_spin_unlock_irqrestore
805 ##### CPU 2 buffer started ####
806 <idle>-0 [002] .N.1 21169.031484: rcu_idle_exit <-cpu_idle
807 <idle>-0 [001] .Ns3 21169.031484: _raw_spin_unlock <-clocksource_watchdog
808 <idle>-0 [001] .Ns3 21169.031485: sub_preempt_count <-_raw_spin_unlock
810 userstacktrace - This option changes the trace. It records a
811 stacktrace of the current userspace thread.
813 sym-userobj - when user stacktrace are enabled, look up which
814 object the address belongs to, and print a
815 relative address. This is especially useful when
816 ASLR is on, otherwise you don't get a chance to
817 resolve the address to object/file/line after
818 the app is no longer running
820 The lookup is performed when you read
821 trace,trace_pipe. Example:
823 a.out-1623 [000] 40874.465068: /root/a.out[+0x480] <-/root/a.out[+0
824 x494] <- /root/a.out[+0x4a8] <- /lib/libc-2.7.so[+0x1e1a6]
827 printk-msg-only - When set, trace_printk()s will only show the format
828 and not their parameters (if trace_bprintk() or
829 trace_bputs() was used to save the trace_printk()).
831 context-info - Show only the event data. Hides the comm, PID,
832 timestamp, CPU, and other useful data.
834 latency-format - This option changes the trace. When
835 it is enabled, the trace displays
836 additional information about the
837 latencies, as described in "Latency
840 sleep-time - When running function graph tracer, to include
841 the time a task schedules out in its function.
842 When enabled, it will account time the task has been
843 scheduled out as part of the function call.
845 graph-time - When running function graph tracer, to include the
846 time to call nested functions. When this is not set,
847 the time reported for the function will only include
848 the time the function itself executed for, not the time
849 for functions that it called.
851 record-cmd - When any event or tracer is enabled, a hook is enabled
852 in the sched_switch trace point to fill comm cache
853 with mapped pids and comms. But this may cause some
854 overhead, and if you only care about pids, and not the
855 name of the task, disabling this option can lower the
858 overwrite - This controls what happens when the trace buffer is
859 full. If "1" (default), the oldest events are
860 discarded and overwritten. If "0", then the newest
861 events are discarded.
862 (see per_cpu/cpu0/stats for overrun and dropped)
864 disable_on_free - When the free_buffer is closed, tracing will
865 stop (tracing_on set to 0).
867 irq-info - Shows the interrupt, preempt count, need resched data.
868 When disabled, the trace looks like:
872 # entries-in-buffer/entries-written: 144405/9452052 #P:4
874 # TASK-PID CPU# TIMESTAMP FUNCTION
876 <idle>-0 [002] 23636.756054: ttwu_do_activate.constprop.89 <-try_to_wake_up
877 <idle>-0 [002] 23636.756054: activate_task <-ttwu_do_activate.constprop.89
878 <idle>-0 [002] 23636.756055: enqueue_task <-activate_task
881 markers - When set, the trace_marker is writable (only by root).
882 When disabled, the trace_marker will error with EINVAL
886 function-trace - The latency tracers will enable function tracing
887 if this option is enabled (default it is). When
888 it is disabled, the latency tracers do not trace
889 functions. This keeps the overhead of the tracer down
890 when performing latency tests.
892 Note: Some tracers have their own options. They only appear
893 when the tracer is active.
900 When interrupts are disabled, the CPU can not react to any other
901 external event (besides NMIs and SMIs). This prevents the timer
902 interrupt from triggering or the mouse interrupt from letting
903 the kernel know of a new mouse event. The result is a latency
904 with the reaction time.
906 The irqsoff tracer tracks the time for which interrupts are
907 disabled. When a new maximum latency is hit, the tracer saves
908 the trace leading up to that latency point so that every time a
909 new maximum is reached, the old saved trace is discarded and the
912 To reset the maximum, echo 0 into tracing_max_latency. Here is
915 # echo 0 > options/function-trace
916 # echo irqsoff > current_tracer
917 # echo 1 > tracing_on
918 # echo 0 > tracing_max_latency
921 # echo 0 > tracing_on
925 # irqsoff latency trace v1.1.5 on 3.8.0-test+
926 # --------------------------------------------------------------------
927 # latency: 16 us, #4/4, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
929 # | task: swapper/0-0 (uid:0 nice:0 policy:0 rt_prio:0)
931 # => started at: run_timer_softirq
932 # => ended at: run_timer_softirq
936 # / _-----=> irqs-off
937 # | / _----=> need-resched
938 # || / _---=> hardirq/softirq
939 # ||| / _--=> preempt-depth
941 # cmd pid ||||| time | caller
943 <idle>-0 0d.s2 0us+: _raw_spin_lock_irq <-run_timer_softirq
944 <idle>-0 0dNs3 17us : _raw_spin_unlock_irq <-run_timer_softirq
945 <idle>-0 0dNs3 17us+: trace_hardirqs_on <-run_timer_softirq
946 <idle>-0 0dNs3 25us : <stack trace>
947 => _raw_spin_unlock_irq
953 => smp_apic_timer_interrupt
954 => apic_timer_interrupt
959 => x86_64_start_reservations
960 => x86_64_start_kernel
962 Here we see that that we had a latency of 16 microseconds (which is
963 very good). The _raw_spin_lock_irq in run_timer_softirq disabled
964 interrupts. The difference between the 16 and the displayed
965 timestamp 25us occurred because the clock was incremented
966 between the time of recording the max latency and the time of
967 recording the function that had that latency.
969 Note the above example had function-trace not set. If we set
970 function-trace, we get a much larger output:
972 with echo 1 > options/function-trace
976 # irqsoff latency trace v1.1.5 on 3.8.0-test+
977 # --------------------------------------------------------------------
978 # latency: 71 us, #168/168, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
980 # | task: bash-2042 (uid:0 nice:0 policy:0 rt_prio:0)
982 # => started at: ata_scsi_queuecmd
983 # => ended at: ata_scsi_queuecmd
987 # / _-----=> irqs-off
988 # | / _----=> need-resched
989 # || / _---=> hardirq/softirq
990 # ||| / _--=> preempt-depth
992 # cmd pid ||||| time | caller
994 bash-2042 3d... 0us : _raw_spin_lock_irqsave <-ata_scsi_queuecmd
995 bash-2042 3d... 0us : add_preempt_count <-_raw_spin_lock_irqsave
996 bash-2042 3d..1 1us : ata_scsi_find_dev <-ata_scsi_queuecmd
997 bash-2042 3d..1 1us : __ata_scsi_find_dev <-ata_scsi_find_dev
998 bash-2042 3d..1 2us : ata_find_dev.part.14 <-__ata_scsi_find_dev
999 bash-2042 3d..1 2us : ata_qc_new_init <-__ata_scsi_queuecmd
1000 bash-2042 3d..1 3us : ata_sg_init <-__ata_scsi_queuecmd
1001 bash-2042 3d..1 4us : ata_scsi_rw_xlat <-__ata_scsi_queuecmd
1002 bash-2042 3d..1 4us : ata_build_rw_tf <-ata_scsi_rw_xlat
1004 bash-2042 3d..1 67us : delay_tsc <-__delay
1005 bash-2042 3d..1 67us : add_preempt_count <-delay_tsc
1006 bash-2042 3d..2 67us : sub_preempt_count <-delay_tsc
1007 bash-2042 3d..1 67us : add_preempt_count <-delay_tsc
1008 bash-2042 3d..2 68us : sub_preempt_count <-delay_tsc
1009 bash-2042 3d..1 68us+: ata_bmdma_start <-ata_bmdma_qc_issue
1010 bash-2042 3d..1 71us : _raw_spin_unlock_irqrestore <-ata_scsi_queuecmd
1011 bash-2042 3d..1 71us : _raw_spin_unlock_irqrestore <-ata_scsi_queuecmd
1012 bash-2042 3d..1 72us+: trace_hardirqs_on <-ata_scsi_queuecmd
1013 bash-2042 3d..1 120us : <stack trace>
1014 => _raw_spin_unlock_irqrestore
1015 => ata_scsi_queuecmd
1016 => scsi_dispatch_cmd
1018 => __blk_run_queue_uncond
1021 => generic_make_request
1024 => __ext3_get_inode_loc
1033 => user_path_at_empty
1038 => system_call_fastpath
1041 Here we traced a 71 microsecond latency. But we also see all the
1042 functions that were called during that time. Note that by
1043 enabling function tracing, we incur an added overhead. This
1044 overhead may extend the latency times. But nevertheless, this
1045 trace has provided some very helpful debugging information.
1051 When preemption is disabled, we may be able to receive
1052 interrupts but the task cannot be preempted and a higher
1053 priority task must wait for preemption to be enabled again
1054 before it can preempt a lower priority task.
1056 The preemptoff tracer traces the places that disable preemption.
1057 Like the irqsoff tracer, it records the maximum latency for
1058 which preemption was disabled. The control of preemptoff tracer
1059 is much like the irqsoff tracer.
1061 # echo 0 > options/function-trace
1062 # echo preemptoff > current_tracer
1063 # echo 1 > tracing_on
1064 # echo 0 > tracing_max_latency
1067 # echo 0 > tracing_on
1069 # tracer: preemptoff
1071 # preemptoff latency trace v1.1.5 on 3.8.0-test+
1072 # --------------------------------------------------------------------
1073 # latency: 46 us, #4/4, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1075 # | task: sshd-1991 (uid:0 nice:0 policy:0 rt_prio:0)
1077 # => started at: do_IRQ
1078 # => ended at: do_IRQ
1082 # / _-----=> irqs-off
1083 # | / _----=> need-resched
1084 # || / _---=> hardirq/softirq
1085 # ||| / _--=> preempt-depth
1087 # cmd pid ||||| time | caller
1089 sshd-1991 1d.h. 0us+: irq_enter <-do_IRQ
1090 sshd-1991 1d..1 46us : irq_exit <-do_IRQ
1091 sshd-1991 1d..1 47us+: trace_preempt_on <-do_IRQ
1092 sshd-1991 1d..1 52us : <stack trace>
1093 => sub_preempt_count
1099 This has some more changes. Preemption was disabled when an
1100 interrupt came in (notice the 'h'), and was enabled on exit.
1101 But we also see that interrupts have been disabled when entering
1102 the preempt off section and leaving it (the 'd'). We do not know if
1103 interrupts were enabled in the mean time or shortly after this
1106 # tracer: preemptoff
1108 # preemptoff latency trace v1.1.5 on 3.8.0-test+
1109 # --------------------------------------------------------------------
1110 # latency: 83 us, #241/241, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1112 # | task: bash-1994 (uid:0 nice:0 policy:0 rt_prio:0)
1114 # => started at: wake_up_new_task
1115 # => ended at: task_rq_unlock
1119 # / _-----=> irqs-off
1120 # | / _----=> need-resched
1121 # || / _---=> hardirq/softirq
1122 # ||| / _--=> preempt-depth
1124 # cmd pid ||||| time | caller
1126 bash-1994 1d..1 0us : _raw_spin_lock_irqsave <-wake_up_new_task
1127 bash-1994 1d..1 0us : select_task_rq_fair <-select_task_rq
1128 bash-1994 1d..1 1us : __rcu_read_lock <-select_task_rq_fair
1129 bash-1994 1d..1 1us : source_load <-select_task_rq_fair
1130 bash-1994 1d..1 1us : source_load <-select_task_rq_fair
1132 bash-1994 1d..1 12us : irq_enter <-smp_apic_timer_interrupt
1133 bash-1994 1d..1 12us : rcu_irq_enter <-irq_enter
1134 bash-1994 1d..1 13us : add_preempt_count <-irq_enter
1135 bash-1994 1d.h1 13us : exit_idle <-smp_apic_timer_interrupt
1136 bash-1994 1d.h1 13us : hrtimer_interrupt <-smp_apic_timer_interrupt
1137 bash-1994 1d.h1 13us : _raw_spin_lock <-hrtimer_interrupt
1138 bash-1994 1d.h1 14us : add_preempt_count <-_raw_spin_lock
1139 bash-1994 1d.h2 14us : ktime_get_update_offsets <-hrtimer_interrupt
1141 bash-1994 1d.h1 35us : lapic_next_event <-clockevents_program_event
1142 bash-1994 1d.h1 35us : irq_exit <-smp_apic_timer_interrupt
1143 bash-1994 1d.h1 36us : sub_preempt_count <-irq_exit
1144 bash-1994 1d..2 36us : do_softirq <-irq_exit
1145 bash-1994 1d..2 36us : __do_softirq <-call_softirq
1146 bash-1994 1d..2 36us : __local_bh_disable <-__do_softirq
1147 bash-1994 1d.s2 37us : add_preempt_count <-_raw_spin_lock_irq
1148 bash-1994 1d.s3 38us : _raw_spin_unlock <-run_timer_softirq
1149 bash-1994 1d.s3 39us : sub_preempt_count <-_raw_spin_unlock
1150 bash-1994 1d.s2 39us : call_timer_fn <-run_timer_softirq
1152 bash-1994 1dNs2 81us : cpu_needs_another_gp <-rcu_process_callbacks
1153 bash-1994 1dNs2 82us : __local_bh_enable <-__do_softirq
1154 bash-1994 1dNs2 82us : sub_preempt_count <-__local_bh_enable
1155 bash-1994 1dN.2 82us : idle_cpu <-irq_exit
1156 bash-1994 1dN.2 83us : rcu_irq_exit <-irq_exit
1157 bash-1994 1dN.2 83us : sub_preempt_count <-irq_exit
1158 bash-1994 1.N.1 84us : _raw_spin_unlock_irqrestore <-task_rq_unlock
1159 bash-1994 1.N.1 84us+: trace_preempt_on <-task_rq_unlock
1160 bash-1994 1.N.1 104us : <stack trace>
1161 => sub_preempt_count
1162 => _raw_spin_unlock_irqrestore
1170 The above is an example of the preemptoff trace with
1171 function-trace set. Here we see that interrupts were not disabled
1172 the entire time. The irq_enter code lets us know that we entered
1173 an interrupt 'h'. Before that, the functions being traced still
1174 show that it is not in an interrupt, but we can see from the
1175 functions themselves that this is not the case.
1180 Knowing the locations that have interrupts disabled or
1181 preemption disabled for the longest times is helpful. But
1182 sometimes we would like to know when either preemption and/or
1183 interrupts are disabled.
1185 Consider the following code:
1187 local_irq_disable();
1188 call_function_with_irqs_off();
1190 call_function_with_irqs_and_preemption_off();
1192 call_function_with_preemption_off();
1195 The irqsoff tracer will record the total length of
1196 call_function_with_irqs_off() and
1197 call_function_with_irqs_and_preemption_off().
1199 The preemptoff tracer will record the total length of
1200 call_function_with_irqs_and_preemption_off() and
1201 call_function_with_preemption_off().
1203 But neither will trace the time that interrupts and/or
1204 preemption is disabled. This total time is the time that we can
1205 not schedule. To record this time, use the preemptirqsoff
1208 Again, using this trace is much like the irqsoff and preemptoff
1211 # echo 0 > options/function-trace
1212 # echo preemptirqsoff > current_tracer
1213 # echo 1 > tracing_on
1214 # echo 0 > tracing_max_latency
1217 # echo 0 > tracing_on
1219 # tracer: preemptirqsoff
1221 # preemptirqsoff latency trace v1.1.5 on 3.8.0-test+
1222 # --------------------------------------------------------------------
1223 # latency: 100 us, #4/4, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1225 # | task: ls-2230 (uid:0 nice:0 policy:0 rt_prio:0)
1227 # => started at: ata_scsi_queuecmd
1228 # => ended at: ata_scsi_queuecmd
1232 # / _-----=> irqs-off
1233 # | / _----=> need-resched
1234 # || / _---=> hardirq/softirq
1235 # ||| / _--=> preempt-depth
1237 # cmd pid ||||| time | caller
1239 ls-2230 3d... 0us+: _raw_spin_lock_irqsave <-ata_scsi_queuecmd
1240 ls-2230 3...1 100us : _raw_spin_unlock_irqrestore <-ata_scsi_queuecmd
1241 ls-2230 3...1 101us+: trace_preempt_on <-ata_scsi_queuecmd
1242 ls-2230 3...1 111us : <stack trace>
1243 => sub_preempt_count
1244 => _raw_spin_unlock_irqrestore
1245 => ata_scsi_queuecmd
1246 => scsi_dispatch_cmd
1248 => __blk_run_queue_uncond
1251 => generic_make_request
1256 => htree_dirblock_to_tree
1257 => ext3_htree_fill_tree
1261 => system_call_fastpath
1264 The trace_hardirqs_off_thunk is called from assembly on x86 when
1265 interrupts are disabled in the assembly code. Without the
1266 function tracing, we do not know if interrupts were enabled
1267 within the preemption points. We do see that it started with
1270 Here is a trace with function-trace set:
1272 # tracer: preemptirqsoff
1274 # preemptirqsoff latency trace v1.1.5 on 3.8.0-test+
1275 # --------------------------------------------------------------------
1276 # latency: 161 us, #339/339, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1278 # | task: ls-2269 (uid:0 nice:0 policy:0 rt_prio:0)
1280 # => started at: schedule
1281 # => ended at: mutex_unlock
1285 # / _-----=> irqs-off
1286 # | / _----=> need-resched
1287 # || / _---=> hardirq/softirq
1288 # ||| / _--=> preempt-depth
1290 # cmd pid ||||| time | caller
1292 kworker/-59 3...1 0us : __schedule <-schedule
1293 kworker/-59 3d..1 0us : rcu_preempt_qs <-rcu_note_context_switch
1294 kworker/-59 3d..1 1us : add_preempt_count <-_raw_spin_lock_irq
1295 kworker/-59 3d..2 1us : deactivate_task <-__schedule
1296 kworker/-59 3d..2 1us : dequeue_task <-deactivate_task
1297 kworker/-59 3d..2 2us : update_rq_clock <-dequeue_task
1298 kworker/-59 3d..2 2us : dequeue_task_fair <-dequeue_task
1299 kworker/-59 3d..2 2us : update_curr <-dequeue_task_fair
1300 kworker/-59 3d..2 2us : update_min_vruntime <-update_curr
1301 kworker/-59 3d..2 3us : cpuacct_charge <-update_curr
1302 kworker/-59 3d..2 3us : __rcu_read_lock <-cpuacct_charge
1303 kworker/-59 3d..2 3us : __rcu_read_unlock <-cpuacct_charge
1304 kworker/-59 3d..2 3us : update_cfs_rq_blocked_load <-dequeue_task_fair
1305 kworker/-59 3d..2 4us : clear_buddies <-dequeue_task_fair
1306 kworker/-59 3d..2 4us : account_entity_dequeue <-dequeue_task_fair
1307 kworker/-59 3d..2 4us : update_min_vruntime <-dequeue_task_fair
1308 kworker/-59 3d..2 4us : update_cfs_shares <-dequeue_task_fair
1309 kworker/-59 3d..2 5us : hrtick_update <-dequeue_task_fair
1310 kworker/-59 3d..2 5us : wq_worker_sleeping <-__schedule
1311 kworker/-59 3d..2 5us : kthread_data <-wq_worker_sleeping
1312 kworker/-59 3d..2 5us : put_prev_task_fair <-__schedule
1313 kworker/-59 3d..2 6us : pick_next_task_fair <-pick_next_task
1314 kworker/-59 3d..2 6us : clear_buddies <-pick_next_task_fair
1315 kworker/-59 3d..2 6us : set_next_entity <-pick_next_task_fair
1316 kworker/-59 3d..2 6us : update_stats_wait_end <-set_next_entity
1317 ls-2269 3d..2 7us : finish_task_switch <-__schedule
1318 ls-2269 3d..2 7us : _raw_spin_unlock_irq <-finish_task_switch
1319 ls-2269 3d..2 8us : do_IRQ <-ret_from_intr
1320 ls-2269 3d..2 8us : irq_enter <-do_IRQ
1321 ls-2269 3d..2 8us : rcu_irq_enter <-irq_enter
1322 ls-2269 3d..2 9us : add_preempt_count <-irq_enter
1323 ls-2269 3d.h2 9us : exit_idle <-do_IRQ
1325 ls-2269 3d.h3 20us : sub_preempt_count <-_raw_spin_unlock
1326 ls-2269 3d.h2 20us : irq_exit <-do_IRQ
1327 ls-2269 3d.h2 21us : sub_preempt_count <-irq_exit
1328 ls-2269 3d..3 21us : do_softirq <-irq_exit
1329 ls-2269 3d..3 21us : __do_softirq <-call_softirq
1330 ls-2269 3d..3 21us+: __local_bh_disable <-__do_softirq
1331 ls-2269 3d.s4 29us : sub_preempt_count <-_local_bh_enable_ip
1332 ls-2269 3d.s5 29us : sub_preempt_count <-_local_bh_enable_ip
1333 ls-2269 3d.s5 31us : do_IRQ <-ret_from_intr
1334 ls-2269 3d.s5 31us : irq_enter <-do_IRQ
1335 ls-2269 3d.s5 31us : rcu_irq_enter <-irq_enter
1337 ls-2269 3d.s5 31us : rcu_irq_enter <-irq_enter
1338 ls-2269 3d.s5 32us : add_preempt_count <-irq_enter
1339 ls-2269 3d.H5 32us : exit_idle <-do_IRQ
1340 ls-2269 3d.H5 32us : handle_irq <-do_IRQ
1341 ls-2269 3d.H5 32us : irq_to_desc <-handle_irq
1342 ls-2269 3d.H5 33us : handle_fasteoi_irq <-handle_irq
1344 ls-2269 3d.s5 158us : _raw_spin_unlock_irqrestore <-rtl8139_poll
1345 ls-2269 3d.s3 158us : net_rps_action_and_irq_enable.isra.65 <-net_rx_action
1346 ls-2269 3d.s3 159us : __local_bh_enable <-__do_softirq
1347 ls-2269 3d.s3 159us : sub_preempt_count <-__local_bh_enable
1348 ls-2269 3d..3 159us : idle_cpu <-irq_exit
1349 ls-2269 3d..3 159us : rcu_irq_exit <-irq_exit
1350 ls-2269 3d..3 160us : sub_preempt_count <-irq_exit
1351 ls-2269 3d... 161us : __mutex_unlock_slowpath <-mutex_unlock
1352 ls-2269 3d... 162us+: trace_hardirqs_on <-mutex_unlock
1353 ls-2269 3d... 186us : <stack trace>
1354 => __mutex_unlock_slowpath
1361 => system_call_fastpath
1363 This is an interesting trace. It started with kworker running and
1364 scheduling out and ls taking over. But as soon as ls released the
1365 rq lock and enabled interrupts (but not preemption) an interrupt
1366 triggered. When the interrupt finished, it started running softirqs.
1367 But while the softirq was running, another interrupt triggered.
1368 When an interrupt is running inside a softirq, the annotation is 'H'.
1374 One common case that people are interested in tracing is the
1375 time it takes for a task that is woken to actually wake up.
1376 Now for non Real-Time tasks, this can be arbitrary. But tracing
1377 it none the less can be interesting.
1379 Without function tracing:
1381 # echo 0 > options/function-trace
1382 # echo wakeup > current_tracer
1383 # echo 1 > tracing_on
1384 # echo 0 > tracing_max_latency
1386 # echo 0 > tracing_on
1390 # wakeup latency trace v1.1.5 on 3.8.0-test+
1391 # --------------------------------------------------------------------
1392 # latency: 15 us, #4/4, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1394 # | task: kworker/3:1H-312 (uid:0 nice:-20 policy:0 rt_prio:0)
1398 # / _-----=> irqs-off
1399 # | / _----=> need-resched
1400 # || / _---=> hardirq/softirq
1401 # ||| / _--=> preempt-depth
1403 # cmd pid ||||| time | caller
1405 <idle>-0 3dNs7 0us : 0:120:R + [003] 312:100:R kworker/3:1H
1406 <idle>-0 3dNs7 1us+: ttwu_do_activate.constprop.87 <-try_to_wake_up
1407 <idle>-0 3d..3 15us : __schedule <-schedule
1408 <idle>-0 3d..3 15us : 0:120:R ==> [003] 312:100:R kworker/3:1H
1410 The tracer only traces the highest priority task in the system
1411 to avoid tracing the normal circumstances. Here we see that
1412 the kworker with a nice priority of -20 (not very nice), took
1413 just 15 microseconds from the time it woke up, to the time it
1416 Non Real-Time tasks are not that interesting. A more interesting
1417 trace is to concentrate only on Real-Time tasks.
1422 In a Real-Time environment it is very important to know the
1423 wakeup time it takes for the highest priority task that is woken
1424 up to the time that it executes. This is also known as "schedule
1425 latency". I stress the point that this is about RT tasks. It is
1426 also important to know the scheduling latency of non-RT tasks,
1427 but the average schedule latency is better for non-RT tasks.
1428 Tools like LatencyTop are more appropriate for such
1431 Real-Time environments are interested in the worst case latency.
1432 That is the longest latency it takes for something to happen,
1433 and not the average. We can have a very fast scheduler that may
1434 only have a large latency once in a while, but that would not
1435 work well with Real-Time tasks. The wakeup_rt tracer was designed
1436 to record the worst case wakeups of RT tasks. Non-RT tasks are
1437 not recorded because the tracer only records one worst case and
1438 tracing non-RT tasks that are unpredictable will overwrite the
1439 worst case latency of RT tasks (just run the normal wakeup
1440 tracer for a while to see that effect).
1442 Since this tracer only deals with RT tasks, we will run this
1443 slightly differently than we did with the previous tracers.
1444 Instead of performing an 'ls', we will run 'sleep 1' under
1445 'chrt' which changes the priority of the task.
1447 # echo 0 > options/function-trace
1448 # echo wakeup_rt > current_tracer
1449 # echo 1 > tracing_on
1450 # echo 0 > tracing_max_latency
1452 # echo 0 > tracing_on
1458 # wakeup_rt latency trace v1.1.5 on 3.8.0-test+
1459 # --------------------------------------------------------------------
1460 # latency: 5 us, #4/4, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1462 # | task: sleep-2389 (uid:0 nice:0 policy:1 rt_prio:5)
1466 # / _-----=> irqs-off
1467 # | / _----=> need-resched
1468 # || / _---=> hardirq/softirq
1469 # ||| / _--=> preempt-depth
1471 # cmd pid ||||| time | caller
1473 <idle>-0 3d.h4 0us : 0:120:R + [003] 2389: 94:R sleep
1474 <idle>-0 3d.h4 1us+: ttwu_do_activate.constprop.87 <-try_to_wake_up
1475 <idle>-0 3d..3 5us : __schedule <-schedule
1476 <idle>-0 3d..3 5us : 0:120:R ==> [003] 2389: 94:R sleep
1479 Running this on an idle system, we see that it only took 5 microseconds
1480 to perform the task switch. Note, since the trace point in the schedule
1481 is before the actual "switch", we stop the tracing when the recorded task
1482 is about to schedule in. This may change if we add a new marker at the
1483 end of the scheduler.
1485 Notice that the recorded task is 'sleep' with the PID of 2389
1486 and it has an rt_prio of 5. This priority is user-space priority
1487 and not the internal kernel priority. The policy is 1 for
1488 SCHED_FIFO and 2 for SCHED_RR.
1490 Note, that the trace data shows the internal priority (99 - rtprio).
1492 <idle>-0 3d..3 5us : 0:120:R ==> [003] 2389: 94:R sleep
1494 The 0:120:R means idle was running with a nice priority of 0 (120 - 20)
1495 and in the running state 'R'. The sleep task was scheduled in with
1496 2389: 94:R. That is the priority is the kernel rtprio (99 - 5 = 94)
1497 and it too is in the running state.
1499 Doing the same with chrt -r 5 and function-trace set.
1501 echo 1 > options/function-trace
1505 # wakeup_rt latency trace v1.1.5 on 3.8.0-test+
1506 # --------------------------------------------------------------------
1507 # latency: 29 us, #85/85, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1509 # | task: sleep-2448 (uid:0 nice:0 policy:1 rt_prio:5)
1513 # / _-----=> irqs-off
1514 # | / _----=> need-resched
1515 # || / _---=> hardirq/softirq
1516 # ||| / _--=> preempt-depth
1518 # cmd pid ||||| time | caller
1520 <idle>-0 3d.h4 1us+: 0:120:R + [003] 2448: 94:R sleep
1521 <idle>-0 3d.h4 2us : ttwu_do_activate.constprop.87 <-try_to_wake_up
1522 <idle>-0 3d.h3 3us : check_preempt_curr <-ttwu_do_wakeup
1523 <idle>-0 3d.h3 3us : resched_curr <-check_preempt_curr
1524 <idle>-0 3dNh3 4us : task_woken_rt <-ttwu_do_wakeup
1525 <idle>-0 3dNh3 4us : _raw_spin_unlock <-try_to_wake_up
1526 <idle>-0 3dNh3 4us : sub_preempt_count <-_raw_spin_unlock
1527 <idle>-0 3dNh2 5us : ttwu_stat <-try_to_wake_up
1528 <idle>-0 3dNh2 5us : _raw_spin_unlock_irqrestore <-try_to_wake_up
1529 <idle>-0 3dNh2 6us : sub_preempt_count <-_raw_spin_unlock_irqrestore
1530 <idle>-0 3dNh1 6us : _raw_spin_lock <-__run_hrtimer
1531 <idle>-0 3dNh1 6us : add_preempt_count <-_raw_spin_lock
1532 <idle>-0 3dNh2 7us : _raw_spin_unlock <-hrtimer_interrupt
1533 <idle>-0 3dNh2 7us : sub_preempt_count <-_raw_spin_unlock
1534 <idle>-0 3dNh1 7us : tick_program_event <-hrtimer_interrupt
1535 <idle>-0 3dNh1 7us : clockevents_program_event <-tick_program_event
1536 <idle>-0 3dNh1 8us : ktime_get <-clockevents_program_event
1537 <idle>-0 3dNh1 8us : lapic_next_event <-clockevents_program_event
1538 <idle>-0 3dNh1 8us : irq_exit <-smp_apic_timer_interrupt
1539 <idle>-0 3dNh1 9us : sub_preempt_count <-irq_exit
1540 <idle>-0 3dN.2 9us : idle_cpu <-irq_exit
1541 <idle>-0 3dN.2 9us : rcu_irq_exit <-irq_exit
1542 <idle>-0 3dN.2 10us : rcu_eqs_enter_common.isra.45 <-rcu_irq_exit
1543 <idle>-0 3dN.2 10us : sub_preempt_count <-irq_exit
1544 <idle>-0 3.N.1 11us : rcu_idle_exit <-cpu_idle
1545 <idle>-0 3dN.1 11us : rcu_eqs_exit_common.isra.43 <-rcu_idle_exit
1546 <idle>-0 3.N.1 11us : tick_nohz_idle_exit <-cpu_idle
1547 <idle>-0 3dN.1 12us : menu_hrtimer_cancel <-tick_nohz_idle_exit
1548 <idle>-0 3dN.1 12us : ktime_get <-tick_nohz_idle_exit
1549 <idle>-0 3dN.1 12us : tick_do_update_jiffies64 <-tick_nohz_idle_exit
1550 <idle>-0 3dN.1 13us : update_cpu_load_nohz <-tick_nohz_idle_exit
1551 <idle>-0 3dN.1 13us : _raw_spin_lock <-update_cpu_load_nohz
1552 <idle>-0 3dN.1 13us : add_preempt_count <-_raw_spin_lock
1553 <idle>-0 3dN.2 13us : __update_cpu_load <-update_cpu_load_nohz
1554 <idle>-0 3dN.2 14us : sched_avg_update <-__update_cpu_load
1555 <idle>-0 3dN.2 14us : _raw_spin_unlock <-update_cpu_load_nohz
1556 <idle>-0 3dN.2 14us : sub_preempt_count <-_raw_spin_unlock
1557 <idle>-0 3dN.1 15us : calc_load_exit_idle <-tick_nohz_idle_exit
1558 <idle>-0 3dN.1 15us : touch_softlockup_watchdog <-tick_nohz_idle_exit
1559 <idle>-0 3dN.1 15us : hrtimer_cancel <-tick_nohz_idle_exit
1560 <idle>-0 3dN.1 15us : hrtimer_try_to_cancel <-hrtimer_cancel
1561 <idle>-0 3dN.1 16us : lock_hrtimer_base.isra.18 <-hrtimer_try_to_cancel
1562 <idle>-0 3dN.1 16us : _raw_spin_lock_irqsave <-lock_hrtimer_base.isra.18
1563 <idle>-0 3dN.1 16us : add_preempt_count <-_raw_spin_lock_irqsave
1564 <idle>-0 3dN.2 17us : __remove_hrtimer <-remove_hrtimer.part.16
1565 <idle>-0 3dN.2 17us : hrtimer_force_reprogram <-__remove_hrtimer
1566 <idle>-0 3dN.2 17us : tick_program_event <-hrtimer_force_reprogram
1567 <idle>-0 3dN.2 18us : clockevents_program_event <-tick_program_event
1568 <idle>-0 3dN.2 18us : ktime_get <-clockevents_program_event
1569 <idle>-0 3dN.2 18us : lapic_next_event <-clockevents_program_event
1570 <idle>-0 3dN.2 19us : _raw_spin_unlock_irqrestore <-hrtimer_try_to_cancel
1571 <idle>-0 3dN.2 19us : sub_preempt_count <-_raw_spin_unlock_irqrestore
1572 <idle>-0 3dN.1 19us : hrtimer_forward <-tick_nohz_idle_exit
1573 <idle>-0 3dN.1 20us : ktime_add_safe <-hrtimer_forward
1574 <idle>-0 3dN.1 20us : ktime_add_safe <-hrtimer_forward
1575 <idle>-0 3dN.1 20us : hrtimer_start_range_ns <-hrtimer_start_expires.constprop.11
1576 <idle>-0 3dN.1 20us : __hrtimer_start_range_ns <-hrtimer_start_range_ns
1577 <idle>-0 3dN.1 21us : lock_hrtimer_base.isra.18 <-__hrtimer_start_range_ns
1578 <idle>-0 3dN.1 21us : _raw_spin_lock_irqsave <-lock_hrtimer_base.isra.18
1579 <idle>-0 3dN.1 21us : add_preempt_count <-_raw_spin_lock_irqsave
1580 <idle>-0 3dN.2 22us : ktime_add_safe <-__hrtimer_start_range_ns
1581 <idle>-0 3dN.2 22us : enqueue_hrtimer <-__hrtimer_start_range_ns
1582 <idle>-0 3dN.2 22us : tick_program_event <-__hrtimer_start_range_ns
1583 <idle>-0 3dN.2 23us : clockevents_program_event <-tick_program_event
1584 <idle>-0 3dN.2 23us : ktime_get <-clockevents_program_event
1585 <idle>-0 3dN.2 23us : lapic_next_event <-clockevents_program_event
1586 <idle>-0 3dN.2 24us : _raw_spin_unlock_irqrestore <-__hrtimer_start_range_ns
1587 <idle>-0 3dN.2 24us : sub_preempt_count <-_raw_spin_unlock_irqrestore
1588 <idle>-0 3dN.1 24us : account_idle_ticks <-tick_nohz_idle_exit
1589 <idle>-0 3dN.1 24us : account_idle_time <-account_idle_ticks
1590 <idle>-0 3.N.1 25us : sub_preempt_count <-cpu_idle
1591 <idle>-0 3.N.. 25us : schedule <-cpu_idle
1592 <idle>-0 3.N.. 25us : __schedule <-preempt_schedule
1593 <idle>-0 3.N.. 26us : add_preempt_count <-__schedule
1594 <idle>-0 3.N.1 26us : rcu_note_context_switch <-__schedule
1595 <idle>-0 3.N.1 26us : rcu_sched_qs <-rcu_note_context_switch
1596 <idle>-0 3dN.1 27us : rcu_preempt_qs <-rcu_note_context_switch
1597 <idle>-0 3.N.1 27us : _raw_spin_lock_irq <-__schedule
1598 <idle>-0 3dN.1 27us : add_preempt_count <-_raw_spin_lock_irq
1599 <idle>-0 3dN.2 28us : put_prev_task_idle <-__schedule
1600 <idle>-0 3dN.2 28us : pick_next_task_stop <-pick_next_task
1601 <idle>-0 3dN.2 28us : pick_next_task_rt <-pick_next_task
1602 <idle>-0 3dN.2 29us : dequeue_pushable_task <-pick_next_task_rt
1603 <idle>-0 3d..3 29us : __schedule <-preempt_schedule
1604 <idle>-0 3d..3 30us : 0:120:R ==> [003] 2448: 94:R sleep
1606 This isn't that big of a trace, even with function tracing enabled,
1607 so I included the entire trace.
1609 The interrupt went off while when the system was idle. Somewhere
1610 before task_woken_rt() was called, the NEED_RESCHED flag was set,
1611 this is indicated by the first occurrence of the 'N' flag.
1613 Latency tracing and events
1614 --------------------------
1615 As function tracing can induce a much larger latency, but without
1616 seeing what happens within the latency it is hard to know what
1617 caused it. There is a middle ground, and that is with enabling
1620 # echo 0 > options/function-trace
1621 # echo wakeup_rt > current_tracer
1622 # echo 1 > events/enable
1623 # echo 1 > tracing_on
1624 # echo 0 > tracing_max_latency
1626 # echo 0 > tracing_on
1630 # wakeup_rt latency trace v1.1.5 on 3.8.0-test+
1631 # --------------------------------------------------------------------
1632 # latency: 6 us, #12/12, CPU#2 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1634 # | task: sleep-5882 (uid:0 nice:0 policy:1 rt_prio:5)
1638 # / _-----=> irqs-off
1639 # | / _----=> need-resched
1640 # || / _---=> hardirq/softirq
1641 # ||| / _--=> preempt-depth
1643 # cmd pid ||||| time | caller
1645 <idle>-0 2d.h4 0us : 0:120:R + [002] 5882: 94:R sleep
1646 <idle>-0 2d.h4 0us : ttwu_do_activate.constprop.87 <-try_to_wake_up
1647 <idle>-0 2d.h4 1us : sched_wakeup: comm=sleep pid=5882 prio=94 success=1 target_cpu=002
1648 <idle>-0 2dNh2 1us : hrtimer_expire_exit: hrtimer=ffff88007796feb8
1649 <idle>-0 2.N.2 2us : power_end: cpu_id=2
1650 <idle>-0 2.N.2 3us : cpu_idle: state=4294967295 cpu_id=2
1651 <idle>-0 2dN.3 4us : hrtimer_cancel: hrtimer=ffff88007d50d5e0
1652 <idle>-0 2dN.3 4us : hrtimer_start: hrtimer=ffff88007d50d5e0 function=tick_sched_timer expires=34311211000000 softexpires=34311211000000
1653 <idle>-0 2.N.2 5us : rcu_utilization: Start context switch
1654 <idle>-0 2.N.2 5us : rcu_utilization: End context switch
1655 <idle>-0 2d..3 6us : __schedule <-schedule
1656 <idle>-0 2d..3 6us : 0:120:R ==> [002] 5882: 94:R sleep
1662 This tracer is the function tracer. Enabling the function tracer
1663 can be done from the debug file system. Make sure the
1664 ftrace_enabled is set; otherwise this tracer is a nop.
1665 See the "ftrace_enabled" section below.
1667 # sysctl kernel.ftrace_enabled=1
1668 # echo function > current_tracer
1669 # echo 1 > tracing_on
1671 # echo 0 > tracing_on
1675 # entries-in-buffer/entries-written: 24799/24799 #P:4
1678 # / _----=> need-resched
1679 # | / _---=> hardirq/softirq
1680 # || / _--=> preempt-depth
1682 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
1684 bash-1994 [002] .... 3082.063030: mutex_unlock <-rb_simple_write
1685 bash-1994 [002] .... 3082.063031: __mutex_unlock_slowpath <-mutex_unlock
1686 bash-1994 [002] .... 3082.063031: __fsnotify_parent <-fsnotify_modify
1687 bash-1994 [002] .... 3082.063032: fsnotify <-fsnotify_modify
1688 bash-1994 [002] .... 3082.063032: __srcu_read_lock <-fsnotify
1689 bash-1994 [002] .... 3082.063032: add_preempt_count <-__srcu_read_lock
1690 bash-1994 [002] ...1 3082.063032: sub_preempt_count <-__srcu_read_lock
1691 bash-1994 [002] .... 3082.063033: __srcu_read_unlock <-fsnotify
1695 Note: function tracer uses ring buffers to store the above
1696 entries. The newest data may overwrite the oldest data.
1697 Sometimes using echo to stop the trace is not sufficient because
1698 the tracing could have overwritten the data that you wanted to
1699 record. For this reason, it is sometimes better to disable
1700 tracing directly from a program. This allows you to stop the
1701 tracing at the point that you hit the part that you are
1702 interested in. To disable the tracing directly from a C program,
1703 something like following code snippet can be used:
1707 int main(int argc, char *argv[]) {
1709 trace_fd = open(tracing_file("tracing_on"), O_WRONLY);
1711 if (condition_hit()) {
1712 write(trace_fd, "0", 1);
1718 Single thread tracing
1719 ---------------------
1721 By writing into set_ftrace_pid you can trace a
1722 single thread. For example:
1724 # cat set_ftrace_pid
1726 # echo 3111 > set_ftrace_pid
1727 # cat set_ftrace_pid
1729 # echo function > current_tracer
1733 # TASK-PID CPU# TIMESTAMP FUNCTION
1735 yum-updatesd-3111 [003] 1637.254676: finish_task_switch <-thread_return
1736 yum-updatesd-3111 [003] 1637.254681: hrtimer_cancel <-schedule_hrtimeout_range
1737 yum-updatesd-3111 [003] 1637.254682: hrtimer_try_to_cancel <-hrtimer_cancel
1738 yum-updatesd-3111 [003] 1637.254683: lock_hrtimer_base <-hrtimer_try_to_cancel
1739 yum-updatesd-3111 [003] 1637.254685: fget_light <-do_sys_poll
1740 yum-updatesd-3111 [003] 1637.254686: pipe_poll <-do_sys_poll
1741 # echo -1 > set_ftrace_pid
1745 # TASK-PID CPU# TIMESTAMP FUNCTION
1747 ##### CPU 3 buffer started ####
1748 yum-updatesd-3111 [003] 1701.957688: free_poll_entry <-poll_freewait
1749 yum-updatesd-3111 [003] 1701.957689: remove_wait_queue <-free_poll_entry
1750 yum-updatesd-3111 [003] 1701.957691: fput <-free_poll_entry
1751 yum-updatesd-3111 [003] 1701.957692: audit_syscall_exit <-sysret_audit
1752 yum-updatesd-3111 [003] 1701.957693: path_put <-audit_syscall_exit
1754 If you want to trace a function when executing, you could use
1755 something like this simple program:
1759 #include <sys/types.h>
1760 #include <sys/stat.h>
1766 #define STR(x) _STR(x)
1767 #define MAX_PATH 256
1769 const char *find_debugfs(void)
1771 static char debugfs[MAX_PATH+1];
1772 static int debugfs_found;
1779 if ((fp = fopen("/proc/mounts","r")) == NULL) {
1780 perror("/proc/mounts");
1784 while (fscanf(fp, "%*s %"
1786 "s %99s %*s %*d %*d\n",
1787 debugfs, type) == 2) {
1788 if (strcmp(type, "debugfs") == 0)
1793 if (strcmp(type, "debugfs") != 0) {
1794 fprintf(stderr, "debugfs not mounted");
1798 strcat(debugfs, "/tracing/");
1804 const char *tracing_file(const char *file_name)
1806 static char trace_file[MAX_PATH+1];
1807 snprintf(trace_file, MAX_PATH, "%s/%s", find_debugfs(), file_name);
1811 int main (int argc, char **argv)
1821 ffd = open(tracing_file("current_tracer"), O_WRONLY);
1824 write(ffd, "nop", 3);
1826 fd = open(tracing_file("set_ftrace_pid"), O_WRONLY);
1827 s = sprintf(line, "%d\n", getpid());
1830 write(ffd, "function", 8);
1835 execvp(argv[1], argv+1);
1841 Or this simple script!
1846 debugfs=`sed -ne 's/^debugfs \(.*\) debugfs.*/\1/p' /proc/mounts`
1847 echo nop > $debugfs/tracing/current_tracer
1848 echo 0 > $debugfs/tracing/tracing_on
1849 echo $$ > $debugfs/tracing/set_ftrace_pid
1850 echo function > $debugfs/tracing/current_tracer
1851 echo 1 > $debugfs/tracing/tracing_on
1856 function graph tracer
1857 ---------------------------
1859 This tracer is similar to the function tracer except that it
1860 probes a function on its entry and its exit. This is done by
1861 using a dynamically allocated stack of return addresses in each
1862 task_struct. On function entry the tracer overwrites the return
1863 address of each function traced to set a custom probe. Thus the
1864 original return address is stored on the stack of return address
1867 Probing on both ends of a function leads to special features
1870 - measure of a function's time execution
1871 - having a reliable call stack to draw function calls graph
1873 This tracer is useful in several situations:
1875 - you want to find the reason of a strange kernel behavior and
1876 need to see what happens in detail on any areas (or specific
1879 - you are experiencing weird latencies but it's difficult to
1882 - you want to find quickly which path is taken by a specific
1885 - you just want to peek inside a working kernel and want to see
1888 # tracer: function_graph
1890 # CPU DURATION FUNCTION CALLS
1894 0) | do_sys_open() {
1896 0) | kmem_cache_alloc() {
1897 0) 1.382 us | __might_sleep();
1899 0) | strncpy_from_user() {
1900 0) | might_fault() {
1901 0) 1.389 us | __might_sleep();
1906 0) 0.668 us | _spin_lock();
1907 0) 0.570 us | expand_files();
1908 0) 0.586 us | _spin_unlock();
1911 There are several columns that can be dynamically
1912 enabled/disabled. You can use every combination of options you
1913 want, depending on your needs.
1915 - The cpu number on which the function executed is default
1916 enabled. It is sometimes better to only trace one cpu (see
1917 tracing_cpu_mask file) or you might sometimes see unordered
1918 function calls while cpu tracing switch.
1920 hide: echo nofuncgraph-cpu > trace_options
1921 show: echo funcgraph-cpu > trace_options
1923 - The duration (function's time of execution) is displayed on
1924 the closing bracket line of a function or on the same line
1925 than the current function in case of a leaf one. It is default
1928 hide: echo nofuncgraph-duration > trace_options
1929 show: echo funcgraph-duration > trace_options
1931 - The overhead field precedes the duration field in case of
1932 reached duration thresholds.
1934 hide: echo nofuncgraph-overhead > trace_options
1935 show: echo funcgraph-overhead > trace_options
1936 depends on: funcgraph-duration
1941 0) 0.646 us | _spin_lock_irqsave();
1942 0) 0.684 us | _spin_unlock_irqrestore();
1944 0) 0.548 us | fput();
1950 0) | kmem_cache_free() {
1951 0) 0.518 us | __phys_addr();
1957 + means that the function exceeded 10 usecs.
1958 ! means that the function exceeded 100 usecs.
1961 - The task/pid field displays the thread cmdline and pid which
1962 executed the function. It is default disabled.
1964 hide: echo nofuncgraph-proc > trace_options
1965 show: echo funcgraph-proc > trace_options
1969 # tracer: function_graph
1971 # CPU TASK/PID DURATION FUNCTION CALLS
1973 0) sh-4802 | | d_free() {
1974 0) sh-4802 | | call_rcu() {
1975 0) sh-4802 | | __call_rcu() {
1976 0) sh-4802 | 0.616 us | rcu_process_gp_end();
1977 0) sh-4802 | 0.586 us | check_for_new_grace_period();
1978 0) sh-4802 | 2.899 us | }
1979 0) sh-4802 | 4.040 us | }
1980 0) sh-4802 | 5.151 us | }
1981 0) sh-4802 | + 49.370 us | }
1984 - The absolute time field is an absolute timestamp given by the
1985 system clock since it started. A snapshot of this time is
1986 given on each entry/exit of functions
1988 hide: echo nofuncgraph-abstime > trace_options
1989 show: echo funcgraph-abstime > trace_options
1994 # TIME CPU DURATION FUNCTION CALLS
1996 360.774522 | 1) 0.541 us | }
1997 360.774522 | 1) 4.663 us | }
1998 360.774523 | 1) 0.541 us | __wake_up_bit();
1999 360.774524 | 1) 6.796 us | }
2000 360.774524 | 1) 7.952 us | }
2001 360.774525 | 1) 9.063 us | }
2002 360.774525 | 1) 0.615 us | journal_mark_dirty();
2003 360.774527 | 1) 0.578 us | __brelse();
2004 360.774528 | 1) | reiserfs_prepare_for_journal() {
2005 360.774528 | 1) | unlock_buffer() {
2006 360.774529 | 1) | wake_up_bit() {
2007 360.774529 | 1) | bit_waitqueue() {
2008 360.774530 | 1) 0.594 us | __phys_addr();
2011 The function name is always displayed after the closing bracket
2012 for a function if the start of that function is not in the
2015 Display of the function name after the closing bracket may be
2016 enabled for functions whose start is in the trace buffer,
2017 allowing easier searching with grep for function durations.
2018 It is default disabled.
2020 hide: echo nofuncgraph-tail > trace_options
2021 show: echo funcgraph-tail > trace_options
2023 Example with nofuncgraph-tail (default):
2025 0) | kmem_cache_free() {
2026 0) 0.518 us | __phys_addr();
2030 Example with funcgraph-tail:
2032 0) | kmem_cache_free() {
2033 0) 0.518 us | __phys_addr();
2034 0) 1.757 us | } /* kmem_cache_free() */
2035 0) 2.861 us | } /* putname() */
2037 You can put some comments on specific functions by using
2038 trace_printk() For example, if you want to put a comment inside
2039 the __might_sleep() function, you just have to include
2040 <linux/ftrace.h> and call trace_printk() inside __might_sleep()
2042 trace_printk("I'm a comment!\n")
2046 1) | __might_sleep() {
2047 1) | /* I'm a comment! */
2051 You might find other useful features for this tracer in the
2052 following "dynamic ftrace" section such as tracing only specific
2058 If CONFIG_DYNAMIC_FTRACE is set, the system will run with
2059 virtually no overhead when function tracing is disabled. The way
2060 this works is the mcount function call (placed at the start of
2061 every kernel function, produced by the -pg switch in gcc),
2062 starts of pointing to a simple return. (Enabling FTRACE will
2063 include the -pg switch in the compiling of the kernel.)
2065 At compile time every C file object is run through the
2066 recordmcount program (located in the scripts directory). This
2067 program will parse the ELF headers in the C object to find all
2068 the locations in the .text section that call mcount. (Note, only
2069 white listed .text sections are processed, since processing other
2070 sections like .init.text may cause races due to those sections
2071 being freed unexpectedly).
2073 A new section called "__mcount_loc" is created that holds
2074 references to all the mcount call sites in the .text section.
2075 The recordmcount program re-links this section back into the
2076 original object. The final linking stage of the kernel will add all these
2077 references into a single table.
2079 On boot up, before SMP is initialized, the dynamic ftrace code
2080 scans this table and updates all the locations into nops. It
2081 also records the locations, which are added to the
2082 available_filter_functions list. Modules are processed as they
2083 are loaded and before they are executed. When a module is
2084 unloaded, it also removes its functions from the ftrace function
2085 list. This is automatic in the module unload code, and the
2086 module author does not need to worry about it.
2088 When tracing is enabled, the process of modifying the function
2089 tracepoints is dependent on architecture. The old method is to use
2090 kstop_machine to prevent races with the CPUs executing code being
2091 modified (which can cause the CPU to do undesirable things, especially
2092 if the modified code crosses cache (or page) boundaries), and the nops are
2093 patched back to calls. But this time, they do not call mcount
2094 (which is just a function stub). They now call into the ftrace
2097 The new method of modifying the function tracepoints is to place
2098 a breakpoint at the location to be modified, sync all CPUs, modify
2099 the rest of the instruction not covered by the breakpoint. Sync
2100 all CPUs again, and then remove the breakpoint with the finished
2101 version to the ftrace call site.
2103 Some archs do not even need to monkey around with the synchronization,
2104 and can just slap the new code on top of the old without any
2105 problems with other CPUs executing it at the same time.
2107 One special side-effect to the recording of the functions being
2108 traced is that we can now selectively choose which functions we
2109 wish to trace and which ones we want the mcount calls to remain
2112 Two files are used, one for enabling and one for disabling the
2113 tracing of specified functions. They are:
2121 A list of available functions that you can add to these files is
2124 available_filter_functions
2126 # cat available_filter_functions
2135 If I am only interested in sys_nanosleep and hrtimer_interrupt:
2137 # echo sys_nanosleep hrtimer_interrupt > set_ftrace_filter
2138 # echo function > current_tracer
2139 # echo 1 > tracing_on
2141 # echo 0 > tracing_on
2145 # entries-in-buffer/entries-written: 5/5 #P:4
2148 # / _----=> need-resched
2149 # | / _---=> hardirq/softirq
2150 # || / _--=> preempt-depth
2152 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
2154 usleep-2665 [001] .... 4186.475355: sys_nanosleep <-system_call_fastpath
2155 <idle>-0 [001] d.h1 4186.475409: hrtimer_interrupt <-smp_apic_timer_interrupt
2156 usleep-2665 [001] d.h1 4186.475426: hrtimer_interrupt <-smp_apic_timer_interrupt
2157 <idle>-0 [003] d.h1 4186.475426: hrtimer_interrupt <-smp_apic_timer_interrupt
2158 <idle>-0 [002] d.h1 4186.475427: hrtimer_interrupt <-smp_apic_timer_interrupt
2160 To see which functions are being traced, you can cat the file:
2162 # cat set_ftrace_filter
2167 Perhaps this is not enough. The filters also allow simple wild
2168 cards. Only the following are currently available
2170 <match>* - will match functions that begin with <match>
2171 *<match> - will match functions that end with <match>
2172 *<match>* - will match functions that have <match> in it
2174 These are the only wild cards which are supported.
2176 <match>*<match> will not work.
2178 Note: It is better to use quotes to enclose the wild cards,
2179 otherwise the shell may expand the parameters into names
2180 of files in the local directory.
2182 # echo 'hrtimer_*' > set_ftrace_filter
2188 # entries-in-buffer/entries-written: 897/897 #P:4
2191 # / _----=> need-resched
2192 # | / _---=> hardirq/softirq
2193 # || / _--=> preempt-depth
2195 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
2197 <idle>-0 [003] dN.1 4228.547803: hrtimer_cancel <-tick_nohz_idle_exit
2198 <idle>-0 [003] dN.1 4228.547804: hrtimer_try_to_cancel <-hrtimer_cancel
2199 <idle>-0 [003] dN.2 4228.547805: hrtimer_force_reprogram <-__remove_hrtimer
2200 <idle>-0 [003] dN.1 4228.547805: hrtimer_forward <-tick_nohz_idle_exit
2201 <idle>-0 [003] dN.1 4228.547805: hrtimer_start_range_ns <-hrtimer_start_expires.constprop.11
2202 <idle>-0 [003] d..1 4228.547858: hrtimer_get_next_event <-get_next_timer_interrupt
2203 <idle>-0 [003] d..1 4228.547859: hrtimer_start <-__tick_nohz_idle_enter
2204 <idle>-0 [003] d..2 4228.547860: hrtimer_force_reprogram <-__rem
2206 Notice that we lost the sys_nanosleep.
2208 # cat set_ftrace_filter
2213 hrtimer_try_to_cancel
2217 hrtimer_force_reprogram
2218 hrtimer_get_next_event
2222 hrtimer_get_remaining
2224 hrtimer_init_sleeper
2227 This is because the '>' and '>>' act just like they do in bash.
2228 To rewrite the filters, use '>'
2229 To append to the filters, use '>>'
2231 To clear out a filter so that all functions will be recorded
2234 # echo > set_ftrace_filter
2235 # cat set_ftrace_filter
2238 Again, now we want to append.
2240 # echo sys_nanosleep > set_ftrace_filter
2241 # cat set_ftrace_filter
2243 # echo 'hrtimer_*' >> set_ftrace_filter
2244 # cat set_ftrace_filter
2249 hrtimer_try_to_cancel
2253 hrtimer_force_reprogram
2254 hrtimer_get_next_event
2259 hrtimer_get_remaining
2261 hrtimer_init_sleeper
2264 The set_ftrace_notrace prevents those functions from being
2267 # echo '*preempt*' '*lock*' > set_ftrace_notrace
2273 # entries-in-buffer/entries-written: 39608/39608 #P:4
2276 # / _----=> need-resched
2277 # | / _---=> hardirq/softirq
2278 # || / _--=> preempt-depth
2280 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
2282 bash-1994 [000] .... 4342.324896: file_ra_state_init <-do_dentry_open
2283 bash-1994 [000] .... 4342.324897: open_check_o_direct <-do_last
2284 bash-1994 [000] .... 4342.324897: ima_file_check <-do_last
2285 bash-1994 [000] .... 4342.324898: process_measurement <-ima_file_check
2286 bash-1994 [000] .... 4342.324898: ima_get_action <-process_measurement
2287 bash-1994 [000] .... 4342.324898: ima_match_policy <-ima_get_action
2288 bash-1994 [000] .... 4342.324899: do_truncate <-do_last
2289 bash-1994 [000] .... 4342.324899: should_remove_suid <-do_truncate
2290 bash-1994 [000] .... 4342.324899: notify_change <-do_truncate
2291 bash-1994 [000] .... 4342.324900: current_fs_time <-notify_change
2292 bash-1994 [000] .... 4342.324900: current_kernel_time <-current_fs_time
2293 bash-1994 [000] .... 4342.324900: timespec_trunc <-current_fs_time
2295 We can see that there's no more lock or preempt tracing.
2298 Dynamic ftrace with the function graph tracer
2299 ---------------------------------------------
2301 Although what has been explained above concerns both the
2302 function tracer and the function-graph-tracer, there are some
2303 special features only available in the function-graph tracer.
2305 If you want to trace only one function and all of its children,
2306 you just have to echo its name into set_graph_function:
2308 echo __do_fault > set_graph_function
2310 will produce the following "expanded" trace of the __do_fault()
2314 0) | filemap_fault() {
2315 0) | find_lock_page() {
2316 0) 0.804 us | find_get_page();
2317 0) | __might_sleep() {
2321 0) 0.653 us | _spin_lock();
2322 0) 0.578 us | page_add_file_rmap();
2323 0) 0.525 us | native_set_pte_at();
2324 0) 0.585 us | _spin_unlock();
2325 0) | unlock_page() {
2326 0) 0.541 us | page_waitqueue();
2327 0) 0.639 us | __wake_up_bit();
2331 0) | filemap_fault() {
2332 0) | find_lock_page() {
2333 0) 0.698 us | find_get_page();
2334 0) | __might_sleep() {
2338 0) 0.631 us | _spin_lock();
2339 0) 0.571 us | page_add_file_rmap();
2340 0) 0.526 us | native_set_pte_at();
2341 0) 0.586 us | _spin_unlock();
2342 0) | unlock_page() {
2343 0) 0.533 us | page_waitqueue();
2344 0) 0.638 us | __wake_up_bit();
2348 You can also expand several functions at once:
2350 echo sys_open > set_graph_function
2351 echo sys_close >> set_graph_function
2353 Now if you want to go back to trace all functions you can clear
2354 this special filter via:
2356 echo > set_graph_function
2362 Note, the proc sysctl ftrace_enable is a big on/off switch for the
2363 function tracer. By default it is enabled (when function tracing is
2364 enabled in the kernel). If it is disabled, all function tracing is
2365 disabled. This includes not only the function tracers for ftrace, but
2366 also for any other uses (perf, kprobes, stack tracing, profiling, etc).
2368 Please disable this with care.
2370 This can be disable (and enabled) with:
2372 sysctl kernel.ftrace_enabled=0
2373 sysctl kernel.ftrace_enabled=1
2377 echo 0 > /proc/sys/kernel/ftrace_enabled
2378 echo 1 > /proc/sys/kernel/ftrace_enabled
2384 A few commands are supported by the set_ftrace_filter interface.
2385 Trace commands have the following format:
2387 <function>:<command>:<parameter>
2389 The following commands are supported:
2392 This command enables function filtering per module. The
2393 parameter defines the module. For example, if only the write*
2394 functions in the ext3 module are desired, run:
2396 echo 'write*:mod:ext3' > set_ftrace_filter
2398 This command interacts with the filter in the same way as
2399 filtering based on function names. Thus, adding more functions
2400 in a different module is accomplished by appending (>>) to the
2401 filter file. Remove specific module functions by prepending
2404 echo '!writeback*:mod:ext3' >> set_ftrace_filter
2407 These commands turn tracing on and off when the specified
2408 functions are hit. The parameter determines how many times the
2409 tracing system is turned on and off. If unspecified, there is
2410 no limit. For example, to disable tracing when a schedule bug
2411 is hit the first 5 times, run:
2413 echo '__schedule_bug:traceoff:5' > set_ftrace_filter
2415 To always disable tracing when __schedule_bug is hit:
2417 echo '__schedule_bug:traceoff' > set_ftrace_filter
2419 These commands are cumulative whether or not they are appended
2420 to set_ftrace_filter. To remove a command, prepend it by '!'
2421 and drop the parameter:
2423 echo '!__schedule_bug:traceoff:0' > set_ftrace_filter
2425 The above removes the traceoff command for __schedule_bug
2426 that have a counter. To remove commands without counters:
2428 echo '!__schedule_bug:traceoff' > set_ftrace_filter
2431 Will cause a snapshot to be triggered when the function is hit.
2433 echo 'native_flush_tlb_others:snapshot' > set_ftrace_filter
2435 To only snapshot once:
2437 echo 'native_flush_tlb_others:snapshot:1' > set_ftrace_filter
2439 To remove the above commands:
2441 echo '!native_flush_tlb_others:snapshot' > set_ftrace_filter
2442 echo '!native_flush_tlb_others:snapshot:0' > set_ftrace_filter
2444 - enable_event/disable_event
2445 These commands can enable or disable a trace event. Note, because
2446 function tracing callbacks are very sensitive, when these commands
2447 are registered, the trace point is activated, but disabled in
2448 a "soft" mode. That is, the tracepoint will be called, but
2449 just will not be traced. The event tracepoint stays in this mode
2450 as long as there's a command that triggers it.
2452 echo 'try_to_wake_up:enable_event:sched:sched_switch:2' > \
2457 <function>:enable_event:<system>:<event>[:count]
2458 <function>:disable_event:<system>:<event>[:count]
2460 To remove the events commands:
2463 echo '!try_to_wake_up:enable_event:sched:sched_switch:0' > \
2465 echo '!schedule:disable_event:sched:sched_switch' > \
2469 When the function is hit, it will dump the contents of the ftrace
2470 ring buffer to the console. This is useful if you need to debug
2471 something, and want to dump the trace when a certain function
2472 is hit. Perhaps its a function that is called before a tripple
2473 fault happens and does not allow you to get a regular dump.
2476 When the function is hit, it will dump the contents of the ftrace
2477 ring buffer for the current CPU to the console. Unlike the "dump"
2478 command, it only prints out the contents of the ring buffer for the
2479 CPU that executed the function that triggered the dump.
2484 The trace_pipe outputs the same content as the trace file, but
2485 the effect on the tracing is different. Every read from
2486 trace_pipe is consumed. This means that subsequent reads will be
2487 different. The trace is live.
2489 # echo function > current_tracer
2490 # cat trace_pipe > /tmp/trace.out &
2492 # echo 1 > tracing_on
2494 # echo 0 > tracing_on
2498 # entries-in-buffer/entries-written: 0/0 #P:4
2501 # / _----=> need-resched
2502 # | / _---=> hardirq/softirq
2503 # || / _--=> preempt-depth
2505 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
2509 # cat /tmp/trace.out
2510 bash-1994 [000] .... 5281.568961: mutex_unlock <-rb_simple_write
2511 bash-1994 [000] .... 5281.568963: __mutex_unlock_slowpath <-mutex_unlock
2512 bash-1994 [000] .... 5281.568963: __fsnotify_parent <-fsnotify_modify
2513 bash-1994 [000] .... 5281.568964: fsnotify <-fsnotify_modify
2514 bash-1994 [000] .... 5281.568964: __srcu_read_lock <-fsnotify
2515 bash-1994 [000] .... 5281.568964: add_preempt_count <-__srcu_read_lock
2516 bash-1994 [000] ...1 5281.568965: sub_preempt_count <-__srcu_read_lock
2517 bash-1994 [000] .... 5281.568965: __srcu_read_unlock <-fsnotify
2518 bash-1994 [000] .... 5281.568967: sys_dup2 <-system_call_fastpath
2521 Note, reading the trace_pipe file will block until more input is
2527 Having too much or not enough data can be troublesome in
2528 diagnosing an issue in the kernel. The file buffer_size_kb is
2529 used to modify the size of the internal trace buffers. The
2530 number listed is the number of entries that can be recorded per
2531 CPU. To know the full size, multiply the number of possible CPUs
2532 with the number of entries.
2534 # cat buffer_size_kb
2535 1408 (units kilobytes)
2537 Or simply read buffer_total_size_kb
2539 # cat buffer_total_size_kb
2542 To modify the buffer, simple echo in a number (in 1024 byte segments).
2544 # echo 10000 > buffer_size_kb
2545 # cat buffer_size_kb
2546 10000 (units kilobytes)
2548 It will try to allocate as much as possible. If you allocate too
2549 much, it can cause Out-Of-Memory to trigger.
2551 # echo 1000000000000 > buffer_size_kb
2552 -bash: echo: write error: Cannot allocate memory
2553 # cat buffer_size_kb
2556 The per_cpu buffers can be changed individually as well:
2558 # echo 10000 > per_cpu/cpu0/buffer_size_kb
2559 # echo 100 > per_cpu/cpu1/buffer_size_kb
2561 When the per_cpu buffers are not the same, the buffer_size_kb
2562 at the top level will just show an X
2564 # cat buffer_size_kb
2567 This is where the buffer_total_size_kb is useful:
2569 # cat buffer_total_size_kb
2572 Writing to the top level buffer_size_kb will reset all the buffers
2573 to be the same again.
2577 CONFIG_TRACER_SNAPSHOT makes a generic snapshot feature
2578 available to all non latency tracers. (Latency tracers which
2579 record max latency, such as "irqsoff" or "wakeup", can't use
2580 this feature, since those are already using the snapshot
2581 mechanism internally.)
2583 Snapshot preserves a current trace buffer at a particular point
2584 in time without stopping tracing. Ftrace swaps the current
2585 buffer with a spare buffer, and tracing continues in the new
2586 current (=previous spare) buffer.
2588 The following debugfs files in "tracing" are related to this
2593 This is used to take a snapshot and to read the output
2594 of the snapshot. Echo 1 into this file to allocate a
2595 spare buffer and to take a snapshot (swap), then read
2596 the snapshot from this file in the same format as
2597 "trace" (described above in the section "The File
2598 System"). Both reads snapshot and tracing are executable
2599 in parallel. When the spare buffer is allocated, echoing
2600 0 frees it, and echoing else (positive) values clear the
2602 More details are shown in the table below.
2604 status\input | 0 | 1 | else |
2605 --------------+------------+------------+------------+
2606 not allocated |(do nothing)| alloc+swap |(do nothing)|
2607 --------------+------------+------------+------------+
2608 allocated | free | swap | clear |
2609 --------------+------------+------------+------------+
2611 Here is an example of using the snapshot feature.
2613 # echo 1 > events/sched/enable
2618 # entries-in-buffer/entries-written: 71/71 #P:8
2621 # / _----=> need-resched
2622 # | / _---=> hardirq/softirq
2623 # || / _--=> preempt-depth
2625 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
2627 <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
2628 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
2630 <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
2635 # entries-in-buffer/entries-written: 77/77 #P:8
2638 # / _----=> need-resched
2639 # | / _---=> hardirq/softirq
2640 # || / _--=> preempt-depth
2642 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
2644 <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
2645 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
2649 If you try to use this snapshot feature when current tracer is
2650 one of the latency tracers, you will get the following results.
2652 # echo wakeup > current_tracer
2654 bash: echo: write error: Device or resource busy
2656 cat: snapshot: Device or resource busy
2661 In the debugfs tracing directory is a directory called "instances".
2662 This directory can have new directories created inside of it using
2663 mkdir, and removing directories with rmdir. The directory created
2664 with mkdir in this directory will already contain files and other
2665 directories after it is created.
2667 # mkdir instances/foo
2669 buffer_size_kb buffer_total_size_kb events free_buffer per_cpu
2670 set_event snapshot trace trace_clock trace_marker trace_options
2671 trace_pipe tracing_on
2673 As you can see, the new directory looks similar to the tracing directory
2674 itself. In fact, it is very similar, except that the buffer and
2675 events are agnostic from the main director, or from any other
2676 instances that are created.
2678 The files in the new directory work just like the files with the
2679 same name in the tracing directory except the buffer that is used
2680 is a separate and new buffer. The files affect that buffer but do not
2681 affect the main buffer with the exception of trace_options. Currently,
2682 the trace_options affect all instances and the top level buffer
2683 the same, but this may change in future releases. That is, options
2684 may become specific to the instance they reside in.
2686 Notice that none of the function tracer files are there, nor is
2687 current_tracer and available_tracers. This is because the buffers
2688 can currently only have events enabled for them.
2690 # mkdir instances/foo
2691 # mkdir instances/bar
2692 # mkdir instances/zoot
2693 # echo 100000 > buffer_size_kb
2694 # echo 1000 > instances/foo/buffer_size_kb
2695 # echo 5000 > instances/bar/per_cpu/cpu1/buffer_size_kb
2696 # echo function > current_trace
2697 # echo 1 > instances/foo/events/sched/sched_wakeup/enable
2698 # echo 1 > instances/foo/events/sched/sched_wakeup_new/enable
2699 # echo 1 > instances/foo/events/sched/sched_switch/enable
2700 # echo 1 > instances/bar/events/irq/enable
2701 # echo 1 > instances/zoot/events/syscalls/enable
2703 CPU:2 [LOST 11745 EVENTS]
2704 bash-2044 [002] .... 10594.481032: _raw_spin_lock_irqsave <-get_page_from_freelist
2705 bash-2044 [002] d... 10594.481032: add_preempt_count <-_raw_spin_lock_irqsave
2706 bash-2044 [002] d..1 10594.481032: __rmqueue <-get_page_from_freelist
2707 bash-2044 [002] d..1 10594.481033: _raw_spin_unlock <-get_page_from_freelist
2708 bash-2044 [002] d..1 10594.481033: sub_preempt_count <-_raw_spin_unlock
2709 bash-2044 [002] d... 10594.481033: get_pageblock_flags_group <-get_pageblock_migratetype
2710 bash-2044 [002] d... 10594.481034: __mod_zone_page_state <-get_page_from_freelist
2711 bash-2044 [002] d... 10594.481034: zone_statistics <-get_page_from_freelist
2712 bash-2044 [002] d... 10594.481034: __inc_zone_state <-zone_statistics
2713 bash-2044 [002] d... 10594.481034: __inc_zone_state <-zone_statistics
2714 bash-2044 [002] .... 10594.481035: arch_dup_task_struct <-copy_process
2717 # cat instances/foo/trace_pipe
2718 bash-1998 [000] d..4 136.676759: sched_wakeup: comm=kworker/0:1 pid=59 prio=120 success=1 target_cpu=000
2719 bash-1998 [000] dN.4 136.676760: sched_wakeup: comm=bash pid=1998 prio=120 success=1 target_cpu=000
2720 <idle>-0 [003] d.h3 136.676906: sched_wakeup: comm=rcu_preempt pid=9 prio=120 success=1 target_cpu=003
2721 <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
2722 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
2723 bash-1998 [000] d..4 136.677014: sched_wakeup: comm=kworker/0:1 pid=59 prio=120 success=1 target_cpu=000
2724 bash-1998 [000] dN.4 136.677016: sched_wakeup: comm=bash pid=1998 prio=120 success=1 target_cpu=000
2725 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
2726 kworker/0:1-59 [000] d..4 136.677022: sched_wakeup: comm=sshd pid=1995 prio=120 success=1 target_cpu=001
2727 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
2730 # cat instances/bar/trace_pipe
2731 migration/1-14 [001] d.h3 138.732674: softirq_raise: vec=3 [action=NET_RX]
2732 <idle>-0 [001] dNh3 138.732725: softirq_raise: vec=3 [action=NET_RX]
2733 bash-1998 [000] d.h1 138.733101: softirq_raise: vec=1 [action=TIMER]
2734 bash-1998 [000] d.h1 138.733102: softirq_raise: vec=9 [action=RCU]
2735 bash-1998 [000] ..s2 138.733105: softirq_entry: vec=1 [action=TIMER]
2736 bash-1998 [000] ..s2 138.733106: softirq_exit: vec=1 [action=TIMER]
2737 bash-1998 [000] ..s2 138.733106: softirq_entry: vec=9 [action=RCU]
2738 bash-1998 [000] ..s2 138.733109: softirq_exit: vec=9 [action=RCU]
2739 sshd-1995 [001] d.h1 138.733278: irq_handler_entry: irq=21 name=uhci_hcd:usb4
2740 sshd-1995 [001] d.h1 138.733280: irq_handler_exit: irq=21 ret=unhandled
2741 sshd-1995 [001] d.h1 138.733281: irq_handler_entry: irq=21 name=eth0
2742 sshd-1995 [001] d.h1 138.733283: irq_handler_exit: irq=21 ret=handled
2745 # cat instances/zoot/trace
2748 # entries-in-buffer/entries-written: 18996/18996 #P:4
2751 # / _----=> need-resched
2752 # | / _---=> hardirq/softirq
2753 # || / _--=> preempt-depth
2755 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
2757 bash-1998 [000] d... 140.733501: sys_write -> 0x2
2758 bash-1998 [000] d... 140.733504: sys_dup2(oldfd: a, newfd: 1)
2759 bash-1998 [000] d... 140.733506: sys_dup2 -> 0x1
2760 bash-1998 [000] d... 140.733508: sys_fcntl(fd: a, cmd: 1, arg: 0)
2761 bash-1998 [000] d... 140.733509: sys_fcntl -> 0x1
2762 bash-1998 [000] d... 140.733510: sys_close(fd: a)
2763 bash-1998 [000] d... 140.733510: sys_close -> 0x0
2764 bash-1998 [000] d... 140.733514: sys_rt_sigprocmask(how: 0, nset: 0, oset: 6e2768, sigsetsize: 8)
2765 bash-1998 [000] d... 140.733515: sys_rt_sigprocmask -> 0x0
2766 bash-1998 [000] d... 140.733516: sys_rt_sigaction(sig: 2, act: 7fff718846f0, oact: 7fff71884650, sigsetsize: 8)
2767 bash-1998 [000] d... 140.733516: sys_rt_sigaction -> 0x0
2769 You can see that the trace of the top most trace buffer shows only
2770 the function tracing. The foo instance displays wakeups and task
2773 To remove the instances, simply delete their directories:
2775 # rmdir instances/foo
2776 # rmdir instances/bar
2777 # rmdir instances/zoot
2779 Note, if a process has a trace file open in one of the instance
2780 directories, the rmdir will fail with EBUSY.
2785 Since the kernel has a fixed sized stack, it is important not to
2786 waste it in functions. A kernel developer must be conscience of
2787 what they allocate on the stack. If they add too much, the system
2788 can be in danger of a stack overflow, and corruption will occur,
2789 usually leading to a system panic.
2791 There are some tools that check this, usually with interrupts
2792 periodically checking usage. But if you can perform a check
2793 at every function call that will become very useful. As ftrace provides
2794 a function tracer, it makes it convenient to check the stack size
2795 at every function call. This is enabled via the stack tracer.
2797 CONFIG_STACK_TRACER enables the ftrace stack tracing functionality.
2798 To enable it, write a '1' into /proc/sys/kernel/stack_tracer_enabled.
2800 # echo 1 > /proc/sys/kernel/stack_tracer_enabled
2802 You can also enable it from the kernel command line to trace
2803 the stack size of the kernel during boot up, by adding "stacktrace"
2804 to the kernel command line parameter.
2806 After running it for a few minutes, the output looks like:
2808 # cat stack_max_size
2812 Depth Size Location (18 entries)
2814 0) 2928 224 update_sd_lb_stats+0xbc/0x4ac
2815 1) 2704 160 find_busiest_group+0x31/0x1f1
2816 2) 2544 256 load_balance+0xd9/0x662
2817 3) 2288 80 idle_balance+0xbb/0x130
2818 4) 2208 128 __schedule+0x26e/0x5b9
2819 5) 2080 16 schedule+0x64/0x66
2820 6) 2064 128 schedule_timeout+0x34/0xe0
2821 7) 1936 112 wait_for_common+0x97/0xf1
2822 8) 1824 16 wait_for_completion+0x1d/0x1f
2823 9) 1808 128 flush_work+0xfe/0x119
2824 10) 1680 16 tty_flush_to_ldisc+0x1e/0x20
2825 11) 1664 48 input_available_p+0x1d/0x5c
2826 12) 1616 48 n_tty_poll+0x6d/0x134
2827 13) 1568 64 tty_poll+0x64/0x7f
2828 14) 1504 880 do_select+0x31e/0x511
2829 15) 624 400 core_sys_select+0x177/0x216
2830 16) 224 96 sys_select+0x91/0xb9
2831 17) 128 128 system_call_fastpath+0x16/0x1b
2833 Note, if -mfentry is being used by gcc, functions get traced before
2834 they set up the stack frame. This means that leaf level functions
2835 are not tested by the stack tracer when -mfentry is used.
2837 Currently, -mfentry is used by gcc 4.6.0 and above on x86 only.
2841 More details can be found in the source code, in the
2842 kernel/trace/*.c files.
projects / cascardo / linux.git / blob