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, it will display the same
112 information every time it is 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 ppc-tb: This uses the powerpc timebase register value.
350 This is in sync across CPUs and can also be used
351 to correlate events across hypervisor/guest if
354 To set a clock, simply echo the clock name into this file.
356 echo global > trace_clock
360 This is a very useful file for synchronizing user space
361 with events happening in the kernel. Writing strings into
362 this file will be written into the ftrace buffer.
364 It is useful in applications to open this file at the start
365 of the application and just reference the file descriptor
368 void trace_write(const char *fmt, ...)
378 n = vsnprintf(buf, 256, fmt, ap);
381 write(trace_fd, buf, n);
386 trace_fd = open("trace_marker", WR_ONLY);
390 Add dynamic tracepoints in programs.
395 Uprobe statistics. See uprobetrace.txt
399 This is a way to make multiple trace buffers where different
400 events can be recorded in different buffers.
401 See "Instances" section below.
405 This is the trace event directory. It holds event tracepoints
406 (also known as static tracepoints) that have been compiled
407 into the kernel. It shows what event tracepoints exist
408 and how they are grouped by system. There are "enable"
409 files at various levels that can enable the tracepoints
410 when a "1" is written to them.
412 See events.txt for more information.
416 This is a directory that contains the trace per_cpu information.
418 per_cpu/cpu0/buffer_size_kb:
420 The ftrace buffer is defined per_cpu. That is, there's a separate
421 buffer for each CPU to allow writes to be done atomically,
422 and free from cache bouncing. These buffers may have different
423 size buffers. This file is similar to the buffer_size_kb
424 file, but it only displays or sets the buffer size for the
425 specific CPU. (here cpu0).
429 This is similar to the "trace" file, but it will only display
430 the data specific for the CPU. If written to, it only clears
431 the specific CPU buffer.
433 per_cpu/cpu0/trace_pipe
435 This is similar to the "trace_pipe" file, and is a consuming
436 read, but it will only display (and consume) the data specific
439 per_cpu/cpu0/trace_pipe_raw
441 For tools that can parse the ftrace ring buffer binary format,
442 the trace_pipe_raw file can be used to extract the data
443 from the ring buffer directly. With the use of the splice()
444 system call, the buffer data can be quickly transferred to
445 a file or to the network where a server is collecting the
448 Like trace_pipe, this is a consuming reader, where multiple
449 reads will always produce different data.
451 per_cpu/cpu0/snapshot:
453 This is similar to the main "snapshot" file, but will only
454 snapshot the current CPU (if supported). It only displays
455 the content of the snapshot for a given CPU, and if
456 written to, only clears this CPU buffer.
458 per_cpu/cpu0/snapshot_raw:
460 Similar to the trace_pipe_raw, but will read the binary format
461 from the snapshot buffer for the given CPU.
465 This displays certain stats about the ring buffer:
467 entries: The number of events that are still in the buffer.
469 overrun: The number of lost events due to overwriting when
472 commit overrun: Should always be zero.
473 This gets set if so many events happened within a nested
474 event (ring buffer is re-entrant), that it fills the
475 buffer and starts dropping events.
477 bytes: Bytes actually read (not overwritten).
479 oldest event ts: The oldest timestamp in the buffer
481 now ts: The current timestamp
483 dropped events: Events lost due to overwrite option being off.
485 read events: The number of events read.
490 Here is the list of current tracers that may be configured.
494 Function call tracer to trace all kernel functions.
498 Similar to the function tracer except that the
499 function tracer probes the functions on their entry
500 whereas the function graph tracer traces on both entry
501 and exit of the functions. It then provides the ability
502 to draw a graph of function calls similar to C code
507 Traces the areas that disable interrupts and saves
508 the trace with the longest max latency.
509 See tracing_max_latency. When a new max is recorded,
510 it replaces the old trace. It is best to view this
511 trace with the latency-format option enabled.
515 Similar to irqsoff but traces and records the amount of
516 time for which preemption is disabled.
520 Similar to irqsoff and preemptoff, but traces and
521 records the largest time for which irqs and/or preemption
526 Traces and records the max latency that it takes for
527 the highest priority task to get scheduled after
528 it has been woken up.
529 Traces all tasks as an average developer would expect.
533 Traces and records the max latency that it takes for just
534 RT tasks (as the current "wakeup" does). This is useful
535 for those interested in wake up timings of RT tasks.
539 This is the "trace nothing" tracer. To remove all
540 tracers from tracing simply echo "nop" into
544 Examples of using the tracer
545 ----------------------------
547 Here are typical examples of using the tracers when controlling
548 them only with the debugfs interface (without using any
549 user-land utilities).
554 Here is an example of the output format of the file "trace"
559 # entries-in-buffer/entries-written: 140080/250280 #P:4
562 # / _----=> need-resched
563 # | / _---=> hardirq/softirq
564 # || / _--=> preempt-depth
566 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
568 bash-1977 [000] .... 17284.993652: sys_close <-system_call_fastpath
569 bash-1977 [000] .... 17284.993653: __close_fd <-sys_close
570 bash-1977 [000] .... 17284.993653: _raw_spin_lock <-__close_fd
571 sshd-1974 [003] .... 17284.993653: __srcu_read_unlock <-fsnotify
572 bash-1977 [000] .... 17284.993654: add_preempt_count <-_raw_spin_lock
573 bash-1977 [000] ...1 17284.993655: _raw_spin_unlock <-__close_fd
574 bash-1977 [000] ...1 17284.993656: sub_preempt_count <-_raw_spin_unlock
575 bash-1977 [000] .... 17284.993657: filp_close <-__close_fd
576 bash-1977 [000] .... 17284.993657: dnotify_flush <-filp_close
577 sshd-1974 [003] .... 17284.993658: sys_select <-system_call_fastpath
580 A header is printed with the tracer name that is represented by
581 the trace. In this case the tracer is "function". Then it shows the
582 number of events in the buffer as well as the total number of entries
583 that were written. The difference is the number of entries that were
584 lost due to the buffer filling up (250280 - 140080 = 110200 events
587 The header explains the content of the events. Task name "bash", the task
588 PID "1977", the CPU that it was running on "000", the latency format
589 (explained below), the timestamp in <secs>.<usecs> format, the
590 function name that was traced "sys_close" and the parent function that
591 called this function "system_call_fastpath". The timestamp is the time
592 at which the function was entered.
597 When the latency-format option is enabled or when one of the latency
598 tracers is set, the trace file gives somewhat more information to see
599 why a latency happened. Here is a typical trace.
603 # irqsoff latency trace v1.1.5 on 3.8.0-test+
604 # --------------------------------------------------------------------
605 # latency: 259 us, #4/4, CPU#2 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
607 # | task: ps-6143 (uid:0 nice:0 policy:0 rt_prio:0)
609 # => started at: __lock_task_sighand
610 # => ended at: _raw_spin_unlock_irqrestore
614 # / _-----=> irqs-off
615 # | / _----=> need-resched
616 # || / _---=> hardirq/softirq
617 # ||| / _--=> preempt-depth
619 # cmd pid ||||| time | caller
621 ps-6143 2d... 0us!: trace_hardirqs_off <-__lock_task_sighand
622 ps-6143 2d..1 259us+: trace_hardirqs_on <-_raw_spin_unlock_irqrestore
623 ps-6143 2d..1 263us+: time_hardirqs_on <-_raw_spin_unlock_irqrestore
624 ps-6143 2d..1 306us : <stack trace>
625 => trace_hardirqs_on_caller
627 => _raw_spin_unlock_irqrestore
634 => system_call_fastpath
637 This shows that the current tracer is "irqsoff" tracing the time
638 for which interrupts were disabled. It gives the trace version (which
639 never changes) and the version of the kernel upon which this was executed on
640 (3.10). Then it displays the max latency in microseconds (259 us). The number
641 of trace entries displayed and the total number (both are four: #4/4).
642 VP, KP, SP, and HP are always zero and are reserved for later use.
643 #P is the number of online CPUs (#P:4).
645 The task is the process that was running when the latency
646 occurred. (ps pid: 6143).
648 The start and stop (the functions in which the interrupts were
649 disabled and enabled respectively) that caused the latencies:
651 __lock_task_sighand is where the interrupts were disabled.
652 _raw_spin_unlock_irqrestore is where they were enabled again.
654 The next lines after the header are the trace itself. The header
655 explains which is which.
657 cmd: The name of the process in the trace.
659 pid: The PID of that process.
661 CPU#: The CPU which the process was running on.
663 irqs-off: 'd' interrupts are disabled. '.' otherwise.
664 Note: If the architecture does not support a way to
665 read the irq flags variable, an 'X' will always
669 'N' both TIF_NEED_RESCHED and PREEMPT_NEED_RESCHED is set,
670 'n' only TIF_NEED_RESCHED is set,
671 'p' only PREEMPT_NEED_RESCHED is set,
675 'H' - hard irq occurred inside a softirq.
676 'h' - hard irq is running
677 's' - soft irq is running
678 '.' - normal context.
680 preempt-depth: The level of preempt_disabled
682 The above is mostly meaningful for kernel developers.
684 time: When the latency-format option is enabled, the trace file
685 output includes a timestamp relative to the start of the
686 trace. This differs from the output when latency-format
687 is disabled, which includes an absolute timestamp.
689 delay: This is just to help catch your eye a bit better. And
690 needs to be fixed to be only relative to the same CPU.
691 The marks are determined by the difference between this
692 current trace and the next trace.
693 '$' - greater than 1 second
694 '@' - greater than 100 milisecond
695 '*' - greater than 10 milisecond
696 '#' - greater than 1000 microsecond
697 '!' - greater than 100 microsecond
698 '+' - greater than 10 microsecond
699 ' ' - less than or equal to 10 microsecond.
701 The rest is the same as the 'trace' file.
703 Note, the latency tracers will usually end with a back trace
704 to easily find where the latency occurred.
709 The trace_options file (or the options directory) is used to control
710 what gets printed in the trace output, or manipulate the tracers.
711 To see what is available, simply cat the file:
741 To disable one of the options, echo in the option prepended with
744 echo noprint-parent > trace_options
746 To enable an option, leave off the "no".
748 echo sym-offset > trace_options
750 Here are the available options:
752 print-parent - On function traces, display the calling (parent)
753 function as well as the function being traced.
756 bash-4000 [01] 1477.606694: simple_strtoul <-kstrtoul
759 bash-4000 [01] 1477.606694: simple_strtoul
762 sym-offset - Display not only the function name, but also the
763 offset in the function. For example, instead of
764 seeing just "ktime_get", you will see
765 "ktime_get+0xb/0x20".
768 bash-4000 [01] 1477.606694: simple_strtoul+0x6/0xa0
770 sym-addr - this will also display the function address as well
771 as the function name.
774 bash-4000 [01] 1477.606694: simple_strtoul <c0339346>
776 verbose - This deals with the trace file when the
777 latency-format option is enabled.
779 bash 4000 1 0 00000000 00010a95 [58127d26] 1720.415ms \
780 (+0.000ms): simple_strtoul (kstrtoul)
782 raw - This will display raw numbers. This option is best for
783 use with user applications that can translate the raw
784 numbers better than having it done in the kernel.
786 hex - Similar to raw, but the numbers will be in a hexadecimal
789 bin - This will print out the formats in raw binary.
791 block - When set, reading trace_pipe will not block when polled.
793 stacktrace - This is one of the options that changes the trace
794 itself. When a trace is recorded, so is the stack
795 of functions. This allows for back traces of
798 trace_printk - Can disable trace_printk() from writing into the buffer.
800 branch - Enable branch tracing with the tracer.
802 annotate - It is sometimes confusing when the CPU buffers are full
803 and one CPU buffer had a lot of events recently, thus
804 a shorter time frame, were another CPU may have only had
805 a few events, which lets it have older events. When
806 the trace is reported, it shows the oldest events first,
807 and it may look like only one CPU ran (the one with the
808 oldest events). When the annotate option is set, it will
809 display when a new CPU buffer started:
811 <idle>-0 [001] dNs4 21169.031481: wake_up_idle_cpu <-add_timer_on
812 <idle>-0 [001] dNs4 21169.031482: _raw_spin_unlock_irqrestore <-add_timer_on
813 <idle>-0 [001] .Ns4 21169.031484: sub_preempt_count <-_raw_spin_unlock_irqrestore
814 ##### CPU 2 buffer started ####
815 <idle>-0 [002] .N.1 21169.031484: rcu_idle_exit <-cpu_idle
816 <idle>-0 [001] .Ns3 21169.031484: _raw_spin_unlock <-clocksource_watchdog
817 <idle>-0 [001] .Ns3 21169.031485: sub_preempt_count <-_raw_spin_unlock
819 userstacktrace - This option changes the trace. It records a
820 stacktrace of the current userspace thread.
822 sym-userobj - when user stacktrace are enabled, look up which
823 object the address belongs to, and print a
824 relative address. This is especially useful when
825 ASLR is on, otherwise you don't get a chance to
826 resolve the address to object/file/line after
827 the app is no longer running
829 The lookup is performed when you read
830 trace,trace_pipe. Example:
832 a.out-1623 [000] 40874.465068: /root/a.out[+0x480] <-/root/a.out[+0
833 x494] <- /root/a.out[+0x4a8] <- /lib/libc-2.7.so[+0x1e1a6]
836 printk-msg-only - When set, trace_printk()s will only show the format
837 and not their parameters (if trace_bprintk() or
838 trace_bputs() was used to save the trace_printk()).
840 context-info - Show only the event data. Hides the comm, PID,
841 timestamp, CPU, and other useful data.
843 latency-format - This option changes the trace. When
844 it is enabled, the trace displays
845 additional information about the
846 latencies, as described in "Latency
849 sleep-time - When running function graph tracer, to include
850 the time a task schedules out in its function.
851 When enabled, it will account time the task has been
852 scheduled out as part of the function call.
854 graph-time - When running function graph tracer, to include the
855 time to call nested functions. When this is not set,
856 the time reported for the function will only include
857 the time the function itself executed for, not the time
858 for functions that it called.
860 record-cmd - When any event or tracer is enabled, a hook is enabled
861 in the sched_switch trace point to fill comm cache
862 with mapped pids and comms. But this may cause some
863 overhead, and if you only care about pids, and not the
864 name of the task, disabling this option can lower the
867 overwrite - This controls what happens when the trace buffer is
868 full. If "1" (default), the oldest events are
869 discarded and overwritten. If "0", then the newest
870 events are discarded.
871 (see per_cpu/cpu0/stats for overrun and dropped)
873 disable_on_free - When the free_buffer is closed, tracing will
874 stop (tracing_on set to 0).
876 irq-info - Shows the interrupt, preempt count, need resched data.
877 When disabled, the trace looks like:
881 # entries-in-buffer/entries-written: 144405/9452052 #P:4
883 # TASK-PID CPU# TIMESTAMP FUNCTION
885 <idle>-0 [002] 23636.756054: ttwu_do_activate.constprop.89 <-try_to_wake_up
886 <idle>-0 [002] 23636.756054: activate_task <-ttwu_do_activate.constprop.89
887 <idle>-0 [002] 23636.756055: enqueue_task <-activate_task
890 markers - When set, the trace_marker is writable (only by root).
891 When disabled, the trace_marker will error with EINVAL
895 function-trace - The latency tracers will enable function tracing
896 if this option is enabled (default it is). When
897 it is disabled, the latency tracers do not trace
898 functions. This keeps the overhead of the tracer down
899 when performing latency tests.
901 Note: Some tracers have their own options. They only appear
902 when the tracer is active.
909 When interrupts are disabled, the CPU can not react to any other
910 external event (besides NMIs and SMIs). This prevents the timer
911 interrupt from triggering or the mouse interrupt from letting
912 the kernel know of a new mouse event. The result is a latency
913 with the reaction time.
915 The irqsoff tracer tracks the time for which interrupts are
916 disabled. When a new maximum latency is hit, the tracer saves
917 the trace leading up to that latency point so that every time a
918 new maximum is reached, the old saved trace is discarded and the
921 To reset the maximum, echo 0 into tracing_max_latency. Here is
924 # echo 0 > options/function-trace
925 # echo irqsoff > current_tracer
926 # echo 1 > tracing_on
927 # echo 0 > tracing_max_latency
930 # echo 0 > tracing_on
934 # irqsoff latency trace v1.1.5 on 3.8.0-test+
935 # --------------------------------------------------------------------
936 # latency: 16 us, #4/4, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
938 # | task: swapper/0-0 (uid:0 nice:0 policy:0 rt_prio:0)
940 # => started at: run_timer_softirq
941 # => ended at: run_timer_softirq
945 # / _-----=> irqs-off
946 # | / _----=> need-resched
947 # || / _---=> hardirq/softirq
948 # ||| / _--=> preempt-depth
950 # cmd pid ||||| time | caller
952 <idle>-0 0d.s2 0us+: _raw_spin_lock_irq <-run_timer_softirq
953 <idle>-0 0dNs3 17us : _raw_spin_unlock_irq <-run_timer_softirq
954 <idle>-0 0dNs3 17us+: trace_hardirqs_on <-run_timer_softirq
955 <idle>-0 0dNs3 25us : <stack trace>
956 => _raw_spin_unlock_irq
962 => smp_apic_timer_interrupt
963 => apic_timer_interrupt
968 => x86_64_start_reservations
969 => x86_64_start_kernel
971 Here we see that that we had a latency of 16 microseconds (which is
972 very good). The _raw_spin_lock_irq in run_timer_softirq disabled
973 interrupts. The difference between the 16 and the displayed
974 timestamp 25us occurred because the clock was incremented
975 between the time of recording the max latency and the time of
976 recording the function that had that latency.
978 Note the above example had function-trace not set. If we set
979 function-trace, we get a much larger output:
981 with echo 1 > options/function-trace
985 # irqsoff latency trace v1.1.5 on 3.8.0-test+
986 # --------------------------------------------------------------------
987 # latency: 71 us, #168/168, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
989 # | task: bash-2042 (uid:0 nice:0 policy:0 rt_prio:0)
991 # => started at: ata_scsi_queuecmd
992 # => ended at: ata_scsi_queuecmd
996 # / _-----=> irqs-off
997 # | / _----=> need-resched
998 # || / _---=> hardirq/softirq
999 # ||| / _--=> preempt-depth
1001 # cmd pid ||||| time | caller
1003 bash-2042 3d... 0us : _raw_spin_lock_irqsave <-ata_scsi_queuecmd
1004 bash-2042 3d... 0us : add_preempt_count <-_raw_spin_lock_irqsave
1005 bash-2042 3d..1 1us : ata_scsi_find_dev <-ata_scsi_queuecmd
1006 bash-2042 3d..1 1us : __ata_scsi_find_dev <-ata_scsi_find_dev
1007 bash-2042 3d..1 2us : ata_find_dev.part.14 <-__ata_scsi_find_dev
1008 bash-2042 3d..1 2us : ata_qc_new_init <-__ata_scsi_queuecmd
1009 bash-2042 3d..1 3us : ata_sg_init <-__ata_scsi_queuecmd
1010 bash-2042 3d..1 4us : ata_scsi_rw_xlat <-__ata_scsi_queuecmd
1011 bash-2042 3d..1 4us : ata_build_rw_tf <-ata_scsi_rw_xlat
1013 bash-2042 3d..1 67us : delay_tsc <-__delay
1014 bash-2042 3d..1 67us : add_preempt_count <-delay_tsc
1015 bash-2042 3d..2 67us : sub_preempt_count <-delay_tsc
1016 bash-2042 3d..1 67us : add_preempt_count <-delay_tsc
1017 bash-2042 3d..2 68us : sub_preempt_count <-delay_tsc
1018 bash-2042 3d..1 68us+: ata_bmdma_start <-ata_bmdma_qc_issue
1019 bash-2042 3d..1 71us : _raw_spin_unlock_irqrestore <-ata_scsi_queuecmd
1020 bash-2042 3d..1 71us : _raw_spin_unlock_irqrestore <-ata_scsi_queuecmd
1021 bash-2042 3d..1 72us+: trace_hardirqs_on <-ata_scsi_queuecmd
1022 bash-2042 3d..1 120us : <stack trace>
1023 => _raw_spin_unlock_irqrestore
1024 => ata_scsi_queuecmd
1025 => scsi_dispatch_cmd
1027 => __blk_run_queue_uncond
1030 => generic_make_request
1033 => __ext3_get_inode_loc
1042 => user_path_at_empty
1047 => system_call_fastpath
1050 Here we traced a 71 microsecond latency. But we also see all the
1051 functions that were called during that time. Note that by
1052 enabling function tracing, we incur an added overhead. This
1053 overhead may extend the latency times. But nevertheless, this
1054 trace has provided some very helpful debugging information.
1060 When preemption is disabled, we may be able to receive
1061 interrupts but the task cannot be preempted and a higher
1062 priority task must wait for preemption to be enabled again
1063 before it can preempt a lower priority task.
1065 The preemptoff tracer traces the places that disable preemption.
1066 Like the irqsoff tracer, it records the maximum latency for
1067 which preemption was disabled. The control of preemptoff tracer
1068 is much like the irqsoff tracer.
1070 # echo 0 > options/function-trace
1071 # echo preemptoff > current_tracer
1072 # echo 1 > tracing_on
1073 # echo 0 > tracing_max_latency
1076 # echo 0 > tracing_on
1078 # tracer: preemptoff
1080 # preemptoff latency trace v1.1.5 on 3.8.0-test+
1081 # --------------------------------------------------------------------
1082 # latency: 46 us, #4/4, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1084 # | task: sshd-1991 (uid:0 nice:0 policy:0 rt_prio:0)
1086 # => started at: do_IRQ
1087 # => ended at: do_IRQ
1091 # / _-----=> irqs-off
1092 # | / _----=> need-resched
1093 # || / _---=> hardirq/softirq
1094 # ||| / _--=> preempt-depth
1096 # cmd pid ||||| time | caller
1098 sshd-1991 1d.h. 0us+: irq_enter <-do_IRQ
1099 sshd-1991 1d..1 46us : irq_exit <-do_IRQ
1100 sshd-1991 1d..1 47us+: trace_preempt_on <-do_IRQ
1101 sshd-1991 1d..1 52us : <stack trace>
1102 => sub_preempt_count
1108 This has some more changes. Preemption was disabled when an
1109 interrupt came in (notice the 'h'), and was enabled on exit.
1110 But we also see that interrupts have been disabled when entering
1111 the preempt off section and leaving it (the 'd'). We do not know if
1112 interrupts were enabled in the mean time or shortly after this
1115 # tracer: preemptoff
1117 # preemptoff latency trace v1.1.5 on 3.8.0-test+
1118 # --------------------------------------------------------------------
1119 # latency: 83 us, #241/241, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1121 # | task: bash-1994 (uid:0 nice:0 policy:0 rt_prio:0)
1123 # => started at: wake_up_new_task
1124 # => ended at: task_rq_unlock
1128 # / _-----=> irqs-off
1129 # | / _----=> need-resched
1130 # || / _---=> hardirq/softirq
1131 # ||| / _--=> preempt-depth
1133 # cmd pid ||||| time | caller
1135 bash-1994 1d..1 0us : _raw_spin_lock_irqsave <-wake_up_new_task
1136 bash-1994 1d..1 0us : select_task_rq_fair <-select_task_rq
1137 bash-1994 1d..1 1us : __rcu_read_lock <-select_task_rq_fair
1138 bash-1994 1d..1 1us : source_load <-select_task_rq_fair
1139 bash-1994 1d..1 1us : source_load <-select_task_rq_fair
1141 bash-1994 1d..1 12us : irq_enter <-smp_apic_timer_interrupt
1142 bash-1994 1d..1 12us : rcu_irq_enter <-irq_enter
1143 bash-1994 1d..1 13us : add_preempt_count <-irq_enter
1144 bash-1994 1d.h1 13us : exit_idle <-smp_apic_timer_interrupt
1145 bash-1994 1d.h1 13us : hrtimer_interrupt <-smp_apic_timer_interrupt
1146 bash-1994 1d.h1 13us : _raw_spin_lock <-hrtimer_interrupt
1147 bash-1994 1d.h1 14us : add_preempt_count <-_raw_spin_lock
1148 bash-1994 1d.h2 14us : ktime_get_update_offsets <-hrtimer_interrupt
1150 bash-1994 1d.h1 35us : lapic_next_event <-clockevents_program_event
1151 bash-1994 1d.h1 35us : irq_exit <-smp_apic_timer_interrupt
1152 bash-1994 1d.h1 36us : sub_preempt_count <-irq_exit
1153 bash-1994 1d..2 36us : do_softirq <-irq_exit
1154 bash-1994 1d..2 36us : __do_softirq <-call_softirq
1155 bash-1994 1d..2 36us : __local_bh_disable <-__do_softirq
1156 bash-1994 1d.s2 37us : add_preempt_count <-_raw_spin_lock_irq
1157 bash-1994 1d.s3 38us : _raw_spin_unlock <-run_timer_softirq
1158 bash-1994 1d.s3 39us : sub_preempt_count <-_raw_spin_unlock
1159 bash-1994 1d.s2 39us : call_timer_fn <-run_timer_softirq
1161 bash-1994 1dNs2 81us : cpu_needs_another_gp <-rcu_process_callbacks
1162 bash-1994 1dNs2 82us : __local_bh_enable <-__do_softirq
1163 bash-1994 1dNs2 82us : sub_preempt_count <-__local_bh_enable
1164 bash-1994 1dN.2 82us : idle_cpu <-irq_exit
1165 bash-1994 1dN.2 83us : rcu_irq_exit <-irq_exit
1166 bash-1994 1dN.2 83us : sub_preempt_count <-irq_exit
1167 bash-1994 1.N.1 84us : _raw_spin_unlock_irqrestore <-task_rq_unlock
1168 bash-1994 1.N.1 84us+: trace_preempt_on <-task_rq_unlock
1169 bash-1994 1.N.1 104us : <stack trace>
1170 => sub_preempt_count
1171 => _raw_spin_unlock_irqrestore
1179 The above is an example of the preemptoff trace with
1180 function-trace set. Here we see that interrupts were not disabled
1181 the entire time. The irq_enter code lets us know that we entered
1182 an interrupt 'h'. Before that, the functions being traced still
1183 show that it is not in an interrupt, but we can see from the
1184 functions themselves that this is not the case.
1189 Knowing the locations that have interrupts disabled or
1190 preemption disabled for the longest times is helpful. But
1191 sometimes we would like to know when either preemption and/or
1192 interrupts are disabled.
1194 Consider the following code:
1196 local_irq_disable();
1197 call_function_with_irqs_off();
1199 call_function_with_irqs_and_preemption_off();
1201 call_function_with_preemption_off();
1204 The irqsoff tracer will record the total length of
1205 call_function_with_irqs_off() and
1206 call_function_with_irqs_and_preemption_off().
1208 The preemptoff tracer will record the total length of
1209 call_function_with_irqs_and_preemption_off() and
1210 call_function_with_preemption_off().
1212 But neither will trace the time that interrupts and/or
1213 preemption is disabled. This total time is the time that we can
1214 not schedule. To record this time, use the preemptirqsoff
1217 Again, using this trace is much like the irqsoff and preemptoff
1220 # echo 0 > options/function-trace
1221 # echo preemptirqsoff > current_tracer
1222 # echo 1 > tracing_on
1223 # echo 0 > tracing_max_latency
1226 # echo 0 > tracing_on
1228 # tracer: preemptirqsoff
1230 # preemptirqsoff latency trace v1.1.5 on 3.8.0-test+
1231 # --------------------------------------------------------------------
1232 # latency: 100 us, #4/4, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1234 # | task: ls-2230 (uid:0 nice:0 policy:0 rt_prio:0)
1236 # => started at: ata_scsi_queuecmd
1237 # => ended at: ata_scsi_queuecmd
1241 # / _-----=> irqs-off
1242 # | / _----=> need-resched
1243 # || / _---=> hardirq/softirq
1244 # ||| / _--=> preempt-depth
1246 # cmd pid ||||| time | caller
1248 ls-2230 3d... 0us+: _raw_spin_lock_irqsave <-ata_scsi_queuecmd
1249 ls-2230 3...1 100us : _raw_spin_unlock_irqrestore <-ata_scsi_queuecmd
1250 ls-2230 3...1 101us+: trace_preempt_on <-ata_scsi_queuecmd
1251 ls-2230 3...1 111us : <stack trace>
1252 => sub_preempt_count
1253 => _raw_spin_unlock_irqrestore
1254 => ata_scsi_queuecmd
1255 => scsi_dispatch_cmd
1257 => __blk_run_queue_uncond
1260 => generic_make_request
1265 => htree_dirblock_to_tree
1266 => ext3_htree_fill_tree
1270 => system_call_fastpath
1273 The trace_hardirqs_off_thunk is called from assembly on x86 when
1274 interrupts are disabled in the assembly code. Without the
1275 function tracing, we do not know if interrupts were enabled
1276 within the preemption points. We do see that it started with
1279 Here is a trace with function-trace set:
1281 # tracer: preemptirqsoff
1283 # preemptirqsoff latency trace v1.1.5 on 3.8.0-test+
1284 # --------------------------------------------------------------------
1285 # latency: 161 us, #339/339, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1287 # | task: ls-2269 (uid:0 nice:0 policy:0 rt_prio:0)
1289 # => started at: schedule
1290 # => ended at: mutex_unlock
1294 # / _-----=> irqs-off
1295 # | / _----=> need-resched
1296 # || / _---=> hardirq/softirq
1297 # ||| / _--=> preempt-depth
1299 # cmd pid ||||| time | caller
1301 kworker/-59 3...1 0us : __schedule <-schedule
1302 kworker/-59 3d..1 0us : rcu_preempt_qs <-rcu_note_context_switch
1303 kworker/-59 3d..1 1us : add_preempt_count <-_raw_spin_lock_irq
1304 kworker/-59 3d..2 1us : deactivate_task <-__schedule
1305 kworker/-59 3d..2 1us : dequeue_task <-deactivate_task
1306 kworker/-59 3d..2 2us : update_rq_clock <-dequeue_task
1307 kworker/-59 3d..2 2us : dequeue_task_fair <-dequeue_task
1308 kworker/-59 3d..2 2us : update_curr <-dequeue_task_fair
1309 kworker/-59 3d..2 2us : update_min_vruntime <-update_curr
1310 kworker/-59 3d..2 3us : cpuacct_charge <-update_curr
1311 kworker/-59 3d..2 3us : __rcu_read_lock <-cpuacct_charge
1312 kworker/-59 3d..2 3us : __rcu_read_unlock <-cpuacct_charge
1313 kworker/-59 3d..2 3us : update_cfs_rq_blocked_load <-dequeue_task_fair
1314 kworker/-59 3d..2 4us : clear_buddies <-dequeue_task_fair
1315 kworker/-59 3d..2 4us : account_entity_dequeue <-dequeue_task_fair
1316 kworker/-59 3d..2 4us : update_min_vruntime <-dequeue_task_fair
1317 kworker/-59 3d..2 4us : update_cfs_shares <-dequeue_task_fair
1318 kworker/-59 3d..2 5us : hrtick_update <-dequeue_task_fair
1319 kworker/-59 3d..2 5us : wq_worker_sleeping <-__schedule
1320 kworker/-59 3d..2 5us : kthread_data <-wq_worker_sleeping
1321 kworker/-59 3d..2 5us : put_prev_task_fair <-__schedule
1322 kworker/-59 3d..2 6us : pick_next_task_fair <-pick_next_task
1323 kworker/-59 3d..2 6us : clear_buddies <-pick_next_task_fair
1324 kworker/-59 3d..2 6us : set_next_entity <-pick_next_task_fair
1325 kworker/-59 3d..2 6us : update_stats_wait_end <-set_next_entity
1326 ls-2269 3d..2 7us : finish_task_switch <-__schedule
1327 ls-2269 3d..2 7us : _raw_spin_unlock_irq <-finish_task_switch
1328 ls-2269 3d..2 8us : do_IRQ <-ret_from_intr
1329 ls-2269 3d..2 8us : irq_enter <-do_IRQ
1330 ls-2269 3d..2 8us : rcu_irq_enter <-irq_enter
1331 ls-2269 3d..2 9us : add_preempt_count <-irq_enter
1332 ls-2269 3d.h2 9us : exit_idle <-do_IRQ
1334 ls-2269 3d.h3 20us : sub_preempt_count <-_raw_spin_unlock
1335 ls-2269 3d.h2 20us : irq_exit <-do_IRQ
1336 ls-2269 3d.h2 21us : sub_preempt_count <-irq_exit
1337 ls-2269 3d..3 21us : do_softirq <-irq_exit
1338 ls-2269 3d..3 21us : __do_softirq <-call_softirq
1339 ls-2269 3d..3 21us+: __local_bh_disable <-__do_softirq
1340 ls-2269 3d.s4 29us : sub_preempt_count <-_local_bh_enable_ip
1341 ls-2269 3d.s5 29us : sub_preempt_count <-_local_bh_enable_ip
1342 ls-2269 3d.s5 31us : do_IRQ <-ret_from_intr
1343 ls-2269 3d.s5 31us : irq_enter <-do_IRQ
1344 ls-2269 3d.s5 31us : rcu_irq_enter <-irq_enter
1346 ls-2269 3d.s5 31us : rcu_irq_enter <-irq_enter
1347 ls-2269 3d.s5 32us : add_preempt_count <-irq_enter
1348 ls-2269 3d.H5 32us : exit_idle <-do_IRQ
1349 ls-2269 3d.H5 32us : handle_irq <-do_IRQ
1350 ls-2269 3d.H5 32us : irq_to_desc <-handle_irq
1351 ls-2269 3d.H5 33us : handle_fasteoi_irq <-handle_irq
1353 ls-2269 3d.s5 158us : _raw_spin_unlock_irqrestore <-rtl8139_poll
1354 ls-2269 3d.s3 158us : net_rps_action_and_irq_enable.isra.65 <-net_rx_action
1355 ls-2269 3d.s3 159us : __local_bh_enable <-__do_softirq
1356 ls-2269 3d.s3 159us : sub_preempt_count <-__local_bh_enable
1357 ls-2269 3d..3 159us : idle_cpu <-irq_exit
1358 ls-2269 3d..3 159us : rcu_irq_exit <-irq_exit
1359 ls-2269 3d..3 160us : sub_preempt_count <-irq_exit
1360 ls-2269 3d... 161us : __mutex_unlock_slowpath <-mutex_unlock
1361 ls-2269 3d... 162us+: trace_hardirqs_on <-mutex_unlock
1362 ls-2269 3d... 186us : <stack trace>
1363 => __mutex_unlock_slowpath
1370 => system_call_fastpath
1372 This is an interesting trace. It started with kworker running and
1373 scheduling out and ls taking over. But as soon as ls released the
1374 rq lock and enabled interrupts (but not preemption) an interrupt
1375 triggered. When the interrupt finished, it started running softirqs.
1376 But while the softirq was running, another interrupt triggered.
1377 When an interrupt is running inside a softirq, the annotation is 'H'.
1383 One common case that people are interested in tracing is the
1384 time it takes for a task that is woken to actually wake up.
1385 Now for non Real-Time tasks, this can be arbitrary. But tracing
1386 it none the less can be interesting.
1388 Without function tracing:
1390 # echo 0 > options/function-trace
1391 # echo wakeup > current_tracer
1392 # echo 1 > tracing_on
1393 # echo 0 > tracing_max_latency
1395 # echo 0 > tracing_on
1399 # wakeup latency trace v1.1.5 on 3.8.0-test+
1400 # --------------------------------------------------------------------
1401 # latency: 15 us, #4/4, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1403 # | task: kworker/3:1H-312 (uid:0 nice:-20 policy:0 rt_prio:0)
1407 # / _-----=> irqs-off
1408 # | / _----=> need-resched
1409 # || / _---=> hardirq/softirq
1410 # ||| / _--=> preempt-depth
1412 # cmd pid ||||| time | caller
1414 <idle>-0 3dNs7 0us : 0:120:R + [003] 312:100:R kworker/3:1H
1415 <idle>-0 3dNs7 1us+: ttwu_do_activate.constprop.87 <-try_to_wake_up
1416 <idle>-0 3d..3 15us : __schedule <-schedule
1417 <idle>-0 3d..3 15us : 0:120:R ==> [003] 312:100:R kworker/3:1H
1419 The tracer only traces the highest priority task in the system
1420 to avoid tracing the normal circumstances. Here we see that
1421 the kworker with a nice priority of -20 (not very nice), took
1422 just 15 microseconds from the time it woke up, to the time it
1425 Non Real-Time tasks are not that interesting. A more interesting
1426 trace is to concentrate only on Real-Time tasks.
1431 In a Real-Time environment it is very important to know the
1432 wakeup time it takes for the highest priority task that is woken
1433 up to the time that it executes. This is also known as "schedule
1434 latency". I stress the point that this is about RT tasks. It is
1435 also important to know the scheduling latency of non-RT tasks,
1436 but the average schedule latency is better for non-RT tasks.
1437 Tools like LatencyTop are more appropriate for such
1440 Real-Time environments are interested in the worst case latency.
1441 That is the longest latency it takes for something to happen,
1442 and not the average. We can have a very fast scheduler that may
1443 only have a large latency once in a while, but that would not
1444 work well with Real-Time tasks. The wakeup_rt tracer was designed
1445 to record the worst case wakeups of RT tasks. Non-RT tasks are
1446 not recorded because the tracer only records one worst case and
1447 tracing non-RT tasks that are unpredictable will overwrite the
1448 worst case latency of RT tasks (just run the normal wakeup
1449 tracer for a while to see that effect).
1451 Since this tracer only deals with RT tasks, we will run this
1452 slightly differently than we did with the previous tracers.
1453 Instead of performing an 'ls', we will run 'sleep 1' under
1454 'chrt' which changes the priority of the task.
1456 # echo 0 > options/function-trace
1457 # echo wakeup_rt > current_tracer
1458 # echo 1 > tracing_on
1459 # echo 0 > tracing_max_latency
1461 # echo 0 > tracing_on
1467 # wakeup_rt latency trace v1.1.5 on 3.8.0-test+
1468 # --------------------------------------------------------------------
1469 # latency: 5 us, #4/4, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1471 # | task: sleep-2389 (uid:0 nice:0 policy:1 rt_prio:5)
1475 # / _-----=> irqs-off
1476 # | / _----=> need-resched
1477 # || / _---=> hardirq/softirq
1478 # ||| / _--=> preempt-depth
1480 # cmd pid ||||| time | caller
1482 <idle>-0 3d.h4 0us : 0:120:R + [003] 2389: 94:R sleep
1483 <idle>-0 3d.h4 1us+: ttwu_do_activate.constprop.87 <-try_to_wake_up
1484 <idle>-0 3d..3 5us : __schedule <-schedule
1485 <idle>-0 3d..3 5us : 0:120:R ==> [003] 2389: 94:R sleep
1488 Running this on an idle system, we see that it only took 5 microseconds
1489 to perform the task switch. Note, since the trace point in the schedule
1490 is before the actual "switch", we stop the tracing when the recorded task
1491 is about to schedule in. This may change if we add a new marker at the
1492 end of the scheduler.
1494 Notice that the recorded task is 'sleep' with the PID of 2389
1495 and it has an rt_prio of 5. This priority is user-space priority
1496 and not the internal kernel priority. The policy is 1 for
1497 SCHED_FIFO and 2 for SCHED_RR.
1499 Note, that the trace data shows the internal priority (99 - rtprio).
1501 <idle>-0 3d..3 5us : 0:120:R ==> [003] 2389: 94:R sleep
1503 The 0:120:R means idle was running with a nice priority of 0 (120 - 20)
1504 and in the running state 'R'. The sleep task was scheduled in with
1505 2389: 94:R. That is the priority is the kernel rtprio (99 - 5 = 94)
1506 and it too is in the running state.
1508 Doing the same with chrt -r 5 and function-trace set.
1510 echo 1 > options/function-trace
1514 # wakeup_rt latency trace v1.1.5 on 3.8.0-test+
1515 # --------------------------------------------------------------------
1516 # latency: 29 us, #85/85, CPU#3 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1518 # | task: sleep-2448 (uid:0 nice:0 policy:1 rt_prio:5)
1522 # / _-----=> irqs-off
1523 # | / _----=> need-resched
1524 # || / _---=> hardirq/softirq
1525 # ||| / _--=> preempt-depth
1527 # cmd pid ||||| time | caller
1529 <idle>-0 3d.h4 1us+: 0:120:R + [003] 2448: 94:R sleep
1530 <idle>-0 3d.h4 2us : ttwu_do_activate.constprop.87 <-try_to_wake_up
1531 <idle>-0 3d.h3 3us : check_preempt_curr <-ttwu_do_wakeup
1532 <idle>-0 3d.h3 3us : resched_curr <-check_preempt_curr
1533 <idle>-0 3dNh3 4us : task_woken_rt <-ttwu_do_wakeup
1534 <idle>-0 3dNh3 4us : _raw_spin_unlock <-try_to_wake_up
1535 <idle>-0 3dNh3 4us : sub_preempt_count <-_raw_spin_unlock
1536 <idle>-0 3dNh2 5us : ttwu_stat <-try_to_wake_up
1537 <idle>-0 3dNh2 5us : _raw_spin_unlock_irqrestore <-try_to_wake_up
1538 <idle>-0 3dNh2 6us : sub_preempt_count <-_raw_spin_unlock_irqrestore
1539 <idle>-0 3dNh1 6us : _raw_spin_lock <-__run_hrtimer
1540 <idle>-0 3dNh1 6us : add_preempt_count <-_raw_spin_lock
1541 <idle>-0 3dNh2 7us : _raw_spin_unlock <-hrtimer_interrupt
1542 <idle>-0 3dNh2 7us : sub_preempt_count <-_raw_spin_unlock
1543 <idle>-0 3dNh1 7us : tick_program_event <-hrtimer_interrupt
1544 <idle>-0 3dNh1 7us : clockevents_program_event <-tick_program_event
1545 <idle>-0 3dNh1 8us : ktime_get <-clockevents_program_event
1546 <idle>-0 3dNh1 8us : lapic_next_event <-clockevents_program_event
1547 <idle>-0 3dNh1 8us : irq_exit <-smp_apic_timer_interrupt
1548 <idle>-0 3dNh1 9us : sub_preempt_count <-irq_exit
1549 <idle>-0 3dN.2 9us : idle_cpu <-irq_exit
1550 <idle>-0 3dN.2 9us : rcu_irq_exit <-irq_exit
1551 <idle>-0 3dN.2 10us : rcu_eqs_enter_common.isra.45 <-rcu_irq_exit
1552 <idle>-0 3dN.2 10us : sub_preempt_count <-irq_exit
1553 <idle>-0 3.N.1 11us : rcu_idle_exit <-cpu_idle
1554 <idle>-0 3dN.1 11us : rcu_eqs_exit_common.isra.43 <-rcu_idle_exit
1555 <idle>-0 3.N.1 11us : tick_nohz_idle_exit <-cpu_idle
1556 <idle>-0 3dN.1 12us : menu_hrtimer_cancel <-tick_nohz_idle_exit
1557 <idle>-0 3dN.1 12us : ktime_get <-tick_nohz_idle_exit
1558 <idle>-0 3dN.1 12us : tick_do_update_jiffies64 <-tick_nohz_idle_exit
1559 <idle>-0 3dN.1 13us : update_cpu_load_nohz <-tick_nohz_idle_exit
1560 <idle>-0 3dN.1 13us : _raw_spin_lock <-update_cpu_load_nohz
1561 <idle>-0 3dN.1 13us : add_preempt_count <-_raw_spin_lock
1562 <idle>-0 3dN.2 13us : __update_cpu_load <-update_cpu_load_nohz
1563 <idle>-0 3dN.2 14us : sched_avg_update <-__update_cpu_load
1564 <idle>-0 3dN.2 14us : _raw_spin_unlock <-update_cpu_load_nohz
1565 <idle>-0 3dN.2 14us : sub_preempt_count <-_raw_spin_unlock
1566 <idle>-0 3dN.1 15us : calc_load_exit_idle <-tick_nohz_idle_exit
1567 <idle>-0 3dN.1 15us : touch_softlockup_watchdog <-tick_nohz_idle_exit
1568 <idle>-0 3dN.1 15us : hrtimer_cancel <-tick_nohz_idle_exit
1569 <idle>-0 3dN.1 15us : hrtimer_try_to_cancel <-hrtimer_cancel
1570 <idle>-0 3dN.1 16us : lock_hrtimer_base.isra.18 <-hrtimer_try_to_cancel
1571 <idle>-0 3dN.1 16us : _raw_spin_lock_irqsave <-lock_hrtimer_base.isra.18
1572 <idle>-0 3dN.1 16us : add_preempt_count <-_raw_spin_lock_irqsave
1573 <idle>-0 3dN.2 17us : __remove_hrtimer <-remove_hrtimer.part.16
1574 <idle>-0 3dN.2 17us : hrtimer_force_reprogram <-__remove_hrtimer
1575 <idle>-0 3dN.2 17us : tick_program_event <-hrtimer_force_reprogram
1576 <idle>-0 3dN.2 18us : clockevents_program_event <-tick_program_event
1577 <idle>-0 3dN.2 18us : ktime_get <-clockevents_program_event
1578 <idle>-0 3dN.2 18us : lapic_next_event <-clockevents_program_event
1579 <idle>-0 3dN.2 19us : _raw_spin_unlock_irqrestore <-hrtimer_try_to_cancel
1580 <idle>-0 3dN.2 19us : sub_preempt_count <-_raw_spin_unlock_irqrestore
1581 <idle>-0 3dN.1 19us : hrtimer_forward <-tick_nohz_idle_exit
1582 <idle>-0 3dN.1 20us : ktime_add_safe <-hrtimer_forward
1583 <idle>-0 3dN.1 20us : ktime_add_safe <-hrtimer_forward
1584 <idle>-0 3dN.1 20us : hrtimer_start_range_ns <-hrtimer_start_expires.constprop.11
1585 <idle>-0 3dN.1 20us : __hrtimer_start_range_ns <-hrtimer_start_range_ns
1586 <idle>-0 3dN.1 21us : lock_hrtimer_base.isra.18 <-__hrtimer_start_range_ns
1587 <idle>-0 3dN.1 21us : _raw_spin_lock_irqsave <-lock_hrtimer_base.isra.18
1588 <idle>-0 3dN.1 21us : add_preempt_count <-_raw_spin_lock_irqsave
1589 <idle>-0 3dN.2 22us : ktime_add_safe <-__hrtimer_start_range_ns
1590 <idle>-0 3dN.2 22us : enqueue_hrtimer <-__hrtimer_start_range_ns
1591 <idle>-0 3dN.2 22us : tick_program_event <-__hrtimer_start_range_ns
1592 <idle>-0 3dN.2 23us : clockevents_program_event <-tick_program_event
1593 <idle>-0 3dN.2 23us : ktime_get <-clockevents_program_event
1594 <idle>-0 3dN.2 23us : lapic_next_event <-clockevents_program_event
1595 <idle>-0 3dN.2 24us : _raw_spin_unlock_irqrestore <-__hrtimer_start_range_ns
1596 <idle>-0 3dN.2 24us : sub_preempt_count <-_raw_spin_unlock_irqrestore
1597 <idle>-0 3dN.1 24us : account_idle_ticks <-tick_nohz_idle_exit
1598 <idle>-0 3dN.1 24us : account_idle_time <-account_idle_ticks
1599 <idle>-0 3.N.1 25us : sub_preempt_count <-cpu_idle
1600 <idle>-0 3.N.. 25us : schedule <-cpu_idle
1601 <idle>-0 3.N.. 25us : __schedule <-preempt_schedule
1602 <idle>-0 3.N.. 26us : add_preempt_count <-__schedule
1603 <idle>-0 3.N.1 26us : rcu_note_context_switch <-__schedule
1604 <idle>-0 3.N.1 26us : rcu_sched_qs <-rcu_note_context_switch
1605 <idle>-0 3dN.1 27us : rcu_preempt_qs <-rcu_note_context_switch
1606 <idle>-0 3.N.1 27us : _raw_spin_lock_irq <-__schedule
1607 <idle>-0 3dN.1 27us : add_preempt_count <-_raw_spin_lock_irq
1608 <idle>-0 3dN.2 28us : put_prev_task_idle <-__schedule
1609 <idle>-0 3dN.2 28us : pick_next_task_stop <-pick_next_task
1610 <idle>-0 3dN.2 28us : pick_next_task_rt <-pick_next_task
1611 <idle>-0 3dN.2 29us : dequeue_pushable_task <-pick_next_task_rt
1612 <idle>-0 3d..3 29us : __schedule <-preempt_schedule
1613 <idle>-0 3d..3 30us : 0:120:R ==> [003] 2448: 94:R sleep
1615 This isn't that big of a trace, even with function tracing enabled,
1616 so I included the entire trace.
1618 The interrupt went off while when the system was idle. Somewhere
1619 before task_woken_rt() was called, the NEED_RESCHED flag was set,
1620 this is indicated by the first occurrence of the 'N' flag.
1622 Latency tracing and events
1623 --------------------------
1624 As function tracing can induce a much larger latency, but without
1625 seeing what happens within the latency it is hard to know what
1626 caused it. There is a middle ground, and that is with enabling
1629 # echo 0 > options/function-trace
1630 # echo wakeup_rt > current_tracer
1631 # echo 1 > events/enable
1632 # echo 1 > tracing_on
1633 # echo 0 > tracing_max_latency
1635 # echo 0 > tracing_on
1639 # wakeup_rt latency trace v1.1.5 on 3.8.0-test+
1640 # --------------------------------------------------------------------
1641 # latency: 6 us, #12/12, CPU#2 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:4)
1643 # | task: sleep-5882 (uid:0 nice:0 policy:1 rt_prio:5)
1647 # / _-----=> irqs-off
1648 # | / _----=> need-resched
1649 # || / _---=> hardirq/softirq
1650 # ||| / _--=> preempt-depth
1652 # cmd pid ||||| time | caller
1654 <idle>-0 2d.h4 0us : 0:120:R + [002] 5882: 94:R sleep
1655 <idle>-0 2d.h4 0us : ttwu_do_activate.constprop.87 <-try_to_wake_up
1656 <idle>-0 2d.h4 1us : sched_wakeup: comm=sleep pid=5882 prio=94 success=1 target_cpu=002
1657 <idle>-0 2dNh2 1us : hrtimer_expire_exit: hrtimer=ffff88007796feb8
1658 <idle>-0 2.N.2 2us : power_end: cpu_id=2
1659 <idle>-0 2.N.2 3us : cpu_idle: state=4294967295 cpu_id=2
1660 <idle>-0 2dN.3 4us : hrtimer_cancel: hrtimer=ffff88007d50d5e0
1661 <idle>-0 2dN.3 4us : hrtimer_start: hrtimer=ffff88007d50d5e0 function=tick_sched_timer expires=34311211000000 softexpires=34311211000000
1662 <idle>-0 2.N.2 5us : rcu_utilization: Start context switch
1663 <idle>-0 2.N.2 5us : rcu_utilization: End context switch
1664 <idle>-0 2d..3 6us : __schedule <-schedule
1665 <idle>-0 2d..3 6us : 0:120:R ==> [002] 5882: 94:R sleep
1671 This tracer is the function tracer. Enabling the function tracer
1672 can be done from the debug file system. Make sure the
1673 ftrace_enabled is set; otherwise this tracer is a nop.
1674 See the "ftrace_enabled" section below.
1676 # sysctl kernel.ftrace_enabled=1
1677 # echo function > current_tracer
1678 # echo 1 > tracing_on
1680 # echo 0 > tracing_on
1684 # entries-in-buffer/entries-written: 24799/24799 #P:4
1687 # / _----=> need-resched
1688 # | / _---=> hardirq/softirq
1689 # || / _--=> preempt-depth
1691 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
1693 bash-1994 [002] .... 3082.063030: mutex_unlock <-rb_simple_write
1694 bash-1994 [002] .... 3082.063031: __mutex_unlock_slowpath <-mutex_unlock
1695 bash-1994 [002] .... 3082.063031: __fsnotify_parent <-fsnotify_modify
1696 bash-1994 [002] .... 3082.063032: fsnotify <-fsnotify_modify
1697 bash-1994 [002] .... 3082.063032: __srcu_read_lock <-fsnotify
1698 bash-1994 [002] .... 3082.063032: add_preempt_count <-__srcu_read_lock
1699 bash-1994 [002] ...1 3082.063032: sub_preempt_count <-__srcu_read_lock
1700 bash-1994 [002] .... 3082.063033: __srcu_read_unlock <-fsnotify
1704 Note: function tracer uses ring buffers to store the above
1705 entries. The newest data may overwrite the oldest data.
1706 Sometimes using echo to stop the trace is not sufficient because
1707 the tracing could have overwritten the data that you wanted to
1708 record. For this reason, it is sometimes better to disable
1709 tracing directly from a program. This allows you to stop the
1710 tracing at the point that you hit the part that you are
1711 interested in. To disable the tracing directly from a C program,
1712 something like following code snippet can be used:
1716 int main(int argc, char *argv[]) {
1718 trace_fd = open(tracing_file("tracing_on"), O_WRONLY);
1720 if (condition_hit()) {
1721 write(trace_fd, "0", 1);
1727 Single thread tracing
1728 ---------------------
1730 By writing into set_ftrace_pid you can trace a
1731 single thread. For example:
1733 # cat set_ftrace_pid
1735 # echo 3111 > set_ftrace_pid
1736 # cat set_ftrace_pid
1738 # echo function > current_tracer
1742 # TASK-PID CPU# TIMESTAMP FUNCTION
1744 yum-updatesd-3111 [003] 1637.254676: finish_task_switch <-thread_return
1745 yum-updatesd-3111 [003] 1637.254681: hrtimer_cancel <-schedule_hrtimeout_range
1746 yum-updatesd-3111 [003] 1637.254682: hrtimer_try_to_cancel <-hrtimer_cancel
1747 yum-updatesd-3111 [003] 1637.254683: lock_hrtimer_base <-hrtimer_try_to_cancel
1748 yum-updatesd-3111 [003] 1637.254685: fget_light <-do_sys_poll
1749 yum-updatesd-3111 [003] 1637.254686: pipe_poll <-do_sys_poll
1750 # echo > set_ftrace_pid
1754 # TASK-PID CPU# TIMESTAMP FUNCTION
1756 ##### CPU 3 buffer started ####
1757 yum-updatesd-3111 [003] 1701.957688: free_poll_entry <-poll_freewait
1758 yum-updatesd-3111 [003] 1701.957689: remove_wait_queue <-free_poll_entry
1759 yum-updatesd-3111 [003] 1701.957691: fput <-free_poll_entry
1760 yum-updatesd-3111 [003] 1701.957692: audit_syscall_exit <-sysret_audit
1761 yum-updatesd-3111 [003] 1701.957693: path_put <-audit_syscall_exit
1763 If you want to trace a function when executing, you could use
1764 something like this simple program:
1768 #include <sys/types.h>
1769 #include <sys/stat.h>
1775 #define STR(x) _STR(x)
1776 #define MAX_PATH 256
1778 const char *find_debugfs(void)
1780 static char debugfs[MAX_PATH+1];
1781 static int debugfs_found;
1788 if ((fp = fopen("/proc/mounts","r")) == NULL) {
1789 perror("/proc/mounts");
1793 while (fscanf(fp, "%*s %"
1795 "s %99s %*s %*d %*d\n",
1796 debugfs, type) == 2) {
1797 if (strcmp(type, "debugfs") == 0)
1802 if (strcmp(type, "debugfs") != 0) {
1803 fprintf(stderr, "debugfs not mounted");
1807 strcat(debugfs, "/tracing/");
1813 const char *tracing_file(const char *file_name)
1815 static char trace_file[MAX_PATH+1];
1816 snprintf(trace_file, MAX_PATH, "%s/%s", find_debugfs(), file_name);
1820 int main (int argc, char **argv)
1830 ffd = open(tracing_file("current_tracer"), O_WRONLY);
1833 write(ffd, "nop", 3);
1835 fd = open(tracing_file("set_ftrace_pid"), O_WRONLY);
1836 s = sprintf(line, "%d\n", getpid());
1839 write(ffd, "function", 8);
1844 execvp(argv[1], argv+1);
1850 Or this simple script!
1855 debugfs=`sed -ne 's/^debugfs \(.*\) debugfs.*/\1/p' /proc/mounts`
1856 echo nop > $debugfs/tracing/current_tracer
1857 echo 0 > $debugfs/tracing/tracing_on
1858 echo $$ > $debugfs/tracing/set_ftrace_pid
1859 echo function > $debugfs/tracing/current_tracer
1860 echo 1 > $debugfs/tracing/tracing_on
1865 function graph tracer
1866 ---------------------------
1868 This tracer is similar to the function tracer except that it
1869 probes a function on its entry and its exit. This is done by
1870 using a dynamically allocated stack of return addresses in each
1871 task_struct. On function entry the tracer overwrites the return
1872 address of each function traced to set a custom probe. Thus the
1873 original return address is stored on the stack of return address
1876 Probing on both ends of a function leads to special features
1879 - measure of a function's time execution
1880 - having a reliable call stack to draw function calls graph
1882 This tracer is useful in several situations:
1884 - you want to find the reason of a strange kernel behavior and
1885 need to see what happens in detail on any areas (or specific
1888 - you are experiencing weird latencies but it's difficult to
1891 - you want to find quickly which path is taken by a specific
1894 - you just want to peek inside a working kernel and want to see
1897 # tracer: function_graph
1899 # CPU DURATION FUNCTION CALLS
1903 0) | do_sys_open() {
1905 0) | kmem_cache_alloc() {
1906 0) 1.382 us | __might_sleep();
1908 0) | strncpy_from_user() {
1909 0) | might_fault() {
1910 0) 1.389 us | __might_sleep();
1915 0) 0.668 us | _spin_lock();
1916 0) 0.570 us | expand_files();
1917 0) 0.586 us | _spin_unlock();
1920 There are several columns that can be dynamically
1921 enabled/disabled. You can use every combination of options you
1922 want, depending on your needs.
1924 - The cpu number on which the function executed is default
1925 enabled. It is sometimes better to only trace one cpu (see
1926 tracing_cpu_mask file) or you might sometimes see unordered
1927 function calls while cpu tracing switch.
1929 hide: echo nofuncgraph-cpu > trace_options
1930 show: echo funcgraph-cpu > trace_options
1932 - The duration (function's time of execution) is displayed on
1933 the closing bracket line of a function or on the same line
1934 than the current function in case of a leaf one. It is default
1937 hide: echo nofuncgraph-duration > trace_options
1938 show: echo funcgraph-duration > trace_options
1940 - The overhead field precedes the duration field in case of
1941 reached duration thresholds.
1943 hide: echo nofuncgraph-overhead > trace_options
1944 show: echo funcgraph-overhead > trace_options
1945 depends on: funcgraph-duration
1949 3) # 1837.709 us | } /* __switch_to */
1950 3) | finish_task_switch() {
1951 3) 0.313 us | _raw_spin_unlock_irq();
1953 3) # 1889.063 us | } /* __schedule */
1954 3) ! 140.417 us | } /* __schedule */
1955 3) # 2034.948 us | } /* schedule */
1956 3) * 33998.59 us | } /* schedule_preempt_disabled */
1960 1) 0.260 us | msecs_to_jiffies();
1961 1) 0.313 us | __rcu_read_unlock();
1964 1) 0.313 us | rcu_bh_qs();
1965 1) 0.313 us | __local_bh_enable();
1967 1) 0.365 us | idle_cpu();
1968 1) | rcu_irq_exit() {
1969 1) 0.417 us | rcu_eqs_enter_common.isra.47();
1973 1) @ 119760.2 us | }
1979 2) 0.417 us | scheduler_ipi();
1987 + means that the function exceeded 10 usecs.
1988 ! means that the function exceeded 100 usecs.
1989 # means that the function exceeded 1000 usecs.
1990 * means that the function exceeded 10 msecs.
1991 @ means that the function exceeded 100 msecs.
1992 $ means that the function exceeded 1 sec.
1995 - The task/pid field displays the thread cmdline and pid which
1996 executed the function. It is default disabled.
1998 hide: echo nofuncgraph-proc > trace_options
1999 show: echo funcgraph-proc > trace_options
2003 # tracer: function_graph
2005 # CPU TASK/PID DURATION FUNCTION CALLS
2007 0) sh-4802 | | d_free() {
2008 0) sh-4802 | | call_rcu() {
2009 0) sh-4802 | | __call_rcu() {
2010 0) sh-4802 | 0.616 us | rcu_process_gp_end();
2011 0) sh-4802 | 0.586 us | check_for_new_grace_period();
2012 0) sh-4802 | 2.899 us | }
2013 0) sh-4802 | 4.040 us | }
2014 0) sh-4802 | 5.151 us | }
2015 0) sh-4802 | + 49.370 us | }
2018 - The absolute time field is an absolute timestamp given by the
2019 system clock since it started. A snapshot of this time is
2020 given on each entry/exit of functions
2022 hide: echo nofuncgraph-abstime > trace_options
2023 show: echo funcgraph-abstime > trace_options
2028 # TIME CPU DURATION FUNCTION CALLS
2030 360.774522 | 1) 0.541 us | }
2031 360.774522 | 1) 4.663 us | }
2032 360.774523 | 1) 0.541 us | __wake_up_bit();
2033 360.774524 | 1) 6.796 us | }
2034 360.774524 | 1) 7.952 us | }
2035 360.774525 | 1) 9.063 us | }
2036 360.774525 | 1) 0.615 us | journal_mark_dirty();
2037 360.774527 | 1) 0.578 us | __brelse();
2038 360.774528 | 1) | reiserfs_prepare_for_journal() {
2039 360.774528 | 1) | unlock_buffer() {
2040 360.774529 | 1) | wake_up_bit() {
2041 360.774529 | 1) | bit_waitqueue() {
2042 360.774530 | 1) 0.594 us | __phys_addr();
2045 The function name is always displayed after the closing bracket
2046 for a function if the start of that function is not in the
2049 Display of the function name after the closing bracket may be
2050 enabled for functions whose start is in the trace buffer,
2051 allowing easier searching with grep for function durations.
2052 It is default disabled.
2054 hide: echo nofuncgraph-tail > trace_options
2055 show: echo funcgraph-tail > trace_options
2057 Example with nofuncgraph-tail (default):
2059 0) | kmem_cache_free() {
2060 0) 0.518 us | __phys_addr();
2064 Example with funcgraph-tail:
2066 0) | kmem_cache_free() {
2067 0) 0.518 us | __phys_addr();
2068 0) 1.757 us | } /* kmem_cache_free() */
2069 0) 2.861 us | } /* putname() */
2071 You can put some comments on specific functions by using
2072 trace_printk() For example, if you want to put a comment inside
2073 the __might_sleep() function, you just have to include
2074 <linux/ftrace.h> and call trace_printk() inside __might_sleep()
2076 trace_printk("I'm a comment!\n")
2080 1) | __might_sleep() {
2081 1) | /* I'm a comment! */
2085 You might find other useful features for this tracer in the
2086 following "dynamic ftrace" section such as tracing only specific
2092 If CONFIG_DYNAMIC_FTRACE is set, the system will run with
2093 virtually no overhead when function tracing is disabled. The way
2094 this works is the mcount function call (placed at the start of
2095 every kernel function, produced by the -pg switch in gcc),
2096 starts of pointing to a simple return. (Enabling FTRACE will
2097 include the -pg switch in the compiling of the kernel.)
2099 At compile time every C file object is run through the
2100 recordmcount program (located in the scripts directory). This
2101 program will parse the ELF headers in the C object to find all
2102 the locations in the .text section that call mcount. (Note, only
2103 white listed .text sections are processed, since processing other
2104 sections like .init.text may cause races due to those sections
2105 being freed unexpectedly).
2107 A new section called "__mcount_loc" is created that holds
2108 references to all the mcount call sites in the .text section.
2109 The recordmcount program re-links this section back into the
2110 original object. The final linking stage of the kernel will add all these
2111 references into a single table.
2113 On boot up, before SMP is initialized, the dynamic ftrace code
2114 scans this table and updates all the locations into nops. It
2115 also records the locations, which are added to the
2116 available_filter_functions list. Modules are processed as they
2117 are loaded and before they are executed. When a module is
2118 unloaded, it also removes its functions from the ftrace function
2119 list. This is automatic in the module unload code, and the
2120 module author does not need to worry about it.
2122 When tracing is enabled, the process of modifying the function
2123 tracepoints is dependent on architecture. The old method is to use
2124 kstop_machine to prevent races with the CPUs executing code being
2125 modified (which can cause the CPU to do undesirable things, especially
2126 if the modified code crosses cache (or page) boundaries), and the nops are
2127 patched back to calls. But this time, they do not call mcount
2128 (which is just a function stub). They now call into the ftrace
2131 The new method of modifying the function tracepoints is to place
2132 a breakpoint at the location to be modified, sync all CPUs, modify
2133 the rest of the instruction not covered by the breakpoint. Sync
2134 all CPUs again, and then remove the breakpoint with the finished
2135 version to the ftrace call site.
2137 Some archs do not even need to monkey around with the synchronization,
2138 and can just slap the new code on top of the old without any
2139 problems with other CPUs executing it at the same time.
2141 One special side-effect to the recording of the functions being
2142 traced is that we can now selectively choose which functions we
2143 wish to trace and which ones we want the mcount calls to remain
2146 Two files are used, one for enabling and one for disabling the
2147 tracing of specified functions. They are:
2155 A list of available functions that you can add to these files is
2158 available_filter_functions
2160 # cat available_filter_functions
2169 If I am only interested in sys_nanosleep and hrtimer_interrupt:
2171 # echo sys_nanosleep hrtimer_interrupt > set_ftrace_filter
2172 # echo function > current_tracer
2173 # echo 1 > tracing_on
2175 # echo 0 > tracing_on
2179 # entries-in-buffer/entries-written: 5/5 #P:4
2182 # / _----=> need-resched
2183 # | / _---=> hardirq/softirq
2184 # || / _--=> preempt-depth
2186 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
2188 usleep-2665 [001] .... 4186.475355: sys_nanosleep <-system_call_fastpath
2189 <idle>-0 [001] d.h1 4186.475409: hrtimer_interrupt <-smp_apic_timer_interrupt
2190 usleep-2665 [001] d.h1 4186.475426: hrtimer_interrupt <-smp_apic_timer_interrupt
2191 <idle>-0 [003] d.h1 4186.475426: hrtimer_interrupt <-smp_apic_timer_interrupt
2192 <idle>-0 [002] d.h1 4186.475427: hrtimer_interrupt <-smp_apic_timer_interrupt
2194 To see which functions are being traced, you can cat the file:
2196 # cat set_ftrace_filter
2201 Perhaps this is not enough. The filters also allow simple wild
2202 cards. Only the following are currently available
2204 <match>* - will match functions that begin with <match>
2205 *<match> - will match functions that end with <match>
2206 *<match>* - will match functions that have <match> in it
2208 These are the only wild cards which are supported.
2210 <match>*<match> will not work.
2212 Note: It is better to use quotes to enclose the wild cards,
2213 otherwise the shell may expand the parameters into names
2214 of files in the local directory.
2216 # echo 'hrtimer_*' > set_ftrace_filter
2222 # entries-in-buffer/entries-written: 897/897 #P:4
2225 # / _----=> need-resched
2226 # | / _---=> hardirq/softirq
2227 # || / _--=> preempt-depth
2229 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
2231 <idle>-0 [003] dN.1 4228.547803: hrtimer_cancel <-tick_nohz_idle_exit
2232 <idle>-0 [003] dN.1 4228.547804: hrtimer_try_to_cancel <-hrtimer_cancel
2233 <idle>-0 [003] dN.2 4228.547805: hrtimer_force_reprogram <-__remove_hrtimer
2234 <idle>-0 [003] dN.1 4228.547805: hrtimer_forward <-tick_nohz_idle_exit
2235 <idle>-0 [003] dN.1 4228.547805: hrtimer_start_range_ns <-hrtimer_start_expires.constprop.11
2236 <idle>-0 [003] d..1 4228.547858: hrtimer_get_next_event <-get_next_timer_interrupt
2237 <idle>-0 [003] d..1 4228.547859: hrtimer_start <-__tick_nohz_idle_enter
2238 <idle>-0 [003] d..2 4228.547860: hrtimer_force_reprogram <-__rem
2240 Notice that we lost the sys_nanosleep.
2242 # cat set_ftrace_filter
2247 hrtimer_try_to_cancel
2251 hrtimer_force_reprogram
2252 hrtimer_get_next_event
2256 hrtimer_get_remaining
2258 hrtimer_init_sleeper
2261 This is because the '>' and '>>' act just like they do in bash.
2262 To rewrite the filters, use '>'
2263 To append to the filters, use '>>'
2265 To clear out a filter so that all functions will be recorded
2268 # echo > set_ftrace_filter
2269 # cat set_ftrace_filter
2272 Again, now we want to append.
2274 # echo sys_nanosleep > set_ftrace_filter
2275 # cat set_ftrace_filter
2277 # echo 'hrtimer_*' >> set_ftrace_filter
2278 # cat set_ftrace_filter
2283 hrtimer_try_to_cancel
2287 hrtimer_force_reprogram
2288 hrtimer_get_next_event
2293 hrtimer_get_remaining
2295 hrtimer_init_sleeper
2298 The set_ftrace_notrace prevents those functions from being
2301 # echo '*preempt*' '*lock*' > set_ftrace_notrace
2307 # entries-in-buffer/entries-written: 39608/39608 #P:4
2310 # / _----=> need-resched
2311 # | / _---=> hardirq/softirq
2312 # || / _--=> preempt-depth
2314 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
2316 bash-1994 [000] .... 4342.324896: file_ra_state_init <-do_dentry_open
2317 bash-1994 [000] .... 4342.324897: open_check_o_direct <-do_last
2318 bash-1994 [000] .... 4342.324897: ima_file_check <-do_last
2319 bash-1994 [000] .... 4342.324898: process_measurement <-ima_file_check
2320 bash-1994 [000] .... 4342.324898: ima_get_action <-process_measurement
2321 bash-1994 [000] .... 4342.324898: ima_match_policy <-ima_get_action
2322 bash-1994 [000] .... 4342.324899: do_truncate <-do_last
2323 bash-1994 [000] .... 4342.324899: should_remove_suid <-do_truncate
2324 bash-1994 [000] .... 4342.324899: notify_change <-do_truncate
2325 bash-1994 [000] .... 4342.324900: current_fs_time <-notify_change
2326 bash-1994 [000] .... 4342.324900: current_kernel_time <-current_fs_time
2327 bash-1994 [000] .... 4342.324900: timespec_trunc <-current_fs_time
2329 We can see that there's no more lock or preempt tracing.
2332 Dynamic ftrace with the function graph tracer
2333 ---------------------------------------------
2335 Although what has been explained above concerns both the
2336 function tracer and the function-graph-tracer, there are some
2337 special features only available in the function-graph tracer.
2339 If you want to trace only one function and all of its children,
2340 you just have to echo its name into set_graph_function:
2342 echo __do_fault > set_graph_function
2344 will produce the following "expanded" trace of the __do_fault()
2348 0) | filemap_fault() {
2349 0) | find_lock_page() {
2350 0) 0.804 us | find_get_page();
2351 0) | __might_sleep() {
2355 0) 0.653 us | _spin_lock();
2356 0) 0.578 us | page_add_file_rmap();
2357 0) 0.525 us | native_set_pte_at();
2358 0) 0.585 us | _spin_unlock();
2359 0) | unlock_page() {
2360 0) 0.541 us | page_waitqueue();
2361 0) 0.639 us | __wake_up_bit();
2365 0) | filemap_fault() {
2366 0) | find_lock_page() {
2367 0) 0.698 us | find_get_page();
2368 0) | __might_sleep() {
2372 0) 0.631 us | _spin_lock();
2373 0) 0.571 us | page_add_file_rmap();
2374 0) 0.526 us | native_set_pte_at();
2375 0) 0.586 us | _spin_unlock();
2376 0) | unlock_page() {
2377 0) 0.533 us | page_waitqueue();
2378 0) 0.638 us | __wake_up_bit();
2382 You can also expand several functions at once:
2384 echo sys_open > set_graph_function
2385 echo sys_close >> set_graph_function
2387 Now if you want to go back to trace all functions you can clear
2388 this special filter via:
2390 echo > set_graph_function
2396 Note, the proc sysctl ftrace_enable is a big on/off switch for the
2397 function tracer. By default it is enabled (when function tracing is
2398 enabled in the kernel). If it is disabled, all function tracing is
2399 disabled. This includes not only the function tracers for ftrace, but
2400 also for any other uses (perf, kprobes, stack tracing, profiling, etc).
2402 Please disable this with care.
2404 This can be disable (and enabled) with:
2406 sysctl kernel.ftrace_enabled=0
2407 sysctl kernel.ftrace_enabled=1
2411 echo 0 > /proc/sys/kernel/ftrace_enabled
2412 echo 1 > /proc/sys/kernel/ftrace_enabled
2418 A few commands are supported by the set_ftrace_filter interface.
2419 Trace commands have the following format:
2421 <function>:<command>:<parameter>
2423 The following commands are supported:
2426 This command enables function filtering per module. The
2427 parameter defines the module. For example, if only the write*
2428 functions in the ext3 module are desired, run:
2430 echo 'write*:mod:ext3' > set_ftrace_filter
2432 This command interacts with the filter in the same way as
2433 filtering based on function names. Thus, adding more functions
2434 in a different module is accomplished by appending (>>) to the
2435 filter file. Remove specific module functions by prepending
2438 echo '!writeback*:mod:ext3' >> set_ftrace_filter
2441 These commands turn tracing on and off when the specified
2442 functions are hit. The parameter determines how many times the
2443 tracing system is turned on and off. If unspecified, there is
2444 no limit. For example, to disable tracing when a schedule bug
2445 is hit the first 5 times, run:
2447 echo '__schedule_bug:traceoff:5' > set_ftrace_filter
2449 To always disable tracing when __schedule_bug is hit:
2451 echo '__schedule_bug:traceoff' > set_ftrace_filter
2453 These commands are cumulative whether or not they are appended
2454 to set_ftrace_filter. To remove a command, prepend it by '!'
2455 and drop the parameter:
2457 echo '!__schedule_bug:traceoff:0' > set_ftrace_filter
2459 The above removes the traceoff command for __schedule_bug
2460 that have a counter. To remove commands without counters:
2462 echo '!__schedule_bug:traceoff' > set_ftrace_filter
2465 Will cause a snapshot to be triggered when the function is hit.
2467 echo 'native_flush_tlb_others:snapshot' > set_ftrace_filter
2469 To only snapshot once:
2471 echo 'native_flush_tlb_others:snapshot:1' > set_ftrace_filter
2473 To remove the above commands:
2475 echo '!native_flush_tlb_others:snapshot' > set_ftrace_filter
2476 echo '!native_flush_tlb_others:snapshot:0' > set_ftrace_filter
2478 - enable_event/disable_event
2479 These commands can enable or disable a trace event. Note, because
2480 function tracing callbacks are very sensitive, when these commands
2481 are registered, the trace point is activated, but disabled in
2482 a "soft" mode. That is, the tracepoint will be called, but
2483 just will not be traced. The event tracepoint stays in this mode
2484 as long as there's a command that triggers it.
2486 echo 'try_to_wake_up:enable_event:sched:sched_switch:2' > \
2491 <function>:enable_event:<system>:<event>[:count]
2492 <function>:disable_event:<system>:<event>[:count]
2494 To remove the events commands:
2497 echo '!try_to_wake_up:enable_event:sched:sched_switch:0' > \
2499 echo '!schedule:disable_event:sched:sched_switch' > \
2503 When the function is hit, it will dump the contents of the ftrace
2504 ring buffer to the console. This is useful if you need to debug
2505 something, and want to dump the trace when a certain function
2506 is hit. Perhaps its a function that is called before a tripple
2507 fault happens and does not allow you to get a regular dump.
2510 When the function is hit, it will dump the contents of the ftrace
2511 ring buffer for the current CPU to the console. Unlike the "dump"
2512 command, it only prints out the contents of the ring buffer for the
2513 CPU that executed the function that triggered the dump.
2518 The trace_pipe outputs the same content as the trace file, but
2519 the effect on the tracing is different. Every read from
2520 trace_pipe is consumed. This means that subsequent reads will be
2521 different. The trace is live.
2523 # echo function > current_tracer
2524 # cat trace_pipe > /tmp/trace.out &
2526 # echo 1 > tracing_on
2528 # echo 0 > tracing_on
2532 # entries-in-buffer/entries-written: 0/0 #P:4
2535 # / _----=> need-resched
2536 # | / _---=> hardirq/softirq
2537 # || / _--=> preempt-depth
2539 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
2543 # cat /tmp/trace.out
2544 bash-1994 [000] .... 5281.568961: mutex_unlock <-rb_simple_write
2545 bash-1994 [000] .... 5281.568963: __mutex_unlock_slowpath <-mutex_unlock
2546 bash-1994 [000] .... 5281.568963: __fsnotify_parent <-fsnotify_modify
2547 bash-1994 [000] .... 5281.568964: fsnotify <-fsnotify_modify
2548 bash-1994 [000] .... 5281.568964: __srcu_read_lock <-fsnotify
2549 bash-1994 [000] .... 5281.568964: add_preempt_count <-__srcu_read_lock
2550 bash-1994 [000] ...1 5281.568965: sub_preempt_count <-__srcu_read_lock
2551 bash-1994 [000] .... 5281.568965: __srcu_read_unlock <-fsnotify
2552 bash-1994 [000] .... 5281.568967: sys_dup2 <-system_call_fastpath
2555 Note, reading the trace_pipe file will block until more input is
2561 Having too much or not enough data can be troublesome in
2562 diagnosing an issue in the kernel. The file buffer_size_kb is
2563 used to modify the size of the internal trace buffers. The
2564 number listed is the number of entries that can be recorded per
2565 CPU. To know the full size, multiply the number of possible CPUs
2566 with the number of entries.
2568 # cat buffer_size_kb
2569 1408 (units kilobytes)
2571 Or simply read buffer_total_size_kb
2573 # cat buffer_total_size_kb
2576 To modify the buffer, simple echo in a number (in 1024 byte segments).
2578 # echo 10000 > buffer_size_kb
2579 # cat buffer_size_kb
2580 10000 (units kilobytes)
2582 It will try to allocate as much as possible. If you allocate too
2583 much, it can cause Out-Of-Memory to trigger.
2585 # echo 1000000000000 > buffer_size_kb
2586 -bash: echo: write error: Cannot allocate memory
2587 # cat buffer_size_kb
2590 The per_cpu buffers can be changed individually as well:
2592 # echo 10000 > per_cpu/cpu0/buffer_size_kb
2593 # echo 100 > per_cpu/cpu1/buffer_size_kb
2595 When the per_cpu buffers are not the same, the buffer_size_kb
2596 at the top level will just show an X
2598 # cat buffer_size_kb
2601 This is where the buffer_total_size_kb is useful:
2603 # cat buffer_total_size_kb
2606 Writing to the top level buffer_size_kb will reset all the buffers
2607 to be the same again.
2611 CONFIG_TRACER_SNAPSHOT makes a generic snapshot feature
2612 available to all non latency tracers. (Latency tracers which
2613 record max latency, such as "irqsoff" or "wakeup", can't use
2614 this feature, since those are already using the snapshot
2615 mechanism internally.)
2617 Snapshot preserves a current trace buffer at a particular point
2618 in time without stopping tracing. Ftrace swaps the current
2619 buffer with a spare buffer, and tracing continues in the new
2620 current (=previous spare) buffer.
2622 The following debugfs files in "tracing" are related to this
2627 This is used to take a snapshot and to read the output
2628 of the snapshot. Echo 1 into this file to allocate a
2629 spare buffer and to take a snapshot (swap), then read
2630 the snapshot from this file in the same format as
2631 "trace" (described above in the section "The File
2632 System"). Both reads snapshot and tracing are executable
2633 in parallel. When the spare buffer is allocated, echoing
2634 0 frees it, and echoing else (positive) values clear the
2636 More details are shown in the table below.
2638 status\input | 0 | 1 | else |
2639 --------------+------------+------------+------------+
2640 not allocated |(do nothing)| alloc+swap |(do nothing)|
2641 --------------+------------+------------+------------+
2642 allocated | free | swap | clear |
2643 --------------+------------+------------+------------+
2645 Here is an example of using the snapshot feature.
2647 # echo 1 > events/sched/enable
2652 # entries-in-buffer/entries-written: 71/71 #P:8
2655 # / _----=> need-resched
2656 # | / _---=> hardirq/softirq
2657 # || / _--=> preempt-depth
2659 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
2661 <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
2662 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
2664 <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
2669 # entries-in-buffer/entries-written: 77/77 #P:8
2672 # / _----=> need-resched
2673 # | / _---=> hardirq/softirq
2674 # || / _--=> preempt-depth
2676 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
2678 <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
2679 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
2683 If you try to use this snapshot feature when current tracer is
2684 one of the latency tracers, you will get the following results.
2686 # echo wakeup > current_tracer
2688 bash: echo: write error: Device or resource busy
2690 cat: snapshot: Device or resource busy
2695 In the debugfs tracing directory is a directory called "instances".
2696 This directory can have new directories created inside of it using
2697 mkdir, and removing directories with rmdir. The directory created
2698 with mkdir in this directory will already contain files and other
2699 directories after it is created.
2701 # mkdir instances/foo
2703 buffer_size_kb buffer_total_size_kb events free_buffer per_cpu
2704 set_event snapshot trace trace_clock trace_marker trace_options
2705 trace_pipe tracing_on
2707 As you can see, the new directory looks similar to the tracing directory
2708 itself. In fact, it is very similar, except that the buffer and
2709 events are agnostic from the main director, or from any other
2710 instances that are created.
2712 The files in the new directory work just like the files with the
2713 same name in the tracing directory except the buffer that is used
2714 is a separate and new buffer. The files affect that buffer but do not
2715 affect the main buffer with the exception of trace_options. Currently,
2716 the trace_options affect all instances and the top level buffer
2717 the same, but this may change in future releases. That is, options
2718 may become specific to the instance they reside in.
2720 Notice that none of the function tracer files are there, nor is
2721 current_tracer and available_tracers. This is because the buffers
2722 can currently only have events enabled for them.
2724 # mkdir instances/foo
2725 # mkdir instances/bar
2726 # mkdir instances/zoot
2727 # echo 100000 > buffer_size_kb
2728 # echo 1000 > instances/foo/buffer_size_kb
2729 # echo 5000 > instances/bar/per_cpu/cpu1/buffer_size_kb
2730 # echo function > current_trace
2731 # echo 1 > instances/foo/events/sched/sched_wakeup/enable
2732 # echo 1 > instances/foo/events/sched/sched_wakeup_new/enable
2733 # echo 1 > instances/foo/events/sched/sched_switch/enable
2734 # echo 1 > instances/bar/events/irq/enable
2735 # echo 1 > instances/zoot/events/syscalls/enable
2737 CPU:2 [LOST 11745 EVENTS]
2738 bash-2044 [002] .... 10594.481032: _raw_spin_lock_irqsave <-get_page_from_freelist
2739 bash-2044 [002] d... 10594.481032: add_preempt_count <-_raw_spin_lock_irqsave
2740 bash-2044 [002] d..1 10594.481032: __rmqueue <-get_page_from_freelist
2741 bash-2044 [002] d..1 10594.481033: _raw_spin_unlock <-get_page_from_freelist
2742 bash-2044 [002] d..1 10594.481033: sub_preempt_count <-_raw_spin_unlock
2743 bash-2044 [002] d... 10594.481033: get_pageblock_flags_group <-get_pageblock_migratetype
2744 bash-2044 [002] d... 10594.481034: __mod_zone_page_state <-get_page_from_freelist
2745 bash-2044 [002] d... 10594.481034: zone_statistics <-get_page_from_freelist
2746 bash-2044 [002] d... 10594.481034: __inc_zone_state <-zone_statistics
2747 bash-2044 [002] d... 10594.481034: __inc_zone_state <-zone_statistics
2748 bash-2044 [002] .... 10594.481035: arch_dup_task_struct <-copy_process
2751 # cat instances/foo/trace_pipe
2752 bash-1998 [000] d..4 136.676759: sched_wakeup: comm=kworker/0:1 pid=59 prio=120 success=1 target_cpu=000
2753 bash-1998 [000] dN.4 136.676760: sched_wakeup: comm=bash pid=1998 prio=120 success=1 target_cpu=000
2754 <idle>-0 [003] d.h3 136.676906: sched_wakeup: comm=rcu_preempt pid=9 prio=120 success=1 target_cpu=003
2755 <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
2756 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
2757 bash-1998 [000] d..4 136.677014: sched_wakeup: comm=kworker/0:1 pid=59 prio=120 success=1 target_cpu=000
2758 bash-1998 [000] dN.4 136.677016: sched_wakeup: comm=bash pid=1998 prio=120 success=1 target_cpu=000
2759 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
2760 kworker/0:1-59 [000] d..4 136.677022: sched_wakeup: comm=sshd pid=1995 prio=120 success=1 target_cpu=001
2761 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
2764 # cat instances/bar/trace_pipe
2765 migration/1-14 [001] d.h3 138.732674: softirq_raise: vec=3 [action=NET_RX]
2766 <idle>-0 [001] dNh3 138.732725: softirq_raise: vec=3 [action=NET_RX]
2767 bash-1998 [000] d.h1 138.733101: softirq_raise: vec=1 [action=TIMER]
2768 bash-1998 [000] d.h1 138.733102: softirq_raise: vec=9 [action=RCU]
2769 bash-1998 [000] ..s2 138.733105: softirq_entry: vec=1 [action=TIMER]
2770 bash-1998 [000] ..s2 138.733106: softirq_exit: vec=1 [action=TIMER]
2771 bash-1998 [000] ..s2 138.733106: softirq_entry: vec=9 [action=RCU]
2772 bash-1998 [000] ..s2 138.733109: softirq_exit: vec=9 [action=RCU]
2773 sshd-1995 [001] d.h1 138.733278: irq_handler_entry: irq=21 name=uhci_hcd:usb4
2774 sshd-1995 [001] d.h1 138.733280: irq_handler_exit: irq=21 ret=unhandled
2775 sshd-1995 [001] d.h1 138.733281: irq_handler_entry: irq=21 name=eth0
2776 sshd-1995 [001] d.h1 138.733283: irq_handler_exit: irq=21 ret=handled
2779 # cat instances/zoot/trace
2782 # entries-in-buffer/entries-written: 18996/18996 #P:4
2785 # / _----=> need-resched
2786 # | / _---=> hardirq/softirq
2787 # || / _--=> preempt-depth
2789 # TASK-PID CPU# |||| TIMESTAMP FUNCTION
2791 bash-1998 [000] d... 140.733501: sys_write -> 0x2
2792 bash-1998 [000] d... 140.733504: sys_dup2(oldfd: a, newfd: 1)
2793 bash-1998 [000] d... 140.733506: sys_dup2 -> 0x1
2794 bash-1998 [000] d... 140.733508: sys_fcntl(fd: a, cmd: 1, arg: 0)
2795 bash-1998 [000] d... 140.733509: sys_fcntl -> 0x1
2796 bash-1998 [000] d... 140.733510: sys_close(fd: a)
2797 bash-1998 [000] d... 140.733510: sys_close -> 0x0
2798 bash-1998 [000] d... 140.733514: sys_rt_sigprocmask(how: 0, nset: 0, oset: 6e2768, sigsetsize: 8)
2799 bash-1998 [000] d... 140.733515: sys_rt_sigprocmask -> 0x0
2800 bash-1998 [000] d... 140.733516: sys_rt_sigaction(sig: 2, act: 7fff718846f0, oact: 7fff71884650, sigsetsize: 8)
2801 bash-1998 [000] d... 140.733516: sys_rt_sigaction -> 0x0
2803 You can see that the trace of the top most trace buffer shows only
2804 the function tracing. The foo instance displays wakeups and task
2807 To remove the instances, simply delete their directories:
2809 # rmdir instances/foo
2810 # rmdir instances/bar
2811 # rmdir instances/zoot
2813 Note, if a process has a trace file open in one of the instance
2814 directories, the rmdir will fail with EBUSY.
2819 Since the kernel has a fixed sized stack, it is important not to
2820 waste it in functions. A kernel developer must be conscience of
2821 what they allocate on the stack. If they add too much, the system
2822 can be in danger of a stack overflow, and corruption will occur,
2823 usually leading to a system panic.
2825 There are some tools that check this, usually with interrupts
2826 periodically checking usage. But if you can perform a check
2827 at every function call that will become very useful. As ftrace provides
2828 a function tracer, it makes it convenient to check the stack size
2829 at every function call. This is enabled via the stack tracer.
2831 CONFIG_STACK_TRACER enables the ftrace stack tracing functionality.
2832 To enable it, write a '1' into /proc/sys/kernel/stack_tracer_enabled.
2834 # echo 1 > /proc/sys/kernel/stack_tracer_enabled
2836 You can also enable it from the kernel command line to trace
2837 the stack size of the kernel during boot up, by adding "stacktrace"
2838 to the kernel command line parameter.
2840 After running it for a few minutes, the output looks like:
2842 # cat stack_max_size
2846 Depth Size Location (18 entries)
2848 0) 2928 224 update_sd_lb_stats+0xbc/0x4ac
2849 1) 2704 160 find_busiest_group+0x31/0x1f1
2850 2) 2544 256 load_balance+0xd9/0x662
2851 3) 2288 80 idle_balance+0xbb/0x130
2852 4) 2208 128 __schedule+0x26e/0x5b9
2853 5) 2080 16 schedule+0x64/0x66
2854 6) 2064 128 schedule_timeout+0x34/0xe0
2855 7) 1936 112 wait_for_common+0x97/0xf1
2856 8) 1824 16 wait_for_completion+0x1d/0x1f
2857 9) 1808 128 flush_work+0xfe/0x119
2858 10) 1680 16 tty_flush_to_ldisc+0x1e/0x20
2859 11) 1664 48 input_available_p+0x1d/0x5c
2860 12) 1616 48 n_tty_poll+0x6d/0x134
2861 13) 1568 64 tty_poll+0x64/0x7f
2862 14) 1504 880 do_select+0x31e/0x511
2863 15) 624 400 core_sys_select+0x177/0x216
2864 16) 224 96 sys_select+0x91/0xb9
2865 17) 128 128 system_call_fastpath+0x16/0x1b
2867 Note, if -mfentry is being used by gcc, functions get traced before
2868 they set up the stack frame. This means that leaf level functions
2869 are not tested by the stack tracer when -mfentry is used.
2871 Currently, -mfentry is used by gcc 4.6.0 and above on x86 only.
2875 More details can be found in the source code, in the
2876 kernel/trace/*.c files.