blob: 5ad2ded8aa6366a42c965049be83ffd0328f438e [file] [log] [blame]
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001 ftrace - Function Tracer
2 ========================
3
4Copyright 2008 Red Hat Inc.
Steven Rostedta41eeba2008-07-14 16:41:12 -04005 Author: Steven Rostedt <srostedt@redhat.com>
6 License: The GNU Free Documentation License, Version 1.2
Steven Rostedta97762a2008-07-31 12:40:52 -04007 (dual licensed under the GPL v2)
Steven Rostedtf2d9c742008-07-15 10:57:33 -04008Reviewers: Elias Oltmanns, Randy Dunlap, Andrew Morton,
9 John Kacur, and David Teigland.
Steven Rostedteb6d42e2008-07-10 12:46:01 -040010
Steven Rostedt42ec6322008-11-03 15:18:56 -050011Written for: 2.6.28-rc2
Steven Rostedteb6d42e2008-07-10 12:46:01 -040012
13Introduction
14------------
15
16Ftrace is an internal tracer designed to help out developers and
17designers of systems to find what is going on inside the kernel.
Ingo Molnar57526742009-02-19 12:54:10 +010018It can be used for debugging or analyzing latencies and
19performance issues that take place outside of user-space.
Steven Rostedteb6d42e2008-07-10 12:46:01 -040020
21Although ftrace is the function tracer, it also includes an
Ingo Molnar57526742009-02-19 12:54:10 +010022infrastructure that allows for other types of tracing. Some of
23the tracers that are currently in ftrace include a tracer to
24trace context switches, the time it takes for a high priority
25task to run after it was woken up, the time interrupts are
26disabled, and more (ftrace allows for tracer plugins, which
27means that the list of tracers can always grow).
Steven Rostedteb6d42e2008-07-10 12:46:01 -040028
29
30The File System
31---------------
32
Ingo Molnar57526742009-02-19 12:54:10 +010033Ftrace uses the debugfs file system to hold the control files as
34well as the files to display output.
Steven Rostedteb6d42e2008-07-10 12:46:01 -040035
36To mount the debugfs system:
37
38 # mkdir /debug
39 # mount -t debugfs nodev /debug
40
Ingo Molnar57526742009-02-19 12:54:10 +010041( Note: it is more common to mount at /sys/kernel/debug, but for
42 simplicity this document will use /debug)
Steven Rostedteb6d42e2008-07-10 12:46:01 -040043
44That's it! (assuming that you have ftrace configured into your kernel)
45
46After mounting the debugfs, you can see a directory called
47"tracing". This directory contains the control and output files
48of ftrace. Here is a list of some of the key files:
49
50
51 Note: all time values are in microseconds.
52
Ingo Molnar57526742009-02-19 12:54:10 +010053 current_tracer:
Steven Rostedteb6d42e2008-07-10 12:46:01 -040054
Ingo Molnar57526742009-02-19 12:54:10 +010055 This is used to set or display the current tracer
56 that is configured.
Steven Rostedteb6d42e2008-07-10 12:46:01 -040057
Ingo Molnar57526742009-02-19 12:54:10 +010058 available_tracers:
Steven Rostedteb6d42e2008-07-10 12:46:01 -040059
Ingo Molnar57526742009-02-19 12:54:10 +010060 This holds the different types of tracers that
61 have been compiled into the kernel. The
62 tracers listed here can be configured by
63 echoing their name into current_tracer.
Steven Rostedteb6d42e2008-07-10 12:46:01 -040064
Ingo Molnar57526742009-02-19 12:54:10 +010065 tracing_enabled:
Steven Rostedteb6d42e2008-07-10 12:46:01 -040066
Ingo Molnar57526742009-02-19 12:54:10 +010067 This sets or displays whether the current_tracer
68 is activated and tracing or not. Echo 0 into this
69 file to disable the tracer or 1 to enable it.
Steven Rostedteb6d42e2008-07-10 12:46:01 -040070
Ingo Molnar57526742009-02-19 12:54:10 +010071 trace:
Steven Rostedteb6d42e2008-07-10 12:46:01 -040072
Ingo Molnar57526742009-02-19 12:54:10 +010073 This file holds the output of the trace in a human
74 readable format (described below).
Steven Rostedteb6d42e2008-07-10 12:46:01 -040075
Ingo Molnar57526742009-02-19 12:54:10 +010076 latency_trace:
Steven Rostedteb6d42e2008-07-10 12:46:01 -040077
Ingo Molnar57526742009-02-19 12:54:10 +010078 This file shows the same trace but the information
79 is organized more to display possible latencies
80 in the system (described below).
Steven Rostedteb6d42e2008-07-10 12:46:01 -040081
Ingo Molnar57526742009-02-19 12:54:10 +010082 trace_pipe:
Steven Rostedteb6d42e2008-07-10 12:46:01 -040083
Ingo Molnar57526742009-02-19 12:54:10 +010084 The output is the same as the "trace" file but this
85 file is meant to be streamed with live tracing.
86 Reads from this file will block until new data
87 is retrieved. Unlike the "trace" and "latency_trace"
88 files, this file is a consumer. This means reading
89 from this file causes sequential reads to display
90 more current data. Once data is read from this
91 file, it is consumed, and will not be read
92 again with a sequential read. The "trace" and
93 "latency_trace" files are static, and if the
94 tracer is not adding more data, they will display
95 the same information every time they are read.
Steven Rostedteb6d42e2008-07-10 12:46:01 -040096
Ingo Molnar57526742009-02-19 12:54:10 +010097 trace_options:
Steven Rostedteb6d42e2008-07-10 12:46:01 -040098
Ingo Molnar57526742009-02-19 12:54:10 +010099 This file lets the user control the amount of data
100 that is displayed in one of the above output
101 files.
Steven Rostedtdf4fc312008-11-26 00:16:23 -0500102
KOSAKI Motohiro42b40b32009-03-07 23:55:09 +0900103 tracing_max_latency:
Frederic Weisbecker985ec202009-02-18 06:35:34 +0100104
Ingo Molnar57526742009-02-19 12:54:10 +0100105 Some of the tracers record the max latency.
106 For example, the time interrupts are disabled.
107 This time is saved in this file. The max trace
108 will also be stored, and displayed by either
109 "trace" or "latency_trace". A new max trace will
110 only be recorded if the latency is greater than
111 the value in this file. (in microseconds)
112
113 buffer_size_kb:
114
115 This sets or displays the number of kilobytes each CPU
116 buffer can hold. The tracer buffers are the same size
117 for each CPU. The displayed number is the size of the
118 CPU buffer and not total size of all buffers. The
119 trace buffers are allocated in pages (blocks of memory
120 that the kernel uses for allocation, usually 4 KB in size).
121 If the last page allocated has room for more bytes
122 than requested, the rest of the page will be used,
123 making the actual allocation bigger than requested.
124 ( Note, the size may not be a multiple of the page size
125 due to buffer managment overhead. )
126
127 This can only be updated when the current_tracer
128 is set to "nop".
129
130 tracing_cpumask:
131
132 This is a mask that lets the user only trace
133 on specified CPUS. The format is a hex string
134 representing the CPUS.
135
136 set_ftrace_filter:
137
138 When dynamic ftrace is configured in (see the
139 section below "dynamic ftrace"), the code is dynamically
140 modified (code text rewrite) to disable calling of the
141 function profiler (mcount). This lets tracing be configured
142 in with practically no overhead in performance. This also
143 has a side effect of enabling or disabling specific functions
144 to be traced. Echoing names of functions into this file
145 will limit the trace to only those functions.
146
147 set_ftrace_notrace:
148
149 This has an effect opposite to that of
150 set_ftrace_filter. Any function that is added here will not
151 be traced. If a function exists in both set_ftrace_filter
152 and set_ftrace_notrace, the function will _not_ be traced.
153
154 set_ftrace_pid:
155
156 Have the function tracer only trace a single thread.
157
158 set_graph_function:
159
160 Set a "trigger" function where tracing should start
161 with the function graph tracer (See the section
162 "dynamic ftrace" for more details).
163
164 available_filter_functions:
165
166 This lists the functions that ftrace
167 has processed and can trace. These are the function
168 names that you can pass to "set_ftrace_filter" or
169 "set_ftrace_notrace". (See the section "dynamic ftrace"
170 below for more details.)
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400171
172
173The Tracers
174-----------
175
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400176Here is the list of current tracers that may be configured.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400177
Ingo Molnar57526742009-02-19 12:54:10 +0100178 "function"
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400179
Ingo Molnar57526742009-02-19 12:54:10 +0100180 Function call tracer to trace all kernel functions.
Frederic Weisbecker985ec202009-02-18 06:35:34 +0100181
Mike Frysingerbc5c6c02009-06-10 04:48:41 -0400182 "function_graph"
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400183
Ingo Molnar57526742009-02-19 12:54:10 +0100184 Similar to the function tracer except that the
185 function tracer probes the functions on their entry
186 whereas the function graph tracer traces on both entry
187 and exit of the functions. It then provides the ability
188 to draw a graph of function calls similar to C code
189 source.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400190
Ingo Molnar57526742009-02-19 12:54:10 +0100191 "sched_switch"
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400192
Ingo Molnar57526742009-02-19 12:54:10 +0100193 Traces the context switches and wakeups between tasks.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400194
Ingo Molnar57526742009-02-19 12:54:10 +0100195 "irqsoff"
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400196
Ingo Molnar57526742009-02-19 12:54:10 +0100197 Traces the areas that disable interrupts and saves
198 the trace with the longest max latency.
199 See tracing_max_latency. When a new max is recorded,
200 it replaces the old trace. It is best to view this
201 trace via the latency_trace file.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400202
Ingo Molnar57526742009-02-19 12:54:10 +0100203 "preemptoff"
204
205 Similar to irqsoff but traces and records the amount of
206 time for which preemption is disabled.
207
208 "preemptirqsoff"
209
210 Similar to irqsoff and preemptoff, but traces and
211 records the largest time for which irqs and/or preemption
212 is disabled.
213
214 "wakeup"
215
216 Traces and records the max latency that it takes for
217 the highest priority task to get scheduled after
218 it has been woken up.
219
220 "hw-branch-tracer"
221
222 Uses the BTS CPU feature on x86 CPUs to traces all
223 branches executed.
224
225 "nop"
226
227 This is the "trace nothing" tracer. To remove all
228 tracers from tracing simply echo "nop" into
229 current_tracer.
Markus Metzgere2ea5392009-01-19 10:35:58 +0100230
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400231
232Examples of using the tracer
233----------------------------
234
Ingo Molnar57526742009-02-19 12:54:10 +0100235Here are typical examples of using the tracers when controlling
236them only with the debugfs interface (without using any
237user-land utilities).
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400238
239Output format:
240--------------
241
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400242Here is an example of the output format of the file "trace"
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400243
244 --------
Steven Rostedt9b803c02008-11-03 15:15:08 -0500245# tracer: function
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400246#
247# TASK-PID CPU# TIMESTAMP FUNCTION
248# | | | | |
249 bash-4251 [01] 10152.583854: path_put <-path_walk
250 bash-4251 [01] 10152.583855: dput <-path_put
251 bash-4251 [01] 10152.583855: _atomic_dec_and_lock <-dput
252 --------
253
Ingo Molnar57526742009-02-19 12:54:10 +0100254A header is printed with the tracer name that is represented by
255the trace. In this case the tracer is "function". Then a header
256showing the format. Task name "bash", the task PID "4251", the
257CPU that it was running on "01", the timestamp in <secs>.<usecs>
258format, the function name that was traced "path_put" and the
259parent function that called this function "path_walk". The
260timestamp is the time at which the function was entered.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400261
Ingo Molnar57526742009-02-19 12:54:10 +0100262The sched_switch tracer also includes tracing of task wakeups
263and context switches.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400264
265 ksoftirqd/1-7 [01] 1453.070013: 7:115:R + 2916:115:S
266 ksoftirqd/1-7 [01] 1453.070013: 7:115:R + 10:115:S
267 ksoftirqd/1-7 [01] 1453.070013: 7:115:R ==> 10:115:R
268 events/1-10 [01] 1453.070013: 10:115:S ==> 2916:115:R
269 kondemand/1-2916 [01] 1453.070013: 2916:115:S ==> 7:115:R
270 ksoftirqd/1-7 [01] 1453.070013: 7:115:S ==> 0:140:R
271
Ingo Molnar57526742009-02-19 12:54:10 +0100272Wake ups are represented by a "+" and the context switches are
273shown as "==>". The format is:
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400274
275 Context switches:
276
277 Previous task Next Task
278
279 <pid>:<prio>:<state> ==> <pid>:<prio>:<state>
280
281 Wake ups:
282
283 Current task Task waking up
284
285 <pid>:<prio>:<state> + <pid>:<prio>:<state>
286
Ingo Molnar57526742009-02-19 12:54:10 +0100287The prio is the internal kernel priority, which is the inverse
288of the priority that is usually displayed by user-space tools.
289Zero represents the highest priority (99). Prio 100 starts the
290"nice" priorities with 100 being equal to nice -20 and 139 being
291nice 19. The prio "140" is reserved for the idle task which is
292the lowest priority thread (pid 0).
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400293
294
295Latency trace format
296--------------------
297
Ingo Molnar57526742009-02-19 12:54:10 +0100298For traces that display latency times, the latency_trace file
299gives somewhat more information to see why a latency happened.
300Here is a typical trace.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400301
302# tracer: irqsoff
303#
304irqsoff latency trace v1.1.5 on 2.6.26-rc8
305--------------------------------------------------------------------
306 latency: 97 us, #3/3, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
307 -----------------
308 | task: swapper-0 (uid:0 nice:0 policy:0 rt_prio:0)
309 -----------------
310 => started at: apic_timer_interrupt
311 => ended at: do_softirq
312
313# _------=> CPU#
314# / _-----=> irqs-off
315# | / _----=> need-resched
316# || / _---=> hardirq/softirq
317# ||| / _--=> preempt-depth
318# |||| /
319# ||||| delay
320# cmd pid ||||| time | caller
321# \ / ||||| \ | /
322 <idle>-0 0d..1 0us+: trace_hardirqs_off_thunk (apic_timer_interrupt)
323 <idle>-0 0d.s. 97us : __do_softirq (do_softirq)
324 <idle>-0 0d.s1 98us : trace_hardirqs_on (do_softirq)
325
326
Ingo Molnar57526742009-02-19 12:54:10 +0100327This shows that the current tracer is "irqsoff" tracing the time
328for which interrupts were disabled. It gives the trace version
329and the version of the kernel upon which this was executed on
330(2.6.26-rc8). Then it displays the max latency in microsecs (97
331us). The number of trace entries displayed and the total number
332recorded (both are three: #3/3). The type of preemption that was
333used (PREEMPT). VP, KP, SP, and HP are always zero and are
334reserved for later use. #P is the number of online CPUS (#P:2).
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400335
Ingo Molnar57526742009-02-19 12:54:10 +0100336The task is the process that was running when the latency
337occurred. (swapper pid: 0).
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400338
Ingo Molnar57526742009-02-19 12:54:10 +0100339The start and stop (the functions in which the interrupts were
340disabled and enabled respectively) that caused the latencies:
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400341
342 apic_timer_interrupt is where the interrupts were disabled.
343 do_softirq is where they were enabled again.
344
345The next lines after the header are the trace itself. The header
346explains which is which.
347
348 cmd: The name of the process in the trace.
349
350 pid: The PID of that process.
351
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400352 CPU#: The CPU which the process was running on.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400353
354 irqs-off: 'd' interrupts are disabled. '.' otherwise.
Steven Rostedt92444892008-10-24 09:42:59 -0400355 Note: If the architecture does not support a way to
356 read the irq flags variable, an 'X' will always
357 be printed here.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400358
359 need-resched: 'N' task need_resched is set, '.' otherwise.
360
361 hardirq/softirq:
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400362 'H' - hard irq occurred inside a softirq.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400363 'h' - hard irq is running
364 's' - soft irq is running
365 '.' - normal context.
366
367 preempt-depth: The level of preempt_disabled
368
369The above is mostly meaningful for kernel developers.
370
Steven Rostedta41eeba2008-07-14 16:41:12 -0400371 time: This differs from the trace file output. The trace file output
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400372 includes an absolute timestamp. The timestamp used by the
Steven Rostedta41eeba2008-07-14 16:41:12 -0400373 latency_trace file is relative to the start of the trace.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400374
375 delay: This is just to help catch your eye a bit better. And
Ingo Molnar57526742009-02-19 12:54:10 +0100376 needs to be fixed to be only relative to the same CPU.
377 The marks are determined by the difference between this
378 current trace and the next trace.
379 '!' - greater than preempt_mark_thresh (default 100)
380 '+' - greater than 1 microsecond
381 ' ' - less than or equal to 1 microsecond.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400382
383 The rest is the same as the 'trace' file.
384
385
Steven Rostedtee6bce52008-11-12 17:52:37 -0500386trace_options
387-------------
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400388
Ingo Molnar57526742009-02-19 12:54:10 +0100389The trace_options file is used to control what gets printed in
390the trace output. To see what is available, simply cat the file:
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400391
Steven Rostedtee6bce52008-11-12 17:52:37 -0500392 cat /debug/tracing/trace_options
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400393 print-parent nosym-offset nosym-addr noverbose noraw nohex nobin \
Ingo Molnar57526742009-02-19 12:54:10 +0100394 noblock nostacktrace nosched-tree nouserstacktrace nosym-userobj
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400395
Ingo Molnar57526742009-02-19 12:54:10 +0100396To disable one of the options, echo in the option prepended with
397"no".
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400398
Steven Rostedtee6bce52008-11-12 17:52:37 -0500399 echo noprint-parent > /debug/tracing/trace_options
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400400
401To enable an option, leave off the "no".
402
Steven Rostedtee6bce52008-11-12 17:52:37 -0500403 echo sym-offset > /debug/tracing/trace_options
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400404
405Here are the available options:
406
Ingo Molnar57526742009-02-19 12:54:10 +0100407 print-parent - On function traces, display the calling (parent)
408 function as well as the function being traced.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400409
410 print-parent:
411 bash-4000 [01] 1477.606694: simple_strtoul <-strict_strtoul
412
413 noprint-parent:
414 bash-4000 [01] 1477.606694: simple_strtoul
415
416
Ingo Molnar57526742009-02-19 12:54:10 +0100417 sym-offset - Display not only the function name, but also the
418 offset in the function. For example, instead of
419 seeing just "ktime_get", you will see
420 "ktime_get+0xb/0x20".
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400421
422 sym-offset:
423 bash-4000 [01] 1477.606694: simple_strtoul+0x6/0xa0
424
Ingo Molnar57526742009-02-19 12:54:10 +0100425 sym-addr - this will also display the function address as well
426 as the function name.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400427
428 sym-addr:
429 bash-4000 [01] 1477.606694: simple_strtoul <c0339346>
430
431 verbose - This deals with the latency_trace file.
432
433 bash 4000 1 0 00000000 00010a95 [58127d26] 1720.415ms \
434 (+0.000ms): simple_strtoul (strict_strtoul)
435
Ingo Molnar57526742009-02-19 12:54:10 +0100436 raw - This will display raw numbers. This option is best for
437 use with user applications that can translate the raw
438 numbers better than having it done in the kernel.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400439
Ingo Molnar57526742009-02-19 12:54:10 +0100440 hex - Similar to raw, but the numbers will be in a hexadecimal
441 format.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400442
443 bin - This will print out the formats in raw binary.
444
445 block - TBD (needs update)
446
Ingo Molnar57526742009-02-19 12:54:10 +0100447 stacktrace - This is one of the options that changes the trace
448 itself. When a trace is recorded, so is the stack
449 of functions. This allows for back traces of
450 trace sites.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400451
Ingo Molnar57526742009-02-19 12:54:10 +0100452 userstacktrace - This option changes the trace. It records a
453 stacktrace of the current userspace thread.
Török Edwin02b67512008-11-22 13:28:47 +0200454
Ingo Molnar57526742009-02-19 12:54:10 +0100455 sym-userobj - when user stacktrace are enabled, look up which
456 object the address belongs to, and print a
457 relative address. This is especially useful when
458 ASLR is on, otherwise you don't get a chance to
459 resolve the address to object/file/line after
460 the app is no longer running
Török Edwinb54d3de2008-11-22 13:28:48 +0200461
Ingo Molnar57526742009-02-19 12:54:10 +0100462 The lookup is performed when you read
463 trace,trace_pipe,latency_trace. Example:
Török Edwinb54d3de2008-11-22 13:28:48 +0200464
465 a.out-1623 [000] 40874.465068: /root/a.out[+0x480] <-/root/a.out[+0
466x494] <- /root/a.out[+0x4a8] <- /lib/libc-2.7.so[+0x1e1a6]
467
Ingo Molnar57526742009-02-19 12:54:10 +0100468 sched-tree - trace all tasks that are on the runqueue, at
469 every scheduling event. Will add overhead if
470 there's a lot of tasks running at once.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400471
472
473sched_switch
474------------
475
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400476This tracer simply records schedule switches. Here is an example
Steven Rostedta41eeba2008-07-14 16:41:12 -0400477of how to use it.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400478
479 # echo sched_switch > /debug/tracing/current_tracer
480 # echo 1 > /debug/tracing/tracing_enabled
481 # sleep 1
482 # echo 0 > /debug/tracing/tracing_enabled
483 # cat /debug/tracing/trace
484
485# tracer: sched_switch
486#
487# TASK-PID CPU# TIMESTAMP FUNCTION
488# | | | | |
489 bash-3997 [01] 240.132281: 3997:120:R + 4055:120:R
490 bash-3997 [01] 240.132284: 3997:120:R ==> 4055:120:R
491 sleep-4055 [01] 240.132371: 4055:120:S ==> 3997:120:R
492 bash-3997 [01] 240.132454: 3997:120:R + 4055:120:S
493 bash-3997 [01] 240.132457: 3997:120:R ==> 4055:120:R
494 sleep-4055 [01] 240.132460: 4055:120:D ==> 3997:120:R
495 bash-3997 [01] 240.132463: 3997:120:R + 4055:120:D
496 bash-3997 [01] 240.132465: 3997:120:R ==> 4055:120:R
497 <idle>-0 [00] 240.132589: 0:140:R + 4:115:S
498 <idle>-0 [00] 240.132591: 0:140:R ==> 4:115:R
499 ksoftirqd/0-4 [00] 240.132595: 4:115:S ==> 0:140:R
500 <idle>-0 [00] 240.132598: 0:140:R + 4:115:S
501 <idle>-0 [00] 240.132599: 0:140:R ==> 4:115:R
502 ksoftirqd/0-4 [00] 240.132603: 4:115:S ==> 0:140:R
503 sleep-4055 [01] 240.133058: 4055:120:S ==> 3997:120:R
504 [...]
505
506
Ingo Molnar57526742009-02-19 12:54:10 +0100507As we have discussed previously about this format, the header
508shows the name of the trace and points to the options. The
509"FUNCTION" is a misnomer since here it represents the wake ups
510and context switches.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400511
Ingo Molnar57526742009-02-19 12:54:10 +0100512The sched_switch file only lists the wake ups (represented with
513'+') and context switches ('==>') with the previous task or
514current task first followed by the next task or task waking up.
515The format for both of these is PID:KERNEL-PRIO:TASK-STATE.
516Remember that the KERNEL-PRIO is the inverse of the actual
517priority with zero (0) being the highest priority and the nice
518values starting at 100 (nice -20). Below is a quick chart to map
519the kernel priority to user land priorities.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400520
521 Kernel priority: 0 to 99 ==> user RT priority 99 to 0
522 Kernel priority: 100 to 139 ==> user nice -20 to 19
523 Kernel priority: 140 ==> idle task priority
524
525The task states are:
526
527 R - running : wants to run, may not actually be running
528 S - sleep : process is waiting to be woken up (handles signals)
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400529 D - disk sleep (uninterruptible sleep) : process must be woken up
530 (ignores signals)
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400531 T - stopped : process suspended
532 t - traced : process is being traced (with something like gdb)
533 Z - zombie : process waiting to be cleaned up
534 X - unknown
535
536
537ftrace_enabled
538--------------
539
Ingo Molnar57526742009-02-19 12:54:10 +0100540The following tracers (listed below) give different output
541depending on whether or not the sysctl ftrace_enabled is set. To
542set ftrace_enabled, one can either use the sysctl function or
543set it via the proc file system interface.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400544
545 sysctl kernel.ftrace_enabled=1
546
547 or
548
549 echo 1 > /proc/sys/kernel/ftrace_enabled
550
Ingo Molnar57526742009-02-19 12:54:10 +0100551To disable ftrace_enabled simply replace the '1' with '0' in the
552above commands.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400553
Ingo Molnar57526742009-02-19 12:54:10 +0100554When ftrace_enabled is set the tracers will also record the
555functions that are within the trace. The descriptions of the
556tracers will also show an example with ftrace enabled.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400557
558
559irqsoff
560-------
561
562When interrupts are disabled, the CPU can not react to any other
563external event (besides NMIs and SMIs). This prevents the timer
Ingo Molnar57526742009-02-19 12:54:10 +0100564interrupt from triggering or the mouse interrupt from letting
565the kernel know of a new mouse event. The result is a latency
566with the reaction time.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400567
Ingo Molnar57526742009-02-19 12:54:10 +0100568The irqsoff tracer tracks the time for which interrupts are
569disabled. When a new maximum latency is hit, the tracer saves
570the trace leading up to that latency point so that every time a
571new maximum is reached, the old saved trace is discarded and the
572new trace is saved.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400573
Ingo Molnar57526742009-02-19 12:54:10 +0100574To reset the maximum, echo 0 into tracing_max_latency. Here is
575an example:
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400576
577 # echo irqsoff > /debug/tracing/current_tracer
578 # echo 0 > /debug/tracing/tracing_max_latency
579 # echo 1 > /debug/tracing/tracing_enabled
580 # ls -ltr
581 [...]
582 # echo 0 > /debug/tracing/tracing_enabled
583 # cat /debug/tracing/latency_trace
584# tracer: irqsoff
585#
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400586irqsoff latency trace v1.1.5 on 2.6.26
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400587--------------------------------------------------------------------
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400588 latency: 12 us, #3/3, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400589 -----------------
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400590 | task: bash-3730 (uid:0 nice:0 policy:0 rt_prio:0)
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400591 -----------------
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400592 => started at: sys_setpgid
593 => ended at: sys_setpgid
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400594
595# _------=> CPU#
596# / _-----=> irqs-off
597# | / _----=> need-resched
598# || / _---=> hardirq/softirq
599# ||| / _--=> preempt-depth
600# |||| /
601# ||||| delay
602# cmd pid ||||| time | caller
603# \ / ||||| \ | /
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400604 bash-3730 1d... 0us : _write_lock_irq (sys_setpgid)
605 bash-3730 1d..1 1us+: _write_unlock_irq (sys_setpgid)
606 bash-3730 1d..2 14us : trace_hardirqs_on (sys_setpgid)
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400607
608
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400609Here we see that that we had a latency of 12 microsecs (which is
Ingo Molnar57526742009-02-19 12:54:10 +0100610very good). The _write_lock_irq in sys_setpgid disabled
611interrupts. The difference between the 12 and the displayed
612timestamp 14us occurred because the clock was incremented
613between the time of recording the max latency and the time of
614recording the function that had that latency.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400615
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400616Note the above example had ftrace_enabled not set. If we set the
617ftrace_enabled, we get a much larger output:
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400618
619# tracer: irqsoff
620#
621irqsoff latency trace v1.1.5 on 2.6.26-rc8
622--------------------------------------------------------------------
623 latency: 50 us, #101/101, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
624 -----------------
625 | task: ls-4339 (uid:0 nice:0 policy:0 rt_prio:0)
626 -----------------
627 => started at: __alloc_pages_internal
628 => ended at: __alloc_pages_internal
629
630# _------=> CPU#
631# / _-----=> irqs-off
632# | / _----=> need-resched
633# || / _---=> hardirq/softirq
634# ||| / _--=> preempt-depth
635# |||| /
636# ||||| delay
637# cmd pid ||||| time | caller
638# \ / ||||| \ | /
639 ls-4339 0...1 0us+: get_page_from_freelist (__alloc_pages_internal)
640 ls-4339 0d..1 3us : rmqueue_bulk (get_page_from_freelist)
641 ls-4339 0d..1 3us : _spin_lock (rmqueue_bulk)
642 ls-4339 0d..1 4us : add_preempt_count (_spin_lock)
643 ls-4339 0d..2 4us : __rmqueue (rmqueue_bulk)
644 ls-4339 0d..2 5us : __rmqueue_smallest (__rmqueue)
645 ls-4339 0d..2 5us : __mod_zone_page_state (__rmqueue_smallest)
646 ls-4339 0d..2 6us : __rmqueue (rmqueue_bulk)
647 ls-4339 0d..2 6us : __rmqueue_smallest (__rmqueue)
648 ls-4339 0d..2 7us : __mod_zone_page_state (__rmqueue_smallest)
649 ls-4339 0d..2 7us : __rmqueue (rmqueue_bulk)
650 ls-4339 0d..2 8us : __rmqueue_smallest (__rmqueue)
651[...]
652 ls-4339 0d..2 46us : __rmqueue_smallest (__rmqueue)
653 ls-4339 0d..2 47us : __mod_zone_page_state (__rmqueue_smallest)
654 ls-4339 0d..2 47us : __rmqueue (rmqueue_bulk)
655 ls-4339 0d..2 48us : __rmqueue_smallest (__rmqueue)
656 ls-4339 0d..2 48us : __mod_zone_page_state (__rmqueue_smallest)
657 ls-4339 0d..2 49us : _spin_unlock (rmqueue_bulk)
658 ls-4339 0d..2 49us : sub_preempt_count (_spin_unlock)
659 ls-4339 0d..1 50us : get_page_from_freelist (__alloc_pages_internal)
660 ls-4339 0d..2 51us : trace_hardirqs_on (__alloc_pages_internal)
661
662
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400663
664Here we traced a 50 microsecond latency. But we also see all the
Ingo Molnar57526742009-02-19 12:54:10 +0100665functions that were called during that time. Note that by
666enabling function tracing, we incur an added overhead. This
667overhead may extend the latency times. But nevertheless, this
668trace has provided some very helpful debugging information.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400669
670
671preemptoff
672----------
673
Ingo Molnar57526742009-02-19 12:54:10 +0100674When preemption is disabled, we may be able to receive
675interrupts but the task cannot be preempted and a higher
676priority task must wait for preemption to be enabled again
677before it can preempt a lower priority task.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400678
Steven Rostedta41eeba2008-07-14 16:41:12 -0400679The preemptoff tracer traces the places that disable preemption.
Ingo Molnar57526742009-02-19 12:54:10 +0100680Like the irqsoff tracer, it records the maximum latency for
681which preemption was disabled. The control of preemptoff tracer
682is much like the irqsoff tracer.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400683
684 # echo preemptoff > /debug/tracing/current_tracer
685 # echo 0 > /debug/tracing/tracing_max_latency
686 # echo 1 > /debug/tracing/tracing_enabled
687 # ls -ltr
688 [...]
689 # echo 0 > /debug/tracing/tracing_enabled
690 # cat /debug/tracing/latency_trace
691# tracer: preemptoff
692#
693preemptoff latency trace v1.1.5 on 2.6.26-rc8
694--------------------------------------------------------------------
695 latency: 29 us, #3/3, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
696 -----------------
697 | task: sshd-4261 (uid:0 nice:0 policy:0 rt_prio:0)
698 -----------------
699 => started at: do_IRQ
700 => ended at: __do_softirq
701
702# _------=> CPU#
703# / _-----=> irqs-off
704# | / _----=> need-resched
705# || / _---=> hardirq/softirq
706# ||| / _--=> preempt-depth
707# |||| /
708# ||||| delay
709# cmd pid ||||| time | caller
710# \ / ||||| \ | /
711 sshd-4261 0d.h. 0us+: irq_enter (do_IRQ)
712 sshd-4261 0d.s. 29us : _local_bh_enable (__do_softirq)
713 sshd-4261 0d.s1 30us : trace_preempt_on (__do_softirq)
714
715
Ingo Molnar57526742009-02-19 12:54:10 +0100716This has some more changes. Preemption was disabled when an
717interrupt came in (notice the 'h'), and was enabled while doing
718a softirq. (notice the 's'). But we also see that interrupts
719have been disabled when entering the preempt off section and
720leaving it (the 'd'). We do not know if interrupts were enabled
721in the mean time.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400722
723# tracer: preemptoff
724#
725preemptoff latency trace v1.1.5 on 2.6.26-rc8
726--------------------------------------------------------------------
727 latency: 63 us, #87/87, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
728 -----------------
729 | task: sshd-4261 (uid:0 nice:0 policy:0 rt_prio:0)
730 -----------------
731 => started at: remove_wait_queue
732 => ended at: __do_softirq
733
734# _------=> CPU#
735# / _-----=> irqs-off
736# | / _----=> need-resched
737# || / _---=> hardirq/softirq
738# ||| / _--=> preempt-depth
739# |||| /
740# ||||| delay
741# cmd pid ||||| time | caller
742# \ / ||||| \ | /
743 sshd-4261 0d..1 0us : _spin_lock_irqsave (remove_wait_queue)
744 sshd-4261 0d..1 1us : _spin_unlock_irqrestore (remove_wait_queue)
745 sshd-4261 0d..1 2us : do_IRQ (common_interrupt)
746 sshd-4261 0d..1 2us : irq_enter (do_IRQ)
747 sshd-4261 0d..1 2us : idle_cpu (irq_enter)
748 sshd-4261 0d..1 3us : add_preempt_count (irq_enter)
749 sshd-4261 0d.h1 3us : idle_cpu (irq_enter)
750 sshd-4261 0d.h. 4us : handle_fasteoi_irq (do_IRQ)
751[...]
752 sshd-4261 0d.h. 12us : add_preempt_count (_spin_lock)
753 sshd-4261 0d.h1 12us : ack_ioapic_quirk_irq (handle_fasteoi_irq)
754 sshd-4261 0d.h1 13us : move_native_irq (ack_ioapic_quirk_irq)
755 sshd-4261 0d.h1 13us : _spin_unlock (handle_fasteoi_irq)
756 sshd-4261 0d.h1 14us : sub_preempt_count (_spin_unlock)
757 sshd-4261 0d.h1 14us : irq_exit (do_IRQ)
758 sshd-4261 0d.h1 15us : sub_preempt_count (irq_exit)
759 sshd-4261 0d..2 15us : do_softirq (irq_exit)
760 sshd-4261 0d... 15us : __do_softirq (do_softirq)
761 sshd-4261 0d... 16us : __local_bh_disable (__do_softirq)
762 sshd-4261 0d... 16us+: add_preempt_count (__local_bh_disable)
763 sshd-4261 0d.s4 20us : add_preempt_count (__local_bh_disable)
764 sshd-4261 0d.s4 21us : sub_preempt_count (local_bh_enable)
765 sshd-4261 0d.s5 21us : sub_preempt_count (local_bh_enable)
766[...]
767 sshd-4261 0d.s6 41us : add_preempt_count (__local_bh_disable)
768 sshd-4261 0d.s6 42us : sub_preempt_count (local_bh_enable)
769 sshd-4261 0d.s7 42us : sub_preempt_count (local_bh_enable)
770 sshd-4261 0d.s5 43us : add_preempt_count (__local_bh_disable)
771 sshd-4261 0d.s5 43us : sub_preempt_count (local_bh_enable_ip)
772 sshd-4261 0d.s6 44us : sub_preempt_count (local_bh_enable_ip)
773 sshd-4261 0d.s5 44us : add_preempt_count (__local_bh_disable)
774 sshd-4261 0d.s5 45us : sub_preempt_count (local_bh_enable)
775[...]
776 sshd-4261 0d.s. 63us : _local_bh_enable (__do_softirq)
777 sshd-4261 0d.s1 64us : trace_preempt_on (__do_softirq)
778
779
Ingo Molnar57526742009-02-19 12:54:10 +0100780The above is an example of the preemptoff trace with
781ftrace_enabled set. Here we see that interrupts were disabled
782the entire time. The irq_enter code lets us know that we entered
783an interrupt 'h'. Before that, the functions being traced still
784show that it is not in an interrupt, but we can see from the
785functions themselves that this is not the case.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400786
Ingo Molnar57526742009-02-19 12:54:10 +0100787Notice that __do_softirq when called does not have a
788preempt_count. It may seem that we missed a preempt enabling.
789What really happened is that the preempt count is held on the
790thread's stack and we switched to the softirq stack (4K stacks
791in effect). The code does not copy the preempt count, but
792because interrupts are disabled, we do not need to worry about
793it. Having a tracer like this is good for letting people know
794what really happens inside the kernel.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400795
796
797preemptirqsoff
798--------------
799
Ingo Molnar57526742009-02-19 12:54:10 +0100800Knowing the locations that have interrupts disabled or
801preemption disabled for the longest times is helpful. But
802sometimes we would like to know when either preemption and/or
803interrupts are disabled.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400804
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400805Consider the following code:
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400806
807 local_irq_disable();
808 call_function_with_irqs_off();
809 preempt_disable();
810 call_function_with_irqs_and_preemption_off();
811 local_irq_enable();
812 call_function_with_preemption_off();
813 preempt_enable();
814
815The irqsoff tracer will record the total length of
816call_function_with_irqs_off() and
817call_function_with_irqs_and_preemption_off().
818
819The preemptoff tracer will record the total length of
820call_function_with_irqs_and_preemption_off() and
821call_function_with_preemption_off().
822
Ingo Molnar57526742009-02-19 12:54:10 +0100823But neither will trace the time that interrupts and/or
824preemption is disabled. This total time is the time that we can
825not schedule. To record this time, use the preemptirqsoff
826tracer.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400827
Ingo Molnar57526742009-02-19 12:54:10 +0100828Again, using this trace is much like the irqsoff and preemptoff
829tracers.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400830
Steven Rostedta41eeba2008-07-14 16:41:12 -0400831 # echo preemptirqsoff > /debug/tracing/current_tracer
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400832 # echo 0 > /debug/tracing/tracing_max_latency
833 # echo 1 > /debug/tracing/tracing_enabled
834 # ls -ltr
835 [...]
836 # echo 0 > /debug/tracing/tracing_enabled
837 # cat /debug/tracing/latency_trace
838# tracer: preemptirqsoff
839#
840preemptirqsoff latency trace v1.1.5 on 2.6.26-rc8
841--------------------------------------------------------------------
842 latency: 293 us, #3/3, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
843 -----------------
844 | task: ls-4860 (uid:0 nice:0 policy:0 rt_prio:0)
845 -----------------
846 => started at: apic_timer_interrupt
847 => ended at: __do_softirq
848
849# _------=> CPU#
850# / _-----=> irqs-off
851# | / _----=> need-resched
852# || / _---=> hardirq/softirq
853# ||| / _--=> preempt-depth
854# |||| /
855# ||||| delay
856# cmd pid ||||| time | caller
857# \ / ||||| \ | /
858 ls-4860 0d... 0us!: trace_hardirqs_off_thunk (apic_timer_interrupt)
859 ls-4860 0d.s. 294us : _local_bh_enable (__do_softirq)
860 ls-4860 0d.s1 294us : trace_preempt_on (__do_softirq)
861
862
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400863
864The trace_hardirqs_off_thunk is called from assembly on x86 when
Ingo Molnar57526742009-02-19 12:54:10 +0100865interrupts are disabled in the assembly code. Without the
866function tracing, we do not know if interrupts were enabled
867within the preemption points. We do see that it started with
868preemption enabled.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400869
870Here is a trace with ftrace_enabled set:
871
872
873# tracer: preemptirqsoff
874#
875preemptirqsoff latency trace v1.1.5 on 2.6.26-rc8
876--------------------------------------------------------------------
877 latency: 105 us, #183/183, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
878 -----------------
879 | task: sshd-4261 (uid:0 nice:0 policy:0 rt_prio:0)
880 -----------------
881 => started at: write_chan
882 => ended at: __do_softirq
883
884# _------=> CPU#
885# / _-----=> irqs-off
886# | / _----=> need-resched
887# || / _---=> hardirq/softirq
888# ||| / _--=> preempt-depth
889# |||| /
890# ||||| delay
891# cmd pid ||||| time | caller
892# \ / ||||| \ | /
893 ls-4473 0.N.. 0us : preempt_schedule (write_chan)
894 ls-4473 0dN.1 1us : _spin_lock (schedule)
895 ls-4473 0dN.1 2us : add_preempt_count (_spin_lock)
896 ls-4473 0d..2 2us : put_prev_task_fair (schedule)
897[...]
898 ls-4473 0d..2 13us : set_normalized_timespec (ktime_get_ts)
899 ls-4473 0d..2 13us : __switch_to (schedule)
900 sshd-4261 0d..2 14us : finish_task_switch (schedule)
901 sshd-4261 0d..2 14us : _spin_unlock_irq (finish_task_switch)
902 sshd-4261 0d..1 15us : add_preempt_count (_spin_lock_irqsave)
903 sshd-4261 0d..2 16us : _spin_unlock_irqrestore (hrtick_set)
904 sshd-4261 0d..2 16us : do_IRQ (common_interrupt)
905 sshd-4261 0d..2 17us : irq_enter (do_IRQ)
906 sshd-4261 0d..2 17us : idle_cpu (irq_enter)
907 sshd-4261 0d..2 18us : add_preempt_count (irq_enter)
908 sshd-4261 0d.h2 18us : idle_cpu (irq_enter)
909 sshd-4261 0d.h. 18us : handle_fasteoi_irq (do_IRQ)
910 sshd-4261 0d.h. 19us : _spin_lock (handle_fasteoi_irq)
911 sshd-4261 0d.h. 19us : add_preempt_count (_spin_lock)
912 sshd-4261 0d.h1 20us : _spin_unlock (handle_fasteoi_irq)
913 sshd-4261 0d.h1 20us : sub_preempt_count (_spin_unlock)
914[...]
915 sshd-4261 0d.h1 28us : _spin_unlock (handle_fasteoi_irq)
916 sshd-4261 0d.h1 29us : sub_preempt_count (_spin_unlock)
917 sshd-4261 0d.h2 29us : irq_exit (do_IRQ)
918 sshd-4261 0d.h2 29us : sub_preempt_count (irq_exit)
919 sshd-4261 0d..3 30us : do_softirq (irq_exit)
920 sshd-4261 0d... 30us : __do_softirq (do_softirq)
921 sshd-4261 0d... 31us : __local_bh_disable (__do_softirq)
922 sshd-4261 0d... 31us+: add_preempt_count (__local_bh_disable)
923 sshd-4261 0d.s4 34us : add_preempt_count (__local_bh_disable)
924[...]
925 sshd-4261 0d.s3 43us : sub_preempt_count (local_bh_enable_ip)
926 sshd-4261 0d.s4 44us : sub_preempt_count (local_bh_enable_ip)
927 sshd-4261 0d.s3 44us : smp_apic_timer_interrupt (apic_timer_interrupt)
928 sshd-4261 0d.s3 45us : irq_enter (smp_apic_timer_interrupt)
929 sshd-4261 0d.s3 45us : idle_cpu (irq_enter)
930 sshd-4261 0d.s3 46us : add_preempt_count (irq_enter)
931 sshd-4261 0d.H3 46us : idle_cpu (irq_enter)
932 sshd-4261 0d.H3 47us : hrtimer_interrupt (smp_apic_timer_interrupt)
933 sshd-4261 0d.H3 47us : ktime_get (hrtimer_interrupt)
934[...]
935 sshd-4261 0d.H3 81us : tick_program_event (hrtimer_interrupt)
936 sshd-4261 0d.H3 82us : ktime_get (tick_program_event)
937 sshd-4261 0d.H3 82us : ktime_get_ts (ktime_get)
938 sshd-4261 0d.H3 83us : getnstimeofday (ktime_get_ts)
939 sshd-4261 0d.H3 83us : set_normalized_timespec (ktime_get_ts)
940 sshd-4261 0d.H3 84us : clockevents_program_event (tick_program_event)
941 sshd-4261 0d.H3 84us : lapic_next_event (clockevents_program_event)
942 sshd-4261 0d.H3 85us : irq_exit (smp_apic_timer_interrupt)
943 sshd-4261 0d.H3 85us : sub_preempt_count (irq_exit)
944 sshd-4261 0d.s4 86us : sub_preempt_count (irq_exit)
945 sshd-4261 0d.s3 86us : add_preempt_count (__local_bh_disable)
946[...]
947 sshd-4261 0d.s1 98us : sub_preempt_count (net_rx_action)
948 sshd-4261 0d.s. 99us : add_preempt_count (_spin_lock_irq)
949 sshd-4261 0d.s1 99us+: _spin_unlock_irq (run_timer_softirq)
950 sshd-4261 0d.s. 104us : _local_bh_enable (__do_softirq)
951 sshd-4261 0d.s. 104us : sub_preempt_count (_local_bh_enable)
952 sshd-4261 0d.s. 105us : _local_bh_enable (__do_softirq)
953 sshd-4261 0d.s1 105us : trace_preempt_on (__do_softirq)
954
955
Ingo Molnar57526742009-02-19 12:54:10 +0100956This is a very interesting trace. It started with the preemption
957of the ls task. We see that the task had the "need_resched" bit
958set via the 'N' in the trace. Interrupts were disabled before
959the spin_lock at the beginning of the trace. We see that a
960schedule took place to run sshd. When the interrupts were
961enabled, we took an interrupt. On return from the interrupt
962handler, the softirq ran. We took another interrupt while
963running the softirq as we see from the capital 'H'.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400964
965
966wakeup
967------
968
Ingo Molnar57526742009-02-19 12:54:10 +0100969In a Real-Time environment it is very important to know the
970wakeup time it takes for the highest priority task that is woken
971up to the time that it executes. This is also known as "schedule
972latency". I stress the point that this is about RT tasks. It is
973also important to know the scheduling latency of non-RT tasks,
974but the average schedule latency is better for non-RT tasks.
975Tools like LatencyTop are more appropriate for such
976measurements.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400977
Steven Rostedta41eeba2008-07-14 16:41:12 -0400978Real-Time environments are interested in the worst case latency.
Ingo Molnar57526742009-02-19 12:54:10 +0100979That is the longest latency it takes for something to happen,
980and not the average. We can have a very fast scheduler that may
981only have a large latency once in a while, but that would not
982work well with Real-Time tasks. The wakeup tracer was designed
983to record the worst case wakeups of RT tasks. Non-RT tasks are
984not recorded because the tracer only records one worst case and
985tracing non-RT tasks that are unpredictable will overwrite the
986worst case latency of RT tasks.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400987
Ingo Molnar57526742009-02-19 12:54:10 +0100988Since this tracer only deals with RT tasks, we will run this
989slightly differently than we did with the previous tracers.
990Instead of performing an 'ls', we will run 'sleep 1' under
991'chrt' which changes the priority of the task.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400992
993 # echo wakeup > /debug/tracing/current_tracer
994 # echo 0 > /debug/tracing/tracing_max_latency
995 # echo 1 > /debug/tracing/tracing_enabled
996 # chrt -f 5 sleep 1
997 # echo 0 > /debug/tracing/tracing_enabled
998 # cat /debug/tracing/latency_trace
999# tracer: wakeup
1000#
1001wakeup latency trace v1.1.5 on 2.6.26-rc8
1002--------------------------------------------------------------------
1003 latency: 4 us, #2/2, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
1004 -----------------
1005 | task: sleep-4901 (uid:0 nice:0 policy:1 rt_prio:5)
1006 -----------------
1007
1008# _------=> CPU#
1009# / _-----=> irqs-off
1010# | / _----=> need-resched
1011# || / _---=> hardirq/softirq
1012# ||| / _--=> preempt-depth
1013# |||| /
1014# ||||| delay
1015# cmd pid ||||| time | caller
1016# \ / ||||| \ | /
1017 <idle>-0 1d.h4 0us+: try_to_wake_up (wake_up_process)
1018 <idle>-0 1d..4 4us : schedule (cpu_idle)
1019
1020
Ingo Molnar57526742009-02-19 12:54:10 +01001021Running this on an idle system, we see that it only took 4
1022microseconds to perform the task switch. Note, since the trace
1023marker in the schedule is before the actual "switch", we stop
1024the tracing when the recorded task is about to schedule in. This
1025may change if we add a new marker at the end of the scheduler.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001026
Ingo Molnar57526742009-02-19 12:54:10 +01001027Notice that the recorded task is 'sleep' with the PID of 4901
1028and it has an rt_prio of 5. This priority is user-space priority
1029and not the internal kernel priority. The policy is 1 for
1030SCHED_FIFO and 2 for SCHED_RR.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001031
1032Doing the same with chrt -r 5 and ftrace_enabled set.
1033
1034# tracer: wakeup
1035#
1036wakeup latency trace v1.1.5 on 2.6.26-rc8
1037--------------------------------------------------------------------
1038 latency: 50 us, #60/60, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
1039 -----------------
1040 | task: sleep-4068 (uid:0 nice:0 policy:2 rt_prio:5)
1041 -----------------
1042
1043# _------=> CPU#
1044# / _-----=> irqs-off
1045# | / _----=> need-resched
1046# || / _---=> hardirq/softirq
1047# ||| / _--=> preempt-depth
1048# |||| /
1049# ||||| delay
1050# cmd pid ||||| time | caller
1051# \ / ||||| \ | /
1052ksoftirq-7 1d.H3 0us : try_to_wake_up (wake_up_process)
1053ksoftirq-7 1d.H4 1us : sub_preempt_count (marker_probe_cb)
1054ksoftirq-7 1d.H3 2us : check_preempt_wakeup (try_to_wake_up)
1055ksoftirq-7 1d.H3 3us : update_curr (check_preempt_wakeup)
1056ksoftirq-7 1d.H3 4us : calc_delta_mine (update_curr)
1057ksoftirq-7 1d.H3 5us : __resched_task (check_preempt_wakeup)
1058ksoftirq-7 1d.H3 6us : task_wake_up_rt (try_to_wake_up)
1059ksoftirq-7 1d.H3 7us : _spin_unlock_irqrestore (try_to_wake_up)
1060[...]
1061ksoftirq-7 1d.H2 17us : irq_exit (smp_apic_timer_interrupt)
1062ksoftirq-7 1d.H2 18us : sub_preempt_count (irq_exit)
1063ksoftirq-7 1d.s3 19us : sub_preempt_count (irq_exit)
1064ksoftirq-7 1..s2 20us : rcu_process_callbacks (__do_softirq)
1065[...]
1066ksoftirq-7 1..s2 26us : __rcu_process_callbacks (rcu_process_callbacks)
1067ksoftirq-7 1d.s2 27us : _local_bh_enable (__do_softirq)
1068ksoftirq-7 1d.s2 28us : sub_preempt_count (_local_bh_enable)
1069ksoftirq-7 1.N.3 29us : sub_preempt_count (ksoftirqd)
1070ksoftirq-7 1.N.2 30us : _cond_resched (ksoftirqd)
1071ksoftirq-7 1.N.2 31us : __cond_resched (_cond_resched)
1072ksoftirq-7 1.N.2 32us : add_preempt_count (__cond_resched)
1073ksoftirq-7 1.N.2 33us : schedule (__cond_resched)
1074ksoftirq-7 1.N.2 33us : add_preempt_count (schedule)
1075ksoftirq-7 1.N.3 34us : hrtick_clear (schedule)
1076ksoftirq-7 1dN.3 35us : _spin_lock (schedule)
1077ksoftirq-7 1dN.3 36us : add_preempt_count (_spin_lock)
1078ksoftirq-7 1d..4 37us : put_prev_task_fair (schedule)
1079ksoftirq-7 1d..4 38us : update_curr (put_prev_task_fair)
1080[...]
1081ksoftirq-7 1d..5 47us : _spin_trylock (tracing_record_cmdline)
1082ksoftirq-7 1d..5 48us : add_preempt_count (_spin_trylock)
1083ksoftirq-7 1d..6 49us : _spin_unlock (tracing_record_cmdline)
1084ksoftirq-7 1d..6 49us : sub_preempt_count (_spin_unlock)
1085ksoftirq-7 1d..4 50us : schedule (__cond_resched)
1086
Ingo Molnar57526742009-02-19 12:54:10 +01001087The interrupt went off while running ksoftirqd. This task runs
1088at SCHED_OTHER. Why did not we see the 'N' set early? This may
1089be a harmless bug with x86_32 and 4K stacks. On x86_32 with 4K
1090stacks configured, the interrupt and softirq run with their own
1091stack. Some information is held on the top of the task's stack
1092(need_resched and preempt_count are both stored there). The
1093setting of the NEED_RESCHED bit is done directly to the task's
1094stack, but the reading of the NEED_RESCHED is done by looking at
1095the current stack, which in this case is the stack for the hard
1096interrupt. This hides the fact that NEED_RESCHED has been set.
1097We do not see the 'N' until we switch back to the task's
Steven Rostedta41eeba2008-07-14 16:41:12 -04001098assigned stack.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001099
Steven Rostedt9b803c02008-11-03 15:15:08 -05001100function
1101--------
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001102
Steven Rostedt9b803c02008-11-03 15:15:08 -05001103This tracer is the function tracer. Enabling the function tracer
Ingo Molnar57526742009-02-19 12:54:10 +01001104can be done from the debug file system. Make sure the
1105ftrace_enabled is set; otherwise this tracer is a nop.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001106
1107 # sysctl kernel.ftrace_enabled=1
Steven Rostedt9b803c02008-11-03 15:15:08 -05001108 # echo function > /debug/tracing/current_tracer
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001109 # echo 1 > /debug/tracing/tracing_enabled
1110 # usleep 1
1111 # echo 0 > /debug/tracing/tracing_enabled
1112 # cat /debug/tracing/trace
Steven Rostedt9b803c02008-11-03 15:15:08 -05001113# tracer: function
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001114#
1115# TASK-PID CPU# TIMESTAMP FUNCTION
1116# | | | | |
1117 bash-4003 [00] 123.638713: finish_task_switch <-schedule
1118 bash-4003 [00] 123.638714: _spin_unlock_irq <-finish_task_switch
1119 bash-4003 [00] 123.638714: sub_preempt_count <-_spin_unlock_irq
1120 bash-4003 [00] 123.638715: hrtick_set <-schedule
1121 bash-4003 [00] 123.638715: _spin_lock_irqsave <-hrtick_set
1122 bash-4003 [00] 123.638716: add_preempt_count <-_spin_lock_irqsave
1123 bash-4003 [00] 123.638716: _spin_unlock_irqrestore <-hrtick_set
1124 bash-4003 [00] 123.638717: sub_preempt_count <-_spin_unlock_irqrestore
1125 bash-4003 [00] 123.638717: hrtick_clear <-hrtick_set
1126 bash-4003 [00] 123.638718: sub_preempt_count <-schedule
1127 bash-4003 [00] 123.638718: sub_preempt_count <-preempt_schedule
1128 bash-4003 [00] 123.638719: wait_for_completion <-__stop_machine_run
1129 bash-4003 [00] 123.638719: wait_for_common <-wait_for_completion
1130 bash-4003 [00] 123.638720: _spin_lock_irq <-wait_for_common
1131 bash-4003 [00] 123.638720: add_preempt_count <-_spin_lock_irq
1132[...]
1133
1134
Ingo Molnar57526742009-02-19 12:54:10 +01001135Note: function tracer uses ring buffers to store the above
1136entries. The newest data may overwrite the oldest data.
1137Sometimes using echo to stop the trace is not sufficient because
1138the tracing could have overwritten the data that you wanted to
1139record. For this reason, it is sometimes better to disable
1140tracing directly from a program. This allows you to stop the
1141tracing at the point that you hit the part that you are
1142interested in. To disable the tracing directly from a C program,
1143something like following code snippet can be used:
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001144
1145int trace_fd;
1146[...]
1147int main(int argc, char *argv[]) {
1148 [...]
1149 trace_fd = open("/debug/tracing/tracing_enabled", O_WRONLY);
1150 [...]
1151 if (condition_hit()) {
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001152 write(trace_fd, "0", 1);
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001153 }
1154 [...]
1155}
1156
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001157Note: Here we hard coded the path name. The debugfs mount is not
Ingo Molnar57526742009-02-19 12:54:10 +01001158guaranteed to be at /debug (and is more commonly at
1159/sys/kernel/debug). For simple one time traces, the above is
1160sufficent. For anything else, a search through /proc/mounts may
1161be needed to find where the debugfs file-system is mounted.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001162
Steven Rostedtdf4fc312008-11-26 00:16:23 -05001163
1164Single thread tracing
1165---------------------
1166
1167By writing into /debug/tracing/set_ftrace_pid you can trace a
1168single thread. For example:
1169
1170# cat /debug/tracing/set_ftrace_pid
1171no pid
1172# echo 3111 > /debug/tracing/set_ftrace_pid
1173# cat /debug/tracing/set_ftrace_pid
11743111
1175# echo function > /debug/tracing/current_tracer
1176# cat /debug/tracing/trace | head
1177 # tracer: function
1178 #
1179 # TASK-PID CPU# TIMESTAMP FUNCTION
1180 # | | | | |
1181 yum-updatesd-3111 [003] 1637.254676: finish_task_switch <-thread_return
1182 yum-updatesd-3111 [003] 1637.254681: hrtimer_cancel <-schedule_hrtimeout_range
1183 yum-updatesd-3111 [003] 1637.254682: hrtimer_try_to_cancel <-hrtimer_cancel
1184 yum-updatesd-3111 [003] 1637.254683: lock_hrtimer_base <-hrtimer_try_to_cancel
1185 yum-updatesd-3111 [003] 1637.254685: fget_light <-do_sys_poll
1186 yum-updatesd-3111 [003] 1637.254686: pipe_poll <-do_sys_poll
1187# echo -1 > /debug/tracing/set_ftrace_pid
1188# cat /debug/tracing/trace |head
1189 # tracer: function
1190 #
1191 # TASK-PID CPU# TIMESTAMP FUNCTION
1192 # | | | | |
1193 ##### CPU 3 buffer started ####
1194 yum-updatesd-3111 [003] 1701.957688: free_poll_entry <-poll_freewait
1195 yum-updatesd-3111 [003] 1701.957689: remove_wait_queue <-free_poll_entry
1196 yum-updatesd-3111 [003] 1701.957691: fput <-free_poll_entry
1197 yum-updatesd-3111 [003] 1701.957692: audit_syscall_exit <-sysret_audit
1198 yum-updatesd-3111 [003] 1701.957693: path_put <-audit_syscall_exit
1199
1200If you want to trace a function when executing, you could use
1201something like this simple program:
1202
1203#include <stdio.h>
1204#include <stdlib.h>
1205#include <sys/types.h>
1206#include <sys/stat.h>
1207#include <fcntl.h>
1208#include <unistd.h>
1209
1210int main (int argc, char **argv)
1211{
1212 if (argc < 1)
1213 exit(-1);
1214
1215 if (fork() > 0) {
1216 int fd, ffd;
1217 char line[64];
1218 int s;
1219
1220 ffd = open("/debug/tracing/current_tracer", O_WRONLY);
1221 if (ffd < 0)
1222 exit(-1);
1223 write(ffd, "nop", 3);
1224
1225 fd = open("/debug/tracing/set_ftrace_pid", O_WRONLY);
1226 s = sprintf(line, "%d\n", getpid());
1227 write(fd, line, s);
1228
1229 write(ffd, "function", 8);
1230
1231 close(fd);
1232 close(ffd);
1233
1234 execvp(argv[1], argv+1);
1235 }
1236
1237 return 0;
1238}
1239
Markus Metzgere2ea5392009-01-19 10:35:58 +01001240
1241hw-branch-tracer (x86 only)
1242---------------------------
1243
1244This tracer uses the x86 last branch tracing hardware feature to
1245collect a branch trace on all cpus with relatively low overhead.
1246
1247The tracer uses a fixed-size circular buffer per cpu and only
1248traces ring 0 branches. The trace file dumps that buffer in the
1249following format:
1250
1251# tracer: hw-branch-tracer
1252#
1253# CPU# TO <- FROM
1254 0 scheduler_tick+0xb5/0x1bf <- task_tick_idle+0x5/0x6
1255 2 run_posix_cpu_timers+0x2b/0x72a <- run_posix_cpu_timers+0x25/0x72a
1256 0 scheduler_tick+0x139/0x1bf <- scheduler_tick+0xed/0x1bf
1257 0 scheduler_tick+0x17c/0x1bf <- scheduler_tick+0x148/0x1bf
1258 2 run_posix_cpu_timers+0x9e/0x72a <- run_posix_cpu_timers+0x5e/0x72a
1259 0 scheduler_tick+0x1b6/0x1bf <- scheduler_tick+0x1aa/0x1bf
1260
1261
Ingo Molnar57526742009-02-19 12:54:10 +01001262The tracer may be used to dump the trace for the oops'ing cpu on
1263a kernel oops into the system log. To enable this,
1264ftrace_dump_on_oops must be set. To set ftrace_dump_on_oops, one
1265can either use the sysctl function or set it via the proc system
1266interface.
Markus Metzgere2ea5392009-01-19 10:35:58 +01001267
1268 sysctl kernel.ftrace_dump_on_oops=1
1269
1270or
1271
1272 echo 1 > /proc/sys/kernel/ftrace_dump_on_oops
1273
1274
Ingo Molnar57526742009-02-19 12:54:10 +01001275Here's an example of such a dump after a null pointer
1276dereference in a kernel module:
Markus Metzgere2ea5392009-01-19 10:35:58 +01001277
1278[57848.105921] BUG: unable to handle kernel NULL pointer dereference at 0000000000000000
1279[57848.106019] IP: [<ffffffffa0000006>] open+0x6/0x14 [oops]
1280[57848.106019] PGD 2354e9067 PUD 2375e7067 PMD 0
1281[57848.106019] Oops: 0002 [#1] SMP
1282[57848.106019] last sysfs file: /sys/devices/pci0000:00/0000:00:1e.0/0000:20:05.0/local_cpus
1283[57848.106019] Dumping ftrace buffer:
1284[57848.106019] ---------------------------------
1285[...]
1286[57848.106019] 0 chrdev_open+0xe6/0x165 <- cdev_put+0x23/0x24
1287[57848.106019] 0 chrdev_open+0x117/0x165 <- chrdev_open+0xfa/0x165
1288[57848.106019] 0 chrdev_open+0x120/0x165 <- chrdev_open+0x11c/0x165
1289[57848.106019] 0 chrdev_open+0x134/0x165 <- chrdev_open+0x12b/0x165
1290[57848.106019] 0 open+0x0/0x14 [oops] <- chrdev_open+0x144/0x165
1291[57848.106019] 0 page_fault+0x0/0x30 <- open+0x6/0x14 [oops]
1292[57848.106019] 0 error_entry+0x0/0x5b <- page_fault+0x4/0x30
1293[57848.106019] 0 error_kernelspace+0x0/0x31 <- error_entry+0x59/0x5b
1294[57848.106019] 0 error_sti+0x0/0x1 <- error_kernelspace+0x2d/0x31
1295[57848.106019] 0 page_fault+0x9/0x30 <- error_sti+0x0/0x1
1296[57848.106019] 0 do_page_fault+0x0/0x881 <- page_fault+0x1a/0x30
1297[...]
1298[57848.106019] 0 do_page_fault+0x66b/0x881 <- is_prefetch+0x1ee/0x1f2
1299[57848.106019] 0 do_page_fault+0x6e0/0x881 <- do_page_fault+0x67a/0x881
1300[57848.106019] 0 oops_begin+0x0/0x96 <- do_page_fault+0x6e0/0x881
1301[57848.106019] 0 trace_hw_branch_oops+0x0/0x2d <- oops_begin+0x9/0x96
1302[...]
1303[57848.106019] 0 ds_suspend_bts+0x2a/0xe3 <- ds_suspend_bts+0x1a/0xe3
1304[57848.106019] ---------------------------------
1305[57848.106019] CPU 0
1306[57848.106019] Modules linked in: oops
1307[57848.106019] Pid: 5542, comm: cat Tainted: G W 2.6.28 #23
1308[57848.106019] RIP: 0010:[<ffffffffa0000006>] [<ffffffffa0000006>] open+0x6/0x14 [oops]
1309[57848.106019] RSP: 0018:ffff880235457d48 EFLAGS: 00010246
1310[...]
1311
1312
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001313function graph tracer
1314---------------------------
1315
Ingo Molnar57526742009-02-19 12:54:10 +01001316This tracer is similar to the function tracer except that it
1317probes a function on its entry and its exit. This is done by
1318using a dynamically allocated stack of return addresses in each
1319task_struct. On function entry the tracer overwrites the return
1320address of each function traced to set a custom probe. Thus the
1321original return address is stored on the stack of return address
1322in the task_struct.
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001323
Ingo Molnar57526742009-02-19 12:54:10 +01001324Probing on both ends of a function leads to special features
1325such as:
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001326
Ingo Molnar57526742009-02-19 12:54:10 +01001327- measure of a function's time execution
1328- having a reliable call stack to draw function calls graph
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001329
1330This tracer is useful in several situations:
1331
Ingo Molnar57526742009-02-19 12:54:10 +01001332- you want to find the reason of a strange kernel behavior and
1333 need to see what happens in detail on any areas (or specific
1334 ones).
1335
1336- you are experiencing weird latencies but it's difficult to
1337 find its origin.
1338
1339- you want to find quickly which path is taken by a specific
1340 function
1341
1342- you just want to peek inside a working kernel and want to see
1343 what happens there.
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001344
1345# tracer: function_graph
1346#
1347# CPU DURATION FUNCTION CALLS
1348# | | | | | | |
1349
1350 0) | sys_open() {
1351 0) | do_sys_open() {
1352 0) | getname() {
1353 0) | kmem_cache_alloc() {
1354 0) 1.382 us | __might_sleep();
1355 0) 2.478 us | }
1356 0) | strncpy_from_user() {
1357 0) | might_fault() {
1358 0) 1.389 us | __might_sleep();
1359 0) 2.553 us | }
1360 0) 3.807 us | }
1361 0) 7.876 us | }
1362 0) | alloc_fd() {
1363 0) 0.668 us | _spin_lock();
1364 0) 0.570 us | expand_files();
1365 0) 0.586 us | _spin_unlock();
1366
1367
Ingo Molnar57526742009-02-19 12:54:10 +01001368There are several columns that can be dynamically
1369enabled/disabled. You can use every combination of options you
1370want, depending on your needs.
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001371
Ingo Molnar57526742009-02-19 12:54:10 +01001372- The cpu number on which the function executed is default
1373 enabled. It is sometimes better to only trace one cpu (see
1374 tracing_cpu_mask file) or you might sometimes see unordered
1375 function calls while cpu tracing switch.
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001376
1377 hide: echo nofuncgraph-cpu > /debug/tracing/trace_options
1378 show: echo funcgraph-cpu > /debug/tracing/trace_options
1379
Ingo Molnar57526742009-02-19 12:54:10 +01001380- The duration (function's time of execution) is displayed on
1381 the closing bracket line of a function or on the same line
1382 than the current function in case of a leaf one. It is default
1383 enabled.
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001384
1385 hide: echo nofuncgraph-duration > /debug/tracing/trace_options
1386 show: echo funcgraph-duration > /debug/tracing/trace_options
1387
Ingo Molnar57526742009-02-19 12:54:10 +01001388- The overhead field precedes the duration field in case of
1389 reached duration thresholds.
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001390
1391 hide: echo nofuncgraph-overhead > /debug/tracing/trace_options
1392 show: echo funcgraph-overhead > /debug/tracing/trace_options
1393 depends on: funcgraph-duration
1394
1395 ie:
1396
1397 0) | up_write() {
1398 0) 0.646 us | _spin_lock_irqsave();
1399 0) 0.684 us | _spin_unlock_irqrestore();
1400 0) 3.123 us | }
1401 0) 0.548 us | fput();
1402 0) + 58.628 us | }
1403
1404 [...]
1405
1406 0) | putname() {
1407 0) | kmem_cache_free() {
1408 0) 0.518 us | __phys_addr();
1409 0) 1.757 us | }
1410 0) 2.861 us | }
1411 0) ! 115.305 us | }
1412 0) ! 116.402 us | }
1413
1414 + means that the function exceeded 10 usecs.
1415 ! means that the function exceeded 100 usecs.
1416
1417
Ingo Molnar57526742009-02-19 12:54:10 +01001418- The task/pid field displays the thread cmdline and pid which
1419 executed the function. It is default disabled.
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001420
1421 hide: echo nofuncgraph-proc > /debug/tracing/trace_options
1422 show: echo funcgraph-proc > /debug/tracing/trace_options
1423
1424 ie:
1425
1426 # tracer: function_graph
1427 #
1428 # CPU TASK/PID DURATION FUNCTION CALLS
1429 # | | | | | | | | |
1430 0) sh-4802 | | d_free() {
1431 0) sh-4802 | | call_rcu() {
1432 0) sh-4802 | | __call_rcu() {
1433 0) sh-4802 | 0.616 us | rcu_process_gp_end();
1434 0) sh-4802 | 0.586 us | check_for_new_grace_period();
1435 0) sh-4802 | 2.899 us | }
1436 0) sh-4802 | 4.040 us | }
1437 0) sh-4802 | 5.151 us | }
1438 0) sh-4802 | + 49.370 us | }
1439
1440
Ingo Molnar57526742009-02-19 12:54:10 +01001441- The absolute time field is an absolute timestamp given by the
1442 system clock since it started. A snapshot of this time is
1443 given on each entry/exit of functions
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001444
1445 hide: echo nofuncgraph-abstime > /debug/tracing/trace_options
1446 show: echo funcgraph-abstime > /debug/tracing/trace_options
1447
1448 ie:
1449
1450 #
1451 # TIME CPU DURATION FUNCTION CALLS
1452 # | | | | | | | |
1453 360.774522 | 1) 0.541 us | }
1454 360.774522 | 1) 4.663 us | }
1455 360.774523 | 1) 0.541 us | __wake_up_bit();
1456 360.774524 | 1) 6.796 us | }
1457 360.774524 | 1) 7.952 us | }
1458 360.774525 | 1) 9.063 us | }
1459 360.774525 | 1) 0.615 us | journal_mark_dirty();
1460 360.774527 | 1) 0.578 us | __brelse();
1461 360.774528 | 1) | reiserfs_prepare_for_journal() {
1462 360.774528 | 1) | unlock_buffer() {
1463 360.774529 | 1) | wake_up_bit() {
1464 360.774529 | 1) | bit_waitqueue() {
1465 360.774530 | 1) 0.594 us | __phys_addr();
1466
1467
Ingo Molnar57526742009-02-19 12:54:10 +01001468You can put some comments on specific functions by using
Ingo Molnar5e1607a2009-03-05 10:24:48 +01001469trace_printk() For example, if you want to put a comment inside
Ingo Molnar57526742009-02-19 12:54:10 +01001470the __might_sleep() function, you just have to include
Ingo Molnar5e1607a2009-03-05 10:24:48 +01001471<linux/ftrace.h> and call trace_printk() inside __might_sleep()
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001472
Ingo Molnar5e1607a2009-03-05 10:24:48 +01001473trace_printk("I'm a comment!\n")
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001474
1475will produce:
1476
1477 1) | __might_sleep() {
1478 1) | /* I'm a comment! */
1479 1) 1.449 us | }
1480
1481
Ingo Molnar57526742009-02-19 12:54:10 +01001482You might find other useful features for this tracer in the
1483following "dynamic ftrace" section such as tracing only specific
1484functions or tasks.
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001485
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001486dynamic ftrace
1487--------------
1488
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001489If CONFIG_DYNAMIC_FTRACE is set, the system will run with
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001490virtually no overhead when function tracing is disabled. The way
1491this works is the mcount function call (placed at the start of
Ingo Molnar57526742009-02-19 12:54:10 +01001492every kernel function, produced by the -pg switch in gcc),
1493starts of pointing to a simple return. (Enabling FTRACE will
1494include the -pg switch in the compiling of the kernel.)
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001495
Steven Rostedt9b803c02008-11-03 15:15:08 -05001496At compile time every C file object is run through the
1497recordmcount.pl script (located in the scripts directory). This
1498script will process the C object using objdump to find all the
Ingo Molnar57526742009-02-19 12:54:10 +01001499locations in the .text section that call mcount. (Note, only the
1500.text section is processed, since processing other sections like
1501.init.text may cause races due to those sections being freed).
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001502
Ingo Molnar57526742009-02-19 12:54:10 +01001503A new section called "__mcount_loc" is created that holds
1504references to all the mcount call sites in the .text section.
1505This section is compiled back into the original object. The
1506final linker will add all these references into a single table.
Steven Rostedt9b803c02008-11-03 15:15:08 -05001507
1508On boot up, before SMP is initialized, the dynamic ftrace code
Ingo Molnar57526742009-02-19 12:54:10 +01001509scans this table and updates all the locations into nops. It
1510also records the locations, which are added to the
1511available_filter_functions list. Modules are processed as they
1512are loaded and before they are executed. When a module is
1513unloaded, it also removes its functions from the ftrace function
1514list. This is automatic in the module unload code, and the
1515module author does not need to worry about it.
Steven Rostedt9b803c02008-11-03 15:15:08 -05001516
Ingo Molnar57526742009-02-19 12:54:10 +01001517When tracing is enabled, kstop_machine is called to prevent
1518races with the CPUS executing code being modified (which can
1519cause the CPU to do undesireable things), and the nops are
1520patched back to calls. But this time, they do not call mcount
1521(which is just a function stub). They now call into the ftrace
1522infrastructure.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001523
1524One special side-effect to the recording of the functions being
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001525traced is that we can now selectively choose which functions we
Ingo Molnar57526742009-02-19 12:54:10 +01001526wish to trace and which ones we want the mcount calls to remain
1527as nops.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001528
Ingo Molnar57526742009-02-19 12:54:10 +01001529Two files are used, one for enabling and one for disabling the
1530tracing of specified functions. They are:
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001531
1532 set_ftrace_filter
1533
1534and
1535
1536 set_ftrace_notrace
1537
Ingo Molnar57526742009-02-19 12:54:10 +01001538A list of available functions that you can add to these files is
1539listed in:
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001540
1541 available_filter_functions
1542
1543 # cat /debug/tracing/available_filter_functions
1544put_prev_task_idle
1545kmem_cache_create
1546pick_next_task_rt
1547get_online_cpus
1548pick_next_task_fair
1549mutex_lock
1550[...]
1551
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001552If I am only interested in sys_nanosleep and hrtimer_interrupt:
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001553
1554 # echo sys_nanosleep hrtimer_interrupt \
1555 > /debug/tracing/set_ftrace_filter
1556 # echo ftrace > /debug/tracing/current_tracer
1557 # echo 1 > /debug/tracing/tracing_enabled
1558 # usleep 1
1559 # echo 0 > /debug/tracing/tracing_enabled
1560 # cat /debug/tracing/trace
1561# tracer: ftrace
1562#
1563# TASK-PID CPU# TIMESTAMP FUNCTION
1564# | | | | |
1565 usleep-4134 [00] 1317.070017: hrtimer_interrupt <-smp_apic_timer_interrupt
1566 usleep-4134 [00] 1317.070111: sys_nanosleep <-syscall_call
1567 <idle>-0 [00] 1317.070115: hrtimer_interrupt <-smp_apic_timer_interrupt
1568
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001569To see which functions are being traced, you can cat the file:
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001570
1571 # cat /debug/tracing/set_ftrace_filter
1572hrtimer_interrupt
1573sys_nanosleep
1574
1575
Ingo Molnar57526742009-02-19 12:54:10 +01001576Perhaps this is not enough. The filters also allow simple wild
1577cards. Only the following are currently available
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001578
Steven Rostedta41eeba2008-07-14 16:41:12 -04001579 <match>* - will match functions that begin with <match>
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001580 *<match> - will match functions that end with <match>
1581 *<match>* - will match functions that have <match> in it
1582
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001583These are the only wild cards which are supported.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001584
1585 <match>*<match> will not work.
1586
Ingo Molnar57526742009-02-19 12:54:10 +01001587Note: It is better to use quotes to enclose the wild cards,
1588 otherwise the shell may expand the parameters into names
1589 of files in the local directory.
walimisc072c242008-11-28 12:21:19 +08001590
1591 # echo 'hrtimer_*' > /debug/tracing/set_ftrace_filter
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001592
1593Produces:
1594
1595# tracer: ftrace
1596#
1597# TASK-PID CPU# TIMESTAMP FUNCTION
1598# | | | | |
1599 bash-4003 [00] 1480.611794: hrtimer_init <-copy_process
1600 bash-4003 [00] 1480.611941: hrtimer_start <-hrtick_set
1601 bash-4003 [00] 1480.611956: hrtimer_cancel <-hrtick_clear
1602 bash-4003 [00] 1480.611956: hrtimer_try_to_cancel <-hrtimer_cancel
1603 <idle>-0 [00] 1480.612019: hrtimer_get_next_event <-get_next_timer_interrupt
1604 <idle>-0 [00] 1480.612025: hrtimer_get_next_event <-get_next_timer_interrupt
1605 <idle>-0 [00] 1480.612032: hrtimer_get_next_event <-get_next_timer_interrupt
1606 <idle>-0 [00] 1480.612037: hrtimer_get_next_event <-get_next_timer_interrupt
1607 <idle>-0 [00] 1480.612382: hrtimer_get_next_event <-get_next_timer_interrupt
1608
1609
1610Notice that we lost the sys_nanosleep.
1611
1612 # cat /debug/tracing/set_ftrace_filter
1613hrtimer_run_queues
1614hrtimer_run_pending
1615hrtimer_init
1616hrtimer_cancel
1617hrtimer_try_to_cancel
1618hrtimer_forward
1619hrtimer_start
1620hrtimer_reprogram
1621hrtimer_force_reprogram
1622hrtimer_get_next_event
1623hrtimer_interrupt
1624hrtimer_nanosleep
1625hrtimer_wakeup
1626hrtimer_get_remaining
1627hrtimer_get_res
1628hrtimer_init_sleeper
1629
1630
1631This is because the '>' and '>>' act just like they do in bash.
1632To rewrite the filters, use '>'
1633To append to the filters, use '>>'
1634
Ingo Molnar57526742009-02-19 12:54:10 +01001635To clear out a filter so that all functions will be recorded
1636again:
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001637
1638 # echo > /debug/tracing/set_ftrace_filter
1639 # cat /debug/tracing/set_ftrace_filter
1640 #
1641
1642Again, now we want to append.
1643
1644 # echo sys_nanosleep > /debug/tracing/set_ftrace_filter
1645 # cat /debug/tracing/set_ftrace_filter
1646sys_nanosleep
walimisc072c242008-11-28 12:21:19 +08001647 # echo 'hrtimer_*' >> /debug/tracing/set_ftrace_filter
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001648 # cat /debug/tracing/set_ftrace_filter
1649hrtimer_run_queues
1650hrtimer_run_pending
1651hrtimer_init
1652hrtimer_cancel
1653hrtimer_try_to_cancel
1654hrtimer_forward
1655hrtimer_start
1656hrtimer_reprogram
1657hrtimer_force_reprogram
1658hrtimer_get_next_event
1659hrtimer_interrupt
1660sys_nanosleep
1661hrtimer_nanosleep
1662hrtimer_wakeup
1663hrtimer_get_remaining
1664hrtimer_get_res
1665hrtimer_init_sleeper
1666
1667
Ingo Molnar57526742009-02-19 12:54:10 +01001668The set_ftrace_notrace prevents those functions from being
1669traced.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001670
1671 # echo '*preempt*' '*lock*' > /debug/tracing/set_ftrace_notrace
1672
1673Produces:
1674
1675# tracer: ftrace
1676#
1677# TASK-PID CPU# TIMESTAMP FUNCTION
1678# | | | | |
1679 bash-4043 [01] 115.281644: finish_task_switch <-schedule
1680 bash-4043 [01] 115.281645: hrtick_set <-schedule
1681 bash-4043 [01] 115.281645: hrtick_clear <-hrtick_set
1682 bash-4043 [01] 115.281646: wait_for_completion <-__stop_machine_run
1683 bash-4043 [01] 115.281647: wait_for_common <-wait_for_completion
1684 bash-4043 [01] 115.281647: kthread_stop <-stop_machine_run
1685 bash-4043 [01] 115.281648: init_waitqueue_head <-kthread_stop
1686 bash-4043 [01] 115.281648: wake_up_process <-kthread_stop
1687 bash-4043 [01] 115.281649: try_to_wake_up <-wake_up_process
1688
1689We can see that there's no more lock or preempt tracing.
1690
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001691
Ingo Molnar57526742009-02-19 12:54:10 +01001692Dynamic ftrace with the function graph tracer
1693---------------------------------------------
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001694
Ingo Molnar57526742009-02-19 12:54:10 +01001695Although what has been explained above concerns both the
1696function tracer and the function-graph-tracer, there are some
1697special features only available in the function-graph tracer.
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001698
Ingo Molnar57526742009-02-19 12:54:10 +01001699If you want to trace only one function and all of its children,
1700you just have to echo its name into set_graph_function:
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001701
Ingo Molnar57526742009-02-19 12:54:10 +01001702 echo __do_fault > set_graph_function
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001703
Ingo Molnar57526742009-02-19 12:54:10 +01001704will produce the following "expanded" trace of the __do_fault()
1705function:
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001706
1707 0) | __do_fault() {
1708 0) | filemap_fault() {
1709 0) | find_lock_page() {
1710 0) 0.804 us | find_get_page();
1711 0) | __might_sleep() {
1712 0) 1.329 us | }
1713 0) 3.904 us | }
1714 0) 4.979 us | }
1715 0) 0.653 us | _spin_lock();
1716 0) 0.578 us | page_add_file_rmap();
1717 0) 0.525 us | native_set_pte_at();
1718 0) 0.585 us | _spin_unlock();
1719 0) | unlock_page() {
1720 0) 0.541 us | page_waitqueue();
1721 0) 0.639 us | __wake_up_bit();
1722 0) 2.786 us | }
1723 0) + 14.237 us | }
1724 0) | __do_fault() {
1725 0) | filemap_fault() {
1726 0) | find_lock_page() {
1727 0) 0.698 us | find_get_page();
1728 0) | __might_sleep() {
1729 0) 1.412 us | }
1730 0) 3.950 us | }
1731 0) 5.098 us | }
1732 0) 0.631 us | _spin_lock();
1733 0) 0.571 us | page_add_file_rmap();
1734 0) 0.526 us | native_set_pte_at();
1735 0) 0.586 us | _spin_unlock();
1736 0) | unlock_page() {
1737 0) 0.533 us | page_waitqueue();
1738 0) 0.638 us | __wake_up_bit();
1739 0) 2.793 us | }
1740 0) + 14.012 us | }
1741
Ingo Molnar57526742009-02-19 12:54:10 +01001742You can also expand several functions at once:
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001743
Ingo Molnar57526742009-02-19 12:54:10 +01001744 echo sys_open > set_graph_function
1745 echo sys_close >> set_graph_function
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001746
Ingo Molnar57526742009-02-19 12:54:10 +01001747Now if you want to go back to trace all functions you can clear
1748this special filter via:
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001749
Ingo Molnar57526742009-02-19 12:54:10 +01001750 echo > set_graph_function
Frederic Weisbecker985ec202009-02-18 06:35:34 +01001751
1752
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001753trace_pipe
1754----------
1755
Ingo Molnar57526742009-02-19 12:54:10 +01001756The trace_pipe outputs the same content as the trace file, but
1757the effect on the tracing is different. Every read from
1758trace_pipe is consumed. This means that subsequent reads will be
1759different. The trace is live.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001760
Steven Rostedt9b803c02008-11-03 15:15:08 -05001761 # echo function > /debug/tracing/current_tracer
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001762 # cat /debug/tracing/trace_pipe > /tmp/trace.out &
1763[1] 4153
1764 # echo 1 > /debug/tracing/tracing_enabled
1765 # usleep 1
1766 # echo 0 > /debug/tracing/tracing_enabled
1767 # cat /debug/tracing/trace
Steven Rostedt9b803c02008-11-03 15:15:08 -05001768# tracer: function
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001769#
1770# TASK-PID CPU# TIMESTAMP FUNCTION
1771# | | | | |
1772
1773 #
1774 # cat /tmp/trace.out
1775 bash-4043 [00] 41.267106: finish_task_switch <-schedule
1776 bash-4043 [00] 41.267106: hrtick_set <-schedule
1777 bash-4043 [00] 41.267107: hrtick_clear <-hrtick_set
1778 bash-4043 [00] 41.267108: wait_for_completion <-__stop_machine_run
1779 bash-4043 [00] 41.267108: wait_for_common <-wait_for_completion
1780 bash-4043 [00] 41.267109: kthread_stop <-stop_machine_run
1781 bash-4043 [00] 41.267109: init_waitqueue_head <-kthread_stop
1782 bash-4043 [00] 41.267110: wake_up_process <-kthread_stop
1783 bash-4043 [00] 41.267110: try_to_wake_up <-wake_up_process
1784 bash-4043 [00] 41.267111: select_task_rq_rt <-try_to_wake_up
1785
1786
Ingo Molnar57526742009-02-19 12:54:10 +01001787Note, reading the trace_pipe file will block until more input is
1788added. By changing the tracer, trace_pipe will issue an EOF. We
1789needed to set the function tracer _before_ we "cat" the
1790trace_pipe file.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001791
1792
1793trace entries
1794-------------
1795
Ingo Molnar57526742009-02-19 12:54:10 +01001796Having too much or not enough data can be troublesome in
1797diagnosing an issue in the kernel. The file buffer_size_kb is
1798used to modify the size of the internal trace buffers. The
1799number listed is the number of entries that can be recorded per
1800CPU. To know the full size, multiply the number of possible CPUS
1801with the number of entries.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001802
Steven Rostedta94c80e2008-11-12 17:52:36 -05001803 # cat /debug/tracing/buffer_size_kb
Steven Rostedt1696b2b2008-11-13 00:09:35 -050018041408 (units kilobytes)
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001805
Ingo Molnar57526742009-02-19 12:54:10 +01001806Note, to modify this, you must have tracing completely disabled.
1807To do that, echo "nop" into the current_tracer. If the
1808current_tracer is not set to "nop", an EINVAL error will be
1809returned.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001810
Steven Rostedt9b803c02008-11-03 15:15:08 -05001811 # echo nop > /debug/tracing/current_tracer
Steven Rostedt1696b2b2008-11-13 00:09:35 -05001812 # echo 10000 > /debug/tracing/buffer_size_kb
Steven Rostedta94c80e2008-11-12 17:52:36 -05001813 # cat /debug/tracing/buffer_size_kb
Steven Rostedt1696b2b2008-11-13 00:09:35 -0500181410000 (units kilobytes)
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001815
Ingo Molnar57526742009-02-19 12:54:10 +01001816The number of pages which will be allocated is limited to a
1817percentage of available memory. Allocating too much will produce
1818an error.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001819
Steven Rostedta94c80e2008-11-12 17:52:36 -05001820 # echo 1000000000000 > /debug/tracing/buffer_size_kb
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001821-bash: echo: write error: Cannot allocate memory
Steven Rostedta94c80e2008-11-12 17:52:36 -05001822 # cat /debug/tracing/buffer_size_kb
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400182385
1824
Ingo Molnar57526742009-02-19 12:54:10 +01001825-----------
1826
1827More details can be found in the source code, in the
1828kernel/tracing/*.c files.