1TRACE-CMD-RECORD(1) TRACE-CMD-RECORD(1)
2
3
4
6 trace-cmd-record - record a trace from the Ftrace Linux internal tracer
7
9 trace-cmd record [OPTIONS] [command]
10
12 The trace-cmd(1) record command will set up the Ftrace Linux kernel
13 tracer to record the specified plugins or events that happen while the
14 command executes. If no command is given, then it will record until the
15 user hits Ctrl-C.
16
17 The record command of trace-cmd will set up the Ftrace tracer to start
18 tracing the various events or plugins that are given on the command
19 line. It will then create a number of tracing processes (one per CPU)
20 that will start recording from the kernel ring buffer straight into
21 temporary files. When the command is complete (or Ctrl-C is hit) all
22 the files will be combined into a trace.dat file that can later be read
23 (see trace-cmd-report(1)).
24
26 -p tracer
27 Specify a tracer. Tracers usually do more than just trace an event.
28 Common tracers are: function, function_graph, preemptirqsoff,
29 irqsoff, preemptoff and wakeup. A tracer must be supported by the
30 running kernel. To see a list of available tracers, see
31 trace-cmd-list(1).
32
33 -e event
34 Specify an event to trace. Various static trace points have been
35 added to the Linux kernel. They are grouped by subsystem where you
36 can enable all events of a given subsystem or specify specific
37 events to be enabled. The event is of the format
38 "subsystem:event-name". You can also just specify the subsystem
39 without the :event-name or the event-name without the "subsystem:".
40 Using "-e sched_switch" will enable the "sched_switch" event where
41 as, "-e sched" will enable all events under the "sched" subsystem.
42
43 The 'event' can also contain glob expressions. That is, "*stat*" will
44 select all events (or subsystems) that have the characters "stat" in their
45 names.
46
47 The keyword 'all' can be used to enable all events.
48
49 -a
50 Every event that is being recorded has its output format file saved
51 in the output file to be able to display it later. But if other
52 events are enabled in the trace without trace-cmd’s knowledge, the
53 formats of those events will not be recorded and trace-cmd report
54 will not be able to display them. If this is the case, then specify
55 the -a option and the format for all events in the system will be
56 saved.
57
58 -T
59 Enable a stacktrace on each event. For example:
60
61 <idle>-0 [003] 58549.289091: sched_switch: kworker/0:1:0 [120] R ==> trace-cmd:2603 [120]
62 <idle>-0 [003] 58549.289092: kernel_stack: <stack trace>
63 => schedule (ffffffff814b260e)
64 => cpu_idle (ffffffff8100a38c)
65 => start_secondary (ffffffff814ab828)
66
67 --func-stack
68 Enable a stack trace on all functions. Note this is only applicable
69 for the "function" plugin tracer, and will only take effect if the
70 -l option is used and succeeds in limiting functions. If the
71 function tracer is not filtered, and the stack trace is enabled,
72 you can live lock the machine.
73
74 -f filter
75 Specify a filter for the previous event. This must come after a -e.
76 This will filter what events get recorded based on the content of
77 the event. Filtering is passed to the kernel directly so what
78 filtering is allowed may depend on what version of the kernel you
79 have. Basically, it will let you use C notation to check if an
80 event should be processed or not.
81
82 ==, >=, <=, >, <, &, |, && and ||
83
84 The above are usually safe to use to compare fields.
85
86 -R trigger
87 Specify a trigger for the previous event. This must come after a
88 -e. This will add a given trigger to the given event. To only
89 enable the trigger and not the event itself, then place the event
90 after the -v option.
91
92 See Documentation/trace/events.txt in the Linux kernel source for more
93 information on triggers.
94
95 -v
96 This will cause all events specified after it on the command line
97 to not be traced. This is useful for selecting a subsystem to be
98 traced but to leave out various events. For Example: "-e sched -v
99 -e "*stat\*"" will enable all events in the sched subsystem except
100 those that have "stat" in their names.
101
102 Note: the *-v* option was taken from the way grep(1) inverts the following
103 matches.
104
105 -F
106 This will filter only the executable that is given on the command
107 line. If no command is given, then it will filter itself (pretty
108 pointless). Using -F will let you trace only events that are caused
109 by the given command.
110
111 -P pid
112 Similar to -F but lets you specify a process ID to trace.
113
114 -c
115 Used with either -F (or -P if kernel supports it) to trace the
116 process' children too.
117
118 -C clock
119 Set the trace clock to "clock".
120
121 Use trace-cmd(1) list -C to see what clocks are available.
122
123 -o output-file
124 By default, trace-cmd report will create a trace.dat file. You can
125 specify a different file to write to with the -o option.
126
127 -l function-name
128 This will limit the function and function_graph tracers to only
129 trace the given function name. More than one -l may be specified on
130 the command line to trace more than one function. The limited use
131 of glob expressions are also allowed. These are match* to only
132 filter functions that start with match. *match to only filter
133 functions that end with match. *match\* to only filter on
134 functions that contain match.
135
136 -g function-name
137 This option is for the function_graph plugin. It will graph the
138 given function. That is, it will only trace the function and all
139 functions that it calls. You can have more than one -g on the
140 command line.
141
142 -n function-name
143 This has the opposite effect of -l. The function given with the -n
144 option will not be traced. This takes precedence, that is, if you
145 include the same function for both -n and -l, it will not be
146 traced.
147
148 -d
149 Some tracer plugins enable the function tracer by default. Like the
150 latency tracers. This option prevents the function tracer from
151 being enabled at start up.
152
153 -D
154 The option -d will try to use the function-trace option to disable
155 the function tracer (if available), otherwise it defaults to the
156 proc file: /proc/sys/kernel/ftrace_enabled, but will not touch it
157 if the function-trace option is available. The -D option will
158 disable both the ftrace_enabled proc file as well as the
159 function-trace option if it exists.
160
161 Note, this disable function tracing for all users, which includes users
162 outside of ftrace tracers (stack_tracer, perf, etc).
163
164 -O option
165 Ftrace has various options that can be enabled or disabled. This
166 allows you to set them. Appending the text no to an option disables
167 it. For example: "-O nograph-time" will disable the "graph-time"
168 Ftrace option.
169
170 -s interval
171 The processes that trace-cmd creates to record from the ring buffer
172 need to wake up to do the recording. Setting the interval to zero
173 will cause the processes to wakeup every time new data is written
174 into the buffer. But since Ftrace is recording kernel activity, the
175 act of this processes going back to sleep may cause new events into
176 the ring buffer which will wake the process back up. This will
177 needlessly add extra data into the ring buffer.
178
179 The 'interval' metric is microseconds. The default is set to 1000 (1 ms).
180 This is the time each recording process will sleep before waking up to
181 record any new data that was written to the ring buffer.
182
183 -r priority
184 The priority to run the capture threads at. In a busy system the
185 trace capturing threads may be staved and events can be lost. This
186 increases the priority of those threads to the real time (FIFO)
187 priority. But use this option with care, it can also change the
188 behaviour of the system being traced.
189
190 -b size
191 This sets the ring buffer size to size kilobytes. Because the
192 Ftrace ring buffer is per CPU, this size is the size of each per
193 CPU ring buffer inside the kernel. Using "-b 10000" on a machine
194 with 4 CPUs will make Ftrace have a total buffer size of 40 Megs.
195
196 -B buffer-name
197 If the kernel supports multiple buffers, this will add a buffer
198 with the given name. If the buffer name already exists, that buffer
199 is just reset and will not be deleted at the end of record
200 execution. If the buffer is created, it will be removed at the end
201 of execution (unless the -k is set, or start command was used).
202
203 After a buffer name is stated, all events added after that will be
204 associated with that buffer. If no buffer is specified, or an event
205 is specified before a buffer name, it will be associated with the
206 main (toplevel) buffer.
207
208 trace-cmd record -e sched -B block -e block -B time -e timer sleep 1
209
210 The above is will enable all sched events in the main buffer. It will
211 then create a 'block' buffer instance and enable all block events within
212 that buffer. A 'time' buffer instance is created and all timer events
213 will be enabled for that event.
214
215 -m size
216 The max size in kilobytes that a per cpu buffer should be. Note,
217 due to rounding to page size, the number may not be totally
218 correct. Also, this is performed by switching between two buffers
219 that are half the given size thus the output may not be of the
220 given size even if much more was written.
221
222 Use this to prevent running out of diskspace for long runs.
223
224 -M cpumask
225 Set the cpumask for to trace. It only affects the last buffer
226 instance given. If supplied before any buffer instance, then it
227 affects the main buffer. The value supplied must be a hex number.
228
229 trace-cmd record -p function -M c -B events13 -e all -M 5
230
231 If the -M is left out, then the mask stays the same. To enable all
232 CPUs, pass in a value of '-1'.
233
234 -k
235 By default, when trace-cmd is finished tracing, it will reset the
236 buffers and disable all the tracing that it enabled. This option
237 keeps trace-cmd from disabling the tracer and reseting the buffer.
238 This option is useful for debugging trace-cmd.
239
240 Note: usually trace-cmd will set the "tracing_on" file back to what it
241 was before it was called. This option will leave that file set to zero.
242
243 -i
244 By default, if an event is listed that trace-cmd does not find, it
245 will exit with an error. This option will just ignore events that
246 are listed on the command line but are not found on the system.
247
248 -N host:port
249 If another machine is running "trace-cmd listen", this option is
250 used to have the data sent to that machine with UDP packets.
251 Instead of writing to an output file, the data is sent off to a
252 remote box. This is ideal for embedded machines with little
253 storage, or having a single machine that will keep all the data in
254 a single repository.
255
256 Note: This option is not supported with latency tracer plugins:
257 wakeup, wakeup_rt, irqsoff, preemptoff and preemptirqsoff
258
259 -t
260 This option is used with -N, when there’s a need to send the live
261 data with TCP packets instead of UDP. Although TCP is not nearly as
262 fast as sending the UDP packets, but it may be needed if the
263 network is not that reliable, the amount of data is not that
264 intensive, and a guarantee is needed that all traced information is
265 transfered successfully.
266
267 -q | --quiet
268 For use with recording an application. Suppresses normal output
269 (except for errors) to allow only the application’s output to be
270 displayed.
271
272 --date
273 With the --date option, "trace-cmd" will write timestamps into the
274 trace buffer after it has finished recording. It will then map the
275 timestamp to gettimeofday which will allow wall time output from
276 the timestamps reading the created trace.dat file.
277
278 --max-graph-depth depth
279 Set the maximum depth the function_graph tracer will trace into a
280 function. A value of one will only show where userspace enters the
281 kernel but not any functions called in the kernel. The default is
282 zero, which means no limit.
283
284 --module module
285 Filter a module’s name in function tracing. It is equivalent to
286 adding :mod:module after all other functions being filtered. If no
287 other function filter is listed, then all modules functions will be
288 filtered in the filter.
289
290 '--module snd' is equivalent to '-l :mod:snd'
291
292 '--module snd -l "*jack*"' is equivalent to '-l "*jack*:mod:snd"'
293
294 '--module snd -n "*"' is equivalent to '-n :mod:snd'
295
296 --profile
297 With the --profile option, "trace-cmd" will enable tracing that can
298 be used with trace-cmd-report(1) --profile option. If a tracer -p
299 is not set, and function graph depth is supported by the kernel,
300 then the function_graph tracer will be enabled with a depth of one
301 (only show where userspace enters into the kernel). It will also
302 enable various tracepoints with stack tracing such that the report
303 can show where tasks have been blocked for the longest time.
304
305 See trace-cmd-profile(1) for more details and examples.
306
307 -H event-hooks
308 Add custom event matching to connect any two events together. When
309 not used with --profile, it will save the parameter and this will
310 be used by trace-cmd report --profile, too. That is:
311
312 trace-cmd record -H hrtimer_expire_entry,hrtimer/hrtimer_expire_exit,hrtimer,sp
313 trace-cmd report --profile
314
315 Will profile hrtimer_expire_entry and hrtimer_expire_ext times.
316
317 See trace-cmd-profile(1) for format.
318
319 -S
320 (for --profile only) Only enable the tracer or events speficied on
321 the command line. With this option, the function_graph tracer is
322 not enabled, nor are any events (like sched_switch), unless they
323 are specifically specified on the command line (i.e. -p function -e
324 sched_switch -e sched_wakeup)
325
326 --ts-offset offset
327 Add an offset for the timestamp in the trace.dat file. This will
328 add a offset option into the trace.dat file such that a trace-cmd
329 report will offset all the timestamps of the events by the given
330 offset. The offset is in raw units. That is, if the event
331 timestamps are in nanoseconds the offset will also be in
332 nanoseconds even if the displayed units are in microseconds.
333
334 --stderr
335 Have output go to stderr instead of stdout, but the output of the
336 command executed will not be changed. This is useful if you want to
337 monitor the output of the command being executed, but not see the
338 output from trace-cmd.
339
341 The basic way to trace all events:
342
343 # trace-cmd record -e all ls > /dev/null
344 # trace-cmd report
345 trace-cmd-13541 [003] 106260.693809: filemap_fault: address=0x128122 offset=0xce
346 trace-cmd-13543 [001] 106260.693809: kmalloc: call_site=81128dd4 ptr=0xffff88003dd83800 bytes_req=768 bytes_alloc=1024 gfp_flags=GFP_KERNEL|GFP_ZERO
347 ls-13545 [002] 106260.693809: kfree: call_site=810a7abb ptr=0x0
348 ls-13545 [002] 106260.693818: sys_exit_write: 0x1
349
350 To use the function tracer with sched switch tracing:
351
352 # trace-cmd record -p function -e sched_switch ls > /dev/null
353 # trace-cmd report
354 ls-13587 [002] 106467.860310: function: hrtick_start_fair <-- pick_next_task_fair
355 ls-13587 [002] 106467.860313: sched_switch: prev_comm=trace-cmd prev_pid=13587 prev_prio=120 prev_state=R ==> next_comm=trace-cmd next_pid=13583 next_prio=120
356 trace-cmd-13585 [001] 106467.860314: function: native_set_pte_at <-- __do_fault
357 trace-cmd-13586 [003] 106467.860314: function: up_read <-- do_page_fault
358 ls-13587 [002] 106467.860317: function: __phys_addr <-- schedule
359 trace-cmd-13585 [001] 106467.860318: function: _raw_spin_unlock <-- __do_fault
360 ls-13587 [002] 106467.860320: function: native_load_sp0 <-- __switch_to
361 trace-cmd-13586 [003] 106467.860322: function: down_read_trylock <-- do_page_fault
362
363 Here is a nice way to find what interrupts have the highest latency:
364
365 # trace-cmd record -p function_graph -e irq_handler_entry -l do_IRQ sleep 10
366 # trace-cmd report
367 <idle>-0 [000] 157412.933969: funcgraph_entry: | do_IRQ() {
368 <idle>-0 [000] 157412.933974: irq_handler_entry: irq=48 name=eth0
369 <idle>-0 [000] 157412.934004: funcgraph_exit: + 36.358 us | }
370 <idle>-0 [000] 157413.895004: funcgraph_entry: | do_IRQ() {
371 <idle>-0 [000] 157413.895011: irq_handler_entry: irq=48 name=eth0
372 <idle>-0 [000] 157413.895026: funcgraph_exit: + 24.014 us | }
373 <idle>-0 [000] 157415.891762: funcgraph_entry: | do_IRQ() {
374 <idle>-0 [000] 157415.891769: irq_handler_entry: irq=48 name=eth0
375 <idle>-0 [000] 157415.891784: funcgraph_exit: + 22.928 us | }
376 <idle>-0 [000] 157415.934869: funcgraph_entry: | do_IRQ() {
377 <idle>-0 [000] 157415.934874: irq_handler_entry: irq=48 name=eth0
378 <idle>-0 [000] 157415.934906: funcgraph_exit: + 37.512 us | }
379 <idle>-0 [000] 157417.888373: funcgraph_entry: | do_IRQ() {
380 <idle>-0 [000] 157417.888381: irq_handler_entry: irq=48 name=eth0
381 <idle>-0 [000] 157417.888398: funcgraph_exit: + 25.943 us | }
382
383 An example of the profile:
384
385 # trace-cmd record --profile sleep 1
386 # trace-cmd report --profile --comm sleep
387 task: sleep-21611
388 Event: sched_switch:R (1) Total: 99442 Avg: 99442 Max: 99442 Min:99442
389 <stack> 1 total:99442 min:99442 max:99442 avg=99442
390 => ftrace_raw_event_sched_switch (0xffffffff8105f812)
391 => __schedule (0xffffffff8150810a)
392 => preempt_schedule (0xffffffff8150842e)
393 => ___preempt_schedule (0xffffffff81273354)
394 => cpu_stop_queue_work (0xffffffff810b03c5)
395 => stop_one_cpu (0xffffffff810b063b)
396 => sched_exec (0xffffffff8106136d)
397 => do_execve_common.isra.27 (0xffffffff81148c89)
398 => do_execve (0xffffffff811490b0)
399 => SyS_execve (0xffffffff811492c4)
400 => return_to_handler (0xffffffff8150e3c8)
401 => stub_execve (0xffffffff8150c699)
402 Event: sched_switch:S (1) Total: 1000506680 Avg: 1000506680 Max: 1000506680 Min:1000506680
403 <stack> 1 total:1000506680 min:1000506680 max:1000506680 avg=1000506680
404 => ftrace_raw_event_sched_switch (0xffffffff8105f812)
405 => __schedule (0xffffffff8150810a)
406 => schedule (0xffffffff815084b8)
407 => do_nanosleep (0xffffffff8150b22c)
408 => hrtimer_nanosleep (0xffffffff8108d647)
409 => SyS_nanosleep (0xffffffff8108d72c)
410 => return_to_handler (0xffffffff8150e3c8)
411 => tracesys_phase2 (0xffffffff8150c304)
412 Event: sched_wakeup:21611 (1) Total: 30326 Avg: 30326 Max: 30326 Min:30326
413 <stack> 1 total:30326 min:30326 max:30326 avg=30326
414 => ftrace_raw_event_sched_wakeup_template (0xffffffff8105f653)
415 => ttwu_do_wakeup (0xffffffff810606eb)
416 => ttwu_do_activate.constprop.124 (0xffffffff810607c8)
417 => try_to_wake_up (0xffffffff8106340a)
418
420 trace-cmd(1), trace-cmd-report(1), trace-cmd-start(1),
421 trace-cmd-stop(1), trace-cmd-extract(1), trace-cmd-reset(1),
422 trace-cmd-split(1), trace-cmd-list(1), trace-cmd-listen(1),
423 trace-cmd-profile(1)
424
426 Written by Steven Rostedt, <rostedt@goodmis.org[1]>
427
429 git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/trace-cmd.git
430
432 Copyright (C) 2010 Red Hat, Inc. Free use of this software is granted
433 under the terms of the GNU Public License (GPL).
434
436 1. rostedt@goodmis.org
437 mailto:rostedt@goodmis.org
438
439
440
441 08/07/2019 TRACE-CMD-RECORD(1)