1TRACE-CMD-REPORT(1) TRACE-CMD-REPORT(1)
2
6 trace-cmd-report - show in ASCII a trace created by trace-cmd record
7
9 trace-cmd report [OPTIONS] [input-file]
10
12 The trace-cmd(1) report command will output a human readable report of
13 a trace created by trace-cmd record.
14
16 -i input-file
17 By default, trace-cmd report will read the file trace.dat. But the
18 -i option open up the given input-file instead. Note, the input
19 file may also be specified as the last item on the command line.
20 -e
22 This outputs the endianess of the file. trace-cmd report is smart
23 enough to be able to read big endian files on little endian
24 machines, and vise versa.
25 -f
27 This outputs the list of all functions that have been mapped in the
28 trace.dat file. Note, this list may contain functions that may not
29 appear in the trace, as it is the list of mappings to translate
30 function addresses into function names.
31 -P
33 This outputs the list of "trace_printk()" data. The raw trace data
34 points to static pointers in the kernel. This must be stored in the
35 trace.dat file.
36 -E
38 This lists the possible events in the file (but this list is not
39 necessarily the list of events in the file).
40 --events
42 This will list the event formats that are stored in the trace.dat
43 file.
44 --event regex
46 This will print events that match the given regex. If a colon is
47 specified, then the characters before the colon will be used to
48 match the system and the characters after the colon will match the
49 event.
50 trace-cmd report --event sys:read
52 The above will only match events where the system name contains "sys"
54 and the event name contains "read".
55 trace-cmd report --event read
57 The above will match all events that contain "read" in its name. Also it
59 may list all events of a system that contains "read" as well.
60 --check-events
62 This will parse the event format strings that are stored in the
63 trace.dat file and return whether the formats can be parsed
64 correctly. It will load plugins unless -N is specified.
65 -t
67 Print the full timestamp. The timestamps in the data file are
68 usually recorded to the nanosecond. But the default display of the
69 timestamp is only to the microsecond. To see the full timestamp,
70 add the -t option.
71 -F filter
73 Add a filter to limit what events are displayed. The format of the
74 filter is:
75 <events> ':' <filter>
77 <events> = SYSTEM'/'EVENT | SYSTEM | EVENT | <events> ',' <events>
78 <filter> = EVENT_FIELD <op> <value> | <filter> '&&' <filter> |
79 <filter> '||' <filter> | '(' <filter> ')' | '!' <filter>
80 <op> = '==' | '!=' | '>=' | '<=' | '>' | '<' | '&' | '|' | '^' |
81 '+' | '-' | '*' | '/' | '%'
82 <value> = NUM | STRING | EVENT_FIELD
83 SYSTEM is the name of the system to filter on. If the EVENT is left out,
85 then it applies to all events under the SYSTEM. If only one string is used
86 without the '/' to deliminate between SYSTEM and EVENT, then the filter
87 will be applied to all systems and events that match the given string.
88 Whitespace is ignored, such that "sched:next_pid==123" is equivalent to
90 "sched : next_pid == 123".
91 STRING is defined with single or double quotes (single quote must end with
93 single quote, and double with double). Whitespace within quotes are not
94 ignored.
95 The representation of a SYSTEM or EVENT may also be a regular expression
97 as defined by 'regcomp(3)'.
98 The EVENT_FIELD is the name of the field of an event that is being
100 filtered. If the event does not contain the EVENT_FIELD, that part of the
101 equation will be considered false.
102 -F 'sched : bogus == 1 || common_pid == 2'
104 The "bogus == 1" will always evaluate to FALSE because no event has a
106 field called "bogus", but the "common_pid == 2" will still be evaluated
107 since all events have the field "common_pid". Any "sched" event that was
108 traced by the process with the PID of 2 will be shown.
109 Note, the EVENT_FIELD is the field name as shown by an events format
111 (as displayed with *--events*), and not what is found in the output.
112 If the output shows "ID:foo" but the field that "foo" belongs to was
113 called "name" in the event format, then "name" must be used in the filter.
114 The same is true about values. If the value that is displayed is converted
115 by to a string symbol, the filter checks the original value and not the
116 value displayed. For example, to filter on all tasks that were in the
117 running state at a context switch:
118 -F 'sched/sched_switch : prev_state==0'
120 Although the output displays 'R', having 'prev_stat=="R"' will not work.
122 Note: You can also specify 'COMM' as an EVENT_FIELD. This will use the
124 task name (or comm) of the record to compare. For example, to filter out
125 all of the "trace-cmd" tasks:
126 -F '.*:COMM != "trace-cmd"'
128 -I
130 Do not print events where the HARDIRQ latency flag is set. This
131 will filter out most events that are from interrupt context. Note,
132 it may not filter out function traced functions that are in
133 interrupt context but were called before the kernel "in interrupt"
134 flag was set.
135 -S
137 Do not print events where the SOFTIRQ latency flag is set. This
138 will filter out most events that are from soft interrupt context.
139 -v
141 This causes the following filters of -F to filter out the matching
142 events.
143 -v -F 'sched/sched_switch : prev_state == 0'
145 Will not display any sched_switch events that have a prev_state of 0.
147 Removing the *-v* will only print out those events.
148 -T
150 Test the filters of -F. After processing a filter string, the
151 resulting filter will be displayed for each event. This is useful
152 for using a filter for more than one event where a field may not
153 exist in all events. Also it can be used to make sure there are no
154 misspelled event field names, as they will simply be ignored. -T
155 is ignored if -F is not specified.
156 -V
158 Show the plugins that are loaded.
159 -L
161 This will not load system wide plugins. It loads "local only". That
162 is what it finds in the ~/.trace-cmd/plugins directory.
163 -N
165 This will not load any plugins.
166 -n event-re
168 This will cause all events that match the option to ignore any
169 registered handler (by the plugins) to print the event. The normal
170 event will be printed instead. The event-re is a regular expression
171 as defined by regcomp(3).
172 --profile
174 With the --profile option, "trace-cmd report" will process all the
175 events first, and then output a format showing where tasks have
176 spent their time in the kernel, as well as where they are blocked
177 the most, and where wake up latencies are.
178 See trace-cmd-profile(1) for more details and examples.
180 -G
182 Set interrupt (soft and hard) events as global (associated to CPU
183 instead of tasks). Only works for --profile.
184 -H event-hooks
186 Add custom event matching to connect any two events together.
187 See trace-cmd-profile(1) for format.
189 -R
191 This will show the events in "raw" format. That is, it will ignore
192 the event’s print formatting and just print the contents of each
193 field.
194 -r event-re
196 This will cause all events that match the option to print its raw
197 fields. The event-re is a regular expression as defined by
198 regcomp(3).
199 -l
201 This adds a "latency output" format. Information about interrupts
202 being disabled, soft irq being disabled, the "need_resched" flag
203 being set, preempt count, and big kernel lock are all being
204 recorded with every event. But the default display does not show
205 this information. This option will set display this information
206 with 6 characters. When one of the fields is zero or N/A a '.\' is
207 shown.
208 <idle>-0 0d.h1. 106467.859747: function: ktime_get <-- tick_check_idle
210 The 0d.h1. denotes this information. The first character is never a '.'
212 and represents what CPU the trace was recorded on (CPU 0). The 'd' denotes
213 that interrupts were disabled. The 'h' means that this was called inside
214 an interrupt handler. The '1' is the preemption disabled (preempt_count)
215 was set to one. The two '.'s are "need_resched" flag and kernel lock
216 counter. If the "need_resched" flag is set, then that character would be a
217 'N'.
218 -w
220 If both the sched_switch and sched_wakeup events are enabled, then
221 this option will report the latency between the time the task was
222 first woken, and the time it was scheduled in.
223 -q
225 Quiet non critical warnings.
226 -O
228 Pass options to the trace-cmd plugins that are loaded.
229 -O plugin:var=value
231 The 'plugin:' and '=value' are optional. Value may be left off for options
233 that are boolean. If the 'plugin:' is left off, then any variable that matches
234 in all plugins will be set.
235 Example: -O fgraph:tailprint
237 --cpu <cpu list>
239 List of CPUs, separated by "," or ":", used for filtering the
240 events. A range of CPUs can be specified using "cpuX-cpuY"
241 notation, where all CPUs in the range between cpuX and cpuY will be
242 included in the list. The order of CPUs in the list must be from
243 lower to greater.
244 Example: "--cpu 0,3" - show events from CPUs 0 and 3
246 "--cpu 2-4" - show events from CPUs 2, 3 and 4
247 --stat
249 If the trace.dat file recorded the final stats (outputed at the end
250 of record) the --stat option can be used to retrieve them.
251 --uname
253 If the trace.dat file recorded uname during the run, this will
254 retrieve that information.
255 --version
257 If the trace.dat file recorded the version of the executable used
258 to create it, report that version.
259 --ts-offset offset
261 Add (or subtract if negative) an offset for all timestamps of the
262 previous data file specified with -i. This is useful to merge sort
263 multiple trace.dat files where the difference in the timestamp is
264 known. For example if a trace is done on a virtual guest, and
265 another trace is done on the host. If the host timestamp is 1000
266 units ahead of the guest, the following can be done:
267 trace-cmd report -i host.dat --ts-offset -1000 -i guest.dat
269 This will subtract 1000 timestamp units from all the host events as it merges
271 with the guest.dat events. Note, the units is for the raw units recorded in
272 the trace. If the units are nanoseconds, the addition (or subtraction) from
273 the offset will be nanoseconds even if the displayed units are microseconds.
274 --ts2secs HZ
276 Convert the current clock source into a second (nanosecond
277 resolution) output. When using clocks like x86-tsc, if the
278 frequency is known, by passing in the clock frequency, this will
279 convert the time to seconds.
280 This option affects any trace.dat file given with *-i* proceeding it.
282 If this option comes before any *-i* option, then that value becomes
283 the default conversion for all other trace.dat files. If another
284 --ts2secs option appears after a *-i* trace.dat file, than that option
285 will override the default value.
286 Example: On a 3.4 GHz machine
288 trace-cmd record -p function -C x86-tsc
290 trace-cmd report --ts2ns 3400000000
292 The report will convert the cycles timestamps into a readable second
294 display. The default display resolution is microseconds, unless *-t*
295 is used.
296 The value of --ts-offset must still be in the raw timestamp units, even
298 with this option. The offset will be converted as well.
299 --ts-diff
301 Show the time differences between events. The difference will
302 appear in parenthesis just after the timestamp.
303 --ts-check
305 Make sure no timestamp goes backwards, and if it does, print out a
306 warning message of the fact.
307
309 Using a trace.dat file that was created with:
310 # trace-cmd record -p function -e all sleep 5
312 The default report shows:
314 # trace-cmd report
316 trace-cmd-16129 [002] 158126.498411: function: __mutex_unlock_slowpath <-- mutex_unlock
317 trace-cmd-16131 [000] 158126.498411: kmem_cache_alloc: call_site=811223c5 ptr=0xffff88003ecf2b40 bytes_req=272 bytes_alloc=320 gfp_flags=GFP_KERNEL|GFP_ZERO
318 trace-cmd-16130 [003] 158126.498411: function: do_splice_to <-- sys_splice
319 sleep-16133 [001] 158126.498412: function: inotify_inode_queue_event <-- vfs_write
320 trace-cmd-16129 [002] 158126.498420: lock_release: 0xffff88003f1fa4f8 &sb->s_type->i_mutex_key
321 trace-cmd-16131 [000] 158126.498421: function: security_file_alloc <-- get_empty_filp
322 sleep-16133 [001] 158126.498422: function: __fsnotify_parent <-- vfs_write
323 trace-cmd-16130 [003] 158126.498422: function: rw_verify_area <-- do_splice_to
324 trace-cmd-16131 [000] 158126.498424: function: cap_file_alloc_security <-- security_file_alloc
325 trace-cmd-16129 [002] 158126.498425: function: syscall_trace_leave <-- int_check_syscall_exit_work
326 sleep-16133 [001] 158126.498426: function: inotify_dentry_parent_queue_event <-- vfs_write
327 trace-cmd-16130 [003] 158126.498426: function: security_file_permission <-- rw_verify_area
328 trace-cmd-16129 [002] 158126.498428: function: audit_syscall_exit <-- syscall_trace_leave
329 [...]
330 To see everything but the function traces:
332 # trace-cmd report -v -F 'function'
334 trace-cmd-16131 [000] 158126.498411: kmem_cache_alloc: call_site=811223c5 ptr=0xffff88003ecf2b40 bytes_req=272 bytes_alloc=320 gfp_flags=GFP_KERNEL|GFP_ZERO
335 trace-cmd-16129 [002] 158126.498420: lock_release: 0xffff88003f1fa4f8 &sb->s_type->i_mutex_key
336 trace-cmd-16130 [003] 158126.498436: lock_acquire: 0xffffffff8166bf78 read all_cpu_access_lock
337 trace-cmd-16131 [000] 158126.498438: lock_acquire: 0xffff88003df5b520 read &fs->lock
338 trace-cmd-16129 [002] 158126.498446: kfree: call_site=810a7abb ptr=0x0
339 trace-cmd-16130 [003] 158126.498448: lock_acquire: 0xffff880002250a80 &per_cpu(cpu_access_lock, cpu)
340 trace-cmd-16129 [002] 158126.498450: sys_exit_splice: 0xfffffff5
341 trace-cmd-16131 [000] 158126.498454: lock_release: 0xffff88003df5b520 &fs->lock
342 sleep-16133 [001] 158126.498456: kfree: call_site=810a7abb ptr=0x0
343 sleep-16133 [001] 158126.498460: sys_exit_write: 0x1
344 trace-cmd-16130 [003] 158126.498462: kmalloc: call_site=810bf95b ptr=0xffff88003dedc040 bytes_req=24 bytes_alloc=32 gfp_flags=GFP_KERNEL|GFP_ZERO
345 To see only the kmalloc calls that were greater than 1000 bytes:
347 #trace-cmd report -F 'kmalloc: bytes_req > 1000'
349 <idle>-0 [000] 158128.126641: kmalloc: call_site=81330635 ptr=0xffff88003c2fd000 bytes_req=2096 bytes_alloc=4096 gfp_flags=GFP_ATOMIC
350 To see wakeups and sched switches that left the previous task in the
352 running state:
353 # trace-cmd report -F 'sched: prev_state == 0 || (success == 1)'
355 trace-cmd-16132 [002] 158126.499951: sched_wakeup: comm=trace-cmd pid=16129 prio=120 success=1 target_cpu=002
356 trace-cmd-16132 [002] 158126.500401: sched_switch: prev_comm=trace-cmd prev_pid=16132 prev_prio=120 prev_state=R ==> next_comm=trace-cmd next_pid=16129 next_prio=120
357 <idle>-0 [003] 158126.500585: sched_wakeup: comm=trace-cmd pid=16130 prio=120 success=1 target_cpu=003
358 <idle>-0 [003] 158126.501241: sched_switch: prev_comm=swapper prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=trace-cmd next_pid=16130 next_prio=120
359 trace-cmd-16132 [000] 158126.502475: sched_wakeup: comm=trace-cmd pid=16131 prio=120 success=1 target_cpu=000
360 trace-cmd-16131 [002] 158126.506516: sched_wakeup: comm=trace-cmd pid=16129 prio=120 success=1 target_cpu=002
361 <idle>-0 [003] 158126.550110: sched_switch: prev_comm=swapper prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=trace-cmd next_pid=16130 next_prio=120
362 trace-cmd-16131 [003] 158126.570243: sched_wakeup: comm=trace-cmd pid=16129 prio=120 success=1 target_cpu=003
363 trace-cmd-16130 [002] 158126.618202: sched_switch: prev_comm=trace-cmd prev_pid=16130 prev_prio=120 prev_state=R ==> next_comm=yum-updatesd next_pid=3088 next_prio=1 20
364 trace-cmd-16129 [003] 158126.622379: sched_wakeup: comm=trace-cmd pid=16131 prio=120 success=1 target_cpu=003
365 trace-cmd-16129 [000] 158126.649287: sched_wakeup: comm=trace-cmd pid=16131 prio=120 success=1 target_cpu=000
366 The above needs a little explanation. The filter specifies the "sched"
368 subsystem, which includes both sched_switch and sched_wakeup events.
369 Any event that does not have the format field "prev_state" or
370 "success", will evaluate those expressions as FALSE, and will not
371 produce a match. Using "||" will have the "prev_state" test happen for
372 the "sched_switch" event and the "success" test happen for the
373 "sched_wakeup" event.
374 # trace-cmd report -w -F 'sched_switch, sched_wakeup.*'
376 [...]
377 trace-cmd-16130 [003] 158131.580616: sched_wakeup: comm=trace-cmd pid=16131 prio=120 success=1 target_cpu=003
378 trace-cmd-16129 [000] 158131.581502: sched_switch: prev_comm=trace-cmd prev_pid=16129 prev_prio=120 prev_state=S ==> next_comm=trace-cmd next_pid=16131 next_prio=120 Latency: 885.901 usecs
379 trace-cmd-16131 [000] 158131.582414: sched_wakeup: comm=trace-cmd pid=16129 prio=120 success=1 target_cpu=000
380 trace-cmd-16132 [001] 158131.583219: sched_switch: prev_comm=trace-cmd prev_pid=16132 prev_prio=120 prev_state=S ==> next_comm=trace-cmd next_pid=16129 next_prio=120 Latency: 804.809 usecs
381 sleep-16133 [002] 158131.584121: sched_wakeup: comm=trace-cmd pid=16120 prio=120 success=1 target_cpu=002
382 trace-cmd-16129 [001] 158131.584128: sched_wakeup: comm=trace-cmd pid=16132 prio=120 success=1 target_cpu=001
383 sleep-16133 [002] 158131.584275: sched_switch: prev_comm=sleep prev_pid=16133 prev_prio=120 prev_state=R ==> next_comm=trace-cmd next_pid=16120 next_prio=120 Latency: 153.915 usecs
384 trace-cmd-16130 [003] 158131.585284: sched_switch: prev_comm=trace-cmd prev_pid=16130 prev_prio=120 prev_state=S ==> next_comm=trace-cmd next_pid=16132 next_prio=120 Latency: 1155.677 usecs
385 Average wakeup latency: 26626.656 usecs
387 The above trace produces the wakeup latencies of the tasks. The
389 "sched_switch" event reports each individual latency after writing the
390 event information. At the end of the report, the average wakeup latency
391 is reported.
392 # trace-cmd report -w -F 'sched_switch, sched_wakeup.*: prio < 100 || next_prio < 100'
394 <idle>-0 [003] 158131.516753: sched_wakeup: comm=ksoftirqd/3 pid=13 prio=49 success=1 target_cpu=003
395 <idle>-0 [003] 158131.516855: sched_switch: prev_comm=swapper prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=ksoftirqd/3 next_pid=13 next_prio=49 Latency: 101.244 usecs
396 <idle>-0 [003] 158131.533781: sched_wakeup: comm=ksoftirqd/3 pid=13 prio=49 success=1 target_cpu=003
397 <idle>-0 [003] 158131.533897: sched_switch: prev_comm=swapper prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=ksoftirqd/3 next_pid=13 next_prio=49 Latency: 115.608 usecs
398 <idle>-0 [003] 158131.569730: sched_wakeup: comm=ksoftirqd/3 pid=13 prio=49 success=1 target_cpu=003
399 <idle>-0 [003] 158131.569851: sched_switch: prev_comm=swapper prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=ksoftirqd/3 next_pid=13 next_prio=49 Latency: 121.024 usecs
400 Average wakeup latency: 110.021 usecs
402 The above version will only show the wakeups and context switches of
404 Real Time tasks. The prio used inside the kernel starts at 0 for
405 highest priority. That is prio 0 is equivalent to user space real time
406 priority 99, and priority 98 is equivalent to user space real time
407 priority 1. Prios less than 100 represent Real Time tasks.
408 An example of the profile:
410 # trace-cmd record --profile sleep 1
412 # trace-cmd report --profile --comm sleep
413 task: sleep-21611
414 Event: sched_switch:R (1) Total: 99442 Avg: 99442 Max: 99442 Min:99442
415 <stack> 1 total:99442 min:99442 max:99442 avg=99442
416 => ftrace_raw_event_sched_switch (0xffffffff8105f812)
417 => __schedule (0xffffffff8150810a)
418 => preempt_schedule (0xffffffff8150842e)
419 => ___preempt_schedule (0xffffffff81273354)
420 => cpu_stop_queue_work (0xffffffff810b03c5)
421 => stop_one_cpu (0xffffffff810b063b)
422 => sched_exec (0xffffffff8106136d)
423 => do_execve_common.isra.27 (0xffffffff81148c89)
424 => do_execve (0xffffffff811490b0)
425 => SyS_execve (0xffffffff811492c4)
426 => return_to_handler (0xffffffff8150e3c8)
427 => stub_execve (0xffffffff8150c699)
428 Event: sched_switch:S (1) Total: 1000506680 Avg: 1000506680 Max: 1000506680 Min:1000506680
429 <stack> 1 total:1000506680 min:1000506680 max:1000506680 avg=1000506680
430 => ftrace_raw_event_sched_switch (0xffffffff8105f812)
431 => __schedule (0xffffffff8150810a)
432 => schedule (0xffffffff815084b8)
433 => do_nanosleep (0xffffffff8150b22c)
434 => hrtimer_nanosleep (0xffffffff8108d647)
435 => SyS_nanosleep (0xffffffff8108d72c)
436 => return_to_handler (0xffffffff8150e3c8)
437 => tracesys_phase2 (0xffffffff8150c304)
438 Event: sched_wakeup:21611 (1) Total: 30326 Avg: 30326 Max: 30326 Min:30326
439 <stack> 1 total:30326 min:30326 max:30326 avg=30326
440 => ftrace_raw_event_sched_wakeup_template (0xffffffff8105f653)
441 => ttwu_do_wakeup (0xffffffff810606eb)
442 => ttwu_do_activate.constprop.124 (0xffffffff810607c8)
443 => try_to_wake_up (0xffffffff8106340a)
444
446 trace-cmd(1), trace-cmd-record(1), trace-cmd-start(1),
447 trace-cmd-stop(1), trace-cmd-extract(1), trace-cmd-reset(1),
448 trace-cmd-split(1), trace-cmd-list(1), trace-cmd-listen(1),
449 trace-cmd-profile(1)
450
452 Written by Steven Rostedt, <rostedt@goodmis.org[1]>
453
455 https://git.kernel.org/pub/scm/utils/trace-cmd/trace-cmd.git/
456
458 Copyright (C) 2010 Red Hat, Inc. Free use of this software is granted
459 under the terms of the GNU Public License (GPL).
460
462 1. rostedt@goodmis.org
463 mailto:rostedt@goodmis.org
464 07/23/2021 TRACE-CMD-REPORT(1)