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 functions that have been recorded in the
28 file.
29 -P
31 This outputs the list of "trace_printk()" data. The raw trace data
32 points to static pointers in the kernel. This must be stored in the
33 trace.dat file.
34 -E
36 This lists the possible events in the file (but this list is not
37 necessarily the list of events in the file).
38 --events
40 This will list the event formats that are stored in the trace.dat
41 file.
42 --event regex
44 This will print events that match the given regex. If a colon is
45 specified, then the characters before the colon will be used to
46 match the system and the characters after the colon will match the
47 event.
48 trace-cmd report --event sys:read
50 The above will only match events where the system name contains "sys"
52 and the event name contains "read".
53 trace-cmd report --event read
55 The above will match all events that contain "read" in its name. Also it
57 may list all events of a system that contains "read" as well.
58 --check-events
60 This will parse the event format strings that are stored in the
61 trace.dat file and return whether the formats can be parsed
62 correctly. It will load plugins unless -N is specified.
63 -t
65 Print the full timestamp. The timestamps in the data file are
66 usually recorded to the nanosecond. But the default display of the
67 timestamp is only to the microsecond. To see the full timestamp,
68 add the -t option.
69 -F filter
71 Add a filter to limit what events are displayed. The format of the
72 filter is:
73 <events> ':' <filter>
75 <events> = SYSTEM'/'EVENT | SYSTEM | EVENT | <events> ',' <events>
76 <filter> = EVENT_FIELD <op> <value> | <filter> '&&' <filter> |
77 <filter> '||' <filter> | '(' <filter> ')' | '!' <filter>
78 <op> = '==' | '!=' | '>=' | '<=' | '>' | '<' | '&' | '|' | '^' |
79 '+' | '-' | '*' | '/' | '%'
80 <value> = NUM | STRING | EVENT_FIELD
81 SYSTEM is the name of the system to filter on. If the EVENT is left out,
83 then it applies to all events under the SYSTEM. If only one string is used
84 without the '/' to deliminate between SYSTEM and EVENT, then the filter
85 will be applied to all systems and events that match the given string.
86 Whitespace is ignored, such that "sched:next_pid==123" is equivalent to
88 "sched : next_pid == 123".
89 STRING is defined with single or double quotes (single quote must end with
91 single quote, and double with double). Whitespace within quotes are not
92 ignored.
93 The representation of a SYSTEM or EVENT may also be a regular expression
95 as defined by 'regcomp(3)'.
96 The EVENT_FIELD is the name of the field of an event that is being
98 filtered. If the event does not contain the EVENT_FIELD, that part of the
99 equation will be considered false.
100 -F 'sched : bogus == 1 || common_pid == 2'
102 The "bogus == 1" will always evaluate to FALSE because no event has a
104 field called "bogus", but the "common_pid == 2" will still be evaluated
105 since all events have the field "common_pid". Any "sched" event that was
106 traced by the process with the PID of 2 will be shown.
107 Note, the EVENT_FIELD is the field name as shown by an events format
109 (as displayed with *--events*), and not what is found in the output.
110 If the output shows "ID:foo" but the field that "foo" belongs to was
111 called "name" in the event format, then "name" must be used in the filter.
112 The same is true about values. If the value that is displayed is converted
113 by to a string symbol, the filter checks the original value and not the
114 value displayed. For example, to filter on all tasks that were in the
115 running state at a context switch:
116 -F 'sched/sched_switch : prev_state==0'
118 Although the output displays 'R', having 'prev_stat=="R"' will not work.
120 Note: You can also specify 'COMM' as an EVENT_FIELD. This will use the
122 task name (or comm) of the record to compare. For example, to filter out
123 all of the "trace-cmd" tasks:
124 -F '.*:COMM != "trace-cmd"'
126 -I
128 Do not print events where the HARDIRQ latency flag is set. This
129 will filter out most events that are from interrupt context. Note,
130 it may not filter out function traced functions that are in
131 interrupt context but were called before the kernel "in interrupt"
132 flag was set.
133 -S
135 Do not print events where the SOFTIRQ latency flag is set. This
136 will filter out most events that are from soft interrupt context.
137 -v
139 This causes the following filters of -F to filter out the matching
140 events.
141 -v -F 'sched/sched_switch : prev_state == 0'
143 Will not display any sched_switch events that have a prev_state of 0.
145 Removing the *-v* will only print out those events.
146 -T
148 Test the filters of -F. After processing a filter string, the
149 resulting filter will be displayed for each event. This is useful
150 for using a filter for more than one event where a field may not
151 exist in all events. Also it can be used to make sure there are no
152 misspelled event field names, as they will simply be ignored. -T
153 is ignored if -F is not specified.
154 -V
156 Show the plugins that are loaded.
157 -L
159 This will not load system wide plugins. It loads "local only". That
160 is what it finds in the ~/.trace-cmd/plugins directory.
161 -N
163 This will not load any plugins.
164 -n event-re
166 This will cause all events that match the option to ignore any
167 registered handler (by the plugins) to print the event. The normal
168 event will be printed instead. The event-re is a regular expression
169 as defined by regcomp(3).
170 --profile
172 With the --profile option, "trace-cmd report" will process all the
173 events first, and then output a format showing where tasks have
174 spent their time in the kernel, as well as where they are blocked
175 the most, and where wake up latencies are.
176 See trace-cmd-profile(1) for more details and examples.
178 -G
180 Set interrupt (soft and hard) events as global (associated to CPU
181 instead of tasks). Only works for --profile.
182 -H event-hooks
184 Add custom event matching to connect any two events together.
185 See trace-cmd-profile(1) for format.
187 -R
189 This will show the events in "raw" format. That is, it will ignore
190 the event’s print formatting and just print the contents of each
191 field.
192 -r event-re
194 This will cause all events that match the option to print its raw
195 fields. The event-re is a regular expression as defined by
196 regcomp(3).
197 -l
199 This adds a "latency output" format. Information about interrupts
200 being disabled, soft irq being disabled, the "need_resched" flag
201 being set, preempt count, and big kernel lock are all being
202 recorded with every event. But the default display does not show
203 this information. This option will set display this information
204 with 6 characters. When one of the fields is zero or N/A a '.\' is
205 shown.
206 <idle>-0 0d.h1. 106467.859747: function: ktime_get <-- tick_check_idle
208 The 0d.h1. denotes this information. The first character is never a '.'
210 and represents what CPU the trace was recorded on (CPU 0). The 'd' denotes
211 that interrupts were disabled. The 'h' means that this was called inside
212 an interrupt handler. The '1' is the preemption disabled (preempt_count)
213 was set to one. The two '.'s are "need_resched" flag and kernel lock
214 counter. If the "need_resched" flag is set, then that character would be a
215 'N'.
216 -w
218 If both the sched_switch and sched_wakeup events are enabled, then
219 this option will report the latency between the time the task was
220 first woken, and the time it was scheduled in.
221 -q
223 Quiet non critical warnings.
224 -O
226 Pass options to the trace-cmd plugins that are loaded.
227 -O plugin:var=value
229 The 'plugin:' and '=value' are optional. Value may be left off for options
231 that are boolean. If the 'plugin:' is left off, then any variable that matches
232 in all plugins will be set.
233 Example: -O fgraph:tailprint
235 --stat
237 If the trace.dat file recorded the final stats (outputed at the end
238 of record) the --stat option can be used to retrieve them.
239 --uname
241 If the trace.dat file recorded uname during the run, this will
242 retrieve that information.
243 --ts-offset offset
245 Add (or subtract if negative) an offset for all timestamps of the
246 previous data file specified with -i. This is useful to merge sort
247 multiple trace.dat files where the difference in the timestamp is
248 known. For example if a trace is done on a virtual guest, and
249 another trace is done on the host. If the host timestamp is 1000
250 units ahead of the guest, the following can be done:
251 trace-cmd report -i host.dat --ts-offset -1000 -i guest.dat
253 This will subtract 1000 timestamp units from all the host events as it merges
255 with the guest.dat events. Note, the units is for the raw units recorded in
256 the trace. If the units are nanoseconds, the addition (or subtraction) from
257 the offset will be nanoseconds even if the displayed units are microseconds.
258 --ts2secs HZ
260 Convert the current clock source into a second (nanosecond
261 resolution) output. When using clocks like x86-tsc, if the
262 frequency is known, by passing in the clock frequency, this will
263 convert the time to seconds.
264 This option affects any trace.dat file given with *-i* proceeding it.
266 If this option comes before any *-i* option, then that value becomes
267 the default conversion for all other trace.dat files. If another
268 --ts2secs option appears after a *-i* trace.dat file, than that option
269 will override the default value.
270 Example: On a 3.4 GHz machine
272 trace-cmd record -p function -C x86-tsc
274 trace-cmd report --ts2ns 3400000000
276 The report will convert the cycles timestamps into a readable second
278 display. The default display resolution is microseconds, unless *-t*
279 is used.
280 The value of --ts-offset must still be in the raw timestamp units, even
282 with this option. The offset will be converted as well.
283 --ts-diff
285 Show the time differences between events. The difference will
286 appear in parenthesis just after the timestamp.
287
289 Using a trace.dat file that was created with:
290 # trace-cmd record -p function -e all sleep 5
292 The default report shows:
294 # trace-cmd report
296 trace-cmd-16129 [002] 158126.498411: function: __mutex_unlock_slowpath <-- mutex_unlock
297 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
298 trace-cmd-16130 [003] 158126.498411: function: do_splice_to <-- sys_splice
299 sleep-16133 [001] 158126.498412: function: inotify_inode_queue_event <-- vfs_write
300 trace-cmd-16129 [002] 158126.498420: lock_release: 0xffff88003f1fa4f8 &sb->s_type->i_mutex_key
301 trace-cmd-16131 [000] 158126.498421: function: security_file_alloc <-- get_empty_filp
302 sleep-16133 [001] 158126.498422: function: __fsnotify_parent <-- vfs_write
303 trace-cmd-16130 [003] 158126.498422: function: rw_verify_area <-- do_splice_to
304 trace-cmd-16131 [000] 158126.498424: function: cap_file_alloc_security <-- security_file_alloc
305 trace-cmd-16129 [002] 158126.498425: function: syscall_trace_leave <-- int_check_syscall_exit_work
306 sleep-16133 [001] 158126.498426: function: inotify_dentry_parent_queue_event <-- vfs_write
307 trace-cmd-16130 [003] 158126.498426: function: security_file_permission <-- rw_verify_area
308 trace-cmd-16129 [002] 158126.498428: function: audit_syscall_exit <-- syscall_trace_leave
309 [...]
310 To see everything but the function traces:
312 # trace-cmd report -v -F 'function'
314 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
315 trace-cmd-16129 [002] 158126.498420: lock_release: 0xffff88003f1fa4f8 &sb->s_type->i_mutex_key
316 trace-cmd-16130 [003] 158126.498436: lock_acquire: 0xffffffff8166bf78 read all_cpu_access_lock
317 trace-cmd-16131 [000] 158126.498438: lock_acquire: 0xffff88003df5b520 read &fs->lock
318 trace-cmd-16129 [002] 158126.498446: kfree: call_site=810a7abb ptr=0x0
319 trace-cmd-16130 [003] 158126.498448: lock_acquire: 0xffff880002250a80 &per_cpu(cpu_access_lock, cpu)
320 trace-cmd-16129 [002] 158126.498450: sys_exit_splice: 0xfffffff5
321 trace-cmd-16131 [000] 158126.498454: lock_release: 0xffff88003df5b520 &fs->lock
322 sleep-16133 [001] 158126.498456: kfree: call_site=810a7abb ptr=0x0
323 sleep-16133 [001] 158126.498460: sys_exit_write: 0x1
324 trace-cmd-16130 [003] 158126.498462: kmalloc: call_site=810bf95b ptr=0xffff88003dedc040 bytes_req=24 bytes_alloc=32 gfp_flags=GFP_KERNEL|GFP_ZERO
325 To see only the kmalloc calls that were greater than 1000 bytes:
327 #trace-cmd report -F 'kmalloc: bytes_req > 1000'
329 <idle>-0 [000] 158128.126641: kmalloc: call_site=81330635 ptr=0xffff88003c2fd000 bytes_req=2096 bytes_alloc=4096 gfp_flags=GFP_ATOMIC
330 To see wakeups and sched switches that left the previous task in the
332 running state:
333 # trace-cmd report -F 'sched: prev_state == 0 || (success == 1)'
335 trace-cmd-16132 [002] 158126.499951: sched_wakeup: comm=trace-cmd pid=16129 prio=120 success=1 target_cpu=002
336 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
337 <idle>-0 [003] 158126.500585: sched_wakeup: comm=trace-cmd pid=16130 prio=120 success=1 target_cpu=003
338 <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
339 trace-cmd-16132 [000] 158126.502475: sched_wakeup: comm=trace-cmd pid=16131 prio=120 success=1 target_cpu=000
340 trace-cmd-16131 [002] 158126.506516: sched_wakeup: comm=trace-cmd pid=16129 prio=120 success=1 target_cpu=002
341 <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
342 trace-cmd-16131 [003] 158126.570243: sched_wakeup: comm=trace-cmd pid=16129 prio=120 success=1 target_cpu=003
343 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
344 trace-cmd-16129 [003] 158126.622379: sched_wakeup: comm=trace-cmd pid=16131 prio=120 success=1 target_cpu=003
345 trace-cmd-16129 [000] 158126.649287: sched_wakeup: comm=trace-cmd pid=16131 prio=120 success=1 target_cpu=000
346 The above needs a little explanation. The filter specifies the "sched"
348 subsystem, which includes both sched_switch and sched_wakeup events.
349 Any event that does not have the format field "prev_state" or
350 "success", will evaluate those expressions as FALSE, and will not
351 produce a match. Using "||" will have the "prev_state" test happen for
352 the "sched_switch" event and the "success" test happen for the
353 "sched_wakeup" event.
354 # trace-cmd report -w -F 'sched_switch, sched_wakeup.*'
356 [...]
357 trace-cmd-16130 [003] 158131.580616: sched_wakeup: comm=trace-cmd pid=16131 prio=120 success=1 target_cpu=003
358 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
359 trace-cmd-16131 [000] 158131.582414: sched_wakeup: comm=trace-cmd pid=16129 prio=120 success=1 target_cpu=000
360 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
361 sleep-16133 [002] 158131.584121: sched_wakeup: comm=trace-cmd pid=16120 prio=120 success=1 target_cpu=002
362 trace-cmd-16129 [001] 158131.584128: sched_wakeup: comm=trace-cmd pid=16132 prio=120 success=1 target_cpu=001
363 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
364 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
365 Average wakeup latency: 26626.656 usecs
367 The above trace produces the wakeup latencies of the tasks. The
369 "sched_switch" event reports each individual latency after writing the
370 event information. At the end of the report, the average wakeup latency
371 is reported.
372 # trace-cmd report -w -F 'sched_switch, sched_wakeup.*: prio < 100 || next_prio < 100'
374 <idle>-0 [003] 158131.516753: sched_wakeup: comm=ksoftirqd/3 pid=13 prio=49 success=1 target_cpu=003
375 <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
376 <idle>-0 [003] 158131.533781: sched_wakeup: comm=ksoftirqd/3 pid=13 prio=49 success=1 target_cpu=003
377 <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
378 <idle>-0 [003] 158131.569730: sched_wakeup: comm=ksoftirqd/3 pid=13 prio=49 success=1 target_cpu=003
379 <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
380 Average wakeup latency: 110.021 usecs
382 The above version will only show the wakeups and context switches of
384 Real Time tasks. The prio used inside the kernel starts at 0 for
385 highest priority. That is prio 0 is equivalent to user space real time
386 priority 99, and priority 98 is equivalent to user space real time
387 priority 1. Prios less than 100 represent Real Time tasks.
388 An example of the profile:
390 # trace-cmd record --profile sleep 1
392 # trace-cmd report --profile --comm sleep
393 task: sleep-21611
394 Event: sched_switch:R (1) Total: 99442 Avg: 99442 Max: 99442 Min:99442
395 <stack> 1 total:99442 min:99442 max:99442 avg=99442
396 => ftrace_raw_event_sched_switch (0xffffffff8105f812)
397 => __schedule (0xffffffff8150810a)
398 => preempt_schedule (0xffffffff8150842e)
399 => ___preempt_schedule (0xffffffff81273354)
400 => cpu_stop_queue_work (0xffffffff810b03c5)
401 => stop_one_cpu (0xffffffff810b063b)
402 => sched_exec (0xffffffff8106136d)
403 => do_execve_common.isra.27 (0xffffffff81148c89)
404 => do_execve (0xffffffff811490b0)
405 => SyS_execve (0xffffffff811492c4)
406 => return_to_handler (0xffffffff8150e3c8)
407 => stub_execve (0xffffffff8150c699)
408 Event: sched_switch:S (1) Total: 1000506680 Avg: 1000506680 Max: 1000506680 Min:1000506680
409 <stack> 1 total:1000506680 min:1000506680 max:1000506680 avg=1000506680
410 => ftrace_raw_event_sched_switch (0xffffffff8105f812)
411 => __schedule (0xffffffff8150810a)
412 => schedule (0xffffffff815084b8)
413 => do_nanosleep (0xffffffff8150b22c)
414 => hrtimer_nanosleep (0xffffffff8108d647)
415 => SyS_nanosleep (0xffffffff8108d72c)
416 => return_to_handler (0xffffffff8150e3c8)
417 => tracesys_phase2 (0xffffffff8150c304)
418 Event: sched_wakeup:21611 (1) Total: 30326 Avg: 30326 Max: 30326 Min:30326
419 <stack> 1 total:30326 min:30326 max:30326 avg=30326
420 => ftrace_raw_event_sched_wakeup_template (0xffffffff8105f653)
421 => ttwu_do_wakeup (0xffffffff810606eb)
422 => ttwu_do_activate.constprop.124 (0xffffffff810607c8)
423 => try_to_wake_up (0xffffffff8106340a)
424
426 trace-cmd(1), trace-cmd-record(1), trace-cmd-start(1),
427 trace-cmd-stop(1), trace-cmd-extract(1), trace-cmd-reset(1),
428 trace-cmd-split(1), trace-cmd-list(1), trace-cmd-listen(1),
429 trace-cmd-profile(1)
430
432 Written by Steven Rostedt, <rostedt@goodmis.org[1]>
433
435 git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/trace-cmd.git
436
438 Copyright (C) 2010 Red Hat, Inc. Free use of this software is granted
439 under the terms of the GNU Public License (GPL).
440
442 1. rostedt@goodmis.org
443 mailto:rostedt@goodmis.org
444 07/18/2018 TRACE-CMD-REPORT(1)