1 perf-trace(1) 1 perf-trace(1)
2 ============= 2 =============
3 3
4 NAME 4 NAME
5 ---- 5 ----
6 perf-trace - strace inspired tool 6 perf-trace - strace inspired tool
7 7
8 SYNOPSIS 8 SYNOPSIS
9 -------- 9 --------
10 [verse] 10 [verse]
11 'perf trace' 11 'perf trace'
12 'perf trace record' 12 'perf trace record'
13 13
14 DESCRIPTION 14 DESCRIPTION
15 ----------- 15 -----------
16 This command will show the events associated w 16 This command will show the events associated with the target, initially
17 syscalls, but other system events like pagefau 17 syscalls, but other system events like pagefaults, task lifetime events,
18 scheduling events, etc. 18 scheduling events, etc.
19 19
20 This is a live mode tool in addition to workin 20 This is a live mode tool in addition to working with perf.data files like
21 the other perf tools. Files can be generated u 21 the other perf tools. Files can be generated using the 'perf record' command
22 but the session needs to include the raw_sysca 22 but the session needs to include the raw_syscalls events (-e 'raw_syscalls:*').
23 Alternatively, 'perf trace record' can be used 23 Alternatively, 'perf trace record' can be used as a shortcut to
24 automatically include the raw_syscalls events 24 automatically include the raw_syscalls events when writing events to a file.
25 25
26 The following options apply to perf trace; opt 26 The following options apply to perf trace; options to perf trace record are
27 found in the perf record man page. 27 found in the perf record man page.
28 28
29 OPTIONS 29 OPTIONS
30 ------- 30 -------
31 31
32 -a:: 32 -a::
33 --all-cpus:: 33 --all-cpus::
34 System-wide collection from all CPUs. 34 System-wide collection from all CPUs.
35 35
36 -e:: 36 -e::
37 --expr:: 37 --expr::
38 --event:: 38 --event::
39 List of syscalls and other perf events 39 List of syscalls and other perf events (tracepoints, HW cache events,
40 etc) to show. Globbing is supported, e 40 etc) to show. Globbing is supported, e.g.: "epoll_*", "*msg*", etc.
41 See 'perf list' for a complete list of 41 See 'perf list' for a complete list of events.
42 Prefixing with ! shows all syscalls bu 42 Prefixing with ! shows all syscalls but the ones specified. You may
43 need to escape it. 43 need to escape it.
44 44
45 --filter=<filter>:: 45 --filter=<filter>::
46 Event filter. This option should follo 46 Event filter. This option should follow an event selector (-e) which
47 selects tracepoint event(s). 47 selects tracepoint event(s).
48 48
49 49
50 -D msecs:: 50 -D msecs::
51 --delay msecs:: 51 --delay msecs::
52 After starting the program, wait msecs before 52 After starting the program, wait msecs before measuring. This is useful to
53 filter out the startup phase of the program, w 53 filter out the startup phase of the program, which is often very different.
54 54
55 -o:: 55 -o::
56 --output=:: 56 --output=::
57 Output file name. 57 Output file name.
58 58
59 -p:: 59 -p::
60 --pid=:: 60 --pid=::
61 Record events on existing process ID ( 61 Record events on existing process ID (comma separated list).
62 62
63 -t:: 63 -t::
64 --tid=:: 64 --tid=::
65 Record events on existing thread ID (c 65 Record events on existing thread ID (comma separated list).
66 66
67 -u:: 67 -u::
68 --uid=:: 68 --uid=::
69 Record events in threads owned by uid. 69 Record events in threads owned by uid. Name or number.
70 70
71 -G:: 71 -G::
72 --cgroup:: 72 --cgroup::
73 Record events in threads in a cgroup. 73 Record events in threads in a cgroup.
74 74
75 Look for cgroups to set at the /sys/fs 75 Look for cgroups to set at the /sys/fs/cgroup/perf_event directory, then
76 remove the /sys/fs/cgroup/perf_event/ 76 remove the /sys/fs/cgroup/perf_event/ part and try:
77 77
78 perf trace -G A -e sched:*swit 78 perf trace -G A -e sched:*switch
79 79
80 Will set all raw_syscalls:sys_{enter,e 80 Will set all raw_syscalls:sys_{enter,exit}, pgfault, vfs_getname, etc
81 _and_ sched:sched_switch to the 'A' cg 81 _and_ sched:sched_switch to the 'A' cgroup, while:
82 82
83 perf trace -e sched:*switch -G 83 perf trace -e sched:*switch -G A
84 84
85 will only set the sched:sched_switch e 85 will only set the sched:sched_switch event to the 'A' cgroup, all the
86 other events (raw_syscalls:sys_{enter, 86 other events (raw_syscalls:sys_{enter,exit}, etc are left "without"
87 a cgroup (on the root cgroup, sys wide 87 a cgroup (on the root cgroup, sys wide, etc).
88 88
89 Multiple cgroups: 89 Multiple cgroups:
90 90
91 perf trace -G A -e sched:*swit 91 perf trace -G A -e sched:*switch -G B
92 92
93 the syscall ones go to the 'A' cgroup, 93 the syscall ones go to the 'A' cgroup, the sched:sched_switch goes
94 to the 'B' cgroup. 94 to the 'B' cgroup.
95 95
96 --filter-pids=:: 96 --filter-pids=::
97 Filter out events for these pids and f 97 Filter out events for these pids and for 'trace' itself (comma separated list).
98 98
99 -v:: 99 -v::
100 --verbose:: 100 --verbose::
101 Increase the verbosity level. 101 Increase the verbosity level.
102 102
103 --no-inherit:: 103 --no-inherit::
104 Child tasks do not inherit counters. 104 Child tasks do not inherit counters.
105 105
106 -m:: 106 -m::
107 --mmap-pages=:: 107 --mmap-pages=::
108 Number of mmap data pages (must be a p 108 Number of mmap data pages (must be a power of two) or size
109 specification in bytes with appended u !! 109 specification with appended unit character - B/K/M/G. The
110 The size is rounded up to the nearest !! 110 size is rounded up to have nearest pages power of two value.
111 111
112 -C:: 112 -C::
113 --cpu:: 113 --cpu::
114 Collect samples only on the list of CPUs provi 114 Collect samples only on the list of CPUs provided. Multiple CPUs can be provided as a
115 comma-separated list with no space: 0,1. Range 115 comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2.
116 In per-thread mode with inheritance mode on (d 116 In per-thread mode with inheritance mode on (default), Events are captured only when
117 the thread executes on the designated CPUs. De 117 the thread executes on the designated CPUs. Default is to monitor all CPUs.
118 118
119 --duration:: 119 --duration::
120 Show only events that had a duration g 120 Show only events that had a duration greater than N.M ms.
121 121
122 --sched:: 122 --sched::
123 Accrue thread runtime and provide a su 123 Accrue thread runtime and provide a summary at the end of the session.
124 124
125 --failure:: 125 --failure::
126 Show only syscalls that failed, i.e. t 126 Show only syscalls that failed, i.e. that returned < 0.
127 127
128 -i:: 128 -i::
129 --input:: 129 --input::
130 Process events from a given perf data 130 Process events from a given perf data file.
131 131
132 -T:: 132 -T::
133 --time:: 133 --time::
134 Print full timestamp rather time relat 134 Print full timestamp rather time relative to first sample.
135 135
136 --comm:: 136 --comm::
137 Show process COMM right beside its ID, 137 Show process COMM right beside its ID, on by default, disable with --no-comm.
138 138
139 -s:: 139 -s::
140 --summary:: 140 --summary::
141 Show only a summary of syscalls by thr 141 Show only a summary of syscalls by thread with min, max, and average times
142 (in msec) and relative stddev. 142 (in msec) and relative stddev.
143 143
144 -S:: 144 -S::
145 --with-summary:: 145 --with-summary::
146 Show all syscalls followed by a summar 146 Show all syscalls followed by a summary by thread with min, max, and
147 average times (in msec) and relative stdde 147 average times (in msec) and relative stddev.
148 148
149 --errno-summary:: 149 --errno-summary::
150 To be used with -s or -S, to show stat 150 To be used with -s or -S, to show stats for the errnos experienced by
151 syscalls, using only this option will 151 syscalls, using only this option will trigger --summary.
152 152
153 --tool_stats:: 153 --tool_stats::
154 Show tool stats such as number of time 154 Show tool stats such as number of times fd->pathname was discovered thru
155 hooking the open syscall return + vfs_ 155 hooking the open syscall return + vfs_getname or via reading /proc/pid/fd, etc.
156 156
157 -f:: 157 -f::
158 --force:: 158 --force::
159 Don't complain, do it. 159 Don't complain, do it.
160 160
161 -F=[all|min|maj]:: 161 -F=[all|min|maj]::
162 --pf=[all|min|maj]:: 162 --pf=[all|min|maj]::
163 Trace pagefaults. Optionally, you can 163 Trace pagefaults. Optionally, you can specify whether you want minor,
164 major or all pagefaults. Default value 164 major or all pagefaults. Default value is maj.
165 165
166 --syscalls:: 166 --syscalls::
167 Trace system calls. This options is en 167 Trace system calls. This options is enabled by default, disable with
168 --no-syscalls. 168 --no-syscalls.
169 169
170 --call-graph [mode,type,min[,limit],order[,key 170 --call-graph [mode,type,min[,limit],order[,key][,branch]]::
171 Setup and enable call-graph (stack cha 171 Setup and enable call-graph (stack chain/backtrace) recording.
172 See `--call-graph` section in perf-rec 172 See `--call-graph` section in perf-record and perf-report
173 man pages for details. The ones that a 173 man pages for details. The ones that are most useful in 'perf trace'
174 are 'dwarf' and 'lbr', where available 174 are 'dwarf' and 'lbr', where available, try: 'perf trace --call-graph dwarf'.
175 175
176 Using this will, for the root user, bu 176 Using this will, for the root user, bump the value of --mmap-pages to 4
177 times the maximum for non-root users, 177 times the maximum for non-root users, based on the kernel.perf_event_mlock_kb
178 sysctl. This is done only if the user 178 sysctl. This is done only if the user doesn't specify a --mmap-pages value.
179 179
180 --kernel-syscall-graph:: 180 --kernel-syscall-graph::
181 Show the kernel callchains on the sys 181 Show the kernel callchains on the syscall exit path.
182 182
183 --max-events=N:: 183 --max-events=N::
184 Stop after processing N events. Note t 184 Stop after processing N events. Note that strace-like events are considered
185 only at exit time or when a syscall is 185 only at exit time or when a syscall is interrupted, i.e. in those cases this
186 option is equivalent to the number of 186 option is equivalent to the number of lines printed.
187 187
188 --switch-on EVENT_NAME:: 188 --switch-on EVENT_NAME::
189 Only consider events after this event 189 Only consider events after this event is found.
190 190
191 --switch-off EVENT_NAME:: 191 --switch-off EVENT_NAME::
192 Stop considering events after this eve 192 Stop considering events after this event is found.
193 193
194 --show-on-off-events:: 194 --show-on-off-events::
195 Show the --switch-on/off events too. 195 Show the --switch-on/off events too.
196 196
197 --max-stack:: 197 --max-stack::
198 Set the stack depth limit when parsing 198 Set the stack depth limit when parsing the callchain, anything
199 beyond the specified depth will be ign 199 beyond the specified depth will be ignored. Note that at this point
200 this is just about the presentation pa 200 this is just about the presentation part, i.e. the kernel is still
201 not limiting, the overhead of callchai 201 not limiting, the overhead of callchains needs to be set via the
202 knobs in --call-graph dwarf. 202 knobs in --call-graph dwarf.
203 203
204 Implies '--call-graph dwarf' when --ca 204 Implies '--call-graph dwarf' when --call-graph not present on the
205 command line, on systems where DWARF u 205 command line, on systems where DWARF unwinding was built in.
206 206
207 Default: /proc/sys/kernel/perf_event_m 207 Default: /proc/sys/kernel/perf_event_max_stack when present for
208 live sessions (without --inpu 208 live sessions (without --input/-i), 127 otherwise.
209 209
210 --min-stack:: 210 --min-stack::
211 Set the stack depth limit when parsing 211 Set the stack depth limit when parsing the callchain, anything
212 below the specified depth will be igno 212 below the specified depth will be ignored. Disabled by default.
213 213
214 Implies '--call-graph dwarf' when --ca 214 Implies '--call-graph dwarf' when --call-graph not present on the
215 command line, on systems where DWARF u 215 command line, on systems where DWARF unwinding was built in.
216 216
217 --print-sample:: 217 --print-sample::
218 Print the PERF_RECORD_SAMPLE PERF_SAMP 218 Print the PERF_RECORD_SAMPLE PERF_SAMPLE_ info for the
219 raw_syscalls:sys_{enter,exit} tracepoi 219 raw_syscalls:sys_{enter,exit} tracepoints, for debugging.
220 220
221 --proc-map-timeout:: 221 --proc-map-timeout::
222 When processing pre-existing threads / 222 When processing pre-existing threads /proc/XXX/mmap, it may take a long time,
223 because the file may be huge. A time o 223 because the file may be huge. A time out is needed in such cases.
224 This option sets the time out limit. T 224 This option sets the time out limit. The default value is 500 ms.
225 225
226 --sort-events:: 226 --sort-events::
227 Do sorting on batches of events, use w 227 Do sorting on batches of events, use when noticing out of order events that
228 may happen, for instance, when a threa 228 may happen, for instance, when a thread gets migrated to a different CPU
229 while processing a syscall. 229 while processing a syscall.
230 230
231 --libtraceevent_print:: 231 --libtraceevent_print::
232 Use libtraceevent to print tracepoint 232 Use libtraceevent to print tracepoint arguments. By default 'perf trace' uses
233 the same beautifiers used in the strac 233 the same beautifiers used in the strace-like enter+exit lines to augment the
234 tracepoint arguments. 234 tracepoint arguments.
235 235
236 --map-dump:: 236 --map-dump::
237 Dump BPF maps setup by events passed v 237 Dump BPF maps setup by events passed via -e, for instance the augmented_raw_syscalls
238 living in tools/perf/examples/bpf/augm 238 living in tools/perf/examples/bpf/augmented_raw_syscalls.c. For now this
239 dumps just boolean map values and inte 239 dumps just boolean map values and integer keys, in time this will print in hex
240 by default and use BTF when available, 240 by default and use BTF when available, as well as use functions to do pretty
241 printing using the existing 'perf trac 241 printing using the existing 'perf trace' syscall arg beautifiers to map integer
242 arguments to strings (pid to comm, sys 242 arguments to strings (pid to comm, syscall id to syscall name, etc).
243 243
244 244
245 PAGEFAULTS 245 PAGEFAULTS
246 ---------- 246 ----------
247 247
248 When tracing pagefaults, the format of the tra 248 When tracing pagefaults, the format of the trace is as follows:
249 249
250 <min|maj>fault [<ip.symbol>+<ip.offset>] => <ad 250 <min|maj>fault [<ip.symbol>+<ip.offset>] => <addr.dso@addr.offset> (<map type><addr level>).
251 251
252 - min/maj indicates whether fault event is min 252 - min/maj indicates whether fault event is minor or major;
253 - ip.symbol shows symbol for instruction point 253 - ip.symbol shows symbol for instruction pointer (the code that generated the
254 fault); if no debug symbols available, perf 254 fault); if no debug symbols available, perf trace will print raw IP;
255 - addr.dso shows DSO for the faulted address; 255 - addr.dso shows DSO for the faulted address;
256 - map type is either 'd' for non-executable ma 256 - map type is either 'd' for non-executable maps or 'x' for executable maps;
257 - addr level is either 'k' for kernel dso or ' 257 - addr level is either 'k' for kernel dso or '.' for user dso.
258 258
259 For symbols resolution you may need to install 259 For symbols resolution you may need to install debugging symbols.
260 260
261 Please be aware that duration is currently alw 261 Please be aware that duration is currently always 0 and doesn't reflect actual
262 time it took for fault to be handled! 262 time it took for fault to be handled!
263 263
264 When --verbose specified, perf trace tries to 264 When --verbose specified, perf trace tries to print all available information
265 for both IP and fault address in the form of d 265 for both IP and fault address in the form of dso@symbol+offset.
266 266
267 EXAMPLES 267 EXAMPLES
268 -------- 268 --------
269 269
270 Trace only major pagefaults: 270 Trace only major pagefaults:
271 271
272 $ perf trace --no-syscalls -F 272 $ perf trace --no-syscalls -F
273 273
274 Trace syscalls, major and minor pagefaults: 274 Trace syscalls, major and minor pagefaults:
275 275
276 $ perf trace -F all 276 $ perf trace -F all
277 277
278 1416.547 ( 0.000 ms): python/20235 majfault 278 1416.547 ( 0.000 ms): python/20235 majfault [CRYPTO_push_info_+0x0] => /lib/x86_64-linux-gnu/libcrypto.so.1.0.0@0x61be0 (x.)
279 279
280 As you can see, there was major pagefault in 280 As you can see, there was major pagefault in python process, from
281 CRYPTO_push_info_ routine which faulted some 281 CRYPTO_push_info_ routine which faulted somewhere in libcrypto.so.
282 282
283 Trace the first 4 open, openat or open_by_hand 283 Trace the first 4 open, openat or open_by_handle_at syscalls (in the future more syscalls may match here):
284 284
285 $ perf trace -e open* --max-events 4 285 $ perf trace -e open* --max-events 4
286 [root@jouet perf]# trace -e open* --max-even 286 [root@jouet perf]# trace -e open* --max-events 4
287 2272.992 ( 0.037 ms): gnome-shell/1370 opena 287 2272.992 ( 0.037 ms): gnome-shell/1370 openat(dfd: CWD, filename: /proc/self/stat) = 31
288 2277.481 ( 0.139 ms): gnome-shell/3039 opena 288 2277.481 ( 0.139 ms): gnome-shell/3039 openat(dfd: CWD, filename: /proc/self/stat) = 65
289 3026.398 ( 0.076 ms): gnome-shell/3039 opena 289 3026.398 ( 0.076 ms): gnome-shell/3039 openat(dfd: CWD, filename: /proc/self/stat) = 65
290 4294.665 ( 0.015 ms): sed/15879 openat(dfd: 290 4294.665 ( 0.015 ms): sed/15879 openat(dfd: CWD, filename: /etc/ld.so.cache, flags: CLOEXEC) = 3
291 $ 291 $
292 292
293 Trace the first minor page fault when running 293 Trace the first minor page fault when running a workload:
294 294
295 # perf trace -F min --max-stack=7 --max-even 295 # perf trace -F min --max-stack=7 --max-events 1 sleep 1
296 0.000 ( 0.000 ms): sleep/18006 minfault [ 296 0.000 ( 0.000 ms): sleep/18006 minfault [__clear_user+0x1a] => 0x5626efa56080 (?k)
297 __clear 297 __clear_user ([kernel.kallsyms])
298 load_el 298 load_elf_binary ([kernel.kallsyms])
299 search_ 299 search_binary_handler ([kernel.kallsyms])
300 __do_ex 300 __do_execve_file.isra.33 ([kernel.kallsyms])
301 __x64_s 301 __x64_sys_execve ([kernel.kallsyms])
302 do_sysc 302 do_syscall_64 ([kernel.kallsyms])
303 entry_S 303 entry_SYSCALL_64 ([kernel.kallsyms])
304 # 304 #
305 305
306 Trace the next min page page fault to take pla 306 Trace the next min page page fault to take place on the first CPU:
307 307
308 # perf trace -F min --call-graph=dwarf --max 308 # perf trace -F min --call-graph=dwarf --max-events 1 --cpu 0
309 0.000 ( 0.000 ms): Web Content/17136 minf 309 0.000 ( 0.000 ms): Web Content/17136 minfault [js::gc::Chunk::fetchNextDecommittedArena+0x4b] => 0x7fbe6181b000 (?.)
310 js::gc: 310 js::gc::FreeSpan::initAsEmpty (inlined)
311 js::gc: 311 js::gc::Arena::setAsNotAllocated (inlined)
312 js::gc: 312 js::gc::Chunk::fetchNextDecommittedArena (/usr/lib64/firefox/libxul.so)
313 js::gc: 313 js::gc::Chunk::allocateArena (/usr/lib64/firefox/libxul.so)
314 js::gc: 314 js::gc::GCRuntime::allocateArena (/usr/lib64/firefox/libxul.so)
315 js::gc: 315 js::gc::ArenaLists::allocateFromArena (/usr/lib64/firefox/libxul.so)
316 js::gc: 316 js::gc::GCRuntime::tryNewTenuredThing<JSString, (js::AllowGC)1> (inlined)
317 js::All 317 js::AllocateString<JSString, (js::AllowGC)1> (/usr/lib64/firefox/libxul.so)
318 js::All 318 js::Allocate<JSThinInlineString, (js::AllowGC)1> (inlined)
319 JSThinI 319 JSThinInlineString::new_<(js::AllowGC)1> (inlined)
320 Allocat 320 AllocateInlineString<(js::AllowGC)1, unsigned char> (inlined)
321 js::Con 321 js::ConcatStrings<(js::AllowGC)1> (/usr/lib64/firefox/libxul.so)
322 [0x18b2 322 [0x18b26e6bc2bd] (/tmp/perf-17136.map)
323 # 323 #
324 324
325 Trace the next two sched:sched_switch events, 325 Trace the next two sched:sched_switch events, four block:*_plug events, the
326 next block:*_unplug and the next three net:*de 326 next block:*_unplug and the next three net:*dev_queue events, this last one
327 with a backtrace of at most 16 entries, system 327 with a backtrace of at most 16 entries, system wide:
328 328
329 # perf trace -e sched:*switch/nr=2/,block:*_ 329 # perf trace -e sched:*switch/nr=2/,block:*_plug/nr=4/,block:*_unplug/nr=1/,net:*dev_queue/nr=3,max-stack=16/
330 0.000 :0/0 sched:sched_switch:swapper/2:0 330 0.000 :0/0 sched:sched_switch:swapper/2:0 [120] S ==> rcu_sched:10 [120]
331 0.015 rcu_sched/10 sched:sched_switch:rcu 331 0.015 rcu_sched/10 sched:sched_switch:rcu_sched:10 [120] R ==> swapper/2:0 [120]
332 254.198 irq/50-iwlwifi/680 net:net_dev_queu 332 254.198 irq/50-iwlwifi/680 net:net_dev_queue:dev=wlp3s0 skbaddr=0xffff93498051f600 len=66
333 __dev_q 333 __dev_queue_xmit ([kernel.kallsyms])
334 273.977 :0/0 net:net_dev_queue:dev=wlp3s0 s 334 273.977 :0/0 net:net_dev_queue:dev=wlp3s0 skbaddr=0xffff93498051f600 len=78
335 __dev_q 335 __dev_queue_xmit ([kernel.kallsyms])
336 274.007 :0/0 net:net_dev_queue:dev=wlp3s0 s 336 274.007 :0/0 net:net_dev_queue:dev=wlp3s0 skbaddr=0xffff93498051ff00 len=78
337 __dev_q 337 __dev_queue_xmit ([kernel.kallsyms])
338 2930.140 kworker/u16:58/2722 block:block_plu 338 2930.140 kworker/u16:58/2722 block:block_plug:[kworker/u16:58]
339 2930.162 kworker/u16:58/2722 block:block_unp 339 2930.162 kworker/u16:58/2722 block:block_unplug:[kworker/u16:58] 1
340 4466.094 jbd2/dm-2-8/748 block:block_plug:[j 340 4466.094 jbd2/dm-2-8/748 block:block_plug:[jbd2/dm-2-8]
341 8050.123 kworker/u16:30/2694 block:block_plu 341 8050.123 kworker/u16:30/2694 block:block_plug:[kworker/u16:30]
342 8050.271 kworker/u16:30/2694 block:block_plu 342 8050.271 kworker/u16:30/2694 block:block_plug:[kworker/u16:30]
343 # 343 #
344 344
345 SEE ALSO 345 SEE ALSO
346 -------- 346 --------
347 linkperf:perf-record[1], linkperf:perf-script[ 347 linkperf:perf-record[1], linkperf:perf-script[1]
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.