1perf-trace(1) 2============= 3 4NAME 5---- 6perf-trace - strace inspired tool 7 8SYNOPSIS 9-------- 10[verse] 11'perf trace' 12'perf trace record' 13 14DESCRIPTION 15----------- 16This command will show the events associated with the target, initially 17syscalls, but other system events like pagefaults, task lifetime events, 18scheduling events, etc. 19 20This is a live mode tool in addition to working with perf.data files like 21the other perf tools. Files can be generated using the 'perf record' command 22but the session needs to include the raw_syscalls events (-e 'raw_syscalls:*'). 23Alternatively, 'perf trace record' can be used as a shortcut to 24automatically include the raw_syscalls events when writing events to a file. 25 26The following options apply to perf trace; options to perf trace record are 27found in the perf record man page. 28 29OPTIONS 30------- 31 32-a:: 33--all-cpus:: 34 System-wide collection from all CPUs. 35 36-e:: 37--expr:: 38--event:: 39 List of syscalls and other perf events (tracepoints, HW cache events, 40 etc) to show. Globbing is supported, e.g.: "epoll_*", "*msg*", etc. 41 See 'perf list' for a complete list of events. 42 Prefixing with ! shows all syscalls but the ones specified. You may 43 need to escape it. 44 45--filter=<filter>:: 46 Event filter. This option should follow an event selector (-e) which 47 selects tracepoint event(s). 48 49 50-D msecs:: 51--delay msecs:: 52After starting the program, wait msecs before measuring. This is useful to 53filter out the startup phase of the program, which is often very different. 54 55-o:: 56--output=:: 57 Output file name. 58 59-p:: 60--pid=:: 61 Record events on existing process ID (comma separated list). 62 63-t:: 64--tid=:: 65 Record events on existing thread ID (comma separated list). 66 67-u:: 68--uid=:: 69 Record events in threads owned by uid. Name or number. 70 71-G:: 72--cgroup:: 73 Record events in threads in a cgroup. 74 75 Look for cgroups to set at the /sys/fs/cgroup/perf_event directory, then 76 remove the /sys/fs/cgroup/perf_event/ part and try: 77 78 perf trace -G A -e sched:*switch 79 80 Will set all raw_syscalls:sys_{enter,exit}, pgfault, vfs_getname, etc 81 _and_ sched:sched_switch to the 'A' cgroup, while: 82 83 perf trace -e sched:*switch -G A 84 85 will only set the sched:sched_switch event to the 'A' cgroup, all the 86 other events (raw_syscalls:sys_{enter,exit}, etc are left "without" 87 a cgroup (on the root cgroup, sys wide, etc). 88 89 Multiple cgroups: 90 91 perf trace -G A -e sched:*switch -G B 92 93 the syscall ones go to the 'A' cgroup, the sched:sched_switch goes 94 to the 'B' cgroup. 95 96--filter-pids=:: 97 Filter out events for these pids and for 'trace' itself (comma separated list). 98 99-v:: 100--verbose:: 101 Increase the verbosity level. 102 103--no-inherit:: 104 Child tasks do not inherit counters. 105 106-m:: 107--mmap-pages=:: 108 Number of mmap data pages (must be a power of two) or size 109 specification in bytes with appended unit character - B/K/M/G. 110 The size is rounded up to the nearest power-of-two page value. 111 112-C:: 113--cpu:: 114Collect samples only on the list of CPUs provided. Multiple CPUs can be provided as a 115comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2. 116In per-thread mode with inheritance mode on (default), Events are captured only when 117the thread executes on the designated CPUs. Default is to monitor all CPUs. 118 119--duration:: 120 Show only events that had a duration greater than N.M ms. 121 122--sched:: 123 Accrue thread runtime and provide a summary at the end of the session. 124 125--failure:: 126 Show only syscalls that failed, i.e. that returned < 0. 127 128-i:: 129--input:: 130 Process events from a given perf data file. 131 132-T:: 133--time:: 134 Print full timestamp rather time relative to first sample. 135 136--comm:: 137 Show process COMM right beside its ID, on by default, disable with --no-comm. 138 139-s:: 140--summary:: 141 Show only a summary of syscalls by thread with min, max, and average times 142 (in msec) and relative stddev. 143 144-S:: 145--with-summary:: 146 Show all syscalls followed by a summary by thread with min, max, and 147 average times (in msec) and relative stddev. 148 149--errno-summary:: 150 To be used with -s or -S, to show stats for the errnos experienced by 151 syscalls, using only this option will trigger --summary. 152 153--summary-mode=mode:: 154 To be used with -s or -S, to select how to show summary. By default it'll 155 show the syscall summary by thread. Possible values are: thread, total, 156 cgroup. 157 158--tool_stats:: 159 Show tool stats such as number of times fd->pathname was discovered thru 160 hooking the open syscall return + vfs_getname or via reading /proc/pid/fd, etc. 161 162-f:: 163--force:: 164 Don't complain, do it. 165 166-F=[all|min|maj]:: 167--pf=[all|min|maj]:: 168 Trace pagefaults. Optionally, you can specify whether you want minor, 169 major or all pagefaults. Default value is maj. 170 171--syscalls:: 172 Trace system calls. This options is enabled by default, disable with 173 --no-syscalls. 174 175--call-graph [mode,type,min[,limit],order[,key][,branch]]:: 176 Setup and enable call-graph (stack chain/backtrace) recording. 177 See `--call-graph` section in perf-record and perf-report 178 man pages for details. The ones that are most useful in 'perf trace' 179 are 'dwarf' and 'lbr', where available, try: 'perf trace --call-graph dwarf'. 180 181 Using this will, for the root user, bump the value of --mmap-pages to 4 182 times the maximum for non-root users, based on the kernel.perf_event_mlock_kb 183 sysctl. This is done only if the user doesn't specify a --mmap-pages value. 184 185--kernel-syscall-graph:: 186 Show the kernel callchains on the syscall exit path. 187 188--max-events=N:: 189 Stop after processing N events. Note that strace-like events are considered 190 only at exit time or when a syscall is interrupted, i.e. in those cases this 191 option is equivalent to the number of lines printed. 192 193--switch-on EVENT_NAME:: 194 Only consider events after this event is found. 195 196--switch-off EVENT_NAME:: 197 Stop considering events after this event is found. 198 199--show-on-off-events:: 200 Show the --switch-on/off events too. 201 202--max-stack:: 203 Set the stack depth limit when parsing the callchain, anything 204 beyond the specified depth will be ignored. Note that at this point 205 this is just about the presentation part, i.e. the kernel is still 206 not limiting, the overhead of callchains needs to be set via the 207 knobs in --call-graph dwarf. 208 209 Implies '--call-graph dwarf' when --call-graph not present on the 210 command line, on systems where DWARF unwinding was built in. 211 212 Default: /proc/sys/kernel/perf_event_max_stack when present for 213 live sessions (without --input/-i), 127 otherwise. 214 215--min-stack:: 216 Set the stack depth limit when parsing the callchain, anything 217 below the specified depth will be ignored. Disabled by default. 218 219 Implies '--call-graph dwarf' when --call-graph not present on the 220 command line, on systems where DWARF unwinding was built in. 221 222--print-sample:: 223 Print the PERF_RECORD_SAMPLE PERF_SAMPLE_ info for the 224 raw_syscalls:sys_{enter,exit} tracepoints, for debugging. 225 226--proc-map-timeout:: 227 When processing pre-existing threads /proc/XXX/mmap, it may take a long time, 228 because the file may be huge. A time out is needed in such cases. 229 This option sets the time out limit. The default value is 500 ms. 230 231--sort-events:: 232 Do sorting on batches of events, use when noticing out of order events that 233 may happen, for instance, when a thread gets migrated to a different CPU 234 while processing a syscall. 235 236--libtraceevent_print:: 237 Use libtraceevent to print tracepoint arguments. By default 'perf trace' uses 238 the same beautifiers used in the strace-like enter+exit lines to augment the 239 tracepoint arguments. 240 241--force-btf:: 242 Use btf_dump to pretty print syscall argument data, instead of using hand-crafted pretty 243 printers. This option is intended for testing BTF integration in perf trace. btf_dump-based 244 pretty-printing serves as a fallback to hand-crafted pretty printers, as the latter can 245 better pretty-print integer flags and struct pointers. 246 247--bpf-summary:: 248 Collect system call statistics in BPF. This is only for live mode and 249 works well with -s/--summary option where no argument information is 250 required. 251 252--max-summary=N:: 253 Maximum number of lines in the summary mode. Note that this applies to 254 each entry (thread or cgroup). 255 256 257PAGEFAULTS 258---------- 259 260When tracing pagefaults, the format of the trace is as follows: 261 262<min|maj>fault [<ip.symbol>+<ip.offset>] => <addr.dso@addr.offset> (<map type><addr level>). 263 264- min/maj indicates whether fault event is minor or major; 265- ip.symbol shows symbol for instruction pointer (the code that generated the 266 fault); if no debug symbols available, perf trace will print raw IP; 267- addr.dso shows DSO for the faulted address; 268- map type is either 'd' for non-executable maps or 'x' for executable maps; 269- addr level is either 'k' for kernel dso or '.' for user dso. 270 271For symbols resolution you may need to install debugging symbols. 272 273Please be aware that duration is currently always 0 and doesn't reflect actual 274time it took for fault to be handled! 275 276When --verbose specified, perf trace tries to print all available information 277for both IP and fault address in the form of dso@symbol+offset. 278 279EXAMPLES 280-------- 281 282Trace only major pagefaults: 283 284 $ perf trace --no-syscalls -F 285 286Trace syscalls, major and minor pagefaults: 287 288 $ perf trace -F all 289 290 1416.547 ( 0.000 ms): python/20235 majfault [CRYPTO_push_info_+0x0] => /lib/x86_64-linux-gnu/libcrypto.so.1.0.0@0x61be0 (x.) 291 292 As you can see, there was major pagefault in python process, from 293 CRYPTO_push_info_ routine which faulted somewhere in libcrypto.so. 294 295Trace the first 4 open, openat or open_by_handle_at syscalls (in the future more syscalls may match here): 296 297 $ perf trace -e open* --max-events 4 298 [root@jouet perf]# trace -e open* --max-events 4 299 2272.992 ( 0.037 ms): gnome-shell/1370 openat(dfd: CWD, filename: /proc/self/stat) = 31 300 2277.481 ( 0.139 ms): gnome-shell/3039 openat(dfd: CWD, filename: /proc/self/stat) = 65 301 3026.398 ( 0.076 ms): gnome-shell/3039 openat(dfd: CWD, filename: /proc/self/stat) = 65 302 4294.665 ( 0.015 ms): sed/15879 openat(dfd: CWD, filename: /etc/ld.so.cache, flags: CLOEXEC) = 3 303 $ 304 305Trace the first minor page fault when running a workload: 306 307 # perf trace -F min --max-stack=7 --max-events 1 sleep 1 308 0.000 ( 0.000 ms): sleep/18006 minfault [__clear_user+0x1a] => 0x5626efa56080 (?k) 309 __clear_user ([kernel.kallsyms]) 310 load_elf_binary ([kernel.kallsyms]) 311 search_binary_handler ([kernel.kallsyms]) 312 __do_execve_file.isra.33 ([kernel.kallsyms]) 313 __x64_sys_execve ([kernel.kallsyms]) 314 do_syscall_64 ([kernel.kallsyms]) 315 entry_SYSCALL_64 ([kernel.kallsyms]) 316 # 317 318Trace the next min page page fault to take place on the first CPU: 319 320 # perf trace -F min --call-graph=dwarf --max-events 1 --cpu 0 321 0.000 ( 0.000 ms): Web Content/17136 minfault [js::gc::Chunk::fetchNextDecommittedArena+0x4b] => 0x7fbe6181b000 (?.) 322 js::gc::FreeSpan::initAsEmpty (inlined) 323 js::gc::Arena::setAsNotAllocated (inlined) 324 js::gc::Chunk::fetchNextDecommittedArena (/usr/lib64/firefox/libxul.so) 325 js::gc::Chunk::allocateArena (/usr/lib64/firefox/libxul.so) 326 js::gc::GCRuntime::allocateArena (/usr/lib64/firefox/libxul.so) 327 js::gc::ArenaLists::allocateFromArena (/usr/lib64/firefox/libxul.so) 328 js::gc::GCRuntime::tryNewTenuredThing<JSString, (js::AllowGC)1> (inlined) 329 js::AllocateString<JSString, (js::AllowGC)1> (/usr/lib64/firefox/libxul.so) 330 js::Allocate<JSThinInlineString, (js::AllowGC)1> (inlined) 331 JSThinInlineString::new_<(js::AllowGC)1> (inlined) 332 AllocateInlineString<(js::AllowGC)1, unsigned char> (inlined) 333 js::ConcatStrings<(js::AllowGC)1> (/usr/lib64/firefox/libxul.so) 334 [0x18b26e6bc2bd] (/tmp/perf-17136.map) 335 # 336 337Trace the next two sched:sched_switch events, four block:*_plug events, the 338next block:*_unplug and the next three net:*dev_queue events, this last one 339with a backtrace of at most 16 entries, system wide: 340 341 # perf trace -e sched:*switch/nr=2/,block:*_plug/nr=4/,block:*_unplug/nr=1/,net:*dev_queue/nr=3,max-stack=16/ 342 0.000 :0/0 sched:sched_switch:swapper/2:0 [120] S ==> rcu_sched:10 [120] 343 0.015 rcu_sched/10 sched:sched_switch:rcu_sched:10 [120] R ==> swapper/2:0 [120] 344 254.198 irq/50-iwlwifi/680 net:net_dev_queue:dev=wlp3s0 skbaddr=0xffff93498051f600 len=66 345 __dev_queue_xmit ([kernel.kallsyms]) 346 273.977 :0/0 net:net_dev_queue:dev=wlp3s0 skbaddr=0xffff93498051f600 len=78 347 __dev_queue_xmit ([kernel.kallsyms]) 348 274.007 :0/0 net:net_dev_queue:dev=wlp3s0 skbaddr=0xffff93498051ff00 len=78 349 __dev_queue_xmit ([kernel.kallsyms]) 350 2930.140 kworker/u16:58/2722 block:block_plug:[kworker/u16:58] 351 2930.162 kworker/u16:58/2722 block:block_unplug:[kworker/u16:58] 1 352 4466.094 jbd2/dm-2-8/748 block:block_plug:[jbd2/dm-2-8] 353 8050.123 kworker/u16:30/2694 block:block_plug:[kworker/u16:30] 354 8050.271 kworker/u16:30/2694 block:block_plug:[kworker/u16:30] 355 # 356 357SEE ALSO 358-------- 359linkperf:perf-record[1], linkperf:perf-script[1] 360