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