6 perf-trace - strace inspired tool
16 This command will show the events associated with the target, initially
17 syscalls, but other system events like pagefaults, task lifetime events,
18 scheduling events, etc.
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 using the 'perf record' command
22 but the session needs to include the raw_syscalls events (-e 'raw_syscalls:*').
23 Alternatively, 'perf trace record' can be used as a shortcut to
24 automatically include the raw_syscalls events when writing events to a file.
26 The following options apply to perf trace; options to perf trace record are
27 found in the perf record man page.
34 System-wide collection from all CPUs.
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
47 After starting the program, wait msecs before measuring. This is useful to
48 filter out the startup phase of the program, which is often very different.
56 Record events on existing process ID (comma separated list).
60 Record events on existing thread ID (comma separated list).
64 Record events in threads owned by uid. Name or number.
68 Record events in threads in a cgroup.
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:
73 perf trace -G A -e sched:*switch
75 Will set all raw_syscalls:sys_{enter,exit}, pgfault, vfs_getname, etc
76 _and_ sched:sched_switch to the 'A' cgroup, while:
78 perf trace -e sched:*switch -G A
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).
86 perf trace -G A -e sched:*switch -G B
88 the syscall ones go to the 'A' cgroup, the sched:sched_switch goes
92 Filter out events for these pids and for 'trace' itself (comma separated list).
99 Child tasks do not inherit counters.
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.
109 Collect samples only on the list of CPUs provided. Multiple CPUs can be provided as a
110 comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2.
111 In per-thread mode with inheritance mode on (default), Events are captured only when
112 the thread executes on the designated CPUs. Default is to monitor all CPUs.
115 Show only events that had a duration greater than N.M ms.
118 Accrue thread runtime and provide a summary at the end of the session.
121 Show only syscalls that failed, i.e. that returned < 0.
125 Process events from a given perf data file.
129 Print full timestamp rather time relative to first sample.
132 Show process COMM right beside its ID, on by default, disable with --no-comm.
136 Show only a summary of syscalls by thread with min, max, and average times
137 (in msec) and relative stddev.
141 Show all syscalls followed by a summary by thread with min, max, and
142 average times (in msec) and relative stddev.
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.
150 Don't complain, do it.
154 Trace pagefaults. Optionally, you can specify whether you want minor,
155 major or all pagefaults. Default value is maj.
158 Trace system calls. This options is enabled by default, disable with
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'.
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.
171 --kernel-syscall-graph::
172 Show the kernel callchains on the syscall exit path.
175 Set the stack depth limit when parsing the callchain, anything
176 beyond the specified depth will be ignored. Note that at this point
177 this is just about the presentation part, i.e. the kernel is still
178 not limiting, the overhead of callchains needs to be set via the
179 knobs in --call-graph dwarf.
181 Implies '--call-graph dwarf' when --call-graph not present on the
182 command line, on systems where DWARF unwinding was built in.
184 Default: /proc/sys/kernel/perf_event_max_stack when present for
185 live sessions (without --input/-i), 127 otherwise.
188 Set the stack depth limit when parsing the callchain, anything
189 below the specified depth will be ignored. Disabled by default.
191 Implies '--call-graph dwarf' when --call-graph not present on the
192 command line, on systems where DWARF unwinding was built in.
195 Print the PERF_RECORD_SAMPLE PERF_SAMPLE_ info for the
196 raw_syscalls:sys_{enter,exit} tracepoints, for debugging.
199 When processing pre-existing threads /proc/XXX/mmap, it may take a long time,
200 because the file may be huge. A time out is needed in such cases.
201 This option sets the time out limit. The default value is 500 ms.
206 When tracing pagefaults, the format of the trace is as follows:
208 <min|maj>fault [<ip.symbol>+<ip.offset>] => <addr.dso@addr.offset> (<map type><addr level>).
210 - min/maj indicates whether fault event is minor or major;
211 - ip.symbol shows symbol for instruction pointer (the code that generated the
212 fault); if no debug symbols available, perf trace will print raw IP;
213 - addr.dso shows DSO for the faulted address;
214 - map type is either 'd' for non-executable maps or 'x' for executable maps;
215 - addr level is either 'k' for kernel dso or '.' for user dso.
217 For symbols resolution you may need to install debugging symbols.
219 Please be aware that duration is currently always 0 and doesn't reflect actual
220 time it took for fault to be handled!
222 When --verbose specified, perf trace tries to print all available information
223 for both IP and fault address in the form of dso@symbol+offset.
228 Trace only major pagefaults:
230 $ perf trace --no-syscalls -F
232 Trace syscalls, major and minor pagefaults:
236 1416.547 ( 0.000 ms): python/20235 majfault [CRYPTO_push_info_+0x0] => /lib/x86_64-linux-gnu/libcrypto.so.1.0.0@0x61be0 (x.)
238 As you can see, there was major pagefault in python process, from
239 CRYPTO_push_info_ routine which faulted somewhere in libcrypto.so.
243 linkperf:perf-record[1], linkperf:perf-script[1]