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,
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.
67 Filter out events for these pids and for 'trace' itself (comma separated list).
74 Child tasks do not inherit counters.
78 Number of mmap data pages (must be a power of two) or size
79 specification with appended unit character - B/K/M/G. The
80 size is rounded up to have nearest pages power of two value.
84 Collect samples only on the list of CPUs provided. Multiple CPUs can be provided as a
85 comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2.
86 In per-thread mode with inheritance mode on (default), Events are captured only when
87 the thread executes on the designated CPUs. Default is to monitor all CPUs.
90 Show only events that had a duration greater than N.M ms.
93 Accrue thread runtime and provide a summary at the end of the session.
97 Process events from a given perf data file.
101 Print full timestamp rather time relative to first sample.
104 Show process COMM right beside its ID, on by default, disable with --no-comm.
108 Show only a summary of syscalls by thread with min, max, and average times
109 (in msec) and relative stddev.
113 Show all syscalls followed by a summary by thread with min, max, and
114 average times (in msec) and relative stddev.
117 Show tool stats such as number of times fd->pathname was discovered thru
118 hooking the open syscall return + vfs_getname or via reading /proc/pid/fd, etc.
122 Trace pagefaults. Optionally, you can specify whether you want minor,
123 major or all pagefaults. Default value is maj.
126 Trace system calls. This options is enabled by default, disable with
129 --call-graph [mode,type,min[,limit],order[,key][,branch]]::
130 Setup and enable call-graph (stack chain/backtrace) recording.
131 See `--call-graph` section in perf-record and perf-report
132 man pages for details. The ones that are most useful in 'perf trace'
133 are 'dwarf' and 'lbr', where available, try: 'perf trace --call-graph dwarf'.
135 Using this will, for the root user, bump the value of --mmap-pages to 4
136 times the maximum for non-root users, based on the kernel.perf_event_mlock_kb
137 sysctl. This is done only if the user doesn't specify a --mmap-pages value.
139 --kernel-syscall-graph::
140 Show the kernel callchains on the syscall exit path.
143 Set the stack depth limit when parsing the callchain, anything
144 beyond the specified depth will be ignored. Note that at this point
145 this is just about the presentation part, i.e. the kernel is still
146 not limiting, the overhead of callchains needs to be set via the
147 knobs in --call-graph dwarf.
149 Implies '--call-graph dwarf' when --call-graph not present on the
150 command line, on systems where DWARF unwinding was built in.
152 Default: /proc/sys/kernel/perf_event_max_stack when present for
153 live sessions (without --input/-i), 127 otherwise.
156 Set the stack depth limit when parsing the callchain, anything
157 below the specified depth will be ignored. Disabled by default.
159 Implies '--call-graph dwarf' when --call-graph not present on the
160 command line, on systems where DWARF unwinding was built in.
163 When processing pre-existing threads /proc/XXX/mmap, it may take a long time,
164 because the file may be huge. A time out is needed in such cases.
165 This option sets the time out limit. The default value is 500 ms.
170 When tracing pagefaults, the format of the trace is as follows:
172 <min|maj>fault [<ip.symbol>+<ip.offset>] => <addr.dso@addr.offset> (<map type><addr level>).
174 - min/maj indicates whether fault event is minor or major;
175 - ip.symbol shows symbol for instruction pointer (the code that generated the
176 fault); if no debug symbols available, perf trace will print raw IP;
177 - addr.dso shows DSO for the faulted address;
178 - map type is either 'd' for non-executable maps or 'x' for executable maps;
179 - addr level is either 'k' for kernel dso or '.' for user dso.
181 For symbols resolution you may need to install debugging symbols.
183 Please be aware that duration is currently always 0 and doesn't reflect actual
184 time it took for fault to be handled!
186 When --verbose specified, perf trace tries to print all available information
187 for both IP and fault address in the form of dso@symbol+offset.
192 Trace only major pagefaults:
194 $ perf trace --no-syscalls -F
196 Trace syscalls, major and minor pagefaults:
200 1416.547 ( 0.000 ms): python/20235 majfault [CRYPTO_push_info_+0x0] => /lib/x86_64-linux-gnu/libcrypto.so.1.0.0@0x61be0 (x.)
202 As you can see, there was major pagefault in python process, from
203 CRYPTO_push_info_ routine which faulted somewhere in libcrypto.so.
207 linkperf:perf-record[1], linkperf:perf-script[1]