6 perf-inject - Filter to augment the events stream with additional information
11 'perf inject <options>'
15 perf-inject reads a perf-record event stream and repipes it to stdout. At any
16 point the processing code can inject other events into the event stream - in
17 this case build-ids (-b option) are read and injected as needed into the event
20 Build-ids are just the first user of perf-inject - potentially anything that
21 needs userspace processing to augment the events stream with additional
22 information could make use of this facility.
28 Inject build-ids of DSOs hit by samples into the output stream.
29 This means it needs to process all SAMPLE records to find the DSOs.
32 Inject build-ids of all DSOs into the output stream regardless of hits
33 and skip SAMPLE processing.
36 Override build-ids to inject using these comma-separated pairs of
37 build-id and path. Understands file://filename to read these pairs
38 from a file, which can be generated with perf buildid-list.
45 Input file name. (default: stdin)
48 Output file name. (default: stdout)
51 Merge sched_stat and sched_switch for getting events where and how long
52 tasks slept. sched_switch contains a callchain where a task slept and
53 sched_stat contains a timeslice how long a task slept.
66 Decode Instruction Tracing data, replacing it with synthesized events.
72 Use with --itrace to strip out non-synthesized events.
76 Process jitdump files by injecting the mmap records corresponding to jitted
77 functions. This option also generates the ELF images for each jitted function
78 found in the jitdumps files captured in the input perf.data file. Use this option
79 if you are monitoring environment using JIT runtimes, such as Java, DART or V8.
83 Don't complain, do it.
85 --vm-time-correlation[=OPTIONS]::
86 Some architectures may capture AUX area data which contains timestamps
87 affected by virtualization. This option will update those timestamps
88 in place, to correlate with host timestamps. The in-place update means
89 that an output file is not specified, and instead the input file is
90 modified. The options are architecture specific, except that they may
91 start with "dry-run" which will cause the file to be processed but
92 without updating it. Currently this option is supported only by
93 Intel PT, refer linkperf:perf-intel-pt[1]
95 --guest-data=<path>,<pid>[,<time offset>[,<time scale>]]::
96 Insert events from a perf.data file recorded in a virtual machine at
97 the same time as the input perf.data file was recorded on the host.
98 The Process ID (PID) of the QEMU hypervisor process must be provided,
99 and the time offset and time scale (multiplier) will likely be needed
100 to convert guest time stamps into host time stamps. For example, for
101 x86 the TSC Offset and Multiplier could be provided for a virtual machine
102 using Linux command line option no-kvmclock.
103 Currently only mmap, mmap2, comm, task, context_switch, ksymbol,
104 and text_poke events are inserted, as well as build ID information.
105 The QEMU option -name debug-threads=on is needed so that thread names
106 can be used to determine which thread is running which VCPU. Note
107 libvirt seems to use this by default.
108 When using perf record in the guest, option --sample-identifier
109 should be used, and also --buildid-all and --switch-events may be
112 :GMEXAMPLECMD: inject
114 include::guestmount.txt[]
118 linkperf:perf-record[1], linkperf:perf-report[1], linkperf:perf-archive[1],
119 linkperf:perf-intel-pt[1]