1perf-inject(1) 2============== 3 4NAME 5---- 6perf-inject - Filter to augment the events stream with additional information 7 8SYNOPSIS 9-------- 10[verse] 11'perf inject <options>' 12 13DESCRIPTION 14----------- 15perf-inject reads a perf-record event stream and repipes it to stdout. At any 16point the processing code can inject other events into the event stream - in 17this case build-ids (-b option) are read and injected as needed into the event 18stream. 19 20Build-ids are just the first user of perf-inject - potentially anything that 21needs userspace processing to augment the events stream with additional 22information could make use of this facility. 23 24OPTIONS 25------- 26-b:: 27--build-ids:: 28 Inject build-ids into the output stream 29 30--buildid-all: 31 Inject build-ids of all DSOs into the output stream 32 33-v:: 34--verbose:: 35 Be more verbose. 36-i:: 37--input=:: 38 Input file name. (default: stdin) 39-o:: 40--output=:: 41 Output file name. (default: stdout) 42-s:: 43--sched-stat:: 44 Merge sched_stat and sched_switch for getting events where and how long 45 tasks slept. sched_switch contains a callchain where a task slept and 46 sched_stat contains a timeslice how long a task slept. 47 48-k:: 49--vmlinux=<file>:: 50 vmlinux pathname 51 52--ignore-vmlinux:: 53 Ignore vmlinux files. 54 55--kallsyms=<file>:: 56 kallsyms pathname 57 58--itrace:: 59 Decode Instruction Tracing data, replacing it with synthesized events. 60 Options are: 61 62include::itrace.txt[] 63 64--strip:: 65 Use with --itrace to strip out non-synthesized events. 66 67-j:: 68--jit:: 69 Process jitdump files by injecting the mmap records corresponding to jitted 70 functions. This option also generates the ELF images for each jitted function 71 found in the jitdumps files captured in the input perf.data file. Use this option 72 if you are monitoring environment using JIT runtimes, such as Java, DART or V8. 73 74-f:: 75--force:: 76 Don't complain, do it. 77 78--vm-time-correlation[=OPTIONS]:: 79 Some architectures may capture AUX area data which contains timestamps 80 affected by virtualization. This option will update those timestamps 81 in place, to correlate with host timestamps. The in-place update means 82 that an output file is not specified, and instead the input file is 83 modified. The options are architecture specific, except that they may 84 start with "dry-run" which will cause the file to be processed but 85 without updating it. Currently this option is supported only by 86 Intel PT, refer linkperf:perf-intel-pt[1] 87 88--guest-data=<path>,<pid>[,<time offset>[,<time scale>]]:: 89 Insert events from a perf.data file recorded in a virtual machine at 90 the same time as the input perf.data file was recorded on the host. 91 The Process ID (PID) of the QEMU hypervisor process must be provided, 92 and the time offset and time scale (multiplier) will likely be needed 93 to convert guest time stamps into host time stamps. For example, for 94 x86 the TSC Offset and Multiplier could be provided for a virtual machine 95 using Linux command line option no-kvmclock. 96 Currently only mmap, mmap2, comm, task, context_switch, ksymbol, 97 and text_poke events are inserted, as well as build ID information. 98 The QEMU option -name debug-threads=on is needed so that thread names 99 can be used to determine which thread is running which VCPU. Note 100 libvirt seems to use this by default. 101 When using perf record in the guest, option --sample-identifier 102 should be used, and also --buildid-all and --switch-events may be 103 useful. 104 105:GMEXAMPLECMD: inject 106:GMEXAMPLESUBCMD: 107include::guestmount.txt[] 108 109SEE ALSO 110-------- 111linkperf:perf-record[1], linkperf:perf-report[1], linkperf:perf-archive[1], 112linkperf:perf-intel-pt[1] 113