1perf-record(1) 2============== 3 4NAME 5---- 6perf-record - Run a command and record its profile into perf.data 7 8SYNOPSIS 9-------- 10[verse] 11'perf record' [-e <EVENT> | --event=EVENT] [-l] [-a] <command> 12'perf record' [-e <EVENT> | --event=EVENT] [-l] [-a] -- <command> [<options>] 13 14DESCRIPTION 15----------- 16This command runs a command and gathers a performance counter profile 17from it, into perf.data - without displaying anything. 18 19This file can then be inspected later on, using 'perf report'. 20 21 22OPTIONS 23------- 24<command>...:: 25 Any command you can specify in a shell. 26 27-e:: 28--event=:: 29 Select the PMU event. Selection can be: 30 31 - a symbolic event name (use 'perf list' to list all events) 32 33 - a raw PMU event (eventsel+umask) in the form of rNNN where NNN is a 34 hexadecimal event descriptor. 35 36 - a symbolically formed PMU event like 'pmu/param1=0x3,param2/' where 37 'param1', 'param2', etc are defined as formats for the PMU in 38 /sys/bus/event_source/devices/<pmu>/format/*. 39 40 - a symbolically formed event like 'pmu/config=M,config1=N,config3=K/' 41 42 where M, N, K are numbers (in decimal, hex, octal format). Acceptable 43 values for each of 'config', 'config1' and 'config2' are defined by 44 corresponding entries in /sys/bus/event_source/devices/<pmu>/format/* 45 param1 and param2 are defined as formats for the PMU in: 46 /sys/bus/event_source/devices/<pmu>/format/* 47 48 There are also some params which are not defined in .../<pmu>/format/*. 49 These params can be used to overload default config values per event. 50 Here is a list of the params. 51 - 'period': Set event sampling period 52 - 'freq': Set event sampling frequency 53 - 'time': Disable/enable time stamping. Acceptable values are 1 for 54 enabling time stamping. 0 for disabling time stamping. 55 The default is 1. 56 - 'call-graph': Disable/enable callgraph. Acceptable str are "fp" for 57 FP mode, "dwarf" for DWARF mode, "lbr" for LBR mode and 58 "no" for disable callgraph. 59 - 'stack-size': user stack size for dwarf mode 60 Note: If user explicitly sets options which conflict with the params, 61 the value set by the params will be overridden. 62 63 Also not defined in .../<pmu>/format/* are PMU driver specific 64 configuration parameters. Any configuration parameter preceded by 65 the letter '@' is not interpreted in user space and sent down directly 66 to the PMU driver. For example: 67 68 perf record -e some_event/@cfg1,@cfg2=config/ ... 69 70 will see 'cfg1' and 'cfg2=config' pushed to the PMU driver associated 71 with the event for further processing. There is no restriction on 72 what the configuration parameters are, as long as their semantic is 73 understood and supported by the PMU driver. 74 75 - a hardware breakpoint event in the form of '\mem:addr[/len][:access]' 76 where addr is the address in memory you want to break in. 77 Access is the memory access type (read, write, execute) it can 78 be passed as follows: '\mem:addr[:[r][w][x]]'. len is the range, 79 number of bytes from specified addr, which the breakpoint will cover. 80 If you want to profile read-write accesses in 0x1000, just set 81 'mem:0x1000:rw'. 82 If you want to profile write accesses in [0x1000~1008), just set 83 'mem:0x1000/8:w'. 84 85 - a group of events surrounded by a pair of brace ("{event1,event2,...}"). 86 Each event is separated by commas and the group should be quoted to 87 prevent the shell interpretation. You also need to use --group on 88 "perf report" to view group events together. 89 90--filter=<filter>:: 91 Event filter. This option should follow a event selector (-e) which 92 selects tracepoint event(s). Multiple '--filter' options are combined 93 using '&&'. 94 95--exclude-perf:: 96 Don't record events issued by perf itself. This option should follow 97 a event selector (-e) which selects tracepoint event(s). It adds a 98 filter expression 'common_pid != $PERFPID' to filters. If other 99 '--filter' exists, the new filter expression will be combined with 100 them by '&&'. 101 102-a:: 103--all-cpus:: 104 System-wide collection from all CPUs. 105 106-p:: 107--pid=:: 108 Record events on existing process ID (comma separated list). 109 110-t:: 111--tid=:: 112 Record events on existing thread ID (comma separated list). 113 This option also disables inheritance by default. Enable it by adding 114 --inherit. 115 116-u:: 117--uid=:: 118 Record events in threads owned by uid. Name or number. 119 120-r:: 121--realtime=:: 122 Collect data with this RT SCHED_FIFO priority. 123 124--no-buffering:: 125 Collect data without buffering. 126 127-c:: 128--count=:: 129 Event period to sample. 130 131-o:: 132--output=:: 133 Output file name. 134 135-i:: 136--no-inherit:: 137 Child tasks do not inherit counters. 138-F:: 139--freq=:: 140 Profile at this frequency. 141 142-m:: 143--mmap-pages=:: 144 Number of mmap data pages (must be a power of two) or size 145 specification with appended unit character - B/K/M/G. The 146 size is rounded up to have nearest pages power of two value. 147 Also, by adding a comma, the number of mmap pages for AUX 148 area tracing can be specified. 149 150--group:: 151 Put all events in a single event group. This precedes the --event 152 option and remains only for backward compatibility. See --event. 153 154-g:: 155 Enables call-graph (stack chain/backtrace) recording. 156 157--call-graph:: 158 Setup and enable call-graph (stack chain/backtrace) recording, 159 implies -g. Default is "fp". 160 161 Allows specifying "fp" (frame pointer) or "dwarf" 162 (DWARF's CFI - Call Frame Information) or "lbr" 163 (Hardware Last Branch Record facility) as the method to collect 164 the information used to show the call graphs. 165 166 In some systems, where binaries are build with gcc 167 --fomit-frame-pointer, using the "fp" method will produce bogus 168 call graphs, using "dwarf", if available (perf tools linked to 169 the libunwind or libdw library) should be used instead. 170 Using the "lbr" method doesn't require any compiler options. It 171 will produce call graphs from the hardware LBR registers. The 172 main limition is that it is only available on new Intel 173 platforms, such as Haswell. It can only get user call chain. It 174 doesn't work with branch stack sampling at the same time. 175 176 When "dwarf" recording is used, perf also records (user) stack dump 177 when sampled. Default size of the stack dump is 8192 (bytes). 178 User can change the size by passing the size after comma like 179 "--call-graph dwarf,4096". 180 181-q:: 182--quiet:: 183 Don't print any message, useful for scripting. 184 185-v:: 186--verbose:: 187 Be more verbose (show counter open errors, etc). 188 189-s:: 190--stat:: 191 Record per-thread event counts. Use it with 'perf report -T' to see 192 the values. 193 194-d:: 195--data:: 196 Record the sample addresses. 197 198-T:: 199--timestamp:: 200 Record the sample timestamps. Use it with 'perf report -D' to see the 201 timestamps, for instance. 202 203-P:: 204--period:: 205 Record the sample period. 206 207--sample-cpu:: 208 Record the sample cpu. 209 210-n:: 211--no-samples:: 212 Don't sample. 213 214-R:: 215--raw-samples:: 216Collect raw sample records from all opened counters (default for tracepoint counters). 217 218-C:: 219--cpu:: 220Collect samples only on the list of CPUs provided. Multiple CPUs can be provided as a 221comma-separated list with no space: 0,1. Ranges of CPUs are specified with -: 0-2. 222In per-thread mode with inheritance mode on (default), samples are captured only when 223the thread executes on the designated CPUs. Default is to monitor all CPUs. 224 225-B:: 226--no-buildid:: 227Do not save the build ids of binaries in the perf.data files. This skips 228post processing after recording, which sometimes makes the final step in 229the recording process to take a long time, as it needs to process all 230events looking for mmap records. The downside is that it can misresolve 231symbols if the workload binaries used when recording get locally rebuilt 232or upgraded, because the only key available in this case is the 233pathname. You can also set the "record.build-id" config variable to 234'skip to have this behaviour permanently. 235 236-N:: 237--no-buildid-cache:: 238Do not update the buildid cache. This saves some overhead in situations 239where the information in the perf.data file (which includes buildids) 240is sufficient. You can also set the "record.build-id" config variable to 241'no-cache' to have the same effect. 242 243-G name,...:: 244--cgroup name,...:: 245monitor only in the container (cgroup) called "name". This option is available only 246in per-cpu mode. The cgroup filesystem must be mounted. All threads belonging to 247container "name" are monitored when they run on the monitored CPUs. Multiple cgroups 248can be provided. Each cgroup is applied to the corresponding event, i.e., first cgroup 249to first event, second cgroup to second event and so on. It is possible to provide 250an empty cgroup (monitor all the time) using, e.g., -G foo,,bar. Cgroups must have 251corresponding events, i.e., they always refer to events defined earlier on the command 252line. 253 254-b:: 255--branch-any:: 256Enable taken branch stack sampling. Any type of taken branch may be sampled. 257This is a shortcut for --branch-filter any. See --branch-filter for more infos. 258 259-j:: 260--branch-filter:: 261Enable taken branch stack sampling. Each sample captures a series of consecutive 262taken branches. The number of branches captured with each sample depends on the 263underlying hardware, the type of branches of interest, and the executed code. 264It is possible to select the types of branches captured by enabling filters. The 265following filters are defined: 266 267 - any: any type of branches 268 - any_call: any function call or system call 269 - any_ret: any function return or system call return 270 - ind_call: any indirect branch 271 - call: direct calls, including far (to/from kernel) calls 272 - u: only when the branch target is at the user level 273 - k: only when the branch target is in the kernel 274 - hv: only when the target is at the hypervisor level 275 - in_tx: only when the target is in a hardware transaction 276 - no_tx: only when the target is not in a hardware transaction 277 - abort_tx: only when the target is a hardware transaction abort 278 - cond: conditional branches 279 280+ 281The option requires at least one branch type among any, any_call, any_ret, ind_call, cond. 282The privilege levels may be omitted, in which case, the privilege levels of the associated 283event are applied to the branch filter. Both kernel (k) and hypervisor (hv) privilege 284levels are subject to permissions. When sampling on multiple events, branch stack sampling 285is enabled for all the sampling events. The sampled branch type is the same for all events. 286The various filters must be specified as a comma separated list: --branch-filter any_ret,u,k 287Note that this feature may not be available on all processors. 288 289--weight:: 290Enable weightened sampling. An additional weight is recorded per sample and can be 291displayed with the weight and local_weight sort keys. This currently works for TSX 292abort events and some memory events in precise mode on modern Intel CPUs. 293 294--transaction:: 295Record transaction flags for transaction related events. 296 297--per-thread:: 298Use per-thread mmaps. By default per-cpu mmaps are created. This option 299overrides that and uses per-thread mmaps. A side-effect of that is that 300inheritance is automatically disabled. --per-thread is ignored with a warning 301if combined with -a or -C options. 302 303-D:: 304--delay=:: 305After starting the program, wait msecs before measuring. This is useful to 306filter out the startup phase of the program, which is often very different. 307 308-I:: 309--intr-regs:: 310Capture machine state (registers) at interrupt, i.e., on counter overflows for 311each sample. List of captured registers depends on the architecture. This option 312is off by default. It is possible to select the registers to sample using their 313symbolic names, e.g. on x86, ax, si. To list the available registers use 314--intr-regs=\?. To name registers, pass a comma separated list such as 315--intr-regs=ax,bx. The list of register is architecture dependent. 316 317 318--running-time:: 319Record running and enabled time for read events (:S) 320 321-k:: 322--clockid:: 323Sets the clock id to use for the various time fields in the perf_event_type 324records. See clock_gettime(). In particular CLOCK_MONOTONIC and 325CLOCK_MONOTONIC_RAW are supported, some events might also allow 326CLOCK_BOOTTIME, CLOCK_REALTIME and CLOCK_TAI. 327 328-S:: 329--snapshot:: 330Select AUX area tracing Snapshot Mode. This option is valid only with an 331AUX area tracing event. Optionally the number of bytes to capture per 332snapshot can be specified. In Snapshot Mode, trace data is captured only when 333signal SIGUSR2 is received. 334 335--proc-map-timeout:: 336When processing pre-existing threads /proc/XXX/mmap, it may take a long time, 337because the file may be huge. A time out is needed in such cases. 338This option sets the time out limit. The default value is 500 ms. 339 340--switch-events:: 341Record context switch events i.e. events of type PERF_RECORD_SWITCH or 342PERF_RECORD_SWITCH_CPU_WIDE. 343 344--clang-path=PATH:: 345Path to clang binary to use for compiling BPF scriptlets. 346(enabled when BPF support is on) 347 348--clang-opt=OPTIONS:: 349Options passed to clang when compiling BPF scriptlets. 350(enabled when BPF support is on) 351 352--vmlinux=PATH:: 353Specify vmlinux path which has debuginfo. 354(enabled when BPF prologue is on) 355 356--buildid-all:: 357Record build-id of all DSOs regardless whether it's actually hit or not. 358 359--all-kernel:: 360Configure all used events to run in kernel space. 361 362--all-user:: 363Configure all used events to run in user space. 364 365--timestamp-filename 366Append timestamp to output file name. 367 368--switch-output:: 369Generate multiple perf.data files, timestamp prefixed, switching to a new one 370when receiving a SIGUSR2. 371 372A possible use case is to, given an external event, slice the perf.data file 373that gets then processed, possibly via a perf script, to decide if that 374particular perf.data snapshot should be kept or not. 375 376Implies --timestamp-filename, --no-buildid and --no-buildid-cache. 377 378--dry-run:: 379Parse options then exit. --dry-run can be used to detect errors in cmdline 380options. 381 382'perf record --dry-run -e' can act as a BPF script compiler if llvm.dump-obj 383in config file is set to true. 384 385--tail-synthesize:: 386Instead of collecting non-sample events (for example, fork, comm, mmap) at 387the beginning of record, collect them during finalizing an output file. 388The collected non-sample events reflects the status of the system when 389record is finished. 390 391--overwrite:: 392Makes all events use an overwritable ring buffer. An overwritable ring 393buffer works like a flight recorder: when it gets full, the kernel will 394overwrite the oldest records, that thus will never make it to the 395perf.data file. 396 397When '--overwrite' and '--switch-output' are used perf records and drops 398events until it receives a signal, meaning that something unusual was 399detected that warrants taking a snapshot of the most current events, 400those fitting in the ring buffer at that moment. 401 402'overwrite' attribute can also be set or canceled for an event using 403config terms. For example: 'cycles/overwrite/' and 'instructions/no-overwrite/'. 404 405Implies --tail-synthesize. 406 407SEE ALSO 408-------- 409linkperf:perf-stat[1], linkperf:perf-list[1] 410