3072 lines · plain
1================2Event Histograms3================4 5Documentation written by Tom Zanussi6 71. Introduction8===============9 10 Histogram triggers are special event triggers that can be used to11 aggregate trace event data into histograms. For information on12 trace events and event triggers, see Documentation/trace/events.rst.13 14 152. Histogram Trigger Command16============================17 18 A histogram trigger command is an event trigger command that19 aggregates event hits into a hash table keyed on one or more trace20 event format fields (or stacktrace) and a set of running totals21 derived from one or more trace event format fields and/or event22 counts (hitcount).23 24 The format of a hist trigger is as follows::25 26 hist:keys=<field1[,field2,...]>[:values=<field1[,field2,...]>]27 [:sort=<field1[,field2,...]>][:size=#entries][:pause][:continue]28 [:clear][:name=histname1][:nohitcount][:<handler>.<action>] [if <filter>]29 30 When a matching event is hit, an entry is added to a hash table31 using the key(s) and value(s) named. Keys and values correspond to32 fields in the event's format description. Values must correspond to33 numeric fields - on an event hit, the value(s) will be added to a34 sum kept for that field. The special string 'hitcount' can be used35 in place of an explicit value field - this is simply a count of36 event hits. If 'values' isn't specified, an implicit 'hitcount'37 value will be automatically created and used as the only value.38 Keys can be any field, or the special string 'common_stacktrace', which39 will use the event's kernel stacktrace as the key. The keywords40 'keys' or 'key' can be used to specify keys, and the keywords41 'values', 'vals', or 'val' can be used to specify values. Compound42 keys consisting of up to three fields can be specified by the 'keys'43 keyword. Hashing a compound key produces a unique entry in the44 table for each unique combination of component keys, and can be45 useful for providing more fine-grained summaries of event data.46 Additionally, sort keys consisting of up to two fields can be47 specified by the 'sort' keyword. If more than one field is48 specified, the result will be a 'sort within a sort': the first key49 is taken to be the primary sort key and the second the secondary50 key. If a hist trigger is given a name using the 'name' parameter,51 its histogram data will be shared with other triggers of the same52 name, and trigger hits will update this common data. Only triggers53 with 'compatible' fields can be combined in this way; triggers are54 'compatible' if the fields named in the trigger share the same55 number and type of fields and those fields also have the same names.56 Note that any two events always share the compatible 'hitcount' and57 'common_stacktrace' fields and can therefore be combined using those58 fields, however pointless that may be.59 60 'hist' triggers add a 'hist' file to each event's subdirectory.61 Reading the 'hist' file for the event will dump the hash table in62 its entirety to stdout. If there are multiple hist triggers63 attached to an event, there will be a table for each trigger in the64 output. The table displayed for a named trigger will be the same as65 any other instance having the same name. Each printed hash table66 entry is a simple list of the keys and values comprising the entry;67 keys are printed first and are delineated by curly braces, and are68 followed by the set of value fields for the entry. By default,69 numeric fields are displayed as base-10 integers. This can be70 modified by appending any of the following modifiers to the field71 name:72 73 ============= =================================================74 .hex display a number as a hex value75 .sym display an address as a symbol76 .sym-offset display an address as a symbol and offset77 .syscall display a syscall id as a system call name78 .execname display a common_pid as a program name79 .log2 display log2 value rather than raw number80 .buckets=size display grouping of values rather than raw number81 .usecs display a common_timestamp in microseconds82 .percent display a number of percentage value83 .graph display a bar-graph of a value84 .stacktrace display as a stacktrace (must by a long[] type)85 ============= =================================================86 87 Note that in general the semantics of a given field aren't88 interpreted when applying a modifier to it, but there are some89 restrictions to be aware of in this regard:90 91 - only the 'hex' modifier can be used for values (because values92 are essentially sums, and the other modifiers don't make sense93 in that context).94 - the 'execname' modifier can only be used on a 'common_pid'. The95 reason for this is that the execname is simply the 'comm' value96 saved for the 'current' process when an event was triggered,97 which is the same as the common_pid value saved by the event98 tracing code. Trying to apply that comm value to other pid99 values wouldn't be correct, and typically events that care save100 pid-specific comm fields in the event itself.101 102 A typical usage scenario would be the following to enable a hist103 trigger, read its current contents, and then turn it off::104 105 # echo 'hist:keys=skbaddr.hex:vals=len' > \106 /sys/kernel/tracing/events/net/netif_rx/trigger107 108 # cat /sys/kernel/tracing/events/net/netif_rx/hist109 110 # echo '!hist:keys=skbaddr.hex:vals=len' > \111 /sys/kernel/tracing/events/net/netif_rx/trigger112 113 The trigger file itself can be read to show the details of the114 currently attached hist trigger. This information is also displayed115 at the top of the 'hist' file when read.116 117 By default, the size of the hash table is 2048 entries. The 'size'118 parameter can be used to specify more or fewer than that. The units119 are in terms of hashtable entries - if a run uses more entries than120 specified, the results will show the number of 'drops', the number121 of hits that were ignored. The size should be a power of 2 between122 128 and 131072 (any non- power-of-2 number specified will be rounded123 up).124 125 The 'sort' parameter can be used to specify a value field to sort126 on. The default if unspecified is 'hitcount' and the default sort127 order is 'ascending'. To sort in the opposite direction, append128 .descending' to the sort key.129 130 The 'pause' parameter can be used to pause an existing hist trigger131 or to start a hist trigger but not log any events until told to do132 so. 'continue' or 'cont' can be used to start or restart a paused133 hist trigger.134 135 The 'clear' parameter will clear the contents of a running hist136 trigger and leave its current paused/active state.137 138 Note that the 'pause', 'cont', and 'clear' parameters should be139 applied using 'append' shell operator ('>>') if applied to an140 existing trigger, rather than via the '>' operator, which will cause141 the trigger to be removed through truncation.142 143 The 'nohitcount' (or NOHC) parameter will suppress display of144 raw hitcount in the histogram. This option requires at least one145 value field which is not a 'raw hitcount'. For example,146 'hist:...:vals=hitcount:nohitcount' is rejected, but147 'hist:...:vals=hitcount.percent:nohitcount' is OK.148 149- enable_hist/disable_hist150 151 The enable_hist and disable_hist triggers can be used to have one152 event conditionally start and stop another event's already-attached153 hist trigger. Any number of enable_hist and disable_hist triggers154 can be attached to a given event, allowing that event to kick off155 and stop aggregations on a host of other events.156 157 The format is very similar to the enable/disable_event triggers::158 159 enable_hist:<system>:<event>[:count]160 disable_hist:<system>:<event>[:count]161 162 Instead of enabling or disabling the tracing of the target event163 into the trace buffer as the enable/disable_event triggers do, the164 enable/disable_hist triggers enable or disable the aggregation of165 the target event into a hash table.166 167 A typical usage scenario for the enable_hist/disable_hist triggers168 would be to first set up a paused hist trigger on some event,169 followed by an enable_hist/disable_hist pair that turns the hist170 aggregation on and off when conditions of interest are hit::171 172 # echo 'hist:keys=skbaddr.hex:vals=len:pause' > \173 /sys/kernel/tracing/events/net/netif_receive_skb/trigger174 175 # echo 'enable_hist:net:netif_receive_skb if filename==/usr/bin/wget' > \176 /sys/kernel/tracing/events/sched/sched_process_exec/trigger177 178 # echo 'disable_hist:net:netif_receive_skb if comm==wget' > \179 /sys/kernel/tracing/events/sched/sched_process_exit/trigger180 181 The above sets up an initially paused hist trigger which is unpaused182 and starts aggregating events when a given program is executed, and183 which stops aggregating when the process exits and the hist trigger184 is paused again.185 186 The examples below provide a more concrete illustration of the187 concepts and typical usage patterns discussed above.188 189'special' event fields190------------------------191 192 There are a number of 'special event fields' available for use as193 keys or values in a hist trigger. These look like and behave as if194 they were actual event fields, but aren't really part of the event's195 field definition or format file. They are however available for any196 event, and can be used anywhere an actual event field could be.197 They are:198 199 ====================== ==== =======================================200 common_timestamp u64 timestamp (from ring buffer) associated201 with the event, in nanoseconds. May be202 modified by .usecs to have timestamps203 interpreted as microseconds.204 common_cpu int the cpu on which the event occurred.205 ====================== ==== =======================================206 207Extended error information208--------------------------209 210 For some error conditions encountered when invoking a hist trigger211 command, extended error information is available via the212 tracing/error_log file. See Error Conditions in213 :file:`Documentation/trace/ftrace.rst` for details.214 2156.2 'hist' trigger examples216---------------------------217 218 The first set of examples creates aggregations using the kmalloc219 event. The fields that can be used for the hist trigger are listed220 in the kmalloc event's format file::221 222 # cat /sys/kernel/tracing/events/kmem/kmalloc/format223 name: kmalloc224 ID: 374225 format:226 field:unsigned short common_type; offset:0; size:2; signed:0;227 field:unsigned char common_flags; offset:2; size:1; signed:0;228 field:unsigned char common_preempt_count; offset:3; size:1; signed:0;229 field:int common_pid; offset:4; size:4; signed:1;230 231 field:unsigned long call_site; offset:8; size:8; signed:0;232 field:const void * ptr; offset:16; size:8; signed:0;233 field:size_t bytes_req; offset:24; size:8; signed:0;234 field:size_t bytes_alloc; offset:32; size:8; signed:0;235 field:gfp_t gfp_flags; offset:40; size:4; signed:0;236 237 We'll start by creating a hist trigger that generates a simple table238 that lists the total number of bytes requested for each function in239 the kernel that made one or more calls to kmalloc::240 241 # echo 'hist:key=call_site:val=bytes_req.buckets=32' > \242 /sys/kernel/tracing/events/kmem/kmalloc/trigger243 244 This tells the tracing system to create a 'hist' trigger using the245 call_site field of the kmalloc event as the key for the table, which246 just means that each unique call_site address will have an entry247 created for it in the table. The 'val=bytes_req' parameter tells248 the hist trigger that for each unique entry (call_site) in the249 table, it should keep a running total of the number of bytes250 requested by that call_site.251 252 We'll let it run for awhile and then dump the contents of the 'hist'253 file in the kmalloc event's subdirectory (for readability, a number254 of entries have been omitted)::255 256 # cat /sys/kernel/tracing/events/kmem/kmalloc/hist257 # trigger info: hist:keys=call_site:vals=bytes_req:sort=hitcount:size=2048 [active]258 259 { call_site: 18446744072106379007 } hitcount: 1 bytes_req: 176260 { call_site: 18446744071579557049 } hitcount: 1 bytes_req: 1024261 { call_site: 18446744071580608289 } hitcount: 1 bytes_req: 16384262 { call_site: 18446744071581827654 } hitcount: 1 bytes_req: 24263 { call_site: 18446744071580700980 } hitcount: 1 bytes_req: 8264 { call_site: 18446744071579359876 } hitcount: 1 bytes_req: 152265 { call_site: 18446744071580795365 } hitcount: 3 bytes_req: 144266 { call_site: 18446744071581303129 } hitcount: 3 bytes_req: 144267 { call_site: 18446744071580713234 } hitcount: 4 bytes_req: 2560268 { call_site: 18446744071580933750 } hitcount: 4 bytes_req: 736269 .270 .271 .272 { call_site: 18446744072106047046 } hitcount: 69 bytes_req: 5576273 { call_site: 18446744071582116407 } hitcount: 73 bytes_req: 2336274 { call_site: 18446744072106054684 } hitcount: 136 bytes_req: 140504275 { call_site: 18446744072106224230 } hitcount: 136 bytes_req: 19584276 { call_site: 18446744072106078074 } hitcount: 153 bytes_req: 2448277 { call_site: 18446744072106062406 } hitcount: 153 bytes_req: 36720278 { call_site: 18446744071582507929 } hitcount: 153 bytes_req: 37088279 { call_site: 18446744072102520590 } hitcount: 273 bytes_req: 10920280 { call_site: 18446744071582143559 } hitcount: 358 bytes_req: 716281 { call_site: 18446744072106465852 } hitcount: 417 bytes_req: 56712282 { call_site: 18446744072102523378 } hitcount: 485 bytes_req: 27160283 { call_site: 18446744072099568646 } hitcount: 1676 bytes_req: 33520284 285 Totals:286 Hits: 4610287 Entries: 45288 Dropped: 0289 290 The output displays a line for each entry, beginning with the key291 specified in the trigger, followed by the value(s) also specified in292 the trigger. At the beginning of the output is a line that displays293 the trigger info, which can also be displayed by reading the294 'trigger' file::295 296 # cat /sys/kernel/tracing/events/kmem/kmalloc/trigger297 hist:keys=call_site:vals=bytes_req:sort=hitcount:size=2048 [active]298 299 At the end of the output are a few lines that display the overall300 totals for the run. The 'Hits' field shows the total number of301 times the event trigger was hit, the 'Entries' field shows the total302 number of used entries in the hash table, and the 'Dropped' field303 shows the number of hits that were dropped because the number of304 used entries for the run exceeded the maximum number of entries305 allowed for the table (normally 0, but if not a hint that you may306 want to increase the size of the table using the 'size' parameter).307 308 Notice in the above output that there's an extra field, 'hitcount',309 which wasn't specified in the trigger. Also notice that in the310 trigger info output, there's a parameter, 'sort=hitcount', which311 wasn't specified in the trigger either. The reason for that is that312 every trigger implicitly keeps a count of the total number of hits313 attributed to a given entry, called the 'hitcount'. That hitcount314 information is explicitly displayed in the output, and in the315 absence of a user-specified sort parameter, is used as the default316 sort field.317 318 The value 'hitcount' can be used in place of an explicit value in319 the 'values' parameter if you don't really need to have any320 particular field summed and are mainly interested in hit321 frequencies.322 323 To turn the hist trigger off, simply call up the trigger in the324 command history and re-execute it with a '!' prepended::325 326 # echo '!hist:key=call_site:val=bytes_req' > \327 /sys/kernel/tracing/events/kmem/kmalloc/trigger328 329 Finally, notice that the call_site as displayed in the output above330 isn't really very useful. It's an address, but normally addresses331 are displayed in hex. To have a numeric field displayed as a hex332 value, simply append '.hex' to the field name in the trigger::333 334 # echo 'hist:key=call_site.hex:val=bytes_req' > \335 /sys/kernel/tracing/events/kmem/kmalloc/trigger336 337 # cat /sys/kernel/tracing/events/kmem/kmalloc/hist338 # trigger info: hist:keys=call_site.hex:vals=bytes_req:sort=hitcount:size=2048 [active]339 340 { call_site: ffffffffa026b291 } hitcount: 1 bytes_req: 433341 { call_site: ffffffffa07186ff } hitcount: 1 bytes_req: 176342 { call_site: ffffffff811ae721 } hitcount: 1 bytes_req: 16384343 { call_site: ffffffff811c5134 } hitcount: 1 bytes_req: 8344 { call_site: ffffffffa04a9ebb } hitcount: 1 bytes_req: 511345 { call_site: ffffffff8122e0a6 } hitcount: 1 bytes_req: 12346 { call_site: ffffffff8107da84 } hitcount: 1 bytes_req: 152347 { call_site: ffffffff812d8246 } hitcount: 1 bytes_req: 24348 { call_site: ffffffff811dc1e5 } hitcount: 3 bytes_req: 144349 { call_site: ffffffffa02515e8 } hitcount: 3 bytes_req: 648350 { call_site: ffffffff81258159 } hitcount: 3 bytes_req: 144351 { call_site: ffffffff811c80f4 } hitcount: 4 bytes_req: 544352 .353 .354 .355 { call_site: ffffffffa06c7646 } hitcount: 106 bytes_req: 8024356 { call_site: ffffffffa06cb246 } hitcount: 132 bytes_req: 31680357 { call_site: ffffffffa06cef7a } hitcount: 132 bytes_req: 2112358 { call_site: ffffffff8137e399 } hitcount: 132 bytes_req: 23232359 { call_site: ffffffffa06c941c } hitcount: 185 bytes_req: 171360360 { call_site: ffffffffa06f2a66 } hitcount: 185 bytes_req: 26640361 { call_site: ffffffffa036a70e } hitcount: 265 bytes_req: 10600362 { call_site: ffffffff81325447 } hitcount: 292 bytes_req: 584363 { call_site: ffffffffa072da3c } hitcount: 446 bytes_req: 60656364 { call_site: ffffffffa036b1f2 } hitcount: 526 bytes_req: 29456365 { call_site: ffffffffa0099c06 } hitcount: 1780 bytes_req: 35600366 367 Totals:368 Hits: 4775369 Entries: 46370 Dropped: 0371 372 Even that's only marginally more useful - while hex values do look373 more like addresses, what users are typically more interested in374 when looking at text addresses are the corresponding symbols375 instead. To have an address displayed as symbolic value instead,376 simply append '.sym' or '.sym-offset' to the field name in the377 trigger::378 379 # echo 'hist:key=call_site.sym:val=bytes_req' > \380 /sys/kernel/tracing/events/kmem/kmalloc/trigger381 382 # cat /sys/kernel/tracing/events/kmem/kmalloc/hist383 # trigger info: hist:keys=call_site.sym:vals=bytes_req:sort=hitcount:size=2048 [active]384 385 { call_site: [ffffffff810adcb9] syslog_print_all } hitcount: 1 bytes_req: 1024386 { call_site: [ffffffff8154bc62] usb_control_msg } hitcount: 1 bytes_req: 8387 { call_site: [ffffffffa00bf6fe] hidraw_send_report [hid] } hitcount: 1 bytes_req: 7388 { call_site: [ffffffff8154acbe] usb_alloc_urb } hitcount: 1 bytes_req: 192389 { call_site: [ffffffffa00bf1ca] hidraw_report_event [hid] } hitcount: 1 bytes_req: 7390 { call_site: [ffffffff811e3a25] __seq_open_private } hitcount: 1 bytes_req: 40391 { call_site: [ffffffff8109524a] alloc_fair_sched_group } hitcount: 2 bytes_req: 128392 { call_site: [ffffffff811febd5] fsnotify_alloc_group } hitcount: 2 bytes_req: 528393 { call_site: [ffffffff81440f58] __tty_buffer_request_room } hitcount: 2 bytes_req: 2624394 { call_site: [ffffffff81200ba6] inotify_new_group } hitcount: 2 bytes_req: 96395 { call_site: [ffffffffa05e19af] ieee80211_start_tx_ba_session [mac80211] } hitcount: 2 bytes_req: 464396 { call_site: [ffffffff81672406] tcp_get_metrics } hitcount: 2 bytes_req: 304397 { call_site: [ffffffff81097ec2] alloc_rt_sched_group } hitcount: 2 bytes_req: 128398 { call_site: [ffffffff81089b05] sched_create_group } hitcount: 2 bytes_req: 1424399 .400 .401 .402 { call_site: [ffffffffa04a580c] intel_crtc_page_flip [i915] } hitcount: 1185 bytes_req: 123240403 { call_site: [ffffffffa0287592] drm_mode_page_flip_ioctl [drm] } hitcount: 1185 bytes_req: 104280404 { call_site: [ffffffffa04c4a3c] intel_plane_duplicate_state [i915] } hitcount: 1402 bytes_req: 190672405 { call_site: [ffffffff812891ca] ext4_find_extent } hitcount: 1518 bytes_req: 146208406 { call_site: [ffffffffa029070e] drm_vma_node_allow [drm] } hitcount: 1746 bytes_req: 69840407 { call_site: [ffffffffa045e7c4] i915_gem_do_execbuffer.isra.23 [i915] } hitcount: 2021 bytes_req: 792312408 { call_site: [ffffffffa02911f2] drm_modeset_lock_crtc [drm] } hitcount: 2592 bytes_req: 145152409 { call_site: [ffffffffa0489a66] intel_ring_begin [i915] } hitcount: 2629 bytes_req: 378576410 { call_site: [ffffffffa046041c] i915_gem_execbuffer2 [i915] } hitcount: 2629 bytes_req: 3783248411 { call_site: [ffffffff81325607] apparmor_file_alloc_security } hitcount: 5192 bytes_req: 10384412 { call_site: [ffffffffa00b7c06] hid_report_raw_event [hid] } hitcount: 5529 bytes_req: 110584413 { call_site: [ffffffff8131ebf7] aa_alloc_task_context } hitcount: 21943 bytes_req: 702176414 { call_site: [ffffffff8125847d] ext4_htree_store_dirent } hitcount: 55759 bytes_req: 5074265415 416 Totals:417 Hits: 109928418 Entries: 71419 Dropped: 0420 421 Because the default sort key above is 'hitcount', the above shows a422 the list of call_sites by increasing hitcount, so that at the bottom423 we see the functions that made the most kmalloc calls during the424 run. If instead we wanted to see the top kmalloc callers in425 terms of the number of bytes requested rather than the number of426 calls, and we wanted the top caller to appear at the top, we can use427 the 'sort' parameter, along with the 'descending' modifier::428 429 # echo 'hist:key=call_site.sym:val=bytes_req:sort=bytes_req.descending' > \430 /sys/kernel/tracing/events/kmem/kmalloc/trigger431 432 # cat /sys/kernel/tracing/events/kmem/kmalloc/hist433 # trigger info: hist:keys=call_site.sym:vals=bytes_req:sort=bytes_req.descending:size=2048 [active]434 435 { call_site: [ffffffffa046041c] i915_gem_execbuffer2 [i915] } hitcount: 2186 bytes_req: 3397464436 { call_site: [ffffffffa045e7c4] i915_gem_do_execbuffer.isra.23 [i915] } hitcount: 1790 bytes_req: 712176437 { call_site: [ffffffff8125847d] ext4_htree_store_dirent } hitcount: 8132 bytes_req: 513135438 { call_site: [ffffffff811e2a1b] seq_buf_alloc } hitcount: 106 bytes_req: 440128439 { call_site: [ffffffffa0489a66] intel_ring_begin [i915] } hitcount: 2186 bytes_req: 314784440 { call_site: [ffffffff812891ca] ext4_find_extent } hitcount: 2174 bytes_req: 208992441 { call_site: [ffffffff811ae8e1] __kmalloc } hitcount: 8 bytes_req: 131072442 { call_site: [ffffffffa04c4a3c] intel_plane_duplicate_state [i915] } hitcount: 859 bytes_req: 116824443 { call_site: [ffffffffa02911f2] drm_modeset_lock_crtc [drm] } hitcount: 1834 bytes_req: 102704444 { call_site: [ffffffffa04a580c] intel_crtc_page_flip [i915] } hitcount: 972 bytes_req: 101088445 { call_site: [ffffffffa0287592] drm_mode_page_flip_ioctl [drm] } hitcount: 972 bytes_req: 85536446 { call_site: [ffffffffa00b7c06] hid_report_raw_event [hid] } hitcount: 3333 bytes_req: 66664447 { call_site: [ffffffff8137e559] sg_kmalloc } hitcount: 209 bytes_req: 61632448 .449 .450 .451 { call_site: [ffffffff81095225] alloc_fair_sched_group } hitcount: 2 bytes_req: 128452 { call_site: [ffffffff81097ec2] alloc_rt_sched_group } hitcount: 2 bytes_req: 128453 { call_site: [ffffffff812d8406] copy_semundo } hitcount: 2 bytes_req: 48454 { call_site: [ffffffff81200ba6] inotify_new_group } hitcount: 1 bytes_req: 48455 { call_site: [ffffffffa027121a] drm_getmagic [drm] } hitcount: 1 bytes_req: 48456 { call_site: [ffffffff811e3a25] __seq_open_private } hitcount: 1 bytes_req: 40457 { call_site: [ffffffff811c52f4] bprm_change_interp } hitcount: 2 bytes_req: 16458 { call_site: [ffffffff8154bc62] usb_control_msg } hitcount: 1 bytes_req: 8459 { call_site: [ffffffffa00bf1ca] hidraw_report_event [hid] } hitcount: 1 bytes_req: 7460 { call_site: [ffffffffa00bf6fe] hidraw_send_report [hid] } hitcount: 1 bytes_req: 7461 462 Totals:463 Hits: 32133464 Entries: 81465 Dropped: 0466 467 To display the offset and size information in addition to the symbol468 name, just use 'sym-offset' instead::469 470 # echo 'hist:key=call_site.sym-offset:val=bytes_req:sort=bytes_req.descending' > \471 /sys/kernel/tracing/events/kmem/kmalloc/trigger472 473 # cat /sys/kernel/tracing/events/kmem/kmalloc/hist474 # trigger info: hist:keys=call_site.sym-offset:vals=bytes_req:sort=bytes_req.descending:size=2048 [active]475 476 { call_site: [ffffffffa046041c] i915_gem_execbuffer2+0x6c/0x2c0 [i915] } hitcount: 4569 bytes_req: 3163720477 { call_site: [ffffffffa0489a66] intel_ring_begin+0xc6/0x1f0 [i915] } hitcount: 4569 bytes_req: 657936478 { call_site: [ffffffffa045e7c4] i915_gem_do_execbuffer.isra.23+0x694/0x1020 [i915] } hitcount: 1519 bytes_req: 472936479 { call_site: [ffffffffa045e646] i915_gem_do_execbuffer.isra.23+0x516/0x1020 [i915] } hitcount: 3050 bytes_req: 211832480 { call_site: [ffffffff811e2a1b] seq_buf_alloc+0x1b/0x50 } hitcount: 34 bytes_req: 148384481 { call_site: [ffffffffa04a580c] intel_crtc_page_flip+0xbc/0x870 [i915] } hitcount: 1385 bytes_req: 144040482 { call_site: [ffffffff811ae8e1] __kmalloc+0x191/0x1b0 } hitcount: 8 bytes_req: 131072483 { call_site: [ffffffffa0287592] drm_mode_page_flip_ioctl+0x282/0x360 [drm] } hitcount: 1385 bytes_req: 121880484 { call_site: [ffffffffa02911f2] drm_modeset_lock_crtc+0x32/0x100 [drm] } hitcount: 1848 bytes_req: 103488485 { call_site: [ffffffffa04c4a3c] intel_plane_duplicate_state+0x2c/0xa0 [i915] } hitcount: 461 bytes_req: 62696486 { call_site: [ffffffffa029070e] drm_vma_node_allow+0x2e/0xd0 [drm] } hitcount: 1541 bytes_req: 61640487 { call_site: [ffffffff815f8d7b] sk_prot_alloc+0xcb/0x1b0 } hitcount: 57 bytes_req: 57456488 .489 .490 .491 { call_site: [ffffffff8109524a] alloc_fair_sched_group+0x5a/0x1a0 } hitcount: 2 bytes_req: 128492 { call_site: [ffffffffa027b921] drm_vm_open_locked+0x31/0xa0 [drm] } hitcount: 3 bytes_req: 96493 { call_site: [ffffffff8122e266] proc_self_follow_link+0x76/0xb0 } hitcount: 8 bytes_req: 96494 { call_site: [ffffffff81213e80] load_elf_binary+0x240/0x1650 } hitcount: 3 bytes_req: 84495 { call_site: [ffffffff8154bc62] usb_control_msg+0x42/0x110 } hitcount: 1 bytes_req: 8496 { call_site: [ffffffffa00bf6fe] hidraw_send_report+0x7e/0x1a0 [hid] } hitcount: 1 bytes_req: 7497 { call_site: [ffffffffa00bf1ca] hidraw_report_event+0x8a/0x120 [hid] } hitcount: 1 bytes_req: 7498 499 Totals:500 Hits: 26098501 Entries: 64502 Dropped: 0503 504 We can also add multiple fields to the 'values' parameter. For505 example, we might want to see the total number of bytes allocated506 alongside bytes requested, and display the result sorted by bytes507 allocated in a descending order::508 509 # echo 'hist:keys=call_site.sym:values=bytes_req,bytes_alloc:sort=bytes_alloc.descending' > \510 /sys/kernel/tracing/events/kmem/kmalloc/trigger511 512 # cat /sys/kernel/tracing/events/kmem/kmalloc/hist513 # trigger info: hist:keys=call_site.sym:vals=bytes_req,bytes_alloc:sort=bytes_alloc.descending:size=2048 [active]514 515 { call_site: [ffffffffa046041c] i915_gem_execbuffer2 [i915] } hitcount: 7403 bytes_req: 4084360 bytes_alloc: 5958016516 { call_site: [ffffffff811e2a1b] seq_buf_alloc } hitcount: 541 bytes_req: 2213968 bytes_alloc: 2228224517 { call_site: [ffffffffa0489a66] intel_ring_begin [i915] } hitcount: 7404 bytes_req: 1066176 bytes_alloc: 1421568518 { call_site: [ffffffffa045e7c4] i915_gem_do_execbuffer.isra.23 [i915] } hitcount: 1565 bytes_req: 557368 bytes_alloc: 1037760519 { call_site: [ffffffff8125847d] ext4_htree_store_dirent } hitcount: 9557 bytes_req: 595778 bytes_alloc: 695744520 { call_site: [ffffffffa045e646] i915_gem_do_execbuffer.isra.23 [i915] } hitcount: 5839 bytes_req: 430680 bytes_alloc: 470400521 { call_site: [ffffffffa04c4a3c] intel_plane_duplicate_state [i915] } hitcount: 2388 bytes_req: 324768 bytes_alloc: 458496522 { call_site: [ffffffffa02911f2] drm_modeset_lock_crtc [drm] } hitcount: 3911 bytes_req: 219016 bytes_alloc: 250304523 { call_site: [ffffffff815f8d7b] sk_prot_alloc } hitcount: 235 bytes_req: 236880 bytes_alloc: 240640524 { call_site: [ffffffff8137e559] sg_kmalloc } hitcount: 557 bytes_req: 169024 bytes_alloc: 221760525 { call_site: [ffffffffa00b7c06] hid_report_raw_event [hid] } hitcount: 9378 bytes_req: 187548 bytes_alloc: 206312526 { call_site: [ffffffffa04a580c] intel_crtc_page_flip [i915] } hitcount: 1519 bytes_req: 157976 bytes_alloc: 194432527 .528 .529 .530 { call_site: [ffffffff8109bd3b] sched_autogroup_create_attach } hitcount: 2 bytes_req: 144 bytes_alloc: 192531 { call_site: [ffffffff81097ee8] alloc_rt_sched_group } hitcount: 2 bytes_req: 128 bytes_alloc: 128532 { call_site: [ffffffff8109524a] alloc_fair_sched_group } hitcount: 2 bytes_req: 128 bytes_alloc: 128533 { call_site: [ffffffff81095225] alloc_fair_sched_group } hitcount: 2 bytes_req: 128 bytes_alloc: 128534 { call_site: [ffffffff81097ec2] alloc_rt_sched_group } hitcount: 2 bytes_req: 128 bytes_alloc: 128535 { call_site: [ffffffff81213e80] load_elf_binary } hitcount: 3 bytes_req: 84 bytes_alloc: 96536 { call_site: [ffffffff81079a2e] kthread_create_on_node } hitcount: 1 bytes_req: 56 bytes_alloc: 64537 { call_site: [ffffffffa00bf6fe] hidraw_send_report [hid] } hitcount: 1 bytes_req: 7 bytes_alloc: 8538 { call_site: [ffffffff8154bc62] usb_control_msg } hitcount: 1 bytes_req: 8 bytes_alloc: 8539 { call_site: [ffffffffa00bf1ca] hidraw_report_event [hid] } hitcount: 1 bytes_req: 7 bytes_alloc: 8540 541 Totals:542 Hits: 66598543 Entries: 65544 Dropped: 0545 546 Finally, to finish off our kmalloc example, instead of simply having547 the hist trigger display symbolic call_sites, we can have the hist548 trigger additionally display the complete set of kernel stack traces549 that led to each call_site. To do that, we simply use the special550 value 'common_stacktrace' for the key parameter::551 552 # echo 'hist:keys=common_stacktrace:values=bytes_req,bytes_alloc:sort=bytes_alloc' > \553 /sys/kernel/tracing/events/kmem/kmalloc/trigger554 555 The above trigger will use the kernel stack trace in effect when an556 event is triggered as the key for the hash table. This allows the557 enumeration of every kernel callpath that led up to a particular558 event, along with a running total of any of the event fields for559 that event. Here we tally bytes requested and bytes allocated for560 every callpath in the system that led up to a kmalloc (in this case561 every callpath to a kmalloc for a kernel compile)::562 563 # cat /sys/kernel/tracing/events/kmem/kmalloc/hist564 # trigger info: hist:keys=common_stacktrace:vals=bytes_req,bytes_alloc:sort=bytes_alloc:size=2048 [active]565 566 { common_stacktrace:567 __kmalloc_track_caller+0x10b/0x1a0568 kmemdup+0x20/0x50569 hidraw_report_event+0x8a/0x120 [hid]570 hid_report_raw_event+0x3ea/0x440 [hid]571 hid_input_report+0x112/0x190 [hid]572 hid_irq_in+0xc2/0x260 [usbhid]573 __usb_hcd_giveback_urb+0x72/0x120574 usb_giveback_urb_bh+0x9e/0xe0575 tasklet_hi_action+0xf8/0x100576 __do_softirq+0x114/0x2c0577 irq_exit+0xa5/0xb0578 do_IRQ+0x5a/0xf0579 ret_from_intr+0x0/0x30580 cpuidle_enter+0x17/0x20581 cpu_startup_entry+0x315/0x3e0582 rest_init+0x7c/0x80583 } hitcount: 3 bytes_req: 21 bytes_alloc: 24584 { common_stacktrace:585 __kmalloc_track_caller+0x10b/0x1a0586 kmemdup+0x20/0x50587 hidraw_report_event+0x8a/0x120 [hid]588 hid_report_raw_event+0x3ea/0x440 [hid]589 hid_input_report+0x112/0x190 [hid]590 hid_irq_in+0xc2/0x260 [usbhid]591 __usb_hcd_giveback_urb+0x72/0x120592 usb_giveback_urb_bh+0x9e/0xe0593 tasklet_hi_action+0xf8/0x100594 __do_softirq+0x114/0x2c0595 irq_exit+0xa5/0xb0596 do_IRQ+0x5a/0xf0597 ret_from_intr+0x0/0x30598 } hitcount: 3 bytes_req: 21 bytes_alloc: 24599 { common_stacktrace:600 kmem_cache_alloc_trace+0xeb/0x150601 aa_alloc_task_context+0x27/0x40602 apparmor_cred_prepare+0x1f/0x50603 security_prepare_creds+0x16/0x20604 prepare_creds+0xdf/0x1a0605 SyS_capset+0xb5/0x200606 system_call_fastpath+0x12/0x6a607 } hitcount: 1 bytes_req: 32 bytes_alloc: 32608 .609 .610 .611 { common_stacktrace:612 __kmalloc+0x11b/0x1b0613 i915_gem_execbuffer2+0x6c/0x2c0 [i915]614 drm_ioctl+0x349/0x670 [drm]615 do_vfs_ioctl+0x2f0/0x4f0616 SyS_ioctl+0x81/0xa0617 system_call_fastpath+0x12/0x6a618 } hitcount: 17726 bytes_req: 13944120 bytes_alloc: 19593808619 { common_stacktrace:620 __kmalloc+0x11b/0x1b0621 load_elf_phdrs+0x76/0xa0622 load_elf_binary+0x102/0x1650623 search_binary_handler+0x97/0x1d0624 do_execveat_common.isra.34+0x551/0x6e0625 SyS_execve+0x3a/0x50626 return_from_execve+0x0/0x23627 } hitcount: 33348 bytes_req: 17152128 bytes_alloc: 20226048628 { common_stacktrace:629 kmem_cache_alloc_trace+0xeb/0x150630 apparmor_file_alloc_security+0x27/0x40631 security_file_alloc+0x16/0x20632 get_empty_filp+0x93/0x1c0633 path_openat+0x31/0x5f0634 do_filp_open+0x3a/0x90635 do_sys_open+0x128/0x220636 SyS_open+0x1e/0x20637 system_call_fastpath+0x12/0x6a638 } hitcount: 4766422 bytes_req: 9532844 bytes_alloc: 38131376639 { common_stacktrace:640 __kmalloc+0x11b/0x1b0641 seq_buf_alloc+0x1b/0x50642 seq_read+0x2cc/0x370643 proc_reg_read+0x3d/0x80644 __vfs_read+0x28/0xe0645 vfs_read+0x86/0x140646 SyS_read+0x46/0xb0647 system_call_fastpath+0x12/0x6a648 } hitcount: 19133 bytes_req: 78368768 bytes_alloc: 78368768649 650 Totals:651 Hits: 6085872652 Entries: 253653 Dropped: 0654 655 If you key a hist trigger on common_pid, in order for example to656 gather and display sorted totals for each process, you can use the657 special .execname modifier to display the executable names for the658 processes in the table rather than raw pids. The example below659 keeps a per-process sum of total bytes read::660 661 # echo 'hist:key=common_pid.execname:val=count:sort=count.descending' > \662 /sys/kernel/tracing/events/syscalls/sys_enter_read/trigger663 664 # cat /sys/kernel/tracing/events/syscalls/sys_enter_read/hist665 # trigger info: hist:keys=common_pid.execname:vals=count:sort=count.descending:size=2048 [active]666 667 { common_pid: gnome-terminal [ 3196] } hitcount: 280 count: 1093512668 { common_pid: Xorg [ 1309] } hitcount: 525 count: 256640669 { common_pid: compiz [ 2889] } hitcount: 59 count: 254400670 { common_pid: bash [ 8710] } hitcount: 3 count: 66369671 { common_pid: dbus-daemon-lau [ 8703] } hitcount: 49 count: 47739672 { common_pid: irqbalance [ 1252] } hitcount: 27 count: 27648673 { common_pid: 01ifupdown [ 8705] } hitcount: 3 count: 17216674 { common_pid: dbus-daemon [ 772] } hitcount: 10 count: 12396675 { common_pid: Socket Thread [ 8342] } hitcount: 11 count: 11264676 { common_pid: nm-dhcp-client. [ 8701] } hitcount: 6 count: 7424677 { common_pid: gmain [ 1315] } hitcount: 18 count: 6336678 .679 .680 .681 { common_pid: postgres [ 1892] } hitcount: 2 count: 32682 { common_pid: postgres [ 1891] } hitcount: 2 count: 32683 { common_pid: gmain [ 8704] } hitcount: 2 count: 32684 { common_pid: upstart-dbus-br [ 2740] } hitcount: 21 count: 21685 { common_pid: nm-dispatcher.a [ 8696] } hitcount: 1 count: 16686 { common_pid: indicator-datet [ 2904] } hitcount: 1 count: 16687 { common_pid: gdbus [ 2998] } hitcount: 1 count: 16688 { common_pid: rtkit-daemon [ 2052] } hitcount: 1 count: 8689 { common_pid: init [ 1] } hitcount: 2 count: 2690 691 Totals:692 Hits: 2116693 Entries: 51694 Dropped: 0695 696 Similarly, if you key a hist trigger on syscall id, for example to697 gather and display a list of systemwide syscall hits, you can use698 the special .syscall modifier to display the syscall names rather699 than raw ids. The example below keeps a running total of syscall700 counts for the system during the run::701 702 # echo 'hist:key=id.syscall:val=hitcount' > \703 /sys/kernel/tracing/events/raw_syscalls/sys_enter/trigger704 705 # cat /sys/kernel/tracing/events/raw_syscalls/sys_enter/hist706 # trigger info: hist:keys=id.syscall:vals=hitcount:sort=hitcount:size=2048 [active]707 708 { id: sys_fsync [ 74] } hitcount: 1709 { id: sys_newuname [ 63] } hitcount: 1710 { id: sys_prctl [157] } hitcount: 1711 { id: sys_statfs [137] } hitcount: 1712 { id: sys_symlink [ 88] } hitcount: 1713 { id: sys_sendmmsg [307] } hitcount: 1714 { id: sys_semctl [ 66] } hitcount: 1715 { id: sys_readlink [ 89] } hitcount: 3716 { id: sys_bind [ 49] } hitcount: 3717 { id: sys_getsockname [ 51] } hitcount: 3718 { id: sys_unlink [ 87] } hitcount: 3719 { id: sys_rename [ 82] } hitcount: 4720 { id: unknown_syscall [ 58] } hitcount: 4721 { id: sys_connect [ 42] } hitcount: 4722 { id: sys_getpid [ 39] } hitcount: 4723 .724 .725 .726 { id: sys_rt_sigprocmask [ 14] } hitcount: 952727 { id: sys_futex [202] } hitcount: 1534728 { id: sys_write [ 1] } hitcount: 2689729 { id: sys_setitimer [ 38] } hitcount: 2797730 { id: sys_read [ 0] } hitcount: 3202731 { id: sys_select [ 23] } hitcount: 3773732 { id: sys_writev [ 20] } hitcount: 4531733 { id: sys_poll [ 7] } hitcount: 8314734 { id: sys_recvmsg [ 47] } hitcount: 13738735 { id: sys_ioctl [ 16] } hitcount: 21843736 737 Totals:738 Hits: 67612739 Entries: 72740 Dropped: 0741 742 The syscall counts above provide a rough overall picture of system743 call activity on the system; we can see for example that the most744 popular system call on this system was the 'sys_ioctl' system call.745 746 We can use 'compound' keys to refine that number and provide some747 further insight as to which processes exactly contribute to the748 overall ioctl count.749 750 The command below keeps a hitcount for every unique combination of751 system call id and pid - the end result is essentially a table752 that keeps a per-pid sum of system call hits. The results are753 sorted using the system call id as the primary key, and the754 hitcount sum as the secondary key::755 756 # echo 'hist:key=id.syscall,common_pid.execname:val=hitcount:sort=id,hitcount' > \757 /sys/kernel/tracing/events/raw_syscalls/sys_enter/trigger758 759 # cat /sys/kernel/tracing/events/raw_syscalls/sys_enter/hist760 # trigger info: hist:keys=id.syscall,common_pid.execname:vals=hitcount:sort=id.syscall,hitcount:size=2048 [active]761 762 { id: sys_read [ 0], common_pid: rtkit-daemon [ 1877] } hitcount: 1763 { id: sys_read [ 0], common_pid: gdbus [ 2976] } hitcount: 1764 { id: sys_read [ 0], common_pid: console-kit-dae [ 3400] } hitcount: 1765 { id: sys_read [ 0], common_pid: postgres [ 1865] } hitcount: 1766 { id: sys_read [ 0], common_pid: deja-dup-monito [ 3543] } hitcount: 2767 { id: sys_read [ 0], common_pid: NetworkManager [ 890] } hitcount: 2768 { id: sys_read [ 0], common_pid: evolution-calen [ 3048] } hitcount: 2769 { id: sys_read [ 0], common_pid: postgres [ 1864] } hitcount: 2770 { id: sys_read [ 0], common_pid: nm-applet [ 3022] } hitcount: 2771 { id: sys_read [ 0], common_pid: whoopsie [ 1212] } hitcount: 2772 .773 .774 .775 { id: sys_ioctl [ 16], common_pid: bash [ 8479] } hitcount: 1776 { id: sys_ioctl [ 16], common_pid: bash [ 3472] } hitcount: 12777 { id: sys_ioctl [ 16], common_pid: gnome-terminal [ 3199] } hitcount: 16778 { id: sys_ioctl [ 16], common_pid: Xorg [ 1267] } hitcount: 1808779 { id: sys_ioctl [ 16], common_pid: compiz [ 2994] } hitcount: 5580780 .781 .782 .783 { id: sys_waitid [247], common_pid: upstart-dbus-br [ 2690] } hitcount: 3784 { id: sys_waitid [247], common_pid: upstart-dbus-br [ 2688] } hitcount: 16785 { id: sys_inotify_add_watch [254], common_pid: gmain [ 975] } hitcount: 2786 { id: sys_inotify_add_watch [254], common_pid: gmain [ 3204] } hitcount: 4787 { id: sys_inotify_add_watch [254], common_pid: gmain [ 2888] } hitcount: 4788 { id: sys_inotify_add_watch [254], common_pid: gmain [ 3003] } hitcount: 4789 { id: sys_inotify_add_watch [254], common_pid: gmain [ 2873] } hitcount: 4790 { id: sys_inotify_add_watch [254], common_pid: gmain [ 3196] } hitcount: 6791 { id: sys_openat [257], common_pid: java [ 2623] } hitcount: 2792 { id: sys_eventfd2 [290], common_pid: ibus-ui-gtk3 [ 2760] } hitcount: 4793 { id: sys_eventfd2 [290], common_pid: compiz [ 2994] } hitcount: 6794 795 Totals:796 Hits: 31536797 Entries: 323798 Dropped: 0799 800 The above list does give us a breakdown of the ioctl syscall by801 pid, but it also gives us quite a bit more than that, which we802 don't really care about at the moment. Since we know the syscall803 id for sys_ioctl (16, displayed next to the sys_ioctl name), we804 can use that to filter out all the other syscalls::805 806 # echo 'hist:key=id.syscall,common_pid.execname:val=hitcount:sort=id,hitcount if id == 16' > \807 /sys/kernel/tracing/events/raw_syscalls/sys_enter/trigger808 809 # cat /sys/kernel/tracing/events/raw_syscalls/sys_enter/hist810 # trigger info: hist:keys=id.syscall,common_pid.execname:vals=hitcount:sort=id.syscall,hitcount:size=2048 if id == 16 [active]811 812 { id: sys_ioctl [ 16], common_pid: gmain [ 2769] } hitcount: 1813 { id: sys_ioctl [ 16], common_pid: evolution-addre [ 8571] } hitcount: 1814 { id: sys_ioctl [ 16], common_pid: gmain [ 3003] } hitcount: 1815 { id: sys_ioctl [ 16], common_pid: gmain [ 2781] } hitcount: 1816 { id: sys_ioctl [ 16], common_pid: gmain [ 2829] } hitcount: 1817 { id: sys_ioctl [ 16], common_pid: bash [ 8726] } hitcount: 1818 { id: sys_ioctl [ 16], common_pid: bash [ 8508] } hitcount: 1819 { id: sys_ioctl [ 16], common_pid: gmain [ 2970] } hitcount: 1820 { id: sys_ioctl [ 16], common_pid: gmain [ 2768] } hitcount: 1821 .822 .823 .824 { id: sys_ioctl [ 16], common_pid: pool [ 8559] } hitcount: 45825 { id: sys_ioctl [ 16], common_pid: pool [ 8555] } hitcount: 48826 { id: sys_ioctl [ 16], common_pid: pool [ 8551] } hitcount: 48827 { id: sys_ioctl [ 16], common_pid: avahi-daemon [ 896] } hitcount: 66828 { id: sys_ioctl [ 16], common_pid: Xorg [ 1267] } hitcount: 26674829 { id: sys_ioctl [ 16], common_pid: compiz [ 2994] } hitcount: 73443830 831 Totals:832 Hits: 101162833 Entries: 103834 Dropped: 0835 836 The above output shows that 'compiz' and 'Xorg' are far and away837 the heaviest ioctl callers (which might lead to questions about838 whether they really need to be making all those calls and to839 possible avenues for further investigation.)840 841 The compound key examples used a key and a sum value (hitcount) to842 sort the output, but we can just as easily use two keys instead.843 Here's an example where we use a compound key composed of the the844 common_pid and size event fields. Sorting with pid as the primary845 key and 'size' as the secondary key allows us to display an846 ordered summary of the recvfrom sizes, with counts, received by847 each process::848 849 # echo 'hist:key=common_pid.execname,size:val=hitcount:sort=common_pid,size' > \850 /sys/kernel/tracing/events/syscalls/sys_enter_recvfrom/trigger851 852 # cat /sys/kernel/tracing/events/syscalls/sys_enter_recvfrom/hist853 # trigger info: hist:keys=common_pid.execname,size:vals=hitcount:sort=common_pid.execname,size:size=2048 [active]854 855 { common_pid: smbd [ 784], size: 4 } hitcount: 1856 { common_pid: dnsmasq [ 1412], size: 4096 } hitcount: 672857 { common_pid: postgres [ 1796], size: 1000 } hitcount: 6858 { common_pid: postgres [ 1867], size: 1000 } hitcount: 10859 { common_pid: bamfdaemon [ 2787], size: 28 } hitcount: 2860 { common_pid: bamfdaemon [ 2787], size: 14360 } hitcount: 1861 { common_pid: compiz [ 2994], size: 8 } hitcount: 1862 { common_pid: compiz [ 2994], size: 20 } hitcount: 11863 { common_pid: gnome-terminal [ 3199], size: 4 } hitcount: 2864 { common_pid: firefox [ 8817], size: 4 } hitcount: 1865 { common_pid: firefox [ 8817], size: 8 } hitcount: 5866 { common_pid: firefox [ 8817], size: 588 } hitcount: 2867 { common_pid: firefox [ 8817], size: 628 } hitcount: 1868 { common_pid: firefox [ 8817], size: 6944 } hitcount: 1869 { common_pid: firefox [ 8817], size: 408880 } hitcount: 2870 { common_pid: firefox [ 8822], size: 8 } hitcount: 2871 { common_pid: firefox [ 8822], size: 160 } hitcount: 2872 { common_pid: firefox [ 8822], size: 320 } hitcount: 2873 { common_pid: firefox [ 8822], size: 352 } hitcount: 1874 .875 .876 .877 { common_pid: pool [ 8923], size: 1960 } hitcount: 10878 { common_pid: pool [ 8923], size: 2048 } hitcount: 10879 { common_pid: pool [ 8924], size: 1960 } hitcount: 10880 { common_pid: pool [ 8924], size: 2048 } hitcount: 10881 { common_pid: pool [ 8928], size: 1964 } hitcount: 4882 { common_pid: pool [ 8928], size: 1965 } hitcount: 2883 { common_pid: pool [ 8928], size: 2048 } hitcount: 6884 { common_pid: pool [ 8929], size: 1982 } hitcount: 1885 { common_pid: pool [ 8929], size: 2048 } hitcount: 1886 887 Totals:888 Hits: 2016889 Entries: 224890 Dropped: 0891 892 The above example also illustrates the fact that although a compound893 key is treated as a single entity for hashing purposes, the sub-keys894 it's composed of can be accessed independently.895 896 The next example uses a string field as the hash key and897 demonstrates how you can manually pause and continue a hist trigger.898 In this example, we'll aggregate fork counts and don't expect a899 large number of entries in the hash table, so we'll drop it to a900 much smaller number, say 256::901 902 # echo 'hist:key=child_comm:val=hitcount:size=256' > \903 /sys/kernel/tracing/events/sched/sched_process_fork/trigger904 905 # cat /sys/kernel/tracing/events/sched/sched_process_fork/hist906 # trigger info: hist:keys=child_comm:vals=hitcount:sort=hitcount:size=256 [active]907 908 { child_comm: dconf worker } hitcount: 1909 { child_comm: ibus-daemon } hitcount: 1910 { child_comm: whoopsie } hitcount: 1911 { child_comm: smbd } hitcount: 1912 { child_comm: gdbus } hitcount: 1913 { child_comm: kthreadd } hitcount: 1914 { child_comm: dconf worker } hitcount: 1915 { child_comm: evolution-alarm } hitcount: 2916 { child_comm: Socket Thread } hitcount: 2917 { child_comm: postgres } hitcount: 2918 { child_comm: bash } hitcount: 3919 { child_comm: compiz } hitcount: 3920 { child_comm: evolution-sourc } hitcount: 4921 { child_comm: dhclient } hitcount: 4922 { child_comm: pool } hitcount: 5923 { child_comm: nm-dispatcher.a } hitcount: 8924 { child_comm: firefox } hitcount: 8925 { child_comm: dbus-daemon } hitcount: 8926 { child_comm: glib-pacrunner } hitcount: 10927 { child_comm: evolution } hitcount: 23928 929 Totals:930 Hits: 89931 Entries: 20932 Dropped: 0933 934 If we want to pause the hist trigger, we can simply append :pause to935 the command that started the trigger. Notice that the trigger info936 displays as [paused]::937 938 # echo 'hist:key=child_comm:val=hitcount:size=256:pause' >> \939 /sys/kernel/tracing/events/sched/sched_process_fork/trigger940 941 # cat /sys/kernel/tracing/events/sched/sched_process_fork/hist942 # trigger info: hist:keys=child_comm:vals=hitcount:sort=hitcount:size=256 [paused]943 944 { child_comm: dconf worker } hitcount: 1945 { child_comm: kthreadd } hitcount: 1946 { child_comm: dconf worker } hitcount: 1947 { child_comm: gdbus } hitcount: 1948 { child_comm: ibus-daemon } hitcount: 1949 { child_comm: Socket Thread } hitcount: 2950 { child_comm: evolution-alarm } hitcount: 2951 { child_comm: smbd } hitcount: 2952 { child_comm: bash } hitcount: 3953 { child_comm: whoopsie } hitcount: 3954 { child_comm: compiz } hitcount: 3955 { child_comm: evolution-sourc } hitcount: 4956 { child_comm: pool } hitcount: 5957 { child_comm: postgres } hitcount: 6958 { child_comm: firefox } hitcount: 8959 { child_comm: dhclient } hitcount: 10960 { child_comm: emacs } hitcount: 12961 { child_comm: dbus-daemon } hitcount: 20962 { child_comm: nm-dispatcher.a } hitcount: 20963 { child_comm: evolution } hitcount: 35964 { child_comm: glib-pacrunner } hitcount: 59965 966 Totals:967 Hits: 199968 Entries: 21969 Dropped: 0970 971 To manually continue having the trigger aggregate events, append972 :cont instead. Notice that the trigger info displays as [active]973 again, and the data has changed::974 975 # echo 'hist:key=child_comm:val=hitcount:size=256:cont' >> \976 /sys/kernel/tracing/events/sched/sched_process_fork/trigger977 978 # cat /sys/kernel/tracing/events/sched/sched_process_fork/hist979 # trigger info: hist:keys=child_comm:vals=hitcount:sort=hitcount:size=256 [active]980 981 { child_comm: dconf worker } hitcount: 1982 { child_comm: dconf worker } hitcount: 1983 { child_comm: kthreadd } hitcount: 1984 { child_comm: gdbus } hitcount: 1985 { child_comm: ibus-daemon } hitcount: 1986 { child_comm: Socket Thread } hitcount: 2987 { child_comm: evolution-alarm } hitcount: 2988 { child_comm: smbd } hitcount: 2989 { child_comm: whoopsie } hitcount: 3990 { child_comm: compiz } hitcount: 3991 { child_comm: evolution-sourc } hitcount: 4992 { child_comm: bash } hitcount: 5993 { child_comm: pool } hitcount: 5994 { child_comm: postgres } hitcount: 6995 { child_comm: firefox } hitcount: 8996 { child_comm: dhclient } hitcount: 11997 { child_comm: emacs } hitcount: 12998 { child_comm: dbus-daemon } hitcount: 22999 { child_comm: nm-dispatcher.a } hitcount: 221000 { child_comm: evolution } hitcount: 351001 { child_comm: glib-pacrunner } hitcount: 591002 1003 Totals:1004 Hits: 2061005 Entries: 211006 Dropped: 01007 1008 The previous example showed how to start and stop a hist trigger by1009 appending 'pause' and 'continue' to the hist trigger command. A1010 hist trigger can also be started in a paused state by initially1011 starting the trigger with ':pause' appended. This allows you to1012 start the trigger only when you're ready to start collecting data1013 and not before. For example, you could start the trigger in a1014 paused state, then unpause it and do something you want to measure,1015 then pause the trigger again when done.1016 1017 Of course, doing this manually can be difficult and error-prone, but1018 it is possible to automatically start and stop a hist trigger based1019 on some condition, via the enable_hist and disable_hist triggers.1020 1021 For example, suppose we wanted to take a look at the relative1022 weights in terms of skb length for each callpath that leads to a1023 netif_receive_skb event when downloading a decent-sized file using1024 wget.1025 1026 First we set up an initially paused stacktrace trigger on the1027 netif_receive_skb event::1028 1029 # echo 'hist:key=common_stacktrace:vals=len:pause' > \1030 /sys/kernel/tracing/events/net/netif_receive_skb/trigger1031 1032 Next, we set up an 'enable_hist' trigger on the sched_process_exec1033 event, with an 'if filename==/usr/bin/wget' filter. The effect of1034 this new trigger is that it will 'unpause' the hist trigger we just1035 set up on netif_receive_skb if and only if it sees a1036 sched_process_exec event with a filename of '/usr/bin/wget'. When1037 that happens, all netif_receive_skb events are aggregated into a1038 hash table keyed on stacktrace::1039 1040 # echo 'enable_hist:net:netif_receive_skb if filename==/usr/bin/wget' > \1041 /sys/kernel/tracing/events/sched/sched_process_exec/trigger1042 1043 The aggregation continues until the netif_receive_skb is paused1044 again, which is what the following disable_hist event does by1045 creating a similar setup on the sched_process_exit event, using the1046 filter 'comm==wget'::1047 1048 # echo 'disable_hist:net:netif_receive_skb if comm==wget' > \1049 /sys/kernel/tracing/events/sched/sched_process_exit/trigger1050 1051 Whenever a process exits and the comm field of the disable_hist1052 trigger filter matches 'comm==wget', the netif_receive_skb hist1053 trigger is disabled.1054 1055 The overall effect is that netif_receive_skb events are aggregated1056 into the hash table for only the duration of the wget. Executing a1057 wget command and then listing the 'hist' file will display the1058 output generated by the wget command::1059 1060 $ wget https://www.kernel.org/pub/linux/kernel/v3.x/patch-3.19.xz1061 1062 # cat /sys/kernel/tracing/events/net/netif_receive_skb/hist1063 # trigger info: hist:keys=common_stacktrace:vals=len:sort=hitcount:size=2048 [paused]1064 1065 { common_stacktrace:1066 __netif_receive_skb_core+0x46d/0x9901067 __netif_receive_skb+0x18/0x601068 netif_receive_skb_internal+0x23/0x901069 napi_gro_receive+0xc8/0x1001070 ieee80211_deliver_skb+0xd6/0x270 [mac80211]1071 ieee80211_rx_handlers+0xccf/0x22f0 [mac80211]1072 ieee80211_prepare_and_rx_handle+0x4e7/0xc40 [mac80211]1073 ieee80211_rx+0x31d/0x900 [mac80211]1074 iwlagn_rx_reply_rx+0x3db/0x6f0 [iwldvm]1075 iwl_rx_dispatch+0x8e/0xf0 [iwldvm]1076 iwl_pcie_irq_handler+0xe3c/0x12f0 [iwlwifi]1077 irq_thread_fn+0x20/0x501078 irq_thread+0x11f/0x1501079 kthread+0xd2/0xf01080 ret_from_fork+0x42/0x701081 } hitcount: 85 len: 288841082 { common_stacktrace:1083 __netif_receive_skb_core+0x46d/0x9901084 __netif_receive_skb+0x18/0x601085 netif_receive_skb_internal+0x23/0x901086 napi_gro_complete+0xa4/0xe01087 dev_gro_receive+0x23a/0x3601088 napi_gro_receive+0x30/0x1001089 ieee80211_deliver_skb+0xd6/0x270 [mac80211]1090 ieee80211_rx_handlers+0xccf/0x22f0 [mac80211]1091 ieee80211_prepare_and_rx_handle+0x4e7/0xc40 [mac80211]1092 ieee80211_rx+0x31d/0x900 [mac80211]1093 iwlagn_rx_reply_rx+0x3db/0x6f0 [iwldvm]1094 iwl_rx_dispatch+0x8e/0xf0 [iwldvm]1095 iwl_pcie_irq_handler+0xe3c/0x12f0 [iwlwifi]1096 irq_thread_fn+0x20/0x501097 irq_thread+0x11f/0x1501098 kthread+0xd2/0xf01099 } hitcount: 98 len: 6643291100 { common_stacktrace:1101 __netif_receive_skb_core+0x46d/0x9901102 __netif_receive_skb+0x18/0x601103 process_backlog+0xa8/0x1501104 net_rx_action+0x15d/0x3401105 __do_softirq+0x114/0x2c01106 do_softirq_own_stack+0x1c/0x301107 do_softirq+0x65/0x701108 __local_bh_enable_ip+0xb5/0xc01109 ip_finish_output+0x1f4/0x8401110 ip_output+0x6b/0xc01111 ip_local_out_sk+0x31/0x401112 ip_send_skb+0x1a/0x501113 udp_send_skb+0x173/0x2a01114 udp_sendmsg+0x2bf/0x9f01115 inet_sendmsg+0x64/0xa01116 sock_sendmsg+0x3d/0x501117 } hitcount: 115 len: 130301118 { common_stacktrace:1119 __netif_receive_skb_core+0x46d/0x9901120 __netif_receive_skb+0x18/0x601121 netif_receive_skb_internal+0x23/0x901122 napi_gro_complete+0xa4/0xe01123 napi_gro_flush+0x6d/0x901124 iwl_pcie_irq_handler+0x92a/0x12f0 [iwlwifi]1125 irq_thread_fn+0x20/0x501126 irq_thread+0x11f/0x1501127 kthread+0xd2/0xf01128 ret_from_fork+0x42/0x701129 } hitcount: 934 len: 55122121130 1131 Totals:1132 Hits: 12321133 Entries: 41134 Dropped: 01135 1136 The above shows all the netif_receive_skb callpaths and their total1137 lengths for the duration of the wget command.1138 1139 The 'clear' hist trigger param can be used to clear the hash table.1140 Suppose we wanted to try another run of the previous example but1141 this time also wanted to see the complete list of events that went1142 into the histogram. In order to avoid having to set everything up1143 again, we can just clear the histogram first::1144 1145 # echo 'hist:key=common_stacktrace:vals=len:clear' >> \1146 /sys/kernel/tracing/events/net/netif_receive_skb/trigger1147 1148 Just to verify that it is in fact cleared, here's what we now see in1149 the hist file::1150 1151 # cat /sys/kernel/tracing/events/net/netif_receive_skb/hist1152 # trigger info: hist:keys=common_stacktrace:vals=len:sort=hitcount:size=2048 [paused]1153 1154 Totals:1155 Hits: 01156 Entries: 01157 Dropped: 01158 1159 Since we want to see the detailed list of every netif_receive_skb1160 event occurring during the new run, which are in fact the same1161 events being aggregated into the hash table, we add some additional1162 'enable_event' events to the triggering sched_process_exec and1163 sched_process_exit events as such::1164 1165 # echo 'enable_event:net:netif_receive_skb if filename==/usr/bin/wget' > \1166 /sys/kernel/tracing/events/sched/sched_process_exec/trigger1167 1168 # echo 'disable_event:net:netif_receive_skb if comm==wget' > \1169 /sys/kernel/tracing/events/sched/sched_process_exit/trigger1170 1171 If you read the trigger files for the sched_process_exec and1172 sched_process_exit triggers, you should see two triggers for each:1173 one enabling/disabling the hist aggregation and the other1174 enabling/disabling the logging of events::1175 1176 # cat /sys/kernel/tracing/events/sched/sched_process_exec/trigger1177 enable_event:net:netif_receive_skb:unlimited if filename==/usr/bin/wget1178 enable_hist:net:netif_receive_skb:unlimited if filename==/usr/bin/wget1179 1180 # cat /sys/kernel/tracing/events/sched/sched_process_exit/trigger1181 enable_event:net:netif_receive_skb:unlimited if comm==wget1182 disable_hist:net:netif_receive_skb:unlimited if comm==wget1183 1184 In other words, whenever either of the sched_process_exec or1185 sched_process_exit events is hit and matches 'wget', it enables or1186 disables both the histogram and the event log, and what you end up1187 with is a hash table and set of events just covering the specified1188 duration. Run the wget command again::1189 1190 $ wget https://www.kernel.org/pub/linux/kernel/v3.x/patch-3.19.xz1191 1192 Displaying the 'hist' file should show something similar to what you1193 saw in the last run, but this time you should also see the1194 individual events in the trace file::1195 1196 # cat /sys/kernel/tracing/trace1197 1198 # tracer: nop1199 #1200 # entries-in-buffer/entries-written: 183/1426 #P:41201 #1202 # _-----=> irqs-off1203 # / _----=> need-resched1204 # | / _---=> hardirq/softirq1205 # || / _--=> preempt-depth1206 # ||| / delay1207 # TASK-PID CPU# |||| TIMESTAMP FUNCTION1208 # | | | |||| | |1209 wget-15108 [000] ..s1 31769.606929: netif_receive_skb: dev=lo skbaddr=ffff88009c353100 len=601210 wget-15108 [000] ..s1 31769.606999: netif_receive_skb: dev=lo skbaddr=ffff88009c353200 len=601211 dnsmasq-1382 [000] ..s1 31769.677652: netif_receive_skb: dev=lo skbaddr=ffff88009c352b00 len=1301212 dnsmasq-1382 [000] ..s1 31769.685917: netif_receive_skb: dev=lo skbaddr=ffff88009c352200 len=1381213 ##### CPU 2 buffer started ####1214 irq/29-iwlwifi-559 [002] ..s. 31772.031529: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d433d00 len=29481215 irq/29-iwlwifi-559 [002] ..s. 31772.031572: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d432200 len=15001216 irq/29-iwlwifi-559 [002] ..s. 31772.032196: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d433100 len=29481217 irq/29-iwlwifi-559 [002] ..s. 31772.032761: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d433000 len=29481218 irq/29-iwlwifi-559 [002] ..s. 31772.033220: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d432e00 len=15001219 .1220 .1221 .1222 1223 The following example demonstrates how multiple hist triggers can be1224 attached to a given event. This capability can be useful for1225 creating a set of different summaries derived from the same set of1226 events, or for comparing the effects of different filters, among1227 other things::1228 1229 # echo 'hist:keys=skbaddr.hex:vals=len if len < 0' >> \1230 /sys/kernel/tracing/events/net/netif_receive_skb/trigger1231 # echo 'hist:keys=skbaddr.hex:vals=len if len > 4096' >> \1232 /sys/kernel/tracing/events/net/netif_receive_skb/trigger1233 # echo 'hist:keys=skbaddr.hex:vals=len if len == 256' >> \1234 /sys/kernel/tracing/events/net/netif_receive_skb/trigger1235 # echo 'hist:keys=skbaddr.hex:vals=len' >> \1236 /sys/kernel/tracing/events/net/netif_receive_skb/trigger1237 # echo 'hist:keys=len:vals=common_preempt_count' >> \1238 /sys/kernel/tracing/events/net/netif_receive_skb/trigger1239 1240 The above set of commands create four triggers differing only in1241 their filters, along with a completely different though fairly1242 nonsensical trigger. Note that in order to append multiple hist1243 triggers to the same file, you should use the '>>' operator to1244 append them ('>' will also add the new hist trigger, but will remove1245 any existing hist triggers beforehand).1246 1247 Displaying the contents of the 'hist' file for the event shows the1248 contents of all five histograms::1249 1250 # cat /sys/kernel/tracing/events/net/netif_receive_skb/hist1251 1252 # event histogram1253 #1254 # trigger info: hist:keys=len:vals=hitcount,common_preempt_count:sort=hitcount:size=2048 [active]1255 #1256 1257 { len: 176 } hitcount: 1 common_preempt_count: 01258 { len: 223 } hitcount: 1 common_preempt_count: 01259 { len: 4854 } hitcount: 1 common_preempt_count: 01260 { len: 395 } hitcount: 1 common_preempt_count: 01261 { len: 177 } hitcount: 1 common_preempt_count: 01262 { len: 446 } hitcount: 1 common_preempt_count: 01263 { len: 1601 } hitcount: 1 common_preempt_count: 01264 .1265 .1266 .1267 { len: 1280 } hitcount: 66 common_preempt_count: 01268 { len: 116 } hitcount: 81 common_preempt_count: 401269 { len: 708 } hitcount: 112 common_preempt_count: 01270 { len: 46 } hitcount: 221 common_preempt_count: 01271 { len: 1264 } hitcount: 458 common_preempt_count: 01272 1273 Totals:1274 Hits: 14281275 Entries: 1471276 Dropped: 01277 1278 1279 # event histogram1280 #1281 # trigger info: hist:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 [active]1282 #1283 1284 { skbaddr: ffff8800baee5e00 } hitcount: 1 len: 1301285 { skbaddr: ffff88005f3d5600 } hitcount: 1 len: 12801286 { skbaddr: ffff88005f3d4900 } hitcount: 1 len: 12801287 { skbaddr: ffff88009fed6300 } hitcount: 1 len: 1151288 { skbaddr: ffff88009fe0ad00 } hitcount: 1 len: 1151289 { skbaddr: ffff88008cdb1900 } hitcount: 1 len: 461290 { skbaddr: ffff880064b5ef00 } hitcount: 1 len: 1181291 { skbaddr: ffff880044e3c700 } hitcount: 1 len: 601292 { skbaddr: ffff880100065900 } hitcount: 1 len: 461293 { skbaddr: ffff8800d46bd500 } hitcount: 1 len: 1161294 { skbaddr: ffff88005f3d5f00 } hitcount: 1 len: 12801295 { skbaddr: ffff880100064700 } hitcount: 1 len: 3651296 { skbaddr: ffff8800badb6f00 } hitcount: 1 len: 601297 .1298 .1299 .1300 { skbaddr: ffff88009fe0be00 } hitcount: 27 len: 246771301 { skbaddr: ffff88009fe0a400 } hitcount: 27 len: 230521302 { skbaddr: ffff88009fe0b700 } hitcount: 31 len: 255891303 { skbaddr: ffff88009fe0b600 } hitcount: 32 len: 273261304 { skbaddr: ffff88006a462800 } hitcount: 68 len: 716781305 { skbaddr: ffff88006a463700 } hitcount: 70 len: 726781306 { skbaddr: ffff88006a462b00 } hitcount: 71 len: 775891307 { skbaddr: ffff88006a463600 } hitcount: 73 len: 713071308 { skbaddr: ffff88006a462200 } hitcount: 81 len: 810321309 1310 Totals:1311 Hits: 14511312 Entries: 3181313 Dropped: 01314 1315 1316 # event histogram1317 #1318 # trigger info: hist:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 if len == 256 [active]1319 #1320 1321 1322 Totals:1323 Hits: 01324 Entries: 01325 Dropped: 01326 1327 1328 # event histogram1329 #1330 # trigger info: hist:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 if len > 4096 [active]1331 #1332 1333 { skbaddr: ffff88009fd2c300 } hitcount: 1 len: 72121334 { skbaddr: ffff8800d2bcce00 } hitcount: 1 len: 72121335 { skbaddr: ffff8800d2bcd700 } hitcount: 1 len: 72121336 { skbaddr: ffff8800d2bcda00 } hitcount: 1 len: 214921337 { skbaddr: ffff8800ae2e2d00 } hitcount: 1 len: 72121338 { skbaddr: ffff8800d2bcdb00 } hitcount: 1 len: 72121339 { skbaddr: ffff88006a4df500 } hitcount: 1 len: 48541340 { skbaddr: ffff88008ce47b00 } hitcount: 1 len: 186361341 { skbaddr: ffff8800ae2e2200 } hitcount: 1 len: 129241342 { skbaddr: ffff88005f3e1000 } hitcount: 1 len: 43561343 { skbaddr: ffff8800d2bcdc00 } hitcount: 2 len: 244201344 { skbaddr: ffff8800d2bcc200 } hitcount: 2 len: 129961345 1346 Totals:1347 Hits: 141348 Entries: 121349 Dropped: 01350 1351 1352 # event histogram1353 #1354 # trigger info: hist:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 if len < 0 [active]1355 #1356 1357 1358 Totals:1359 Hits: 01360 Entries: 01361 Dropped: 01362 1363 Named triggers can be used to have triggers share a common set of1364 histogram data. This capability is mostly useful for combining the1365 output of events generated by tracepoints contained inside inline1366 functions, but names can be used in a hist trigger on any event.1367 For example, these two triggers when hit will update the same 'len'1368 field in the shared 'foo' histogram data::1369 1370 # echo 'hist:name=foo:keys=skbaddr.hex:vals=len' > \1371 /sys/kernel/tracing/events/net/netif_receive_skb/trigger1372 # echo 'hist:name=foo:keys=skbaddr.hex:vals=len' > \1373 /sys/kernel/tracing/events/net/netif_rx/trigger1374 1375 You can see that they're updating common histogram data by reading1376 each event's hist files at the same time::1377 1378 # cat /sys/kernel/tracing/events/net/netif_receive_skb/hist;1379 cat /sys/kernel/tracing/events/net/netif_rx/hist1380 1381 # event histogram1382 #1383 # trigger info: hist:name=foo:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 [active]1384 #1385 1386 { skbaddr: ffff88000ad53500 } hitcount: 1 len: 461387 { skbaddr: ffff8800af5a1500 } hitcount: 1 len: 761388 { skbaddr: ffff8800d62a1900 } hitcount: 1 len: 461389 { skbaddr: ffff8800d2bccb00 } hitcount: 1 len: 4681390 { skbaddr: ffff8800d3c69900 } hitcount: 1 len: 461391 { skbaddr: ffff88009ff09100 } hitcount: 1 len: 521392 { skbaddr: ffff88010f13ab00 } hitcount: 1 len: 1681393 { skbaddr: ffff88006a54f400 } hitcount: 1 len: 461394 { skbaddr: ffff8800d2bcc500 } hitcount: 1 len: 2601395 { skbaddr: ffff880064505000 } hitcount: 1 len: 461396 { skbaddr: ffff8800baf24e00 } hitcount: 1 len: 321397 { skbaddr: ffff88009fe0ad00 } hitcount: 1 len: 461398 { skbaddr: ffff8800d3edff00 } hitcount: 1 len: 441399 { skbaddr: ffff88009fe0b400 } hitcount: 1 len: 1681400 { skbaddr: ffff8800a1c55a00 } hitcount: 1 len: 401401 { skbaddr: ffff8800d2bcd100 } hitcount: 1 len: 401402 { skbaddr: ffff880064505f00 } hitcount: 1 len: 1741403 { skbaddr: ffff8800a8bff200 } hitcount: 1 len: 1601404 { skbaddr: ffff880044e3cc00 } hitcount: 1 len: 761405 { skbaddr: ffff8800a8bfe700 } hitcount: 1 len: 461406 { skbaddr: ffff8800d2bcdc00 } hitcount: 1 len: 321407 { skbaddr: ffff8800a1f64800 } hitcount: 1 len: 461408 { skbaddr: ffff8800d2bcde00 } hitcount: 1 len: 9881409 { skbaddr: ffff88006a5dea00 } hitcount: 1 len: 461410 { skbaddr: ffff88002e37a200 } hitcount: 1 len: 441411 { skbaddr: ffff8800a1f32c00 } hitcount: 2 len: 6761412 { skbaddr: ffff88000ad52600 } hitcount: 2 len: 1071413 { skbaddr: ffff8800a1f91e00 } hitcount: 2 len: 921414 { skbaddr: ffff8800af5a0200 } hitcount: 2 len: 1421415 { skbaddr: ffff8800d2bcc600 } hitcount: 2 len: 2201416 { skbaddr: ffff8800ba36f500 } hitcount: 2 len: 921417 { skbaddr: ffff8800d021f800 } hitcount: 2 len: 921418 { skbaddr: ffff8800a1f33600 } hitcount: 2 len: 6751419 { skbaddr: ffff8800a8bfff00 } hitcount: 3 len: 1381420 { skbaddr: ffff8800d62a1300 } hitcount: 3 len: 1381421 { skbaddr: ffff88002e37a100 } hitcount: 4 len: 1841422 { skbaddr: ffff880064504400 } hitcount: 4 len: 1841423 { skbaddr: ffff8800a8bfec00 } hitcount: 4 len: 1841424 { skbaddr: ffff88000ad53700 } hitcount: 5 len: 2301425 { skbaddr: ffff8800d2bcdb00 } hitcount: 5 len: 1961426 { skbaddr: ffff8800a1f90000 } hitcount: 6 len: 2761427 { skbaddr: ffff88006a54f900 } hitcount: 6 len: 2761428 1429 Totals:1430 Hits: 811431 Entries: 421432 Dropped: 01433 # event histogram1434 #1435 # trigger info: hist:name=foo:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 [active]1436 #1437 1438 { skbaddr: ffff88000ad53500 } hitcount: 1 len: 461439 { skbaddr: ffff8800af5a1500 } hitcount: 1 len: 761440 { skbaddr: ffff8800d62a1900 } hitcount: 1 len: 461441 { skbaddr: ffff8800d2bccb00 } hitcount: 1 len: 4681442 { skbaddr: ffff8800d3c69900 } hitcount: 1 len: 461443 { skbaddr: ffff88009ff09100 } hitcount: 1 len: 521444 { skbaddr: ffff88010f13ab00 } hitcount: 1 len: 1681445 { skbaddr: ffff88006a54f400 } hitcount: 1 len: 461446 { skbaddr: ffff8800d2bcc500 } hitcount: 1 len: 2601447 { skbaddr: ffff880064505000 } hitcount: 1 len: 461448 { skbaddr: ffff8800baf24e00 } hitcount: 1 len: 321449 { skbaddr: ffff88009fe0ad00 } hitcount: 1 len: 461450 { skbaddr: ffff8800d3edff00 } hitcount: 1 len: 441451 { skbaddr: ffff88009fe0b400 } hitcount: 1 len: 1681452 { skbaddr: ffff8800a1c55a00 } hitcount: 1 len: 401453 { skbaddr: ffff8800d2bcd100 } hitcount: 1 len: 401454 { skbaddr: ffff880064505f00 } hitcount: 1 len: 1741455 { skbaddr: ffff8800a8bff200 } hitcount: 1 len: 1601456 { skbaddr: ffff880044e3cc00 } hitcount: 1 len: 761457 { skbaddr: ffff8800a8bfe700 } hitcount: 1 len: 461458 { skbaddr: ffff8800d2bcdc00 } hitcount: 1 len: 321459 { skbaddr: ffff8800a1f64800 } hitcount: 1 len: 461460 { skbaddr: ffff8800d2bcde00 } hitcount: 1 len: 9881461 { skbaddr: ffff88006a5dea00 } hitcount: 1 len: 461462 { skbaddr: ffff88002e37a200 } hitcount: 1 len: 441463 { skbaddr: ffff8800a1f32c00 } hitcount: 2 len: 6761464 { skbaddr: ffff88000ad52600 } hitcount: 2 len: 1071465 { skbaddr: ffff8800a1f91e00 } hitcount: 2 len: 921466 { skbaddr: ffff8800af5a0200 } hitcount: 2 len: 1421467 { skbaddr: ffff8800d2bcc600 } hitcount: 2 len: 2201468 { skbaddr: ffff8800ba36f500 } hitcount: 2 len: 921469 { skbaddr: ffff8800d021f800 } hitcount: 2 len: 921470 { skbaddr: ffff8800a1f33600 } hitcount: 2 len: 6751471 { skbaddr: ffff8800a8bfff00 } hitcount: 3 len: 1381472 { skbaddr: ffff8800d62a1300 } hitcount: 3 len: 1381473 { skbaddr: ffff88002e37a100 } hitcount: 4 len: 1841474 { skbaddr: ffff880064504400 } hitcount: 4 len: 1841475 { skbaddr: ffff8800a8bfec00 } hitcount: 4 len: 1841476 { skbaddr: ffff88000ad53700 } hitcount: 5 len: 2301477 { skbaddr: ffff8800d2bcdb00 } hitcount: 5 len: 1961478 { skbaddr: ffff8800a1f90000 } hitcount: 6 len: 2761479 { skbaddr: ffff88006a54f900 } hitcount: 6 len: 2761480 1481 Totals:1482 Hits: 811483 Entries: 421484 Dropped: 01485 1486 And here's an example that shows how to combine histogram data from1487 any two events even if they don't share any 'compatible' fields1488 other than 'hitcount' and 'common_stacktrace'. These commands create a1489 couple of triggers named 'bar' using those fields::1490 1491 # echo 'hist:name=bar:key=common_stacktrace:val=hitcount' > \1492 /sys/kernel/tracing/events/sched/sched_process_fork/trigger1493 # echo 'hist:name=bar:key=common_stacktrace:val=hitcount' > \1494 /sys/kernel/tracing/events/net/netif_rx/trigger1495 1496 And displaying the output of either shows some interesting if1497 somewhat confusing output::1498 1499 # cat /sys/kernel/tracing/events/sched/sched_process_fork/hist1500 # cat /sys/kernel/tracing/events/net/netif_rx/hist1501 1502 # event histogram1503 #1504 # trigger info: hist:name=bar:keys=common_stacktrace:vals=hitcount:sort=hitcount:size=2048 [active]1505 #1506 1507 { common_stacktrace:1508 kernel_clone+0x18e/0x3301509 kernel_thread+0x29/0x301510 kthreadd+0x154/0x1b01511 ret_from_fork+0x3f/0x701512 } hitcount: 11513 { common_stacktrace:1514 netif_rx_internal+0xb2/0xd01515 netif_rx_ni+0x20/0x701516 dev_loopback_xmit+0xaa/0xd01517 ip_mc_output+0x126/0x2401518 ip_local_out_sk+0x31/0x401519 igmp_send_report+0x1e9/0x2301520 igmp_timer_expire+0xe9/0x1201521 call_timer_fn+0x39/0xf01522 run_timer_softirq+0x1e1/0x2901523 __do_softirq+0xfd/0x2901524 irq_exit+0x98/0xb01525 smp_apic_timer_interrupt+0x4a/0x601526 apic_timer_interrupt+0x6d/0x801527 cpuidle_enter+0x17/0x201528 call_cpuidle+0x3b/0x601529 cpu_startup_entry+0x22d/0x3101530 } hitcount: 11531 { common_stacktrace:1532 netif_rx_internal+0xb2/0xd01533 netif_rx_ni+0x20/0x701534 dev_loopback_xmit+0xaa/0xd01535 ip_mc_output+0x17f/0x2401536 ip_local_out_sk+0x31/0x401537 ip_send_skb+0x1a/0x501538 udp_send_skb+0x13e/0x2701539 udp_sendmsg+0x2bf/0x9801540 inet_sendmsg+0x67/0xa01541 sock_sendmsg+0x38/0x501542 SYSC_sendto+0xef/0x1701543 SyS_sendto+0xe/0x101544 entry_SYSCALL_64_fastpath+0x12/0x6a1545 } hitcount: 21546 { common_stacktrace:1547 netif_rx_internal+0xb2/0xd01548 netif_rx+0x1c/0x601549 loopback_xmit+0x6c/0xb01550 dev_hard_start_xmit+0x219/0x3a01551 __dev_queue_xmit+0x415/0x4f01552 dev_queue_xmit_sk+0x13/0x201553 ip_finish_output2+0x237/0x3401554 ip_finish_output+0x113/0x1d01555 ip_output+0x66/0xc01556 ip_local_out_sk+0x31/0x401557 ip_send_skb+0x1a/0x501558 udp_send_skb+0x16d/0x2701559 udp_sendmsg+0x2bf/0x9801560 inet_sendmsg+0x67/0xa01561 sock_sendmsg+0x38/0x501562 ___sys_sendmsg+0x14e/0x2701563 } hitcount: 761564 { common_stacktrace:1565 netif_rx_internal+0xb2/0xd01566 netif_rx+0x1c/0x601567 loopback_xmit+0x6c/0xb01568 dev_hard_start_xmit+0x219/0x3a01569 __dev_queue_xmit+0x415/0x4f01570 dev_queue_xmit_sk+0x13/0x201571 ip_finish_output2+0x237/0x3401572 ip_finish_output+0x113/0x1d01573 ip_output+0x66/0xc01574 ip_local_out_sk+0x31/0x401575 ip_send_skb+0x1a/0x501576 udp_send_skb+0x16d/0x2701577 udp_sendmsg+0x2bf/0x9801578 inet_sendmsg+0x67/0xa01579 sock_sendmsg+0x38/0x501580 ___sys_sendmsg+0x269/0x2701581 } hitcount: 771582 { common_stacktrace:1583 netif_rx_internal+0xb2/0xd01584 netif_rx+0x1c/0x601585 loopback_xmit+0x6c/0xb01586 dev_hard_start_xmit+0x219/0x3a01587 __dev_queue_xmit+0x415/0x4f01588 dev_queue_xmit_sk+0x13/0x201589 ip_finish_output2+0x237/0x3401590 ip_finish_output+0x113/0x1d01591 ip_output+0x66/0xc01592 ip_local_out_sk+0x31/0x401593 ip_send_skb+0x1a/0x501594 udp_send_skb+0x16d/0x2701595 udp_sendmsg+0x2bf/0x9801596 inet_sendmsg+0x67/0xa01597 sock_sendmsg+0x38/0x501598 SYSC_sendto+0xef/0x1701599 } hitcount: 881600 { common_stacktrace:1601 kernel_clone+0x18e/0x3301602 SyS_clone+0x19/0x201603 entry_SYSCALL_64_fastpath+0x12/0x6a1604 } hitcount: 2441605 1606 Totals:1607 Hits: 4891608 Entries: 71609 Dropped: 01610 16112.2 Inter-event hist triggers1612-----------------------------1613 1614Inter-event hist triggers are hist triggers that combine values from1615one or more other events and create a histogram using that data. Data1616from an inter-event histogram can in turn become the source for1617further combined histograms, thus providing a chain of related1618histograms, which is important for some applications.1619 1620The most important example of an inter-event quantity that can be used1621in this manner is latency, which is simply a difference in timestamps1622between two events. Although latency is the most important1623inter-event quantity, note that because the support is completely1624general across the trace event subsystem, any event field can be used1625in an inter-event quantity.1626 1627An example of a histogram that combines data from other histograms1628into a useful chain would be a 'wakeupswitch latency' histogram that1629combines a 'wakeup latency' histogram and a 'switch latency'1630histogram.1631 1632Normally, a hist trigger specification consists of a (possibly1633compound) key along with one or more numeric values, which are1634continually updated sums associated with that key. A histogram1635specification in this case consists of individual key and value1636specifications that refer to trace event fields associated with a1637single event type.1638 1639The inter-event hist trigger extension allows fields from multiple1640events to be referenced and combined into a multi-event histogram1641specification. In support of this overall goal, a few enabling1642features have been added to the hist trigger support:1643 1644 - In order to compute an inter-event quantity, a value from one1645 event needs to saved and then referenced from another event. This1646 requires the introduction of support for histogram 'variables'.1647 1648 - The computation of inter-event quantities and their combination1649 require some minimal amount of support for applying simple1650 expressions to variables (+ and -).1651 1652 - A histogram consisting of inter-event quantities isn't logically a1653 histogram on either event (so having the 'hist' file for either1654 event host the histogram output doesn't really make sense). To1655 address the idea that the histogram is associated with a1656 combination of events, support is added allowing the creation of1657 'synthetic' events that are events derived from other events.1658 These synthetic events are full-fledged events just like any other1659 and can be used as such, as for instance to create the1660 'combination' histograms mentioned previously.1661 1662 - A set of 'actions' can be associated with histogram entries -1663 these can be used to generate the previously mentioned synthetic1664 events, but can also be used for other purposes, such as for1665 example saving context when a 'max' latency has been hit.1666 1667 - Trace events don't have a 'timestamp' associated with them, but1668 there is an implicit timestamp saved along with an event in the1669 underlying ftrace ring buffer. This timestamp is now exposed as a1670 a synthetic field named 'common_timestamp' which can be used in1671 histograms as if it were any other event field; it isn't an actual1672 field in the trace format but rather is a synthesized value that1673 nonetheless can be used as if it were an actual field. By default1674 it is in units of nanoseconds; appending '.usecs' to a1675 common_timestamp field changes the units to microseconds.1676 1677A note on inter-event timestamps: If common_timestamp is used in a1678histogram, the trace buffer is automatically switched over to using1679absolute timestamps and the "global" trace clock, in order to avoid1680bogus timestamp differences with other clocks that aren't coherent1681across CPUs. This can be overridden by specifying one of the other1682trace clocks instead, using the "clock=XXX" hist trigger attribute,1683where XXX is any of the clocks listed in the tracing/trace_clock1684pseudo-file.1685 1686These features are described in more detail in the following sections.1687 16882.2.1 Histogram Variables1689-------------------------1690 1691Variables are simply named locations used for saving and retrieving1692values between matching events. A 'matching' event is defined as an1693event that has a matching key - if a variable is saved for a histogram1694entry corresponding to that key, any subsequent event with a matching1695key can access that variable.1696 1697A variable's value is normally available to any subsequent event until1698it is set to something else by a subsequent event. The one exception1699to that rule is that any variable used in an expression is essentially1700'read-once' - once it's used by an expression in a subsequent event,1701it's reset to its 'unset' state, which means it can't be used again1702unless it's set again. This ensures not only that an event doesn't1703use an uninitialized variable in a calculation, but that that variable1704is used only once and not for any unrelated subsequent match.1705 1706The basic syntax for saving a variable is to simply prefix a unique1707variable name not corresponding to any keyword along with an '=' sign1708to any event field.1709 1710Either keys or values can be saved and retrieved in this way. This1711creates a variable named 'ts0' for a histogram entry with the key1712'next_pid'::1713 1714 # echo 'hist:keys=next_pid:vals=$ts0:ts0=common_timestamp ... >> \1715 event/trigger1716 1717The ts0 variable can be accessed by any subsequent event having the1718same pid as 'next_pid'.1719 1720Variable references are formed by prepending the variable name with1721the '$' sign. Thus for example, the ts0 variable above would be1722referenced as '$ts0' in expressions.1723 1724Because 'vals=' is used, the common_timestamp variable value above1725will also be summed as a normal histogram value would (though for a1726timestamp it makes little sense).1727 1728The below shows that a key value can also be saved in the same way::1729 1730 # echo 'hist:timer_pid=common_pid:key=timer_pid ...' >> event/trigger1731 1732If a variable isn't a key variable or prefixed with 'vals=', the1733associated event field will be saved in a variable but won't be summed1734as a value::1735 1736 # echo 'hist:keys=next_pid:ts1=common_timestamp ...' >> event/trigger1737 1738Multiple variables can be assigned at the same time. The below would1739result in both ts0 and b being created as variables, with both1740common_timestamp and field1 additionally being summed as values::1741 1742 # echo 'hist:keys=pid:vals=$ts0,$b:ts0=common_timestamp,b=field1 ...' >> \1743 event/trigger1744 1745Note that variable assignments can appear either preceding or1746following their use. The command below behaves identically to the1747command above::1748 1749 # echo 'hist:keys=pid:ts0=common_timestamp,b=field1:vals=$ts0,$b ...' >> \1750 event/trigger1751 1752Any number of variables not bound to a 'vals=' prefix can also be1753assigned by simply separating them with colons. Below is the same1754thing but without the values being summed in the histogram::1755 1756 # echo 'hist:keys=pid:ts0=common_timestamp:b=field1 ...' >> event/trigger1757 1758Variables set as above can be referenced and used in expressions on1759another event.1760 1761For example, here's how a latency can be calculated::1762 1763 # echo 'hist:keys=pid,prio:ts0=common_timestamp ...' >> event1/trigger1764 # echo 'hist:keys=next_pid:wakeup_lat=common_timestamp-$ts0 ...' >> event2/trigger1765 1766In the first line above, the event's timestamp is saved into the1767variable ts0. In the next line, ts0 is subtracted from the second1768event's timestamp to produce the latency, which is then assigned into1769yet another variable, 'wakeup_lat'. The hist trigger below in turn1770makes use of the wakeup_lat variable to compute a combined latency1771using the same key and variable from yet another event::1772 1773 # echo 'hist:key=pid:wakeupswitch_lat=$wakeup_lat+$switchtime_lat ...' >> event3/trigger1774 1775Expressions support the use of addition, subtraction, multiplication and1776division operators (+-\*/).1777 1778Note if division by zero cannot be detected at parse time (i.e. the1779divisor is not a constant), the result will be -1.1780 1781Numeric constants can also be used directly in an expression::1782 1783 # echo 'hist:keys=next_pid:timestamp_secs=common_timestamp/1000000 ...' >> event/trigger1784 1785or assigned to a variable and referenced in a subsequent expression::1786 1787 # echo 'hist:keys=next_pid:us_per_sec=1000000 ...' >> event/trigger1788 # echo 'hist:keys=next_pid:timestamp_secs=common_timestamp/$us_per_sec ...' >> event/trigger1789 1790Variables can even hold stacktraces, which are useful with synthetic events.1791 17922.2.2 Synthetic Events1793----------------------1794 1795Synthetic events are user-defined events generated from hist trigger1796variables or fields associated with one or more other events. Their1797purpose is to provide a mechanism for displaying data spanning1798multiple events consistent with the existing and already familiar1799usage for normal events.1800 1801To define a synthetic event, the user writes a simple specification1802consisting of the name of the new event along with one or more1803variables and their types, which can be any valid field type,1804separated by semicolons, to the tracing/synthetic_events file.1805 1806See synth_field_size() for available types.1807 1808If field_name contains [n], the field is considered to be a static array.1809 1810If field_names contains[] (no subscript), the field is considered to1811be a dynamic array, which will only take as much space in the event as1812is required to hold the array.1813 1814A string field can be specified using either the static notation:1815 1816 char name[32];1817 1818Or the dynamic:1819 1820 char name[];1821 1822The size limit for either is 256.1823 1824For instance, the following creates a new event named 'wakeup_latency'1825with 3 fields: lat, pid, and prio. Each of those fields is simply a1826variable reference to a variable on another event::1827 1828 # echo 'wakeup_latency \1829 u64 lat; \1830 pid_t pid; \1831 int prio' >> \1832 /sys/kernel/tracing/synthetic_events1833 1834Reading the tracing/synthetic_events file lists all the currently1835defined synthetic events, in this case the event defined above::1836 1837 # cat /sys/kernel/tracing/synthetic_events1838 wakeup_latency u64 lat; pid_t pid; int prio1839 1840An existing synthetic event definition can be removed by prepending1841the command that defined it with a '!'::1842 1843 # echo '!wakeup_latency u64 lat pid_t pid int prio' >> \1844 /sys/kernel/tracing/synthetic_events1845 1846At this point, there isn't yet an actual 'wakeup_latency' event1847instantiated in the event subsystem - for this to happen, a 'hist1848trigger action' needs to be instantiated and bound to actual fields1849and variables defined on other events (see Section 2.2.3 below on1850how that is done using hist trigger 'onmatch' action). Once that is1851done, the 'wakeup_latency' synthetic event instance is created.1852 1853The new event is created under the tracing/events/synthetic/ directory1854and looks and behaves just like any other event::1855 1856 # ls /sys/kernel/tracing/events/synthetic/wakeup_latency1857 enable filter format hist id trigger1858 1859A histogram can now be defined for the new synthetic event::1860 1861 # echo 'hist:keys=pid,prio,lat.log2:sort=lat' >> \1862 /sys/kernel/tracing/events/synthetic/wakeup_latency/trigger1863 1864The above shows the latency "lat" in a power of 2 grouping.1865 1866Like any other event, once a histogram is enabled for the event, the1867output can be displayed by reading the event's 'hist' file::1868 1869 # cat /sys/kernel/tracing/events/synthetic/wakeup_latency/hist1870 1871 # event histogram1872 #1873 # trigger info: hist:keys=pid,prio,lat.log2:vals=hitcount:sort=lat.log2:size=2048 [active]1874 #1875 1876 { pid: 2035, prio: 9, lat: ~ 2^2 } hitcount: 431877 { pid: 2034, prio: 9, lat: ~ 2^2 } hitcount: 601878 { pid: 2029, prio: 9, lat: ~ 2^2 } hitcount: 9651879 { pid: 2034, prio: 120, lat: ~ 2^2 } hitcount: 91880 { pid: 2033, prio: 120, lat: ~ 2^2 } hitcount: 51881 { pid: 2030, prio: 9, lat: ~ 2^2 } hitcount: 3351882 { pid: 2030, prio: 120, lat: ~ 2^2 } hitcount: 101883 { pid: 2032, prio: 120, lat: ~ 2^2 } hitcount: 11884 { pid: 2035, prio: 120, lat: ~ 2^2 } hitcount: 21885 { pid: 2031, prio: 9, lat: ~ 2^2 } hitcount: 1761886 { pid: 2028, prio: 120, lat: ~ 2^2 } hitcount: 151887 { pid: 2033, prio: 9, lat: ~ 2^2 } hitcount: 911888 { pid: 2032, prio: 9, lat: ~ 2^2 } hitcount: 1251889 { pid: 2029, prio: 120, lat: ~ 2^2 } hitcount: 41890 { pid: 2031, prio: 120, lat: ~ 2^2 } hitcount: 31891 { pid: 2029, prio: 120, lat: ~ 2^3 } hitcount: 21892 { pid: 2035, prio: 9, lat: ~ 2^3 } hitcount: 411893 { pid: 2030, prio: 120, lat: ~ 2^3 } hitcount: 11894 { pid: 2032, prio: 9, lat: ~ 2^3 } hitcount: 321895 { pid: 2031, prio: 9, lat: ~ 2^3 } hitcount: 441896 { pid: 2034, prio: 9, lat: ~ 2^3 } hitcount: 401897 { pid: 2030, prio: 9, lat: ~ 2^3 } hitcount: 291898 { pid: 2033, prio: 9, lat: ~ 2^3 } hitcount: 311899 { pid: 2029, prio: 9, lat: ~ 2^3 } hitcount: 311900 { pid: 2028, prio: 120, lat: ~ 2^3 } hitcount: 181901 { pid: 2031, prio: 120, lat: ~ 2^3 } hitcount: 21902 { pid: 2028, prio: 120, lat: ~ 2^4 } hitcount: 11903 { pid: 2029, prio: 9, lat: ~ 2^4 } hitcount: 41904 { pid: 2031, prio: 120, lat: ~ 2^7 } hitcount: 11905 { pid: 2032, prio: 120, lat: ~ 2^7 } hitcount: 11906 1907 Totals:1908 Hits: 21221909 Entries: 301910 Dropped: 01911 1912 1913The latency values can also be grouped linearly by a given size with1914the ".buckets" modifier and specify a size (in this case groups of 10)::1915 1916 # echo 'hist:keys=pid,prio,lat.buckets=10:sort=lat' >> \1917 /sys/kernel/tracing/events/synthetic/wakeup_latency/trigger1918 1919 # event histogram1920 #1921 # trigger info: hist:keys=pid,prio,lat.buckets=10:vals=hitcount:sort=lat.buckets=10:size=2048 [active]1922 #1923 1924 { pid: 2067, prio: 9, lat: ~ 0-9 } hitcount: 2201925 { pid: 2068, prio: 9, lat: ~ 0-9 } hitcount: 1571926 { pid: 2070, prio: 9, lat: ~ 0-9 } hitcount: 1001927 { pid: 2067, prio: 120, lat: ~ 0-9 } hitcount: 61928 { pid: 2065, prio: 120, lat: ~ 0-9 } hitcount: 21929 { pid: 2066, prio: 120, lat: ~ 0-9 } hitcount: 21930 { pid: 2069, prio: 9, lat: ~ 0-9 } hitcount: 1221931 { pid: 2069, prio: 120, lat: ~ 0-9 } hitcount: 81932 { pid: 2070, prio: 120, lat: ~ 0-9 } hitcount: 11933 { pid: 2068, prio: 120, lat: ~ 0-9 } hitcount: 71934 { pid: 2066, prio: 9, lat: ~ 0-9 } hitcount: 3651935 { pid: 2064, prio: 120, lat: ~ 0-9 } hitcount: 351936 { pid: 2065, prio: 9, lat: ~ 0-9 } hitcount: 9981937 { pid: 2071, prio: 9, lat: ~ 0-9 } hitcount: 851938 { pid: 2065, prio: 9, lat: ~ 10-19 } hitcount: 21939 { pid: 2064, prio: 120, lat: ~ 10-19 } hitcount: 21940 1941 Totals:1942 Hits: 21121943 Entries: 161944 Dropped: 01945 1946To save stacktraces, create a synthetic event with a field of type "unsigned long[]"1947or even just "long[]". For example, to see how long a task is blocked in an1948uninterruptible state::1949 1950 # cd /sys/kernel/tracing1951 # echo 's:block_lat pid_t pid; u64 delta; unsigned long[] stack;' > dynamic_events1952 # echo 'hist:keys=next_pid:ts=common_timestamp.usecs,st=common_stacktrace if prev_state == 2' >> events/sched/sched_switch/trigger1953 # echo 'hist:keys=prev_pid:delta=common_timestamp.usecs-$ts,s=$st:onmax($delta).trace(block_lat,prev_pid,$delta,$s)' >> events/sched/sched_switch/trigger1954 # echo 1 > events/synthetic/block_lat/enable1955 # cat trace1956 1957 # tracer: nop1958 #1959 # entries-in-buffer/entries-written: 2/2 #P:81960 #1961 # _-----=> irqs-off/BH-disabled1962 # / _----=> need-resched1963 # | / _---=> hardirq/softirq1964 # || / _--=> preempt-depth1965 # ||| / _-=> migrate-disable1966 # |||| / delay1967 # TASK-PID CPU# ||||| TIMESTAMP FUNCTION1968 # | | | ||||| | |1969 <idle>-0 [005] d..4. 521.164922: block_lat: pid=0 delta=8322 stack=STACK:1970 => __schedule+0x448/0x7b01971 => schedule+0x5a/0xb01972 => io_schedule+0x42/0x701973 => bit_wait_io+0xd/0x601974 => __wait_on_bit+0x4b/0x1401975 => out_of_line_wait_on_bit+0x91/0xb01976 => jbd2_journal_commit_transaction+0x1679/0x1a701977 => kjournald2+0xa9/0x2801978 => kthread+0xe9/0x1101979 => ret_from_fork+0x2c/0x501980 1981 <...>-2 [004] d..4. 525.184257: block_lat: pid=2 delta=76 stack=STACK:1982 => __schedule+0x448/0x7b01983 => schedule+0x5a/0xb01984 => schedule_timeout+0x11a/0x1501985 => wait_for_completion_killable+0x144/0x1f01986 => __kthread_create_on_node+0xe7/0x1e01987 => kthread_create_on_node+0x51/0x701988 => create_worker+0xcc/0x1a01989 => worker_thread+0x2ad/0x3801990 => kthread+0xe9/0x1101991 => ret_from_fork+0x2c/0x501992 1993A synthetic event that has a stacktrace field may use it as a key in1994histogram::1995 1996 # echo 'hist:keys=delta.buckets=100,stack.stacktrace:sort=delta' > events/synthetic/block_lat/trigger1997 # cat events/synthetic/block_lat/hist1998 1999 # event histogram2000 #2001 # trigger info: hist:keys=delta.buckets=100,stack.stacktrace:vals=hitcount:sort=delta.buckets=100:size=2048 [active]2002 #2003 { delta: ~ 0-99, stack.stacktrace __schedule+0xa19/0x15202004 schedule+0x6b/0x1102005 io_schedule+0x46/0x802006 bit_wait_io+0x11/0x802007 __wait_on_bit+0x4e/0x1202008 out_of_line_wait_on_bit+0x8d/0xb02009 __wait_on_buffer+0x33/0x402010 jbd2_journal_commit_transaction+0x155a/0x19b02011 kjournald2+0xab/0x2702012 kthread+0xfa/0x1302013 ret_from_fork+0x29/0x502014 } hitcount: 12015 { delta: ~ 0-99, stack.stacktrace __schedule+0xa19/0x15202016 schedule+0x6b/0x1102017 io_schedule+0x46/0x802018 rq_qos_wait+0xd0/0x1702019 wbt_wait+0x9e/0xf02020 __rq_qos_throttle+0x25/0x402021 blk_mq_submit_bio+0x2c3/0x5b02022 __submit_bio+0xff/0x1902023 submit_bio_noacct_nocheck+0x25b/0x2b02024 submit_bio_noacct+0x20b/0x6002025 submit_bio+0x28/0x902026 ext4_bio_write_page+0x1e0/0x8c02027 mpage_submit_page+0x60/0x802028 mpage_process_page_bufs+0x16c/0x1802029 mpage_prepare_extent_to_map+0x23f/0x5302030 } hitcount: 12031 { delta: ~ 0-99, stack.stacktrace __schedule+0xa19/0x15202032 schedule+0x6b/0x1102033 schedule_hrtimeout_range_clock+0x97/0x1102034 schedule_hrtimeout_range+0x13/0x202035 usleep_range_state+0x65/0x902036 __intel_wait_for_register+0x1c1/0x230 [i915]2037 intel_psr_wait_for_idle_locked+0x171/0x2a0 [i915]2038 intel_pipe_update_start+0x169/0x360 [i915]2039 intel_update_crtc+0x112/0x490 [i915]2040 skl_commit_modeset_enables+0x199/0x600 [i915]2041 intel_atomic_commit_tail+0x7c4/0x1080 [i915]2042 intel_atomic_commit_work+0x12/0x20 [i915]2043 process_one_work+0x21c/0x3f02044 worker_thread+0x50/0x3e02045 kthread+0xfa/0x1302046 } hitcount: 32047 { delta: ~ 0-99, stack.stacktrace __schedule+0xa19/0x15202048 schedule+0x6b/0x1102049 schedule_timeout+0x11e/0x1602050 __wait_for_common+0x8f/0x1902051 wait_for_completion+0x24/0x302052 __flush_work.isra.0+0x1cc/0x3602053 flush_work+0xe/0x202054 drm_mode_rmfb+0x18b/0x1d0 [drm]2055 drm_mode_rmfb_ioctl+0x10/0x20 [drm]2056 drm_ioctl_kernel+0xb8/0x150 [drm]2057 drm_ioctl+0x243/0x560 [drm]2058 __x64_sys_ioctl+0x92/0xd02059 do_syscall_64+0x59/0x902060 entry_SYSCALL_64_after_hwframe+0x72/0xdc2061 } hitcount: 12062 { delta: ~ 0-99, stack.stacktrace __schedule+0xa19/0x15202063 schedule+0x6b/0x1102064 schedule_timeout+0x87/0x1602065 __wait_for_common+0x8f/0x1902066 wait_for_completion_timeout+0x1d/0x302067 drm_atomic_helper_wait_for_flip_done+0x57/0x90 [drm_kms_helper]2068 intel_atomic_commit_tail+0x8ce/0x1080 [i915]2069 intel_atomic_commit_work+0x12/0x20 [i915]2070 process_one_work+0x21c/0x3f02071 worker_thread+0x50/0x3e02072 kthread+0xfa/0x1302073 ret_from_fork+0x29/0x502074 } hitcount: 12075 { delta: ~ 100-199, stack.stacktrace __schedule+0xa19/0x15202076 schedule+0x6b/0x1102077 schedule_hrtimeout_range_clock+0x97/0x1102078 schedule_hrtimeout_range+0x13/0x202079 usleep_range_state+0x65/0x902080 pci_set_low_power_state+0x17f/0x1f02081 pci_set_power_state+0x49/0x2502082 pci_finish_runtime_suspend+0x4a/0x902083 pci_pm_runtime_suspend+0xcb/0x1b02084 __rpm_callback+0x48/0x1202085 rpm_callback+0x67/0x702086 rpm_suspend+0x167/0x7802087 rpm_idle+0x25a/0x3802088 pm_runtime_work+0x93/0xc02089 process_one_work+0x21c/0x3f02090 } hitcount: 12091 2092 Totals:2093 Hits: 102094 Entries: 72095 Dropped: 02096 20972.2.3 Hist trigger 'handlers' and 'actions'2098-------------------------------------------2099 2100A hist trigger 'action' is a function that's executed (in most cases2101conditionally) whenever a histogram entry is added or updated.2102 2103When a histogram entry is added or updated, a hist trigger 'handler'2104is what decides whether the corresponding action is actually invoked2105or not.2106 2107Hist trigger handlers and actions are paired together in the general2108form:2109 2110 <handler>.<action>2111 2112To specify a handler.action pair for a given event, simply specify2113that handler.action pair between colons in the hist trigger2114specification.2115 2116In theory, any handler can be combined with any action, but in2117practice, not every handler.action combination is currently supported;2118if a given handler.action combination isn't supported, the hist2119trigger will fail with -EINVAL;2120 2121The default 'handler.action' if none is explicitly specified is as it2122always has been, to simply update the set of values associated with an2123entry. Some applications, however, may want to perform additional2124actions at that point, such as generate another event, or compare and2125save a maximum.2126 2127The supported handlers and actions are listed below, and each is2128described in more detail in the following paragraphs, in the context2129of descriptions of some common and useful handler.action combinations.2130 2131The available handlers are:2132 2133 - onmatch(matching.event) - invoke action on any addition or update2134 - onmax(var) - invoke action if var exceeds current max2135 - onchange(var) - invoke action if var changes2136 2137The available actions are:2138 2139 - trace(<synthetic_event_name>,param list) - generate synthetic event2140 - save(field,...) - save current event fields2141 - snapshot() - snapshot the trace buffer2142 2143The following commonly-used handler.action pairs are available:2144 2145 - onmatch(matching.event).trace(<synthetic_event_name>,param list)2146 2147 The 'onmatch(matching.event).trace(<synthetic_event_name>,param2148 list)' hist trigger action is invoked whenever an event matches2149 and the histogram entry would be added or updated. It causes the2150 named synthetic event to be generated with the values given in the2151 'param list'. The result is the generation of a synthetic event2152 that consists of the values contained in those variables at the2153 time the invoking event was hit. For example, if the synthetic2154 event name is 'wakeup_latency', a wakeup_latency event is2155 generated using onmatch(event).trace(wakeup_latency,arg1,arg2).2156 2157 There is also an equivalent alternative form available for2158 generating synthetic events. In this form, the synthetic event2159 name is used as if it were a function name. For example, using2160 the 'wakeup_latency' synthetic event name again, the2161 wakeup_latency event would be generated by invoking it as if it2162 were a function call, with the event field values passed in as2163 arguments: onmatch(event).wakeup_latency(arg1,arg2). The syntax2164 for this form is:2165 2166 onmatch(matching.event).<synthetic_event_name>(param list)2167 2168 In either case, the 'param list' consists of one or more2169 parameters which may be either variables or fields defined on2170 either the 'matching.event' or the target event. The variables or2171 fields specified in the param list may be either fully-qualified2172 or unqualified. If a variable is specified as unqualified, it2173 must be unique between the two events. A field name used as a2174 param can be unqualified if it refers to the target event, but2175 must be fully qualified if it refers to the matching event. A2176 fully-qualified name is of the form 'system.event_name.$var_name'2177 or 'system.event_name.field'.2178 2179 The 'matching.event' specification is simply the fully qualified2180 event name of the event that matches the target event for the2181 onmatch() functionality, in the form 'system.event_name'. Histogram2182 keys of both events are compared to find if events match. In case2183 multiple histogram keys are used, they all must match in the specified2184 order.2185 2186 Finally, the number and type of variables/fields in the 'param2187 list' must match the number and types of the fields in the2188 synthetic event being generated.2189 2190 As an example the below defines a simple synthetic event and uses2191 a variable defined on the sched_wakeup_new event as a parameter2192 when invoking the synthetic event. Here we define the synthetic2193 event::2194 2195 # echo 'wakeup_new_test pid_t pid' >> \2196 /sys/kernel/tracing/synthetic_events2197 2198 # cat /sys/kernel/tracing/synthetic_events2199 wakeup_new_test pid_t pid2200 2201 The following hist trigger both defines the missing testpid2202 variable and specifies an onmatch() action that generates a2203 wakeup_new_test synthetic event whenever a sched_wakeup_new event2204 occurs, which because of the 'if comm == "cyclictest"' filter only2205 happens when the executable is cyclictest::2206 2207 # echo 'hist:keys=$testpid:testpid=pid:onmatch(sched.sched_wakeup_new).\2208 wakeup_new_test($testpid) if comm=="cyclictest"' >> \2209 /sys/kernel/tracing/events/sched/sched_wakeup_new/trigger2210 2211 Or, equivalently, using the 'trace' keyword syntax::2212 2213 # echo 'hist:keys=$testpid:testpid=pid:onmatch(sched.sched_wakeup_new).\2214 trace(wakeup_new_test,$testpid) if comm=="cyclictest"' >> \2215 /sys/kernel/tracing/events/sched/sched_wakeup_new/trigger2216 2217 Creating and displaying a histogram based on those events is now2218 just a matter of using the fields and new synthetic event in the2219 tracing/events/synthetic directory, as usual::2220 2221 # echo 'hist:keys=pid:sort=pid' >> \2222 /sys/kernel/tracing/events/synthetic/wakeup_new_test/trigger2223 2224 Running 'cyclictest' should cause wakeup_new events to generate2225 wakeup_new_test synthetic events which should result in histogram2226 output in the wakeup_new_test event's hist file::2227 2228 # cat /sys/kernel/tracing/events/synthetic/wakeup_new_test/hist2229 2230 A more typical usage would be to use two events to calculate a2231 latency. The following example uses a set of hist triggers to2232 produce a 'wakeup_latency' histogram.2233 2234 First, we define a 'wakeup_latency' synthetic event::2235 2236 # echo 'wakeup_latency u64 lat; pid_t pid; int prio' >> \2237 /sys/kernel/tracing/synthetic_events2238 2239 Next, we specify that whenever we see a sched_waking event for a2240 cyclictest thread, save the timestamp in a 'ts0' variable::2241 2242 # echo 'hist:keys=$saved_pid:saved_pid=pid:ts0=common_timestamp.usecs \2243 if comm=="cyclictest"' >> \2244 /sys/kernel/tracing/events/sched/sched_waking/trigger2245 2246 Then, when the corresponding thread is actually scheduled onto the2247 CPU by a sched_switch event (saved_pid matches next_pid), calculate2248 the latency and use that along with another variable and an event field2249 to generate a wakeup_latency synthetic event::2250 2251 # echo 'hist:keys=next_pid:wakeup_lat=common_timestamp.usecs-$ts0:\2252 onmatch(sched.sched_waking).wakeup_latency($wakeup_lat,\2253 $saved_pid,next_prio) if next_comm=="cyclictest"' >> \2254 /sys/kernel/tracing/events/sched/sched_switch/trigger2255 2256 We also need to create a histogram on the wakeup_latency synthetic2257 event in order to aggregate the generated synthetic event data::2258 2259 # echo 'hist:keys=pid,prio,lat:sort=pid,lat' >> \2260 /sys/kernel/tracing/events/synthetic/wakeup_latency/trigger2261 2262 Finally, once we've run cyclictest to actually generate some2263 events, we can see the output by looking at the wakeup_latency2264 synthetic event's hist file::2265 2266 # cat /sys/kernel/tracing/events/synthetic/wakeup_latency/hist2267 2268 - onmax(var).save(field,.. .)2269 2270 The 'onmax(var).save(field,...)' hist trigger action is invoked2271 whenever the value of 'var' associated with a histogram entry2272 exceeds the current maximum contained in that variable.2273 2274 The end result is that the trace event fields specified as the2275 onmax.save() params will be saved if 'var' exceeds the current2276 maximum for that hist trigger entry. This allows context from the2277 event that exhibited the new maximum to be saved for later2278 reference. When the histogram is displayed, additional fields2279 displaying the saved values will be printed.2280 2281 As an example the below defines a couple of hist triggers, one for2282 sched_waking and another for sched_switch, keyed on pid. Whenever2283 a sched_waking occurs, the timestamp is saved in the entry2284 corresponding to the current pid, and when the scheduler switches2285 back to that pid, the timestamp difference is calculated. If the2286 resulting latency, stored in wakeup_lat, exceeds the current2287 maximum latency, the values specified in the save() fields are2288 recorded::2289 2290 # echo 'hist:keys=pid:ts0=common_timestamp.usecs \2291 if comm=="cyclictest"' >> \2292 /sys/kernel/tracing/events/sched/sched_waking/trigger2293 2294 # echo 'hist:keys=next_pid:\2295 wakeup_lat=common_timestamp.usecs-$ts0:\2296 onmax($wakeup_lat).save(next_comm,prev_pid,prev_prio,prev_comm) \2297 if next_comm=="cyclictest"' >> \2298 /sys/kernel/tracing/events/sched/sched_switch/trigger2299 2300 When the histogram is displayed, the max value and the saved2301 values corresponding to the max are displayed following the rest2302 of the fields::2303 2304 # cat /sys/kernel/tracing/events/sched/sched_switch/hist2305 { next_pid: 2255 } hitcount: 2392306 common_timestamp-ts0: 02307 max: 272308 next_comm: cyclictest2309 prev_pid: 0 prev_prio: 120 prev_comm: swapper/12310 2311 { next_pid: 2256 } hitcount: 23552312 common_timestamp-ts0: 02313 max: 49 next_comm: cyclictest2314 prev_pid: 0 prev_prio: 120 prev_comm: swapper/02315 2316 Totals:2317 Hits: 129702318 Entries: 22319 Dropped: 02320 2321 - onmax(var).snapshot()2322 2323 The 'onmax(var).snapshot()' hist trigger action is invoked2324 whenever the value of 'var' associated with a histogram entry2325 exceeds the current maximum contained in that variable.2326 2327 The end result is that a global snapshot of the trace buffer will2328 be saved in the tracing/snapshot file if 'var' exceeds the current2329 maximum for any hist trigger entry.2330 2331 Note that in this case the maximum is a global maximum for the2332 current trace instance, which is the maximum across all buckets of2333 the histogram. The key of the specific trace event that caused2334 the global maximum and the global maximum itself are displayed,2335 along with a message stating that a snapshot has been taken and2336 where to find it. The user can use the key information displayed2337 to locate the corresponding bucket in the histogram for even more2338 detail.2339 2340 As an example the below defines a couple of hist triggers, one for2341 sched_waking and another for sched_switch, keyed on pid. Whenever2342 a sched_waking event occurs, the timestamp is saved in the entry2343 corresponding to the current pid, and when the scheduler switches2344 back to that pid, the timestamp difference is calculated. If the2345 resulting latency, stored in wakeup_lat, exceeds the current2346 maximum latency, a snapshot is taken. As part of the setup, all2347 the scheduler events are also enabled, which are the events that2348 will show up in the snapshot when it is taken at some point::2349 2350 # echo 1 > /sys/kernel/tracing/events/sched/enable2351 2352 # echo 'hist:keys=pid:ts0=common_timestamp.usecs \2353 if comm=="cyclictest"' >> \2354 /sys/kernel/tracing/events/sched/sched_waking/trigger2355 2356 # echo 'hist:keys=next_pid:wakeup_lat=common_timestamp.usecs-$ts0: \2357 onmax($wakeup_lat).save(next_prio,next_comm,prev_pid,prev_prio, \2358 prev_comm):onmax($wakeup_lat).snapshot() \2359 if next_comm=="cyclictest"' >> \2360 /sys/kernel/tracing/events/sched/sched_switch/trigger2361 2362 When the histogram is displayed, for each bucket the max value2363 and the saved values corresponding to the max are displayed2364 following the rest of the fields.2365 2366 If a snapshot was taken, there is also a message indicating that,2367 along with the value and event that triggered the global maximum::2368 2369 # cat /sys/kernel/tracing/events/sched/sched_switch/hist2370 { next_pid: 2101 } hitcount: 2002371 max: 52 next_prio: 120 next_comm: cyclictest \2372 prev_pid: 0 prev_prio: 120 prev_comm: swapper/62373 2374 { next_pid: 2103 } hitcount: 13262375 max: 572 next_prio: 19 next_comm: cyclictest \2376 prev_pid: 0 prev_prio: 120 prev_comm: swapper/12377 2378 { next_pid: 2102 } hitcount: 1982 \2379 max: 74 next_prio: 19 next_comm: cyclictest \2380 prev_pid: 0 prev_prio: 120 prev_comm: swapper/52381 2382 Snapshot taken (see tracing/snapshot). Details:2383 triggering value { onmax($wakeup_lat) }: 572 \2384 triggered by event with key: { next_pid: 2103 }2385 2386 Totals:2387 Hits: 35082388 Entries: 32389 Dropped: 02390 2391 In the above case, the event that triggered the global maximum has2392 the key with next_pid == 2103. If you look at the bucket that has2393 2103 as the key, you'll find the additional values save()'d along2394 with the local maximum for that bucket, which should be the same2395 as the global maximum (since that was the same value that2396 triggered the global snapshot).2397 2398 And finally, looking at the snapshot data should show at or near2399 the end the event that triggered the snapshot (in this case you2400 can verify the timestamps between the sched_waking and2401 sched_switch events, which should match the time displayed in the2402 global maximum)::2403 2404 # cat /sys/kernel/tracing/snapshot2405 2406 <...>-2103 [005] d..3 309.873125: sched_switch: prev_comm=cyclictest prev_pid=2103 prev_prio=19 prev_state=D ==> next_comm=swapper/5 next_pid=0 next_prio=1202407 <idle>-0 [005] d.h3 309.873611: sched_waking: comm=cyclictest pid=2102 prio=19 target_cpu=0052408 <idle>-0 [005] dNh4 309.873613: sched_wakeup: comm=cyclictest pid=2102 prio=19 target_cpu=0052409 <idle>-0 [005] d..3 309.873616: sched_switch: prev_comm=swapper/5 prev_pid=0 prev_prio=120 prev_state=S ==> next_comm=cyclictest next_pid=2102 next_prio=192410 <...>-2102 [005] d..3 309.873625: sched_switch: prev_comm=cyclictest prev_pid=2102 prev_prio=19 prev_state=D ==> next_comm=swapper/5 next_pid=0 next_prio=1202411 <idle>-0 [005] d.h3 309.874624: sched_waking: comm=cyclictest pid=2102 prio=19 target_cpu=0052412 <idle>-0 [005] dNh4 309.874626: sched_wakeup: comm=cyclictest pid=2102 prio=19 target_cpu=0052413 <idle>-0 [005] dNh3 309.874628: sched_waking: comm=cyclictest pid=2103 prio=19 target_cpu=0052414 <idle>-0 [005] dNh4 309.874630: sched_wakeup: comm=cyclictest pid=2103 prio=19 target_cpu=0052415 <idle>-0 [005] d..3 309.874633: sched_switch: prev_comm=swapper/5 prev_pid=0 prev_prio=120 prev_state=S ==> next_comm=cyclictest next_pid=2102 next_prio=192416 <idle>-0 [004] d.h3 309.874757: sched_waking: comm=gnome-terminal- pid=1699 prio=120 target_cpu=0042417 <idle>-0 [004] dNh4 309.874762: sched_wakeup: comm=gnome-terminal- pid=1699 prio=120 target_cpu=0042418 <idle>-0 [004] d..3 309.874766: sched_switch: prev_comm=swapper/4 prev_pid=0 prev_prio=120 prev_state=S ==> next_comm=gnome-terminal- next_pid=1699 next_prio=1202419 gnome-terminal--1699 [004] d.h2 309.874941: sched_stat_runtime: comm=gnome-terminal- pid=1699 runtime=180706 [ns] vruntime=1126870572 [ns]2420 <idle>-0 [003] d.s4 309.874956: sched_waking: comm=rcu_sched pid=9 prio=120 target_cpu=0072421 <idle>-0 [003] d.s5 309.874960: sched_wake_idle_without_ipi: cpu=72422 <idle>-0 [003] d.s5 309.874961: sched_wakeup: comm=rcu_sched pid=9 prio=120 target_cpu=0072423 <idle>-0 [007] d..3 309.874963: sched_switch: prev_comm=swapper/7 prev_pid=0 prev_prio=120 prev_state=S ==> next_comm=rcu_sched next_pid=9 next_prio=1202424 rcu_sched-9 [007] d..3 309.874973: sched_stat_runtime: comm=rcu_sched pid=9 runtime=13646 [ns] vruntime=22531430286 [ns]2425 rcu_sched-9 [007] d..3 309.874978: sched_switch: prev_comm=rcu_sched prev_pid=9 prev_prio=120 prev_state=R+ ==> next_comm=swapper/7 next_pid=0 next_prio=1202426 <...>-2102 [005] d..4 309.874994: sched_migrate_task: comm=cyclictest pid=2103 prio=19 orig_cpu=5 dest_cpu=12427 <...>-2102 [005] d..4 309.875185: sched_wake_idle_without_ipi: cpu=12428 <idle>-0 [001] d..3 309.875200: sched_switch: prev_comm=swapper/1 prev_pid=0 prev_prio=120 prev_state=S ==> next_comm=cyclictest next_pid=2103 next_prio=192429 2430 - onchange(var).save(field,.. .)2431 2432 The 'onchange(var).save(field,...)' hist trigger action is invoked2433 whenever the value of 'var' associated with a histogram entry2434 changes.2435 2436 The end result is that the trace event fields specified as the2437 onchange.save() params will be saved if 'var' changes for that2438 hist trigger entry. This allows context from the event that2439 changed the value to be saved for later reference. When the2440 histogram is displayed, additional fields displaying the saved2441 values will be printed.2442 2443 - onchange(var).snapshot()2444 2445 The 'onchange(var).snapshot()' hist trigger action is invoked2446 whenever the value of 'var' associated with a histogram entry2447 changes.2448 2449 The end result is that a global snapshot of the trace buffer will2450 be saved in the tracing/snapshot file if 'var' changes for any2451 hist trigger entry.2452 2453 Note that in this case the changed value is a global variable2454 associated with current trace instance. The key of the specific2455 trace event that caused the value to change and the global value2456 itself are displayed, along with a message stating that a snapshot2457 has been taken and where to find it. The user can use the key2458 information displayed to locate the corresponding bucket in the2459 histogram for even more detail.2460 2461 As an example the below defines a hist trigger on the tcp_probe2462 event, keyed on dport. Whenever a tcp_probe event occurs, the2463 cwnd field is checked against the current value stored in the2464 $cwnd variable. If the value has changed, a snapshot is taken.2465 As part of the setup, all the scheduler and tcp events are also2466 enabled, which are the events that will show up in the snapshot2467 when it is taken at some point::2468 2469 # echo 1 > /sys/kernel/tracing/events/sched/enable2470 # echo 1 > /sys/kernel/tracing/events/tcp/enable2471 2472 # echo 'hist:keys=dport:cwnd=snd_cwnd: \2473 onchange($cwnd).save(snd_wnd,srtt,rcv_wnd): \2474 onchange($cwnd).snapshot()' >> \2475 /sys/kernel/tracing/events/tcp/tcp_probe/trigger2476 2477 When the histogram is displayed, for each bucket the tracked value2478 and the saved values corresponding to that value are displayed2479 following the rest of the fields.2480 2481 If a snapshot was taken, there is also a message indicating that,2482 along with the value and event that triggered the snapshot::2483 2484 # cat /sys/kernel/tracing/events/tcp/tcp_probe/hist2485 2486 { dport: 1521 } hitcount: 82487 changed: 10 snd_wnd: 35456 srtt: 154262 rcv_wnd: 421122488 2489 { dport: 80 } hitcount: 232490 changed: 10 snd_wnd: 28960 srtt: 19604 rcv_wnd: 293122491 2492 { dport: 9001 } hitcount: 1722493 changed: 10 snd_wnd: 48384 srtt: 260444 rcv_wnd: 551682494 2495 { dport: 443 } hitcount: 2112496 changed: 10 snd_wnd: 26960 srtt: 17379 rcv_wnd: 288002497 2498 Snapshot taken (see tracing/snapshot). Details:2499 2500 triggering value { onchange($cwnd) }: 102501 triggered by event with key: { dport: 80 }2502 2503 Totals:2504 Hits: 4142505 Entries: 42506 Dropped: 02507 2508 In the above case, the event that triggered the snapshot has the2509 key with dport == 80. If you look at the bucket that has 80 as2510 the key, you'll find the additional values save()'d along with the2511 changed value for that bucket, which should be the same as the2512 global changed value (since that was the same value that triggered2513 the global snapshot).2514 2515 And finally, looking at the snapshot data should show at or near2516 the end the event that triggered the snapshot::2517 2518 # cat /sys/kernel/tracing/snapshot2519 2520 gnome-shell-1261 [006] dN.3 49.823113: sched_stat_runtime: comm=gnome-shell pid=1261 runtime=49347 [ns] vruntime=1835730389 [ns]2521 kworker/u16:4-773 [003] d..3 49.823114: sched_switch: prev_comm=kworker/u16:4 prev_pid=773 prev_prio=120 prev_state=R+ ==> next_comm=kworker/3:2 next_pid=135 next_prio=1202522 gnome-shell-1261 [006] d..3 49.823114: sched_switch: prev_comm=gnome-shell prev_pid=1261 prev_prio=120 prev_state=R+ ==> next_comm=kworker/6:2 next_pid=387 next_prio=1202523 kworker/3:2-135 [003] d..3 49.823118: sched_stat_runtime: comm=kworker/3:2 pid=135 runtime=5339 [ns] vruntime=17815800388 [ns]2524 kworker/6:2-387 [006] d..3 49.823120: sched_stat_runtime: comm=kworker/6:2 pid=387 runtime=9594 [ns] vruntime=14589605367 [ns]2525 kworker/6:2-387 [006] d..3 49.823122: sched_switch: prev_comm=kworker/6:2 prev_pid=387 prev_prio=120 prev_state=R+ ==> next_comm=gnome-shell next_pid=1261 next_prio=1202526 kworker/3:2-135 [003] d..3 49.823123: sched_switch: prev_comm=kworker/3:2 prev_pid=135 prev_prio=120 prev_state=T ==> next_comm=swapper/3 next_pid=0 next_prio=1202527 <idle>-0 [004] ..s7 49.823798: tcp_probe: src=10.0.0.10:54326 dest=23.215.104.193:80 mark=0x0 length=32 snd_nxt=0xe3ae2ff5 snd_una=0xe3ae2ecd snd_cwnd=10 ssthresh=2147483647 snd_wnd=28960 srtt=19604 rcv_wnd=293122528 25293. User space creating a trigger2530--------------------------------2531 2532Writing into /sys/kernel/tracing/trace_marker writes into the ftrace2533ring buffer. This can also act like an event, by writing into the trigger2534file located in /sys/kernel/tracing/events/ftrace/print/2535 2536Modifying cyclictest to write into the trace_marker file before it sleeps2537and after it wakes up, something like this::2538 2539 static void traceputs(char *str)2540 {2541 /* tracemark_fd is the trace_marker file descriptor */2542 if (tracemark_fd < 0)2543 return;2544 /* write the tracemark message */2545 write(tracemark_fd, str, strlen(str));2546 }2547 2548And later add something like::2549 2550 traceputs("start");2551 clock_nanosleep(...);2552 traceputs("end");2553 2554We can make a histogram from this::2555 2556 # cd /sys/kernel/tracing2557 # echo 'latency u64 lat' > synthetic_events2558 # echo 'hist:keys=common_pid:ts0=common_timestamp.usecs if buf == "start"' > events/ftrace/print/trigger2559 # echo 'hist:keys=common_pid:lat=common_timestamp.usecs-$ts0:onmatch(ftrace.print).latency($lat) if buf == "end"' >> events/ftrace/print/trigger2560 # echo 'hist:keys=lat,common_pid:sort=lat' > events/synthetic/latency/trigger2561 2562The above created a synthetic event called "latency" and two histograms2563against the trace_marker, one gets triggered when "start" is written into the2564trace_marker file and the other when "end" is written. If the pids match, then2565it will call the "latency" synthetic event with the calculated latency as its2566parameter. Finally, a histogram is added to the latency synthetic event to2567record the calculated latency along with the pid.2568 2569Now running cyclictest with::2570 2571 # ./cyclictest -p80 -d0 -i250 -n -a -t --tracemark -b 10002572 2573 -p80 : run threads at priority 802574 -d0 : have all threads run at the same interval2575 -i250 : start the interval at 250 microseconds (all threads will do this)2576 -n : sleep with nanosleep2577 -a : affine all threads to a separate CPU2578 -t : one thread per available CPU2579 --tracemark : enable trace mark writing2580 -b 1000 : stop if any latency is greater than 1000 microseconds2581 2582Note, the -b 1000 is used just to make --tracemark available.2583 2584Then we can see the histogram created by this with::2585 2586 # cat events/synthetic/latency/hist2587 # event histogram2588 #2589 # trigger info: hist:keys=lat,common_pid:vals=hitcount:sort=lat:size=2048 [active]2590 #2591 2592 { lat: 107, common_pid: 2039 } hitcount: 12593 { lat: 122, common_pid: 2041 } hitcount: 12594 { lat: 166, common_pid: 2039 } hitcount: 12595 { lat: 174, common_pid: 2039 } hitcount: 12596 { lat: 194, common_pid: 2041 } hitcount: 12597 { lat: 196, common_pid: 2036 } hitcount: 12598 { lat: 197, common_pid: 2038 } hitcount: 12599 { lat: 198, common_pid: 2039 } hitcount: 12600 { lat: 199, common_pid: 2039 } hitcount: 12601 { lat: 200, common_pid: 2041 } hitcount: 12602 { lat: 201, common_pid: 2039 } hitcount: 22603 { lat: 202, common_pid: 2038 } hitcount: 12604 { lat: 202, common_pid: 2043 } hitcount: 12605 { lat: 203, common_pid: 2039 } hitcount: 12606 { lat: 203, common_pid: 2036 } hitcount: 12607 { lat: 203, common_pid: 2041 } hitcount: 12608 { lat: 206, common_pid: 2038 } hitcount: 22609 { lat: 207, common_pid: 2039 } hitcount: 12610 { lat: 207, common_pid: 2036 } hitcount: 12611 { lat: 208, common_pid: 2040 } hitcount: 12612 { lat: 209, common_pid: 2043 } hitcount: 12613 { lat: 210, common_pid: 2039 } hitcount: 12614 { lat: 211, common_pid: 2039 } hitcount: 42615 { lat: 212, common_pid: 2043 } hitcount: 12616 { lat: 212, common_pid: 2039 } hitcount: 22617 { lat: 213, common_pid: 2039 } hitcount: 12618 { lat: 214, common_pid: 2038 } hitcount: 12619 { lat: 214, common_pid: 2039 } hitcount: 22620 { lat: 214, common_pid: 2042 } hitcount: 12621 { lat: 215, common_pid: 2039 } hitcount: 12622 { lat: 217, common_pid: 2036 } hitcount: 12623 { lat: 217, common_pid: 2040 } hitcount: 12624 { lat: 217, common_pid: 2039 } hitcount: 12625 { lat: 218, common_pid: 2039 } hitcount: 62626 { lat: 219, common_pid: 2039 } hitcount: 92627 { lat: 220, common_pid: 2039 } hitcount: 112628 { lat: 221, common_pid: 2039 } hitcount: 52629 { lat: 221, common_pid: 2042 } hitcount: 12630 { lat: 222, common_pid: 2039 } hitcount: 72631 { lat: 223, common_pid: 2036 } hitcount: 12632 { lat: 223, common_pid: 2039 } hitcount: 32633 { lat: 224, common_pid: 2039 } hitcount: 42634 { lat: 224, common_pid: 2037 } hitcount: 12635 { lat: 224, common_pid: 2036 } hitcount: 22636 { lat: 225, common_pid: 2039 } hitcount: 52637 { lat: 225, common_pid: 2042 } hitcount: 12638 { lat: 226, common_pid: 2039 } hitcount: 72639 { lat: 226, common_pid: 2036 } hitcount: 42640 { lat: 227, common_pid: 2039 } hitcount: 62641 { lat: 227, common_pid: 2036 } hitcount: 122642 { lat: 227, common_pid: 2043 } hitcount: 12643 { lat: 228, common_pid: 2039 } hitcount: 72644 { lat: 228, common_pid: 2036 } hitcount: 142645 { lat: 229, common_pid: 2039 } hitcount: 92646 { lat: 229, common_pid: 2036 } hitcount: 82647 { lat: 229, common_pid: 2038 } hitcount: 12648 { lat: 230, common_pid: 2039 } hitcount: 112649 { lat: 230, common_pid: 2036 } hitcount: 62650 { lat: 230, common_pid: 2043 } hitcount: 12651 { lat: 230, common_pid: 2042 } hitcount: 22652 { lat: 231, common_pid: 2041 } hitcount: 12653 { lat: 231, common_pid: 2036 } hitcount: 62654 { lat: 231, common_pid: 2043 } hitcount: 12655 { lat: 231, common_pid: 2039 } hitcount: 82656 { lat: 232, common_pid: 2037 } hitcount: 12657 { lat: 232, common_pid: 2039 } hitcount: 62658 { lat: 232, common_pid: 2040 } hitcount: 22659 { lat: 232, common_pid: 2036 } hitcount: 52660 { lat: 232, common_pid: 2043 } hitcount: 12661 { lat: 233, common_pid: 2036 } hitcount: 52662 { lat: 233, common_pid: 2039 } hitcount: 112663 { lat: 234, common_pid: 2039 } hitcount: 42664 { lat: 234, common_pid: 2038 } hitcount: 22665 { lat: 234, common_pid: 2043 } hitcount: 22666 { lat: 234, common_pid: 2036 } hitcount: 112667 { lat: 234, common_pid: 2040 } hitcount: 12668 { lat: 235, common_pid: 2037 } hitcount: 22669 { lat: 235, common_pid: 2036 } hitcount: 82670 { lat: 235, common_pid: 2043 } hitcount: 22671 { lat: 235, common_pid: 2039 } hitcount: 52672 { lat: 235, common_pid: 2042 } hitcount: 22673 { lat: 235, common_pid: 2040 } hitcount: 42674 { lat: 235, common_pid: 2041 } hitcount: 12675 { lat: 236, common_pid: 2036 } hitcount: 72676 { lat: 236, common_pid: 2037 } hitcount: 12677 { lat: 236, common_pid: 2041 } hitcount: 52678 { lat: 236, common_pid: 2039 } hitcount: 32679 { lat: 236, common_pid: 2043 } hitcount: 92680 { lat: 236, common_pid: 2040 } hitcount: 72681 { lat: 237, common_pid: 2037 } hitcount: 12682 { lat: 237, common_pid: 2040 } hitcount: 12683 { lat: 237, common_pid: 2036 } hitcount: 92684 { lat: 237, common_pid: 2039 } hitcount: 32685 { lat: 237, common_pid: 2043 } hitcount: 82686 { lat: 237, common_pid: 2042 } hitcount: 22687 { lat: 237, common_pid: 2041 } hitcount: 22688 { lat: 238, common_pid: 2043 } hitcount: 102689 { lat: 238, common_pid: 2040 } hitcount: 12690 { lat: 238, common_pid: 2037 } hitcount: 92691 { lat: 238, common_pid: 2038 } hitcount: 12692 { lat: 238, common_pid: 2039 } hitcount: 12693 { lat: 238, common_pid: 2042 } hitcount: 32694 { lat: 238, common_pid: 2036 } hitcount: 72695 { lat: 239, common_pid: 2041 } hitcount: 12696 { lat: 239, common_pid: 2043 } hitcount: 112697 { lat: 239, common_pid: 2037 } hitcount: 112698 { lat: 239, common_pid: 2038 } hitcount: 62699 { lat: 239, common_pid: 2036 } hitcount: 72700 { lat: 239, common_pid: 2040 } hitcount: 12701 { lat: 239, common_pid: 2042 } hitcount: 92702 { lat: 240, common_pid: 2037 } hitcount: 292703 { lat: 240, common_pid: 2043 } hitcount: 152704 { lat: 240, common_pid: 2040 } hitcount: 442705 { lat: 240, common_pid: 2039 } hitcount: 12706 { lat: 240, common_pid: 2041 } hitcount: 22707 { lat: 240, common_pid: 2038 } hitcount: 12708 { lat: 240, common_pid: 2036 } hitcount: 102709 { lat: 240, common_pid: 2042 } hitcount: 132710 { lat: 241, common_pid: 2036 } hitcount: 212711 { lat: 241, common_pid: 2041 } hitcount: 362712 { lat: 241, common_pid: 2037 } hitcount: 342713 { lat: 241, common_pid: 2042 } hitcount: 142714 { lat: 241, common_pid: 2040 } hitcount: 942715 { lat: 241, common_pid: 2039 } hitcount: 122716 { lat: 241, common_pid: 2038 } hitcount: 22717 { lat: 241, common_pid: 2043 } hitcount: 282718 { lat: 242, common_pid: 2040 } hitcount: 1092719 { lat: 242, common_pid: 2041 } hitcount: 5062720 { lat: 242, common_pid: 2039 } hitcount: 1552721 { lat: 242, common_pid: 2042 } hitcount: 212722 { lat: 242, common_pid: 2037 } hitcount: 522723 { lat: 242, common_pid: 2043 } hitcount: 212724 { lat: 242, common_pid: 2036 } hitcount: 162725 { lat: 242, common_pid: 2038 } hitcount: 1562726 { lat: 243, common_pid: 2037 } hitcount: 462727 { lat: 243, common_pid: 2039 } hitcount: 402728 { lat: 243, common_pid: 2042 } hitcount: 1192729 { lat: 243, common_pid: 2041 } hitcount: 6112730 { lat: 243, common_pid: 2036 } hitcount: 692731 { lat: 243, common_pid: 2038 } hitcount: 7842732 { lat: 243, common_pid: 2040 } hitcount: 3232733 { lat: 243, common_pid: 2043 } hitcount: 142734 { lat: 244, common_pid: 2043 } hitcount: 352735 { lat: 244, common_pid: 2042 } hitcount: 3052736 { lat: 244, common_pid: 2039 } hitcount: 82737 { lat: 244, common_pid: 2040 } hitcount: 45152738 { lat: 244, common_pid: 2038 } hitcount: 3712739 { lat: 244, common_pid: 2037 } hitcount: 312740 { lat: 244, common_pid: 2036 } hitcount: 1142741 { lat: 244, common_pid: 2041 } hitcount: 33962742 { lat: 245, common_pid: 2036 } hitcount: 7002743 { lat: 245, common_pid: 2041 } hitcount: 27722744 { lat: 245, common_pid: 2037 } hitcount: 2682745 { lat: 245, common_pid: 2039 } hitcount: 4722746 { lat: 245, common_pid: 2038 } hitcount: 27582747 { lat: 245, common_pid: 2042 } hitcount: 38332748 { lat: 245, common_pid: 2040 } hitcount: 31052749 { lat: 245, common_pid: 2043 } hitcount: 6452750 { lat: 246, common_pid: 2038 } hitcount: 34512751 { lat: 246, common_pid: 2041 } hitcount: 1422752 { lat: 246, common_pid: 2037 } hitcount: 51012753 { lat: 246, common_pid: 2040 } hitcount: 682754 { lat: 246, common_pid: 2043 } hitcount: 50992755 { lat: 246, common_pid: 2039 } hitcount: 56082756 { lat: 246, common_pid: 2042 } hitcount: 37232757 { lat: 246, common_pid: 2036 } hitcount: 47382758 { lat: 247, common_pid: 2042 } hitcount: 3122759 { lat: 247, common_pid: 2043 } hitcount: 23852760 { lat: 247, common_pid: 2041 } hitcount: 4522761 { lat: 247, common_pid: 2038 } hitcount: 7922762 { lat: 247, common_pid: 2040 } hitcount: 782763 { lat: 247, common_pid: 2036 } hitcount: 23752764 { lat: 247, common_pid: 2039 } hitcount: 18342765 { lat: 247, common_pid: 2037 } hitcount: 26552766 { lat: 248, common_pid: 2037 } hitcount: 362767 { lat: 248, common_pid: 2042 } hitcount: 112768 { lat: 248, common_pid: 2038 } hitcount: 1222769 { lat: 248, common_pid: 2036 } hitcount: 1352770 { lat: 248, common_pid: 2039 } hitcount: 262771 { lat: 248, common_pid: 2041 } hitcount: 5032772 { lat: 248, common_pid: 2043 } hitcount: 662773 { lat: 248, common_pid: 2040 } hitcount: 462774 { lat: 249, common_pid: 2037 } hitcount: 292775 { lat: 249, common_pid: 2038 } hitcount: 12776 { lat: 249, common_pid: 2043 } hitcount: 292777 { lat: 249, common_pid: 2039 } hitcount: 82778 { lat: 249, common_pid: 2042 } hitcount: 562779 { lat: 249, common_pid: 2040 } hitcount: 272780 { lat: 249, common_pid: 2041 } hitcount: 112781 { lat: 249, common_pid: 2036 } hitcount: 272782 { lat: 250, common_pid: 2038 } hitcount: 12783 { lat: 250, common_pid: 2036 } hitcount: 302784 { lat: 250, common_pid: 2040 } hitcount: 192785 { lat: 250, common_pid: 2043 } hitcount: 222786 { lat: 250, common_pid: 2042 } hitcount: 202787 { lat: 250, common_pid: 2041 } hitcount: 12788 { lat: 250, common_pid: 2039 } hitcount: 62789 { lat: 250, common_pid: 2037 } hitcount: 482790 { lat: 251, common_pid: 2037 } hitcount: 432791 { lat: 251, common_pid: 2039 } hitcount: 12792 { lat: 251, common_pid: 2036 } hitcount: 122793 { lat: 251, common_pid: 2042 } hitcount: 22794 { lat: 251, common_pid: 2041 } hitcount: 12795 { lat: 251, common_pid: 2043 } hitcount: 152796 { lat: 251, common_pid: 2040 } hitcount: 32797 { lat: 252, common_pid: 2040 } hitcount: 12798 { lat: 252, common_pid: 2036 } hitcount: 122799 { lat: 252, common_pid: 2037 } hitcount: 212800 { lat: 252, common_pid: 2043 } hitcount: 142801 { lat: 253, common_pid: 2037 } hitcount: 212802 { lat: 253, common_pid: 2039 } hitcount: 22803 { lat: 253, common_pid: 2036 } hitcount: 92804 { lat: 253, common_pid: 2043 } hitcount: 62805 { lat: 253, common_pid: 2040 } hitcount: 12806 { lat: 254, common_pid: 2036 } hitcount: 82807 { lat: 254, common_pid: 2043 } hitcount: 32808 { lat: 254, common_pid: 2041 } hitcount: 12809 { lat: 254, common_pid: 2042 } hitcount: 12810 { lat: 254, common_pid: 2039 } hitcount: 12811 { lat: 254, common_pid: 2037 } hitcount: 122812 { lat: 255, common_pid: 2043 } hitcount: 12813 { lat: 255, common_pid: 2037 } hitcount: 22814 { lat: 255, common_pid: 2036 } hitcount: 22815 { lat: 255, common_pid: 2039 } hitcount: 82816 { lat: 256, common_pid: 2043 } hitcount: 12817 { lat: 256, common_pid: 2036 } hitcount: 42818 { lat: 256, common_pid: 2039 } hitcount: 62819 { lat: 257, common_pid: 2039 } hitcount: 52820 { lat: 257, common_pid: 2036 } hitcount: 42821 { lat: 258, common_pid: 2039 } hitcount: 52822 { lat: 258, common_pid: 2036 } hitcount: 22823 { lat: 259, common_pid: 2036 } hitcount: 72824 { lat: 259, common_pid: 2039 } hitcount: 72825 { lat: 260, common_pid: 2036 } hitcount: 82826 { lat: 260, common_pid: 2039 } hitcount: 62827 { lat: 261, common_pid: 2036 } hitcount: 52828 { lat: 261, common_pid: 2039 } hitcount: 72829 { lat: 262, common_pid: 2039 } hitcount: 52830 { lat: 262, common_pid: 2036 } hitcount: 52831 { lat: 263, common_pid: 2039 } hitcount: 72832 { lat: 263, common_pid: 2036 } hitcount: 72833 { lat: 264, common_pid: 2039 } hitcount: 92834 { lat: 264, common_pid: 2036 } hitcount: 92835 { lat: 265, common_pid: 2036 } hitcount: 52836 { lat: 265, common_pid: 2039 } hitcount: 12837 { lat: 266, common_pid: 2036 } hitcount: 12838 { lat: 266, common_pid: 2039 } hitcount: 32839 { lat: 267, common_pid: 2036 } hitcount: 12840 { lat: 267, common_pid: 2039 } hitcount: 32841 { lat: 268, common_pid: 2036 } hitcount: 12842 { lat: 268, common_pid: 2039 } hitcount: 62843 { lat: 269, common_pid: 2036 } hitcount: 12844 { lat: 269, common_pid: 2043 } hitcount: 12845 { lat: 269, common_pid: 2039 } hitcount: 22846 { lat: 270, common_pid: 2040 } hitcount: 12847 { lat: 270, common_pid: 2039 } hitcount: 62848 { lat: 271, common_pid: 2041 } hitcount: 12849 { lat: 271, common_pid: 2039 } hitcount: 52850 { lat: 272, common_pid: 2039 } hitcount: 102851 { lat: 273, common_pid: 2039 } hitcount: 82852 { lat: 274, common_pid: 2039 } hitcount: 22853 { lat: 275, common_pid: 2039 } hitcount: 12854 { lat: 276, common_pid: 2039 } hitcount: 22855 { lat: 276, common_pid: 2037 } hitcount: 12856 { lat: 276, common_pid: 2038 } hitcount: 12857 { lat: 277, common_pid: 2039 } hitcount: 12858 { lat: 277, common_pid: 2042 } hitcount: 12859 { lat: 278, common_pid: 2039 } hitcount: 12860 { lat: 279, common_pid: 2039 } hitcount: 42861 { lat: 279, common_pid: 2043 } hitcount: 12862 { lat: 280, common_pid: 2039 } hitcount: 32863 { lat: 283, common_pid: 2036 } hitcount: 22864 { lat: 284, common_pid: 2039 } hitcount: 12865 { lat: 284, common_pid: 2043 } hitcount: 12866 { lat: 288, common_pid: 2039 } hitcount: 12867 { lat: 289, common_pid: 2039 } hitcount: 12868 { lat: 300, common_pid: 2039 } hitcount: 12869 { lat: 384, common_pid: 2039 } hitcount: 12870 2871 Totals:2872 Hits: 676252873 Entries: 2782874 Dropped: 02875 2876Note, the writes are around the sleep, so ideally they will all be of 2502877microseconds. If you are wondering how there are several that are under2878250 microseconds, that is because the way cyclictest works, is if one2879iteration comes in late, the next one will set the timer to wake up less that2880250. That is, if an iteration came in 50 microseconds late, the next wake up2881will be at 200 microseconds.2882 2883But this could easily be done in userspace. To make this even more2884interesting, we can mix the histogram between events that happened in the2885kernel with trace_marker::2886 2887 # cd /sys/kernel/tracing2888 # echo 'latency u64 lat' > synthetic_events2889 # echo 'hist:keys=pid:ts0=common_timestamp.usecs' > events/sched/sched_waking/trigger2890 # echo 'hist:keys=common_pid:lat=common_timestamp.usecs-$ts0:onmatch(sched.sched_waking).latency($lat) if buf == "end"' > events/ftrace/print/trigger2891 # echo 'hist:keys=lat,common_pid:sort=lat' > events/synthetic/latency/trigger2892 2893The difference this time is that instead of using the trace_marker to start2894the latency, the sched_waking event is used, matching the common_pid for the2895trace_marker write with the pid that is being woken by sched_waking.2896 2897After running cyclictest again with the same parameters, we now have::2898 2899 # cat events/synthetic/latency/hist2900 # event histogram2901 #2902 # trigger info: hist:keys=lat,common_pid:vals=hitcount:sort=lat:size=2048 [active]2903 #2904 2905 { lat: 7, common_pid: 2302 } hitcount: 6402906 { lat: 7, common_pid: 2299 } hitcount: 422907 { lat: 7, common_pid: 2303 } hitcount: 182908 { lat: 7, common_pid: 2305 } hitcount: 1662909 { lat: 7, common_pid: 2306 } hitcount: 12910 { lat: 7, common_pid: 2301 } hitcount: 912911 { lat: 7, common_pid: 2300 } hitcount: 172912 { lat: 8, common_pid: 2303 } hitcount: 82962913 { lat: 8, common_pid: 2304 } hitcount: 68642914 { lat: 8, common_pid: 2305 } hitcount: 94642915 { lat: 8, common_pid: 2301 } hitcount: 92132916 { lat: 8, common_pid: 2306 } hitcount: 62462917 { lat: 8, common_pid: 2302 } hitcount: 87972918 { lat: 8, common_pid: 2299 } hitcount: 87712919 { lat: 8, common_pid: 2300 } hitcount: 81192920 { lat: 9, common_pid: 2305 } hitcount: 15192921 { lat: 9, common_pid: 2299 } hitcount: 23462922 { lat: 9, common_pid: 2303 } hitcount: 28412923 { lat: 9, common_pid: 2301 } hitcount: 18462924 { lat: 9, common_pid: 2304 } hitcount: 38612925 { lat: 9, common_pid: 2302 } hitcount: 12102926 { lat: 9, common_pid: 2300 } hitcount: 27622927 { lat: 9, common_pid: 2306 } hitcount: 42472928 { lat: 10, common_pid: 2299 } hitcount: 162929 { lat: 10, common_pid: 2306 } hitcount: 3332930 { lat: 10, common_pid: 2303 } hitcount: 162931 { lat: 10, common_pid: 2304 } hitcount: 1682932 { lat: 10, common_pid: 2302 } hitcount: 2402933 { lat: 10, common_pid: 2301 } hitcount: 282934 { lat: 10, common_pid: 2300 } hitcount: 952935 { lat: 10, common_pid: 2305 } hitcount: 182936 { lat: 11, common_pid: 2303 } hitcount: 52937 { lat: 11, common_pid: 2305 } hitcount: 82938 { lat: 11, common_pid: 2306 } hitcount: 2212939 { lat: 11, common_pid: 2302 } hitcount: 762940 { lat: 11, common_pid: 2304 } hitcount: 262941 { lat: 11, common_pid: 2300 } hitcount: 1252942 { lat: 11, common_pid: 2299 } hitcount: 22943 { lat: 12, common_pid: 2305 } hitcount: 32944 { lat: 12, common_pid: 2300 } hitcount: 62945 { lat: 12, common_pid: 2306 } hitcount: 902946 { lat: 12, common_pid: 2302 } hitcount: 42947 { lat: 12, common_pid: 2303 } hitcount: 12948 { lat: 12, common_pid: 2304 } hitcount: 1222949 { lat: 13, common_pid: 2300 } hitcount: 122950 { lat: 13, common_pid: 2301 } hitcount: 12951 { lat: 13, common_pid: 2306 } hitcount: 322952 { lat: 13, common_pid: 2302 } hitcount: 52953 { lat: 13, common_pid: 2305 } hitcount: 12954 { lat: 13, common_pid: 2303 } hitcount: 12955 { lat: 13, common_pid: 2304 } hitcount: 612956 { lat: 14, common_pid: 2303 } hitcount: 42957 { lat: 14, common_pid: 2306 } hitcount: 52958 { lat: 14, common_pid: 2305 } hitcount: 42959 { lat: 14, common_pid: 2304 } hitcount: 622960 { lat: 14, common_pid: 2302 } hitcount: 192961 { lat: 14, common_pid: 2300 } hitcount: 332962 { lat: 14, common_pid: 2299 } hitcount: 12963 { lat: 14, common_pid: 2301 } hitcount: 42964 { lat: 15, common_pid: 2305 } hitcount: 12965 { lat: 15, common_pid: 2302 } hitcount: 252966 { lat: 15, common_pid: 2300 } hitcount: 112967 { lat: 15, common_pid: 2299 } hitcount: 52968 { lat: 15, common_pid: 2301 } hitcount: 12969 { lat: 15, common_pid: 2304 } hitcount: 82970 { lat: 15, common_pid: 2303 } hitcount: 12971 { lat: 15, common_pid: 2306 } hitcount: 62972 { lat: 16, common_pid: 2302 } hitcount: 312973 { lat: 16, common_pid: 2306 } hitcount: 32974 { lat: 16, common_pid: 2300 } hitcount: 52975 { lat: 17, common_pid: 2302 } hitcount: 62976 { lat: 17, common_pid: 2303 } hitcount: 12977 { lat: 18, common_pid: 2304 } hitcount: 12978 { lat: 18, common_pid: 2302 } hitcount: 82979 { lat: 18, common_pid: 2299 } hitcount: 12980 { lat: 18, common_pid: 2301 } hitcount: 12981 { lat: 19, common_pid: 2303 } hitcount: 42982 { lat: 19, common_pid: 2304 } hitcount: 52983 { lat: 19, common_pid: 2302 } hitcount: 42984 { lat: 19, common_pid: 2299 } hitcount: 32985 { lat: 19, common_pid: 2306 } hitcount: 12986 { lat: 19, common_pid: 2300 } hitcount: 42987 { lat: 19, common_pid: 2305 } hitcount: 52988 { lat: 20, common_pid: 2299 } hitcount: 22989 { lat: 20, common_pid: 2302 } hitcount: 32990 { lat: 20, common_pid: 2305 } hitcount: 12991 { lat: 20, common_pid: 2300 } hitcount: 22992 { lat: 20, common_pid: 2301 } hitcount: 22993 { lat: 20, common_pid: 2303 } hitcount: 32994 { lat: 21, common_pid: 2305 } hitcount: 12995 { lat: 21, common_pid: 2299 } hitcount: 52996 { lat: 21, common_pid: 2303 } hitcount: 42997 { lat: 21, common_pid: 2302 } hitcount: 72998 { lat: 21, common_pid: 2300 } hitcount: 12999 { lat: 21, common_pid: 2301 } hitcount: 53000 { lat: 21, common_pid: 2304 } hitcount: 23001 { lat: 22, common_pid: 2302 } hitcount: 53002 { lat: 22, common_pid: 2303 } hitcount: 13003 { lat: 22, common_pid: 2306 } hitcount: 33004 { lat: 22, common_pid: 2301 } hitcount: 23005 { lat: 22, common_pid: 2300 } hitcount: 13006 { lat: 22, common_pid: 2299 } hitcount: 13007 { lat: 22, common_pid: 2305 } hitcount: 13008 { lat: 22, common_pid: 2304 } hitcount: 13009 { lat: 23, common_pid: 2299 } hitcount: 13010 { lat: 23, common_pid: 2306 } hitcount: 23011 { lat: 23, common_pid: 2302 } hitcount: 63012 { lat: 24, common_pid: 2302 } hitcount: 33013 { lat: 24, common_pid: 2300 } hitcount: 13014 { lat: 24, common_pid: 2306 } hitcount: 23015 { lat: 24, common_pid: 2305 } hitcount: 13016 { lat: 24, common_pid: 2299 } hitcount: 13017 { lat: 25, common_pid: 2300 } hitcount: 13018 { lat: 25, common_pid: 2302 } hitcount: 43019 { lat: 26, common_pid: 2302 } hitcount: 23020 { lat: 27, common_pid: 2305 } hitcount: 13021 { lat: 27, common_pid: 2300 } hitcount: 13022 { lat: 27, common_pid: 2302 } hitcount: 33023 { lat: 28, common_pid: 2306 } hitcount: 13024 { lat: 28, common_pid: 2302 } hitcount: 43025 { lat: 29, common_pid: 2302 } hitcount: 13026 { lat: 29, common_pid: 2300 } hitcount: 23027 { lat: 29, common_pid: 2306 } hitcount: 13028 { lat: 29, common_pid: 2304 } hitcount: 13029 { lat: 30, common_pid: 2302 } hitcount: 43030 { lat: 31, common_pid: 2302 } hitcount: 63031 { lat: 32, common_pid: 2302 } hitcount: 13032 { lat: 33, common_pid: 2299 } hitcount: 13033 { lat: 33, common_pid: 2302 } hitcount: 33034 { lat: 34, common_pid: 2302 } hitcount: 23035 { lat: 35, common_pid: 2302 } hitcount: 13036 { lat: 35, common_pid: 2304 } hitcount: 13037 { lat: 36, common_pid: 2302 } hitcount: 43038 { lat: 37, common_pid: 2302 } hitcount: 63039 { lat: 38, common_pid: 2302 } hitcount: 23040 { lat: 39, common_pid: 2302 } hitcount: 23041 { lat: 39, common_pid: 2304 } hitcount: 13042 { lat: 40, common_pid: 2304 } hitcount: 23043 { lat: 40, common_pid: 2302 } hitcount: 53044 { lat: 41, common_pid: 2304 } hitcount: 13045 { lat: 41, common_pid: 2302 } hitcount: 83046 { lat: 42, common_pid: 2302 } hitcount: 63047 { lat: 42, common_pid: 2304 } hitcount: 13048 { lat: 43, common_pid: 2302 } hitcount: 33049 { lat: 43, common_pid: 2304 } hitcount: 43050 { lat: 44, common_pid: 2302 } hitcount: 63051 { lat: 45, common_pid: 2302 } hitcount: 53052 { lat: 46, common_pid: 2302 } hitcount: 53053 { lat: 47, common_pid: 2302 } hitcount: 73054 { lat: 48, common_pid: 2301 } hitcount: 13055 { lat: 48, common_pid: 2302 } hitcount: 93056 { lat: 49, common_pid: 2302 } hitcount: 33057 { lat: 50, common_pid: 2302 } hitcount: 13058 { lat: 50, common_pid: 2301 } hitcount: 13059 { lat: 51, common_pid: 2302 } hitcount: 23060 { lat: 51, common_pid: 2301 } hitcount: 13061 { lat: 61, common_pid: 2302 } hitcount: 13062 { lat: 110, common_pid: 2302 } hitcount: 13063 3064 Totals:3065 Hits: 895653066 Entries: 1583067 Dropped: 03068 3069This doesn't tell us any information about how late cyclictest may have3070woken up, but it does show us a nice histogram of how long it took from3071the time that cyclictest was woken to the time it made it into user space.3072