1================ 2Event Histograms 3================ 4 5Documentation written by Tom Zanussi 6 71. Introduction 8=============== 9 10 Histogram triggers are special event triggers that can be used to 11 aggregate trace event data into histograms. For information on 12 trace events and event triggers, see Documentation/trace/events.rst. 13 14 152. Histogram Trigger Command 16============================ 17 18 A histogram trigger command is an event trigger command that 19 aggregates event hits into a hash table keyed on one or more trace 20 event format fields (or stacktrace) and a set of running totals 21 derived from one or more trace event format fields and/or event 22 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][:<handler>.<action>] [if <filter>] 29 30 When a matching event is hit, an entry is added to a hash table 31 using the key(s) and value(s) named. Keys and values correspond to 32 fields in the event's format description. Values must correspond to 33 numeric fields - on an event hit, the value(s) will be added to a 34 sum kept for that field. The special string 'hitcount' can be used 35 in place of an explicit value field - this is simply a count of 36 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 'stacktrace', which 39 will use the event's kernel stacktrace as the key. The keywords 40 'keys' or 'key' can be used to specify keys, and the keywords 41 'values', 'vals', or 'val' can be used to specify values. Compound 42 keys consisting of up to two fields can be specified by the 'keys' 43 keyword. Hashing a compound key produces a unique entry in the 44 table for each unique combination of component keys, and can be 45 useful for providing more fine-grained summaries of event data. 46 Additionally, sort keys consisting of up to two fields can be 47 specified by the 'sort' keyword. If more than one field is 48 specified, the result will be a 'sort within a sort': the first key 49 is taken to be the primary sort key and the second the secondary 50 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 same 52 name, and trigger hits will update this common data. Only triggers 53 with 'compatible' fields can be combined in this way; triggers are 54 'compatible' if the fields named in the trigger share the same 55 number and type of fields and those fields also have the same names. 56 Note that any two events always share the compatible 'hitcount' and 57 'stacktrace' fields and can therefore be combined using those 58 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 in 62 its entirety to stdout. If there are multiple hist triggers 63 attached to an event, there will be a table for each trigger in the 64 output. The table displayed for a named trigger will be the same as 65 any other instance having the same name. Each printed hash table 66 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 are 68 followed by the set of value fields for the entry. By default, 69 numeric fields are displayed as base-10 integers. This can be 70 modified by appending any of the following modifiers to the field 71 name: 72 73 =========== ========================================== 74 .hex display a number as a hex value 75 .sym display an address as a symbol 76 .sym-offset display an address as a symbol and offset 77 .syscall display a syscall id as a system call name 78 .execname display a common_pid as a program name 79 .log2 display log2 value rather than raw number 80 .usecs display a common_timestamp in microseconds 81 =========== ========================================== 82 83 Note that in general the semantics of a given field aren't 84 interpreted when applying a modifier to it, but there are some 85 restrictions to be aware of in this regard: 86 87 - only the 'hex' modifier can be used for values (because values 88 are essentially sums, and the other modifiers don't make sense 89 in that context). 90 - the 'execname' modifier can only be used on a 'common_pid'. The 91 reason for this is that the execname is simply the 'comm' value 92 saved for the 'current' process when an event was triggered, 93 which is the same as the common_pid value saved by the event 94 tracing code. Trying to apply that comm value to other pid 95 values wouldn't be correct, and typically events that care save 96 pid-specific comm fields in the event itself. 97 98 A typical usage scenario would be the following to enable a hist 99 trigger, read its current contents, and then turn it off:: 100 101 # echo 'hist:keys=skbaddr.hex:vals=len' > \ 102 /sys/kernel/debug/tracing/events/net/netif_rx/trigger 103 104 # cat /sys/kernel/debug/tracing/events/net/netif_rx/hist 105 106 # echo '!hist:keys=skbaddr.hex:vals=len' > \ 107 /sys/kernel/debug/tracing/events/net/netif_rx/trigger 108 109 The trigger file itself can be read to show the details of the 110 currently attached hist trigger. This information is also displayed 111 at the top of the 'hist' file when read. 112 113 By default, the size of the hash table is 2048 entries. The 'size' 114 parameter can be used to specify more or fewer than that. The units 115 are in terms of hashtable entries - if a run uses more entries than 116 specified, the results will show the number of 'drops', the number 117 of hits that were ignored. The size should be a power of 2 between 118 128 and 131072 (any non- power-of-2 number specified will be rounded 119 up). 120 121 The 'sort' parameter can be used to specify a value field to sort 122 on. The default if unspecified is 'hitcount' and the default sort 123 order is 'ascending'. To sort in the opposite direction, append 124 .descending' to the sort key. 125 126 The 'pause' parameter can be used to pause an existing hist trigger 127 or to start a hist trigger but not log any events until told to do 128 so. 'continue' or 'cont' can be used to start or restart a paused 129 hist trigger. 130 131 The 'clear' parameter will clear the contents of a running hist 132 trigger and leave its current paused/active state. 133 134 Note that the 'pause', 'cont', and 'clear' parameters should be 135 applied using 'append' shell operator ('>>') if applied to an 136 existing trigger, rather than via the '>' operator, which will cause 137 the trigger to be removed through truncation. 138 139- enable_hist/disable_hist 140 141 The enable_hist and disable_hist triggers can be used to have one 142 event conditionally start and stop another event's already-attached 143 hist trigger. Any number of enable_hist and disable_hist triggers 144 can be attached to a given event, allowing that event to kick off 145 and stop aggregations on a host of other events. 146 147 The format is very similar to the enable/disable_event triggers:: 148 149 enable_hist:<system>:<event>[:count] 150 disable_hist:<system>:<event>[:count] 151 152 Instead of enabling or disabling the tracing of the target event 153 into the trace buffer as the enable/disable_event triggers do, the 154 enable/disable_hist triggers enable or disable the aggregation of 155 the target event into a hash table. 156 157 A typical usage scenario for the enable_hist/disable_hist triggers 158 would be to first set up a paused hist trigger on some event, 159 followed by an enable_hist/disable_hist pair that turns the hist 160 aggregation on and off when conditions of interest are hit:: 161 162 # echo 'hist:keys=skbaddr.hex:vals=len:pause' > \ 163 /sys/kernel/debug/tracing/events/net/netif_receive_skb/trigger 164 165 # echo 'enable_hist:net:netif_receive_skb if filename==/usr/bin/wget' > \ 166 /sys/kernel/debug/tracing/events/sched/sched_process_exec/trigger 167 168 # echo 'disable_hist:net:netif_receive_skb if comm==wget' > \ 169 /sys/kernel/debug/tracing/events/sched/sched_process_exit/trigger 170 171 The above sets up an initially paused hist trigger which is unpaused 172 and starts aggregating events when a given program is executed, and 173 which stops aggregating when the process exits and the hist trigger 174 is paused again. 175 176 The examples below provide a more concrete illustration of the 177 concepts and typical usage patterns discussed above. 178 179'special' event fields 180------------------------ 181 182 There are a number of 'special event fields' available for use as 183 keys or values in a hist trigger. These look like and behave as if 184 they were actual event fields, but aren't really part of the event's 185 field definition or format file. They are however available for any 186 event, and can be used anywhere an actual event field could be. 187 They are: 188 189 ====================== ==== ======================================= 190 common_timestamp u64 timestamp (from ring buffer) associated 191 with the event, in nanoseconds. May be 192 modified by .usecs to have timestamps 193 interpreted as microseconds. 194 cpu int the cpu on which the event occurred. 195 ====================== ==== ======================================= 196 197Extended error information 198-------------------------- 199 200 For some error conditions encountered when invoking a hist trigger 201 command, extended error information is available via the 202 tracing/error_log file. See Error Conditions in 203 :file:`Documentation/trace/ftrace.rst` for details. 204 2056.2 'hist' trigger examples 206--------------------------- 207 208 The first set of examples creates aggregations using the kmalloc 209 event. The fields that can be used for the hist trigger are listed 210 in the kmalloc event's format file:: 211 212 # cat /sys/kernel/debug/tracing/events/kmem/kmalloc/format 213 name: kmalloc 214 ID: 374 215 format: 216 field:unsigned short common_type; offset:0; size:2; signed:0; 217 field:unsigned char common_flags; offset:2; size:1; signed:0; 218 field:unsigned char common_preempt_count; offset:3; size:1; signed:0; 219 field:int common_pid; offset:4; size:4; signed:1; 220 221 field:unsigned long call_site; offset:8; size:8; signed:0; 222 field:const void * ptr; offset:16; size:8; signed:0; 223 field:size_t bytes_req; offset:24; size:8; signed:0; 224 field:size_t bytes_alloc; offset:32; size:8; signed:0; 225 field:gfp_t gfp_flags; offset:40; size:4; signed:0; 226 227 We'll start by creating a hist trigger that generates a simple table 228 that lists the total number of bytes requested for each function in 229 the kernel that made one or more calls to kmalloc:: 230 231 # echo 'hist:key=call_site:val=bytes_req' > \ 232 /sys/kernel/debug/tracing/events/kmem/kmalloc/trigger 233 234 This tells the tracing system to create a 'hist' trigger using the 235 call_site field of the kmalloc event as the key for the table, which 236 just means that each unique call_site address will have an entry 237 created for it in the table. The 'val=bytes_req' parameter tells 238 the hist trigger that for each unique entry (call_site) in the 239 table, it should keep a running total of the number of bytes 240 requested by that call_site. 241 242 We'll let it run for awhile and then dump the contents of the 'hist' 243 file in the kmalloc event's subdirectory (for readability, a number 244 of entries have been omitted):: 245 246 # cat /sys/kernel/debug/tracing/events/kmem/kmalloc/hist 247 # trigger info: hist:keys=call_site:vals=bytes_req:sort=hitcount:size=2048 [active] 248 249 { call_site: 18446744072106379007 } hitcount: 1 bytes_req: 176 250 { call_site: 18446744071579557049 } hitcount: 1 bytes_req: 1024 251 { call_site: 18446744071580608289 } hitcount: 1 bytes_req: 16384 252 { call_site: 18446744071581827654 } hitcount: 1 bytes_req: 24 253 { call_site: 18446744071580700980 } hitcount: 1 bytes_req: 8 254 { call_site: 18446744071579359876 } hitcount: 1 bytes_req: 152 255 { call_site: 18446744071580795365 } hitcount: 3 bytes_req: 144 256 { call_site: 18446744071581303129 } hitcount: 3 bytes_req: 144 257 { call_site: 18446744071580713234 } hitcount: 4 bytes_req: 2560 258 { call_site: 18446744071580933750 } hitcount: 4 bytes_req: 736 259 . 260 . 261 . 262 { call_site: 18446744072106047046 } hitcount: 69 bytes_req: 5576 263 { call_site: 18446744071582116407 } hitcount: 73 bytes_req: 2336 264 { call_site: 18446744072106054684 } hitcount: 136 bytes_req: 140504 265 { call_site: 18446744072106224230 } hitcount: 136 bytes_req: 19584 266 { call_site: 18446744072106078074 } hitcount: 153 bytes_req: 2448 267 { call_site: 18446744072106062406 } hitcount: 153 bytes_req: 36720 268 { call_site: 18446744071582507929 } hitcount: 153 bytes_req: 37088 269 { call_site: 18446744072102520590 } hitcount: 273 bytes_req: 10920 270 { call_site: 18446744071582143559 } hitcount: 358 bytes_req: 716 271 { call_site: 18446744072106465852 } hitcount: 417 bytes_req: 56712 272 { call_site: 18446744072102523378 } hitcount: 485 bytes_req: 27160 273 { call_site: 18446744072099568646 } hitcount: 1676 bytes_req: 33520 274 275 Totals: 276 Hits: 4610 277 Entries: 45 278 Dropped: 0 279 280 The output displays a line for each entry, beginning with the key 281 specified in the trigger, followed by the value(s) also specified in 282 the trigger. At the beginning of the output is a line that displays 283 the trigger info, which can also be displayed by reading the 284 'trigger' file:: 285 286 # cat /sys/kernel/debug/tracing/events/kmem/kmalloc/trigger 287 hist:keys=call_site:vals=bytes_req:sort=hitcount:size=2048 [active] 288 289 At the end of the output are a few lines that display the overall 290 totals for the run. The 'Hits' field shows the total number of 291 times the event trigger was hit, the 'Entries' field shows the total 292 number of used entries in the hash table, and the 'Dropped' field 293 shows the number of hits that were dropped because the number of 294 used entries for the run exceeded the maximum number of entries 295 allowed for the table (normally 0, but if not a hint that you may 296 want to increase the size of the table using the 'size' parameter). 297 298 Notice in the above output that there's an extra field, 'hitcount', 299 which wasn't specified in the trigger. Also notice that in the 300 trigger info output, there's a parameter, 'sort=hitcount', which 301 wasn't specified in the trigger either. The reason for that is that 302 every trigger implicitly keeps a count of the total number of hits 303 attributed to a given entry, called the 'hitcount'. That hitcount 304 information is explicitly displayed in the output, and in the 305 absence of a user-specified sort parameter, is used as the default 306 sort field. 307 308 The value 'hitcount' can be used in place of an explicit value in 309 the 'values' parameter if you don't really need to have any 310 particular field summed and are mainly interested in hit 311 frequencies. 312 313 To turn the hist trigger off, simply call up the trigger in the 314 command history and re-execute it with a '!' prepended:: 315 316 # echo '!hist:key=call_site:val=bytes_req' > \ 317 /sys/kernel/debug/tracing/events/kmem/kmalloc/trigger 318 319 Finally, notice that the call_site as displayed in the output above 320 isn't really very useful. It's an address, but normally addresses 321 are displayed in hex. To have a numeric field displayed as a hex 322 value, simply append '.hex' to the field name in the trigger:: 323 324 # echo 'hist:key=call_site.hex:val=bytes_req' > \ 325 /sys/kernel/debug/tracing/events/kmem/kmalloc/trigger 326 327 # cat /sys/kernel/debug/tracing/events/kmem/kmalloc/hist 328 # trigger info: hist:keys=call_site.hex:vals=bytes_req:sort=hitcount:size=2048 [active] 329 330 { call_site: ffffffffa026b291 } hitcount: 1 bytes_req: 433 331 { call_site: ffffffffa07186ff } hitcount: 1 bytes_req: 176 332 { call_site: ffffffff811ae721 } hitcount: 1 bytes_req: 16384 333 { call_site: ffffffff811c5134 } hitcount: 1 bytes_req: 8 334 { call_site: ffffffffa04a9ebb } hitcount: 1 bytes_req: 511 335 { call_site: ffffffff8122e0a6 } hitcount: 1 bytes_req: 12 336 { call_site: ffffffff8107da84 } hitcount: 1 bytes_req: 152 337 { call_site: ffffffff812d8246 } hitcount: 1 bytes_req: 24 338 { call_site: ffffffff811dc1e5 } hitcount: 3 bytes_req: 144 339 { call_site: ffffffffa02515e8 } hitcount: 3 bytes_req: 648 340 { call_site: ffffffff81258159 } hitcount: 3 bytes_req: 144 341 { call_site: ffffffff811c80f4 } hitcount: 4 bytes_req: 544 342 . 343 . 344 . 345 { call_site: ffffffffa06c7646 } hitcount: 106 bytes_req: 8024 346 { call_site: ffffffffa06cb246 } hitcount: 132 bytes_req: 31680 347 { call_site: ffffffffa06cef7a } hitcount: 132 bytes_req: 2112 348 { call_site: ffffffff8137e399 } hitcount: 132 bytes_req: 23232 349 { call_site: ffffffffa06c941c } hitcount: 185 bytes_req: 171360 350 { call_site: ffffffffa06f2a66 } hitcount: 185 bytes_req: 26640 351 { call_site: ffffffffa036a70e } hitcount: 265 bytes_req: 10600 352 { call_site: ffffffff81325447 } hitcount: 292 bytes_req: 584 353 { call_site: ffffffffa072da3c } hitcount: 446 bytes_req: 60656 354 { call_site: ffffffffa036b1f2 } hitcount: 526 bytes_req: 29456 355 { call_site: ffffffffa0099c06 } hitcount: 1780 bytes_req: 35600 356 357 Totals: 358 Hits: 4775 359 Entries: 46 360 Dropped: 0 361 362 Even that's only marginally more useful - while hex values do look 363 more like addresses, what users are typically more interested in 364 when looking at text addresses are the corresponding symbols 365 instead. To have an address displayed as symbolic value instead, 366 simply append '.sym' or '.sym-offset' to the field name in the 367 trigger:: 368 369 # echo 'hist:key=call_site.sym:val=bytes_req' > \ 370 /sys/kernel/debug/tracing/events/kmem/kmalloc/trigger 371 372 # cat /sys/kernel/debug/tracing/events/kmem/kmalloc/hist 373 # trigger info: hist:keys=call_site.sym:vals=bytes_req:sort=hitcount:size=2048 [active] 374 375 { call_site: [ffffffff810adcb9] syslog_print_all } hitcount: 1 bytes_req: 1024 376 { call_site: [ffffffff8154bc62] usb_control_msg } hitcount: 1 bytes_req: 8 377 { call_site: [ffffffffa00bf6fe] hidraw_send_report [hid] } hitcount: 1 bytes_req: 7 378 { call_site: [ffffffff8154acbe] usb_alloc_urb } hitcount: 1 bytes_req: 192 379 { call_site: [ffffffffa00bf1ca] hidraw_report_event [hid] } hitcount: 1 bytes_req: 7 380 { call_site: [ffffffff811e3a25] __seq_open_private } hitcount: 1 bytes_req: 40 381 { call_site: [ffffffff8109524a] alloc_fair_sched_group } hitcount: 2 bytes_req: 128 382 { call_site: [ffffffff811febd5] fsnotify_alloc_group } hitcount: 2 bytes_req: 528 383 { call_site: [ffffffff81440f58] __tty_buffer_request_room } hitcount: 2 bytes_req: 2624 384 { call_site: [ffffffff81200ba6] inotify_new_group } hitcount: 2 bytes_req: 96 385 { call_site: [ffffffffa05e19af] ieee80211_start_tx_ba_session [mac80211] } hitcount: 2 bytes_req: 464 386 { call_site: [ffffffff81672406] tcp_get_metrics } hitcount: 2 bytes_req: 304 387 { call_site: [ffffffff81097ec2] alloc_rt_sched_group } hitcount: 2 bytes_req: 128 388 { call_site: [ffffffff81089b05] sched_create_group } hitcount: 2 bytes_req: 1424 389 . 390 . 391 . 392 { call_site: [ffffffffa04a580c] intel_crtc_page_flip [i915] } hitcount: 1185 bytes_req: 123240 393 { call_site: [ffffffffa0287592] drm_mode_page_flip_ioctl [drm] } hitcount: 1185 bytes_req: 104280 394 { call_site: [ffffffffa04c4a3c] intel_plane_duplicate_state [i915] } hitcount: 1402 bytes_req: 190672 395 { call_site: [ffffffff812891ca] ext4_find_extent } hitcount: 1518 bytes_req: 146208 396 { call_site: [ffffffffa029070e] drm_vma_node_allow [drm] } hitcount: 1746 bytes_req: 69840 397 { call_site: [ffffffffa045e7c4] i915_gem_do_execbuffer.isra.23 [i915] } hitcount: 2021 bytes_req: 792312 398 { call_site: [ffffffffa02911f2] drm_modeset_lock_crtc [drm] } hitcount: 2592 bytes_req: 145152 399 { call_site: [ffffffffa0489a66] intel_ring_begin [i915] } hitcount: 2629 bytes_req: 378576 400 { call_site: [ffffffffa046041c] i915_gem_execbuffer2 [i915] } hitcount: 2629 bytes_req: 3783248 401 { call_site: [ffffffff81325607] apparmor_file_alloc_security } hitcount: 5192 bytes_req: 10384 402 { call_site: [ffffffffa00b7c06] hid_report_raw_event [hid] } hitcount: 5529 bytes_req: 110584 403 { call_site: [ffffffff8131ebf7] aa_alloc_task_context } hitcount: 21943 bytes_req: 702176 404 { call_site: [ffffffff8125847d] ext4_htree_store_dirent } hitcount: 55759 bytes_req: 5074265 405 406 Totals: 407 Hits: 109928 408 Entries: 71 409 Dropped: 0 410 411 Because the default sort key above is 'hitcount', the above shows a 412 the list of call_sites by increasing hitcount, so that at the bottom 413 we see the functions that made the most kmalloc calls during the 414 run. If instead we we wanted to see the top kmalloc callers in 415 terms of the number of bytes requested rather than the number of 416 calls, and we wanted the top caller to appear at the top, we can use 417 the 'sort' parameter, along with the 'descending' modifier:: 418 419 # echo 'hist:key=call_site.sym:val=bytes_req:sort=bytes_req.descending' > \ 420 /sys/kernel/debug/tracing/events/kmem/kmalloc/trigger 421 422 # cat /sys/kernel/debug/tracing/events/kmem/kmalloc/hist 423 # trigger info: hist:keys=call_site.sym:vals=bytes_req:sort=bytes_req.descending:size=2048 [active] 424 425 { call_site: [ffffffffa046041c] i915_gem_execbuffer2 [i915] } hitcount: 2186 bytes_req: 3397464 426 { call_site: [ffffffffa045e7c4] i915_gem_do_execbuffer.isra.23 [i915] } hitcount: 1790 bytes_req: 712176 427 { call_site: [ffffffff8125847d] ext4_htree_store_dirent } hitcount: 8132 bytes_req: 513135 428 { call_site: [ffffffff811e2a1b] seq_buf_alloc } hitcount: 106 bytes_req: 440128 429 { call_site: [ffffffffa0489a66] intel_ring_begin [i915] } hitcount: 2186 bytes_req: 314784 430 { call_site: [ffffffff812891ca] ext4_find_extent } hitcount: 2174 bytes_req: 208992 431 { call_site: [ffffffff811ae8e1] __kmalloc } hitcount: 8 bytes_req: 131072 432 { call_site: [ffffffffa04c4a3c] intel_plane_duplicate_state [i915] } hitcount: 859 bytes_req: 116824 433 { call_site: [ffffffffa02911f2] drm_modeset_lock_crtc [drm] } hitcount: 1834 bytes_req: 102704 434 { call_site: [ffffffffa04a580c] intel_crtc_page_flip [i915] } hitcount: 972 bytes_req: 101088 435 { call_site: [ffffffffa0287592] drm_mode_page_flip_ioctl [drm] } hitcount: 972 bytes_req: 85536 436 { call_site: [ffffffffa00b7c06] hid_report_raw_event [hid] } hitcount: 3333 bytes_req: 66664 437 { call_site: [ffffffff8137e559] sg_kmalloc } hitcount: 209 bytes_req: 61632 438 . 439 . 440 . 441 { call_site: [ffffffff81095225] alloc_fair_sched_group } hitcount: 2 bytes_req: 128 442 { call_site: [ffffffff81097ec2] alloc_rt_sched_group } hitcount: 2 bytes_req: 128 443 { call_site: [ffffffff812d8406] copy_semundo } hitcount: 2 bytes_req: 48 444 { call_site: [ffffffff81200ba6] inotify_new_group } hitcount: 1 bytes_req: 48 445 { call_site: [ffffffffa027121a] drm_getmagic [drm] } hitcount: 1 bytes_req: 48 446 { call_site: [ffffffff811e3a25] __seq_open_private } hitcount: 1 bytes_req: 40 447 { call_site: [ffffffff811c52f4] bprm_change_interp } hitcount: 2 bytes_req: 16 448 { call_site: [ffffffff8154bc62] usb_control_msg } hitcount: 1 bytes_req: 8 449 { call_site: [ffffffffa00bf1ca] hidraw_report_event [hid] } hitcount: 1 bytes_req: 7 450 { call_site: [ffffffffa00bf6fe] hidraw_send_report [hid] } hitcount: 1 bytes_req: 7 451 452 Totals: 453 Hits: 32133 454 Entries: 81 455 Dropped: 0 456 457 To display the offset and size information in addition to the symbol 458 name, just use 'sym-offset' instead:: 459 460 # echo 'hist:key=call_site.sym-offset:val=bytes_req:sort=bytes_req.descending' > \ 461 /sys/kernel/debug/tracing/events/kmem/kmalloc/trigger 462 463 # cat /sys/kernel/debug/tracing/events/kmem/kmalloc/hist 464 # trigger info: hist:keys=call_site.sym-offset:vals=bytes_req:sort=bytes_req.descending:size=2048 [active] 465 466 { call_site: [ffffffffa046041c] i915_gem_execbuffer2+0x6c/0x2c0 [i915] } hitcount: 4569 bytes_req: 3163720 467 { call_site: [ffffffffa0489a66] intel_ring_begin+0xc6/0x1f0 [i915] } hitcount: 4569 bytes_req: 657936 468 { call_site: [ffffffffa045e7c4] i915_gem_do_execbuffer.isra.23+0x694/0x1020 [i915] } hitcount: 1519 bytes_req: 472936 469 { call_site: [ffffffffa045e646] i915_gem_do_execbuffer.isra.23+0x516/0x1020 [i915] } hitcount: 3050 bytes_req: 211832 470 { call_site: [ffffffff811e2a1b] seq_buf_alloc+0x1b/0x50 } hitcount: 34 bytes_req: 148384 471 { call_site: [ffffffffa04a580c] intel_crtc_page_flip+0xbc/0x870 [i915] } hitcount: 1385 bytes_req: 144040 472 { call_site: [ffffffff811ae8e1] __kmalloc+0x191/0x1b0 } hitcount: 8 bytes_req: 131072 473 { call_site: [ffffffffa0287592] drm_mode_page_flip_ioctl+0x282/0x360 [drm] } hitcount: 1385 bytes_req: 121880 474 { call_site: [ffffffffa02911f2] drm_modeset_lock_crtc+0x32/0x100 [drm] } hitcount: 1848 bytes_req: 103488 475 { call_site: [ffffffffa04c4a3c] intel_plane_duplicate_state+0x2c/0xa0 [i915] } hitcount: 461 bytes_req: 62696 476 { call_site: [ffffffffa029070e] drm_vma_node_allow+0x2e/0xd0 [drm] } hitcount: 1541 bytes_req: 61640 477 { call_site: [ffffffff815f8d7b] sk_prot_alloc+0xcb/0x1b0 } hitcount: 57 bytes_req: 57456 478 . 479 . 480 . 481 { call_site: [ffffffff8109524a] alloc_fair_sched_group+0x5a/0x1a0 } hitcount: 2 bytes_req: 128 482 { call_site: [ffffffffa027b921] drm_vm_open_locked+0x31/0xa0 [drm] } hitcount: 3 bytes_req: 96 483 { call_site: [ffffffff8122e266] proc_self_follow_link+0x76/0xb0 } hitcount: 8 bytes_req: 96 484 { call_site: [ffffffff81213e80] load_elf_binary+0x240/0x1650 } hitcount: 3 bytes_req: 84 485 { call_site: [ffffffff8154bc62] usb_control_msg+0x42/0x110 } hitcount: 1 bytes_req: 8 486 { call_site: [ffffffffa00bf6fe] hidraw_send_report+0x7e/0x1a0 [hid] } hitcount: 1 bytes_req: 7 487 { call_site: [ffffffffa00bf1ca] hidraw_report_event+0x8a/0x120 [hid] } hitcount: 1 bytes_req: 7 488 489 Totals: 490 Hits: 26098 491 Entries: 64 492 Dropped: 0 493 494 We can also add multiple fields to the 'values' parameter. For 495 example, we might want to see the total number of bytes allocated 496 alongside bytes requested, and display the result sorted by bytes 497 allocated in a descending order:: 498 499 # echo 'hist:keys=call_site.sym:values=bytes_req,bytes_alloc:sort=bytes_alloc.descending' > \ 500 /sys/kernel/debug/tracing/events/kmem/kmalloc/trigger 501 502 # cat /sys/kernel/debug/tracing/events/kmem/kmalloc/hist 503 # trigger info: hist:keys=call_site.sym:vals=bytes_req,bytes_alloc:sort=bytes_alloc.descending:size=2048 [active] 504 505 { call_site: [ffffffffa046041c] i915_gem_execbuffer2 [i915] } hitcount: 7403 bytes_req: 4084360 bytes_alloc: 5958016 506 { call_site: [ffffffff811e2a1b] seq_buf_alloc } hitcount: 541 bytes_req: 2213968 bytes_alloc: 2228224 507 { call_site: [ffffffffa0489a66] intel_ring_begin [i915] } hitcount: 7404 bytes_req: 1066176 bytes_alloc: 1421568 508 { call_site: [ffffffffa045e7c4] i915_gem_do_execbuffer.isra.23 [i915] } hitcount: 1565 bytes_req: 557368 bytes_alloc: 1037760 509 { call_site: [ffffffff8125847d] ext4_htree_store_dirent } hitcount: 9557 bytes_req: 595778 bytes_alloc: 695744 510 { call_site: [ffffffffa045e646] i915_gem_do_execbuffer.isra.23 [i915] } hitcount: 5839 bytes_req: 430680 bytes_alloc: 470400 511 { call_site: [ffffffffa04c4a3c] intel_plane_duplicate_state [i915] } hitcount: 2388 bytes_req: 324768 bytes_alloc: 458496 512 { call_site: [ffffffffa02911f2] drm_modeset_lock_crtc [drm] } hitcount: 3911 bytes_req: 219016 bytes_alloc: 250304 513 { call_site: [ffffffff815f8d7b] sk_prot_alloc } hitcount: 235 bytes_req: 236880 bytes_alloc: 240640 514 { call_site: [ffffffff8137e559] sg_kmalloc } hitcount: 557 bytes_req: 169024 bytes_alloc: 221760 515 { call_site: [ffffffffa00b7c06] hid_report_raw_event [hid] } hitcount: 9378 bytes_req: 187548 bytes_alloc: 206312 516 { call_site: [ffffffffa04a580c] intel_crtc_page_flip [i915] } hitcount: 1519 bytes_req: 157976 bytes_alloc: 194432 517 . 518 . 519 . 520 { call_site: [ffffffff8109bd3b] sched_autogroup_create_attach } hitcount: 2 bytes_req: 144 bytes_alloc: 192 521 { call_site: [ffffffff81097ee8] alloc_rt_sched_group } hitcount: 2 bytes_req: 128 bytes_alloc: 128 522 { call_site: [ffffffff8109524a] alloc_fair_sched_group } hitcount: 2 bytes_req: 128 bytes_alloc: 128 523 { call_site: [ffffffff81095225] alloc_fair_sched_group } hitcount: 2 bytes_req: 128 bytes_alloc: 128 524 { call_site: [ffffffff81097ec2] alloc_rt_sched_group } hitcount: 2 bytes_req: 128 bytes_alloc: 128 525 { call_site: [ffffffff81213e80] load_elf_binary } hitcount: 3 bytes_req: 84 bytes_alloc: 96 526 { call_site: [ffffffff81079a2e] kthread_create_on_node } hitcount: 1 bytes_req: 56 bytes_alloc: 64 527 { call_site: [ffffffffa00bf6fe] hidraw_send_report [hid] } hitcount: 1 bytes_req: 7 bytes_alloc: 8 528 { call_site: [ffffffff8154bc62] usb_control_msg } hitcount: 1 bytes_req: 8 bytes_alloc: 8 529 { call_site: [ffffffffa00bf1ca] hidraw_report_event [hid] } hitcount: 1 bytes_req: 7 bytes_alloc: 8 530 531 Totals: 532 Hits: 66598 533 Entries: 65 534 Dropped: 0 535 536 Finally, to finish off our kmalloc example, instead of simply having 537 the hist trigger display symbolic call_sites, we can have the hist 538 trigger additionally display the complete set of kernel stack traces 539 that led to each call_site. To do that, we simply use the special 540 value 'stacktrace' for the key parameter:: 541 542 # echo 'hist:keys=stacktrace:values=bytes_req,bytes_alloc:sort=bytes_alloc' > \ 543 /sys/kernel/debug/tracing/events/kmem/kmalloc/trigger 544 545 The above trigger will use the kernel stack trace in effect when an 546 event is triggered as the key for the hash table. This allows the 547 enumeration of every kernel callpath that led up to a particular 548 event, along with a running total of any of the event fields for 549 that event. Here we tally bytes requested and bytes allocated for 550 every callpath in the system that led up to a kmalloc (in this case 551 every callpath to a kmalloc for a kernel compile):: 552 553 # cat /sys/kernel/debug/tracing/events/kmem/kmalloc/hist 554 # trigger info: hist:keys=stacktrace:vals=bytes_req,bytes_alloc:sort=bytes_alloc:size=2048 [active] 555 556 { stacktrace: 557 __kmalloc_track_caller+0x10b/0x1a0 558 kmemdup+0x20/0x50 559 hidraw_report_event+0x8a/0x120 [hid] 560 hid_report_raw_event+0x3ea/0x440 [hid] 561 hid_input_report+0x112/0x190 [hid] 562 hid_irq_in+0xc2/0x260 [usbhid] 563 __usb_hcd_giveback_urb+0x72/0x120 564 usb_giveback_urb_bh+0x9e/0xe0 565 tasklet_hi_action+0xf8/0x100 566 __do_softirq+0x114/0x2c0 567 irq_exit+0xa5/0xb0 568 do_IRQ+0x5a/0xf0 569 ret_from_intr+0x0/0x30 570 cpuidle_enter+0x17/0x20 571 cpu_startup_entry+0x315/0x3e0 572 rest_init+0x7c/0x80 573 } hitcount: 3 bytes_req: 21 bytes_alloc: 24 574 { stacktrace: 575 __kmalloc_track_caller+0x10b/0x1a0 576 kmemdup+0x20/0x50 577 hidraw_report_event+0x8a/0x120 [hid] 578 hid_report_raw_event+0x3ea/0x440 [hid] 579 hid_input_report+0x112/0x190 [hid] 580 hid_irq_in+0xc2/0x260 [usbhid] 581 __usb_hcd_giveback_urb+0x72/0x120 582 usb_giveback_urb_bh+0x9e/0xe0 583 tasklet_hi_action+0xf8/0x100 584 __do_softirq+0x114/0x2c0 585 irq_exit+0xa5/0xb0 586 do_IRQ+0x5a/0xf0 587 ret_from_intr+0x0/0x30 588 } hitcount: 3 bytes_req: 21 bytes_alloc: 24 589 { stacktrace: 590 kmem_cache_alloc_trace+0xeb/0x150 591 aa_alloc_task_context+0x27/0x40 592 apparmor_cred_prepare+0x1f/0x50 593 security_prepare_creds+0x16/0x20 594 prepare_creds+0xdf/0x1a0 595 SyS_capset+0xb5/0x200 596 system_call_fastpath+0x12/0x6a 597 } hitcount: 1 bytes_req: 32 bytes_alloc: 32 598 . 599 . 600 . 601 { stacktrace: 602 __kmalloc+0x11b/0x1b0 603 i915_gem_execbuffer2+0x6c/0x2c0 [i915] 604 drm_ioctl+0x349/0x670 [drm] 605 do_vfs_ioctl+0x2f0/0x4f0 606 SyS_ioctl+0x81/0xa0 607 system_call_fastpath+0x12/0x6a 608 } hitcount: 17726 bytes_req: 13944120 bytes_alloc: 19593808 609 { stacktrace: 610 __kmalloc+0x11b/0x1b0 611 load_elf_phdrs+0x76/0xa0 612 load_elf_binary+0x102/0x1650 613 search_binary_handler+0x97/0x1d0 614 do_execveat_common.isra.34+0x551/0x6e0 615 SyS_execve+0x3a/0x50 616 return_from_execve+0x0/0x23 617 } hitcount: 33348 bytes_req: 17152128 bytes_alloc: 20226048 618 { stacktrace: 619 kmem_cache_alloc_trace+0xeb/0x150 620 apparmor_file_alloc_security+0x27/0x40 621 security_file_alloc+0x16/0x20 622 get_empty_filp+0x93/0x1c0 623 path_openat+0x31/0x5f0 624 do_filp_open+0x3a/0x90 625 do_sys_open+0x128/0x220 626 SyS_open+0x1e/0x20 627 system_call_fastpath+0x12/0x6a 628 } hitcount: 4766422 bytes_req: 9532844 bytes_alloc: 38131376 629 { stacktrace: 630 __kmalloc+0x11b/0x1b0 631 seq_buf_alloc+0x1b/0x50 632 seq_read+0x2cc/0x370 633 proc_reg_read+0x3d/0x80 634 __vfs_read+0x28/0xe0 635 vfs_read+0x86/0x140 636 SyS_read+0x46/0xb0 637 system_call_fastpath+0x12/0x6a 638 } hitcount: 19133 bytes_req: 78368768 bytes_alloc: 78368768 639 640 Totals: 641 Hits: 6085872 642 Entries: 253 643 Dropped: 0 644 645 If you key a hist trigger on common_pid, in order for example to 646 gather and display sorted totals for each process, you can use the 647 special .execname modifier to display the executable names for the 648 processes in the table rather than raw pids. The example below 649 keeps a per-process sum of total bytes read:: 650 651 # echo 'hist:key=common_pid.execname:val=count:sort=count.descending' > \ 652 /sys/kernel/debug/tracing/events/syscalls/sys_enter_read/trigger 653 654 # cat /sys/kernel/debug/tracing/events/syscalls/sys_enter_read/hist 655 # trigger info: hist:keys=common_pid.execname:vals=count:sort=count.descending:size=2048 [active] 656 657 { common_pid: gnome-terminal [ 3196] } hitcount: 280 count: 1093512 658 { common_pid: Xorg [ 1309] } hitcount: 525 count: 256640 659 { common_pid: compiz [ 2889] } hitcount: 59 count: 254400 660 { common_pid: bash [ 8710] } hitcount: 3 count: 66369 661 { common_pid: dbus-daemon-lau [ 8703] } hitcount: 49 count: 47739 662 { common_pid: irqbalance [ 1252] } hitcount: 27 count: 27648 663 { common_pid: 01ifupdown [ 8705] } hitcount: 3 count: 17216 664 { common_pid: dbus-daemon [ 772] } hitcount: 10 count: 12396 665 { common_pid: Socket Thread [ 8342] } hitcount: 11 count: 11264 666 { common_pid: nm-dhcp-client. [ 8701] } hitcount: 6 count: 7424 667 { common_pid: gmain [ 1315] } hitcount: 18 count: 6336 668 . 669 . 670 . 671 { common_pid: postgres [ 1892] } hitcount: 2 count: 32 672 { common_pid: postgres [ 1891] } hitcount: 2 count: 32 673 { common_pid: gmain [ 8704] } hitcount: 2 count: 32 674 { common_pid: upstart-dbus-br [ 2740] } hitcount: 21 count: 21 675 { common_pid: nm-dispatcher.a [ 8696] } hitcount: 1 count: 16 676 { common_pid: indicator-datet [ 2904] } hitcount: 1 count: 16 677 { common_pid: gdbus [ 2998] } hitcount: 1 count: 16 678 { common_pid: rtkit-daemon [ 2052] } hitcount: 1 count: 8 679 { common_pid: init [ 1] } hitcount: 2 count: 2 680 681 Totals: 682 Hits: 2116 683 Entries: 51 684 Dropped: 0 685 686 Similarly, if you key a hist trigger on syscall id, for example to 687 gather and display a list of systemwide syscall hits, you can use 688 the special .syscall modifier to display the syscall names rather 689 than raw ids. The example below keeps a running total of syscall 690 counts for the system during the run:: 691 692 # echo 'hist:key=id.syscall:val=hitcount' > \ 693 /sys/kernel/debug/tracing/events/raw_syscalls/sys_enter/trigger 694 695 # cat /sys/kernel/debug/tracing/events/raw_syscalls/sys_enter/hist 696 # trigger info: hist:keys=id.syscall:vals=hitcount:sort=hitcount:size=2048 [active] 697 698 { id: sys_fsync [ 74] } hitcount: 1 699 { id: sys_newuname [ 63] } hitcount: 1 700 { id: sys_prctl [157] } hitcount: 1 701 { id: sys_statfs [137] } hitcount: 1 702 { id: sys_symlink [ 88] } hitcount: 1 703 { id: sys_sendmmsg [307] } hitcount: 1 704 { id: sys_semctl [ 66] } hitcount: 1 705 { id: sys_readlink [ 89] } hitcount: 3 706 { id: sys_bind [ 49] } hitcount: 3 707 { id: sys_getsockname [ 51] } hitcount: 3 708 { id: sys_unlink [ 87] } hitcount: 3 709 { id: sys_rename [ 82] } hitcount: 4 710 { id: unknown_syscall [ 58] } hitcount: 4 711 { id: sys_connect [ 42] } hitcount: 4 712 { id: sys_getpid [ 39] } hitcount: 4 713 . 714 . 715 . 716 { id: sys_rt_sigprocmask [ 14] } hitcount: 952 717 { id: sys_futex [202] } hitcount: 1534 718 { id: sys_write [ 1] } hitcount: 2689 719 { id: sys_setitimer [ 38] } hitcount: 2797 720 { id: sys_read [ 0] } hitcount: 3202 721 { id: sys_select [ 23] } hitcount: 3773 722 { id: sys_writev [ 20] } hitcount: 4531 723 { id: sys_poll [ 7] } hitcount: 8314 724 { id: sys_recvmsg [ 47] } hitcount: 13738 725 { id: sys_ioctl [ 16] } hitcount: 21843 726 727 Totals: 728 Hits: 67612 729 Entries: 72 730 Dropped: 0 731 732 The syscall counts above provide a rough overall picture of system 733 call activity on the system; we can see for example that the most 734 popular system call on this system was the 'sys_ioctl' system call. 735 736 We can use 'compound' keys to refine that number and provide some 737 further insight as to which processes exactly contribute to the 738 overall ioctl count. 739 740 The command below keeps a hitcount for every unique combination of 741 system call id and pid - the end result is essentially a table 742 that keeps a per-pid sum of system call hits. The results are 743 sorted using the system call id as the primary key, and the 744 hitcount sum as the secondary key:: 745 746 # echo 'hist:key=id.syscall,common_pid.execname:val=hitcount:sort=id,hitcount' > \ 747 /sys/kernel/debug/tracing/events/raw_syscalls/sys_enter/trigger 748 749 # cat /sys/kernel/debug/tracing/events/raw_syscalls/sys_enter/hist 750 # trigger info: hist:keys=id.syscall,common_pid.execname:vals=hitcount:sort=id.syscall,hitcount:size=2048 [active] 751 752 { id: sys_read [ 0], common_pid: rtkit-daemon [ 1877] } hitcount: 1 753 { id: sys_read [ 0], common_pid: gdbus [ 2976] } hitcount: 1 754 { id: sys_read [ 0], common_pid: console-kit-dae [ 3400] } hitcount: 1 755 { id: sys_read [ 0], common_pid: postgres [ 1865] } hitcount: 1 756 { id: sys_read [ 0], common_pid: deja-dup-monito [ 3543] } hitcount: 2 757 { id: sys_read [ 0], common_pid: NetworkManager [ 890] } hitcount: 2 758 { id: sys_read [ 0], common_pid: evolution-calen [ 3048] } hitcount: 2 759 { id: sys_read [ 0], common_pid: postgres [ 1864] } hitcount: 2 760 { id: sys_read [ 0], common_pid: nm-applet [ 3022] } hitcount: 2 761 { id: sys_read [ 0], common_pid: whoopsie [ 1212] } hitcount: 2 762 . 763 . 764 . 765 { id: sys_ioctl [ 16], common_pid: bash [ 8479] } hitcount: 1 766 { id: sys_ioctl [ 16], common_pid: bash [ 3472] } hitcount: 12 767 { id: sys_ioctl [ 16], common_pid: gnome-terminal [ 3199] } hitcount: 16 768 { id: sys_ioctl [ 16], common_pid: Xorg [ 1267] } hitcount: 1808 769 { id: sys_ioctl [ 16], common_pid: compiz [ 2994] } hitcount: 5580 770 . 771 . 772 . 773 { id: sys_waitid [247], common_pid: upstart-dbus-br [ 2690] } hitcount: 3 774 { id: sys_waitid [247], common_pid: upstart-dbus-br [ 2688] } hitcount: 16 775 { id: sys_inotify_add_watch [254], common_pid: gmain [ 975] } hitcount: 2 776 { id: sys_inotify_add_watch [254], common_pid: gmain [ 3204] } hitcount: 4 777 { id: sys_inotify_add_watch [254], common_pid: gmain [ 2888] } hitcount: 4 778 { id: sys_inotify_add_watch [254], common_pid: gmain [ 3003] } hitcount: 4 779 { id: sys_inotify_add_watch [254], common_pid: gmain [ 2873] } hitcount: 4 780 { id: sys_inotify_add_watch [254], common_pid: gmain [ 3196] } hitcount: 6 781 { id: sys_openat [257], common_pid: java [ 2623] } hitcount: 2 782 { id: sys_eventfd2 [290], common_pid: ibus-ui-gtk3 [ 2760] } hitcount: 4 783 { id: sys_eventfd2 [290], common_pid: compiz [ 2994] } hitcount: 6 784 785 Totals: 786 Hits: 31536 787 Entries: 323 788 Dropped: 0 789 790 The above list does give us a breakdown of the ioctl syscall by 791 pid, but it also gives us quite a bit more than that, which we 792 don't really care about at the moment. Since we know the syscall 793 id for sys_ioctl (16, displayed next to the sys_ioctl name), we 794 can use that to filter out all the other syscalls:: 795 796 # echo 'hist:key=id.syscall,common_pid.execname:val=hitcount:sort=id,hitcount if id == 16' > \ 797 /sys/kernel/debug/tracing/events/raw_syscalls/sys_enter/trigger 798 799 # cat /sys/kernel/debug/tracing/events/raw_syscalls/sys_enter/hist 800 # trigger info: hist:keys=id.syscall,common_pid.execname:vals=hitcount:sort=id.syscall,hitcount:size=2048 if id == 16 [active] 801 802 { id: sys_ioctl [ 16], common_pid: gmain [ 2769] } hitcount: 1 803 { id: sys_ioctl [ 16], common_pid: evolution-addre [ 8571] } hitcount: 1 804 { id: sys_ioctl [ 16], common_pid: gmain [ 3003] } hitcount: 1 805 { id: sys_ioctl [ 16], common_pid: gmain [ 2781] } hitcount: 1 806 { id: sys_ioctl [ 16], common_pid: gmain [ 2829] } hitcount: 1 807 { id: sys_ioctl [ 16], common_pid: bash [ 8726] } hitcount: 1 808 { id: sys_ioctl [ 16], common_pid: bash [ 8508] } hitcount: 1 809 { id: sys_ioctl [ 16], common_pid: gmain [ 2970] } hitcount: 1 810 { id: sys_ioctl [ 16], common_pid: gmain [ 2768] } hitcount: 1 811 . 812 . 813 . 814 { id: sys_ioctl [ 16], common_pid: pool [ 8559] } hitcount: 45 815 { id: sys_ioctl [ 16], common_pid: pool [ 8555] } hitcount: 48 816 { id: sys_ioctl [ 16], common_pid: pool [ 8551] } hitcount: 48 817 { id: sys_ioctl [ 16], common_pid: avahi-daemon [ 896] } hitcount: 66 818 { id: sys_ioctl [ 16], common_pid: Xorg [ 1267] } hitcount: 26674 819 { id: sys_ioctl [ 16], common_pid: compiz [ 2994] } hitcount: 73443 820 821 Totals: 822 Hits: 101162 823 Entries: 103 824 Dropped: 0 825 826 The above output shows that 'compiz' and 'Xorg' are far and away 827 the heaviest ioctl callers (which might lead to questions about 828 whether they really need to be making all those calls and to 829 possible avenues for further investigation.) 830 831 The compound key examples used a key and a sum value (hitcount) to 832 sort the output, but we can just as easily use two keys instead. 833 Here's an example where we use a compound key composed of the the 834 common_pid and size event fields. Sorting with pid as the primary 835 key and 'size' as the secondary key allows us to display an 836 ordered summary of the recvfrom sizes, with counts, received by 837 each process:: 838 839 # echo 'hist:key=common_pid.execname,size:val=hitcount:sort=common_pid,size' > \ 840 /sys/kernel/debug/tracing/events/syscalls/sys_enter_recvfrom/trigger 841 842 # cat /sys/kernel/debug/tracing/events/syscalls/sys_enter_recvfrom/hist 843 # trigger info: hist:keys=common_pid.execname,size:vals=hitcount:sort=common_pid.execname,size:size=2048 [active] 844 845 { common_pid: smbd [ 784], size: 4 } hitcount: 1 846 { common_pid: dnsmasq [ 1412], size: 4096 } hitcount: 672 847 { common_pid: postgres [ 1796], size: 1000 } hitcount: 6 848 { common_pid: postgres [ 1867], size: 1000 } hitcount: 10 849 { common_pid: bamfdaemon [ 2787], size: 28 } hitcount: 2 850 { common_pid: bamfdaemon [ 2787], size: 14360 } hitcount: 1 851 { common_pid: compiz [ 2994], size: 8 } hitcount: 1 852 { common_pid: compiz [ 2994], size: 20 } hitcount: 11 853 { common_pid: gnome-terminal [ 3199], size: 4 } hitcount: 2 854 { common_pid: firefox [ 8817], size: 4 } hitcount: 1 855 { common_pid: firefox [ 8817], size: 8 } hitcount: 5 856 { common_pid: firefox [ 8817], size: 588 } hitcount: 2 857 { common_pid: firefox [ 8817], size: 628 } hitcount: 1 858 { common_pid: firefox [ 8817], size: 6944 } hitcount: 1 859 { common_pid: firefox [ 8817], size: 408880 } hitcount: 2 860 { common_pid: firefox [ 8822], size: 8 } hitcount: 2 861 { common_pid: firefox [ 8822], size: 160 } hitcount: 2 862 { common_pid: firefox [ 8822], size: 320 } hitcount: 2 863 { common_pid: firefox [ 8822], size: 352 } hitcount: 1 864 . 865 . 866 . 867 { common_pid: pool [ 8923], size: 1960 } hitcount: 10 868 { common_pid: pool [ 8923], size: 2048 } hitcount: 10 869 { common_pid: pool [ 8924], size: 1960 } hitcount: 10 870 { common_pid: pool [ 8924], size: 2048 } hitcount: 10 871 { common_pid: pool [ 8928], size: 1964 } hitcount: 4 872 { common_pid: pool [ 8928], size: 1965 } hitcount: 2 873 { common_pid: pool [ 8928], size: 2048 } hitcount: 6 874 { common_pid: pool [ 8929], size: 1982 } hitcount: 1 875 { common_pid: pool [ 8929], size: 2048 } hitcount: 1 876 877 Totals: 878 Hits: 2016 879 Entries: 224 880 Dropped: 0 881 882 The above example also illustrates the fact that although a compound 883 key is treated as a single entity for hashing purposes, the sub-keys 884 it's composed of can be accessed independently. 885 886 The next example uses a string field as the hash key and 887 demonstrates how you can manually pause and continue a hist trigger. 888 In this example, we'll aggregate fork counts and don't expect a 889 large number of entries in the hash table, so we'll drop it to a 890 much smaller number, say 256:: 891 892 # echo 'hist:key=child_comm:val=hitcount:size=256' > \ 893 /sys/kernel/debug/tracing/events/sched/sched_process_fork/trigger 894 895 # cat /sys/kernel/debug/tracing/events/sched/sched_process_fork/hist 896 # trigger info: hist:keys=child_comm:vals=hitcount:sort=hitcount:size=256 [active] 897 898 { child_comm: dconf worker } hitcount: 1 899 { child_comm: ibus-daemon } hitcount: 1 900 { child_comm: whoopsie } hitcount: 1 901 { child_comm: smbd } hitcount: 1 902 { child_comm: gdbus } hitcount: 1 903 { child_comm: kthreadd } hitcount: 1 904 { child_comm: dconf worker } hitcount: 1 905 { child_comm: evolution-alarm } hitcount: 2 906 { child_comm: Socket Thread } hitcount: 2 907 { child_comm: postgres } hitcount: 2 908 { child_comm: bash } hitcount: 3 909 { child_comm: compiz } hitcount: 3 910 { child_comm: evolution-sourc } hitcount: 4 911 { child_comm: dhclient } hitcount: 4 912 { child_comm: pool } hitcount: 5 913 { child_comm: nm-dispatcher.a } hitcount: 8 914 { child_comm: firefox } hitcount: 8 915 { child_comm: dbus-daemon } hitcount: 8 916 { child_comm: glib-pacrunner } hitcount: 10 917 { child_comm: evolution } hitcount: 23 918 919 Totals: 920 Hits: 89 921 Entries: 20 922 Dropped: 0 923 924 If we want to pause the hist trigger, we can simply append :pause to 925 the command that started the trigger. Notice that the trigger info 926 displays as [paused]:: 927 928 # echo 'hist:key=child_comm:val=hitcount:size=256:pause' >> \ 929 /sys/kernel/debug/tracing/events/sched/sched_process_fork/trigger 930 931 # cat /sys/kernel/debug/tracing/events/sched/sched_process_fork/hist 932 # trigger info: hist:keys=child_comm:vals=hitcount:sort=hitcount:size=256 [paused] 933 934 { child_comm: dconf worker } hitcount: 1 935 { child_comm: kthreadd } hitcount: 1 936 { child_comm: dconf worker } hitcount: 1 937 { child_comm: gdbus } hitcount: 1 938 { child_comm: ibus-daemon } hitcount: 1 939 { child_comm: Socket Thread } hitcount: 2 940 { child_comm: evolution-alarm } hitcount: 2 941 { child_comm: smbd } hitcount: 2 942 { child_comm: bash } hitcount: 3 943 { child_comm: whoopsie } hitcount: 3 944 { child_comm: compiz } hitcount: 3 945 { child_comm: evolution-sourc } hitcount: 4 946 { child_comm: pool } hitcount: 5 947 { child_comm: postgres } hitcount: 6 948 { child_comm: firefox } hitcount: 8 949 { child_comm: dhclient } hitcount: 10 950 { child_comm: emacs } hitcount: 12 951 { child_comm: dbus-daemon } hitcount: 20 952 { child_comm: nm-dispatcher.a } hitcount: 20 953 { child_comm: evolution } hitcount: 35 954 { child_comm: glib-pacrunner } hitcount: 59 955 956 Totals: 957 Hits: 199 958 Entries: 21 959 Dropped: 0 960 961 To manually continue having the trigger aggregate events, append 962 :cont instead. Notice that the trigger info displays as [active] 963 again, and the data has changed:: 964 965 # echo 'hist:key=child_comm:val=hitcount:size=256:cont' >> \ 966 /sys/kernel/debug/tracing/events/sched/sched_process_fork/trigger 967 968 # cat /sys/kernel/debug/tracing/events/sched/sched_process_fork/hist 969 # trigger info: hist:keys=child_comm:vals=hitcount:sort=hitcount:size=256 [active] 970 971 { child_comm: dconf worker } hitcount: 1 972 { child_comm: dconf worker } hitcount: 1 973 { child_comm: kthreadd } hitcount: 1 974 { child_comm: gdbus } hitcount: 1 975 { child_comm: ibus-daemon } hitcount: 1 976 { child_comm: Socket Thread } hitcount: 2 977 { child_comm: evolution-alarm } hitcount: 2 978 { child_comm: smbd } hitcount: 2 979 { child_comm: whoopsie } hitcount: 3 980 { child_comm: compiz } hitcount: 3 981 { child_comm: evolution-sourc } hitcount: 4 982 { child_comm: bash } hitcount: 5 983 { child_comm: pool } hitcount: 5 984 { child_comm: postgres } hitcount: 6 985 { child_comm: firefox } hitcount: 8 986 { child_comm: dhclient } hitcount: 11 987 { child_comm: emacs } hitcount: 12 988 { child_comm: dbus-daemon } hitcount: 22 989 { child_comm: nm-dispatcher.a } hitcount: 22 990 { child_comm: evolution } hitcount: 35 991 { child_comm: glib-pacrunner } hitcount: 59 992 993 Totals: 994 Hits: 206 995 Entries: 21 996 Dropped: 0 997 998 The previous example showed how to start and stop a hist trigger by 999 appending 'pause' and 'continue' to the hist trigger command. A 1000 hist trigger can also be started in a paused state by initially 1001 starting the trigger with ':pause' appended. This allows you to 1002 start the trigger only when you're ready to start collecting data 1003 and not before. For example, you could start the trigger in a 1004 paused state, then unpause it and do something you want to measure, 1005 then pause the trigger again when done. 1006 1007 Of course, doing this manually can be difficult and error-prone, but 1008 it is possible to automatically start and stop a hist trigger based 1009 on some condition, via the enable_hist and disable_hist triggers. 1010 1011 For example, suppose we wanted to take a look at the relative 1012 weights in terms of skb length for each callpath that leads to a 1013 netif_receive_skb event when downloading a decent-sized file using 1014 wget. 1015 1016 First we set up an initially paused stacktrace trigger on the 1017 netif_receive_skb event:: 1018 1019 # echo 'hist:key=stacktrace:vals=len:pause' > \ 1020 /sys/kernel/debug/tracing/events/net/netif_receive_skb/trigger 1021 1022 Next, we set up an 'enable_hist' trigger on the sched_process_exec 1023 event, with an 'if filename==/usr/bin/wget' filter. The effect of 1024 this new trigger is that it will 'unpause' the hist trigger we just 1025 set up on netif_receive_skb if and only if it sees a 1026 sched_process_exec event with a filename of '/usr/bin/wget'. When 1027 that happens, all netif_receive_skb events are aggregated into a 1028 hash table keyed on stacktrace:: 1029 1030 # echo 'enable_hist:net:netif_receive_skb if filename==/usr/bin/wget' > \ 1031 /sys/kernel/debug/tracing/events/sched/sched_process_exec/trigger 1032 1033 The aggregation continues until the netif_receive_skb is paused 1034 again, which is what the following disable_hist event does by 1035 creating a similar setup on the sched_process_exit event, using the 1036 filter 'comm==wget':: 1037 1038 # echo 'disable_hist:net:netif_receive_skb if comm==wget' > \ 1039 /sys/kernel/debug/tracing/events/sched/sched_process_exit/trigger 1040 1041 Whenever a process exits and the comm field of the disable_hist 1042 trigger filter matches 'comm==wget', the netif_receive_skb hist 1043 trigger is disabled. 1044 1045 The overall effect is that netif_receive_skb events are aggregated 1046 into the hash table for only the duration of the wget. Executing a 1047 wget command and then listing the 'hist' file will display the 1048 output generated by the wget command:: 1049 1050 $ wget https://www.kernel.org/pub/linux/kernel/v3.x/patch-3.19.xz 1051 1052 # cat /sys/kernel/debug/tracing/events/net/netif_receive_skb/hist 1053 # trigger info: hist:keys=stacktrace:vals=len:sort=hitcount:size=2048 [paused] 1054 1055 { stacktrace: 1056 __netif_receive_skb_core+0x46d/0x990 1057 __netif_receive_skb+0x18/0x60 1058 netif_receive_skb_internal+0x23/0x90 1059 napi_gro_receive+0xc8/0x100 1060 ieee80211_deliver_skb+0xd6/0x270 [mac80211] 1061 ieee80211_rx_handlers+0xccf/0x22f0 [mac80211] 1062 ieee80211_prepare_and_rx_handle+0x4e7/0xc40 [mac80211] 1063 ieee80211_rx+0x31d/0x900 [mac80211] 1064 iwlagn_rx_reply_rx+0x3db/0x6f0 [iwldvm] 1065 iwl_rx_dispatch+0x8e/0xf0 [iwldvm] 1066 iwl_pcie_irq_handler+0xe3c/0x12f0 [iwlwifi] 1067 irq_thread_fn+0x20/0x50 1068 irq_thread+0x11f/0x150 1069 kthread+0xd2/0xf0 1070 ret_from_fork+0x42/0x70 1071 } hitcount: 85 len: 28884 1072 { stacktrace: 1073 __netif_receive_skb_core+0x46d/0x990 1074 __netif_receive_skb+0x18/0x60 1075 netif_receive_skb_internal+0x23/0x90 1076 napi_gro_complete+0xa4/0xe0 1077 dev_gro_receive+0x23a/0x360 1078 napi_gro_receive+0x30/0x100 1079 ieee80211_deliver_skb+0xd6/0x270 [mac80211] 1080 ieee80211_rx_handlers+0xccf/0x22f0 [mac80211] 1081 ieee80211_prepare_and_rx_handle+0x4e7/0xc40 [mac80211] 1082 ieee80211_rx+0x31d/0x900 [mac80211] 1083 iwlagn_rx_reply_rx+0x3db/0x6f0 [iwldvm] 1084 iwl_rx_dispatch+0x8e/0xf0 [iwldvm] 1085 iwl_pcie_irq_handler+0xe3c/0x12f0 [iwlwifi] 1086 irq_thread_fn+0x20/0x50 1087 irq_thread+0x11f/0x150 1088 kthread+0xd2/0xf0 1089 } hitcount: 98 len: 664329 1090 { stacktrace: 1091 __netif_receive_skb_core+0x46d/0x990 1092 __netif_receive_skb+0x18/0x60 1093 process_backlog+0xa8/0x150 1094 net_rx_action+0x15d/0x340 1095 __do_softirq+0x114/0x2c0 1096 do_softirq_own_stack+0x1c/0x30 1097 do_softirq+0x65/0x70 1098 __local_bh_enable_ip+0xb5/0xc0 1099 ip_finish_output+0x1f4/0x840 1100 ip_output+0x6b/0xc0 1101 ip_local_out_sk+0x31/0x40 1102 ip_send_skb+0x1a/0x50 1103 udp_send_skb+0x173/0x2a0 1104 udp_sendmsg+0x2bf/0x9f0 1105 inet_sendmsg+0x64/0xa0 1106 sock_sendmsg+0x3d/0x50 1107 } hitcount: 115 len: 13030 1108 { stacktrace: 1109 __netif_receive_skb_core+0x46d/0x990 1110 __netif_receive_skb+0x18/0x60 1111 netif_receive_skb_internal+0x23/0x90 1112 napi_gro_complete+0xa4/0xe0 1113 napi_gro_flush+0x6d/0x90 1114 iwl_pcie_irq_handler+0x92a/0x12f0 [iwlwifi] 1115 irq_thread_fn+0x20/0x50 1116 irq_thread+0x11f/0x150 1117 kthread+0xd2/0xf0 1118 ret_from_fork+0x42/0x70 1119 } hitcount: 934 len: 5512212 1120 1121 Totals: 1122 Hits: 1232 1123 Entries: 4 1124 Dropped: 0 1125 1126 The above shows all the netif_receive_skb callpaths and their total 1127 lengths for the duration of the wget command. 1128 1129 The 'clear' hist trigger param can be used to clear the hash table. 1130 Suppose we wanted to try another run of the previous example but 1131 this time also wanted to see the complete list of events that went 1132 into the histogram. In order to avoid having to set everything up 1133 again, we can just clear the histogram first:: 1134 1135 # echo 'hist:key=stacktrace:vals=len:clear' >> \ 1136 /sys/kernel/debug/tracing/events/net/netif_receive_skb/trigger 1137 1138 Just to verify that it is in fact cleared, here's what we now see in 1139 the hist file:: 1140 1141 # cat /sys/kernel/debug/tracing/events/net/netif_receive_skb/hist 1142 # trigger info: hist:keys=stacktrace:vals=len:sort=hitcount:size=2048 [paused] 1143 1144 Totals: 1145 Hits: 0 1146 Entries: 0 1147 Dropped: 0 1148 1149 Since we want to see the detailed list of every netif_receive_skb 1150 event occurring during the new run, which are in fact the same 1151 events being aggregated into the hash table, we add some additional 1152 'enable_event' events to the triggering sched_process_exec and 1153 sched_process_exit events as such:: 1154 1155 # echo 'enable_event:net:netif_receive_skb if filename==/usr/bin/wget' > \ 1156 /sys/kernel/debug/tracing/events/sched/sched_process_exec/trigger 1157 1158 # echo 'disable_event:net:netif_receive_skb if comm==wget' > \ 1159 /sys/kernel/debug/tracing/events/sched/sched_process_exit/trigger 1160 1161 If you read the trigger files for the sched_process_exec and 1162 sched_process_exit triggers, you should see two triggers for each: 1163 one enabling/disabling the hist aggregation and the other 1164 enabling/disabling the logging of events:: 1165 1166 # cat /sys/kernel/debug/tracing/events/sched/sched_process_exec/trigger 1167 enable_event:net:netif_receive_skb:unlimited if filename==/usr/bin/wget 1168 enable_hist:net:netif_receive_skb:unlimited if filename==/usr/bin/wget 1169 1170 # cat /sys/kernel/debug/tracing/events/sched/sched_process_exit/trigger 1171 enable_event:net:netif_receive_skb:unlimited if comm==wget 1172 disable_hist:net:netif_receive_skb:unlimited if comm==wget 1173 1174 In other words, whenever either of the sched_process_exec or 1175 sched_process_exit events is hit and matches 'wget', it enables or 1176 disables both the histogram and the event log, and what you end up 1177 with is a hash table and set of events just covering the specified 1178 duration. Run the wget command again:: 1179 1180 $ wget https://www.kernel.org/pub/linux/kernel/v3.x/patch-3.19.xz 1181 1182 Displaying the 'hist' file should show something similar to what you 1183 saw in the last run, but this time you should also see the 1184 individual events in the trace file:: 1185 1186 # cat /sys/kernel/debug/tracing/trace 1187 1188 # tracer: nop 1189 # 1190 # entries-in-buffer/entries-written: 183/1426 #P:4 1191 # 1192 # _-----=> irqs-off 1193 # / _----=> need-resched 1194 # | / _---=> hardirq/softirq 1195 # || / _--=> preempt-depth 1196 # ||| / delay 1197 # TASK-PID CPU# |||| TIMESTAMP FUNCTION 1198 # | | | |||| | | 1199 wget-15108 [000] ..s1 31769.606929: netif_receive_skb: dev=lo skbaddr=ffff88009c353100 len=60 1200 wget-15108 [000] ..s1 31769.606999: netif_receive_skb: dev=lo skbaddr=ffff88009c353200 len=60 1201 dnsmasq-1382 [000] ..s1 31769.677652: netif_receive_skb: dev=lo skbaddr=ffff88009c352b00 len=130 1202 dnsmasq-1382 [000] ..s1 31769.685917: netif_receive_skb: dev=lo skbaddr=ffff88009c352200 len=138 1203 ##### CPU 2 buffer started #### 1204 irq/29-iwlwifi-559 [002] ..s. 31772.031529: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d433d00 len=2948 1205 irq/29-iwlwifi-559 [002] ..s. 31772.031572: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d432200 len=1500 1206 irq/29-iwlwifi-559 [002] ..s. 31772.032196: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d433100 len=2948 1207 irq/29-iwlwifi-559 [002] ..s. 31772.032761: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d433000 len=2948 1208 irq/29-iwlwifi-559 [002] ..s. 31772.033220: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d432e00 len=1500 1209 . 1210 . 1211 . 1212 1213 The following example demonstrates how multiple hist triggers can be 1214 attached to a given event. This capability can be useful for 1215 creating a set of different summaries derived from the same set of 1216 events, or for comparing the effects of different filters, among 1217 other things:: 1218 1219 # echo 'hist:keys=skbaddr.hex:vals=len if len < 0' >> \ 1220 /sys/kernel/debug/tracing/events/net/netif_receive_skb/trigger 1221 # echo 'hist:keys=skbaddr.hex:vals=len if len > 4096' >> \ 1222 /sys/kernel/debug/tracing/events/net/netif_receive_skb/trigger 1223 # echo 'hist:keys=skbaddr.hex:vals=len if len == 256' >> \ 1224 /sys/kernel/debug/tracing/events/net/netif_receive_skb/trigger 1225 # echo 'hist:keys=skbaddr.hex:vals=len' >> \ 1226 /sys/kernel/debug/tracing/events/net/netif_receive_skb/trigger 1227 # echo 'hist:keys=len:vals=common_preempt_count' >> \ 1228 /sys/kernel/debug/tracing/events/net/netif_receive_skb/trigger 1229 1230 The above set of commands create four triggers differing only in 1231 their filters, along with a completely different though fairly 1232 nonsensical trigger. Note that in order to append multiple hist 1233 triggers to the same file, you should use the '>>' operator to 1234 append them ('>' will also add the new hist trigger, but will remove 1235 any existing hist triggers beforehand). 1236 1237 Displaying the contents of the 'hist' file for the event shows the 1238 contents of all five histograms:: 1239 1240 # cat /sys/kernel/debug/tracing/events/net/netif_receive_skb/hist 1241 1242 # event histogram 1243 # 1244 # trigger info: hist:keys=len:vals=hitcount,common_preempt_count:sort=hitcount:size=2048 [active] 1245 # 1246 1247 { len: 176 } hitcount: 1 common_preempt_count: 0 1248 { len: 223 } hitcount: 1 common_preempt_count: 0 1249 { len: 4854 } hitcount: 1 common_preempt_count: 0 1250 { len: 395 } hitcount: 1 common_preempt_count: 0 1251 { len: 177 } hitcount: 1 common_preempt_count: 0 1252 { len: 446 } hitcount: 1 common_preempt_count: 0 1253 { len: 1601 } hitcount: 1 common_preempt_count: 0 1254 . 1255 . 1256 . 1257 { len: 1280 } hitcount: 66 common_preempt_count: 0 1258 { len: 116 } hitcount: 81 common_preempt_count: 40 1259 { len: 708 } hitcount: 112 common_preempt_count: 0 1260 { len: 46 } hitcount: 221 common_preempt_count: 0 1261 { len: 1264 } hitcount: 458 common_preempt_count: 0 1262 1263 Totals: 1264 Hits: 1428 1265 Entries: 147 1266 Dropped: 0 1267 1268 1269 # event histogram 1270 # 1271 # trigger info: hist:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 [active] 1272 # 1273 1274 { skbaddr: ffff8800baee5e00 } hitcount: 1 len: 130 1275 { skbaddr: ffff88005f3d5600 } hitcount: 1 len: 1280 1276 { skbaddr: ffff88005f3d4900 } hitcount: 1 len: 1280 1277 { skbaddr: ffff88009fed6300 } hitcount: 1 len: 115 1278 { skbaddr: ffff88009fe0ad00 } hitcount: 1 len: 115 1279 { skbaddr: ffff88008cdb1900 } hitcount: 1 len: 46 1280 { skbaddr: ffff880064b5ef00 } hitcount: 1 len: 118 1281 { skbaddr: ffff880044e3c700 } hitcount: 1 len: 60 1282 { skbaddr: ffff880100065900 } hitcount: 1 len: 46 1283 { skbaddr: ffff8800d46bd500 } hitcount: 1 len: 116 1284 { skbaddr: ffff88005f3d5f00 } hitcount: 1 len: 1280 1285 { skbaddr: ffff880100064700 } hitcount: 1 len: 365 1286 { skbaddr: ffff8800badb6f00 } hitcount: 1 len: 60 1287 . 1288 . 1289 . 1290 { skbaddr: ffff88009fe0be00 } hitcount: 27 len: 24677 1291 { skbaddr: ffff88009fe0a400 } hitcount: 27 len: 23052 1292 { skbaddr: ffff88009fe0b700 } hitcount: 31 len: 25589 1293 { skbaddr: ffff88009fe0b600 } hitcount: 32 len: 27326 1294 { skbaddr: ffff88006a462800 } hitcount: 68 len: 71678 1295 { skbaddr: ffff88006a463700 } hitcount: 70 len: 72678 1296 { skbaddr: ffff88006a462b00 } hitcount: 71 len: 77589 1297 { skbaddr: ffff88006a463600 } hitcount: 73 len: 71307 1298 { skbaddr: ffff88006a462200 } hitcount: 81 len: 81032 1299 1300 Totals: 1301 Hits: 1451 1302 Entries: 318 1303 Dropped: 0 1304 1305 1306 # event histogram 1307 # 1308 # trigger info: hist:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 if len == 256 [active] 1309 # 1310 1311 1312 Totals: 1313 Hits: 0 1314 Entries: 0 1315 Dropped: 0 1316 1317 1318 # event histogram 1319 # 1320 # trigger info: hist:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 if len > 4096 [active] 1321 # 1322 1323 { skbaddr: ffff88009fd2c300 } hitcount: 1 len: 7212 1324 { skbaddr: ffff8800d2bcce00 } hitcount: 1 len: 7212 1325 { skbaddr: ffff8800d2bcd700 } hitcount: 1 len: 7212 1326 { skbaddr: ffff8800d2bcda00 } hitcount: 1 len: 21492 1327 { skbaddr: ffff8800ae2e2d00 } hitcount: 1 len: 7212 1328 { skbaddr: ffff8800d2bcdb00 } hitcount: 1 len: 7212 1329 { skbaddr: ffff88006a4df500 } hitcount: 1 len: 4854 1330 { skbaddr: ffff88008ce47b00 } hitcount: 1 len: 18636 1331 { skbaddr: ffff8800ae2e2200 } hitcount: 1 len: 12924 1332 { skbaddr: ffff88005f3e1000 } hitcount: 1 len: 4356 1333 { skbaddr: ffff8800d2bcdc00 } hitcount: 2 len: 24420 1334 { skbaddr: ffff8800d2bcc200 } hitcount: 2 len: 12996 1335 1336 Totals: 1337 Hits: 14 1338 Entries: 12 1339 Dropped: 0 1340 1341 1342 # event histogram 1343 # 1344 # trigger info: hist:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 if len < 0 [active] 1345 # 1346 1347 1348 Totals: 1349 Hits: 0 1350 Entries: 0 1351 Dropped: 0 1352 1353 Named triggers can be used to have triggers share a common set of 1354 histogram data. This capability is mostly useful for combining the 1355 output of events generated by tracepoints contained inside inline 1356 functions, but names can be used in a hist trigger on any event. 1357 For example, these two triggers when hit will update the same 'len' 1358 field in the shared 'foo' histogram data:: 1359 1360 # echo 'hist:name=foo:keys=skbaddr.hex:vals=len' > \ 1361 /sys/kernel/debug/tracing/events/net/netif_receive_skb/trigger 1362 # echo 'hist:name=foo:keys=skbaddr.hex:vals=len' > \ 1363 /sys/kernel/debug/tracing/events/net/netif_rx/trigger 1364 1365 You can see that they're updating common histogram data by reading 1366 each event's hist files at the same time:: 1367 1368 # cat /sys/kernel/debug/tracing/events/net/netif_receive_skb/hist; 1369 cat /sys/kernel/debug/tracing/events/net/netif_rx/hist 1370 1371 # event histogram 1372 # 1373 # trigger info: hist:name=foo:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 [active] 1374 # 1375 1376 { skbaddr: ffff88000ad53500 } hitcount: 1 len: 46 1377 { skbaddr: ffff8800af5a1500 } hitcount: 1 len: 76 1378 { skbaddr: ffff8800d62a1900 } hitcount: 1 len: 46 1379 { skbaddr: ffff8800d2bccb00 } hitcount: 1 len: 468 1380 { skbaddr: ffff8800d3c69900 } hitcount: 1 len: 46 1381 { skbaddr: ffff88009ff09100 } hitcount: 1 len: 52 1382 { skbaddr: ffff88010f13ab00 } hitcount: 1 len: 168 1383 { skbaddr: ffff88006a54f400 } hitcount: 1 len: 46 1384 { skbaddr: ffff8800d2bcc500 } hitcount: 1 len: 260 1385 { skbaddr: ffff880064505000 } hitcount: 1 len: 46 1386 { skbaddr: ffff8800baf24e00 } hitcount: 1 len: 32 1387 { skbaddr: ffff88009fe0ad00 } hitcount: 1 len: 46 1388 { skbaddr: ffff8800d3edff00 } hitcount: 1 len: 44 1389 { skbaddr: ffff88009fe0b400 } hitcount: 1 len: 168 1390 { skbaddr: ffff8800a1c55a00 } hitcount: 1 len: 40 1391 { skbaddr: ffff8800d2bcd100 } hitcount: 1 len: 40 1392 { skbaddr: ffff880064505f00 } hitcount: 1 len: 174 1393 { skbaddr: ffff8800a8bff200 } hitcount: 1 len: 160 1394 { skbaddr: ffff880044e3cc00 } hitcount: 1 len: 76 1395 { skbaddr: ffff8800a8bfe700 } hitcount: 1 len: 46 1396 { skbaddr: ffff8800d2bcdc00 } hitcount: 1 len: 32 1397 { skbaddr: ffff8800a1f64800 } hitcount: 1 len: 46 1398 { skbaddr: ffff8800d2bcde00 } hitcount: 1 len: 988 1399 { skbaddr: ffff88006a5dea00 } hitcount: 1 len: 46 1400 { skbaddr: ffff88002e37a200 } hitcount: 1 len: 44 1401 { skbaddr: ffff8800a1f32c00 } hitcount: 2 len: 676 1402 { skbaddr: ffff88000ad52600 } hitcount: 2 len: 107 1403 { skbaddr: ffff8800a1f91e00 } hitcount: 2 len: 92 1404 { skbaddr: ffff8800af5a0200 } hitcount: 2 len: 142 1405 { skbaddr: ffff8800d2bcc600 } hitcount: 2 len: 220 1406 { skbaddr: ffff8800ba36f500 } hitcount: 2 len: 92 1407 { skbaddr: ffff8800d021f800 } hitcount: 2 len: 92 1408 { skbaddr: ffff8800a1f33600 } hitcount: 2 len: 675 1409 { skbaddr: ffff8800a8bfff00 } hitcount: 3 len: 138 1410 { skbaddr: ffff8800d62a1300 } hitcount: 3 len: 138 1411 { skbaddr: ffff88002e37a100 } hitcount: 4 len: 184 1412 { skbaddr: ffff880064504400 } hitcount: 4 len: 184 1413 { skbaddr: ffff8800a8bfec00 } hitcount: 4 len: 184 1414 { skbaddr: ffff88000ad53700 } hitcount: 5 len: 230 1415 { skbaddr: ffff8800d2bcdb00 } hitcount: 5 len: 196 1416 { skbaddr: ffff8800a1f90000 } hitcount: 6 len: 276 1417 { skbaddr: ffff88006a54f900 } hitcount: 6 len: 276 1418 1419 Totals: 1420 Hits: 81 1421 Entries: 42 1422 Dropped: 0 1423 # event histogram 1424 # 1425 # trigger info: hist:name=foo:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 [active] 1426 # 1427 1428 { skbaddr: ffff88000ad53500 } hitcount: 1 len: 46 1429 { skbaddr: ffff8800af5a1500 } hitcount: 1 len: 76 1430 { skbaddr: ffff8800d62a1900 } hitcount: 1 len: 46 1431 { skbaddr: ffff8800d2bccb00 } hitcount: 1 len: 468 1432 { skbaddr: ffff8800d3c69900 } hitcount: 1 len: 46 1433 { skbaddr: ffff88009ff09100 } hitcount: 1 len: 52 1434 { skbaddr: ffff88010f13ab00 } hitcount: 1 len: 168 1435 { skbaddr: ffff88006a54f400 } hitcount: 1 len: 46 1436 { skbaddr: ffff8800d2bcc500 } hitcount: 1 len: 260 1437 { skbaddr: ffff880064505000 } hitcount: 1 len: 46 1438 { skbaddr: ffff8800baf24e00 } hitcount: 1 len: 32 1439 { skbaddr: ffff88009fe0ad00 } hitcount: 1 len: 46 1440 { skbaddr: ffff8800d3edff00 } hitcount: 1 len: 44 1441 { skbaddr: ffff88009fe0b400 } hitcount: 1 len: 168 1442 { skbaddr: ffff8800a1c55a00 } hitcount: 1 len: 40 1443 { skbaddr: ffff8800d2bcd100 } hitcount: 1 len: 40 1444 { skbaddr: ffff880064505f00 } hitcount: 1 len: 174 1445 { skbaddr: ffff8800a8bff200 } hitcount: 1 len: 160 1446 { skbaddr: ffff880044e3cc00 } hitcount: 1 len: 76 1447 { skbaddr: ffff8800a8bfe700 } hitcount: 1 len: 46 1448 { skbaddr: ffff8800d2bcdc00 } hitcount: 1 len: 32 1449 { skbaddr: ffff8800a1f64800 } hitcount: 1 len: 46 1450 { skbaddr: ffff8800d2bcde00 } hitcount: 1 len: 988 1451 { skbaddr: ffff88006a5dea00 } hitcount: 1 len: 46 1452 { skbaddr: ffff88002e37a200 } hitcount: 1 len: 44 1453 { skbaddr: ffff8800a1f32c00 } hitcount: 2 len: 676 1454 { skbaddr: ffff88000ad52600 } hitcount: 2 len: 107 1455 { skbaddr: ffff8800a1f91e00 } hitcount: 2 len: 92 1456 { skbaddr: ffff8800af5a0200 } hitcount: 2 len: 142 1457 { skbaddr: ffff8800d2bcc600 } hitcount: 2 len: 220 1458 { skbaddr: ffff8800ba36f500 } hitcount: 2 len: 92 1459 { skbaddr: ffff8800d021f800 } hitcount: 2 len: 92 1460 { skbaddr: ffff8800a1f33600 } hitcount: 2 len: 675 1461 { skbaddr: ffff8800a8bfff00 } hitcount: 3 len: 138 1462 { skbaddr: ffff8800d62a1300 } hitcount: 3 len: 138 1463 { skbaddr: ffff88002e37a100 } hitcount: 4 len: 184 1464 { skbaddr: ffff880064504400 } hitcount: 4 len: 184 1465 { skbaddr: ffff8800a8bfec00 } hitcount: 4 len: 184 1466 { skbaddr: ffff88000ad53700 } hitcount: 5 len: 230 1467 { skbaddr: ffff8800d2bcdb00 } hitcount: 5 len: 196 1468 { skbaddr: ffff8800a1f90000 } hitcount: 6 len: 276 1469 { skbaddr: ffff88006a54f900 } hitcount: 6 len: 276 1470 1471 Totals: 1472 Hits: 81 1473 Entries: 42 1474 Dropped: 0 1475 1476 And here's an example that shows how to combine histogram data from 1477 any two events even if they don't share any 'compatible' fields 1478 other than 'hitcount' and 'stacktrace'. These commands create a 1479 couple of triggers named 'bar' using those fields:: 1480 1481 # echo 'hist:name=bar:key=stacktrace:val=hitcount' > \ 1482 /sys/kernel/debug/tracing/events/sched/sched_process_fork/trigger 1483 # echo 'hist:name=bar:key=stacktrace:val=hitcount' > \ 1484 /sys/kernel/debug/tracing/events/net/netif_rx/trigger 1485 1486 And displaying the output of either shows some interesting if 1487 somewhat confusing output:: 1488 1489 # cat /sys/kernel/debug/tracing/events/sched/sched_process_fork/hist 1490 # cat /sys/kernel/debug/tracing/events/net/netif_rx/hist 1491 1492 # event histogram 1493 # 1494 # trigger info: hist:name=bar:keys=stacktrace:vals=hitcount:sort=hitcount:size=2048 [active] 1495 # 1496 1497 { stacktrace: 1498 _do_fork+0x18e/0x330 1499 kernel_thread+0x29/0x30 1500 kthreadd+0x154/0x1b0 1501 ret_from_fork+0x3f/0x70 1502 } hitcount: 1 1503 { stacktrace: 1504 netif_rx_internal+0xb2/0xd0 1505 netif_rx_ni+0x20/0x70 1506 dev_loopback_xmit+0xaa/0xd0 1507 ip_mc_output+0x126/0x240 1508 ip_local_out_sk+0x31/0x40 1509 igmp_send_report+0x1e9/0x230 1510 igmp_timer_expire+0xe9/0x120 1511 call_timer_fn+0x39/0xf0 1512 run_timer_softirq+0x1e1/0x290 1513 __do_softirq+0xfd/0x290 1514 irq_exit+0x98/0xb0 1515 smp_apic_timer_interrupt+0x4a/0x60 1516 apic_timer_interrupt+0x6d/0x80 1517 cpuidle_enter+0x17/0x20 1518 call_cpuidle+0x3b/0x60 1519 cpu_startup_entry+0x22d/0x310 1520 } hitcount: 1 1521 { stacktrace: 1522 netif_rx_internal+0xb2/0xd0 1523 netif_rx_ni+0x20/0x70 1524 dev_loopback_xmit+0xaa/0xd0 1525 ip_mc_output+0x17f/0x240 1526 ip_local_out_sk+0x31/0x40 1527 ip_send_skb+0x1a/0x50 1528 udp_send_skb+0x13e/0x270 1529 udp_sendmsg+0x2bf/0x980 1530 inet_sendmsg+0x67/0xa0 1531 sock_sendmsg+0x38/0x50 1532 SYSC_sendto+0xef/0x170 1533 SyS_sendto+0xe/0x10 1534 entry_SYSCALL_64_fastpath+0x12/0x6a 1535 } hitcount: 2 1536 { stacktrace: 1537 netif_rx_internal+0xb2/0xd0 1538 netif_rx+0x1c/0x60 1539 loopback_xmit+0x6c/0xb0 1540 dev_hard_start_xmit+0x219/0x3a0 1541 __dev_queue_xmit+0x415/0x4f0 1542 dev_queue_xmit_sk+0x13/0x20 1543 ip_finish_output2+0x237/0x340 1544 ip_finish_output+0x113/0x1d0 1545 ip_output+0x66/0xc0 1546 ip_local_out_sk+0x31/0x40 1547 ip_send_skb+0x1a/0x50 1548 udp_send_skb+0x16d/0x270 1549 udp_sendmsg+0x2bf/0x980 1550 inet_sendmsg+0x67/0xa0 1551 sock_sendmsg+0x38/0x50 1552 ___sys_sendmsg+0x14e/0x270 1553 } hitcount: 76 1554 { stacktrace: 1555 netif_rx_internal+0xb2/0xd0 1556 netif_rx+0x1c/0x60 1557 loopback_xmit+0x6c/0xb0 1558 dev_hard_start_xmit+0x219/0x3a0 1559 __dev_queue_xmit+0x415/0x4f0 1560 dev_queue_xmit_sk+0x13/0x20 1561 ip_finish_output2+0x237/0x340 1562 ip_finish_output+0x113/0x1d0 1563 ip_output+0x66/0xc0 1564 ip_local_out_sk+0x31/0x40 1565 ip_send_skb+0x1a/0x50 1566 udp_send_skb+0x16d/0x270 1567 udp_sendmsg+0x2bf/0x980 1568 inet_sendmsg+0x67/0xa0 1569 sock_sendmsg+0x38/0x50 1570 ___sys_sendmsg+0x269/0x270 1571 } hitcount: 77 1572 { stacktrace: 1573 netif_rx_internal+0xb2/0xd0 1574 netif_rx+0x1c/0x60 1575 loopback_xmit+0x6c/0xb0 1576 dev_hard_start_xmit+0x219/0x3a0 1577 __dev_queue_xmit+0x415/0x4f0 1578 dev_queue_xmit_sk+0x13/0x20 1579 ip_finish_output2+0x237/0x340 1580 ip_finish_output+0x113/0x1d0 1581 ip_output+0x66/0xc0 1582 ip_local_out_sk+0x31/0x40 1583 ip_send_skb+0x1a/0x50 1584 udp_send_skb+0x16d/0x270 1585 udp_sendmsg+0x2bf/0x980 1586 inet_sendmsg+0x67/0xa0 1587 sock_sendmsg+0x38/0x50 1588 SYSC_sendto+0xef/0x170 1589 } hitcount: 88 1590 { stacktrace: 1591 _do_fork+0x18e/0x330 1592 SyS_clone+0x19/0x20 1593 entry_SYSCALL_64_fastpath+0x12/0x6a 1594 } hitcount: 244 1595 1596 Totals: 1597 Hits: 489 1598 Entries: 7 1599 Dropped: 0 1600 16012.2 Inter-event hist triggers 1602----------------------------- 1603 1604Inter-event hist triggers are hist triggers that combine values from 1605one or more other events and create a histogram using that data. Data 1606from an inter-event histogram can in turn become the source for 1607further combined histograms, thus providing a chain of related 1608histograms, which is important for some applications. 1609 1610The most important example of an inter-event quantity that can be used 1611in this manner is latency, which is simply a difference in timestamps 1612between two events. Although latency is the most important 1613inter-event quantity, note that because the support is completely 1614general across the trace event subsystem, any event field can be used 1615in an inter-event quantity. 1616 1617An example of a histogram that combines data from other histograms 1618into a useful chain would be a 'wakeupswitch latency' histogram that 1619combines a 'wakeup latency' histogram and a 'switch latency' 1620histogram. 1621 1622Normally, a hist trigger specification consists of a (possibly 1623compound) key along with one or more numeric values, which are 1624continually updated sums associated with that key. A histogram 1625specification in this case consists of individual key and value 1626specifications that refer to trace event fields associated with a 1627single event type. 1628 1629The inter-event hist trigger extension allows fields from multiple 1630events to be referenced and combined into a multi-event histogram 1631specification. In support of this overall goal, a few enabling 1632features have been added to the hist trigger support: 1633 1634 - In order to compute an inter-event quantity, a value from one 1635 event needs to saved and then referenced from another event. This 1636 requires the introduction of support for histogram 'variables'. 1637 1638 - The computation of inter-event quantities and their combination 1639 require some minimal amount of support for applying simple 1640 expressions to variables (+ and -). 1641 1642 - A histogram consisting of inter-event quantities isn't logically a 1643 histogram on either event (so having the 'hist' file for either 1644 event host the histogram output doesn't really make sense). To 1645 address the idea that the histogram is associated with a 1646 combination of events, support is added allowing the creation of 1647 'synthetic' events that are events derived from other events. 1648 These synthetic events are full-fledged events just like any other 1649 and can be used as such, as for instance to create the 1650 'combination' histograms mentioned previously. 1651 1652 - A set of 'actions' can be associated with histogram entries - 1653 these can be used to generate the previously mentioned synthetic 1654 events, but can also be used for other purposes, such as for 1655 example saving context when a 'max' latency has been hit. 1656 1657 - Trace events don't have a 'timestamp' associated with them, but 1658 there is an implicit timestamp saved along with an event in the 1659 underlying ftrace ring buffer. This timestamp is now exposed as a 1660 a synthetic field named 'common_timestamp' which can be used in 1661 histograms as if it were any other event field; it isn't an actual 1662 field in the trace format but rather is a synthesized value that 1663 nonetheless can be used as if it were an actual field. By default 1664 it is in units of nanoseconds; appending '.usecs' to a 1665 common_timestamp field changes the units to microseconds. 1666 1667A note on inter-event timestamps: If common_timestamp is used in a 1668histogram, the trace buffer is automatically switched over to using 1669absolute timestamps and the "global" trace clock, in order to avoid 1670bogus timestamp differences with other clocks that aren't coherent 1671across CPUs. This can be overridden by specifying one of the other 1672trace clocks instead, using the "clock=XXX" hist trigger attribute, 1673where XXX is any of the clocks listed in the tracing/trace_clock 1674pseudo-file. 1675 1676These features are described in more detail in the following sections. 1677 16782.2.1 Histogram Variables 1679------------------------- 1680 1681Variables are simply named locations used for saving and retrieving 1682values between matching events. A 'matching' event is defined as an 1683event that has a matching key - if a variable is saved for a histogram 1684entry corresponding to that key, any subsequent event with a matching 1685key can access that variable. 1686 1687A variable's value is normally available to any subsequent event until 1688it is set to something else by a subsequent event. The one exception 1689to that rule is that any variable used in an expression is essentially 1690'read-once' - once it's used by an expression in a subsequent event, 1691it's reset to its 'unset' state, which means it can't be used again 1692unless it's set again. This ensures not only that an event doesn't 1693use an uninitialized variable in a calculation, but that that variable 1694is used only once and not for any unrelated subsequent match. 1695 1696The basic syntax for saving a variable is to simply prefix a unique 1697variable name not corresponding to any keyword along with an '=' sign 1698to any event field. 1699 1700Either keys or values can be saved and retrieved in this way. This 1701creates a variable named 'ts0' for a histogram entry with the key 1702'next_pid':: 1703 1704 # echo 'hist:keys=next_pid:vals=$ts0:ts0=common_timestamp ... >> \ 1705 event/trigger 1706 1707The ts0 variable can be accessed by any subsequent event having the 1708same pid as 'next_pid'. 1709 1710Variable references are formed by prepending the variable name with 1711the '$' sign. Thus for example, the ts0 variable above would be 1712referenced as '$ts0' in expressions. 1713 1714Because 'vals=' is used, the common_timestamp variable value above 1715will also be summed as a normal histogram value would (though for a 1716timestamp it makes little sense). 1717 1718The below shows that a key value can also be saved in the same way:: 1719 1720 # echo 'hist:timer_pid=common_pid:key=timer_pid ...' >> event/trigger 1721 1722If a variable isn't a key variable or prefixed with 'vals=', the 1723associated event field will be saved in a variable but won't be summed 1724as a value:: 1725 1726 # echo 'hist:keys=next_pid:ts1=common_timestamp ...' >> event/trigger 1727 1728Multiple variables can be assigned at the same time. The below would 1729result in both ts0 and b being created as variables, with both 1730common_timestamp and field1 additionally being summed as values:: 1731 1732 # echo 'hist:keys=pid:vals=$ts0,$b:ts0=common_timestamp,b=field1 ...' >> \ 1733 event/trigger 1734 1735Note that variable assignments can appear either preceding or 1736following their use. The command below behaves identically to the 1737command above:: 1738 1739 # echo 'hist:keys=pid:ts0=common_timestamp,b=field1:vals=$ts0,$b ...' >> \ 1740 event/trigger 1741 1742Any number of variables not bound to a 'vals=' prefix can also be 1743assigned by simply separating them with colons. Below is the same 1744thing but without the values being summed in the histogram:: 1745 1746 # echo 'hist:keys=pid:ts0=common_timestamp:b=field1 ...' >> event/trigger 1747 1748Variables set as above can be referenced and used in expressions on 1749another event. 1750 1751For example, here's how a latency can be calculated:: 1752 1753 # echo 'hist:keys=pid,prio:ts0=common_timestamp ...' >> event1/trigger 1754 # echo 'hist:keys=next_pid:wakeup_lat=common_timestamp-$ts0 ...' >> event2/trigger 1755 1756In the first line above, the event's timestamp is saved into the 1757variable ts0. In the next line, ts0 is subtracted from the second 1758event's timestamp to produce the latency, which is then assigned into 1759yet another variable, 'wakeup_lat'. The hist trigger below in turn 1760makes use of the wakeup_lat variable to compute a combined latency 1761using the same key and variable from yet another event:: 1762 1763 # echo 'hist:key=pid:wakeupswitch_lat=$wakeup_lat+$switchtime_lat ...' >> event3/trigger 1764 17652.2.2 Synthetic Events 1766---------------------- 1767 1768Synthetic events are user-defined events generated from hist trigger 1769variables or fields associated with one or more other events. Their 1770purpose is to provide a mechanism for displaying data spanning 1771multiple events consistent with the existing and already familiar 1772usage for normal events. 1773 1774To define a synthetic event, the user writes a simple specification 1775consisting of the name of the new event along with one or more 1776variables and their types, which can be any valid field type, 1777separated by semicolons, to the tracing/synthetic_events file. 1778 1779For instance, the following creates a new event named 'wakeup_latency' 1780with 3 fields: lat, pid, and prio. Each of those fields is simply a 1781variable reference to a variable on another event:: 1782 1783 # echo 'wakeup_latency \ 1784 u64 lat; \ 1785 pid_t pid; \ 1786 int prio' >> \ 1787 /sys/kernel/debug/tracing/synthetic_events 1788 1789Reading the tracing/synthetic_events file lists all the currently 1790defined synthetic events, in this case the event defined above:: 1791 1792 # cat /sys/kernel/debug/tracing/synthetic_events 1793 wakeup_latency u64 lat; pid_t pid; int prio 1794 1795An existing synthetic event definition can be removed by prepending 1796the command that defined it with a '!':: 1797 1798 # echo '!wakeup_latency u64 lat pid_t pid int prio' >> \ 1799 /sys/kernel/debug/tracing/synthetic_events 1800 1801At this point, there isn't yet an actual 'wakeup_latency' event 1802instantiated in the event subsystem - for this to happen, a 'hist 1803trigger action' needs to be instantiated and bound to actual fields 1804and variables defined on other events (see Section 2.2.3 below on 1805how that is done using hist trigger 'onmatch' action). Once that is 1806done, the 'wakeup_latency' synthetic event instance is created. 1807 1808A histogram can now be defined for the new synthetic event:: 1809 1810 # echo 'hist:keys=pid,prio,lat.log2:sort=pid,lat' >> \ 1811 /sys/kernel/debug/tracing/events/synthetic/wakeup_latency/trigger 1812 1813The new event is created under the tracing/events/synthetic/ directory 1814and looks and behaves just like any other event:: 1815 1816 # ls /sys/kernel/debug/tracing/events/synthetic/wakeup_latency 1817 enable filter format hist id trigger 1818 1819Like any other event, once a histogram is enabled for the event, the 1820output can be displayed by reading the event's 'hist' file. 1821 18222.2.3 Hist trigger 'handlers' and 'actions' 1823------------------------------------------- 1824 1825A hist trigger 'action' is a function that's executed (in most cases 1826conditionally) whenever a histogram entry is added or updated. 1827 1828When a histogram entry is added or updated, a hist trigger 'handler' 1829is what decides whether the corresponding action is actually invoked 1830or not. 1831 1832Hist trigger handlers and actions are paired together in the general 1833form: 1834 1835 <handler>.<action> 1836 1837To specify a handler.action pair for a given event, simply specify 1838that handler.action pair between colons in the hist trigger 1839specification. 1840 1841In theory, any handler can be combined with any action, but in 1842practice, not every handler.action combination is currently supported; 1843if a given handler.action combination isn't supported, the hist 1844trigger will fail with -EINVAL; 1845 1846The default 'handler.action' if none is explicitly specified is as it 1847always has been, to simply update the set of values associated with an 1848entry. Some applications, however, may want to perform additional 1849actions at that point, such as generate another event, or compare and 1850save a maximum. 1851 1852The supported handlers and actions are listed below, and each is 1853described in more detail in the following paragraphs, in the context 1854of descriptions of some common and useful handler.action combinations. 1855 1856The available handlers are: 1857 1858 - onmatch(matching.event) - invoke action on any addition or update 1859 - onmax(var) - invoke action if var exceeds current max 1860 - onchange(var) - invoke action if var changes 1861 1862The available actions are: 1863 1864 - trace(<synthetic_event_name>,param list) - generate synthetic event 1865 - save(field,...) - save current event fields 1866 - snapshot() - snapshot the trace buffer 1867 1868The following commonly-used handler.action pairs are available: 1869 1870 - onmatch(matching.event).trace(<synthetic_event_name>,param list) 1871 1872 The 'onmatch(matching.event).trace(<synthetic_event_name>,param 1873 list)' hist trigger action is invoked whenever an event matches 1874 and the histogram entry would be added or updated. It causes the 1875 named synthetic event to be generated with the values given in the 1876 'param list'. The result is the generation of a synthetic event 1877 that consists of the values contained in those variables at the 1878 time the invoking event was hit. For example, if the synthetic 1879 event name is 'wakeup_latency', a wakeup_latency event is 1880 generated using onmatch(event).trace(wakeup_latency,arg1,arg2). 1881 1882 There is also an equivalent alternative form available for 1883 generating synthetic events. In this form, the synthetic event 1884 name is used as if it were a function name. For example, using 1885 the 'wakeup_latency' synthetic event name again, the 1886 wakeup_latency event would be generated by invoking it as if it 1887 were a function call, with the event field values passed in as 1888 arguments: onmatch(event).wakeup_latency(arg1,arg2). The syntax 1889 for this form is: 1890 1891 onmatch(matching.event).<synthetic_event_name>(param list) 1892 1893 In either case, the 'param list' consists of one or more 1894 parameters which may be either variables or fields defined on 1895 either the 'matching.event' or the target event. The variables or 1896 fields specified in the param list may be either fully-qualified 1897 or unqualified. If a variable is specified as unqualified, it 1898 must be unique between the two events. A field name used as a 1899 param can be unqualified if it refers to the target event, but 1900 must be fully qualified if it refers to the matching event. A 1901 fully-qualified name is of the form 'system.event_name.$var_name' 1902 or 'system.event_name.field'. 1903 1904 The 'matching.event' specification is simply the fully qualified 1905 event name of the event that matches the target event for the 1906 onmatch() functionality, in the form 'system.event_name'. Histogram 1907 keys of both events are compared to find if events match. In case 1908 multiple histogram keys are used, they all must match in the specified 1909 order. 1910 1911 Finally, the number and type of variables/fields in the 'param 1912 list' must match the number and types of the fields in the 1913 synthetic event being generated. 1914 1915 As an example the below defines a simple synthetic event and uses 1916 a variable defined on the sched_wakeup_new event as a parameter 1917 when invoking the synthetic event. Here we define the synthetic 1918 event:: 1919 1920 # echo 'wakeup_new_test pid_t pid' >> \ 1921 /sys/kernel/debug/tracing/synthetic_events 1922 1923 # cat /sys/kernel/debug/tracing/synthetic_events 1924 wakeup_new_test pid_t pid 1925 1926 The following hist trigger both defines the missing testpid 1927 variable and specifies an onmatch() action that generates a 1928 wakeup_new_test synthetic event whenever a sched_wakeup_new event 1929 occurs, which because of the 'if comm == "cyclictest"' filter only 1930 happens when the executable is cyclictest:: 1931 1932 # echo 'hist:keys=$testpid:testpid=pid:onmatch(sched.sched_wakeup_new).\ 1933 wakeup_new_test($testpid) if comm=="cyclictest"' >> \ 1934 /sys/kernel/debug/tracing/events/sched/sched_wakeup_new/trigger 1935 1936 Or, equivalently, using the 'trace' keyword syntax: 1937 1938 # echo 'hist:keys=$testpid:testpid=pid:onmatch(sched.sched_wakeup_new).\ 1939 trace(wakeup_new_test,$testpid) if comm=="cyclictest"' >> \ 1940 /sys/kernel/debug/tracing/events/sched/sched_wakeup_new/trigger 1941 1942 Creating and displaying a histogram based on those events is now 1943 just a matter of using the fields and new synthetic event in the 1944 tracing/events/synthetic directory, as usual:: 1945 1946 # echo 'hist:keys=pid:sort=pid' >> \ 1947 /sys/kernel/debug/tracing/events/synthetic/wakeup_new_test/trigger 1948 1949 Running 'cyclictest' should cause wakeup_new events to generate 1950 wakeup_new_test synthetic events which should result in histogram 1951 output in the wakeup_new_test event's hist file:: 1952 1953 # cat /sys/kernel/debug/tracing/events/synthetic/wakeup_new_test/hist 1954 1955 A more typical usage would be to use two events to calculate a 1956 latency. The following example uses a set of hist triggers to 1957 produce a 'wakeup_latency' histogram. 1958 1959 First, we define a 'wakeup_latency' synthetic event:: 1960 1961 # echo 'wakeup_latency u64 lat; pid_t pid; int prio' >> \ 1962 /sys/kernel/debug/tracing/synthetic_events 1963 1964 Next, we specify that whenever we see a sched_waking event for a 1965 cyclictest thread, save the timestamp in a 'ts0' variable:: 1966 1967 # echo 'hist:keys=$saved_pid:saved_pid=pid:ts0=common_timestamp.usecs \ 1968 if comm=="cyclictest"' >> \ 1969 /sys/kernel/debug/tracing/events/sched/sched_waking/trigger 1970 1971 Then, when the corresponding thread is actually scheduled onto the 1972 CPU by a sched_switch event (saved_pid matches next_pid), calculate 1973 the latency and use that along with another variable and an event field 1974 to generate a wakeup_latency synthetic event:: 1975 1976 # echo 'hist:keys=next_pid:wakeup_lat=common_timestamp.usecs-$ts0:\ 1977 onmatch(sched.sched_waking).wakeup_latency($wakeup_lat,\ 1978 $saved_pid,next_prio) if next_comm=="cyclictest"' >> \ 1979 /sys/kernel/debug/tracing/events/sched/sched_switch/trigger 1980 1981 We also need to create a histogram on the wakeup_latency synthetic 1982 event in order to aggregate the generated synthetic event data:: 1983 1984 # echo 'hist:keys=pid,prio,lat:sort=pid,lat' >> \ 1985 /sys/kernel/debug/tracing/events/synthetic/wakeup_latency/trigger 1986 1987 Finally, once we've run cyclictest to actually generate some 1988 events, we can see the output by looking at the wakeup_latency 1989 synthetic event's hist file:: 1990 1991 # cat /sys/kernel/debug/tracing/events/synthetic/wakeup_latency/hist 1992 1993 - onmax(var).save(field,.. .) 1994 1995 The 'onmax(var).save(field,...)' hist trigger action is invoked 1996 whenever the value of 'var' associated with a histogram entry 1997 exceeds the current maximum contained in that variable. 1998 1999 The end result is that the trace event fields specified as the 2000 onmax.save() params will be saved if 'var' exceeds the current 2001 maximum for that hist trigger entry. This allows context from the 2002 event that exhibited the new maximum to be saved for later 2003 reference. When the histogram is displayed, additional fields 2004 displaying the saved values will be printed. 2005 2006 As an example the below defines a couple of hist triggers, one for 2007 sched_waking and another for sched_switch, keyed on pid. Whenever 2008 a sched_waking occurs, the timestamp is saved in the entry 2009 corresponding to the current pid, and when the scheduler switches 2010 back to that pid, the timestamp difference is calculated. If the 2011 resulting latency, stored in wakeup_lat, exceeds the current 2012 maximum latency, the values specified in the save() fields are 2013 recorded:: 2014 2015 # echo 'hist:keys=pid:ts0=common_timestamp.usecs \ 2016 if comm=="cyclictest"' >> \ 2017 /sys/kernel/debug/tracing/events/sched/sched_waking/trigger 2018 2019 # echo 'hist:keys=next_pid:\ 2020 wakeup_lat=common_timestamp.usecs-$ts0:\ 2021 onmax($wakeup_lat).save(next_comm,prev_pid,prev_prio,prev_comm) \ 2022 if next_comm=="cyclictest"' >> \ 2023 /sys/kernel/debug/tracing/events/sched/sched_switch/trigger 2024 2025 When the histogram is displayed, the max value and the saved 2026 values corresponding to the max are displayed following the rest 2027 of the fields:: 2028 2029 # cat /sys/kernel/debug/tracing/events/sched/sched_switch/hist 2030 { next_pid: 2255 } hitcount: 239 2031 common_timestamp-ts0: 0 2032 max: 27 2033 next_comm: cyclictest 2034 prev_pid: 0 prev_prio: 120 prev_comm: swapper/1 2035 2036 { next_pid: 2256 } hitcount: 2355 2037 common_timestamp-ts0: 0 2038 max: 49 next_comm: cyclictest 2039 prev_pid: 0 prev_prio: 120 prev_comm: swapper/0 2040 2041 Totals: 2042 Hits: 12970 2043 Entries: 2 2044 Dropped: 0 2045 2046 - onmax(var).snapshot() 2047 2048 The 'onmax(var).snapshot()' hist trigger action is invoked 2049 whenever the value of 'var' associated with a histogram entry 2050 exceeds the current maximum contained in that variable. 2051 2052 The end result is that a global snapshot of the trace buffer will 2053 be saved in the tracing/snapshot file if 'var' exceeds the current 2054 maximum for any hist trigger entry. 2055 2056 Note that in this case the maximum is a global maximum for the 2057 current trace instance, which is the maximum across all buckets of 2058 the histogram. The key of the specific trace event that caused 2059 the global maximum and the global maximum itself are displayed, 2060 along with a message stating that a snapshot has been taken and 2061 where to find it. The user can use the key information displayed 2062 to locate the corresponding bucket in the histogram for even more 2063 detail. 2064 2065 As an example the below defines a couple of hist triggers, one for 2066 sched_waking and another for sched_switch, keyed on pid. Whenever 2067 a sched_waking event occurs, the timestamp is saved in the entry 2068 corresponding to the current pid, and when the scheduler switches 2069 back to that pid, the timestamp difference is calculated. If the 2070 resulting latency, stored in wakeup_lat, exceeds the current 2071 maximum latency, a snapshot is taken. As part of the setup, all 2072 the scheduler events are also enabled, which are the events that 2073 will show up in the snapshot when it is taken at some point: 2074 2075 # echo 1 > /sys/kernel/debug/tracing/events/sched/enable 2076 2077 # echo 'hist:keys=pid:ts0=common_timestamp.usecs \ 2078 if comm=="cyclictest"' >> \ 2079 /sys/kernel/debug/tracing/events/sched/sched_waking/trigger 2080 2081 # echo 'hist:keys=next_pid:wakeup_lat=common_timestamp.usecs-$ts0: \ 2082 onmax($wakeup_lat).save(next_prio,next_comm,prev_pid,prev_prio, \ 2083 prev_comm):onmax($wakeup_lat).snapshot() \ 2084 if next_comm=="cyclictest"' >> \ 2085 /sys/kernel/debug/tracing/events/sched/sched_switch/trigger 2086 2087 When the histogram is displayed, for each bucket the max value 2088 and the saved values corresponding to the max are displayed 2089 following the rest of the fields. 2090 2091 If a snapshot was taken, there is also a message indicating that, 2092 along with the value and event that triggered the global maximum: 2093 2094 # cat /sys/kernel/debug/tracing/events/sched/sched_switch/hist 2095 { next_pid: 2101 } hitcount: 200 2096 max: 52 next_prio: 120 next_comm: cyclictest \ 2097 prev_pid: 0 prev_prio: 120 prev_comm: swapper/6 2098 2099 { next_pid: 2103 } hitcount: 1326 2100 max: 572 next_prio: 19 next_comm: cyclictest \ 2101 prev_pid: 0 prev_prio: 120 prev_comm: swapper/1 2102 2103 { next_pid: 2102 } hitcount: 1982 \ 2104 max: 74 next_prio: 19 next_comm: cyclictest \ 2105 prev_pid: 0 prev_prio: 120 prev_comm: swapper/5 2106 2107 Snapshot taken (see tracing/snapshot). Details: 2108 triggering value { onmax($wakeup_lat) }: 572 \ 2109 triggered by event with key: { next_pid: 2103 } 2110 2111 Totals: 2112 Hits: 3508 2113 Entries: 3 2114 Dropped: 0 2115 2116 In the above case, the event that triggered the global maximum has 2117 the key with next_pid == 2103. If you look at the bucket that has 2118 2103 as the key, you'll find the additional values save()'d along 2119 with the local maximum for that bucket, which should be the same 2120 as the global maximum (since that was the same value that 2121 triggered the global snapshot). 2122 2123 And finally, looking at the snapshot data should show at or near 2124 the end the event that triggered the snapshot (in this case you 2125 can verify the timestamps between the sched_waking and 2126 sched_switch events, which should match the time displayed in the 2127 global maximum):: 2128 2129 # cat /sys/kernel/debug/tracing/snapshot 2130 2131 <...>-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=120 2132 <idle>-0 [005] d.h3 309.873611: sched_waking: comm=cyclictest pid=2102 prio=19 target_cpu=005 2133 <idle>-0 [005] dNh4 309.873613: sched_wakeup: comm=cyclictest pid=2102 prio=19 target_cpu=005 2134 <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=19 2135 <...>-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=120 2136 <idle>-0 [005] d.h3 309.874624: sched_waking: comm=cyclictest pid=2102 prio=19 target_cpu=005 2137 <idle>-0 [005] dNh4 309.874626: sched_wakeup: comm=cyclictest pid=2102 prio=19 target_cpu=005 2138 <idle>-0 [005] dNh3 309.874628: sched_waking: comm=cyclictest pid=2103 prio=19 target_cpu=005 2139 <idle>-0 [005] dNh4 309.874630: sched_wakeup: comm=cyclictest pid=2103 prio=19 target_cpu=005 2140 <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=19 2141 <idle>-0 [004] d.h3 309.874757: sched_waking: comm=gnome-terminal- pid=1699 prio=120 target_cpu=004 2142 <idle>-0 [004] dNh4 309.874762: sched_wakeup: comm=gnome-terminal- pid=1699 prio=120 target_cpu=004 2143 <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=120 2144 gnome-terminal--1699 [004] d.h2 309.874941: sched_stat_runtime: comm=gnome-terminal- pid=1699 runtime=180706 [ns] vruntime=1126870572 [ns] 2145 <idle>-0 [003] d.s4 309.874956: sched_waking: comm=rcu_sched pid=9 prio=120 target_cpu=007 2146 <idle>-0 [003] d.s5 309.874960: sched_wake_idle_without_ipi: cpu=7 2147 <idle>-0 [003] d.s5 309.874961: sched_wakeup: comm=rcu_sched pid=9 prio=120 target_cpu=007 2148 <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=120 2149 rcu_sched-9 [007] d..3 309.874973: sched_stat_runtime: comm=rcu_sched pid=9 runtime=13646 [ns] vruntime=22531430286 [ns] 2150 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=120 2151 <...>-2102 [005] d..4 309.874994: sched_migrate_task: comm=cyclictest pid=2103 prio=19 orig_cpu=5 dest_cpu=1 2152 <...>-2102 [005] d..4 309.875185: sched_wake_idle_without_ipi: cpu=1 2153 <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=19 2154 2155 - onchange(var).save(field,.. .) 2156 2157 The 'onchange(var).save(field,...)' hist trigger action is invoked 2158 whenever the value of 'var' associated with a histogram entry 2159 changes. 2160 2161 The end result is that the trace event fields specified as the 2162 onchange.save() params will be saved if 'var' changes for that 2163 hist trigger entry. This allows context from the event that 2164 changed the value to be saved for later reference. When the 2165 histogram is displayed, additional fields displaying the saved 2166 values will be printed. 2167 2168 - onchange(var).snapshot() 2169 2170 The 'onchange(var).snapshot()' hist trigger action is invoked 2171 whenever the value of 'var' associated with a histogram entry 2172 changes. 2173 2174 The end result is that a global snapshot of the trace buffer will 2175 be saved in the tracing/snapshot file if 'var' changes for any 2176 hist trigger entry. 2177 2178 Note that in this case the changed value is a global variable 2179 associated with current trace instance. The key of the specific 2180 trace event that caused the value to change and the global value 2181 itself are displayed, along with a message stating that a snapshot 2182 has been taken and where to find it. The user can use the key 2183 information displayed to locate the corresponding bucket in the 2184 histogram for even more detail. 2185 2186 As an example the below defines a hist trigger on the tcp_probe 2187 event, keyed on dport. Whenever a tcp_probe event occurs, the 2188 cwnd field is checked against the current value stored in the 2189 $cwnd variable. If the value has changed, a snapshot is taken. 2190 As part of the setup, all the scheduler and tcp events are also 2191 enabled, which are the events that will show up in the snapshot 2192 when it is taken at some point: 2193 2194 # echo 1 > /sys/kernel/debug/tracing/events/sched/enable 2195 # echo 1 > /sys/kernel/debug/tracing/events/tcp/enable 2196 2197 # echo 'hist:keys=dport:cwnd=snd_cwnd: \ 2198 onchange($cwnd).save(snd_wnd,srtt,rcv_wnd): \ 2199 onchange($cwnd).snapshot()' >> \ 2200 /sys/kernel/debug/tracing/events/tcp/tcp_probe/trigger 2201 2202 When the histogram is displayed, for each bucket the tracked value 2203 and the saved values corresponding to that value are displayed 2204 following the rest of the fields. 2205 2206 If a snapshot was taken, there is also a message indicating that, 2207 along with the value and event that triggered the snapshot:: 2208 2209 # cat /sys/kernel/debug/tracing/events/tcp/tcp_probe/hist 2210 2211 { dport: 1521 } hitcount: 8 2212 changed: 10 snd_wnd: 35456 srtt: 154262 rcv_wnd: 42112 2213 2214 { dport: 80 } hitcount: 23 2215 changed: 10 snd_wnd: 28960 srtt: 19604 rcv_wnd: 29312 2216 2217 { dport: 9001 } hitcount: 172 2218 changed: 10 snd_wnd: 48384 srtt: 260444 rcv_wnd: 55168 2219 2220 { dport: 443 } hitcount: 211 2221 changed: 10 snd_wnd: 26960 srtt: 17379 rcv_wnd: 28800 2222 2223 Snapshot taken (see tracing/snapshot). Details:: 2224 2225 triggering value { onchange($cwnd) }: 10 2226 triggered by event with key: { dport: 80 } 2227 2228 Totals: 2229 Hits: 414 2230 Entries: 4 2231 Dropped: 0 2232 2233 In the above case, the event that triggered the snapshot has the 2234 key with dport == 80. If you look at the bucket that has 80 as 2235 the key, you'll find the additional values save()'d along with the 2236 changed value for that bucket, which should be the same as the 2237 global changed value (since that was the same value that triggered 2238 the global snapshot). 2239 2240 And finally, looking at the snapshot data should show at or near 2241 the end the event that triggered the snapshot:: 2242 2243 # cat /sys/kernel/debug/tracing/snapshot 2244 2245 gnome-shell-1261 [006] dN.3 49.823113: sched_stat_runtime: comm=gnome-shell pid=1261 runtime=49347 [ns] vruntime=1835730389 [ns] 2246 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=120 2247 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=120 2248 kworker/3:2-135 [003] d..3 49.823118: sched_stat_runtime: comm=kworker/3:2 pid=135 runtime=5339 [ns] vruntime=17815800388 [ns] 2249 kworker/6:2-387 [006] d..3 49.823120: sched_stat_runtime: comm=kworker/6:2 pid=387 runtime=9594 [ns] vruntime=14589605367 [ns] 2250 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=120 2251 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=120 2252 <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=29312 2253 22543. User space creating a trigger 2255-------------------------------- 2256 2257Writing into /sys/kernel/tracing/trace_marker writes into the ftrace 2258ring buffer. This can also act like an event, by writing into the trigger 2259file located in /sys/kernel/tracing/events/ftrace/print/ 2260 2261Modifying cyclictest to write into the trace_marker file before it sleeps 2262and after it wakes up, something like this:: 2263 2264 static void traceputs(char *str) 2265 { 2266 /* tracemark_fd is the trace_marker file descriptor */ 2267 if (tracemark_fd < 0) 2268 return; 2269 /* write the tracemark message */ 2270 write(tracemark_fd, str, strlen(str)); 2271 } 2272 2273And later add something like:: 2274 2275 traceputs("start"); 2276 clock_nanosleep(...); 2277 traceputs("end"); 2278 2279We can make a histogram from this:: 2280 2281 # cd /sys/kernel/tracing 2282 # echo 'latency u64 lat' > synthetic_events 2283 # echo 'hist:keys=common_pid:ts0=common_timestamp.usecs if buf == "start"' > events/ftrace/print/trigger 2284 # echo 'hist:keys=common_pid:lat=common_timestamp.usecs-$ts0:onmatch(ftrace.print).latency($lat) if buf == "end"' >> events/ftrace/print/trigger 2285 # echo 'hist:keys=lat,common_pid:sort=lat' > events/synthetic/latency/trigger 2286 2287The above created a synthetic event called "latency" and two histograms 2288against the trace_marker, one gets triggered when "start" is written into the 2289trace_marker file and the other when "end" is written. If the pids match, then 2290it will call the "latency" synthetic event with the calculated latency as its 2291parameter. Finally, a histogram is added to the latency synthetic event to 2292record the calculated latency along with the pid. 2293 2294Now running cyclictest with:: 2295 2296 # ./cyclictest -p80 -d0 -i250 -n -a -t --tracemark -b 1000 2297 2298 -p80 : run threads at priority 80 2299 -d0 : have all threads run at the same interval 2300 -i250 : start the interval at 250 microseconds (all threads will do this) 2301 -n : sleep with nanosleep 2302 -a : affine all threads to a separate CPU 2303 -t : one thread per available CPU 2304 --tracemark : enable trace mark writing 2305 -b 1000 : stop if any latency is greater than 1000 microseconds 2306 2307Note, the -b 1000 is used just to make --tracemark available. 2308 2309Then we can see the histogram created by this with:: 2310 2311 # cat events/synthetic/latency/hist 2312 # event histogram 2313 # 2314 # trigger info: hist:keys=lat,common_pid:vals=hitcount:sort=lat:size=2048 [active] 2315 # 2316 2317 { lat: 107, common_pid: 2039 } hitcount: 1 2318 { lat: 122, common_pid: 2041 } hitcount: 1 2319 { lat: 166, common_pid: 2039 } hitcount: 1 2320 { lat: 174, common_pid: 2039 } hitcount: 1 2321 { lat: 194, common_pid: 2041 } hitcount: 1 2322 { lat: 196, common_pid: 2036 } hitcount: 1 2323 { lat: 197, common_pid: 2038 } hitcount: 1 2324 { lat: 198, common_pid: 2039 } hitcount: 1 2325 { lat: 199, common_pid: 2039 } hitcount: 1 2326 { lat: 200, common_pid: 2041 } hitcount: 1 2327 { lat: 201, common_pid: 2039 } hitcount: 2 2328 { lat: 202, common_pid: 2038 } hitcount: 1 2329 { lat: 202, common_pid: 2043 } hitcount: 1 2330 { lat: 203, common_pid: 2039 } hitcount: 1 2331 { lat: 203, common_pid: 2036 } hitcount: 1 2332 { lat: 203, common_pid: 2041 } hitcount: 1 2333 { lat: 206, common_pid: 2038 } hitcount: 2 2334 { lat: 207, common_pid: 2039 } hitcount: 1 2335 { lat: 207, common_pid: 2036 } hitcount: 1 2336 { lat: 208, common_pid: 2040 } hitcount: 1 2337 { lat: 209, common_pid: 2043 } hitcount: 1 2338 { lat: 210, common_pid: 2039 } hitcount: 1 2339 { lat: 211, common_pid: 2039 } hitcount: 4 2340 { lat: 212, common_pid: 2043 } hitcount: 1 2341 { lat: 212, common_pid: 2039 } hitcount: 2 2342 { lat: 213, common_pid: 2039 } hitcount: 1 2343 { lat: 214, common_pid: 2038 } hitcount: 1 2344 { lat: 214, common_pid: 2039 } hitcount: 2 2345 { lat: 214, common_pid: 2042 } hitcount: 1 2346 { lat: 215, common_pid: 2039 } hitcount: 1 2347 { lat: 217, common_pid: 2036 } hitcount: 1 2348 { lat: 217, common_pid: 2040 } hitcount: 1 2349 { lat: 217, common_pid: 2039 } hitcount: 1 2350 { lat: 218, common_pid: 2039 } hitcount: 6 2351 { lat: 219, common_pid: 2039 } hitcount: 9 2352 { lat: 220, common_pid: 2039 } hitcount: 11 2353 { lat: 221, common_pid: 2039 } hitcount: 5 2354 { lat: 221, common_pid: 2042 } hitcount: 1 2355 { lat: 222, common_pid: 2039 } hitcount: 7 2356 { lat: 223, common_pid: 2036 } hitcount: 1 2357 { lat: 223, common_pid: 2039 } hitcount: 3 2358 { lat: 224, common_pid: 2039 } hitcount: 4 2359 { lat: 224, common_pid: 2037 } hitcount: 1 2360 { lat: 224, common_pid: 2036 } hitcount: 2 2361 { lat: 225, common_pid: 2039 } hitcount: 5 2362 { lat: 225, common_pid: 2042 } hitcount: 1 2363 { lat: 226, common_pid: 2039 } hitcount: 7 2364 { lat: 226, common_pid: 2036 } hitcount: 4 2365 { lat: 227, common_pid: 2039 } hitcount: 6 2366 { lat: 227, common_pid: 2036 } hitcount: 12 2367 { lat: 227, common_pid: 2043 } hitcount: 1 2368 { lat: 228, common_pid: 2039 } hitcount: 7 2369 { lat: 228, common_pid: 2036 } hitcount: 14 2370 { lat: 229, common_pid: 2039 } hitcount: 9 2371 { lat: 229, common_pid: 2036 } hitcount: 8 2372 { lat: 229, common_pid: 2038 } hitcount: 1 2373 { lat: 230, common_pid: 2039 } hitcount: 11 2374 { lat: 230, common_pid: 2036 } hitcount: 6 2375 { lat: 230, common_pid: 2043 } hitcount: 1 2376 { lat: 230, common_pid: 2042 } hitcount: 2 2377 { lat: 231, common_pid: 2041 } hitcount: 1 2378 { lat: 231, common_pid: 2036 } hitcount: 6 2379 { lat: 231, common_pid: 2043 } hitcount: 1 2380 { lat: 231, common_pid: 2039 } hitcount: 8 2381 { lat: 232, common_pid: 2037 } hitcount: 1 2382 { lat: 232, common_pid: 2039 } hitcount: 6 2383 { lat: 232, common_pid: 2040 } hitcount: 2 2384 { lat: 232, common_pid: 2036 } hitcount: 5 2385 { lat: 232, common_pid: 2043 } hitcount: 1 2386 { lat: 233, common_pid: 2036 } hitcount: 5 2387 { lat: 233, common_pid: 2039 } hitcount: 11 2388 { lat: 234, common_pid: 2039 } hitcount: 4 2389 { lat: 234, common_pid: 2038 } hitcount: 2 2390 { lat: 234, common_pid: 2043 } hitcount: 2 2391 { lat: 234, common_pid: 2036 } hitcount: 11 2392 { lat: 234, common_pid: 2040 } hitcount: 1 2393 { lat: 235, common_pid: 2037 } hitcount: 2 2394 { lat: 235, common_pid: 2036 } hitcount: 8 2395 { lat: 235, common_pid: 2043 } hitcount: 2 2396 { lat: 235, common_pid: 2039 } hitcount: 5 2397 { lat: 235, common_pid: 2042 } hitcount: 2 2398 { lat: 235, common_pid: 2040 } hitcount: 4 2399 { lat: 235, common_pid: 2041 } hitcount: 1 2400 { lat: 236, common_pid: 2036 } hitcount: 7 2401 { lat: 236, common_pid: 2037 } hitcount: 1 2402 { lat: 236, common_pid: 2041 } hitcount: 5 2403 { lat: 236, common_pid: 2039 } hitcount: 3 2404 { lat: 236, common_pid: 2043 } hitcount: 9 2405 { lat: 236, common_pid: 2040 } hitcount: 7 2406 { lat: 237, common_pid: 2037 } hitcount: 1 2407 { lat: 237, common_pid: 2040 } hitcount: 1 2408 { lat: 237, common_pid: 2036 } hitcount: 9 2409 { lat: 237, common_pid: 2039 } hitcount: 3 2410 { lat: 237, common_pid: 2043 } hitcount: 8 2411 { lat: 237, common_pid: 2042 } hitcount: 2 2412 { lat: 237, common_pid: 2041 } hitcount: 2 2413 { lat: 238, common_pid: 2043 } hitcount: 10 2414 { lat: 238, common_pid: 2040 } hitcount: 1 2415 { lat: 238, common_pid: 2037 } hitcount: 9 2416 { lat: 238, common_pid: 2038 } hitcount: 1 2417 { lat: 238, common_pid: 2039 } hitcount: 1 2418 { lat: 238, common_pid: 2042 } hitcount: 3 2419 { lat: 238, common_pid: 2036 } hitcount: 7 2420 { lat: 239, common_pid: 2041 } hitcount: 1 2421 { lat: 239, common_pid: 2043 } hitcount: 11 2422 { lat: 239, common_pid: 2037 } hitcount: 11 2423 { lat: 239, common_pid: 2038 } hitcount: 6 2424 { lat: 239, common_pid: 2036 } hitcount: 7 2425 { lat: 239, common_pid: 2040 } hitcount: 1 2426 { lat: 239, common_pid: 2042 } hitcount: 9 2427 { lat: 240, common_pid: 2037 } hitcount: 29 2428 { lat: 240, common_pid: 2043 } hitcount: 15 2429 { lat: 240, common_pid: 2040 } hitcount: 44 2430 { lat: 240, common_pid: 2039 } hitcount: 1 2431 { lat: 240, common_pid: 2041 } hitcount: 2 2432 { lat: 240, common_pid: 2038 } hitcount: 1 2433 { lat: 240, common_pid: 2036 } hitcount: 10 2434 { lat: 240, common_pid: 2042 } hitcount: 13 2435 { lat: 241, common_pid: 2036 } hitcount: 21 2436 { lat: 241, common_pid: 2041 } hitcount: 36 2437 { lat: 241, common_pid: 2037 } hitcount: 34 2438 { lat: 241, common_pid: 2042 } hitcount: 14 2439 { lat: 241, common_pid: 2040 } hitcount: 94 2440 { lat: 241, common_pid: 2039 } hitcount: 12 2441 { lat: 241, common_pid: 2038 } hitcount: 2 2442 { lat: 241, common_pid: 2043 } hitcount: 28 2443 { lat: 242, common_pid: 2040 } hitcount: 109 2444 { lat: 242, common_pid: 2041 } hitcount: 506 2445 { lat: 242, common_pid: 2039 } hitcount: 155 2446 { lat: 242, common_pid: 2042 } hitcount: 21 2447 { lat: 242, common_pid: 2037 } hitcount: 52 2448 { lat: 242, common_pid: 2043 } hitcount: 21 2449 { lat: 242, common_pid: 2036 } hitcount: 16 2450 { lat: 242, common_pid: 2038 } hitcount: 156 2451 { lat: 243, common_pid: 2037 } hitcount: 46 2452 { lat: 243, common_pid: 2039 } hitcount: 40 2453 { lat: 243, common_pid: 2042 } hitcount: 119 2454 { lat: 243, common_pid: 2041 } hitcount: 611 2455 { lat: 243, common_pid: 2036 } hitcount: 69 2456 { lat: 243, common_pid: 2038 } hitcount: 784 2457 { lat: 243, common_pid: 2040 } hitcount: 323 2458 { lat: 243, common_pid: 2043 } hitcount: 14 2459 { lat: 244, common_pid: 2043 } hitcount: 35 2460 { lat: 244, common_pid: 2042 } hitcount: 305 2461 { lat: 244, common_pid: 2039 } hitcount: 8 2462 { lat: 244, common_pid: 2040 } hitcount: 4515 2463 { lat: 244, common_pid: 2038 } hitcount: 371 2464 { lat: 244, common_pid: 2037 } hitcount: 31 2465 { lat: 244, common_pid: 2036 } hitcount: 114 2466 { lat: 244, common_pid: 2041 } hitcount: 3396 2467 { lat: 245, common_pid: 2036 } hitcount: 700 2468 { lat: 245, common_pid: 2041 } hitcount: 2772 2469 { lat: 245, common_pid: 2037 } hitcount: 268 2470 { lat: 245, common_pid: 2039 } hitcount: 472 2471 { lat: 245, common_pid: 2038 } hitcount: 2758 2472 { lat: 245, common_pid: 2042 } hitcount: 3833 2473 { lat: 245, common_pid: 2040 } hitcount: 3105 2474 { lat: 245, common_pid: 2043 } hitcount: 645 2475 { lat: 246, common_pid: 2038 } hitcount: 3451 2476 { lat: 246, common_pid: 2041 } hitcount: 142 2477 { lat: 246, common_pid: 2037 } hitcount: 5101 2478 { lat: 246, common_pid: 2040 } hitcount: 68 2479 { lat: 246, common_pid: 2043 } hitcount: 5099 2480 { lat: 246, common_pid: 2039 } hitcount: 5608 2481 { lat: 246, common_pid: 2042 } hitcount: 3723 2482 { lat: 246, common_pid: 2036 } hitcount: 4738 2483 { lat: 247, common_pid: 2042 } hitcount: 312 2484 { lat: 247, common_pid: 2043 } hitcount: 2385 2485 { lat: 247, common_pid: 2041 } hitcount: 452 2486 { lat: 247, common_pid: 2038 } hitcount: 792 2487 { lat: 247, common_pid: 2040 } hitcount: 78 2488 { lat: 247, common_pid: 2036 } hitcount: 2375 2489 { lat: 247, common_pid: 2039 } hitcount: 1834 2490 { lat: 247, common_pid: 2037 } hitcount: 2655 2491 { lat: 248, common_pid: 2037 } hitcount: 36 2492 { lat: 248, common_pid: 2042 } hitcount: 11 2493 { lat: 248, common_pid: 2038 } hitcount: 122 2494 { lat: 248, common_pid: 2036 } hitcount: 135 2495 { lat: 248, common_pid: 2039 } hitcount: 26 2496 { lat: 248, common_pid: 2041 } hitcount: 503 2497 { lat: 248, common_pid: 2043 } hitcount: 66 2498 { lat: 248, common_pid: 2040 } hitcount: 46 2499 { lat: 249, common_pid: 2037 } hitcount: 29 2500 { lat: 249, common_pid: 2038 } hitcount: 1 2501 { lat: 249, common_pid: 2043 } hitcount: 29 2502 { lat: 249, common_pid: 2039 } hitcount: 8 2503 { lat: 249, common_pid: 2042 } hitcount: 56 2504 { lat: 249, common_pid: 2040 } hitcount: 27 2505 { lat: 249, common_pid: 2041 } hitcount: 11 2506 { lat: 249, common_pid: 2036 } hitcount: 27 2507 { lat: 250, common_pid: 2038 } hitcount: 1 2508 { lat: 250, common_pid: 2036 } hitcount: 30 2509 { lat: 250, common_pid: 2040 } hitcount: 19 2510 { lat: 250, common_pid: 2043 } hitcount: 22 2511 { lat: 250, common_pid: 2042 } hitcount: 20 2512 { lat: 250, common_pid: 2041 } hitcount: 1 2513 { lat: 250, common_pid: 2039 } hitcount: 6 2514 { lat: 250, common_pid: 2037 } hitcount: 48 2515 { lat: 251, common_pid: 2037 } hitcount: 43 2516 { lat: 251, common_pid: 2039 } hitcount: 1 2517 { lat: 251, common_pid: 2036 } hitcount: 12 2518 { lat: 251, common_pid: 2042 } hitcount: 2 2519 { lat: 251, common_pid: 2041 } hitcount: 1 2520 { lat: 251, common_pid: 2043 } hitcount: 15 2521 { lat: 251, common_pid: 2040 } hitcount: 3 2522 { lat: 252, common_pid: 2040 } hitcount: 1 2523 { lat: 252, common_pid: 2036 } hitcount: 12 2524 { lat: 252, common_pid: 2037 } hitcount: 21 2525 { lat: 252, common_pid: 2043 } hitcount: 14 2526 { lat: 253, common_pid: 2037 } hitcount: 21 2527 { lat: 253, common_pid: 2039 } hitcount: 2 2528 { lat: 253, common_pid: 2036 } hitcount: 9 2529 { lat: 253, common_pid: 2043 } hitcount: 6 2530 { lat: 253, common_pid: 2040 } hitcount: 1 2531 { lat: 254, common_pid: 2036 } hitcount: 8 2532 { lat: 254, common_pid: 2043 } hitcount: 3 2533 { lat: 254, common_pid: 2041 } hitcount: 1 2534 { lat: 254, common_pid: 2042 } hitcount: 1 2535 { lat: 254, common_pid: 2039 } hitcount: 1 2536 { lat: 254, common_pid: 2037 } hitcount: 12 2537 { lat: 255, common_pid: 2043 } hitcount: 1 2538 { lat: 255, common_pid: 2037 } hitcount: 2 2539 { lat: 255, common_pid: 2036 } hitcount: 2 2540 { lat: 255, common_pid: 2039 } hitcount: 8 2541 { lat: 256, common_pid: 2043 } hitcount: 1 2542 { lat: 256, common_pid: 2036 } hitcount: 4 2543 { lat: 256, common_pid: 2039 } hitcount: 6 2544 { lat: 257, common_pid: 2039 } hitcount: 5 2545 { lat: 257, common_pid: 2036 } hitcount: 4 2546 { lat: 258, common_pid: 2039 } hitcount: 5 2547 { lat: 258, common_pid: 2036 } hitcount: 2 2548 { lat: 259, common_pid: 2036 } hitcount: 7 2549 { lat: 259, common_pid: 2039 } hitcount: 7 2550 { lat: 260, common_pid: 2036 } hitcount: 8 2551 { lat: 260, common_pid: 2039 } hitcount: 6 2552 { lat: 261, common_pid: 2036 } hitcount: 5 2553 { lat: 261, common_pid: 2039 } hitcount: 7 2554 { lat: 262, common_pid: 2039 } hitcount: 5 2555 { lat: 262, common_pid: 2036 } hitcount: 5 2556 { lat: 263, common_pid: 2039 } hitcount: 7 2557 { lat: 263, common_pid: 2036 } hitcount: 7 2558 { lat: 264, common_pid: 2039 } hitcount: 9 2559 { lat: 264, common_pid: 2036 } hitcount: 9 2560 { lat: 265, common_pid: 2036 } hitcount: 5 2561 { lat: 265, common_pid: 2039 } hitcount: 1 2562 { lat: 266, common_pid: 2036 } hitcount: 1 2563 { lat: 266, common_pid: 2039 } hitcount: 3 2564 { lat: 267, common_pid: 2036 } hitcount: 1 2565 { lat: 267, common_pid: 2039 } hitcount: 3 2566 { lat: 268, common_pid: 2036 } hitcount: 1 2567 { lat: 268, common_pid: 2039 } hitcount: 6 2568 { lat: 269, common_pid: 2036 } hitcount: 1 2569 { lat: 269, common_pid: 2043 } hitcount: 1 2570 { lat: 269, common_pid: 2039 } hitcount: 2 2571 { lat: 270, common_pid: 2040 } hitcount: 1 2572 { lat: 270, common_pid: 2039 } hitcount: 6 2573 { lat: 271, common_pid: 2041 } hitcount: 1 2574 { lat: 271, common_pid: 2039 } hitcount: 5 2575 { lat: 272, common_pid: 2039 } hitcount: 10 2576 { lat: 273, common_pid: 2039 } hitcount: 8 2577 { lat: 274, common_pid: 2039 } hitcount: 2 2578 { lat: 275, common_pid: 2039 } hitcount: 1 2579 { lat: 276, common_pid: 2039 } hitcount: 2 2580 { lat: 276, common_pid: 2037 } hitcount: 1 2581 { lat: 276, common_pid: 2038 } hitcount: 1 2582 { lat: 277, common_pid: 2039 } hitcount: 1 2583 { lat: 277, common_pid: 2042 } hitcount: 1 2584 { lat: 278, common_pid: 2039 } hitcount: 1 2585 { lat: 279, common_pid: 2039 } hitcount: 4 2586 { lat: 279, common_pid: 2043 } hitcount: 1 2587 { lat: 280, common_pid: 2039 } hitcount: 3 2588 { lat: 283, common_pid: 2036 } hitcount: 2 2589 { lat: 284, common_pid: 2039 } hitcount: 1 2590 { lat: 284, common_pid: 2043 } hitcount: 1 2591 { lat: 288, common_pid: 2039 } hitcount: 1 2592 { lat: 289, common_pid: 2039 } hitcount: 1 2593 { lat: 300, common_pid: 2039 } hitcount: 1 2594 { lat: 384, common_pid: 2039 } hitcount: 1 2595 2596 Totals: 2597 Hits: 67625 2598 Entries: 278 2599 Dropped: 0 2600 2601Note, the writes are around the sleep, so ideally they will all be of 250 2602microseconds. If you are wondering how there are several that are under 2603250 microseconds, that is because the way cyclictest works, is if one 2604iteration comes in late, the next one will set the timer to wake up less that 2605250. That is, if an iteration came in 50 microseconds late, the next wake up 2606will be at 200 microseconds. 2607 2608But this could easily be done in userspace. To make this even more 2609interesting, we can mix the histogram between events that happened in the 2610kernel with trace_marker:: 2611 2612 # cd /sys/kernel/tracing 2613 # echo 'latency u64 lat' > synthetic_events 2614 # echo 'hist:keys=pid:ts0=common_timestamp.usecs' > events/sched/sched_waking/trigger 2615 # echo 'hist:keys=common_pid:lat=common_timestamp.usecs-$ts0:onmatch(sched.sched_waking).latency($lat) if buf == "end"' > events/ftrace/print/trigger 2616 # echo 'hist:keys=lat,common_pid:sort=lat' > events/synthetic/latency/trigger 2617 2618The difference this time is that instead of using the trace_marker to start 2619the latency, the sched_waking event is used, matching the common_pid for the 2620trace_marker write with the pid that is being woken by sched_waking. 2621 2622After running cyclictest again with the same parameters, we now have:: 2623 2624 # cat events/synthetic/latency/hist 2625 # event histogram 2626 # 2627 # trigger info: hist:keys=lat,common_pid:vals=hitcount:sort=lat:size=2048 [active] 2628 # 2629 2630 { lat: 7, common_pid: 2302 } hitcount: 640 2631 { lat: 7, common_pid: 2299 } hitcount: 42 2632 { lat: 7, common_pid: 2303 } hitcount: 18 2633 { lat: 7, common_pid: 2305 } hitcount: 166 2634 { lat: 7, common_pid: 2306 } hitcount: 1 2635 { lat: 7, common_pid: 2301 } hitcount: 91 2636 { lat: 7, common_pid: 2300 } hitcount: 17 2637 { lat: 8, common_pid: 2303 } hitcount: 8296 2638 { lat: 8, common_pid: 2304 } hitcount: 6864 2639 { lat: 8, common_pid: 2305 } hitcount: 9464 2640 { lat: 8, common_pid: 2301 } hitcount: 9213 2641 { lat: 8, common_pid: 2306 } hitcount: 6246 2642 { lat: 8, common_pid: 2302 } hitcount: 8797 2643 { lat: 8, common_pid: 2299 } hitcount: 8771 2644 { lat: 8, common_pid: 2300 } hitcount: 8119 2645 { lat: 9, common_pid: 2305 } hitcount: 1519 2646 { lat: 9, common_pid: 2299 } hitcount: 2346 2647 { lat: 9, common_pid: 2303 } hitcount: 2841 2648 { lat: 9, common_pid: 2301 } hitcount: 1846 2649 { lat: 9, common_pid: 2304 } hitcount: 3861 2650 { lat: 9, common_pid: 2302 } hitcount: 1210 2651 { lat: 9, common_pid: 2300 } hitcount: 2762 2652 { lat: 9, common_pid: 2306 } hitcount: 4247 2653 { lat: 10, common_pid: 2299 } hitcount: 16 2654 { lat: 10, common_pid: 2306 } hitcount: 333 2655 { lat: 10, common_pid: 2303 } hitcount: 16 2656 { lat: 10, common_pid: 2304 } hitcount: 168 2657 { lat: 10, common_pid: 2302 } hitcount: 240 2658 { lat: 10, common_pid: 2301 } hitcount: 28 2659 { lat: 10, common_pid: 2300 } hitcount: 95 2660 { lat: 10, common_pid: 2305 } hitcount: 18 2661 { lat: 11, common_pid: 2303 } hitcount: 5 2662 { lat: 11, common_pid: 2305 } hitcount: 8 2663 { lat: 11, common_pid: 2306 } hitcount: 221 2664 { lat: 11, common_pid: 2302 } hitcount: 76 2665 { lat: 11, common_pid: 2304 } hitcount: 26 2666 { lat: 11, common_pid: 2300 } hitcount: 125 2667 { lat: 11, common_pid: 2299 } hitcount: 2 2668 { lat: 12, common_pid: 2305 } hitcount: 3 2669 { lat: 12, common_pid: 2300 } hitcount: 6 2670 { lat: 12, common_pid: 2306 } hitcount: 90 2671 { lat: 12, common_pid: 2302 } hitcount: 4 2672 { lat: 12, common_pid: 2303 } hitcount: 1 2673 { lat: 12, common_pid: 2304 } hitcount: 122 2674 { lat: 13, common_pid: 2300 } hitcount: 12 2675 { lat: 13, common_pid: 2301 } hitcount: 1 2676 { lat: 13, common_pid: 2306 } hitcount: 32 2677 { lat: 13, common_pid: 2302 } hitcount: 5 2678 { lat: 13, common_pid: 2305 } hitcount: 1 2679 { lat: 13, common_pid: 2303 } hitcount: 1 2680 { lat: 13, common_pid: 2304 } hitcount: 61 2681 { lat: 14, common_pid: 2303 } hitcount: 4 2682 { lat: 14, common_pid: 2306 } hitcount: 5 2683 { lat: 14, common_pid: 2305 } hitcount: 4 2684 { lat: 14, common_pid: 2304 } hitcount: 62 2685 { lat: 14, common_pid: 2302 } hitcount: 19 2686 { lat: 14, common_pid: 2300 } hitcount: 33 2687 { lat: 14, common_pid: 2299 } hitcount: 1 2688 { lat: 14, common_pid: 2301 } hitcount: 4 2689 { lat: 15, common_pid: 2305 } hitcount: 1 2690 { lat: 15, common_pid: 2302 } hitcount: 25 2691 { lat: 15, common_pid: 2300 } hitcount: 11 2692 { lat: 15, common_pid: 2299 } hitcount: 5 2693 { lat: 15, common_pid: 2301 } hitcount: 1 2694 { lat: 15, common_pid: 2304 } hitcount: 8 2695 { lat: 15, common_pid: 2303 } hitcount: 1 2696 { lat: 15, common_pid: 2306 } hitcount: 6 2697 { lat: 16, common_pid: 2302 } hitcount: 31 2698 { lat: 16, common_pid: 2306 } hitcount: 3 2699 { lat: 16, common_pid: 2300 } hitcount: 5 2700 { lat: 17, common_pid: 2302 } hitcount: 6 2701 { lat: 17, common_pid: 2303 } hitcount: 1 2702 { lat: 18, common_pid: 2304 } hitcount: 1 2703 { lat: 18, common_pid: 2302 } hitcount: 8 2704 { lat: 18, common_pid: 2299 } hitcount: 1 2705 { lat: 18, common_pid: 2301 } hitcount: 1 2706 { lat: 19, common_pid: 2303 } hitcount: 4 2707 { lat: 19, common_pid: 2304 } hitcount: 5 2708 { lat: 19, common_pid: 2302 } hitcount: 4 2709 { lat: 19, common_pid: 2299 } hitcount: 3 2710 { lat: 19, common_pid: 2306 } hitcount: 1 2711 { lat: 19, common_pid: 2300 } hitcount: 4 2712 { lat: 19, common_pid: 2305 } hitcount: 5 2713 { lat: 20, common_pid: 2299 } hitcount: 2 2714 { lat: 20, common_pid: 2302 } hitcount: 3 2715 { lat: 20, common_pid: 2305 } hitcount: 1 2716 { lat: 20, common_pid: 2300 } hitcount: 2 2717 { lat: 20, common_pid: 2301 } hitcount: 2 2718 { lat: 20, common_pid: 2303 } hitcount: 3 2719 { lat: 21, common_pid: 2305 } hitcount: 1 2720 { lat: 21, common_pid: 2299 } hitcount: 5 2721 { lat: 21, common_pid: 2303 } hitcount: 4 2722 { lat: 21, common_pid: 2302 } hitcount: 7 2723 { lat: 21, common_pid: 2300 } hitcount: 1 2724 { lat: 21, common_pid: 2301 } hitcount: 5 2725 { lat: 21, common_pid: 2304 } hitcount: 2 2726 { lat: 22, common_pid: 2302 } hitcount: 5 2727 { lat: 22, common_pid: 2303 } hitcount: 1 2728 { lat: 22, common_pid: 2306 } hitcount: 3 2729 { lat: 22, common_pid: 2301 } hitcount: 2 2730 { lat: 22, common_pid: 2300 } hitcount: 1 2731 { lat: 22, common_pid: 2299 } hitcount: 1 2732 { lat: 22, common_pid: 2305 } hitcount: 1 2733 { lat: 22, common_pid: 2304 } hitcount: 1 2734 { lat: 23, common_pid: 2299 } hitcount: 1 2735 { lat: 23, common_pid: 2306 } hitcount: 2 2736 { lat: 23, common_pid: 2302 } hitcount: 6 2737 { lat: 24, common_pid: 2302 } hitcount: 3 2738 { lat: 24, common_pid: 2300 } hitcount: 1 2739 { lat: 24, common_pid: 2306 } hitcount: 2 2740 { lat: 24, common_pid: 2305 } hitcount: 1 2741 { lat: 24, common_pid: 2299 } hitcount: 1 2742 { lat: 25, common_pid: 2300 } hitcount: 1 2743 { lat: 25, common_pid: 2302 } hitcount: 4 2744 { lat: 26, common_pid: 2302 } hitcount: 2 2745 { lat: 27, common_pid: 2305 } hitcount: 1 2746 { lat: 27, common_pid: 2300 } hitcount: 1 2747 { lat: 27, common_pid: 2302 } hitcount: 3 2748 { lat: 28, common_pid: 2306 } hitcount: 1 2749 { lat: 28, common_pid: 2302 } hitcount: 4 2750 { lat: 29, common_pid: 2302 } hitcount: 1 2751 { lat: 29, common_pid: 2300 } hitcount: 2 2752 { lat: 29, common_pid: 2306 } hitcount: 1 2753 { lat: 29, common_pid: 2304 } hitcount: 1 2754 { lat: 30, common_pid: 2302 } hitcount: 4 2755 { lat: 31, common_pid: 2302 } hitcount: 6 2756 { lat: 32, common_pid: 2302 } hitcount: 1 2757 { lat: 33, common_pid: 2299 } hitcount: 1 2758 { lat: 33, common_pid: 2302 } hitcount: 3 2759 { lat: 34, common_pid: 2302 } hitcount: 2 2760 { lat: 35, common_pid: 2302 } hitcount: 1 2761 { lat: 35, common_pid: 2304 } hitcount: 1 2762 { lat: 36, common_pid: 2302 } hitcount: 4 2763 { lat: 37, common_pid: 2302 } hitcount: 6 2764 { lat: 38, common_pid: 2302 } hitcount: 2 2765 { lat: 39, common_pid: 2302 } hitcount: 2 2766 { lat: 39, common_pid: 2304 } hitcount: 1 2767 { lat: 40, common_pid: 2304 } hitcount: 2 2768 { lat: 40, common_pid: 2302 } hitcount: 5 2769 { lat: 41, common_pid: 2304 } hitcount: 1 2770 { lat: 41, common_pid: 2302 } hitcount: 8 2771 { lat: 42, common_pid: 2302 } hitcount: 6 2772 { lat: 42, common_pid: 2304 } hitcount: 1 2773 { lat: 43, common_pid: 2302 } hitcount: 3 2774 { lat: 43, common_pid: 2304 } hitcount: 4 2775 { lat: 44, common_pid: 2302 } hitcount: 6 2776 { lat: 45, common_pid: 2302 } hitcount: 5 2777 { lat: 46, common_pid: 2302 } hitcount: 5 2778 { lat: 47, common_pid: 2302 } hitcount: 7 2779 { lat: 48, common_pid: 2301 } hitcount: 1 2780 { lat: 48, common_pid: 2302 } hitcount: 9 2781 { lat: 49, common_pid: 2302 } hitcount: 3 2782 { lat: 50, common_pid: 2302 } hitcount: 1 2783 { lat: 50, common_pid: 2301 } hitcount: 1 2784 { lat: 51, common_pid: 2302 } hitcount: 2 2785 { lat: 51, common_pid: 2301 } hitcount: 1 2786 { lat: 61, common_pid: 2302 } hitcount: 1 2787 { lat: 110, common_pid: 2302 } hitcount: 1 2788 2789 Totals: 2790 Hits: 89565 2791 Entries: 158 2792 Dropped: 0 2793 2794This doesn't tell us any information about how late cyclictest may have 2795woken up, but it does show us a nice histogram of how long it took from 2796the time that cyclictest was woken to the time it made it into user space. 2797