요약·해설과 원문, 전문 번역을 서로 분리했습니다. API 이름, symbol, source path는 원문 표기를 사용합니다.
1. 요약·해설
원문의 핵심 논리와 kernel programming 관점의 보충 설명입니다. 아래의 전문 번역과는 별도로 작성했습니다.
2. 영어 원문 전체
번역 기준이 된 Linux v6.18.37 원문입니다. 줄 번호는 이 버전의 파일 좌표입니다.
원문 전체 펼치기
1
================
2
Event Histograms
3
================
5
Documentation written by Tom Zanussi
7
1. Introduction
8
===============
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.
15
2. Histogram Trigger Command
16
============================
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).
24
The format of a hist trigger is as follows::
26
hist:keys=<field1[,field2,...]>[:values=<field1[,field2,...]>]
27
[:sort=<field1[,field2,...]>][:size=#entries][:pause][:continue]
28
[:clear][:name=histname1][:nohitcount][:<handler>.<action>] [if <filter>]
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 'common_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 three 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
'common_stacktrace' fields and can therefore be combined using those
58
fields, however pointless that may be.
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:
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
.buckets=size display grouping of values rather than raw number
81
.usecs display a common_timestamp in microseconds
82
.percent display a number of percentage value
83
.graph display a bar-graph of a value
84
.stacktrace display as a stacktrace (must be a long[] type)
85
============= =================================================
87
Note that in general the semantics of a given field aren't
88
interpreted when applying a modifier to it, but there are some
89
restrictions to be aware of in this regard:
91
- only the 'hex' modifier can be used for values (because values
92
are essentially sums, and the other modifiers don't make sense
93
in that context).
94
- the 'execname' modifier can only be used on a 'common_pid'. The
95
reason for this is that the execname is simply the 'comm' value
96
saved for the 'current' process when an event was triggered,
97
which is the same as the common_pid value saved by the event
98
tracing code. Trying to apply that comm value to other pid
99
values wouldn't be correct, and typically events that care save
100
pid-specific comm fields in the event itself.
102
A typical usage scenario would be the following to enable a hist
103
trigger, read its current contents, and then turn it off::
105
# echo 'hist:keys=skbaddr.hex:vals=len' > \
106
/sys/kernel/tracing/events/net/netif_rx/trigger
108
# cat /sys/kernel/tracing/events/net/netif_rx/hist
110
# echo '!hist:keys=skbaddr.hex:vals=len' > \
111
/sys/kernel/tracing/events/net/netif_rx/trigger
113
The trigger file itself can be read to show the details of the
114
currently attached hist trigger. This information is also displayed
115
at the top of the 'hist' file when read.
117
By default, the size of the hash table is 2048 entries. The 'size'
118
parameter can be used to specify more or fewer than that. The units
119
are in terms of hashtable entries - if a run uses more entries than
120
specified, the results will show the number of 'drops', the number
121
of hits that were ignored. The size should be a power of 2 between
122
128 and 131072 (any non- power-of-2 number specified will be rounded
123
up).
125
The 'sort' parameter can be used to specify a value field to sort
126
on. The default if unspecified is 'hitcount' and the default sort
127
order is 'ascending'. To sort in the opposite direction, append
128
.descending' to the sort key.
130
The 'pause' parameter can be used to pause an existing hist trigger
131
or to start a hist trigger but not log any events until told to do
132
so. 'continue' or 'cont' can be used to start or restart a paused
133
hist trigger.
135
The 'clear' parameter will clear the contents of a running hist
136
trigger and leave its current paused/active state.
138
Note that the 'pause', 'cont', and 'clear' parameters should be
139
applied using 'append' shell operator ('>>') if applied to an
140
existing trigger, rather than via the '>' operator, which will cause
141
the trigger to be removed through truncation.
143
The 'nohitcount' (or NOHC) parameter will suppress display of
144
raw hitcount in the histogram. This option requires at least one
145
value field which is not a 'raw hitcount'. For example,
146
'hist:...:vals=hitcount:nohitcount' is rejected, but
147
'hist:...:vals=hitcount.percent:nohitcount' is OK.
149
- enable_hist/disable_hist
151
The enable_hist and disable_hist triggers can be used to have one
152
event conditionally start and stop another event's already-attached
153
hist trigger. Any number of enable_hist and disable_hist triggers
154
can be attached to a given event, allowing that event to kick off
155
and stop aggregations on a host of other events.
157
The format is very similar to the enable/disable_event triggers::
159
enable_hist:<system>:<event>[:count]
160
disable_hist:<system>:<event>[:count]
162
Instead of enabling or disabling the tracing of the target event
163
into the trace buffer as the enable/disable_event triggers do, the
164
enable/disable_hist triggers enable or disable the aggregation of
165
the target event into a hash table.
167
A typical usage scenario for the enable_hist/disable_hist triggers
168
would be to first set up a paused hist trigger on some event,
169
followed by an enable_hist/disable_hist pair that turns the hist
170
aggregation on and off when conditions of interest are hit::
172
# echo 'hist:keys=skbaddr.hex:vals=len:pause' > \
173
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
175
# echo 'enable_hist:net:netif_receive_skb if filename==/usr/bin/wget' > \
176
/sys/kernel/tracing/events/sched/sched_process_exec/trigger
178
# echo 'disable_hist:net:netif_receive_skb if comm==wget' > \
179
/sys/kernel/tracing/events/sched/sched_process_exit/trigger
181
The above sets up an initially paused hist trigger which is unpaused
182
and starts aggregating events when a given program is executed, and
183
which stops aggregating when the process exits and the hist trigger
184
is paused again.
186
The examples below provide a more concrete illustration of the
187
concepts and typical usage patterns discussed above.
189
2.1. 'special' event fields
190
---------------------------
192
There are a number of 'special event fields' available for use as
193
keys or values in a hist trigger. These look like and behave as if
194
they were actual event fields, but aren't really part of the event's
195
field definition or format file. They are however available for any
196
event, and can be used anywhere an actual event field could be.
197
They are:
199
====================== ==== =======================================
200
common_timestamp u64 timestamp (from ring buffer) associated
201
with the event, in nanoseconds. May be
202
modified by .usecs to have timestamps
203
interpreted as microseconds.
204
common_cpu int the cpu on which the event occurred.
205
====================== ==== =======================================
207
2.2. Extended error information
208
-------------------------------
210
For some error conditions encountered when invoking a hist trigger
211
command, extended error information is available via the
212
tracing/error_log file. See "Error conditions" section in
213
Documentation/trace/ftrace.rst for details.
215
2.3. 'hist' trigger examples
216
----------------------------
218
The first set of examples creates aggregations using the kmalloc
219
event. The fields that can be used for the hist trigger are listed
220
in the kmalloc event's format file::
222
# cat /sys/kernel/tracing/events/kmem/kmalloc/format
223
name: kmalloc
224
ID: 374
225
format:
226
field:unsigned short common_type; offset:0; size:2; signed:0;
227
field:unsigned char common_flags; offset:2; size:1; signed:0;
228
field:unsigned char common_preempt_count; offset:3; size:1; signed:0;
229
field:int common_pid; offset:4; size:4; signed:1;
231
field:unsigned long call_site; offset:8; size:8; signed:0;
232
field:const void * ptr; offset:16; size:8; signed:0;
233
field:size_t bytes_req; offset:24; size:8; signed:0;
234
field:size_t bytes_alloc; offset:32; size:8; signed:0;
235
field:gfp_t gfp_flags; offset:40; size:4; signed:0;
237
We'll start by creating a hist trigger that generates a simple table
238
that lists the total number of bytes requested for each function in
239
the kernel that made one or more calls to kmalloc::
241
# echo 'hist:key=call_site:val=bytes_req.buckets=32' > \
242
/sys/kernel/tracing/events/kmem/kmalloc/trigger
244
This tells the tracing system to create a 'hist' trigger using the
245
call_site field of the kmalloc event as the key for the table, which
246
just means that each unique call_site address will have an entry
247
created for it in the table. The 'val=bytes_req' parameter tells
248
the hist trigger that for each unique entry (call_site) in the
249
table, it should keep a running total of the number of bytes
250
requested by that call_site.
252
We'll let it run for a while and then dump the contents of the 'hist'
253
file in the kmalloc event's subdirectory (for readability, a number
254
of entries have been omitted)::
256
# cat /sys/kernel/tracing/events/kmem/kmalloc/hist
257
# trigger info: hist:keys=call_site:vals=bytes_req:sort=hitcount:size=2048 [active]
259
{ call_site: 18446744072106379007 } hitcount: 1 bytes_req: 176
260
{ call_site: 18446744071579557049 } hitcount: 1 bytes_req: 1024
261
{ call_site: 18446744071580608289 } hitcount: 1 bytes_req: 16384
262
{ call_site: 18446744071581827654 } hitcount: 1 bytes_req: 24
263
{ call_site: 18446744071580700980 } hitcount: 1 bytes_req: 8
264
{ call_site: 18446744071579359876 } hitcount: 1 bytes_req: 152
265
{ call_site: 18446744071580795365 } hitcount: 3 bytes_req: 144
266
{ call_site: 18446744071581303129 } hitcount: 3 bytes_req: 144
267
{ call_site: 18446744071580713234 } hitcount: 4 bytes_req: 2560
268
{ call_site: 18446744071580933750 } hitcount: 4 bytes_req: 736
269
.
270
.
271
.
272
{ call_site: 18446744072106047046 } hitcount: 69 bytes_req: 5576
273
{ call_site: 18446744071582116407 } hitcount: 73 bytes_req: 2336
274
{ call_site: 18446744072106054684 } hitcount: 136 bytes_req: 140504
275
{ call_site: 18446744072106224230 } hitcount: 136 bytes_req: 19584
276
{ call_site: 18446744072106078074 } hitcount: 153 bytes_req: 2448
277
{ call_site: 18446744072106062406 } hitcount: 153 bytes_req: 36720
278
{ call_site: 18446744071582507929 } hitcount: 153 bytes_req: 37088
279
{ call_site: 18446744072102520590 } hitcount: 273 bytes_req: 10920
280
{ call_site: 18446744071582143559 } hitcount: 358 bytes_req: 716
281
{ call_site: 18446744072106465852 } hitcount: 417 bytes_req: 56712
282
{ call_site: 18446744072102523378 } hitcount: 485 bytes_req: 27160
283
{ call_site: 18446744072099568646 } hitcount: 1676 bytes_req: 33520
285
Totals:
286
Hits: 4610
287
Entries: 45
288
Dropped: 0
290
The output displays a line for each entry, beginning with the key
291
specified in the trigger, followed by the value(s) also specified in
292
the trigger. At the beginning of the output is a line that displays
293
the trigger info, which can also be displayed by reading the
294
'trigger' file::
296
# cat /sys/kernel/tracing/events/kmem/kmalloc/trigger
297
hist:keys=call_site:vals=bytes_req:sort=hitcount:size=2048 [active]
299
At the end of the output are a few lines that display the overall
300
totals for the run. The 'Hits' field shows the total number of
301
times the event trigger was hit, the 'Entries' field shows the total
302
number of used entries in the hash table, and the 'Dropped' field
303
shows the number of hits that were dropped because the number of
304
used entries for the run exceeded the maximum number of entries
305
allowed for the table (normally 0, but if not a hint that you may
306
want to increase the size of the table using the 'size' parameter).
308
Notice in the above output that there's an extra field, 'hitcount',
309
which wasn't specified in the trigger. Also notice that in the
310
trigger info output, there's a parameter, 'sort=hitcount', which
311
wasn't specified in the trigger either. The reason for that is that
312
every trigger implicitly keeps a count of the total number of hits
313
attributed to a given entry, called the 'hitcount'. That hitcount
314
information is explicitly displayed in the output, and in the
315
absence of a user-specified sort parameter, is used as the default
316
sort field.
318
The value 'hitcount' can be used in place of an explicit value in
319
the 'values' parameter if you don't really need to have any
320
particular field summed and are mainly interested in hit
321
frequencies.
323
To turn the hist trigger off, simply call up the trigger in the
324
command history and re-execute it with a '!' prepended::
326
# echo '!hist:key=call_site:val=bytes_req' > \
327
/sys/kernel/tracing/events/kmem/kmalloc/trigger
329
Finally, notice that the call_site as displayed in the output above
330
isn't really very useful. It's an address, but normally addresses
331
are displayed in hex. To have a numeric field displayed as a hex
332
value, simply append '.hex' to the field name in the trigger::
334
# echo 'hist:key=call_site.hex:val=bytes_req' > \
335
/sys/kernel/tracing/events/kmem/kmalloc/trigger
337
# cat /sys/kernel/tracing/events/kmem/kmalloc/hist
338
# trigger info: hist:keys=call_site.hex:vals=bytes_req:sort=hitcount:size=2048 [active]
340
{ call_site: ffffffffa026b291 } hitcount: 1 bytes_req: 433
341
{ call_site: ffffffffa07186ff } hitcount: 1 bytes_req: 176
342
{ call_site: ffffffff811ae721 } hitcount: 1 bytes_req: 16384
343
{ call_site: ffffffff811c5134 } hitcount: 1 bytes_req: 8
344
{ call_site: ffffffffa04a9ebb } hitcount: 1 bytes_req: 511
345
{ call_site: ffffffff8122e0a6 } hitcount: 1 bytes_req: 12
346
{ call_site: ffffffff8107da84 } hitcount: 1 bytes_req: 152
347
{ call_site: ffffffff812d8246 } hitcount: 1 bytes_req: 24
348
{ call_site: ffffffff811dc1e5 } hitcount: 3 bytes_req: 144
349
{ call_site: ffffffffa02515e8 } hitcount: 3 bytes_req: 648
350
{ call_site: ffffffff81258159 } hitcount: 3 bytes_req: 144
351
{ call_site: ffffffff811c80f4 } hitcount: 4 bytes_req: 544
352
.
353
.
354
.
355
{ call_site: ffffffffa06c7646 } hitcount: 106 bytes_req: 8024
356
{ call_site: ffffffffa06cb246 } hitcount: 132 bytes_req: 31680
357
{ call_site: ffffffffa06cef7a } hitcount: 132 bytes_req: 2112
358
{ call_site: ffffffff8137e399 } hitcount: 132 bytes_req: 23232
359
{ call_site: ffffffffa06c941c } hitcount: 185 bytes_req: 171360
360
{ call_site: ffffffffa06f2a66 } hitcount: 185 bytes_req: 26640
361
{ call_site: ffffffffa036a70e } hitcount: 265 bytes_req: 10600
362
{ call_site: ffffffff81325447 } hitcount: 292 bytes_req: 584
363
{ call_site: ffffffffa072da3c } hitcount: 446 bytes_req: 60656
364
{ call_site: ffffffffa036b1f2 } hitcount: 526 bytes_req: 29456
365
{ call_site: ffffffffa0099c06 } hitcount: 1780 bytes_req: 35600
367
Totals:
368
Hits: 4775
369
Entries: 46
370
Dropped: 0
372
Even that's only marginally more useful - while hex values do look
373
more like addresses, what users are typically more interested in
374
when looking at text addresses are the corresponding symbols
375
instead. To have an address displayed as symbolic value instead,
376
simply append '.sym' or '.sym-offset' to the field name in the
377
trigger::
379
# echo 'hist:key=call_site.sym:val=bytes_req' > \
380
/sys/kernel/tracing/events/kmem/kmalloc/trigger
382
# cat /sys/kernel/tracing/events/kmem/kmalloc/hist
383
# trigger info: hist:keys=call_site.sym:vals=bytes_req:sort=hitcount:size=2048 [active]
385
{ call_site: [ffffffff810adcb9] syslog_print_all } hitcount: 1 bytes_req: 1024
386
{ call_site: [ffffffff8154bc62] usb_control_msg } hitcount: 1 bytes_req: 8
387
{ call_site: [ffffffffa00bf6fe] hidraw_send_report [hid] } hitcount: 1 bytes_req: 7
388
{ call_site: [ffffffff8154acbe] usb_alloc_urb } hitcount: 1 bytes_req: 192
389
{ call_site: [ffffffffa00bf1ca] hidraw_report_event [hid] } hitcount: 1 bytes_req: 7
390
{ call_site: [ffffffff811e3a25] __seq_open_private } hitcount: 1 bytes_req: 40
391
{ call_site: [ffffffff8109524a] alloc_fair_sched_group } hitcount: 2 bytes_req: 128
392
{ call_site: [ffffffff811febd5] fsnotify_alloc_group } hitcount: 2 bytes_req: 528
393
{ call_site: [ffffffff81440f58] __tty_buffer_request_room } hitcount: 2 bytes_req: 2624
394
{ call_site: [ffffffff81200ba6] inotify_new_group } hitcount: 2 bytes_req: 96
395
{ call_site: [ffffffffa05e19af] ieee80211_start_tx_ba_session [mac80211] } hitcount: 2 bytes_req: 464
396
{ call_site: [ffffffff81672406] tcp_get_metrics } hitcount: 2 bytes_req: 304
397
{ call_site: [ffffffff81097ec2] alloc_rt_sched_group } hitcount: 2 bytes_req: 128
398
{ call_site: [ffffffff81089b05] sched_create_group } hitcount: 2 bytes_req: 1424
399
.
400
.
401
.
402
{ call_site: [ffffffffa04a580c] intel_crtc_page_flip [i915] } hitcount: 1185 bytes_req: 123240
403
{ call_site: [ffffffffa0287592] drm_mode_page_flip_ioctl [drm] } hitcount: 1185 bytes_req: 104280
404
{ call_site: [ffffffffa04c4a3c] intel_plane_duplicate_state [i915] } hitcount: 1402 bytes_req: 190672
405
{ call_site: [ffffffff812891ca] ext4_find_extent } hitcount: 1518 bytes_req: 146208
406
{ call_site: [ffffffffa029070e] drm_vma_node_allow [drm] } hitcount: 1746 bytes_req: 69840
407
{ call_site: [ffffffffa045e7c4] i915_gem_do_execbuffer.isra.23 [i915] } hitcount: 2021 bytes_req: 792312
408
{ call_site: [ffffffffa02911f2] drm_modeset_lock_crtc [drm] } hitcount: 2592 bytes_req: 145152
409
{ call_site: [ffffffffa0489a66] intel_ring_begin [i915] } hitcount: 2629 bytes_req: 378576
410
{ call_site: [ffffffffa046041c] i915_gem_execbuffer2 [i915] } hitcount: 2629 bytes_req: 3783248
411
{ call_site: [ffffffff81325607] apparmor_file_alloc_security } hitcount: 5192 bytes_req: 10384
412
{ call_site: [ffffffffa00b7c06] hid_report_raw_event [hid] } hitcount: 5529 bytes_req: 110584
413
{ call_site: [ffffffff8131ebf7] aa_alloc_task_context } hitcount: 21943 bytes_req: 702176
414
{ call_site: [ffffffff8125847d] ext4_htree_store_dirent } hitcount: 55759 bytes_req: 5074265
416
Totals:
417
Hits: 109928
418
Entries: 71
419
Dropped: 0
421
Because the default sort key above is 'hitcount', the above shows a
422
the list of call_sites by increasing hitcount, so that at the bottom
423
we see the functions that made the most kmalloc calls during the
424
run. If instead we wanted to see the top kmalloc callers in
425
terms of the number of bytes requested rather than the number of
426
calls, and we wanted the top caller to appear at the top, we can use
427
the 'sort' parameter, along with the 'descending' modifier::
429
# echo 'hist:key=call_site.sym:val=bytes_req:sort=bytes_req.descending' > \
430
/sys/kernel/tracing/events/kmem/kmalloc/trigger
432
# cat /sys/kernel/tracing/events/kmem/kmalloc/hist
433
# trigger info: hist:keys=call_site.sym:vals=bytes_req:sort=bytes_req.descending:size=2048 [active]
435
{ call_site: [ffffffffa046041c] i915_gem_execbuffer2 [i915] } hitcount: 2186 bytes_req: 3397464
436
{ call_site: [ffffffffa045e7c4] i915_gem_do_execbuffer.isra.23 [i915] } hitcount: 1790 bytes_req: 712176
437
{ call_site: [ffffffff8125847d] ext4_htree_store_dirent } hitcount: 8132 bytes_req: 513135
438
{ call_site: [ffffffff811e2a1b] seq_buf_alloc } hitcount: 106 bytes_req: 440128
439
{ call_site: [ffffffffa0489a66] intel_ring_begin [i915] } hitcount: 2186 bytes_req: 314784
440
{ call_site: [ffffffff812891ca] ext4_find_extent } hitcount: 2174 bytes_req: 208992
441
{ call_site: [ffffffff811ae8e1] __kmalloc } hitcount: 8 bytes_req: 131072
442
{ call_site: [ffffffffa04c4a3c] intel_plane_duplicate_state [i915] } hitcount: 859 bytes_req: 116824
443
{ call_site: [ffffffffa02911f2] drm_modeset_lock_crtc [drm] } hitcount: 1834 bytes_req: 102704
444
{ call_site: [ffffffffa04a580c] intel_crtc_page_flip [i915] } hitcount: 972 bytes_req: 101088
445
{ call_site: [ffffffffa0287592] drm_mode_page_flip_ioctl [drm] } hitcount: 972 bytes_req: 85536
446
{ call_site: [ffffffffa00b7c06] hid_report_raw_event [hid] } hitcount: 3333 bytes_req: 66664
447
{ call_site: [ffffffff8137e559] sg_kmalloc } hitcount: 209 bytes_req: 61632
448
.
449
.
450
.
451
{ call_site: [ffffffff81095225] alloc_fair_sched_group } hitcount: 2 bytes_req: 128
452
{ call_site: [ffffffff81097ec2] alloc_rt_sched_group } hitcount: 2 bytes_req: 128
453
{ call_site: [ffffffff812d8406] copy_semundo } hitcount: 2 bytes_req: 48
454
{ call_site: [ffffffff81200ba6] inotify_new_group } hitcount: 1 bytes_req: 48
455
{ call_site: [ffffffffa027121a] drm_getmagic [drm] } hitcount: 1 bytes_req: 48
456
{ call_site: [ffffffff811e3a25] __seq_open_private } hitcount: 1 bytes_req: 40
457
{ call_site: [ffffffff811c52f4] bprm_change_interp } hitcount: 2 bytes_req: 16
458
{ call_site: [ffffffff8154bc62] usb_control_msg } hitcount: 1 bytes_req: 8
459
{ call_site: [ffffffffa00bf1ca] hidraw_report_event [hid] } hitcount: 1 bytes_req: 7
460
{ call_site: [ffffffffa00bf6fe] hidraw_send_report [hid] } hitcount: 1 bytes_req: 7
462
Totals:
463
Hits: 32133
464
Entries: 81
465
Dropped: 0
467
To display the offset and size information in addition to the symbol
468
name, just use 'sym-offset' instead::
470
# echo 'hist:key=call_site.sym-offset:val=bytes_req:sort=bytes_req.descending' > \
471
/sys/kernel/tracing/events/kmem/kmalloc/trigger
473
# cat /sys/kernel/tracing/events/kmem/kmalloc/hist
474
# trigger info: hist:keys=call_site.sym-offset:vals=bytes_req:sort=bytes_req.descending:size=2048 [active]
476
{ call_site: [ffffffffa046041c] i915_gem_execbuffer2+0x6c/0x2c0 [i915] } hitcount: 4569 bytes_req: 3163720
477
{ call_site: [ffffffffa0489a66] intel_ring_begin+0xc6/0x1f0 [i915] } hitcount: 4569 bytes_req: 657936
478
{ call_site: [ffffffffa045e7c4] i915_gem_do_execbuffer.isra.23+0x694/0x1020 [i915] } hitcount: 1519 bytes_req: 472936
479
{ call_site: [ffffffffa045e646] i915_gem_do_execbuffer.isra.23+0x516/0x1020 [i915] } hitcount: 3050 bytes_req: 211832
480
{ call_site: [ffffffff811e2a1b] seq_buf_alloc+0x1b/0x50 } hitcount: 34 bytes_req: 148384
481
{ call_site: [ffffffffa04a580c] intel_crtc_page_flip+0xbc/0x870 [i915] } hitcount: 1385 bytes_req: 144040
482
{ call_site: [ffffffff811ae8e1] __kmalloc+0x191/0x1b0 } hitcount: 8 bytes_req: 131072
483
{ call_site: [ffffffffa0287592] drm_mode_page_flip_ioctl+0x282/0x360 [drm] } hitcount: 1385 bytes_req: 121880
484
{ call_site: [ffffffffa02911f2] drm_modeset_lock_crtc+0x32/0x100 [drm] } hitcount: 1848 bytes_req: 103488
485
{ call_site: [ffffffffa04c4a3c] intel_plane_duplicate_state+0x2c/0xa0 [i915] } hitcount: 461 bytes_req: 62696
486
{ call_site: [ffffffffa029070e] drm_vma_node_allow+0x2e/0xd0 [drm] } hitcount: 1541 bytes_req: 61640
487
{ call_site: [ffffffff815f8d7b] sk_prot_alloc+0xcb/0x1b0 } hitcount: 57 bytes_req: 57456
488
.
489
.
490
.
491
{ call_site: [ffffffff8109524a] alloc_fair_sched_group+0x5a/0x1a0 } hitcount: 2 bytes_req: 128
492
{ call_site: [ffffffffa027b921] drm_vm_open_locked+0x31/0xa0 [drm] } hitcount: 3 bytes_req: 96
493
{ call_site: [ffffffff8122e266] proc_self_follow_link+0x76/0xb0 } hitcount: 8 bytes_req: 96
494
{ call_site: [ffffffff81213e80] load_elf_binary+0x240/0x1650 } hitcount: 3 bytes_req: 84
495
{ call_site: [ffffffff8154bc62] usb_control_msg+0x42/0x110 } hitcount: 1 bytes_req: 8
496
{ call_site: [ffffffffa00bf6fe] hidraw_send_report+0x7e/0x1a0 [hid] } hitcount: 1 bytes_req: 7
497
{ call_site: [ffffffffa00bf1ca] hidraw_report_event+0x8a/0x120 [hid] } hitcount: 1 bytes_req: 7
499
Totals:
500
Hits: 26098
501
Entries: 64
502
Dropped: 0
504
We can also add multiple fields to the 'values' parameter. For
505
example, we might want to see the total number of bytes allocated
506
alongside bytes requested, and display the result sorted by bytes
507
allocated in a descending order::
509
# echo 'hist:keys=call_site.sym:values=bytes_req,bytes_alloc:sort=bytes_alloc.descending' > \
510
/sys/kernel/tracing/events/kmem/kmalloc/trigger
512
# cat /sys/kernel/tracing/events/kmem/kmalloc/hist
513
# trigger info: hist:keys=call_site.sym:vals=bytes_req,bytes_alloc:sort=bytes_alloc.descending:size=2048 [active]
515
{ call_site: [ffffffffa046041c] i915_gem_execbuffer2 [i915] } hitcount: 7403 bytes_req: 4084360 bytes_alloc: 5958016
516
{ call_site: [ffffffff811e2a1b] seq_buf_alloc } hitcount: 541 bytes_req: 2213968 bytes_alloc: 2228224
517
{ call_site: [ffffffffa0489a66] intel_ring_begin [i915] } hitcount: 7404 bytes_req: 1066176 bytes_alloc: 1421568
518
{ call_site: [ffffffffa045e7c4] i915_gem_do_execbuffer.isra.23 [i915] } hitcount: 1565 bytes_req: 557368 bytes_alloc: 1037760
519
{ call_site: [ffffffff8125847d] ext4_htree_store_dirent } hitcount: 9557 bytes_req: 595778 bytes_alloc: 695744
520
{ call_site: [ffffffffa045e646] i915_gem_do_execbuffer.isra.23 [i915] } hitcount: 5839 bytes_req: 430680 bytes_alloc: 470400
521
{ call_site: [ffffffffa04c4a3c] intel_plane_duplicate_state [i915] } hitcount: 2388 bytes_req: 324768 bytes_alloc: 458496
522
{ call_site: [ffffffffa02911f2] drm_modeset_lock_crtc [drm] } hitcount: 3911 bytes_req: 219016 bytes_alloc: 250304
523
{ call_site: [ffffffff815f8d7b] sk_prot_alloc } hitcount: 235 bytes_req: 236880 bytes_alloc: 240640
524
{ call_site: [ffffffff8137e559] sg_kmalloc } hitcount: 557 bytes_req: 169024 bytes_alloc: 221760
525
{ call_site: [ffffffffa00b7c06] hid_report_raw_event [hid] } hitcount: 9378 bytes_req: 187548 bytes_alloc: 206312
526
{ call_site: [ffffffffa04a580c] intel_crtc_page_flip [i915] } hitcount: 1519 bytes_req: 157976 bytes_alloc: 194432
527
.
528
.
529
.
530
{ call_site: [ffffffff8109bd3b] sched_autogroup_create_attach } hitcount: 2 bytes_req: 144 bytes_alloc: 192
531
{ call_site: [ffffffff81097ee8] alloc_rt_sched_group } hitcount: 2 bytes_req: 128 bytes_alloc: 128
532
{ call_site: [ffffffff8109524a] alloc_fair_sched_group } hitcount: 2 bytes_req: 128 bytes_alloc: 128
533
{ call_site: [ffffffff81095225] alloc_fair_sched_group } hitcount: 2 bytes_req: 128 bytes_alloc: 128
534
{ call_site: [ffffffff81097ec2] alloc_rt_sched_group } hitcount: 2 bytes_req: 128 bytes_alloc: 128
535
{ call_site: [ffffffff81213e80] load_elf_binary } hitcount: 3 bytes_req: 84 bytes_alloc: 96
536
{ call_site: [ffffffff81079a2e] kthread_create_on_node } hitcount: 1 bytes_req: 56 bytes_alloc: 64
537
{ call_site: [ffffffffa00bf6fe] hidraw_send_report [hid] } hitcount: 1 bytes_req: 7 bytes_alloc: 8
538
{ call_site: [ffffffff8154bc62] usb_control_msg } hitcount: 1 bytes_req: 8 bytes_alloc: 8
539
{ call_site: [ffffffffa00bf1ca] hidraw_report_event [hid] } hitcount: 1 bytes_req: 7 bytes_alloc: 8
541
Totals:
542
Hits: 66598
543
Entries: 65
544
Dropped: 0
546
Finally, to finish off our kmalloc example, instead of simply having
547
the hist trigger display symbolic call_sites, we can have the hist
548
trigger additionally display the complete set of kernel stack traces
549
that led to each call_site. To do that, we simply use the special
550
value 'common_stacktrace' for the key parameter::
552
# echo 'hist:keys=common_stacktrace:values=bytes_req,bytes_alloc:sort=bytes_alloc' > \
553
/sys/kernel/tracing/events/kmem/kmalloc/trigger
555
The above trigger will use the kernel stack trace in effect when an
556
event is triggered as the key for the hash table. This allows the
557
enumeration of every kernel callpath that led up to a particular
558
event, along with a running total of any of the event fields for
559
that event. Here we tally bytes requested and bytes allocated for
560
every callpath in the system that led up to a kmalloc (in this case
561
every callpath to a kmalloc for a kernel compile)::
563
# cat /sys/kernel/tracing/events/kmem/kmalloc/hist
564
# trigger info: hist:keys=common_stacktrace:vals=bytes_req,bytes_alloc:sort=bytes_alloc:size=2048 [active]
566
{ common_stacktrace:
567
__kmalloc_track_caller+0x10b/0x1a0
568
kmemdup+0x20/0x50
569
hidraw_report_event+0x8a/0x120 [hid]
570
hid_report_raw_event+0x3ea/0x440 [hid]
571
hid_input_report+0x112/0x190 [hid]
572
hid_irq_in+0xc2/0x260 [usbhid]
573
__usb_hcd_giveback_urb+0x72/0x120
574
usb_giveback_urb_bh+0x9e/0xe0
575
tasklet_hi_action+0xf8/0x100
576
__do_softirq+0x114/0x2c0
577
irq_exit+0xa5/0xb0
578
do_IRQ+0x5a/0xf0
579
ret_from_intr+0x0/0x30
580
cpuidle_enter+0x17/0x20
581
cpu_startup_entry+0x315/0x3e0
582
rest_init+0x7c/0x80
583
} hitcount: 3 bytes_req: 21 bytes_alloc: 24
584
{ common_stacktrace:
585
__kmalloc_track_caller+0x10b/0x1a0
586
kmemdup+0x20/0x50
587
hidraw_report_event+0x8a/0x120 [hid]
588
hid_report_raw_event+0x3ea/0x440 [hid]
589
hid_input_report+0x112/0x190 [hid]
590
hid_irq_in+0xc2/0x260 [usbhid]
591
__usb_hcd_giveback_urb+0x72/0x120
592
usb_giveback_urb_bh+0x9e/0xe0
593
tasklet_hi_action+0xf8/0x100
594
__do_softirq+0x114/0x2c0
595
irq_exit+0xa5/0xb0
596
do_IRQ+0x5a/0xf0
597
ret_from_intr+0x0/0x30
598
} hitcount: 3 bytes_req: 21 bytes_alloc: 24
599
{ common_stacktrace:
600
kmem_cache_alloc_trace+0xeb/0x150
601
aa_alloc_task_context+0x27/0x40
602
apparmor_cred_prepare+0x1f/0x50
603
security_prepare_creds+0x16/0x20
604
prepare_creds+0xdf/0x1a0
605
SyS_capset+0xb5/0x200
606
system_call_fastpath+0x12/0x6a
607
} hitcount: 1 bytes_req: 32 bytes_alloc: 32
608
.
609
.
610
.
611
{ common_stacktrace:
612
__kmalloc+0x11b/0x1b0
613
i915_gem_execbuffer2+0x6c/0x2c0 [i915]
614
drm_ioctl+0x349/0x670 [drm]
615
do_vfs_ioctl+0x2f0/0x4f0
616
SyS_ioctl+0x81/0xa0
617
system_call_fastpath+0x12/0x6a
618
} hitcount: 17726 bytes_req: 13944120 bytes_alloc: 19593808
619
{ common_stacktrace:
620
__kmalloc+0x11b/0x1b0
621
load_elf_phdrs+0x76/0xa0
622
load_elf_binary+0x102/0x1650
623
search_binary_handler+0x97/0x1d0
624
do_execveat_common.isra.34+0x551/0x6e0
625
SyS_execve+0x3a/0x50
626
return_from_execve+0x0/0x23
627
} hitcount: 33348 bytes_req: 17152128 bytes_alloc: 20226048
628
{ common_stacktrace:
629
kmem_cache_alloc_trace+0xeb/0x150
630
apparmor_file_alloc_security+0x27/0x40
631
security_file_alloc+0x16/0x20
632
get_empty_filp+0x93/0x1c0
633
path_openat+0x31/0x5f0
634
do_filp_open+0x3a/0x90
635
do_sys_open+0x128/0x220
636
SyS_open+0x1e/0x20
637
system_call_fastpath+0x12/0x6a
638
} hitcount: 4766422 bytes_req: 9532844 bytes_alloc: 38131376
639
{ common_stacktrace:
640
__kmalloc+0x11b/0x1b0
641
seq_buf_alloc+0x1b/0x50
642
seq_read+0x2cc/0x370
643
proc_reg_read+0x3d/0x80
644
__vfs_read+0x28/0xe0
645
vfs_read+0x86/0x140
646
SyS_read+0x46/0xb0
647
system_call_fastpath+0x12/0x6a
648
} hitcount: 19133 bytes_req: 78368768 bytes_alloc: 78368768
650
Totals:
651
Hits: 6085872
652
Entries: 253
653
Dropped: 0
655
If you key a hist trigger on common_pid, in order for example to
656
gather and display sorted totals for each process, you can use the
657
special .execname modifier to display the executable names for the
658
processes in the table rather than raw pids. The example below
659
keeps a per-process sum of total bytes read::
661
# echo 'hist:key=common_pid.execname:val=count:sort=count.descending' > \
662
/sys/kernel/tracing/events/syscalls/sys_enter_read/trigger
664
# cat /sys/kernel/tracing/events/syscalls/sys_enter_read/hist
665
# trigger info: hist:keys=common_pid.execname:vals=count:sort=count.descending:size=2048 [active]
667
{ common_pid: gnome-terminal [ 3196] } hitcount: 280 count: 1093512
668
{ common_pid: Xorg [ 1309] } hitcount: 525 count: 256640
669
{ common_pid: compiz [ 2889] } hitcount: 59 count: 254400
670
{ common_pid: bash [ 8710] } hitcount: 3 count: 66369
671
{ common_pid: dbus-daemon-lau [ 8703] } hitcount: 49 count: 47739
672
{ common_pid: irqbalance [ 1252] } hitcount: 27 count: 27648
673
{ common_pid: 01ifupdown [ 8705] } hitcount: 3 count: 17216
674
{ common_pid: dbus-daemon [ 772] } hitcount: 10 count: 12396
675
{ common_pid: Socket Thread [ 8342] } hitcount: 11 count: 11264
676
{ common_pid: nm-dhcp-client. [ 8701] } hitcount: 6 count: 7424
677
{ common_pid: gmain [ 1315] } hitcount: 18 count: 6336
678
.
679
.
680
.
681
{ common_pid: postgres [ 1892] } hitcount: 2 count: 32
682
{ common_pid: postgres [ 1891] } hitcount: 2 count: 32
683
{ common_pid: gmain [ 8704] } hitcount: 2 count: 32
684
{ common_pid: upstart-dbus-br [ 2740] } hitcount: 21 count: 21
685
{ common_pid: nm-dispatcher.a [ 8696] } hitcount: 1 count: 16
686
{ common_pid: indicator-datet [ 2904] } hitcount: 1 count: 16
687
{ common_pid: gdbus [ 2998] } hitcount: 1 count: 16
688
{ common_pid: rtkit-daemon [ 2052] } hitcount: 1 count: 8
689
{ common_pid: init [ 1] } hitcount: 2 count: 2
691
Totals:
692
Hits: 2116
693
Entries: 51
694
Dropped: 0
696
Similarly, if you key a hist trigger on syscall id, for example to
697
gather and display a list of systemwide syscall hits, you can use
698
the special .syscall modifier to display the syscall names rather
699
than raw ids. The example below keeps a running total of syscall
700
counts for the system during the run::
702
# echo 'hist:key=id.syscall:val=hitcount' > \
703
/sys/kernel/tracing/events/raw_syscalls/sys_enter/trigger
705
# cat /sys/kernel/tracing/events/raw_syscalls/sys_enter/hist
706
# trigger info: hist:keys=id.syscall:vals=hitcount:sort=hitcount:size=2048 [active]
708
{ id: sys_fsync [ 74] } hitcount: 1
709
{ id: sys_newuname [ 63] } hitcount: 1
710
{ id: sys_prctl [157] } hitcount: 1
711
{ id: sys_statfs [137] } hitcount: 1
712
{ id: sys_symlink [ 88] } hitcount: 1
713
{ id: sys_sendmmsg [307] } hitcount: 1
714
{ id: sys_semctl [ 66] } hitcount: 1
715
{ id: sys_readlink [ 89] } hitcount: 3
716
{ id: sys_bind [ 49] } hitcount: 3
717
{ id: sys_getsockname [ 51] } hitcount: 3
718
{ id: sys_unlink [ 87] } hitcount: 3
719
{ id: sys_rename [ 82] } hitcount: 4
720
{ id: unknown_syscall [ 58] } hitcount: 4
721
{ id: sys_connect [ 42] } hitcount: 4
722
{ id: sys_getpid [ 39] } hitcount: 4
723
.
724
.
725
.
726
{ id: sys_rt_sigprocmask [ 14] } hitcount: 952
727
{ id: sys_futex [202] } hitcount: 1534
728
{ id: sys_write [ 1] } hitcount: 2689
729
{ id: sys_setitimer [ 38] } hitcount: 2797
730
{ id: sys_read [ 0] } hitcount: 3202
731
{ id: sys_select [ 23] } hitcount: 3773
732
{ id: sys_writev [ 20] } hitcount: 4531
733
{ id: sys_poll [ 7] } hitcount: 8314
734
{ id: sys_recvmsg [ 47] } hitcount: 13738
735
{ id: sys_ioctl [ 16] } hitcount: 21843
737
Totals:
738
Hits: 67612
739
Entries: 72
740
Dropped: 0
742
The syscall counts above provide a rough overall picture of system
743
call activity on the system; we can see for example that the most
744
popular system call on this system was the 'sys_ioctl' system call.
746
We can use 'compound' keys to refine that number and provide some
747
further insight as to which processes exactly contribute to the
748
overall ioctl count.
750
The command below keeps a hitcount for every unique combination of
751
system call id and pid - the end result is essentially a table
752
that keeps a per-pid sum of system call hits. The results are
753
sorted using the system call id as the primary key, and the
754
hitcount sum as the secondary key::
756
# echo 'hist:key=id.syscall,common_pid.execname:val=hitcount:sort=id,hitcount' > \
757
/sys/kernel/tracing/events/raw_syscalls/sys_enter/trigger
759
# cat /sys/kernel/tracing/events/raw_syscalls/sys_enter/hist
760
# trigger info: hist:keys=id.syscall,common_pid.execname:vals=hitcount:sort=id.syscall,hitcount:size=2048 [active]
762
{ id: sys_read [ 0], common_pid: rtkit-daemon [ 1877] } hitcount: 1
763
{ id: sys_read [ 0], common_pid: gdbus [ 2976] } hitcount: 1
764
{ id: sys_read [ 0], common_pid: console-kit-dae [ 3400] } hitcount: 1
765
{ id: sys_read [ 0], common_pid: postgres [ 1865] } hitcount: 1
766
{ id: sys_read [ 0], common_pid: deja-dup-monito [ 3543] } hitcount: 2
767
{ id: sys_read [ 0], common_pid: NetworkManager [ 890] } hitcount: 2
768
{ id: sys_read [ 0], common_pid: evolution-calen [ 3048] } hitcount: 2
769
{ id: sys_read [ 0], common_pid: postgres [ 1864] } hitcount: 2
770
{ id: sys_read [ 0], common_pid: nm-applet [ 3022] } hitcount: 2
771
{ id: sys_read [ 0], common_pid: whoopsie [ 1212] } hitcount: 2
772
.
773
.
774
.
775
{ id: sys_ioctl [ 16], common_pid: bash [ 8479] } hitcount: 1
776
{ id: sys_ioctl [ 16], common_pid: bash [ 3472] } hitcount: 12
777
{ id: sys_ioctl [ 16], common_pid: gnome-terminal [ 3199] } hitcount: 16
778
{ id: sys_ioctl [ 16], common_pid: Xorg [ 1267] } hitcount: 1808
779
{ id: sys_ioctl [ 16], common_pid: compiz [ 2994] } hitcount: 5580
780
.
781
.
782
.
783
{ id: sys_waitid [247], common_pid: upstart-dbus-br [ 2690] } hitcount: 3
784
{ id: sys_waitid [247], common_pid: upstart-dbus-br [ 2688] } hitcount: 16
785
{ id: sys_inotify_add_watch [254], common_pid: gmain [ 975] } hitcount: 2
786
{ id: sys_inotify_add_watch [254], common_pid: gmain [ 3204] } hitcount: 4
787
{ id: sys_inotify_add_watch [254], common_pid: gmain [ 2888] } hitcount: 4
788
{ id: sys_inotify_add_watch [254], common_pid: gmain [ 3003] } hitcount: 4
789
{ id: sys_inotify_add_watch [254], common_pid: gmain [ 2873] } hitcount: 4
790
{ id: sys_inotify_add_watch [254], common_pid: gmain [ 3196] } hitcount: 6
791
{ id: sys_openat [257], common_pid: java [ 2623] } hitcount: 2
792
{ id: sys_eventfd2 [290], common_pid: ibus-ui-gtk3 [ 2760] } hitcount: 4
793
{ id: sys_eventfd2 [290], common_pid: compiz [ 2994] } hitcount: 6
795
Totals:
796
Hits: 31536
797
Entries: 323
798
Dropped: 0
800
The above list does give us a breakdown of the ioctl syscall by
801
pid, but it also gives us quite a bit more than that, which we
802
don't really care about at the moment. Since we know the syscall
803
id for sys_ioctl (16, displayed next to the sys_ioctl name), we
804
can use that to filter out all the other syscalls::
806
# echo 'hist:key=id.syscall,common_pid.execname:val=hitcount:sort=id,hitcount if id == 16' > \
807
/sys/kernel/tracing/events/raw_syscalls/sys_enter/trigger
809
# cat /sys/kernel/tracing/events/raw_syscalls/sys_enter/hist
810
# trigger info: hist:keys=id.syscall,common_pid.execname:vals=hitcount:sort=id.syscall,hitcount:size=2048 if id == 16 [active]
812
{ id: sys_ioctl [ 16], common_pid: gmain [ 2769] } hitcount: 1
813
{ id: sys_ioctl [ 16], common_pid: evolution-addre [ 8571] } hitcount: 1
814
{ id: sys_ioctl [ 16], common_pid: gmain [ 3003] } hitcount: 1
815
{ id: sys_ioctl [ 16], common_pid: gmain [ 2781] } hitcount: 1
816
{ id: sys_ioctl [ 16], common_pid: gmain [ 2829] } hitcount: 1
817
{ id: sys_ioctl [ 16], common_pid: bash [ 8726] } hitcount: 1
818
{ id: sys_ioctl [ 16], common_pid: bash [ 8508] } hitcount: 1
819
{ id: sys_ioctl [ 16], common_pid: gmain [ 2970] } hitcount: 1
820
{ id: sys_ioctl [ 16], common_pid: gmain [ 2768] } hitcount: 1
821
.
822
.
823
.
824
{ id: sys_ioctl [ 16], common_pid: pool [ 8559] } hitcount: 45
825
{ id: sys_ioctl [ 16], common_pid: pool [ 8555] } hitcount: 48
826
{ id: sys_ioctl [ 16], common_pid: pool [ 8551] } hitcount: 48
827
{ id: sys_ioctl [ 16], common_pid: avahi-daemon [ 896] } hitcount: 66
828
{ id: sys_ioctl [ 16], common_pid: Xorg [ 1267] } hitcount: 26674
829
{ id: sys_ioctl [ 16], common_pid: compiz [ 2994] } hitcount: 73443
831
Totals:
832
Hits: 101162
833
Entries: 103
834
Dropped: 0
836
The above output shows that 'compiz' and 'Xorg' are far and away
837
the heaviest ioctl callers (which might lead to questions about
838
whether they really need to be making all those calls and to
839
possible avenues for further investigation.)
841
The compound key examples used a key and a sum value (hitcount) to
842
sort the output, but we can just as easily use two keys instead.
843
Here's an example where we use a compound key composed of the
844
common_pid and size event fields. Sorting with pid as the primary
845
key and 'size' as the secondary key allows us to display an
846
ordered summary of the recvfrom sizes, with counts, received by
847
each process::
849
# echo 'hist:key=common_pid.execname,size:val=hitcount:sort=common_pid,size' > \
850
/sys/kernel/tracing/events/syscalls/sys_enter_recvfrom/trigger
852
# cat /sys/kernel/tracing/events/syscalls/sys_enter_recvfrom/hist
853
# trigger info: hist:keys=common_pid.execname,size:vals=hitcount:sort=common_pid.execname,size:size=2048 [active]
855
{ common_pid: smbd [ 784], size: 4 } hitcount: 1
856
{ common_pid: dnsmasq [ 1412], size: 4096 } hitcount: 672
857
{ common_pid: postgres [ 1796], size: 1000 } hitcount: 6
858
{ common_pid: postgres [ 1867], size: 1000 } hitcount: 10
859
{ common_pid: bamfdaemon [ 2787], size: 28 } hitcount: 2
860
{ common_pid: bamfdaemon [ 2787], size: 14360 } hitcount: 1
861
{ common_pid: compiz [ 2994], size: 8 } hitcount: 1
862
{ common_pid: compiz [ 2994], size: 20 } hitcount: 11
863
{ common_pid: gnome-terminal [ 3199], size: 4 } hitcount: 2
864
{ common_pid: firefox [ 8817], size: 4 } hitcount: 1
865
{ common_pid: firefox [ 8817], size: 8 } hitcount: 5
866
{ common_pid: firefox [ 8817], size: 588 } hitcount: 2
867
{ common_pid: firefox [ 8817], size: 628 } hitcount: 1
868
{ common_pid: firefox [ 8817], size: 6944 } hitcount: 1
869
{ common_pid: firefox [ 8817], size: 408880 } hitcount: 2
870
{ common_pid: firefox [ 8822], size: 8 } hitcount: 2
871
{ common_pid: firefox [ 8822], size: 160 } hitcount: 2
872
{ common_pid: firefox [ 8822], size: 320 } hitcount: 2
873
{ common_pid: firefox [ 8822], size: 352 } hitcount: 1
874
.
875
.
876
.
877
{ common_pid: pool [ 8923], size: 1960 } hitcount: 10
878
{ common_pid: pool [ 8923], size: 2048 } hitcount: 10
879
{ common_pid: pool [ 8924], size: 1960 } hitcount: 10
880
{ common_pid: pool [ 8924], size: 2048 } hitcount: 10
881
{ common_pid: pool [ 8928], size: 1964 } hitcount: 4
882
{ common_pid: pool [ 8928], size: 1965 } hitcount: 2
883
{ common_pid: pool [ 8928], size: 2048 } hitcount: 6
884
{ common_pid: pool [ 8929], size: 1982 } hitcount: 1
885
{ common_pid: pool [ 8929], size: 2048 } hitcount: 1
887
Totals:
888
Hits: 2016
889
Entries: 224
890
Dropped: 0
892
The above example also illustrates the fact that although a compound
893
key is treated as a single entity for hashing purposes, the sub-keys
894
it's composed of can be accessed independently.
896
The next example uses a string field as the hash key and
897
demonstrates how you can manually pause and continue a hist trigger.
898
In this example, we'll aggregate fork counts and don't expect a
899
large number of entries in the hash table, so we'll drop it to a
900
much smaller number, say 256::
902
# echo 'hist:key=child_comm:val=hitcount:size=256' > \
903
/sys/kernel/tracing/events/sched/sched_process_fork/trigger
905
# cat /sys/kernel/tracing/events/sched/sched_process_fork/hist
906
# trigger info: hist:keys=child_comm:vals=hitcount:sort=hitcount:size=256 [active]
908
{ child_comm: dconf worker } hitcount: 1
909
{ child_comm: ibus-daemon } hitcount: 1
910
{ child_comm: whoopsie } hitcount: 1
911
{ child_comm: smbd } hitcount: 1
912
{ child_comm: gdbus } hitcount: 1
913
{ child_comm: kthreadd } hitcount: 1
914
{ child_comm: dconf worker } hitcount: 1
915
{ child_comm: evolution-alarm } hitcount: 2
916
{ child_comm: Socket Thread } hitcount: 2
917
{ child_comm: postgres } hitcount: 2
918
{ child_comm: bash } hitcount: 3
919
{ child_comm: compiz } hitcount: 3
920
{ child_comm: evolution-sourc } hitcount: 4
921
{ child_comm: dhclient } hitcount: 4
922
{ child_comm: pool } hitcount: 5
923
{ child_comm: nm-dispatcher.a } hitcount: 8
924
{ child_comm: firefox } hitcount: 8
925
{ child_comm: dbus-daemon } hitcount: 8
926
{ child_comm: glib-pacrunner } hitcount: 10
927
{ child_comm: evolution } hitcount: 23
929
Totals:
930
Hits: 89
931
Entries: 20
932
Dropped: 0
934
If we want to pause the hist trigger, we can simply append :pause to
935
the command that started the trigger. Notice that the trigger info
936
displays as [paused]::
938
# echo 'hist:key=child_comm:val=hitcount:size=256:pause' >> \
939
/sys/kernel/tracing/events/sched/sched_process_fork/trigger
941
# cat /sys/kernel/tracing/events/sched/sched_process_fork/hist
942
# trigger info: hist:keys=child_comm:vals=hitcount:sort=hitcount:size=256 [paused]
944
{ child_comm: dconf worker } hitcount: 1
945
{ child_comm: kthreadd } hitcount: 1
946
{ child_comm: dconf worker } hitcount: 1
947
{ child_comm: gdbus } hitcount: 1
948
{ child_comm: ibus-daemon } hitcount: 1
949
{ child_comm: Socket Thread } hitcount: 2
950
{ child_comm: evolution-alarm } hitcount: 2
951
{ child_comm: smbd } hitcount: 2
952
{ child_comm: bash } hitcount: 3
953
{ child_comm: whoopsie } hitcount: 3
954
{ child_comm: compiz } hitcount: 3
955
{ child_comm: evolution-sourc } hitcount: 4
956
{ child_comm: pool } hitcount: 5
957
{ child_comm: postgres } hitcount: 6
958
{ child_comm: firefox } hitcount: 8
959
{ child_comm: dhclient } hitcount: 10
960
{ child_comm: emacs } hitcount: 12
961
{ child_comm: dbus-daemon } hitcount: 20
962
{ child_comm: nm-dispatcher.a } hitcount: 20
963
{ child_comm: evolution } hitcount: 35
964
{ child_comm: glib-pacrunner } hitcount: 59
966
Totals:
967
Hits: 199
968
Entries: 21
969
Dropped: 0
971
To manually continue having the trigger aggregate events, append
972
:cont instead. Notice that the trigger info displays as [active]
973
again, and the data has changed::
975
# echo 'hist:key=child_comm:val=hitcount:size=256:cont' >> \
976
/sys/kernel/tracing/events/sched/sched_process_fork/trigger
978
# cat /sys/kernel/tracing/events/sched/sched_process_fork/hist
979
# trigger info: hist:keys=child_comm:vals=hitcount:sort=hitcount:size=256 [active]
981
{ child_comm: dconf worker } hitcount: 1
982
{ child_comm: dconf worker } hitcount: 1
983
{ child_comm: kthreadd } hitcount: 1
984
{ child_comm: gdbus } hitcount: 1
985
{ child_comm: ibus-daemon } hitcount: 1
986
{ child_comm: Socket Thread } hitcount: 2
987
{ child_comm: evolution-alarm } hitcount: 2
988
{ child_comm: smbd } hitcount: 2
989
{ child_comm: whoopsie } hitcount: 3
990
{ child_comm: compiz } hitcount: 3
991
{ child_comm: evolution-sourc } hitcount: 4
992
{ child_comm: bash } hitcount: 5
993
{ child_comm: pool } hitcount: 5
994
{ child_comm: postgres } hitcount: 6
995
{ child_comm: firefox } hitcount: 8
996
{ child_comm: dhclient } hitcount: 11
997
{ child_comm: emacs } hitcount: 12
998
{ child_comm: dbus-daemon } hitcount: 22
999
{ child_comm: nm-dispatcher.a } hitcount: 22
1000
{ child_comm: evolution } hitcount: 35
1001
{ child_comm: glib-pacrunner } hitcount: 59
1003
Totals:
1004
Hits: 206
1005
Entries: 21
1006
Dropped: 0
1008
The previous example showed how to start and stop a hist trigger by
1009
appending 'pause' and 'continue' to the hist trigger command. A
1010
hist trigger can also be started in a paused state by initially
1011
starting the trigger with ':pause' appended. This allows you to
1012
start the trigger only when you're ready to start collecting data
1013
and not before. For example, you could start the trigger in a
1014
paused state, then unpause it and do something you want to measure,
1015
then pause the trigger again when done.
1017
Of course, doing this manually can be difficult and error-prone, but
1018
it is possible to automatically start and stop a hist trigger based
1019
on some condition, via the enable_hist and disable_hist triggers.
1021
For example, suppose we wanted to take a look at the relative
1022
weights in terms of skb length for each callpath that leads to a
1023
netif_receive_skb event when downloading a decent-sized file using
1024
wget.
1026
First we set up an initially paused stacktrace trigger on the
1027
netif_receive_skb event::
1029
# echo 'hist:key=common_stacktrace:vals=len:pause' > \
1030
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
1032
Next, we set up an 'enable_hist' trigger on the sched_process_exec
1033
event, with an 'if filename==/usr/bin/wget' filter. The effect of
1034
this new trigger is that it will 'unpause' the hist trigger we just
1035
set up on netif_receive_skb if and only if it sees a
1036
sched_process_exec event with a filename of '/usr/bin/wget'. When
1037
that happens, all netif_receive_skb events are aggregated into a
1038
hash table keyed on stacktrace::
1040
# echo 'enable_hist:net:netif_receive_skb if filename==/usr/bin/wget' > \
1041
/sys/kernel/tracing/events/sched/sched_process_exec/trigger
1043
The aggregation continues until the netif_receive_skb is paused
1044
again, which is what the following disable_hist event does by
1045
creating a similar setup on the sched_process_exit event, using the
1046
filter 'comm==wget'::
1048
# echo 'disable_hist:net:netif_receive_skb if comm==wget' > \
1049
/sys/kernel/tracing/events/sched/sched_process_exit/trigger
1051
Whenever a process exits and the comm field of the disable_hist
1052
trigger filter matches 'comm==wget', the netif_receive_skb hist
1053
trigger is disabled.
1055
The overall effect is that netif_receive_skb events are aggregated
1056
into the hash table for only the duration of the wget. Executing a
1057
wget command and then listing the 'hist' file will display the
1058
output generated by the wget command::
1060
$ wget https://www.kernel.org/pub/linux/kernel/v3.x/patch-3.19.xz
1062
# cat /sys/kernel/tracing/events/net/netif_receive_skb/hist
1063
# trigger info: hist:keys=common_stacktrace:vals=len:sort=hitcount:size=2048 [paused]
1065
{ common_stacktrace:
1066
__netif_receive_skb_core+0x46d/0x990
1067
__netif_receive_skb+0x18/0x60
1068
netif_receive_skb_internal+0x23/0x90
1069
napi_gro_receive+0xc8/0x100
1070
ieee80211_deliver_skb+0xd6/0x270 [mac80211]
1071
ieee80211_rx_handlers+0xccf/0x22f0 [mac80211]
1072
ieee80211_prepare_and_rx_handle+0x4e7/0xc40 [mac80211]
1073
ieee80211_rx+0x31d/0x900 [mac80211]
1074
iwlagn_rx_reply_rx+0x3db/0x6f0 [iwldvm]
1075
iwl_rx_dispatch+0x8e/0xf0 [iwldvm]
1076
iwl_pcie_irq_handler+0xe3c/0x12f0 [iwlwifi]
1077
irq_thread_fn+0x20/0x50
1078
irq_thread+0x11f/0x150
1079
kthread+0xd2/0xf0
1080
ret_from_fork+0x42/0x70
1081
} hitcount: 85 len: 28884
1082
{ common_stacktrace:
1083
__netif_receive_skb_core+0x46d/0x990
1084
__netif_receive_skb+0x18/0x60
1085
netif_receive_skb_internal+0x23/0x90
1086
napi_gro_complete+0xa4/0xe0
1087
dev_gro_receive+0x23a/0x360
1088
napi_gro_receive+0x30/0x100
1089
ieee80211_deliver_skb+0xd6/0x270 [mac80211]
1090
ieee80211_rx_handlers+0xccf/0x22f0 [mac80211]
1091
ieee80211_prepare_and_rx_handle+0x4e7/0xc40 [mac80211]
1092
ieee80211_rx+0x31d/0x900 [mac80211]
1093
iwlagn_rx_reply_rx+0x3db/0x6f0 [iwldvm]
1094
iwl_rx_dispatch+0x8e/0xf0 [iwldvm]
1095
iwl_pcie_irq_handler+0xe3c/0x12f0 [iwlwifi]
1096
irq_thread_fn+0x20/0x50
1097
irq_thread+0x11f/0x150
1098
kthread+0xd2/0xf0
1099
} hitcount: 98 len: 664329
1100
{ common_stacktrace:
1101
__netif_receive_skb_core+0x46d/0x990
1102
__netif_receive_skb+0x18/0x60
1103
process_backlog+0xa8/0x150
1104
net_rx_action+0x15d/0x340
1105
__do_softirq+0x114/0x2c0
1106
do_softirq_own_stack+0x1c/0x30
1107
do_softirq+0x65/0x70
1108
__local_bh_enable_ip+0xb5/0xc0
1109
ip_finish_output+0x1f4/0x840
1110
ip_output+0x6b/0xc0
1111
ip_local_out_sk+0x31/0x40
1112
ip_send_skb+0x1a/0x50
1113
udp_send_skb+0x173/0x2a0
1114
udp_sendmsg+0x2bf/0x9f0
1115
inet_sendmsg+0x64/0xa0
1116
sock_sendmsg+0x3d/0x50
1117
} hitcount: 115 len: 13030
1118
{ common_stacktrace:
1119
__netif_receive_skb_core+0x46d/0x990
1120
__netif_receive_skb+0x18/0x60
1121
netif_receive_skb_internal+0x23/0x90
1122
napi_gro_complete+0xa4/0xe0
1123
napi_gro_flush+0x6d/0x90
1124
iwl_pcie_irq_handler+0x92a/0x12f0 [iwlwifi]
1125
irq_thread_fn+0x20/0x50
1126
irq_thread+0x11f/0x150
1127
kthread+0xd2/0xf0
1128
ret_from_fork+0x42/0x70
1129
} hitcount: 934 len: 5512212
1131
Totals:
1132
Hits: 1232
1133
Entries: 4
1134
Dropped: 0
1136
The above shows all the netif_receive_skb callpaths and their total
1137
lengths for the duration of the wget command.
1139
The 'clear' hist trigger param can be used to clear the hash table.
1140
Suppose we wanted to try another run of the previous example but
1141
this time also wanted to see the complete list of events that went
1142
into the histogram. In order to avoid having to set everything up
1143
again, we can just clear the histogram first::
1145
# echo 'hist:key=common_stacktrace:vals=len:clear' >> \
1146
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
1148
Just to verify that it is in fact cleared, here's what we now see in
1149
the hist file::
1151
# cat /sys/kernel/tracing/events/net/netif_receive_skb/hist
1152
# trigger info: hist:keys=common_stacktrace:vals=len:sort=hitcount:size=2048 [paused]
1154
Totals:
1155
Hits: 0
1156
Entries: 0
1157
Dropped: 0
1159
Since we want to see the detailed list of every netif_receive_skb
1160
event occurring during the new run, which are in fact the same
1161
events being aggregated into the hash table, we add some additional
1162
'enable_event' events to the triggering sched_process_exec and
1163
sched_process_exit events as such::
1165
# echo 'enable_event:net:netif_receive_skb if filename==/usr/bin/wget' > \
1166
/sys/kernel/tracing/events/sched/sched_process_exec/trigger
1168
# echo 'disable_event:net:netif_receive_skb if comm==wget' > \
1169
/sys/kernel/tracing/events/sched/sched_process_exit/trigger
1171
If you read the trigger files for the sched_process_exec and
1172
sched_process_exit triggers, you should see two triggers for each:
1173
one enabling/disabling the hist aggregation and the other
1174
enabling/disabling the logging of events::
1176
# cat /sys/kernel/tracing/events/sched/sched_process_exec/trigger
1177
enable_event:net:netif_receive_skb:unlimited if filename==/usr/bin/wget
1178
enable_hist:net:netif_receive_skb:unlimited if filename==/usr/bin/wget
1180
# cat /sys/kernel/tracing/events/sched/sched_process_exit/trigger
1181
enable_event:net:netif_receive_skb:unlimited if comm==wget
1182
disable_hist:net:netif_receive_skb:unlimited if comm==wget
1184
In other words, whenever either of the sched_process_exec or
1185
sched_process_exit events is hit and matches 'wget', it enables or
1186
disables both the histogram and the event log, and what you end up
1187
with is a hash table and set of events just covering the specified
1188
duration. Run the wget command again::
1190
$ wget https://www.kernel.org/pub/linux/kernel/v3.x/patch-3.19.xz
1192
Displaying the 'hist' file should show something similar to what you
1193
saw in the last run, but this time you should also see the
1194
individual events in the trace file::
1196
# cat /sys/kernel/tracing/trace
1198
# tracer: nop
1199
#
1200
# entries-in-buffer/entries-written: 183/1426 #P:4
1201
#
1202
# _-----=> irqs-off
1203
# / _----=> need-resched
1204
# | / _---=> hardirq/softirq
1205
# || / _--=> preempt-depth
1206
# ||| / delay
1207
# TASK-PID CPU# |||| TIMESTAMP FUNCTION
1208
# | | | |||| | |
1209
wget-15108 [000] ..s1 31769.606929: netif_receive_skb: dev=lo skbaddr=ffff88009c353100 len=60
1210
wget-15108 [000] ..s1 31769.606999: netif_receive_skb: dev=lo skbaddr=ffff88009c353200 len=60
1211
dnsmasq-1382 [000] ..s1 31769.677652: netif_receive_skb: dev=lo skbaddr=ffff88009c352b00 len=130
1212
dnsmasq-1382 [000] ..s1 31769.685917: netif_receive_skb: dev=lo skbaddr=ffff88009c352200 len=138
1213
##### CPU 2 buffer started ####
1214
irq/29-iwlwifi-559 [002] ..s. 31772.031529: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d433d00 len=2948
1215
irq/29-iwlwifi-559 [002] ..s. 31772.031572: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d432200 len=1500
1216
irq/29-iwlwifi-559 [002] ..s. 31772.032196: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d433100 len=2948
1217
irq/29-iwlwifi-559 [002] ..s. 31772.032761: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d433000 len=2948
1218
irq/29-iwlwifi-559 [002] ..s. 31772.033220: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d432e00 len=1500
1219
.
1220
.
1221
.
1223
The following example demonstrates how multiple hist triggers can be
1224
attached to a given event. This capability can be useful for
1225
creating a set of different summaries derived from the same set of
1226
events, or for comparing the effects of different filters, among
1227
other things::
1229
# echo 'hist:keys=skbaddr.hex:vals=len if len < 0' >> \
1230
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
1231
# echo 'hist:keys=skbaddr.hex:vals=len if len > 4096' >> \
1232
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
1233
# echo 'hist:keys=skbaddr.hex:vals=len if len == 256' >> \
1234
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
1235
# echo 'hist:keys=skbaddr.hex:vals=len' >> \
1236
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
1237
# echo 'hist:keys=len:vals=common_preempt_count' >> \
1238
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
1240
The above set of commands create four triggers differing only in
1241
their filters, along with a completely different though fairly
1242
nonsensical trigger. Note that in order to append multiple hist
1243
triggers to the same file, you should use the '>>' operator to
1244
append them ('>' will also add the new hist trigger, but will remove
1245
any existing hist triggers beforehand).
1247
Displaying the contents of the 'hist' file for the event shows the
1248
contents of all five histograms::
1250
# cat /sys/kernel/tracing/events/net/netif_receive_skb/hist
1252
# event histogram
1253
#
1254
# trigger info: hist:keys=len:vals=hitcount,common_preempt_count:sort=hitcount:size=2048 [active]
1255
#
1257
{ len: 176 } hitcount: 1 common_preempt_count: 0
1258
{ len: 223 } hitcount: 1 common_preempt_count: 0
1259
{ len: 4854 } hitcount: 1 common_preempt_count: 0
1260
{ len: 395 } hitcount: 1 common_preempt_count: 0
1261
{ len: 177 } hitcount: 1 common_preempt_count: 0
1262
{ len: 446 } hitcount: 1 common_preempt_count: 0
1263
{ len: 1601 } hitcount: 1 common_preempt_count: 0
1264
.
1265
.
1266
.
1267
{ len: 1280 } hitcount: 66 common_preempt_count: 0
1268
{ len: 116 } hitcount: 81 common_preempt_count: 40
1269
{ len: 708 } hitcount: 112 common_preempt_count: 0
1270
{ len: 46 } hitcount: 221 common_preempt_count: 0
1271
{ len: 1264 } hitcount: 458 common_preempt_count: 0
1273
Totals:
1274
Hits: 1428
1275
Entries: 147
1276
Dropped: 0
1279
# event histogram
1280
#
1281
# trigger info: hist:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 [active]
1282
#
1284
{ skbaddr: ffff8800baee5e00 } hitcount: 1 len: 130
1285
{ skbaddr: ffff88005f3d5600 } hitcount: 1 len: 1280
1286
{ skbaddr: ffff88005f3d4900 } hitcount: 1 len: 1280
1287
{ skbaddr: ffff88009fed6300 } hitcount: 1 len: 115
1288
{ skbaddr: ffff88009fe0ad00 } hitcount: 1 len: 115
1289
{ skbaddr: ffff88008cdb1900 } hitcount: 1 len: 46
1290
{ skbaddr: ffff880064b5ef00 } hitcount: 1 len: 118
1291
{ skbaddr: ffff880044e3c700 } hitcount: 1 len: 60
1292
{ skbaddr: ffff880100065900 } hitcount: 1 len: 46
1293
{ skbaddr: ffff8800d46bd500 } hitcount: 1 len: 116
1294
{ skbaddr: ffff88005f3d5f00 } hitcount: 1 len: 1280
1295
{ skbaddr: ffff880100064700 } hitcount: 1 len: 365
1296
{ skbaddr: ffff8800badb6f00 } hitcount: 1 len: 60
1297
.
1298
.
1299
.
1300
{ skbaddr: ffff88009fe0be00 } hitcount: 27 len: 24677
1301
{ skbaddr: ffff88009fe0a400 } hitcount: 27 len: 23052
1302
{ skbaddr: ffff88009fe0b700 } hitcount: 31 len: 25589
1303
{ skbaddr: ffff88009fe0b600 } hitcount: 32 len: 27326
1304
{ skbaddr: ffff88006a462800 } hitcount: 68 len: 71678
1305
{ skbaddr: ffff88006a463700 } hitcount: 70 len: 72678
1306
{ skbaddr: ffff88006a462b00 } hitcount: 71 len: 77589
1307
{ skbaddr: ffff88006a463600 } hitcount: 73 len: 71307
1308
{ skbaddr: ffff88006a462200 } hitcount: 81 len: 81032
1310
Totals:
1311
Hits: 1451
1312
Entries: 318
1313
Dropped: 0
1316
# event histogram
1317
#
1318
# trigger info: hist:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 if len == 256 [active]
1319
#
1322
Totals:
1323
Hits: 0
1324
Entries: 0
1325
Dropped: 0
1328
# event histogram
1329
#
1330
# trigger info: hist:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 if len > 4096 [active]
1331
#
1333
{ skbaddr: ffff88009fd2c300 } hitcount: 1 len: 7212
1334
{ skbaddr: ffff8800d2bcce00 } hitcount: 1 len: 7212
1335
{ skbaddr: ffff8800d2bcd700 } hitcount: 1 len: 7212
1336
{ skbaddr: ffff8800d2bcda00 } hitcount: 1 len: 21492
1337
{ skbaddr: ffff8800ae2e2d00 } hitcount: 1 len: 7212
1338
{ skbaddr: ffff8800d2bcdb00 } hitcount: 1 len: 7212
1339
{ skbaddr: ffff88006a4df500 } hitcount: 1 len: 4854
1340
{ skbaddr: ffff88008ce47b00 } hitcount: 1 len: 18636
1341
{ skbaddr: ffff8800ae2e2200 } hitcount: 1 len: 12924
1342
{ skbaddr: ffff88005f3e1000 } hitcount: 1 len: 4356
1343
{ skbaddr: ffff8800d2bcdc00 } hitcount: 2 len: 24420
1344
{ skbaddr: ffff8800d2bcc200 } hitcount: 2 len: 12996
1346
Totals:
1347
Hits: 14
1348
Entries: 12
1349
Dropped: 0
1352
# event histogram
1353
#
1354
# trigger info: hist:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 if len < 0 [active]
1355
#
1358
Totals:
1359
Hits: 0
1360
Entries: 0
1361
Dropped: 0
1363
Named triggers can be used to have triggers share a common set of
1364
histogram data. This capability is mostly useful for combining the
1365
output of events generated by tracepoints contained inside inline
1366
functions, but names can be used in a hist trigger on any event.
1367
For example, these two triggers when hit will update the same 'len'
1368
field in the shared 'foo' histogram data::
1370
# echo 'hist:name=foo:keys=skbaddr.hex:vals=len' > \
1371
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
1372
# echo 'hist:name=foo:keys=skbaddr.hex:vals=len' > \
1373
/sys/kernel/tracing/events/net/netif_rx/trigger
1375
You can see that they're updating common histogram data by reading
1376
each event's hist files at the same time::
1378
# cat /sys/kernel/tracing/events/net/netif_receive_skb/hist;
1379
cat /sys/kernel/tracing/events/net/netif_rx/hist
1381
# event histogram
1382
#
1383
# trigger info: hist:name=foo:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 [active]
1384
#
1386
{ skbaddr: ffff88000ad53500 } hitcount: 1 len: 46
1387
{ skbaddr: ffff8800af5a1500 } hitcount: 1 len: 76
1388
{ skbaddr: ffff8800d62a1900 } hitcount: 1 len: 46
1389
{ skbaddr: ffff8800d2bccb00 } hitcount: 1 len: 468
1390
{ skbaddr: ffff8800d3c69900 } hitcount: 1 len: 46
1391
{ skbaddr: ffff88009ff09100 } hitcount: 1 len: 52
1392
{ skbaddr: ffff88010f13ab00 } hitcount: 1 len: 168
1393
{ skbaddr: ffff88006a54f400 } hitcount: 1 len: 46
1394
{ skbaddr: ffff8800d2bcc500 } hitcount: 1 len: 260
1395
{ skbaddr: ffff880064505000 } hitcount: 1 len: 46
1396
{ skbaddr: ffff8800baf24e00 } hitcount: 1 len: 32
1397
{ skbaddr: ffff88009fe0ad00 } hitcount: 1 len: 46
1398
{ skbaddr: ffff8800d3edff00 } hitcount: 1 len: 44
1399
{ skbaddr: ffff88009fe0b400 } hitcount: 1 len: 168
1400
{ skbaddr: ffff8800a1c55a00 } hitcount: 1 len: 40
1401
{ skbaddr: ffff8800d2bcd100 } hitcount: 1 len: 40
1402
{ skbaddr: ffff880064505f00 } hitcount: 1 len: 174
1403
{ skbaddr: ffff8800a8bff200 } hitcount: 1 len: 160
1404
{ skbaddr: ffff880044e3cc00 } hitcount: 1 len: 76
1405
{ skbaddr: ffff8800a8bfe700 } hitcount: 1 len: 46
1406
{ skbaddr: ffff8800d2bcdc00 } hitcount: 1 len: 32
1407
{ skbaddr: ffff8800a1f64800 } hitcount: 1 len: 46
1408
{ skbaddr: ffff8800d2bcde00 } hitcount: 1 len: 988
1409
{ skbaddr: ffff88006a5dea00 } hitcount: 1 len: 46
1410
{ skbaddr: ffff88002e37a200 } hitcount: 1 len: 44
1411
{ skbaddr: ffff8800a1f32c00 } hitcount: 2 len: 676
1412
{ skbaddr: ffff88000ad52600 } hitcount: 2 len: 107
1413
{ skbaddr: ffff8800a1f91e00 } hitcount: 2 len: 92
1414
{ skbaddr: ffff8800af5a0200 } hitcount: 2 len: 142
1415
{ skbaddr: ffff8800d2bcc600 } hitcount: 2 len: 220
1416
{ skbaddr: ffff8800ba36f500 } hitcount: 2 len: 92
1417
{ skbaddr: ffff8800d021f800 } hitcount: 2 len: 92
1418
{ skbaddr: ffff8800a1f33600 } hitcount: 2 len: 675
1419
{ skbaddr: ffff8800a8bfff00 } hitcount: 3 len: 138
1420
{ skbaddr: ffff8800d62a1300 } hitcount: 3 len: 138
1421
{ skbaddr: ffff88002e37a100 } hitcount: 4 len: 184
1422
{ skbaddr: ffff880064504400 } hitcount: 4 len: 184
1423
{ skbaddr: ffff8800a8bfec00 } hitcount: 4 len: 184
1424
{ skbaddr: ffff88000ad53700 } hitcount: 5 len: 230
1425
{ skbaddr: ffff8800d2bcdb00 } hitcount: 5 len: 196
1426
{ skbaddr: ffff8800a1f90000 } hitcount: 6 len: 276
1427
{ skbaddr: ffff88006a54f900 } hitcount: 6 len: 276
1429
Totals:
1430
Hits: 81
1431
Entries: 42
1432
Dropped: 0
1433
# event histogram
1434
#
1435
# trigger info: hist:name=foo:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 [active]
1436
#
1438
{ skbaddr: ffff88000ad53500 } hitcount: 1 len: 46
1439
{ skbaddr: ffff8800af5a1500 } hitcount: 1 len: 76
1440
{ skbaddr: ffff8800d62a1900 } hitcount: 1 len: 46
1441
{ skbaddr: ffff8800d2bccb00 } hitcount: 1 len: 468
1442
{ skbaddr: ffff8800d3c69900 } hitcount: 1 len: 46
1443
{ skbaddr: ffff88009ff09100 } hitcount: 1 len: 52
1444
{ skbaddr: ffff88010f13ab00 } hitcount: 1 len: 168
1445
{ skbaddr: ffff88006a54f400 } hitcount: 1 len: 46
1446
{ skbaddr: ffff8800d2bcc500 } hitcount: 1 len: 260
1447
{ skbaddr: ffff880064505000 } hitcount: 1 len: 46
1448
{ skbaddr: ffff8800baf24e00 } hitcount: 1 len: 32
1449
{ skbaddr: ffff88009fe0ad00 } hitcount: 1 len: 46
1450
{ skbaddr: ffff8800d3edff00 } hitcount: 1 len: 44
1451
{ skbaddr: ffff88009fe0b400 } hitcount: 1 len: 168
1452
{ skbaddr: ffff8800a1c55a00 } hitcount: 1 len: 40
1453
{ skbaddr: ffff8800d2bcd100 } hitcount: 1 len: 40
1454
{ skbaddr: ffff880064505f00 } hitcount: 1 len: 174
1455
{ skbaddr: ffff8800a8bff200 } hitcount: 1 len: 160
1456
{ skbaddr: ffff880044e3cc00 } hitcount: 1 len: 76
1457
{ skbaddr: ffff8800a8bfe700 } hitcount: 1 len: 46
1458
{ skbaddr: ffff8800d2bcdc00 } hitcount: 1 len: 32
1459
{ skbaddr: ffff8800a1f64800 } hitcount: 1 len: 46
1460
{ skbaddr: ffff8800d2bcde00 } hitcount: 1 len: 988
1461
{ skbaddr: ffff88006a5dea00 } hitcount: 1 len: 46
1462
{ skbaddr: ffff88002e37a200 } hitcount: 1 len: 44
1463
{ skbaddr: ffff8800a1f32c00 } hitcount: 2 len: 676
1464
{ skbaddr: ffff88000ad52600 } hitcount: 2 len: 107
1465
{ skbaddr: ffff8800a1f91e00 } hitcount: 2 len: 92
1466
{ skbaddr: ffff8800af5a0200 } hitcount: 2 len: 142
1467
{ skbaddr: ffff8800d2bcc600 } hitcount: 2 len: 220
1468
{ skbaddr: ffff8800ba36f500 } hitcount: 2 len: 92
1469
{ skbaddr: ffff8800d021f800 } hitcount: 2 len: 92
1470
{ skbaddr: ffff8800a1f33600 } hitcount: 2 len: 675
1471
{ skbaddr: ffff8800a8bfff00 } hitcount: 3 len: 138
1472
{ skbaddr: ffff8800d62a1300 } hitcount: 3 len: 138
1473
{ skbaddr: ffff88002e37a100 } hitcount: 4 len: 184
1474
{ skbaddr: ffff880064504400 } hitcount: 4 len: 184
1475
{ skbaddr: ffff8800a8bfec00 } hitcount: 4 len: 184
1476
{ skbaddr: ffff88000ad53700 } hitcount: 5 len: 230
1477
{ skbaddr: ffff8800d2bcdb00 } hitcount: 5 len: 196
1478
{ skbaddr: ffff8800a1f90000 } hitcount: 6 len: 276
1479
{ skbaddr: ffff88006a54f900 } hitcount: 6 len: 276
1481
Totals:
1482
Hits: 81
1483
Entries: 42
1484
Dropped: 0
1486
And here's an example that shows how to combine histogram data from
1487
any two events even if they don't share any 'compatible' fields
1488
other than 'hitcount' and 'common_stacktrace'. These commands create a
1489
couple of triggers named 'bar' using those fields::
1491
# echo 'hist:name=bar:key=common_stacktrace:val=hitcount' > \
1492
/sys/kernel/tracing/events/sched/sched_process_fork/trigger
1493
# echo 'hist:name=bar:key=common_stacktrace:val=hitcount' > \
1494
/sys/kernel/tracing/events/net/netif_rx/trigger
1496
And displaying the output of either shows some interesting if
1497
somewhat confusing output::
1499
# cat /sys/kernel/tracing/events/sched/sched_process_fork/hist
1500
# cat /sys/kernel/tracing/events/net/netif_rx/hist
1502
# event histogram
1503
#
1504
# trigger info: hist:name=bar:keys=common_stacktrace:vals=hitcount:sort=hitcount:size=2048 [active]
1505
#
1507
{ common_stacktrace:
1508
kernel_clone+0x18e/0x330
1509
kernel_thread+0x29/0x30
1510
kthreadd+0x154/0x1b0
1511
ret_from_fork+0x3f/0x70
1512
} hitcount: 1
1513
{ common_stacktrace:
1514
netif_rx_internal+0xb2/0xd0
1515
netif_rx_ni+0x20/0x70
1516
dev_loopback_xmit+0xaa/0xd0
1517
ip_mc_output+0x126/0x240
1518
ip_local_out_sk+0x31/0x40
1519
igmp_send_report+0x1e9/0x230
1520
igmp_timer_expire+0xe9/0x120
1521
call_timer_fn+0x39/0xf0
1522
run_timer_softirq+0x1e1/0x290
1523
__do_softirq+0xfd/0x290
1524
irq_exit+0x98/0xb0
1525
smp_apic_timer_interrupt+0x4a/0x60
1526
apic_timer_interrupt+0x6d/0x80
1527
cpuidle_enter+0x17/0x20
1528
call_cpuidle+0x3b/0x60
1529
cpu_startup_entry+0x22d/0x310
1530
} hitcount: 1
1531
{ common_stacktrace:
1532
netif_rx_internal+0xb2/0xd0
1533
netif_rx_ni+0x20/0x70
1534
dev_loopback_xmit+0xaa/0xd0
1535
ip_mc_output+0x17f/0x240
1536
ip_local_out_sk+0x31/0x40
1537
ip_send_skb+0x1a/0x50
1538
udp_send_skb+0x13e/0x270
1539
udp_sendmsg+0x2bf/0x980
1540
inet_sendmsg+0x67/0xa0
1541
sock_sendmsg+0x38/0x50
1542
SYSC_sendto+0xef/0x170
1543
SyS_sendto+0xe/0x10
1544
entry_SYSCALL_64_fastpath+0x12/0x6a
1545
} hitcount: 2
1546
{ common_stacktrace:
1547
netif_rx_internal+0xb2/0xd0
1548
netif_rx+0x1c/0x60
1549
loopback_xmit+0x6c/0xb0
1550
dev_hard_start_xmit+0x219/0x3a0
1551
__dev_queue_xmit+0x415/0x4f0
1552
dev_queue_xmit_sk+0x13/0x20
1553
ip_finish_output2+0x237/0x340
1554
ip_finish_output+0x113/0x1d0
1555
ip_output+0x66/0xc0
1556
ip_local_out_sk+0x31/0x40
1557
ip_send_skb+0x1a/0x50
1558
udp_send_skb+0x16d/0x270
1559
udp_sendmsg+0x2bf/0x980
1560
inet_sendmsg+0x67/0xa0
1561
sock_sendmsg+0x38/0x50
1562
___sys_sendmsg+0x14e/0x270
1563
} hitcount: 76
1564
{ common_stacktrace:
1565
netif_rx_internal+0xb2/0xd0
1566
netif_rx+0x1c/0x60
1567
loopback_xmit+0x6c/0xb0
1568
dev_hard_start_xmit+0x219/0x3a0
1569
__dev_queue_xmit+0x415/0x4f0
1570
dev_queue_xmit_sk+0x13/0x20
1571
ip_finish_output2+0x237/0x340
1572
ip_finish_output+0x113/0x1d0
1573
ip_output+0x66/0xc0
1574
ip_local_out_sk+0x31/0x40
1575
ip_send_skb+0x1a/0x50
1576
udp_send_skb+0x16d/0x270
1577
udp_sendmsg+0x2bf/0x980
1578
inet_sendmsg+0x67/0xa0
1579
sock_sendmsg+0x38/0x50
1580
___sys_sendmsg+0x269/0x270
1581
} hitcount: 77
1582
{ common_stacktrace:
1583
netif_rx_internal+0xb2/0xd0
1584
netif_rx+0x1c/0x60
1585
loopback_xmit+0x6c/0xb0
1586
dev_hard_start_xmit+0x219/0x3a0
1587
__dev_queue_xmit+0x415/0x4f0
1588
dev_queue_xmit_sk+0x13/0x20
1589
ip_finish_output2+0x237/0x340
1590
ip_finish_output+0x113/0x1d0
1591
ip_output+0x66/0xc0
1592
ip_local_out_sk+0x31/0x40
1593
ip_send_skb+0x1a/0x50
1594
udp_send_skb+0x16d/0x270
1595
udp_sendmsg+0x2bf/0x980
1596
inet_sendmsg+0x67/0xa0
1597
sock_sendmsg+0x38/0x50
1598
SYSC_sendto+0xef/0x170
1599
} hitcount: 88
1600
{ common_stacktrace:
1601
kernel_clone+0x18e/0x330
1602
SyS_clone+0x19/0x20
1603
entry_SYSCALL_64_fastpath+0x12/0x6a
1604
} hitcount: 244
1606
Totals:
1607
Hits: 489
1608
Entries: 7
1609
Dropped: 0
1611
2.4. Inter-event hist triggers
1612
------------------------------
1614
Inter-event hist triggers are hist triggers that combine values from
1615
one or more other events and create a histogram using that data. Data
1616
from an inter-event histogram can in turn become the source for
1617
further combined histograms, thus providing a chain of related
1618
histograms, which is important for some applications.
1620
The most important example of an inter-event quantity that can be used
1621
in this manner is latency, which is simply a difference in timestamps
1622
between two events. Although latency is the most important
1623
inter-event quantity, note that because the support is completely
1624
general across the trace event subsystem, any event field can be used
1625
in an inter-event quantity.
1627
An example of a histogram that combines data from other histograms
1628
into a useful chain would be a 'wakeupswitch latency' histogram that
1629
combines a 'wakeup latency' histogram and a 'switch latency'
1630
histogram.
1632
Normally, a hist trigger specification consists of a (possibly
1633
compound) key along with one or more numeric values, which are
1634
continually updated sums associated with that key. A histogram
1635
specification in this case consists of individual key and value
1636
specifications that refer to trace event fields associated with a
1637
single event type.
1639
The inter-event hist trigger extension allows fields from multiple
1640
events to be referenced and combined into a multi-event histogram
1641
specification. In support of this overall goal, a few enabling
1642
features have been added to the hist trigger support:
1644
- In order to compute an inter-event quantity, a value from one
1645
event needs to saved and then referenced from another event. This
1646
requires the introduction of support for histogram 'variables'.
1648
- The computation of inter-event quantities and their combination
1649
require some minimal amount of support for applying simple
1650
expressions to variables (+ and -).
1652
- A histogram consisting of inter-event quantities isn't logically a
1653
histogram on either event (so having the 'hist' file for either
1654
event host the histogram output doesn't really make sense). To
1655
address the idea that the histogram is associated with a
1656
combination of events, support is added allowing the creation of
1657
'synthetic' events that are events derived from other events.
1658
These synthetic events are full-fledged events just like any other
1659
and can be used as such, as for instance to create the
1660
'combination' histograms mentioned previously.
1662
- A set of 'actions' can be associated with histogram entries -
1663
these can be used to generate the previously mentioned synthetic
1664
events, but can also be used for other purposes, such as for
1665
example saving context when a 'max' latency has been hit.
1667
- Trace events don't have a 'timestamp' associated with them, but
1668
there is an implicit timestamp saved along with an event in the
1669
underlying ftrace ring buffer. This timestamp is now exposed as a
1670
a synthetic field named 'common_timestamp' which can be used in
1671
histograms as if it were any other event field; it isn't an actual
1672
field in the trace format but rather is a synthesized value that
1673
nonetheless can be used as if it were an actual field. By default
1674
it is in units of nanoseconds; appending '.usecs' to a
1675
common_timestamp field changes the units to microseconds.
1677
A note on inter-event timestamps: If common_timestamp is used in a
1678
histogram, the trace buffer is automatically switched over to using
1679
absolute timestamps and the "global" trace clock, in order to avoid
1680
bogus timestamp differences with other clocks that aren't coherent
1681
across CPUs. This can be overridden by specifying one of the other
1682
trace clocks instead, using the "clock=XXX" hist trigger attribute,
1683
where XXX is any of the clocks listed in the tracing/trace_clock
1684
pseudo-file.
1686
These features are described in more detail in the following sections.
1688
2.5. Histogram Variables
1689
------------------------
1691
Variables are simply named locations used for saving and retrieving
1692
values between matching events. A 'matching' event is defined as an
1693
event that has a matching key - if a variable is saved for a histogram
1694
entry corresponding to that key, any subsequent event with a matching
1695
key can access that variable.
1697
A variable's value is normally available to any subsequent event until
1698
it is set to something else by a subsequent event. The one exception
1699
to that rule is that any variable used in an expression is essentially
1700
'read-once' - once it's used by an expression in a subsequent event,
1701
it's reset to its 'unset' state, which means it can't be used again
1702
unless it's set again. This ensures not only that an event doesn't
1703
use an uninitialized variable in a calculation, but that that variable
1704
is used only once and not for any unrelated subsequent match.
1706
The basic syntax for saving a variable is to simply prefix a unique
1707
variable name not corresponding to any keyword along with an '=' sign
1708
to any event field.
1710
Either keys or values can be saved and retrieved in this way. This
1711
creates a variable named 'ts0' for a histogram entry with the key
1712
'next_pid'::
1714
# echo 'hist:keys=next_pid:vals=$ts0:ts0=common_timestamp ... >> \
1715
event/trigger
1717
The ts0 variable can be accessed by any subsequent event having the
1718
same pid as 'next_pid'.
1720
Variable references are formed by prepending the variable name with
1721
the '$' sign. Thus for example, the ts0 variable above would be
1722
referenced as '$ts0' in expressions.
1724
Because 'vals=' is used, the common_timestamp variable value above
1725
will also be summed as a normal histogram value would (though for a
1726
timestamp it makes little sense).
1728
The below shows that a key value can also be saved in the same way::
1730
# echo 'hist:timer_pid=common_pid:key=timer_pid ...' >> event/trigger
1732
If a variable isn't a key variable or prefixed with 'vals=', the
1733
associated event field will be saved in a variable but won't be summed
1734
as a value::
1736
# echo 'hist:keys=next_pid:ts1=common_timestamp ...' >> event/trigger
1738
Multiple variables can be assigned at the same time. The below would
1739
result in both ts0 and b being created as variables, with both
1740
common_timestamp and field1 additionally being summed as values::
1742
# echo 'hist:keys=pid:vals=$ts0,$b:ts0=common_timestamp,b=field1 ...' >> \
1743
event/trigger
1745
Note that variable assignments can appear either preceding or
1746
following their use. The command below behaves identically to the
1747
command above::
1749
# echo 'hist:keys=pid:ts0=common_timestamp,b=field1:vals=$ts0,$b ...' >> \
1750
event/trigger
1752
Any number of variables not bound to a 'vals=' prefix can also be
1753
assigned by simply separating them with colons. Below is the same
1754
thing but without the values being summed in the histogram::
1756
# echo 'hist:keys=pid:ts0=common_timestamp:b=field1 ...' >> event/trigger
1758
Variables set as above can be referenced and used in expressions on
1759
another event.
1761
For example, here's how a latency can be calculated::
1763
# echo 'hist:keys=pid,prio:ts0=common_timestamp ...' >> event1/trigger
1764
# echo 'hist:keys=next_pid:wakeup_lat=common_timestamp-$ts0 ...' >> event2/trigger
1766
In the first line above, the event's timestamp is saved into the
1767
variable ts0. In the next line, ts0 is subtracted from the second
1768
event's timestamp to produce the latency, which is then assigned into
1769
yet another variable, 'wakeup_lat'. The hist trigger below in turn
1770
makes use of the wakeup_lat variable to compute a combined latency
1771
using the same key and variable from yet another event::
1773
# echo 'hist:key=pid:wakeupswitch_lat=$wakeup_lat+$switchtime_lat ...' >> event3/trigger
1775
Expressions support the use of addition, subtraction, multiplication and
1776
division operators (+-\*/).
1778
Note if division by zero cannot be detected at parse time (i.e. the
1779
divisor is not a constant), the result will be -1.
1781
Numeric constants can also be used directly in an expression::
1783
# echo 'hist:keys=next_pid:timestamp_secs=common_timestamp/1000000 ...' >> event/trigger
1785
or assigned to a variable and referenced in a subsequent expression::
1787
# echo 'hist:keys=next_pid:us_per_sec=1000000 ...' >> event/trigger
1788
# echo 'hist:keys=next_pid:timestamp_secs=common_timestamp/$us_per_sec ...' >> event/trigger
1790
Variables can even hold stacktraces, which are useful with synthetic events.
1792
2.6. Synthetic Events
1793
---------------------
1795
Synthetic events are user-defined events generated from hist trigger
1796
variables or fields associated with one or more other events. Their
1797
purpose is to provide a mechanism for displaying data spanning
1798
multiple events consistent with the existing and already familiar
1799
usage for normal events.
1801
To define a synthetic event, the user writes a simple specification
1802
consisting of the name of the new event along with one or more
1803
variables and their types, which can be any valid field type,
1804
separated by semicolons, to the tracing/synthetic_events file.
1806
See synth_field_size() for available types.
1808
If field_name contains [n], the field is considered to be a static array.
1810
If field_names contains[] (no subscript), the field is considered to
1811
be a dynamic array, which will only take as much space in the event as
1812
is required to hold the array.
1814
A string field can be specified using either the static notation:
1816
char name[32];
1818
Or the dynamic:
1820
char name[];
1822
The size limit for either is 256.
1824
For instance, the following creates a new event named 'wakeup_latency'
1825
with 3 fields: lat, pid, and prio. Each of those fields is simply a
1826
variable reference to a variable on another event::
1828
# echo 'wakeup_latency \
1829
u64 lat; \
1830
pid_t pid; \
1831
int prio' >> \
1832
/sys/kernel/tracing/synthetic_events
1834
Reading the tracing/synthetic_events file lists all the currently
1835
defined synthetic events, in this case the event defined above::
1837
# cat /sys/kernel/tracing/synthetic_events
1838
wakeup_latency u64 lat; pid_t pid; int prio
1840
An existing synthetic event definition can be removed by prepending
1841
the command that defined it with a '!'::
1843
# echo '!wakeup_latency u64 lat pid_t pid int prio' >> \
1844
/sys/kernel/tracing/synthetic_events
1846
At this point, there isn't yet an actual 'wakeup_latency' event
1847
instantiated in the event subsystem - for this to happen, a 'hist
1848
trigger action' needs to be instantiated and bound to actual fields
1849
and variables defined on other events (see Section 2.7. below on
1850
how that is done using hist trigger 'onmatch' action). Once that is
1851
done, the 'wakeup_latency' synthetic event instance is created.
1853
The new event is created under the tracing/events/synthetic/ directory
1854
and looks and behaves just like any other event::
1856
# ls /sys/kernel/tracing/events/synthetic/wakeup_latency
1857
enable filter format hist id trigger
1859
A histogram can now be defined for the new synthetic event::
1861
# echo 'hist:keys=pid,prio,lat.log2:sort=lat' >> \
1862
/sys/kernel/tracing/events/synthetic/wakeup_latency/trigger
1864
The above shows the latency "lat" in a power of 2 grouping.
1866
Like any other event, once a histogram is enabled for the event, the
1867
output can be displayed by reading the event's 'hist' file::
1869
# cat /sys/kernel/tracing/events/synthetic/wakeup_latency/hist
1871
# event histogram
1872
#
1873
# trigger info: hist:keys=pid,prio,lat.log2:vals=hitcount:sort=lat.log2:size=2048 [active]
1874
#
1876
{ pid: 2035, prio: 9, lat: ~ 2^2 } hitcount: 43
1877
{ pid: 2034, prio: 9, lat: ~ 2^2 } hitcount: 60
1878
{ pid: 2029, prio: 9, lat: ~ 2^2 } hitcount: 965
1879
{ pid: 2034, prio: 120, lat: ~ 2^2 } hitcount: 9
1880
{ pid: 2033, prio: 120, lat: ~ 2^2 } hitcount: 5
1881
{ pid: 2030, prio: 9, lat: ~ 2^2 } hitcount: 335
1882
{ pid: 2030, prio: 120, lat: ~ 2^2 } hitcount: 10
1883
{ pid: 2032, prio: 120, lat: ~ 2^2 } hitcount: 1
1884
{ pid: 2035, prio: 120, lat: ~ 2^2 } hitcount: 2
1885
{ pid: 2031, prio: 9, lat: ~ 2^2 } hitcount: 176
1886
{ pid: 2028, prio: 120, lat: ~ 2^2 } hitcount: 15
1887
{ pid: 2033, prio: 9, lat: ~ 2^2 } hitcount: 91
1888
{ pid: 2032, prio: 9, lat: ~ 2^2 } hitcount: 125
1889
{ pid: 2029, prio: 120, lat: ~ 2^2 } hitcount: 4
1890
{ pid: 2031, prio: 120, lat: ~ 2^2 } hitcount: 3
1891
{ pid: 2029, prio: 120, lat: ~ 2^3 } hitcount: 2
1892
{ pid: 2035, prio: 9, lat: ~ 2^3 } hitcount: 41
1893
{ pid: 2030, prio: 120, lat: ~ 2^3 } hitcount: 1
1894
{ pid: 2032, prio: 9, lat: ~ 2^3 } hitcount: 32
1895
{ pid: 2031, prio: 9, lat: ~ 2^3 } hitcount: 44
1896
{ pid: 2034, prio: 9, lat: ~ 2^3 } hitcount: 40
1897
{ pid: 2030, prio: 9, lat: ~ 2^3 } hitcount: 29
1898
{ pid: 2033, prio: 9, lat: ~ 2^3 } hitcount: 31
1899
{ pid: 2029, prio: 9, lat: ~ 2^3 } hitcount: 31
1900
{ pid: 2028, prio: 120, lat: ~ 2^3 } hitcount: 18
1901
{ pid: 2031, prio: 120, lat: ~ 2^3 } hitcount: 2
1902
{ pid: 2028, prio: 120, lat: ~ 2^4 } hitcount: 1
1903
{ pid: 2029, prio: 9, lat: ~ 2^4 } hitcount: 4
1904
{ pid: 2031, prio: 120, lat: ~ 2^7 } hitcount: 1
1905
{ pid: 2032, prio: 120, lat: ~ 2^7 } hitcount: 1
1907
Totals:
1908
Hits: 2122
1909
Entries: 30
1910
Dropped: 0
1913
The latency values can also be grouped linearly by a given size with
1914
the ".buckets" modifier and specify a size (in this case groups of 10)::
1916
# echo 'hist:keys=pid,prio,lat.buckets=10:sort=lat' >> \
1917
/sys/kernel/tracing/events/synthetic/wakeup_latency/trigger
1919
# event histogram
1920
#
1921
# trigger info: hist:keys=pid,prio,lat.buckets=10:vals=hitcount:sort=lat.buckets=10:size=2048 [active]
1922
#
1924
{ pid: 2067, prio: 9, lat: ~ 0-9 } hitcount: 220
1925
{ pid: 2068, prio: 9, lat: ~ 0-9 } hitcount: 157
1926
{ pid: 2070, prio: 9, lat: ~ 0-9 } hitcount: 100
1927
{ pid: 2067, prio: 120, lat: ~ 0-9 } hitcount: 6
1928
{ pid: 2065, prio: 120, lat: ~ 0-9 } hitcount: 2
1929
{ pid: 2066, prio: 120, lat: ~ 0-9 } hitcount: 2
1930
{ pid: 2069, prio: 9, lat: ~ 0-9 } hitcount: 122
1931
{ pid: 2069, prio: 120, lat: ~ 0-9 } hitcount: 8
1932
{ pid: 2070, prio: 120, lat: ~ 0-9 } hitcount: 1
1933
{ pid: 2068, prio: 120, lat: ~ 0-9 } hitcount: 7
1934
{ pid: 2066, prio: 9, lat: ~ 0-9 } hitcount: 365
1935
{ pid: 2064, prio: 120, lat: ~ 0-9 } hitcount: 35
1936
{ pid: 2065, prio: 9, lat: ~ 0-9 } hitcount: 998
1937
{ pid: 2071, prio: 9, lat: ~ 0-9 } hitcount: 85
1938
{ pid: 2065, prio: 9, lat: ~ 10-19 } hitcount: 2
1939
{ pid: 2064, prio: 120, lat: ~ 10-19 } hitcount: 2
1941
Totals:
1942
Hits: 2112
1943
Entries: 16
1944
Dropped: 0
1946
To save stacktraces, create a synthetic event with a field of type "unsigned long[]"
1947
or even just "long[]". For example, to see how long a task is blocked in an
1948
uninterruptible state::
1950
# cd /sys/kernel/tracing
1951
# echo 's:block_lat pid_t pid; u64 delta; unsigned long[] stack;' > dynamic_events
1952
# echo 'hist:keys=next_pid:ts=common_timestamp.usecs,st=common_stacktrace if prev_state == 2' >> events/sched/sched_switch/trigger
1953
# echo 'hist:keys=prev_pid:delta=common_timestamp.usecs-$ts,s=$st:onmax($delta).trace(block_lat,prev_pid,$delta,$s)' >> events/sched/sched_switch/trigger
1954
# echo 1 > events/synthetic/block_lat/enable
1955
# cat trace
1957
# tracer: nop
1958
#
1959
# entries-in-buffer/entries-written: 2/2 #P:8
1960
#
1961
# _-----=> irqs-off/BH-disabled
1962
# / _----=> need-resched
1963
# | / _---=> hardirq/softirq
1964
# || / _--=> preempt-depth
1965
# ||| / _-=> migrate-disable
1966
# |||| / delay
1967
# TASK-PID CPU# ||||| TIMESTAMP FUNCTION
1968
# | | | ||||| | |
1969
<idle>-0 [005] d..4. 521.164922: block_lat: pid=0 delta=8322 stack=STACK:
1970
=> __schedule+0x448/0x7b0
1971
=> schedule+0x5a/0xb0
1972
=> io_schedule+0x42/0x70
1973
=> bit_wait_io+0xd/0x60
1974
=> __wait_on_bit+0x4b/0x140
1975
=> out_of_line_wait_on_bit+0x91/0xb0
1976
=> jbd2_journal_commit_transaction+0x1679/0x1a70
1977
=> kjournald2+0xa9/0x280
1978
=> kthread+0xe9/0x110
1979
=> ret_from_fork+0x2c/0x50
1981
<...>-2 [004] d..4. 525.184257: block_lat: pid=2 delta=76 stack=STACK:
1982
=> __schedule+0x448/0x7b0
1983
=> schedule+0x5a/0xb0
1984
=> schedule_timeout+0x11a/0x150
1985
=> wait_for_completion_killable+0x144/0x1f0
1986
=> __kthread_create_on_node+0xe7/0x1e0
1987
=> kthread_create_on_node+0x51/0x70
1988
=> create_worker+0xcc/0x1a0
1989
=> worker_thread+0x2ad/0x380
1990
=> kthread+0xe9/0x110
1991
=> ret_from_fork+0x2c/0x50
1993
A synthetic event that has a stacktrace field may use it as a key in
1994
histogram::
1996
# echo 'hist:keys=delta.buckets=100,stack.stacktrace:sort=delta' > events/synthetic/block_lat/trigger
1997
# cat events/synthetic/block_lat/hist
1999
# event histogram
2000
#
2001
# trigger info: hist:keys=delta.buckets=100,stack.stacktrace:vals=hitcount:sort=delta.buckets=100:size=2048 [active]
2002
#
2003
{ delta: ~ 0-99, stack.stacktrace __schedule+0xa19/0x1520
2004
schedule+0x6b/0x110
2005
io_schedule+0x46/0x80
2006
bit_wait_io+0x11/0x80
2007
__wait_on_bit+0x4e/0x120
2008
out_of_line_wait_on_bit+0x8d/0xb0
2009
__wait_on_buffer+0x33/0x40
2010
jbd2_journal_commit_transaction+0x155a/0x19b0
2011
kjournald2+0xab/0x270
2012
kthread+0xfa/0x130
2013
ret_from_fork+0x29/0x50
2014
} hitcount: 1
2015
{ delta: ~ 0-99, stack.stacktrace __schedule+0xa19/0x1520
2016
schedule+0x6b/0x110
2017
io_schedule+0x46/0x80
2018
rq_qos_wait+0xd0/0x170
2019
wbt_wait+0x9e/0xf0
2020
__rq_qos_throttle+0x25/0x40
2021
blk_mq_submit_bio+0x2c3/0x5b0
2022
__submit_bio+0xff/0x190
2023
submit_bio_noacct_nocheck+0x25b/0x2b0
2024
submit_bio_noacct+0x20b/0x600
2025
submit_bio+0x28/0x90
2026
ext4_bio_write_page+0x1e0/0x8c0
2027
mpage_submit_page+0x60/0x80
2028
mpage_process_page_bufs+0x16c/0x180
2029
mpage_prepare_extent_to_map+0x23f/0x530
2030
} hitcount: 1
2031
{ delta: ~ 0-99, stack.stacktrace __schedule+0xa19/0x1520
2032
schedule+0x6b/0x110
2033
schedule_hrtimeout_range_clock+0x97/0x110
2034
schedule_hrtimeout_range+0x13/0x20
2035
usleep_range_state+0x65/0x90
2036
__intel_wait_for_register+0x1c1/0x230 [i915]
2037
intel_psr_wait_for_idle_locked+0x171/0x2a0 [i915]
2038
intel_pipe_update_start+0x169/0x360 [i915]
2039
intel_update_crtc+0x112/0x490 [i915]
2040
skl_commit_modeset_enables+0x199/0x600 [i915]
2041
intel_atomic_commit_tail+0x7c4/0x1080 [i915]
2042
intel_atomic_commit_work+0x12/0x20 [i915]
2043
process_one_work+0x21c/0x3f0
2044
worker_thread+0x50/0x3e0
2045
kthread+0xfa/0x130
2046
} hitcount: 3
2047
{ delta: ~ 0-99, stack.stacktrace __schedule+0xa19/0x1520
2048
schedule+0x6b/0x110
2049
schedule_timeout+0x11e/0x160
2050
__wait_for_common+0x8f/0x190
2051
wait_for_completion+0x24/0x30
2052
__flush_work.isra.0+0x1cc/0x360
2053
flush_work+0xe/0x20
2054
drm_mode_rmfb+0x18b/0x1d0 [drm]
2055
drm_mode_rmfb_ioctl+0x10/0x20 [drm]
2056
drm_ioctl_kernel+0xb8/0x150 [drm]
2057
drm_ioctl+0x243/0x560 [drm]
2058
__x64_sys_ioctl+0x92/0xd0
2059
do_syscall_64+0x59/0x90
2060
entry_SYSCALL_64_after_hwframe+0x72/0xdc
2061
} hitcount: 1
2062
{ delta: ~ 0-99, stack.stacktrace __schedule+0xa19/0x1520
2063
schedule+0x6b/0x110
2064
schedule_timeout+0x87/0x160
2065
__wait_for_common+0x8f/0x190
2066
wait_for_completion_timeout+0x1d/0x30
2067
drm_atomic_helper_wait_for_flip_done+0x57/0x90 [drm_kms_helper]
2068
intel_atomic_commit_tail+0x8ce/0x1080 [i915]
2069
intel_atomic_commit_work+0x12/0x20 [i915]
2070
process_one_work+0x21c/0x3f0
2071
worker_thread+0x50/0x3e0
2072
kthread+0xfa/0x130
2073
ret_from_fork+0x29/0x50
2074
} hitcount: 1
2075
{ delta: ~ 100-199, stack.stacktrace __schedule+0xa19/0x1520
2076
schedule+0x6b/0x110
2077
schedule_hrtimeout_range_clock+0x97/0x110
2078
schedule_hrtimeout_range+0x13/0x20
2079
usleep_range_state+0x65/0x90
2080
pci_set_low_power_state+0x17f/0x1f0
2081
pci_set_power_state+0x49/0x250
2082
pci_finish_runtime_suspend+0x4a/0x90
2083
pci_pm_runtime_suspend+0xcb/0x1b0
2084
__rpm_callback+0x48/0x120
2085
rpm_callback+0x67/0x70
2086
rpm_suspend+0x167/0x780
2087
rpm_idle+0x25a/0x380
2088
pm_runtime_work+0x93/0xc0
2089
process_one_work+0x21c/0x3f0
2090
} hitcount: 1
2092
Totals:
2093
Hits: 10
2094
Entries: 7
2095
Dropped: 0
2097
2.7. Hist trigger 'handlers' and 'actions'
2098
------------------------------------------
2100
A hist trigger 'action' is a function that's executed (in most cases
2101
conditionally) whenever a histogram entry is added or updated.
2103
When a histogram entry is added or updated, a hist trigger 'handler'
2104
is what decides whether the corresponding action is actually invoked
2105
or not.
2107
Hist trigger handlers and actions are paired together in the general
2108
form:
2110
<handler>.<action>
2112
To specify a handler.action pair for a given event, simply specify
2113
that handler.action pair between colons in the hist trigger
2114
specification.
2116
In theory, any handler can be combined with any action, but in
2117
practice, not every handler.action combination is currently supported;
2118
if a given handler.action combination isn't supported, the hist
2119
trigger will fail with -EINVAL;
2121
The default 'handler.action' if none is explicitly specified is as it
2122
always has been, to simply update the set of values associated with an
2123
entry. Some applications, however, may want to perform additional
2124
actions at that point, such as generate another event, or compare and
2125
save a maximum.
2127
The supported handlers and actions are listed below, and each is
2128
described in more detail in the following paragraphs, in the context
2129
of descriptions of some common and useful handler.action combinations.
2131
The available handlers are:
2133
- onmatch(matching.event) - invoke action on any addition or update
2134
- onmax(var) - invoke action if var exceeds current max
2135
- onchange(var) - invoke action if var changes
2137
The available actions are:
2139
- trace(<synthetic_event_name>,param list) - generate synthetic event
2140
- save(field,...) - save current event fields
2141
- snapshot() - snapshot the trace buffer
2143
The following commonly-used handler.action pairs are available:
2145
- onmatch(matching.event).trace(<synthetic_event_name>,param list)
2147
The 'onmatch(matching.event).trace(<synthetic_event_name>,param
2148
list)' hist trigger action is invoked whenever an event matches
2149
and the histogram entry would be added or updated. It causes the
2150
named synthetic event to be generated with the values given in the
2151
'param list'. The result is the generation of a synthetic event
2152
that consists of the values contained in those variables at the
2153
time the invoking event was hit. For example, if the synthetic
2154
event name is 'wakeup_latency', a wakeup_latency event is
2155
generated using onmatch(event).trace(wakeup_latency,arg1,arg2).
2157
There is also an equivalent alternative form available for
2158
generating synthetic events. In this form, the synthetic event
2159
name is used as if it were a function name. For example, using
2160
the 'wakeup_latency' synthetic event name again, the
2161
wakeup_latency event would be generated by invoking it as if it
2162
were a function call, with the event field values passed in as
2163
arguments: onmatch(event).wakeup_latency(arg1,arg2). The syntax
2164
for this form is:
2166
onmatch(matching.event).<synthetic_event_name>(param list)
2168
In either case, the 'param list' consists of one or more
2169
parameters which may be either variables or fields defined on
2170
either the 'matching.event' or the target event. The variables or
2171
fields specified in the param list may be either fully-qualified
2172
or unqualified. If a variable is specified as unqualified, it
2173
must be unique between the two events. A field name used as a
2174
param can be unqualified if it refers to the target event, but
2175
must be fully qualified if it refers to the matching event. A
2176
fully-qualified name is of the form 'system.event_name.$var_name'
2177
or 'system.event_name.field'.
2179
The 'matching.event' specification is simply the fully qualified
2180
event name of the event that matches the target event for the
2181
onmatch() functionality, in the form 'system.event_name'. Histogram
2182
keys of both events are compared to find if events match. In case
2183
multiple histogram keys are used, they all must match in the specified
2184
order.
2186
Finally, the number and type of variables/fields in the 'param
2187
list' must match the number and types of the fields in the
2188
synthetic event being generated.
2190
As an example the below defines a simple synthetic event and uses
2191
a variable defined on the sched_wakeup_new event as a parameter
2192
when invoking the synthetic event. Here we define the synthetic
2193
event::
2195
# echo 'wakeup_new_test pid_t pid' >> \
2196
/sys/kernel/tracing/synthetic_events
2198
# cat /sys/kernel/tracing/synthetic_events
2199
wakeup_new_test pid_t pid
2201
The following hist trigger both defines the missing testpid
2202
variable and specifies an onmatch() action that generates a
2203
wakeup_new_test synthetic event whenever a sched_wakeup_new event
2204
occurs, which because of the 'if comm == "cyclictest"' filter only
2205
happens when the executable is cyclictest::
2207
# echo 'hist:keys=$testpid:testpid=pid:onmatch(sched.sched_wakeup_new).\
2208
wakeup_new_test($testpid) if comm=="cyclictest"' >> \
2209
/sys/kernel/tracing/events/sched/sched_wakeup_new/trigger
2211
Or, equivalently, using the 'trace' keyword syntax::
2213
# echo 'hist:keys=$testpid:testpid=pid:onmatch(sched.sched_wakeup_new).\
2214
trace(wakeup_new_test,$testpid) if comm=="cyclictest"' >> \
2215
/sys/kernel/tracing/events/sched/sched_wakeup_new/trigger
2217
Creating and displaying a histogram based on those events is now
2218
just a matter of using the fields and new synthetic event in the
2219
tracing/events/synthetic directory, as usual::
2221
# echo 'hist:keys=pid:sort=pid' >> \
2222
/sys/kernel/tracing/events/synthetic/wakeup_new_test/trigger
2224
Running 'cyclictest' should cause wakeup_new events to generate
2225
wakeup_new_test synthetic events which should result in histogram
2226
output in the wakeup_new_test event's hist file::
2228
# cat /sys/kernel/tracing/events/synthetic/wakeup_new_test/hist
2230
A more typical usage would be to use two events to calculate a
2231
latency. The following example uses a set of hist triggers to
2232
produce a 'wakeup_latency' histogram.
2234
First, we define a 'wakeup_latency' synthetic event::
2236
# echo 'wakeup_latency u64 lat; pid_t pid; int prio' >> \
2237
/sys/kernel/tracing/synthetic_events
2239
Next, we specify that whenever we see a sched_waking event for a
2240
cyclictest thread, save the timestamp in a 'ts0' variable::
2242
# echo 'hist:keys=$saved_pid:saved_pid=pid:ts0=common_timestamp.usecs \
2243
if comm=="cyclictest"' >> \
2244
/sys/kernel/tracing/events/sched/sched_waking/trigger
2246
Then, when the corresponding thread is actually scheduled onto the
2247
CPU by a sched_switch event (saved_pid matches next_pid), calculate
2248
the latency and use that along with another variable and an event field
2249
to generate a wakeup_latency synthetic event::
2251
# echo 'hist:keys=next_pid:wakeup_lat=common_timestamp.usecs-$ts0:\
2252
onmatch(sched.sched_waking).wakeup_latency($wakeup_lat,\
2253
$saved_pid,next_prio) if next_comm=="cyclictest"' >> \
2254
/sys/kernel/tracing/events/sched/sched_switch/trigger
2256
We also need to create a histogram on the wakeup_latency synthetic
2257
event in order to aggregate the generated synthetic event data::
2259
# echo 'hist:keys=pid,prio,lat:sort=pid,lat' >> \
2260
/sys/kernel/tracing/events/synthetic/wakeup_latency/trigger
2262
Finally, once we've run cyclictest to actually generate some
2263
events, we can see the output by looking at the wakeup_latency
2264
synthetic event's hist file::
2266
# cat /sys/kernel/tracing/events/synthetic/wakeup_latency/hist
2268
- onmax(var).save(field,.. .)
2270
The 'onmax(var).save(field,...)' hist trigger action is invoked
2271
whenever the value of 'var' associated with a histogram entry
2272
exceeds the current maximum contained in that variable.
2274
The end result is that the trace event fields specified as the
2275
onmax.save() params will be saved if 'var' exceeds the current
2276
maximum for that hist trigger entry. This allows context from the
2277
event that exhibited the new maximum to be saved for later
2278
reference. When the histogram is displayed, additional fields
2279
displaying the saved values will be printed.
2281
As an example the below defines a couple of hist triggers, one for
2282
sched_waking and another for sched_switch, keyed on pid. Whenever
2283
a sched_waking occurs, the timestamp is saved in the entry
2284
corresponding to the current pid, and when the scheduler switches
2285
back to that pid, the timestamp difference is calculated. If the
2286
resulting latency, stored in wakeup_lat, exceeds the current
2287
maximum latency, the values specified in the save() fields are
2288
recorded::
2290
# echo 'hist:keys=pid:ts0=common_timestamp.usecs \
2291
if comm=="cyclictest"' >> \
2292
/sys/kernel/tracing/events/sched/sched_waking/trigger
2294
# echo 'hist:keys=next_pid:\
2295
wakeup_lat=common_timestamp.usecs-$ts0:\
2296
onmax($wakeup_lat).save(next_comm,prev_pid,prev_prio,prev_comm) \
2297
if next_comm=="cyclictest"' >> \
2298
/sys/kernel/tracing/events/sched/sched_switch/trigger
2300
When the histogram is displayed, the max value and the saved
2301
values corresponding to the max are displayed following the rest
2302
of the fields::
2304
# cat /sys/kernel/tracing/events/sched/sched_switch/hist
2305
{ next_pid: 2255 } hitcount: 239
2306
common_timestamp-ts0: 0
2307
max: 27
2308
next_comm: cyclictest
2309
prev_pid: 0 prev_prio: 120 prev_comm: swapper/1
2311
{ next_pid: 2256 } hitcount: 2355
2312
common_timestamp-ts0: 0
2313
max: 49 next_comm: cyclictest
2314
prev_pid: 0 prev_prio: 120 prev_comm: swapper/0
2316
Totals:
2317
Hits: 12970
2318
Entries: 2
2319
Dropped: 0
2321
- onmax(var).snapshot()
2323
The 'onmax(var).snapshot()' hist trigger action is invoked
2324
whenever the value of 'var' associated with a histogram entry
2325
exceeds the current maximum contained in that variable.
2327
The end result is that a global snapshot of the trace buffer will
2328
be saved in the tracing/snapshot file if 'var' exceeds the current
2329
maximum for any hist trigger entry.
2331
Note that in this case the maximum is a global maximum for the
2332
current trace instance, which is the maximum across all buckets of
2333
the histogram. The key of the specific trace event that caused
2334
the global maximum and the global maximum itself are displayed,
2335
along with a message stating that a snapshot has been taken and
2336
where to find it. The user can use the key information displayed
2337
to locate the corresponding bucket in the histogram for even more
2338
detail.
2340
As an example the below defines a couple of hist triggers, one for
2341
sched_waking and another for sched_switch, keyed on pid. Whenever
2342
a sched_waking event occurs, the timestamp is saved in the entry
2343
corresponding to the current pid, and when the scheduler switches
2344
back to that pid, the timestamp difference is calculated. If the
2345
resulting latency, stored in wakeup_lat, exceeds the current
2346
maximum latency, a snapshot is taken. As part of the setup, all
2347
the scheduler events are also enabled, which are the events that
2348
will show up in the snapshot when it is taken at some point::
2350
# echo 1 > /sys/kernel/tracing/events/sched/enable
2352
# echo 'hist:keys=pid:ts0=common_timestamp.usecs \
2353
if comm=="cyclictest"' >> \
2354
/sys/kernel/tracing/events/sched/sched_waking/trigger
2356
# echo 'hist:keys=next_pid:wakeup_lat=common_timestamp.usecs-$ts0: \
2357
onmax($wakeup_lat).save(next_prio,next_comm,prev_pid,prev_prio, \
2358
prev_comm):onmax($wakeup_lat).snapshot() \
2359
if next_comm=="cyclictest"' >> \
2360
/sys/kernel/tracing/events/sched/sched_switch/trigger
2362
When the histogram is displayed, for each bucket the max value
2363
and the saved values corresponding to the max are displayed
2364
following the rest of the fields.
2366
If a snapshot was taken, there is also a message indicating that,
2367
along with the value and event that triggered the global maximum::
2369
# cat /sys/kernel/tracing/events/sched/sched_switch/hist
2370
{ next_pid: 2101 } hitcount: 200
2371
max: 52 next_prio: 120 next_comm: cyclictest \
2372
prev_pid: 0 prev_prio: 120 prev_comm: swapper/6
2374
{ next_pid: 2103 } hitcount: 1326
2375
max: 572 next_prio: 19 next_comm: cyclictest \
2376
prev_pid: 0 prev_prio: 120 prev_comm: swapper/1
2378
{ next_pid: 2102 } hitcount: 1982 \
2379
max: 74 next_prio: 19 next_comm: cyclictest \
2380
prev_pid: 0 prev_prio: 120 prev_comm: swapper/5
2382
Snapshot taken (see tracing/snapshot). Details:
2383
triggering value { onmax($wakeup_lat) }: 572 \
2384
triggered by event with key: { next_pid: 2103 }
2386
Totals:
2387
Hits: 3508
2388
Entries: 3
2389
Dropped: 0
2391
In the above case, the event that triggered the global maximum has
2392
the key with next_pid == 2103. If you look at the bucket that has
2393
2103 as the key, you'll find the additional values save()'d along
2394
with the local maximum for that bucket, which should be the same
2395
as the global maximum (since that was the same value that
2396
triggered the global snapshot).
2398
And finally, looking at the snapshot data should show at or near
2399
the end the event that triggered the snapshot (in this case you
2400
can verify the timestamps between the sched_waking and
2401
sched_switch events, which should match the time displayed in the
2402
global maximum)::
2404
# cat /sys/kernel/tracing/snapshot
2406
<...>-2103 [005] d..3 309.873125: sched_switch: prev_comm=cyclictest prev_pid=2103 prev_prio=19 prev_state=D ==> next_comm=swapper/5 next_pid=0 next_prio=120
2407
<idle>-0 [005] d.h3 309.873611: sched_waking: comm=cyclictest pid=2102 prio=19 target_cpu=005
2408
<idle>-0 [005] dNh4 309.873613: sched_wakeup: comm=cyclictest pid=2102 prio=19 target_cpu=005
2409
<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
2410
<...>-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
2411
<idle>-0 [005] d.h3 309.874624: sched_waking: comm=cyclictest pid=2102 prio=19 target_cpu=005
2412
<idle>-0 [005] dNh4 309.874626: sched_wakeup: comm=cyclictest pid=2102 prio=19 target_cpu=005
2413
<idle>-0 [005] dNh3 309.874628: sched_waking: comm=cyclictest pid=2103 prio=19 target_cpu=005
2414
<idle>-0 [005] dNh4 309.874630: sched_wakeup: comm=cyclictest pid=2103 prio=19 target_cpu=005
2415
<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
2416
<idle>-0 [004] d.h3 309.874757: sched_waking: comm=gnome-terminal- pid=1699 prio=120 target_cpu=004
2417
<idle>-0 [004] dNh4 309.874762: sched_wakeup: comm=gnome-terminal- pid=1699 prio=120 target_cpu=004
2418
<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
2419
gnome-terminal--1699 [004] d.h2 309.874941: sched_stat_runtime: comm=gnome-terminal- pid=1699 runtime=180706 [ns] vruntime=1126870572 [ns]
2420
<idle>-0 [003] d.s4 309.874956: sched_waking: comm=rcu_sched pid=9 prio=120 target_cpu=007
2421
<idle>-0 [003] d.s5 309.874960: sched_wake_idle_without_ipi: cpu=7
2422
<idle>-0 [003] d.s5 309.874961: sched_wakeup: comm=rcu_sched pid=9 prio=120 target_cpu=007
2423
<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
2424
rcu_sched-9 [007] d..3 309.874973: sched_stat_runtime: comm=rcu_sched pid=9 runtime=13646 [ns] vruntime=22531430286 [ns]
2425
rcu_sched-9 [007] d..3 309.874978: sched_switch: prev_comm=rcu_sched prev_pid=9 prev_prio=120 prev_state=R+ ==> next_comm=swapper/7 next_pid=0 next_prio=120
2426
<...>-2102 [005] d..4 309.874994: sched_migrate_task: comm=cyclictest pid=2103 prio=19 orig_cpu=5 dest_cpu=1
2427
<...>-2102 [005] d..4 309.875185: sched_wake_idle_without_ipi: cpu=1
2428
<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
2430
- onchange(var).save(field,.. .)
2432
The 'onchange(var).save(field,...)' hist trigger action is invoked
2433
whenever the value of 'var' associated with a histogram entry
2434
changes.
2436
The end result is that the trace event fields specified as the
2437
onchange.save() params will be saved if 'var' changes for that
2438
hist trigger entry. This allows context from the event that
2439
changed the value to be saved for later reference. When the
2440
histogram is displayed, additional fields displaying the saved
2441
values will be printed.
2443
- onchange(var).snapshot()
2445
The 'onchange(var).snapshot()' hist trigger action is invoked
2446
whenever the value of 'var' associated with a histogram entry
2447
changes.
2449
The end result is that a global snapshot of the trace buffer will
2450
be saved in the tracing/snapshot file if 'var' changes for any
2451
hist trigger entry.
2453
Note that in this case the changed value is a global variable
2454
associated with current trace instance. The key of the specific
2455
trace event that caused the value to change and the global value
2456
itself are displayed, along with a message stating that a snapshot
2457
has been taken and where to find it. The user can use the key
2458
information displayed to locate the corresponding bucket in the
2459
histogram for even more detail.
2461
As an example the below defines a hist trigger on the tcp_probe
2462
event, keyed on dport. Whenever a tcp_probe event occurs, the
2463
cwnd field is checked against the current value stored in the
2464
$cwnd variable. If the value has changed, a snapshot is taken.
2465
As part of the setup, all the scheduler and tcp events are also
2466
enabled, which are the events that will show up in the snapshot
2467
when it is taken at some point::
2469
# echo 1 > /sys/kernel/tracing/events/sched/enable
2470
# echo 1 > /sys/kernel/tracing/events/tcp/enable
2472
# echo 'hist:keys=dport:cwnd=snd_cwnd: \
2473
onchange($cwnd).save(snd_wnd,srtt,rcv_wnd): \
2474
onchange($cwnd).snapshot()' >> \
2475
/sys/kernel/tracing/events/tcp/tcp_probe/trigger
2477
When the histogram is displayed, for each bucket the tracked value
2478
and the saved values corresponding to that value are displayed
2479
following the rest of the fields.
2481
If a snapshot was taken, there is also a message indicating that,
2482
along with the value and event that triggered the snapshot::
2484
# cat /sys/kernel/tracing/events/tcp/tcp_probe/hist
2486
{ dport: 1521 } hitcount: 8
2487
changed: 10 snd_wnd: 35456 srtt: 154262 rcv_wnd: 42112
2489
{ dport: 80 } hitcount: 23
2490
changed: 10 snd_wnd: 28960 srtt: 19604 rcv_wnd: 29312
2492
{ dport: 9001 } hitcount: 172
2493
changed: 10 snd_wnd: 48384 srtt: 260444 rcv_wnd: 55168
2495
{ dport: 443 } hitcount: 211
2496
changed: 10 snd_wnd: 26960 srtt: 17379 rcv_wnd: 28800
2498
Snapshot taken (see tracing/snapshot). Details:
2500
triggering value { onchange($cwnd) }: 10
2501
triggered by event with key: { dport: 80 }
2503
Totals:
2504
Hits: 414
2505
Entries: 4
2506
Dropped: 0
2508
In the above case, the event that triggered the snapshot has the
2509
key with dport == 80. If you look at the bucket that has 80 as
2510
the key, you'll find the additional values save()'d along with the
2511
changed value for that bucket, which should be the same as the
2512
global changed value (since that was the same value that triggered
2513
the global snapshot).
2515
And finally, looking at the snapshot data should show at or near
2516
the end the event that triggered the snapshot::
2518
# cat /sys/kernel/tracing/snapshot
2520
gnome-shell-1261 [006] dN.3 49.823113: sched_stat_runtime: comm=gnome-shell pid=1261 runtime=49347 [ns] vruntime=1835730389 [ns]
2521
kworker/u16:4-773 [003] d..3 49.823114: sched_switch: prev_comm=kworker/u16:4 prev_pid=773 prev_prio=120 prev_state=R+ ==> next_comm=kworker/3:2 next_pid=135 next_prio=120
2522
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
2523
kworker/3:2-135 [003] d..3 49.823118: sched_stat_runtime: comm=kworker/3:2 pid=135 runtime=5339 [ns] vruntime=17815800388 [ns]
2524
kworker/6:2-387 [006] d..3 49.823120: sched_stat_runtime: comm=kworker/6:2 pid=387 runtime=9594 [ns] vruntime=14589605367 [ns]
2525
kworker/6:2-387 [006] d..3 49.823122: sched_switch: prev_comm=kworker/6:2 prev_pid=387 prev_prio=120 prev_state=R+ ==> next_comm=gnome-shell next_pid=1261 next_prio=120
2526
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
2527
<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
2529
2.8. User space creating a trigger
2530
----------------------------------
2532
Writing into /sys/kernel/tracing/trace_marker writes into the ftrace
2533
ring buffer. This can also act like an event, by writing into the trigger
2534
file located in /sys/kernel/tracing/events/ftrace/print/
2536
Modifying cyclictest to write into the trace_marker file before it sleeps
2537
and after it wakes up, something like this::
2539
static void traceputs(char *str)
2540
{
2541
/* tracemark_fd is the trace_marker file descriptor */
2542
if (tracemark_fd < 0)
2543
return;
2544
/* write the tracemark message */
2545
write(tracemark_fd, str, strlen(str));
2546
}
2548
And later add something like::
2550
traceputs("start");
2551
clock_nanosleep(...);
2552
traceputs("end");
2554
We can make a histogram from this::
2556
# cd /sys/kernel/tracing
2557
# echo 'latency u64 lat' > synthetic_events
2558
# echo 'hist:keys=common_pid:ts0=common_timestamp.usecs if buf == "start"' > events/ftrace/print/trigger
2559
# echo 'hist:keys=common_pid:lat=common_timestamp.usecs-$ts0:onmatch(ftrace.print).latency($lat) if buf == "end"' >> events/ftrace/print/trigger
2560
# echo 'hist:keys=lat,common_pid:sort=lat' > events/synthetic/latency/trigger
2562
The above created a synthetic event called "latency" and two histograms
2563
against the trace_marker, one gets triggered when "start" is written into the
2564
trace_marker file and the other when "end" is written. If the pids match, then
2565
it will call the "latency" synthetic event with the calculated latency as its
2566
parameter. Finally, a histogram is added to the latency synthetic event to
2567
record the calculated latency along with the pid.
2569
Now running cyclictest with::
2571
# ./cyclictest -p80 -d0 -i250 -n -a -t --tracemark -b 1000
2573
-p80 : run threads at priority 80
2574
-d0 : have all threads run at the same interval
2575
-i250 : start the interval at 250 microseconds (all threads will do this)
2576
-n : sleep with nanosleep
2577
-a : affine all threads to a separate CPU
2578
-t : one thread per available CPU
2579
--tracemark : enable trace mark writing
2580
-b 1000 : stop if any latency is greater than 1000 microseconds
2582
Note, the -b 1000 is used just to make --tracemark available.
2584
Then we can see the histogram created by this with::
2586
# cat events/synthetic/latency/hist
2587
# event histogram
2588
#
2589
# trigger info: hist:keys=lat,common_pid:vals=hitcount:sort=lat:size=2048 [active]
2590
#
2592
{ lat: 107, common_pid: 2039 } hitcount: 1
2593
{ lat: 122, common_pid: 2041 } hitcount: 1
2594
{ lat: 166, common_pid: 2039 } hitcount: 1
2595
{ lat: 174, common_pid: 2039 } hitcount: 1
2596
{ lat: 194, common_pid: 2041 } hitcount: 1
2597
{ lat: 196, common_pid: 2036 } hitcount: 1
2598
{ lat: 197, common_pid: 2038 } hitcount: 1
2599
{ lat: 198, common_pid: 2039 } hitcount: 1
2600
{ lat: 199, common_pid: 2039 } hitcount: 1
2601
{ lat: 200, common_pid: 2041 } hitcount: 1
2602
{ lat: 201, common_pid: 2039 } hitcount: 2
2603
{ lat: 202, common_pid: 2038 } hitcount: 1
2604
{ lat: 202, common_pid: 2043 } hitcount: 1
2605
{ lat: 203, common_pid: 2039 } hitcount: 1
2606
{ lat: 203, common_pid: 2036 } hitcount: 1
2607
{ lat: 203, common_pid: 2041 } hitcount: 1
2608
{ lat: 206, common_pid: 2038 } hitcount: 2
2609
{ lat: 207, common_pid: 2039 } hitcount: 1
2610
{ lat: 207, common_pid: 2036 } hitcount: 1
2611
{ lat: 208, common_pid: 2040 } hitcount: 1
2612
{ lat: 209, common_pid: 2043 } hitcount: 1
2613
{ lat: 210, common_pid: 2039 } hitcount: 1
2614
{ lat: 211, common_pid: 2039 } hitcount: 4
2615
{ lat: 212, common_pid: 2043 } hitcount: 1
2616
{ lat: 212, common_pid: 2039 } hitcount: 2
2617
{ lat: 213, common_pid: 2039 } hitcount: 1
2618
{ lat: 214, common_pid: 2038 } hitcount: 1
2619
{ lat: 214, common_pid: 2039 } hitcount: 2
2620
{ lat: 214, common_pid: 2042 } hitcount: 1
2621
{ lat: 215, common_pid: 2039 } hitcount: 1
2622
{ lat: 217, common_pid: 2036 } hitcount: 1
2623
{ lat: 217, common_pid: 2040 } hitcount: 1
2624
{ lat: 217, common_pid: 2039 } hitcount: 1
2625
{ lat: 218, common_pid: 2039 } hitcount: 6
2626
{ lat: 219, common_pid: 2039 } hitcount: 9
2627
{ lat: 220, common_pid: 2039 } hitcount: 11
2628
{ lat: 221, common_pid: 2039 } hitcount: 5
2629
{ lat: 221, common_pid: 2042 } hitcount: 1
2630
{ lat: 222, common_pid: 2039 } hitcount: 7
2631
{ lat: 223, common_pid: 2036 } hitcount: 1
2632
{ lat: 223, common_pid: 2039 } hitcount: 3
2633
{ lat: 224, common_pid: 2039 } hitcount: 4
2634
{ lat: 224, common_pid: 2037 } hitcount: 1
2635
{ lat: 224, common_pid: 2036 } hitcount: 2
2636
{ lat: 225, common_pid: 2039 } hitcount: 5
2637
{ lat: 225, common_pid: 2042 } hitcount: 1
2638
{ lat: 226, common_pid: 2039 } hitcount: 7
2639
{ lat: 226, common_pid: 2036 } hitcount: 4
2640
{ lat: 227, common_pid: 2039 } hitcount: 6
2641
{ lat: 227, common_pid: 2036 } hitcount: 12
2642
{ lat: 227, common_pid: 2043 } hitcount: 1
2643
{ lat: 228, common_pid: 2039 } hitcount: 7
2644
{ lat: 228, common_pid: 2036 } hitcount: 14
2645
{ lat: 229, common_pid: 2039 } hitcount: 9
2646
{ lat: 229, common_pid: 2036 } hitcount: 8
2647
{ lat: 229, common_pid: 2038 } hitcount: 1
2648
{ lat: 230, common_pid: 2039 } hitcount: 11
2649
{ lat: 230, common_pid: 2036 } hitcount: 6
2650
{ lat: 230, common_pid: 2043 } hitcount: 1
2651
{ lat: 230, common_pid: 2042 } hitcount: 2
2652
{ lat: 231, common_pid: 2041 } hitcount: 1
2653
{ lat: 231, common_pid: 2036 } hitcount: 6
2654
{ lat: 231, common_pid: 2043 } hitcount: 1
2655
{ lat: 231, common_pid: 2039 } hitcount: 8
2656
{ lat: 232, common_pid: 2037 } hitcount: 1
2657
{ lat: 232, common_pid: 2039 } hitcount: 6
2658
{ lat: 232, common_pid: 2040 } hitcount: 2
2659
{ lat: 232, common_pid: 2036 } hitcount: 5
2660
{ lat: 232, common_pid: 2043 } hitcount: 1
2661
{ lat: 233, common_pid: 2036 } hitcount: 5
2662
{ lat: 233, common_pid: 2039 } hitcount: 11
2663
{ lat: 234, common_pid: 2039 } hitcount: 4
2664
{ lat: 234, common_pid: 2038 } hitcount: 2
2665
{ lat: 234, common_pid: 2043 } hitcount: 2
2666
{ lat: 234, common_pid: 2036 } hitcount: 11
2667
{ lat: 234, common_pid: 2040 } hitcount: 1
2668
{ lat: 235, common_pid: 2037 } hitcount: 2
2669
{ lat: 235, common_pid: 2036 } hitcount: 8
2670
{ lat: 235, common_pid: 2043 } hitcount: 2
2671
{ lat: 235, common_pid: 2039 } hitcount: 5
2672
{ lat: 235, common_pid: 2042 } hitcount: 2
2673
{ lat: 235, common_pid: 2040 } hitcount: 4
2674
{ lat: 235, common_pid: 2041 } hitcount: 1
2675
{ lat: 236, common_pid: 2036 } hitcount: 7
2676
{ lat: 236, common_pid: 2037 } hitcount: 1
2677
{ lat: 236, common_pid: 2041 } hitcount: 5
2678
{ lat: 236, common_pid: 2039 } hitcount: 3
2679
{ lat: 236, common_pid: 2043 } hitcount: 9
2680
{ lat: 236, common_pid: 2040 } hitcount: 7
2681
{ lat: 237, common_pid: 2037 } hitcount: 1
2682
{ lat: 237, common_pid: 2040 } hitcount: 1
2683
{ lat: 237, common_pid: 2036 } hitcount: 9
2684
{ lat: 237, common_pid: 2039 } hitcount: 3
2685
{ lat: 237, common_pid: 2043 } hitcount: 8
2686
{ lat: 237, common_pid: 2042 } hitcount: 2
2687
{ lat: 237, common_pid: 2041 } hitcount: 2
2688
{ lat: 238, common_pid: 2043 } hitcount: 10
2689
{ lat: 238, common_pid: 2040 } hitcount: 1
2690
{ lat: 238, common_pid: 2037 } hitcount: 9
2691
{ lat: 238, common_pid: 2038 } hitcount: 1
2692
{ lat: 238, common_pid: 2039 } hitcount: 1
2693
{ lat: 238, common_pid: 2042 } hitcount: 3
2694
{ lat: 238, common_pid: 2036 } hitcount: 7
2695
{ lat: 239, common_pid: 2041 } hitcount: 1
2696
{ lat: 239, common_pid: 2043 } hitcount: 11
2697
{ lat: 239, common_pid: 2037 } hitcount: 11
2698
{ lat: 239, common_pid: 2038 } hitcount: 6
2699
{ lat: 239, common_pid: 2036 } hitcount: 7
2700
{ lat: 239, common_pid: 2040 } hitcount: 1
2701
{ lat: 239, common_pid: 2042 } hitcount: 9
2702
{ lat: 240, common_pid: 2037 } hitcount: 29
2703
{ lat: 240, common_pid: 2043 } hitcount: 15
2704
{ lat: 240, common_pid: 2040 } hitcount: 44
2705
{ lat: 240, common_pid: 2039 } hitcount: 1
2706
{ lat: 240, common_pid: 2041 } hitcount: 2
2707
{ lat: 240, common_pid: 2038 } hitcount: 1
2708
{ lat: 240, common_pid: 2036 } hitcount: 10
2709
{ lat: 240, common_pid: 2042 } hitcount: 13
2710
{ lat: 241, common_pid: 2036 } hitcount: 21
2711
{ lat: 241, common_pid: 2041 } hitcount: 36
2712
{ lat: 241, common_pid: 2037 } hitcount: 34
2713
{ lat: 241, common_pid: 2042 } hitcount: 14
2714
{ lat: 241, common_pid: 2040 } hitcount: 94
2715
{ lat: 241, common_pid: 2039 } hitcount: 12
2716
{ lat: 241, common_pid: 2038 } hitcount: 2
2717
{ lat: 241, common_pid: 2043 } hitcount: 28
2718
{ lat: 242, common_pid: 2040 } hitcount: 109
2719
{ lat: 242, common_pid: 2041 } hitcount: 506
2720
{ lat: 242, common_pid: 2039 } hitcount: 155
2721
{ lat: 242, common_pid: 2042 } hitcount: 21
2722
{ lat: 242, common_pid: 2037 } hitcount: 52
2723
{ lat: 242, common_pid: 2043 } hitcount: 21
2724
{ lat: 242, common_pid: 2036 } hitcount: 16
2725
{ lat: 242, common_pid: 2038 } hitcount: 156
2726
{ lat: 243, common_pid: 2037 } hitcount: 46
2727
{ lat: 243, common_pid: 2039 } hitcount: 40
2728
{ lat: 243, common_pid: 2042 } hitcount: 119
2729
{ lat: 243, common_pid: 2041 } hitcount: 611
2730
{ lat: 243, common_pid: 2036 } hitcount: 69
2731
{ lat: 243, common_pid: 2038 } hitcount: 784
2732
{ lat: 243, common_pid: 2040 } hitcount: 323
2733
{ lat: 243, common_pid: 2043 } hitcount: 14
2734
{ lat: 244, common_pid: 2043 } hitcount: 35
2735
{ lat: 244, common_pid: 2042 } hitcount: 305
2736
{ lat: 244, common_pid: 2039 } hitcount: 8
2737
{ lat: 244, common_pid: 2040 } hitcount: 4515
2738
{ lat: 244, common_pid: 2038 } hitcount: 371
2739
{ lat: 244, common_pid: 2037 } hitcount: 31
2740
{ lat: 244, common_pid: 2036 } hitcount: 114
2741
{ lat: 244, common_pid: 2041 } hitcount: 3396
2742
{ lat: 245, common_pid: 2036 } hitcount: 700
2743
{ lat: 245, common_pid: 2041 } hitcount: 2772
2744
{ lat: 245, common_pid: 2037 } hitcount: 268
2745
{ lat: 245, common_pid: 2039 } hitcount: 472
2746
{ lat: 245, common_pid: 2038 } hitcount: 2758
2747
{ lat: 245, common_pid: 2042 } hitcount: 3833
2748
{ lat: 245, common_pid: 2040 } hitcount: 3105
2749
{ lat: 245, common_pid: 2043 } hitcount: 645
2750
{ lat: 246, common_pid: 2038 } hitcount: 3451
2751
{ lat: 246, common_pid: 2041 } hitcount: 142
2752
{ lat: 246, common_pid: 2037 } hitcount: 5101
2753
{ lat: 246, common_pid: 2040 } hitcount: 68
2754
{ lat: 246, common_pid: 2043 } hitcount: 5099
2755
{ lat: 246, common_pid: 2039 } hitcount: 5608
2756
{ lat: 246, common_pid: 2042 } hitcount: 3723
2757
{ lat: 246, common_pid: 2036 } hitcount: 4738
2758
{ lat: 247, common_pid: 2042 } hitcount: 312
2759
{ lat: 247, common_pid: 2043 } hitcount: 2385
2760
{ lat: 247, common_pid: 2041 } hitcount: 452
2761
{ lat: 247, common_pid: 2038 } hitcount: 792
2762
{ lat: 247, common_pid: 2040 } hitcount: 78
2763
{ lat: 247, common_pid: 2036 } hitcount: 2375
2764
{ lat: 247, common_pid: 2039 } hitcount: 1834
2765
{ lat: 247, common_pid: 2037 } hitcount: 2655
2766
{ lat: 248, common_pid: 2037 } hitcount: 36
2767
{ lat: 248, common_pid: 2042 } hitcount: 11
2768
{ lat: 248, common_pid: 2038 } hitcount: 122
2769
{ lat: 248, common_pid: 2036 } hitcount: 135
2770
{ lat: 248, common_pid: 2039 } hitcount: 26
2771
{ lat: 248, common_pid: 2041 } hitcount: 503
2772
{ lat: 248, common_pid: 2043 } hitcount: 66
2773
{ lat: 248, common_pid: 2040 } hitcount: 46
2774
{ lat: 249, common_pid: 2037 } hitcount: 29
2775
{ lat: 249, common_pid: 2038 } hitcount: 1
2776
{ lat: 249, common_pid: 2043 } hitcount: 29
2777
{ lat: 249, common_pid: 2039 } hitcount: 8
2778
{ lat: 249, common_pid: 2042 } hitcount: 56
2779
{ lat: 249, common_pid: 2040 } hitcount: 27
2780
{ lat: 249, common_pid: 2041 } hitcount: 11
2781
{ lat: 249, common_pid: 2036 } hitcount: 27
2782
{ lat: 250, common_pid: 2038 } hitcount: 1
2783
{ lat: 250, common_pid: 2036 } hitcount: 30
2784
{ lat: 250, common_pid: 2040 } hitcount: 19
2785
{ lat: 250, common_pid: 2043 } hitcount: 22
2786
{ lat: 250, common_pid: 2042 } hitcount: 20
2787
{ lat: 250, common_pid: 2041 } hitcount: 1
2788
{ lat: 250, common_pid: 2039 } hitcount: 6
2789
{ lat: 250, common_pid: 2037 } hitcount: 48
2790
{ lat: 251, common_pid: 2037 } hitcount: 43
2791
{ lat: 251, common_pid: 2039 } hitcount: 1
2792
{ lat: 251, common_pid: 2036 } hitcount: 12
2793
{ lat: 251, common_pid: 2042 } hitcount: 2
2794
{ lat: 251, common_pid: 2041 } hitcount: 1
2795
{ lat: 251, common_pid: 2043 } hitcount: 15
2796
{ lat: 251, common_pid: 2040 } hitcount: 3
2797
{ lat: 252, common_pid: 2040 } hitcount: 1
2798
{ lat: 252, common_pid: 2036 } hitcount: 12
2799
{ lat: 252, common_pid: 2037 } hitcount: 21
2800
{ lat: 252, common_pid: 2043 } hitcount: 14
2801
{ lat: 253, common_pid: 2037 } hitcount: 21
2802
{ lat: 253, common_pid: 2039 } hitcount: 2
2803
{ lat: 253, common_pid: 2036 } hitcount: 9
2804
{ lat: 253, common_pid: 2043 } hitcount: 6
2805
{ lat: 253, common_pid: 2040 } hitcount: 1
2806
{ lat: 254, common_pid: 2036 } hitcount: 8
2807
{ lat: 254, common_pid: 2043 } hitcount: 3
2808
{ lat: 254, common_pid: 2041 } hitcount: 1
2809
{ lat: 254, common_pid: 2042 } hitcount: 1
2810
{ lat: 254, common_pid: 2039 } hitcount: 1
2811
{ lat: 254, common_pid: 2037 } hitcount: 12
2812
{ lat: 255, common_pid: 2043 } hitcount: 1
2813
{ lat: 255, common_pid: 2037 } hitcount: 2
2814
{ lat: 255, common_pid: 2036 } hitcount: 2
2815
{ lat: 255, common_pid: 2039 } hitcount: 8
2816
{ lat: 256, common_pid: 2043 } hitcount: 1
2817
{ lat: 256, common_pid: 2036 } hitcount: 4
2818
{ lat: 256, common_pid: 2039 } hitcount: 6
2819
{ lat: 257, common_pid: 2039 } hitcount: 5
2820
{ lat: 257, common_pid: 2036 } hitcount: 4
2821
{ lat: 258, common_pid: 2039 } hitcount: 5
2822
{ lat: 258, common_pid: 2036 } hitcount: 2
2823
{ lat: 259, common_pid: 2036 } hitcount: 7
2824
{ lat: 259, common_pid: 2039 } hitcount: 7
2825
{ lat: 260, common_pid: 2036 } hitcount: 8
2826
{ lat: 260, common_pid: 2039 } hitcount: 6
2827
{ lat: 261, common_pid: 2036 } hitcount: 5
2828
{ lat: 261, common_pid: 2039 } hitcount: 7
2829
{ lat: 262, common_pid: 2039 } hitcount: 5
2830
{ lat: 262, common_pid: 2036 } hitcount: 5
2831
{ lat: 263, common_pid: 2039 } hitcount: 7
2832
{ lat: 263, common_pid: 2036 } hitcount: 7
2833
{ lat: 264, common_pid: 2039 } hitcount: 9
2834
{ lat: 264, common_pid: 2036 } hitcount: 9
2835
{ lat: 265, common_pid: 2036 } hitcount: 5
2836
{ lat: 265, common_pid: 2039 } hitcount: 1
2837
{ lat: 266, common_pid: 2036 } hitcount: 1
2838
{ lat: 266, common_pid: 2039 } hitcount: 3
2839
{ lat: 267, common_pid: 2036 } hitcount: 1
2840
{ lat: 267, common_pid: 2039 } hitcount: 3
2841
{ lat: 268, common_pid: 2036 } hitcount: 1
2842
{ lat: 268, common_pid: 2039 } hitcount: 6
2843
{ lat: 269, common_pid: 2036 } hitcount: 1
2844
{ lat: 269, common_pid: 2043 } hitcount: 1
2845
{ lat: 269, common_pid: 2039 } hitcount: 2
2846
{ lat: 270, common_pid: 2040 } hitcount: 1
2847
{ lat: 270, common_pid: 2039 } hitcount: 6
2848
{ lat: 271, common_pid: 2041 } hitcount: 1
2849
{ lat: 271, common_pid: 2039 } hitcount: 5
2850
{ lat: 272, common_pid: 2039 } hitcount: 10
2851
{ lat: 273, common_pid: 2039 } hitcount: 8
2852
{ lat: 274, common_pid: 2039 } hitcount: 2
2853
{ lat: 275, common_pid: 2039 } hitcount: 1
2854
{ lat: 276, common_pid: 2039 } hitcount: 2
2855
{ lat: 276, common_pid: 2037 } hitcount: 1
2856
{ lat: 276, common_pid: 2038 } hitcount: 1
2857
{ lat: 277, common_pid: 2039 } hitcount: 1
2858
{ lat: 277, common_pid: 2042 } hitcount: 1
2859
{ lat: 278, common_pid: 2039 } hitcount: 1
2860
{ lat: 279, common_pid: 2039 } hitcount: 4
2861
{ lat: 279, common_pid: 2043 } hitcount: 1
2862
{ lat: 280, common_pid: 2039 } hitcount: 3
2863
{ lat: 283, common_pid: 2036 } hitcount: 2
2864
{ lat: 284, common_pid: 2039 } hitcount: 1
2865
{ lat: 284, common_pid: 2043 } hitcount: 1
2866
{ lat: 288, common_pid: 2039 } hitcount: 1
2867
{ lat: 289, common_pid: 2039 } hitcount: 1
2868
{ lat: 300, common_pid: 2039 } hitcount: 1
2869
{ lat: 384, common_pid: 2039 } hitcount: 1
2871
Totals:
2872
Hits: 67625
2873
Entries: 278
2874
Dropped: 0
2876
Note, the writes are around the sleep, so ideally they will all be of 250
2877
microseconds. If you are wondering how there are several that are under
2878
250 microseconds, that is because the way cyclictest works, is if one
2879
iteration comes in late, the next one will set the timer to wake up less that
2880
250. That is, if an iteration came in 50 microseconds late, the next wake up
2881
will be at 200 microseconds.
2883
But this could easily be done in userspace. To make this even more
2884
interesting, we can mix the histogram between events that happened in the
2885
kernel with trace_marker::
2887
# cd /sys/kernel/tracing
2888
# echo 'latency u64 lat' > synthetic_events
2889
# echo 'hist:keys=pid:ts0=common_timestamp.usecs' > events/sched/sched_waking/trigger
2890
# echo 'hist:keys=common_pid:lat=common_timestamp.usecs-$ts0:onmatch(sched.sched_waking).latency($lat) if buf == "end"' > events/ftrace/print/trigger
2891
# echo 'hist:keys=lat,common_pid:sort=lat' > events/synthetic/latency/trigger
2893
The difference this time is that instead of using the trace_marker to start
2894
the latency, the sched_waking event is used, matching the common_pid for the
2895
trace_marker write with the pid that is being woken by sched_waking.
2897
After running cyclictest again with the same parameters, we now have::
2899
# cat events/synthetic/latency/hist
2900
# event histogram
2901
#
2902
# trigger info: hist:keys=lat,common_pid:vals=hitcount:sort=lat:size=2048 [active]
2903
#
2905
{ lat: 7, common_pid: 2302 } hitcount: 640
2906
{ lat: 7, common_pid: 2299 } hitcount: 42
2907
{ lat: 7, common_pid: 2303 } hitcount: 18
2908
{ lat: 7, common_pid: 2305 } hitcount: 166
2909
{ lat: 7, common_pid: 2306 } hitcount: 1
2910
{ lat: 7, common_pid: 2301 } hitcount: 91
2911
{ lat: 7, common_pid: 2300 } hitcount: 17
2912
{ lat: 8, common_pid: 2303 } hitcount: 8296
2913
{ lat: 8, common_pid: 2304 } hitcount: 6864
2914
{ lat: 8, common_pid: 2305 } hitcount: 9464
2915
{ lat: 8, common_pid: 2301 } hitcount: 9213
2916
{ lat: 8, common_pid: 2306 } hitcount: 6246
2917
{ lat: 8, common_pid: 2302 } hitcount: 8797
2918
{ lat: 8, common_pid: 2299 } hitcount: 8771
2919
{ lat: 8, common_pid: 2300 } hitcount: 8119
2920
{ lat: 9, common_pid: 2305 } hitcount: 1519
2921
{ lat: 9, common_pid: 2299 } hitcount: 2346
2922
{ lat: 9, common_pid: 2303 } hitcount: 2841
2923
{ lat: 9, common_pid: 2301 } hitcount: 1846
2924
{ lat: 9, common_pid: 2304 } hitcount: 3861
2925
{ lat: 9, common_pid: 2302 } hitcount: 1210
2926
{ lat: 9, common_pid: 2300 } hitcount: 2762
2927
{ lat: 9, common_pid: 2306 } hitcount: 4247
2928
{ lat: 10, common_pid: 2299 } hitcount: 16
2929
{ lat: 10, common_pid: 2306 } hitcount: 333
2930
{ lat: 10, common_pid: 2303 } hitcount: 16
2931
{ lat: 10, common_pid: 2304 } hitcount: 168
2932
{ lat: 10, common_pid: 2302 } hitcount: 240
2933
{ lat: 10, common_pid: 2301 } hitcount: 28
2934
{ lat: 10, common_pid: 2300 } hitcount: 95
2935
{ lat: 10, common_pid: 2305 } hitcount: 18
2936
{ lat: 11, common_pid: 2303 } hitcount: 5
2937
{ lat: 11, common_pid: 2305 } hitcount: 8
2938
{ lat: 11, common_pid: 2306 } hitcount: 221
2939
{ lat: 11, common_pid: 2302 } hitcount: 76
2940
{ lat: 11, common_pid: 2304 } hitcount: 26
2941
{ lat: 11, common_pid: 2300 } hitcount: 125
2942
{ lat: 11, common_pid: 2299 } hitcount: 2
2943
{ lat: 12, common_pid: 2305 } hitcount: 3
2944
{ lat: 12, common_pid: 2300 } hitcount: 6
2945
{ lat: 12, common_pid: 2306 } hitcount: 90
2946
{ lat: 12, common_pid: 2302 } hitcount: 4
2947
{ lat: 12, common_pid: 2303 } hitcount: 1
2948
{ lat: 12, common_pid: 2304 } hitcount: 122
2949
{ lat: 13, common_pid: 2300 } hitcount: 12
2950
{ lat: 13, common_pid: 2301 } hitcount: 1
2951
{ lat: 13, common_pid: 2306 } hitcount: 32
2952
{ lat: 13, common_pid: 2302 } hitcount: 5
2953
{ lat: 13, common_pid: 2305 } hitcount: 1
2954
{ lat: 13, common_pid: 2303 } hitcount: 1
2955
{ lat: 13, common_pid: 2304 } hitcount: 61
2956
{ lat: 14, common_pid: 2303 } hitcount: 4
2957
{ lat: 14, common_pid: 2306 } hitcount: 5
2958
{ lat: 14, common_pid: 2305 } hitcount: 4
2959
{ lat: 14, common_pid: 2304 } hitcount: 62
2960
{ lat: 14, common_pid: 2302 } hitcount: 19
2961
{ lat: 14, common_pid: 2300 } hitcount: 33
2962
{ lat: 14, common_pid: 2299 } hitcount: 1
2963
{ lat: 14, common_pid: 2301 } hitcount: 4
2964
{ lat: 15, common_pid: 2305 } hitcount: 1
2965
{ lat: 15, common_pid: 2302 } hitcount: 25
2966
{ lat: 15, common_pid: 2300 } hitcount: 11
2967
{ lat: 15, common_pid: 2299 } hitcount: 5
2968
{ lat: 15, common_pid: 2301 } hitcount: 1
2969
{ lat: 15, common_pid: 2304 } hitcount: 8
2970
{ lat: 15, common_pid: 2303 } hitcount: 1
2971
{ lat: 15, common_pid: 2306 } hitcount: 6
2972
{ lat: 16, common_pid: 2302 } hitcount: 31
2973
{ lat: 16, common_pid: 2306 } hitcount: 3
2974
{ lat: 16, common_pid: 2300 } hitcount: 5
2975
{ lat: 17, common_pid: 2302 } hitcount: 6
2976
{ lat: 17, common_pid: 2303 } hitcount: 1
2977
{ lat: 18, common_pid: 2304 } hitcount: 1
2978
{ lat: 18, common_pid: 2302 } hitcount: 8
2979
{ lat: 18, common_pid: 2299 } hitcount: 1
2980
{ lat: 18, common_pid: 2301 } hitcount: 1
2981
{ lat: 19, common_pid: 2303 } hitcount: 4
2982
{ lat: 19, common_pid: 2304 } hitcount: 5
2983
{ lat: 19, common_pid: 2302 } hitcount: 4
2984
{ lat: 19, common_pid: 2299 } hitcount: 3
2985
{ lat: 19, common_pid: 2306 } hitcount: 1
2986
{ lat: 19, common_pid: 2300 } hitcount: 4
2987
{ lat: 19, common_pid: 2305 } hitcount: 5
2988
{ lat: 20, common_pid: 2299 } hitcount: 2
2989
{ lat: 20, common_pid: 2302 } hitcount: 3
2990
{ lat: 20, common_pid: 2305 } hitcount: 1
2991
{ lat: 20, common_pid: 2300 } hitcount: 2
2992
{ lat: 20, common_pid: 2301 } hitcount: 2
2993
{ lat: 20, common_pid: 2303 } hitcount: 3
2994
{ lat: 21, common_pid: 2305 } hitcount: 1
2995
{ lat: 21, common_pid: 2299 } hitcount: 5
2996
{ lat: 21, common_pid: 2303 } hitcount: 4
2997
{ lat: 21, common_pid: 2302 } hitcount: 7
2998
{ lat: 21, common_pid: 2300 } hitcount: 1
2999
{ lat: 21, common_pid: 2301 } hitcount: 5
3000
{ lat: 21, common_pid: 2304 } hitcount: 2
3001
{ lat: 22, common_pid: 2302 } hitcount: 5
3002
{ lat: 22, common_pid: 2303 } hitcount: 1
3003
{ lat: 22, common_pid: 2306 } hitcount: 3
3004
{ lat: 22, common_pid: 2301 } hitcount: 2
3005
{ lat: 22, common_pid: 2300 } hitcount: 1
3006
{ lat: 22, common_pid: 2299 } hitcount: 1
3007
{ lat: 22, common_pid: 2305 } hitcount: 1
3008
{ lat: 22, common_pid: 2304 } hitcount: 1
3009
{ lat: 23, common_pid: 2299 } hitcount: 1
3010
{ lat: 23, common_pid: 2306 } hitcount: 2
3011
{ lat: 23, common_pid: 2302 } hitcount: 6
3012
{ lat: 24, common_pid: 2302 } hitcount: 3
3013
{ lat: 24, common_pid: 2300 } hitcount: 1
3014
{ lat: 24, common_pid: 2306 } hitcount: 2
3015
{ lat: 24, common_pid: 2305 } hitcount: 1
3016
{ lat: 24, common_pid: 2299 } hitcount: 1
3017
{ lat: 25, common_pid: 2300 } hitcount: 1
3018
{ lat: 25, common_pid: 2302 } hitcount: 4
3019
{ lat: 26, common_pid: 2302 } hitcount: 2
3020
{ lat: 27, common_pid: 2305 } hitcount: 1
3021
{ lat: 27, common_pid: 2300 } hitcount: 1
3022
{ lat: 27, common_pid: 2302 } hitcount: 3
3023
{ lat: 28, common_pid: 2306 } hitcount: 1
3024
{ lat: 28, common_pid: 2302 } hitcount: 4
3025
{ lat: 29, common_pid: 2302 } hitcount: 1
3026
{ lat: 29, common_pid: 2300 } hitcount: 2
3027
{ lat: 29, common_pid: 2306 } hitcount: 1
3028
{ lat: 29, common_pid: 2304 } hitcount: 1
3029
{ lat: 30, common_pid: 2302 } hitcount: 4
3030
{ lat: 31, common_pid: 2302 } hitcount: 6
3031
{ lat: 32, common_pid: 2302 } hitcount: 1
3032
{ lat: 33, common_pid: 2299 } hitcount: 1
3033
{ lat: 33, common_pid: 2302 } hitcount: 3
3034
{ lat: 34, common_pid: 2302 } hitcount: 2
3035
{ lat: 35, common_pid: 2302 } hitcount: 1
3036
{ lat: 35, common_pid: 2304 } hitcount: 1
3037
{ lat: 36, common_pid: 2302 } hitcount: 4
3038
{ lat: 37, common_pid: 2302 } hitcount: 6
3039
{ lat: 38, common_pid: 2302 } hitcount: 2
3040
{ lat: 39, common_pid: 2302 } hitcount: 2
3041
{ lat: 39, common_pid: 2304 } hitcount: 1
3042
{ lat: 40, common_pid: 2304 } hitcount: 2
3043
{ lat: 40, common_pid: 2302 } hitcount: 5
3044
{ lat: 41, common_pid: 2304 } hitcount: 1
3045
{ lat: 41, common_pid: 2302 } hitcount: 8
3046
{ lat: 42, common_pid: 2302 } hitcount: 6
3047
{ lat: 42, common_pid: 2304 } hitcount: 1
3048
{ lat: 43, common_pid: 2302 } hitcount: 3
3049
{ lat: 43, common_pid: 2304 } hitcount: 4
3050
{ lat: 44, common_pid: 2302 } hitcount: 6
3051
{ lat: 45, common_pid: 2302 } hitcount: 5
3052
{ lat: 46, common_pid: 2302 } hitcount: 5
3053
{ lat: 47, common_pid: 2302 } hitcount: 7
3054
{ lat: 48, common_pid: 2301 } hitcount: 1
3055
{ lat: 48, common_pid: 2302 } hitcount: 9
3056
{ lat: 49, common_pid: 2302 } hitcount: 3
3057
{ lat: 50, common_pid: 2302 } hitcount: 1
3058
{ lat: 50, common_pid: 2301 } hitcount: 1
3059
{ lat: 51, common_pid: 2302 } hitcount: 2
3060
{ lat: 51, common_pid: 2301 } hitcount: 1
3061
{ lat: 61, common_pid: 2302 } hitcount: 1
3062
{ lat: 110, common_pid: 2302 } hitcount: 1
3064
Totals:
3065
Hits: 89565
3066
Entries: 158
3067
Dropped: 0
3069
This doesn't tell us any information about how late cyclictest may have
3070
woken up, but it does show us a nice histogram of how long it took from
3071
the time that cyclictest was woken to the time it made it into user space.
3. 한국어 전문 번역
영어 원문의 문단 순서와 의미를 유지한 전체 번역입니다. 코드, 함수명, symbol과 URL은 원문 표기를 유지합니다.
Event Histogram 소개
1-14이 `Event Histograms` 문서는 Tom Zanussi가 작성했다.
Histogram trigger는 trace event 데이터를 histogram으로 집계하는 특수 event trigger다. trace event와 event trigger의 기본 내용은 `Documentation/trace/events.rst`를 참조한다.
================
Event Histograms
================
Documentation written by Tom Zanussi
1. Introduction
===============
Histogram triggers are special event triggers that can be used to
aggregate trace event data into histograms. For information on
trace events and event triggers, see Documentation/trace/events.rst.
Histogram trigger 명령
15-148histogram trigger 명령은 event 적중을 hash table에 집계한다. 하나 이상의 trace event format field 또는 stacktrace가 key가 되고, 하나 이상의 numeric event field 합계와 event 적중 수인 `hitcount`가 계속 누적되는 value가 된다.
hist trigger의 전체 형식은 다음과 같다.
hist:keys=<field1[,field2,...]>[:values=<field1[,field2,...]>]
[:sort=<field1[,field2,...]>][:size=#entries][:pause][:continue]
[:clear][:name=histname1][:nohitcount][:<handler>.<action>] [if <filter>]event가 filter와 일치하면 지정한 key와 value로 hash table entry를 갱신한다. value는 numeric field여야 하며 event가 적중할 때마다 해당 field 합계에 더해진다. 명시적 value 대신 `hitcount`를 쓸 수 있고 `values`를 생략하면 유일한 value로 암시적 `hitcount`가 만들어진다. key는 일반 field 또는 kernel stacktrace를 쓰는 `common_stacktrace`가 될 수 있다. `key`와 `keys`, `val`·`vals`·`values`는 각각 같은 용도의 별칭이다. compound key는 최대 세 field를 조합하고 각 고유 조합마다 별도 entry를 만든다. `sort`는 최대 두 field를 받아 첫 field를 primary, 둘째를 secondary로 하는 중첩 정렬을 수행한다. `name`이 같은 trigger는 histogram 데이터를 공유하지만 field 수·type·이름이 모두 호환돼야 한다. 모든 event가 공통으로 가진 `hitcount`와 `common_stacktrace`는 언제나 서로 호환된다.
parameter기능제약/기본값
`keys` / `key`entry를 구분하는 field 또는 `common_stacktrace`compound key 최대 3개
`values` / `vals` / `val`entry별 numeric 합계생략 시 `hitcount`
`sort`출력 정렬 field최대 2개, 기본 `hitcount` ascending
`size`hash table entry 수기본 2048, 128~131072의 2제곱
`name`여러 compatible trigger의 데이터 공유field 수·type·이름 일치
`pause` / `continue` / `clear`집계 상태 제어기존 trigger에는 `>>` 사용
`nohitcount` / `NOHC`raw hitcount 출력 숨김raw hitcount 외 value 필요
명령의 각 parameter가 hash table 구성과 실행 상태를 제어한다.
`hist` trigger가 연결되면 event 하위 directory에 `hist` 파일이 생긴다. 파일을 읽으면 각 trigger의 hash table 전체가 출력되며, 같은 이름을 공유하는 trigger instance는 같은 table을 보여 준다. 각 entry는 중괄호 안의 key를 먼저 출력한 뒤 value field를 표시한다. numeric field는 기본적으로 10진 정수로 출력하며 field 이름 뒤 modifier로 표시 방식을 바꿀 수 있다.
modifier표시 방식
`.hex`number를 hexadecimal value로 표시
`.sym`address를 symbol로 표시
`.sym-offset`address를 symbol과 offset으로 표시
`.syscall`syscall id를 system call 이름으로 표시
`.execname``common_pid`를 program 이름으로 표시
`.log2`raw number 대신 log2 value 표시
`.buckets=size`raw number 대신 size 단위 grouping 표시
`.usecs``common_timestamp`를 microseconds로 표시
`.percent`number를 percentage value로 표시
`.graph`value를 bar graph로 표시
`.stacktrace``long[]` type을 stacktrace로 표시
원문의 modifier 표를 용도별로 구조화했다.
일반적으로 modifier를 적용할 때 field 의미를 해석하지는 않지만 몇 가지 제한이 있다.
대상허용/제한이유
value field`.hex`만 사용 가능value는 합계이므로 다른 modifier 의미가 맞지 않음
`.execname``common_pid`에만 사용 가능event가 trigger될 당시 current process의 `comm`과 대응
다른 pid field`.execname` 사용 불가pid별 comm은 event 자체의 전용 field를 사용해야 정확함
value 합계와 process 이름의 의미를 보존하기 위한 제약이다.
일반적인 사용 흐름은 trigger를 활성화하고 현재 histogram을 읽은 뒤 같은 명령 앞에 `!`를 붙여 비활성화하는 것이다.
# echo 'hist:keys=skbaddr.hex:vals=len' > \
/sys/kernel/tracing/events/net/netif_rx/trigger
# cat /sys/kernel/tracing/events/net/netif_rx/hist
# echo '!hist:keys=skbaddr.hex:vals=len' > \
/sys/kernel/tracing/events/net/netif_rx/triggerevent의 `trigger` 파일을 읽으면 현재 연결된 hist trigger 세부 정보가 나오며, 이 정보는 `hist` 파일 맨 위에도 표시된다.
hash table 기본 크기는 2048 entry다. `size`로 늘리거나 줄일 수 있으며 128~131072 범위의 2제곱이어야 한다. 2제곱이 아닌 수는 올림된다. 실행 중 사용 entry가 크기를 넘으면 무시된 적중 수가 `drops`로 보고된다.
`sort`는 정렬할 value field를 지정한다. 생략하면 `hitcount`를 ascending으로 정렬하며 반대 방향은 sort key에 `.descending`을 붙인다.
`pause`는 기존 trigger를 멈추거나 처음부터 event를 기록하지 않는 상태로 시작한다. `continue` 또는 `cont`는 paused trigger를 시작하거나 재개한다.
`clear`는 실행 중 histogram 내용을 비우되 현재 paused/active 상태는 유지한다.
기존 trigger에 `pause`, `cont`, `clear`를 적용할 때는 append 연산자 `>>`를 사용해야 한다. `>`로 쓰면 파일 truncation 때문에 trigger가 제거된다.
`nohitcount` 또는 `NOHC`는 raw hitcount 출력을 숨긴다. raw hitcount가 아닌 value field가 최소 하나 필요하므로 `vals=hitcount:nohitcount`는 거부되지만 `vals=hitcount.percent:nohitcount`는 허용된다.
2. Histogram Trigger Command
============================
A histogram trigger command is an event trigger command that
aggregates event hits into a hash table keyed on one or more trace
event format fields (or stacktrace) and a set of running totals
derived from one or more trace event format fields and/or event
counts (hitcount).
The format of a hist trigger is as follows::
hist:keys=<field1[,field2,...]>[:values=<field1[,field2,...]>]
[:sort=<field1[,field2,...]>][:size=#entries][:pause][:continue]
[:clear][:name=histname1][:nohitcount][:<handler>.<action>] [if <filter>]
When a matching event is hit, an entry is added to a hash table
using the key(s) and value(s) named. Keys and values correspond to
fields in the event's format description. Values must correspond to
numeric fields - on an event hit, the value(s) will be added to a
sum kept for that field. The special string 'hitcount' can be used
in place of an explicit value field - this is simply a count of
event hits. If 'values' isn't specified, an implicit 'hitcount'
value will be automatically created and used as the only value.
Keys can be any field, or the special string 'common_stacktrace', which
will use the event's kernel stacktrace as the key. The keywords
'keys' or 'key' can be used to specify keys, and the keywords
'values', 'vals', or 'val' can be used to specify values. Compound
keys consisting of up to three fields can be specified by the 'keys'
keyword. Hashing a compound key produces a unique entry in the
table for each unique combination of component keys, and can be
useful for providing more fine-grained summaries of event data.
Additionally, sort keys consisting of up to two fields can be
specified by the 'sort' keyword. If more than one field is
specified, the result will be a 'sort within a sort': the first key
is taken to be the primary sort key and the second the secondary
key. If a hist trigger is given a name using the 'name' parameter,
its histogram data will be shared with other triggers of the same
name, and trigger hits will update this common data. Only triggers
with 'compatible' fields can be combined in this way; triggers are
'compatible' if the fields named in the trigger share the same
number and type of fields and those fields also have the same names.
Note that any two events always share the compatible 'hitcount' and
'common_stacktrace' fields and can therefore be combined using those
fields, however pointless that may be.
'hist' triggers add a 'hist' file to each event's subdirectory.
Reading the 'hist' file for the event will dump the hash table in
its entirety to stdout. If there are multiple hist triggers
attached to an event, there will be a table for each trigger in the
output. The table displayed for a named trigger will be the same as
any other instance having the same name. Each printed hash table
entry is a simple list of the keys and values comprising the entry;
keys are printed first and are delineated by curly braces, and are
followed by the set of value fields for the entry. By default,
numeric fields are displayed as base-10 integers. This can be
modified by appending any of the following modifiers to the field
name:
============= =================================================
.hex display a number as a hex value
.sym display an address as a symbol
.sym-offset display an address as a symbol and offset
.syscall display a syscall id as a system call name
.execname display a common_pid as a program name
.log2 display log2 value rather than raw number
.buckets=size display grouping of values rather than raw number
.usecs display a common_timestamp in microseconds
.percent display a number of percentage value
.graph display a bar-graph of a value
.stacktrace display as a stacktrace (must be a long[] type)
============= =================================================
Note that in general the semantics of a given field aren't
interpreted when applying a modifier to it, but there are some
restrictions to be aware of in this regard:
- only the 'hex' modifier can be used for values (because values
are essentially sums, and the other modifiers don't make sense
in that context).
- the 'execname' modifier can only be used on a 'common_pid'. The
reason for this is that the execname is simply the 'comm' value
saved for the 'current' process when an event was triggered,
which is the same as the common_pid value saved by the event
tracing code. Trying to apply that comm value to other pid
values wouldn't be correct, and typically events that care save
pid-specific comm fields in the event itself.
A typical usage scenario would be the following to enable a hist
trigger, read its current contents, and then turn it off::
# echo 'hist:keys=skbaddr.hex:vals=len' > \
/sys/kernel/tracing/events/net/netif_rx/trigger
# cat /sys/kernel/tracing/events/net/netif_rx/hist
# echo '!hist:keys=skbaddr.hex:vals=len' > \
/sys/kernel/tracing/events/net/netif_rx/trigger
The trigger file itself can be read to show the details of the
currently attached hist trigger. This information is also displayed
at the top of the 'hist' file when read.
By default, the size of the hash table is 2048 entries. The 'size'
parameter can be used to specify more or fewer than that. The units
are in terms of hashtable entries - if a run uses more entries than
specified, the results will show the number of 'drops', the number
of hits that were ignored. The size should be a power of 2 between
128 and 131072 (any non- power-of-2 number specified will be rounded
up).
The 'sort' parameter can be used to specify a value field to sort
on. The default if unspecified is 'hitcount' and the default sort
order is 'ascending'. To sort in the opposite direction, append
.descending' to the sort key.
The 'pause' parameter can be used to pause an existing hist trigger
or to start a hist trigger but not log any events until told to do
so. 'continue' or 'cont' can be used to start or restart a paused
hist trigger.
The 'clear' parameter will clear the contents of a running hist
trigger and leave its current paused/active state.
Note that the 'pause', 'cont', and 'clear' parameters should be
applied using 'append' shell operator ('>>') if applied to an
existing trigger, rather than via the '>' operator, which will cause
the trigger to be removed through truncation.
The 'nohitcount' (or NOHC) parameter will suppress display of
raw hitcount in the histogram. This option requires at least one
value field which is not a 'raw hitcount'. For example,
'hist:...:vals=hitcount:nohitcount' is rejected, but
'hist:...:vals=hitcount.percent:nohitcount' is OK.
`enable_hist`와 `disable_hist`
149-188`enable_hist`와 `disable_hist` trigger는 한 event가 다른 event에 이미 연결된 hist trigger를 조건부로 시작하거나 중지하게 한다. 한 event에 여러 trigger를 붙여 여러 event의 집계를 함께 제어할 수 있다.
형식은 `enable_event`와 `disable_event` trigger와 매우 비슷하다.
enable_hist:<system>:<event>[:count]
disable_hist:<system>:<event>[:count]`enable_event`와 `disable_event`가 target event를 trace buffer에 기록할지 제어하는 것과 달리, `enable_hist`와 `disable_hist`는 target event를 hash table에 집계할지 제어한다.
일반적인 사용법은 target event에 paused hist trigger를 먼저 만든 뒤, 관심 조건이 발생할 때 집계를 켜고 끄는 `enable_hist`/`disable_hist` 쌍을 다른 event에 연결하는 것이다.
# echo 'hist:keys=skbaddr.hex:vals=len:pause' > \
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
# echo 'enable_hist:net:netif_receive_skb if filename==/usr/bin/wget' > \
/sys/kernel/tracing/events/sched/sched_process_exec/trigger
# echo 'disable_hist:net:netif_receive_skb if comm==wget' > \
/sys/kernel/tracing/events/sched/sched_process_exit/trigger위 예제는 특정 program이 실행되면 paused histogram을 재개해 event 집계를 시작하고, process가 종료되면 다시 pause해 집계를 멈춘다.
뒤의 예제에서는 여기서 설명한 개념과 일반적인 사용 패턴을 더 구체적으로 보여 준다.
netif_receive_skb hist trigger→pause 상태
sched_process_exec + filename filter→enable_hist→network 집계 시작
sched_process_exit + comm filter→disable_hist→network 집계 중지
histogram data→중지 후에도 보존
process 실행과 종료 event가 network histogram의 paused 상태를 제어한다.
- enable_hist/disable_hist
The enable_hist and disable_hist triggers can be used to have one
event conditionally start and stop another event's already-attached
hist trigger. Any number of enable_hist and disable_hist triggers
can be attached to a given event, allowing that event to kick off
and stop aggregations on a host of other events.
The format is very similar to the enable/disable_event triggers::
enable_hist:<system>:<event>[:count]
disable_hist:<system>:<event>[:count]
Instead of enabling or disabling the tracing of the target event
into the trace buffer as the enable/disable_event triggers do, the
enable/disable_hist triggers enable or disable the aggregation of
the target event into a hash table.
A typical usage scenario for the enable_hist/disable_hist triggers
would be to first set up a paused hist trigger on some event,
followed by an enable_hist/disable_hist pair that turns the hist
aggregation on and off when conditions of interest are hit::
# echo 'hist:keys=skbaddr.hex:vals=len:pause' > \
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
# echo 'enable_hist:net:netif_receive_skb if filename==/usr/bin/wget' > \
/sys/kernel/tracing/events/sched/sched_process_exec/trigger
# echo 'disable_hist:net:netif_receive_skb if comm==wget' > \
/sys/kernel/tracing/events/sched/sched_process_exit/trigger
The above sets up an initially paused hist trigger which is unpaused
and starts aggregating events when a given program is executed, and
which stops aggregating when the process exits and the hist trigger
is paused again.
The examples below provide a more concrete illustration of the
concepts and typical usage patterns discussed above.
특수 event field
189-206hist trigger의 key나 value로 사용할 수 있는 특수 event field가 있다. 실제 event format field처럼 보이고 동작하지만 event의 field 정의나 format 파일에는 포함되지 않는다. 모든 event에서 일반 field를 쓸 수 있는 위치에 사용할 수 있다.
fieldtype의미
`common_timestamp``u64`event와 연관된 ring buffer timestamp(ns), `.usecs`로 microseconds 해석
`common_cpu``int`event가 발생한 CPU
ring buffer metadata에서 얻는 timestamp와 CPU를 histogram field처럼 제공한다.
2.1. 'special' event fields
---------------------------
There are a number of 'special event fields' available for use as
keys or values in a hist trigger. These look like and behave as if
they were actual event fields, but aren't really part of the event's
field definition or format file. They are however available for any
event, and can be used anywhere an actual event field could be.
They are:
====================== ==== =======================================
common_timestamp u64 timestamp (from ring buffer) associated
with the event, in nanoseconds. May be
modified by .usecs to have timestamps
interpreted as microseconds.
common_cpu int the cpu on which the event occurred.
====================== ==== =======================================
확장 오류 정보
207-214hist trigger 명령 실행 중 일부 오류 조건에서는 `tracing/error_log` 파일에서 확장 오류 정보를 확인할 수 있다. 자세한 내용은 `Documentation/trace/ftrace.rst`의 `Error conditions` 절을 참조한다.
2.2. Extended error information
-------------------------------
For some error conditions encountered when invoking a hist trigger
command, extended error information is available via the
tracing/error_log file. See "Error conditions" section in
Documentation/trace/ftrace.rst for details.
`kmalloc` hist trigger 예제
215-654첫 번째 예제군은 `kmalloc` event를 사용해 집계를 만든다. hist trigger에서 쓸 수 있는 field는 `kmalloc` event의 format 파일에 나열되어 있다.
# cat /sys/kernel/tracing/events/kmem/kmalloc/format
name: kmalloc
ID: 374
format:
field:unsigned short common_type; offset:0; size:2; signed:0;
field:unsigned char common_flags; offset:2; size:1; signed:0;
field:unsigned char common_preempt_count; offset:3; size:1; signed:0;
field:int common_pid; offset:4; size:4; signed:1;
field:unsigned long call_site; offset:8; size:8; signed:0;
field:const void * ptr; offset:16; size:8; signed:0;
field:size_t bytes_req; offset:24; size:8; signed:0;
field:size_t bytes_alloc; offset:32; size:8; signed:0;
field:gfp_t gfp_flags; offset:40; size:4; signed:0;먼저 kernel에서 `kmalloc`을 한 번 이상 호출한 각 function별로 요청한 전체 byte 수를 보여 주는 간단한 table을 생성한다.
# echo 'hist:key=call_site:val=bytes_req.buckets=32' > \
/sys/kernel/tracing/events/kmem/kmalloc/trigger이 명령은 `kmalloc` event의 `call_site` field를 table key로 사용한다. 따라서 고유한 `call_site` address마다 entry 하나가 생긴다. `val=bytes_req`는 각 고유 entry에 그 call site가 요청한 byte 수의 누적 합계를 유지하라는 뜻이다. 명령에 적힌 `bytes_req.buckets=32`의 bucket 지정은 값의 집계 단위를 설정하며, 출력의 trigger 정보에서는 정규화된 value 이름이 `bytes_req`로 나타난다.
kmalloc event→call_site 읽기
call_site key→고유 hash entry 선택
bytes_req→entry 누적 합계에 더하기
implicit hitcount→entry 적중 횟수 1 증가
hist file→정렬된 entry와 totals 출력
event 한 건이 들어올 때 key를 찾고 해당 entry의 합계와 적중 수를 갱신한다.
trigger를 잠시 실행한 뒤 `kmalloc` event 하위 directory의 `hist` 파일을 읽는다. 원문 출력은 읽기 쉽도록 일부 entry를 생략했다.
# cat /sys/kernel/tracing/events/kmem/kmalloc/hist
# trigger info: hist:keys=call_site:vals=bytes_req:sort=hitcount:size=2048 [active]
{ call_site: 18446744072106379007 } hitcount: 1 bytes_req: 176
{ call_site: 18446744071579557049 } hitcount: 1 bytes_req: 1024
{ call_site: 18446744071580608289 } hitcount: 1 bytes_req: 16384
{ call_site: 18446744071581827654 } hitcount: 1 bytes_req: 24
{ call_site: 18446744071580700980 } hitcount: 1 bytes_req: 8
{ call_site: 18446744071579359876 } hitcount: 1 bytes_req: 152
{ call_site: 18446744071580795365 } hitcount: 3 bytes_req: 144
{ call_site: 18446744071581303129 } hitcount: 3 bytes_req: 144
{ call_site: 18446744071580713234 } hitcount: 4 bytes_req: 2560
{ call_site: 18446744071580933750 } hitcount: 4 bytes_req: 736
.
.
.
{ call_site: 18446744072106047046 } hitcount: 69 bytes_req: 5576
{ call_site: 18446744071582116407 } hitcount: 73 bytes_req: 2336
{ call_site: 18446744072106054684 } hitcount: 136 bytes_req: 140504
{ call_site: 18446744072106224230 } hitcount: 136 bytes_req: 19584
{ call_site: 18446744072106078074 } hitcount: 153 bytes_req: 2448
{ call_site: 18446744072106062406 } hitcount: 153 bytes_req: 36720
{ call_site: 18446744071582507929 } hitcount: 153 bytes_req: 37088
{ call_site: 18446744072102520590 } hitcount: 273 bytes_req: 10920
{ call_site: 18446744071582143559 } hitcount: 358 bytes_req: 716
{ call_site: 18446744072106465852 } hitcount: 417 bytes_req: 56712
{ call_site: 18446744072102523378 } hitcount: 485 bytes_req: 27160
{ call_site: 18446744072099568646 } hitcount: 1676 bytes_req: 33520
Totals:
Hits: 4610
Entries: 45
Dropped: 0출력은 entry마다 한 줄을 사용한다. trigger에서 지정한 key가 먼저 나오고 그 뒤에 지정한 value가 이어진다. 출력 맨 앞에는 trigger 정보가 있으며, 같은 정보는 `trigger` 파일을 읽어도 확인할 수 있다.
# cat /sys/kernel/tracing/events/kmem/kmalloc/trigger
hist:keys=call_site:vals=bytes_req:sort=hitcount:size=2048 [active]출력 끝의 전체 통계에서 `Hits`는 event trigger가 적중한 총횟수, `Entries`는 hash table에서 사용한 entry 수, `Dropped`는 사용 entry가 table의 최대 허용 수를 넘어 버린 적중 수다. `Dropped`는 보통 0이며, 0이 아니라면 `size` parameter로 table을 늘려야 할 수 있다.
출력 항목범위의미
`call_site`entry고유 allocation 호출 위치를 식별하는 key
`hitcount`entry해당 key에 귀속된 event 적중 횟수
`bytes_req`entry해당 key가 요청한 byte의 누적 합계
`Hits`전체trigger가 적중한 총횟수
`Entries`전체사용 중인 hash entry 수
`Dropped`전체table 용량 초과로 집계하지 못한 적중 수
entry 행과 마지막 totals가 서로 다른 범위의 집계 정보를 제공한다.
위 출력에는 trigger에 직접 지정하지 않은 `hitcount` field와 `sort=hitcount` parameter가 추가되어 있다. 모든 trigger는 각 entry에 귀속된 총 적중 수인 `hitcount`를 암시적으로 유지하고 출력한다. 사용자가 `sort`를 지정하지 않으면 이 값이 기본 정렬 field가 된다.
특정 field의 합계가 필요하지 않고 적중 빈도만 알고 싶다면 `values` parameter에서 명시적인 value 대신 `hitcount`를 사용할 수 있다.
hist trigger를 끄려면 command history에서 trigger 명령을 다시 불러오고 앞에 `!`를 붙여 실행한다.
# echo '!hist:key=call_site:val=bytes_req' > \
/sys/kernel/tracing/events/kmem/kmalloc/trigger처음 출력의 `call_site`는 10진 address라서 실용성이 낮다. numeric field 이름 뒤에 `.hex`를 붙이면 hexadecimal value로 표시할 수 있다.
# echo 'hist:key=call_site.hex:val=bytes_req' > \
/sys/kernel/tracing/events/kmem/kmalloc/trigger
# cat /sys/kernel/tracing/events/kmem/kmalloc/hist
# trigger info: hist:keys=call_site.hex:vals=bytes_req:sort=hitcount:size=2048 [active]
{ call_site: ffffffffa026b291 } hitcount: 1 bytes_req: 433
{ call_site: ffffffffa07186ff } hitcount: 1 bytes_req: 176
{ call_site: ffffffff811ae721 } hitcount: 1 bytes_req: 16384
{ call_site: ffffffff811c5134 } hitcount: 1 bytes_req: 8
{ call_site: ffffffffa04a9ebb } hitcount: 1 bytes_req: 511
{ call_site: ffffffff8122e0a6 } hitcount: 1 bytes_req: 12
{ call_site: ffffffff8107da84 } hitcount: 1 bytes_req: 152
{ call_site: ffffffff812d8246 } hitcount: 1 bytes_req: 24
{ call_site: ffffffff811dc1e5 } hitcount: 3 bytes_req: 144
{ call_site: ffffffffa02515e8 } hitcount: 3 bytes_req: 648
{ call_site: ffffffff81258159 } hitcount: 3 bytes_req: 144
{ call_site: ffffffff811c80f4 } hitcount: 4 bytes_req: 544
.
.
.
{ call_site: ffffffffa06c7646 } hitcount: 106 bytes_req: 8024
{ call_site: ffffffffa06cb246 } hitcount: 132 bytes_req: 31680
{ call_site: ffffffffa06cef7a } hitcount: 132 bytes_req: 2112
{ call_site: ffffffff8137e399 } hitcount: 132 bytes_req: 23232
{ call_site: ffffffffa06c941c } hitcount: 185 bytes_req: 171360
{ call_site: ffffffffa06f2a66 } hitcount: 185 bytes_req: 26640
{ call_site: ffffffffa036a70e } hitcount: 265 bytes_req: 10600
{ call_site: ffffffff81325447 } hitcount: 292 bytes_req: 584
{ call_site: ffffffffa072da3c } hitcount: 446 bytes_req: 60656
{ call_site: ffffffffa036b1f2 } hitcount: 526 bytes_req: 29456
{ call_site: ffffffffa0099c06 } hitcount: 1780 bytes_req: 35600
Totals:
Hits: 4775
Entries: 46
Dropped: 0hexadecimal address도 제한적인 정보만 준다. text address를 볼 때는 대응하는 symbol이 대개 더 유용하므로 field 이름에 `.sym` 또는 `.sym-offset`을 붙여 symbolic value로 표시할 수 있다. 다음 `.sym` 예제는 각 address를 function 또는 module symbol로 바꾼다.
# echo 'hist:key=call_site.sym:val=bytes_req' > \
/sys/kernel/tracing/events/kmem/kmalloc/trigger
# cat /sys/kernel/tracing/events/kmem/kmalloc/hist
# trigger info: hist:keys=call_site.sym:vals=bytes_req:sort=hitcount:size=2048 [active]
{ call_site: [ffffffff810adcb9] syslog_print_all } hitcount: 1 bytes_req: 1024
{ call_site: [ffffffff8154bc62] usb_control_msg } hitcount: 1 bytes_req: 8
{ call_site: [ffffffffa00bf6fe] hidraw_send_report [hid] } hitcount: 1 bytes_req: 7
{ call_site: [ffffffff8154acbe] usb_alloc_urb } hitcount: 1 bytes_req: 192
{ call_site: [ffffffffa00bf1ca] hidraw_report_event [hid] } hitcount: 1 bytes_req: 7
{ call_site: [ffffffff811e3a25] __seq_open_private } hitcount: 1 bytes_req: 40
{ call_site: [ffffffff8109524a] alloc_fair_sched_group } hitcount: 2 bytes_req: 128
{ call_site: [ffffffff811febd5] fsnotify_alloc_group } hitcount: 2 bytes_req: 528
{ call_site: [ffffffff81440f58] __tty_buffer_request_room } hitcount: 2 bytes_req: 2624
{ call_site: [ffffffff81200ba6] inotify_new_group } hitcount: 2 bytes_req: 96
{ call_site: [ffffffffa05e19af] ieee80211_start_tx_ba_session [mac80211] } hitcount: 2 bytes_req: 464
{ call_site: [ffffffff81672406] tcp_get_metrics } hitcount: 2 bytes_req: 304
{ call_site: [ffffffff81097ec2] alloc_rt_sched_group } hitcount: 2 bytes_req: 128
{ call_site: [ffffffff81089b05] sched_create_group } hitcount: 2 bytes_req: 1424
.
.
.
{ call_site: [ffffffffa04a580c] intel_crtc_page_flip [i915] } hitcount: 1185 bytes_req: 123240
{ call_site: [ffffffffa0287592] drm_mode_page_flip_ioctl [drm] } hitcount: 1185 bytes_req: 104280
{ call_site: [ffffffffa04c4a3c] intel_plane_duplicate_state [i915] } hitcount: 1402 bytes_req: 190672
{ call_site: [ffffffff812891ca] ext4_find_extent } hitcount: 1518 bytes_req: 146208
{ call_site: [ffffffffa029070e] drm_vma_node_allow [drm] } hitcount: 1746 bytes_req: 69840
{ call_site: [ffffffffa045e7c4] i915_gem_do_execbuffer.isra.23 [i915] } hitcount: 2021 bytes_req: 792312
{ call_site: [ffffffffa02911f2] drm_modeset_lock_crtc [drm] } hitcount: 2592 bytes_req: 145152
{ call_site: [ffffffffa0489a66] intel_ring_begin [i915] } hitcount: 2629 bytes_req: 378576
{ call_site: [ffffffffa046041c] i915_gem_execbuffer2 [i915] } hitcount: 2629 bytes_req: 3783248
{ call_site: [ffffffff81325607] apparmor_file_alloc_security } hitcount: 5192 bytes_req: 10384
{ call_site: [ffffffffa00b7c06] hid_report_raw_event [hid] } hitcount: 5529 bytes_req: 110584
{ call_site: [ffffffff8131ebf7] aa_alloc_task_context } hitcount: 21943 bytes_req: 702176
{ call_site: [ffffffff8125847d] ext4_htree_store_dirent } hitcount: 55759 bytes_req: 5074265
Totals:
Hits: 109928
Entries: 71
Dropped: 0위 결과는 기본 sort key인 `hitcount`를 오름차순으로 사용하므로 아래쪽에 실행 중 `kmalloc`을 가장 많이 호출한 function이 놓인다. 호출 횟수 대신 요청 byte 수가 큰 caller를 보고 가장 큰 값을 위에 놓으려면 `sort` parameter와 `.descending` modifier를 함께 사용한다.
# echo 'hist:key=call_site.sym:val=bytes_req:sort=bytes_req.descending' > \
/sys/kernel/tracing/events/kmem/kmalloc/trigger
# cat /sys/kernel/tracing/events/kmem/kmalloc/hist
# trigger info: hist:keys=call_site.sym:vals=bytes_req:sort=bytes_req.descending:size=2048 [active]
{ call_site: [ffffffffa046041c] i915_gem_execbuffer2 [i915] } hitcount: 2186 bytes_req: 3397464
{ call_site: [ffffffffa045e7c4] i915_gem_do_execbuffer.isra.23 [i915] } hitcount: 1790 bytes_req: 712176
{ call_site: [ffffffff8125847d] ext4_htree_store_dirent } hitcount: 8132 bytes_req: 513135
{ call_site: [ffffffff811e2a1b] seq_buf_alloc } hitcount: 106 bytes_req: 440128
{ call_site: [ffffffffa0489a66] intel_ring_begin [i915] } hitcount: 2186 bytes_req: 314784
{ call_site: [ffffffff812891ca] ext4_find_extent } hitcount: 2174 bytes_req: 208992
{ call_site: [ffffffff811ae8e1] __kmalloc } hitcount: 8 bytes_req: 131072
{ call_site: [ffffffffa04c4a3c] intel_plane_duplicate_state [i915] } hitcount: 859 bytes_req: 116824
{ call_site: [ffffffffa02911f2] drm_modeset_lock_crtc [drm] } hitcount: 1834 bytes_req: 102704
{ call_site: [ffffffffa04a580c] intel_crtc_page_flip [i915] } hitcount: 972 bytes_req: 101088
{ call_site: [ffffffffa0287592] drm_mode_page_flip_ioctl [drm] } hitcount: 972 bytes_req: 85536
{ call_site: [ffffffffa00b7c06] hid_report_raw_event [hid] } hitcount: 3333 bytes_req: 66664
{ call_site: [ffffffff8137e559] sg_kmalloc } hitcount: 209 bytes_req: 61632
.
.
.
{ call_site: [ffffffff81095225] alloc_fair_sched_group } hitcount: 2 bytes_req: 128
{ call_site: [ffffffff81097ec2] alloc_rt_sched_group } hitcount: 2 bytes_req: 128
{ call_site: [ffffffff812d8406] copy_semundo } hitcount: 2 bytes_req: 48
{ call_site: [ffffffff81200ba6] inotify_new_group } hitcount: 1 bytes_req: 48
{ call_site: [ffffffffa027121a] drm_getmagic [drm] } hitcount: 1 bytes_req: 48
{ call_site: [ffffffff811e3a25] __seq_open_private } hitcount: 1 bytes_req: 40
{ call_site: [ffffffff811c52f4] bprm_change_interp } hitcount: 2 bytes_req: 16
{ call_site: [ffffffff8154bc62] usb_control_msg } hitcount: 1 bytes_req: 8
{ call_site: [ffffffffa00bf1ca] hidraw_report_event [hid] } hitcount: 1 bytes_req: 7
{ call_site: [ffffffffa00bf6fe] hidraw_send_report [hid] } hitcount: 1 bytes_req: 7
Totals:
Hits: 32133
Entries: 81
Dropped: 0symbol 이름뿐 아니라 function 내부 offset과 function size도 표시하려면 `.sym` 대신 `.sym-offset`을 사용한다.
# echo 'hist:key=call_site.sym-offset:val=bytes_req:sort=bytes_req.descending' > \
/sys/kernel/tracing/events/kmem/kmalloc/trigger
# cat /sys/kernel/tracing/events/kmem/kmalloc/hist
# trigger info: hist:keys=call_site.sym-offset:vals=bytes_req:sort=bytes_req.descending:size=2048 [active]
{ call_site: [ffffffffa046041c] i915_gem_execbuffer2+0x6c/0x2c0 [i915] } hitcount: 4569 bytes_req: 3163720
{ call_site: [ffffffffa0489a66] intel_ring_begin+0xc6/0x1f0 [i915] } hitcount: 4569 bytes_req: 657936
{ call_site: [ffffffffa045e7c4] i915_gem_do_execbuffer.isra.23+0x694/0x1020 [i915] } hitcount: 1519 bytes_req: 472936
{ call_site: [ffffffffa045e646] i915_gem_do_execbuffer.isra.23+0x516/0x1020 [i915] } hitcount: 3050 bytes_req: 211832
{ call_site: [ffffffff811e2a1b] seq_buf_alloc+0x1b/0x50 } hitcount: 34 bytes_req: 148384
{ call_site: [ffffffffa04a580c] intel_crtc_page_flip+0xbc/0x870 [i915] } hitcount: 1385 bytes_req: 144040
{ call_site: [ffffffff811ae8e1] __kmalloc+0x191/0x1b0 } hitcount: 8 bytes_req: 131072
{ call_site: [ffffffffa0287592] drm_mode_page_flip_ioctl+0x282/0x360 [drm] } hitcount: 1385 bytes_req: 121880
{ call_site: [ffffffffa02911f2] drm_modeset_lock_crtc+0x32/0x100 [drm] } hitcount: 1848 bytes_req: 103488
{ call_site: [ffffffffa04c4a3c] intel_plane_duplicate_state+0x2c/0xa0 [i915] } hitcount: 461 bytes_req: 62696
{ call_site: [ffffffffa029070e] drm_vma_node_allow+0x2e/0xd0 [drm] } hitcount: 1541 bytes_req: 61640
{ call_site: [ffffffff815f8d7b] sk_prot_alloc+0xcb/0x1b0 } hitcount: 57 bytes_req: 57456
.
.
.
{ call_site: [ffffffff8109524a] alloc_fair_sched_group+0x5a/0x1a0 } hitcount: 2 bytes_req: 128
{ call_site: [ffffffffa027b921] drm_vm_open_locked+0x31/0xa0 [drm] } hitcount: 3 bytes_req: 96
{ call_site: [ffffffff8122e266] proc_self_follow_link+0x76/0xb0 } hitcount: 8 bytes_req: 96
{ call_site: [ffffffff81213e80] load_elf_binary+0x240/0x1650 } hitcount: 3 bytes_req: 84
{ call_site: [ffffffff8154bc62] usb_control_msg+0x42/0x110 } hitcount: 1 bytes_req: 8
{ call_site: [ffffffffa00bf6fe] hidraw_send_report+0x7e/0x1a0 [hid] } hitcount: 1 bytes_req: 7
{ call_site: [ffffffffa00bf1ca] hidraw_report_event+0x8a/0x120 [hid] } hitcount: 1 bytes_req: 7
Totals:
Hits: 26098
Entries: 64
Dropped: 0설정key 표시정렬 결과
`call_site`unsigned decimal address기본 `hitcount` ascending
`call_site.hex`hexadecimal address기본 `hitcount` ascending
`call_site.sym`symbol 이름기본 `hitcount` ascending
`call_site.sym:sort=bytes_req.descending`symbol 이름요청 byte 합계 descending
`call_site.sym-offset`symbol + offset/function size지정한 value 순서
같은 key를 유지하면서 표시 형식과 결과 순서를 독립적으로 바꿀 수 있다.
`values` parameter에는 여러 field를 넣을 수도 있다. 다음 예제는 요청 byte의 합계와 함께 실제 할당 byte의 합계를 보여 주고, `bytes_alloc`을 내림차순으로 정렬한다.
# echo 'hist:keys=call_site.sym:values=bytes_req,bytes_alloc:sort=bytes_alloc.descending' > \
/sys/kernel/tracing/events/kmem/kmalloc/trigger
# cat /sys/kernel/tracing/events/kmem/kmalloc/hist
# trigger info: hist:keys=call_site.sym:vals=bytes_req,bytes_alloc:sort=bytes_alloc.descending:size=2048 [active]
{ call_site: [ffffffffa046041c] i915_gem_execbuffer2 [i915] } hitcount: 7403 bytes_req: 4084360 bytes_alloc: 5958016
{ call_site: [ffffffff811e2a1b] seq_buf_alloc } hitcount: 541 bytes_req: 2213968 bytes_alloc: 2228224
{ call_site: [ffffffffa0489a66] intel_ring_begin [i915] } hitcount: 7404 bytes_req: 1066176 bytes_alloc: 1421568
{ call_site: [ffffffffa045e7c4] i915_gem_do_execbuffer.isra.23 [i915] } hitcount: 1565 bytes_req: 557368 bytes_alloc: 1037760
{ call_site: [ffffffff8125847d] ext4_htree_store_dirent } hitcount: 9557 bytes_req: 595778 bytes_alloc: 695744
{ call_site: [ffffffffa045e646] i915_gem_do_execbuffer.isra.23 [i915] } hitcount: 5839 bytes_req: 430680 bytes_alloc: 470400
{ call_site: [ffffffffa04c4a3c] intel_plane_duplicate_state [i915] } hitcount: 2388 bytes_req: 324768 bytes_alloc: 458496
{ call_site: [ffffffffa02911f2] drm_modeset_lock_crtc [drm] } hitcount: 3911 bytes_req: 219016 bytes_alloc: 250304
{ call_site: [ffffffff815f8d7b] sk_prot_alloc } hitcount: 235 bytes_req: 236880 bytes_alloc: 240640
{ call_site: [ffffffff8137e559] sg_kmalloc } hitcount: 557 bytes_req: 169024 bytes_alloc: 221760
{ call_site: [ffffffffa00b7c06] hid_report_raw_event [hid] } hitcount: 9378 bytes_req: 187548 bytes_alloc: 206312
{ call_site: [ffffffffa04a580c] intel_crtc_page_flip [i915] } hitcount: 1519 bytes_req: 157976 bytes_alloc: 194432
.
.
.
{ call_site: [ffffffff8109bd3b] sched_autogroup_create_attach } hitcount: 2 bytes_req: 144 bytes_alloc: 192
{ call_site: [ffffffff81097ee8] alloc_rt_sched_group } hitcount: 2 bytes_req: 128 bytes_alloc: 128
{ call_site: [ffffffff8109524a] alloc_fair_sched_group } hitcount: 2 bytes_req: 128 bytes_alloc: 128
{ call_site: [ffffffff81095225] alloc_fair_sched_group } hitcount: 2 bytes_req: 128 bytes_alloc: 128
{ call_site: [ffffffff81097ec2] alloc_rt_sched_group } hitcount: 2 bytes_req: 128 bytes_alloc: 128
{ call_site: [ffffffff81213e80] load_elf_binary } hitcount: 3 bytes_req: 84 bytes_alloc: 96
{ call_site: [ffffffff81079a2e] kthread_create_on_node } hitcount: 1 bytes_req: 56 bytes_alloc: 64
{ call_site: [ffffffffa00bf6fe] hidraw_send_report [hid] } hitcount: 1 bytes_req: 7 bytes_alloc: 8
{ call_site: [ffffffff8154bc62] usb_control_msg } hitcount: 1 bytes_req: 8 bytes_alloc: 8
{ call_site: [ffffffffa00bf1ca] hidraw_report_event [hid] } hitcount: 1 bytes_req: 7 bytes_alloc: 8
Totals:
Hits: 66598
Entries: 65
Dropped: 0마지막 `kmalloc` 예제는 symbolic `call_site`만 보여 주는 대신 각 call site로 이어진 kernel stack trace 전체를 함께 구분한다. key parameter에 특수 값 `common_stacktrace`를 사용하면 된다.
# echo 'hist:keys=common_stacktrace:values=bytes_req,bytes_alloc:sort=bytes_alloc' > \
/sys/kernel/tracing/events/kmem/kmalloc/trigger이 trigger는 event가 발생했을 때의 kernel stack trace를 hash table key로 사용한다. 따라서 특정 event에 이른 모든 kernel call path를 열거하고, 각 경로에 대해 event field의 누적 합계를 유지할 수 있다. 여기서는 kernel compile 중 `kmalloc`에 이른 각 call path별 요청 byte와 할당 byte를 집계한다.
kmalloc 발생→현재 kernel stack capture
stack frame 전체→common_stacktrace key
동일 call path→같은 hash entry에 병합
bytes_req + bytes_alloc→경로별 누적
hist output→경로, hitcount, 두 합계 표시
address 하나가 아니라 stack frame 전체 조합이 hash key가 된다.
# cat /sys/kernel/tracing/events/kmem/kmalloc/hist
# trigger info: hist:keys=common_stacktrace:vals=bytes_req,bytes_alloc:sort=bytes_alloc:size=2048 [active]
{ common_stacktrace:
__kmalloc_track_caller+0x10b/0x1a0
kmemdup+0x20/0x50
hidraw_report_event+0x8a/0x120 [hid]
hid_report_raw_event+0x3ea/0x440 [hid]
hid_input_report+0x112/0x190 [hid]
hid_irq_in+0xc2/0x260 [usbhid]
__usb_hcd_giveback_urb+0x72/0x120
usb_giveback_urb_bh+0x9e/0xe0
tasklet_hi_action+0xf8/0x100
__do_softirq+0x114/0x2c0
irq_exit+0xa5/0xb0
do_IRQ+0x5a/0xf0
ret_from_intr+0x0/0x30
cpuidle_enter+0x17/0x20
cpu_startup_entry+0x315/0x3e0
rest_init+0x7c/0x80
} hitcount: 3 bytes_req: 21 bytes_alloc: 24
{ common_stacktrace:
__kmalloc_track_caller+0x10b/0x1a0
kmemdup+0x20/0x50
hidraw_report_event+0x8a/0x120 [hid]
hid_report_raw_event+0x3ea/0x440 [hid]
hid_input_report+0x112/0x190 [hid]
hid_irq_in+0xc2/0x260 [usbhid]
__usb_hcd_giveback_urb+0x72/0x120
usb_giveback_urb_bh+0x9e/0xe0
tasklet_hi_action+0xf8/0x100
__do_softirq+0x114/0x2c0
irq_exit+0xa5/0xb0
do_IRQ+0x5a/0xf0
ret_from_intr+0x0/0x30
} hitcount: 3 bytes_req: 21 bytes_alloc: 24
{ common_stacktrace:
kmem_cache_alloc_trace+0xeb/0x150
aa_alloc_task_context+0x27/0x40
apparmor_cred_prepare+0x1f/0x50
security_prepare_creds+0x16/0x20
prepare_creds+0xdf/0x1a0
SyS_capset+0xb5/0x200
system_call_fastpath+0x12/0x6a
} hitcount: 1 bytes_req: 32 bytes_alloc: 32
.
.
.
{ common_stacktrace:
__kmalloc+0x11b/0x1b0
i915_gem_execbuffer2+0x6c/0x2c0 [i915]
drm_ioctl+0x349/0x670 [drm]
do_vfs_ioctl+0x2f0/0x4f0
SyS_ioctl+0x81/0xa0
system_call_fastpath+0x12/0x6a
} hitcount: 17726 bytes_req: 13944120 bytes_alloc: 19593808
{ common_stacktrace:
__kmalloc+0x11b/0x1b0
load_elf_phdrs+0x76/0xa0
load_elf_binary+0x102/0x1650
search_binary_handler+0x97/0x1d0
do_execveat_common.isra.34+0x551/0x6e0
SyS_execve+0x3a/0x50
return_from_execve+0x0/0x23
} hitcount: 33348 bytes_req: 17152128 bytes_alloc: 20226048
{ common_stacktrace:
kmem_cache_alloc_trace+0xeb/0x150
apparmor_file_alloc_security+0x27/0x40
security_file_alloc+0x16/0x20
get_empty_filp+0x93/0x1c0
path_openat+0x31/0x5f0
do_filp_open+0x3a/0x90
do_sys_open+0x128/0x220
SyS_open+0x1e/0x20
system_call_fastpath+0x12/0x6a
} hitcount: 4766422 bytes_req: 9532844 bytes_alloc: 38131376
{ common_stacktrace:
__kmalloc+0x11b/0x1b0
seq_buf_alloc+0x1b/0x50
seq_read+0x2cc/0x370
proc_reg_read+0x3d/0x80
__vfs_read+0x28/0xe0
vfs_read+0x86/0x140
SyS_read+0x46/0xb0
system_call_fastpath+0x12/0x6a
} hitcount: 19133 bytes_req: 78368768 bytes_alloc: 78368768
Totals:
Hits: 6085872
Entries: 253
Dropped: 02.3. 'hist' trigger examples
----------------------------
The first set of examples creates aggregations using the kmalloc
event. The fields that can be used for the hist trigger are listed
in the kmalloc event's format file::
# cat /sys/kernel/tracing/events/kmem/kmalloc/format
name: kmalloc
ID: 374
format:
field:unsigned short common_type; offset:0; size:2; signed:0;
field:unsigned char common_flags; offset:2; size:1; signed:0;
field:unsigned char common_preempt_count; offset:3; size:1; signed:0;
field:int common_pid; offset:4; size:4; signed:1;
field:unsigned long call_site; offset:8; size:8; signed:0;
field:const void * ptr; offset:16; size:8; signed:0;
field:size_t bytes_req; offset:24; size:8; signed:0;
field:size_t bytes_alloc; offset:32; size:8; signed:0;
field:gfp_t gfp_flags; offset:40; size:4; signed:0;
We'll start by creating a hist trigger that generates a simple table
that lists the total number of bytes requested for each function in
the kernel that made one or more calls to kmalloc::
# echo 'hist:key=call_site:val=bytes_req.buckets=32' > \
/sys/kernel/tracing/events/kmem/kmalloc/trigger
This tells the tracing system to create a 'hist' trigger using the
call_site field of the kmalloc event as the key for the table, which
just means that each unique call_site address will have an entry
created for it in the table. The 'val=bytes_req' parameter tells
the hist trigger that for each unique entry (call_site) in the
table, it should keep a running total of the number of bytes
requested by that call_site.
We'll let it run for a while and then dump the contents of the 'hist'
file in the kmalloc event's subdirectory (for readability, a number
of entries have been omitted)::
# cat /sys/kernel/tracing/events/kmem/kmalloc/hist
# trigger info: hist:keys=call_site:vals=bytes_req:sort=hitcount:size=2048 [active]
{ call_site: 18446744072106379007 } hitcount: 1 bytes_req: 176
{ call_site: 18446744071579557049 } hitcount: 1 bytes_req: 1024
{ call_site: 18446744071580608289 } hitcount: 1 bytes_req: 16384
{ call_site: 18446744071581827654 } hitcount: 1 bytes_req: 24
{ call_site: 18446744071580700980 } hitcount: 1 bytes_req: 8
{ call_site: 18446744071579359876 } hitcount: 1 bytes_req: 152
{ call_site: 18446744071580795365 } hitcount: 3 bytes_req: 144
{ call_site: 18446744071581303129 } hitcount: 3 bytes_req: 144
{ call_site: 18446744071580713234 } hitcount: 4 bytes_req: 2560
{ call_site: 18446744071580933750 } hitcount: 4 bytes_req: 736
.
.
.
{ call_site: 18446744072106047046 } hitcount: 69 bytes_req: 5576
{ call_site: 18446744071582116407 } hitcount: 73 bytes_req: 2336
{ call_site: 18446744072106054684 } hitcount: 136 bytes_req: 140504
{ call_site: 18446744072106224230 } hitcount: 136 bytes_req: 19584
{ call_site: 18446744072106078074 } hitcount: 153 bytes_req: 2448
{ call_site: 18446744072106062406 } hitcount: 153 bytes_req: 36720
{ call_site: 18446744071582507929 } hitcount: 153 bytes_req: 37088
{ call_site: 18446744072102520590 } hitcount: 273 bytes_req: 10920
{ call_site: 18446744071582143559 } hitcount: 358 bytes_req: 716
{ call_site: 18446744072106465852 } hitcount: 417 bytes_req: 56712
{ call_site: 18446744072102523378 } hitcount: 485 bytes_req: 27160
{ call_site: 18446744072099568646 } hitcount: 1676 bytes_req: 33520
Totals:
Hits: 4610
Entries: 45
Dropped: 0
The output displays a line for each entry, beginning with the key
specified in the trigger, followed by the value(s) also specified in
the trigger. At the beginning of the output is a line that displays
the trigger info, which can also be displayed by reading the
'trigger' file::
# cat /sys/kernel/tracing/events/kmem/kmalloc/trigger
hist:keys=call_site:vals=bytes_req:sort=hitcount:size=2048 [active]
At the end of the output are a few lines that display the overall
totals for the run. The 'Hits' field shows the total number of
times the event trigger was hit, the 'Entries' field shows the total
number of used entries in the hash table, and the 'Dropped' field
shows the number of hits that were dropped because the number of
used entries for the run exceeded the maximum number of entries
allowed for the table (normally 0, but if not a hint that you may
want to increase the size of the table using the 'size' parameter).
Notice in the above output that there's an extra field, 'hitcount',
which wasn't specified in the trigger. Also notice that in the
trigger info output, there's a parameter, 'sort=hitcount', which
wasn't specified in the trigger either. The reason for that is that
every trigger implicitly keeps a count of the total number of hits
attributed to a given entry, called the 'hitcount'. That hitcount
information is explicitly displayed in the output, and in the
absence of a user-specified sort parameter, is used as the default
sort field.
The value 'hitcount' can be used in place of an explicit value in
the 'values' parameter if you don't really need to have any
particular field summed and are mainly interested in hit
frequencies.
To turn the hist trigger off, simply call up the trigger in the
command history and re-execute it with a '!' prepended::
# echo '!hist:key=call_site:val=bytes_req' > \
/sys/kernel/tracing/events/kmem/kmalloc/trigger
Finally, notice that the call_site as displayed in the output above
isn't really very useful. It's an address, but normally addresses
are displayed in hex. To have a numeric field displayed as a hex
value, simply append '.hex' to the field name in the trigger::
# echo 'hist:key=call_site.hex:val=bytes_req' > \
/sys/kernel/tracing/events/kmem/kmalloc/trigger
# cat /sys/kernel/tracing/events/kmem/kmalloc/hist
# trigger info: hist:keys=call_site.hex:vals=bytes_req:sort=hitcount:size=2048 [active]
{ call_site: ffffffffa026b291 } hitcount: 1 bytes_req: 433
{ call_site: ffffffffa07186ff } hitcount: 1 bytes_req: 176
{ call_site: ffffffff811ae721 } hitcount: 1 bytes_req: 16384
{ call_site: ffffffff811c5134 } hitcount: 1 bytes_req: 8
{ call_site: ffffffffa04a9ebb } hitcount: 1 bytes_req: 511
{ call_site: ffffffff8122e0a6 } hitcount: 1 bytes_req: 12
{ call_site: ffffffff8107da84 } hitcount: 1 bytes_req: 152
{ call_site: ffffffff812d8246 } hitcount: 1 bytes_req: 24
{ call_site: ffffffff811dc1e5 } hitcount: 3 bytes_req: 144
{ call_site: ffffffffa02515e8 } hitcount: 3 bytes_req: 648
{ call_site: ffffffff81258159 } hitcount: 3 bytes_req: 144
{ call_site: ffffffff811c80f4 } hitcount: 4 bytes_req: 544
.
.
.
{ call_site: ffffffffa06c7646 } hitcount: 106 bytes_req: 8024
{ call_site: ffffffffa06cb246 } hitcount: 132 bytes_req: 31680
{ call_site: ffffffffa06cef7a } hitcount: 132 bytes_req: 2112
{ call_site: ffffffff8137e399 } hitcount: 132 bytes_req: 23232
{ call_site: ffffffffa06c941c } hitcount: 185 bytes_req: 171360
{ call_site: ffffffffa06f2a66 } hitcount: 185 bytes_req: 26640
{ call_site: ffffffffa036a70e } hitcount: 265 bytes_req: 10600
{ call_site: ffffffff81325447 } hitcount: 292 bytes_req: 584
{ call_site: ffffffffa072da3c } hitcount: 446 bytes_req: 60656
{ call_site: ffffffffa036b1f2 } hitcount: 526 bytes_req: 29456
{ call_site: ffffffffa0099c06 } hitcount: 1780 bytes_req: 35600
Totals:
Hits: 4775
Entries: 46
Dropped: 0
Even that's only marginally more useful - while hex values do look
more like addresses, what users are typically more interested in
when looking at text addresses are the corresponding symbols
instead. To have an address displayed as symbolic value instead,
simply append '.sym' or '.sym-offset' to the field name in the
trigger::
# echo 'hist:key=call_site.sym:val=bytes_req' > \
/sys/kernel/tracing/events/kmem/kmalloc/trigger
# cat /sys/kernel/tracing/events/kmem/kmalloc/hist
# trigger info: hist:keys=call_site.sym:vals=bytes_req:sort=hitcount:size=2048 [active]
{ call_site: [ffffffff810adcb9] syslog_print_all } hitcount: 1 bytes_req: 1024
{ call_site: [ffffffff8154bc62] usb_control_msg } hitcount: 1 bytes_req: 8
{ call_site: [ffffffffa00bf6fe] hidraw_send_report [hid] } hitcount: 1 bytes_req: 7
{ call_site: [ffffffff8154acbe] usb_alloc_urb } hitcount: 1 bytes_req: 192
{ call_site: [ffffffffa00bf1ca] hidraw_report_event [hid] } hitcount: 1 bytes_req: 7
{ call_site: [ffffffff811e3a25] __seq_open_private } hitcount: 1 bytes_req: 40
{ call_site: [ffffffff8109524a] alloc_fair_sched_group } hitcount: 2 bytes_req: 128
{ call_site: [ffffffff811febd5] fsnotify_alloc_group } hitcount: 2 bytes_req: 528
{ call_site: [ffffffff81440f58] __tty_buffer_request_room } hitcount: 2 bytes_req: 2624
{ call_site: [ffffffff81200ba6] inotify_new_group } hitcount: 2 bytes_req: 96
{ call_site: [ffffffffa05e19af] ieee80211_start_tx_ba_session [mac80211] } hitcount: 2 bytes_req: 464
{ call_site: [ffffffff81672406] tcp_get_metrics } hitcount: 2 bytes_req: 304
{ call_site: [ffffffff81097ec2] alloc_rt_sched_group } hitcount: 2 bytes_req: 128
{ call_site: [ffffffff81089b05] sched_create_group } hitcount: 2 bytes_req: 1424
.
.
.
{ call_site: [ffffffffa04a580c] intel_crtc_page_flip [i915] } hitcount: 1185 bytes_req: 123240
{ call_site: [ffffffffa0287592] drm_mode_page_flip_ioctl [drm] } hitcount: 1185 bytes_req: 104280
{ call_site: [ffffffffa04c4a3c] intel_plane_duplicate_state [i915] } hitcount: 1402 bytes_req: 190672
{ call_site: [ffffffff812891ca] ext4_find_extent } hitcount: 1518 bytes_req: 146208
{ call_site: [ffffffffa029070e] drm_vma_node_allow [drm] } hitcount: 1746 bytes_req: 69840
{ call_site: [ffffffffa045e7c4] i915_gem_do_execbuffer.isra.23 [i915] } hitcount: 2021 bytes_req: 792312
{ call_site: [ffffffffa02911f2] drm_modeset_lock_crtc [drm] } hitcount: 2592 bytes_req: 145152
{ call_site: [ffffffffa0489a66] intel_ring_begin [i915] } hitcount: 2629 bytes_req: 378576
{ call_site: [ffffffffa046041c] i915_gem_execbuffer2 [i915] } hitcount: 2629 bytes_req: 3783248
{ call_site: [ffffffff81325607] apparmor_file_alloc_security } hitcount: 5192 bytes_req: 10384
{ call_site: [ffffffffa00b7c06] hid_report_raw_event [hid] } hitcount: 5529 bytes_req: 110584
{ call_site: [ffffffff8131ebf7] aa_alloc_task_context } hitcount: 21943 bytes_req: 702176
{ call_site: [ffffffff8125847d] ext4_htree_store_dirent } hitcount: 55759 bytes_req: 5074265
Totals:
Hits: 109928
Entries: 71
Dropped: 0
Because the default sort key above is 'hitcount', the above shows a
the list of call_sites by increasing hitcount, so that at the bottom
we see the functions that made the most kmalloc calls during the
run. If instead we wanted to see the top kmalloc callers in
terms of the number of bytes requested rather than the number of
calls, and we wanted the top caller to appear at the top, we can use
the 'sort' parameter, along with the 'descending' modifier::
# echo 'hist:key=call_site.sym:val=bytes_req:sort=bytes_req.descending' > \
/sys/kernel/tracing/events/kmem/kmalloc/trigger
# cat /sys/kernel/tracing/events/kmem/kmalloc/hist
# trigger info: hist:keys=call_site.sym:vals=bytes_req:sort=bytes_req.descending:size=2048 [active]
{ call_site: [ffffffffa046041c] i915_gem_execbuffer2 [i915] } hitcount: 2186 bytes_req: 3397464
{ call_site: [ffffffffa045e7c4] i915_gem_do_execbuffer.isra.23 [i915] } hitcount: 1790 bytes_req: 712176
{ call_site: [ffffffff8125847d] ext4_htree_store_dirent } hitcount: 8132 bytes_req: 513135
{ call_site: [ffffffff811e2a1b] seq_buf_alloc } hitcount: 106 bytes_req: 440128
{ call_site: [ffffffffa0489a66] intel_ring_begin [i915] } hitcount: 2186 bytes_req: 314784
{ call_site: [ffffffff812891ca] ext4_find_extent } hitcount: 2174 bytes_req: 208992
{ call_site: [ffffffff811ae8e1] __kmalloc } hitcount: 8 bytes_req: 131072
{ call_site: [ffffffffa04c4a3c] intel_plane_duplicate_state [i915] } hitcount: 859 bytes_req: 116824
{ call_site: [ffffffffa02911f2] drm_modeset_lock_crtc [drm] } hitcount: 1834 bytes_req: 102704
{ call_site: [ffffffffa04a580c] intel_crtc_page_flip [i915] } hitcount: 972 bytes_req: 101088
{ call_site: [ffffffffa0287592] drm_mode_page_flip_ioctl [drm] } hitcount: 972 bytes_req: 85536
{ call_site: [ffffffffa00b7c06] hid_report_raw_event [hid] } hitcount: 3333 bytes_req: 66664
{ call_site: [ffffffff8137e559] sg_kmalloc } hitcount: 209 bytes_req: 61632
.
.
.
{ call_site: [ffffffff81095225] alloc_fair_sched_group } hitcount: 2 bytes_req: 128
{ call_site: [ffffffff81097ec2] alloc_rt_sched_group } hitcount: 2 bytes_req: 128
{ call_site: [ffffffff812d8406] copy_semundo } hitcount: 2 bytes_req: 48
{ call_site: [ffffffff81200ba6] inotify_new_group } hitcount: 1 bytes_req: 48
{ call_site: [ffffffffa027121a] drm_getmagic [drm] } hitcount: 1 bytes_req: 48
{ call_site: [ffffffff811e3a25] __seq_open_private } hitcount: 1 bytes_req: 40
{ call_site: [ffffffff811c52f4] bprm_change_interp } hitcount: 2 bytes_req: 16
{ call_site: [ffffffff8154bc62] usb_control_msg } hitcount: 1 bytes_req: 8
{ call_site: [ffffffffa00bf1ca] hidraw_report_event [hid] } hitcount: 1 bytes_req: 7
{ call_site: [ffffffffa00bf6fe] hidraw_send_report [hid] } hitcount: 1 bytes_req: 7
Totals:
Hits: 32133
Entries: 81
Dropped: 0
To display the offset and size information in addition to the symbol
name, just use 'sym-offset' instead::
# echo 'hist:key=call_site.sym-offset:val=bytes_req:sort=bytes_req.descending' > \
/sys/kernel/tracing/events/kmem/kmalloc/trigger
# cat /sys/kernel/tracing/events/kmem/kmalloc/hist
# trigger info: hist:keys=call_site.sym-offset:vals=bytes_req:sort=bytes_req.descending:size=2048 [active]
{ call_site: [ffffffffa046041c] i915_gem_execbuffer2+0x6c/0x2c0 [i915] } hitcount: 4569 bytes_req: 3163720
{ call_site: [ffffffffa0489a66] intel_ring_begin+0xc6/0x1f0 [i915] } hitcount: 4569 bytes_req: 657936
{ call_site: [ffffffffa045e7c4] i915_gem_do_execbuffer.isra.23+0x694/0x1020 [i915] } hitcount: 1519 bytes_req: 472936
{ call_site: [ffffffffa045e646] i915_gem_do_execbuffer.isra.23+0x516/0x1020 [i915] } hitcount: 3050 bytes_req: 211832
{ call_site: [ffffffff811e2a1b] seq_buf_alloc+0x1b/0x50 } hitcount: 34 bytes_req: 148384
{ call_site: [ffffffffa04a580c] intel_crtc_page_flip+0xbc/0x870 [i915] } hitcount: 1385 bytes_req: 144040
{ call_site: [ffffffff811ae8e1] __kmalloc+0x191/0x1b0 } hitcount: 8 bytes_req: 131072
{ call_site: [ffffffffa0287592] drm_mode_page_flip_ioctl+0x282/0x360 [drm] } hitcount: 1385 bytes_req: 121880
{ call_site: [ffffffffa02911f2] drm_modeset_lock_crtc+0x32/0x100 [drm] } hitcount: 1848 bytes_req: 103488
{ call_site: [ffffffffa04c4a3c] intel_plane_duplicate_state+0x2c/0xa0 [i915] } hitcount: 461 bytes_req: 62696
{ call_site: [ffffffffa029070e] drm_vma_node_allow+0x2e/0xd0 [drm] } hitcount: 1541 bytes_req: 61640
{ call_site: [ffffffff815f8d7b] sk_prot_alloc+0xcb/0x1b0 } hitcount: 57 bytes_req: 57456
.
.
.
{ call_site: [ffffffff8109524a] alloc_fair_sched_group+0x5a/0x1a0 } hitcount: 2 bytes_req: 128
{ call_site: [ffffffffa027b921] drm_vm_open_locked+0x31/0xa0 [drm] } hitcount: 3 bytes_req: 96
{ call_site: [ffffffff8122e266] proc_self_follow_link+0x76/0xb0 } hitcount: 8 bytes_req: 96
{ call_site: [ffffffff81213e80] load_elf_binary+0x240/0x1650 } hitcount: 3 bytes_req: 84
{ call_site: [ffffffff8154bc62] usb_control_msg+0x42/0x110 } hitcount: 1 bytes_req: 8
{ call_site: [ffffffffa00bf6fe] hidraw_send_report+0x7e/0x1a0 [hid] } hitcount: 1 bytes_req: 7
{ call_site: [ffffffffa00bf1ca] hidraw_report_event+0x8a/0x120 [hid] } hitcount: 1 bytes_req: 7
Totals:
Hits: 26098
Entries: 64
Dropped: 0
We can also add multiple fields to the 'values' parameter. For
example, we might want to see the total number of bytes allocated
alongside bytes requested, and display the result sorted by bytes
allocated in a descending order::
# echo 'hist:keys=call_site.sym:values=bytes_req,bytes_alloc:sort=bytes_alloc.descending' > \
/sys/kernel/tracing/events/kmem/kmalloc/trigger
# cat /sys/kernel/tracing/events/kmem/kmalloc/hist
# trigger info: hist:keys=call_site.sym:vals=bytes_req,bytes_alloc:sort=bytes_alloc.descending:size=2048 [active]
{ call_site: [ffffffffa046041c] i915_gem_execbuffer2 [i915] } hitcount: 7403 bytes_req: 4084360 bytes_alloc: 5958016
{ call_site: [ffffffff811e2a1b] seq_buf_alloc } hitcount: 541 bytes_req: 2213968 bytes_alloc: 2228224
{ call_site: [ffffffffa0489a66] intel_ring_begin [i915] } hitcount: 7404 bytes_req: 1066176 bytes_alloc: 1421568
{ call_site: [ffffffffa045e7c4] i915_gem_do_execbuffer.isra.23 [i915] } hitcount: 1565 bytes_req: 557368 bytes_alloc: 1037760
{ call_site: [ffffffff8125847d] ext4_htree_store_dirent } hitcount: 9557 bytes_req: 595778 bytes_alloc: 695744
{ call_site: [ffffffffa045e646] i915_gem_do_execbuffer.isra.23 [i915] } hitcount: 5839 bytes_req: 430680 bytes_alloc: 470400
{ call_site: [ffffffffa04c4a3c] intel_plane_duplicate_state [i915] } hitcount: 2388 bytes_req: 324768 bytes_alloc: 458496
{ call_site: [ffffffffa02911f2] drm_modeset_lock_crtc [drm] } hitcount: 3911 bytes_req: 219016 bytes_alloc: 250304
{ call_site: [ffffffff815f8d7b] sk_prot_alloc } hitcount: 235 bytes_req: 236880 bytes_alloc: 240640
{ call_site: [ffffffff8137e559] sg_kmalloc } hitcount: 557 bytes_req: 169024 bytes_alloc: 221760
{ call_site: [ffffffffa00b7c06] hid_report_raw_event [hid] } hitcount: 9378 bytes_req: 187548 bytes_alloc: 206312
{ call_site: [ffffffffa04a580c] intel_crtc_page_flip [i915] } hitcount: 1519 bytes_req: 157976 bytes_alloc: 194432
.
.
.
{ call_site: [ffffffff8109bd3b] sched_autogroup_create_attach } hitcount: 2 bytes_req: 144 bytes_alloc: 192
{ call_site: [ffffffff81097ee8] alloc_rt_sched_group } hitcount: 2 bytes_req: 128 bytes_alloc: 128
{ call_site: [ffffffff8109524a] alloc_fair_sched_group } hitcount: 2 bytes_req: 128 bytes_alloc: 128
{ call_site: [ffffffff81095225] alloc_fair_sched_group } hitcount: 2 bytes_req: 128 bytes_alloc: 128
{ call_site: [ffffffff81097ec2] alloc_rt_sched_group } hitcount: 2 bytes_req: 128 bytes_alloc: 128
{ call_site: [ffffffff81213e80] load_elf_binary } hitcount: 3 bytes_req: 84 bytes_alloc: 96
{ call_site: [ffffffff81079a2e] kthread_create_on_node } hitcount: 1 bytes_req: 56 bytes_alloc: 64
{ call_site: [ffffffffa00bf6fe] hidraw_send_report [hid] } hitcount: 1 bytes_req: 7 bytes_alloc: 8
{ call_site: [ffffffff8154bc62] usb_control_msg } hitcount: 1 bytes_req: 8 bytes_alloc: 8
{ call_site: [ffffffffa00bf1ca] hidraw_report_event [hid] } hitcount: 1 bytes_req: 7 bytes_alloc: 8
Totals:
Hits: 66598
Entries: 65
Dropped: 0
Finally, to finish off our kmalloc example, instead of simply having
the hist trigger display symbolic call_sites, we can have the hist
trigger additionally display the complete set of kernel stack traces
that led to each call_site. To do that, we simply use the special
value 'common_stacktrace' for the key parameter::
# echo 'hist:keys=common_stacktrace:values=bytes_req,bytes_alloc:sort=bytes_alloc' > \
/sys/kernel/tracing/events/kmem/kmalloc/trigger
The above trigger will use the kernel stack trace in effect when an
event is triggered as the key for the hash table. This allows the
enumeration of every kernel callpath that led up to a particular
event, along with a running total of any of the event fields for
that event. Here we tally bytes requested and bytes allocated for
every callpath in the system that led up to a kmalloc (in this case
every callpath to a kmalloc for a kernel compile)::
# cat /sys/kernel/tracing/events/kmem/kmalloc/hist
# trigger info: hist:keys=common_stacktrace:vals=bytes_req,bytes_alloc:sort=bytes_alloc:size=2048 [active]
{ common_stacktrace:
__kmalloc_track_caller+0x10b/0x1a0
kmemdup+0x20/0x50
hidraw_report_event+0x8a/0x120 [hid]
hid_report_raw_event+0x3ea/0x440 [hid]
hid_input_report+0x112/0x190 [hid]
hid_irq_in+0xc2/0x260 [usbhid]
__usb_hcd_giveback_urb+0x72/0x120
usb_giveback_urb_bh+0x9e/0xe0
tasklet_hi_action+0xf8/0x100
__do_softirq+0x114/0x2c0
irq_exit+0xa5/0xb0
do_IRQ+0x5a/0xf0
ret_from_intr+0x0/0x30
cpuidle_enter+0x17/0x20
cpu_startup_entry+0x315/0x3e0
rest_init+0x7c/0x80
} hitcount: 3 bytes_req: 21 bytes_alloc: 24
{ common_stacktrace:
__kmalloc_track_caller+0x10b/0x1a0
kmemdup+0x20/0x50
hidraw_report_event+0x8a/0x120 [hid]
hid_report_raw_event+0x3ea/0x440 [hid]
hid_input_report+0x112/0x190 [hid]
hid_irq_in+0xc2/0x260 [usbhid]
__usb_hcd_giveback_urb+0x72/0x120
usb_giveback_urb_bh+0x9e/0xe0
tasklet_hi_action+0xf8/0x100
__do_softirq+0x114/0x2c0
irq_exit+0xa5/0xb0
do_IRQ+0x5a/0xf0
ret_from_intr+0x0/0x30
} hitcount: 3 bytes_req: 21 bytes_alloc: 24
{ common_stacktrace:
kmem_cache_alloc_trace+0xeb/0x150
aa_alloc_task_context+0x27/0x40
apparmor_cred_prepare+0x1f/0x50
security_prepare_creds+0x16/0x20
prepare_creds+0xdf/0x1a0
SyS_capset+0xb5/0x200
system_call_fastpath+0x12/0x6a
} hitcount: 1 bytes_req: 32 bytes_alloc: 32
.
.
.
{ common_stacktrace:
__kmalloc+0x11b/0x1b0
i915_gem_execbuffer2+0x6c/0x2c0 [i915]
drm_ioctl+0x349/0x670 [drm]
do_vfs_ioctl+0x2f0/0x4f0
SyS_ioctl+0x81/0xa0
system_call_fastpath+0x12/0x6a
} hitcount: 17726 bytes_req: 13944120 bytes_alloc: 19593808
{ common_stacktrace:
__kmalloc+0x11b/0x1b0
load_elf_phdrs+0x76/0xa0
load_elf_binary+0x102/0x1650
search_binary_handler+0x97/0x1d0
do_execveat_common.isra.34+0x551/0x6e0
SyS_execve+0x3a/0x50
return_from_execve+0x0/0x23
} hitcount: 33348 bytes_req: 17152128 bytes_alloc: 20226048
{ common_stacktrace:
kmem_cache_alloc_trace+0xeb/0x150
apparmor_file_alloc_security+0x27/0x40
security_file_alloc+0x16/0x20
get_empty_filp+0x93/0x1c0
path_openat+0x31/0x5f0
do_filp_open+0x3a/0x90
do_sys_open+0x128/0x220
SyS_open+0x1e/0x20
system_call_fastpath+0x12/0x6a
} hitcount: 4766422 bytes_req: 9532844 bytes_alloc: 38131376
{ common_stacktrace:
__kmalloc+0x11b/0x1b0
seq_buf_alloc+0x1b/0x50
seq_read+0x2cc/0x370
proc_reg_read+0x3d/0x80
__vfs_read+0x28/0xe0
vfs_read+0x86/0x140
SyS_read+0x46/0xb0
system_call_fastpath+0x12/0x6a
} hitcount: 19133 bytes_req: 78368768 bytes_alloc: 78368768
Totals:
Hits: 6085872
Entries: 253
Dropped: 0
Process와 syscall 이름 표시
655-745process별 정렬 합계를 모아 표시하는 경우처럼 `common_pid`를 hist trigger key로 사용할 때는 특수 `.execname` modifier를 붙여 raw pid 대신 process의 executable 이름을 table에 표시할 수 있다. 다음 예제는 process마다 읽은 전체 byte 수의 합계를 유지한다.
# echo 'hist:key=common_pid.execname:val=count:sort=count.descending' > \
/sys/kernel/tracing/events/syscalls/sys_enter_read/trigger
# cat /sys/kernel/tracing/events/syscalls/sys_enter_read/hist
# trigger info: hist:keys=common_pid.execname:vals=count:sort=count.descending:size=2048 [active]
{ common_pid: gnome-terminal [ 3196] } hitcount: 280 count: 1093512
{ common_pid: Xorg [ 1309] } hitcount: 525 count: 256640
{ common_pid: compiz [ 2889] } hitcount: 59 count: 254400
{ common_pid: bash [ 8710] } hitcount: 3 count: 66369
{ common_pid: dbus-daemon-lau [ 8703] } hitcount: 49 count: 47739
{ common_pid: irqbalance [ 1252] } hitcount: 27 count: 27648
{ common_pid: 01ifupdown [ 8705] } hitcount: 3 count: 17216
{ common_pid: dbus-daemon [ 772] } hitcount: 10 count: 12396
{ common_pid: Socket Thread [ 8342] } hitcount: 11 count: 11264
{ common_pid: nm-dhcp-client. [ 8701] } hitcount: 6 count: 7424
{ common_pid: gmain [ 1315] } hitcount: 18 count: 6336
.
.
.
{ common_pid: postgres [ 1892] } hitcount: 2 count: 32
{ common_pid: postgres [ 1891] } hitcount: 2 count: 32
{ common_pid: gmain [ 8704] } hitcount: 2 count: 32
{ common_pid: upstart-dbus-br [ 2740] } hitcount: 21 count: 21
{ common_pid: nm-dispatcher.a [ 8696] } hitcount: 1 count: 16
{ common_pid: indicator-datet [ 2904] } hitcount: 1 count: 16
{ common_pid: gdbus [ 2998] } hitcount: 1 count: 16
{ common_pid: rtkit-daemon [ 2052] } hitcount: 1 count: 8
{ common_pid: init [ 1] } hitcount: 2 count: 2
Totals:
Hits: 2116
Entries: 51
Dropped: 0마찬가지로 system 전체의 syscall 적중 목록을 모으는 경우 syscall id를 key로 삼고 특수 `.syscall` modifier를 사용하면 raw id 대신 syscall 이름을 표시할 수 있다. 다음 예제는 실행 중 system의 syscall별 누적 적중 수를 유지한다.
# echo 'hist:key=id.syscall:val=hitcount' > \
/sys/kernel/tracing/events/raw_syscalls/sys_enter/trigger
# cat /sys/kernel/tracing/events/raw_syscalls/sys_enter/hist
# trigger info: hist:keys=id.syscall:vals=hitcount:sort=hitcount:size=2048 [active]
{ id: sys_fsync [ 74] } hitcount: 1
{ id: sys_newuname [ 63] } hitcount: 1
{ id: sys_prctl [157] } hitcount: 1
{ id: sys_statfs [137] } hitcount: 1
{ id: sys_symlink [ 88] } hitcount: 1
{ id: sys_sendmmsg [307] } hitcount: 1
{ id: sys_semctl [ 66] } hitcount: 1
{ id: sys_readlink [ 89] } hitcount: 3
{ id: sys_bind [ 49] } hitcount: 3
{ id: sys_getsockname [ 51] } hitcount: 3
{ id: sys_unlink [ 87] } hitcount: 3
{ id: sys_rename [ 82] } hitcount: 4
{ id: unknown_syscall [ 58] } hitcount: 4
{ id: sys_connect [ 42] } hitcount: 4
{ id: sys_getpid [ 39] } hitcount: 4
.
.
.
{ id: sys_rt_sigprocmask [ 14] } hitcount: 952
{ id: sys_futex [202] } hitcount: 1534
{ id: sys_write [ 1] } hitcount: 2689
{ id: sys_setitimer [ 38] } hitcount: 2797
{ id: sys_read [ 0] } hitcount: 3202
{ id: sys_select [ 23] } hitcount: 3773
{ id: sys_writev [ 20] } hitcount: 4531
{ id: sys_poll [ 7] } hitcount: 8314
{ id: sys_recvmsg [ 47] } hitcount: 13738
{ id: sys_ioctl [ 16] } hitcount: 21843
Totals:
Hits: 67612
Entries: 72
Dropped: 0이 syscall 집계는 system call 활동의 대략적인 전체 모습을 보여 준다. 예제 system에서는 `sys_ioctl`이 가장 많이 호출된 system call임을 알 수 있다.
key 설정표시 결과예제 집계
`common_pid.execname`executable 이름과 pidprocess별 `count` 합계
`id.syscall`syscall 이름과 syscall idsystem 전체 `hitcount`
숫자 식별자를 사람이 읽을 수 있는 이름으로 바꾸되 원래 값도 대괄호 안에 함께 보여 준다.
If you key a hist trigger on common_pid, in order for example to
gather and display sorted totals for each process, you can use the
special .execname modifier to display the executable names for the
processes in the table rather than raw pids. The example below
keeps a per-process sum of total bytes read::
# echo 'hist:key=common_pid.execname:val=count:sort=count.descending' > \
/sys/kernel/tracing/events/syscalls/sys_enter_read/trigger
# cat /sys/kernel/tracing/events/syscalls/sys_enter_read/hist
# trigger info: hist:keys=common_pid.execname:vals=count:sort=count.descending:size=2048 [active]
{ common_pid: gnome-terminal [ 3196] } hitcount: 280 count: 1093512
{ common_pid: Xorg [ 1309] } hitcount: 525 count: 256640
{ common_pid: compiz [ 2889] } hitcount: 59 count: 254400
{ common_pid: bash [ 8710] } hitcount: 3 count: 66369
{ common_pid: dbus-daemon-lau [ 8703] } hitcount: 49 count: 47739
{ common_pid: irqbalance [ 1252] } hitcount: 27 count: 27648
{ common_pid: 01ifupdown [ 8705] } hitcount: 3 count: 17216
{ common_pid: dbus-daemon [ 772] } hitcount: 10 count: 12396
{ common_pid: Socket Thread [ 8342] } hitcount: 11 count: 11264
{ common_pid: nm-dhcp-client. [ 8701] } hitcount: 6 count: 7424
{ common_pid: gmain [ 1315] } hitcount: 18 count: 6336
.
.
.
{ common_pid: postgres [ 1892] } hitcount: 2 count: 32
{ common_pid: postgres [ 1891] } hitcount: 2 count: 32
{ common_pid: gmain [ 8704] } hitcount: 2 count: 32
{ common_pid: upstart-dbus-br [ 2740] } hitcount: 21 count: 21
{ common_pid: nm-dispatcher.a [ 8696] } hitcount: 1 count: 16
{ common_pid: indicator-datet [ 2904] } hitcount: 1 count: 16
{ common_pid: gdbus [ 2998] } hitcount: 1 count: 16
{ common_pid: rtkit-daemon [ 2052] } hitcount: 1 count: 8
{ common_pid: init [ 1] } hitcount: 2 count: 2
Totals:
Hits: 2116
Entries: 51
Dropped: 0
Similarly, if you key a hist trigger on syscall id, for example to
gather and display a list of systemwide syscall hits, you can use
the special .syscall modifier to display the syscall names rather
than raw ids. The example below keeps a running total of syscall
counts for the system during the run::
# echo 'hist:key=id.syscall:val=hitcount' > \
/sys/kernel/tracing/events/raw_syscalls/sys_enter/trigger
# cat /sys/kernel/tracing/events/raw_syscalls/sys_enter/hist
# trigger info: hist:keys=id.syscall:vals=hitcount:sort=hitcount:size=2048 [active]
{ id: sys_fsync [ 74] } hitcount: 1
{ id: sys_newuname [ 63] } hitcount: 1
{ id: sys_prctl [157] } hitcount: 1
{ id: sys_statfs [137] } hitcount: 1
{ id: sys_symlink [ 88] } hitcount: 1
{ id: sys_sendmmsg [307] } hitcount: 1
{ id: sys_semctl [ 66] } hitcount: 1
{ id: sys_readlink [ 89] } hitcount: 3
{ id: sys_bind [ 49] } hitcount: 3
{ id: sys_getsockname [ 51] } hitcount: 3
{ id: sys_unlink [ 87] } hitcount: 3
{ id: sys_rename [ 82] } hitcount: 4
{ id: unknown_syscall [ 58] } hitcount: 4
{ id: sys_connect [ 42] } hitcount: 4
{ id: sys_getpid [ 39] } hitcount: 4
.
.
.
{ id: sys_rt_sigprocmask [ 14] } hitcount: 952
{ id: sys_futex [202] } hitcount: 1534
{ id: sys_write [ 1] } hitcount: 2689
{ id: sys_setitimer [ 38] } hitcount: 2797
{ id: sys_read [ 0] } hitcount: 3202
{ id: sys_select [ 23] } hitcount: 3773
{ id: sys_writev [ 20] } hitcount: 4531
{ id: sys_poll [ 7] } hitcount: 8314
{ id: sys_recvmsg [ 47] } hitcount: 13738
{ id: sys_ioctl [ 16] } hitcount: 21843
Totals:
Hits: 67612
Entries: 72
Dropped: 0
The syscall counts above provide a rough overall picture of system
call activity on the system; we can see for example that the most
popular system call on this system was the 'sys_ioctl' system call.
Compound key와 filter
746-895`compound` key를 사용하면 syscall 집계를 더 세분화해 전체 `ioctl` 횟수에 정확히 어떤 process가 기여했는지 확인할 수 있다.
다음 명령은 system call id와 pid의 고유한 조합마다 `hitcount`를 유지한다. 결과는 사실상 pid별 system call 적중 합계를 저장하는 table이며, system call id를 primary key로 하고 `hitcount` 합계를 secondary key로 정렬한다.
# echo 'hist:key=id.syscall,common_pid.execname:val=hitcount:sort=id,hitcount' > \
/sys/kernel/tracing/events/raw_syscalls/sys_enter/trigger
# cat /sys/kernel/tracing/events/raw_syscalls/sys_enter/hist
# trigger info: hist:keys=id.syscall,common_pid.execname:vals=hitcount:sort=id.syscall,hitcount:size=2048 [active]
{ id: sys_read [ 0], common_pid: rtkit-daemon [ 1877] } hitcount: 1
{ id: sys_read [ 0], common_pid: gdbus [ 2976] } hitcount: 1
{ id: sys_read [ 0], common_pid: console-kit-dae [ 3400] } hitcount: 1
{ id: sys_read [ 0], common_pid: postgres [ 1865] } hitcount: 1
{ id: sys_read [ 0], common_pid: deja-dup-monito [ 3543] } hitcount: 2
{ id: sys_read [ 0], common_pid: NetworkManager [ 890] } hitcount: 2
{ id: sys_read [ 0], common_pid: evolution-calen [ 3048] } hitcount: 2
{ id: sys_read [ 0], common_pid: postgres [ 1864] } hitcount: 2
{ id: sys_read [ 0], common_pid: nm-applet [ 3022] } hitcount: 2
{ id: sys_read [ 0], common_pid: whoopsie [ 1212] } hitcount: 2
.
.
.
{ id: sys_ioctl [ 16], common_pid: bash [ 8479] } hitcount: 1
{ id: sys_ioctl [ 16], common_pid: bash [ 3472] } hitcount: 12
{ id: sys_ioctl [ 16], common_pid: gnome-terminal [ 3199] } hitcount: 16
{ id: sys_ioctl [ 16], common_pid: Xorg [ 1267] } hitcount: 1808
{ id: sys_ioctl [ 16], common_pid: compiz [ 2994] } hitcount: 5580
.
.
.
{ id: sys_waitid [247], common_pid: upstart-dbus-br [ 2690] } hitcount: 3
{ id: sys_waitid [247], common_pid: upstart-dbus-br [ 2688] } hitcount: 16
{ id: sys_inotify_add_watch [254], common_pid: gmain [ 975] } hitcount: 2
{ id: sys_inotify_add_watch [254], common_pid: gmain [ 3204] } hitcount: 4
{ id: sys_inotify_add_watch [254], common_pid: gmain [ 2888] } hitcount: 4
{ id: sys_inotify_add_watch [254], common_pid: gmain [ 3003] } hitcount: 4
{ id: sys_inotify_add_watch [254], common_pid: gmain [ 2873] } hitcount: 4
{ id: sys_inotify_add_watch [254], common_pid: gmain [ 3196] } hitcount: 6
{ id: sys_openat [257], common_pid: java [ 2623] } hitcount: 2
{ id: sys_eventfd2 [290], common_pid: ibus-ui-gtk3 [ 2760] } hitcount: 4
{ id: sys_eventfd2 [290], common_pid: compiz [ 2994] } hitcount: 6
Totals:
Hits: 31536
Entries: 323
Dropped: 0이 목록은 pid별 `ioctl` 분석을 제공하지만 지금 필요하지 않은 다른 syscall도 모두 포함한다. `sys_ioctl`의 syscall id가 16임을 알고 있으므로 `if id == 16` filter를 붙여 나머지 syscall을 제외할 수 있다.
# echo 'hist:key=id.syscall,common_pid.execname:val=hitcount:sort=id,hitcount if id == 16' > \
/sys/kernel/tracing/events/raw_syscalls/sys_enter/trigger
# cat /sys/kernel/tracing/events/raw_syscalls/sys_enter/hist
# trigger info: hist:keys=id.syscall,common_pid.execname:vals=hitcount:sort=id.syscall,hitcount:size=2048 if id == 16 [active]
{ id: sys_ioctl [ 16], common_pid: gmain [ 2769] } hitcount: 1
{ id: sys_ioctl [ 16], common_pid: evolution-addre [ 8571] } hitcount: 1
{ id: sys_ioctl [ 16], common_pid: gmain [ 3003] } hitcount: 1
{ id: sys_ioctl [ 16], common_pid: gmain [ 2781] } hitcount: 1
{ id: sys_ioctl [ 16], common_pid: gmain [ 2829] } hitcount: 1
{ id: sys_ioctl [ 16], common_pid: bash [ 8726] } hitcount: 1
{ id: sys_ioctl [ 16], common_pid: bash [ 8508] } hitcount: 1
{ id: sys_ioctl [ 16], common_pid: gmain [ 2970] } hitcount: 1
{ id: sys_ioctl [ 16], common_pid: gmain [ 2768] } hitcount: 1
.
.
.
{ id: sys_ioctl [ 16], common_pid: pool [ 8559] } hitcount: 45
{ id: sys_ioctl [ 16], common_pid: pool [ 8555] } hitcount: 48
{ id: sys_ioctl [ 16], common_pid: pool [ 8551] } hitcount: 48
{ id: sys_ioctl [ 16], common_pid: avahi-daemon [ 896] } hitcount: 66
{ id: sys_ioctl [ 16], common_pid: Xorg [ 1267] } hitcount: 26674
{ id: sys_ioctl [ 16], common_pid: compiz [ 2994] } hitcount: 73443
Totals:
Hits: 101162
Entries: 103
Dropped: 0filter 적용 결과에서는 `compiz`와 `Xorg`가 다른 process보다 훨씬 많은 `ioctl`을 호출한다. 이 차이는 호출이 모두 필요한지 살피는 후속 조사의 출발점이 될 수 있다.
앞의 compound key 예제는 key 하나와 합계 value인 `hitcount`로 정렬했지만 두 key를 직접 사용할 수도 있다. 다음 예제는 `common_pid`와 `size` event field를 compound key로 구성한다. pid를 primary key, `size`를 secondary key로 정렬하여 각 process가 받은 `recvfrom` 크기와 횟수를 순서대로 요약한다.
# echo 'hist:key=common_pid.execname,size:val=hitcount:sort=common_pid,size' > \
/sys/kernel/tracing/events/syscalls/sys_enter_recvfrom/trigger
# cat /sys/kernel/tracing/events/syscalls/sys_enter_recvfrom/hist
# trigger info: hist:keys=common_pid.execname,size:vals=hitcount:sort=common_pid.execname,size:size=2048 [active]
{ common_pid: smbd [ 784], size: 4 } hitcount: 1
{ common_pid: dnsmasq [ 1412], size: 4096 } hitcount: 672
{ common_pid: postgres [ 1796], size: 1000 } hitcount: 6
{ common_pid: postgres [ 1867], size: 1000 } hitcount: 10
{ common_pid: bamfdaemon [ 2787], size: 28 } hitcount: 2
{ common_pid: bamfdaemon [ 2787], size: 14360 } hitcount: 1
{ common_pid: compiz [ 2994], size: 8 } hitcount: 1
{ common_pid: compiz [ 2994], size: 20 } hitcount: 11
{ common_pid: gnome-terminal [ 3199], size: 4 } hitcount: 2
{ common_pid: firefox [ 8817], size: 4 } hitcount: 1
{ common_pid: firefox [ 8817], size: 8 } hitcount: 5
{ common_pid: firefox [ 8817], size: 588 } hitcount: 2
{ common_pid: firefox [ 8817], size: 628 } hitcount: 1
{ common_pid: firefox [ 8817], size: 6944 } hitcount: 1
{ common_pid: firefox [ 8817], size: 408880 } hitcount: 2
{ common_pid: firefox [ 8822], size: 8 } hitcount: 2
{ common_pid: firefox [ 8822], size: 160 } hitcount: 2
{ common_pid: firefox [ 8822], size: 320 } hitcount: 2
{ common_pid: firefox [ 8822], size: 352 } hitcount: 1
.
.
.
{ common_pid: pool [ 8923], size: 1960 } hitcount: 10
{ common_pid: pool [ 8923], size: 2048 } hitcount: 10
{ common_pid: pool [ 8924], size: 1960 } hitcount: 10
{ common_pid: pool [ 8924], size: 2048 } hitcount: 10
{ common_pid: pool [ 8928], size: 1964 } hitcount: 4
{ common_pid: pool [ 8928], size: 1965 } hitcount: 2
{ common_pid: pool [ 8928], size: 2048 } hitcount: 6
{ common_pid: pool [ 8929], size: 1982 } hitcount: 1
{ common_pid: pool [ 8929], size: 2048 } hitcount: 1
Totals:
Hits: 2016
Entries: 224
Dropped: 0compound key는 hash 계산에서 하나의 entity로 취급되지만, 그 안의 sub-key는 각각 독립적으로 접근하고 정렬할 수 있다.
raw_syscalls:sys_enter→id + common_pid 읽기
if id == 16→sys_ioctl만 통과
id.syscall,common_pid.execname→compound hash key
hitcount→고유 syscall/process 조합별 누적
sort=id,hitcount→primary/secondary 정렬
여러 field 조합으로 entry를 구분하고 filter와 중첩 정렬로 관심 범위를 좁힌다.
We can use 'compound' keys to refine that number and provide some
further insight as to which processes exactly contribute to the
overall ioctl count.
The command below keeps a hitcount for every unique combination of
system call id and pid - the end result is essentially a table
that keeps a per-pid sum of system call hits. The results are
sorted using the system call id as the primary key, and the
hitcount sum as the secondary key::
# echo 'hist:key=id.syscall,common_pid.execname:val=hitcount:sort=id,hitcount' > \
/sys/kernel/tracing/events/raw_syscalls/sys_enter/trigger
# cat /sys/kernel/tracing/events/raw_syscalls/sys_enter/hist
# trigger info: hist:keys=id.syscall,common_pid.execname:vals=hitcount:sort=id.syscall,hitcount:size=2048 [active]
{ id: sys_read [ 0], common_pid: rtkit-daemon [ 1877] } hitcount: 1
{ id: sys_read [ 0], common_pid: gdbus [ 2976] } hitcount: 1
{ id: sys_read [ 0], common_pid: console-kit-dae [ 3400] } hitcount: 1
{ id: sys_read [ 0], common_pid: postgres [ 1865] } hitcount: 1
{ id: sys_read [ 0], common_pid: deja-dup-monito [ 3543] } hitcount: 2
{ id: sys_read [ 0], common_pid: NetworkManager [ 890] } hitcount: 2
{ id: sys_read [ 0], common_pid: evolution-calen [ 3048] } hitcount: 2
{ id: sys_read [ 0], common_pid: postgres [ 1864] } hitcount: 2
{ id: sys_read [ 0], common_pid: nm-applet [ 3022] } hitcount: 2
{ id: sys_read [ 0], common_pid: whoopsie [ 1212] } hitcount: 2
.
.
.
{ id: sys_ioctl [ 16], common_pid: bash [ 8479] } hitcount: 1
{ id: sys_ioctl [ 16], common_pid: bash [ 3472] } hitcount: 12
{ id: sys_ioctl [ 16], common_pid: gnome-terminal [ 3199] } hitcount: 16
{ id: sys_ioctl [ 16], common_pid: Xorg [ 1267] } hitcount: 1808
{ id: sys_ioctl [ 16], common_pid: compiz [ 2994] } hitcount: 5580
.
.
.
{ id: sys_waitid [247], common_pid: upstart-dbus-br [ 2690] } hitcount: 3
{ id: sys_waitid [247], common_pid: upstart-dbus-br [ 2688] } hitcount: 16
{ id: sys_inotify_add_watch [254], common_pid: gmain [ 975] } hitcount: 2
{ id: sys_inotify_add_watch [254], common_pid: gmain [ 3204] } hitcount: 4
{ id: sys_inotify_add_watch [254], common_pid: gmain [ 2888] } hitcount: 4
{ id: sys_inotify_add_watch [254], common_pid: gmain [ 3003] } hitcount: 4
{ id: sys_inotify_add_watch [254], common_pid: gmain [ 2873] } hitcount: 4
{ id: sys_inotify_add_watch [254], common_pid: gmain [ 3196] } hitcount: 6
{ id: sys_openat [257], common_pid: java [ 2623] } hitcount: 2
{ id: sys_eventfd2 [290], common_pid: ibus-ui-gtk3 [ 2760] } hitcount: 4
{ id: sys_eventfd2 [290], common_pid: compiz [ 2994] } hitcount: 6
Totals:
Hits: 31536
Entries: 323
Dropped: 0
The above list does give us a breakdown of the ioctl syscall by
pid, but it also gives us quite a bit more than that, which we
don't really care about at the moment. Since we know the syscall
id for sys_ioctl (16, displayed next to the sys_ioctl name), we
can use that to filter out all the other syscalls::
# echo 'hist:key=id.syscall,common_pid.execname:val=hitcount:sort=id,hitcount if id == 16' > \
/sys/kernel/tracing/events/raw_syscalls/sys_enter/trigger
# cat /sys/kernel/tracing/events/raw_syscalls/sys_enter/hist
# trigger info: hist:keys=id.syscall,common_pid.execname:vals=hitcount:sort=id.syscall,hitcount:size=2048 if id == 16 [active]
{ id: sys_ioctl [ 16], common_pid: gmain [ 2769] } hitcount: 1
{ id: sys_ioctl [ 16], common_pid: evolution-addre [ 8571] } hitcount: 1
{ id: sys_ioctl [ 16], common_pid: gmain [ 3003] } hitcount: 1
{ id: sys_ioctl [ 16], common_pid: gmain [ 2781] } hitcount: 1
{ id: sys_ioctl [ 16], common_pid: gmain [ 2829] } hitcount: 1
{ id: sys_ioctl [ 16], common_pid: bash [ 8726] } hitcount: 1
{ id: sys_ioctl [ 16], common_pid: bash [ 8508] } hitcount: 1
{ id: sys_ioctl [ 16], common_pid: gmain [ 2970] } hitcount: 1
{ id: sys_ioctl [ 16], common_pid: gmain [ 2768] } hitcount: 1
.
.
.
{ id: sys_ioctl [ 16], common_pid: pool [ 8559] } hitcount: 45
{ id: sys_ioctl [ 16], common_pid: pool [ 8555] } hitcount: 48
{ id: sys_ioctl [ 16], common_pid: pool [ 8551] } hitcount: 48
{ id: sys_ioctl [ 16], common_pid: avahi-daemon [ 896] } hitcount: 66
{ id: sys_ioctl [ 16], common_pid: Xorg [ 1267] } hitcount: 26674
{ id: sys_ioctl [ 16], common_pid: compiz [ 2994] } hitcount: 73443
Totals:
Hits: 101162
Entries: 103
Dropped: 0
The above output shows that 'compiz' and 'Xorg' are far and away
the heaviest ioctl callers (which might lead to questions about
whether they really need to be making all those calls and to
possible avenues for further investigation.)
The compound key examples used a key and a sum value (hitcount) to
sort the output, but we can just as easily use two keys instead.
Here's an example where we use a compound key composed of the
common_pid and size event fields. Sorting with pid as the primary
key and 'size' as the secondary key allows us to display an
ordered summary of the recvfrom sizes, with counts, received by
each process::
# echo 'hist:key=common_pid.execname,size:val=hitcount:sort=common_pid,size' > \
/sys/kernel/tracing/events/syscalls/sys_enter_recvfrom/trigger
# cat /sys/kernel/tracing/events/syscalls/sys_enter_recvfrom/hist
# trigger info: hist:keys=common_pid.execname,size:vals=hitcount:sort=common_pid.execname,size:size=2048 [active]
{ common_pid: smbd [ 784], size: 4 } hitcount: 1
{ common_pid: dnsmasq [ 1412], size: 4096 } hitcount: 672
{ common_pid: postgres [ 1796], size: 1000 } hitcount: 6
{ common_pid: postgres [ 1867], size: 1000 } hitcount: 10
{ common_pid: bamfdaemon [ 2787], size: 28 } hitcount: 2
{ common_pid: bamfdaemon [ 2787], size: 14360 } hitcount: 1
{ common_pid: compiz [ 2994], size: 8 } hitcount: 1
{ common_pid: compiz [ 2994], size: 20 } hitcount: 11
{ common_pid: gnome-terminal [ 3199], size: 4 } hitcount: 2
{ common_pid: firefox [ 8817], size: 4 } hitcount: 1
{ common_pid: firefox [ 8817], size: 8 } hitcount: 5
{ common_pid: firefox [ 8817], size: 588 } hitcount: 2
{ common_pid: firefox [ 8817], size: 628 } hitcount: 1
{ common_pid: firefox [ 8817], size: 6944 } hitcount: 1
{ common_pid: firefox [ 8817], size: 408880 } hitcount: 2
{ common_pid: firefox [ 8822], size: 8 } hitcount: 2
{ common_pid: firefox [ 8822], size: 160 } hitcount: 2
{ common_pid: firefox [ 8822], size: 320 } hitcount: 2
{ common_pid: firefox [ 8822], size: 352 } hitcount: 1
.
.
.
{ common_pid: pool [ 8923], size: 1960 } hitcount: 10
{ common_pid: pool [ 8923], size: 2048 } hitcount: 10
{ common_pid: pool [ 8924], size: 1960 } hitcount: 10
{ common_pid: pool [ 8924], size: 2048 } hitcount: 10
{ common_pid: pool [ 8928], size: 1964 } hitcount: 4
{ common_pid: pool [ 8928], size: 1965 } hitcount: 2
{ common_pid: pool [ 8928], size: 2048 } hitcount: 6
{ common_pid: pool [ 8929], size: 1982 } hitcount: 1
{ common_pid: pool [ 8929], size: 2048 } hitcount: 1
Totals:
Hits: 2016
Entries: 224
Dropped: 0
The above example also illustrates the fact that although a compound
key is treated as a single entity for hashing purposes, the sub-keys
it's composed of can be accessed independently.
문자열 key와 수동 pause/continue
896-1007다음 예제는 string field를 hash key로 사용하면서 hist trigger를 수동으로 pause하고 continue하는 방법을 보여 준다. fork 횟수를 집계하며 hash table entry가 많지 않을 것으로 예상하므로 table 크기를 256으로 줄인다.
# echo 'hist:key=child_comm:val=hitcount:size=256' > \
/sys/kernel/tracing/events/sched/sched_process_fork/trigger
# cat /sys/kernel/tracing/events/sched/sched_process_fork/hist
# trigger info: hist:keys=child_comm:vals=hitcount:sort=hitcount:size=256 [active]
{ child_comm: dconf worker } hitcount: 1
{ child_comm: ibus-daemon } hitcount: 1
{ child_comm: whoopsie } hitcount: 1
{ child_comm: smbd } hitcount: 1
{ child_comm: gdbus } hitcount: 1
{ child_comm: kthreadd } hitcount: 1
{ child_comm: dconf worker } hitcount: 1
{ child_comm: evolution-alarm } hitcount: 2
{ child_comm: Socket Thread } hitcount: 2
{ child_comm: postgres } hitcount: 2
{ child_comm: bash } hitcount: 3
{ child_comm: compiz } hitcount: 3
{ child_comm: evolution-sourc } hitcount: 4
{ child_comm: dhclient } hitcount: 4
{ child_comm: pool } hitcount: 5
{ child_comm: nm-dispatcher.a } hitcount: 8
{ child_comm: firefox } hitcount: 8
{ child_comm: dbus-daemon } hitcount: 8
{ child_comm: glib-pacrunner } hitcount: 10
{ child_comm: evolution } hitcount: 23
Totals:
Hits: 89
Entries: 20
Dropped: 0hist trigger를 일시 중지하려면 trigger를 시작한 명령에 `:pause`를 붙여 append한다. trigger 정보의 상태가 `[paused]`로 바뀌며, 기존 histogram data는 그대로 남는다.
# echo 'hist:key=child_comm:val=hitcount:size=256:pause' >> \
/sys/kernel/tracing/events/sched/sched_process_fork/trigger
# cat /sys/kernel/tracing/events/sched/sched_process_fork/hist
# trigger info: hist:keys=child_comm:vals=hitcount:sort=hitcount:size=256 [paused]
{ child_comm: dconf worker } hitcount: 1
{ child_comm: kthreadd } hitcount: 1
{ child_comm: dconf worker } hitcount: 1
{ child_comm: gdbus } hitcount: 1
{ child_comm: ibus-daemon } hitcount: 1
{ child_comm: Socket Thread } hitcount: 2
{ child_comm: evolution-alarm } hitcount: 2
{ child_comm: smbd } hitcount: 2
{ child_comm: bash } hitcount: 3
{ child_comm: whoopsie } hitcount: 3
{ child_comm: compiz } hitcount: 3
{ child_comm: evolution-sourc } hitcount: 4
{ child_comm: pool } hitcount: 5
{ child_comm: postgres } hitcount: 6
{ child_comm: firefox } hitcount: 8
{ child_comm: dhclient } hitcount: 10
{ child_comm: emacs } hitcount: 12
{ child_comm: dbus-daemon } hitcount: 20
{ child_comm: nm-dispatcher.a } hitcount: 20
{ child_comm: evolution } hitcount: 35
{ child_comm: glib-pacrunner } hitcount: 59
Totals:
Hits: 199
Entries: 21
Dropped: 0trigger가 event를 다시 집계하게 하려면 `:cont`를 append한다. trigger 정보가 다시 `[active]`가 되고 새 event가 기존 data에 누적된다.
# echo 'hist:key=child_comm:val=hitcount:size=256:cont' >> \
/sys/kernel/tracing/events/sched/sched_process_fork/trigger
# cat /sys/kernel/tracing/events/sched/sched_process_fork/hist
# trigger info: hist:keys=child_comm:vals=hitcount:sort=hitcount:size=256 [active]
{ child_comm: dconf worker } hitcount: 1
{ child_comm: dconf worker } hitcount: 1
{ child_comm: kthreadd } hitcount: 1
{ child_comm: gdbus } hitcount: 1
{ child_comm: ibus-daemon } hitcount: 1
{ child_comm: Socket Thread } hitcount: 2
{ child_comm: evolution-alarm } hitcount: 2
{ child_comm: smbd } hitcount: 2
{ child_comm: whoopsie } hitcount: 3
{ child_comm: compiz } hitcount: 3
{ child_comm: evolution-sourc } hitcount: 4
{ child_comm: bash } hitcount: 5
{ child_comm: pool } hitcount: 5
{ child_comm: postgres } hitcount: 6
{ child_comm: firefox } hitcount: 8
{ child_comm: dhclient } hitcount: 11
{ child_comm: emacs } hitcount: 12
{ child_comm: dbus-daemon } hitcount: 22
{ child_comm: nm-dispatcher.a } hitcount: 22
{ child_comm: evolution } hitcount: 35
{ child_comm: glib-pacrunner } hitcount: 59
Totals:
Hits: 206
Entries: 21
Dropped: 0hist 생성→[active]→event 집계
명령 + :pause→[paused]→집계 중지, data 유지
명령 + :cont→[active]→기존 data에 집계 재개
같은 trigger 명령에 상태 action을 append하여 data를 보존한 채 집계만 제어한다.
The next example uses a string field as the hash key and
demonstrates how you can manually pause and continue a hist trigger.
In this example, we'll aggregate fork counts and don't expect a
large number of entries in the hash table, so we'll drop it to a
much smaller number, say 256::
# echo 'hist:key=child_comm:val=hitcount:size=256' > \
/sys/kernel/tracing/events/sched/sched_process_fork/trigger
# cat /sys/kernel/tracing/events/sched/sched_process_fork/hist
# trigger info: hist:keys=child_comm:vals=hitcount:sort=hitcount:size=256 [active]
{ child_comm: dconf worker } hitcount: 1
{ child_comm: ibus-daemon } hitcount: 1
{ child_comm: whoopsie } hitcount: 1
{ child_comm: smbd } hitcount: 1
{ child_comm: gdbus } hitcount: 1
{ child_comm: kthreadd } hitcount: 1
{ child_comm: dconf worker } hitcount: 1
{ child_comm: evolution-alarm } hitcount: 2
{ child_comm: Socket Thread } hitcount: 2
{ child_comm: postgres } hitcount: 2
{ child_comm: bash } hitcount: 3
{ child_comm: compiz } hitcount: 3
{ child_comm: evolution-sourc } hitcount: 4
{ child_comm: dhclient } hitcount: 4
{ child_comm: pool } hitcount: 5
{ child_comm: nm-dispatcher.a } hitcount: 8
{ child_comm: firefox } hitcount: 8
{ child_comm: dbus-daemon } hitcount: 8
{ child_comm: glib-pacrunner } hitcount: 10
{ child_comm: evolution } hitcount: 23
Totals:
Hits: 89
Entries: 20
Dropped: 0
If we want to pause the hist trigger, we can simply append :pause to
the command that started the trigger. Notice that the trigger info
displays as [paused]::
# echo 'hist:key=child_comm:val=hitcount:size=256:pause' >> \
/sys/kernel/tracing/events/sched/sched_process_fork/trigger
# cat /sys/kernel/tracing/events/sched/sched_process_fork/hist
# trigger info: hist:keys=child_comm:vals=hitcount:sort=hitcount:size=256 [paused]
{ child_comm: dconf worker } hitcount: 1
{ child_comm: kthreadd } hitcount: 1
{ child_comm: dconf worker } hitcount: 1
{ child_comm: gdbus } hitcount: 1
{ child_comm: ibus-daemon } hitcount: 1
{ child_comm: Socket Thread } hitcount: 2
{ child_comm: evolution-alarm } hitcount: 2
{ child_comm: smbd } hitcount: 2
{ child_comm: bash } hitcount: 3
{ child_comm: whoopsie } hitcount: 3
{ child_comm: compiz } hitcount: 3
{ child_comm: evolution-sourc } hitcount: 4
{ child_comm: pool } hitcount: 5
{ child_comm: postgres } hitcount: 6
{ child_comm: firefox } hitcount: 8
{ child_comm: dhclient } hitcount: 10
{ child_comm: emacs } hitcount: 12
{ child_comm: dbus-daemon } hitcount: 20
{ child_comm: nm-dispatcher.a } hitcount: 20
{ child_comm: evolution } hitcount: 35
{ child_comm: glib-pacrunner } hitcount: 59
Totals:
Hits: 199
Entries: 21
Dropped: 0
To manually continue having the trigger aggregate events, append
:cont instead. Notice that the trigger info displays as [active]
again, and the data has changed::
# echo 'hist:key=child_comm:val=hitcount:size=256:cont' >> \
/sys/kernel/tracing/events/sched/sched_process_fork/trigger
# cat /sys/kernel/tracing/events/sched/sched_process_fork/hist
# trigger info: hist:keys=child_comm:vals=hitcount:sort=hitcount:size=256 [active]
{ child_comm: dconf worker } hitcount: 1
{ child_comm: dconf worker } hitcount: 1
{ child_comm: kthreadd } hitcount: 1
{ child_comm: gdbus } hitcount: 1
{ child_comm: ibus-daemon } hitcount: 1
{ child_comm: Socket Thread } hitcount: 2
{ child_comm: evolution-alarm } hitcount: 2
{ child_comm: smbd } hitcount: 2
{ child_comm: whoopsie } hitcount: 3
{ child_comm: compiz } hitcount: 3
{ child_comm: evolution-sourc } hitcount: 4
{ child_comm: bash } hitcount: 5
{ child_comm: pool } hitcount: 5
{ child_comm: postgres } hitcount: 6
{ child_comm: firefox } hitcount: 8
{ child_comm: dhclient } hitcount: 11
{ child_comm: emacs } hitcount: 12
{ child_comm: dbus-daemon } hitcount: 22
{ child_comm: nm-dispatcher.a } hitcount: 22
{ child_comm: evolution } hitcount: 35
{ child_comm: glib-pacrunner } hitcount: 59
Totals:
Hits: 206
Entries: 21
Dropped: 0
Paused 상태로 시작하기
1008-1019앞 예제는 hist trigger 명령에 `pause`와 `continue`를 붙여 실행 중인 집계를 중지하고 재개했다. trigger를 처음 만들 때 `:pause`를 붙이면 paused 상태로 시작할 수도 있다. 이렇게 하면 준비가 끝나기 전에 data를 모으지 않고, 측정 직전에 unpause한 뒤 작업이 끝났을 때 다시 pause할 수 있다.
이 과정을 수동으로 수행하면 어렵고 실수하기 쉽다. 대신 `enable_hist`와 `disable_hist` trigger를 사용하면 어떤 조건에 따라 hist trigger를 자동으로 시작하고 중지할 수 있다.
The previous example showed how to start and stop a hist trigger by
appending 'pause' and 'continue' to the hist trigger command. A
hist trigger can also be started in a paused state by initially
starting the trigger with ':pause' appended. This allows you to
start the trigger only when you're ready to start collecting data
and not before. For example, you could start the trigger in a
paused state, then unpause it and do something you want to measure,
then pause the trigger again when done.
Of course, doing this manually can be difficult and error-prone, but
it is possible to automatically start and stop a hist trigger based
on some condition, via the enable_hist and disable_hist triggers.조건부 histogram 측정 구간
1020-1138예를 들어 `wget`으로 충분히 큰 파일을 내려받는 동안 `netif_receive_skb` event에 이르는 각 call path가 skb 길이 측면에서 차지하는 상대적 비중을 살펴본다고 하자.
먼저 `netif_receive_skb` event에 처음부터 paused 상태인 stacktrace trigger를 설정한다.
# echo 'hist:key=common_stacktrace:vals=len:pause' > \
/sys/kernel/tracing/events/net/netif_receive_skb/trigger다음으로 `sched_process_exec` event에 `if filename==/usr/bin/wget` filter가 있는 `enable_hist` trigger를 설정한다. 이 trigger는 filename이 `/usr/bin/wget`인 `sched_process_exec` event를 볼 때만 앞서 만든 `netif_receive_skb` hist trigger의 pause를 해제한다. 그 순간부터 모든 `netif_receive_skb` event가 stacktrace를 key로 하는 hash table에 집계된다.
# echo 'enable_hist:net:netif_receive_skb if filename==/usr/bin/wget' > \
/sys/kernel/tracing/events/sched/sched_process_exec/trigger집계는 `netif_receive_skb`가 다시 paused 상태가 될 때까지 계속된다. 다음 `disable_hist` event는 `sched_process_exit` event에 `comm==wget` filter를 사용해 그 동작을 구성한다.
# echo 'disable_hist:net:netif_receive_skb if comm==wget' > \
/sys/kernel/tracing/events/sched/sched_process_exit/triggerprocess가 종료되고 `disable_hist` trigger filter의 `comm` field가 `comm==wget`과 일치할 때마다 `netif_receive_skb` hist trigger가 중지된다.
전체 효과는 `wget`이 실행되는 동안에만 `netif_receive_skb` event가 hash table에 집계되는 것이다. `wget` 명령을 실행한 다음 `hist` 파일을 읽으면 그 실행이 만든 결과가 표시된다.
$ wget https://www.kernel.org/pub/linux/kernel/v3.x/patch-3.19.xz
# cat /sys/kernel/tracing/events/net/netif_receive_skb/hist
# trigger info: hist:keys=common_stacktrace:vals=len:sort=hitcount:size=2048 [paused]
{ common_stacktrace:
__netif_receive_skb_core+0x46d/0x990
__netif_receive_skb+0x18/0x60
netif_receive_skb_internal+0x23/0x90
napi_gro_receive+0xc8/0x100
ieee80211_deliver_skb+0xd6/0x270 [mac80211]
ieee80211_rx_handlers+0xccf/0x22f0 [mac80211]
ieee80211_prepare_and_rx_handle+0x4e7/0xc40 [mac80211]
ieee80211_rx+0x31d/0x900 [mac80211]
iwlagn_rx_reply_rx+0x3db/0x6f0 [iwldvm]
iwl_rx_dispatch+0x8e/0xf0 [iwldvm]
iwl_pcie_irq_handler+0xe3c/0x12f0 [iwlwifi]
irq_thread_fn+0x20/0x50
irq_thread+0x11f/0x150
kthread+0xd2/0xf0
ret_from_fork+0x42/0x70
} hitcount: 85 len: 28884
{ common_stacktrace:
__netif_receive_skb_core+0x46d/0x990
__netif_receive_skb+0x18/0x60
netif_receive_skb_internal+0x23/0x90
napi_gro_complete+0xa4/0xe0
dev_gro_receive+0x23a/0x360
napi_gro_receive+0x30/0x100
ieee80211_deliver_skb+0xd6/0x270 [mac80211]
ieee80211_rx_handlers+0xccf/0x22f0 [mac80211]
ieee80211_prepare_and_rx_handle+0x4e7/0xc40 [mac80211]
ieee80211_rx+0x31d/0x900 [mac80211]
iwlagn_rx_reply_rx+0x3db/0x6f0 [iwldvm]
iwl_rx_dispatch+0x8e/0xf0 [iwldvm]
iwl_pcie_irq_handler+0xe3c/0x12f0 [iwlwifi]
irq_thread_fn+0x20/0x50
irq_thread+0x11f/0x150
kthread+0xd2/0xf0
} hitcount: 98 len: 664329
{ common_stacktrace:
__netif_receive_skb_core+0x46d/0x990
__netif_receive_skb+0x18/0x60
process_backlog+0xa8/0x150
net_rx_action+0x15d/0x340
__do_softirq+0x114/0x2c0
do_softirq_own_stack+0x1c/0x30
do_softirq+0x65/0x70
__local_bh_enable_ip+0xb5/0xc0
ip_finish_output+0x1f4/0x840
ip_output+0x6b/0xc0
ip_local_out_sk+0x31/0x40
ip_send_skb+0x1a/0x50
udp_send_skb+0x173/0x2a0
udp_sendmsg+0x2bf/0x9f0
inet_sendmsg+0x64/0xa0
sock_sendmsg+0x3d/0x50
} hitcount: 115 len: 13030
{ common_stacktrace:
__netif_receive_skb_core+0x46d/0x990
__netif_receive_skb+0x18/0x60
netif_receive_skb_internal+0x23/0x90
napi_gro_complete+0xa4/0xe0
napi_gro_flush+0x6d/0x90
iwl_pcie_irq_handler+0x92a/0x12f0 [iwlwifi]
irq_thread_fn+0x20/0x50
irq_thread+0x11f/0x150
kthread+0xd2/0xf0
ret_from_fork+0x42/0x70
} hitcount: 934 len: 5512212
Totals:
Hits: 1232
Entries: 4
Dropped: 0출력에는 `wget` 실행 구간에 발생한 모든 `netif_receive_skb` call path와 각 경로의 전체 길이 합계가 나타난다.
netif_receive_skb hist→common_stacktrace key→초기 [paused]
sched_process_exec→filename == /usr/bin/wget→enable_hist
wget 실행 중→stacktrace별 len 누적
sched_process_exit→comm == wget→disable_hist
netif_receive_skb hist→[paused]→측정 결과 보존
process lifecycle event가 network histogram의 집계 시간을 wget 실행 구간으로 제한한다.
For example, suppose we wanted to take a look at the relative
weights in terms of skb length for each callpath that leads to a
netif_receive_skb event when downloading a decent-sized file using
wget.
First we set up an initially paused stacktrace trigger on the
netif_receive_skb event::
# echo 'hist:key=common_stacktrace:vals=len:pause' > \
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
Next, we set up an 'enable_hist' trigger on the sched_process_exec
event, with an 'if filename==/usr/bin/wget' filter. The effect of
this new trigger is that it will 'unpause' the hist trigger we just
set up on netif_receive_skb if and only if it sees a
sched_process_exec event with a filename of '/usr/bin/wget'. When
that happens, all netif_receive_skb events are aggregated into a
hash table keyed on stacktrace::
# echo 'enable_hist:net:netif_receive_skb if filename==/usr/bin/wget' > \
/sys/kernel/tracing/events/sched/sched_process_exec/trigger
The aggregation continues until the netif_receive_skb is paused
again, which is what the following disable_hist event does by
creating a similar setup on the sched_process_exit event, using the
filter 'comm==wget'::
# echo 'disable_hist:net:netif_receive_skb if comm==wget' > \
/sys/kernel/tracing/events/sched/sched_process_exit/trigger
Whenever a process exits and the comm field of the disable_hist
trigger filter matches 'comm==wget', the netif_receive_skb hist
trigger is disabled.
The overall effect is that netif_receive_skb events are aggregated
into the hash table for only the duration of the wget. Executing a
wget command and then listing the 'hist' file will display the
output generated by the wget command::
$ wget https://www.kernel.org/pub/linux/kernel/v3.x/patch-3.19.xz
# cat /sys/kernel/tracing/events/net/netif_receive_skb/hist
# trigger info: hist:keys=common_stacktrace:vals=len:sort=hitcount:size=2048 [paused]
{ common_stacktrace:
__netif_receive_skb_core+0x46d/0x990
__netif_receive_skb+0x18/0x60
netif_receive_skb_internal+0x23/0x90
napi_gro_receive+0xc8/0x100
ieee80211_deliver_skb+0xd6/0x270 [mac80211]
ieee80211_rx_handlers+0xccf/0x22f0 [mac80211]
ieee80211_prepare_and_rx_handle+0x4e7/0xc40 [mac80211]
ieee80211_rx+0x31d/0x900 [mac80211]
iwlagn_rx_reply_rx+0x3db/0x6f0 [iwldvm]
iwl_rx_dispatch+0x8e/0xf0 [iwldvm]
iwl_pcie_irq_handler+0xe3c/0x12f0 [iwlwifi]
irq_thread_fn+0x20/0x50
irq_thread+0x11f/0x150
kthread+0xd2/0xf0
ret_from_fork+0x42/0x70
} hitcount: 85 len: 28884
{ common_stacktrace:
__netif_receive_skb_core+0x46d/0x990
__netif_receive_skb+0x18/0x60
netif_receive_skb_internal+0x23/0x90
napi_gro_complete+0xa4/0xe0
dev_gro_receive+0x23a/0x360
napi_gro_receive+0x30/0x100
ieee80211_deliver_skb+0xd6/0x270 [mac80211]
ieee80211_rx_handlers+0xccf/0x22f0 [mac80211]
ieee80211_prepare_and_rx_handle+0x4e7/0xc40 [mac80211]
ieee80211_rx+0x31d/0x900 [mac80211]
iwlagn_rx_reply_rx+0x3db/0x6f0 [iwldvm]
iwl_rx_dispatch+0x8e/0xf0 [iwldvm]
iwl_pcie_irq_handler+0xe3c/0x12f0 [iwlwifi]
irq_thread_fn+0x20/0x50
irq_thread+0x11f/0x150
kthread+0xd2/0xf0
} hitcount: 98 len: 664329
{ common_stacktrace:
__netif_receive_skb_core+0x46d/0x990
__netif_receive_skb+0x18/0x60
process_backlog+0xa8/0x150
net_rx_action+0x15d/0x340
__do_softirq+0x114/0x2c0
do_softirq_own_stack+0x1c/0x30
do_softirq+0x65/0x70
__local_bh_enable_ip+0xb5/0xc0
ip_finish_output+0x1f4/0x840
ip_output+0x6b/0xc0
ip_local_out_sk+0x31/0x40
ip_send_skb+0x1a/0x50
udp_send_skb+0x173/0x2a0
udp_sendmsg+0x2bf/0x9f0
inet_sendmsg+0x64/0xa0
sock_sendmsg+0x3d/0x50
} hitcount: 115 len: 13030
{ common_stacktrace:
__netif_receive_skb_core+0x46d/0x990
__netif_receive_skb+0x18/0x60
netif_receive_skb_internal+0x23/0x90
napi_gro_complete+0xa4/0xe0
napi_gro_flush+0x6d/0x90
iwl_pcie_irq_handler+0x92a/0x12f0 [iwlwifi]
irq_thread_fn+0x20/0x50
irq_thread+0x11f/0x150
kthread+0xd2/0xf0
ret_from_fork+0x42/0x70
} hitcount: 934 len: 5512212
Totals:
Hits: 1232
Entries: 4
Dropped: 0
The above shows all the netif_receive_skb callpaths and their total
lengths for the duration of the wget command.
Histogram 초기화와 event log 동기화
1139-1222`clear` hist trigger parameter는 hash table을 비운다. 앞 예제를 다시 실행하면서 histogram에 들어간 event 전체도 함께 보고 싶다면 모든 trigger를 다시 설정할 필요 없이 먼저 histogram만 초기화하면 된다.
# echo 'hist:key=common_stacktrace:vals=len:clear' >> \
/sys/kernel/tracing/events/net/netif_receive_skb/trigger실제로 비워졌는지 확인하면 `hist` 파일의 trigger 상태는 `[paused]`로 유지되지만 `Hits`, `Entries`, `Dropped`가 모두 0으로 표시된다.
# cat /sys/kernel/tracing/events/net/netif_receive_skb/hist
# trigger info: hist:keys=common_stacktrace:vals=len:sort=hitcount:size=2048 [paused]
Totals:
Hits: 0
Entries: 0
Dropped: 0새 실행 중 발생하는 모든 `netif_receive_skb` event의 상세 목록은 hash table에 집계되는 바로 그 event들이다. 이를 trace log에도 남기기 위해 trigger 역할을 하는 `sched_process_exec`와 `sched_process_exit` event에 `enable_event`와 `disable_event` trigger를 추가한다.
# echo 'enable_event:net:netif_receive_skb if filename==/usr/bin/wget' > \
/sys/kernel/tracing/events/sched/sched_process_exec/trigger
# echo 'disable_event:net:netif_receive_skb if comm==wget' > \
/sys/kernel/tracing/events/sched/sched_process_exit/trigger`sched_process_exec`와 `sched_process_exit`의 trigger 파일을 읽으면 각각 trigger가 두 개씩 보인다. 하나는 histogram 집계를 시작하거나 중지하고, 다른 하나는 event 기록을 시작하거나 중지한다.
# cat /sys/kernel/tracing/events/sched/sched_process_exec/trigger
enable_event:net:netif_receive_skb:unlimited if filename==/usr/bin/wget
enable_hist:net:netif_receive_skb:unlimited if filename==/usr/bin/wget
# cat /sys/kernel/tracing/events/sched/sched_process_exit/trigger
enable_event:net:netif_receive_skb:unlimited if comm==wget
disable_hist:net:netif_receive_skb:unlimited if comm==wget즉 두 sched event 중 하나가 적중해 `wget` 조건과 일치할 때 histogram과 event log가 함께 활성화되거나 비활성화된다. 그 결과 지정한 구간만 포함하는 hash table과 event 집합을 얻는다. 이제 `wget`을 다시 실행한다.
$ wget https://www.kernel.org/pub/linux/kernel/v3.x/patch-3.19.xz`hist` 파일은 앞 실행과 비슷한 집계를 보여 주며, 이번에는 trace 파일에서 개별 event도 확인할 수 있다.
# cat /sys/kernel/tracing/trace
# tracer: nop
#
# entries-in-buffer/entries-written: 183/1426 #P:4
#
# _-----=> irqs-off
# / _----=> need-resched
# | / _---=> hardirq/softirq
# || / _--=> preempt-depth
# ||| / delay
# TASK-PID CPU# |||| TIMESTAMP FUNCTION
# | | | |||| | |
wget-15108 [000] ..s1 31769.606929: netif_receive_skb: dev=lo skbaddr=ffff88009c353100 len=60
wget-15108 [000] ..s1 31769.606999: netif_receive_skb: dev=lo skbaddr=ffff88009c353200 len=60
dnsmasq-1382 [000] ..s1 31769.677652: netif_receive_skb: dev=lo skbaddr=ffff88009c352b00 len=130
dnsmasq-1382 [000] ..s1 31769.685917: netif_receive_skb: dev=lo skbaddr=ffff88009c352200 len=138
##### CPU 2 buffer started ####
irq/29-iwlwifi-559 [002] ..s. 31772.031529: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d433d00 len=2948
irq/29-iwlwifi-559 [002] ..s. 31772.031572: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d432200 len=1500
irq/29-iwlwifi-559 [002] ..s. 31772.032196: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d433100 len=2948
irq/29-iwlwifi-559 [002] ..s. 31772.032761: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d433000 len=2948
irq/29-iwlwifi-559 [002] ..s. 31772.033220: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d432e00 len=1500
.
.
.lifecycle eventhistogram actionevent-log action측정 경계
`sched_process_exec``enable_hist``enable_event`wget 시작
`sched_process_exit``disable_hist``disable_event`wget 종료
같은 process 조건을 사용하되 집계와 raw event 기록은 별도 trigger로 제어한다.
기존 hist trigger→[paused], 이전 data 보유
명령 + :clear→hash table 초기화
hist 확인→Hits 0, Entries 0
wget 재실행→histogram + raw event 동시 수집
hist / trace 읽기→요약과 개별 event 대조
기존 trigger 배치를 유지하면서 data만 비우고 다음 측정을 수행한다.
The 'clear' hist trigger param can be used to clear the hash table.
Suppose we wanted to try another run of the previous example but
this time also wanted to see the complete list of events that went
into the histogram. In order to avoid having to set everything up
again, we can just clear the histogram first::
# echo 'hist:key=common_stacktrace:vals=len:clear' >> \
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
Just to verify that it is in fact cleared, here's what we now see in
the hist file::
# cat /sys/kernel/tracing/events/net/netif_receive_skb/hist
# trigger info: hist:keys=common_stacktrace:vals=len:sort=hitcount:size=2048 [paused]
Totals:
Hits: 0
Entries: 0
Dropped: 0
Since we want to see the detailed list of every netif_receive_skb
event occurring during the new run, which are in fact the same
events being aggregated into the hash table, we add some additional
'enable_event' events to the triggering sched_process_exec and
sched_process_exit events as such::
# echo 'enable_event:net:netif_receive_skb if filename==/usr/bin/wget' > \
/sys/kernel/tracing/events/sched/sched_process_exec/trigger
# echo 'disable_event:net:netif_receive_skb if comm==wget' > \
/sys/kernel/tracing/events/sched/sched_process_exit/trigger
If you read the trigger files for the sched_process_exec and
sched_process_exit triggers, you should see two triggers for each:
one enabling/disabling the hist aggregation and the other
enabling/disabling the logging of events::
# cat /sys/kernel/tracing/events/sched/sched_process_exec/trigger
enable_event:net:netif_receive_skb:unlimited if filename==/usr/bin/wget
enable_hist:net:netif_receive_skb:unlimited if filename==/usr/bin/wget
# cat /sys/kernel/tracing/events/sched/sched_process_exit/trigger
enable_event:net:netif_receive_skb:unlimited if comm==wget
disable_hist:net:netif_receive_skb:unlimited if comm==wget
In other words, whenever either of the sched_process_exec or
sched_process_exit events is hit and matches 'wget', it enables or
disables both the histogram and the event log, and what you end up
with is a hash table and set of events just covering the specified
duration. Run the wget command again::
$ wget https://www.kernel.org/pub/linux/kernel/v3.x/patch-3.19.xz
Displaying the 'hist' file should show something similar to what you
saw in the last run, but this time you should also see the
individual events in the trace file::
# cat /sys/kernel/tracing/trace
# tracer: nop
#
# entries-in-buffer/entries-written: 183/1426 #P:4
#
# _-----=> irqs-off
# / _----=> need-resched
# | / _---=> hardirq/softirq
# || / _--=> preempt-depth
# ||| / delay
# TASK-PID CPU# |||| TIMESTAMP FUNCTION
# | | | |||| | |
wget-15108 [000] ..s1 31769.606929: netif_receive_skb: dev=lo skbaddr=ffff88009c353100 len=60
wget-15108 [000] ..s1 31769.606999: netif_receive_skb: dev=lo skbaddr=ffff88009c353200 len=60
dnsmasq-1382 [000] ..s1 31769.677652: netif_receive_skb: dev=lo skbaddr=ffff88009c352b00 len=130
dnsmasq-1382 [000] ..s1 31769.685917: netif_receive_skb: dev=lo skbaddr=ffff88009c352200 len=138
##### CPU 2 buffer started ####
irq/29-iwlwifi-559 [002] ..s. 31772.031529: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d433d00 len=2948
irq/29-iwlwifi-559 [002] ..s. 31772.031572: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d432200 len=1500
irq/29-iwlwifi-559 [002] ..s. 31772.032196: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d433100 len=2948
irq/29-iwlwifi-559 [002] ..s. 31772.032761: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d433000 len=2948
irq/29-iwlwifi-559 [002] ..s. 31772.033220: netif_receive_skb: dev=wlan0 skbaddr=ffff88009d432e00 len=1500
.
.
.
한 event에 여러 hist trigger 연결
1223-1362다음 예제는 한 event에 여러 hist trigger를 연결하는 방법을 보여 준다. 같은 event 집합에서 서로 다른 요약을 만들거나 서로 다른 filter의 효과를 비교할 때 유용하다.
# echo 'hist:keys=skbaddr.hex:vals=len if len < 0' >> \
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
# echo 'hist:keys=skbaddr.hex:vals=len if len > 4096' >> \
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
# echo 'hist:keys=skbaddr.hex:vals=len if len == 256' >> \
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
# echo 'hist:keys=skbaddr.hex:vals=len' >> \
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
# echo 'hist:keys=len:vals=common_preempt_count' >> \
/sys/kernel/tracing/events/net/netif_receive_skb/trigger이 명령들은 filter만 다른 trigger 네 개와, 별개의 다소 비현실적인 trigger 하나를 만든다. 같은 파일에 여러 hist trigger를 추가하려면 append 연산자 `>>`를 사용해야 한다. `>`도 새 trigger를 추가하지만 그 전에 기존 hist trigger를 모두 제거한다.
해당 event의 `hist` 파일을 읽으면 다섯 histogram의 내용이 차례로 표시된다.
# cat /sys/kernel/tracing/events/net/netif_receive_skb/hist
# event histogram
#
# trigger info: hist:keys=len:vals=hitcount,common_preempt_count:sort=hitcount:size=2048 [active]
#
{ len: 176 } hitcount: 1 common_preempt_count: 0
{ len: 223 } hitcount: 1 common_preempt_count: 0
{ len: 4854 } hitcount: 1 common_preempt_count: 0
{ len: 395 } hitcount: 1 common_preempt_count: 0
{ len: 177 } hitcount: 1 common_preempt_count: 0
{ len: 446 } hitcount: 1 common_preempt_count: 0
{ len: 1601 } hitcount: 1 common_preempt_count: 0
.
.
.
{ len: 1280 } hitcount: 66 common_preempt_count: 0
{ len: 116 } hitcount: 81 common_preempt_count: 40
{ len: 708 } hitcount: 112 common_preempt_count: 0
{ len: 46 } hitcount: 221 common_preempt_count: 0
{ len: 1264 } hitcount: 458 common_preempt_count: 0
Totals:
Hits: 1428
Entries: 147
Dropped: 0
# event histogram
#
# trigger info: hist:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 [active]
#
{ skbaddr: ffff8800baee5e00 } hitcount: 1 len: 130
{ skbaddr: ffff88005f3d5600 } hitcount: 1 len: 1280
{ skbaddr: ffff88005f3d4900 } hitcount: 1 len: 1280
{ skbaddr: ffff88009fed6300 } hitcount: 1 len: 115
{ skbaddr: ffff88009fe0ad00 } hitcount: 1 len: 115
{ skbaddr: ffff88008cdb1900 } hitcount: 1 len: 46
{ skbaddr: ffff880064b5ef00 } hitcount: 1 len: 118
{ skbaddr: ffff880044e3c700 } hitcount: 1 len: 60
{ skbaddr: ffff880100065900 } hitcount: 1 len: 46
{ skbaddr: ffff8800d46bd500 } hitcount: 1 len: 116
{ skbaddr: ffff88005f3d5f00 } hitcount: 1 len: 1280
{ skbaddr: ffff880100064700 } hitcount: 1 len: 365
{ skbaddr: ffff8800badb6f00 } hitcount: 1 len: 60
.
.
.
{ skbaddr: ffff88009fe0be00 } hitcount: 27 len: 24677
{ skbaddr: ffff88009fe0a400 } hitcount: 27 len: 23052
{ skbaddr: ffff88009fe0b700 } hitcount: 31 len: 25589
{ skbaddr: ffff88009fe0b600 } hitcount: 32 len: 27326
{ skbaddr: ffff88006a462800 } hitcount: 68 len: 71678
{ skbaddr: ffff88006a463700 } hitcount: 70 len: 72678
{ skbaddr: ffff88006a462b00 } hitcount: 71 len: 77589
{ skbaddr: ffff88006a463600 } hitcount: 73 len: 71307
{ skbaddr: ffff88006a462200 } hitcount: 81 len: 81032
Totals:
Hits: 1451
Entries: 318
Dropped: 0
# event histogram
#
# trigger info: hist:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 if len == 256 [active]
#
Totals:
Hits: 0
Entries: 0
Dropped: 0
# event histogram
#
# trigger info: hist:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 if len > 4096 [active]
#
{ skbaddr: ffff88009fd2c300 } hitcount: 1 len: 7212
{ skbaddr: ffff8800d2bcce00 } hitcount: 1 len: 7212
{ skbaddr: ffff8800d2bcd700 } hitcount: 1 len: 7212
{ skbaddr: ffff8800d2bcda00 } hitcount: 1 len: 21492
{ skbaddr: ffff8800ae2e2d00 } hitcount: 1 len: 7212
{ skbaddr: ffff8800d2bcdb00 } hitcount: 1 len: 7212
{ skbaddr: ffff88006a4df500 } hitcount: 1 len: 4854
{ skbaddr: ffff88008ce47b00 } hitcount: 1 len: 18636
{ skbaddr: ffff8800ae2e2200 } hitcount: 1 len: 12924
{ skbaddr: ffff88005f3e1000 } hitcount: 1 len: 4356
{ skbaddr: ffff8800d2bcdc00 } hitcount: 2 len: 24420
{ skbaddr: ffff8800d2bcc200 } hitcount: 2 len: 12996
Totals:
Hits: 14
Entries: 12
Dropped: 0
# event histogram
#
# trigger info: hist:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 if len < 0 [active]
#
Totals:
Hits: 0
Entries: 0
Dropped: 0trigger조건예제 결과
`keys=len:vals=common_preempt_count`없음길이별 preempt count 합계
`keys=skbaddr.hex:vals=len`없음모든 skb address 집계
같은 skb trigger`len == 256`일치 event가 없어 Hits 0
같은 skb trigger`len > 4096`큰 packet만 집계
같은 skb trigger`len < 0`일치 event가 없어 Hits 0
동일한 netif_receive_skb stream을 각 trigger가 독립적인 hash table과 조건으로 집계한다.
The following example demonstrates how multiple hist triggers can be
attached to a given event. This capability can be useful for
creating a set of different summaries derived from the same set of
events, or for comparing the effects of different filters, among
other things::
# echo 'hist:keys=skbaddr.hex:vals=len if len < 0' >> \
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
# echo 'hist:keys=skbaddr.hex:vals=len if len > 4096' >> \
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
# echo 'hist:keys=skbaddr.hex:vals=len if len == 256' >> \
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
# echo 'hist:keys=skbaddr.hex:vals=len' >> \
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
# echo 'hist:keys=len:vals=common_preempt_count' >> \
/sys/kernel/tracing/events/net/netif_receive_skb/trigger
The above set of commands create four triggers differing only in
their filters, along with a completely different though fairly
nonsensical trigger. Note that in order to append multiple hist
triggers to the same file, you should use the '>>' operator to
append them ('>' will also add the new hist trigger, but will remove
any existing hist triggers beforehand).
Displaying the contents of the 'hist' file for the event shows the
contents of all five histograms::
# cat /sys/kernel/tracing/events/net/netif_receive_skb/hist
# event histogram
#
# trigger info: hist:keys=len:vals=hitcount,common_preempt_count:sort=hitcount:size=2048 [active]
#
{ len: 176 } hitcount: 1 common_preempt_count: 0
{ len: 223 } hitcount: 1 common_preempt_count: 0
{ len: 4854 } hitcount: 1 common_preempt_count: 0
{ len: 395 } hitcount: 1 common_preempt_count: 0
{ len: 177 } hitcount: 1 common_preempt_count: 0
{ len: 446 } hitcount: 1 common_preempt_count: 0
{ len: 1601 } hitcount: 1 common_preempt_count: 0
.
.
.
{ len: 1280 } hitcount: 66 common_preempt_count: 0
{ len: 116 } hitcount: 81 common_preempt_count: 40
{ len: 708 } hitcount: 112 common_preempt_count: 0
{ len: 46 } hitcount: 221 common_preempt_count: 0
{ len: 1264 } hitcount: 458 common_preempt_count: 0
Totals:
Hits: 1428
Entries: 147
Dropped: 0
# event histogram
#
# trigger info: hist:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 [active]
#
{ skbaddr: ffff8800baee5e00 } hitcount: 1 len: 130
{ skbaddr: ffff88005f3d5600 } hitcount: 1 len: 1280
{ skbaddr: ffff88005f3d4900 } hitcount: 1 len: 1280
{ skbaddr: ffff88009fed6300 } hitcount: 1 len: 115
{ skbaddr: ffff88009fe0ad00 } hitcount: 1 len: 115
{ skbaddr: ffff88008cdb1900 } hitcount: 1 len: 46
{ skbaddr: ffff880064b5ef00 } hitcount: 1 len: 118
{ skbaddr: ffff880044e3c700 } hitcount: 1 len: 60
{ skbaddr: ffff880100065900 } hitcount: 1 len: 46
{ skbaddr: ffff8800d46bd500 } hitcount: 1 len: 116
{ skbaddr: ffff88005f3d5f00 } hitcount: 1 len: 1280
{ skbaddr: ffff880100064700 } hitcount: 1 len: 365
{ skbaddr: ffff8800badb6f00 } hitcount: 1 len: 60
.
.
.
{ skbaddr: ffff88009fe0be00 } hitcount: 27 len: 24677
{ skbaddr: ffff88009fe0a400 } hitcount: 27 len: 23052
{ skbaddr: ffff88009fe0b700 } hitcount: 31 len: 25589
{ skbaddr: ffff88009fe0b600 } hitcount: 32 len: 27326
{ skbaddr: ffff88006a462800 } hitcount: 68 len: 71678
{ skbaddr: ffff88006a463700 } hitcount: 70 len: 72678
{ skbaddr: ffff88006a462b00 } hitcount: 71 len: 77589
{ skbaddr: ffff88006a463600 } hitcount: 73 len: 71307
{ skbaddr: ffff88006a462200 } hitcount: 81 len: 81032
Totals:
Hits: 1451
Entries: 318
Dropped: 0
# event histogram
#
# trigger info: hist:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 if len == 256 [active]
#
Totals:
Hits: 0
Entries: 0
Dropped: 0
# event histogram
#
# trigger info: hist:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 if len > 4096 [active]
#
{ skbaddr: ffff88009fd2c300 } hitcount: 1 len: 7212
{ skbaddr: ffff8800d2bcce00 } hitcount: 1 len: 7212
{ skbaddr: ffff8800d2bcd700 } hitcount: 1 len: 7212
{ skbaddr: ffff8800d2bcda00 } hitcount: 1 len: 21492
{ skbaddr: ffff8800ae2e2d00 } hitcount: 1 len: 7212
{ skbaddr: ffff8800d2bcdb00 } hitcount: 1 len: 7212
{ skbaddr: ffff88006a4df500 } hitcount: 1 len: 4854
{ skbaddr: ffff88008ce47b00 } hitcount: 1 len: 18636
{ skbaddr: ffff8800ae2e2200 } hitcount: 1 len: 12924
{ skbaddr: ffff88005f3e1000 } hitcount: 1 len: 4356
{ skbaddr: ffff8800d2bcdc00 } hitcount: 2 len: 24420
{ skbaddr: ffff8800d2bcc200 } hitcount: 2 len: 12996
Totals:
Hits: 14
Entries: 12
Dropped: 0
# event histogram
#
# trigger info: hist:keys=skbaddr.hex:vals=hitcount,len:sort=hitcount:size=2048 if len < 0 [active]
#
Totals:
Hits: 0
Entries: 0
Dropped: 0
공통 호환 field로 event 결합
1486-1610두 event 사이에 `hitcount`와 `common_stacktrace` 외에는 호환되는 field가 없어도 두 field를 이용해 histogram data를 결합할 수 있다. 다음 명령은 이 field들로 `bar`라는 trigger 두 개를 만든다.
# echo 'hist:name=bar:key=common_stacktrace:val=hitcount' > \
/sys/kernel/tracing/events/sched/sched_process_fork/trigger
# echo 'hist:name=bar:key=common_stacktrace:val=hitcount' > \
/sys/kernel/tracing/events/net/netif_rx/trigger어느 event의 출력을 읽어도 두 event에서 온 call path가 섞인, 흥미롭지만 다소 해석하기 어려운 공통 결과가 나타난다.
# cat /sys/kernel/tracing/events/sched/sched_process_fork/hist
# cat /sys/kernel/tracing/events/net/netif_rx/hist
# event histogram
#
# trigger info: hist:name=bar:keys=common_stacktrace:vals=hitcount:sort=hitcount:size=2048 [active]
#
{ common_stacktrace:
kernel_clone+0x18e/0x330
kernel_thread+0x29/0x30
kthreadd+0x154/0x1b0
ret_from_fork+0x3f/0x70
} hitcount: 1
{ common_stacktrace:
netif_rx_internal+0xb2/0xd0
netif_rx_ni+0x20/0x70
dev_loopback_xmit+0xaa/0xd0
ip_mc_output+0x126/0x240
ip_local_out_sk+0x31/0x40
igmp_send_report+0x1e9/0x230
igmp_timer_expire+0xe9/0x120
call_timer_fn+0x39/0xf0
run_timer_softirq+0x1e1/0x290
__do_softirq+0xfd/0x290
irq_exit+0x98/0xb0
smp_apic_timer_interrupt+0x4a/0x60
apic_timer_interrupt+0x6d/0x80
cpuidle_enter+0x17/0x20
call_cpuidle+0x3b/0x60
cpu_startup_entry+0x22d/0x310
} hitcount: 1
{ common_stacktrace:
netif_rx_internal+0xb2/0xd0
netif_rx_ni+0x20/0x70
dev_loopback_xmit+0xaa/0xd0
ip_mc_output+0x17f/0x240
ip_local_out_sk+0x31/0x40
ip_send_skb+0x1a/0x50
udp_send_skb+0x13e/0x270
udp_sendmsg+0x2bf/0x980
inet_sendmsg+0x67/0xa0
sock_sendmsg+0x38/0x50
SYSC_sendto+0xef/0x170
SyS_sendto+0xe/0x10
entry_SYSCALL_64_fastpath+0x12/0x6a
} hitcount: 2
{ common_stacktrace:
netif_rx_internal+0xb2/0xd0
netif_rx+0x1c/0x60
loopback_xmit+0x6c/0xb0
dev_hard_start_xmit+0x219/0x3a0
__dev_queue_xmit+0x415/0x4f0
dev_queue_xmit_sk+0x13/0x20
ip_finish_output2+0x237/0x340
ip_finish_output+0x113/0x1d0
ip_output+0x66/0xc0
ip_local_out_sk+0x31/0x40
ip_send_skb+0x1a/0x50
udp_send_skb+0x16d/0x270
udp_sendmsg+0x2bf/0x980
inet_sendmsg+0x67/0xa0
sock_sendmsg+0x38/0x50
___sys_sendmsg+0x14e/0x270
} hitcount: 76
{ common_stacktrace:
netif_rx_internal+0xb2/0xd0
netif_rx+0x1c/0x60
loopback_xmit+0x6c/0xb0
dev_hard_start_xmit+0x219/0x3a0
__dev_queue_xmit+0x415/0x4f0
dev_queue_xmit_sk+0x13/0x20
ip_finish_output2+0x237/0x340
ip_finish_output+0x113/0x1d0
ip_output+0x66/0xc0
ip_local_out_sk+0x31/0x40
ip_send_skb+0x1a/0x50
udp_send_skb+0x16d/0x270
udp_sendmsg+0x2bf/0x980
inet_sendmsg+0x67/0xa0
sock_sendmsg+0x38/0x50
___sys_sendmsg+0x269/0x270
} hitcount: 77
{ common_stacktrace:
netif_rx_internal+0xb2/0xd0
netif_rx+0x1c/0x60
loopback_xmit+0x6c/0xb0
dev_hard_start_xmit+0x219/0x3a0
__dev_queue_xmit+0x415/0x4f0
dev_queue_xmit_sk+0x13/0x20
ip_finish_output2+0x237/0x340
ip_finish_output+0x113/0x1d0
ip_output+0x66/0xc0
ip_local_out_sk+0x31/0x40
ip_send_skb+0x1a/0x50
udp_send_skb+0x16d/0x270
udp_sendmsg+0x2bf/0x980
inet_sendmsg+0x67/0xa0
sock_sendmsg+0x38/0x50
SYSC_sendto+0xef/0x170
} hitcount: 88
{ common_stacktrace:
kernel_clone+0x18e/0x330
SyS_clone+0x19/0x20
entry_SYSCALL_64_fastpath+0x12/0x6a
} hitcount: 244
Totals:
Hits: 489
Entries: 7
Dropped: 0공유 구성가능 여부예제
같은 key/value 이름과 type가능`foo`: `skbaddr.hex`, `len`
모든 event의 암시적 적중 수가능`hitcount`
모든 event에서 얻는 kernel 호출 경로가능`common_stacktrace`
event별로만 존재하는 비호환 field불가공유 trigger 구조에서 제외
공유 histogram은 참여 trigger가 같은 구조로 해석할 수 있는 field만 결합한다.
And here's an example that shows how to combine histogram data from
any two events even if they don't share any 'compatible' fields
other than 'hitcount' and 'common_stacktrace'. These commands create a
couple of triggers named 'bar' using those fields::
# echo 'hist:name=bar:key=common_stacktrace:val=hitcount' > \
/sys/kernel/tracing/events/sched/sched_process_fork/trigger
# echo 'hist:name=bar:key=common_stacktrace:val=hitcount' > \
/sys/kernel/tracing/events/net/netif_rx/trigger
And displaying the output of either shows some interesting if
somewhat confusing output::
# cat /sys/kernel/tracing/events/sched/sched_process_fork/hist
# cat /sys/kernel/tracing/events/net/netif_rx/hist
# event histogram
#
# trigger info: hist:name=bar:keys=common_stacktrace:vals=hitcount:sort=hitcount:size=2048 [active]
#
{ common_stacktrace:
kernel_clone+0x18e/0x330
kernel_thread+0x29/0x30
kthreadd+0x154/0x1b0
ret_from_fork+0x3f/0x70
} hitcount: 1
{ common_stacktrace:
netif_rx_internal+0xb2/0xd0
netif_rx_ni+0x20/0x70
dev_loopback_xmit+0xaa/0xd0
ip_mc_output+0x126/0x240
ip_local_out_sk+0x31/0x40
igmp_send_report+0x1e9/0x230
igmp_timer_expire+0xe9/0x120
call_timer_fn+0x39/0xf0
run_timer_softirq+0x1e1/0x290
__do_softirq+0xfd/0x290
irq_exit+0x98/0xb0
smp_apic_timer_interrupt+0x4a/0x60
apic_timer_interrupt+0x6d/0x80
cpuidle_enter+0x17/0x20
call_cpuidle+0x3b/0x60
cpu_startup_entry+0x22d/0x310
} hitcount: 1
{ common_stacktrace:
netif_rx_internal+0xb2/0xd0
netif_rx_ni+0x20/0x70
dev_loopback_xmit+0xaa/0xd0
ip_mc_output+0x17f/0x240
ip_local_out_sk+0x31/0x40
ip_send_skb+0x1a/0x50
udp_send_skb+0x13e/0x270
udp_sendmsg+0x2bf/0x980
inet_sendmsg+0x67/0xa0
sock_sendmsg+0x38/0x50
SYSC_sendto+0xef/0x170
SyS_sendto+0xe/0x10
entry_SYSCALL_64_fastpath+0x12/0x6a
} hitcount: 2
{ common_stacktrace:
netif_rx_internal+0xb2/0xd0
netif_rx+0x1c/0x60
loopback_xmit+0x6c/0xb0
dev_hard_start_xmit+0x219/0x3a0
__dev_queue_xmit+0x415/0x4f0
dev_queue_xmit_sk+0x13/0x20
ip_finish_output2+0x237/0x340
ip_finish_output+0x113/0x1d0
ip_output+0x66/0xc0
ip_local_out_sk+0x31/0x40
ip_send_skb+0x1a/0x50
udp_send_skb+0x16d/0x270
udp_sendmsg+0x2bf/0x980
inet_sendmsg+0x67/0xa0
sock_sendmsg+0x38/0x50
___sys_sendmsg+0x14e/0x270
} hitcount: 76
{ common_stacktrace:
netif_rx_internal+0xb2/0xd0
netif_rx+0x1c/0x60
loopback_xmit+0x6c/0xb0
dev_hard_start_xmit+0x219/0x3a0
__dev_queue_xmit+0x415/0x4f0
dev_queue_xmit_sk+0x13/0x20
ip_finish_output2+0x237/0x340
ip_finish_output+0x113/0x1d0
ip_output+0x66/0xc0
ip_local_out_sk+0x31/0x40
ip_send_skb+0x1a/0x50
udp_send_skb+0x16d/0x270
udp_sendmsg+0x2bf/0x980
inet_sendmsg+0x67/0xa0
sock_sendmsg+0x38/0x50
___sys_sendmsg+0x269/0x270
} hitcount: 77
{ common_stacktrace:
netif_rx_internal+0xb2/0xd0
netif_rx+0x1c/0x60
loopback_xmit+0x6c/0xb0
dev_hard_start_xmit+0x219/0x3a0
__dev_queue_xmit+0x415/0x4f0
dev_queue_xmit_sk+0x13/0x20
ip_finish_output2+0x237/0x340
ip_finish_output+0x113/0x1d0
ip_output+0x66/0xc0
ip_local_out_sk+0x31/0x40
ip_send_skb+0x1a/0x50
udp_send_skb+0x16d/0x270
udp_sendmsg+0x2bf/0x980
inet_sendmsg+0x67/0xa0
sock_sendmsg+0x38/0x50
SYSC_sendto+0xef/0x170
} hitcount: 88
{ common_stacktrace:
kernel_clone+0x18e/0x330
SyS_clone+0x19/0x20
entry_SYSCALL_64_fastpath+0x12/0x6a
} hitcount: 244
Totals:
Hits: 489
Entries: 7
Dropped: 0
Inter-event hist trigger
1611-1687inter-event hist trigger는 하나 이상의 다른 event에서 값을 결합하고 그 data로 histogram을 만든다. 이렇게 만든 inter-event histogram의 data를 다시 다음 결합 histogram의 source로 삼을 수 있으므로, 일부 application에 중요한 연관 histogram chain을 구성할 수 있다.
이 방식으로 사용할 수 있는 가장 중요한 inter-event quantity는 두 event timestamp의 차이인 latency다. 그러나 trace event subsystem 전체에 일반적으로 적용되는 기능이므로 latency뿐 아니라 어떤 event field도 inter-event quantity에 사용할 수 있다.
다른 histogram의 data를 유용한 chain으로 결합하는 예로는 `wakeup latency` histogram과 `switch latency` histogram을 합친 `wakeupswitch latency` histogram이 있다.
일반 hist trigger specification은 하나의 event type에 속한 trace event field를 참조하는 단일 또는 compound key와, 그 key에 연결되어 계속 합산되는 하나 이상의 numeric value로 구성된다.
inter-event hist trigger 확장은 여러 event의 field를 참조해 하나의 multi-event histogram specification으로 결합한다. 이를 가능하게 하기 위해 hist trigger에 다음 기능이 추가되었다.
기능역할
histogram variable한 event의 값을 저장한 뒤 다른 matching event에서 참조
simple expressionvariable에 `+`, `-` 등의 연산을 적용해 inter-event quantity 계산
synthetic event어느 한 source event에도 속하지 않는 결합 결과를 독립적인 정식 event로 표현
histogram actionsynthetic event 생성 또는 최대 latency 시점의 context 저장
`common_timestamp`ftrace ring buffer의 암시적 timestamp를 synthetic field로 노출
여러 event의 값을 저장·계산·결합하고 결과를 event로 다시 노출하기 위한 구성 요소다.
`common_timestamp`는 trace format의 실제 field가 아니지만 일반 event field처럼 histogram에서 사용할 수 있다. 기본 단위는 nanosecond이며 `.usecs`를 붙이면 microsecond로 바뀐다.
inter-event timestamp에 `common_timestamp`를 사용하면 CPU 사이에서 일관되지 않은 clock으로 인한 잘못된 차이를 피하기 위해 trace buffer가 absolute timestamp와 `global` trace clock을 자동 사용한다. 필요하면 `clock=XXX` hist trigger attribute로 `tracing/trace_clock` pseudo-file에 나열된 다른 clock을 지정해 재정의할 수 있다.
이 기반 기능들은 다음 절에서 더 자세히 설명한다.
2.4. Inter-event hist triggers
------------------------------
Inter-event hist triggers are hist triggers that combine values from
one or more other events and create a histogram using that data. Data
from an inter-event histogram can in turn become the source for
further combined histograms, thus providing a chain of related
histograms, which is important for some applications.
The most important example of an inter-event quantity that can be used
in this manner is latency, which is simply a difference in timestamps
between two events. Although latency is the most important
inter-event quantity, note that because the support is completely
general across the trace event subsystem, any event field can be used
in an inter-event quantity.
An example of a histogram that combines data from other histograms
into a useful chain would be a 'wakeupswitch latency' histogram that
combines a 'wakeup latency' histogram and a 'switch latency'
histogram.
Normally, a hist trigger specification consists of a (possibly
compound) key along with one or more numeric values, which are
continually updated sums associated with that key. A histogram
specification in this case consists of individual key and value
specifications that refer to trace event fields associated with a
single event type.
The inter-event hist trigger extension allows fields from multiple
events to be referenced and combined into a multi-event histogram
specification. In support of this overall goal, a few enabling
features have been added to the hist trigger support:
- In order to compute an inter-event quantity, a value from one
event needs to saved and then referenced from another event. This
requires the introduction of support for histogram 'variables'.
- The computation of inter-event quantities and their combination
require some minimal amount of support for applying simple
expressions to variables (+ and -).
- A histogram consisting of inter-event quantities isn't logically a
histogram on either event (so having the 'hist' file for either
event host the histogram output doesn't really make sense). To
address the idea that the histogram is associated with a
combination of events, support is added allowing the creation of
'synthetic' events that are events derived from other events.
These synthetic events are full-fledged events just like any other
and can be used as such, as for instance to create the
'combination' histograms mentioned previously.
- A set of 'actions' can be associated with histogram entries -
these can be used to generate the previously mentioned synthetic
events, but can also be used for other purposes, such as for
example saving context when a 'max' latency has been hit.
- Trace events don't have a 'timestamp' associated with them, but
there is an implicit timestamp saved along with an event in the
underlying ftrace ring buffer. This timestamp is now exposed as a
a synthetic field named 'common_timestamp' which can be used in
histograms as if it were any other event field; it isn't an actual
field in the trace format but rather is a synthesized value that
nonetheless can be used as if it were an actual field. By default
it is in units of nanoseconds; appending '.usecs' to a
common_timestamp field changes the units to microseconds.
A note on inter-event timestamps: If common_timestamp is used in a
histogram, the trace buffer is automatically switched over to using
absolute timestamps and the "global" trace clock, in order to avoid
bogus timestamp differences with other clocks that aren't coherent
across CPUs. This can be overridden by specifying one of the other
trace clocks instead, using the "clock=XXX" hist trigger attribute,
where XXX is any of the clocks listed in the tracing/trace_clock
pseudo-file.
These features are described in more detail in the following sections.
Histogram variable
1688-1791variable은 matching event 사이에서 값을 저장하고 가져오는 이름 있는 위치다. matching event는 key가 일치하는 event를 뜻한다. 어떤 key에 해당하는 histogram entry에 variable을 저장하면 이후 같은 key를 가진 event가 그 variable에 접근할 수 있다.
variable 값은 보통 뒤따르는 event가 다른 값으로 덮어쓸 때까지 사용할 수 있다. 예외는 expression에서 사용한 variable로, 사실상 `read-once`다. 뒤 event의 expression에서 한 번 사용되면 `unset` 상태로 재설정되어 다시 값을 저장하기 전에는 사용할 수 없다. 이 규칙은 초기화되지 않은 variable을 계산에 쓰는 일을 막고, 관련 없는 후속 match에서 같은 값을 재사용하는 일도 방지한다.
variable을 저장하는 기본 문법은 keyword와 겹치지 않는 고유 variable 이름과 `=`를 event field 앞에 붙이는 것이다.
key와 value 모두 이 방식으로 저장하고 검색할 수 있다. 다음 명령은 `next_pid` key를 가진 histogram entry에 `ts0` variable을 만든다.
# echo 'hist:keys=next_pid:vals=$ts0:ts0=common_timestamp ... >> \
event/trigger`ts0` variable은 이후 `next_pid`와 같은 pid를 가진 어떤 event에서도 접근할 수 있다.
variable reference는 이름 앞에 `$`를 붙여 만든다. 따라서 위의 `ts0`는 expression에서 `$ts0`로 참조한다.
위 명령은 `vals=`를 사용했으므로 `common_timestamp` variable 값이 일반 histogram value처럼 합산되기도 한다. timestamp를 합산하는 것은 실용적인 의미가 거의 없다.
key 값도 같은 방식으로 저장할 수 있다.
# echo 'hist:timer_pid=common_pid:key=timer_pid ...' >> event/triggervariable이 key variable이 아니고 `vals=` prefix에도 포함되지 않았다면 연결된 event field는 variable에 저장되지만 value로 합산되지는 않는다.
# echo 'hist:keys=next_pid:ts1=common_timestamp ...' >> event/trigger여러 variable을 동시에 할당할 수 있다. 다음 명령은 `ts0`와 `b`를 만들고 `common_timestamp`와 `field1`을 value로도 합산한다.
# echo 'hist:keys=pid:vals=$ts0,$b:ts0=common_timestamp,b=field1 ...' >> \
event/triggervariable assignment는 사용 위치의 앞이나 뒤 어느 쪽에 있어도 된다. 다음 명령은 앞 명령과 동일하게 동작한다.
# echo 'hist:keys=pid:ts0=common_timestamp,b=field1:vals=$ts0,$b ...' >> \
event/trigger`vals=`에 묶이지 않은 variable은 colon으로 구분해 필요한 만큼 할당할 수 있다. 다음 명령은 같은 field를 저장하지만 histogram value로 합산하지 않는다.
# echo 'hist:keys=pid:ts0=common_timestamp:b=field1 ...' >> event/trigger형식variable 저장histogram 합산
`vals=$ts0:ts0=common_timestamp`예예
`timer_pid=common_pid:key=timer_pid`예, key로 사용key는 value 합산 아님
`keys=next_pid:ts1=common_timestamp`예아니요
`ts0=...:b=...:vals=$ts0,$b`예예, assignment 순서 무관
`ts0=...:b=...`예아니요
할당 위치와 `vals=` 포함 여부가 저장과 합산 동작을 구분한다.
이렇게 설정한 variable은 다른 event에서 참조해 expression에 사용할 수 있다.
예를 들어 다음과 같이 latency를 계산한다.
# echo 'hist:keys=pid,prio:ts0=common_timestamp ...' >> event1/trigger
# echo 'hist:keys=next_pid:wakeup_lat=common_timestamp-$ts0 ...' >> event2/trigger첫 명령은 event timestamp를 `ts0`에 저장한다. 다음 명령은 두 번째 event timestamp에서 `$ts0`를 빼 latency를 만들고 이를 `wakeup_lat` variable에 할당한다. 이어지는 hist trigger는 같은 key와 또 다른 event의 variable을 이용해 결합 latency를 계산한다.
# echo 'hist:key=pid:wakeupswitch_lat=$wakeup_lat+$switchtime_lat ...' >> event3/triggerevent1→ts0 = common_timestamp
matching pid→event2
wakeup_lat = common_timestamp - $ts0→ts0는 read-once 후 unset
event3→같은 pid key
wakeupswitch_lat = $wakeup_lat + $switchtime_lat→결합 latency
matching key를 따라 timestamp 차이와 다음 단계 latency를 순차적으로 전달한다.
expression은 덧셈, 뺄셈, 곱셈, 나눗셈 연산자 `+`, `-`, `*`, `/`를 지원한다.
0으로 나누는 상황을 parse 시점에 알 수 없다면, 즉 divisor가 constant가 아니라면 결과는 `-1`이다.
numeric constant도 expression에서 직접 사용할 수 있다.
# echo 'hist:keys=next_pid:timestamp_secs=common_timestamp/1000000 ...' >> event/triggerconstant를 variable에 할당하고 다음 expression에서 참조할 수도 있다.
# echo 'hist:keys=next_pid:us_per_sec=1000000 ...' >> event/trigger
# echo 'hist:keys=next_pid:timestamp_secs=common_timestamp/$us_per_sec ...' >> event/triggervariable에는 stacktrace도 저장할 수 있으며 synthetic event와 함께 사용할 때 유용하다.
matching event→variable 저장
후속 matching key→variable 참조 가능
일반 참조→덮어쓸 때까지 유지
expression 참조→한 번 사용
unset→다시 저장될 때까지 사용 불가
저장된 값은 matching key를 통해 전달되고 expression 사용 여부에 따라 유지 또는 해제된다.
2.5. Histogram Variables
------------------------
Variables are simply named locations used for saving and retrieving
values between matching events. A 'matching' event is defined as an
event that has a matching key - if a variable is saved for a histogram
entry corresponding to that key, any subsequent event with a matching
key can access that variable.
A variable's value is normally available to any subsequent event until
it is set to something else by a subsequent event. The one exception
to that rule is that any variable used in an expression is essentially
'read-once' - once it's used by an expression in a subsequent event,
it's reset to its 'unset' state, which means it can't be used again
unless it's set again. This ensures not only that an event doesn't
use an uninitialized variable in a calculation, but that that variable
is used only once and not for any unrelated subsequent match.
The basic syntax for saving a variable is to simply prefix a unique
variable name not corresponding to any keyword along with an '=' sign
to any event field.
Either keys or values can be saved and retrieved in this way. This
creates a variable named 'ts0' for a histogram entry with the key
'next_pid'::
# echo 'hist:keys=next_pid:vals=$ts0:ts0=common_timestamp ... >> \
event/trigger
The ts0 variable can be accessed by any subsequent event having the
same pid as 'next_pid'.
Variable references are formed by prepending the variable name with
the '$' sign. Thus for example, the ts0 variable above would be
referenced as '$ts0' in expressions.
Because 'vals=' is used, the common_timestamp variable value above
will also be summed as a normal histogram value would (though for a
timestamp it makes little sense).
The below shows that a key value can also be saved in the same way::
# echo 'hist:timer_pid=common_pid:key=timer_pid ...' >> event/trigger
If a variable isn't a key variable or prefixed with 'vals=', the
associated event field will be saved in a variable but won't be summed
as a value::
# echo 'hist:keys=next_pid:ts1=common_timestamp ...' >> event/trigger
Multiple variables can be assigned at the same time. The below would
result in both ts0 and b being created as variables, with both
common_timestamp and field1 additionally being summed as values::
# echo 'hist:keys=pid:vals=$ts0,$b:ts0=common_timestamp,b=field1 ...' >> \
event/trigger
Note that variable assignments can appear either preceding or
following their use. The command below behaves identically to the
command above::
# echo 'hist:keys=pid:ts0=common_timestamp,b=field1:vals=$ts0,$b ...' >> \
event/trigger
Any number of variables not bound to a 'vals=' prefix can also be
assigned by simply separating them with colons. Below is the same
thing but without the values being summed in the histogram::
# echo 'hist:keys=pid:ts0=common_timestamp:b=field1 ...' >> event/trigger
Variables set as above can be referenced and used in expressions on
another event.
For example, here's how a latency can be calculated::
# echo 'hist:keys=pid,prio:ts0=common_timestamp ...' >> event1/trigger
# echo 'hist:keys=next_pid:wakeup_lat=common_timestamp-$ts0 ...' >> event2/trigger
In the first line above, the event's timestamp is saved into the
variable ts0. In the next line, ts0 is subtracted from the second
event's timestamp to produce the latency, which is then assigned into
yet another variable, 'wakeup_lat'. The hist trigger below in turn
makes use of the wakeup_lat variable to compute a combined latency
using the same key and variable from yet another event::
# echo 'hist:key=pid:wakeupswitch_lat=$wakeup_lat+$switchtime_lat ...' >> event3/trigger
Expressions support the use of addition, subtraction, multiplication and
division operators (+-\*/).
Note if division by zero cannot be detected at parse time (i.e. the
divisor is not a constant), the result will be -1.
Numeric constants can also be used directly in an expression::
# echo 'hist:keys=next_pid:timestamp_secs=common_timestamp/1000000 ...' >> event/trigger
or assigned to a variable and referenced in a subsequent expression::
# echo 'hist:keys=next_pid:us_per_sec=1000000 ...' >> event/trigger
# echo 'hist:keys=next_pid:timestamp_secs=common_timestamp/$us_per_sec ...' >> event/trigger
Variables can even hold stacktraces, which are useful with synthetic events.
Synthetic event 정의와 생성
1792-1865synthetic event는 하나 이상의 다른 event에 연결된 hist trigger variable 또는 field에서 생성하는 사용자 정의 event다. 여러 event에 걸친 data를 기존 normal event와 동일하고 익숙한 방식으로 표시할 수 있게 한다.
synthetic event를 정의하려면 새 event 이름과 하나 이상의 variable 및 type으로 된 간단한 specification을 `tracing/synthetic_events` 파일에 쓴다. field는 semicolon으로 구분하며 type은 유효한 어떤 field type도 사용할 수 있다.
사용 가능한 type은 `synth_field_size()`를 참조한다.
`field_name`에 `[n]`이 있으면 static array로 간주한다.
`field_name`에 subscript 없는 `[]`가 있으면 dynamic array로 간주하며, event 안에서 실제 array를 담는 데 필요한 공간만 차지한다.
string field는 static notation으로 지정할 수 있다.
char name[32];또는 dynamic notation을 사용할 수 있다.
char name[];두 string 형식 모두 크기 제한은 256이다.
형식분류공간/제한
일반 scalar typescalar`synth_field_size()`가 지원하는 type
`type field[n]`static array고정 길이 n
`type field[]`dynamic array실제 data에 필요한 공간만 사용
`char name[32]`static string최대 256
`char name[]`dynamic string최대 256
field 이름의 array notation이 저장 공간과 길이 처리 방식을 결정한다.
다음 명령은 `lat`, `pid`, `prio` field 세 개를 가진 `wakeup_latency` event를 만든다. 각 field는 다른 event의 variable을 받게 될 자리다.
# echo 'wakeup_latency \
u64 lat; \
pid_t pid; \
int prio' >> \
/sys/kernel/tracing/synthetic_events`tracing/synthetic_events`를 읽으면 현재 정의된 synthetic event가 모두 나열되며 여기서는 방금 정의한 event가 표시된다.
# cat /sys/kernel/tracing/synthetic_events
wakeup_latency u64 lat; pid_t pid; int prio기존 synthetic event definition은 정의 명령 앞에 `!`를 붙여 제거할 수 있다.
# echo '!wakeup_latency u64 lat pid_t pid int prio' >> \
/sys/kernel/tracing/synthetic_events이 시점에는 event subsystem에 실제 `wakeup_latency` event instance가 아직 없다. instance를 만들려면 hist trigger action을 생성하고 다른 event에 정의된 실제 field와 variable에 연결해야 한다. 2.7절의 hist trigger `onmatch` action이 그 방법을 설명한다. 연결이 완료되면 `wakeup_latency` synthetic event instance가 생성된다.
새 event는 `tracing/events/synthetic/` directory 아래에 만들어지고 일반 event와 똑같이 보이며 동작한다.
# ls /sys/kernel/tracing/events/synthetic/wakeup_latency
enable filter format hist id trigger이제 새 synthetic event에도 histogram을 정의할 수 있다.
# echo 'hist:keys=pid,prio,lat.log2:sort=lat' >> \
/sys/kernel/tracing/events/synthetic/wakeup_latency/trigger위 명령은 latency `lat`를 2의 거듭제곱 단위로 grouping한다.
synthetic_events→event 이름 + field type 정의
definition 등록→아직 instance 없음
source event hist variable→실제 값 저장
onmatch action→field/variable binding
events/synthetic/wakeup_latency→일반 event처럼 사용
definition만 등록한 상태에서 hist action이 실제 source field를 연결하면 정식 event instance가 된다.
2.6. Synthetic Events
---------------------
Synthetic events are user-defined events generated from hist trigger
variables or fields associated with one or more other events. Their
purpose is to provide a mechanism for displaying data spanning
multiple events consistent with the existing and already familiar
usage for normal events.
To define a synthetic event, the user writes a simple specification
consisting of the name of the new event along with one or more
variables and their types, which can be any valid field type,
separated by semicolons, to the tracing/synthetic_events file.
See synth_field_size() for available types.
If field_name contains [n], the field is considered to be a static array.
If field_names contains[] (no subscript), the field is considered to
be a dynamic array, which will only take as much space in the event as
is required to hold the array.
A string field can be specified using either the static notation:
char name[32];
Or the dynamic:
char name[];
The size limit for either is 256.
For instance, the following creates a new event named 'wakeup_latency'
with 3 fields: lat, pid, and prio. Each of those fields is simply a
variable reference to a variable on another event::
# echo 'wakeup_latency \
u64 lat; \
pid_t pid; \
int prio' >> \
/sys/kernel/tracing/synthetic_events
Reading the tracing/synthetic_events file lists all the currently
defined synthetic events, in this case the event defined above::
# cat /sys/kernel/tracing/synthetic_events
wakeup_latency u64 lat; pid_t pid; int prio
An existing synthetic event definition can be removed by prepending
the command that defined it with a '!'::
# echo '!wakeup_latency u64 lat pid_t pid int prio' >> \
/sys/kernel/tracing/synthetic_events
At this point, there isn't yet an actual 'wakeup_latency' event
instantiated in the event subsystem - for this to happen, a 'hist
trigger action' needs to be instantiated and bound to actual fields
and variables defined on other events (see Section 2.7. below on
how that is done using hist trigger 'onmatch' action). Once that is
done, the 'wakeup_latency' synthetic event instance is created.
The new event is created under the tracing/events/synthetic/ directory
and looks and behaves just like any other event::
# ls /sys/kernel/tracing/events/synthetic/wakeup_latency
enable filter format hist id trigger
A histogram can now be defined for the new synthetic event::
# echo 'hist:keys=pid,prio,lat.log2:sort=lat' >> \
/sys/kernel/tracing/events/synthetic/wakeup_latency/trigger
The above shows the latency "lat" in a power of 2 grouping.
Synthetic latency histogram
1866-1945다른 event와 마찬가지로 synthetic event에 histogram을 활성화하면 해당 event의 `hist` 파일을 읽어 결과를 표시할 수 있다. 첫 출력은 `pid`, `prio`, `lat.log2`를 key로 사용해 latency를 2의 거듭제곱 구간으로 집계한다.
# cat /sys/kernel/tracing/events/synthetic/wakeup_latency/hist
# event histogram
#
# trigger info: hist:keys=pid,prio,lat.log2:vals=hitcount:sort=lat.log2:size=2048 [active]
#
{ pid: 2035, prio: 9, lat: ~ 2^2 } hitcount: 43
{ pid: 2034, prio: 9, lat: ~ 2^2 } hitcount: 60
{ pid: 2029, prio: 9, lat: ~ 2^2 } hitcount: 965
{ pid: 2034, prio: 120, lat: ~ 2^2 } hitcount: 9
{ pid: 2033, prio: 120, lat: ~ 2^2 } hitcount: 5
{ pid: 2030, prio: 9, lat: ~ 2^2 } hitcount: 335
{ pid: 2030, prio: 120, lat: ~ 2^2 } hitcount: 10
{ pid: 2032, prio: 120, lat: ~ 2^2 } hitcount: 1
{ pid: 2035, prio: 120, lat: ~ 2^2 } hitcount: 2
{ pid: 2031, prio: 9, lat: ~ 2^2 } hitcount: 176
{ pid: 2028, prio: 120, lat: ~ 2^2 } hitcount: 15
{ pid: 2033, prio: 9, lat: ~ 2^2 } hitcount: 91
{ pid: 2032, prio: 9, lat: ~ 2^2 } hitcount: 125
{ pid: 2029, prio: 120, lat: ~ 2^2 } hitcount: 4
{ pid: 2031, prio: 120, lat: ~ 2^2 } hitcount: 3
{ pid: 2029, prio: 120, lat: ~ 2^3 } hitcount: 2
{ pid: 2035, prio: 9, lat: ~ 2^3 } hitcount: 41
{ pid: 2030, prio: 120, lat: ~ 2^3 } hitcount: 1
{ pid: 2032, prio: 9, lat: ~ 2^3 } hitcount: 32
{ pid: 2031, prio: 9, lat: ~ 2^3 } hitcount: 44
{ pid: 2034, prio: 9, lat: ~ 2^3 } hitcount: 40
{ pid: 2030, prio: 9, lat: ~ 2^3 } hitcount: 29
{ pid: 2033, prio: 9, lat: ~ 2^3 } hitcount: 31
{ pid: 2029, prio: 9, lat: ~ 2^3 } hitcount: 31
{ pid: 2028, prio: 120, lat: ~ 2^3 } hitcount: 18
{ pid: 2031, prio: 120, lat: ~ 2^3 } hitcount: 2
{ pid: 2028, prio: 120, lat: ~ 2^4 } hitcount: 1
{ pid: 2029, prio: 9, lat: ~ 2^4 } hitcount: 4
{ pid: 2031, prio: 120, lat: ~ 2^7 } hitcount: 1
{ pid: 2032, prio: 120, lat: ~ 2^7 } hitcount: 1
Totals:
Hits: 2122
Entries: 30
Dropped: 0latency 값은 `.buckets` modifier와 크기를 지정해 선형 구간으로도 grouping할 수 있다. 다음 예제는 크기 10인 구간을 사용한다.
# echo 'hist:keys=pid,prio,lat.buckets=10:sort=lat' >> \
/sys/kernel/tracing/events/synthetic/wakeup_latency/trigger
# event histogram
#
# trigger info: hist:keys=pid,prio,lat.buckets=10:vals=hitcount:sort=lat.buckets=10:size=2048 [active]
#
{ pid: 2067, prio: 9, lat: ~ 0-9 } hitcount: 220
{ pid: 2068, prio: 9, lat: ~ 0-9 } hitcount: 157
{ pid: 2070, prio: 9, lat: ~ 0-9 } hitcount: 100
{ pid: 2067, prio: 120, lat: ~ 0-9 } hitcount: 6
{ pid: 2065, prio: 120, lat: ~ 0-9 } hitcount: 2
{ pid: 2066, prio: 120, lat: ~ 0-9 } hitcount: 2
{ pid: 2069, prio: 9, lat: ~ 0-9 } hitcount: 122
{ pid: 2069, prio: 120, lat: ~ 0-9 } hitcount: 8
{ pid: 2070, prio: 120, lat: ~ 0-9 } hitcount: 1
{ pid: 2068, prio: 120, lat: ~ 0-9 } hitcount: 7
{ pid: 2066, prio: 9, lat: ~ 0-9 } hitcount: 365
{ pid: 2064, prio: 120, lat: ~ 0-9 } hitcount: 35
{ pid: 2065, prio: 9, lat: ~ 0-9 } hitcount: 998
{ pid: 2071, prio: 9, lat: ~ 0-9 } hitcount: 85
{ pid: 2065, prio: 9, lat: ~ 10-19 } hitcount: 2
{ pid: 2064, prio: 120, lat: ~ 10-19 } hitcount: 2
Totals:
Hits: 2112
Entries: 16
Dropped: 0modifier구간 형태예제 출력
`lat.log2`2의 거듭제곱`~ 2^2`, `~ 2^3`
`lat.buckets=10`폭 10의 선형 구간`~ 0-9`, `~ 10-19`
같은 latency field를 분석 목적에 따라 logarithmic 또는 linear bucket으로 표시한다.
Like any other event, once a histogram is enabled for the event, the
output can be displayed by reading the event's 'hist' file::
# cat /sys/kernel/tracing/events/synthetic/wakeup_latency/hist
# event histogram
#
# trigger info: hist:keys=pid,prio,lat.log2:vals=hitcount:sort=lat.log2:size=2048 [active]
#
{ pid: 2035, prio: 9, lat: ~ 2^2 } hitcount: 43
{ pid: 2034, prio: 9, lat: ~ 2^2 } hitcount: 60
{ pid: 2029, prio: 9, lat: ~ 2^2 } hitcount: 965
{ pid: 2034, prio: 120, lat: ~ 2^2 } hitcount: 9
{ pid: 2033, prio: 120, lat: ~ 2^2 } hitcount: 5
{ pid: 2030, prio: 9, lat: ~ 2^2 } hitcount: 335
{ pid: 2030, prio: 120, lat: ~ 2^2 } hitcount: 10
{ pid: 2032, prio: 120, lat: ~ 2^2 } hitcount: 1
{ pid: 2035, prio: 120, lat: ~ 2^2 } hitcount: 2
{ pid: 2031, prio: 9, lat: ~ 2^2 } hitcount: 176
{ pid: 2028, prio: 120, lat: ~ 2^2 } hitcount: 15
{ pid: 2033, prio: 9, lat: ~ 2^2 } hitcount: 91
{ pid: 2032, prio: 9, lat: ~ 2^2 } hitcount: 125
{ pid: 2029, prio: 120, lat: ~ 2^2 } hitcount: 4
{ pid: 2031, prio: 120, lat: ~ 2^2 } hitcount: 3
{ pid: 2029, prio: 120, lat: ~ 2^3 } hitcount: 2
{ pid: 2035, prio: 9, lat: ~ 2^3 } hitcount: 41
{ pid: 2030, prio: 120, lat: ~ 2^3 } hitcount: 1
{ pid: 2032, prio: 9, lat: ~ 2^3 } hitcount: 32
{ pid: 2031, prio: 9, lat: ~ 2^3 } hitcount: 44
{ pid: 2034, prio: 9, lat: ~ 2^3 } hitcount: 40
{ pid: 2030, prio: 9, lat: ~ 2^3 } hitcount: 29
{ pid: 2033, prio: 9, lat: ~ 2^3 } hitcount: 31
{ pid: 2029, prio: 9, lat: ~ 2^3 } hitcount: 31
{ pid: 2028, prio: 120, lat: ~ 2^3 } hitcount: 18
{ pid: 2031, prio: 120, lat: ~ 2^3 } hitcount: 2
{ pid: 2028, prio: 120, lat: ~ 2^4 } hitcount: 1
{ pid: 2029, prio: 9, lat: ~ 2^4 } hitcount: 4
{ pid: 2031, prio: 120, lat: ~ 2^7 } hitcount: 1
{ pid: 2032, prio: 120, lat: ~ 2^7 } hitcount: 1
Totals:
Hits: 2122
Entries: 30
Dropped: 0
The latency values can also be grouped linearly by a given size with
the ".buckets" modifier and specify a size (in this case groups of 10)::
# echo 'hist:keys=pid,prio,lat.buckets=10:sort=lat' >> \
/sys/kernel/tracing/events/synthetic/wakeup_latency/trigger
# event histogram
#
# trigger info: hist:keys=pid,prio,lat.buckets=10:vals=hitcount:sort=lat.buckets=10:size=2048 [active]
#
{ pid: 2067, prio: 9, lat: ~ 0-9 } hitcount: 220
{ pid: 2068, prio: 9, lat: ~ 0-9 } hitcount: 157
{ pid: 2070, prio: 9, lat: ~ 0-9 } hitcount: 100
{ pid: 2067, prio: 120, lat: ~ 0-9 } hitcount: 6
{ pid: 2065, prio: 120, lat: ~ 0-9 } hitcount: 2
{ pid: 2066, prio: 120, lat: ~ 0-9 } hitcount: 2
{ pid: 2069, prio: 9, lat: ~ 0-9 } hitcount: 122
{ pid: 2069, prio: 120, lat: ~ 0-9 } hitcount: 8
{ pid: 2070, prio: 120, lat: ~ 0-9 } hitcount: 1
{ pid: 2068, prio: 120, lat: ~ 0-9 } hitcount: 7
{ pid: 2066, prio: 9, lat: ~ 0-9 } hitcount: 365
{ pid: 2064, prio: 120, lat: ~ 0-9 } hitcount: 35
{ pid: 2065, prio: 9, lat: ~ 0-9 } hitcount: 998
{ pid: 2071, prio: 9, lat: ~ 0-9 } hitcount: 85
{ pid: 2065, prio: 9, lat: ~ 10-19 } hitcount: 2
{ pid: 2064, prio: 120, lat: ~ 10-19 } hitcount: 2
Totals:
Hits: 2112
Entries: 16
Dropped: 0
Stacktrace를 담는 synthetic event
1946-2096stacktrace를 저장하려면 `unsigned long[]` 또는 간단히 `long[]` type field가 있는 synthetic event를 만든다. 다음 예제는 task가 uninterruptible state에서 block된 시간을 측정한다.
# cd /sys/kernel/tracing
# echo 's:block_lat pid_t pid; u64 delta; unsigned long[] stack;' > dynamic_events
# echo 'hist:keys=next_pid:ts=common_timestamp.usecs,st=common_stacktrace if prev_state == 2' >> events/sched/sched_switch/trigger
# echo 'hist:keys=prev_pid:delta=common_timestamp.usecs-$ts,s=$st:onmax($delta).trace(block_lat,prev_pid,$delta,$s)' >> events/sched/sched_switch/trigger
# echo 1 > events/synthetic/block_lat/enable
# cat trace
# tracer: nop
#
# entries-in-buffer/entries-written: 2/2 #P:8
#
# _-----=> irqs-off/BH-disabled
# / _----=> need-resched
# | / _---=> hardirq/softirq
# || / _--=> preempt-depth
# ||| / _-=> migrate-disable
# |||| / delay
# TASK-PID CPU# ||||| TIMESTAMP FUNCTION
# | | | ||||| | |
<idle>-0 [005] d..4. 521.164922: block_lat: pid=0 delta=8322 stack=STACK:
=> __schedule+0x448/0x7b0
=> schedule+0x5a/0xb0
=> io_schedule+0x42/0x70
=> bit_wait_io+0xd/0x60
=> __wait_on_bit+0x4b/0x140
=> out_of_line_wait_on_bit+0x91/0xb0
=> jbd2_journal_commit_transaction+0x1679/0x1a70
=> kjournald2+0xa9/0x280
=> kthread+0xe9/0x110
=> ret_from_fork+0x2c/0x50
<...>-2 [004] d..4. 525.184257: block_lat: pid=2 delta=76 stack=STACK:
=> __schedule+0x448/0x7b0
=> schedule+0x5a/0xb0
=> schedule_timeout+0x11a/0x150
=> wait_for_completion_killable+0x144/0x1f0
=> __kthread_create_on_node+0xe7/0x1e0
=> kthread_create_on_node+0x51/0x70
=> create_worker+0xcc/0x1a0
=> worker_thread+0x2ad/0x380
=> kthread+0xe9/0x110
=> ret_from_fork+0x2c/0x50stacktrace field가 있는 synthetic event는 그 field를 histogram key로 사용할 수 있다. 다음 histogram은 `delta`를 100 단위로 grouping하고 `stack.stacktrace` modifier로 각 호출 경로를 펼쳐 표시한다.
# echo 'hist:keys=delta.buckets=100,stack.stacktrace:sort=delta' > events/synthetic/block_lat/trigger
# cat events/synthetic/block_lat/hist
# event histogram
#
# trigger info: hist:keys=delta.buckets=100,stack.stacktrace:vals=hitcount:sort=delta.buckets=100:size=2048 [active]
#
{ delta: ~ 0-99, stack.stacktrace __schedule+0xa19/0x1520
schedule+0x6b/0x110
io_schedule+0x46/0x80
bit_wait_io+0x11/0x80
__wait_on_bit+0x4e/0x120
out_of_line_wait_on_bit+0x8d/0xb0
__wait_on_buffer+0x33/0x40
jbd2_journal_commit_transaction+0x155a/0x19b0
kjournald2+0xab/0x270
kthread+0xfa/0x130
ret_from_fork+0x29/0x50
} hitcount: 1
{ delta: ~ 0-99, stack.stacktrace __schedule+0xa19/0x1520
schedule+0x6b/0x110
io_schedule+0x46/0x80
rq_qos_wait+0xd0/0x170
wbt_wait+0x9e/0xf0
__rq_qos_throttle+0x25/0x40
blk_mq_submit_bio+0x2c3/0x5b0
__submit_bio+0xff/0x190
submit_bio_noacct_nocheck+0x25b/0x2b0
submit_bio_noacct+0x20b/0x600
submit_bio+0x28/0x90
ext4_bio_write_page+0x1e0/0x8c0
mpage_submit_page+0x60/0x80
mpage_process_page_bufs+0x16c/0x180
mpage_prepare_extent_to_map+0x23f/0x530
} hitcount: 1
{ delta: ~ 0-99, stack.stacktrace __schedule+0xa19/0x1520
schedule+0x6b/0x110
schedule_hrtimeout_range_clock+0x97/0x110
schedule_hrtimeout_range+0x13/0x20
usleep_range_state+0x65/0x90
__intel_wait_for_register+0x1c1/0x230 [i915]
intel_psr_wait_for_idle_locked+0x171/0x2a0 [i915]
intel_pipe_update_start+0x169/0x360 [i915]
intel_update_crtc+0x112/0x490 [i915]
skl_commit_modeset_enables+0x199/0x600 [i915]
intel_atomic_commit_tail+0x7c4/0x1080 [i915]
intel_atomic_commit_work+0x12/0x20 [i915]
process_one_work+0x21c/0x3f0
worker_thread+0x50/0x3e0
kthread+0xfa/0x130
} hitcount: 3
{ delta: ~ 0-99, stack.stacktrace __schedule+0xa19/0x1520
schedule+0x6b/0x110
schedule_timeout+0x11e/0x160
__wait_for_common+0x8f/0x190
wait_for_completion+0x24/0x30
__flush_work.isra.0+0x1cc/0x360
flush_work+0xe/0x20
drm_mode_rmfb+0x18b/0x1d0 [drm]
drm_mode_rmfb_ioctl+0x10/0x20 [drm]
drm_ioctl_kernel+0xb8/0x150 [drm]
drm_ioctl+0x243/0x560 [drm]
__x64_sys_ioctl+0x92/0xd0
do_syscall_64+0x59/0x90
entry_SYSCALL_64_after_hwframe+0x72/0xdc
} hitcount: 1
{ delta: ~ 0-99, stack.stacktrace __schedule+0xa19/0x1520
schedule+0x6b/0x110
schedule_timeout+0x87/0x160
__wait_for_common+0x8f/0x190
wait_for_completion_timeout+0x1d/0x30
drm_atomic_helper_wait_for_flip_done+0x57/0x90 [drm_kms_helper]
intel_atomic_commit_tail+0x8ce/0x1080 [i915]
intel_atomic_commit_work+0x12/0x20 [i915]
process_one_work+0x21c/0x3f0
worker_thread+0x50/0x3e0
kthread+0xfa/0x130
ret_from_fork+0x29/0x50
} hitcount: 1
{ delta: ~ 100-199, stack.stacktrace __schedule+0xa19/0x1520
schedule+0x6b/0x110
schedule_hrtimeout_range_clock+0x97/0x110
schedule_hrtimeout_range+0x13/0x20
usleep_range_state+0x65/0x90
pci_set_low_power_state+0x17f/0x1f0
pci_set_power_state+0x49/0x250
pci_finish_runtime_suspend+0x4a/0x90
pci_pm_runtime_suspend+0xcb/0x1b0
__rpm_callback+0x48/0x120
rpm_callback+0x67/0x70
rpm_suspend+0x167/0x780
rpm_idle+0x25a/0x380
pm_runtime_work+0x93/0xc0
process_one_work+0x21c/0x3f0
} hitcount: 1
Totals:
Hits: 10
Entries: 7
Dropped: 0sched_switch out→prev_state == 2
ts = common_timestamp.usecs→st = common_stacktrace
matching task switch in→delta = now - $ts
onmax($delta).trace→block_lat(pid, delta, stack)
synthetic/block_lat→trace와 histogram에서 분석
두 sched_switch 관측점 사이의 시간과 첫 지점의 stacktrace를 결합해 최대 block latency event를 만든다.
단계field 표현역할
synthetic definition`unsigned long[] stack`가변 길이 stack frame 저장
source variable`st=common_stacktrace`block 시작 시 호출 경로 capture
synthetic action`$s`저장한 stack을 `block_lat`로 전달
histogram key`stack.stacktrace`frame 목록을 호출 경로 key로 표시
동적 stack array를 synthetic event와 histogram에서 단계별로 해석한다.
To save stacktraces, create a synthetic event with a field of type "unsigned long[]"
or even just "long[]". For example, to see how long a task is blocked in an
uninterruptible state::
# cd /sys/kernel/tracing
# echo 's:block_lat pid_t pid; u64 delta; unsigned long[] stack;' > dynamic_events
# echo 'hist:keys=next_pid:ts=common_timestamp.usecs,st=common_stacktrace if prev_state == 2' >> events/sched/sched_switch/trigger
# echo 'hist:keys=prev_pid:delta=common_timestamp.usecs-$ts,s=$st:onmax($delta).trace(block_lat,prev_pid,$delta,$s)' >> events/sched/sched_switch/trigger
# echo 1 > events/synthetic/block_lat/enable
# cat trace
# tracer: nop
#
# entries-in-buffer/entries-written: 2/2 #P:8
#
# _-----=> irqs-off/BH-disabled
# / _----=> need-resched
# | / _---=> hardirq/softirq
# || / _--=> preempt-depth
# ||| / _-=> migrate-disable
# |||| / delay
# TASK-PID CPU# ||||| TIMESTAMP FUNCTION
# | | | ||||| | |
<idle>-0 [005] d..4. 521.164922: block_lat: pid=0 delta=8322 stack=STACK:
=> __schedule+0x448/0x7b0
=> schedule+0x5a/0xb0
=> io_schedule+0x42/0x70
=> bit_wait_io+0xd/0x60
=> __wait_on_bit+0x4b/0x140
=> out_of_line_wait_on_bit+0x91/0xb0
=> jbd2_journal_commit_transaction+0x1679/0x1a70
=> kjournald2+0xa9/0x280
=> kthread+0xe9/0x110
=> ret_from_fork+0x2c/0x50
<...>-2 [004] d..4. 525.184257: block_lat: pid=2 delta=76 stack=STACK:
=> __schedule+0x448/0x7b0
=> schedule+0x5a/0xb0
=> schedule_timeout+0x11a/0x150
=> wait_for_completion_killable+0x144/0x1f0
=> __kthread_create_on_node+0xe7/0x1e0
=> kthread_create_on_node+0x51/0x70
=> create_worker+0xcc/0x1a0
=> worker_thread+0x2ad/0x380
=> kthread+0xe9/0x110
=> ret_from_fork+0x2c/0x50
A synthetic event that has a stacktrace field may use it as a key in
histogram::
# echo 'hist:keys=delta.buckets=100,stack.stacktrace:sort=delta' > events/synthetic/block_lat/trigger
# cat events/synthetic/block_lat/hist
# event histogram
#
# trigger info: hist:keys=delta.buckets=100,stack.stacktrace:vals=hitcount:sort=delta.buckets=100:size=2048 [active]
#
{ delta: ~ 0-99, stack.stacktrace __schedule+0xa19/0x1520
schedule+0x6b/0x110
io_schedule+0x46/0x80
bit_wait_io+0x11/0x80
__wait_on_bit+0x4e/0x120
out_of_line_wait_on_bit+0x8d/0xb0
__wait_on_buffer+0x33/0x40
jbd2_journal_commit_transaction+0x155a/0x19b0
kjournald2+0xab/0x270
kthread+0xfa/0x130
ret_from_fork+0x29/0x50
} hitcount: 1
{ delta: ~ 0-99, stack.stacktrace __schedule+0xa19/0x1520
schedule+0x6b/0x110
io_schedule+0x46/0x80
rq_qos_wait+0xd0/0x170
wbt_wait+0x9e/0xf0
__rq_qos_throttle+0x25/0x40
blk_mq_submit_bio+0x2c3/0x5b0
__submit_bio+0xff/0x190
submit_bio_noacct_nocheck+0x25b/0x2b0
submit_bio_noacct+0x20b/0x600
submit_bio+0x28/0x90
ext4_bio_write_page+0x1e0/0x8c0
mpage_submit_page+0x60/0x80
mpage_process_page_bufs+0x16c/0x180
mpage_prepare_extent_to_map+0x23f/0x530
} hitcount: 1
{ delta: ~ 0-99, stack.stacktrace __schedule+0xa19/0x1520
schedule+0x6b/0x110
schedule_hrtimeout_range_clock+0x97/0x110
schedule_hrtimeout_range+0x13/0x20
usleep_range_state+0x65/0x90
__intel_wait_for_register+0x1c1/0x230 [i915]
intel_psr_wait_for_idle_locked+0x171/0x2a0 [i915]
intel_pipe_update_start+0x169/0x360 [i915]
intel_update_crtc+0x112/0x490 [i915]
skl_commit_modeset_enables+0x199/0x600 [i915]
intel_atomic_commit_tail+0x7c4/0x1080 [i915]
intel_atomic_commit_work+0x12/0x20 [i915]
process_one_work+0x21c/0x3f0
worker_thread+0x50/0x3e0
kthread+0xfa/0x130
} hitcount: 3
{ delta: ~ 0-99, stack.stacktrace __schedule+0xa19/0x1520
schedule+0x6b/0x110
schedule_timeout+0x11e/0x160
__wait_for_common+0x8f/0x190
wait_for_completion+0x24/0x30
__flush_work.isra.0+0x1cc/0x360
flush_work+0xe/0x20
drm_mode_rmfb+0x18b/0x1d0 [drm]
drm_mode_rmfb_ioctl+0x10/0x20 [drm]
drm_ioctl_kernel+0xb8/0x150 [drm]
drm_ioctl+0x243/0x560 [drm]
__x64_sys_ioctl+0x92/0xd0
do_syscall_64+0x59/0x90
entry_SYSCALL_64_after_hwframe+0x72/0xdc
} hitcount: 1
{ delta: ~ 0-99, stack.stacktrace __schedule+0xa19/0x1520
schedule+0x6b/0x110
schedule_timeout+0x87/0x160
__wait_for_common+0x8f/0x190
wait_for_completion_timeout+0x1d/0x30
drm_atomic_helper_wait_for_flip_done+0x57/0x90 [drm_kms_helper]
intel_atomic_commit_tail+0x8ce/0x1080 [i915]
intel_atomic_commit_work+0x12/0x20 [i915]
process_one_work+0x21c/0x3f0
worker_thread+0x50/0x3e0
kthread+0xfa/0x130
ret_from_fork+0x29/0x50
} hitcount: 1
{ delta: ~ 100-199, stack.stacktrace __schedule+0xa19/0x1520
schedule+0x6b/0x110
schedule_hrtimeout_range_clock+0x97/0x110
schedule_hrtimeout_range+0x13/0x20
usleep_range_state+0x65/0x90
pci_set_low_power_state+0x17f/0x1f0
pci_set_power_state+0x49/0x250
pci_finish_runtime_suspend+0x4a/0x90
pci_pm_runtime_suspend+0xcb/0x1b0
__rpm_callback+0x48/0x120
rpm_callback+0x67/0x70
rpm_suspend+0x167/0x780
rpm_idle+0x25a/0x380
pm_runtime_work+0x93/0xc0
process_one_work+0x21c/0x3f0
} hitcount: 1
Totals:
Hits: 10
Entries: 7
Dropped: 0
Hist trigger handler와 action
2097-2189hist trigger `action`은 histogram entry가 추가되거나 갱신될 때마다, 대부분 조건부로 실행되는 function이다.
histogram entry가 추가되거나 갱신될 때 대응 action을 실제 호출할지는 hist trigger `handler`가 결정한다.
handler와 action은 다음 일반 형식으로 짝을 이룬다.
<handler>.<action>특정 event에 handler.action 쌍을 지정하려면 hist trigger specification의 colon 사이에 그 쌍을 넣는다.
이론적으로는 어떤 handler와 action도 결합할 수 있지만 현재 모든 조합을 지원하지는 않는다. 지원하지 않는 handler.action 조합을 쓰면 hist trigger가 `-EINVAL`로 실패한다.
명시하지 않았을 때의 기본 handler.action은 기존처럼 entry에 연결된 value 집합을 갱신하는 것이다. 그러나 어떤 application은 그 시점에 다른 event를 생성하거나 최댓값을 비교해 저장하는 추가 동작이 필요하다.
지원하는 handler와 action은 다음과 같으며, 뒤의 일반적이고 유용한 조합 예제에서 자세히 설명한다.
handler호출 조건
`onmatch(matching.event)`entry가 추가되거나 갱신되는 모든 matching event
`onmax(var)`variable이 현재 maximum을 초과
`onchange(var)`variable 값이 변경
entry 갱신 시 action을 호출할 조건을 정한다.
action효과
`trace(synthetic_event, params)`synthetic event 생성
`save(field, ...)`현재 event field 저장
`snapshot()`trace buffer snapshot 저장
handler 조건이 충족됐을 때 수행할 동작이다.
`onmatch(matching.event).trace(synthetic_event, param list)`는 event가 match해 histogram entry가 추가되거나 갱신될 때 호출된다. 지정한 parameter 값으로 synthetic event를 생성하므로, invoking event가 적중한 시점의 variable 값으로 구성된 event가 나온다. 예를 들어 `onmatch(event).trace(wakeup_latency,arg1,arg2)`는 `wakeup_latency` event를 생성한다.
synthetic event 이름을 function 이름처럼 호출하는 동등한 대체 형식도 있다. `onmatch(event).wakeup_latency(arg1,arg2)`처럼 event field 값을 argument로 전달하며 일반 문법은 다음과 같다.
onmatch(matching.event).<synthetic_event_name>(param list)두 형식 모두 parameter list에는 `matching.event`나 target event에 정의된 variable 또는 field를 하나 이상 넣는다. variable과 field는 fully-qualified 또는 unqualified일 수 있다. unqualified variable은 두 event 사이에서 고유해야 한다. target event field는 unqualified로 쓸 수 있지만 matching event field는 `system.event_name.$var_name` 또는 `system.event_name.field` 형식의 fully-qualified name이어야 한다.
`matching.event`는 `system.event_name` 형식의 fully-qualified event 이름이다. 두 event의 histogram key를 비교해 match 여부를 판단하며, key가 여러 개라면 지정된 순서로 모두 일치해야 한다.
parameter list의 variable/field 수와 type은 생성할 synthetic event의 field 수와 type에 일치해야 한다.
2.7. Hist trigger 'handlers' and 'actions'
------------------------------------------
A hist trigger 'action' is a function that's executed (in most cases
conditionally) whenever a histogram entry is added or updated.
When a histogram entry is added or updated, a hist trigger 'handler'
is what decides whether the corresponding action is actually invoked
or not.
Hist trigger handlers and actions are paired together in the general
form:
<handler>.<action>
To specify a handler.action pair for a given event, simply specify
that handler.action pair between colons in the hist trigger
specification.
In theory, any handler can be combined with any action, but in
practice, not every handler.action combination is currently supported;
if a given handler.action combination isn't supported, the hist
trigger will fail with -EINVAL;
The default 'handler.action' if none is explicitly specified is as it
always has been, to simply update the set of values associated with an
entry. Some applications, however, may want to perform additional
actions at that point, such as generate another event, or compare and
save a maximum.
The supported handlers and actions are listed below, and each is
described in more detail in the following paragraphs, in the context
of descriptions of some common and useful handler.action combinations.
The available handlers are:
- onmatch(matching.event) - invoke action on any addition or update
- onmax(var) - invoke action if var exceeds current max
- onchange(var) - invoke action if var changes
The available actions are:
- trace(<synthetic_event_name>,param list) - generate synthetic event
- save(field,...) - save current event fields
- snapshot() - snapshot the trace buffer
The following commonly-used handler.action pairs are available:
- onmatch(matching.event).trace(<synthetic_event_name>,param list)
The 'onmatch(matching.event).trace(<synthetic_event_name>,param
list)' hist trigger action is invoked whenever an event matches
and the histogram entry would be added or updated. It causes the
named synthetic event to be generated with the values given in the
'param list'. The result is the generation of a synthetic event
that consists of the values contained in those variables at the
time the invoking event was hit. For example, if the synthetic
event name is 'wakeup_latency', a wakeup_latency event is
generated using onmatch(event).trace(wakeup_latency,arg1,arg2).
There is also an equivalent alternative form available for
generating synthetic events. In this form, the synthetic event
name is used as if it were a function name. For example, using
the 'wakeup_latency' synthetic event name again, the
wakeup_latency event would be generated by invoking it as if it
were a function call, with the event field values passed in as
arguments: onmatch(event).wakeup_latency(arg1,arg2). The syntax
for this form is:
onmatch(matching.event).<synthetic_event_name>(param list)
In either case, the 'param list' consists of one or more
parameters which may be either variables or fields defined on
either the 'matching.event' or the target event. The variables or
fields specified in the param list may be either fully-qualified
or unqualified. If a variable is specified as unqualified, it
must be unique between the two events. A field name used as a
param can be unqualified if it refers to the target event, but
must be fully qualified if it refers to the matching event. A
fully-qualified name is of the form 'system.event_name.$var_name'
or 'system.event_name.field'.
The 'matching.event' specification is simply the fully qualified
event name of the event that matches the target event for the
onmatch() functionality, in the form 'system.event_name'. Histogram
keys of both events are compared to find if events match. In case
multiple histogram keys are used, they all must match in the specified
order.
Finally, the number and type of variables/fields in the 'param
list' must match the number and types of the fields in the
synthetic event being generated.
`onmatch` synthetic event 예제
2190-2267다음 예제는 간단한 synthetic event를 정의하고 `sched_wakeup_new` event의 variable을 호출 parameter로 사용한다. 먼저 synthetic event를 정의한다.
# echo 'wakeup_new_test pid_t pid' >> \
/sys/kernel/tracing/synthetic_events
# cat /sys/kernel/tracing/synthetic_events
wakeup_new_test pid_t pid다음 hist trigger는 빠져 있던 `testpid` variable을 정의하고, `sched_wakeup_new` event가 발생할 때마다 `wakeup_new_test` synthetic event를 만드는 `onmatch()` action을 지정한다. `if comm=="cyclictest"` filter 때문에 executable이 `cyclictest`일 때만 생성된다.
# echo 'hist:keys=$testpid:testpid=pid:onmatch(sched.sched_wakeup_new).\
wakeup_new_test($testpid) if comm=="cyclictest"' >> \
/sys/kernel/tracing/events/sched/sched_wakeup_new/trigger`trace` keyword 문법을 사용해도 동등하게 표현할 수 있다.
# echo 'hist:keys=$testpid:testpid=pid:onmatch(sched.sched_wakeup_new).\
trace(wakeup_new_test,$testpid) if comm=="cyclictest"' >> \
/sys/kernel/tracing/events/sched/sched_wakeup_new/trigger이 event를 기반으로 histogram을 만들고 표시하는 일은 `tracing/events/synthetic` directory의 새 synthetic event와 field를 일반 event처럼 사용하면 된다.
# echo 'hist:keys=pid:sort=pid' >> \
/sys/kernel/tracing/events/synthetic/wakeup_new_test/trigger`cyclictest`를 실행하면 wakeup_new event가 `wakeup_new_test` synthetic event를 만들고, 그 결과가 `wakeup_new_test` event의 `hist` 파일에 집계된다.
# cat /sys/kernel/tracing/events/synthetic/wakeup_new_test/hist더 일반적인 사용법은 event 두 개로 latency를 계산하는 것이다. 다음 hist trigger 집합은 `wakeup_latency` histogram을 만든다.
먼저 `wakeup_latency` synthetic event를 정의한다.
# echo 'wakeup_latency u64 lat; pid_t pid; int prio' >> \
/sys/kernel/tracing/synthetic_events`cyclictest` thread의 `sched_waking` event를 볼 때마다 timestamp를 `ts0` variable에 저장한다.
# echo 'hist:keys=$saved_pid:saved_pid=pid:ts0=common_timestamp.usecs \
if comm=="cyclictest"' >> \
/sys/kernel/tracing/events/sched/sched_waking/trigger대응 thread가 `sched_switch` event로 실제 CPU에 schedule될 때 `saved_pid`와 `next_pid`가 일치한다. 이때 latency를 계산하고 다른 variable 및 event field와 함께 `wakeup_latency` synthetic event를 생성한다.
# echo 'hist:keys=next_pid:wakeup_lat=common_timestamp.usecs-$ts0:\
onmatch(sched.sched_waking).wakeup_latency($wakeup_lat,\
$saved_pid,next_prio) if next_comm=="cyclictest"' >> \
/sys/kernel/tracing/events/sched/sched_switch/trigger생성된 synthetic event data를 집계하려면 `wakeup_latency` event에도 histogram을 만들어야 한다.
# echo 'hist:keys=pid,prio,lat:sort=pid,lat' >> \
/sys/kernel/tracing/events/synthetic/wakeup_latency/trigger마지막으로 `cyclictest`를 실행해 event를 생성한 뒤 `wakeup_latency` synthetic event의 `hist` 파일에서 결과를 확인한다.
# cat /sys/kernel/tracing/events/synthetic/wakeup_latency/histsched_waking→saved_pid + ts0 저장
sched_switch→next_pid와 key match
wakeup_lat = now - $ts0→latency 계산
onmatch(...).wakeup_latency→lat, pid, prio 전달
synthetic/wakeup_latency hist→생성 event 집계
두 scheduler event가 matching pid를 통해 timestamp와 context를 synthetic event로 결합한다.
As an example the below defines a simple synthetic event and uses
a variable defined on the sched_wakeup_new event as a parameter
when invoking the synthetic event. Here we define the synthetic
event::
# echo 'wakeup_new_test pid_t pid' >> \
/sys/kernel/tracing/synthetic_events
# cat /sys/kernel/tracing/synthetic_events
wakeup_new_test pid_t pid
The following hist trigger both defines the missing testpid
variable and specifies an onmatch() action that generates a
wakeup_new_test synthetic event whenever a sched_wakeup_new event
occurs, which because of the 'if comm == "cyclictest"' filter only
happens when the executable is cyclictest::
# echo 'hist:keys=$testpid:testpid=pid:onmatch(sched.sched_wakeup_new).\
wakeup_new_test($testpid) if comm=="cyclictest"' >> \
/sys/kernel/tracing/events/sched/sched_wakeup_new/trigger
Or, equivalently, using the 'trace' keyword syntax::
# echo 'hist:keys=$testpid:testpid=pid:onmatch(sched.sched_wakeup_new).\
trace(wakeup_new_test,$testpid) if comm=="cyclictest"' >> \
/sys/kernel/tracing/events/sched/sched_wakeup_new/trigger
Creating and displaying a histogram based on those events is now
just a matter of using the fields and new synthetic event in the
tracing/events/synthetic directory, as usual::
# echo 'hist:keys=pid:sort=pid' >> \
/sys/kernel/tracing/events/synthetic/wakeup_new_test/trigger
Running 'cyclictest' should cause wakeup_new events to generate
wakeup_new_test synthetic events which should result in histogram
output in the wakeup_new_test event's hist file::
# cat /sys/kernel/tracing/events/synthetic/wakeup_new_test/hist
A more typical usage would be to use two events to calculate a
latency. The following example uses a set of hist triggers to
produce a 'wakeup_latency' histogram.
First, we define a 'wakeup_latency' synthetic event::
# echo 'wakeup_latency u64 lat; pid_t pid; int prio' >> \
/sys/kernel/tracing/synthetic_events
Next, we specify that whenever we see a sched_waking event for a
cyclictest thread, save the timestamp in a 'ts0' variable::
# echo 'hist:keys=$saved_pid:saved_pid=pid:ts0=common_timestamp.usecs \
if comm=="cyclictest"' >> \
/sys/kernel/tracing/events/sched/sched_waking/trigger
Then, when the corresponding thread is actually scheduled onto the
CPU by a sched_switch event (saved_pid matches next_pid), calculate
the latency and use that along with another variable and an event field
to generate a wakeup_latency synthetic event::
# echo 'hist:keys=next_pid:wakeup_lat=common_timestamp.usecs-$ts0:\
onmatch(sched.sched_waking).wakeup_latency($wakeup_lat,\
$saved_pid,next_prio) if next_comm=="cyclictest"' >> \
/sys/kernel/tracing/events/sched/sched_switch/trigger
We also need to create a histogram on the wakeup_latency synthetic
event in order to aggregate the generated synthetic event data::
# echo 'hist:keys=pid,prio,lat:sort=pid,lat' >> \
/sys/kernel/tracing/events/synthetic/wakeup_latency/trigger
Finally, once we've run cyclictest to actually generate some
events, we can see the output by looking at the wakeup_latency
synthetic event's hist file::
# cat /sys/kernel/tracing/events/synthetic/wakeup_latency/hist
`onmax().save()`
2268-2320`onmax(var).save(field,...)` action은 histogram entry에 연결된 `var` 값이 그 variable에 저장된 현재 maximum을 초과할 때마다 호출된다.
`var`가 해당 entry의 현재 maximum을 넘으면 `save()` parameter로 지정한 trace event field를 저장한다. 새 maximum을 만든 event의 context를 나중에 참조할 수 있으며 histogram 출력에는 저장한 값이 추가 field로 표시된다.
다음 예제는 pid를 key로 하는 `sched_waking`과 `sched_switch` hist trigger를 정의한다. `sched_waking` 때 현재 pid entry에 timestamp를 저장하고 scheduler가 그 pid로 돌아오면 timestamp 차이를 계산한다. 결과 latency인 `wakeup_lat`가 현재 maximum을 넘으면 `save()` field 값을 기록한다.
# echo 'hist:keys=pid:ts0=common_timestamp.usecs \
if comm=="cyclictest"' >> \
/sys/kernel/tracing/events/sched/sched_waking/trigger
# echo 'hist:keys=next_pid:\
wakeup_lat=common_timestamp.usecs-$ts0:\
onmax($wakeup_lat).save(next_comm,prev_pid,prev_prio,prev_comm) \
if next_comm=="cyclictest"' >> \
/sys/kernel/tracing/events/sched/sched_switch/triggerhistogram을 표시하면 각 entry의 나머지 field 뒤에 maximum과 그 maximum에 대응해 저장된 값이 나온다.
# cat /sys/kernel/tracing/events/sched/sched_switch/hist
{ next_pid: 2255 } hitcount: 239
common_timestamp-ts0: 0
max: 27
next_comm: cyclictest
prev_pid: 0 prev_prio: 120 prev_comm: swapper/1
{ next_pid: 2256 } hitcount: 2355
common_timestamp-ts0: 0
max: 49 next_comm: cyclictest
prev_pid: 0 prev_prio: 120 prev_comm: swapper/0
Totals:
Hits: 12970
Entries: 2
Dropped: 0항목범위예제
비교 variableentry별`$wakeup_lat`
maximumentry별각 `next_pid`의 `max`
saved fields새 maximum을 만든 event`next_comm`, `prev_pid`, `prev_prio`, `prev_comm`
maximum과 context는 histogram key별 entry에 귀속된다.
- onmax(var).save(field,.. .)
The 'onmax(var).save(field,...)' hist trigger action is invoked
whenever the value of 'var' associated with a histogram entry
exceeds the current maximum contained in that variable.
The end result is that the trace event fields specified as the
onmax.save() params will be saved if 'var' exceeds the current
maximum for that hist trigger entry. This allows context from the
event that exhibited the new maximum to be saved for later
reference. When the histogram is displayed, additional fields
displaying the saved values will be printed.
As an example the below defines a couple of hist triggers, one for
sched_waking and another for sched_switch, keyed on pid. Whenever
a sched_waking occurs, the timestamp is saved in the entry
corresponding to the current pid, and when the scheduler switches
back to that pid, the timestamp difference is calculated. If the
resulting latency, stored in wakeup_lat, exceeds the current
maximum latency, the values specified in the save() fields are
recorded::
# echo 'hist:keys=pid:ts0=common_timestamp.usecs \
if comm=="cyclictest"' >> \
/sys/kernel/tracing/events/sched/sched_waking/trigger
# echo 'hist:keys=next_pid:\
wakeup_lat=common_timestamp.usecs-$ts0:\
onmax($wakeup_lat).save(next_comm,prev_pid,prev_prio,prev_comm) \
if next_comm=="cyclictest"' >> \
/sys/kernel/tracing/events/sched/sched_switch/trigger
When the histogram is displayed, the max value and the saved
values corresponding to the max are displayed following the rest
of the fields::
# cat /sys/kernel/tracing/events/sched/sched_switch/hist
{ next_pid: 2255 } hitcount: 239
common_timestamp-ts0: 0
max: 27
next_comm: cyclictest
prev_pid: 0 prev_prio: 120 prev_comm: swapper/1
{ next_pid: 2256 } hitcount: 2355
common_timestamp-ts0: 0
max: 49 next_comm: cyclictest
prev_pid: 0 prev_prio: 120 prev_comm: swapper/0
Totals:
Hits: 12970
Entries: 2
Dropped: 0
`onmax().snapshot()`
2321-2429`onmax(var).snapshot()` action은 histogram entry에 연결된 `var` 값이 현재 maximum을 넘을 때마다 호출된다.
어떤 hist trigger entry의 `var`가 current maximum을 초과하면 trace buffer의 global snapshot을 `tracing/snapshot` 파일에 저장한다.
이 경우 maximum은 current trace instance의 global maximum, 즉 histogram 모든 bucket 중 maximum이다. 이를 만든 trace event의 key와 global maximum, snapshot이 저장됐다는 메시지와 위치가 표시된다. key로 대응 histogram bucket을 찾으면 더 자세한 context를 볼 수 있다.
다음 예제는 pid key의 `sched_waking`과 `sched_switch` trigger로 latency를 계산한다. `wakeup_lat`가 현재 maximum을 넘으면 snapshot을 만든다. snapshot에 scheduler event가 들어오도록 설정 과정에서 scheduler event 전체도 활성화한다.
# echo 1 > /sys/kernel/tracing/events/sched/enable
# echo 'hist:keys=pid:ts0=common_timestamp.usecs \
if comm=="cyclictest"' >> \
/sys/kernel/tracing/events/sched/sched_waking/trigger
# echo 'hist:keys=next_pid:wakeup_lat=common_timestamp.usecs-$ts0: \
onmax($wakeup_lat).save(next_prio,next_comm,prev_pid,prev_prio, \
prev_comm):onmax($wakeup_lat).snapshot() \
if next_comm=="cyclictest"' >> \
/sys/kernel/tracing/events/sched/sched_switch/triggerhistogram에는 각 bucket의 maximum과 이에 대응해 저장된 값이 나머지 field 뒤에 표시된다.
snapshot이 생성됐다면 그 사실과 global maximum을 발생시킨 value 및 event도 함께 표시된다.
# cat /sys/kernel/tracing/events/sched/sched_switch/hist
{ next_pid: 2101 } hitcount: 200
max: 52 next_prio: 120 next_comm: cyclictest \
prev_pid: 0 prev_prio: 120 prev_comm: swapper/6
{ next_pid: 2103 } hitcount: 1326
max: 572 next_prio: 19 next_comm: cyclictest \
prev_pid: 0 prev_prio: 120 prev_comm: swapper/1
{ next_pid: 2102 } hitcount: 1982 \
max: 74 next_prio: 19 next_comm: cyclictest \
prev_pid: 0 prev_prio: 120 prev_comm: swapper/5
Snapshot taken (see tracing/snapshot). Details:
triggering value { onmax($wakeup_lat) }: 572 \
triggered by event with key: { next_pid: 2103 }
Totals:
Hits: 3508
Entries: 3
Dropped: 0예제에서 global maximum을 만든 event key는 `next_pid == 2103`이다. key가 2103인 bucket에는 local maximum과 함께 `save()`한 추가 값이 있다. global snapshot을 일으킨 바로 그 값이므로 local maximum은 global maximum과 같다.
마지막으로 snapshot data의 끝부분 근처에는 snapshot을 발생시킨 event가 보여야 한다. `sched_waking`과 `sched_switch` 사이 timestamp 차이가 표시된 global maximum과 일치하는지 확인할 수 있다.
# cat /sys/kernel/tracing/snapshot
<...>-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
<idle>-0 [005] d.h3 309.873611: sched_waking: comm=cyclictest pid=2102 prio=19 target_cpu=005
<idle>-0 [005] dNh4 309.873613: sched_wakeup: comm=cyclictest pid=2102 prio=19 target_cpu=005
<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
<...>-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
<idle>-0 [005] d.h3 309.874624: sched_waking: comm=cyclictest pid=2102 prio=19 target_cpu=005
<idle>-0 [005] dNh4 309.874626: sched_wakeup: comm=cyclictest pid=2102 prio=19 target_cpu=005
<idle>-0 [005] dNh3 309.874628: sched_waking: comm=cyclictest pid=2103 prio=19 target_cpu=005
<idle>-0 [005] dNh4 309.874630: sched_wakeup: comm=cyclictest pid=2103 prio=19 target_cpu=005
<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
<idle>-0 [004] d.h3 309.874757: sched_waking: comm=gnome-terminal- pid=1699 prio=120 target_cpu=004
<idle>-0 [004] dNh4 309.874762: sched_wakeup: comm=gnome-terminal- pid=1699 prio=120 target_cpu=004
<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
gnome-terminal--1699 [004] d.h2 309.874941: sched_stat_runtime: comm=gnome-terminal- pid=1699 runtime=180706 [ns] vruntime=1126870572 [ns]
<idle>-0 [003] d.s4 309.874956: sched_waking: comm=rcu_sched pid=9 prio=120 target_cpu=007
<idle>-0 [003] d.s5 309.874960: sched_wake_idle_without_ipi: cpu=7
<idle>-0 [003] d.s5 309.874961: sched_wakeup: comm=rcu_sched pid=9 prio=120 target_cpu=007
<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
rcu_sched-9 [007] d..3 309.874973: sched_stat_runtime: comm=rcu_sched pid=9 runtime=13646 [ns] vruntime=22531430286 [ns]
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
<...>-2102 [005] d..4 309.874994: sched_migrate_task: comm=cyclictest pid=2103 prio=19 orig_cpu=5 dest_cpu=1
<...>-2102 [005] d..4 309.875185: sched_wake_idle_without_ipi: cpu=1
<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=19각 pid bucket→wakeup_lat 계산
entry local max 갱신→save context
모든 bucket 비교→global max
onmax().snapshot()→tracing/snapshot 저장
triggering key→hist bucket과 snapshot event 연결
entry별 latency 비교에서 trace instance 전체의 새 maximum이 생길 때 buffer를 고정한다.
- onmax(var).snapshot()
The 'onmax(var).snapshot()' hist trigger action is invoked
whenever the value of 'var' associated with a histogram entry
exceeds the current maximum contained in that variable.
The end result is that a global snapshot of the trace buffer will
be saved in the tracing/snapshot file if 'var' exceeds the current
maximum for any hist trigger entry.
Note that in this case the maximum is a global maximum for the
current trace instance, which is the maximum across all buckets of
the histogram. The key of the specific trace event that caused
the global maximum and the global maximum itself are displayed,
along with a message stating that a snapshot has been taken and
where to find it. The user can use the key information displayed
to locate the corresponding bucket in the histogram for even more
detail.
As an example the below defines a couple of hist triggers, one for
sched_waking and another for sched_switch, keyed on pid. Whenever
a sched_waking event occurs, the timestamp is saved in the entry
corresponding to the current pid, and when the scheduler switches
back to that pid, the timestamp difference is calculated. If the
resulting latency, stored in wakeup_lat, exceeds the current
maximum latency, a snapshot is taken. As part of the setup, all
the scheduler events are also enabled, which are the events that
will show up in the snapshot when it is taken at some point::
# echo 1 > /sys/kernel/tracing/events/sched/enable
# echo 'hist:keys=pid:ts0=common_timestamp.usecs \
if comm=="cyclictest"' >> \
/sys/kernel/tracing/events/sched/sched_waking/trigger
# echo 'hist:keys=next_pid:wakeup_lat=common_timestamp.usecs-$ts0: \
onmax($wakeup_lat).save(next_prio,next_comm,prev_pid,prev_prio, \
prev_comm):onmax($wakeup_lat).snapshot() \
if next_comm=="cyclictest"' >> \
/sys/kernel/tracing/events/sched/sched_switch/trigger
When the histogram is displayed, for each bucket the max value
and the saved values corresponding to the max are displayed
following the rest of the fields.
If a snapshot was taken, there is also a message indicating that,
along with the value and event that triggered the global maximum::
# cat /sys/kernel/tracing/events/sched/sched_switch/hist
{ next_pid: 2101 } hitcount: 200
max: 52 next_prio: 120 next_comm: cyclictest \
prev_pid: 0 prev_prio: 120 prev_comm: swapper/6
{ next_pid: 2103 } hitcount: 1326
max: 572 next_prio: 19 next_comm: cyclictest \
prev_pid: 0 prev_prio: 120 prev_comm: swapper/1
{ next_pid: 2102 } hitcount: 1982 \
max: 74 next_prio: 19 next_comm: cyclictest \
prev_pid: 0 prev_prio: 120 prev_comm: swapper/5
Snapshot taken (see tracing/snapshot). Details:
triggering value { onmax($wakeup_lat) }: 572 \
triggered by event with key: { next_pid: 2103 }
Totals:
Hits: 3508
Entries: 3
Dropped: 0
In the above case, the event that triggered the global maximum has
the key with next_pid == 2103. If you look at the bucket that has
2103 as the key, you'll find the additional values save()'d along
with the local maximum for that bucket, which should be the same
as the global maximum (since that was the same value that
triggered the global snapshot).
And finally, looking at the snapshot data should show at or near
the end the event that triggered the snapshot (in this case you
can verify the timestamps between the sched_waking and
sched_switch events, which should match the time displayed in the
global maximum)::
# cat /sys/kernel/tracing/snapshot
<...>-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
<idle>-0 [005] d.h3 309.873611: sched_waking: comm=cyclictest pid=2102 prio=19 target_cpu=005
<idle>-0 [005] dNh4 309.873613: sched_wakeup: comm=cyclictest pid=2102 prio=19 target_cpu=005
<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
<...>-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
<idle>-0 [005] d.h3 309.874624: sched_waking: comm=cyclictest pid=2102 prio=19 target_cpu=005
<idle>-0 [005] dNh4 309.874626: sched_wakeup: comm=cyclictest pid=2102 prio=19 target_cpu=005
<idle>-0 [005] dNh3 309.874628: sched_waking: comm=cyclictest pid=2103 prio=19 target_cpu=005
<idle>-0 [005] dNh4 309.874630: sched_wakeup: comm=cyclictest pid=2103 prio=19 target_cpu=005
<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
<idle>-0 [004] d.h3 309.874757: sched_waking: comm=gnome-terminal- pid=1699 prio=120 target_cpu=004
<idle>-0 [004] dNh4 309.874762: sched_wakeup: comm=gnome-terminal- pid=1699 prio=120 target_cpu=004
<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
gnome-terminal--1699 [004] d.h2 309.874941: sched_stat_runtime: comm=gnome-terminal- pid=1699 runtime=180706 [ns] vruntime=1126870572 [ns]
<idle>-0 [003] d.s4 309.874956: sched_waking: comm=rcu_sched pid=9 prio=120 target_cpu=007
<idle>-0 [003] d.s5 309.874960: sched_wake_idle_without_ipi: cpu=7
<idle>-0 [003] d.s5 309.874961: sched_wakeup: comm=rcu_sched pid=9 prio=120 target_cpu=007
<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
rcu_sched-9 [007] d..3 309.874973: sched_stat_runtime: comm=rcu_sched pid=9 runtime=13646 [ns] vruntime=22531430286 [ns]
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
<...>-2102 [005] d..4 309.874994: sched_migrate_task: comm=cyclictest pid=2103 prio=19 orig_cpu=5 dest_cpu=1
<...>-2102 [005] d..4 309.875185: sched_wake_idle_without_ipi: cpu=1
<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
`onchange` save와 snapshot
2430-2528`onchange(var).save(field,...)` action은 histogram entry에 연결된 `var` 값이 바뀔 때마다 호출된다.
해당 entry에서 `var`가 변하면 `save()` parameter로 지정한 trace event field를 저장한다. 값을 바꾼 event의 context를 나중에 참조할 수 있으며 histogram에 저장된 값을 보여 주는 추가 field가 표시된다.
`onchange(var).snapshot()` action도 histogram entry의 `var` 값이 바뀔 때마다 호출된다.
어떤 hist trigger entry에서든 `var`가 바뀌면 trace buffer의 global snapshot을 `tracing/snapshot` 파일에 저장한다.
이때 changed value는 current trace instance에 연결된 global variable이다. 변경을 일으킨 trace event의 key와 global value, snapshot 생성 메시지와 위치가 표시된다. key를 사용해 대응 histogram bucket을 찾으면 세부 context를 볼 수 있다.
다음 예제는 `tcp_probe` event에 dport key의 hist trigger를 정의한다. event가 발생할 때마다 `snd_cwnd`를 `$cwnd`의 현재 값과 비교하고 바뀌면 snapshot을 만든다. snapshot에 포함시키기 위해 scheduler와 tcp event도 모두 활성화한다.
# echo 1 > /sys/kernel/tracing/events/sched/enable
# echo 1 > /sys/kernel/tracing/events/tcp/enable
# echo 'hist:keys=dport:cwnd=snd_cwnd: \
onchange($cwnd).save(snd_wnd,srtt,rcv_wnd): \
onchange($cwnd).snapshot()' >> \
/sys/kernel/tracing/events/tcp/tcp_probe/triggerhistogram에는 각 bucket의 tracked value와 그 값에 대응해 저장된 field가 나머지 field 뒤에 표시된다.
snapshot이 생성됐다면 이를 알리는 메시지와 변경을 일으킨 value 및 event도 나타난다.
# cat /sys/kernel/tracing/events/tcp/tcp_probe/hist
{ dport: 1521 } hitcount: 8
changed: 10 snd_wnd: 35456 srtt: 154262 rcv_wnd: 42112
{ dport: 80 } hitcount: 23
changed: 10 snd_wnd: 28960 srtt: 19604 rcv_wnd: 29312
{ dport: 9001 } hitcount: 172
changed: 10 snd_wnd: 48384 srtt: 260444 rcv_wnd: 55168
{ dport: 443 } hitcount: 211
changed: 10 snd_wnd: 26960 srtt: 17379 rcv_wnd: 28800
Snapshot taken (see tracing/snapshot). Details:
triggering value { onchange($cwnd) }: 10
triggered by event with key: { dport: 80 }
Totals:
Hits: 414
Entries: 4
Dropped: 0예제에서 snapshot을 만든 event key는 `dport == 80`이다. key가 80인 bucket에는 해당 bucket의 changed value와 함께 `save()`한 추가 값이 있다. global snapshot을 발생시킨 값과 같으므로 bucket의 changed value도 global changed value와 같다.
snapshot data 끝부분 근처에는 snapshot을 발생시킨 `tcp_probe` event가 보여야 한다.
# cat /sys/kernel/tracing/snapshot
gnome-shell-1261 [006] dN.3 49.823113: sched_stat_runtime: comm=gnome-shell pid=1261 runtime=49347 [ns] vruntime=1835730389 [ns]
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
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
kworker/3:2-135 [003] d..3 49.823118: sched_stat_runtime: comm=kworker/3:2 pid=135 runtime=5339 [ns] vruntime=17815800388 [ns]
kworker/6:2-387 [006] d..3 49.823120: sched_stat_runtime: comm=kworker/6:2 pid=387 runtime=9594 [ns] vruntime=14589605367 [ns]
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
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
<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=29312handler호출 조건대표 용도
`onmax(var)`현재 maximum 초과최악 latency context와 snapshot
`onchange(var)`이전 값과 달라짐TCP cwnd 변화 context와 snapshot
두 handler는 context 저장과 snapshot action을 공유하지만 호출 조건이 다르다.
tcp_probe→dport key
cwnd = snd_cwnd→이전 $cwnd와 비교
값 변경→snd_wnd, srtt, rcv_wnd 저장
onchange().snapshot()→global snapshot
dport key→hist bucket과 tcp_probe event 대조
connection별 cwnd 변화가 context 저장과 global trace snapshot을 함께 유발한다.
- onchange(var).save(field,.. .)
The 'onchange(var).save(field,...)' hist trigger action is invoked
whenever the value of 'var' associated with a histogram entry
changes.
The end result is that the trace event fields specified as the
onchange.save() params will be saved if 'var' changes for that
hist trigger entry. This allows context from the event that
changed the value to be saved for later reference. When the
histogram is displayed, additional fields displaying the saved
values will be printed.
- onchange(var).snapshot()
The 'onchange(var).snapshot()' hist trigger action is invoked
whenever the value of 'var' associated with a histogram entry
changes.
The end result is that a global snapshot of the trace buffer will
be saved in the tracing/snapshot file if 'var' changes for any
hist trigger entry.
Note that in this case the changed value is a global variable
associated with current trace instance. The key of the specific
trace event that caused the value to change and the global value
itself are displayed, along with a message stating that a snapshot
has been taken and where to find it. The user can use the key
information displayed to locate the corresponding bucket in the
histogram for even more detail.
As an example the below defines a hist trigger on the tcp_probe
event, keyed on dport. Whenever a tcp_probe event occurs, the
cwnd field is checked against the current value stored in the
$cwnd variable. If the value has changed, a snapshot is taken.
As part of the setup, all the scheduler and tcp events are also
enabled, which are the events that will show up in the snapshot
when it is taken at some point::
# echo 1 > /sys/kernel/tracing/events/sched/enable
# echo 1 > /sys/kernel/tracing/events/tcp/enable
# echo 'hist:keys=dport:cwnd=snd_cwnd: \
onchange($cwnd).save(snd_wnd,srtt,rcv_wnd): \
onchange($cwnd).snapshot()' >> \
/sys/kernel/tracing/events/tcp/tcp_probe/trigger
When the histogram is displayed, for each bucket the tracked value
and the saved values corresponding to that value are displayed
following the rest of the fields.
If a snapshot was taken, there is also a message indicating that,
along with the value and event that triggered the snapshot::
# cat /sys/kernel/tracing/events/tcp/tcp_probe/hist
{ dport: 1521 } hitcount: 8
changed: 10 snd_wnd: 35456 srtt: 154262 rcv_wnd: 42112
{ dport: 80 } hitcount: 23
changed: 10 snd_wnd: 28960 srtt: 19604 rcv_wnd: 29312
{ dport: 9001 } hitcount: 172
changed: 10 snd_wnd: 48384 srtt: 260444 rcv_wnd: 55168
{ dport: 443 } hitcount: 211
changed: 10 snd_wnd: 26960 srtt: 17379 rcv_wnd: 28800
Snapshot taken (see tracing/snapshot). Details:
triggering value { onchange($cwnd) }: 10
triggered by event with key: { dport: 80 }
Totals:
Hits: 414
Entries: 4
Dropped: 0
In the above case, the event that triggered the snapshot has the
key with dport == 80. If you look at the bucket that has 80 as
the key, you'll find the additional values save()'d along with the
changed value for that bucket, which should be the same as the
global changed value (since that was the same value that triggered
the global snapshot).
And finally, looking at the snapshot data should show at or near
the end the event that triggered the snapshot::
# cat /sys/kernel/tracing/snapshot
gnome-shell-1261 [006] dN.3 49.823113: sched_stat_runtime: comm=gnome-shell pid=1261 runtime=49347 [ns] vruntime=1835730389 [ns]
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
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
kworker/3:2-135 [003] d..3 49.823118: sched_stat_runtime: comm=kworker/3:2 pid=135 runtime=5339 [ns] vruntime=17815800388 [ns]
kworker/6:2-387 [006] d..3 49.823120: sched_stat_runtime: comm=kworker/6:2 pid=387 runtime=9594 [ns] vruntime=14589605367 [ns]
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
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
<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
사용자 공간에서 trigger 만들기
2529-3071`/sys/kernel/tracing/trace_marker`에 쓴 내용은 ftrace ring buffer에 기록된다. `/sys/kernel/tracing/events/ftrace/print/` 아래의 `trigger` 파일을 이용하면 이 쓰기를 event처럼 동작하게 할 수도 있다.
예를 들어 `cyclictest`가 잠들기 전과 깨어난 뒤 `trace_marker` 파일에 문자열을 쓰도록 다음과 비슷한 함수를 추가한다.
static void traceputs(char *str)
{
/* tracemark_fd is the trace_marker file descriptor */
if (tracemark_fd < 0)
return;
/* write the tracemark message */
write(tracemark_fd, str, strlen(str));
}그리고 실제 sleep 호출 앞뒤에 다음과 같은 표시를 넣는다.
traceputs("start");
clock_nanosleep(...);
traceputs("end");이 두 표시로 latency histogram을 만들 수 있다.
# cd /sys/kernel/tracing
# echo 'latency u64 lat' > synthetic_events
# echo 'hist:keys=common_pid:ts0=common_timestamp.usecs if buf == "start"' > events/ftrace/print/trigger
# echo 'hist:keys=common_pid:lat=common_timestamp.usecs-$ts0:onmatch(ftrace.print).latency($lat) if buf == "end"' >> events/ftrace/print/trigger
# echo 'hist:keys=lat,common_pid:sort=lat' > events/synthetic/latency/trigger위 명령은 `latency`라는 synthetic event와 `trace_marker`를 대상으로 하는 histogram 두 개를 만든다. 하나는 `start`가 쓰일 때, 다른 하나는 `end`가 쓰일 때 작동한다. pid가 일치하면 계산한 latency를 parameter로 넘겨 `latency` synthetic event를 호출한다. 마지막 histogram은 계산된 latency와 pid를 함께 기록한다.
trace_marker: start→pid별 ts0 저장
trace_marker: end→현재 시각 - $ts0
onmatch(ftrace.print)→latency($lat) 호출
synthetic/latency→lat + common_pid 집계
동일 pid의 start와 end 표시를 histogram variable과 synthetic event로 연결한다.
이제 다음 parameter로 `cyclictest`를 실행한다.
# ./cyclictest -p80 -d0 -i250 -n -a -t --tracemark -b 1000option의미
`-p80`thread를 priority 80으로 실행
`-d0`모든 thread가 같은 interval을 사용
`-i250`모든 thread의 interval을 250 microseconds로 시작
`-n``nanosleep`으로 sleep
`-a`각 thread를 별도 CPU에 고정
`-t`사용 가능한 CPU마다 thread 하나를 생성
`--tracemark`trace mark 쓰기를 활성화
`-b 1000`latency가 1000 microseconds를 넘으면 중지
예제 명령의 각 option이 thread 배치와 측정 조건을 정한다.
여기서 `-b 1000`은 `--tracemark`를 사용할 수 있게 하기 위해 지정한다.
생성된 histogram은 다음과 같이 확인한다.
# cat events/synthetic/latency/hist
# event histogram
#
# trigger info: hist:keys=lat,common_pid:vals=hitcount:sort=lat:size=2048 [active]
#
{ lat: 107, common_pid: 2039 } hitcount: 1
{ lat: 122, common_pid: 2041 } hitcount: 1
{ lat: 166, common_pid: 2039 } hitcount: 1
{ lat: 174, common_pid: 2039 } hitcount: 1
{ lat: 194, common_pid: 2041 } hitcount: 1
{ lat: 196, common_pid: 2036 } hitcount: 1
{ lat: 197, common_pid: 2038 } hitcount: 1
{ lat: 198, common_pid: 2039 } hitcount: 1
{ lat: 199, common_pid: 2039 } hitcount: 1
{ lat: 200, common_pid: 2041 } hitcount: 1
{ lat: 201, common_pid: 2039 } hitcount: 2
{ lat: 202, common_pid: 2038 } hitcount: 1
{ lat: 202, common_pid: 2043 } hitcount: 1
{ lat: 203, common_pid: 2039 } hitcount: 1
{ lat: 203, common_pid: 2036 } hitcount: 1
{ lat: 203, common_pid: 2041 } hitcount: 1
{ lat: 206, common_pid: 2038 } hitcount: 2
{ lat: 207, common_pid: 2039 } hitcount: 1
{ lat: 207, common_pid: 2036 } hitcount: 1
{ lat: 208, common_pid: 2040 } hitcount: 1
{ lat: 209, common_pid: 2043 } hitcount: 1
{ lat: 210, common_pid: 2039 } hitcount: 1
{ lat: 211, common_pid: 2039 } hitcount: 4
{ lat: 212, common_pid: 2043 } hitcount: 1
{ lat: 212, common_pid: 2039 } hitcount: 2
{ lat: 213, common_pid: 2039 } hitcount: 1
{ lat: 214, common_pid: 2038 } hitcount: 1
{ lat: 214, common_pid: 2039 } hitcount: 2
{ lat: 214, common_pid: 2042 } hitcount: 1
{ lat: 215, common_pid: 2039 } hitcount: 1
{ lat: 217, common_pid: 2036 } hitcount: 1
{ lat: 217, common_pid: 2040 } hitcount: 1
{ lat: 217, common_pid: 2039 } hitcount: 1
{ lat: 218, common_pid: 2039 } hitcount: 6
{ lat: 219, common_pid: 2039 } hitcount: 9
{ lat: 220, common_pid: 2039 } hitcount: 11
{ lat: 221, common_pid: 2039 } hitcount: 5
{ lat: 221, common_pid: 2042 } hitcount: 1
{ lat: 222, common_pid: 2039 } hitcount: 7
{ lat: 223, common_pid: 2036 } hitcount: 1
{ lat: 223, common_pid: 2039 } hitcount: 3
{ lat: 224, common_pid: 2039 } hitcount: 4
{ lat: 224, common_pid: 2037 } hitcount: 1
{ lat: 224, common_pid: 2036 } hitcount: 2
{ lat: 225, common_pid: 2039 } hitcount: 5
{ lat: 225, common_pid: 2042 } hitcount: 1
{ lat: 226, common_pid: 2039 } hitcount: 7
{ lat: 226, common_pid: 2036 } hitcount: 4
{ lat: 227, common_pid: 2039 } hitcount: 6
{ lat: 227, common_pid: 2036 } hitcount: 12
{ lat: 227, common_pid: 2043 } hitcount: 1
{ lat: 228, common_pid: 2039 } hitcount: 7
{ lat: 228, common_pid: 2036 } hitcount: 14
{ lat: 229, common_pid: 2039 } hitcount: 9
{ lat: 229, common_pid: 2036 } hitcount: 8
{ lat: 229, common_pid: 2038 } hitcount: 1
{ lat: 230, common_pid: 2039 } hitcount: 11
{ lat: 230, common_pid: 2036 } hitcount: 6
{ lat: 230, common_pid: 2043 } hitcount: 1
{ lat: 230, common_pid: 2042 } hitcount: 2
{ lat: 231, common_pid: 2041 } hitcount: 1
{ lat: 231, common_pid: 2036 } hitcount: 6
{ lat: 231, common_pid: 2043 } hitcount: 1
{ lat: 231, common_pid: 2039 } hitcount: 8
{ lat: 232, common_pid: 2037 } hitcount: 1
{ lat: 232, common_pid: 2039 } hitcount: 6
{ lat: 232, common_pid: 2040 } hitcount: 2
{ lat: 232, common_pid: 2036 } hitcount: 5
{ lat: 232, common_pid: 2043 } hitcount: 1
{ lat: 233, common_pid: 2036 } hitcount: 5
{ lat: 233, common_pid: 2039 } hitcount: 11
{ lat: 234, common_pid: 2039 } hitcount: 4
{ lat: 234, common_pid: 2038 } hitcount: 2
{ lat: 234, common_pid: 2043 } hitcount: 2
{ lat: 234, common_pid: 2036 } hitcount: 11
{ lat: 234, common_pid: 2040 } hitcount: 1
{ lat: 235, common_pid: 2037 } hitcount: 2
{ lat: 235, common_pid: 2036 } hitcount: 8
{ lat: 235, common_pid: 2043 } hitcount: 2
{ lat: 235, common_pid: 2039 } hitcount: 5
{ lat: 235, common_pid: 2042 } hitcount: 2
{ lat: 235, common_pid: 2040 } hitcount: 4
{ lat: 235, common_pid: 2041 } hitcount: 1
{ lat: 236, common_pid: 2036 } hitcount: 7
{ lat: 236, common_pid: 2037 } hitcount: 1
{ lat: 236, common_pid: 2041 } hitcount: 5
{ lat: 236, common_pid: 2039 } hitcount: 3
{ lat: 236, common_pid: 2043 } hitcount: 9
{ lat: 236, common_pid: 2040 } hitcount: 7
{ lat: 237, common_pid: 2037 } hitcount: 1
{ lat: 237, common_pid: 2040 } hitcount: 1
{ lat: 237, common_pid: 2036 } hitcount: 9
{ lat: 237, common_pid: 2039 } hitcount: 3
{ lat: 237, common_pid: 2043 } hitcount: 8
{ lat: 237, common_pid: 2042 } hitcount: 2
{ lat: 237, common_pid: 2041 } hitcount: 2
{ lat: 238, common_pid: 2043 } hitcount: 10
{ lat: 238, common_pid: 2040 } hitcount: 1
{ lat: 238, common_pid: 2037 } hitcount: 9
{ lat: 238, common_pid: 2038 } hitcount: 1
{ lat: 238, common_pid: 2039 } hitcount: 1
{ lat: 238, common_pid: 2042 } hitcount: 3
{ lat: 238, common_pid: 2036 } hitcount: 7
{ lat: 239, common_pid: 2041 } hitcount: 1
{ lat: 239, common_pid: 2043 } hitcount: 11
{ lat: 239, common_pid: 2037 } hitcount: 11
{ lat: 239, common_pid: 2038 } hitcount: 6
{ lat: 239, common_pid: 2036 } hitcount: 7
{ lat: 239, common_pid: 2040 } hitcount: 1
{ lat: 239, common_pid: 2042 } hitcount: 9
{ lat: 240, common_pid: 2037 } hitcount: 29
{ lat: 240, common_pid: 2043 } hitcount: 15
{ lat: 240, common_pid: 2040 } hitcount: 44
{ lat: 240, common_pid: 2039 } hitcount: 1
{ lat: 240, common_pid: 2041 } hitcount: 2
{ lat: 240, common_pid: 2038 } hitcount: 1
{ lat: 240, common_pid: 2036 } hitcount: 10
{ lat: 240, common_pid: 2042 } hitcount: 13
{ lat: 241, common_pid: 2036 } hitcount: 21
{ lat: 241, common_pid: 2041 } hitcount: 36
{ lat: 241, common_pid: 2037 } hitcount: 34
{ lat: 241, common_pid: 2042 } hitcount: 14
{ lat: 241, common_pid: 2040 } hitcount: 94
{ lat: 241, common_pid: 2039 } hitcount: 12
{ lat: 241, common_pid: 2038 } hitcount: 2
{ lat: 241, common_pid: 2043 } hitcount: 28
{ lat: 242, common_pid: 2040 } hitcount: 109
{ lat: 242, common_pid: 2041 } hitcount: 506
{ lat: 242, common_pid: 2039 } hitcount: 155
{ lat: 242, common_pid: 2042 } hitcount: 21
{ lat: 242, common_pid: 2037 } hitcount: 52
{ lat: 242, common_pid: 2043 } hitcount: 21
{ lat: 242, common_pid: 2036 } hitcount: 16
{ lat: 242, common_pid: 2038 } hitcount: 156
{ lat: 243, common_pid: 2037 } hitcount: 46
{ lat: 243, common_pid: 2039 } hitcount: 40
{ lat: 243, common_pid: 2042 } hitcount: 119
{ lat: 243, common_pid: 2041 } hitcount: 611
{ lat: 243, common_pid: 2036 } hitcount: 69
{ lat: 243, common_pid: 2038 } hitcount: 784
{ lat: 243, common_pid: 2040 } hitcount: 323
{ lat: 243, common_pid: 2043 } hitcount: 14
{ lat: 244, common_pid: 2043 } hitcount: 35
{ lat: 244, common_pid: 2042 } hitcount: 305
{ lat: 244, common_pid: 2039 } hitcount: 8
{ lat: 244, common_pid: 2040 } hitcount: 4515
{ lat: 244, common_pid: 2038 } hitcount: 371
{ lat: 244, common_pid: 2037 } hitcount: 31
{ lat: 244, common_pid: 2036 } hitcount: 114
{ lat: 244, common_pid: 2041 } hitcount: 3396
{ lat: 245, common_pid: 2036 } hitcount: 700
{ lat: 245, common_pid: 2041 } hitcount: 2772
{ lat: 245, common_pid: 2037 } hitcount: 268
{ lat: 245, common_pid: 2039 } hitcount: 472
{ lat: 245, common_pid: 2038 } hitcount: 2758
{ lat: 245, common_pid: 2042 } hitcount: 3833
{ lat: 245, common_pid: 2040 } hitcount: 3105
{ lat: 245, common_pid: 2043 } hitcount: 645
{ lat: 246, common_pid: 2038 } hitcount: 3451
{ lat: 246, common_pid: 2041 } hitcount: 142
{ lat: 246, common_pid: 2037 } hitcount: 5101
{ lat: 246, common_pid: 2040 } hitcount: 68
{ lat: 246, common_pid: 2043 } hitcount: 5099
{ lat: 246, common_pid: 2039 } hitcount: 5608
{ lat: 246, common_pid: 2042 } hitcount: 3723
{ lat: 246, common_pid: 2036 } hitcount: 4738
{ lat: 247, common_pid: 2042 } hitcount: 312
{ lat: 247, common_pid: 2043 } hitcount: 2385
{ lat: 247, common_pid: 2041 } hitcount: 452
{ lat: 247, common_pid: 2038 } hitcount: 792
{ lat: 247, common_pid: 2040 } hitcount: 78
{ lat: 247, common_pid: 2036 } hitcount: 2375
{ lat: 247, common_pid: 2039 } hitcount: 1834
{ lat: 247, common_pid: 2037 } hitcount: 2655
{ lat: 248, common_pid: 2037 } hitcount: 36
{ lat: 248, common_pid: 2042 } hitcount: 11
{ lat: 248, common_pid: 2038 } hitcount: 122
{ lat: 248, common_pid: 2036 } hitcount: 135
{ lat: 248, common_pid: 2039 } hitcount: 26
{ lat: 248, common_pid: 2041 } hitcount: 503
{ lat: 248, common_pid: 2043 } hitcount: 66
{ lat: 248, common_pid: 2040 } hitcount: 46
{ lat: 249, common_pid: 2037 } hitcount: 29
{ lat: 249, common_pid: 2038 } hitcount: 1
{ lat: 249, common_pid: 2043 } hitcount: 29
{ lat: 249, common_pid: 2039 } hitcount: 8
{ lat: 249, common_pid: 2042 } hitcount: 56
{ lat: 249, common_pid: 2040 } hitcount: 27
{ lat: 249, common_pid: 2041 } hitcount: 11
{ lat: 249, common_pid: 2036 } hitcount: 27
{ lat: 250, common_pid: 2038 } hitcount: 1
{ lat: 250, common_pid: 2036 } hitcount: 30
{ lat: 250, common_pid: 2040 } hitcount: 19
{ lat: 250, common_pid: 2043 } hitcount: 22
{ lat: 250, common_pid: 2042 } hitcount: 20
{ lat: 250, common_pid: 2041 } hitcount: 1
{ lat: 250, common_pid: 2039 } hitcount: 6
{ lat: 250, common_pid: 2037 } hitcount: 48
{ lat: 251, common_pid: 2037 } hitcount: 43
{ lat: 251, common_pid: 2039 } hitcount: 1
{ lat: 251, common_pid: 2036 } hitcount: 12
{ lat: 251, common_pid: 2042 } hitcount: 2
{ lat: 251, common_pid: 2041 } hitcount: 1
{ lat: 251, common_pid: 2043 } hitcount: 15
{ lat: 251, common_pid: 2040 } hitcount: 3
{ lat: 252, common_pid: 2040 } hitcount: 1
{ lat: 252, common_pid: 2036 } hitcount: 12
{ lat: 252, common_pid: 2037 } hitcount: 21
{ lat: 252, common_pid: 2043 } hitcount: 14
{ lat: 253, common_pid: 2037 } hitcount: 21
{ lat: 253, common_pid: 2039 } hitcount: 2
{ lat: 253, common_pid: 2036 } hitcount: 9
{ lat: 253, common_pid: 2043 } hitcount: 6
{ lat: 253, common_pid: 2040 } hitcount: 1
{ lat: 254, common_pid: 2036 } hitcount: 8
{ lat: 254, common_pid: 2043 } hitcount: 3
{ lat: 254, common_pid: 2041 } hitcount: 1
{ lat: 254, common_pid: 2042 } hitcount: 1
{ lat: 254, common_pid: 2039 } hitcount: 1
{ lat: 254, common_pid: 2037 } hitcount: 12
{ lat: 255, common_pid: 2043 } hitcount: 1
{ lat: 255, common_pid: 2037 } hitcount: 2
{ lat: 255, common_pid: 2036 } hitcount: 2
{ lat: 255, common_pid: 2039 } hitcount: 8
{ lat: 256, common_pid: 2043 } hitcount: 1
{ lat: 256, common_pid: 2036 } hitcount: 4
{ lat: 256, common_pid: 2039 } hitcount: 6
{ lat: 257, common_pid: 2039 } hitcount: 5
{ lat: 257, common_pid: 2036 } hitcount: 4
{ lat: 258, common_pid: 2039 } hitcount: 5
{ lat: 258, common_pid: 2036 } hitcount: 2
{ lat: 259, common_pid: 2036 } hitcount: 7
{ lat: 259, common_pid: 2039 } hitcount: 7
{ lat: 260, common_pid: 2036 } hitcount: 8
{ lat: 260, common_pid: 2039 } hitcount: 6
{ lat: 261, common_pid: 2036 } hitcount: 5
{ lat: 261, common_pid: 2039 } hitcount: 7
{ lat: 262, common_pid: 2039 } hitcount: 5
{ lat: 262, common_pid: 2036 } hitcount: 5
{ lat: 263, common_pid: 2039 } hitcount: 7
{ lat: 263, common_pid: 2036 } hitcount: 7
{ lat: 264, common_pid: 2039 } hitcount: 9
{ lat: 264, common_pid: 2036 } hitcount: 9
{ lat: 265, common_pid: 2036 } hitcount: 5
{ lat: 265, common_pid: 2039 } hitcount: 1
{ lat: 266, common_pid: 2036 } hitcount: 1
{ lat: 266, common_pid: 2039 } hitcount: 3
{ lat: 267, common_pid: 2036 } hitcount: 1
{ lat: 267, common_pid: 2039 } hitcount: 3
{ lat: 268, common_pid: 2036 } hitcount: 1
{ lat: 268, common_pid: 2039 } hitcount: 6
{ lat: 269, common_pid: 2036 } hitcount: 1
{ lat: 269, common_pid: 2043 } hitcount: 1
{ lat: 269, common_pid: 2039 } hitcount: 2
{ lat: 270, common_pid: 2040 } hitcount: 1
{ lat: 270, common_pid: 2039 } hitcount: 6
{ lat: 271, common_pid: 2041 } hitcount: 1
{ lat: 271, common_pid: 2039 } hitcount: 5
{ lat: 272, common_pid: 2039 } hitcount: 10
{ lat: 273, common_pid: 2039 } hitcount: 8
{ lat: 274, common_pid: 2039 } hitcount: 2
{ lat: 275, common_pid: 2039 } hitcount: 1
{ lat: 276, common_pid: 2039 } hitcount: 2
{ lat: 276, common_pid: 2037 } hitcount: 1
{ lat: 276, common_pid: 2038 } hitcount: 1
{ lat: 277, common_pid: 2039 } hitcount: 1
{ lat: 277, common_pid: 2042 } hitcount: 1
{ lat: 278, common_pid: 2039 } hitcount: 1
{ lat: 279, common_pid: 2039 } hitcount: 4
{ lat: 279, common_pid: 2043 } hitcount: 1
{ lat: 280, common_pid: 2039 } hitcount: 3
{ lat: 283, common_pid: 2036 } hitcount: 2
{ lat: 284, common_pid: 2039 } hitcount: 1
{ lat: 284, common_pid: 2043 } hitcount: 1
{ lat: 288, common_pid: 2039 } hitcount: 1
{ lat: 289, common_pid: 2039 } hitcount: 1
{ lat: 300, common_pid: 2039 } hitcount: 1
{ lat: 384, common_pid: 2039 } hitcount: 1
Totals:
Hits: 67625
Entries: 278
Dropped: 0쓰기는 sleep 바로 앞뒤에서 일어나므로 이상적으로 모든 latency는 250 microseconds여야 한다. 250보다 작은 값도 있는 이유는 `cyclictest`의 동작 방식 때문이다. 한 iteration이 늦게 끝나면 다음 timer는 그만큼 더 일찍 깨도록 설정된다. 예를 들어 한 iteration이 50 microseconds 늦었다면 다음 wakeup은 200 microseconds 뒤에 일어난다.
이 측정은 사용자 공간만으로도 할 수 있다. 더 흥미로운 구성으로, kernel에서 발생한 event와 `trace_marker`를 한 histogram에서 결합할 수 있다.
# cd /sys/kernel/tracing
# echo 'latency u64 lat' > synthetic_events
# echo 'hist:keys=pid:ts0=common_timestamp.usecs' > events/sched/sched_waking/trigger
# echo 'hist:keys=common_pid:lat=common_timestamp.usecs-$ts0:onmatch(sched.sched_waking).latency($lat) if buf == "end"' > events/ftrace/print/trigger
# echo 'hist:keys=lat,common_pid:sort=lat' > events/synthetic/latency/trigger이번에는 `trace_marker`로 latency 시작 시각을 잡는 대신 `sched_waking` event를 사용한다. `sched_waking`이 깨우려는 pid와 `trace_marker` 쓰기의 `common_pid`를 맞춘다.
같은 parameter로 `cyclictest`를 다시 실행하면 다음 결과를 얻는다.
# cat events/synthetic/latency/hist
# event histogram
#
# trigger info: hist:keys=lat,common_pid:vals=hitcount:sort=lat:size=2048 [active]
#
{ lat: 7, common_pid: 2302 } hitcount: 640
{ lat: 7, common_pid: 2299 } hitcount: 42
{ lat: 7, common_pid: 2303 } hitcount: 18
{ lat: 7, common_pid: 2305 } hitcount: 166
{ lat: 7, common_pid: 2306 } hitcount: 1
{ lat: 7, common_pid: 2301 } hitcount: 91
{ lat: 7, common_pid: 2300 } hitcount: 17
{ lat: 8, common_pid: 2303 } hitcount: 8296
{ lat: 8, common_pid: 2304 } hitcount: 6864
{ lat: 8, common_pid: 2305 } hitcount: 9464
{ lat: 8, common_pid: 2301 } hitcount: 9213
{ lat: 8, common_pid: 2306 } hitcount: 6246
{ lat: 8, common_pid: 2302 } hitcount: 8797
{ lat: 8, common_pid: 2299 } hitcount: 8771
{ lat: 8, common_pid: 2300 } hitcount: 8119
{ lat: 9, common_pid: 2305 } hitcount: 1519
{ lat: 9, common_pid: 2299 } hitcount: 2346
{ lat: 9, common_pid: 2303 } hitcount: 2841
{ lat: 9, common_pid: 2301 } hitcount: 1846
{ lat: 9, common_pid: 2304 } hitcount: 3861
{ lat: 9, common_pid: 2302 } hitcount: 1210
{ lat: 9, common_pid: 2300 } hitcount: 2762
{ lat: 9, common_pid: 2306 } hitcount: 4247
{ lat: 10, common_pid: 2299 } hitcount: 16
{ lat: 10, common_pid: 2306 } hitcount: 333
{ lat: 10, common_pid: 2303 } hitcount: 16
{ lat: 10, common_pid: 2304 } hitcount: 168
{ lat: 10, common_pid: 2302 } hitcount: 240
{ lat: 10, common_pid: 2301 } hitcount: 28
{ lat: 10, common_pid: 2300 } hitcount: 95
{ lat: 10, common_pid: 2305 } hitcount: 18
{ lat: 11, common_pid: 2303 } hitcount: 5
{ lat: 11, common_pid: 2305 } hitcount: 8
{ lat: 11, common_pid: 2306 } hitcount: 221
{ lat: 11, common_pid: 2302 } hitcount: 76
{ lat: 11, common_pid: 2304 } hitcount: 26
{ lat: 11, common_pid: 2300 } hitcount: 125
{ lat: 11, common_pid: 2299 } hitcount: 2
{ lat: 12, common_pid: 2305 } hitcount: 3
{ lat: 12, common_pid: 2300 } hitcount: 6
{ lat: 12, common_pid: 2306 } hitcount: 90
{ lat: 12, common_pid: 2302 } hitcount: 4
{ lat: 12, common_pid: 2303 } hitcount: 1
{ lat: 12, common_pid: 2304 } hitcount: 122
{ lat: 13, common_pid: 2300 } hitcount: 12
{ lat: 13, common_pid: 2301 } hitcount: 1
{ lat: 13, common_pid: 2306 } hitcount: 32
{ lat: 13, common_pid: 2302 } hitcount: 5
{ lat: 13, common_pid: 2305 } hitcount: 1
{ lat: 13, common_pid: 2303 } hitcount: 1
{ lat: 13, common_pid: 2304 } hitcount: 61
{ lat: 14, common_pid: 2303 } hitcount: 4
{ lat: 14, common_pid: 2306 } hitcount: 5
{ lat: 14, common_pid: 2305 } hitcount: 4
{ lat: 14, common_pid: 2304 } hitcount: 62
{ lat: 14, common_pid: 2302 } hitcount: 19
{ lat: 14, common_pid: 2300 } hitcount: 33
{ lat: 14, common_pid: 2299 } hitcount: 1
{ lat: 14, common_pid: 2301 } hitcount: 4
{ lat: 15, common_pid: 2305 } hitcount: 1
{ lat: 15, common_pid: 2302 } hitcount: 25
{ lat: 15, common_pid: 2300 } hitcount: 11
{ lat: 15, common_pid: 2299 } hitcount: 5
{ lat: 15, common_pid: 2301 } hitcount: 1
{ lat: 15, common_pid: 2304 } hitcount: 8
{ lat: 15, common_pid: 2303 } hitcount: 1
{ lat: 15, common_pid: 2306 } hitcount: 6
{ lat: 16, common_pid: 2302 } hitcount: 31
{ lat: 16, common_pid: 2306 } hitcount: 3
{ lat: 16, common_pid: 2300 } hitcount: 5
{ lat: 17, common_pid: 2302 } hitcount: 6
{ lat: 17, common_pid: 2303 } hitcount: 1
{ lat: 18, common_pid: 2304 } hitcount: 1
{ lat: 18, common_pid: 2302 } hitcount: 8
{ lat: 18, common_pid: 2299 } hitcount: 1
{ lat: 18, common_pid: 2301 } hitcount: 1
{ lat: 19, common_pid: 2303 } hitcount: 4
{ lat: 19, common_pid: 2304 } hitcount: 5
{ lat: 19, common_pid: 2302 } hitcount: 4
{ lat: 19, common_pid: 2299 } hitcount: 3
{ lat: 19, common_pid: 2306 } hitcount: 1
{ lat: 19, common_pid: 2300 } hitcount: 4
{ lat: 19, common_pid: 2305 } hitcount: 5
{ lat: 20, common_pid: 2299 } hitcount: 2
{ lat: 20, common_pid: 2302 } hitcount: 3
{ lat: 20, common_pid: 2305 } hitcount: 1
{ lat: 20, common_pid: 2300 } hitcount: 2
{ lat: 20, common_pid: 2301 } hitcount: 2
{ lat: 20, common_pid: 2303 } hitcount: 3
{ lat: 21, common_pid: 2305 } hitcount: 1
{ lat: 21, common_pid: 2299 } hitcount: 5
{ lat: 21, common_pid: 2303 } hitcount: 4
{ lat: 21, common_pid: 2302 } hitcount: 7
{ lat: 21, common_pid: 2300 } hitcount: 1
{ lat: 21, common_pid: 2301 } hitcount: 5
{ lat: 21, common_pid: 2304 } hitcount: 2
{ lat: 22, common_pid: 2302 } hitcount: 5
{ lat: 22, common_pid: 2303 } hitcount: 1
{ lat: 22, common_pid: 2306 } hitcount: 3
{ lat: 22, common_pid: 2301 } hitcount: 2
{ lat: 22, common_pid: 2300 } hitcount: 1
{ lat: 22, common_pid: 2299 } hitcount: 1
{ lat: 22, common_pid: 2305 } hitcount: 1
{ lat: 22, common_pid: 2304 } hitcount: 1
{ lat: 23, common_pid: 2299 } hitcount: 1
{ lat: 23, common_pid: 2306 } hitcount: 2
{ lat: 23, common_pid: 2302 } hitcount: 6
{ lat: 24, common_pid: 2302 } hitcount: 3
{ lat: 24, common_pid: 2300 } hitcount: 1
{ lat: 24, common_pid: 2306 } hitcount: 2
{ lat: 24, common_pid: 2305 } hitcount: 1
{ lat: 24, common_pid: 2299 } hitcount: 1
{ lat: 25, common_pid: 2300 } hitcount: 1
{ lat: 25, common_pid: 2302 } hitcount: 4
{ lat: 26, common_pid: 2302 } hitcount: 2
{ lat: 27, common_pid: 2305 } hitcount: 1
{ lat: 27, common_pid: 2300 } hitcount: 1
{ lat: 27, common_pid: 2302 } hitcount: 3
{ lat: 28, common_pid: 2306 } hitcount: 1
{ lat: 28, common_pid: 2302 } hitcount: 4
{ lat: 29, common_pid: 2302 } hitcount: 1
{ lat: 29, common_pid: 2300 } hitcount: 2
{ lat: 29, common_pid: 2306 } hitcount: 1
{ lat: 29, common_pid: 2304 } hitcount: 1
{ lat: 30, common_pid: 2302 } hitcount: 4
{ lat: 31, common_pid: 2302 } hitcount: 6
{ lat: 32, common_pid: 2302 } hitcount: 1
{ lat: 33, common_pid: 2299 } hitcount: 1
{ lat: 33, common_pid: 2302 } hitcount: 3
{ lat: 34, common_pid: 2302 } hitcount: 2
{ lat: 35, common_pid: 2302 } hitcount: 1
{ lat: 35, common_pid: 2304 } hitcount: 1
{ lat: 36, common_pid: 2302 } hitcount: 4
{ lat: 37, common_pid: 2302 } hitcount: 6
{ lat: 38, common_pid: 2302 } hitcount: 2
{ lat: 39, common_pid: 2302 } hitcount: 2
{ lat: 39, common_pid: 2304 } hitcount: 1
{ lat: 40, common_pid: 2304 } hitcount: 2
{ lat: 40, common_pid: 2302 } hitcount: 5
{ lat: 41, common_pid: 2304 } hitcount: 1
{ lat: 41, common_pid: 2302 } hitcount: 8
{ lat: 42, common_pid: 2302 } hitcount: 6
{ lat: 42, common_pid: 2304 } hitcount: 1
{ lat: 43, common_pid: 2302 } hitcount: 3
{ lat: 43, common_pid: 2304 } hitcount: 4
{ lat: 44, common_pid: 2302 } hitcount: 6
{ lat: 45, common_pid: 2302 } hitcount: 5
{ lat: 46, common_pid: 2302 } hitcount: 5
{ lat: 47, common_pid: 2302 } hitcount: 7
{ lat: 48, common_pid: 2301 } hitcount: 1
{ lat: 48, common_pid: 2302 } hitcount: 9
{ lat: 49, common_pid: 2302 } hitcount: 3
{ lat: 50, common_pid: 2302 } hitcount: 1
{ lat: 50, common_pid: 2301 } hitcount: 1
{ lat: 51, common_pid: 2302 } hitcount: 2
{ lat: 51, common_pid: 2301 } hitcount: 1
{ lat: 61, common_pid: 2302 } hitcount: 1
{ lat: 110, common_pid: 2302 } hitcount: 1
Totals:
Hits: 89565
Entries: 158
Dropped: 0이 결과만으로 `cyclictest`가 예정 시각보다 얼마나 늦게 깨어났는지는 알 수 없다. 다만 `cyclictest`가 깨어난 시점부터 사용자 공간에 실제로 진입할 때까지 걸린 시간을 잘 보여 주는 histogram이다.
2.8. User space creating a trigger
----------------------------------
Writing into /sys/kernel/tracing/trace_marker writes into the ftrace
ring buffer. This can also act like an event, by writing into the trigger
file located in /sys/kernel/tracing/events/ftrace/print/
Modifying cyclictest to write into the trace_marker file before it sleeps
and after it wakes up, something like this::
static void traceputs(char *str)
{
/* tracemark_fd is the trace_marker file descriptor */
if (tracemark_fd < 0)
return;
/* write the tracemark message */
write(tracemark_fd, str, strlen(str));
}
And later add something like::
traceputs("start");
clock_nanosleep(...);
traceputs("end");
We can make a histogram from this::
# cd /sys/kernel/tracing
# echo 'latency u64 lat' > synthetic_events
# echo 'hist:keys=common_pid:ts0=common_timestamp.usecs if buf == "start"' > events/ftrace/print/trigger
# echo 'hist:keys=common_pid:lat=common_timestamp.usecs-$ts0:onmatch(ftrace.print).latency($lat) if buf == "end"' >> events/ftrace/print/trigger
# echo 'hist:keys=lat,common_pid:sort=lat' > events/synthetic/latency/trigger
The above created a synthetic event called "latency" and two histograms
against the trace_marker, one gets triggered when "start" is written into the
trace_marker file and the other when "end" is written. If the pids match, then
it will call the "latency" synthetic event with the calculated latency as its
parameter. Finally, a histogram is added to the latency synthetic event to
record the calculated latency along with the pid.
Now running cyclictest with::
# ./cyclictest -p80 -d0 -i250 -n -a -t --tracemark -b 1000
-p80 : run threads at priority 80
-d0 : have all threads run at the same interval
-i250 : start the interval at 250 microseconds (all threads will do this)
-n : sleep with nanosleep
-a : affine all threads to a separate CPU
-t : one thread per available CPU
--tracemark : enable trace mark writing
-b 1000 : stop if any latency is greater than 1000 microseconds
Note, the -b 1000 is used just to make --tracemark available.
Then we can see the histogram created by this with::
# cat events/synthetic/latency/hist
# event histogram
#
# trigger info: hist:keys=lat,common_pid:vals=hitcount:sort=lat:size=2048 [active]
#
{ lat: 107, common_pid: 2039 } hitcount: 1
{ lat: 122, common_pid: 2041 } hitcount: 1
{ lat: 166, common_pid: 2039 } hitcount: 1
{ lat: 174, common_pid: 2039 } hitcount: 1
{ lat: 194, common_pid: 2041 } hitcount: 1
{ lat: 196, common_pid: 2036 } hitcount: 1
{ lat: 197, common_pid: 2038 } hitcount: 1
{ lat: 198, common_pid: 2039 } hitcount: 1
{ lat: 199, common_pid: 2039 } hitcount: 1
{ lat: 200, common_pid: 2041 } hitcount: 1
{ lat: 201, common_pid: 2039 } hitcount: 2
{ lat: 202, common_pid: 2038 } hitcount: 1
{ lat: 202, common_pid: 2043 } hitcount: 1
{ lat: 203, common_pid: 2039 } hitcount: 1
{ lat: 203, common_pid: 2036 } hitcount: 1
{ lat: 203, common_pid: 2041 } hitcount: 1
{ lat: 206, common_pid: 2038 } hitcount: 2
{ lat: 207, common_pid: 2039 } hitcount: 1
{ lat: 207, common_pid: 2036 } hitcount: 1
{ lat: 208, common_pid: 2040 } hitcount: 1
{ lat: 209, common_pid: 2043 } hitcount: 1
{ lat: 210, common_pid: 2039 } hitcount: 1
{ lat: 211, common_pid: 2039 } hitcount: 4
{ lat: 212, common_pid: 2043 } hitcount: 1
{ lat: 212, common_pid: 2039 } hitcount: 2
{ lat: 213, common_pid: 2039 } hitcount: 1
{ lat: 214, common_pid: 2038 } hitcount: 1
{ lat: 214, common_pid: 2039 } hitcount: 2
{ lat: 214, common_pid: 2042 } hitcount: 1
{ lat: 215, common_pid: 2039 } hitcount: 1
{ lat: 217, common_pid: 2036 } hitcount: 1
{ lat: 217, common_pid: 2040 } hitcount: 1
{ lat: 217, common_pid: 2039 } hitcount: 1
{ lat: 218, common_pid: 2039 } hitcount: 6
{ lat: 219, common_pid: 2039 } hitcount: 9
{ lat: 220, common_pid: 2039 } hitcount: 11
{ lat: 221, common_pid: 2039 } hitcount: 5
{ lat: 221, common_pid: 2042 } hitcount: 1
{ lat: 222, common_pid: 2039 } hitcount: 7
{ lat: 223, common_pid: 2036 } hitcount: 1
{ lat: 223, common_pid: 2039 } hitcount: 3
{ lat: 224, common_pid: 2039 } hitcount: 4
{ lat: 224, common_pid: 2037 } hitcount: 1
{ lat: 224, common_pid: 2036 } hitcount: 2
{ lat: 225, common_pid: 2039 } hitcount: 5
{ lat: 225, common_pid: 2042 } hitcount: 1
{ lat: 226, common_pid: 2039 } hitcount: 7
{ lat: 226, common_pid: 2036 } hitcount: 4
{ lat: 227, common_pid: 2039 } hitcount: 6
{ lat: 227, common_pid: 2036 } hitcount: 12
{ lat: 227, common_pid: 2043 } hitcount: 1
{ lat: 228, common_pid: 2039 } hitcount: 7
{ lat: 228, common_pid: 2036 } hitcount: 14
{ lat: 229, common_pid: 2039 } hitcount: 9
{ lat: 229, common_pid: 2036 } hitcount: 8
{ lat: 229, common_pid: 2038 } hitcount: 1
{ lat: 230, common_pid: 2039 } hitcount: 11
{ lat: 230, common_pid: 2036 } hitcount: 6
{ lat: 230, common_pid: 2043 } hitcount: 1
{ lat: 230, common_pid: 2042 } hitcount: 2
{ lat: 231, common_pid: 2041 } hitcount: 1
{ lat: 231, common_pid: 2036 } hitcount: 6
{ lat: 231, common_pid: 2043 } hitcount: 1
{ lat: 231, common_pid: 2039 } hitcount: 8
{ lat: 232, common_pid: 2037 } hitcount: 1
{ lat: 232, common_pid: 2039 } hitcount: 6
{ lat: 232, common_pid: 2040 } hitcount: 2
{ lat: 232, common_pid: 2036 } hitcount: 5
{ lat: 232, common_pid: 2043 } hitcount: 1
{ lat: 233, common_pid: 2036 } hitcount: 5
{ lat: 233, common_pid: 2039 } hitcount: 11
{ lat: 234, common_pid: 2039 } hitcount: 4
{ lat: 234, common_pid: 2038 } hitcount: 2
{ lat: 234, common_pid: 2043 } hitcount: 2
{ lat: 234, common_pid: 2036 } hitcount: 11
{ lat: 234, common_pid: 2040 } hitcount: 1
{ lat: 235, common_pid: 2037 } hitcount: 2
{ lat: 235, common_pid: 2036 } hitcount: 8
{ lat: 235, common_pid: 2043 } hitcount: 2
{ lat: 235, common_pid: 2039 } hitcount: 5
{ lat: 235, common_pid: 2042 } hitcount: 2
{ lat: 235, common_pid: 2040 } hitcount: 4
{ lat: 235, common_pid: 2041 } hitcount: 1
{ lat: 236, common_pid: 2036 } hitcount: 7
{ lat: 236, common_pid: 2037 } hitcount: 1
{ lat: 236, common_pid: 2041 } hitcount: 5
{ lat: 236, common_pid: 2039 } hitcount: 3
{ lat: 236, common_pid: 2043 } hitcount: 9
{ lat: 236, common_pid: 2040 } hitcount: 7
{ lat: 237, common_pid: 2037 } hitcount: 1
{ lat: 237, common_pid: 2040 } hitcount: 1
{ lat: 237, common_pid: 2036 } hitcount: 9
{ lat: 237, common_pid: 2039 } hitcount: 3
{ lat: 237, common_pid: 2043 } hitcount: 8
{ lat: 237, common_pid: 2042 } hitcount: 2
{ lat: 237, common_pid: 2041 } hitcount: 2
{ lat: 238, common_pid: 2043 } hitcount: 10
{ lat: 238, common_pid: 2040 } hitcount: 1
{ lat: 238, common_pid: 2037 } hitcount: 9
{ lat: 238, common_pid: 2038 } hitcount: 1
{ lat: 238, common_pid: 2039 } hitcount: 1
{ lat: 238, common_pid: 2042 } hitcount: 3
{ lat: 238, common_pid: 2036 } hitcount: 7
{ lat: 239, common_pid: 2041 } hitcount: 1
{ lat: 239, common_pid: 2043 } hitcount: 11
{ lat: 239, common_pid: 2037 } hitcount: 11
{ lat: 239, common_pid: 2038 } hitcount: 6
{ lat: 239, common_pid: 2036 } hitcount: 7
{ lat: 239, common_pid: 2040 } hitcount: 1
{ lat: 239, common_pid: 2042 } hitcount: 9
{ lat: 240, common_pid: 2037 } hitcount: 29
{ lat: 240, common_pid: 2043 } hitcount: 15
{ lat: 240, common_pid: 2040 } hitcount: 44
{ lat: 240, common_pid: 2039 } hitcount: 1
{ lat: 240, common_pid: 2041 } hitcount: 2
{ lat: 240, common_pid: 2038 } hitcount: 1
{ lat: 240, common_pid: 2036 } hitcount: 10
{ lat: 240, common_pid: 2042 } hitcount: 13
{ lat: 241, common_pid: 2036 } hitcount: 21
{ lat: 241, common_pid: 2041 } hitcount: 36
{ lat: 241, common_pid: 2037 } hitcount: 34
{ lat: 241, common_pid: 2042 } hitcount: 14
{ lat: 241, common_pid: 2040 } hitcount: 94
{ lat: 241, common_pid: 2039 } hitcount: 12
{ lat: 241, common_pid: 2038 } hitcount: 2
{ lat: 241, common_pid: 2043 } hitcount: 28
{ lat: 242, common_pid: 2040 } hitcount: 109
{ lat: 242, common_pid: 2041 } hitcount: 506
{ lat: 242, common_pid: 2039 } hitcount: 155
{ lat: 242, common_pid: 2042 } hitcount: 21
{ lat: 242, common_pid: 2037 } hitcount: 52
{ lat: 242, common_pid: 2043 } hitcount: 21
{ lat: 242, common_pid: 2036 } hitcount: 16
{ lat: 242, common_pid: 2038 } hitcount: 156
{ lat: 243, common_pid: 2037 } hitcount: 46
{ lat: 243, common_pid: 2039 } hitcount: 40
{ lat: 243, common_pid: 2042 } hitcount: 119
{ lat: 243, common_pid: 2041 } hitcount: 611
{ lat: 243, common_pid: 2036 } hitcount: 69
{ lat: 243, common_pid: 2038 } hitcount: 784
{ lat: 243, common_pid: 2040 } hitcount: 323
{ lat: 243, common_pid: 2043 } hitcount: 14
{ lat: 244, common_pid: 2043 } hitcount: 35
{ lat: 244, common_pid: 2042 } hitcount: 305
{ lat: 244, common_pid: 2039 } hitcount: 8
{ lat: 244, common_pid: 2040 } hitcount: 4515
{ lat: 244, common_pid: 2038 } hitcount: 371
{ lat: 244, common_pid: 2037 } hitcount: 31
{ lat: 244, common_pid: 2036 } hitcount: 114
{ lat: 244, common_pid: 2041 } hitcount: 3396
{ lat: 245, common_pid: 2036 } hitcount: 700
{ lat: 245, common_pid: 2041 } hitcount: 2772
{ lat: 245, common_pid: 2037 } hitcount: 268
{ lat: 245, common_pid: 2039 } hitcount: 472
{ lat: 245, common_pid: 2038 } hitcount: 2758
{ lat: 245, common_pid: 2042 } hitcount: 3833
{ lat: 245, common_pid: 2040 } hitcount: 3105
{ lat: 245, common_pid: 2043 } hitcount: 645
{ lat: 246, common_pid: 2038 } hitcount: 3451
{ lat: 246, common_pid: 2041 } hitcount: 142
{ lat: 246, common_pid: 2037 } hitcount: 5101
{ lat: 246, common_pid: 2040 } hitcount: 68
{ lat: 246, common_pid: 2043 } hitcount: 5099
{ lat: 246, common_pid: 2039 } hitcount: 5608
{ lat: 246, common_pid: 2042 } hitcount: 3723
{ lat: 246, common_pid: 2036 } hitcount: 4738
{ lat: 247, common_pid: 2042 } hitcount: 312
{ lat: 247, common_pid: 2043 } hitcount: 2385
{ lat: 247, common_pid: 2041 } hitcount: 452
{ lat: 247, common_pid: 2038 } hitcount: 792
{ lat: 247, common_pid: 2040 } hitcount: 78
{ lat: 247, common_pid: 2036 } hitcount: 2375
{ lat: 247, common_pid: 2039 } hitcount: 1834
{ lat: 247, common_pid: 2037 } hitcount: 2655
{ lat: 248, common_pid: 2037 } hitcount: 36
{ lat: 248, common_pid: 2042 } hitcount: 11
{ lat: 248, common_pid: 2038 } hitcount: 122
{ lat: 248, common_pid: 2036 } hitcount: 135
{ lat: 248, common_pid: 2039 } hitcount: 26
{ lat: 248, common_pid: 2041 } hitcount: 503
{ lat: 248, common_pid: 2043 } hitcount: 66
{ lat: 248, common_pid: 2040 } hitcount: 46
{ lat: 249, common_pid: 2037 } hitcount: 29
{ lat: 249, common_pid: 2038 } hitcount: 1
{ lat: 249, common_pid: 2043 } hitcount: 29
{ lat: 249, common_pid: 2039 } hitcount: 8
{ lat: 249, common_pid: 2042 } hitcount: 56
{ lat: 249, common_pid: 2040 } hitcount: 27
{ lat: 249, common_pid: 2041 } hitcount: 11
{ lat: 249, common_pid: 2036 } hitcount: 27
{ lat: 250, common_pid: 2038 } hitcount: 1
{ lat: 250, common_pid: 2036 } hitcount: 30
{ lat: 250, common_pid: 2040 } hitcount: 19
{ lat: 250, common_pid: 2043 } hitcount: 22
{ lat: 250, common_pid: 2042 } hitcount: 20
{ lat: 250, common_pid: 2041 } hitcount: 1
{ lat: 250, common_pid: 2039 } hitcount: 6
{ lat: 250, common_pid: 2037 } hitcount: 48
{ lat: 251, common_pid: 2037 } hitcount: 43
{ lat: 251, common_pid: 2039 } hitcount: 1
{ lat: 251, common_pid: 2036 } hitcount: 12
{ lat: 251, common_pid: 2042 } hitcount: 2
{ lat: 251, common_pid: 2041 } hitcount: 1
{ lat: 251, common_pid: 2043 } hitcount: 15
{ lat: 251, common_pid: 2040 } hitcount: 3
{ lat: 252, common_pid: 2040 } hitcount: 1
{ lat: 252, common_pid: 2036 } hitcount: 12
{ lat: 252, common_pid: 2037 } hitcount: 21
{ lat: 252, common_pid: 2043 } hitcount: 14
{ lat: 253, common_pid: 2037 } hitcount: 21
{ lat: 253, common_pid: 2039 } hitcount: 2
{ lat: 253, common_pid: 2036 } hitcount: 9
{ lat: 253, common_pid: 2043 } hitcount: 6
{ lat: 253, common_pid: 2040 } hitcount: 1
{ lat: 254, common_pid: 2036 } hitcount: 8
{ lat: 254, common_pid: 2043 } hitcount: 3
{ lat: 254, common_pid: 2041 } hitcount: 1
{ lat: 254, common_pid: 2042 } hitcount: 1
{ lat: 254, common_pid: 2039 } hitcount: 1
{ lat: 254, common_pid: 2037 } hitcount: 12
{ lat: 255, common_pid: 2043 } hitcount: 1
{ lat: 255, common_pid: 2037 } hitcount: 2
{ lat: 255, common_pid: 2036 } hitcount: 2
{ lat: 255, common_pid: 2039 } hitcount: 8
{ lat: 256, common_pid: 2043 } hitcount: 1
{ lat: 256, common_pid: 2036 } hitcount: 4
{ lat: 256, common_pid: 2039 } hitcount: 6
{ lat: 257, common_pid: 2039 } hitcount: 5
{ lat: 257, common_pid: 2036 } hitcount: 4
{ lat: 258, common_pid: 2039 } hitcount: 5
{ lat: 258, common_pid: 2036 } hitcount: 2
{ lat: 259, common_pid: 2036 } hitcount: 7
{ lat: 259, common_pid: 2039 } hitcount: 7
{ lat: 260, common_pid: 2036 } hitcount: 8
{ lat: 260, common_pid: 2039 } hitcount: 6
{ lat: 261, common_pid: 2036 } hitcount: 5
{ lat: 261, common_pid: 2039 } hitcount: 7
{ lat: 262, common_pid: 2039 } hitcount: 5
{ lat: 262, common_pid: 2036 } hitcount: 5
{ lat: 263, common_pid: 2039 } hitcount: 7
{ lat: 263, common_pid: 2036 } hitcount: 7
{ lat: 264, common_pid: 2039 } hitcount: 9
{ lat: 264, common_pid: 2036 } hitcount: 9
{ lat: 265, common_pid: 2036 } hitcount: 5
{ lat: 265, common_pid: 2039 } hitcount: 1
{ lat: 266, common_pid: 2036 } hitcount: 1
{ lat: 266, common_pid: 2039 } hitcount: 3
{ lat: 267, common_pid: 2036 } hitcount: 1
{ lat: 267, common_pid: 2039 } hitcount: 3
{ lat: 268, common_pid: 2036 } hitcount: 1
{ lat: 268, common_pid: 2039 } hitcount: 6
{ lat: 269, common_pid: 2036 } hitcount: 1
{ lat: 269, common_pid: 2043 } hitcount: 1
{ lat: 269, common_pid: 2039 } hitcount: 2
{ lat: 270, common_pid: 2040 } hitcount: 1
{ lat: 270, common_pid: 2039 } hitcount: 6
{ lat: 271, common_pid: 2041 } hitcount: 1
{ lat: 271, common_pid: 2039 } hitcount: 5
{ lat: 272, common_pid: 2039 } hitcount: 10
{ lat: 273, common_pid: 2039 } hitcount: 8
{ lat: 274, common_pid: 2039 } hitcount: 2
{ lat: 275, common_pid: 2039 } hitcount: 1
{ lat: 276, common_pid: 2039 } hitcount: 2
{ lat: 276, common_pid: 2037 } hitcount: 1
{ lat: 276, common_pid: 2038 } hitcount: 1
{ lat: 277, common_pid: 2039 } hitcount: 1
{ lat: 277, common_pid: 2042 } hitcount: 1
{ lat: 278, common_pid: 2039 } hitcount: 1
{ lat: 279, common_pid: 2039 } hitcount: 4
{ lat: 279, common_pid: 2043 } hitcount: 1
{ lat: 280, common_pid: 2039 } hitcount: 3
{ lat: 283, common_pid: 2036 } hitcount: 2
{ lat: 284, common_pid: 2039 } hitcount: 1
{ lat: 284, common_pid: 2043 } hitcount: 1
{ lat: 288, common_pid: 2039 } hitcount: 1
{ lat: 289, common_pid: 2039 } hitcount: 1
{ lat: 300, common_pid: 2039 } hitcount: 1
{ lat: 384, common_pid: 2039 } hitcount: 1
Totals:
Hits: 67625
Entries: 278
Dropped: 0
Note, the writes are around the sleep, so ideally they will all be of 250
microseconds. If you are wondering how there are several that are under
250 microseconds, that is because the way cyclictest works, is if one
iteration comes in late, the next one will set the timer to wake up less that
250. That is, if an iteration came in 50 microseconds late, the next wake up
will be at 200 microseconds.
But this could easily be done in userspace. To make this even more
interesting, we can mix the histogram between events that happened in the
kernel with trace_marker::
# cd /sys/kernel/tracing
# echo 'latency u64 lat' > synthetic_events
# echo 'hist:keys=pid:ts0=common_timestamp.usecs' > events/sched/sched_waking/trigger
# echo 'hist:keys=common_pid:lat=common_timestamp.usecs-$ts0:onmatch(sched.sched_waking).latency($lat) if buf == "end"' > events/ftrace/print/trigger
# echo 'hist:keys=lat,common_pid:sort=lat' > events/synthetic/latency/trigger
The difference this time is that instead of using the trace_marker to start
the latency, the sched_waking event is used, matching the common_pid for the
trace_marker write with the pid that is being woken by sched_waking.
After running cyclictest again with the same parameters, we now have::
# cat events/synthetic/latency/hist
# event histogram
#
# trigger info: hist:keys=lat,common_pid:vals=hitcount:sort=lat:size=2048 [active]
#
{ lat: 7, common_pid: 2302 } hitcount: 640
{ lat: 7, common_pid: 2299 } hitcount: 42
{ lat: 7, common_pid: 2303 } hitcount: 18
{ lat: 7, common_pid: 2305 } hitcount: 166
{ lat: 7, common_pid: 2306 } hitcount: 1
{ lat: 7, common_pid: 2301 } hitcount: 91
{ lat: 7, common_pid: 2300 } hitcount: 17
{ lat: 8, common_pid: 2303 } hitcount: 8296
{ lat: 8, common_pid: 2304 } hitcount: 6864
{ lat: 8, common_pid: 2305 } hitcount: 9464
{ lat: 8, common_pid: 2301 } hitcount: 9213
{ lat: 8, common_pid: 2306 } hitcount: 6246
{ lat: 8, common_pid: 2302 } hitcount: 8797
{ lat: 8, common_pid: 2299 } hitcount: 8771
{ lat: 8, common_pid: 2300 } hitcount: 8119
{ lat: 9, common_pid: 2305 } hitcount: 1519
{ lat: 9, common_pid: 2299 } hitcount: 2346
{ lat: 9, common_pid: 2303 } hitcount: 2841
{ lat: 9, common_pid: 2301 } hitcount: 1846
{ lat: 9, common_pid: 2304 } hitcount: 3861
{ lat: 9, common_pid: 2302 } hitcount: 1210
{ lat: 9, common_pid: 2300 } hitcount: 2762
{ lat: 9, common_pid: 2306 } hitcount: 4247
{ lat: 10, common_pid: 2299 } hitcount: 16
{ lat: 10, common_pid: 2306 } hitcount: 333
{ lat: 10, common_pid: 2303 } hitcount: 16
{ lat: 10, common_pid: 2304 } hitcount: 168
{ lat: 10, common_pid: 2302 } hitcount: 240
{ lat: 10, common_pid: 2301 } hitcount: 28
{ lat: 10, common_pid: 2300 } hitcount: 95
{ lat: 10, common_pid: 2305 } hitcount: 18
{ lat: 11, common_pid: 2303 } hitcount: 5
{ lat: 11, common_pid: 2305 } hitcount: 8
{ lat: 11, common_pid: 2306 } hitcount: 221
{ lat: 11, common_pid: 2302 } hitcount: 76
{ lat: 11, common_pid: 2304 } hitcount: 26
{ lat: 11, common_pid: 2300 } hitcount: 125
{ lat: 11, common_pid: 2299 } hitcount: 2
{ lat: 12, common_pid: 2305 } hitcount: 3
{ lat: 12, common_pid: 2300 } hitcount: 6
{ lat: 12, common_pid: 2306 } hitcount: 90
{ lat: 12, common_pid: 2302 } hitcount: 4
{ lat: 12, common_pid: 2303 } hitcount: 1
{ lat: 12, common_pid: 2304 } hitcount: 122
{ lat: 13, common_pid: 2300 } hitcount: 12
{ lat: 13, common_pid: 2301 } hitcount: 1
{ lat: 13, common_pid: 2306 } hitcount: 32
{ lat: 13, common_pid: 2302 } hitcount: 5
{ lat: 13, common_pid: 2305 } hitcount: 1
{ lat: 13, common_pid: 2303 } hitcount: 1
{ lat: 13, common_pid: 2304 } hitcount: 61
{ lat: 14, common_pid: 2303 } hitcount: 4
{ lat: 14, common_pid: 2306 } hitcount: 5
{ lat: 14, common_pid: 2305 } hitcount: 4
{ lat: 14, common_pid: 2304 } hitcount: 62
{ lat: 14, common_pid: 2302 } hitcount: 19
{ lat: 14, common_pid: 2300 } hitcount: 33
{ lat: 14, common_pid: 2299 } hitcount: 1
{ lat: 14, common_pid: 2301 } hitcount: 4
{ lat: 15, common_pid: 2305 } hitcount: 1
{ lat: 15, common_pid: 2302 } hitcount: 25
{ lat: 15, common_pid: 2300 } hitcount: 11
{ lat: 15, common_pid: 2299 } hitcount: 5
{ lat: 15, common_pid: 2301 } hitcount: 1
{ lat: 15, common_pid: 2304 } hitcount: 8
{ lat: 15, common_pid: 2303 } hitcount: 1
{ lat: 15, common_pid: 2306 } hitcount: 6
{ lat: 16, common_pid: 2302 } hitcount: 31
{ lat: 16, common_pid: 2306 } hitcount: 3
{ lat: 16, common_pid: 2300 } hitcount: 5
{ lat: 17, common_pid: 2302 } hitcount: 6
{ lat: 17, common_pid: 2303 } hitcount: 1
{ lat: 18, common_pid: 2304 } hitcount: 1
{ lat: 18, common_pid: 2302 } hitcount: 8
{ lat: 18, common_pid: 2299 } hitcount: 1
{ lat: 18, common_pid: 2301 } hitcount: 1
{ lat: 19, common_pid: 2303 } hitcount: 4
{ lat: 19, common_pid: 2304 } hitcount: 5
{ lat: 19, common_pid: 2302 } hitcount: 4
{ lat: 19, common_pid: 2299 } hitcount: 3
{ lat: 19, common_pid: 2306 } hitcount: 1
{ lat: 19, common_pid: 2300 } hitcount: 4
{ lat: 19, common_pid: 2305 } hitcount: 5
{ lat: 20, common_pid: 2299 } hitcount: 2
{ lat: 20, common_pid: 2302 } hitcount: 3
{ lat: 20, common_pid: 2305 } hitcount: 1
{ lat: 20, common_pid: 2300 } hitcount: 2
{ lat: 20, common_pid: 2301 } hitcount: 2
{ lat: 20, common_pid: 2303 } hitcount: 3
{ lat: 21, common_pid: 2305 } hitcount: 1
{ lat: 21, common_pid: 2299 } hitcount: 5
{ lat: 21, common_pid: 2303 } hitcount: 4
{ lat: 21, common_pid: 2302 } hitcount: 7
{ lat: 21, common_pid: 2300 } hitcount: 1
{ lat: 21, common_pid: 2301 } hitcount: 5
{ lat: 21, common_pid: 2304 } hitcount: 2
{ lat: 22, common_pid: 2302 } hitcount: 5
{ lat: 22, common_pid: 2303 } hitcount: 1
{ lat: 22, common_pid: 2306 } hitcount: 3
{ lat: 22, common_pid: 2301 } hitcount: 2
{ lat: 22, common_pid: 2300 } hitcount: 1
{ lat: 22, common_pid: 2299 } hitcount: 1
{ lat: 22, common_pid: 2305 } hitcount: 1
{ lat: 22, common_pid: 2304 } hitcount: 1
{ lat: 23, common_pid: 2299 } hitcount: 1
{ lat: 23, common_pid: 2306 } hitcount: 2
{ lat: 23, common_pid: 2302 } hitcount: 6
{ lat: 24, common_pid: 2302 } hitcount: 3
{ lat: 24, common_pid: 2300 } hitcount: 1
{ lat: 24, common_pid: 2306 } hitcount: 2
{ lat: 24, common_pid: 2305 } hitcount: 1
{ lat: 24, common_pid: 2299 } hitcount: 1
{ lat: 25, common_pid: 2300 } hitcount: 1
{ lat: 25, common_pid: 2302 } hitcount: 4
{ lat: 26, common_pid: 2302 } hitcount: 2
{ lat: 27, common_pid: 2305 } hitcount: 1
{ lat: 27, common_pid: 2300 } hitcount: 1
{ lat: 27, common_pid: 2302 } hitcount: 3
{ lat: 28, common_pid: 2306 } hitcount: 1
{ lat: 28, common_pid: 2302 } hitcount: 4
{ lat: 29, common_pid: 2302 } hitcount: 1
{ lat: 29, common_pid: 2300 } hitcount: 2
{ lat: 29, common_pid: 2306 } hitcount: 1
{ lat: 29, common_pid: 2304 } hitcount: 1
{ lat: 30, common_pid: 2302 } hitcount: 4
{ lat: 31, common_pid: 2302 } hitcount: 6
{ lat: 32, common_pid: 2302 } hitcount: 1
{ lat: 33, common_pid: 2299 } hitcount: 1
{ lat: 33, common_pid: 2302 } hitcount: 3
{ lat: 34, common_pid: 2302 } hitcount: 2
{ lat: 35, common_pid: 2302 } hitcount: 1
{ lat: 35, common_pid: 2304 } hitcount: 1
{ lat: 36, common_pid: 2302 } hitcount: 4
{ lat: 37, common_pid: 2302 } hitcount: 6
{ lat: 38, common_pid: 2302 } hitcount: 2
{ lat: 39, common_pid: 2302 } hitcount: 2
{ lat: 39, common_pid: 2304 } hitcount: 1
{ lat: 40, common_pid: 2304 } hitcount: 2
{ lat: 40, common_pid: 2302 } hitcount: 5
{ lat: 41, common_pid: 2304 } hitcount: 1
{ lat: 41, common_pid: 2302 } hitcount: 8
{ lat: 42, common_pid: 2302 } hitcount: 6
{ lat: 42, common_pid: 2304 } hitcount: 1
{ lat: 43, common_pid: 2302 } hitcount: 3
{ lat: 43, common_pid: 2304 } hitcount: 4
{ lat: 44, common_pid: 2302 } hitcount: 6
{ lat: 45, common_pid: 2302 } hitcount: 5
{ lat: 46, common_pid: 2302 } hitcount: 5
{ lat: 47, common_pid: 2302 } hitcount: 7
{ lat: 48, common_pid: 2301 } hitcount: 1
{ lat: 48, common_pid: 2302 } hitcount: 9
{ lat: 49, common_pid: 2302 } hitcount: 3
{ lat: 50, common_pid: 2302 } hitcount: 1
{ lat: 50, common_pid: 2301 } hitcount: 1
{ lat: 51, common_pid: 2302 } hitcount: 2
{ lat: 51, common_pid: 2301 } hitcount: 1
{ lat: 61, common_pid: 2302 } hitcount: 1
{ lat: 110, common_pid: 2302 } hitcount: 1
Totals:
Hits: 89565
Entries: 158
Dropped: 0
This doesn't tell us any information about how late cyclictest may have
woken up, but it does show us a nice histogram of how long it took from
the time that cyclictest was woken to the time it made it into user space.
요약·해설
histogram.rst:1-3071ftrace event histogram의 명령 형식, key와 value, field modifier, variable, synthetic event, handler와 action, 사용자 공간 trace_marker 연계를 Linux v6.18.37 원문 전체에 맞춰 설명합니다.
hist trigger는 event field 또는 stack trace를 key로 삼아 hash table entry를 만들고, `hitcount`와 numeric field 합계를 value로 누적한다. field modifier는 address, symbol, syscall, executable name, timestamp, percentage, graph, stacktrace 표시를 제어하며 filter, named histogram, pause/continue/clear를 조합할 수 있다.
여러 event 사이의 timestamp와 field를 histogram variable로 전달하면 `onmatch()`가 synthetic event를 만들 수 있다. `onmax()`와 `onchange()`는 새 maximum 또는 값 변화가 생긴 순간의 field를 저장하거나 trace buffer snapshot을 남긴다. 사용자 공간에서는 `trace_marker`의 `ftrace/print` event를 이용해 같은 기법으로 구간 latency를 측정한다.
측정 목적에 따라 핵심 구성 요소와 결과를 구분한다.