blob: 803b1318b13da11242a3558123d28293d3ee4378 [file] [log] [blame]
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001 ftrace - Function Tracer
2 ========================
3
4Copyright 2008 Red Hat Inc.
Steven Rostedta41eeba2008-07-14 16:41:12 -04005 Author: Steven Rostedt <srostedt@redhat.com>
6 License: The GNU Free Documentation License, Version 1.2
Steven Rostedta97762a2008-07-31 12:40:52 -04007 (dual licensed under the GPL v2)
Steven Rostedtf2d9c742008-07-15 10:57:33 -04008Reviewers: Elias Oltmanns, Randy Dunlap, Andrew Morton,
9 John Kacur, and David Teigland.
Steven Rostedteb6d42e2008-07-10 12:46:01 -040010
Steven Rostedt42ec6322008-11-03 15:18:56 -050011Written for: 2.6.28-rc2
Steven Rostedteb6d42e2008-07-10 12:46:01 -040012
13Introduction
14------------
15
16Ftrace is an internal tracer designed to help out developers and
17designers of systems to find what is going on inside the kernel.
18It can be used for debugging or analyzing latencies and performance
19issues that take place outside of user-space.
20
21Although ftrace is the function tracer, it also includes an
22infrastructure that allows for other types of tracing. Some of the
Steven Rostedtf2d9c742008-07-15 10:57:33 -040023tracers that are currently in ftrace include a tracer to trace
Steven Rostedteb6d42e2008-07-10 12:46:01 -040024context switches, the time it takes for a high priority task to
25run after it was woken up, the time interrupts are disabled, and
Steven Rostedtf2d9c742008-07-15 10:57:33 -040026more (ftrace allows for tracer plugins, which means that the list of
27tracers can always grow).
Steven Rostedteb6d42e2008-07-10 12:46:01 -040028
29
30The File System
31---------------
32
33Ftrace uses the debugfs file system to hold the control files as well
34as the files to display output.
35
36To mount the debugfs system:
37
38 # mkdir /debug
39 # mount -t debugfs nodev /debug
40
Steven Rostedtf2d9c742008-07-15 10:57:33 -040041(Note: it is more common to mount at /sys/kernel/debug, but for simplicity
42 this document will use /debug)
Steven Rostedteb6d42e2008-07-10 12:46:01 -040043
44That's it! (assuming that you have ftrace configured into your kernel)
45
46After mounting the debugfs, you can see a directory called
47"tracing". This directory contains the control and output files
48of ftrace. Here is a list of some of the key files:
49
50
51 Note: all time values are in microseconds.
52
Steven Rostedt9b803c02008-11-03 15:15:08 -050053 current_tracer: This is used to set or display the current tracer
Steven Rostedteb6d42e2008-07-10 12:46:01 -040054 that is configured.
55
Steven Rostedt9b803c02008-11-03 15:15:08 -050056 available_tracers: This holds the different types of tracers that
Steven Rostedta41eeba2008-07-14 16:41:12 -040057 have been compiled into the kernel. The tracers
Steven Rostedtf2d9c742008-07-15 10:57:33 -040058 listed here can be configured by echoing their name
59 into current_tracer.
Steven Rostedteb6d42e2008-07-10 12:46:01 -040060
Steven Rostedt9b803c02008-11-03 15:15:08 -050061 tracing_enabled: This sets or displays whether the current_tracer
Steven Rostedteb6d42e2008-07-10 12:46:01 -040062 is activated and tracing or not. Echo 0 into this
Steven Rostedtf2d9c742008-07-15 10:57:33 -040063 file to disable the tracer or 1 to enable it.
Steven Rostedteb6d42e2008-07-10 12:46:01 -040064
Steven Rostedt9b803c02008-11-03 15:15:08 -050065 trace: This file holds the output of the trace in a human readable
Steven Rostedtf2d9c742008-07-15 10:57:33 -040066 format (described below).
Steven Rostedteb6d42e2008-07-10 12:46:01 -040067
Steven Rostedt9b803c02008-11-03 15:15:08 -050068 latency_trace: This file shows the same trace but the information
Steven Rostedteb6d42e2008-07-10 12:46:01 -040069 is organized more to display possible latencies
Steven Rostedtf2d9c742008-07-15 10:57:33 -040070 in the system (described below).
Steven Rostedteb6d42e2008-07-10 12:46:01 -040071
Steven Rostedt9b803c02008-11-03 15:15:08 -050072 trace_pipe: The output is the same as the "trace" file but this
Steven Rostedteb6d42e2008-07-10 12:46:01 -040073 file is meant to be streamed with live tracing.
74 Reads from this file will block until new data
75 is retrieved. Unlike the "trace" and "latency_trace"
76 files, this file is a consumer. This means reading
77 from this file causes sequential reads to display
78 more current data. Once data is read from this
79 file, it is consumed, and will not be read
80 again with a sequential read. The "trace" and
81 "latency_trace" files are static, and if the
Steven Rostedtf2d9c742008-07-15 10:57:33 -040082 tracer is not adding more data, they will display
Steven Rostedteb6d42e2008-07-10 12:46:01 -040083 the same information every time they are read.
84
Steven Rostedtee6bce52008-11-12 17:52:37 -050085 trace_options: This file lets the user control the amount of data
Steven Rostedteb6d42e2008-07-10 12:46:01 -040086 that is displayed in one of the above output
87 files.
88
Steven Rostedt9b803c02008-11-03 15:15:08 -050089 trace_max_latency: Some of the tracers record the max latency.
Steven Rostedteb6d42e2008-07-10 12:46:01 -040090 For example, the time interrupts are disabled.
91 This time is saved in this file. The max trace
92 will also be stored, and displayed by either
93 "trace" or "latency_trace". A new max trace will
94 only be recorded if the latency is greater than
95 the value in this file. (in microseconds)
96
Steven Rostedt1696b2b2008-11-13 00:09:35 -050097 buffer_size_kb: This sets or displays the number of kilobytes each CPU
Steven Rostedt9b803c02008-11-03 15:15:08 -050098 buffer can hold. The tracer buffers are the same size
99 for each CPU. The displayed number is the size of the
Steven Rostedt1696b2b2008-11-13 00:09:35 -0500100 CPU buffer and not total size of all buffers. The
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400101 trace buffers are allocated in pages (blocks of memory
102 that the kernel uses for allocation, usually 4 KB in size).
Steven Rostedt9b803c02008-11-03 15:15:08 -0500103 If the last page allocated has room for more bytes
104 than requested, the rest of the page will be used,
105 making the actual allocation bigger than requested.
106 (Note, the size may not be a multiple of the page size due
107 to buffer managment overhead.)
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400108
109 This can only be updated when the current_tracer
Steven Rostedt9b803c02008-11-03 15:15:08 -0500110 is set to "nop".
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400111
Steven Rostedt9b803c02008-11-03 15:15:08 -0500112 tracing_cpumask: This is a mask that lets the user only trace
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400113 on specified CPUS. The format is a hex string
114 representing the CPUS.
115
Steven Rostedt9b803c02008-11-03 15:15:08 -0500116 set_ftrace_filter: When dynamic ftrace is configured in (see the
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400117 section below "dynamic ftrace"), the code is dynamically
118 modified (code text rewrite) to disable calling of the
119 function profiler (mcount). This lets tracing be configured
120 in with practically no overhead in performance. This also
121 has a side effect of enabling or disabling specific functions
122 to be traced. Echoing names of functions into this file
123 will limit the trace to only those functions.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400124
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400125 set_ftrace_notrace: This has an effect opposite to that of
126 set_ftrace_filter. Any function that is added here will not
127 be traced. If a function exists in both set_ftrace_filter
128 and set_ftrace_notrace, the function will _not_ be traced.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400129
Steven Rostedtdf4fc312008-11-26 00:16:23 -0500130 set_ftrace_pid: Have the function tracer only trace a single thread.
131
Steven Rostedt9b803c02008-11-03 15:15:08 -0500132 available_filter_functions: This lists the functions that ftrace
133 has processed and can trace. These are the function
134 names that you can pass to "set_ftrace_filter" or
135 "set_ftrace_notrace". (See the section "dynamic ftrace"
136 below for more details.)
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400137
138
139The Tracers
140-----------
141
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400142Here is the list of current tracers that may be configured.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400143
Steven Rostedt9b803c02008-11-03 15:15:08 -0500144 function - function tracer that uses mcount to trace all functions.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400145
146 sched_switch - traces the context switches between tasks.
147
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400148 irqsoff - traces the areas that disable interrupts and saves
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400149 the trace with the longest max latency.
150 See tracing_max_latency. When a new max is recorded,
151 it replaces the old trace. It is best to view this
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400152 trace via the latency_trace file.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400153
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400154 preemptoff - Similar to irqsoff but traces and records the amount of
155 time for which preemption is disabled.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400156
157 preemptirqsoff - Similar to irqsoff and preemptoff, but traces and
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400158 records the largest time for which irqs and/or preemption
159 is disabled.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400160
161 wakeup - Traces and records the max latency that it takes for
162 the highest priority task to get scheduled after
163 it has been woken up.
164
Steven Rostedt9b803c02008-11-03 15:15:08 -0500165 nop - This is not a tracer. To remove all tracers from tracing
166 simply echo "nop" into current_tracer.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400167
168
169Examples of using the tracer
170----------------------------
171
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400172Here are typical examples of using the tracers when controlling them only
173with the debugfs interface (without using any user-land utilities).
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400174
175Output format:
176--------------
177
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400178Here is an example of the output format of the file "trace"
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400179
180 --------
Steven Rostedt9b803c02008-11-03 15:15:08 -0500181# tracer: function
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400182#
183# TASK-PID CPU# TIMESTAMP FUNCTION
184# | | | | |
185 bash-4251 [01] 10152.583854: path_put <-path_walk
186 bash-4251 [01] 10152.583855: dput <-path_put
187 bash-4251 [01] 10152.583855: _atomic_dec_and_lock <-dput
188 --------
189
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400190A header is printed with the tracer name that is represented by the trace.
Steven Rostedt9b803c02008-11-03 15:15:08 -0500191In this case the tracer is "function". Then a header showing the format. Task
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400192name "bash", the task PID "4251", the CPU that it was running on
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400193"01", the timestamp in <secs>.<usecs> format, the function name that was
194traced "path_put" and the parent function that called this function
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400195"path_walk". The timestamp is the time at which the function was
196entered.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400197
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400198The sched_switch tracer also includes tracing of task wakeups and
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400199context switches.
200
201 ksoftirqd/1-7 [01] 1453.070013: 7:115:R + 2916:115:S
202 ksoftirqd/1-7 [01] 1453.070013: 7:115:R + 10:115:S
203 ksoftirqd/1-7 [01] 1453.070013: 7:115:R ==> 10:115:R
204 events/1-10 [01] 1453.070013: 10:115:S ==> 2916:115:R
205 kondemand/1-2916 [01] 1453.070013: 2916:115:S ==> 7:115:R
206 ksoftirqd/1-7 [01] 1453.070013: 7:115:S ==> 0:140:R
207
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400208Wake ups are represented by a "+" and the context switches are shown as
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400209"==>". The format is:
210
211 Context switches:
212
213 Previous task Next Task
214
215 <pid>:<prio>:<state> ==> <pid>:<prio>:<state>
216
217 Wake ups:
218
219 Current task Task waking up
220
221 <pid>:<prio>:<state> + <pid>:<prio>:<state>
222
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400223The prio is the internal kernel priority, which is the inverse of the
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400224priority that is usually displayed by user-space tools. Zero represents
225the highest priority (99). Prio 100 starts the "nice" priorities with
226100 being equal to nice -20 and 139 being nice 19. The prio "140" is
227reserved for the idle task which is the lowest priority thread (pid 0).
228
229
230Latency trace format
231--------------------
232
233For traces that display latency times, the latency_trace file gives
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400234somewhat more information to see why a latency happened. Here is a typical
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400235trace.
236
237# tracer: irqsoff
238#
239irqsoff latency trace v1.1.5 on 2.6.26-rc8
240--------------------------------------------------------------------
241 latency: 97 us, #3/3, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
242 -----------------
243 | task: swapper-0 (uid:0 nice:0 policy:0 rt_prio:0)
244 -----------------
245 => started at: apic_timer_interrupt
246 => ended at: do_softirq
247
248# _------=> CPU#
249# / _-----=> irqs-off
250# | / _----=> need-resched
251# || / _---=> hardirq/softirq
252# ||| / _--=> preempt-depth
253# |||| /
254# ||||| delay
255# cmd pid ||||| time | caller
256# \ / ||||| \ | /
257 <idle>-0 0d..1 0us+: trace_hardirqs_off_thunk (apic_timer_interrupt)
258 <idle>-0 0d.s. 97us : __do_softirq (do_softirq)
259 <idle>-0 0d.s1 98us : trace_hardirqs_on (do_softirq)
260
261
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400262
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400263This shows that the current tracer is "irqsoff" tracing the time for which
264interrupts were disabled. It gives the trace version and the version
265of the kernel upon which this was executed on (2.6.26-rc8). Then it displays
266the max latency in microsecs (97 us). The number of trace entries displayed
267and the total number recorded (both are three: #3/3). The type of
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400268preemption that was used (PREEMPT). VP, KP, SP, and HP are always zero
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400269and are reserved for later use. #P is the number of online CPUS (#P:2).
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400270
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400271The task is the process that was running when the latency occurred.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400272(swapper pid: 0).
273
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400274The start and stop (the functions in which the interrupts were disabled and
275enabled respectively) that caused the latencies:
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400276
277 apic_timer_interrupt is where the interrupts were disabled.
278 do_softirq is where they were enabled again.
279
280The next lines after the header are the trace itself. The header
281explains which is which.
282
283 cmd: The name of the process in the trace.
284
285 pid: The PID of that process.
286
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400287 CPU#: The CPU which the process was running on.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400288
289 irqs-off: 'd' interrupts are disabled. '.' otherwise.
Steven Rostedt92444892008-10-24 09:42:59 -0400290 Note: If the architecture does not support a way to
291 read the irq flags variable, an 'X' will always
292 be printed here.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400293
294 need-resched: 'N' task need_resched is set, '.' otherwise.
295
296 hardirq/softirq:
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400297 'H' - hard irq occurred inside a softirq.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400298 'h' - hard irq is running
299 's' - soft irq is running
300 '.' - normal context.
301
302 preempt-depth: The level of preempt_disabled
303
304The above is mostly meaningful for kernel developers.
305
Steven Rostedta41eeba2008-07-14 16:41:12 -0400306 time: This differs from the trace file output. The trace file output
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400307 includes an absolute timestamp. The timestamp used by the
Steven Rostedta41eeba2008-07-14 16:41:12 -0400308 latency_trace file is relative to the start of the trace.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400309
310 delay: This is just to help catch your eye a bit better. And
311 needs to be fixed to be only relative to the same CPU.
Steven Rostedta41eeba2008-07-14 16:41:12 -0400312 The marks are determined by the difference between this
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400313 current trace and the next trace.
314 '!' - greater than preempt_mark_thresh (default 100)
315 '+' - greater than 1 microsecond
316 ' ' - less than or equal to 1 microsecond.
317
318 The rest is the same as the 'trace' file.
319
320
Steven Rostedtee6bce52008-11-12 17:52:37 -0500321trace_options
322-------------
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400323
Steven Rostedtee6bce52008-11-12 17:52:37 -0500324The trace_options file is used to control what gets printed in the trace
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400325output. To see what is available, simply cat the file:
326
Steven Rostedtee6bce52008-11-12 17:52:37 -0500327 cat /debug/tracing/trace_options
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400328 print-parent nosym-offset nosym-addr noverbose noraw nohex nobin \
Török Edwinb54d3de2008-11-22 13:28:48 +0200329 noblock nostacktrace nosched-tree nouserstacktrace nosym-userobj
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400330
Steven Rostedta41eeba2008-07-14 16:41:12 -0400331To disable one of the options, echo in the option prepended with "no".
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400332
Steven Rostedtee6bce52008-11-12 17:52:37 -0500333 echo noprint-parent > /debug/tracing/trace_options
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400334
335To enable an option, leave off the "no".
336
Steven Rostedtee6bce52008-11-12 17:52:37 -0500337 echo sym-offset > /debug/tracing/trace_options
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400338
339Here are the available options:
340
341 print-parent - On function traces, display the calling function
342 as well as the function being traced.
343
344 print-parent:
345 bash-4000 [01] 1477.606694: simple_strtoul <-strict_strtoul
346
347 noprint-parent:
348 bash-4000 [01] 1477.606694: simple_strtoul
349
350
351 sym-offset - Display not only the function name, but also the offset
352 in the function. For example, instead of seeing just
Steven Rostedta41eeba2008-07-14 16:41:12 -0400353 "ktime_get", you will see "ktime_get+0xb/0x20".
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400354
355 sym-offset:
356 bash-4000 [01] 1477.606694: simple_strtoul+0x6/0xa0
357
358 sym-addr - this will also display the function address as well as
359 the function name.
360
361 sym-addr:
362 bash-4000 [01] 1477.606694: simple_strtoul <c0339346>
363
364 verbose - This deals with the latency_trace file.
365
366 bash 4000 1 0 00000000 00010a95 [58127d26] 1720.415ms \
367 (+0.000ms): simple_strtoul (strict_strtoul)
368
369 raw - This will display raw numbers. This option is best for use with
370 user applications that can translate the raw numbers better than
371 having it done in the kernel.
372
Steven Rostedta41eeba2008-07-14 16:41:12 -0400373 hex - Similar to raw, but the numbers will be in a hexadecimal format.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400374
375 bin - This will print out the formats in raw binary.
376
377 block - TBD (needs update)
378
379 stacktrace - This is one of the options that changes the trace itself.
380 When a trace is recorded, so is the stack of functions.
381 This allows for back traces of trace sites.
382
Török Edwin02b67512008-11-22 13:28:47 +0200383 userstacktrace - This option changes the trace.
384 It records a stacktrace of the current userspace thread.
385
Török Edwinb54d3de2008-11-22 13:28:48 +0200386 sym-userobj - when user stacktrace are enabled, look up which object the
387 address belongs to, and print a relative address
388 This is especially useful when ASLR is on, otherwise you don't
389 get a chance to resolve the address to object/file/line after the app is no
390 longer running
391
392 The lookup is performed when you read trace,trace_pipe,latency_trace. Example:
393
394 a.out-1623 [000] 40874.465068: /root/a.out[+0x480] <-/root/a.out[+0
395x494] <- /root/a.out[+0x4a8] <- /lib/libc-2.7.so[+0x1e1a6]
396
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400397 sched-tree - TBD (any users??)
398
399
400sched_switch
401------------
402
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400403This tracer simply records schedule switches. Here is an example
Steven Rostedta41eeba2008-07-14 16:41:12 -0400404of how to use it.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400405
406 # echo sched_switch > /debug/tracing/current_tracer
407 # echo 1 > /debug/tracing/tracing_enabled
408 # sleep 1
409 # echo 0 > /debug/tracing/tracing_enabled
410 # cat /debug/tracing/trace
411
412# tracer: sched_switch
413#
414# TASK-PID CPU# TIMESTAMP FUNCTION
415# | | | | |
416 bash-3997 [01] 240.132281: 3997:120:R + 4055:120:R
417 bash-3997 [01] 240.132284: 3997:120:R ==> 4055:120:R
418 sleep-4055 [01] 240.132371: 4055:120:S ==> 3997:120:R
419 bash-3997 [01] 240.132454: 3997:120:R + 4055:120:S
420 bash-3997 [01] 240.132457: 3997:120:R ==> 4055:120:R
421 sleep-4055 [01] 240.132460: 4055:120:D ==> 3997:120:R
422 bash-3997 [01] 240.132463: 3997:120:R + 4055:120:D
423 bash-3997 [01] 240.132465: 3997:120:R ==> 4055:120:R
424 <idle>-0 [00] 240.132589: 0:140:R + 4:115:S
425 <idle>-0 [00] 240.132591: 0:140:R ==> 4:115:R
426 ksoftirqd/0-4 [00] 240.132595: 4:115:S ==> 0:140:R
427 <idle>-0 [00] 240.132598: 0:140:R + 4:115:S
428 <idle>-0 [00] 240.132599: 0:140:R ==> 4:115:R
429 ksoftirqd/0-4 [00] 240.132603: 4:115:S ==> 0:140:R
430 sleep-4055 [01] 240.133058: 4055:120:S ==> 3997:120:R
431 [...]
432
433
434As we have discussed previously about this format, the header shows
435the name of the trace and points to the options. The "FUNCTION"
436is a misnomer since here it represents the wake ups and context
437switches.
438
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400439The sched_switch file only lists the wake ups (represented with '+')
440and context switches ('==>') with the previous task or current task
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400441first followed by the next task or task waking up. The format for both
442of these is PID:KERNEL-PRIO:TASK-STATE. Remember that the KERNEL-PRIO
443is the inverse of the actual priority with zero (0) being the highest
444priority and the nice values starting at 100 (nice -20). Below is
445a quick chart to map the kernel priority to user land priorities.
446
447 Kernel priority: 0 to 99 ==> user RT priority 99 to 0
448 Kernel priority: 100 to 139 ==> user nice -20 to 19
449 Kernel priority: 140 ==> idle task priority
450
451The task states are:
452
453 R - running : wants to run, may not actually be running
454 S - sleep : process is waiting to be woken up (handles signals)
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400455 D - disk sleep (uninterruptible sleep) : process must be woken up
456 (ignores signals)
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400457 T - stopped : process suspended
458 t - traced : process is being traced (with something like gdb)
459 Z - zombie : process waiting to be cleaned up
460 X - unknown
461
462
463ftrace_enabled
464--------------
465
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400466The following tracers (listed below) give different output depending
467on whether or not the sysctl ftrace_enabled is set. To set ftrace_enabled,
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400468one can either use the sysctl function or set it via the proc
469file system interface.
470
471 sysctl kernel.ftrace_enabled=1
472
473 or
474
475 echo 1 > /proc/sys/kernel/ftrace_enabled
476
477To disable ftrace_enabled simply replace the '1' with '0' in
478the above commands.
479
480When ftrace_enabled is set the tracers will also record the functions
481that are within the trace. The descriptions of the tracers
482will also show an example with ftrace enabled.
483
484
485irqsoff
486-------
487
488When interrupts are disabled, the CPU can not react to any other
489external event (besides NMIs and SMIs). This prevents the timer
490interrupt from triggering or the mouse interrupt from letting the
491kernel know of a new mouse event. The result is a latency with the
492reaction time.
493
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400494The irqsoff tracer tracks the time for which interrupts are disabled.
495When a new maximum latency is hit, the tracer saves the trace leading up
496to that latency point so that every time a new maximum is reached, the old
497saved trace is discarded and the new trace is saved.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400498
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400499To reset the maximum, echo 0 into tracing_max_latency. Here is an
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400500example:
501
502 # echo irqsoff > /debug/tracing/current_tracer
503 # echo 0 > /debug/tracing/tracing_max_latency
504 # echo 1 > /debug/tracing/tracing_enabled
505 # ls -ltr
506 [...]
507 # echo 0 > /debug/tracing/tracing_enabled
508 # cat /debug/tracing/latency_trace
509# tracer: irqsoff
510#
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400511irqsoff latency trace v1.1.5 on 2.6.26
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400512--------------------------------------------------------------------
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400513 latency: 12 us, #3/3, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400514 -----------------
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400515 | task: bash-3730 (uid:0 nice:0 policy:0 rt_prio:0)
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400516 -----------------
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400517 => started at: sys_setpgid
518 => ended at: sys_setpgid
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400519
520# _------=> CPU#
521# / _-----=> irqs-off
522# | / _----=> need-resched
523# || / _---=> hardirq/softirq
524# ||| / _--=> preempt-depth
525# |||| /
526# ||||| delay
527# cmd pid ||||| time | caller
528# \ / ||||| \ | /
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400529 bash-3730 1d... 0us : _write_lock_irq (sys_setpgid)
530 bash-3730 1d..1 1us+: _write_unlock_irq (sys_setpgid)
531 bash-3730 1d..2 14us : trace_hardirqs_on (sys_setpgid)
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400532
533
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400534Here we see that that we had a latency of 12 microsecs (which is
535very good). The _write_lock_irq in sys_setpgid disabled interrupts.
536The difference between the 12 and the displayed timestamp 14us occurred
537because the clock was incremented between the time of recording the max
538latency and the time of recording the function that had that latency.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400539
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400540Note the above example had ftrace_enabled not set. If we set the
541ftrace_enabled, we get a much larger output:
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400542
543# tracer: irqsoff
544#
545irqsoff latency trace v1.1.5 on 2.6.26-rc8
546--------------------------------------------------------------------
547 latency: 50 us, #101/101, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
548 -----------------
549 | task: ls-4339 (uid:0 nice:0 policy:0 rt_prio:0)
550 -----------------
551 => started at: __alloc_pages_internal
552 => ended at: __alloc_pages_internal
553
554# _------=> CPU#
555# / _-----=> irqs-off
556# | / _----=> need-resched
557# || / _---=> hardirq/softirq
558# ||| / _--=> preempt-depth
559# |||| /
560# ||||| delay
561# cmd pid ||||| time | caller
562# \ / ||||| \ | /
563 ls-4339 0...1 0us+: get_page_from_freelist (__alloc_pages_internal)
564 ls-4339 0d..1 3us : rmqueue_bulk (get_page_from_freelist)
565 ls-4339 0d..1 3us : _spin_lock (rmqueue_bulk)
566 ls-4339 0d..1 4us : add_preempt_count (_spin_lock)
567 ls-4339 0d..2 4us : __rmqueue (rmqueue_bulk)
568 ls-4339 0d..2 5us : __rmqueue_smallest (__rmqueue)
569 ls-4339 0d..2 5us : __mod_zone_page_state (__rmqueue_smallest)
570 ls-4339 0d..2 6us : __rmqueue (rmqueue_bulk)
571 ls-4339 0d..2 6us : __rmqueue_smallest (__rmqueue)
572 ls-4339 0d..2 7us : __mod_zone_page_state (__rmqueue_smallest)
573 ls-4339 0d..2 7us : __rmqueue (rmqueue_bulk)
574 ls-4339 0d..2 8us : __rmqueue_smallest (__rmqueue)
575[...]
576 ls-4339 0d..2 46us : __rmqueue_smallest (__rmqueue)
577 ls-4339 0d..2 47us : __mod_zone_page_state (__rmqueue_smallest)
578 ls-4339 0d..2 47us : __rmqueue (rmqueue_bulk)
579 ls-4339 0d..2 48us : __rmqueue_smallest (__rmqueue)
580 ls-4339 0d..2 48us : __mod_zone_page_state (__rmqueue_smallest)
581 ls-4339 0d..2 49us : _spin_unlock (rmqueue_bulk)
582 ls-4339 0d..2 49us : sub_preempt_count (_spin_unlock)
583 ls-4339 0d..1 50us : get_page_from_freelist (__alloc_pages_internal)
584 ls-4339 0d..2 51us : trace_hardirqs_on (__alloc_pages_internal)
585
586
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400587
588Here we traced a 50 microsecond latency. But we also see all the
Steven Rostedta41eeba2008-07-14 16:41:12 -0400589functions that were called during that time. Note that by enabling
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400590function tracing, we incur an added overhead. This overhead may
Steven Rostedta41eeba2008-07-14 16:41:12 -0400591extend the latency times. But nevertheless, this trace has provided
592some very helpful debugging information.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400593
594
595preemptoff
596----------
597
Steven Rostedta41eeba2008-07-14 16:41:12 -0400598When preemption is disabled, we may be able to receive interrupts but
599the task cannot be preempted and a higher priority task must wait
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400600for preemption to be enabled again before it can preempt a lower
601priority task.
602
Steven Rostedta41eeba2008-07-14 16:41:12 -0400603The preemptoff tracer traces the places that disable preemption.
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400604Like the irqsoff tracer, it records the maximum latency for which preemption
605was disabled. The control of preemptoff tracer is much like the irqsoff
606tracer.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400607
608 # echo preemptoff > /debug/tracing/current_tracer
609 # echo 0 > /debug/tracing/tracing_max_latency
610 # echo 1 > /debug/tracing/tracing_enabled
611 # ls -ltr
612 [...]
613 # echo 0 > /debug/tracing/tracing_enabled
614 # cat /debug/tracing/latency_trace
615# tracer: preemptoff
616#
617preemptoff latency trace v1.1.5 on 2.6.26-rc8
618--------------------------------------------------------------------
619 latency: 29 us, #3/3, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
620 -----------------
621 | task: sshd-4261 (uid:0 nice:0 policy:0 rt_prio:0)
622 -----------------
623 => started at: do_IRQ
624 => ended at: __do_softirq
625
626# _------=> CPU#
627# / _-----=> irqs-off
628# | / _----=> need-resched
629# || / _---=> hardirq/softirq
630# ||| / _--=> preempt-depth
631# |||| /
632# ||||| delay
633# cmd pid ||||| time | caller
634# \ / ||||| \ | /
635 sshd-4261 0d.h. 0us+: irq_enter (do_IRQ)
636 sshd-4261 0d.s. 29us : _local_bh_enable (__do_softirq)
637 sshd-4261 0d.s1 30us : trace_preempt_on (__do_softirq)
638
639
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400640This has some more changes. Preemption was disabled when an interrupt
641came in (notice the 'h'), and was enabled while doing a softirq.
642(notice the 's'). But we also see that interrupts have been disabled
643when entering the preempt off section and leaving it (the 'd').
644We do not know if interrupts were enabled in the mean time.
645
646# tracer: preemptoff
647#
648preemptoff latency trace v1.1.5 on 2.6.26-rc8
649--------------------------------------------------------------------
650 latency: 63 us, #87/87, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
651 -----------------
652 | task: sshd-4261 (uid:0 nice:0 policy:0 rt_prio:0)
653 -----------------
654 => started at: remove_wait_queue
655 => ended at: __do_softirq
656
657# _------=> CPU#
658# / _-----=> irqs-off
659# | / _----=> need-resched
660# || / _---=> hardirq/softirq
661# ||| / _--=> preempt-depth
662# |||| /
663# ||||| delay
664# cmd pid ||||| time | caller
665# \ / ||||| \ | /
666 sshd-4261 0d..1 0us : _spin_lock_irqsave (remove_wait_queue)
667 sshd-4261 0d..1 1us : _spin_unlock_irqrestore (remove_wait_queue)
668 sshd-4261 0d..1 2us : do_IRQ (common_interrupt)
669 sshd-4261 0d..1 2us : irq_enter (do_IRQ)
670 sshd-4261 0d..1 2us : idle_cpu (irq_enter)
671 sshd-4261 0d..1 3us : add_preempt_count (irq_enter)
672 sshd-4261 0d.h1 3us : idle_cpu (irq_enter)
673 sshd-4261 0d.h. 4us : handle_fasteoi_irq (do_IRQ)
674[...]
675 sshd-4261 0d.h. 12us : add_preempt_count (_spin_lock)
676 sshd-4261 0d.h1 12us : ack_ioapic_quirk_irq (handle_fasteoi_irq)
677 sshd-4261 0d.h1 13us : move_native_irq (ack_ioapic_quirk_irq)
678 sshd-4261 0d.h1 13us : _spin_unlock (handle_fasteoi_irq)
679 sshd-4261 0d.h1 14us : sub_preempt_count (_spin_unlock)
680 sshd-4261 0d.h1 14us : irq_exit (do_IRQ)
681 sshd-4261 0d.h1 15us : sub_preempt_count (irq_exit)
682 sshd-4261 0d..2 15us : do_softirq (irq_exit)
683 sshd-4261 0d... 15us : __do_softirq (do_softirq)
684 sshd-4261 0d... 16us : __local_bh_disable (__do_softirq)
685 sshd-4261 0d... 16us+: add_preempt_count (__local_bh_disable)
686 sshd-4261 0d.s4 20us : add_preempt_count (__local_bh_disable)
687 sshd-4261 0d.s4 21us : sub_preempt_count (local_bh_enable)
688 sshd-4261 0d.s5 21us : sub_preempt_count (local_bh_enable)
689[...]
690 sshd-4261 0d.s6 41us : add_preempt_count (__local_bh_disable)
691 sshd-4261 0d.s6 42us : sub_preempt_count (local_bh_enable)
692 sshd-4261 0d.s7 42us : sub_preempt_count (local_bh_enable)
693 sshd-4261 0d.s5 43us : add_preempt_count (__local_bh_disable)
694 sshd-4261 0d.s5 43us : sub_preempt_count (local_bh_enable_ip)
695 sshd-4261 0d.s6 44us : sub_preempt_count (local_bh_enable_ip)
696 sshd-4261 0d.s5 44us : add_preempt_count (__local_bh_disable)
697 sshd-4261 0d.s5 45us : sub_preempt_count (local_bh_enable)
698[...]
699 sshd-4261 0d.s. 63us : _local_bh_enable (__do_softirq)
700 sshd-4261 0d.s1 64us : trace_preempt_on (__do_softirq)
701
702
703The above is an example of the preemptoff trace with ftrace_enabled
704set. Here we see that interrupts were disabled the entire time.
705The irq_enter code lets us know that we entered an interrupt 'h'.
706Before that, the functions being traced still show that it is not
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400707in an interrupt, but we can see from the functions themselves that
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400708this is not the case.
709
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400710Notice that __do_softirq when called does not have a preempt_count.
711It may seem that we missed a preempt enabling. What really happened
712is that the preempt count is held on the thread's stack and we
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400713switched to the softirq stack (4K stacks in effect). The code
Steven Rostedta41eeba2008-07-14 16:41:12 -0400714does not copy the preempt count, but because interrupts are disabled,
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400715we do not need to worry about it. Having a tracer like this is good
716for letting people know what really happens inside the kernel.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400717
718
719preemptirqsoff
720--------------
721
722Knowing the locations that have interrupts disabled or preemption
723disabled for the longest times is helpful. But sometimes we would
724like to know when either preemption and/or interrupts are disabled.
725
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400726Consider the following code:
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400727
728 local_irq_disable();
729 call_function_with_irqs_off();
730 preempt_disable();
731 call_function_with_irqs_and_preemption_off();
732 local_irq_enable();
733 call_function_with_preemption_off();
734 preempt_enable();
735
736The irqsoff tracer will record the total length of
737call_function_with_irqs_off() and
738call_function_with_irqs_and_preemption_off().
739
740The preemptoff tracer will record the total length of
741call_function_with_irqs_and_preemption_off() and
742call_function_with_preemption_off().
743
744But neither will trace the time that interrupts and/or preemption
745is disabled. This total time is the time that we can not schedule.
746To record this time, use the preemptirqsoff tracer.
747
748Again, using this trace is much like the irqsoff and preemptoff tracers.
749
Steven Rostedta41eeba2008-07-14 16:41:12 -0400750 # echo preemptirqsoff > /debug/tracing/current_tracer
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400751 # echo 0 > /debug/tracing/tracing_max_latency
752 # echo 1 > /debug/tracing/tracing_enabled
753 # ls -ltr
754 [...]
755 # echo 0 > /debug/tracing/tracing_enabled
756 # cat /debug/tracing/latency_trace
757# tracer: preemptirqsoff
758#
759preemptirqsoff latency trace v1.1.5 on 2.6.26-rc8
760--------------------------------------------------------------------
761 latency: 293 us, #3/3, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
762 -----------------
763 | task: ls-4860 (uid:0 nice:0 policy:0 rt_prio:0)
764 -----------------
765 => started at: apic_timer_interrupt
766 => ended at: __do_softirq
767
768# _------=> CPU#
769# / _-----=> irqs-off
770# | / _----=> need-resched
771# || / _---=> hardirq/softirq
772# ||| / _--=> preempt-depth
773# |||| /
774# ||||| delay
775# cmd pid ||||| time | caller
776# \ / ||||| \ | /
777 ls-4860 0d... 0us!: trace_hardirqs_off_thunk (apic_timer_interrupt)
778 ls-4860 0d.s. 294us : _local_bh_enable (__do_softirq)
779 ls-4860 0d.s1 294us : trace_preempt_on (__do_softirq)
780
781
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400782
783The trace_hardirqs_off_thunk is called from assembly on x86 when
784interrupts are disabled in the assembly code. Without the function
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400785tracing, we do not know if interrupts were enabled within the preemption
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400786points. We do see that it started with preemption enabled.
787
788Here is a trace with ftrace_enabled set:
789
790
791# tracer: preemptirqsoff
792#
793preemptirqsoff latency trace v1.1.5 on 2.6.26-rc8
794--------------------------------------------------------------------
795 latency: 105 us, #183/183, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
796 -----------------
797 | task: sshd-4261 (uid:0 nice:0 policy:0 rt_prio:0)
798 -----------------
799 => started at: write_chan
800 => ended at: __do_softirq
801
802# _------=> CPU#
803# / _-----=> irqs-off
804# | / _----=> need-resched
805# || / _---=> hardirq/softirq
806# ||| / _--=> preempt-depth
807# |||| /
808# ||||| delay
809# cmd pid ||||| time | caller
810# \ / ||||| \ | /
811 ls-4473 0.N.. 0us : preempt_schedule (write_chan)
812 ls-4473 0dN.1 1us : _spin_lock (schedule)
813 ls-4473 0dN.1 2us : add_preempt_count (_spin_lock)
814 ls-4473 0d..2 2us : put_prev_task_fair (schedule)
815[...]
816 ls-4473 0d..2 13us : set_normalized_timespec (ktime_get_ts)
817 ls-4473 0d..2 13us : __switch_to (schedule)
818 sshd-4261 0d..2 14us : finish_task_switch (schedule)
819 sshd-4261 0d..2 14us : _spin_unlock_irq (finish_task_switch)
820 sshd-4261 0d..1 15us : add_preempt_count (_spin_lock_irqsave)
821 sshd-4261 0d..2 16us : _spin_unlock_irqrestore (hrtick_set)
822 sshd-4261 0d..2 16us : do_IRQ (common_interrupt)
823 sshd-4261 0d..2 17us : irq_enter (do_IRQ)
824 sshd-4261 0d..2 17us : idle_cpu (irq_enter)
825 sshd-4261 0d..2 18us : add_preempt_count (irq_enter)
826 sshd-4261 0d.h2 18us : idle_cpu (irq_enter)
827 sshd-4261 0d.h. 18us : handle_fasteoi_irq (do_IRQ)
828 sshd-4261 0d.h. 19us : _spin_lock (handle_fasteoi_irq)
829 sshd-4261 0d.h. 19us : add_preempt_count (_spin_lock)
830 sshd-4261 0d.h1 20us : _spin_unlock (handle_fasteoi_irq)
831 sshd-4261 0d.h1 20us : sub_preempt_count (_spin_unlock)
832[...]
833 sshd-4261 0d.h1 28us : _spin_unlock (handle_fasteoi_irq)
834 sshd-4261 0d.h1 29us : sub_preempt_count (_spin_unlock)
835 sshd-4261 0d.h2 29us : irq_exit (do_IRQ)
836 sshd-4261 0d.h2 29us : sub_preempt_count (irq_exit)
837 sshd-4261 0d..3 30us : do_softirq (irq_exit)
838 sshd-4261 0d... 30us : __do_softirq (do_softirq)
839 sshd-4261 0d... 31us : __local_bh_disable (__do_softirq)
840 sshd-4261 0d... 31us+: add_preempt_count (__local_bh_disable)
841 sshd-4261 0d.s4 34us : add_preempt_count (__local_bh_disable)
842[...]
843 sshd-4261 0d.s3 43us : sub_preempt_count (local_bh_enable_ip)
844 sshd-4261 0d.s4 44us : sub_preempt_count (local_bh_enable_ip)
845 sshd-4261 0d.s3 44us : smp_apic_timer_interrupt (apic_timer_interrupt)
846 sshd-4261 0d.s3 45us : irq_enter (smp_apic_timer_interrupt)
847 sshd-4261 0d.s3 45us : idle_cpu (irq_enter)
848 sshd-4261 0d.s3 46us : add_preempt_count (irq_enter)
849 sshd-4261 0d.H3 46us : idle_cpu (irq_enter)
850 sshd-4261 0d.H3 47us : hrtimer_interrupt (smp_apic_timer_interrupt)
851 sshd-4261 0d.H3 47us : ktime_get (hrtimer_interrupt)
852[...]
853 sshd-4261 0d.H3 81us : tick_program_event (hrtimer_interrupt)
854 sshd-4261 0d.H3 82us : ktime_get (tick_program_event)
855 sshd-4261 0d.H3 82us : ktime_get_ts (ktime_get)
856 sshd-4261 0d.H3 83us : getnstimeofday (ktime_get_ts)
857 sshd-4261 0d.H3 83us : set_normalized_timespec (ktime_get_ts)
858 sshd-4261 0d.H3 84us : clockevents_program_event (tick_program_event)
859 sshd-4261 0d.H3 84us : lapic_next_event (clockevents_program_event)
860 sshd-4261 0d.H3 85us : irq_exit (smp_apic_timer_interrupt)
861 sshd-4261 0d.H3 85us : sub_preempt_count (irq_exit)
862 sshd-4261 0d.s4 86us : sub_preempt_count (irq_exit)
863 sshd-4261 0d.s3 86us : add_preempt_count (__local_bh_disable)
864[...]
865 sshd-4261 0d.s1 98us : sub_preempt_count (net_rx_action)
866 sshd-4261 0d.s. 99us : add_preempt_count (_spin_lock_irq)
867 sshd-4261 0d.s1 99us+: _spin_unlock_irq (run_timer_softirq)
868 sshd-4261 0d.s. 104us : _local_bh_enable (__do_softirq)
869 sshd-4261 0d.s. 104us : sub_preempt_count (_local_bh_enable)
870 sshd-4261 0d.s. 105us : _local_bh_enable (__do_softirq)
871 sshd-4261 0d.s1 105us : trace_preempt_on (__do_softirq)
872
873
874This is a very interesting trace. It started with the preemption of
875the ls task. We see that the task had the "need_resched" bit set
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400876via the 'N' in the trace. Interrupts were disabled before the spin_lock
877at the beginning of the trace. We see that a schedule took place to run
Steven Rostedta41eeba2008-07-14 16:41:12 -0400878sshd. When the interrupts were enabled, we took an interrupt.
879On return from the interrupt handler, the softirq ran. We took another
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400880interrupt while running the softirq as we see from the capital 'H'.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400881
882
883wakeup
884------
885
Steven Rostedtf2d9c742008-07-15 10:57:33 -0400886In a Real-Time environment it is very important to know the wakeup
887time it takes for the highest priority task that is woken up to the
888time that it executes. This is also known as "schedule latency".
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400889I stress the point that this is about RT tasks. It is also important
890to know the scheduling latency of non-RT tasks, but the average
891schedule latency is better for non-RT tasks. Tools like
Steven Rostedta41eeba2008-07-14 16:41:12 -0400892LatencyTop are more appropriate for such measurements.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400893
Steven Rostedta41eeba2008-07-14 16:41:12 -0400894Real-Time environments are interested in the worst case latency.
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400895That is the longest latency it takes for something to happen, and
896not the average. We can have a very fast scheduler that may only
897have a large latency once in a while, but that would not work well
898with Real-Time tasks. The wakeup tracer was designed to record
899the worst case wakeups of RT tasks. Non-RT tasks are not recorded
900because the tracer only records one worst case and tracing non-RT
901tasks that are unpredictable will overwrite the worst case latency
902of RT tasks.
903
904Since this tracer only deals with RT tasks, we will run this slightly
Steven Rostedta41eeba2008-07-14 16:41:12 -0400905differently than we did with the previous tracers. Instead of performing
906an 'ls', we will run 'sleep 1' under 'chrt' which changes the
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400907priority of the task.
908
909 # echo wakeup > /debug/tracing/current_tracer
910 # echo 0 > /debug/tracing/tracing_max_latency
911 # echo 1 > /debug/tracing/tracing_enabled
912 # chrt -f 5 sleep 1
913 # echo 0 > /debug/tracing/tracing_enabled
914 # cat /debug/tracing/latency_trace
915# tracer: wakeup
916#
917wakeup latency trace v1.1.5 on 2.6.26-rc8
918--------------------------------------------------------------------
919 latency: 4 us, #2/2, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
920 -----------------
921 | task: sleep-4901 (uid:0 nice:0 policy:1 rt_prio:5)
922 -----------------
923
924# _------=> CPU#
925# / _-----=> irqs-off
926# | / _----=> need-resched
927# || / _---=> hardirq/softirq
928# ||| / _--=> preempt-depth
929# |||| /
930# ||||| delay
931# cmd pid ||||| time | caller
932# \ / ||||| \ | /
933 <idle>-0 1d.h4 0us+: try_to_wake_up (wake_up_process)
934 <idle>-0 1d..4 4us : schedule (cpu_idle)
935
936
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400937
Steven Rostedta41eeba2008-07-14 16:41:12 -0400938Running this on an idle system, we see that it only took 4 microseconds
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400939to perform the task switch. Note, since the trace marker in the
Steven Rostedta41eeba2008-07-14 16:41:12 -0400940schedule is before the actual "switch", we stop the tracing when
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400941the recorded task is about to schedule in. This may change if
942we add a new marker at the end of the scheduler.
943
944Notice that the recorded task is 'sleep' with the PID of 4901 and it
945has an rt_prio of 5. This priority is user-space priority and not
946the internal kernel priority. The policy is 1 for SCHED_FIFO and 2
947for SCHED_RR.
948
949Doing the same with chrt -r 5 and ftrace_enabled set.
950
951# tracer: wakeup
952#
953wakeup latency trace v1.1.5 on 2.6.26-rc8
954--------------------------------------------------------------------
955 latency: 50 us, #60/60, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)
956 -----------------
957 | task: sleep-4068 (uid:0 nice:0 policy:2 rt_prio:5)
958 -----------------
959
960# _------=> CPU#
961# / _-----=> irqs-off
962# | / _----=> need-resched
963# || / _---=> hardirq/softirq
964# ||| / _--=> preempt-depth
965# |||| /
966# ||||| delay
967# cmd pid ||||| time | caller
968# \ / ||||| \ | /
969ksoftirq-7 1d.H3 0us : try_to_wake_up (wake_up_process)
970ksoftirq-7 1d.H4 1us : sub_preempt_count (marker_probe_cb)
971ksoftirq-7 1d.H3 2us : check_preempt_wakeup (try_to_wake_up)
972ksoftirq-7 1d.H3 3us : update_curr (check_preempt_wakeup)
973ksoftirq-7 1d.H3 4us : calc_delta_mine (update_curr)
974ksoftirq-7 1d.H3 5us : __resched_task (check_preempt_wakeup)
975ksoftirq-7 1d.H3 6us : task_wake_up_rt (try_to_wake_up)
976ksoftirq-7 1d.H3 7us : _spin_unlock_irqrestore (try_to_wake_up)
977[...]
978ksoftirq-7 1d.H2 17us : irq_exit (smp_apic_timer_interrupt)
979ksoftirq-7 1d.H2 18us : sub_preempt_count (irq_exit)
980ksoftirq-7 1d.s3 19us : sub_preempt_count (irq_exit)
981ksoftirq-7 1..s2 20us : rcu_process_callbacks (__do_softirq)
982[...]
983ksoftirq-7 1..s2 26us : __rcu_process_callbacks (rcu_process_callbacks)
984ksoftirq-7 1d.s2 27us : _local_bh_enable (__do_softirq)
985ksoftirq-7 1d.s2 28us : sub_preempt_count (_local_bh_enable)
986ksoftirq-7 1.N.3 29us : sub_preempt_count (ksoftirqd)
987ksoftirq-7 1.N.2 30us : _cond_resched (ksoftirqd)
988ksoftirq-7 1.N.2 31us : __cond_resched (_cond_resched)
989ksoftirq-7 1.N.2 32us : add_preempt_count (__cond_resched)
990ksoftirq-7 1.N.2 33us : schedule (__cond_resched)
991ksoftirq-7 1.N.2 33us : add_preempt_count (schedule)
992ksoftirq-7 1.N.3 34us : hrtick_clear (schedule)
993ksoftirq-7 1dN.3 35us : _spin_lock (schedule)
994ksoftirq-7 1dN.3 36us : add_preempt_count (_spin_lock)
995ksoftirq-7 1d..4 37us : put_prev_task_fair (schedule)
996ksoftirq-7 1d..4 38us : update_curr (put_prev_task_fair)
997[...]
998ksoftirq-7 1d..5 47us : _spin_trylock (tracing_record_cmdline)
999ksoftirq-7 1d..5 48us : add_preempt_count (_spin_trylock)
1000ksoftirq-7 1d..6 49us : _spin_unlock (tracing_record_cmdline)
1001ksoftirq-7 1d..6 49us : sub_preempt_count (_spin_unlock)
1002ksoftirq-7 1d..4 50us : schedule (__cond_resched)
1003
1004The interrupt went off while running ksoftirqd. This task runs at
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001005SCHED_OTHER. Why did not we see the 'N' set early? This may be
Steven Rostedta41eeba2008-07-14 16:41:12 -04001006a harmless bug with x86_32 and 4K stacks. On x86_32 with 4K stacks
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001007configured, the interrupt and softirq run with their own stack.
Steven Rostedta41eeba2008-07-14 16:41:12 -04001008Some information is held on the top of the task's stack (need_resched
1009and preempt_count are both stored there). The setting of the NEED_RESCHED
1010bit is done directly to the task's stack, but the reading of the
1011NEED_RESCHED is done by looking at the current stack, which in this case
1012is the stack for the hard interrupt. This hides the fact that NEED_RESCHED
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001013has been set. We do not see the 'N' until we switch back to the task's
Steven Rostedta41eeba2008-07-14 16:41:12 -04001014assigned stack.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001015
Steven Rostedt9b803c02008-11-03 15:15:08 -05001016function
1017--------
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001018
Steven Rostedt9b803c02008-11-03 15:15:08 -05001019This tracer is the function tracer. Enabling the function tracer
1020can be done from the debug file system. Make sure the ftrace_enabled is
1021set; otherwise this tracer is a nop.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001022
1023 # sysctl kernel.ftrace_enabled=1
Steven Rostedt9b803c02008-11-03 15:15:08 -05001024 # echo function > /debug/tracing/current_tracer
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001025 # echo 1 > /debug/tracing/tracing_enabled
1026 # usleep 1
1027 # echo 0 > /debug/tracing/tracing_enabled
1028 # cat /debug/tracing/trace
Steven Rostedt9b803c02008-11-03 15:15:08 -05001029# tracer: function
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001030#
1031# TASK-PID CPU# TIMESTAMP FUNCTION
1032# | | | | |
1033 bash-4003 [00] 123.638713: finish_task_switch <-schedule
1034 bash-4003 [00] 123.638714: _spin_unlock_irq <-finish_task_switch
1035 bash-4003 [00] 123.638714: sub_preempt_count <-_spin_unlock_irq
1036 bash-4003 [00] 123.638715: hrtick_set <-schedule
1037 bash-4003 [00] 123.638715: _spin_lock_irqsave <-hrtick_set
1038 bash-4003 [00] 123.638716: add_preempt_count <-_spin_lock_irqsave
1039 bash-4003 [00] 123.638716: _spin_unlock_irqrestore <-hrtick_set
1040 bash-4003 [00] 123.638717: sub_preempt_count <-_spin_unlock_irqrestore
1041 bash-4003 [00] 123.638717: hrtick_clear <-hrtick_set
1042 bash-4003 [00] 123.638718: sub_preempt_count <-schedule
1043 bash-4003 [00] 123.638718: sub_preempt_count <-preempt_schedule
1044 bash-4003 [00] 123.638719: wait_for_completion <-__stop_machine_run
1045 bash-4003 [00] 123.638719: wait_for_common <-wait_for_completion
1046 bash-4003 [00] 123.638720: _spin_lock_irq <-wait_for_common
1047 bash-4003 [00] 123.638720: add_preempt_count <-_spin_lock_irq
1048[...]
1049
1050
Steven Rostedt9b803c02008-11-03 15:15:08 -05001051Note: function tracer uses ring buffers to store the above entries.
1052The newest data may overwrite the oldest data. Sometimes using echo to
1053stop the trace is not sufficient because the tracing could have overwritten
1054the data that you wanted to record. For this reason, it is sometimes better to
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001055disable tracing directly from a program. This allows you to stop the
1056tracing at the point that you hit the part that you are interested in.
1057To disable the tracing directly from a C program, something like following
1058code snippet can be used:
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001059
1060int trace_fd;
1061[...]
1062int main(int argc, char *argv[]) {
1063 [...]
1064 trace_fd = open("/debug/tracing/tracing_enabled", O_WRONLY);
1065 [...]
1066 if (condition_hit()) {
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001067 write(trace_fd, "0", 1);
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001068 }
1069 [...]
1070}
1071
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001072Note: Here we hard coded the path name. The debugfs mount is not
1073guaranteed to be at /debug (and is more commonly at /sys/kernel/debug).
1074For simple one time traces, the above is sufficent. For anything else,
1075a search through /proc/mounts may be needed to find where the debugfs
1076file-system is mounted.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001077
Steven Rostedtdf4fc312008-11-26 00:16:23 -05001078
1079Single thread tracing
1080---------------------
1081
1082By writing into /debug/tracing/set_ftrace_pid you can trace a
1083single thread. For example:
1084
1085# cat /debug/tracing/set_ftrace_pid
1086no pid
1087# echo 3111 > /debug/tracing/set_ftrace_pid
1088# cat /debug/tracing/set_ftrace_pid
10893111
1090# echo function > /debug/tracing/current_tracer
1091# cat /debug/tracing/trace | head
1092 # tracer: function
1093 #
1094 # TASK-PID CPU# TIMESTAMP FUNCTION
1095 # | | | | |
1096 yum-updatesd-3111 [003] 1637.254676: finish_task_switch <-thread_return
1097 yum-updatesd-3111 [003] 1637.254681: hrtimer_cancel <-schedule_hrtimeout_range
1098 yum-updatesd-3111 [003] 1637.254682: hrtimer_try_to_cancel <-hrtimer_cancel
1099 yum-updatesd-3111 [003] 1637.254683: lock_hrtimer_base <-hrtimer_try_to_cancel
1100 yum-updatesd-3111 [003] 1637.254685: fget_light <-do_sys_poll
1101 yum-updatesd-3111 [003] 1637.254686: pipe_poll <-do_sys_poll
1102# echo -1 > /debug/tracing/set_ftrace_pid
1103# cat /debug/tracing/trace |head
1104 # tracer: function
1105 #
1106 # TASK-PID CPU# TIMESTAMP FUNCTION
1107 # | | | | |
1108 ##### CPU 3 buffer started ####
1109 yum-updatesd-3111 [003] 1701.957688: free_poll_entry <-poll_freewait
1110 yum-updatesd-3111 [003] 1701.957689: remove_wait_queue <-free_poll_entry
1111 yum-updatesd-3111 [003] 1701.957691: fput <-free_poll_entry
1112 yum-updatesd-3111 [003] 1701.957692: audit_syscall_exit <-sysret_audit
1113 yum-updatesd-3111 [003] 1701.957693: path_put <-audit_syscall_exit
1114
1115If you want to trace a function when executing, you could use
1116something like this simple program:
1117
1118#include <stdio.h>
1119#include <stdlib.h>
1120#include <sys/types.h>
1121#include <sys/stat.h>
1122#include <fcntl.h>
1123#include <unistd.h>
1124
1125int main (int argc, char **argv)
1126{
1127 if (argc < 1)
1128 exit(-1);
1129
1130 if (fork() > 0) {
1131 int fd, ffd;
1132 char line[64];
1133 int s;
1134
1135 ffd = open("/debug/tracing/current_tracer", O_WRONLY);
1136 if (ffd < 0)
1137 exit(-1);
1138 write(ffd, "nop", 3);
1139
1140 fd = open("/debug/tracing/set_ftrace_pid", O_WRONLY);
1141 s = sprintf(line, "%d\n", getpid());
1142 write(fd, line, s);
1143
1144 write(ffd, "function", 8);
1145
1146 close(fd);
1147 close(ffd);
1148
1149 execvp(argv[1], argv+1);
1150 }
1151
1152 return 0;
1153}
1154
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001155dynamic ftrace
1156--------------
1157
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001158If CONFIG_DYNAMIC_FTRACE is set, the system will run with
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001159virtually no overhead when function tracing is disabled. The way
1160this works is the mcount function call (placed at the start of
1161every kernel function, produced by the -pg switch in gcc), starts
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001162of pointing to a simple return. (Enabling FTRACE will include the
1163-pg switch in the compiling of the kernel.)
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001164
Steven Rostedt9b803c02008-11-03 15:15:08 -05001165At compile time every C file object is run through the
1166recordmcount.pl script (located in the scripts directory). This
1167script will process the C object using objdump to find all the
1168locations in the .text section that call mcount. (Note, only
1169the .text section is processed, since processing other sections
1170like .init.text may cause races due to those sections being freed).
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001171
Steven Rostedt9b803c02008-11-03 15:15:08 -05001172A new section called "__mcount_loc" is created that holds references
1173to all the mcount call sites in the .text section. This section is
1174compiled back into the original object. The final linker will add
1175all these references into a single table.
1176
1177On boot up, before SMP is initialized, the dynamic ftrace code
1178scans this table and updates all the locations into nops. It also
1179records the locations, which are added to the available_filter_functions
1180list. Modules are processed as they are loaded and before they are
1181executed. When a module is unloaded, it also removes its functions from
1182the ftrace function list. This is automatic in the module unload
1183code, and the module author does not need to worry about it.
1184
1185When tracing is enabled, kstop_machine is called to prevent races
1186with the CPUS executing code being modified (which can cause the
1187CPU to do undesireable things), and the nops are patched back
1188to calls. But this time, they do not call mcount (which is just
1189a function stub). They now call into the ftrace infrastructure.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001190
1191One special side-effect to the recording of the functions being
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001192traced is that we can now selectively choose which functions we
1193wish to trace and which ones we want the mcount calls to remain as
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001194nops.
1195
Steven Rostedta41eeba2008-07-14 16:41:12 -04001196Two files are used, one for enabling and one for disabling the tracing
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001197of specified functions. They are:
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001198
1199 set_ftrace_filter
1200
1201and
1202
1203 set_ftrace_notrace
1204
Steven Rostedta41eeba2008-07-14 16:41:12 -04001205A list of available functions that you can add to these files is listed
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001206in:
1207
1208 available_filter_functions
1209
1210 # cat /debug/tracing/available_filter_functions
1211put_prev_task_idle
1212kmem_cache_create
1213pick_next_task_rt
1214get_online_cpus
1215pick_next_task_fair
1216mutex_lock
1217[...]
1218
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001219If I am only interested in sys_nanosleep and hrtimer_interrupt:
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001220
1221 # echo sys_nanosleep hrtimer_interrupt \
1222 > /debug/tracing/set_ftrace_filter
1223 # echo ftrace > /debug/tracing/current_tracer
1224 # echo 1 > /debug/tracing/tracing_enabled
1225 # usleep 1
1226 # echo 0 > /debug/tracing/tracing_enabled
1227 # cat /debug/tracing/trace
1228# tracer: ftrace
1229#
1230# TASK-PID CPU# TIMESTAMP FUNCTION
1231# | | | | |
1232 usleep-4134 [00] 1317.070017: hrtimer_interrupt <-smp_apic_timer_interrupt
1233 usleep-4134 [00] 1317.070111: sys_nanosleep <-syscall_call
1234 <idle>-0 [00] 1317.070115: hrtimer_interrupt <-smp_apic_timer_interrupt
1235
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001236To see which functions are being traced, you can cat the file:
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001237
1238 # cat /debug/tracing/set_ftrace_filter
1239hrtimer_interrupt
1240sys_nanosleep
1241
1242
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001243Perhaps this is not enough. The filters also allow simple wild cards.
Steven Rostedta41eeba2008-07-14 16:41:12 -04001244Only the following are currently available
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001245
Steven Rostedta41eeba2008-07-14 16:41:12 -04001246 <match>* - will match functions that begin with <match>
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001247 *<match> - will match functions that end with <match>
1248 *<match>* - will match functions that have <match> in it
1249
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001250These are the only wild cards which are supported.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001251
1252 <match>*<match> will not work.
1253
walimisc072c242008-11-28 12:21:19 +08001254Note: It is better to use quotes to enclose the wild cards, otherwise
1255 the shell may expand the parameters into names of files in the local
1256 directory.
1257
1258 # echo 'hrtimer_*' > /debug/tracing/set_ftrace_filter
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001259
1260Produces:
1261
1262# tracer: ftrace
1263#
1264# TASK-PID CPU# TIMESTAMP FUNCTION
1265# | | | | |
1266 bash-4003 [00] 1480.611794: hrtimer_init <-copy_process
1267 bash-4003 [00] 1480.611941: hrtimer_start <-hrtick_set
1268 bash-4003 [00] 1480.611956: hrtimer_cancel <-hrtick_clear
1269 bash-4003 [00] 1480.611956: hrtimer_try_to_cancel <-hrtimer_cancel
1270 <idle>-0 [00] 1480.612019: hrtimer_get_next_event <-get_next_timer_interrupt
1271 <idle>-0 [00] 1480.612025: hrtimer_get_next_event <-get_next_timer_interrupt
1272 <idle>-0 [00] 1480.612032: hrtimer_get_next_event <-get_next_timer_interrupt
1273 <idle>-0 [00] 1480.612037: hrtimer_get_next_event <-get_next_timer_interrupt
1274 <idle>-0 [00] 1480.612382: hrtimer_get_next_event <-get_next_timer_interrupt
1275
1276
1277Notice that we lost the sys_nanosleep.
1278
1279 # cat /debug/tracing/set_ftrace_filter
1280hrtimer_run_queues
1281hrtimer_run_pending
1282hrtimer_init
1283hrtimer_cancel
1284hrtimer_try_to_cancel
1285hrtimer_forward
1286hrtimer_start
1287hrtimer_reprogram
1288hrtimer_force_reprogram
1289hrtimer_get_next_event
1290hrtimer_interrupt
1291hrtimer_nanosleep
1292hrtimer_wakeup
1293hrtimer_get_remaining
1294hrtimer_get_res
1295hrtimer_init_sleeper
1296
1297
1298This is because the '>' and '>>' act just like they do in bash.
1299To rewrite the filters, use '>'
1300To append to the filters, use '>>'
1301
Steven Rostedta41eeba2008-07-14 16:41:12 -04001302To clear out a filter so that all functions will be recorded again:
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001303
1304 # echo > /debug/tracing/set_ftrace_filter
1305 # cat /debug/tracing/set_ftrace_filter
1306 #
1307
1308Again, now we want to append.
1309
1310 # echo sys_nanosleep > /debug/tracing/set_ftrace_filter
1311 # cat /debug/tracing/set_ftrace_filter
1312sys_nanosleep
walimisc072c242008-11-28 12:21:19 +08001313 # echo 'hrtimer_*' >> /debug/tracing/set_ftrace_filter
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001314 # cat /debug/tracing/set_ftrace_filter
1315hrtimer_run_queues
1316hrtimer_run_pending
1317hrtimer_init
1318hrtimer_cancel
1319hrtimer_try_to_cancel
1320hrtimer_forward
1321hrtimer_start
1322hrtimer_reprogram
1323hrtimer_force_reprogram
1324hrtimer_get_next_event
1325hrtimer_interrupt
1326sys_nanosleep
1327hrtimer_nanosleep
1328hrtimer_wakeup
1329hrtimer_get_remaining
1330hrtimer_get_res
1331hrtimer_init_sleeper
1332
1333
1334The set_ftrace_notrace prevents those functions from being traced.
1335
1336 # echo '*preempt*' '*lock*' > /debug/tracing/set_ftrace_notrace
1337
1338Produces:
1339
1340# tracer: ftrace
1341#
1342# TASK-PID CPU# TIMESTAMP FUNCTION
1343# | | | | |
1344 bash-4043 [01] 115.281644: finish_task_switch <-schedule
1345 bash-4043 [01] 115.281645: hrtick_set <-schedule
1346 bash-4043 [01] 115.281645: hrtick_clear <-hrtick_set
1347 bash-4043 [01] 115.281646: wait_for_completion <-__stop_machine_run
1348 bash-4043 [01] 115.281647: wait_for_common <-wait_for_completion
1349 bash-4043 [01] 115.281647: kthread_stop <-stop_machine_run
1350 bash-4043 [01] 115.281648: init_waitqueue_head <-kthread_stop
1351 bash-4043 [01] 115.281648: wake_up_process <-kthread_stop
1352 bash-4043 [01] 115.281649: try_to_wake_up <-wake_up_process
1353
1354We can see that there's no more lock or preempt tracing.
1355
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001356trace_pipe
1357----------
1358
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001359The trace_pipe outputs the same content as the trace file, but the effect
1360on the tracing is different. Every read from trace_pipe is consumed.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001361This means that subsequent reads will be different. The trace
1362is live.
1363
Steven Rostedt9b803c02008-11-03 15:15:08 -05001364 # echo function > /debug/tracing/current_tracer
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001365 # cat /debug/tracing/trace_pipe > /tmp/trace.out &
1366[1] 4153
1367 # echo 1 > /debug/tracing/tracing_enabled
1368 # usleep 1
1369 # echo 0 > /debug/tracing/tracing_enabled
1370 # cat /debug/tracing/trace
Steven Rostedt9b803c02008-11-03 15:15:08 -05001371# tracer: function
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001372#
1373# TASK-PID CPU# TIMESTAMP FUNCTION
1374# | | | | |
1375
1376 #
1377 # cat /tmp/trace.out
1378 bash-4043 [00] 41.267106: finish_task_switch <-schedule
1379 bash-4043 [00] 41.267106: hrtick_set <-schedule
1380 bash-4043 [00] 41.267107: hrtick_clear <-hrtick_set
1381 bash-4043 [00] 41.267108: wait_for_completion <-__stop_machine_run
1382 bash-4043 [00] 41.267108: wait_for_common <-wait_for_completion
1383 bash-4043 [00] 41.267109: kthread_stop <-stop_machine_run
1384 bash-4043 [00] 41.267109: init_waitqueue_head <-kthread_stop
1385 bash-4043 [00] 41.267110: wake_up_process <-kthread_stop
1386 bash-4043 [00] 41.267110: try_to_wake_up <-wake_up_process
1387 bash-4043 [00] 41.267111: select_task_rq_rt <-try_to_wake_up
1388
1389
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001390Note, reading the trace_pipe file will block until more input is added.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001391By changing the tracer, trace_pipe will issue an EOF. We needed
Steven Rostedt9b803c02008-11-03 15:15:08 -05001392to set the function tracer _before_ we "cat" the trace_pipe file.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001393
1394
1395trace entries
1396-------------
1397
1398Having too much or not enough data can be troublesome in diagnosing
Steven Rostedta94c80e2008-11-12 17:52:36 -05001399an issue in the kernel. The file buffer_size_kb is used to modify
Steven Rostedta41eeba2008-07-14 16:41:12 -04001400the size of the internal trace buffers. The number listed
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001401is the number of entries that can be recorded per CPU. To know
1402the full size, multiply the number of possible CPUS with the
1403number of entries.
1404
Steven Rostedta94c80e2008-11-12 17:52:36 -05001405 # cat /debug/tracing/buffer_size_kb
Steven Rostedt1696b2b2008-11-13 00:09:35 -050014061408 (units kilobytes)
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001407
Steven Rostedta41eeba2008-07-14 16:41:12 -04001408Note, to modify this, you must have tracing completely disabled. To do that,
Steven Rostedt9b803c02008-11-03 15:15:08 -05001409echo "nop" into the current_tracer. If the current_tracer is not set
1410to "nop", an EINVAL error will be returned.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001411
Steven Rostedt9b803c02008-11-03 15:15:08 -05001412 # echo nop > /debug/tracing/current_tracer
Steven Rostedt1696b2b2008-11-13 00:09:35 -05001413 # echo 10000 > /debug/tracing/buffer_size_kb
Steven Rostedta94c80e2008-11-12 17:52:36 -05001414 # cat /debug/tracing/buffer_size_kb
Steven Rostedt1696b2b2008-11-13 00:09:35 -0500141510000 (units kilobytes)
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001416
Steven Rostedtf2d9c742008-07-15 10:57:33 -04001417The number of pages which will be allocated is limited to a percentage
1418of available memory. Allocating too much will produce an error.
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001419
Steven Rostedta94c80e2008-11-12 17:52:36 -05001420 # echo 1000000000000 > /debug/tracing/buffer_size_kb
Steven Rostedteb6d42e2008-07-10 12:46:01 -04001421-bash: echo: write error: Cannot allocate memory
Steven Rostedta94c80e2008-11-12 17:52:36 -05001422 # cat /debug/tracing/buffer_size_kb
Steven Rostedteb6d42e2008-07-10 12:46:01 -0400142385
1424