1

ftrace Function Tracer ======================== Copyright 2008 Red Hat Inc. Author: Steven Rostedt <[email protected]> License: The GNU Free Documentation License, Version 1.2 (dual licensed under the GPL v2) Reviewers: Elias Oltmanns, Randy Dunlap, Andrew Morton, John Kacur, and David Teigland. Written for: 2.6.28 rc2 Introduction

10

15

Ftrace is an internal tracer designed to help out developers and designers of systems to nd what is going on inside the kernel. It can be used for debugging or analyzing latencies and performance issues that take place outside of user space. Although ftrace is the function tracer, it also includes an infrastructure that allows for other types of tracing. Some of the tracers that are currently in ftrace include a tracer to trace context switches, the time it takes for a high priority task to run after it was woken up, the time interrupts are disabled, and more (ftrace allows for tracer plugins, which means that the list of tracers can always grow).

20

25

Implementation Details
30

See ftrace design.txt for details for arch porters and such.

35

The File System

Ftrace uses the debugfs le system to hold the control les as well as the les to display output.
40

When debugfs is congured into the kernel (which selecting any ftrace option will do) the directory /sys/kernel/debug will be created. To mount this directory, you can add to your /etc/fstab le:
45

debugfs

/sys/kernel/debug

debugfs defaults

Or you can mount it at run time with: mount t debugfs nodev /sys/kernel/debug
50

For quicker access to that directory you may want to make a soft link to it: ln s /sys/kernel/debug /debug
55

60

Any selected ftrace option will also create a directory called tracing within the debugfs. The rest of the document will assume that you are in the ftrace directory (cd /sys/kernel/debug/tracing) and will only concentrate on the les within that directory and not distract from the content with the extended "/sys/kernel/debug/tracing" path name. Thats it! (assuming that you have ftrace congured into your kernel)

linux/Documentation/trace/ftrace.txt (v2.6.39)

65

After mounting the debugfs, you can see a directory called "tracing". This directory contains the control and output les of ftrace. Here is a list of some of the key les:

Note: all time values are in microseconds.

current_tracer: This is used to set or display the current tracer that is congured.
75

available_tracers: This holds the dierent types of tracers that have been compiled into the kernel. The tracers listed here can be congured by echoing their name into current_tracer. tracing_on:
85

80

This sets or displays whether writing to the trace ring buer is enabled. Echo 0 into this le to disable the tracer or 1 to enable it. trace:

90

This le holds the output of the trace in a human readable format (described below). trace_pipe:
95

100

105

The output is the same as the "trace" le but this le is meant to be streamed with live tracing. Reads from this le will block until new data is retrieved. Unlike the "trace" le, this le is a consumer. This means reading from this le causes sequential reads to display more current data. Once data is read from this le, it is consumed, and will not be read again with a sequential read. The "trace" le is static, and if the tracer is not adding more data,they will display the same information every time they are read. trace_options:

110

This le lets the user control the amount of data that is displayed in one of the above output les. tracing_max_latency:

115

120

Some of the tracers record the max latency. For example, the time interrupts are disabled. This time is saved in this le. The max trace will also be stored, and displayed by "trace". A new max trace will only be recorded if the latency is greater than the value in this le. (in microseconds) buer_size_kb:

125

This sets or displays the number of kilobytes each CPU

linux/Documentation/trace/ftrace.txt (v2.6.39)

130

135

buer can hold. The tracer buers are the same size for each CPU. The displayed number is the size of the CPU buer and not total size of all buers. The trace buers are allocated in pages (blocks of memory that the kernel uses for allocation, usually 4 KB in size). If the last page allocated has room for more bytes than requested, the rest of the page will be used, making the actual allocation bigger than requested. ( Note, the size may not be a multiple of the page size due to buer management overhead. ) This can only be updated when the current_tracer is set to "nop".

140

tracing_cpumask: This is a mask that lets the user only trace on specied CPUS. The format is a hex string representing the CPUS. set_ftrace_lter: When dynamic ftrace is congured in (see the section below "dynamic ftrace"), the code is dynamically modied (code text rewrite) to disable calling of the function proler (mcount). This lets tracing be congured in with practically no overhead in performance. This also has a side eect of enabling or disabling specic functions to be traced. Echoing names of functions into this le will limit the trace to only those functions. This interface also allows for commands to be used. See the "Filter commands" section for more details.
160

145

150

155

set_ftrace_notrace: This has an eect opposite to that of set_ftrace_lter. Any function that is added here will not be traced. If a function exists in both set_ftrace_lter and set_ftrace_notrace, the function will _not_ be traced. set_ftrace_pid:
170

165

Have the function tracer only trace a single thread. set_graph_function: Set a "trigger" function where tracing should start with the function graph tracer (See the section "dynamic ftrace" for more details). available_lter_functions:

175

180

This lists the functions that ftrace has processed and can trace. These are the function names that you can pass to "set_ftrace_lter" or "set_ftrace_notrace". (See the section "dynamic ftrace" below for more details.)

185

The Tracers

linux/Documentation/trace/ftrace.txt (v2.6.39)

190

Here is the list of current tracers that may be congured. "function" Function call tracer to trace all kernel functions.

195

"function_graph" Similar to the function tracer except that the function tracer probes the functions on their entry whereas the function graph tracer traces on both entry and exit of the functions. It then provides the ability to draw a graph of function calls similar to C code source. "irqso" Traces the areas that disable interrupts and saves the trace with the longest max latency. See tracing_max_latency. When a new max is recorded, it replaces the old trace. It is best to view this trace with the latency format option enabled. "preempto"
215

200

205

210

Similar to irqso but traces and records the amount of time for which preemption is disabled. "preemptirqso"

220

Similar to irqso and preempto, but traces and records the largest time for which irqs and/or preemption is disabled. "wakeup"

225

Traces and records the max latency that it takes for the highest priority task to get scheduled after it has been woken up.
230

"hw branch tracer" Uses the BTS CPU feature on x86 CPUs to traces all branches executed.

235

"nop" This is the "trace nothing" tracer. To remove all tracers from tracing simply echo "nop" into current_tracer.

240

Examples of using the tracer

245

Here are typical examples of using the tracers when controlling them only with the debugfs interface (without using any user land utilities). Output format:

250

Here is an example of the output format of the le "trace"

linux/Documentation/trace/ftrace.txt (v2.6.39)

255

260

# tracer: function # # TASK PID # || bash 4251 [01] bash 4251 [01] bash 4251 [01]

CPU# TIMESTAMP FUNCTION | | | 10152.583854: path_put < path_walk 10152.583855: dput < path_put 10152.583855: _atomic_dec_and_lock < dput

265

270

A header is printed with the tracer name that is represented by the trace. In this case the tracer is "function". Then a header showing the format. Task name "bash", the task PID "4251", the CPU that it was running on "01", the timestamp in <secs>.<usecs> format, the function name that was traced "path_put" and the parent function that called this function "path_walk". The timestamp is the time at which the function was entered. Latency trace format

275

When the latency format option is enabled, the trace le gives somewhat more information to see why a latency happened. Here is a typical trace. # tracer: irqso # irqso latency trace v1.1.5 on 2.6.26 rc8 latency: 97 us, #3/3, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)

280

285

| task: swapper 0 (uid:0 nice:0 policy:0 rt_prio:0) => started at: apic_timer_interrupt => ended at: do_softirq

290

295

300

# # # # # # # # cmd # <idle> <idle> <idle>

pid \ / 0 0 0

_ /_ |/_ || / _ ||| / _ |||| / ||||| ||||| time ||||| 0d..1 0d.s. 0d.s1

=> CPU# => irqs o => need resched => hardirq/softirq => preempt depth delay | caller \ | / 0us+: trace_hardirqs_o_thunk (apic_timer_interrupt) 97us : __do_softirq (do_softirq) 98us : trace_hardirqs_on (do_softirq)

305

310

This shows that the current tracer is "irqso" tracing the time for which interrupts were disabled. It gives the trace version and the version of the kernel upon which this was executed on (2.6.26 rc8). Then it displays the max latency in microsecs (97 us). The number of trace entries displayed and the total number recorded (both are three: #3/3). The type of preemption that was used (PREEMPT). VP, KP, SP, and HP are always zero and are reserved for later use. #P is the number of online CPUS (#P:2). The task is the process that was running when the latency occurred. (swapper pid: 0).

315

linux/Documentation/trace/ftrace.txt (v2.6.39)

The start and stop (the functions in which the interrupts were disabled and enabled respectively) that caused the latencies: apic_timer_interrupt is where the interrupts were disabled. do_softirq is where they were enabled again. The next lines after the header are the trace itself. The header explains which is which.
325

320

cmd: The name of the process in the trace. pid: The PID of that process. CPU#: The CPU which the process was running on.

330

irqs o: d interrupts are disabled. . otherwise. Note: If the architecture does not support a way to read the irq ags variable, an X will always be printed here.
335

need resched: N task need_resched is set, . otherwise. hardirq/softirq: H h s .

340

hard irq occurred inside a softirq. hard irq is running soft irq is running normal context.

preempt depth: The level of preempt_disabled

The above is mostly meaningful for kernel developers. time: When the latency format option is enabled, the trace le output includes a timestamp relative to the start of the trace. This diers from the output when latency format is disabled, which includes an absolute timestamp. delay: This is just to help catch your eye a bit better. And needs to be xed to be only relative to the same CPU. The marks are determined by the dierence between this current trace and the next trace. ! greater than preempt_mark_thresh (default 100) + greater than 1 microsecond less than or equal to 1 microsecond. The rest is the same as the trace le.

350

355

360

trace_options
365

The trace_options le is used to control what gets printed in the trace output. To see what is available, simply cat the le:
370

cat trace_options print parent nosym oset nosym addr noverbose noraw nohex nobin \ noblock nostacktrace nosched tree nouserstacktrace nosym userobj To disable one of the options, echo in the option prepended with "no". echo noprint parent > trace_options

375

linux/Documentation/trace/ftrace.txt (v2.6.39)

To enable an option, leave o the "no".

echo sym oset > trace_options Here are the available options:
385

print parent

On function traces, display the calling (parent) function as well as the function being traced.

print parent: bash 4000 [01] 1477.606694: simple_strtoul < strict_strtoul

noprint parent: bash 4000 [01] 1477.606694: simple_strtoul

395

sym oset

Display not only the function name, but also the oset in the function. For example, instead of seeing just "ktime_get", you will see "ktime_get+0xb/0x20".

400

sym oset: bash 4000 [01] 1477.606694: simple_strtoul+0x6/0xa0 sym addr this will also display the function address as well as the function name.

405

sym addr: bash 4000 [01] 1477.606694: simple_strtoul <c0339346> verbose

This deals with the trace le when the latency format option is enabled.

bash 4000 1 0 00000000 00010a95 [58127d26] 1720.415ms \ (+0.000ms): simple_strtoul (strict_strtoul)

raw

This will display raw numbers. This option is best for use with user applications that can translate the raw numbers better than having it done in the kernel. Similar to raw, but the numbers will be in a hexadecimal format. This will print out the formats in raw binary. TBD (needs update) This is one of the options that changes the trace itself. When a trace is recorded, so is the stack of functions. This allows for back traces of trace sites. This option changes the trace. It records a stacktrace of the current userspace thread. when user stacktrace are enabled, look up which object the address belongs to, and print a relative address. This is especially useful when ASLR is on, otherwise you dont get a chance to resolve the address to object/le/line after the app is no longer running The lookup is performed when you read

hex
420

bin block
425

stacktrace

430

userstacktrace

sym userobj
435

440

linux/Documentation/trace/ftrace.txt (v2.6.39)

trace,trace_pipe. Example: a.out 1623 [000] 40874.465068: /root/a.out[+0x480] < /root/a.out[+0 x494] < /root/a.out[+0x4a8] < /lib/libc 2.7.so[+0x1e1a6] sched tree trace all tasks that are on the runqueue, at every scheduling event. Will add overhead if theres a lot of tasks running at once. This option changes the trace. When it is enabled, the trace displays additional information about the latencies, as described in "Latency trace format".

445

450

latency format

455

overwrite

460

This controls what happens when the trace buer is full. If "1" (default), the oldest events are discarded and overwritten. If "0", then the newest events are discarded.

ftrace_enabled

465

The following tracers (listed below) give dierent output depending on whether or not the sysctl ftrace_enabled is set. To set ftrace_enabled, one can either use the sysctl function or set it via the proc le system interface. sysctl kernel.ftrace_enabled=1 or echo 1 > /proc/sys/kernel/ftrace_enabled

470

475

To disable ftrace_enabled simply replace the 1 with 0 in the above commands. When ftrace_enabled is set the tracers will also record the functions that are within the trace. The descriptions of the tracers will also show an example with ftrace enabled.

480

irqso
485

490

When interrupts are disabled, the CPU can not react to any other external event (besides NMIs and SMIs). This prevents the timer interrupt from triggering or the mouse interrupt from letting the kernel know of a new mouse event. The result is a latency with the reaction time. The irqso tracer tracks the time for which interrupts are disabled. When a new maximum latency is hit, the tracer saves the trace leading up to that latency point so that every time a new maximum is reached, the old saved trace is discarded and the new trace is saved. To reset the maximum, echo 0 into tracing_max_latency. Here is an example: # echo irqso > current_tracer # echo latency format > trace_options # echo 0 > tracing_max_latency

495

500

linux/Documentation/trace/ftrace.txt (v2.6.39)

505

510

# echo 1 > tracing_on # ls ltr [...] # echo 0 > tracing_on # cat trace # tracer: irqso # irqso latency trace v1.1.5 on 2.6.26 latency: 12 us, #3/3, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)

515

| task: bash 3730 (uid:0 nice:0 policy:0 rt_prio:0) => started at: sys_setpgid => ended at: sys_setpgid
520

525

530

# # # # # # # # cmd # \ bash bash bash

pid / 3730 3730 3730

_ /_ |/_ || / _ ||| / _ |||| / ||||| ||||| time ||||| 1d... 1d..1 1d..2

=> CPU# => irqs o => need resched => hardirq/softirq => preempt depth delay | caller \ | / 0us : _write_lock_irq (sys_setpgid) 1us+: _write_unlock_irq (sys_setpgid) 14us : trace_hardirqs_on (sys_setpgid)

535

540

Here we see that that we had a latency of 12 microsecs (which is very good). The _write_lock_irq in sys_setpgid disabled interrupts. The dierence between the 12 and the displayed timestamp 14us occurred because the clock was incremented between the time of recording the max latency and the time of recording the function that had that latency. Note the above example had ftrace_enabled not set. If we set the ftrace_enabled, we get a much larger output:

545

# tracer: irqso # irqso latency trace v1.1.5 on 2.6.26 rc8 latency: 50 us, #101/101, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)

550

| task: ls 4339 (uid:0 nice:0 policy:0 rt_prio:0) => started at: __alloc_pages_internal => ended at: __alloc_pages_internal
555

560

565

# # # # # # # # cmd # \ ls ls ls

pid / 4339 4339 4339

_ /_ |/_ || / _ ||| / _ |||| / ||||| ||||| time ||||| 0...1 0d..1 0d..1

=> CPU# => irqs o => need resched => hardirq/softirq => preempt depth delay | caller \ | / 0us+: get_page_from_freelist (__alloc_pages_internal) 3us : rmqueue_bulk (get_page_from_freelist) 3us : _spin_lock (rmqueue_bulk)

linux/Documentation/trace/ftrace.txt (v2.6.39)

570

575

ls ls ls ls ls ls ls ls ls [...] ls ls ls ls ls ls ls ls ls

4339 4339 4339 4339 4339 4339 4339 4339 4339 4339 4339 4339 4339 4339 4339 4339 4339 4339 0d..2 0d..2 0d..2 0d..2 0d..2 0d..2 0d..2 0d..1 0d..2

0d..1 0d..2 0d..2 0d..2 0d..2 0d..2 0d..2 0d..2 0d..2

4us 4us 5us 5us 6us 6us 7us 7us 8us 46us 47us 47us 48us 48us 49us 49us 50us 51us

: : : : : : : : : : : : : : : : : :

add_preempt_count (_spin_lock) __rmqueue (rmqueue_bulk) __rmqueue_smallest (__rmqueue) __mod_zone_page_state (__rmqueue_smallest) __rmqueue (rmqueue_bulk) __rmqueue_smallest (__rmqueue) __mod_zone_page_state (__rmqueue_smallest) __rmqueue (rmqueue_bulk) __rmqueue_smallest (__rmqueue) __rmqueue_smallest (__rmqueue) __mod_zone_page_state (__rmqueue_smallest) __rmqueue (rmqueue_bulk) __rmqueue_smallest (__rmqueue) __mod_zone_page_state (__rmqueue_smallest) _spin_unlock (rmqueue_bulk) sub_preempt_count (_spin_unlock) get_page_from_freelist (__alloc_pages_internal) trace_hardirqs_on (__alloc_pages_internal)

580

585

590

Here we traced a 50 microsecond latency. But we also see all the functions that were called during that time. Note that by enabling function tracing, we incur an added overhead. This overhead may extend the latency times. But nevertheless, this trace has provided some very helpful debugging information.

595

preempto

600

When preemption is disabled, we may be able to receive interrupts but the task cannot be preempted and a higher priority task must wait for preemption to be enabled again before it can preempt a lower priority task. The preempto tracer traces the places that disable preemption. Like the irqso tracer, it records the maximum latency for which preemption was disabled. The control of preempto tracer is much like the irqso tracer. # echo preempto > current_tracer # echo latency format > trace_options # echo 0 > tracing_max_latency # echo 1 > tracing_on # ls ltr [...] # echo 0 > tracing_on # cat trace # tracer: preempto # preempto latency trace v1.1.5 on 2.6.26 rc8 latency: 29 us, #3/3, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2) | task: sshd 4261 (uid:0 nice:0 policy:0 rt_prio:0)

605

610

615

620

625

=> started at: do_IRQ => ended at: __do_softirq # # _ /_ => CPU# => irqs o

630

10

linux/Documentation/trace/ftrace.txt (v2.6.39)

635

640

# # # # # # cmd # \ sshd sshd sshd

pid / 4261 4261 4261

|/_ || / _ ||| / _ |||| / ||||| ||||| time ||||| 0d.h. 0d.s. 0d.s1

=> need resched => hardirq/softirq => preempt depth delay | caller \ | / 0us+: irq_enter (do_IRQ) 29us : _local_bh_enable (__do_softirq) 30us : trace_preempt_on (__do_softirq)

645

This has some more changes. Preemption was disabled when an interrupt came in (notice the h), and was enabled while doing a softirq. (notice the s). But we also see that interrupts have been disabled when entering the preempt o section and leaving it (the d). We do not know if interrupts were enabled in the mean time. # tracer: preempto # preempto latency trace v1.1.5 on 2.6.26 rc8 latency: 63 us, #87/87, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2)

650

655

| task: sshd 4261 (uid:0 nice:0 policy:0 rt_prio:0) => started at: remove_wait_queue => ended at: __do_softirq
660

665

670

675

680

685

690

# # # # # # # # cmd # \ sshd sshd sshd sshd sshd sshd sshd sshd [...] sshd sshd sshd sshd sshd sshd sshd sshd sshd sshd sshd sshd sshd sshd [...]

pid / 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261

_ /_ |/_ || / _ ||| / _ |||| / ||||| ||||| time ||||| 0d..1 0d..1 0d..1 0d..1 0d..1 0d..1 0d.h1 0d.h.

=> CPU# => irqs o => need resched => hardirq/softirq => preempt depth delay | \ | 0us 1us 2us 2us 2us 3us 3us 4us : : : : : : : : caller / _spin_lock_irqsave (remove_wait_queue) _spin_unlock_irqrestore (remove_wait_queue) do_IRQ (common_interrupt) irq_enter (do_IRQ) idle_cpu (irq_enter) add_preempt_count (irq_enter) idle_cpu (irq_enter) handle_fasteoi_irq (do_IRQ)

0d.h. 0d.h1 0d.h1 0d.h1 0d.h1 0d.h1 0d.h1 0d..2 0d... 0d... 0d... 0d.s4 0d.s4 0d.s5

12us : add_preempt_count (_spin_lock) 12us : ack_ioapic_quirk_irq (handle_fasteoi_irq) 13us : move_native_irq (ack_ioapic_quirk_irq) 13us : _spin_unlock (handle_fasteoi_irq) 14us : sub_preempt_count (_spin_unlock) 14us : irq_exit (do_IRQ) 15us : sub_preempt_count (irq_exit) 15us : do_softirq (irq_exit) 15us : __do_softirq (do_softirq) 16us : __local_bh_disable (__do_softirq) 16us+: add_preempt_count (__local_bh_disable) 20us : add_preempt_count (__local_bh_disable) 21us : sub_preempt_count (local_bh_enable) 21us : sub_preempt_count (local_bh_enable)

linux/Documentation/trace/ftrace.txt (v2.6.39)

11

695

700

sshd sshd sshd sshd sshd sshd sshd sshd [...]

4261 4261 4261 4261 4261 4261 4261 4261

0d.s6 0d.s6 0d.s7 0d.s5 0d.s5 0d.s6 0d.s5 0d.s5

41us 42us 42us 43us 43us 44us 44us 45us

: : : : : : : :

add_preempt_count (__local_bh_disable) sub_preempt_count (local_bh_enable) sub_preempt_count (local_bh_enable) add_preempt_count (__local_bh_disable) sub_preempt_count (local_bh_enable_ip) sub_preempt_count (local_bh_enable_ip) add_preempt_count (__local_bh_disable) sub_preempt_count (local_bh_enable)

sshd 4261 0d.s. sshd 4261 0d.s1

63us : _local_bh_enable (__do_softirq) 64us : trace_preempt_on (__do_softirq)

710

The above is an example of the preempto trace with ftrace_enabled set. Here we see that interrupts were disabled the entire time. The irq_enter code lets us know that we entered an interrupt h. Before that, the functions being traced still show that it is not in an interrupt, but we can see from the functions themselves that this is not the case. Notice that __do_softirq when called does not have a preempt_count. It may seem that we missed a preempt enabling. What really happened is that the preempt count is held on the threads stack and we switched to the softirq stack (4K stacks in eect). The code does not copy the preempt count, but because interrupts are disabled, we do not need to worry about it. Having a tracer like this is good for letting people know what really happens inside the kernel.

715

720

preemptirqso
725

730

Knowing the locations that have interrupts disabled or preemption disabled for the longest times is helpful. But sometimes we would like to know when either preemption and/or interrupts are disabled. Consider the following code: local_irq_disable(); call_function_with_irqs_o(); preempt_disable(); call_function_with_irqs_and_preemption_o(); local_irq_enable(); call_function_with_preemption_o(); preempt_enable(); The irqso tracer will record the total length of call_function_with_irqs_o() and call_function_with_irqs_and_preemption_o().

735

740

745

The preempto tracer will record the total length of call_function_with_irqs_and_preemption_o() and call_function_with_preemption_o().
750

But neither will trace the time that interrupts and/or preemption is disabled. This total time is the time that we can not schedule. To record this time, use the preemptirqso tracer. Again, using this trace is much like the irqso and preempto tracers.

755

12

linux/Documentation/trace/ftrace.txt (v2.6.39)

760

765

# echo preemptirqso > current_tracer # echo latency format > trace_options # echo 0 > tracing_max_latency # echo 1 > tracing_on # ls ltr [...] # echo 0 > tracing_on # cat trace # tracer: preemptirqso # preemptirqso latency trace v1.1.5 on 2.6.26 rc8 latency: 293 us, #3/3, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2) | task: ls 4860 (uid:0 nice:0 policy:0 rt_prio:0) => started at: apic_timer_interrupt => ended at: __do_softirq # # # # # # # # cmd # \ ls ls ls => CPU# _ /_ => irqs o => need resched |/_ || / _ => hardirq/softirq ||| / _ => preempt depth |||| / ||||| delay ||||| time | caller ||||| \ | / 0d... 0us!: trace_hardirqs_o_thunk (apic_timer_interrupt) 0d.s. 294us : _local_bh_enable (__do_softirq) 0d.s1 294us : trace_preempt_on (__do_softirq)

770

775

780

785

pid / 4860 4860 4860

790

795

The trace_hardirqs_o_thunk is called from assembly on x86 when interrupts are disabled in the assembly code. Without the function tracing, we do not know if interrupts were enabled within the preemption points. We do see that it started with preemption enabled. Here is a trace with ftrace_enabled set:

800

# tracer: preemptirqso # preemptirqso latency trace v1.1.5 on 2.6.26 rc8

latency: 105 us, #183/183, CPU#0 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2) | task: sshd 4261 (uid:0 nice:0 policy:0 rt_prio:0) => started at: write_chan => ended at: __do_softirq # # # # # # # # cmd _ /_ |/_ || / _ ||| / _ |||| / ||||| ||||| time => CPU# => irqs o => need resched => hardirq/softirq => preempt depth delay | caller

810

815

pid

linux/Documentation/trace/ftrace.txt (v2.6.39)

13

820

\ ls ls ls ls ls ls sshd sshd sshd sshd sshd sshd sshd sshd sshd sshd sshd sshd sshd sshd

/ 4473 4473 4473 4473

||||| 0.N.. 0dN.1 0dN.1 0d..2

\ | 0us 1us 2us 2us 13us 13us 14us 14us 15us 16us 16us 17us 17us 18us 18us 18us 19us 19us 20us 20us

: : : : : : : : : : : : : : : : : : : :

/ preempt_schedule (write_chan) _spin_lock (schedule) add_preempt_count (_spin_lock) put_prev_task_fair (schedule) set_normalized_timespec (ktime_get_ts) __switch_to (schedule) nish_task_switch (schedule) _spin_unlock_irq (nish_task_switch) add_preempt_count (_spin_lock_irqsave) _spin_unlock_irqrestore (hrtick_set) do_IRQ (common_interrupt) irq_enter (do_IRQ) idle_cpu (irq_enter) add_preempt_count (irq_enter) idle_cpu (irq_enter) handle_fasteoi_irq (do_IRQ) _spin_lock (handle_fasteoi_irq) add_preempt_count (_spin_lock) _spin_unlock (handle_fasteoi_irq) sub_preempt_count (_spin_unlock)

825

[...] 4473 0d..2 4473 0d..2 4261 0d..2 4261 0d..2 4261 0d..1 4261 0d..2 4261 0d..2 4261 0d..2 4261 0d..2 4261 0d..2 4261 0d.h2 4261 0d.h. 4261 0d.h. 4261 0d.h. 4261 0d.h1 4261 0d.h1 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 4261 0d.h1 0d.h1 0d.h2 0d.h2 0d..3 0d... 0d... 0d... 0d.s4 0d.s3 0d.s4 0d.s3 0d.s3 0d.s3 0d.s3 0d.H3 0d.H3 0d.H3 0d.H3 0d.H3 0d.H3 0d.H3 0d.H3 0d.H3 0d.H3 0d.H3 0d.H3 0d.s4 0d.s3

830

835

840

[...] sshd sshd sshd sshd sshd sshd sshd sshd sshd [...] sshd sshd sshd sshd sshd sshd sshd sshd sshd [...] sshd sshd sshd sshd sshd sshd sshd sshd sshd sshd sshd [...]
875

845

850

28us : _spin_unlock (handle_fasteoi_irq) 29us : sub_preempt_count (_spin_unlock) 29us : irq_exit (do_IRQ) 29us : sub_preempt_count (irq_exit) 30us : do_softirq (irq_exit) 30us : __do_softirq (do_softirq) 31us : __local_bh_disable (__do_softirq) 31us+: add_preempt_count (__local_bh_disable) 34us : add_preempt_count (__local_bh_disable) 43us 44us 44us 45us 45us 46us 46us 47us 47us 81us 82us 82us 83us 83us 84us 84us 85us 85us 86us 86us : : : : : : : : : : : : : : : : : : : : sub_preempt_count (local_bh_enable_ip) sub_preempt_count (local_bh_enable_ip) smp_apic_timer_interrupt (apic_timer_interrupt) irq_enter (smp_apic_timer_interrupt) idle_cpu (irq_enter) add_preempt_count (irq_enter) idle_cpu (irq_enter) hrtimer_interrupt (smp_apic_timer_interrupt) ktime_get (hrtimer_interrupt) tick_program_event (hrtimer_interrupt) ktime_get (tick_program_event) ktime_get_ts (ktime_get) getnstimeofday (ktime_get_ts) set_normalized_timespec (ktime_get_ts) clockevents_program_event (tick_program_event) lapic_next_event (clockevents_program_event) irq_exit (smp_apic_timer_interrupt) sub_preempt_count (irq_exit) sub_preempt_count (irq_exit) add_preempt_count (__local_bh_disable)

855

860

865

870

880

sshd sshd sshd sshd sshd sshd sshd

0d.s1 98us : sub_preempt_count (net_rx_action) 0d.s. 99us : add_preempt_count (_spin_lock_irq) 0d.s1 99us+: _spin_unlock_irq (run_timer_softirq) 0d.s. 104us : _local_bh_enable (__do_softirq) 0d.s. 104us : sub_preempt_count (_local_bh_enable) 0d.s. 105us : _local_bh_enable (__do_softirq) 0d.s1 105us : trace_preempt_on (__do_softirq)

14

linux/Documentation/trace/ftrace.txt (v2.6.39)

885

890

This is a very interesting trace. It started with the preemption of the ls task. We see that the task had the "need_resched" bit set via the N in the trace. Interrupts were disabled before the spin_lock at the beginning of the trace. We see that a schedule took place to run sshd. When the interrupts were enabled, we took an interrupt. On return from the interrupt handler, the softirq ran. We took another interrupt while running the softirq as we see from the capital H.

wakeup
895

900

In a Real Time environment it is very important to know the wakeup time it takes for the highest priority task that is woken up to the time that it executes. This is also known as "schedule latency". I stress the point that this is about RT tasks. It is also important to know the scheduling latency of non RT tasks, but the average schedule latency is better for non RT tasks. Tools like LatencyTop are more appropriate for such measurements. Real Time environments are interested in the worst case latency. That is the longest latency it takes for something to happen, and not the average. We can have a very fast scheduler that may only have a large latency once in a while, but that would not work well with Real Time tasks. The wakeup tracer was designed to record the worst case wakeups of RT tasks. Non RT tasks are not recorded because the tracer only records one worst case and tracing non RT tasks that are unpredictable will overwrite the worst case latency of RT tasks. Since this tracer only deals with RT tasks, we will run this slightly dierently than we did with the previous tracers. Instead of performing an ls, we will run sleep 1 under chrt which changes the priority of the task.

905

910

915

920

925

930

# echo wakeup > current_tracer # echo latency format > trace_options # echo 0 > tracing_max_latency # echo 1 > tracing_on # chrt f 5 sleep 1 # echo 0 > tracing_on # cat trace # tracer: wakeup # wakeup latency trace v1.1.5 on 2.6.26 rc8 latency: 4 us, #2/2, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2) | task: sleep 4901 (uid:0 nice:0 policy:1 rt_prio:5)

935

940

945

# # # # # # # # cmd # \ /

pid

_ /_ |/_ || / _ ||| / _ |||| / ||||| ||||| time |||||

=> CPU# => irqs o => need resched => hardirq/softirq => preempt depth delay | \ | caller /

linux/Documentation/trace/ftrace.txt (v2.6.39)

15

<idle> 0 <idle> 0

1d.h4 1d..4

0us+: try_to_wake_up (wake_up_process) 4us : schedule (cpu_idle)

950

Running this on an idle system, we see that it only took 4 microseconds to perform the task switch. Note, since the trace marker in the schedule is before the actual "switch", we stop the tracing when the recorded task is about to schedule in. This may change if we add a new marker at the end of the scheduler. Notice that the recorded task is sleep with the PID of 4901 and it has an rt_prio of 5. This priority is user space priority and not the internal kernel priority. The policy is 1 for SCHED_FIFO and 2 for SCHED_RR.

955

960

Doing the same with chrt r 5 and ftrace_enabled set. # tracer: wakeup # wakeup latency trace v1.1.5 on 2.6.26 rc8 latency: 50 us, #60/60, CPU#1 | (M:preempt VP:0, KP:0, SP:0 HP:0 #P:2) | task: sleep 4068 (uid:0 nice:0 policy:2 rt_prio:5)
970

965

975

980

985

990

995

1000

1005

# # # # # # # # cmd # ksoftirq ksoftirq ksoftirq ksoftirq ksoftirq ksoftirq ksoftirq ksoftirq [...] ksoftirq ksoftirq ksoftirq ksoftirq [...] ksoftirq ksoftirq ksoftirq ksoftirq ksoftirq ksoftirq ksoftirq ksoftirq ksoftirq ksoftirq ksoftirq ksoftirq ksoftirq ksoftirq

pid \ / 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7 7

_ /_ |/_ || / _ ||| / _ |||| / ||||| ||||| time ||||| 1d.H3 1d.H4 1d.H3 1d.H3 1d.H3 1d.H3 1d.H3 1d.H3 1d.H2 1d.H2 1d.s3 1..s2 1..s2 1d.s2 1d.s2 1.N.3 1.N.2 1.N.2 1.N.2 1.N.2 1.N.2 1.N.3 1dN.3 1dN.3 1d..4 1d..4

=> CPU# => irqs o => need resched => hardirq/softirq => preempt depth delay | \ | 0us 1us 2us 3us 4us 5us 6us 7us 17us 18us 19us 20us 26us 27us 28us 29us 30us 31us 32us 33us 33us 34us 35us 36us 37us 38us : : : : : : : : : : : : : : : : : : : : : : : : : : caller / try_to_wake_up (wake_up_process) sub_preempt_count (marker_probe_cb) check_preempt_wakeup (try_to_wake_up) update_curr (check_preempt_wakeup) calc_delta_mine (update_curr) __resched_task (check_preempt_wakeup) task_wake_up_rt (try_to_wake_up) _spin_unlock_irqrestore (try_to_wake_up) irq_exit (smp_apic_timer_interrupt) sub_preempt_count (irq_exit) sub_preempt_count (irq_exit) rcu_process_callbacks (__do_softirq) __rcu_process_callbacks (rcu_process_callbacks) _local_bh_enable (__do_softirq) sub_preempt_count (_local_bh_enable) sub_preempt_count (ksoftirqd) _cond_resched (ksoftirqd) __cond_resched (_cond_resched) add_preempt_count (__cond_resched) schedule (__cond_resched) add_preempt_count (schedule) hrtick_clear (schedule) _spin_lock (schedule) add_preempt_count (_spin_lock) put_prev_task_fair (schedule) update_curr (put_prev_task_fair)

16

linux/Documentation/trace/ftrace.txt (v2.6.39)

1010

[...] ksoftirq ksoftirq ksoftirq ksoftirq ksoftirq

7 7 7 7 7

1d..5 1d..5 1d..6 1d..6 1d..4

47us 48us 49us 49us 50us

: : : : :

_spin_trylock (tracing_record_cmdline) add_preempt_count (_spin_trylock) _spin_unlock (tracing_record_cmdline) sub_preempt_count (_spin_unlock) schedule (__cond_resched)

1015

1020

1025

The interrupt went o while running ksoftirqd. This task runs at SCHED_OTHER. Why did not we see the N set early? This may be a harmless bug with x86_32 and 4K stacks. On x86_32 with 4K stacks congured, the interrupt and softirq run with their own stack. Some information is held on the top of the tasks stack (need_resched and preempt_count are both stored there). The setting of the NEED_RESCHED bit is done directly to the tasks stack, but the reading of the NEED_RESCHED is done by looking at the current stack, which in this case is the stack for the hard interrupt. This hides the fact that NEED_RESCHED has been set. We do not see the N until we switch back to the tasks assigned stack. function

1030

This tracer is the function tracer. Enabling the function tracer can be done from the debug le system. Make sure the ftrace_enabled is set; otherwise this tracer is a nop.
1035

1040

1045

1050

1055

1060

# sysctl kernel.ftrace_enabled=1 # echo function > current_tracer # echo 1 > tracing_on # usleep 1 # echo 0 > tracing_on # cat trace # tracer: function # CPU# # TASK PID # || | bash 4003 [00] bash 4003 [00] bash 4003 [00] bash 4003 [00] bash 4003 [00] bash 4003 [00] bash 4003 [00] bash 4003 [00] bash 4003 [00] bash 4003 [00] bash 4003 [00] bash 4003 [00] bash 4003 [00] bash 4003 [00] bash 4003 [00] [...]

TIMESTAMP FUNCTION | | 123.638713: nish_task_switch < schedule 123.638714: _spin_unlock_irq < nish_task_switch 123.638714: sub_preempt_count < _spin_unlock_irq 123.638715: hrtick_set < schedule 123.638715: _spin_lock_irqsave < hrtick_set 123.638716: add_preempt_count < _spin_lock_irqsave 123.638716: _spin_unlock_irqrestore < hrtick_set 123.638717: sub_preempt_count < _spin_unlock_irqrestore 123.638717: hrtick_clear < hrtick_set 123.638718: sub_preempt_count < schedule 123.638718: sub_preempt_count < preempt_schedule 123.638719: wait_for_completion < __stop_machine_run 123.638719: wait_for_common < wait_for_completion 123.638720: _spin_lock_irq < wait_for_common 123.638720: add_preempt_count < _spin_lock_irq

1065

1070

Note: function tracer uses ring buers to store the above entries. The newest data may overwrite the oldest data. Sometimes using echo to stop the trace is not sucient because the tracing could have overwritten the data that you wanted to record. For this reason, it is sometimes better to disable tracing directly from a program. This allows you to stop the tracing at the point that you hit the part that you are interested in. To disable the tracing directly from a C program,

linux/Documentation/trace/ftrace.txt (v2.6.39)

17

something like following code snippet can be used: int trace_fd; [...] int main(int argc, char argv[]) { [...] trace_fd = open(tracing_le("tracing_on"), O_WRONLY); [...] if (condition_hit()) { write(trace_fd, "0", 1); } [...] }

1075

1080

1085

Single thread tracing

1090

By writing into set_ftrace_pid you can trace a single thread. For example: # cat set_ftrace_pid no pid # echo 3111 > set_ftrace_pid # cat set_ftrace_pid 3111 # echo function > current_tracer # cat trace | head # tracer: function # CPU# # TASK PID # || | yum updatesd 3111 [003] 1637.254676: yum updatesd 3111 [003] 1637.254681: yum updatesd 3111 [003] 1637.254682: yum updatesd 3111 [003] 1637.254683: yum updatesd 3111 [003] 1637.254685: yum updatesd 3111 [003] 1637.254686: # echo 1 > set_ftrace_pid # cat trace |head # tracer: function # # TASK PID CPU# # || | ##### CPU 3 buer started #### yum updatesd 3111 [003] 1701.957688: yum updatesd 3111 [003] 1701.957689: yum updatesd 3111 [003] 1701.957691: yum updatesd 3111 [003] 1701.957692: yum updatesd 3111 [003] 1701.957693:

1095

1100

1105

TIMESTAMP FUNCTION | | nish_task_switch < thread_return hrtimer_cancel < schedule_hrtimeout_range hrtimer_try_to_cancel < hrtimer_cancel lock_hrtimer_base < hrtimer_try_to_cancel fget_light < do_sys_poll pipe_poll < do_sys_poll

1110

1115

TIMESTAMP FUNCTION | | free_poll_entry < poll_freewait remove_wait_queue < free_poll_entry fput < free_poll_entry audit_syscall_exit < sysret_audit path_put < audit_syscall_exit

1120

If you want to trace a function when executing, you could use something like this simple program:
1125

1130

#include #include #include #include #include #include #include

<stdio.h> <stdlib.h> <sys/types.h> <sys/stat.h> <fcntl.h> <unistd.h> <string.h>

#dene _STR(x) #x

18

linux/Documentation/trace/ftrace.txt (v2.6.39)

1135

#dene STR(x) _STR(x) #dene MAX_PATH 256 const char nd_debugfs(void) { static char debugfs[MAX_PATH+1]; static int debugfs_found; char type[100]; FILE fp; if (debugfs_found) return debugfs; if ((fp = fopen("/proc/mounts","r")) == NULL) { perror("/proc/mounts"); return NULL; } while (fscanf(fp, "% s %" STR(MAX_PATH) "s %99s % s % d % d\n", debugfs, type) == 2) { if (strcmp(type, "debugfs") == 0) break; } fclose(fp); if (strcmp(type, "debugfs") != 0) { fprintf(stderr, "debugfs not mounted"); return NULL; } strcat(debugfs, "/tracing/"); debugfs_found = 1;

1140

1145

1150

1155

1160

1165

1170

return debugfs; } const char tracing_le(const char le_name) { static char trace_le[MAX_PATH+1]; snprintf(trace_le, MAX_PATH, "%s/%s", nd_debugfs(), le_name); return trace_le; } int main (int argc, char argv) { if (argc < 1) exit( 1); if (fork() > 0) { int fd, d; char line[64]; int s; d = open(tracing_le("current_tracer"), O_WRONLY); if (d < 0) exit( 1); write(d, "nop", 3); fd = open(tracing_le("set_ftrace_pid"), O_WRONLY); s = sprintf(line, "%d\n", getpid()); write(fd, line, s);

1175

1180

1185

1190

1195

linux/Documentation/trace/ftrace.txt (v2.6.39)

19

write(d, "function", 8);

close(fd); close(d); execvp(argv[1], argv+1);

} return 0; }

1210

hw branch tracer (x86 only)

1215

This tracer uses the x86 last branch tracing hardware feature to collect a branch trace on all cpus with relatively low overhead. The tracer uses a xed size circular buer per cpu and only traces ring 0 branches. The trace le dumps that buer in the following format:

1220

1225

# tracer: hw branch tracer # # CPU# TO < FROM 0 scheduler_tick+0xb5/0x1bf 2 run_posix_cpu_timers+0x2b/0x72a 0 scheduler_tick+0x139/0x1bf 0 scheduler_tick+0x17c/0x1bf 2 run_posix_cpu_timers+0x9e/0x72a 0 scheduler_tick+0x1b6/0x1bf

< task_tick_idle+0x5/0x6 < run_posix_cpu_timers+0x25/0x72a < scheduler_tick+0xed/0x1bf < scheduler_tick+0x148/0x1bf < run_posix_cpu_timers+0x5e/0x72a < scheduler_tick+0x1aa/0x1bf

1230

1235

The tracer may be used to dump the trace for the oopsing cpu on a kernel oops into the system log. To enable this, ftrace_dump_on_oops must be set. To set ftrace_dump_on_oops, one can either use the sysctl function or set it via the proc system interface. sysctl kernel.ftrace_dump_on_oops=n

1240

or echo n > /proc/sys/kernel/ftrace_dump_on_oops If n = 1, ftrace will dump buers of all CPUs, if n = 2 ftrace will only dump the buer of the CPU that triggered the oops. Heres an example of such a dump after a null pointer dereference in a kernel module:

1245

1250

1255

1260

[57848.105921] [57848.106019] [57848.106019] [57848.106019] [57848.106019] [57848.106019] [57848.106019] [...] [57848.106019] [57848.106019] [57848.106019]

BUG: unable to handle kernel NULL pointer dereference at 0000000000000000 IP: [<a0000006>] open+0x6/0x14 [oops] PGD 2354e9067 PUD 2375e7067 PMD 0 Oops: 0002 [#1] SMP last sysfs le: /sys/devices/pci0000:00/0000:00:1e.0/0000:20:05.0/local_cpus Dumping ftrace buer:

0 chrdev_open+0xe6/0x165 0 chrdev_open+0x117/0x165 0 chrdev_open+0x120/0x165

< cdev_put+0x23/0x24 < chrdev_open+0xfa/0x165 < chrdev_open+0x11c/0x165

20

linux/Documentation/trace/ftrace.txt (v2.6.39)

1265

1270

1275

1280

[57848.106019] [57848.106019] [57848.106019] [57848.106019] [57848.106019] [57848.106019] [57848.106019] [57848.106019] [...] [57848.106019] [57848.106019] [57848.106019] [57848.106019] [...] [57848.106019] [57848.106019] [57848.106019] [57848.106019] [57848.106019] [57848.106019] [57848.106019] [...]

0 0 0 0 0 0 0 0 0 0 0 0

chrdev_open+0x134/0x165 open+0x0/0x14 [oops] page_fault+0x0/0x30 error_entry+0x0/0x5b error_kernelspace+0x0/0x31 error_sti+0x0/0x1 page_fault+0x9/0x30 do_page_fault+0x0/0x881

< chrdev_open+0x12b/0x165 < chrdev_open+0x144/0x165 < open+0x6/0x14 [oops] < page_fault+0x4/0x30 < error_entry+0x59/0x5b < error_kernelspace+0x2d/0x31 < error_sti+0x0/0x1 < page_fault+0x1a/0x30 < is_prefetch+0x1ee/0x1f2 < do_page_fault+0x67a/0x881 < do_page_fault+0x6e0/0x881 < oops_begin+0x9/0x96 < ds_suspend_bts+0x1a/0xe3

do_page_fault+0x66b/0x881 do_page_fault+0x6e0/0x881 oops_begin+0x0/0x96 trace_hw_branch_oops+0x0/0x2d

0 ds_suspend_bts+0x2a/0xe3

CPU 0 Modules linked in: oops Pid: 5542, comm: cat Tainted: G W 2.6.28 #23 RIP: 0010:[<a0000006>] [<a0000006>] open+0x6/0x14 [oops] RSP: 0018:880235457d48 EFLAGS: 00010246

1285

function graph tracer

1290

This tracer is similar to the function tracer except that it probes a function on its entry and its exit. This is done by using a dynamically allocated stack of return addresses in each task_struct. On function entry the tracer overwrites the return address of each function traced to set a custom probe. Thus the original return address is stored on the stack of return address in the task_struct. Probing on both ends of a function leads to special features such as: measure of a functions time execution having a reliable call stack to draw function calls graph This tracer is useful in several situations: you want to nd the reason of a strange kernel behavior and need to see what happens in detail on any areas (or specic ones). you are experiencing weird latencies but its dicult to nd its origin.

1295

1300

1305

1310

you want to nd quickly which path is taken by a specic function you just want to peek inside a working kernel and want to see what happens there. # tracer: function_graph # # CPU DURATION #| | | 0) 0)

1315

1320

FUNCTION CALLS | | | | | sys_open() { | do_sys_open() {

linux/Documentation/trace/ftrace.txt (v2.6.39)

21

1325

1330

1335

0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0)

1.382 us 2.478 us

1.389 2.553 3.807 7.876

us us us us

0.668 us 0.570 us 0.586 us

| | | | | | | | | | | | | |

getname() { kmem_cache_alloc() { __might_sleep(); } strncpy_from_user() { might_fault() { __might_sleep(); } } } alloc_fd() { _spin_lock(); expand_les(); _spin_unlock();

1340

There are several columns that can be dynamically enabled/disabled. You can use every combination of options you want, depending on your needs. The cpu number on which the function executed is default enabled. It is sometimes better to only trace one cpu (see tracing_cpu_mask le) or you might sometimes see unordered function calls while cpu tracing switch. hide: echo nofuncgraph cpu > trace_options show: echo funcgraph cpu > trace_options The duration (functions time of execution) is displayed on the closing bracket line of a function or on the same line than the current function in case of a leaf one. It is default enabled. hide: echo nofuncgraph duration > trace_options show: echo funcgraph duration > trace_options

1345

1350

1355

1360

The overhead eld precedes the duration eld in case of reached duration thresholds. hide: echo nofuncgraph overhead > trace_options show: echo funcgraph overhead > trace_options depends on: funcgraph duration ie: 0) 0) 0) 0) 0) 0) + [...] 0) 0) 0) 0) 0) 0) ! 0) ! | | | | | | |} putname() { kmem_cache_free() { __phys_addr(); } } } | | | | | |} up_write() { _spin_lock_irqsave(); _spin_unlock_irqrestore(); } fput();

1365

1370

0.646 us 0.684 us 3.123 us 0.548 us 58.628 us

1375

1380

0.518 us 1.757 us 2.861 us 115.305 us 116.402 us

1385

+ means that the function exceeded 10 usecs.

22

linux/Documentation/trace/ftrace.txt (v2.6.39)

! means that the function exceeded 100 usecs.

1390

The task/pid eld displays the thread cmdline and pid which executed the function. It is default disabled. hide: echo nofuncgraph proc > trace_options show: echo funcgraph proc > trace_options

1395

ie: # tracer: function_graph # # CPU TASK/PID #| | | 0) sh 4802 0) sh 4802 0) sh 4802 0) sh 4802 0) sh 4802 period(); 0) sh 4802 0) sh 4802 0) sh 4802 0) sh 4802

1400

DURATION | | | | | | | | | | |+ | | | | | | | | |

1405

y
1410

0.616 us 0.586 us 2.899 us 4.040 us 5.151 us 49.370 us

FUNCTION CALLS | | | | d_free() { call_rcu() { __call_rcu() { rcu_process_gp_end(); check_for_new_grace_ } } } }

1415

The absolute time eld is an absolute timestamp given by the system clock since it started. A snapshot of this time is given on each entry/exit of functions hide: echo nofuncgraph abstime > trace_options show: echo funcgraph abstime > trace_options

1420

ie: # # TIME # | 360.774522 | 360.774522 | 360.774523 | _up_bit(); 360.774524 | 360.774524 | 360.774525 | 360.774525 | dirty(); 360.774527 | 360.774528 | for_journal() { 360.774528 | () { 360.774529 | bit() { 360.774529 | waitqueue() { 360.774530 | phys_addr();

1425

y
1430

1) 1) 1) 1) 1) 1) 1) 1) 1) 1) 1) 1) 1)

CPU DURATION | | | 0.541 us 4.663 us 0.541 us 6.796 7.952 9.063 0.615 us us us us

FUNCTION CALLS | | | | | | | | | | | | | | | | } } } journal_mark_ __brelse(); reiserfs_prepare_ } } __wake

w w

0.578 us

1435

y y y y y

0.594 us

w unlock_buer w wake_up_ w bit_ w __ w

1440

You can put some comments on specic functions by using trace_printk() For example, if you want to put a comment inside

linux/Documentation/trace/ftrace.txt (v2.6.39)

23

the __might_sleep() function, you just have to include <linux/ftrace.h> and call trace_printk() inside __might_sleep()
1445

trace_printk("Im a comment!\n") will produce: 1) 1) 1) | | | __might_sleep() { / Im a comment! / }

1450

1.449 us

1455

You might nd other useful features for this tracer in the following "dynamic ftrace" section such as tracing only specic functions or tasks. dynamic ftrace

1460

1465

If CONFIG_DYNAMIC_FTRACE is set, the system will run with virtually no overhead when function tracing is disabled. The way this works is the mcount function call (placed at the start of every kernel function, produced by the pg switch in gcc), starts of pointing to a simple return. (Enabling FTRACE will include the pg switch in the compiling of the kernel.) At compile time every C le object is run through the recordmcount.pl script (located in the scripts directory). This script will process the C object using objdump to nd all the locations in the .text section that call mcount. (Note, only the .text section is processed, since processing other sections like .init.text may cause races due to those sections being freed). A new section called "__mcount_loc" is created that holds references to all the mcount call sites in the .text section. This section is compiled back into the original object. The nal linker will add all these references into a single table. On boot up, before SMP is initialized, the dynamic ftrace code scans this table and updates all the locations into nops. It also records the locations, which are added to the available_lter_functions list. Modules are processed as they are loaded and before they are executed. When a module is unloaded, it also removes its functions from the ftrace function list. This is automatic in the module unload code, and the module author does not need to worry about it. When tracing is enabled, kstop_machine is called to prevent races with the CPUS executing code being modied (which can cause the CPU to do undesirable things), and the nops are patched back to calls. But this time, they do not call mcount (which is just a function stub). They now call into the ftrace infrastructure. One special side eect to the recording of the functions being traced is that we can now selectively choose which functions we wish to trace and which ones we want the mcount calls to remain as nops.

1470

1475

1480

1485

1490

1495

1500

Two les are used, one for enabling and one for disabling the tracing of specied functions. They are: set_ftrace_lter

24

linux/Documentation/trace/ftrace.txt (v2.6.39)

1505

and set_ftrace_notrace
1510

A list of available functions that you can add to these les is listed in: available_lter_functions

1515

1520

# cat available_lter_functions put_prev_task_idle kmem_cache_create pick_next_task_rt get_online_cpus pick_next_task_fair mutex_lock [...] If I am only interested in sys_nanosleep and hrtimer_interrupt:

1525

1530

1535

# echo sys_nanosleep hrtimer_interrupt \ > set_ftrace_lter # echo function > current_tracer # echo 1 > tracing_on # usleep 1 # echo 0 > tracing_on # cat trace # tracer: ftrace # CPU# TIMESTAMP FUNCTION # TASK PID # || | | | usleep 4134 [00] 1317.070017: hrtimer_interrupt < smp_apic_timer_interrupt usleep 4134 [00] 1317.070111: sys_nanosleep < syscall_call <idle> 0 [00] 1317.070115: hrtimer_interrupt < smp_apic_timer_interrupt To see which functions are being traced, you can cat the le: # cat set_ftrace_lter hrtimer_interrupt sys_nanosleep

1540

1545

Perhaps this is not enough. The lters also allow simple wild cards. Only the following are currently available
1550

<match> will match functions that begin with <match> <match> will match functions that end with <match> <match> will match functions that have <match> in it
1555

These are the only wild cards which are supported. <match> <match> will not work. Note: It is better to use quotes to enclose the wild cards, otherwise the shell may expand the parameters into names of les in the local directory. # echo hrtimer_ > set_ftrace_lter

1560

1565

Produces: # tracer: ftrace

linux/Documentation/trace/ftrace.txt (v2.6.39)

25

1570

# # #

1575

TASK PID || bash 4003 [00] bash 4003 [00] bash 4003 [00] bash 4003 [00] <idle> 0 <idle> 0 <idle> 0 <idle> 0 <idle> 0

CPU# TIMESTAMP FUNCTION | | | 1480.611794: hrtimer_init < copy_process 1480.611941: hrtimer_start < hrtick_set 1480.611956: hrtimer_cancel < hrtick_clear 1480.611956: hrtimer_try_to_cancel < hrtimer_cancel [00] 1480.612019: hrtimer_get_next_event < get_next_timer_interrupt [00] 1480.612025: hrtimer_get_next_event < get_next_timer_interrupt [00] 1480.612032: hrtimer_get_next_event < get_next_timer_interrupt [00] 1480.612037: hrtimer_get_next_event < get_next_timer_interrupt [00] 1480.612382: hrtimer_get_next_event < get_next_timer_interrupt

1580

Notice that we lost the sys_nanosleep. # cat set_ftrace_lter hrtimer_run_queues hrtimer_run_pending hrtimer_init hrtimer_cancel hrtimer_try_to_cancel hrtimer_forward hrtimer_start hrtimer_reprogram hrtimer_force_reprogram hrtimer_get_next_event hrtimer_interrupt hrtimer_nanosleep hrtimer_wakeup hrtimer_get_remaining hrtimer_get_res hrtimer_init_sleeper

1585

1590

1595

1600

1605

This is because the > and >> act just like they do in bash. To rewrite the lters, use > To append to the lters, use >> To clear out a lter so that all functions will be recorded again:

1610

# echo > set_ftrace_lter # cat set_ftrace_lter # Again, now we want to append.

1615

1620

1625

1630

# echo sys_nanosleep > set_ftrace_lter # cat set_ftrace_lter sys_nanosleep # echo hrtimer_ >> set_ftrace_lter # cat set_ftrace_lter hrtimer_run_queues hrtimer_run_pending hrtimer_init hrtimer_cancel hrtimer_try_to_cancel hrtimer_forward hrtimer_start hrtimer_reprogram hrtimer_force_reprogram hrtimer_get_next_event

26

linux/Documentation/trace/ftrace.txt (v2.6.39)

1635

hrtimer_interrupt sys_nanosleep hrtimer_nanosleep hrtimer_wakeup hrtimer_get_remaining hrtimer_get_res hrtimer_init_sleeper

1640

The set_ftrace_notrace prevents those functions from being traced. # echo preempt lock > set_ftrace_notrace

1645

Produces: # tracer: ftrace # # #

1650

1655

CPU# TASK PID || | bash 4043 [01] bash 4043 [01] bash 4043 [01] bash 4043 [01] bash 4043 [01] bash 4043 [01] bash 4043 [01] bash 4043 [01] bash 4043 [01]

TIMESTAMP FUNCTION | | 115.281644: nish_task_switch < schedule 115.281645: hrtick_set < schedule 115.281645: hrtick_clear < hrtick_set 115.281646: wait_for_completion < __stop_machine_run 115.281647: wait_for_common < wait_for_completion 115.281647: kthread_stop < stop_machine_run 115.281648: init_waitqueue_head < kthread_stop 115.281648: wake_up_process < kthread_stop 115.281649: try_to_wake_up < wake_up_process

1660

We can see that theres no more lock or preempt tracing.

Dynamic ftrace with the function graph tracer

Although what has been explained above concerns both the function tracer and the function graph tracer, there are some special features only available in the function graph tracer.
1670

If you want to trace only one function and all of its children, you just have to echo its name into set_graph_function: echo __do_fault > set_graph_function
1675

will produce the following "expanded" trace of the __do_fault() function: 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) | __do_fault() { | lemap_fault() { | nd_lock_page() { | nd_get_page(); | __might_sleep() { | } | } | } | _spin_lock(); | page_add_le_rmap(); | native_set_pte_at(); | _spin_unlock(); | unlock_page() { | page_waitqueue(); | __wake_up_bit();

1680

0.804 us 1.329 3.904 4.979 0.653 0.578 0.525 0.585 us us us us us us us

1685

1690

0.541 us 0.639 us

linux/Documentation/trace/ftrace.txt (v2.6.39)

27

1695

1700

1705

1710

0) 0) + 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) 0) +

2.786 us 14.237 us

0.698 us 1.412 3.950 5.098 0.631 0.571 0.526 0.586 us us us us us us us

0.533 us 0.638 us 2.793 us 14.012 us

| } |} | __do_fault() { | lemap_fault() { | nd_lock_page() { | nd_get_page(); | __might_sleep() { | } | } | } | _spin_lock(); | page_add_le_rmap(); | native_set_pte_at(); | _spin_unlock(); | unlock_page() { | page_waitqueue(); | __wake_up_bit(); | } |}

You can also expand several functions at once:

echo sys_open > set_graph_function echo sys_close >> set_graph_function Now if you want to go back to trace all functions you can clear this special lter via: echo > set_graph_function

1720

1725

Filter commands

A few commands are supported by the set_ftrace_lter interface. Trace commands have the following format:
1730

<function>:<command>:<parameter> The following commands are supported:

mod This command enables function ltering per module. The parameter denes the module. For example, if only the write functions in the ext3 module are desired, run: echo write :mod:ext3 > set_ftrace_lter This command interacts with the lter in the same way as ltering based on function names. Thus, adding more functions in a dierent module is accomplished by appending (>>) to the lter le. Remove specic module functions by prepending !: echo !writeback :mod:ext3 >> set_ftrace_lter

1740

1745

1750

1755

traceon/traceo These commands turn tracing on and o when the specied functions are hit. The parameter determines how many times the tracing system is turned on and o. If unspecied, there is no limit. For example, to disable tracing when a schedule bug is hit the rst 5 times, run:

28

linux/Documentation/trace/ftrace.txt (v2.6.39)

echo __schedule_bug:traceo:5 > set_ftrace_lter These commands are cumulative whether or not they are appended to set_ftrace_lter. To remove a command, prepend it by ! and drop the parameter: echo !__schedule_bug:traceo > set_ftrace_lter
1765

1760

trace_pipe

1770

The trace_pipe outputs the same content as the trace le, but the eect on the tracing is dierent. Every read from trace_pipe is consumed. This means that subsequent reads will be dierent. The trace is live. # echo function > current_tracer # cat trace_pipe > /tmp/trace.out & [1] 4153 # echo 1 > tracing_on # usleep 1 # echo 0 > tracing_on # cat trace # tracer: function # CPU# # TASK PID # || | # # cat /tmp/trace.out bash bash bash bash bash bash bash bash bash bash

1775

1780

TIMESTAMP FUNCTION | |

1785

1790

1795

4043 4043 4043 4043 4043 4043 4043 4043 4043 4043

[00] [00] [00] [00] [00] [00] [00] [00] [00] [00]

41.267106: 41.267106: 41.267107: 41.267108: 41.267108: 41.267109: 41.267109: 41.267110: 41.267110: 41.267111:

nish_task_switch < schedule hrtick_set < schedule hrtick_clear < hrtick_set wait_for_completion < __stop_machine_run wait_for_common < wait_for_completion kthread_stop < stop_machine_run init_waitqueue_head < kthread_stop wake_up_process < kthread_stop try_to_wake_up < wake_up_process select_task_rq_rt < try_to_wake_up

1800

Note, reading the trace_pipe le will block until more input is added. By changing the tracer, trace_pipe will issue an EOF. We needed to set the function tracer _before_ we "cat" the trace_pipe le.

1805

trace entries

1810

Having too much or not enough data can be troublesome in diagnosing an issue in the kernel. The le buer_size_kb is used to modify the size of the internal trace buers. The number listed is the number of entries that can be recorded per CPU. To know the full size, multiply the number of possible CPUS with the number of entries. # cat buer_size_kb 1408 (units kilobytes) Note, to modify this, you must have tracing completely disabled.

1815

linux/Documentation/trace/ftrace.txt (v2.6.39)

29

1820

To do that, echo "nop" into the current_tracer. If the current_tracer is not set to "nop", an EINVAL error will be returned. # echo nop > current_tracer # echo 10000 > buer_size_kb # cat buer_size_kb 10000 (units kilobytes) The number of pages which will be allocated is limited to a percentage of available memory. Allocating too much will produce an error. # echo 1000000000000 > buer_size_kb bash: echo: write error: Cannot allocate memory # cat buer_size_kb 85

1825

1830

1835

1840 1841

More details can be found in the source code, in the kernel/trace/ .c les.

30

linux/Documentation/trace/ftrace.txt (v2.6.39)