2 # Architectures that offer an FUNCTION_TRACER implementation should
3 # select HAVE_FUNCTION_TRACER:
6 config USER_STACKTRACE_SUPPORT
12 config HAVE_FTRACE_NMI_ENTER
15 See Documentation/trace/ftrace-design.rst
17 config HAVE_FUNCTION_TRACER
20 See Documentation/trace/ftrace-design.rst
22 config HAVE_FUNCTION_GRAPH_TRACER
25 See Documentation/trace/ftrace-design.rst
27 config HAVE_DYNAMIC_FTRACE
30 See Documentation/trace/ftrace-design.rst
32 config HAVE_DYNAMIC_FTRACE_WITH_REGS
35 config HAVE_FTRACE_MCOUNT_RECORD
38 See Documentation/trace/ftrace-design.rst
40 config HAVE_SYSCALL_TRACEPOINTS
43 See Documentation/trace/ftrace-design.rst
48 Arch supports the gcc options -pg with -mfentry
50 config HAVE_NOP_MCOUNT
53 Arch supports the gcc options -pg with -mrecord-mcount and -nop-mcount
55 config HAVE_C_RECORDMCOUNT
58 C version of recordmcount available?
60 config TRACER_MAX_TRACE
71 config FTRACE_NMI_ENTER
73 depends on HAVE_FTRACE_NMI_ENTER
77 select CONTEXT_SWITCH_TRACER
81 config CONTEXT_SWITCH_TRACER
84 config RING_BUFFER_ALLOW_SWAP
87 Allow the use of ring_buffer_swap_cpu.
88 Adds a very slight overhead to tracing when enabled.
90 config PREEMPTIRQ_TRACEPOINTS
92 depends on TRACE_PREEMPT_TOGGLE || TRACE_IRQFLAGS
96 Create preempt/irq toggle tracepoints if needed, so that other parts
97 of the kernel can use them to generate or add hooks to them.
99 # All tracer options should select GENERIC_TRACER. For those options that are
100 # enabled by all tracers (context switch and event tracer) they select TRACING.
101 # This allows those options to appear when no other tracer is selected. But the
102 # options do not appear when something else selects it. We need the two options
103 # GENERIC_TRACER and TRACING to avoid circular dependencies to accomplish the
104 # hiding of the automatic options.
110 select STACKTRACE if STACKTRACE_SUPPORT
117 config GENERIC_TRACER
122 # Minimum requirements an architecture has to meet for us to
123 # be able to offer generic tracing facilities:
125 config TRACING_SUPPORT
127 depends on TRACE_IRQFLAGS_SUPPORT
128 depends on STACKTRACE_SUPPORT
135 default y if DEBUG_KERNEL
137 Enable the kernel tracing infrastructure.
141 config FUNCTION_TRACER
142 bool "Kernel Function Tracer"
143 depends on HAVE_FUNCTION_TRACER
145 select GENERIC_TRACER
146 select CONTEXT_SWITCH_TRACER
148 select TASKS_RCU if PREEMPT
150 Enable the kernel to trace every kernel function. This is done
151 by using a compiler feature to insert a small, 5-byte No-Operation
152 instruction at the beginning of every kernel function, which NOP
153 sequence is then dynamically patched into a tracer call when
154 tracing is enabled by the administrator. If it's runtime disabled
155 (the bootup default), then the overhead of the instructions is very
156 small and not measurable even in micro-benchmarks.
158 config FUNCTION_GRAPH_TRACER
159 bool "Kernel Function Graph Tracer"
160 depends on HAVE_FUNCTION_GRAPH_TRACER
161 depends on FUNCTION_TRACER
162 depends on !X86_32 || !CC_OPTIMIZE_FOR_SIZE
165 Enable the kernel to trace a function at both its return
167 Its first purpose is to trace the duration of functions and
168 draw a call graph for each thread with some information like
169 the return value. This is done by setting the current return
170 address on the current task structure into a stack of calls.
172 config TRACE_PREEMPT_TOGGLE
175 Enables hooks which will be called when preemption is first disabled,
178 config PREEMPTIRQ_EVENTS
179 bool "Enable trace events for preempt and irq disable/enable"
180 select TRACE_IRQFLAGS
181 select TRACE_PREEMPT_TOGGLE if PREEMPT
182 select GENERIC_TRACER
185 Enable tracing of disable and enable events for preemption and irqs.
187 config IRQSOFF_TRACER
188 bool "Interrupts-off Latency Tracer"
190 depends on TRACE_IRQFLAGS_SUPPORT
191 depends on !ARCH_USES_GETTIMEOFFSET
192 select TRACE_IRQFLAGS
193 select GENERIC_TRACER
194 select TRACER_MAX_TRACE
195 select RING_BUFFER_ALLOW_SWAP
196 select TRACER_SNAPSHOT
197 select TRACER_SNAPSHOT_PER_CPU_SWAP
199 This option measures the time spent in irqs-off critical
200 sections, with microsecond accuracy.
202 The default measurement method is a maximum search, which is
203 disabled by default and can be runtime (re-)started
206 echo 0 > /sys/kernel/debug/tracing/tracing_max_latency
208 (Note that kernel size and overhead increase with this option
209 enabled. This option and the preempt-off timing option can be
210 used together or separately.)
212 config PREEMPT_TRACER
213 bool "Preemption-off Latency Tracer"
215 depends on !ARCH_USES_GETTIMEOFFSET
217 select GENERIC_TRACER
218 select TRACER_MAX_TRACE
219 select RING_BUFFER_ALLOW_SWAP
220 select TRACER_SNAPSHOT
221 select TRACER_SNAPSHOT_PER_CPU_SWAP
222 select TRACE_PREEMPT_TOGGLE
224 This option measures the time spent in preemption-off critical
225 sections, with microsecond accuracy.
227 The default measurement method is a maximum search, which is
228 disabled by default and can be runtime (re-)started
231 echo 0 > /sys/kernel/debug/tracing/tracing_max_latency
233 (Note that kernel size and overhead increase with this option
234 enabled. This option and the irqs-off timing option can be
235 used together or separately.)
238 bool "Scheduling Latency Tracer"
239 select GENERIC_TRACER
240 select CONTEXT_SWITCH_TRACER
241 select TRACER_MAX_TRACE
242 select TRACER_SNAPSHOT
244 This tracer tracks the latency of the highest priority task
245 to be scheduled in, starting from the point it has woken up.
248 bool "Tracer to detect hardware latencies (like SMIs)"
249 select GENERIC_TRACER
251 This tracer, when enabled will create one or more kernel threads,
252 depending on what the cpumask file is set to, which each thread
253 spinning in a loop looking for interruptions caused by
254 something other than the kernel. For example, if a
255 System Management Interrupt (SMI) takes a noticeable amount of
256 time, this tracer will detect it. This is useful for testing
257 if a system is reliable for Real Time tasks.
259 Some files are created in the tracing directory when this
262 hwlat_detector/width - time in usecs for how long to spin for
263 hwlat_detector/window - time in usecs between the start of each
266 A kernel thread is created that will spin with interrupts disabled
267 for "width" microseconds in every "window" cycle. It will not spin
268 for "window - width" microseconds, where the system can
271 The output will appear in the trace and trace_pipe files.
273 When the tracer is not running, it has no affect on the system,
274 but when it is running, it can cause the system to be
275 periodically non responsive. Do not run this tracer on a
278 To enable this tracer, echo in "hwlat" into the current_tracer
279 file. Every time a latency is greater than tracing_thresh, it will
280 be recorded into the ring buffer.
282 config ENABLE_DEFAULT_TRACERS
283 bool "Trace process context switches and events"
284 depends on !GENERIC_TRACER
287 This tracer hooks to various trace points in the kernel,
288 allowing the user to pick and choose which trace point they
289 want to trace. It also includes the sched_switch tracer plugin.
291 config FTRACE_SYSCALLS
292 bool "Trace syscalls"
293 depends on HAVE_SYSCALL_TRACEPOINTS
294 select GENERIC_TRACER
297 Basic tracer to catch the syscall entry and exit events.
299 config TRACER_SNAPSHOT
300 bool "Create a snapshot trace buffer"
301 select TRACER_MAX_TRACE
303 Allow tracing users to take snapshot of the current buffer using the
304 ftrace interface, e.g.:
306 echo 1 > /sys/kernel/debug/tracing/snapshot
309 config TRACER_SNAPSHOT_PER_CPU_SWAP
310 bool "Allow snapshot to swap per CPU"
311 depends on TRACER_SNAPSHOT
312 select RING_BUFFER_ALLOW_SWAP
314 Allow doing a snapshot of a single CPU buffer instead of a
315 full swap (all buffers). If this is set, then the following is
318 echo 1 > /sys/kernel/debug/tracing/per_cpu/cpu2/snapshot
320 After which, only the tracing buffer for CPU 2 was swapped with
321 the main tracing buffer, and the other CPU buffers remain the same.
323 When this is enabled, this adds a little more overhead to the
324 trace recording, as it needs to add some checks to synchronize
325 recording with swaps. But this does not affect the performance
326 of the overall system. This is enabled by default when the preempt
327 or irq latency tracers are enabled, as those need to swap as well
328 and already adds the overhead (plus a lot more).
330 config TRACE_BRANCH_PROFILING
332 select GENERIC_TRACER
335 prompt "Branch Profiling"
336 default BRANCH_PROFILE_NONE
338 The branch profiling is a software profiler. It will add hooks
339 into the C conditionals to test which path a branch takes.
341 The likely/unlikely profiler only looks at the conditions that
342 are annotated with a likely or unlikely macro.
344 The "all branch" profiler will profile every if-statement in the
345 kernel. This profiler will also enable the likely/unlikely
348 Either of the above profilers adds a bit of overhead to the system.
349 If unsure, choose "No branch profiling".
351 config BRANCH_PROFILE_NONE
352 bool "No branch profiling"
354 No branch profiling. Branch profiling adds a bit of overhead.
355 Only enable it if you want to analyse the branching behavior.
356 Otherwise keep it disabled.
358 config PROFILE_ANNOTATED_BRANCHES
359 bool "Trace likely/unlikely profiler"
360 select TRACE_BRANCH_PROFILING
362 This tracer profiles all likely and unlikely macros
363 in the kernel. It will display the results in:
365 /sys/kernel/debug/tracing/trace_stat/branch_annotated
367 Note: this will add a significant overhead; only turn this
368 on if you need to profile the system's use of these macros.
370 config PROFILE_ALL_BRANCHES
371 bool "Profile all if conditionals" if !FORTIFY_SOURCE
372 select TRACE_BRANCH_PROFILING
373 imply CC_DISABLE_WARN_MAYBE_UNINITIALIZED # avoid false positives
375 This tracer profiles all branch conditions. Every if ()
376 taken in the kernel is recorded whether it hit or miss.
377 The results will be displayed in:
379 /sys/kernel/debug/tracing/trace_stat/branch_all
381 This option also enables the likely/unlikely profiler.
383 This configuration, when enabled, will impose a great overhead
384 on the system. This should only be enabled when the system
385 is to be analyzed in much detail.
388 config TRACING_BRANCHES
391 Selected by tracers that will trace the likely and unlikely
392 conditions. This prevents the tracers themselves from being
393 profiled. Profiling the tracing infrastructure can only happen
394 when the likelys and unlikelys are not being traced.
397 bool "Trace likely/unlikely instances"
398 depends on TRACE_BRANCH_PROFILING
399 select TRACING_BRANCHES
401 This traces the events of likely and unlikely condition
402 calls in the kernel. The difference between this and the
403 "Trace likely/unlikely profiler" is that this is not a
404 histogram of the callers, but actually places the calling
405 events into a running trace buffer to see when and where the
406 events happened, as well as their results.
411 bool "Trace max stack"
412 depends on HAVE_FUNCTION_TRACER
413 select FUNCTION_TRACER
417 This special tracer records the maximum stack footprint of the
418 kernel and displays it in /sys/kernel/debug/tracing/stack_trace.
420 This tracer works by hooking into every function call that the
421 kernel executes, and keeping a maximum stack depth value and
422 stack-trace saved. If this is configured with DYNAMIC_FTRACE
423 then it will not have any overhead while the stack tracer
426 To enable the stack tracer on bootup, pass in 'stacktrace'
427 on the kernel command line.
429 The stack tracer can also be enabled or disabled via the
430 sysctl kernel.stack_tracer_enabled
434 config BLK_DEV_IO_TRACE
435 bool "Support for tracing block IO actions"
441 select GENERIC_TRACER
444 Say Y here if you want to be able to trace the block layer actions
445 on a given queue. Tracing allows you to see any traffic happening
446 on a block device queue. For more information (and the userspace
447 support tools needed), fetch the blktrace tools from:
449 git://git.kernel.dk/blktrace.git
451 Tracing also is possible using the ftrace interface, e.g.:
453 echo 1 > /sys/block/sda/sda1/trace/enable
454 echo blk > /sys/kernel/debug/tracing/current_tracer
455 cat /sys/kernel/debug/tracing/trace_pipe
461 depends on HAVE_REGS_AND_STACK_ACCESS_API
462 bool "Enable kprobes-based dynamic events"
465 select DYNAMIC_EVENTS
468 This allows the user to add tracing events (similar to tracepoints)
469 on the fly via the ftrace interface. See
470 Documentation/trace/kprobetrace.rst for more details.
472 Those events can be inserted wherever kprobes can probe, and record
473 various register and memory values.
475 This option is also required by perf-probe subcommand of perf tools.
476 If you want to use perf tools, this option is strongly recommended.
478 config KPROBE_EVENTS_ON_NOTRACE
479 bool "Do NOT protect notrace function from kprobe events"
480 depends on KPROBE_EVENTS
481 depends on KPROBES_ON_FTRACE
484 This is only for the developers who want to debug ftrace itself
487 If kprobes can use ftrace instead of breakpoint, ftrace related
488 functions are protected from kprobe-events to prevent an infinit
489 recursion or any unexpected execution path which leads to a kernel
492 This option disables such protection and allows you to put kprobe
493 events on ftrace functions for debugging ftrace by itself.
494 Note that this might let you shoot yourself in the foot.
499 bool "Enable uprobes-based dynamic events"
500 depends on ARCH_SUPPORTS_UPROBES
502 depends on PERF_EVENTS
505 select DYNAMIC_EVENTS
509 This allows the user to add tracing events on top of userspace
510 dynamic events (similar to tracepoints) on the fly via the trace
511 events interface. Those events can be inserted wherever uprobes
512 can probe, and record various registers.
513 This option is required if you plan to use perf-probe subcommand
514 of perf tools on user space applications.
517 depends on BPF_SYSCALL
518 depends on (KPROBE_EVENTS || UPROBE_EVENTS) && PERF_EVENTS
522 This allows the user to attach BPF programs to kprobe events.
524 config DYNAMIC_EVENTS
530 config DYNAMIC_FTRACE
531 bool "enable/disable function tracing dynamically"
532 depends on FUNCTION_TRACER
533 depends on HAVE_DYNAMIC_FTRACE
536 This option will modify all the calls to function tracing
537 dynamically (will patch them out of the binary image and
538 replace them with a No-Op instruction) on boot up. During
539 compile time, a table is made of all the locations that ftrace
540 can function trace, and this table is linked into the kernel
541 image. When this is enabled, functions can be individually
542 enabled, and the functions not enabled will not affect
543 performance of the system.
545 See the files in /sys/kernel/debug/tracing:
546 available_filter_functions
550 This way a CONFIG_FUNCTION_TRACER kernel is slightly larger, but
551 otherwise has native performance as long as no tracing is active.
553 config DYNAMIC_FTRACE_WITH_REGS
555 depends on DYNAMIC_FTRACE
556 depends on HAVE_DYNAMIC_FTRACE_WITH_REGS
558 config FUNCTION_PROFILER
559 bool "Kernel function profiler"
560 depends on FUNCTION_TRACER
563 This option enables the kernel function profiler. A file is created
564 in debugfs called function_profile_enabled which defaults to zero.
565 When a 1 is echoed into this file profiling begins, and when a
566 zero is entered, profiling stops. A "functions" file is created in
567 the trace_stat directory; this file shows the list of functions that
568 have been hit and their counters.
572 config BPF_KPROBE_OVERRIDE
573 bool "Enable BPF programs to override a kprobed function"
574 depends on BPF_EVENTS
575 depends on FUNCTION_ERROR_INJECTION
578 Allows BPF to override the execution of a probed function and
579 set a different return value. This is used for error injection.
581 config FTRACE_MCOUNT_RECORD
583 depends on DYNAMIC_FTRACE
584 depends on HAVE_FTRACE_MCOUNT_RECORD
586 config FTRACE_SELFTEST
589 config FTRACE_STARTUP_TEST
590 bool "Perform a startup test on ftrace"
591 depends on GENERIC_TRACER
592 select FTRACE_SELFTEST
594 This option performs a series of startup tests on ftrace. On bootup
595 a series of tests are made to verify that the tracer is
596 functioning properly. It will do tests on all the configured
599 config EVENT_TRACE_TEST_SYSCALLS
600 bool "Run selftest on syscall events"
601 depends on FTRACE_STARTUP_TEST
603 This option will also enable testing every syscall event.
604 It only enables the event and disables it and runs various loads
605 with the event enabled. This adds a bit more time for kernel boot
606 up since it runs this on every system call defined.
608 TBD - enable a way to actually call the syscalls as we test their
612 bool "Memory mapped IO tracing"
613 depends on HAVE_MMIOTRACE_SUPPORT && PCI
614 select GENERIC_TRACER
616 Mmiotrace traces Memory Mapped I/O access and is meant for
617 debugging and reverse engineering. It is called from the ioremap
618 implementation and works via page faults. Tracing is disabled by
619 default and can be enabled at run-time.
621 See Documentation/trace/mmiotrace.rst.
622 If you are not helping to develop drivers, say N.
626 depends on ARCH_HAVE_NMI_SAFE_CMPXCHG
628 tracing_map is a special-purpose lock-free map for tracing,
629 separated out as a stand-alone facility in order to allow it
630 to be shared between multiple tracers. It isn't meant to be
631 generally used outside of that context, and is normally
632 selected by tracers that use it.
635 bool "Histogram triggers"
636 depends on ARCH_HAVE_NMI_SAFE_CMPXCHG
639 select DYNAMIC_EVENTS
642 Hist triggers allow one or more arbitrary trace event fields
643 to be aggregated into hash tables and dumped to stdout by
644 reading a debugfs/tracefs file. They're useful for
645 gathering quick and dirty (though precise) summaries of
646 event activity as an initial guide for further investigation
647 using more advanced tools.
649 Inter-event tracing of quantities such as latencies is also
650 supported using hist triggers under this option.
652 See Documentation/trace/histogram.rst.
655 config MMIOTRACE_TEST
656 tristate "Test module for mmiotrace"
657 depends on MMIOTRACE && m
659 This is a dumb module for testing mmiotrace. It is very dangerous
660 as it will write garbage to IO memory starting at a given address.
661 However, it should be safe to use on e.g. unused portion of VRAM.
663 Say N, unless you absolutely know what you are doing.
665 config TRACEPOINT_BENCHMARK
666 bool "Add tracepoint that benchmarks tracepoints"
668 This option creates the tracepoint "benchmark:benchmark_event".
669 When the tracepoint is enabled, it kicks off a kernel thread that
670 goes into an infinite loop (calling cond_sched() to let other tasks
671 run), and calls the tracepoint. Each iteration will record the time
672 it took to write to the tracepoint and the next iteration that
673 data will be passed to the tracepoint itself. That is, the tracepoint
674 will report the time it took to do the previous tracepoint.
675 The string written to the tracepoint is a static string of 128 bytes
676 to keep the time the same. The initial string is simply a write of
677 "START". The second string records the cold cache time of the first
678 write which is not added to the rest of the calculations.
680 As it is a tight loop, it benchmarks as hot cache. That's fine because
681 we care most about hot paths that are probably in cache already.
683 An example of the output:
686 first=3672 [COLD CACHED]
687 last=632 first=3672 max=632 min=632 avg=316 std=446 std^2=199712
688 last=278 first=3672 max=632 min=278 avg=303 std=316 std^2=100337
689 last=277 first=3672 max=632 min=277 avg=296 std=258 std^2=67064
690 last=273 first=3672 max=632 min=273 avg=292 std=224 std^2=50411
691 last=273 first=3672 max=632 min=273 avg=288 std=200 std^2=40389
692 last=281 first=3672 max=632 min=273 avg=287 std=183 std^2=33666
695 config RING_BUFFER_BENCHMARK
696 tristate "Ring buffer benchmark stress tester"
697 depends on RING_BUFFER
699 This option creates a test to stress the ring buffer and benchmark it.
700 It creates its own ring buffer such that it will not interfere with
701 any other users of the ring buffer (such as ftrace). It then creates
702 a producer and consumer that will run for 10 seconds and sleep for
703 10 seconds. Each interval it will print out the number of events
704 it recorded and give a rough estimate of how long each iteration took.
706 It does not disable interrupts or raise its priority, so it may be
707 affected by processes that are running.
711 config RING_BUFFER_STARTUP_TEST
712 bool "Ring buffer startup self test"
713 depends on RING_BUFFER
715 Run a simple self test on the ring buffer on boot up. Late in the
716 kernel boot sequence, the test will start that kicks off
717 a thread per cpu. Each thread will write various size events
718 into the ring buffer. Another thread is created to send IPIs
719 to each of the threads, where the IPI handler will also write
720 to the ring buffer, to test/stress the nesting ability.
721 If any anomalies are discovered, a warning will be displayed
722 and all ring buffers will be disabled.
724 The test runs for 10 seconds. This will slow your boot time
725 by at least 10 more seconds.
727 At the end of the test, statics and more checks are done.
728 It will output the stats of each per cpu buffer. What
729 was written, the sizes, what was read, what was lost, and
730 other similar details.
734 config PREEMPTIRQ_DELAY_TEST
735 tristate "Preempt / IRQ disable delay thread to test latency tracers"
738 Select this option to build a test module that can help test latency
739 tracers by executing a preempt or irq disable section with a user
740 configurable delay. The module busy waits for the duration of the
743 For example, the following invocation forces a one-time irq-disabled
744 critical section for 500us:
745 modprobe preemptirq_delay_test test_mode=irq delay=500000
749 config TRACE_EVAL_MAP_FILE
750 bool "Show eval mappings for trace events"
753 The "print fmt" of the trace events will show the enum/sizeof names
754 instead of their values. This can cause problems for user space tools
755 that use this string to parse the raw data as user space does not know
756 how to convert the string to its value.
758 To fix this, there's a special macro in the kernel that can be used
759 to convert an enum/sizeof into its value. If this macro is used, then
760 the print fmt strings will be converted to their values.
762 If something does not get converted properly, this option can be
763 used to show what enums/sizeof the kernel tried to convert.
765 This option is for debugging the conversions. A file is created
766 in the tracing directory called "eval_map" that will show the
767 names matched with their values and what trace event system they
770 Normally, the mapping of the strings to values will be freed after
771 boot up or module load. With this option, they will not be freed, as
772 they are needed for the "eval_map" file. Enabling this option will
773 increase the memory footprint of the running kernel.
777 config GCOV_PROFILE_FTRACE
778 bool "Enable GCOV profiling on ftrace subsystem"
779 depends on GCOV_KERNEL
781 Enable GCOV profiling on ftrace subsystem for checking
782 which functions/lines are tested.
786 Note that on a kernel compiled with this config, ftrace will
787 run significantly slower.
791 endif # TRACING_SUPPORT