1 # SPDX-License-Identifier: GPL-2.0-only
3 # Architectures that offer an FUNCTION_TRACER implementation should
4 # select HAVE_FUNCTION_TRACER:
7 config USER_STACKTRACE_SUPPORT
18 depends on HAVE_RETHOOK
20 Enable generic return hooking feature. This is an internal
21 API, which will be used by other function-entry hooking
22 features like fprobe and kprobes.
24 config HAVE_FUNCTION_TRACER
27 See Documentation/trace/ftrace-design.rst
29 config HAVE_FUNCTION_GRAPH_TRACER
32 See Documentation/trace/ftrace-design.rst
34 config HAVE_DYNAMIC_FTRACE
37 See Documentation/trace/ftrace-design.rst
39 config HAVE_DYNAMIC_FTRACE_WITH_REGS
42 config HAVE_DYNAMIC_FTRACE_WITH_DIRECT_CALLS
45 config HAVE_DYNAMIC_FTRACE_WITH_CALL_OPS
48 config HAVE_DYNAMIC_FTRACE_WITH_ARGS
51 If this is set, then arguments and stack can be found from
52 the ftrace_regs passed into the function callback regs parameter
53 by default, even without setting the REGS flag in the ftrace_ops.
54 This allows for use of ftrace_regs_get_argument() and
55 ftrace_regs_get_stack_pointer().
57 config HAVE_DYNAMIC_FTRACE_NO_PATCHABLE
60 If the architecture generates __patchable_function_entries sections
61 but does not want them included in the ftrace locations.
63 config HAVE_FTRACE_MCOUNT_RECORD
66 See Documentation/trace/ftrace-design.rst
68 config HAVE_SYSCALL_TRACEPOINTS
71 See Documentation/trace/ftrace-design.rst
76 Arch supports the gcc options -pg with -mfentry
78 config HAVE_NOP_MCOUNT
81 Arch supports the gcc options -pg with -mrecord-mcount and -nop-mcount
83 config HAVE_OBJTOOL_MCOUNT
86 Arch supports objtool --mcount
88 config HAVE_OBJTOOL_NOP_MCOUNT
91 Arch supports the objtool options --mcount with --mnop.
92 An architecture can select this if it wants to enable nop'ing
95 config HAVE_C_RECORDMCOUNT
98 C version of recordmcount available?
100 config HAVE_BUILDTIME_MCOUNT_SORT
103 An architecture selects this if it sorts the mcount_loc section
106 config BUILDTIME_MCOUNT_SORT
109 depends on HAVE_BUILDTIME_MCOUNT_SORT && DYNAMIC_FTRACE
111 Sort the mcount_loc section at build time.
113 config TRACER_MAX_TRACE
125 select CONTEXT_SWITCH_TRACER
129 config CONTEXT_SWITCH_TRACER
132 config RING_BUFFER_ALLOW_SWAP
135 Allow the use of ring_buffer_swap_cpu.
136 Adds a very slight overhead to tracing when enabled.
138 config PREEMPTIRQ_TRACEPOINTS
140 depends on TRACE_PREEMPT_TOGGLE || TRACE_IRQFLAGS
144 Create preempt/irq toggle tracepoints if needed, so that other parts
145 of the kernel can use them to generate or add hooks to them.
147 # All tracer options should select GENERIC_TRACER. For those options that are
148 # enabled by all tracers (context switch and event tracer) they select TRACING.
149 # This allows those options to appear when no other tracer is selected. But the
150 # options do not appear when something else selects it. We need the two options
151 # GENERIC_TRACER and TRACING to avoid circular dependencies to accomplish the
152 # hiding of the automatic options.
157 select STACKTRACE if STACKTRACE_SUPPORT
163 select TASKS_RCU if PREEMPTION
165 config GENERIC_TRACER
170 # Minimum requirements an architecture has to meet for us to
171 # be able to offer generic tracing facilities:
173 config TRACING_SUPPORT
175 depends on TRACE_IRQFLAGS_SUPPORT
176 depends on STACKTRACE_SUPPORT
181 depends on TRACING_SUPPORT
182 default y if DEBUG_KERNEL
184 Enable the kernel tracing infrastructure.
188 config BOOTTIME_TRACING
189 bool "Boot-time Tracing support"
193 Enable developer to setup ftrace subsystem via supplemental
194 kernel cmdline at boot time for debugging (tracing) driver
195 initialization and boot process.
197 config FUNCTION_TRACER
198 bool "Kernel Function Tracer"
199 depends on HAVE_FUNCTION_TRACER
201 select GENERIC_TRACER
202 select CONTEXT_SWITCH_TRACER
204 select TASKS_RCU if PREEMPTION
205 select TASKS_RUDE_RCU
207 Enable the kernel to trace every kernel function. This is done
208 by using a compiler feature to insert a small, 5-byte No-Operation
209 instruction at the beginning of every kernel function, which NOP
210 sequence is then dynamically patched into a tracer call when
211 tracing is enabled by the administrator. If it's runtime disabled
212 (the bootup default), then the overhead of the instructions is very
213 small and not measurable even in micro-benchmarks (at least on
214 x86, but may have impact on other architectures).
216 config FUNCTION_GRAPH_TRACER
217 bool "Kernel Function Graph Tracer"
218 depends on HAVE_FUNCTION_GRAPH_TRACER
219 depends on FUNCTION_TRACER
220 depends on !X86_32 || !CC_OPTIMIZE_FOR_SIZE
223 Enable the kernel to trace a function at both its return
225 Its first purpose is to trace the duration of functions and
226 draw a call graph for each thread with some information like
227 the return value. This is done by setting the current return
228 address on the current task structure into a stack of calls.
230 config DYNAMIC_FTRACE
231 bool "enable/disable function tracing dynamically"
232 depends on FUNCTION_TRACER
233 depends on HAVE_DYNAMIC_FTRACE
236 This option will modify all the calls to function tracing
237 dynamically (will patch them out of the binary image and
238 replace them with a No-Op instruction) on boot up. During
239 compile time, a table is made of all the locations that ftrace
240 can function trace, and this table is linked into the kernel
241 image. When this is enabled, functions can be individually
242 enabled, and the functions not enabled will not affect
243 performance of the system.
245 See the files in /sys/kernel/tracing:
246 available_filter_functions
250 This way a CONFIG_FUNCTION_TRACER kernel is slightly larger, but
251 otherwise has native performance as long as no tracing is active.
253 config DYNAMIC_FTRACE_WITH_REGS
255 depends on DYNAMIC_FTRACE
256 depends on HAVE_DYNAMIC_FTRACE_WITH_REGS
258 config DYNAMIC_FTRACE_WITH_DIRECT_CALLS
260 depends on DYNAMIC_FTRACE_WITH_REGS || DYNAMIC_FTRACE_WITH_ARGS
261 depends on HAVE_DYNAMIC_FTRACE_WITH_DIRECT_CALLS
263 config DYNAMIC_FTRACE_WITH_CALL_OPS
265 depends on HAVE_DYNAMIC_FTRACE_WITH_CALL_OPS
267 config DYNAMIC_FTRACE_WITH_ARGS
269 depends on DYNAMIC_FTRACE
270 depends on HAVE_DYNAMIC_FTRACE_WITH_ARGS
273 bool "Kernel Function Probe (fprobe)"
274 depends on FUNCTION_TRACER
275 depends on DYNAMIC_FTRACE_WITH_REGS
276 depends on HAVE_RETHOOK
280 This option enables kernel function probe (fprobe) based on ftrace.
281 The fprobe is similar to kprobes, but probes only for kernel function
282 entries and exits. This also can probe multiple functions by one
287 config FUNCTION_PROFILER
288 bool "Kernel function profiler"
289 depends on FUNCTION_TRACER
292 This option enables the kernel function profiler. A file is created
293 in debugfs called function_profile_enabled which defaults to zero.
294 When a 1 is echoed into this file profiling begins, and when a
295 zero is entered, profiling stops. A "functions" file is created in
296 the trace_stat directory; this file shows the list of functions that
297 have been hit and their counters.
302 bool "Trace max stack"
303 depends on HAVE_FUNCTION_TRACER
304 select FUNCTION_TRACER
308 This special tracer records the maximum stack footprint of the
309 kernel and displays it in /sys/kernel/tracing/stack_trace.
311 This tracer works by hooking into every function call that the
312 kernel executes, and keeping a maximum stack depth value and
313 stack-trace saved. If this is configured with DYNAMIC_FTRACE
314 then it will not have any overhead while the stack tracer
317 To enable the stack tracer on bootup, pass in 'stacktrace'
318 on the kernel command line.
320 The stack tracer can also be enabled or disabled via the
321 sysctl kernel.stack_tracer_enabled
325 config TRACE_PREEMPT_TOGGLE
328 Enables hooks which will be called when preemption is first disabled,
331 config IRQSOFF_TRACER
332 bool "Interrupts-off Latency Tracer"
334 depends on TRACE_IRQFLAGS_SUPPORT
335 select TRACE_IRQFLAGS
336 select GENERIC_TRACER
337 select TRACER_MAX_TRACE
338 select RING_BUFFER_ALLOW_SWAP
339 select TRACER_SNAPSHOT
340 select TRACER_SNAPSHOT_PER_CPU_SWAP
342 This option measures the time spent in irqs-off critical
343 sections, with microsecond accuracy.
345 The default measurement method is a maximum search, which is
346 disabled by default and can be runtime (re-)started
349 echo 0 > /sys/kernel/tracing/tracing_max_latency
351 (Note that kernel size and overhead increase with this option
352 enabled. This option and the preempt-off timing option can be
353 used together or separately.)
355 config PREEMPT_TRACER
356 bool "Preemption-off Latency Tracer"
358 depends on PREEMPTION
359 select GENERIC_TRACER
360 select TRACER_MAX_TRACE
361 select RING_BUFFER_ALLOW_SWAP
362 select TRACER_SNAPSHOT
363 select TRACER_SNAPSHOT_PER_CPU_SWAP
364 select TRACE_PREEMPT_TOGGLE
366 This option measures the time spent in preemption-off critical
367 sections, with microsecond accuracy.
369 The default measurement method is a maximum search, which is
370 disabled by default and can be runtime (re-)started
373 echo 0 > /sys/kernel/tracing/tracing_max_latency
375 (Note that kernel size and overhead increase with this option
376 enabled. This option and the irqs-off timing option can be
377 used together or separately.)
380 bool "Scheduling Latency Tracer"
381 select GENERIC_TRACER
382 select CONTEXT_SWITCH_TRACER
383 select TRACER_MAX_TRACE
384 select TRACER_SNAPSHOT
386 This tracer tracks the latency of the highest priority task
387 to be scheduled in, starting from the point it has woken up.
390 bool "Tracer to detect hardware latencies (like SMIs)"
391 select GENERIC_TRACER
392 select TRACER_MAX_TRACE
394 This tracer, when enabled will create one or more kernel threads,
395 depending on what the cpumask file is set to, which each thread
396 spinning in a loop looking for interruptions caused by
397 something other than the kernel. For example, if a
398 System Management Interrupt (SMI) takes a noticeable amount of
399 time, this tracer will detect it. This is useful for testing
400 if a system is reliable for Real Time tasks.
402 Some files are created in the tracing directory when this
405 hwlat_detector/width - time in usecs for how long to spin for
406 hwlat_detector/window - time in usecs between the start of each
409 A kernel thread is created that will spin with interrupts disabled
410 for "width" microseconds in every "window" cycle. It will not spin
411 for "window - width" microseconds, where the system can
414 The output will appear in the trace and trace_pipe files.
416 When the tracer is not running, it has no affect on the system,
417 but when it is running, it can cause the system to be
418 periodically non responsive. Do not run this tracer on a
421 To enable this tracer, echo in "hwlat" into the current_tracer
422 file. Every time a latency is greater than tracing_thresh, it will
423 be recorded into the ring buffer.
425 config OSNOISE_TRACER
426 bool "OS Noise tracer"
427 select GENERIC_TRACER
428 select TRACER_MAX_TRACE
430 In the context of high-performance computing (HPC), the Operating
431 System Noise (osnoise) refers to the interference experienced by an
432 application due to activities inside the operating system. In the
433 context of Linux, NMIs, IRQs, SoftIRQs, and any other system thread
434 can cause noise to the system. Moreover, hardware-related jobs can
435 also cause noise, for example, via SMIs.
437 The osnoise tracer leverages the hwlat_detector by running a similar
438 loop with preemption, SoftIRQs and IRQs enabled, thus allowing all
439 the sources of osnoise during its execution. The osnoise tracer takes
440 note of the entry and exit point of any source of interferences,
441 increasing a per-cpu interference counter. It saves an interference
442 counter for each source of interference. The interference counter for
443 NMI, IRQs, SoftIRQs, and threads is increased anytime the tool
444 observes these interferences' entry events. When a noise happens
445 without any interference from the operating system level, the
446 hardware noise counter increases, pointing to a hardware-related
447 noise. In this way, osnoise can account for any source of
448 interference. At the end of the period, the osnoise tracer prints
449 the sum of all noise, the max single noise, the percentage of CPU
450 available for the thread, and the counters for the noise sources.
452 In addition to the tracer, a set of tracepoints were added to
453 facilitate the identification of the osnoise source.
455 The output will appear in the trace and trace_pipe files.
457 To enable this tracer, echo in "osnoise" into the current_tracer
460 config TIMERLAT_TRACER
461 bool "Timerlat tracer"
462 select OSNOISE_TRACER
463 select GENERIC_TRACER
465 The timerlat tracer aims to help the preemptive kernel developers
466 to find sources of wakeup latencies of real-time threads.
468 The tracer creates a per-cpu kernel thread with real-time priority.
469 The tracer thread sets a periodic timer to wakeup itself, and goes
470 to sleep waiting for the timer to fire. At the wakeup, the thread
471 then computes a wakeup latency value as the difference between
472 the current time and the absolute time that the timer was set
475 The tracer prints two lines at every activation. The first is the
476 timer latency observed at the hardirq context before the
477 activation of the thread. The second is the timer latency observed
478 by the thread, which is the same level that cyclictest reports. The
479 ACTIVATION ID field serves to relate the irq execution to its
480 respective thread execution.
482 The tracer is build on top of osnoise tracer, and the osnoise:
483 events can be used to trace the source of interference from NMI,
484 IRQs and other threads. It also enables the capture of the
485 stacktrace at the IRQ context, which helps to identify the code
486 path that can cause thread delay.
489 bool "Memory mapped IO tracing"
490 depends on HAVE_MMIOTRACE_SUPPORT && PCI
491 select GENERIC_TRACER
493 Mmiotrace traces Memory Mapped I/O access and is meant for
494 debugging and reverse engineering. It is called from the ioremap
495 implementation and works via page faults. Tracing is disabled by
496 default and can be enabled at run-time.
498 See Documentation/trace/mmiotrace.rst.
499 If you are not helping to develop drivers, say N.
501 config ENABLE_DEFAULT_TRACERS
502 bool "Trace process context switches and events"
503 depends on !GENERIC_TRACER
506 This tracer hooks to various trace points in the kernel,
507 allowing the user to pick and choose which trace point they
508 want to trace. It also includes the sched_switch tracer plugin.
510 config FTRACE_SYSCALLS
511 bool "Trace syscalls"
512 depends on HAVE_SYSCALL_TRACEPOINTS
513 select GENERIC_TRACER
516 Basic tracer to catch the syscall entry and exit events.
518 config TRACER_SNAPSHOT
519 bool "Create a snapshot trace buffer"
520 select TRACER_MAX_TRACE
522 Allow tracing users to take snapshot of the current buffer using the
523 ftrace interface, e.g.:
525 echo 1 > /sys/kernel/tracing/snapshot
528 config TRACER_SNAPSHOT_PER_CPU_SWAP
529 bool "Allow snapshot to swap per CPU"
530 depends on TRACER_SNAPSHOT
531 select RING_BUFFER_ALLOW_SWAP
533 Allow doing a snapshot of a single CPU buffer instead of a
534 full swap (all buffers). If this is set, then the following is
537 echo 1 > /sys/kernel/tracing/per_cpu/cpu2/snapshot
539 After which, only the tracing buffer for CPU 2 was swapped with
540 the main tracing buffer, and the other CPU buffers remain the same.
542 When this is enabled, this adds a little more overhead to the
543 trace recording, as it needs to add some checks to synchronize
544 recording with swaps. But this does not affect the performance
545 of the overall system. This is enabled by default when the preempt
546 or irq latency tracers are enabled, as those need to swap as well
547 and already adds the overhead (plus a lot more).
549 config TRACE_BRANCH_PROFILING
551 select GENERIC_TRACER
554 prompt "Branch Profiling"
555 default BRANCH_PROFILE_NONE
557 The branch profiling is a software profiler. It will add hooks
558 into the C conditionals to test which path a branch takes.
560 The likely/unlikely profiler only looks at the conditions that
561 are annotated with a likely or unlikely macro.
563 The "all branch" profiler will profile every if-statement in the
564 kernel. This profiler will also enable the likely/unlikely
567 Either of the above profilers adds a bit of overhead to the system.
568 If unsure, choose "No branch profiling".
570 config BRANCH_PROFILE_NONE
571 bool "No branch profiling"
573 No branch profiling. Branch profiling adds a bit of overhead.
574 Only enable it if you want to analyse the branching behavior.
575 Otherwise keep it disabled.
577 config PROFILE_ANNOTATED_BRANCHES
578 bool "Trace likely/unlikely profiler"
579 select TRACE_BRANCH_PROFILING
581 This tracer profiles all likely and unlikely macros
582 in the kernel. It will display the results in:
584 /sys/kernel/tracing/trace_stat/branch_annotated
586 Note: this will add a significant overhead; only turn this
587 on if you need to profile the system's use of these macros.
589 config PROFILE_ALL_BRANCHES
590 bool "Profile all if conditionals" if !FORTIFY_SOURCE
591 select TRACE_BRANCH_PROFILING
593 This tracer profiles all branch conditions. Every if ()
594 taken in the kernel is recorded whether it hit or miss.
595 The results will be displayed in:
597 /sys/kernel/tracing/trace_stat/branch_all
599 This option also enables the likely/unlikely profiler.
601 This configuration, when enabled, will impose a great overhead
602 on the system. This should only be enabled when the system
603 is to be analyzed in much detail.
606 config TRACING_BRANCHES
609 Selected by tracers that will trace the likely and unlikely
610 conditions. This prevents the tracers themselves from being
611 profiled. Profiling the tracing infrastructure can only happen
612 when the likelys and unlikelys are not being traced.
615 bool "Trace likely/unlikely instances"
616 depends on TRACE_BRANCH_PROFILING
617 select TRACING_BRANCHES
619 This traces the events of likely and unlikely condition
620 calls in the kernel. The difference between this and the
621 "Trace likely/unlikely profiler" is that this is not a
622 histogram of the callers, but actually places the calling
623 events into a running trace buffer to see when and where the
624 events happened, as well as their results.
628 config BLK_DEV_IO_TRACE
629 bool "Support for tracing block IO actions"
635 select GENERIC_TRACER
638 Say Y here if you want to be able to trace the block layer actions
639 on a given queue. Tracing allows you to see any traffic happening
640 on a block device queue. For more information (and the userspace
641 support tools needed), fetch the blktrace tools from:
643 git://git.kernel.dk/blktrace.git
645 Tracing also is possible using the ftrace interface, e.g.:
647 echo 1 > /sys/block/sda/sda1/trace/enable
648 echo blk > /sys/kernel/tracing/current_tracer
649 cat /sys/kernel/tracing/trace_pipe
655 depends on HAVE_REGS_AND_STACK_ACCESS_API
656 bool "Enable kprobes-based dynamic events"
659 select DYNAMIC_EVENTS
662 This allows the user to add tracing events (similar to tracepoints)
663 on the fly via the ftrace interface. See
664 Documentation/trace/kprobetrace.rst for more details.
666 Those events can be inserted wherever kprobes can probe, and record
667 various register and memory values.
669 This option is also required by perf-probe subcommand of perf tools.
670 If you want to use perf tools, this option is strongly recommended.
672 config KPROBE_EVENTS_ON_NOTRACE
673 bool "Do NOT protect notrace function from kprobe events"
674 depends on KPROBE_EVENTS
675 depends on DYNAMIC_FTRACE
678 This is only for the developers who want to debug ftrace itself
681 If kprobes can use ftrace instead of breakpoint, ftrace related
682 functions are protected from kprobe-events to prevent an infinite
683 recursion or any unexpected execution path which leads to a kernel
686 This option disables such protection and allows you to put kprobe
687 events on ftrace functions for debugging ftrace by itself.
688 Note that this might let you shoot yourself in the foot.
693 bool "Enable uprobes-based dynamic events"
694 depends on ARCH_SUPPORTS_UPROBES
696 depends on PERF_EVENTS
699 select DYNAMIC_EVENTS
703 This allows the user to add tracing events on top of userspace
704 dynamic events (similar to tracepoints) on the fly via the trace
705 events interface. Those events can be inserted wherever uprobes
706 can probe, and record various registers.
707 This option is required if you plan to use perf-probe subcommand
708 of perf tools on user space applications.
711 depends on BPF_SYSCALL
712 depends on (KPROBE_EVENTS || UPROBE_EVENTS) && PERF_EVENTS
716 This allows the user to attach BPF programs to kprobe, uprobe, and
719 config DYNAMIC_EVENTS
725 config BPF_KPROBE_OVERRIDE
726 bool "Enable BPF programs to override a kprobed function"
727 depends on BPF_EVENTS
728 depends on FUNCTION_ERROR_INJECTION
731 Allows BPF to override the execution of a probed function and
732 set a different return value. This is used for error injection.
734 config FTRACE_MCOUNT_RECORD
736 depends on DYNAMIC_FTRACE
737 depends on HAVE_FTRACE_MCOUNT_RECORD
739 config FTRACE_MCOUNT_USE_PATCHABLE_FUNCTION_ENTRY
741 depends on FTRACE_MCOUNT_RECORD
743 config FTRACE_MCOUNT_USE_CC
745 depends on $(cc-option,-mrecord-mcount)
746 depends on !FTRACE_MCOUNT_USE_PATCHABLE_FUNCTION_ENTRY
747 depends on FTRACE_MCOUNT_RECORD
749 config FTRACE_MCOUNT_USE_OBJTOOL
751 depends on HAVE_OBJTOOL_MCOUNT
752 depends on !FTRACE_MCOUNT_USE_PATCHABLE_FUNCTION_ENTRY
753 depends on !FTRACE_MCOUNT_USE_CC
754 depends on FTRACE_MCOUNT_RECORD
757 config FTRACE_MCOUNT_USE_RECORDMCOUNT
759 depends on !FTRACE_MCOUNT_USE_PATCHABLE_FUNCTION_ENTRY
760 depends on !FTRACE_MCOUNT_USE_CC
761 depends on !FTRACE_MCOUNT_USE_OBJTOOL
762 depends on FTRACE_MCOUNT_RECORD
766 depends on ARCH_HAVE_NMI_SAFE_CMPXCHG
768 tracing_map is a special-purpose lock-free map for tracing,
769 separated out as a stand-alone facility in order to allow it
770 to be shared between multiple tracers. It isn't meant to be
771 generally used outside of that context, and is normally
772 selected by tracers that use it.
775 bool "Synthetic trace events"
777 select DYNAMIC_EVENTS
780 Synthetic events are user-defined trace events that can be
781 used to combine data from other trace events or in fact any
782 data source. Synthetic events can be generated indirectly
783 via the trace() action of histogram triggers or directly
784 by way of an in-kernel API.
786 See Documentation/trace/events.rst or
787 Documentation/trace/histogram.rst for details and examples.
792 bool "User trace events"
794 select DYNAMIC_EVENTS
796 User trace events are user-defined trace events that
797 can be used like an existing kernel trace event. User trace
798 events are generated by writing to a tracefs file. User
799 processes can determine if their tracing events should be
800 generated by registering a value and bit with the kernel
801 that reflects when it is enabled or not.
803 See Documentation/trace/user_events.rst.
807 bool "Histogram triggers"
808 depends on ARCH_HAVE_NMI_SAFE_CMPXCHG
811 select DYNAMIC_EVENTS
815 Hist triggers allow one or more arbitrary trace event fields
816 to be aggregated into hash tables and dumped to stdout by
817 reading a debugfs/tracefs file. They're useful for
818 gathering quick and dirty (though precise) summaries of
819 event activity as an initial guide for further investigation
820 using more advanced tools.
822 Inter-event tracing of quantities such as latencies is also
823 supported using hist triggers under this option.
825 See Documentation/trace/histogram.rst.
828 config TRACE_EVENT_INJECT
829 bool "Trace event injection"
832 Allow user-space to inject a specific trace event into the ring
833 buffer. This is mainly used for testing purpose.
837 config TRACEPOINT_BENCHMARK
838 bool "Add tracepoint that benchmarks tracepoints"
840 This option creates the tracepoint "benchmark:benchmark_event".
841 When the tracepoint is enabled, it kicks off a kernel thread that
842 goes into an infinite loop (calling cond_resched() to let other tasks
843 run), and calls the tracepoint. Each iteration will record the time
844 it took to write to the tracepoint and the next iteration that
845 data will be passed to the tracepoint itself. That is, the tracepoint
846 will report the time it took to do the previous tracepoint.
847 The string written to the tracepoint is a static string of 128 bytes
848 to keep the time the same. The initial string is simply a write of
849 "START". The second string records the cold cache time of the first
850 write which is not added to the rest of the calculations.
852 As it is a tight loop, it benchmarks as hot cache. That's fine because
853 we care most about hot paths that are probably in cache already.
855 An example of the output:
858 first=3672 [COLD CACHED]
859 last=632 first=3672 max=632 min=632 avg=316 std=446 std^2=199712
860 last=278 first=3672 max=632 min=278 avg=303 std=316 std^2=100337
861 last=277 first=3672 max=632 min=277 avg=296 std=258 std^2=67064
862 last=273 first=3672 max=632 min=273 avg=292 std=224 std^2=50411
863 last=273 first=3672 max=632 min=273 avg=288 std=200 std^2=40389
864 last=281 first=3672 max=632 min=273 avg=287 std=183 std^2=33666
867 config RING_BUFFER_BENCHMARK
868 tristate "Ring buffer benchmark stress tester"
869 depends on RING_BUFFER
871 This option creates a test to stress the ring buffer and benchmark it.
872 It creates its own ring buffer such that it will not interfere with
873 any other users of the ring buffer (such as ftrace). It then creates
874 a producer and consumer that will run for 10 seconds and sleep for
875 10 seconds. Each interval it will print out the number of events
876 it recorded and give a rough estimate of how long each iteration took.
878 It does not disable interrupts or raise its priority, so it may be
879 affected by processes that are running.
883 config TRACE_EVAL_MAP_FILE
884 bool "Show eval mappings for trace events"
887 The "print fmt" of the trace events will show the enum/sizeof names
888 instead of their values. This can cause problems for user space tools
889 that use this string to parse the raw data as user space does not know
890 how to convert the string to its value.
892 To fix this, there's a special macro in the kernel that can be used
893 to convert an enum/sizeof into its value. If this macro is used, then
894 the print fmt strings will be converted to their values.
896 If something does not get converted properly, this option can be
897 used to show what enums/sizeof the kernel tried to convert.
899 This option is for debugging the conversions. A file is created
900 in the tracing directory called "eval_map" that will show the
901 names matched with their values and what trace event system they
904 Normally, the mapping of the strings to values will be freed after
905 boot up or module load. With this option, they will not be freed, as
906 they are needed for the "eval_map" file. Enabling this option will
907 increase the memory footprint of the running kernel.
911 config FTRACE_RECORD_RECURSION
912 bool "Record functions that recurse in function tracing"
913 depends on FUNCTION_TRACER
915 All callbacks that attach to the function tracing have some sort
916 of protection against recursion. Even though the protection exists,
917 it adds overhead. This option will create a file in the tracefs
918 file system called "recursed_functions" that will list the functions
919 that triggered a recursion.
921 This will add more overhead to cases that have recursion.
925 config FTRACE_RECORD_RECURSION_SIZE
926 int "Max number of recursed functions to record"
928 depends on FTRACE_RECORD_RECURSION
930 This defines the limit of number of functions that can be
931 listed in the "recursed_functions" file, that lists all
932 the functions that caused a recursion to happen.
933 This file can be reset, but the limit can not change in
936 config RING_BUFFER_RECORD_RECURSION
937 bool "Record functions that recurse in the ring buffer"
938 depends on FTRACE_RECORD_RECURSION
939 # default y, because it is coupled with FTRACE_RECORD_RECURSION
942 The ring buffer has its own internal recursion. Although when
943 recursion happens it won't cause harm because of the protection,
944 but it does cause unwanted overhead. Enabling this option will
945 place where recursion was detected into the ftrace "recursed_functions"
948 This will add more overhead to cases that have recursion.
950 config GCOV_PROFILE_FTRACE
951 bool "Enable GCOV profiling on ftrace subsystem"
952 depends on GCOV_KERNEL
954 Enable GCOV profiling on ftrace subsystem for checking
955 which functions/lines are tested.
959 Note that on a kernel compiled with this config, ftrace will
960 run significantly slower.
962 config FTRACE_SELFTEST
965 config FTRACE_STARTUP_TEST
966 bool "Perform a startup test on ftrace"
967 depends on GENERIC_TRACER
968 select FTRACE_SELFTEST
970 This option performs a series of startup tests on ftrace. On bootup
971 a series of tests are made to verify that the tracer is
972 functioning properly. It will do tests on all the configured
975 config EVENT_TRACE_STARTUP_TEST
976 bool "Run selftest on trace events"
977 depends on FTRACE_STARTUP_TEST
980 This option performs a test on all trace events in the system.
981 It basically just enables each event and runs some code that
982 will trigger events (not necessarily the event it enables)
983 This may take some time run as there are a lot of events.
985 config EVENT_TRACE_TEST_SYSCALLS
986 bool "Run selftest on syscall events"
987 depends on EVENT_TRACE_STARTUP_TEST
989 This option will also enable testing every syscall event.
990 It only enables the event and disables it and runs various loads
991 with the event enabled. This adds a bit more time for kernel boot
992 up since it runs this on every system call defined.
994 TBD - enable a way to actually call the syscalls as we test their
997 config FTRACE_SORT_STARTUP_TEST
998 bool "Verify compile time sorting of ftrace functions"
999 depends on DYNAMIC_FTRACE
1000 depends on BUILDTIME_MCOUNT_SORT
1002 Sorting of the mcount_loc sections that is used to find the
1003 where the ftrace knows where to patch functions for tracing
1004 and other callbacks is done at compile time. But if the sort
1005 is not done correctly, it will cause non-deterministic failures.
1006 When this is set, the sorted sections will be verified that they
1007 are in deed sorted and will warn if they are not.
1011 config RING_BUFFER_STARTUP_TEST
1012 bool "Ring buffer startup self test"
1013 depends on RING_BUFFER
1015 Run a simple self test on the ring buffer on boot up. Late in the
1016 kernel boot sequence, the test will start that kicks off
1017 a thread per cpu. Each thread will write various size events
1018 into the ring buffer. Another thread is created to send IPIs
1019 to each of the threads, where the IPI handler will also write
1020 to the ring buffer, to test/stress the nesting ability.
1021 If any anomalies are discovered, a warning will be displayed
1022 and all ring buffers will be disabled.
1024 The test runs for 10 seconds. This will slow your boot time
1025 by at least 10 more seconds.
1027 At the end of the test, statistics and more checks are done.
1028 It will output the stats of each per cpu buffer: What
1029 was written, the sizes, what was read, what was lost, and
1030 other similar details.
1034 config RING_BUFFER_VALIDATE_TIME_DELTAS
1035 bool "Verify ring buffer time stamp deltas"
1036 depends on RING_BUFFER
1038 This will audit the time stamps on the ring buffer sub
1039 buffer to make sure that all the time deltas for the
1040 events on a sub buffer matches the current time stamp.
1041 This audit is performed for every event that is not
1042 interrupted, or interrupting another event. A check
1043 is also made when traversing sub buffers to make sure
1044 that all the deltas on the previous sub buffer do not
1045 add up to be greater than the current time stamp.
1047 NOTE: This adds significant overhead to recording of events,
1048 and should only be used to test the logic of the ring buffer.
1049 Do not use it on production systems.
1051 Only say Y if you understand what this does, and you
1052 still want it enabled. Otherwise say N
1054 config MMIOTRACE_TEST
1055 tristate "Test module for mmiotrace"
1056 depends on MMIOTRACE && m
1058 This is a dumb module for testing mmiotrace. It is very dangerous
1059 as it will write garbage to IO memory starting at a given address.
1060 However, it should be safe to use on e.g. unused portion of VRAM.
1062 Say N, unless you absolutely know what you are doing.
1064 config PREEMPTIRQ_DELAY_TEST
1065 tristate "Test module to create a preempt / IRQ disable delay thread to test latency tracers"
1068 Select this option to build a test module that can help test latency
1069 tracers by executing a preempt or irq disable section with a user
1070 configurable delay. The module busy waits for the duration of the
1073 For example, the following invocation generates a burst of three
1074 irq-disabled critical sections for 500us:
1075 modprobe preemptirq_delay_test test_mode=irq delay=500 burst_size=3
1077 What's more, if you want to attach the test on the cpu which the latency
1078 tracer is running on, specify cpu_affinity=cpu_num at the end of the
1083 config SYNTH_EVENT_GEN_TEST
1084 tristate "Test module for in-kernel synthetic event generation"
1085 depends on SYNTH_EVENTS
1087 This option creates a test module to check the base
1088 functionality of in-kernel synthetic event definition and
1091 To test, insert the module, and then check the trace buffer
1092 for the generated sample events.
1096 config KPROBE_EVENT_GEN_TEST
1097 tristate "Test module for in-kernel kprobe event generation"
1098 depends on KPROBE_EVENTS
1100 This option creates a test module to check the base
1101 functionality of in-kernel kprobe event definition.
1103 To test, insert the module, and then check the trace buffer
1104 for the generated kprobe events.
1108 config HIST_TRIGGERS_DEBUG
1109 bool "Hist trigger debug support"
1110 depends on HIST_TRIGGERS
1112 Add "hist_debug" file for each event, which when read will
1113 dump out a bunch of internal details about the hist triggers
1114 defined on that event.
1116 The hist_debug file serves a couple of purposes:
1118 - Helps developers verify that nothing is broken.
1120 - Provides educational information to support the details
1121 of the hist trigger internals as described by
1122 Documentation/trace/histogram-design.rst.
1124 The hist_debug output only covers the data structures
1125 related to the histogram definitions themselves and doesn't
1126 display the internals of map buckets or variable values of
1131 source "kernel/trace/rv/Kconfig"