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_ARGS
48 If this is set, then arguments and stack can be found from
49 the ftrace_regs passed into the function callback regs parameter
50 by default, even without setting the REGS flag in the ftrace_ops.
51 This allows for use of ftrace_regs_get_argument() and
52 ftrace_regs_get_stack_pointer().
54 config HAVE_DYNAMIC_FTRACE_NO_PATCHABLE
57 If the architecture generates __patchable_function_entries sections
58 but does not want them included in the ftrace locations.
60 config HAVE_FTRACE_MCOUNT_RECORD
63 See Documentation/trace/ftrace-design.rst
65 config HAVE_SYSCALL_TRACEPOINTS
68 See Documentation/trace/ftrace-design.rst
73 Arch supports the gcc options -pg with -mfentry
75 config HAVE_NOP_MCOUNT
78 Arch supports the gcc options -pg with -mrecord-mcount and -nop-mcount
80 config HAVE_OBJTOOL_MCOUNT
83 Arch supports objtool --mcount
85 config HAVE_OBJTOOL_NOP_MCOUNT
88 Arch supports the objtool options --mcount with --mnop.
89 An architecture can select this if it wants to enable nop'ing
92 config HAVE_C_RECORDMCOUNT
95 C version of recordmcount available?
97 config HAVE_BUILDTIME_MCOUNT_SORT
100 An architecture selects this if it sorts the mcount_loc section
103 config BUILDTIME_MCOUNT_SORT
106 depends on HAVE_BUILDTIME_MCOUNT_SORT && DYNAMIC_FTRACE
108 Sort the mcount_loc section at build time.
110 config TRACER_MAX_TRACE
122 select CONTEXT_SWITCH_TRACER
126 config CONTEXT_SWITCH_TRACER
129 config RING_BUFFER_ALLOW_SWAP
132 Allow the use of ring_buffer_swap_cpu.
133 Adds a very slight overhead to tracing when enabled.
135 config PREEMPTIRQ_TRACEPOINTS
137 depends on TRACE_PREEMPT_TOGGLE || TRACE_IRQFLAGS
141 Create preempt/irq toggle tracepoints if needed, so that other parts
142 of the kernel can use them to generate or add hooks to them.
144 # All tracer options should select GENERIC_TRACER. For those options that are
145 # enabled by all tracers (context switch and event tracer) they select TRACING.
146 # This allows those options to appear when no other tracer is selected. But the
147 # options do not appear when something else selects it. We need the two options
148 # GENERIC_TRACER and TRACING to avoid circular dependencies to accomplish the
149 # hiding of the automatic options.
154 select STACKTRACE if STACKTRACE_SUPPORT
160 select TASKS_RCU if PREEMPTION
162 config GENERIC_TRACER
167 # Minimum requirements an architecture has to meet for us to
168 # be able to offer generic tracing facilities:
170 config TRACING_SUPPORT
172 depends on TRACE_IRQFLAGS_SUPPORT
173 depends on STACKTRACE_SUPPORT
178 depends on TRACING_SUPPORT
179 default y if DEBUG_KERNEL
181 Enable the kernel tracing infrastructure.
185 config BOOTTIME_TRACING
186 bool "Boot-time Tracing support"
190 Enable developer to setup ftrace subsystem via supplemental
191 kernel cmdline at boot time for debugging (tracing) driver
192 initialization and boot process.
194 config FUNCTION_TRACER
195 bool "Kernel Function Tracer"
196 depends on HAVE_FUNCTION_TRACER
198 select GENERIC_TRACER
199 select CONTEXT_SWITCH_TRACER
201 select TASKS_RCU if PREEMPTION
202 select TASKS_RUDE_RCU
204 Enable the kernel to trace every kernel function. This is done
205 by using a compiler feature to insert a small, 5-byte No-Operation
206 instruction at the beginning of every kernel function, which NOP
207 sequence is then dynamically patched into a tracer call when
208 tracing is enabled by the administrator. If it's runtime disabled
209 (the bootup default), then the overhead of the instructions is very
210 small and not measurable even in micro-benchmarks (at least on
211 x86, but may have impact on other architectures).
213 config FUNCTION_GRAPH_TRACER
214 bool "Kernel Function Graph Tracer"
215 depends on HAVE_FUNCTION_GRAPH_TRACER
216 depends on FUNCTION_TRACER
217 depends on !X86_32 || !CC_OPTIMIZE_FOR_SIZE
220 Enable the kernel to trace a function at both its return
222 Its first purpose is to trace the duration of functions and
223 draw a call graph for each thread with some information like
224 the return value. This is done by setting the current return
225 address on the current task structure into a stack of calls.
227 config DYNAMIC_FTRACE
228 bool "enable/disable function tracing dynamically"
229 depends on FUNCTION_TRACER
230 depends on HAVE_DYNAMIC_FTRACE
233 This option will modify all the calls to function tracing
234 dynamically (will patch them out of the binary image and
235 replace them with a No-Op instruction) on boot up. During
236 compile time, a table is made of all the locations that ftrace
237 can function trace, and this table is linked into the kernel
238 image. When this is enabled, functions can be individually
239 enabled, and the functions not enabled will not affect
240 performance of the system.
242 See the files in /sys/kernel/debug/tracing:
243 available_filter_functions
247 This way a CONFIG_FUNCTION_TRACER kernel is slightly larger, but
248 otherwise has native performance as long as no tracing is active.
250 config DYNAMIC_FTRACE_WITH_REGS
252 depends on DYNAMIC_FTRACE
253 depends on HAVE_DYNAMIC_FTRACE_WITH_REGS
255 config DYNAMIC_FTRACE_WITH_DIRECT_CALLS
257 depends on DYNAMIC_FTRACE_WITH_REGS
258 depends on HAVE_DYNAMIC_FTRACE_WITH_DIRECT_CALLS
260 config DYNAMIC_FTRACE_WITH_ARGS
262 depends on DYNAMIC_FTRACE
263 depends on HAVE_DYNAMIC_FTRACE_WITH_ARGS
266 bool "Kernel Function Probe (fprobe)"
267 depends on FUNCTION_TRACER
268 depends on DYNAMIC_FTRACE_WITH_REGS
269 depends on HAVE_RETHOOK
273 This option enables kernel function probe (fprobe) based on ftrace.
274 The fprobe is similar to kprobes, but probes only for kernel function
275 entries and exits. This also can probe multiple functions by one
280 config FUNCTION_PROFILER
281 bool "Kernel function profiler"
282 depends on FUNCTION_TRACER
285 This option enables the kernel function profiler. A file is created
286 in debugfs called function_profile_enabled which defaults to zero.
287 When a 1 is echoed into this file profiling begins, and when a
288 zero is entered, profiling stops. A "functions" file is created in
289 the trace_stat directory; this file shows the list of functions that
290 have been hit and their counters.
295 bool "Trace max stack"
296 depends on HAVE_FUNCTION_TRACER
297 select FUNCTION_TRACER
301 This special tracer records the maximum stack footprint of the
302 kernel and displays it in /sys/kernel/debug/tracing/stack_trace.
304 This tracer works by hooking into every function call that the
305 kernel executes, and keeping a maximum stack depth value and
306 stack-trace saved. If this is configured with DYNAMIC_FTRACE
307 then it will not have any overhead while the stack tracer
310 To enable the stack tracer on bootup, pass in 'stacktrace'
311 on the kernel command line.
313 The stack tracer can also be enabled or disabled via the
314 sysctl kernel.stack_tracer_enabled
318 config TRACE_PREEMPT_TOGGLE
321 Enables hooks which will be called when preemption is first disabled,
324 config IRQSOFF_TRACER
325 bool "Interrupts-off Latency Tracer"
327 depends on TRACE_IRQFLAGS_SUPPORT
328 select TRACE_IRQFLAGS
329 select GENERIC_TRACER
330 select TRACER_MAX_TRACE
331 select RING_BUFFER_ALLOW_SWAP
332 select TRACER_SNAPSHOT
333 select TRACER_SNAPSHOT_PER_CPU_SWAP
335 This option measures the time spent in irqs-off critical
336 sections, with microsecond accuracy.
338 The default measurement method is a maximum search, which is
339 disabled by default and can be runtime (re-)started
342 echo 0 > /sys/kernel/debug/tracing/tracing_max_latency
344 (Note that kernel size and overhead increase with this option
345 enabled. This option and the preempt-off timing option can be
346 used together or separately.)
348 config PREEMPT_TRACER
349 bool "Preemption-off Latency Tracer"
351 depends on PREEMPTION
352 select GENERIC_TRACER
353 select TRACER_MAX_TRACE
354 select RING_BUFFER_ALLOW_SWAP
355 select TRACER_SNAPSHOT
356 select TRACER_SNAPSHOT_PER_CPU_SWAP
357 select TRACE_PREEMPT_TOGGLE
359 This option measures the time spent in preemption-off critical
360 sections, with microsecond accuracy.
362 The default measurement method is a maximum search, which is
363 disabled by default and can be runtime (re-)started
366 echo 0 > /sys/kernel/debug/tracing/tracing_max_latency
368 (Note that kernel size and overhead increase with this option
369 enabled. This option and the irqs-off timing option can be
370 used together or separately.)
373 bool "Scheduling Latency Tracer"
374 select GENERIC_TRACER
375 select CONTEXT_SWITCH_TRACER
376 select TRACER_MAX_TRACE
377 select TRACER_SNAPSHOT
379 This tracer tracks the latency of the highest priority task
380 to be scheduled in, starting from the point it has woken up.
383 bool "Tracer to detect hardware latencies (like SMIs)"
384 select GENERIC_TRACER
385 select TRACER_MAX_TRACE
387 This tracer, when enabled will create one or more kernel threads,
388 depending on what the cpumask file is set to, which each thread
389 spinning in a loop looking for interruptions caused by
390 something other than the kernel. For example, if a
391 System Management Interrupt (SMI) takes a noticeable amount of
392 time, this tracer will detect it. This is useful for testing
393 if a system is reliable for Real Time tasks.
395 Some files are created in the tracing directory when this
398 hwlat_detector/width - time in usecs for how long to spin for
399 hwlat_detector/window - time in usecs between the start of each
402 A kernel thread is created that will spin with interrupts disabled
403 for "width" microseconds in every "window" cycle. It will not spin
404 for "window - width" microseconds, where the system can
407 The output will appear in the trace and trace_pipe files.
409 When the tracer is not running, it has no affect on the system,
410 but when it is running, it can cause the system to be
411 periodically non responsive. Do not run this tracer on a
414 To enable this tracer, echo in "hwlat" into the current_tracer
415 file. Every time a latency is greater than tracing_thresh, it will
416 be recorded into the ring buffer.
418 config OSNOISE_TRACER
419 bool "OS Noise tracer"
420 select GENERIC_TRACER
421 select TRACER_MAX_TRACE
423 In the context of high-performance computing (HPC), the Operating
424 System Noise (osnoise) refers to the interference experienced by an
425 application due to activities inside the operating system. In the
426 context of Linux, NMIs, IRQs, SoftIRQs, and any other system thread
427 can cause noise to the system. Moreover, hardware-related jobs can
428 also cause noise, for example, via SMIs.
430 The osnoise tracer leverages the hwlat_detector by running a similar
431 loop with preemption, SoftIRQs and IRQs enabled, thus allowing all
432 the sources of osnoise during its execution. The osnoise tracer takes
433 note of the entry and exit point of any source of interferences,
434 increasing a per-cpu interference counter. It saves an interference
435 counter for each source of interference. The interference counter for
436 NMI, IRQs, SoftIRQs, and threads is increased anytime the tool
437 observes these interferences' entry events. When a noise happens
438 without any interference from the operating system level, the
439 hardware noise counter increases, pointing to a hardware-related
440 noise. In this way, osnoise can account for any source of
441 interference. At the end of the period, the osnoise tracer prints
442 the sum of all noise, the max single noise, the percentage of CPU
443 available for the thread, and the counters for the noise sources.
445 In addition to the tracer, a set of tracepoints were added to
446 facilitate the identification of the osnoise source.
448 The output will appear in the trace and trace_pipe files.
450 To enable this tracer, echo in "osnoise" into the current_tracer
453 config TIMERLAT_TRACER
454 bool "Timerlat tracer"
455 select OSNOISE_TRACER
456 select GENERIC_TRACER
458 The timerlat tracer aims to help the preemptive kernel developers
459 to find sources of wakeup latencies of real-time threads.
461 The tracer creates a per-cpu kernel thread with real-time priority.
462 The tracer thread sets a periodic timer to wakeup itself, and goes
463 to sleep waiting for the timer to fire. At the wakeup, the thread
464 then computes a wakeup latency value as the difference between
465 the current time and the absolute time that the timer was set
468 The tracer prints two lines at every activation. The first is the
469 timer latency observed at the hardirq context before the
470 activation of the thread. The second is the timer latency observed
471 by the thread, which is the same level that cyclictest reports. The
472 ACTIVATION ID field serves to relate the irq execution to its
473 respective thread execution.
475 The tracer is build on top of osnoise tracer, and the osnoise:
476 events can be used to trace the source of interference from NMI,
477 IRQs and other threads. It also enables the capture of the
478 stacktrace at the IRQ context, which helps to identify the code
479 path that can cause thread delay.
482 bool "Memory mapped IO tracing"
483 depends on HAVE_MMIOTRACE_SUPPORT && PCI
484 select GENERIC_TRACER
486 Mmiotrace traces Memory Mapped I/O access and is meant for
487 debugging and reverse engineering. It is called from the ioremap
488 implementation and works via page faults. Tracing is disabled by
489 default and can be enabled at run-time.
491 See Documentation/trace/mmiotrace.rst.
492 If you are not helping to develop drivers, say N.
494 config ENABLE_DEFAULT_TRACERS
495 bool "Trace process context switches and events"
496 depends on !GENERIC_TRACER
499 This tracer hooks to various trace points in the kernel,
500 allowing the user to pick and choose which trace point they
501 want to trace. It also includes the sched_switch tracer plugin.
503 config FTRACE_SYSCALLS
504 bool "Trace syscalls"
505 depends on HAVE_SYSCALL_TRACEPOINTS
506 select GENERIC_TRACER
509 Basic tracer to catch the syscall entry and exit events.
511 config TRACER_SNAPSHOT
512 bool "Create a snapshot trace buffer"
513 select TRACER_MAX_TRACE
515 Allow tracing users to take snapshot of the current buffer using the
516 ftrace interface, e.g.:
518 echo 1 > /sys/kernel/debug/tracing/snapshot
521 config TRACER_SNAPSHOT_PER_CPU_SWAP
522 bool "Allow snapshot to swap per CPU"
523 depends on TRACER_SNAPSHOT
524 select RING_BUFFER_ALLOW_SWAP
526 Allow doing a snapshot of a single CPU buffer instead of a
527 full swap (all buffers). If this is set, then the following is
530 echo 1 > /sys/kernel/debug/tracing/per_cpu/cpu2/snapshot
532 After which, only the tracing buffer for CPU 2 was swapped with
533 the main tracing buffer, and the other CPU buffers remain the same.
535 When this is enabled, this adds a little more overhead to the
536 trace recording, as it needs to add some checks to synchronize
537 recording with swaps. But this does not affect the performance
538 of the overall system. This is enabled by default when the preempt
539 or irq latency tracers are enabled, as those need to swap as well
540 and already adds the overhead (plus a lot more).
542 config TRACE_BRANCH_PROFILING
544 select GENERIC_TRACER
547 prompt "Branch Profiling"
548 default BRANCH_PROFILE_NONE
550 The branch profiling is a software profiler. It will add hooks
551 into the C conditionals to test which path a branch takes.
553 The likely/unlikely profiler only looks at the conditions that
554 are annotated with a likely or unlikely macro.
556 The "all branch" profiler will profile every if-statement in the
557 kernel. This profiler will also enable the likely/unlikely
560 Either of the above profilers adds a bit of overhead to the system.
561 If unsure, choose "No branch profiling".
563 config BRANCH_PROFILE_NONE
564 bool "No branch profiling"
566 No branch profiling. Branch profiling adds a bit of overhead.
567 Only enable it if you want to analyse the branching behavior.
568 Otherwise keep it disabled.
570 config PROFILE_ANNOTATED_BRANCHES
571 bool "Trace likely/unlikely profiler"
572 select TRACE_BRANCH_PROFILING
574 This tracer profiles all likely and unlikely macros
575 in the kernel. It will display the results in:
577 /sys/kernel/debug/tracing/trace_stat/branch_annotated
579 Note: this will add a significant overhead; only turn this
580 on if you need to profile the system's use of these macros.
582 config PROFILE_ALL_BRANCHES
583 bool "Profile all if conditionals" if !FORTIFY_SOURCE
584 select TRACE_BRANCH_PROFILING
586 This tracer profiles all branch conditions. Every if ()
587 taken in the kernel is recorded whether it hit or miss.
588 The results will be displayed in:
590 /sys/kernel/debug/tracing/trace_stat/branch_all
592 This option also enables the likely/unlikely profiler.
594 This configuration, when enabled, will impose a great overhead
595 on the system. This should only be enabled when the system
596 is to be analyzed in much detail.
599 config TRACING_BRANCHES
602 Selected by tracers that will trace the likely and unlikely
603 conditions. This prevents the tracers themselves from being
604 profiled. Profiling the tracing infrastructure can only happen
605 when the likelys and unlikelys are not being traced.
608 bool "Trace likely/unlikely instances"
609 depends on TRACE_BRANCH_PROFILING
610 select TRACING_BRANCHES
612 This traces the events of likely and unlikely condition
613 calls in the kernel. The difference between this and the
614 "Trace likely/unlikely profiler" is that this is not a
615 histogram of the callers, but actually places the calling
616 events into a running trace buffer to see when and where the
617 events happened, as well as their results.
621 config BLK_DEV_IO_TRACE
622 bool "Support for tracing block IO actions"
628 select GENERIC_TRACER
631 Say Y here if you want to be able to trace the block layer actions
632 on a given queue. Tracing allows you to see any traffic happening
633 on a block device queue. For more information (and the userspace
634 support tools needed), fetch the blktrace tools from:
636 git://git.kernel.dk/blktrace.git
638 Tracing also is possible using the ftrace interface, e.g.:
640 echo 1 > /sys/block/sda/sda1/trace/enable
641 echo blk > /sys/kernel/debug/tracing/current_tracer
642 cat /sys/kernel/debug/tracing/trace_pipe
648 depends on HAVE_REGS_AND_STACK_ACCESS_API
649 bool "Enable kprobes-based dynamic events"
652 select DYNAMIC_EVENTS
655 This allows the user to add tracing events (similar to tracepoints)
656 on the fly via the ftrace interface. See
657 Documentation/trace/kprobetrace.rst for more details.
659 Those events can be inserted wherever kprobes can probe, and record
660 various register and memory values.
662 This option is also required by perf-probe subcommand of perf tools.
663 If you want to use perf tools, this option is strongly recommended.
665 config KPROBE_EVENTS_ON_NOTRACE
666 bool "Do NOT protect notrace function from kprobe events"
667 depends on KPROBE_EVENTS
668 depends on DYNAMIC_FTRACE
671 This is only for the developers who want to debug ftrace itself
674 If kprobes can use ftrace instead of breakpoint, ftrace related
675 functions are protected from kprobe-events to prevent an infinite
676 recursion or any unexpected execution path which leads to a kernel
679 This option disables such protection and allows you to put kprobe
680 events on ftrace functions for debugging ftrace by itself.
681 Note that this might let you shoot yourself in the foot.
686 bool "Enable uprobes-based dynamic events"
687 depends on ARCH_SUPPORTS_UPROBES
689 depends on PERF_EVENTS
692 select DYNAMIC_EVENTS
696 This allows the user to add tracing events on top of userspace
697 dynamic events (similar to tracepoints) on the fly via the trace
698 events interface. Those events can be inserted wherever uprobes
699 can probe, and record various registers.
700 This option is required if you plan to use perf-probe subcommand
701 of perf tools on user space applications.
704 depends on BPF_SYSCALL
705 depends on (KPROBE_EVENTS || UPROBE_EVENTS) && PERF_EVENTS
709 This allows the user to attach BPF programs to kprobe, uprobe, and
712 config DYNAMIC_EVENTS
718 config BPF_KPROBE_OVERRIDE
719 bool "Enable BPF programs to override a kprobed function"
720 depends on BPF_EVENTS
721 depends on FUNCTION_ERROR_INJECTION
724 Allows BPF to override the execution of a probed function and
725 set a different return value. This is used for error injection.
727 config FTRACE_MCOUNT_RECORD
729 depends on DYNAMIC_FTRACE
730 depends on HAVE_FTRACE_MCOUNT_RECORD
732 config FTRACE_MCOUNT_USE_PATCHABLE_FUNCTION_ENTRY
734 depends on FTRACE_MCOUNT_RECORD
736 config FTRACE_MCOUNT_USE_CC
738 depends on $(cc-option,-mrecord-mcount)
739 depends on !FTRACE_MCOUNT_USE_PATCHABLE_FUNCTION_ENTRY
740 depends on FTRACE_MCOUNT_RECORD
742 config FTRACE_MCOUNT_USE_OBJTOOL
744 depends on HAVE_OBJTOOL_MCOUNT
745 depends on !FTRACE_MCOUNT_USE_PATCHABLE_FUNCTION_ENTRY
746 depends on !FTRACE_MCOUNT_USE_CC
747 depends on FTRACE_MCOUNT_RECORD
750 config FTRACE_MCOUNT_USE_RECORDMCOUNT
752 depends on !FTRACE_MCOUNT_USE_PATCHABLE_FUNCTION_ENTRY
753 depends on !FTRACE_MCOUNT_USE_CC
754 depends on !FTRACE_MCOUNT_USE_OBJTOOL
755 depends on FTRACE_MCOUNT_RECORD
759 depends on ARCH_HAVE_NMI_SAFE_CMPXCHG
761 tracing_map is a special-purpose lock-free map for tracing,
762 separated out as a stand-alone facility in order to allow it
763 to be shared between multiple tracers. It isn't meant to be
764 generally used outside of that context, and is normally
765 selected by tracers that use it.
768 bool "Synthetic trace events"
770 select DYNAMIC_EVENTS
773 Synthetic events are user-defined trace events that can be
774 used to combine data from other trace events or in fact any
775 data source. Synthetic events can be generated indirectly
776 via the trace() action of histogram triggers or directly
777 by way of an in-kernel API.
779 See Documentation/trace/events.rst or
780 Documentation/trace/histogram.rst for details and examples.
785 bool "User trace events"
787 select DYNAMIC_EVENTS
788 depends on BROKEN || COMPILE_TEST # API needs to be straighten out
790 User trace events are user-defined trace events that
791 can be used like an existing kernel trace event. User trace
792 events are generated by writing to a tracefs file. User
793 processes can determine if their tracing events should be
794 generated by memory mapping a tracefs file and checking for
795 an associated byte being non-zero.
800 bool "Histogram triggers"
801 depends on ARCH_HAVE_NMI_SAFE_CMPXCHG
804 select DYNAMIC_EVENTS
808 Hist triggers allow one or more arbitrary trace event fields
809 to be aggregated into hash tables and dumped to stdout by
810 reading a debugfs/tracefs file. They're useful for
811 gathering quick and dirty (though precise) summaries of
812 event activity as an initial guide for further investigation
813 using more advanced tools.
815 Inter-event tracing of quantities such as latencies is also
816 supported using hist triggers under this option.
818 See Documentation/trace/histogram.rst.
821 config TRACE_EVENT_INJECT
822 bool "Trace event injection"
825 Allow user-space to inject a specific trace event into the ring
826 buffer. This is mainly used for testing purpose.
830 config TRACEPOINT_BENCHMARK
831 bool "Add tracepoint that benchmarks tracepoints"
833 This option creates the tracepoint "benchmark:benchmark_event".
834 When the tracepoint is enabled, it kicks off a kernel thread that
835 goes into an infinite loop (calling cond_resched() to let other tasks
836 run), and calls the tracepoint. Each iteration will record the time
837 it took to write to the tracepoint and the next iteration that
838 data will be passed to the tracepoint itself. That is, the tracepoint
839 will report the time it took to do the previous tracepoint.
840 The string written to the tracepoint is a static string of 128 bytes
841 to keep the time the same. The initial string is simply a write of
842 "START". The second string records the cold cache time of the first
843 write which is not added to the rest of the calculations.
845 As it is a tight loop, it benchmarks as hot cache. That's fine because
846 we care most about hot paths that are probably in cache already.
848 An example of the output:
851 first=3672 [COLD CACHED]
852 last=632 first=3672 max=632 min=632 avg=316 std=446 std^2=199712
853 last=278 first=3672 max=632 min=278 avg=303 std=316 std^2=100337
854 last=277 first=3672 max=632 min=277 avg=296 std=258 std^2=67064
855 last=273 first=3672 max=632 min=273 avg=292 std=224 std^2=50411
856 last=273 first=3672 max=632 min=273 avg=288 std=200 std^2=40389
857 last=281 first=3672 max=632 min=273 avg=287 std=183 std^2=33666
860 config RING_BUFFER_BENCHMARK
861 tristate "Ring buffer benchmark stress tester"
862 depends on RING_BUFFER
864 This option creates a test to stress the ring buffer and benchmark it.
865 It creates its own ring buffer such that it will not interfere with
866 any other users of the ring buffer (such as ftrace). It then creates
867 a producer and consumer that will run for 10 seconds and sleep for
868 10 seconds. Each interval it will print out the number of events
869 it recorded and give a rough estimate of how long each iteration took.
871 It does not disable interrupts or raise its priority, so it may be
872 affected by processes that are running.
876 config TRACE_EVAL_MAP_FILE
877 bool "Show eval mappings for trace events"
880 The "print fmt" of the trace events will show the enum/sizeof names
881 instead of their values. This can cause problems for user space tools
882 that use this string to parse the raw data as user space does not know
883 how to convert the string to its value.
885 To fix this, there's a special macro in the kernel that can be used
886 to convert an enum/sizeof into its value. If this macro is used, then
887 the print fmt strings will be converted to their values.
889 If something does not get converted properly, this option can be
890 used to show what enums/sizeof the kernel tried to convert.
892 This option is for debugging the conversions. A file is created
893 in the tracing directory called "eval_map" that will show the
894 names matched with their values and what trace event system they
897 Normally, the mapping of the strings to values will be freed after
898 boot up or module load. With this option, they will not be freed, as
899 they are needed for the "eval_map" file. Enabling this option will
900 increase the memory footprint of the running kernel.
904 config FTRACE_RECORD_RECURSION
905 bool "Record functions that recurse in function tracing"
906 depends on FUNCTION_TRACER
908 All callbacks that attach to the function tracing have some sort
909 of protection against recursion. Even though the protection exists,
910 it adds overhead. This option will create a file in the tracefs
911 file system called "recursed_functions" that will list the functions
912 that triggered a recursion.
914 This will add more overhead to cases that have recursion.
918 config FTRACE_RECORD_RECURSION_SIZE
919 int "Max number of recursed functions to record"
921 depends on FTRACE_RECORD_RECURSION
923 This defines the limit of number of functions that can be
924 listed in the "recursed_functions" file, that lists all
925 the functions that caused a recursion to happen.
926 This file can be reset, but the limit can not change in
929 config RING_BUFFER_RECORD_RECURSION
930 bool "Record functions that recurse in the ring buffer"
931 depends on FTRACE_RECORD_RECURSION
932 # default y, because it is coupled with FTRACE_RECORD_RECURSION
935 The ring buffer has its own internal recursion. Although when
936 recursion happens it wont cause harm because of the protection,
937 but it does cause an unwanted overhead. Enabling this option will
938 place where recursion was detected into the ftrace "recursed_functions"
941 This will add more overhead to cases that have recursion.
943 config GCOV_PROFILE_FTRACE
944 bool "Enable GCOV profiling on ftrace subsystem"
945 depends on GCOV_KERNEL
947 Enable GCOV profiling on ftrace subsystem for checking
948 which functions/lines are tested.
952 Note that on a kernel compiled with this config, ftrace will
953 run significantly slower.
955 config FTRACE_SELFTEST
958 config FTRACE_STARTUP_TEST
959 bool "Perform a startup test on ftrace"
960 depends on GENERIC_TRACER
961 select FTRACE_SELFTEST
963 This option performs a series of startup tests on ftrace. On bootup
964 a series of tests are made to verify that the tracer is
965 functioning properly. It will do tests on all the configured
968 config EVENT_TRACE_STARTUP_TEST
969 bool "Run selftest on trace events"
970 depends on FTRACE_STARTUP_TEST
973 This option performs a test on all trace events in the system.
974 It basically just enables each event and runs some code that
975 will trigger events (not necessarily the event it enables)
976 This may take some time run as there are a lot of events.
978 config EVENT_TRACE_TEST_SYSCALLS
979 bool "Run selftest on syscall events"
980 depends on EVENT_TRACE_STARTUP_TEST
982 This option will also enable testing every syscall event.
983 It only enables the event and disables it and runs various loads
984 with the event enabled. This adds a bit more time for kernel boot
985 up since it runs this on every system call defined.
987 TBD - enable a way to actually call the syscalls as we test their
990 config FTRACE_SORT_STARTUP_TEST
991 bool "Verify compile time sorting of ftrace functions"
992 depends on DYNAMIC_FTRACE
993 depends on BUILDTIME_MCOUNT_SORT
995 Sorting of the mcount_loc sections that is used to find the
996 where the ftrace knows where to patch functions for tracing
997 and other callbacks is done at compile time. But if the sort
998 is not done correctly, it will cause non-deterministic failures.
999 When this is set, the sorted sections will be verified that they
1000 are in deed sorted and will warn if they are not.
1004 config RING_BUFFER_STARTUP_TEST
1005 bool "Ring buffer startup self test"
1006 depends on RING_BUFFER
1008 Run a simple self test on the ring buffer on boot up. Late in the
1009 kernel boot sequence, the test will start that kicks off
1010 a thread per cpu. Each thread will write various size events
1011 into the ring buffer. Another thread is created to send IPIs
1012 to each of the threads, where the IPI handler will also write
1013 to the ring buffer, to test/stress the nesting ability.
1014 If any anomalies are discovered, a warning will be displayed
1015 and all ring buffers will be disabled.
1017 The test runs for 10 seconds. This will slow your boot time
1018 by at least 10 more seconds.
1020 At the end of the test, statics and more checks are done.
1021 It will output the stats of each per cpu buffer. What
1022 was written, the sizes, what was read, what was lost, and
1023 other similar details.
1027 config RING_BUFFER_VALIDATE_TIME_DELTAS
1028 bool "Verify ring buffer time stamp deltas"
1029 depends on RING_BUFFER
1031 This will audit the time stamps on the ring buffer sub
1032 buffer to make sure that all the time deltas for the
1033 events on a sub buffer matches the current time stamp.
1034 This audit is performed for every event that is not
1035 interrupted, or interrupting another event. A check
1036 is also made when traversing sub buffers to make sure
1037 that all the deltas on the previous sub buffer do not
1038 add up to be greater than the current time stamp.
1040 NOTE: This adds significant overhead to recording of events,
1041 and should only be used to test the logic of the ring buffer.
1042 Do not use it on production systems.
1044 Only say Y if you understand what this does, and you
1045 still want it enabled. Otherwise say N
1047 config MMIOTRACE_TEST
1048 tristate "Test module for mmiotrace"
1049 depends on MMIOTRACE && m
1051 This is a dumb module for testing mmiotrace. It is very dangerous
1052 as it will write garbage to IO memory starting at a given address.
1053 However, it should be safe to use on e.g. unused portion of VRAM.
1055 Say N, unless you absolutely know what you are doing.
1057 config PREEMPTIRQ_DELAY_TEST
1058 tristate "Test module to create a preempt / IRQ disable delay thread to test latency tracers"
1061 Select this option to build a test module that can help test latency
1062 tracers by executing a preempt or irq disable section with a user
1063 configurable delay. The module busy waits for the duration of the
1066 For example, the following invocation generates a burst of three
1067 irq-disabled critical sections for 500us:
1068 modprobe preemptirq_delay_test test_mode=irq delay=500 burst_size=3
1070 What's more, if you want to attach the test on the cpu which the latency
1071 tracer is running on, specify cpu_affinity=cpu_num at the end of the
1076 config SYNTH_EVENT_GEN_TEST
1077 tristate "Test module for in-kernel synthetic event generation"
1078 depends on SYNTH_EVENTS
1080 This option creates a test module to check the base
1081 functionality of in-kernel synthetic event definition and
1084 To test, insert the module, and then check the trace buffer
1085 for the generated sample events.
1089 config KPROBE_EVENT_GEN_TEST
1090 tristate "Test module for in-kernel kprobe event generation"
1091 depends on KPROBE_EVENTS
1093 This option creates a test module to check the base
1094 functionality of in-kernel kprobe event definition.
1096 To test, insert the module, and then check the trace buffer
1097 for the generated kprobe events.
1101 config HIST_TRIGGERS_DEBUG
1102 bool "Hist trigger debug support"
1103 depends on HIST_TRIGGERS
1105 Add "hist_debug" file for each event, which when read will
1106 dump out a bunch of internal details about the hist triggers
1107 defined on that event.
1109 The hist_debug file serves a couple of purposes:
1111 - Helps developers verify that nothing is broken.
1113 - Provides educational information to support the details
1114 of the hist trigger internals as described by
1115 Documentation/trace/histogram-design.rst.
1117 The hist_debug output only covers the data structures
1118 related to the histogram definitions themselves and doesn't
1119 display the internals of map buckets or variable values of
1124 source "kernel/trace/rv/Kconfig"