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 pt_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 regs_get_kernel_argument() and
52 kernel_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_C_RECORDMCOUNT
88 C version of recordmcount available?
90 config HAVE_BUILDTIME_MCOUNT_SORT
93 An architecture selects this if it sorts the mcount_loc section
96 config BUILDTIME_MCOUNT_SORT
99 depends on HAVE_BUILDTIME_MCOUNT_SORT && DYNAMIC_FTRACE
101 Sort the mcount_loc section at build time.
103 config TRACER_MAX_TRACE
115 select CONTEXT_SWITCH_TRACER
119 config CONTEXT_SWITCH_TRACER
122 config RING_BUFFER_ALLOW_SWAP
125 Allow the use of ring_buffer_swap_cpu.
126 Adds a very slight overhead to tracing when enabled.
128 config PREEMPTIRQ_TRACEPOINTS
130 depends on TRACE_PREEMPT_TOGGLE || TRACE_IRQFLAGS
134 Create preempt/irq toggle tracepoints if needed, so that other parts
135 of the kernel can use them to generate or add hooks to them.
137 # All tracer options should select GENERIC_TRACER. For those options that are
138 # enabled by all tracers (context switch and event tracer) they select TRACING.
139 # This allows those options to appear when no other tracer is selected. But the
140 # options do not appear when something else selects it. We need the two options
141 # GENERIC_TRACER and TRACING to avoid circular dependencies to accomplish the
142 # hiding of the automatic options.
147 select STACKTRACE if STACKTRACE_SUPPORT
153 select TASKS_RCU if PREEMPTION
155 config GENERIC_TRACER
160 # Minimum requirements an architecture has to meet for us to
161 # be able to offer generic tracing facilities:
163 config TRACING_SUPPORT
165 depends on TRACE_IRQFLAGS_SUPPORT
166 depends on STACKTRACE_SUPPORT
171 depends on TRACING_SUPPORT
172 default y if DEBUG_KERNEL
174 Enable the kernel tracing infrastructure.
178 config BOOTTIME_TRACING
179 bool "Boot-time Tracing support"
183 Enable developer to setup ftrace subsystem via supplemental
184 kernel cmdline at boot time for debugging (tracing) driver
185 initialization and boot process.
187 config FUNCTION_TRACER
188 bool "Kernel Function Tracer"
189 depends on HAVE_FUNCTION_TRACER
191 select GENERIC_TRACER
192 select CONTEXT_SWITCH_TRACER
194 select TASKS_RCU if PREEMPTION
195 select TASKS_RUDE_RCU
197 Enable the kernel to trace every kernel function. This is done
198 by using a compiler feature to insert a small, 5-byte No-Operation
199 instruction at the beginning of every kernel function, which NOP
200 sequence is then dynamically patched into a tracer call when
201 tracing is enabled by the administrator. If it's runtime disabled
202 (the bootup default), then the overhead of the instructions is very
203 small and not measurable even in micro-benchmarks (at least on
204 x86, but may have impact on other architectures).
206 config FUNCTION_GRAPH_TRACER
207 bool "Kernel Function Graph Tracer"
208 depends on HAVE_FUNCTION_GRAPH_TRACER
209 depends on FUNCTION_TRACER
210 depends on !X86_32 || !CC_OPTIMIZE_FOR_SIZE
213 Enable the kernel to trace a function at both its return
215 Its first purpose is to trace the duration of functions and
216 draw a call graph for each thread with some information like
217 the return value. This is done by setting the current return
218 address on the current task structure into a stack of calls.
220 config DYNAMIC_FTRACE
221 bool "enable/disable function tracing dynamically"
222 depends on FUNCTION_TRACER
223 depends on HAVE_DYNAMIC_FTRACE
226 This option will modify all the calls to function tracing
227 dynamically (will patch them out of the binary image and
228 replace them with a No-Op instruction) on boot up. During
229 compile time, a table is made of all the locations that ftrace
230 can function trace, and this table is linked into the kernel
231 image. When this is enabled, functions can be individually
232 enabled, and the functions not enabled will not affect
233 performance of the system.
235 See the files in /sys/kernel/debug/tracing:
236 available_filter_functions
240 This way a CONFIG_FUNCTION_TRACER kernel is slightly larger, but
241 otherwise has native performance as long as no tracing is active.
243 config DYNAMIC_FTRACE_WITH_REGS
245 depends on DYNAMIC_FTRACE
246 depends on HAVE_DYNAMIC_FTRACE_WITH_REGS
248 config DYNAMIC_FTRACE_WITH_DIRECT_CALLS
250 depends on DYNAMIC_FTRACE_WITH_REGS
251 depends on HAVE_DYNAMIC_FTRACE_WITH_DIRECT_CALLS
253 config DYNAMIC_FTRACE_WITH_ARGS
255 depends on DYNAMIC_FTRACE
256 depends on HAVE_DYNAMIC_FTRACE_WITH_ARGS
259 bool "Kernel Function Probe (fprobe)"
260 depends on FUNCTION_TRACER
261 depends on DYNAMIC_FTRACE_WITH_REGS
262 depends on HAVE_RETHOOK
266 This option enables kernel function probe (fprobe) based on ftrace.
267 The fprobe is similar to kprobes, but probes only for kernel function
268 entries and exits. This also can probe multiple functions by one
273 config FUNCTION_PROFILER
274 bool "Kernel function profiler"
275 depends on FUNCTION_TRACER
278 This option enables the kernel function profiler. A file is created
279 in debugfs called function_profile_enabled which defaults to zero.
280 When a 1 is echoed into this file profiling begins, and when a
281 zero is entered, profiling stops. A "functions" file is created in
282 the trace_stat directory; this file shows the list of functions that
283 have been hit and their counters.
288 bool "Trace max stack"
289 depends on HAVE_FUNCTION_TRACER
290 select FUNCTION_TRACER
294 This special tracer records the maximum stack footprint of the
295 kernel and displays it in /sys/kernel/debug/tracing/stack_trace.
297 This tracer works by hooking into every function call that the
298 kernel executes, and keeping a maximum stack depth value and
299 stack-trace saved. If this is configured with DYNAMIC_FTRACE
300 then it will not have any overhead while the stack tracer
303 To enable the stack tracer on bootup, pass in 'stacktrace'
304 on the kernel command line.
306 The stack tracer can also be enabled or disabled via the
307 sysctl kernel.stack_tracer_enabled
311 config TRACE_PREEMPT_TOGGLE
314 Enables hooks which will be called when preemption is first disabled,
317 config IRQSOFF_TRACER
318 bool "Interrupts-off Latency Tracer"
320 depends on TRACE_IRQFLAGS_SUPPORT
321 select TRACE_IRQFLAGS
322 select GENERIC_TRACER
323 select TRACER_MAX_TRACE
324 select RING_BUFFER_ALLOW_SWAP
325 select TRACER_SNAPSHOT
326 select TRACER_SNAPSHOT_PER_CPU_SWAP
328 This option measures the time spent in irqs-off critical
329 sections, with microsecond accuracy.
331 The default measurement method is a maximum search, which is
332 disabled by default and can be runtime (re-)started
335 echo 0 > /sys/kernel/debug/tracing/tracing_max_latency
337 (Note that kernel size and overhead increase with this option
338 enabled. This option and the preempt-off timing option can be
339 used together or separately.)
341 config PREEMPT_TRACER
342 bool "Preemption-off Latency Tracer"
344 depends on PREEMPTION
345 select GENERIC_TRACER
346 select TRACER_MAX_TRACE
347 select RING_BUFFER_ALLOW_SWAP
348 select TRACER_SNAPSHOT
349 select TRACER_SNAPSHOT_PER_CPU_SWAP
350 select TRACE_PREEMPT_TOGGLE
352 This option measures the time spent in preemption-off critical
353 sections, with microsecond accuracy.
355 The default measurement method is a maximum search, which is
356 disabled by default and can be runtime (re-)started
359 echo 0 > /sys/kernel/debug/tracing/tracing_max_latency
361 (Note that kernel size and overhead increase with this option
362 enabled. This option and the irqs-off timing option can be
363 used together or separately.)
366 bool "Scheduling Latency Tracer"
367 select GENERIC_TRACER
368 select CONTEXT_SWITCH_TRACER
369 select TRACER_MAX_TRACE
370 select TRACER_SNAPSHOT
372 This tracer tracks the latency of the highest priority task
373 to be scheduled in, starting from the point it has woken up.
376 bool "Tracer to detect hardware latencies (like SMIs)"
377 select GENERIC_TRACER
379 This tracer, when enabled will create one or more kernel threads,
380 depending on what the cpumask file is set to, which each thread
381 spinning in a loop looking for interruptions caused by
382 something other than the kernel. For example, if a
383 System Management Interrupt (SMI) takes a noticeable amount of
384 time, this tracer will detect it. This is useful for testing
385 if a system is reliable for Real Time tasks.
387 Some files are created in the tracing directory when this
390 hwlat_detector/width - time in usecs for how long to spin for
391 hwlat_detector/window - time in usecs between the start of each
394 A kernel thread is created that will spin with interrupts disabled
395 for "width" microseconds in every "window" cycle. It will not spin
396 for "window - width" microseconds, where the system can
399 The output will appear in the trace and trace_pipe files.
401 When the tracer is not running, it has no affect on the system,
402 but when it is running, it can cause the system to be
403 periodically non responsive. Do not run this tracer on a
406 To enable this tracer, echo in "hwlat" into the current_tracer
407 file. Every time a latency is greater than tracing_thresh, it will
408 be recorded into the ring buffer.
410 config OSNOISE_TRACER
411 bool "OS Noise tracer"
412 select GENERIC_TRACER
414 In the context of high-performance computing (HPC), the Operating
415 System Noise (osnoise) refers to the interference experienced by an
416 application due to activities inside the operating system. In the
417 context of Linux, NMIs, IRQs, SoftIRQs, and any other system thread
418 can cause noise to the system. Moreover, hardware-related jobs can
419 also cause noise, for example, via SMIs.
421 The osnoise tracer leverages the hwlat_detector by running a similar
422 loop with preemption, SoftIRQs and IRQs enabled, thus allowing all
423 the sources of osnoise during its execution. The osnoise tracer takes
424 note of the entry and exit point of any source of interferences,
425 increasing a per-cpu interference counter. It saves an interference
426 counter for each source of interference. The interference counter for
427 NMI, IRQs, SoftIRQs, and threads is increased anytime the tool
428 observes these interferences' entry events. When a noise happens
429 without any interference from the operating system level, the
430 hardware noise counter increases, pointing to a hardware-related
431 noise. In this way, osnoise can account for any source of
432 interference. At the end of the period, the osnoise tracer prints
433 the sum of all noise, the max single noise, the percentage of CPU
434 available for the thread, and the counters for the noise sources.
436 In addition to the tracer, a set of tracepoints were added to
437 facilitate the identification of the osnoise source.
439 The output will appear in the trace and trace_pipe files.
441 To enable this tracer, echo in "osnoise" into the current_tracer
444 config TIMERLAT_TRACER
445 bool "Timerlat tracer"
446 select OSNOISE_TRACER
447 select GENERIC_TRACER
449 The timerlat tracer aims to help the preemptive kernel developers
450 to find sources of wakeup latencies of real-time threads.
452 The tracer creates a per-cpu kernel thread with real-time priority.
453 The tracer thread sets a periodic timer to wakeup itself, and goes
454 to sleep waiting for the timer to fire. At the wakeup, the thread
455 then computes a wakeup latency value as the difference between
456 the current time and the absolute time that the timer was set
459 The tracer prints two lines at every activation. The first is the
460 timer latency observed at the hardirq context before the
461 activation of the thread. The second is the timer latency observed
462 by the thread, which is the same level that cyclictest reports. The
463 ACTIVATION ID field serves to relate the irq execution to its
464 respective thread execution.
466 The tracer is build on top of osnoise tracer, and the osnoise:
467 events can be used to trace the source of interference from NMI,
468 IRQs and other threads. It also enables the capture of the
469 stacktrace at the IRQ context, which helps to identify the code
470 path that can cause thread delay.
473 bool "Memory mapped IO tracing"
474 depends on HAVE_MMIOTRACE_SUPPORT && PCI
475 select GENERIC_TRACER
477 Mmiotrace traces Memory Mapped I/O access and is meant for
478 debugging and reverse engineering. It is called from the ioremap
479 implementation and works via page faults. Tracing is disabled by
480 default and can be enabled at run-time.
482 See Documentation/trace/mmiotrace.rst.
483 If you are not helping to develop drivers, say N.
485 config ENABLE_DEFAULT_TRACERS
486 bool "Trace process context switches and events"
487 depends on !GENERIC_TRACER
490 This tracer hooks to various trace points in the kernel,
491 allowing the user to pick and choose which trace point they
492 want to trace. It also includes the sched_switch tracer plugin.
494 config FTRACE_SYSCALLS
495 bool "Trace syscalls"
496 depends on HAVE_SYSCALL_TRACEPOINTS
497 select GENERIC_TRACER
500 Basic tracer to catch the syscall entry and exit events.
502 config TRACER_SNAPSHOT
503 bool "Create a snapshot trace buffer"
504 select TRACER_MAX_TRACE
506 Allow tracing users to take snapshot of the current buffer using the
507 ftrace interface, e.g.:
509 echo 1 > /sys/kernel/debug/tracing/snapshot
512 config TRACER_SNAPSHOT_PER_CPU_SWAP
513 bool "Allow snapshot to swap per CPU"
514 depends on TRACER_SNAPSHOT
515 select RING_BUFFER_ALLOW_SWAP
517 Allow doing a snapshot of a single CPU buffer instead of a
518 full swap (all buffers). If this is set, then the following is
521 echo 1 > /sys/kernel/debug/tracing/per_cpu/cpu2/snapshot
523 After which, only the tracing buffer for CPU 2 was swapped with
524 the main tracing buffer, and the other CPU buffers remain the same.
526 When this is enabled, this adds a little more overhead to the
527 trace recording, as it needs to add some checks to synchronize
528 recording with swaps. But this does not affect the performance
529 of the overall system. This is enabled by default when the preempt
530 or irq latency tracers are enabled, as those need to swap as well
531 and already adds the overhead (plus a lot more).
533 config TRACE_BRANCH_PROFILING
535 select GENERIC_TRACER
538 prompt "Branch Profiling"
539 default BRANCH_PROFILE_NONE
541 The branch profiling is a software profiler. It will add hooks
542 into the C conditionals to test which path a branch takes.
544 The likely/unlikely profiler only looks at the conditions that
545 are annotated with a likely or unlikely macro.
547 The "all branch" profiler will profile every if-statement in the
548 kernel. This profiler will also enable the likely/unlikely
551 Either of the above profilers adds a bit of overhead to the system.
552 If unsure, choose "No branch profiling".
554 config BRANCH_PROFILE_NONE
555 bool "No branch profiling"
557 No branch profiling. Branch profiling adds a bit of overhead.
558 Only enable it if you want to analyse the branching behavior.
559 Otherwise keep it disabled.
561 config PROFILE_ANNOTATED_BRANCHES
562 bool "Trace likely/unlikely profiler"
563 select TRACE_BRANCH_PROFILING
565 This tracer profiles all likely and unlikely macros
566 in the kernel. It will display the results in:
568 /sys/kernel/debug/tracing/trace_stat/branch_annotated
570 Note: this will add a significant overhead; only turn this
571 on if you need to profile the system's use of these macros.
573 config PROFILE_ALL_BRANCHES
574 bool "Profile all if conditionals" if !FORTIFY_SOURCE
575 select TRACE_BRANCH_PROFILING
577 This tracer profiles all branch conditions. Every if ()
578 taken in the kernel is recorded whether it hit or miss.
579 The results will be displayed in:
581 /sys/kernel/debug/tracing/trace_stat/branch_all
583 This option also enables the likely/unlikely profiler.
585 This configuration, when enabled, will impose a great overhead
586 on the system. This should only be enabled when the system
587 is to be analyzed in much detail.
590 config TRACING_BRANCHES
593 Selected by tracers that will trace the likely and unlikely
594 conditions. This prevents the tracers themselves from being
595 profiled. Profiling the tracing infrastructure can only happen
596 when the likelys and unlikelys are not being traced.
599 bool "Trace likely/unlikely instances"
600 depends on TRACE_BRANCH_PROFILING
601 select TRACING_BRANCHES
603 This traces the events of likely and unlikely condition
604 calls in the kernel. The difference between this and the
605 "Trace likely/unlikely profiler" is that this is not a
606 histogram of the callers, but actually places the calling
607 events into a running trace buffer to see when and where the
608 events happened, as well as their results.
612 config BLK_DEV_IO_TRACE
613 bool "Support for tracing block IO actions"
619 select GENERIC_TRACER
622 Say Y here if you want to be able to trace the block layer actions
623 on a given queue. Tracing allows you to see any traffic happening
624 on a block device queue. For more information (and the userspace
625 support tools needed), fetch the blktrace tools from:
627 git://git.kernel.dk/blktrace.git
629 Tracing also is possible using the ftrace interface, e.g.:
631 echo 1 > /sys/block/sda/sda1/trace/enable
632 echo blk > /sys/kernel/debug/tracing/current_tracer
633 cat /sys/kernel/debug/tracing/trace_pipe
639 depends on HAVE_REGS_AND_STACK_ACCESS_API
640 bool "Enable kprobes-based dynamic events"
643 select DYNAMIC_EVENTS
646 This allows the user to add tracing events (similar to tracepoints)
647 on the fly via the ftrace interface. See
648 Documentation/trace/kprobetrace.rst for more details.
650 Those events can be inserted wherever kprobes can probe, and record
651 various register and memory values.
653 This option is also required by perf-probe subcommand of perf tools.
654 If you want to use perf tools, this option is strongly recommended.
656 config KPROBE_EVENTS_ON_NOTRACE
657 bool "Do NOT protect notrace function from kprobe events"
658 depends on KPROBE_EVENTS
659 depends on DYNAMIC_FTRACE
662 This is only for the developers who want to debug ftrace itself
665 If kprobes can use ftrace instead of breakpoint, ftrace related
666 functions are protected from kprobe-events to prevent an infinite
667 recursion or any unexpected execution path which leads to a kernel
670 This option disables such protection and allows you to put kprobe
671 events on ftrace functions for debugging ftrace by itself.
672 Note that this might let you shoot yourself in the foot.
677 bool "Enable uprobes-based dynamic events"
678 depends on ARCH_SUPPORTS_UPROBES
680 depends on PERF_EVENTS
683 select DYNAMIC_EVENTS
687 This allows the user to add tracing events on top of userspace
688 dynamic events (similar to tracepoints) on the fly via the trace
689 events interface. Those events can be inserted wherever uprobes
690 can probe, and record various registers.
691 This option is required if you plan to use perf-probe subcommand
692 of perf tools on user space applications.
695 depends on BPF_SYSCALL
696 depends on (KPROBE_EVENTS || UPROBE_EVENTS) && PERF_EVENTS
700 This allows the user to attach BPF programs to kprobe, uprobe, and
703 config DYNAMIC_EVENTS
709 config BPF_KPROBE_OVERRIDE
710 bool "Enable BPF programs to override a kprobed function"
711 depends on BPF_EVENTS
712 depends on FUNCTION_ERROR_INJECTION
715 Allows BPF to override the execution of a probed function and
716 set a different return value. This is used for error injection.
718 config FTRACE_MCOUNT_RECORD
720 depends on DYNAMIC_FTRACE
721 depends on HAVE_FTRACE_MCOUNT_RECORD
723 config FTRACE_MCOUNT_USE_PATCHABLE_FUNCTION_ENTRY
725 depends on FTRACE_MCOUNT_RECORD
727 config FTRACE_MCOUNT_USE_CC
729 depends on $(cc-option,-mrecord-mcount)
730 depends on !FTRACE_MCOUNT_USE_PATCHABLE_FUNCTION_ENTRY
731 depends on FTRACE_MCOUNT_RECORD
733 config FTRACE_MCOUNT_USE_OBJTOOL
735 depends on HAVE_OBJTOOL_MCOUNT
736 depends on !FTRACE_MCOUNT_USE_PATCHABLE_FUNCTION_ENTRY
737 depends on !FTRACE_MCOUNT_USE_CC
738 depends on FTRACE_MCOUNT_RECORD
741 config FTRACE_MCOUNT_USE_RECORDMCOUNT
743 depends on !FTRACE_MCOUNT_USE_PATCHABLE_FUNCTION_ENTRY
744 depends on !FTRACE_MCOUNT_USE_CC
745 depends on !FTRACE_MCOUNT_USE_OBJTOOL
746 depends on FTRACE_MCOUNT_RECORD
750 depends on ARCH_HAVE_NMI_SAFE_CMPXCHG
752 tracing_map is a special-purpose lock-free map for tracing,
753 separated out as a stand-alone facility in order to allow it
754 to be shared between multiple tracers. It isn't meant to be
755 generally used outside of that context, and is normally
756 selected by tracers that use it.
759 bool "Synthetic trace events"
761 select DYNAMIC_EVENTS
764 Synthetic events are user-defined trace events that can be
765 used to combine data from other trace events or in fact any
766 data source. Synthetic events can be generated indirectly
767 via the trace() action of histogram triggers or directly
768 by way of an in-kernel API.
770 See Documentation/trace/events.rst or
771 Documentation/trace/histogram.rst for details and examples.
776 bool "User trace events"
778 select DYNAMIC_EVENTS
779 depends on BROKEN || COMPILE_TEST # API needs to be straighten out
781 User trace events are user-defined trace events that
782 can be used like an existing kernel trace event. User trace
783 events are generated by writing to a tracefs file. User
784 processes can determine if their tracing events should be
785 generated by memory mapping a tracefs file and checking for
786 an associated byte being non-zero.
791 bool "Histogram triggers"
792 depends on ARCH_HAVE_NMI_SAFE_CMPXCHG
795 select DYNAMIC_EVENTS
799 Hist triggers allow one or more arbitrary trace event fields
800 to be aggregated into hash tables and dumped to stdout by
801 reading a debugfs/tracefs file. They're useful for
802 gathering quick and dirty (though precise) summaries of
803 event activity as an initial guide for further investigation
804 using more advanced tools.
806 Inter-event tracing of quantities such as latencies is also
807 supported using hist triggers under this option.
809 See Documentation/trace/histogram.rst.
812 config TRACE_EVENT_INJECT
813 bool "Trace event injection"
816 Allow user-space to inject a specific trace event into the ring
817 buffer. This is mainly used for testing purpose.
821 config TRACEPOINT_BENCHMARK
822 bool "Add tracepoint that benchmarks tracepoints"
824 This option creates the tracepoint "benchmark:benchmark_event".
825 When the tracepoint is enabled, it kicks off a kernel thread that
826 goes into an infinite loop (calling cond_resched() to let other tasks
827 run), and calls the tracepoint. Each iteration will record the time
828 it took to write to the tracepoint and the next iteration that
829 data will be passed to the tracepoint itself. That is, the tracepoint
830 will report the time it took to do the previous tracepoint.
831 The string written to the tracepoint is a static string of 128 bytes
832 to keep the time the same. The initial string is simply a write of
833 "START". The second string records the cold cache time of the first
834 write which is not added to the rest of the calculations.
836 As it is a tight loop, it benchmarks as hot cache. That's fine because
837 we care most about hot paths that are probably in cache already.
839 An example of the output:
842 first=3672 [COLD CACHED]
843 last=632 first=3672 max=632 min=632 avg=316 std=446 std^2=199712
844 last=278 first=3672 max=632 min=278 avg=303 std=316 std^2=100337
845 last=277 first=3672 max=632 min=277 avg=296 std=258 std^2=67064
846 last=273 first=3672 max=632 min=273 avg=292 std=224 std^2=50411
847 last=273 first=3672 max=632 min=273 avg=288 std=200 std^2=40389
848 last=281 first=3672 max=632 min=273 avg=287 std=183 std^2=33666
851 config RING_BUFFER_BENCHMARK
852 tristate "Ring buffer benchmark stress tester"
853 depends on RING_BUFFER
855 This option creates a test to stress the ring buffer and benchmark it.
856 It creates its own ring buffer such that it will not interfere with
857 any other users of the ring buffer (such as ftrace). It then creates
858 a producer and consumer that will run for 10 seconds and sleep for
859 10 seconds. Each interval it will print out the number of events
860 it recorded and give a rough estimate of how long each iteration took.
862 It does not disable interrupts or raise its priority, so it may be
863 affected by processes that are running.
867 config TRACE_EVAL_MAP_FILE
868 bool "Show eval mappings for trace events"
871 The "print fmt" of the trace events will show the enum/sizeof names
872 instead of their values. This can cause problems for user space tools
873 that use this string to parse the raw data as user space does not know
874 how to convert the string to its value.
876 To fix this, there's a special macro in the kernel that can be used
877 to convert an enum/sizeof into its value. If this macro is used, then
878 the print fmt strings will be converted to their values.
880 If something does not get converted properly, this option can be
881 used to show what enums/sizeof the kernel tried to convert.
883 This option is for debugging the conversions. A file is created
884 in the tracing directory called "eval_map" that will show the
885 names matched with their values and what trace event system they
888 Normally, the mapping of the strings to values will be freed after
889 boot up or module load. With this option, they will not be freed, as
890 they are needed for the "eval_map" file. Enabling this option will
891 increase the memory footprint of the running kernel.
895 config FTRACE_RECORD_RECURSION
896 bool "Record functions that recurse in function tracing"
897 depends on FUNCTION_TRACER
899 All callbacks that attach to the function tracing have some sort
900 of protection against recursion. Even though the protection exists,
901 it adds overhead. This option will create a file in the tracefs
902 file system called "recursed_functions" that will list the functions
903 that triggered a recursion.
905 This will add more overhead to cases that have recursion.
909 config FTRACE_RECORD_RECURSION_SIZE
910 int "Max number of recursed functions to record"
912 depends on FTRACE_RECORD_RECURSION
914 This defines the limit of number of functions that can be
915 listed in the "recursed_functions" file, that lists all
916 the functions that caused a recursion to happen.
917 This file can be reset, but the limit can not change in
920 config RING_BUFFER_RECORD_RECURSION
921 bool "Record functions that recurse in the ring buffer"
922 depends on FTRACE_RECORD_RECURSION
923 # default y, because it is coupled with FTRACE_RECORD_RECURSION
926 The ring buffer has its own internal recursion. Although when
927 recursion happens it wont cause harm because of the protection,
928 but it does cause an unwanted overhead. Enabling this option will
929 place where recursion was detected into the ftrace "recursed_functions"
932 This will add more overhead to cases that have recursion.
934 config GCOV_PROFILE_FTRACE
935 bool "Enable GCOV profiling on ftrace subsystem"
936 depends on GCOV_KERNEL
938 Enable GCOV profiling on ftrace subsystem for checking
939 which functions/lines are tested.
943 Note that on a kernel compiled with this config, ftrace will
944 run significantly slower.
946 config FTRACE_SELFTEST
949 config FTRACE_STARTUP_TEST
950 bool "Perform a startup test on ftrace"
951 depends on GENERIC_TRACER
952 select FTRACE_SELFTEST
954 This option performs a series of startup tests on ftrace. On bootup
955 a series of tests are made to verify that the tracer is
956 functioning properly. It will do tests on all the configured
959 config EVENT_TRACE_STARTUP_TEST
960 bool "Run selftest on trace events"
961 depends on FTRACE_STARTUP_TEST
964 This option performs a test on all trace events in the system.
965 It basically just enables each event and runs some code that
966 will trigger events (not necessarily the event it enables)
967 This may take some time run as there are a lot of events.
969 config EVENT_TRACE_TEST_SYSCALLS
970 bool "Run selftest on syscall events"
971 depends on EVENT_TRACE_STARTUP_TEST
973 This option will also enable testing every syscall event.
974 It only enables the event and disables it and runs various loads
975 with the event enabled. This adds a bit more time for kernel boot
976 up since it runs this on every system call defined.
978 TBD - enable a way to actually call the syscalls as we test their
981 config FTRACE_SORT_STARTUP_TEST
982 bool "Verify compile time sorting of ftrace functions"
983 depends on DYNAMIC_FTRACE
984 depends on BUILDTIME_MCOUNT_SORT
986 Sorting of the mcount_loc sections that is used to find the
987 where the ftrace knows where to patch functions for tracing
988 and other callbacks is done at compile time. But if the sort
989 is not done correctly, it will cause non-deterministic failures.
990 When this is set, the sorted sections will be verified that they
991 are in deed sorted and will warn if they are not.
995 config RING_BUFFER_STARTUP_TEST
996 bool "Ring buffer startup self test"
997 depends on RING_BUFFER
999 Run a simple self test on the ring buffer on boot up. Late in the
1000 kernel boot sequence, the test will start that kicks off
1001 a thread per cpu. Each thread will write various size events
1002 into the ring buffer. Another thread is created to send IPIs
1003 to each of the threads, where the IPI handler will also write
1004 to the ring buffer, to test/stress the nesting ability.
1005 If any anomalies are discovered, a warning will be displayed
1006 and all ring buffers will be disabled.
1008 The test runs for 10 seconds. This will slow your boot time
1009 by at least 10 more seconds.
1011 At the end of the test, statics and more checks are done.
1012 It will output the stats of each per cpu buffer. What
1013 was written, the sizes, what was read, what was lost, and
1014 other similar details.
1018 config RING_BUFFER_VALIDATE_TIME_DELTAS
1019 bool "Verify ring buffer time stamp deltas"
1020 depends on RING_BUFFER
1022 This will audit the time stamps on the ring buffer sub
1023 buffer to make sure that all the time deltas for the
1024 events on a sub buffer matches the current time stamp.
1025 This audit is performed for every event that is not
1026 interrupted, or interrupting another event. A check
1027 is also made when traversing sub buffers to make sure
1028 that all the deltas on the previous sub buffer do not
1029 add up to be greater than the current time stamp.
1031 NOTE: This adds significant overhead to recording of events,
1032 and should only be used to test the logic of the ring buffer.
1033 Do not use it on production systems.
1035 Only say Y if you understand what this does, and you
1036 still want it enabled. Otherwise say N
1038 config MMIOTRACE_TEST
1039 tristate "Test module for mmiotrace"
1040 depends on MMIOTRACE && m
1042 This is a dumb module for testing mmiotrace. It is very dangerous
1043 as it will write garbage to IO memory starting at a given address.
1044 However, it should be safe to use on e.g. unused portion of VRAM.
1046 Say N, unless you absolutely know what you are doing.
1048 config PREEMPTIRQ_DELAY_TEST
1049 tristate "Test module to create a preempt / IRQ disable delay thread to test latency tracers"
1052 Select this option to build a test module that can help test latency
1053 tracers by executing a preempt or irq disable section with a user
1054 configurable delay. The module busy waits for the duration of the
1057 For example, the following invocation generates a burst of three
1058 irq-disabled critical sections for 500us:
1059 modprobe preemptirq_delay_test test_mode=irq delay=500 burst_size=3
1061 What's more, if you want to attach the test on the cpu which the latency
1062 tracer is running on, specify cpu_affinity=cpu_num at the end of the
1067 config SYNTH_EVENT_GEN_TEST
1068 tristate "Test module for in-kernel synthetic event generation"
1069 depends on SYNTH_EVENTS
1071 This option creates a test module to check the base
1072 functionality of in-kernel synthetic event definition and
1075 To test, insert the module, and then check the trace buffer
1076 for the generated sample events.
1080 config KPROBE_EVENT_GEN_TEST
1081 tristate "Test module for in-kernel kprobe event generation"
1082 depends on KPROBE_EVENTS
1084 This option creates a test module to check the base
1085 functionality of in-kernel kprobe event definition.
1087 To test, insert the module, and then check the trace buffer
1088 for the generated kprobe events.
1092 config HIST_TRIGGERS_DEBUG
1093 bool "Hist trigger debug support"
1094 depends on HIST_TRIGGERS
1096 Add "hist_debug" file for each event, which when read will
1097 dump out a bunch of internal details about the hist triggers
1098 defined on that event.
1100 The hist_debug file serves a couple of purposes:
1102 - Helps developers verify that nothing is broken.
1104 - Provides educational information to support the details
1105 of the hist trigger internals as described by
1106 Documentation/trace/histogram-design.rst.
1108 The hist_debug output only covers the data structures
1109 related to the histogram definitions themselves and doesn't
1110 display the internals of map buckets or variable values of
1115 source "kernel/trace/rv/Kconfig"