2 # Architectures that offer an FUNCTION_TRACER implementation should
3 # select HAVE_FUNCTION_TRACER:
6 config USER_STACKTRACE_SUPPORT
12 config HAVE_FTRACE_NMI_ENTER
15 See Documentation/trace/ftrace-design.txt
17 config HAVE_FUNCTION_TRACER
20 See Documentation/trace/ftrace-design.txt
22 config HAVE_FUNCTION_GRAPH_TRACER
25 See Documentation/trace/ftrace-design.txt
27 config HAVE_DYNAMIC_FTRACE
30 See Documentation/trace/ftrace-design.txt
32 config HAVE_DYNAMIC_FTRACE_WITH_REGS
35 config HAVE_FTRACE_MCOUNT_RECORD
38 See Documentation/trace/ftrace-design.txt
40 config HAVE_SYSCALL_TRACEPOINTS
43 See Documentation/trace/ftrace-design.txt
48 Arch supports the gcc options -pg with -mfentry
50 config HAVE_C_RECORDMCOUNT
53 C version of recordmcount available?
55 config TRACER_MAX_TRACE
66 config FTRACE_NMI_ENTER
68 depends on HAVE_FTRACE_NMI_ENTER
72 select CONTEXT_SWITCH_TRACER
76 config CONTEXT_SWITCH_TRACER
79 config RING_BUFFER_ALLOW_SWAP
82 Allow the use of ring_buffer_swap_cpu.
83 Adds a very slight overhead to tracing when enabled.
85 # All tracer options should select GENERIC_TRACER. For those options that are
86 # enabled by all tracers (context switch and event tracer) they select TRACING.
87 # This allows those options to appear when no other tracer is selected. But the
88 # options do not appear when something else selects it. We need the two options
89 # GENERIC_TRACER and TRACING to avoid circular dependencies to accomplish the
90 # hiding of the automatic options.
96 select STACKTRACE if STACKTRACE_SUPPORT
103 config GENERIC_TRACER
108 # Minimum requirements an architecture has to meet for us to
109 # be able to offer generic tracing facilities:
111 config TRACING_SUPPORT
113 # PPC32 has no irqflags tracing support, but it can use most of the
114 # tracers anyway, they were tested to build and work. Note that new
115 # exceptions to this list aren't welcomed, better implement the
116 # irqflags tracing for your architecture.
117 depends on TRACE_IRQFLAGS_SUPPORT || PPC32
118 depends on STACKTRACE_SUPPORT
125 default y if DEBUG_KERNEL
127 Enable the kernel tracing infrastructure.
131 config FUNCTION_TRACER
132 bool "Kernel Function Tracer"
133 depends on HAVE_FUNCTION_TRACER
135 select GENERIC_TRACER
136 select CONTEXT_SWITCH_TRACER
138 select TASKS_RCU if PREEMPT
140 Enable the kernel to trace every kernel function. This is done
141 by using a compiler feature to insert a small, 5-byte No-Operation
142 instruction at the beginning of every kernel function, which NOP
143 sequence is then dynamically patched into a tracer call when
144 tracing is enabled by the administrator. If it's runtime disabled
145 (the bootup default), then the overhead of the instructions is very
146 small and not measurable even in micro-benchmarks.
148 config FUNCTION_GRAPH_TRACER
149 bool "Kernel Function Graph Tracer"
150 depends on HAVE_FUNCTION_GRAPH_TRACER
151 depends on FUNCTION_TRACER
152 depends on !X86_32 || !CC_OPTIMIZE_FOR_SIZE
155 Enable the kernel to trace a function at both its return
157 Its first purpose is to trace the duration of functions and
158 draw a call graph for each thread with some information like
159 the return value. This is done by setting the current return
160 address on the current task structure into a stack of calls.
163 config PREEMPTIRQ_EVENTS
164 bool "Enable trace events for preempt and irq disable/enable"
165 select TRACE_IRQFLAGS
166 depends on DEBUG_PREEMPT || !PROVE_LOCKING
170 Enable tracing of disable and enable events for preemption and irqs.
171 For tracing preempt disable/enable events, DEBUG_PREEMPT must be
172 enabled. For tracing irq disable/enable events, PROVE_LOCKING must
175 config IRQSOFF_TRACER
176 bool "Interrupts-off Latency Tracer"
178 depends on TRACE_IRQFLAGS_SUPPORT
179 depends on !ARCH_USES_GETTIMEOFFSET
180 select TRACE_IRQFLAGS
181 select GENERIC_TRACER
182 select TRACER_MAX_TRACE
183 select RING_BUFFER_ALLOW_SWAP
184 select TRACER_SNAPSHOT
185 select TRACER_SNAPSHOT_PER_CPU_SWAP
187 This option measures the time spent in irqs-off critical
188 sections, with microsecond accuracy.
190 The default measurement method is a maximum search, which is
191 disabled by default and can be runtime (re-)started
194 echo 0 > /sys/kernel/debug/tracing/tracing_max_latency
196 (Note that kernel size and overhead increase with this option
197 enabled. This option and the preempt-off timing option can be
198 used together or separately.)
200 config PREEMPT_TRACER
201 bool "Preemption-off Latency Tracer"
203 depends on !ARCH_USES_GETTIMEOFFSET
205 select GENERIC_TRACER
206 select TRACER_MAX_TRACE
207 select RING_BUFFER_ALLOW_SWAP
208 select TRACER_SNAPSHOT
209 select TRACER_SNAPSHOT_PER_CPU_SWAP
211 This option measures the time spent in preemption-off critical
212 sections, with microsecond accuracy.
214 The default measurement method is a maximum search, which is
215 disabled by default and can be runtime (re-)started
218 echo 0 > /sys/kernel/debug/tracing/tracing_max_latency
220 (Note that kernel size and overhead increase with this option
221 enabled. This option and the irqs-off timing option can be
222 used together or separately.)
225 bool "Scheduling Latency Tracer"
226 select GENERIC_TRACER
227 select CONTEXT_SWITCH_TRACER
228 select TRACER_MAX_TRACE
229 select TRACER_SNAPSHOT
231 This tracer tracks the latency of the highest priority task
232 to be scheduled in, starting from the point it has woken up.
235 bool "Tracer to detect hardware latencies (like SMIs)"
236 select GENERIC_TRACER
238 This tracer, when enabled will create one or more kernel threads,
239 depending on what the cpumask file is set to, which each thread
240 spinning in a loop looking for interruptions caused by
241 something other than the kernel. For example, if a
242 System Management Interrupt (SMI) takes a noticeable amount of
243 time, this tracer will detect it. This is useful for testing
244 if a system is reliable for Real Time tasks.
246 Some files are created in the tracing directory when this
249 hwlat_detector/width - time in usecs for how long to spin for
250 hwlat_detector/window - time in usecs between the start of each
253 A kernel thread is created that will spin with interrupts disabled
254 for "width" microseconds in every "window" cycle. It will not spin
255 for "window - width" microseconds, where the system can
258 The output will appear in the trace and trace_pipe files.
260 When the tracer is not running, it has no affect on the system,
261 but when it is running, it can cause the system to be
262 periodically non responsive. Do not run this tracer on a
265 To enable this tracer, echo in "hwlat" into the current_tracer
266 file. Every time a latency is greater than tracing_thresh, it will
267 be recorded into the ring buffer.
269 config ENABLE_DEFAULT_TRACERS
270 bool "Trace process context switches and events"
271 depends on !GENERIC_TRACER
274 This tracer hooks to various trace points in the kernel,
275 allowing the user to pick and choose which trace point they
276 want to trace. It also includes the sched_switch tracer plugin.
278 config FTRACE_SYSCALLS
279 bool "Trace syscalls"
280 depends on HAVE_SYSCALL_TRACEPOINTS
281 select GENERIC_TRACER
284 Basic tracer to catch the syscall entry and exit events.
286 config TRACER_SNAPSHOT
287 bool "Create a snapshot trace buffer"
288 select TRACER_MAX_TRACE
290 Allow tracing users to take snapshot of the current buffer using the
291 ftrace interface, e.g.:
293 echo 1 > /sys/kernel/debug/tracing/snapshot
296 config TRACER_SNAPSHOT_PER_CPU_SWAP
297 bool "Allow snapshot to swap per CPU"
298 depends on TRACER_SNAPSHOT
299 select RING_BUFFER_ALLOW_SWAP
301 Allow doing a snapshot of a single CPU buffer instead of a
302 full swap (all buffers). If this is set, then the following is
305 echo 1 > /sys/kernel/debug/tracing/per_cpu/cpu2/snapshot
307 After which, only the tracing buffer for CPU 2 was swapped with
308 the main tracing buffer, and the other CPU buffers remain the same.
310 When this is enabled, this adds a little more overhead to the
311 trace recording, as it needs to add some checks to synchronize
312 recording with swaps. But this does not affect the performance
313 of the overall system. This is enabled by default when the preempt
314 or irq latency tracers are enabled, as those need to swap as well
315 and already adds the overhead (plus a lot more).
317 config TRACE_BRANCH_PROFILING
319 select GENERIC_TRACER
322 prompt "Branch Profiling"
323 default BRANCH_PROFILE_NONE
325 The branch profiling is a software profiler. It will add hooks
326 into the C conditionals to test which path a branch takes.
328 The likely/unlikely profiler only looks at the conditions that
329 are annotated with a likely or unlikely macro.
331 The "all branch" profiler will profile every if-statement in the
332 kernel. This profiler will also enable the likely/unlikely
335 Either of the above profilers adds a bit of overhead to the system.
336 If unsure, choose "No branch profiling".
338 config BRANCH_PROFILE_NONE
339 bool "No branch profiling"
341 No branch profiling. Branch profiling adds a bit of overhead.
342 Only enable it if you want to analyse the branching behavior.
343 Otherwise keep it disabled.
345 config PROFILE_ANNOTATED_BRANCHES
346 bool "Trace likely/unlikely profiler"
347 select TRACE_BRANCH_PROFILING
349 This tracer profiles all likely and unlikely macros
350 in the kernel. It will display the results in:
352 /sys/kernel/debug/tracing/trace_stat/branch_annotated
354 Note: this will add a significant overhead; only turn this
355 on if you need to profile the system's use of these macros.
357 config PROFILE_ALL_BRANCHES
358 bool "Profile all if conditionals" if !FORTIFY_SOURCE
359 select TRACE_BRANCH_PROFILING
361 This tracer profiles all branch conditions. Every if ()
362 taken in the kernel is recorded whether it hit or miss.
363 The results will be displayed in:
365 /sys/kernel/debug/tracing/trace_stat/branch_all
367 This option also enables the likely/unlikely profiler.
369 This configuration, when enabled, will impose a great overhead
370 on the system. This should only be enabled when the system
371 is to be analyzed in much detail.
374 config TRACING_BRANCHES
377 Selected by tracers that will trace the likely and unlikely
378 conditions. This prevents the tracers themselves from being
379 profiled. Profiling the tracing infrastructure can only happen
380 when the likelys and unlikelys are not being traced.
383 bool "Trace likely/unlikely instances"
384 depends on TRACE_BRANCH_PROFILING
385 select TRACING_BRANCHES
387 This traces the events of likely and unlikely condition
388 calls in the kernel. The difference between this and the
389 "Trace likely/unlikely profiler" is that this is not a
390 histogram of the callers, but actually places the calling
391 events into a running trace buffer to see when and where the
392 events happened, as well as their results.
397 bool "Trace max stack"
398 depends on HAVE_FUNCTION_TRACER
399 select FUNCTION_TRACER
403 This special tracer records the maximum stack footprint of the
404 kernel and displays it in /sys/kernel/debug/tracing/stack_trace.
406 This tracer works by hooking into every function call that the
407 kernel executes, and keeping a maximum stack depth value and
408 stack-trace saved. If this is configured with DYNAMIC_FTRACE
409 then it will not have any overhead while the stack tracer
412 To enable the stack tracer on bootup, pass in 'stacktrace'
413 on the kernel command line.
415 The stack tracer can also be enabled or disabled via the
416 sysctl kernel.stack_tracer_enabled
420 config BLK_DEV_IO_TRACE
421 bool "Support for tracing block IO actions"
427 select GENERIC_TRACER
430 Say Y here if you want to be able to trace the block layer actions
431 on a given queue. Tracing allows you to see any traffic happening
432 on a block device queue. For more information (and the userspace
433 support tools needed), fetch the blktrace tools from:
435 git://git.kernel.dk/blktrace.git
437 Tracing also is possible using the ftrace interface, e.g.:
439 echo 1 > /sys/block/sda/sda1/trace/enable
440 echo blk > /sys/kernel/debug/tracing/current_tracer
441 cat /sys/kernel/debug/tracing/trace_pipe
447 depends on HAVE_REGS_AND_STACK_ACCESS_API
448 bool "Enable kprobes-based dynamic events"
453 This allows the user to add tracing events (similar to tracepoints)
454 on the fly via the ftrace interface. See
455 Documentation/trace/kprobetrace.txt for more details.
457 Those events can be inserted wherever kprobes can probe, and record
458 various register and memory values.
460 This option is also required by perf-probe subcommand of perf tools.
461 If you want to use perf tools, this option is strongly recommended.
464 bool "Enable uprobes-based dynamic events"
465 depends on ARCH_SUPPORTS_UPROBES
467 depends on PERF_EVENTS
473 This allows the user to add tracing events on top of userspace
474 dynamic events (similar to tracepoints) on the fly via the trace
475 events interface. Those events can be inserted wherever uprobes
476 can probe, and record various registers.
477 This option is required if you plan to use perf-probe subcommand
478 of perf tools on user space applications.
481 depends on BPF_SYSCALL
482 depends on (KPROBE_EVENTS || UPROBE_EVENTS) && PERF_EVENTS
486 This allows the user to attach BPF programs to kprobe events.
491 config DYNAMIC_FTRACE
492 bool "enable/disable function tracing dynamically"
493 depends on FUNCTION_TRACER
494 depends on HAVE_DYNAMIC_FTRACE
497 This option will modify all the calls to function tracing
498 dynamically (will patch them out of the binary image and
499 replace them with a No-Op instruction) on boot up. During
500 compile time, a table is made of all the locations that ftrace
501 can function trace, and this table is linked into the kernel
502 image. When this is enabled, functions can be individually
503 enabled, and the functions not enabled will not affect
504 performance of the system.
506 See the files in /sys/kernel/debug/tracing:
507 available_filter_functions
511 This way a CONFIG_FUNCTION_TRACER kernel is slightly larger, but
512 otherwise has native performance as long as no tracing is active.
514 config DYNAMIC_FTRACE_WITH_REGS
516 depends on DYNAMIC_FTRACE
517 depends on HAVE_DYNAMIC_FTRACE_WITH_REGS
519 config FUNCTION_PROFILER
520 bool "Kernel function profiler"
521 depends on FUNCTION_TRACER
524 This option enables the kernel function profiler. A file is created
525 in debugfs called function_profile_enabled which defaults to zero.
526 When a 1 is echoed into this file profiling begins, and when a
527 zero is entered, profiling stops. A "functions" file is created in
528 the trace_stats directory; this file shows the list of functions that
529 have been hit and their counters.
533 config FTRACE_MCOUNT_RECORD
535 depends on DYNAMIC_FTRACE
536 depends on HAVE_FTRACE_MCOUNT_RECORD
538 config FTRACE_SELFTEST
541 config FTRACE_STARTUP_TEST
542 bool "Perform a startup test on ftrace"
543 depends on GENERIC_TRACER
544 select FTRACE_SELFTEST
546 This option performs a series of startup tests on ftrace. On bootup
547 a series of tests are made to verify that the tracer is
548 functioning properly. It will do tests on all the configured
551 config EVENT_TRACE_TEST_SYSCALLS
552 bool "Run selftest on syscall events"
553 depends on FTRACE_STARTUP_TEST
555 This option will also enable testing every syscall event.
556 It only enables the event and disables it and runs various loads
557 with the event enabled. This adds a bit more time for kernel boot
558 up since it runs this on every system call defined.
560 TBD - enable a way to actually call the syscalls as we test their
564 bool "Memory mapped IO tracing"
565 depends on HAVE_MMIOTRACE_SUPPORT && PCI
566 select GENERIC_TRACER
568 Mmiotrace traces Memory Mapped I/O access and is meant for
569 debugging and reverse engineering. It is called from the ioremap
570 implementation and works via page faults. Tracing is disabled by
571 default and can be enabled at run-time.
573 See Documentation/trace/mmiotrace.txt.
574 If you are not helping to develop drivers, say N.
578 depends on ARCH_HAVE_NMI_SAFE_CMPXCHG
580 tracing_map is a special-purpose lock-free map for tracing,
581 separated out as a stand-alone facility in order to allow it
582 to be shared between multiple tracers. It isn't meant to be
583 generally used outside of that context, and is normally
584 selected by tracers that use it.
587 bool "Histogram triggers"
588 depends on ARCH_HAVE_NMI_SAFE_CMPXCHG
593 Hist triggers allow one or more arbitrary trace event fields
594 to be aggregated into hash tables and dumped to stdout by
595 reading a debugfs/tracefs file. They're useful for
596 gathering quick and dirty (though precise) summaries of
597 event activity as an initial guide for further investigation
598 using more advanced tools.
600 See Documentation/trace/events.txt.
603 config MMIOTRACE_TEST
604 tristate "Test module for mmiotrace"
605 depends on MMIOTRACE && m
607 This is a dumb module for testing mmiotrace. It is very dangerous
608 as it will write garbage to IO memory starting at a given address.
609 However, it should be safe to use on e.g. unused portion of VRAM.
611 Say N, unless you absolutely know what you are doing.
613 config TRACEPOINT_BENCHMARK
614 bool "Add tracepoint that benchmarks tracepoints"
616 This option creates the tracepoint "benchmark:benchmark_event".
617 When the tracepoint is enabled, it kicks off a kernel thread that
618 goes into an infinite loop (calling cond_sched() to let other tasks
619 run), and calls the tracepoint. Each iteration will record the time
620 it took to write to the tracepoint and the next iteration that
621 data will be passed to the tracepoint itself. That is, the tracepoint
622 will report the time it took to do the previous tracepoint.
623 The string written to the tracepoint is a static string of 128 bytes
624 to keep the time the same. The initial string is simply a write of
625 "START". The second string records the cold cache time of the first
626 write which is not added to the rest of the calculations.
628 As it is a tight loop, it benchmarks as hot cache. That's fine because
629 we care most about hot paths that are probably in cache already.
631 An example of the output:
634 first=3672 [COLD CACHED]
635 last=632 first=3672 max=632 min=632 avg=316 std=446 std^2=199712
636 last=278 first=3672 max=632 min=278 avg=303 std=316 std^2=100337
637 last=277 first=3672 max=632 min=277 avg=296 std=258 std^2=67064
638 last=273 first=3672 max=632 min=273 avg=292 std=224 std^2=50411
639 last=273 first=3672 max=632 min=273 avg=288 std=200 std^2=40389
640 last=281 first=3672 max=632 min=273 avg=287 std=183 std^2=33666
643 config RING_BUFFER_BENCHMARK
644 tristate "Ring buffer benchmark stress tester"
645 depends on RING_BUFFER
647 This option creates a test to stress the ring buffer and benchmark it.
648 It creates its own ring buffer such that it will not interfere with
649 any other users of the ring buffer (such as ftrace). It then creates
650 a producer and consumer that will run for 10 seconds and sleep for
651 10 seconds. Each interval it will print out the number of events
652 it recorded and give a rough estimate of how long each iteration took.
654 It does not disable interrupts or raise its priority, so it may be
655 affected by processes that are running.
659 config RING_BUFFER_STARTUP_TEST
660 bool "Ring buffer startup self test"
661 depends on RING_BUFFER
663 Run a simple self test on the ring buffer on boot up. Late in the
664 kernel boot sequence, the test will start that kicks off
665 a thread per cpu. Each thread will write various size events
666 into the ring buffer. Another thread is created to send IPIs
667 to each of the threads, where the IPI handler will also write
668 to the ring buffer, to test/stress the nesting ability.
669 If any anomalies are discovered, a warning will be displayed
670 and all ring buffers will be disabled.
672 The test runs for 10 seconds. This will slow your boot time
673 by at least 10 more seconds.
675 At the end of the test, statics and more checks are done.
676 It will output the stats of each per cpu buffer. What
677 was written, the sizes, what was read, what was lost, and
678 other similar details.
682 config TRACE_EVAL_MAP_FILE
683 bool "Show eval mappings for trace events"
686 The "print fmt" of the trace events will show the enum/sizeof names
687 instead of their values. This can cause problems for user space tools
688 that use this string to parse the raw data as user space does not know
689 how to convert the string to its value.
691 To fix this, there's a special macro in the kernel that can be used
692 to convert an enum/sizeof into its value. If this macro is used, then
693 the print fmt strings will be converted to their values.
695 If something does not get converted properly, this option can be
696 used to show what enums/sizeof the kernel tried to convert.
698 This option is for debugging the conversions. A file is created
699 in the tracing directory called "eval_map" that will show the
700 names matched with their values and what trace event system they
703 Normally, the mapping of the strings to values will be freed after
704 boot up or module load. With this option, they will not be freed, as
705 they are needed for the "eval_map" file. Enabling this option will
706 increase the memory footprint of the running kernel.
710 config TRACING_EVENTS_GPIO
711 bool "Trace gpio events"
715 Enable tracing events for gpio subsystem
719 endif # TRACING_SUPPORT