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
75 config CONTEXT_SWITCH_TRACER
78 config RING_BUFFER_ALLOW_SWAP
81 Allow the use of ring_buffer_swap_cpu.
82 Adds a very slight overhead to tracing when enabled.
84 # All tracer options should select GENERIC_TRACER. For those options that are
85 # enabled by all tracers (context switch and event tracer) they select TRACING.
86 # This allows those options to appear when no other tracer is selected. But the
87 # options do not appear when something else selects it. We need the two options
88 # GENERIC_TRACER and TRACING to avoid circular dependencies to accomplish the
89 # hiding of the automatic options.
95 select STACKTRACE if STACKTRACE_SUPPORT
102 config GENERIC_TRACER
107 # Minimum requirements an architecture has to meet for us to
108 # be able to offer generic tracing facilities:
110 config TRACING_SUPPORT
112 # PPC32 has no irqflags tracing support, but it can use most of the
113 # tracers anyway, they were tested to build and work. Note that new
114 # exceptions to this list aren't welcomed, better implement the
115 # irqflags tracing for your architecture.
116 depends on TRACE_IRQFLAGS_SUPPORT || PPC32
117 depends on STACKTRACE_SUPPORT
124 default y if DEBUG_KERNEL
126 Enable the kernel tracing infrastructure.
130 config FUNCTION_TRACER
131 bool "Kernel Function Tracer"
132 depends on HAVE_FUNCTION_TRACER
134 select GENERIC_TRACER
135 select CONTEXT_SWITCH_TRACER
137 Enable the kernel to trace every kernel function. This is done
138 by using a compiler feature to insert a small, 5-byte No-Operation
139 instruction at the beginning of every kernel function, which NOP
140 sequence is then dynamically patched into a tracer call when
141 tracing is enabled by the administrator. If it's runtime disabled
142 (the bootup default), then the overhead of the instructions is very
143 small and not measurable even in micro-benchmarks.
145 config FUNCTION_GRAPH_TRACER
146 bool "Kernel Function Graph Tracer"
147 depends on HAVE_FUNCTION_GRAPH_TRACER
148 depends on FUNCTION_TRACER
149 depends on !X86_32 || !CC_OPTIMIZE_FOR_SIZE
152 Enable the kernel to trace a function at both its return
154 Its first purpose is to trace the duration of functions and
155 draw a call graph for each thread with some information like
156 the return value. This is done by setting the current return
157 address on the current task structure into a stack of calls.
160 config IRQSOFF_TRACER
161 bool "Interrupts-off Latency Tracer"
163 depends on TRACE_IRQFLAGS_SUPPORT
164 depends on !ARCH_USES_GETTIMEOFFSET
165 select TRACE_IRQFLAGS
166 select GENERIC_TRACER
167 select TRACER_MAX_TRACE
168 select RING_BUFFER_ALLOW_SWAP
169 select TRACER_SNAPSHOT
170 select TRACER_SNAPSHOT_PER_CPU_SWAP
172 This option measures the time spent in irqs-off critical
173 sections, with microsecond accuracy.
175 The default measurement method is a maximum search, which is
176 disabled by default and can be runtime (re-)started
179 echo 0 > /sys/kernel/debug/tracing/tracing_max_latency
181 (Note that kernel size and overhead increase with this option
182 enabled. This option and the preempt-off timing option can be
183 used together or separately.)
185 config PREEMPT_TRACER
186 bool "Preemption-off Latency Tracer"
188 depends on !ARCH_USES_GETTIMEOFFSET
190 select GENERIC_TRACER
191 select TRACER_MAX_TRACE
192 select RING_BUFFER_ALLOW_SWAP
193 select TRACER_SNAPSHOT
194 select TRACER_SNAPSHOT_PER_CPU_SWAP
196 This option measures the time spent in preemption-off critical
197 sections, with microsecond accuracy.
199 The default measurement method is a maximum search, which is
200 disabled by default and can be runtime (re-)started
203 echo 0 > /sys/kernel/debug/tracing/tracing_max_latency
205 (Note that kernel size and overhead increase with this option
206 enabled. This option and the irqs-off timing option can be
207 used together or separately.)
210 bool "Scheduling Latency Tracer"
211 select GENERIC_TRACER
212 select CONTEXT_SWITCH_TRACER
213 select TRACER_MAX_TRACE
214 select TRACER_SNAPSHOT
216 This tracer tracks the latency of the highest priority task
217 to be scheduled in, starting from the point it has woken up.
219 config ENABLE_DEFAULT_TRACERS
220 bool "Trace process context switches and events"
221 depends on !GENERIC_TRACER
224 This tracer hooks to various trace points in the kernel,
225 allowing the user to pick and choose which trace point they
226 want to trace. It also includes the sched_switch tracer plugin.
228 config FTRACE_SYSCALLS
229 bool "Trace syscalls"
230 depends on HAVE_SYSCALL_TRACEPOINTS
231 select GENERIC_TRACER
234 Basic tracer to catch the syscall entry and exit events.
236 config TRACER_SNAPSHOT
237 bool "Create a snapshot trace buffer"
238 select TRACER_MAX_TRACE
240 Allow tracing users to take snapshot of the current buffer using the
241 ftrace interface, e.g.:
243 echo 1 > /sys/kernel/debug/tracing/snapshot
246 config TRACER_SNAPSHOT_PER_CPU_SWAP
247 bool "Allow snapshot to swap per CPU"
248 depends on TRACER_SNAPSHOT
249 select RING_BUFFER_ALLOW_SWAP
251 Allow doing a snapshot of a single CPU buffer instead of a
252 full swap (all buffers). If this is set, then the following is
255 echo 1 > /sys/kernel/debug/tracing/per_cpu/cpu2/snapshot
257 After which, only the tracing buffer for CPU 2 was swapped with
258 the main tracing buffer, and the other CPU buffers remain the same.
260 When this is enabled, this adds a little more overhead to the
261 trace recording, as it needs to add some checks to synchronize
262 recording with swaps. But this does not affect the performance
263 of the overall system. This is enabled by default when the preempt
264 or irq latency tracers are enabled, as those need to swap as well
265 and already adds the overhead (plus a lot more).
267 config TRACE_BRANCH_PROFILING
269 select GENERIC_TRACER
272 prompt "Branch Profiling"
273 default BRANCH_PROFILE_NONE
275 The branch profiling is a software profiler. It will add hooks
276 into the C conditionals to test which path a branch takes.
278 The likely/unlikely profiler only looks at the conditions that
279 are annotated with a likely or unlikely macro.
281 The "all branch" profiler will profile every if-statement in the
282 kernel. This profiler will also enable the likely/unlikely
285 Either of the above profilers adds a bit of overhead to the system.
286 If unsure, choose "No branch profiling".
288 config BRANCH_PROFILE_NONE
289 bool "No branch profiling"
291 No branch profiling. Branch profiling adds a bit of overhead.
292 Only enable it if you want to analyse the branching behavior.
293 Otherwise keep it disabled.
295 config PROFILE_ANNOTATED_BRANCHES
296 bool "Trace likely/unlikely profiler"
297 select TRACE_BRANCH_PROFILING
299 This tracer profiles all likely and unlikely macros
300 in the kernel. It will display the results in:
302 /sys/kernel/debug/tracing/trace_stat/branch_annotated
304 Note: this will add a significant overhead; only turn this
305 on if you need to profile the system's use of these macros.
307 config PROFILE_ALL_BRANCHES
308 bool "Profile all if conditionals"
309 select TRACE_BRANCH_PROFILING
311 This tracer profiles all branch conditions. Every if ()
312 taken in the kernel is recorded whether it hit or miss.
313 The results will be displayed in:
315 /sys/kernel/debug/tracing/trace_stat/branch_all
317 This option also enables the likely/unlikely profiler.
319 This configuration, when enabled, will impose a great overhead
320 on the system. This should only be enabled when the system
321 is to be analyzed in much detail.
324 config TRACING_BRANCHES
327 Selected by tracers that will trace the likely and unlikely
328 conditions. This prevents the tracers themselves from being
329 profiled. Profiling the tracing infrastructure can only happen
330 when the likelys and unlikelys are not being traced.
333 bool "Trace likely/unlikely instances"
334 depends on TRACE_BRANCH_PROFILING
335 select TRACING_BRANCHES
337 This traces the events of likely and unlikely condition
338 calls in the kernel. The difference between this and the
339 "Trace likely/unlikely profiler" is that this is not a
340 histogram of the callers, but actually places the calling
341 events into a running trace buffer to see when and where the
342 events happened, as well as their results.
347 bool "Trace max stack"
348 depends on HAVE_FUNCTION_TRACER
349 select FUNCTION_TRACER
353 This special tracer records the maximum stack footprint of the
354 kernel and displays it in /sys/kernel/debug/tracing/stack_trace.
356 This tracer works by hooking into every function call that the
357 kernel executes, and keeping a maximum stack depth value and
358 stack-trace saved. If this is configured with DYNAMIC_FTRACE
359 then it will not have any overhead while the stack tracer
362 To enable the stack tracer on bootup, pass in 'stacktrace'
363 on the kernel command line.
365 The stack tracer can also be enabled or disabled via the
366 sysctl kernel.stack_tracer_enabled
370 config BLK_DEV_IO_TRACE
371 bool "Support for tracing block IO actions"
377 select GENERIC_TRACER
380 Say Y here if you want to be able to trace the block layer actions
381 on a given queue. Tracing allows you to see any traffic happening
382 on a block device queue. For more information (and the userspace
383 support tools needed), fetch the blktrace tools from:
385 git://git.kernel.dk/blktrace.git
387 Tracing also is possible using the ftrace interface, e.g.:
389 echo 1 > /sys/block/sda/sda1/trace/enable
390 echo blk > /sys/kernel/debug/tracing/current_tracer
391 cat /sys/kernel/debug/tracing/trace_pipe
397 depends on HAVE_REGS_AND_STACK_ACCESS_API
398 bool "Enable kprobes-based dynamic events"
403 This allows the user to add tracing events (similar to tracepoints)
404 on the fly via the ftrace interface. See
405 Documentation/trace/kprobetrace.txt for more details.
407 Those events can be inserted wherever kprobes can probe, and record
408 various register and memory values.
410 This option is also required by perf-probe subcommand of perf tools.
411 If you want to use perf tools, this option is strongly recommended.
414 bool "Enable uprobes-based dynamic events"
415 depends on ARCH_SUPPORTS_UPROBES
417 depends on PERF_EVENTS
423 This allows the user to add tracing events on top of userspace
424 dynamic events (similar to tracepoints) on the fly via the trace
425 events interface. Those events can be inserted wherever uprobes
426 can probe, and record various registers.
427 This option is required if you plan to use perf-probe subcommand
428 of perf tools on user space applications.
431 depends on BPF_SYSCALL
432 depends on (KPROBE_EVENT || UPROBE_EVENT) && PERF_EVENTS
436 This allows the user to attach BPF programs to kprobe events.
441 config DYNAMIC_FTRACE
442 bool "enable/disable function tracing dynamically"
443 depends on FUNCTION_TRACER
444 depends on HAVE_DYNAMIC_FTRACE
447 This option will modify all the calls to function tracing
448 dynamically (will patch them out of the binary image and
449 replace them with a No-Op instruction) on boot up. During
450 compile time, a table is made of all the locations that ftrace
451 can function trace, and this table is linked into the kernel
452 image. When this is enabled, functions can be individually
453 enabled, and the functions not enabled will not affect
454 performance of the system.
456 See the files in /sys/kernel/debug/tracing:
457 available_filter_functions
461 This way a CONFIG_FUNCTION_TRACER kernel is slightly larger, but
462 otherwise has native performance as long as no tracing is active.
464 config DYNAMIC_FTRACE_WITH_REGS
466 depends on DYNAMIC_FTRACE
467 depends on HAVE_DYNAMIC_FTRACE_WITH_REGS
469 config FUNCTION_PROFILER
470 bool "Kernel function profiler"
471 depends on FUNCTION_TRACER
474 This option enables the kernel function profiler. A file is created
475 in debugfs called function_profile_enabled which defaults to zero.
476 When a 1 is echoed into this file profiling begins, and when a
477 zero is entered, profiling stops. A "functions" file is created in
478 the trace_stats directory; this file shows the list of functions that
479 have been hit and their counters.
483 config FTRACE_MCOUNT_RECORD
485 depends on DYNAMIC_FTRACE
486 depends on HAVE_FTRACE_MCOUNT_RECORD
488 config FTRACE_SELFTEST
491 config FTRACE_STARTUP_TEST
492 bool "Perform a startup test on ftrace"
493 depends on GENERIC_TRACER
494 select FTRACE_SELFTEST
496 This option performs a series of startup tests on ftrace. On bootup
497 a series of tests are made to verify that the tracer is
498 functioning properly. It will do tests on all the configured
501 config EVENT_TRACE_TEST_SYSCALLS
502 bool "Run selftest on syscall events"
503 depends on FTRACE_STARTUP_TEST
505 This option will also enable testing every syscall event.
506 It only enables the event and disables it and runs various loads
507 with the event enabled. This adds a bit more time for kernel boot
508 up since it runs this on every system call defined.
510 TBD - enable a way to actually call the syscalls as we test their
514 bool "Memory mapped IO tracing"
515 depends on HAVE_MMIOTRACE_SUPPORT && PCI
516 select GENERIC_TRACER
518 Mmiotrace traces Memory Mapped I/O access and is meant for
519 debugging and reverse engineering. It is called from the ioremap
520 implementation and works via page faults. Tracing is disabled by
521 default and can be enabled at run-time.
523 See Documentation/trace/mmiotrace.txt.
524 If you are not helping to develop drivers, say N.
528 depends on ARCH_HAVE_NMI_SAFE_CMPXCHG
530 tracing_map is a special-purpose lock-free map for tracing,
531 separated out as a stand-alone facility in order to allow it
532 to be shared between multiple tracers. It isn't meant to be
533 generally used outside of that context, and is normally
534 selected by tracers that use it.
537 bool "Histogram triggers"
538 depends on ARCH_HAVE_NMI_SAFE_CMPXCHG
543 Hist triggers allow one or more arbitrary trace event fields
544 to be aggregated into hash tables and dumped to stdout by
545 reading a debugfs/tracefs file. They're useful for
546 gathering quick and dirty (though precise) summaries of
547 event activity as an initial guide for further investigation
548 using more advanced tools.
550 See Documentation/trace/events.txt.
553 config MMIOTRACE_TEST
554 tristate "Test module for mmiotrace"
555 depends on MMIOTRACE && m
557 This is a dumb module for testing mmiotrace. It is very dangerous
558 as it will write garbage to IO memory starting at a given address.
559 However, it should be safe to use on e.g. unused portion of VRAM.
561 Say N, unless you absolutely know what you are doing.
563 config TRACEPOINT_BENCHMARK
564 bool "Add tracepoint that benchmarks tracepoints"
566 This option creates the tracepoint "benchmark:benchmark_event".
567 When the tracepoint is enabled, it kicks off a kernel thread that
568 goes into an infinite loop (calling cond_sched() to let other tasks
569 run), and calls the tracepoint. Each iteration will record the time
570 it took to write to the tracepoint and the next iteration that
571 data will be passed to the tracepoint itself. That is, the tracepoint
572 will report the time it took to do the previous tracepoint.
573 The string written to the tracepoint is a static string of 128 bytes
574 to keep the time the same. The initial string is simply a write of
575 "START". The second string records the cold cache time of the first
576 write which is not added to the rest of the calculations.
578 As it is a tight loop, it benchmarks as hot cache. That's fine because
579 we care most about hot paths that are probably in cache already.
581 An example of the output:
584 first=3672 [COLD CACHED]
585 last=632 first=3672 max=632 min=632 avg=316 std=446 std^2=199712
586 last=278 first=3672 max=632 min=278 avg=303 std=316 std^2=100337
587 last=277 first=3672 max=632 min=277 avg=296 std=258 std^2=67064
588 last=273 first=3672 max=632 min=273 avg=292 std=224 std^2=50411
589 last=273 first=3672 max=632 min=273 avg=288 std=200 std^2=40389
590 last=281 first=3672 max=632 min=273 avg=287 std=183 std^2=33666
593 config RING_BUFFER_BENCHMARK
594 tristate "Ring buffer benchmark stress tester"
595 depends on RING_BUFFER
597 This option creates a test to stress the ring buffer and benchmark it.
598 It creates its own ring buffer such that it will not interfere with
599 any other users of the ring buffer (such as ftrace). It then creates
600 a producer and consumer that will run for 10 seconds and sleep for
601 10 seconds. Each interval it will print out the number of events
602 it recorded and give a rough estimate of how long each iteration took.
604 It does not disable interrupts or raise its priority, so it may be
605 affected by processes that are running.
609 config RING_BUFFER_STARTUP_TEST
610 bool "Ring buffer startup self test"
611 depends on RING_BUFFER
613 Run a simple self test on the ring buffer on boot up. Late in the
614 kernel boot sequence, the test will start that kicks off
615 a thread per cpu. Each thread will write various size events
616 into the ring buffer. Another thread is created to send IPIs
617 to each of the threads, where the IPI handler will also write
618 to the ring buffer, to test/stress the nesting ability.
619 If any anomalies are discovered, a warning will be displayed
620 and all ring buffers will be disabled.
622 The test runs for 10 seconds. This will slow your boot time
623 by at least 10 more seconds.
625 At the end of the test, statics and more checks are done.
626 It will output the stats of each per cpu buffer. What
627 was written, the sizes, what was read, what was lost, and
628 other similar details.
632 config TRACE_ENUM_MAP_FILE
633 bool "Show enum mappings for trace events"
636 The "print fmt" of the trace events will show the enum names instead
637 of their values. This can cause problems for user space tools that
638 use this string to parse the raw data as user space does not know
639 how to convert the string to its value.
641 To fix this, there's a special macro in the kernel that can be used
642 to convert the enum into its value. If this macro is used, then the
643 print fmt strings will have the enums converted to their values.
645 If something does not get converted properly, this option can be
646 used to show what enums the kernel tried to convert.
648 This option is for debugging the enum conversions. A file is created
649 in the tracing directory called "enum_map" that will show the enum
650 names matched with their values and what trace event system they
653 Normally, the mapping of the strings to values will be freed after
654 boot up or module load. With this option, they will not be freed, as
655 they are needed for the "enum_map" file. Enabling this option will
656 increase the memory footprint of the running kernel.
660 config TRACING_EVENTS_GPIO
661 bool "Trace gpio events"
665 Enable tracing events for gpio subsystem
669 endif # TRACING_SUPPORT