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_FUNCTION_GRAPH_FP_TEST
30 See Documentation/trace/ftrace-design.txt
32 config HAVE_DYNAMIC_FTRACE
35 See Documentation/trace/ftrace-design.txt
37 config HAVE_DYNAMIC_FTRACE_WITH_REGS
40 config HAVE_FTRACE_MCOUNT_RECORD
43 See Documentation/trace/ftrace-design.txt
45 config HAVE_SYSCALL_TRACEPOINTS
48 See Documentation/trace/ftrace-design.txt
53 Arch supports the gcc options -pg with -mfentry
55 config HAVE_C_RECORDMCOUNT
58 C version of recordmcount available?
60 config TRACER_MAX_TRACE
71 config FTRACE_NMI_ENTER
73 depends on HAVE_FTRACE_NMI_ENTER
77 select CONTEXT_SWITCH_TRACER
80 config GPU_TRACEPOINTS
83 config CONTEXT_SWITCH_TRACER
86 config RING_BUFFER_ALLOW_SWAP
89 Allow the use of ring_buffer_swap_cpu.
90 Adds a very slight overhead to tracing when enabled.
92 # All tracer options should select GENERIC_TRACER. For those options that are
93 # enabled by all tracers (context switch and event tracer) they select TRACING.
94 # This allows those options to appear when no other tracer is selected. But the
95 # options do not appear when something else selects it. We need the two options
96 # GENERIC_TRACER and TRACING to avoid circular dependencies to accomplish the
97 # hiding of the automatic options.
103 select STACKTRACE if STACKTRACE_SUPPORT
110 config GENERIC_TRACER
115 # Minimum requirements an architecture has to meet for us to
116 # be able to offer generic tracing facilities:
118 config TRACING_SUPPORT
120 # PPC32 has no irqflags tracing support, but it can use most of the
121 # tracers anyway, they were tested to build and work. Note that new
122 # exceptions to this list aren't welcomed, better implement the
123 # irqflags tracing for your architecture.
124 depends on TRACE_IRQFLAGS_SUPPORT || PPC32
125 depends on STACKTRACE_SUPPORT
132 default y if DEBUG_KERNEL
134 Enable the kernel tracing infrastructure.
138 config FUNCTION_TRACER
139 bool "Kernel Function Tracer"
140 depends on HAVE_FUNCTION_TRACER
142 select GENERIC_TRACER
143 select CONTEXT_SWITCH_TRACER
145 Enable the kernel to trace every kernel function. This is done
146 by using a compiler feature to insert a small, 5-byte No-Operation
147 instruction at the beginning of every kernel function, which NOP
148 sequence is then dynamically patched into a tracer call when
149 tracing is enabled by the administrator. If it's runtime disabled
150 (the bootup default), then the overhead of the instructions is very
151 small and not measurable even in micro-benchmarks.
153 config FUNCTION_GRAPH_TRACER
154 bool "Kernel Function Graph Tracer"
155 depends on HAVE_FUNCTION_GRAPH_TRACER
156 depends on FUNCTION_TRACER
157 depends on !X86_32 || !CC_OPTIMIZE_FOR_SIZE
160 Enable the kernel to trace a function at both its return
162 Its first purpose is to trace the duration of functions and
163 draw a call graph for each thread with some information like
164 the return value. This is done by setting the current return
165 address on the current task structure into a stack of calls.
168 config IRQSOFF_TRACER
169 bool "Interrupts-off Latency Tracer"
171 depends on TRACE_IRQFLAGS_SUPPORT
172 depends on !ARCH_USES_GETTIMEOFFSET
173 select TRACE_IRQFLAGS
174 select GENERIC_TRACER
175 select TRACER_MAX_TRACE
176 select RING_BUFFER_ALLOW_SWAP
177 select TRACER_SNAPSHOT
178 select TRACER_SNAPSHOT_PER_CPU_SWAP
180 This option measures the time spent in irqs-off critical
181 sections, with microsecond accuracy.
183 The default measurement method is a maximum search, which is
184 disabled by default and can be runtime (re-)started
187 echo 0 > /sys/kernel/debug/tracing/tracing_max_latency
189 (Note that kernel size and overhead increase with this option
190 enabled. This option and the preempt-off timing option can be
191 used together or separately.)
193 config PREEMPT_TRACER
194 bool "Preemption-off Latency Tracer"
196 depends on !ARCH_USES_GETTIMEOFFSET
198 select GENERIC_TRACER
199 select TRACER_MAX_TRACE
200 select RING_BUFFER_ALLOW_SWAP
201 select TRACER_SNAPSHOT
202 select TRACER_SNAPSHOT_PER_CPU_SWAP
204 This option measures the time spent in preemption-off critical
205 sections, with microsecond accuracy.
207 The default measurement method is a maximum search, which is
208 disabled by default and can be runtime (re-)started
211 echo 0 > /sys/kernel/debug/tracing/tracing_max_latency
213 (Note that kernel size and overhead increase with this option
214 enabled. This option and the irqs-off timing option can be
215 used together or separately.)
218 bool "Scheduling Latency Tracer"
219 select GENERIC_TRACER
220 select CONTEXT_SWITCH_TRACER
221 select TRACER_MAX_TRACE
222 select TRACER_SNAPSHOT
224 This tracer tracks the latency of the highest priority task
225 to be scheduled in, starting from the point it has woken up.
227 config ENABLE_DEFAULT_TRACERS
228 bool "Trace process context switches and events"
229 depends on !GENERIC_TRACER
232 This tracer hooks to various trace points in the kernel,
233 allowing the user to pick and choose which trace point they
234 want to trace. It also includes the sched_switch tracer plugin.
236 config FTRACE_SYSCALLS
237 bool "Trace syscalls"
238 depends on HAVE_SYSCALL_TRACEPOINTS
239 select GENERIC_TRACER
242 Basic tracer to catch the syscall entry and exit events.
244 config TRACER_SNAPSHOT
245 bool "Create a snapshot trace buffer"
246 select TRACER_MAX_TRACE
248 Allow tracing users to take snapshot of the current buffer using the
249 ftrace interface, e.g.:
251 echo 1 > /sys/kernel/debug/tracing/snapshot
254 config TRACER_SNAPSHOT_PER_CPU_SWAP
255 bool "Allow snapshot to swap per CPU"
256 depends on TRACER_SNAPSHOT
257 select RING_BUFFER_ALLOW_SWAP
259 Allow doing a snapshot of a single CPU buffer instead of a
260 full swap (all buffers). If this is set, then the following is
263 echo 1 > /sys/kernel/debug/tracing/per_cpu/cpu2/snapshot
265 After which, only the tracing buffer for CPU 2 was swapped with
266 the main tracing buffer, and the other CPU buffers remain the same.
268 When this is enabled, this adds a little more overhead to the
269 trace recording, as it needs to add some checks to synchronize
270 recording with swaps. But this does not affect the performance
271 of the overall system. This is enabled by default when the preempt
272 or irq latency tracers are enabled, as those need to swap as well
273 and already adds the overhead (plus a lot more).
275 config TRACE_BRANCH_PROFILING
277 select GENERIC_TRACER
280 prompt "Branch Profiling"
281 default BRANCH_PROFILE_NONE
283 The branch profiling is a software profiler. It will add hooks
284 into the C conditionals to test which path a branch takes.
286 The likely/unlikely profiler only looks at the conditions that
287 are annotated with a likely or unlikely macro.
289 The "all branch" profiler will profile every if-statement in the
290 kernel. This profiler will also enable the likely/unlikely
293 Either of the above profilers adds a bit of overhead to the system.
294 If unsure, choose "No branch profiling".
296 config BRANCH_PROFILE_NONE
297 bool "No branch profiling"
299 No branch profiling. Branch profiling adds a bit of overhead.
300 Only enable it if you want to analyse the branching behavior.
301 Otherwise keep it disabled.
303 config PROFILE_ANNOTATED_BRANCHES
304 bool "Trace likely/unlikely profiler"
305 select TRACE_BRANCH_PROFILING
307 This tracer profiles all likely and unlikely macros
308 in the kernel. It will display the results in:
310 /sys/kernel/debug/tracing/trace_stat/branch_annotated
312 Note: this will add a significant overhead; only turn this
313 on if you need to profile the system's use of these macros.
315 config PROFILE_ALL_BRANCHES
316 bool "Profile all if conditionals"
317 select TRACE_BRANCH_PROFILING
319 This tracer profiles all branch conditions. Every if ()
320 taken in the kernel is recorded whether it hit or miss.
321 The results will be displayed in:
323 /sys/kernel/debug/tracing/trace_stat/branch_all
325 This option also enables the likely/unlikely profiler.
327 This configuration, when enabled, will impose a great overhead
328 on the system. This should only be enabled when the system
329 is to be analyzed in much detail.
332 config TRACING_BRANCHES
335 Selected by tracers that will trace the likely and unlikely
336 conditions. This prevents the tracers themselves from being
337 profiled. Profiling the tracing infrastructure can only happen
338 when the likelys and unlikelys are not being traced.
341 bool "Trace likely/unlikely instances"
342 depends on TRACE_BRANCH_PROFILING
343 select TRACING_BRANCHES
345 This traces the events of likely and unlikely condition
346 calls in the kernel. The difference between this and the
347 "Trace likely/unlikely profiler" is that this is not a
348 histogram of the callers, but actually places the calling
349 events into a running trace buffer to see when and where the
350 events happened, as well as their results.
355 bool "Trace max stack"
356 depends on HAVE_FUNCTION_TRACER
357 select FUNCTION_TRACER
361 This special tracer records the maximum stack footprint of the
362 kernel and displays it in /sys/kernel/debug/tracing/stack_trace.
364 This tracer works by hooking into every function call that the
365 kernel executes, and keeping a maximum stack depth value and
366 stack-trace saved. If this is configured with DYNAMIC_FTRACE
367 then it will not have any overhead while the stack tracer
370 To enable the stack tracer on bootup, pass in 'stacktrace'
371 on the kernel command line.
373 The stack tracer can also be enabled or disabled via the
374 sysctl kernel.stack_tracer_enabled
378 config BLK_DEV_IO_TRACE
379 bool "Support for tracing block IO actions"
385 select GENERIC_TRACER
388 Say Y here if you want to be able to trace the block layer actions
389 on a given queue. Tracing allows you to see any traffic happening
390 on a block device queue. For more information (and the userspace
391 support tools needed), fetch the blktrace tools from:
393 git://git.kernel.dk/blktrace.git
395 Tracing also is possible using the ftrace interface, e.g.:
397 echo 1 > /sys/block/sda/sda1/trace/enable
398 echo blk > /sys/kernel/debug/tracing/current_tracer
399 cat /sys/kernel/debug/tracing/trace_pipe
405 depends on HAVE_REGS_AND_STACK_ACCESS_API
406 bool "Enable kprobes-based dynamic events"
411 This allows the user to add tracing events (similar to tracepoints)
412 on the fly via the ftrace interface. See
413 Documentation/trace/kprobetrace.txt for more details.
415 Those events can be inserted wherever kprobes can probe, and record
416 various register and memory values.
418 This option is also required by perf-probe subcommand of perf tools.
419 If you want to use perf tools, this option is strongly recommended.
422 bool "Enable uprobes-based dynamic events"
423 depends on ARCH_SUPPORTS_UPROBES
425 depends on PERF_EVENTS
431 This allows the user to add tracing events on top of userspace
432 dynamic events (similar to tracepoints) on the fly via the trace
433 events interface. Those events can be inserted wherever uprobes
434 can probe, and record various registers.
435 This option is required if you plan to use perf-probe subcommand
436 of perf tools on user space applications.
439 depends on BPF_SYSCALL
440 depends on (KPROBE_EVENT || UPROBE_EVENT) && PERF_EVENTS
444 This allows the user to attach BPF programs to kprobe events.
449 config DYNAMIC_FTRACE
450 bool "enable/disable function tracing dynamically"
451 depends on FUNCTION_TRACER
452 depends on HAVE_DYNAMIC_FTRACE
455 This option will modify all the calls to function tracing
456 dynamically (will patch them out of the binary image and
457 replace them with a No-Op instruction) on boot up. During
458 compile time, a table is made of all the locations that ftrace
459 can function trace, and this table is linked into the kernel
460 image. When this is enabled, functions can be individually
461 enabled, and the functions not enabled will not affect
462 performance of the system.
464 See the files in /sys/kernel/debug/tracing:
465 available_filter_functions
469 This way a CONFIG_FUNCTION_TRACER kernel is slightly larger, but
470 otherwise has native performance as long as no tracing is active.
472 config DYNAMIC_FTRACE_WITH_REGS
474 depends on DYNAMIC_FTRACE
475 depends on HAVE_DYNAMIC_FTRACE_WITH_REGS
477 config FUNCTION_PROFILER
478 bool "Kernel function profiler"
479 depends on FUNCTION_TRACER
482 This option enables the kernel function profiler. A file is created
483 in debugfs called function_profile_enabled which defaults to zero.
484 When a 1 is echoed into this file profiling begins, and when a
485 zero is entered, profiling stops. A "functions" file is created in
486 the trace_stats directory; this file shows the list of functions that
487 have been hit and their counters.
491 config FTRACE_MCOUNT_RECORD
493 depends on DYNAMIC_FTRACE
494 depends on HAVE_FTRACE_MCOUNT_RECORD
496 config FTRACE_SELFTEST
499 config FTRACE_STARTUP_TEST
500 bool "Perform a startup test on ftrace"
501 depends on GENERIC_TRACER
502 select FTRACE_SELFTEST
504 This option performs a series of startup tests on ftrace. On bootup
505 a series of tests are made to verify that the tracer is
506 functioning properly. It will do tests on all the configured
509 config EVENT_TRACE_TEST_SYSCALLS
510 bool "Run selftest on syscall events"
511 depends on FTRACE_STARTUP_TEST
513 This option will also enable testing every syscall event.
514 It only enables the event and disables it and runs various loads
515 with the event enabled. This adds a bit more time for kernel boot
516 up since it runs this on every system call defined.
518 TBD - enable a way to actually call the syscalls as we test their
522 bool "Memory mapped IO tracing"
523 depends on HAVE_MMIOTRACE_SUPPORT && PCI
524 select GENERIC_TRACER
526 Mmiotrace traces Memory Mapped I/O access and is meant for
527 debugging and reverse engineering. It is called from the ioremap
528 implementation and works via page faults. Tracing is disabled by
529 default and can be enabled at run-time.
531 See Documentation/trace/mmiotrace.txt.
532 If you are not helping to develop drivers, say N.
534 config MMIOTRACE_TEST
535 tristate "Test module for mmiotrace"
536 depends on MMIOTRACE && m
538 This is a dumb module for testing mmiotrace. It is very dangerous
539 as it will write garbage to IO memory starting at a given address.
540 However, it should be safe to use on e.g. unused portion of VRAM.
542 Say N, unless you absolutely know what you are doing.
544 config TRACEPOINT_BENCHMARK
545 bool "Add tracepoint that benchmarks tracepoints"
547 This option creates the tracepoint "benchmark:benchmark_event".
548 When the tracepoint is enabled, it kicks off a kernel thread that
549 goes into an infinite loop (calling cond_sched() to let other tasks
550 run), and calls the tracepoint. Each iteration will record the time
551 it took to write to the tracepoint and the next iteration that
552 data will be passed to the tracepoint itself. That is, the tracepoint
553 will report the time it took to do the previous tracepoint.
554 The string written to the tracepoint is a static string of 128 bytes
555 to keep the time the same. The initial string is simply a write of
556 "START". The second string records the cold cache time of the first
557 write which is not added to the rest of the calculations.
559 As it is a tight loop, it benchmarks as hot cache. That's fine because
560 we care most about hot paths that are probably in cache already.
562 An example of the output:
565 first=3672 [COLD CACHED]
566 last=632 first=3672 max=632 min=632 avg=316 std=446 std^2=199712
567 last=278 first=3672 max=632 min=278 avg=303 std=316 std^2=100337
568 last=277 first=3672 max=632 min=277 avg=296 std=258 std^2=67064
569 last=273 first=3672 max=632 min=273 avg=292 std=224 std^2=50411
570 last=273 first=3672 max=632 min=273 avg=288 std=200 std^2=40389
571 last=281 first=3672 max=632 min=273 avg=287 std=183 std^2=33666
574 config RING_BUFFER_BENCHMARK
575 tristate "Ring buffer benchmark stress tester"
576 depends on RING_BUFFER
578 This option creates a test to stress the ring buffer and benchmark it.
579 It creates its own ring buffer such that it will not interfere with
580 any other users of the ring buffer (such as ftrace). It then creates
581 a producer and consumer that will run for 10 seconds and sleep for
582 10 seconds. Each interval it will print out the number of events
583 it recorded and give a rough estimate of how long each iteration took.
585 It does not disable interrupts or raise its priority, so it may be
586 affected by processes that are running.
590 config RING_BUFFER_STARTUP_TEST
591 bool "Ring buffer startup self test"
592 depends on RING_BUFFER
594 Run a simple self test on the ring buffer on boot up. Late in the
595 kernel boot sequence, the test will start that kicks off
596 a thread per cpu. Each thread will write various size events
597 into the ring buffer. Another thread is created to send IPIs
598 to each of the threads, where the IPI handler will also write
599 to the ring buffer, to test/stress the nesting ability.
600 If any anomalies are discovered, a warning will be displayed
601 and all ring buffers will be disabled.
603 The test runs for 10 seconds. This will slow your boot time
604 by at least 10 more seconds.
606 At the end of the test, statics and more checks are done.
607 It will output the stats of each per cpu buffer. What
608 was written, the sizes, what was read, what was lost, and
609 other similar details.
613 config TRACE_ENUM_MAP_FILE
614 bool "Show enum mappings for trace events"
617 The "print fmt" of the trace events will show the enum names instead
618 of their values. This can cause problems for user space tools that
619 use this string to parse the raw data as user space does not know
620 how to convert the string to its value.
622 To fix this, there's a special macro in the kernel that can be used
623 to convert the enum into its value. If this macro is used, then the
624 print fmt strings will have the enums converted to their values.
626 If something does not get converted properly, this option can be
627 used to show what enums the kernel tried to convert.
629 This option is for debugging the enum conversions. A file is created
630 in the tracing directory called "enum_map" that will show the enum
631 names matched with their values and what trace event system they
634 Normally, the mapping of the strings to values will be freed after
635 boot up or module load. With this option, they will not be freed, as
636 they are needed for the "enum_map" file. Enabling this option will
637 increase the memory footprint of the running kernel.
641 config TRACING_EVENTS_GPIO
642 bool "Trace gpio events"
646 Enable tracing events for gpio subsystem
650 endif # TRACING_SUPPORT