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 CONTEXT_SWITCH_TRACER
83 config RING_BUFFER_ALLOW_SWAP
86 Allow the use of ring_buffer_swap_cpu.
87 Adds a very slight overhead to tracing when enabled.
89 # All tracer options should select GENERIC_TRACER. For those options that are
90 # enabled by all tracers (context switch and event tracer) they select TRACING.
91 # This allows those options to appear when no other tracer is selected. But the
92 # options do not appear when something else selects it. We need the two options
93 # GENERIC_TRACER and TRACING to avoid circular dependencies to accomplish the
94 # hiding of the automatic options.
100 select STACKTRACE if STACKTRACE_SUPPORT
107 config GENERIC_TRACER
112 # Minimum requirements an architecture has to meet for us to
113 # be able to offer generic tracing facilities:
115 config TRACING_SUPPORT
117 # PPC32 has no irqflags tracing support, but it can use most of the
118 # tracers anyway, they were tested to build and work. Note that new
119 # exceptions to this list aren't welcomed, better implement the
120 # irqflags tracing for your architecture.
121 depends on TRACE_IRQFLAGS_SUPPORT || PPC32
122 depends on STACKTRACE_SUPPORT
129 default y if DEBUG_KERNEL
131 Enable the kernel tracing infrastructure.
135 config FUNCTION_TRACER
136 bool "Kernel Function Tracer"
137 depends on HAVE_FUNCTION_TRACER
139 select GENERIC_TRACER
140 select CONTEXT_SWITCH_TRACER
142 Enable the kernel to trace every kernel function. This is done
143 by using a compiler feature to insert a small, 5-byte No-Operation
144 instruction at the beginning of every kernel function, which NOP
145 sequence is then dynamically patched into a tracer call when
146 tracing is enabled by the administrator. If it's runtime disabled
147 (the bootup default), then the overhead of the instructions is very
148 small and not measurable even in micro-benchmarks.
150 config FUNCTION_GRAPH_TRACER
151 bool "Kernel Function Graph Tracer"
152 depends on HAVE_FUNCTION_GRAPH_TRACER
153 depends on FUNCTION_TRACER
154 depends on !X86_32 || !CC_OPTIMIZE_FOR_SIZE
157 Enable the kernel to trace a function at both its return
159 Its first purpose is to trace the duration of functions and
160 draw a call graph for each thread with some information like
161 the return value. This is done by setting the current return
162 address on the current task structure into a stack of calls.
165 config IRQSOFF_TRACER
166 bool "Interrupts-off Latency Tracer"
168 depends on TRACE_IRQFLAGS_SUPPORT
169 depends on !ARCH_USES_GETTIMEOFFSET
170 select TRACE_IRQFLAGS
171 select GENERIC_TRACER
172 select TRACER_MAX_TRACE
173 select RING_BUFFER_ALLOW_SWAP
174 select TRACER_SNAPSHOT
175 select TRACER_SNAPSHOT_PER_CPU_SWAP
177 This option measures the time spent in irqs-off critical
178 sections, with microsecond accuracy.
180 The default measurement method is a maximum search, which is
181 disabled by default and can be runtime (re-)started
184 echo 0 > /sys/kernel/debug/tracing/tracing_max_latency
186 (Note that kernel size and overhead increase with this option
187 enabled. This option and the preempt-off timing option can be
188 used together or separately.)
190 config PREEMPT_TRACER
191 bool "Preemption-off Latency Tracer"
193 depends on !ARCH_USES_GETTIMEOFFSET
195 select GENERIC_TRACER
196 select TRACER_MAX_TRACE
197 select RING_BUFFER_ALLOW_SWAP
198 select TRACER_SNAPSHOT
199 select TRACER_SNAPSHOT_PER_CPU_SWAP
201 This option measures the time spent in preemption-off critical
202 sections, with microsecond accuracy.
204 The default measurement method is a maximum search, which is
205 disabled by default and can be runtime (re-)started
208 echo 0 > /sys/kernel/debug/tracing/tracing_max_latency
210 (Note that kernel size and overhead increase with this option
211 enabled. This option and the irqs-off timing option can be
212 used together or separately.)
215 bool "Scheduling Latency Tracer"
216 select GENERIC_TRACER
217 select CONTEXT_SWITCH_TRACER
218 select TRACER_MAX_TRACE
219 select TRACER_SNAPSHOT
221 This tracer tracks the latency of the highest priority task
222 to be scheduled in, starting from the point it has woken up.
224 config ENABLE_DEFAULT_TRACERS
225 bool "Trace process context switches and events"
226 depends on !GENERIC_TRACER
229 This tracer hooks to various trace points in the kernel,
230 allowing the user to pick and choose which trace point they
231 want to trace. It also includes the sched_switch tracer plugin.
233 config FTRACE_SYSCALLS
234 bool "Trace syscalls"
235 depends on HAVE_SYSCALL_TRACEPOINTS
236 select GENERIC_TRACER
239 Basic tracer to catch the syscall entry and exit events.
241 config TRACER_SNAPSHOT
242 bool "Create a snapshot trace buffer"
243 select TRACER_MAX_TRACE
245 Allow tracing users to take snapshot of the current buffer using the
246 ftrace interface, e.g.:
248 echo 1 > /sys/kernel/debug/tracing/snapshot
251 config TRACER_SNAPSHOT_PER_CPU_SWAP
252 bool "Allow snapshot to swap per CPU"
253 depends on TRACER_SNAPSHOT
254 select RING_BUFFER_ALLOW_SWAP
256 Allow doing a snapshot of a single CPU buffer instead of a
257 full swap (all buffers). If this is set, then the following is
260 echo 1 > /sys/kernel/debug/tracing/per_cpu/cpu2/snapshot
262 After which, only the tracing buffer for CPU 2 was swapped with
263 the main tracing buffer, and the other CPU buffers remain the same.
265 When this is enabled, this adds a little more overhead to the
266 trace recording, as it needs to add some checks to synchronize
267 recording with swaps. But this does not affect the performance
268 of the overall system. This is enabled by default when the preempt
269 or irq latency tracers are enabled, as those need to swap as well
270 and already adds the overhead (plus a lot more).
272 config TRACE_BRANCH_PROFILING
274 select GENERIC_TRACER
277 prompt "Branch Profiling"
278 default BRANCH_PROFILE_NONE
280 The branch profiling is a software profiler. It will add hooks
281 into the C conditionals to test which path a branch takes.
283 The likely/unlikely profiler only looks at the conditions that
284 are annotated with a likely or unlikely macro.
286 The "all branch" profiler will profile every if-statement in the
287 kernel. This profiler will also enable the likely/unlikely
290 Either of the above profilers adds a bit of overhead to the system.
291 If unsure, choose "No branch profiling".
293 config BRANCH_PROFILE_NONE
294 bool "No branch profiling"
296 No branch profiling. Branch profiling adds a bit of overhead.
297 Only enable it if you want to analyse the branching behavior.
298 Otherwise keep it disabled.
300 config PROFILE_ANNOTATED_BRANCHES
301 bool "Trace likely/unlikely profiler"
302 select TRACE_BRANCH_PROFILING
304 This tracer profiles all likely and unlikely macros
305 in the kernel. It will display the results in:
307 /sys/kernel/debug/tracing/trace_stat/branch_annotated
309 Note: this will add a significant overhead; only turn this
310 on if you need to profile the system's use of these macros.
312 config PROFILE_ALL_BRANCHES
313 bool "Profile all if conditionals"
314 select TRACE_BRANCH_PROFILING
316 This tracer profiles all branch conditions. Every if ()
317 taken in the kernel is recorded whether it hit or miss.
318 The results will be displayed in:
320 /sys/kernel/debug/tracing/trace_stat/branch_all
322 This option also enables the likely/unlikely profiler.
324 This configuration, when enabled, will impose a great overhead
325 on the system. This should only be enabled when the system
326 is to be analyzed in much detail.
329 config TRACING_BRANCHES
332 Selected by tracers that will trace the likely and unlikely
333 conditions. This prevents the tracers themselves from being
334 profiled. Profiling the tracing infrastructure can only happen
335 when the likelys and unlikelys are not being traced.
338 bool "Trace likely/unlikely instances"
339 depends on TRACE_BRANCH_PROFILING
340 select TRACING_BRANCHES
342 This traces the events of likely and unlikely condition
343 calls in the kernel. The difference between this and the
344 "Trace likely/unlikely profiler" is that this is not a
345 histogram of the callers, but actually places the calling
346 events into a running trace buffer to see when and where the
347 events happened, as well as their results.
352 bool "Trace max stack"
353 depends on HAVE_FUNCTION_TRACER
354 select FUNCTION_TRACER
358 This special tracer records the maximum stack footprint of the
359 kernel and displays it in /sys/kernel/debug/tracing/stack_trace.
361 This tracer works by hooking into every function call that the
362 kernel executes, and keeping a maximum stack depth value and
363 stack-trace saved. If this is configured with DYNAMIC_FTRACE
364 then it will not have any overhead while the stack tracer
367 To enable the stack tracer on bootup, pass in 'stacktrace'
368 on the kernel command line.
370 The stack tracer can also be enabled or disabled via the
371 sysctl kernel.stack_tracer_enabled
375 config BLK_DEV_IO_TRACE
376 bool "Support for tracing block IO actions"
382 select GENERIC_TRACER
385 Say Y here if you want to be able to trace the block layer actions
386 on a given queue. Tracing allows you to see any traffic happening
387 on a block device queue. For more information (and the userspace
388 support tools needed), fetch the blktrace tools from:
390 git://git.kernel.dk/blktrace.git
392 Tracing also is possible using the ftrace interface, e.g.:
394 echo 1 > /sys/block/sda/sda1/trace/enable
395 echo blk > /sys/kernel/debug/tracing/current_tracer
396 cat /sys/kernel/debug/tracing/trace_pipe
402 depends on HAVE_REGS_AND_STACK_ACCESS_API
403 bool "Enable kprobes-based dynamic events"
408 This allows the user to add tracing events (similar to tracepoints)
409 on the fly via the ftrace interface. See
410 Documentation/trace/kprobetrace.txt for more details.
412 Those events can be inserted wherever kprobes can probe, and record
413 various register and memory values.
415 This option is also required by perf-probe subcommand of perf tools.
416 If you want to use perf tools, this option is strongly recommended.
419 bool "Enable uprobes-based dynamic events"
420 depends on ARCH_SUPPORTS_UPROBES
422 depends on PERF_EVENTS
428 This allows the user to add tracing events on top of userspace
429 dynamic events (similar to tracepoints) on the fly via the trace
430 events interface. Those events can be inserted wherever uprobes
431 can probe, and record various registers.
432 This option is required if you plan to use perf-probe subcommand
433 of perf tools on user space applications.
436 depends on BPF_SYSCALL
437 depends on KPROBE_EVENT
441 This allows the user to attach BPF programs to kprobe events.
446 config DYNAMIC_FTRACE
447 bool "enable/disable function tracing dynamically"
448 depends on FUNCTION_TRACER
449 depends on HAVE_DYNAMIC_FTRACE
452 This option will modify all the calls to function tracing
453 dynamically (will patch them out of the binary image and
454 replace them with a No-Op instruction) on boot up. During
455 compile time, a table is made of all the locations that ftrace
456 can function trace, and this table is linked into the kernel
457 image. When this is enabled, functions can be individually
458 enabled, and the functions not enabled will not affect
459 performance of the system.
461 See the files in /sys/kernel/debug/tracing:
462 available_filter_functions
466 This way a CONFIG_FUNCTION_TRACER kernel is slightly larger, but
467 otherwise has native performance as long as no tracing is active.
469 config DYNAMIC_FTRACE_WITH_REGS
471 depends on DYNAMIC_FTRACE
472 depends on HAVE_DYNAMIC_FTRACE_WITH_REGS
474 config FUNCTION_PROFILER
475 bool "Kernel function profiler"
476 depends on FUNCTION_TRACER
479 This option enables the kernel function profiler. A file is created
480 in debugfs called function_profile_enabled which defaults to zero.
481 When a 1 is echoed into this file profiling begins, and when a
482 zero is entered, profiling stops. A "functions" file is created in
483 the trace_stats directory; this file shows the list of functions that
484 have been hit and their counters.
488 config FTRACE_MCOUNT_RECORD
490 depends on DYNAMIC_FTRACE
491 depends on HAVE_FTRACE_MCOUNT_RECORD
493 config FTRACE_SELFTEST
496 config FTRACE_STARTUP_TEST
497 bool "Perform a startup test on ftrace"
498 depends on GENERIC_TRACER
499 select FTRACE_SELFTEST
501 This option performs a series of startup tests on ftrace. On bootup
502 a series of tests are made to verify that the tracer is
503 functioning properly. It will do tests on all the configured
506 config EVENT_TRACE_TEST_SYSCALLS
507 bool "Run selftest on syscall events"
508 depends on FTRACE_STARTUP_TEST
510 This option will also enable testing every syscall event.
511 It only enables the event and disables it and runs various loads
512 with the event enabled. This adds a bit more time for kernel boot
513 up since it runs this on every system call defined.
515 TBD - enable a way to actually call the syscalls as we test their
519 bool "Memory mapped IO tracing"
520 depends on HAVE_MMIOTRACE_SUPPORT && PCI
521 select GENERIC_TRACER
523 Mmiotrace traces Memory Mapped I/O access and is meant for
524 debugging and reverse engineering. It is called from the ioremap
525 implementation and works via page faults. Tracing is disabled by
526 default and can be enabled at run-time.
528 See Documentation/trace/mmiotrace.txt.
529 If you are not helping to develop drivers, say N.
531 config MMIOTRACE_TEST
532 tristate "Test module for mmiotrace"
533 depends on MMIOTRACE && m
535 This is a dumb module for testing mmiotrace. It is very dangerous
536 as it will write garbage to IO memory starting at a given address.
537 However, it should be safe to use on e.g. unused portion of VRAM.
539 Say N, unless you absolutely know what you are doing.
541 config TRACEPOINT_BENCHMARK
542 bool "Add tracepoint that benchmarks tracepoints"
544 This option creates the tracepoint "benchmark:benchmark_event".
545 When the tracepoint is enabled, it kicks off a kernel thread that
546 goes into an infinite loop (calling cond_sched() to let other tasks
547 run), and calls the tracepoint. Each iteration will record the time
548 it took to write to the tracepoint and the next iteration that
549 data will be passed to the tracepoint itself. That is, the tracepoint
550 will report the time it took to do the previous tracepoint.
551 The string written to the tracepoint is a static string of 128 bytes
552 to keep the time the same. The initial string is simply a write of
553 "START". The second string records the cold cache time of the first
554 write which is not added to the rest of the calculations.
556 As it is a tight loop, it benchmarks as hot cache. That's fine because
557 we care most about hot paths that are probably in cache already.
559 An example of the output:
562 first=3672 [COLD CACHED]
563 last=632 first=3672 max=632 min=632 avg=316 std=446 std^2=199712
564 last=278 first=3672 max=632 min=278 avg=303 std=316 std^2=100337
565 last=277 first=3672 max=632 min=277 avg=296 std=258 std^2=67064
566 last=273 first=3672 max=632 min=273 avg=292 std=224 std^2=50411
567 last=273 first=3672 max=632 min=273 avg=288 std=200 std^2=40389
568 last=281 first=3672 max=632 min=273 avg=287 std=183 std^2=33666
571 config RING_BUFFER_BENCHMARK
572 tristate "Ring buffer benchmark stress tester"
573 depends on RING_BUFFER
575 This option creates a test to stress the ring buffer and benchmark it.
576 It creates its own ring buffer such that it will not interfere with
577 any other users of the ring buffer (such as ftrace). It then creates
578 a producer and consumer that will run for 10 seconds and sleep for
579 10 seconds. Each interval it will print out the number of events
580 it recorded and give a rough estimate of how long each iteration took.
582 It does not disable interrupts or raise its priority, so it may be
583 affected by processes that are running.
587 config RING_BUFFER_STARTUP_TEST
588 bool "Ring buffer startup self test"
589 depends on RING_BUFFER
591 Run a simple self test on the ring buffer on boot up. Late in the
592 kernel boot sequence, the test will start that kicks off
593 a thread per cpu. Each thread will write various size events
594 into the ring buffer. Another thread is created to send IPIs
595 to each of the threads, where the IPI handler will also write
596 to the ring buffer, to test/stress the nesting ability.
597 If any anomalies are discovered, a warning will be displayed
598 and all ring buffers will be disabled.
600 The test runs for 10 seconds. This will slow your boot time
601 by at least 10 more seconds.
603 At the end of the test, statics and more checks are done.
604 It will output the stats of each per cpu buffer. What
605 was written, the sizes, what was read, what was lost, and
606 other similar details.
610 config TRACE_ENUM_MAP_FILE
611 bool "Show enum mappings for trace events"
614 The "print fmt" of the trace events will show the enum names instead
615 of their values. This can cause problems for user space tools that
616 use this string to parse the raw data as user space does not know
617 how to convert the string to its value.
619 To fix this, there's a special macro in the kernel that can be used
620 to convert the enum into its value. If this macro is used, then the
621 print fmt strings will have the enums converted to their values.
623 If something does not get converted properly, this option can be
624 used to show what enums the kernel tried to convert.
626 This option is for debugging the enum conversions. A file is created
627 in the tracing directory called "enum_map" that will show the enum
628 names matched with their values and what trace event system they
631 Normally, the mapping of the strings to values will be freed after
632 boot up or module load. With this option, they will not be freed, as
633 they are needed for the "enum_map" file. Enabling this option will
634 increase the memory footprint of the running kernel.
640 endif # TRACING_SUPPORT