2 # Architectures that offer an FUNCTION_TRACER implementation should
3 # select HAVE_FUNCTION_TRACER:
9 config HAVE_FUNCTION_TRACER
12 config HAVE_DYNAMIC_FTRACE
15 config HAVE_FTRACE_MCOUNT_RECORD
18 config TRACER_MAX_TRACE
34 config FUNCTION_TRACER
35 bool "Kernel Function Tracer"
36 depends on HAVE_FUNCTION_TRACER
37 depends on DEBUG_KERNEL
40 select CONTEXT_SWITCH_TRACER
42 Enable the kernel to trace every kernel function. This is done
43 by using a compiler feature to insert a small, 5-byte No-Operation
44 instruction to the beginning of every kernel function, which NOP
45 sequence is then dynamically patched into a tracer call when
46 tracing is enabled by the administrator. If it's runtime disabled
47 (the bootup default), then the overhead of the instructions is very
48 small and not measurable even in micro-benchmarks.
51 bool "Interrupts-off Latency Tracer"
53 depends on TRACE_IRQFLAGS_SUPPORT
54 depends on GENERIC_TIME
55 depends on DEBUG_KERNEL
58 select TRACER_MAX_TRACE
60 This option measures the time spent in irqs-off critical
61 sections, with microsecond accuracy.
63 The default measurement method is a maximum search, which is
64 disabled by default and can be runtime (re-)started
67 echo 0 > /debugfs/tracing/tracing_max_latency
69 (Note that kernel size and overhead increases with this option
70 enabled. This option and the preempt-off timing option can be
71 used together or separately.)
74 bool "Preemption-off Latency Tracer"
76 depends on GENERIC_TIME
78 depends on DEBUG_KERNEL
80 select TRACER_MAX_TRACE
82 This option measures the time spent in preemption off critical
83 sections, with microsecond accuracy.
85 The default measurement method is a maximum search, which is
86 disabled by default and can be runtime (re-)started
89 echo 0 > /debugfs/tracing/tracing_max_latency
91 (Note that kernel size and overhead increases with this option
92 enabled. This option and the irqs-off timing option can be
93 used together or separately.)
100 This tracer provides the trace needed by the 'Sysprof' userspace
104 bool "Scheduling Latency Tracer"
105 depends on DEBUG_KERNEL
107 select CONTEXT_SWITCH_TRACER
108 select TRACER_MAX_TRACE
110 This tracer tracks the latency of the highest priority task
111 to be scheduled in, starting from the point it has woken up.
113 config CONTEXT_SWITCH_TRACER
114 bool "Trace process context switches"
115 depends on DEBUG_KERNEL
119 This tracer gets called from the context switch and records
120 all switching of tasks.
123 bool "Trace boot initcalls"
124 depends on DEBUG_KERNEL
126 select CONTEXT_SWITCH_TRACER
128 This tracer helps developers to optimize boot times: it records
129 the timings of the initcalls and traces key events and the identity
130 of tasks that can cause boot delays, such as context-switches.
132 Its aim is to be parsed by the /scripts/bootgraph.pl tool to
133 produce pretty graphics about boot inefficiencies, giving a visual
134 representation of the delays during initcalls - but the raw
135 /debug/tracing/trace text output is readable too.
137 ( Note that tracing self tests can't be enabled if this tracer is
138 selected, because the self-tests are an initcall as well and that
139 would invalidate the boot trace. )
142 bool "Trace max stack"
143 depends on HAVE_FUNCTION_TRACER
144 depends on DEBUG_KERNEL
145 select FUNCTION_TRACER
148 This special tracer records the maximum stack footprint of the
149 kernel and displays it in debugfs/tracing/stack_trace.
151 This tracer works by hooking into every function call that the
152 kernel executes, and keeping a maximum stack depth value and
153 stack-trace saved. Because this logic has to execute in every
154 kernel function, all the time, this option can slow down the
155 kernel measurably and is generally intended for kernel
160 config DYNAMIC_FTRACE
161 bool "enable/disable ftrace tracepoints dynamically"
162 depends on FUNCTION_TRACER
163 depends on HAVE_DYNAMIC_FTRACE
164 depends on DEBUG_KERNEL
167 This option will modify all the calls to ftrace dynamically
168 (will patch them out of the binary image and replaces them
169 with a No-Op instruction) as they are called. A table is
170 created to dynamically enable them again.
172 This way a CONFIG_FUNCTION_TRACER kernel is slightly larger, but otherwise
173 has native performance as long as no tracing is active.
175 The changes to the code are done by a kernel thread that
176 wakes up once a second and checks to see if any ftrace calls
177 were made. If so, it runs stop_machine (stops all CPUS)
178 and modifies the code to jump over the call to ftrace.
180 config FTRACE_MCOUNT_RECORD
182 depends on DYNAMIC_FTRACE
183 depends on HAVE_FTRACE_MCOUNT_RECORD
185 config FTRACE_SELFTEST
188 config FTRACE_STARTUP_TEST
189 bool "Perform a startup test on ftrace"
190 depends on TRACING && DEBUG_KERNEL && !BOOT_TRACER
191 select FTRACE_SELFTEST
193 This option performs a series of startup tests on ftrace. On bootup
194 a series of tests are made to verify that the tracer is
195 functioning properly. It will do tests on all the configured