diff options
author | Frederic Weisbecker <fweisbec@gmail.com> | 2008-11-11 01:14:25 -0500 |
---|---|---|
committer | Ingo Molnar <mingo@elte.hu> | 2008-11-11 04:29:12 -0500 |
commit | 15e6cb3673ea6277999642802406a764b49391b0 (patch) | |
tree | d0a46a918caa64d5fb1ed37be3e026b30e94fe77 | |
parent | caf4b323b02a16c92fba449952ac6515ddc76d7a (diff) |
tracing: add a tracer to catch execution time of kernel functions
Impact: add new tracing plugin which can trace full (entry+exit) function calls
This tracer uses the low level function return ftrace plugin to
measure the execution time of the kernel functions.
The first field is the caller of the function, the second is the
measured function, and the last one is the execution time in
nanoseconds.
- v3:
- HAVE_FUNCTION_RET_TRACER have been added. Each arch that support ftrace return
should enable it.
- ftrace_return_stub becomes ftrace_stub.
- CONFIG_FUNCTION_RET_TRACER depends now on CONFIG_FUNCTION_TRACER
- Return traces printing can be used for other tracers on trace.c
- Adapt to the new tracing API (no more ctrl_update callback)
- Correct the check of "disabled" during insertion.
- Minor changes...
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
-rw-r--r-- | kernel/trace/Kconfig | 14 | ||||
-rw-r--r-- | kernel/trace/Makefile | 1 | ||||
-rw-r--r-- | kernel/trace/ftrace.c | 16 | ||||
-rw-r--r-- | kernel/trace/trace.c | 65 | ||||
-rw-r--r-- | kernel/trace/trace.h | 35 | ||||
-rw-r--r-- | kernel/trace/trace_functions_return.c | 82 |
6 files changed, 205 insertions, 8 deletions
diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig index fc4febc3334a..d986216c8327 100644 --- a/kernel/trace/Kconfig +++ b/kernel/trace/Kconfig | |||
@@ -9,6 +9,9 @@ config NOP_TRACER | |||
9 | config HAVE_FUNCTION_TRACER | 9 | config HAVE_FUNCTION_TRACER |
10 | bool | 10 | bool |
11 | 11 | ||
12 | config HAVE_FUNCTION_RET_TRACER | ||
13 | bool | ||
14 | |||
12 | config HAVE_FUNCTION_TRACE_MCOUNT_TEST | 15 | config HAVE_FUNCTION_TRACE_MCOUNT_TEST |
13 | bool | 16 | bool |
14 | help | 17 | help |
@@ -54,6 +57,17 @@ config FUNCTION_TRACER | |||
54 | (the bootup default), then the overhead of the instructions is very | 57 | (the bootup default), then the overhead of the instructions is very |
55 | small and not measurable even in micro-benchmarks. | 58 | small and not measurable even in micro-benchmarks. |
56 | 59 | ||
60 | config FUNCTION_RET_TRACER | ||
61 | bool "Kernel Function return Tracer" | ||
62 | depends on !DYNAMIC_FTRACE | ||
63 | depends on HAVE_FUNCTION_RET_TRACER | ||
64 | depends on FUNCTION_TRACER | ||
65 | help | ||
66 | Enable the kernel to trace a function at its return. | ||
67 | It's first purpose is to trace the duration of functions. | ||
68 | This is done by setting the current return address on the thread | ||
69 | info structure of the current task. | ||
70 | |||
57 | config IRQSOFF_TRACER | 71 | config IRQSOFF_TRACER |
58 | bool "Interrupts-off Latency Tracer" | 72 | bool "Interrupts-off Latency Tracer" |
59 | default n | 73 | default n |
diff --git a/kernel/trace/Makefile b/kernel/trace/Makefile index c8228b1a49e9..3e1f361bbc17 100644 --- a/kernel/trace/Makefile +++ b/kernel/trace/Makefile | |||
@@ -24,5 +24,6 @@ obj-$(CONFIG_NOP_TRACER) += trace_nop.o | |||
24 | obj-$(CONFIG_STACK_TRACER) += trace_stack.o | 24 | obj-$(CONFIG_STACK_TRACER) += trace_stack.o |
25 | obj-$(CONFIG_MMIOTRACE) += trace_mmiotrace.o | 25 | obj-$(CONFIG_MMIOTRACE) += trace_mmiotrace.o |
26 | obj-$(CONFIG_BOOT_TRACER) += trace_boot.o | 26 | obj-$(CONFIG_BOOT_TRACER) += trace_boot.o |
27 | obj-$(CONFIG_FUNCTION_RET_TRACER) += trace_functions_return.o | ||
27 | 28 | ||
28 | libftrace-y := ftrace.o | 29 | libftrace-y := ftrace.o |
diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c index 4d2e751bfb11..f03fe74ecd67 100644 --- a/kernel/trace/ftrace.c +++ b/kernel/trace/ftrace.c | |||
@@ -1484,3 +1484,19 @@ ftrace_enable_sysctl(struct ctl_table *table, int write, | |||
1484 | return ret; | 1484 | return ret; |
1485 | } | 1485 | } |
1486 | 1486 | ||
1487 | #ifdef CONFIG_FUNCTION_RET_TRACER | ||
1488 | trace_function_return_t ftrace_function_return = | ||
1489 | (trace_function_return_t)ftrace_stub; | ||
1490 | void register_ftrace_return(trace_function_return_t func) | ||
1491 | { | ||
1492 | ftrace_function_return = func; | ||
1493 | } | ||
1494 | |||
1495 | void unregister_ftrace_return(void) | ||
1496 | { | ||
1497 | ftrace_function_return = (trace_function_return_t)ftrace_stub; | ||
1498 | } | ||
1499 | #endif | ||
1500 | |||
1501 | |||
1502 | |||
diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c index 216bbe7547a4..a3f7ae9cd8e1 100644 --- a/kernel/trace/trace.c +++ b/kernel/trace/trace.c | |||
@@ -244,13 +244,6 @@ unsigned long nsecs_to_usecs(unsigned long nsecs) | |||
244 | return nsecs / 1000; | 244 | return nsecs / 1000; |
245 | } | 245 | } |
246 | 246 | ||
247 | /* | ||
248 | * TRACE_ITER_SYM_MASK masks the options in trace_flags that | ||
249 | * control the output of kernel symbols. | ||
250 | */ | ||
251 | #define TRACE_ITER_SYM_MASK \ | ||
252 | (TRACE_ITER_PRINT_PARENT|TRACE_ITER_SYM_OFFSET|TRACE_ITER_SYM_ADDR) | ||
253 | |||
254 | /* These must match the bit postions in trace_iterator_flags */ | 247 | /* These must match the bit postions in trace_iterator_flags */ |
255 | static const char *trace_options[] = { | 248 | static const char *trace_options[] = { |
256 | "print-parent", | 249 | "print-parent", |
@@ -810,6 +803,35 @@ trace_function(struct trace_array *tr, struct trace_array_cpu *data, | |||
810 | ring_buffer_unlock_commit(tr->buffer, event, irq_flags); | 803 | ring_buffer_unlock_commit(tr->buffer, event, irq_flags); |
811 | } | 804 | } |
812 | 805 | ||
806 | #ifdef CONFIG_FUNCTION_RET_TRACER | ||
807 | static void __trace_function_return(struct trace_array *tr, | ||
808 | struct trace_array_cpu *data, | ||
809 | struct ftrace_retfunc *trace, | ||
810 | unsigned long flags, | ||
811 | int pc) | ||
812 | { | ||
813 | struct ring_buffer_event *event; | ||
814 | struct ftrace_ret_entry *entry; | ||
815 | unsigned long irq_flags; | ||
816 | |||
817 | if (unlikely(local_read(&__get_cpu_var(ftrace_cpu_disabled)))) | ||
818 | return; | ||
819 | |||
820 | event = ring_buffer_lock_reserve(global_trace.buffer, sizeof(*entry), | ||
821 | &irq_flags); | ||
822 | if (!event) | ||
823 | return; | ||
824 | entry = ring_buffer_event_data(event); | ||
825 | tracing_generic_entry_update(&entry->ent, flags, pc); | ||
826 | entry->ent.type = TRACE_FN_RET; | ||
827 | entry->ip = trace->func; | ||
828 | entry->parent_ip = trace->ret; | ||
829 | entry->rettime = trace->rettime; | ||
830 | entry->calltime = trace->calltime; | ||
831 | ring_buffer_unlock_commit(global_trace.buffer, event, irq_flags); | ||
832 | } | ||
833 | #endif | ||
834 | |||
813 | void | 835 | void |
814 | ftrace(struct trace_array *tr, struct trace_array_cpu *data, | 836 | ftrace(struct trace_array *tr, struct trace_array_cpu *data, |
815 | unsigned long ip, unsigned long parent_ip, unsigned long flags, | 837 | unsigned long ip, unsigned long parent_ip, unsigned long flags, |
@@ -1038,6 +1060,29 @@ function_trace_call(unsigned long ip, unsigned long parent_ip) | |||
1038 | raw_local_irq_restore(flags); | 1060 | raw_local_irq_restore(flags); |
1039 | } | 1061 | } |
1040 | 1062 | ||
1063 | #ifdef CONFIG_FUNCTION_RET_TRACER | ||
1064 | void trace_function_return(struct ftrace_retfunc *trace) | ||
1065 | { | ||
1066 | struct trace_array *tr = &global_trace; | ||
1067 | struct trace_array_cpu *data; | ||
1068 | unsigned long flags; | ||
1069 | long disabled; | ||
1070 | int cpu; | ||
1071 | int pc; | ||
1072 | |||
1073 | raw_local_irq_save(flags); | ||
1074 | cpu = raw_smp_processor_id(); | ||
1075 | data = tr->data[cpu]; | ||
1076 | disabled = atomic_inc_return(&data->disabled); | ||
1077 | if (likely(disabled == 1)) { | ||
1078 | pc = preempt_count(); | ||
1079 | __trace_function_return(tr, data, trace, flags, pc); | ||
1080 | } | ||
1081 | atomic_dec(&data->disabled); | ||
1082 | raw_local_irq_restore(flags); | ||
1083 | } | ||
1084 | #endif /* CONFIG_FUNCTION_RET_TRACER */ | ||
1085 | |||
1041 | static struct ftrace_ops trace_ops __read_mostly = | 1086 | static struct ftrace_ops trace_ops __read_mostly = |
1042 | { | 1087 | { |
1043 | .func = function_trace_call, | 1088 | .func = function_trace_call, |
@@ -1285,7 +1330,7 @@ seq_print_sym_offset(struct trace_seq *s, const char *fmt, | |||
1285 | # define IP_FMT "%016lx" | 1330 | # define IP_FMT "%016lx" |
1286 | #endif | 1331 | #endif |
1287 | 1332 | ||
1288 | static int | 1333 | int |
1289 | seq_print_ip_sym(struct trace_seq *s, unsigned long ip, unsigned long sym_flags) | 1334 | seq_print_ip_sym(struct trace_seq *s, unsigned long ip, unsigned long sym_flags) |
1290 | { | 1335 | { |
1291 | int ret; | 1336 | int ret; |
@@ -1738,6 +1783,10 @@ static enum print_line_t print_trace_fmt(struct trace_iterator *iter) | |||
1738 | trace_seq_print_cont(s, iter); | 1783 | trace_seq_print_cont(s, iter); |
1739 | break; | 1784 | break; |
1740 | } | 1785 | } |
1786 | case TRACE_FN_RET: { | ||
1787 | return print_return_function(iter); | ||
1788 | break; | ||
1789 | } | ||
1741 | } | 1790 | } |
1742 | return TRACE_TYPE_HANDLED; | 1791 | return TRACE_TYPE_HANDLED; |
1743 | } | 1792 | } |
diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h index 978145088fb8..e40ce0c14690 100644 --- a/kernel/trace/trace.h +++ b/kernel/trace/trace.h | |||
@@ -22,6 +22,7 @@ enum trace_type { | |||
22 | TRACE_MMIO_RW, | 22 | TRACE_MMIO_RW, |
23 | TRACE_MMIO_MAP, | 23 | TRACE_MMIO_MAP, |
24 | TRACE_BOOT, | 24 | TRACE_BOOT, |
25 | TRACE_FN_RET, | ||
25 | 26 | ||
26 | __TRACE_LAST_TYPE | 27 | __TRACE_LAST_TYPE |
27 | }; | 28 | }; |
@@ -48,6 +49,15 @@ struct ftrace_entry { | |||
48 | unsigned long ip; | 49 | unsigned long ip; |
49 | unsigned long parent_ip; | 50 | unsigned long parent_ip; |
50 | }; | 51 | }; |
52 | |||
53 | /* Function return entry */ | ||
54 | struct ftrace_ret_entry { | ||
55 | struct trace_entry ent; | ||
56 | unsigned long ip; | ||
57 | unsigned long parent_ip; | ||
58 | unsigned long long calltime; | ||
59 | unsigned long long rettime; | ||
60 | }; | ||
51 | extern struct tracer boot_tracer; | 61 | extern struct tracer boot_tracer; |
52 | 62 | ||
53 | /* | 63 | /* |
@@ -218,6 +228,7 @@ extern void __ftrace_bad_type(void); | |||
218 | IF_ASSIGN(var, ent, struct trace_mmiotrace_map, \ | 228 | IF_ASSIGN(var, ent, struct trace_mmiotrace_map, \ |
219 | TRACE_MMIO_MAP); \ | 229 | TRACE_MMIO_MAP); \ |
220 | IF_ASSIGN(var, ent, struct trace_boot, TRACE_BOOT); \ | 230 | IF_ASSIGN(var, ent, struct trace_boot, TRACE_BOOT); \ |
231 | IF_ASSIGN(var, ent, struct ftrace_ret_entry, TRACE_FN_RET); \ | ||
221 | __ftrace_bad_type(); \ | 232 | __ftrace_bad_type(); \ |
222 | } while (0) | 233 | } while (0) |
223 | 234 | ||
@@ -321,6 +332,8 @@ void trace_function(struct trace_array *tr, | |||
321 | unsigned long ip, | 332 | unsigned long ip, |
322 | unsigned long parent_ip, | 333 | unsigned long parent_ip, |
323 | unsigned long flags, int pc); | 334 | unsigned long flags, int pc); |
335 | void | ||
336 | trace_function_return(struct ftrace_retfunc *trace); | ||
324 | 337 | ||
325 | void tracing_start_cmdline_record(void); | 338 | void tracing_start_cmdline_record(void); |
326 | void tracing_stop_cmdline_record(void); | 339 | void tracing_stop_cmdline_record(void); |
@@ -393,6 +406,10 @@ extern void *head_page(struct trace_array_cpu *data); | |||
393 | extern int trace_seq_printf(struct trace_seq *s, const char *fmt, ...); | 406 | extern int trace_seq_printf(struct trace_seq *s, const char *fmt, ...); |
394 | extern void trace_seq_print_cont(struct trace_seq *s, | 407 | extern void trace_seq_print_cont(struct trace_seq *s, |
395 | struct trace_iterator *iter); | 408 | struct trace_iterator *iter); |
409 | |||
410 | extern int | ||
411 | seq_print_ip_sym(struct trace_seq *s, unsigned long ip, | ||
412 | unsigned long sym_flags); | ||
396 | extern ssize_t trace_seq_to_user(struct trace_seq *s, char __user *ubuf, | 413 | extern ssize_t trace_seq_to_user(struct trace_seq *s, char __user *ubuf, |
397 | size_t cnt); | 414 | size_t cnt); |
398 | extern long ns2usecs(cycle_t nsec); | 415 | extern long ns2usecs(cycle_t nsec); |
@@ -400,6 +417,17 @@ extern int trace_vprintk(unsigned long ip, const char *fmt, va_list args); | |||
400 | 417 | ||
401 | extern unsigned long trace_flags; | 418 | extern unsigned long trace_flags; |
402 | 419 | ||
420 | /* Standard output formatting function used for function return traces */ | ||
421 | #ifdef CONFIG_FUNCTION_RET_TRACER | ||
422 | extern enum print_line_t print_return_function(struct trace_iterator *iter); | ||
423 | #else | ||
424 | static inline enum print_line_t | ||
425 | print_return_function(struct trace_iterator *iter) | ||
426 | { | ||
427 | return TRACE_TYPE_UNHANDLED; | ||
428 | } | ||
429 | #endif | ||
430 | |||
403 | /* | 431 | /* |
404 | * trace_iterator_flags is an enumeration that defines bit | 432 | * trace_iterator_flags is an enumeration that defines bit |
405 | * positions into trace_flags that controls the output. | 433 | * positions into trace_flags that controls the output. |
@@ -422,6 +450,13 @@ enum trace_iterator_flags { | |||
422 | TRACE_ITER_PREEMPTONLY = 0x800, | 450 | TRACE_ITER_PREEMPTONLY = 0x800, |
423 | }; | 451 | }; |
424 | 452 | ||
453 | /* | ||
454 | * TRACE_ITER_SYM_MASK masks the options in trace_flags that | ||
455 | * control the output of kernel symbols. | ||
456 | */ | ||
457 | #define TRACE_ITER_SYM_MASK \ | ||
458 | (TRACE_ITER_PRINT_PARENT|TRACE_ITER_SYM_OFFSET|TRACE_ITER_SYM_ADDR) | ||
459 | |||
425 | extern struct tracer nop_trace; | 460 | extern struct tracer nop_trace; |
426 | 461 | ||
427 | /** | 462 | /** |
diff --git a/kernel/trace/trace_functions_return.c b/kernel/trace/trace_functions_return.c new file mode 100644 index 000000000000..7680b21537dd --- /dev/null +++ b/kernel/trace/trace_functions_return.c | |||
@@ -0,0 +1,82 @@ | |||
1 | /* | ||
2 | * | ||
3 | * Function return tracer. | ||
4 | * Copyright (c) 2008 Frederic Weisbecker <fweisbec@gmail.com> | ||
5 | * Mostly borrowed from function tracer which | ||
6 | * is Copyright (c) Steven Rostedt <srostedt@redhat.com> | ||
7 | * | ||
8 | */ | ||
9 | #include <linux/debugfs.h> | ||
10 | #include <linux/uaccess.h> | ||
11 | #include <linux/ftrace.h> | ||
12 | #include <linux/fs.h> | ||
13 | |||
14 | #include "trace.h" | ||
15 | |||
16 | |||
17 | static void start_return_trace(struct trace_array *tr) | ||
18 | { | ||
19 | register_ftrace_return(&trace_function_return); | ||
20 | } | ||
21 | |||
22 | static void stop_return_trace(struct trace_array *tr) | ||
23 | { | ||
24 | unregister_ftrace_return(); | ||
25 | } | ||
26 | |||
27 | static void return_trace_init(struct trace_array *tr) | ||
28 | { | ||
29 | int cpu; | ||
30 | for_each_online_cpu(cpu) | ||
31 | tracing_reset(tr, cpu); | ||
32 | |||
33 | start_return_trace(tr); | ||
34 | } | ||
35 | |||
36 | static void return_trace_reset(struct trace_array *tr) | ||
37 | { | ||
38 | stop_return_trace(tr); | ||
39 | } | ||
40 | |||
41 | |||
42 | enum print_line_t | ||
43 | print_return_function(struct trace_iterator *iter) | ||
44 | { | ||
45 | struct trace_seq *s = &iter->seq; | ||
46 | struct trace_entry *entry = iter->ent; | ||
47 | struct ftrace_ret_entry *field; | ||
48 | int ret; | ||
49 | |||
50 | if (entry->type == TRACE_FN_RET) { | ||
51 | trace_assign_type(field, entry); | ||
52 | ret = trace_seq_printf(s, "%pF -> ", (void *)field->parent_ip); | ||
53 | if (!ret) | ||
54 | return TRACE_TYPE_PARTIAL_LINE; | ||
55 | ret = seq_print_ip_sym(s, field->ip, | ||
56 | trace_flags & TRACE_ITER_SYM_MASK); | ||
57 | if (!ret) | ||
58 | return TRACE_TYPE_PARTIAL_LINE; | ||
59 | ret = trace_seq_printf(s, " (%llu ns)\n", | ||
60 | field->rettime - field->calltime); | ||
61 | if (!ret) | ||
62 | return TRACE_TYPE_PARTIAL_LINE; | ||
63 | else | ||
64 | return TRACE_TYPE_HANDLED; | ||
65 | } | ||
66 | return TRACE_TYPE_UNHANDLED; | ||
67 | } | ||
68 | |||
69 | static struct tracer return_trace __read_mostly = | ||
70 | { | ||
71 | .name = "return", | ||
72 | .init = return_trace_init, | ||
73 | .reset = return_trace_reset, | ||
74 | .print_line = print_return_function | ||
75 | }; | ||
76 | |||
77 | static __init int init_return_trace(void) | ||
78 | { | ||
79 | return register_tracer(&return_trace); | ||
80 | } | ||
81 | |||
82 | device_initcall(init_return_trace); | ||