diff options
-rw-r--r-- | include/linux/perf_counter.h | 2 | ||||
-rw-r--r-- | include/linux/syscalls.h | 52 | ||||
-rw-r--r-- | include/trace/syscall.h | 7 | ||||
-rw-r--r-- | kernel/trace/trace_syscalls.c | 121 |
4 files changed, 181 insertions, 1 deletions
diff --git a/include/linux/perf_counter.h b/include/linux/perf_counter.h index a9d823a93fe8..8e6460fb4c02 100644 --- a/include/linux/perf_counter.h +++ b/include/linux/perf_counter.h | |||
@@ -734,6 +734,8 @@ extern int sysctl_perf_counter_mlock; | |||
734 | extern int sysctl_perf_counter_sample_rate; | 734 | extern int sysctl_perf_counter_sample_rate; |
735 | 735 | ||
736 | extern void perf_counter_init(void); | 736 | extern void perf_counter_init(void); |
737 | extern void perf_tpcounter_event(int event_id, u64 addr, u64 count, | ||
738 | void *record, int entry_size); | ||
737 | 739 | ||
738 | #ifndef perf_misc_flags | 740 | #ifndef perf_misc_flags |
739 | #define perf_misc_flags(regs) (user_mode(regs) ? PERF_EVENT_MISC_USER : \ | 741 | #define perf_misc_flags(regs) (user_mode(regs) ? PERF_EVENT_MISC_USER : \ |
diff --git a/include/linux/syscalls.h b/include/linux/syscalls.h index ce4b01c658eb..5541e75e140a 100644 --- a/include/linux/syscalls.h +++ b/include/linux/syscalls.h | |||
@@ -98,6 +98,53 @@ struct perf_counter_attr; | |||
98 | #define __SC_TEST5(t5, a5, ...) __SC_TEST(t5); __SC_TEST4(__VA_ARGS__) | 98 | #define __SC_TEST5(t5, a5, ...) __SC_TEST(t5); __SC_TEST4(__VA_ARGS__) |
99 | #define __SC_TEST6(t6, a6, ...) __SC_TEST(t6); __SC_TEST5(__VA_ARGS__) | 99 | #define __SC_TEST6(t6, a6, ...) __SC_TEST(t6); __SC_TEST5(__VA_ARGS__) |
100 | 100 | ||
101 | #ifdef CONFIG_EVENT_PROFILE | ||
102 | #define TRACE_SYS_ENTER_PROFILE(sname) \ | ||
103 | static int prof_sysenter_enable_##sname(struct ftrace_event_call *event_call) \ | ||
104 | { \ | ||
105 | int ret = 0; \ | ||
106 | if (!atomic_inc_return(&event_enter_##sname.profile_count)) \ | ||
107 | ret = reg_prof_syscall_enter("sys"#sname); \ | ||
108 | return ret; \ | ||
109 | } \ | ||
110 | \ | ||
111 | static void prof_sysenter_disable_##sname(struct ftrace_event_call *event_call)\ | ||
112 | { \ | ||
113 | if (atomic_add_negative(-1, &event_enter_##sname.profile_count)) \ | ||
114 | unreg_prof_syscall_enter("sys"#sname); \ | ||
115 | } | ||
116 | |||
117 | #define TRACE_SYS_EXIT_PROFILE(sname) \ | ||
118 | static int prof_sysexit_enable_##sname(struct ftrace_event_call *event_call) \ | ||
119 | { \ | ||
120 | int ret = 0; \ | ||
121 | if (!atomic_inc_return(&event_exit_##sname.profile_count)) \ | ||
122 | ret = reg_prof_syscall_exit("sys"#sname); \ | ||
123 | return ret; \ | ||
124 | } \ | ||
125 | \ | ||
126 | static void prof_sysexit_disable_##sname(struct ftrace_event_call *event_call) \ | ||
127 | { \ | ||
128 | if (atomic_add_negative(-1, &event_exit_##sname.profile_count)) \ | ||
129 | unreg_prof_syscall_exit("sys"#sname); \ | ||
130 | } | ||
131 | |||
132 | #define TRACE_SYS_ENTER_PROFILE_INIT(sname) \ | ||
133 | .profile_count = ATOMIC_INIT(-1), \ | ||
134 | .profile_enable = prof_sysenter_enable_##sname, \ | ||
135 | .profile_disable = prof_sysenter_disable_##sname, | ||
136 | |||
137 | #define TRACE_SYS_EXIT_PROFILE_INIT(sname) \ | ||
138 | .profile_count = ATOMIC_INIT(-1), \ | ||
139 | .profile_enable = prof_sysexit_enable_##sname, \ | ||
140 | .profile_disable = prof_sysexit_disable_##sname, | ||
141 | #else | ||
142 | #define TRACE_SYS_ENTER_PROFILE(sname) | ||
143 | #define TRACE_SYS_ENTER_PROFILE_INIT(sname) | ||
144 | #define TRACE_SYS_EXIT_PROFILE(sname) | ||
145 | #define TRACE_SYS_EXIT_PROFILE_INIT(sname) | ||
146 | #endif | ||
147 | |||
101 | #ifdef CONFIG_FTRACE_SYSCALLS | 148 | #ifdef CONFIG_FTRACE_SYSCALLS |
102 | #define __SC_STR_ADECL1(t, a) #a | 149 | #define __SC_STR_ADECL1(t, a) #a |
103 | #define __SC_STR_ADECL2(t, a, ...) #a, __SC_STR_ADECL1(__VA_ARGS__) | 150 | #define __SC_STR_ADECL2(t, a, ...) #a, __SC_STR_ADECL1(__VA_ARGS__) |
@@ -113,7 +160,6 @@ struct perf_counter_attr; | |||
113 | #define __SC_STR_TDECL5(t, a, ...) #t, __SC_STR_TDECL4(__VA_ARGS__) | 160 | #define __SC_STR_TDECL5(t, a, ...) #t, __SC_STR_TDECL4(__VA_ARGS__) |
114 | #define __SC_STR_TDECL6(t, a, ...) #t, __SC_STR_TDECL5(__VA_ARGS__) | 161 | #define __SC_STR_TDECL6(t, a, ...) #t, __SC_STR_TDECL5(__VA_ARGS__) |
115 | 162 | ||
116 | |||
117 | #define SYSCALL_TRACE_ENTER_EVENT(sname) \ | 163 | #define SYSCALL_TRACE_ENTER_EVENT(sname) \ |
118 | static struct ftrace_event_call event_enter_##sname; \ | 164 | static struct ftrace_event_call event_enter_##sname; \ |
119 | struct trace_event enter_syscall_print_##sname = { \ | 165 | struct trace_event enter_syscall_print_##sname = { \ |
@@ -134,6 +180,7 @@ struct perf_counter_attr; | |||
134 | init_preds(&event_enter_##sname); \ | 180 | init_preds(&event_enter_##sname); \ |
135 | return 0; \ | 181 | return 0; \ |
136 | } \ | 182 | } \ |
183 | TRACE_SYS_ENTER_PROFILE(sname); \ | ||
137 | static struct ftrace_event_call __used \ | 184 | static struct ftrace_event_call __used \ |
138 | __attribute__((__aligned__(4))) \ | 185 | __attribute__((__aligned__(4))) \ |
139 | __attribute__((section("_ftrace_events"))) \ | 186 | __attribute__((section("_ftrace_events"))) \ |
@@ -145,6 +192,7 @@ struct perf_counter_attr; | |||
145 | .regfunc = reg_event_syscall_enter, \ | 192 | .regfunc = reg_event_syscall_enter, \ |
146 | .unregfunc = unreg_event_syscall_enter, \ | 193 | .unregfunc = unreg_event_syscall_enter, \ |
147 | .data = "sys"#sname, \ | 194 | .data = "sys"#sname, \ |
195 | TRACE_SYS_ENTER_PROFILE_INIT(sname) \ | ||
148 | } | 196 | } |
149 | 197 | ||
150 | #define SYSCALL_TRACE_EXIT_EVENT(sname) \ | 198 | #define SYSCALL_TRACE_EXIT_EVENT(sname) \ |
@@ -167,6 +215,7 @@ struct perf_counter_attr; | |||
167 | init_preds(&event_exit_##sname); \ | 215 | init_preds(&event_exit_##sname); \ |
168 | return 0; \ | 216 | return 0; \ |
169 | } \ | 217 | } \ |
218 | TRACE_SYS_EXIT_PROFILE(sname); \ | ||
170 | static struct ftrace_event_call __used \ | 219 | static struct ftrace_event_call __used \ |
171 | __attribute__((__aligned__(4))) \ | 220 | __attribute__((__aligned__(4))) \ |
172 | __attribute__((section("_ftrace_events"))) \ | 221 | __attribute__((section("_ftrace_events"))) \ |
@@ -178,6 +227,7 @@ struct perf_counter_attr; | |||
178 | .regfunc = reg_event_syscall_exit, \ | 227 | .regfunc = reg_event_syscall_exit, \ |
179 | .unregfunc = unreg_event_syscall_exit, \ | 228 | .unregfunc = unreg_event_syscall_exit, \ |
180 | .data = "sys"#sname, \ | 229 | .data = "sys"#sname, \ |
230 | TRACE_SYS_EXIT_PROFILE_INIT(sname) \ | ||
181 | } | 231 | } |
182 | 232 | ||
183 | #define SYSCALL_METADATA(sname, nb) \ | 233 | #define SYSCALL_METADATA(sname, nb) \ |
diff --git a/include/trace/syscall.h b/include/trace/syscall.h index df628404241a..3ab6dd18fa3a 100644 --- a/include/trace/syscall.h +++ b/include/trace/syscall.h | |||
@@ -58,5 +58,12 @@ extern void unreg_event_syscall_exit(void *ptr); | |||
58 | enum print_line_t print_syscall_enter(struct trace_iterator *iter, int flags); | 58 | enum print_line_t print_syscall_enter(struct trace_iterator *iter, int flags); |
59 | enum print_line_t print_syscall_exit(struct trace_iterator *iter, int flags); | 59 | enum print_line_t print_syscall_exit(struct trace_iterator *iter, int flags); |
60 | #endif | 60 | #endif |
61 | #ifdef CONFIG_EVENT_PROFILE | ||
62 | int reg_prof_syscall_enter(char *name); | ||
63 | void unreg_prof_syscall_enter(char *name); | ||
64 | int reg_prof_syscall_exit(char *name); | ||
65 | void unreg_prof_syscall_exit(char *name); | ||
66 | |||
67 | #endif | ||
61 | 68 | ||
62 | #endif /* _TRACE_SYSCALL_H */ | 69 | #endif /* _TRACE_SYSCALL_H */ |
diff --git a/kernel/trace/trace_syscalls.c b/kernel/trace/trace_syscalls.c index e58a9c11ba85..f4eaec3d559a 100644 --- a/kernel/trace/trace_syscalls.c +++ b/kernel/trace/trace_syscalls.c | |||
@@ -1,6 +1,7 @@ | |||
1 | #include <trace/syscall.h> | 1 | #include <trace/syscall.h> |
2 | #include <linux/kernel.h> | 2 | #include <linux/kernel.h> |
3 | #include <linux/ftrace.h> | 3 | #include <linux/ftrace.h> |
4 | #include <linux/perf_counter.h> | ||
4 | #include <asm/syscall.h> | 5 | #include <asm/syscall.h> |
5 | 6 | ||
6 | #include "trace_output.h" | 7 | #include "trace_output.h" |
@@ -252,3 +253,123 @@ struct trace_event event_syscall_enter = { | |||
252 | struct trace_event event_syscall_exit = { | 253 | struct trace_event event_syscall_exit = { |
253 | .trace = print_syscall_exit, | 254 | .trace = print_syscall_exit, |
254 | }; | 255 | }; |
256 | |||
257 | #ifdef CONFIG_EVENT_PROFILE | ||
258 | static DECLARE_BITMAP(enabled_prof_enter_syscalls, FTRACE_SYSCALL_MAX); | ||
259 | static DECLARE_BITMAP(enabled_prof_exit_syscalls, FTRACE_SYSCALL_MAX); | ||
260 | static int sys_prof_refcount_enter; | ||
261 | static int sys_prof_refcount_exit; | ||
262 | |||
263 | static void prof_syscall_enter(struct pt_regs *regs, long id) | ||
264 | { | ||
265 | struct syscall_metadata *sys_data; | ||
266 | int syscall_nr; | ||
267 | |||
268 | syscall_nr = syscall_get_nr(current, regs); | ||
269 | if (!test_bit(syscall_nr, enabled_prof_enter_syscalls)) | ||
270 | return; | ||
271 | |||
272 | sys_data = syscall_nr_to_meta(syscall_nr); | ||
273 | if (!sys_data) | ||
274 | return; | ||
275 | |||
276 | perf_tpcounter_event(sys_data->enter_id, 0, 1, NULL, 0); | ||
277 | } | ||
278 | |||
279 | int reg_prof_syscall_enter(char *name) | ||
280 | { | ||
281 | int ret = 0; | ||
282 | int num; | ||
283 | |||
284 | num = syscall_name_to_nr(name); | ||
285 | if (num < 0 || num >= FTRACE_SYSCALL_MAX) | ||
286 | return -ENOSYS; | ||
287 | |||
288 | mutex_lock(&syscall_trace_lock); | ||
289 | if (!sys_prof_refcount_enter) | ||
290 | ret = register_trace_syscall_enter(prof_syscall_enter); | ||
291 | if (ret) { | ||
292 | pr_info("event trace: Could not activate" | ||
293 | "syscall entry trace point"); | ||
294 | } else { | ||
295 | set_bit(num, enabled_prof_enter_syscalls); | ||
296 | sys_prof_refcount_enter++; | ||
297 | } | ||
298 | mutex_unlock(&syscall_trace_lock); | ||
299 | return ret; | ||
300 | } | ||
301 | |||
302 | void unreg_prof_syscall_enter(char *name) | ||
303 | { | ||
304 | int num; | ||
305 | |||
306 | num = syscall_name_to_nr(name); | ||
307 | if (num < 0 || num >= FTRACE_SYSCALL_MAX) | ||
308 | return; | ||
309 | |||
310 | mutex_lock(&syscall_trace_lock); | ||
311 | sys_prof_refcount_enter--; | ||
312 | clear_bit(num, enabled_prof_enter_syscalls); | ||
313 | if (!sys_prof_refcount_enter) | ||
314 | unregister_trace_syscall_enter(prof_syscall_enter); | ||
315 | mutex_unlock(&syscall_trace_lock); | ||
316 | } | ||
317 | |||
318 | static void prof_syscall_exit(struct pt_regs *regs, long ret) | ||
319 | { | ||
320 | struct syscall_metadata *sys_data; | ||
321 | int syscall_nr; | ||
322 | |||
323 | syscall_nr = syscall_get_nr(current, regs); | ||
324 | if (!test_bit(syscall_nr, enabled_prof_exit_syscalls)) | ||
325 | return; | ||
326 | |||
327 | sys_data = syscall_nr_to_meta(syscall_nr); | ||
328 | if (!sys_data) | ||
329 | return; | ||
330 | |||
331 | perf_tpcounter_event(sys_data->exit_id, 0, 1, NULL, 0); | ||
332 | } | ||
333 | |||
334 | int reg_prof_syscall_exit(char *name) | ||
335 | { | ||
336 | int ret = 0; | ||
337 | int num; | ||
338 | |||
339 | num = syscall_name_to_nr(name); | ||
340 | if (num < 0 || num >= FTRACE_SYSCALL_MAX) | ||
341 | return -ENOSYS; | ||
342 | |||
343 | mutex_lock(&syscall_trace_lock); | ||
344 | if (!sys_prof_refcount_exit) | ||
345 | ret = register_trace_syscall_exit(prof_syscall_exit); | ||
346 | if (ret) { | ||
347 | pr_info("event trace: Could not activate" | ||
348 | "syscall entry trace point"); | ||
349 | } else { | ||
350 | set_bit(num, enabled_prof_exit_syscalls); | ||
351 | sys_prof_refcount_exit++; | ||
352 | } | ||
353 | mutex_unlock(&syscall_trace_lock); | ||
354 | return ret; | ||
355 | } | ||
356 | |||
357 | void unreg_prof_syscall_exit(char *name) | ||
358 | { | ||
359 | int num; | ||
360 | |||
361 | num = syscall_name_to_nr(name); | ||
362 | if (num < 0 || num >= FTRACE_SYSCALL_MAX) | ||
363 | return; | ||
364 | |||
365 | mutex_lock(&syscall_trace_lock); | ||
366 | sys_prof_refcount_exit--; | ||
367 | clear_bit(num, enabled_prof_exit_syscalls); | ||
368 | if (!sys_prof_refcount_exit) | ||
369 | unregister_trace_syscall_exit(prof_syscall_exit); | ||
370 | mutex_unlock(&syscall_trace_lock); | ||
371 | } | ||
372 | |||
373 | #endif | ||
374 | |||
375 | |||