aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2017-05-03 21:41:21 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2017-05-03 21:41:21 -0400
commit4c174688ee92805aa5df6e06e5b625a3286e415c (patch)
tree78e18b242b31a3a50eda41bfdd4705e07f13647a
parent9c35baf6cee9a5745d55de6f9995916dde642517 (diff)
parent73a757e63114dfd765f1c5d1ff7e994f123d0234 (diff)
Merge tag 'trace-v4.12' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace
Pull tracing updates from Steven Rostedt: "New features for this release: - Pretty much a full rewrite of the processing of function plugins. i.e. echo do_IRQ:stacktrace > set_ftrace_filter - The rewrite was needed to add plugins to be unique to tracing instances. i.e. mkdir instance/foo; cd instances/foo; echo do_IRQ:stacktrace > set_ftrace_filter The old way was written very hacky. This removes a lot of those hacks. - New "function-fork" tracing option. When set, pids in the set_ftrace_pid will have their children added when the processes with their pids listed in the set_ftrace_pid file forks. - Exposure of "maxactive" for kretprobe in kprobe_events - Allow for builtin init functions to be traced by the function tracer (via the kernel command line). Module init function tracing will come in the next release. - Added more selftests, and have selftests also test in an instance" * tag 'trace-v4.12' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace: (60 commits) ring-buffer: Return reader page back into existing ring buffer selftests: ftrace: Allow some event trigger tests to run in an instance selftests: ftrace: Have some basic tests run in a tracing instance too selftests: ftrace: Have event tests also run in an tracing instance selftests: ftrace: Make func_event_triggers and func_traceonoff_triggers tests do instances selftests: ftrace: Allow some tests to be run in a tracing instance tracing/ftrace: Allow for instances to trigger their own stacktrace probes tracing/ftrace: Allow for the traceonoff probe be unique to instances tracing/ftrace: Enable snapshot function trigger to work with instances tracing/ftrace: Allow instances to have their own function probes tracing/ftrace: Add a better way to pass data via the probe functions ftrace: Dynamically create the probe ftrace_ops for the trace_array tracing: Pass the trace_array into ftrace_probe_ops functions tracing: Have the trace_array hold the list of registered func probes ftrace: If the hash for a probe fails to update then free what was initialized ftrace: Have the function probes call their own function ftrace: Have each function probe use its own ftrace_ops ftrace: Have unregister_ftrace_function_probe_func() return a value ftrace: Add helper function ftrace_hash_move_and_update_ops() ftrace: Remove data field from ftrace_func_probe structure ...
Diffstat
-rw-r--r--Documentation/trace/kprobetrace.txt5
-rw-r--r--arch/x86/kernel/ftrace.c8
-rw-r--r--include/linux/ftrace.h92
-rw-r--r--include/linux/init.h4
-rw-r--r--include/linux/rcupdate.h5
-rw-r--r--include/linux/ring_buffer.h2
-rw-r--r--include/linux/trace_events.h11
-rw-r--r--include/linux/tracepoint.h19
-rw-r--r--init/main.c10
-rw-r--r--kernel/rcu/tree.c62
-rw-r--r--kernel/trace/Kconfig3
-rw-r--r--kernel/trace/ftrace.c996
-rw-r--r--kernel/trace/ring_buffer.c40
-rw-r--r--kernel/trace/ring_buffer_benchmark.c2
-rw-r--r--kernel/trace/trace.c242
-rw-r--r--kernel/trace/trace.h77
-rw-r--r--kernel/trace/trace_benchmark.c14
-rw-r--r--kernel/trace/trace_events.c151
-rw-r--r--kernel/trace/trace_functions.c227
-rw-r--r--kernel/trace/trace_kprobe.c39
-rw-r--r--kernel/trace/trace_stack.c35
-rw-r--r--mm/page_alloc.c1
-rw-r--r--scripts/recordmcount.c1
-rwxr-xr-xscripts/recordmcount.pl1
-rwxr-xr-xtools/testing/selftests/ftrace/ftracetest23
-rw-r--r--tools/testing/selftests/ftrace/test.d/00basic/basic2.tc1
-rw-r--r--tools/testing/selftests/ftrace/test.d/00basic/basic3.tc1
-rw-r--r--tools/testing/selftests/ftrace/test.d/event/event-enable.tc1
-rw-r--r--tools/testing/selftests/ftrace/test.d/event/event-pid.tc1
-rw-r--r--tools/testing/selftests/ftrace/test.d/event/subsystem-enable.tc1
-rw-r--r--tools/testing/selftests/ftrace/test.d/ftrace/func_event_triggers.tc114
-rw-r--r--tools/testing/selftests/ftrace/test.d/ftrace/func_set_ftrace_file.tc132
-rw-r--r--tools/testing/selftests/ftrace/test.d/ftrace/func_traceonoff_triggers.tc172
-rw-r--r--tools/testing/selftests/ftrace/test.d/functions21
-rw-r--r--tools/testing/selftests/ftrace/test.d/kprobe/kretprobe_maxactive.tc39
-rw-r--r--tools/testing/selftests/ftrace/test.d/trigger/trigger-eventonoff.tc1
-rw-r--r--tools/testing/selftests/ftrace/test.d/trigger/trigger-filter.tc1
-rw-r--r--tools/testing/selftests/ftrace/test.d/trigger/trigger-hist-mod.tc1
-rw-r--r--tools/testing/selftests/ftrace/test.d/trigger/trigger-hist.tc1
-rw-r--r--tools/testing/selftests/ftrace/test.d/trigger/trigger-multihist.tc1
40 files changed, 1902 insertions, 656 deletions
diff --git a/Documentation/trace/kprobetrace.txt b/Documentation/trace/kprobetrace.txt
index 5ea85059db3b..1a3a3d6bc2a8 100644
--- a/Documentation/trace/kprobetrace.txt
+++ b/Documentation/trace/kprobetrace.txt
@@ -24,7 +24,7 @@ current_tracer. Instead of that, add probe points via
24Synopsis of kprobe_events 24Synopsis of kprobe_events
25------------------------- 25-------------------------
26 p[:[GRP/]EVENT] [MOD:]SYM[+offs]|MEMADDR [FETCHARGS] : Set a probe 26 p[:[GRP/]EVENT] [MOD:]SYM[+offs]|MEMADDR [FETCHARGS] : Set a probe
27 r[:[GRP/]EVENT] [MOD:]SYM[+0] [FETCHARGS] : Set a return probe 27 r[MAXACTIVE][:[GRP/]EVENT] [MOD:]SYM[+0] [FETCHARGS] : Set a return probe
28 -:[GRP/]EVENT : Clear a probe 28 -:[GRP/]EVENT : Clear a probe
29 29
30 GRP : Group name. If omitted, use "kprobes" for it. 30 GRP : Group name. If omitted, use "kprobes" for it.
@@ -33,6 +33,9 @@ Synopsis of kprobe_events
33 MOD : Module name which has given SYM. 33 MOD : Module name which has given SYM.
34 SYM[+offs] : Symbol+offset where the probe is inserted. 34 SYM[+offs] : Symbol+offset where the probe is inserted.
35 MEMADDR : Address where the probe is inserted. 35 MEMADDR : Address where the probe is inserted.
36 MAXACTIVE : Maximum number of instances of the specified function that
37 can be probed simultaneously, or 0 for the default value
38 as defined in Documentation/kprobes.txt section 1.3.1.
36 39
37 FETCHARGS : Arguments. Each probe can have up to 128 args. 40 FETCHARGS : Arguments. Each probe can have up to 128 args.
38 %REG : Fetch register REG 41 %REG : Fetch register REG
diff --git a/arch/x86/kernel/ftrace.c b/arch/x86/kernel/ftrace.c
index 5b7153540727..8ee76dce9140 100644
--- a/arch/x86/kernel/ftrace.c
+++ b/arch/x86/kernel/ftrace.c
@@ -533,7 +533,13 @@ static void do_sync_core(void *data)
533 533
534static void run_sync(void) 534static void run_sync(void)
535{ 535{
536 int enable_irqs = irqs_disabled(); 536 int enable_irqs;
537
538 /* No need to sync if there's only one CPU */
539 if (num_online_cpus() == 1)
540 return;
541
542 enable_irqs = irqs_disabled();
537 543
538 /* We may be called with interrupts disabled (on bootup). */ 544 /* We may be called with interrupts disabled (on bootup). */
539 if (enable_irqs) 545 if (enable_irqs)
diff --git a/include/linux/ftrace.h b/include/linux/ftrace.h
index 3633e8beff39..6d2a63e4ea52 100644
--- a/include/linux/ftrace.h
+++ b/include/linux/ftrace.h
@@ -42,8 +42,10 @@
42/* Main tracing buffer and events set up */ 42/* Main tracing buffer and events set up */
43#ifdef CONFIG_TRACING 43#ifdef CONFIG_TRACING
44void trace_init(void); 44void trace_init(void);
45void early_trace_init(void);
45#else 46#else
46static inline void trace_init(void) { } 47static inline void trace_init(void) { }
48static inline void early_trace_init(void) { }
47#endif 49#endif
48 50
49struct module; 51struct module;
@@ -144,6 +146,10 @@ struct ftrace_ops_hash {
144 struct ftrace_hash *filter_hash; 146 struct ftrace_hash *filter_hash;
145 struct mutex regex_lock; 147 struct mutex regex_lock;
146}; 148};
149
150void ftrace_free_init_mem(void);
151#else
152static inline void ftrace_free_init_mem(void) { }
147#endif 153#endif
148 154
149/* 155/*
@@ -260,6 +266,7 @@ static inline int ftrace_nr_registered_ops(void)
260} 266}
261static inline void clear_ftrace_function(void) { } 267static inline void clear_ftrace_function(void) { }
262static inline void ftrace_kill(void) { } 268static inline void ftrace_kill(void) { }
269static inline void ftrace_free_init_mem(void) { }
263#endif /* CONFIG_FUNCTION_TRACER */ 270#endif /* CONFIG_FUNCTION_TRACER */
264 271
265#ifdef CONFIG_STACK_TRACER 272#ifdef CONFIG_STACK_TRACER
@@ -279,15 +286,45 @@ int
279stack_trace_sysctl(struct ctl_table *table, int write, 286stack_trace_sysctl(struct ctl_table *table, int write,
280 void __user *buffer, size_t *lenp, 287 void __user *buffer, size_t *lenp,
281 loff_t *ppos); 288 loff_t *ppos);
282#endif
283 289
284struct ftrace_func_command { 290/* DO NOT MODIFY THIS VARIABLE DIRECTLY! */
285 struct list_head list; 291DECLARE_PER_CPU(int, disable_stack_tracer);
286 char *name; 292
287 int (*func)(struct ftrace_hash *hash, 293/**
288 char *func, char *cmd, 294 * stack_tracer_disable - temporarily disable the stack tracer
289 char *params, int enable); 295 *
290}; 296 * There's a few locations (namely in RCU) where stack tracing
297 * cannot be executed. This function is used to disable stack
298 * tracing during those critical sections.
299 *
300 * This function must be called with preemption or interrupts
301 * disabled and stack_tracer_enable() must be called shortly after
302 * while preemption or interrupts are still disabled.
303 */
304static inline void stack_tracer_disable(void)
305{
306 /* Preemption or interupts must be disabled */
307 if (IS_ENABLED(CONFIG_PREEMPT_DEBUG))
308 WARN_ON_ONCE(!preempt_count() || !irqs_disabled());
309 this_cpu_inc(disable_stack_tracer);
310}
311
312/**
313 * stack_tracer_enable - re-enable the stack tracer
314 *
315 * After stack_tracer_disable() is called, stack_tracer_enable()
316 * must be called shortly afterward.
317 */
318static inline void stack_tracer_enable(void)
319{
320 if (IS_ENABLED(CONFIG_PREEMPT_DEBUG))
321 WARN_ON_ONCE(!preempt_count() || !irqs_disabled());
322 this_cpu_dec(disable_stack_tracer);
323}
324#else
325static inline void stack_tracer_disable(void) { }
326static inline void stack_tracer_enable(void) { }
327#endif
291 328
292#ifdef CONFIG_DYNAMIC_FTRACE 329#ifdef CONFIG_DYNAMIC_FTRACE
293 330
@@ -315,30 +352,6 @@ void ftrace_bug(int err, struct dyn_ftrace *rec);
315 352
316struct seq_file; 353struct seq_file;
317 354
318struct ftrace_probe_ops {
319 void (*func)(unsigned long ip,
320 unsigned long parent_ip,
321 void **data);
322 int (*init)(struct ftrace_probe_ops *ops,
323 unsigned long ip, void **data);
324 void (*free)(struct ftrace_probe_ops *ops,
325 unsigned long ip, void **data);
326 int (*print)(struct seq_file *m,
327 unsigned long ip,
328 struct ftrace_probe_ops *ops,
329 void *data);
330};
331
332extern int
333register_ftrace_function_probe(char *glob, struct ftrace_probe_ops *ops,
334 void *data);
335extern void
336unregister_ftrace_function_probe(char *glob, struct ftrace_probe_ops *ops,
337 void *data);
338extern void
339unregister_ftrace_function_probe_func(char *glob, struct ftrace_probe_ops *ops);
340extern void unregister_ftrace_function_probe_all(char *glob);
341
342extern int ftrace_text_reserved(const void *start, const void *end); 355extern int ftrace_text_reserved(const void *start, const void *end);
343 356
344extern int ftrace_nr_registered_ops(void); 357extern int ftrace_nr_registered_ops(void);
@@ -400,9 +413,6 @@ void ftrace_set_global_notrace(unsigned char *buf, int len, int reset);
400void ftrace_free_filter(struct ftrace_ops *ops); 413void ftrace_free_filter(struct ftrace_ops *ops);
401void ftrace_ops_set_global_filter(struct ftrace_ops *ops); 414void ftrace_ops_set_global_filter(struct ftrace_ops *ops);
402 415
403int register_ftrace_command(struct ftrace_func_command *cmd);
404int unregister_ftrace_command(struct ftrace_func_command *cmd);
405
406enum { 416enum {
407 FTRACE_UPDATE_CALLS = (1 << 0), 417 FTRACE_UPDATE_CALLS = (1 << 0),
408 FTRACE_DISABLE_CALLS = (1 << 1), 418 FTRACE_DISABLE_CALLS = (1 << 1),
@@ -433,8 +443,8 @@ enum {
433 FTRACE_ITER_FILTER = (1 << 0), 443 FTRACE_ITER_FILTER = (1 << 0),
434 FTRACE_ITER_NOTRACE = (1 << 1), 444 FTRACE_ITER_NOTRACE = (1 << 1),
435 FTRACE_ITER_PRINTALL = (1 << 2), 445 FTRACE_ITER_PRINTALL = (1 << 2),
436 FTRACE_ITER_DO_HASH = (1 << 3), 446 FTRACE_ITER_DO_PROBES = (1 << 3),
437 FTRACE_ITER_HASH = (1 << 4), 447 FTRACE_ITER_PROBE = (1 << 4),
438 FTRACE_ITER_ENABLED = (1 << 5), 448 FTRACE_ITER_ENABLED = (1 << 5),
439}; 449};
440 450
@@ -618,14 +628,6 @@ static inline void ftrace_enable_daemon(void) { }
618static inline void ftrace_module_init(struct module *mod) { } 628static inline void ftrace_module_init(struct module *mod) { }
619static inline void ftrace_module_enable(struct module *mod) { } 629static inline void ftrace_module_enable(struct module *mod) { }
620static inline void ftrace_release_mod(struct module *mod) { } 630static inline void ftrace_release_mod(struct module *mod) { }
621static inline __init int register_ftrace_command(struct ftrace_func_command *cmd)
622{
623 return -EINVAL;
624}
625static inline __init int unregister_ftrace_command(char *cmd_name)
626{
627 return -EINVAL;
628}
629static inline int ftrace_text_reserved(const void *start, const void *end) 631static inline int ftrace_text_reserved(const void *start, const void *end)
630{ 632{
631 return 0; 633 return 0;
diff --git a/include/linux/init.h b/include/linux/init.h
index 79af0962fd52..94769d687cf0 100644
--- a/include/linux/init.h
+++ b/include/linux/init.h
@@ -39,7 +39,7 @@
39 39
40/* These are for everybody (although not all archs will actually 40/* These are for everybody (although not all archs will actually
41 discard it in modules) */ 41 discard it in modules) */
42#define __init __section(.init.text) __cold notrace __latent_entropy 42#define __init __section(.init.text) __cold __inittrace __latent_entropy
43#define __initdata __section(.init.data) 43#define __initdata __section(.init.data)
44#define __initconst __section(.init.rodata) 44#define __initconst __section(.init.rodata)
45#define __exitdata __section(.exit.data) 45#define __exitdata __section(.exit.data)
@@ -68,8 +68,10 @@
68 68
69#ifdef MODULE 69#ifdef MODULE
70#define __exitused 70#define __exitused
71#define __inittrace notrace
71#else 72#else
72#define __exitused __used 73#define __exitused __used
74#define __inittrace
73#endif 75#endif
74 76
75#define __exit __section(.exit.text) __exitused __cold notrace 77#define __exit __section(.exit.text) __exitused __cold notrace
diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
index de88b33c0974..dea8f17b2fe3 100644
--- a/include/linux/rcupdate.h
+++ b/include/linux/rcupdate.h
@@ -97,6 +97,7 @@ void do_trace_rcu_torture_read(const char *rcutorturename,
97 unsigned long secs, 97 unsigned long secs,
98 unsigned long c_old, 98 unsigned long c_old,
99 unsigned long c); 99 unsigned long c);
100bool rcu_irq_enter_disabled(void);
100#else 101#else
101static inline void rcutorture_get_gp_data(enum rcutorture_type test_type, 102static inline void rcutorture_get_gp_data(enum rcutorture_type test_type,
102 int *flags, 103 int *flags,
@@ -113,6 +114,10 @@ static inline void rcutorture_record_test_transition(void)
113static inline void rcutorture_record_progress(unsigned long vernum) 114static inline void rcutorture_record_progress(unsigned long vernum)
114{ 115{
115} 116}
117static inline bool rcu_irq_enter_disabled(void)
118{
119 return false;
120}
116#ifdef CONFIG_RCU_TRACE 121#ifdef CONFIG_RCU_TRACE
117void do_trace_rcu_torture_read(const char *rcutorturename, 122void do_trace_rcu_torture_read(const char *rcutorturename,
118 struct rcu_head *rhp, 123 struct rcu_head *rhp,
diff --git a/include/linux/ring_buffer.h b/include/linux/ring_buffer.h
index b6d4568795a7..ee9b461af095 100644
--- a/include/linux/ring_buffer.h
+++ b/include/linux/ring_buffer.h
@@ -185,7 +185,7 @@ size_t ring_buffer_page_len(void *page);
185 185
186 186
187void *ring_buffer_alloc_read_page(struct ring_buffer *buffer, int cpu); 187void *ring_buffer_alloc_read_page(struct ring_buffer *buffer, int cpu);
188void ring_buffer_free_read_page(struct ring_buffer *buffer, void *data); 188void ring_buffer_free_read_page(struct ring_buffer *buffer, int cpu, void *data);
189int ring_buffer_read_page(struct ring_buffer *buffer, void **data_page, 189int ring_buffer_read_page(struct ring_buffer *buffer, void **data_page,
190 size_t len, int cpu, int full); 190 size_t len, int cpu, int full);
191 191
diff --git a/include/linux/trace_events.h b/include/linux/trace_events.h
index 0af63c4381b9..a556805eff8a 100644
--- a/include/linux/trace_events.h
+++ b/include/linux/trace_events.h
@@ -138,16 +138,7 @@ enum print_line_t {
138 TRACE_TYPE_NO_CONSUME = 3 /* Handled but ask to not consume */ 138 TRACE_TYPE_NO_CONSUME = 3 /* Handled but ask to not consume */
139}; 139};
140 140
141/* 141enum print_line_t trace_handle_return(struct trace_seq *s);
142 * Several functions return TRACE_TYPE_PARTIAL_LINE if the trace_seq
143 * overflowed, and TRACE_TYPE_HANDLED otherwise. This helper function
144 * simplifies those functions and keeps them in sync.
145 */
146static inline enum print_line_t trace_handle_return(struct trace_seq *s)
147{
148 return trace_seq_has_overflowed(s) ?
149 TRACE_TYPE_PARTIAL_LINE : TRACE_TYPE_HANDLED;
150}
151 142
152void tracing_generic_entry_update(struct trace_entry *entry, 143void tracing_generic_entry_update(struct trace_entry *entry,
153 unsigned long flags, 144 unsigned long flags,
diff --git a/include/linux/tracepoint.h b/include/linux/tracepoint.h
index f72fcfe0e66a..cc48cb2ce209 100644
--- a/include/linux/tracepoint.h
+++ b/include/linux/tracepoint.h
@@ -128,7 +128,7 @@ extern void syscall_unregfunc(void);
128 * as "(void *, void)". The DECLARE_TRACE_NOARGS() will pass in just 128 * as "(void *, void)". The DECLARE_TRACE_NOARGS() will pass in just
129 * "void *data", where as the DECLARE_TRACE() will pass in "void *data, proto". 129 * "void *data", where as the DECLARE_TRACE() will pass in "void *data, proto".
130 */ 130 */
131#define __DO_TRACE(tp, proto, args, cond, prercu, postrcu) \ 131#define __DO_TRACE(tp, proto, args, cond, rcucheck) \
132 do { \ 132 do { \
133 struct tracepoint_func *it_func_ptr; \ 133 struct tracepoint_func *it_func_ptr; \
134 void *it_func; \ 134 void *it_func; \
@@ -136,7 +136,11 @@ extern void syscall_unregfunc(void);
136 \ 136 \
137 if (!(cond)) \ 137 if (!(cond)) \
138 return; \ 138 return; \
139 prercu; \ 139 if (rcucheck) { \
140 if (WARN_ON_ONCE(rcu_irq_enter_disabled())) \
141 return; \
142 rcu_irq_enter_irqson(); \
143 } \
140 rcu_read_lock_sched_notrace(); \ 144 rcu_read_lock_sched_notrace(); \
141 it_func_ptr = rcu_dereference_sched((tp)->funcs); \ 145 it_func_ptr = rcu_dereference_sched((tp)->funcs); \
142 if (it_func_ptr) { \ 146 if (it_func_ptr) { \
@@ -147,20 +151,19 @@ extern void syscall_unregfunc(void);
147 } while ((++it_func_ptr)->func); \ 151 } while ((++it_func_ptr)->func); \
148 } \ 152 } \
149 rcu_read_unlock_sched_notrace(); \ 153 rcu_read_unlock_sched_notrace(); \
150 postrcu; \ 154 if (rcucheck) \
155 rcu_irq_exit_irqson(); \
151 } while (0) 156 } while (0)
152 157
153#ifndef MODULE 158#ifndef MODULE
154#define __DECLARE_TRACE_RCU(name, proto, args, cond, data_proto, data_args) \ 159#define __DECLARE_TRACE_RCU(name, proto, args, cond, data_proto, data_args) \
155 static inline void trace_##name##_rcuidle(proto) \ 160 static inline void trace_##name##_rcuidle(proto) \
156 { \ 161 { \
157 if (static_key_false(&__tracepoint_##name.key)) \ 162 if (static_key_false(&__tracepoint_##name.key)) \
158 __DO_TRACE(&__tracepoint_##name, \ 163 __DO_TRACE(&__tracepoint_##name, \
159 TP_PROTO(data_proto), \ 164 TP_PROTO(data_proto), \
160 TP_ARGS(data_args), \ 165 TP_ARGS(data_args), \
161 TP_CONDITION(cond), \ 166 TP_CONDITION(cond), 1); \
162 rcu_irq_enter_irqson(), \
163 rcu_irq_exit_irqson()); \
164 } 167 }
165#else 168#else
166#define __DECLARE_TRACE_RCU(name, proto, args, cond, data_proto, data_args) 169#define __DECLARE_TRACE_RCU(name, proto, args, cond, data_proto, data_args)
@@ -186,7 +189,7 @@ extern void syscall_unregfunc(void);
186 __DO_TRACE(&__tracepoint_##name, \ 189 __DO_TRACE(&__tracepoint_##name, \
187 TP_PROTO(data_proto), \ 190 TP_PROTO(data_proto), \
188 TP_ARGS(data_args), \ 191 TP_ARGS(data_args), \
189 TP_CONDITION(cond),,); \ 192 TP_CONDITION(cond), 0); \
190 if (IS_ENABLED(CONFIG_LOCKDEP) && (cond)) { \ 193 if (IS_ENABLED(CONFIG_LOCKDEP) && (cond)) { \
191 rcu_read_lock_sched_notrace(); \ 194 rcu_read_lock_sched_notrace(); \
192 rcu_dereference_sched(__tracepoint_##name.funcs);\ 195 rcu_dereference_sched(__tracepoint_##name.funcs);\
diff --git a/init/main.c b/init/main.c
index b1b9dbf7622f..cc48053bb39f 100644
--- a/init/main.c
+++ b/init/main.c
@@ -545,6 +545,11 @@ asmlinkage __visible void __init start_kernel(void)
545 trap_init(); 545 trap_init();
546 mm_init(); 546 mm_init();
547 547
548 ftrace_init();
549
550 /* trace_printk can be enabled here */
551 early_trace_init();
552
548 /* 553 /*
549 * Set up the scheduler prior starting any interrupts (such as the 554 * Set up the scheduler prior starting any interrupts (such as the
550 * timer interrupt). Full topology setup happens at smp_init() 555 * timer interrupt). Full topology setup happens at smp_init()
@@ -570,7 +575,7 @@ asmlinkage __visible void __init start_kernel(void)
570 575
571 rcu_init(); 576 rcu_init();
572 577
573 /* trace_printk() and trace points may be used after this */ 578 /* Trace events are available after this */
574 trace_init(); 579 trace_init();
575 580
576 context_tracking_init(); 581 context_tracking_init();
@@ -670,8 +675,6 @@ asmlinkage __visible void __init start_kernel(void)
670 efi_free_boot_services(); 675 efi_free_boot_services();
671 } 676 }
672 677
673 ftrace_init();
674
675 /* Do the rest non-__init'ed, we're now alive */ 678 /* Do the rest non-__init'ed, we're now alive */
676 rest_init(); 679 rest_init();
677} 680}
@@ -959,6 +962,7 @@ static int __ref kernel_init(void *unused)
959 kernel_init_freeable(); 962 kernel_init_freeable();
960 /* need to finish all async __init code before freeing the memory */ 963 /* need to finish all async __init code before freeing the memory */
961 async_synchronize_full(); 964 async_synchronize_full();
965 ftrace_free_init_mem();
962 free_initmem(); 966 free_initmem();
963 mark_readonly(); 967 mark_readonly();
964 system_state = SYSTEM_RUNNING; 968 system_state = SYSTEM_RUNNING;
diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index 50fee7689e71..a6dcf3bd244f 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -57,6 +57,7 @@
57#include <linux/random.h> 57#include <linux/random.h>
58#include <linux/trace_events.h> 58#include <linux/trace_events.h>
59#include <linux/suspend.h> 59#include <linux/suspend.h>
60#include <linux/ftrace.h>
60 61
61#include "tree.h" 62#include "tree.h"
62#include "rcu.h" 63#include "rcu.h"
@@ -284,6 +285,20 @@ static DEFINE_PER_CPU(struct rcu_dynticks, rcu_dynticks) = {
284}; 285};
285 286
286/* 287/*
288 * There's a few places, currently just in the tracing infrastructure,
289 * that uses rcu_irq_enter() to make sure RCU is watching. But there's
290 * a small location where that will not even work. In those cases
291 * rcu_irq_enter_disabled() needs to be checked to make sure rcu_irq_enter()
292 * can be called.
293 */
294static DEFINE_PER_CPU(bool, disable_rcu_irq_enter);
295
296bool rcu_irq_enter_disabled(void)
297{
298 return this_cpu_read(disable_rcu_irq_enter);
299}
300
301/*
287 * Record entry into an extended quiescent state. This is only to be 302 * Record entry into an extended quiescent state. This is only to be
288 * called when not already in an extended quiescent state. 303 * called when not already in an extended quiescent state.
289 */ 304 */
@@ -771,25 +786,24 @@ cpu_needs_another_gp(struct rcu_state *rsp, struct rcu_data *rdp)
771} 786}
772 787
773/* 788/*
774 * rcu_eqs_enter_common - current CPU is moving towards extended quiescent state 789 * rcu_eqs_enter_common - current CPU is entering an extended quiescent state
775 * 790 *
776 * If the new value of the ->dynticks_nesting counter now is zero, 791 * Enter idle, doing appropriate accounting. The caller must have
777 * we really have entered idle, and must do the appropriate accounting. 792 * disabled interrupts.
778 * The caller must have disabled interrupts.
779 */ 793 */
780static void rcu_eqs_enter_common(long long oldval, bool user) 794static void rcu_eqs_enter_common(bool user)
781{ 795{
782 struct rcu_state *rsp; 796 struct rcu_state *rsp;
783 struct rcu_data *rdp; 797 struct rcu_data *rdp;
784 RCU_TRACE(struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);) 798 struct rcu_dynticks *rdtp = this_cpu_ptr(&rcu_dynticks);
785 799
786 trace_rcu_dyntick(TPS("Start"), oldval, rdtp->dynticks_nesting); 800 trace_rcu_dyntick(TPS("Start"), rdtp->dynticks_nesting, 0);
787 if (IS_ENABLED(CONFIG_RCU_EQS_DEBUG) && 801 if (IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
788 !user && !is_idle_task(current)) { 802 !user && !is_idle_task(current)) {
789 struct task_struct *idle __maybe_unused = 803 struct task_struct *idle __maybe_unused =
790 idle_task(smp_processor_id()); 804 idle_task(smp_processor_id());
791 805
792 trace_rcu_dyntick(TPS("Error on entry: not idle task"), oldval, 0); 806 trace_rcu_dyntick(TPS("Error on entry: not idle task"), rdtp->dynticks_nesting, 0);
793 rcu_ftrace_dump(DUMP_ORIG); 807 rcu_ftrace_dump(DUMP_ORIG);
794 WARN_ONCE(1, "Current pid: %d comm: %s / Idle pid: %d comm: %s", 808 WARN_ONCE(1, "Current pid: %d comm: %s / Idle pid: %d comm: %s",
795 current->pid, current->comm, 809 current->pid, current->comm,
@@ -800,7 +814,10 @@ static void rcu_eqs_enter_common(long long oldval, bool user)
800 do_nocb_deferred_wakeup(rdp); 814 do_nocb_deferred_wakeup(rdp);
801 } 815 }
802 rcu_prepare_for_idle(); 816 rcu_prepare_for_idle();
803 rcu_dynticks_eqs_enter(); 817 __this_cpu_inc(disable_rcu_irq_enter);
818 rdtp->dynticks_nesting = 0; /* Breaks tracing momentarily. */
819 rcu_dynticks_eqs_enter(); /* After this, tracing works again. */
820 __this_cpu_dec(disable_rcu_irq_enter);
804 rcu_dynticks_task_enter(); 821 rcu_dynticks_task_enter();
805 822
806 /* 823 /*
@@ -821,19 +838,15 @@ static void rcu_eqs_enter_common(long long oldval, bool user)
821 */ 838 */
822static void rcu_eqs_enter(bool user) 839static void rcu_eqs_enter(bool user)
823{ 840{
824 long long oldval;
825 struct rcu_dynticks *rdtp; 841 struct rcu_dynticks *rdtp;
826 842
827 rdtp = this_cpu_ptr(&rcu_dynticks); 843 rdtp = this_cpu_ptr(&rcu_dynticks);
828 oldval = rdtp->dynticks_nesting;
829 WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) && 844 WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
830 (oldval & DYNTICK_TASK_NEST_MASK) == 0); 845 (rdtp->dynticks_nesting & DYNTICK_TASK_NEST_MASK) == 0);
831 if ((oldval & DYNTICK_TASK_NEST_MASK) == DYNTICK_TASK_NEST_VALUE) { 846 if ((rdtp->dynticks_nesting & DYNTICK_TASK_NEST_MASK) == DYNTICK_TASK_NEST_VALUE)
832 rdtp->dynticks_nesting = 0; 847 rcu_eqs_enter_common(user);
833 rcu_eqs_enter_common(oldval, user); 848 else
834 } else {
835 rdtp->dynticks_nesting -= DYNTICK_TASK_NEST_VALUE; 849 rdtp->dynticks_nesting -= DYNTICK_TASK_NEST_VALUE;
836 }
837} 850}
838 851
839/** 852/**
@@ -892,19 +905,18 @@ void rcu_user_enter(void)
892 */ 905 */
893void rcu_irq_exit(void) 906void rcu_irq_exit(void)
894{ 907{
895 long long oldval;
896 struct rcu_dynticks *rdtp; 908 struct rcu_dynticks *rdtp;
897 909
898 RCU_LOCKDEP_WARN(!irqs_disabled(), "rcu_irq_exit() invoked with irqs enabled!!!"); 910 RCU_LOCKDEP_WARN(!irqs_disabled(), "rcu_irq_exit() invoked with irqs enabled!!!");
899 rdtp = this_cpu_ptr(&rcu_dynticks); 911 rdtp = this_cpu_ptr(&rcu_dynticks);
900 oldval = rdtp->dynticks_nesting;
901 rdtp->dynticks_nesting--;
902 WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) && 912 WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) &&
903 rdtp->dynticks_nesting < 0); 913 rdtp->dynticks_nesting < 1);
904 if (rdtp->dynticks_nesting) 914 if (rdtp->dynticks_nesting <= 1) {
905 trace_rcu_dyntick(TPS("--="), oldval, rdtp->dynticks_nesting); 915 rcu_eqs_enter_common(true);
906 else 916 } else {
907 rcu_eqs_enter_common(oldval, true); 917 trace_rcu_dyntick(TPS("--="), rdtp->dynticks_nesting, rdtp->dynticks_nesting - 1);
918 rdtp->dynticks_nesting--;
919 }
908 rcu_sysidle_enter(1); 920 rcu_sysidle_enter(1);
909} 921}
910 922
diff --git a/kernel/trace/Kconfig b/kernel/trace/Kconfig
index 9619b5768e4b..7e06f04e98fe 100644
--- a/kernel/trace/Kconfig
+++ b/kernel/trace/Kconfig
@@ -134,7 +134,8 @@ config FUNCTION_TRACER
134 select KALLSYMS 134 select KALLSYMS
135 select GENERIC_TRACER 135 select GENERIC_TRACER
136 select CONTEXT_SWITCH_TRACER 136 select CONTEXT_SWITCH_TRACER
137 select GLOB 137 select GLOB
138 select TASKS_RCU if PREEMPT
138 help 139 help
139 Enable the kernel to trace every kernel function. This is done 140 Enable the kernel to trace every kernel function. This is done
140 by using a compiler feature to insert a small, 5-byte No-Operation 141 by using a compiler feature to insert a small, 5-byte No-Operation
diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c
index dd3e91d68dc7..00077a57b746 100644
--- a/kernel/trace/ftrace.c
+++ b/kernel/trace/ftrace.c
@@ -36,6 +36,7 @@
36 36
37#include <trace/events/sched.h> 37#include <trace/events/sched.h>
38 38
39#include <asm/sections.h>
39#include <asm/setup.h> 40#include <asm/setup.h>
40 41
41#include "trace_output.h" 42#include "trace_output.h"
@@ -1095,22 +1096,20 @@ static bool update_all_ops;
1095# error Dynamic ftrace depends on MCOUNT_RECORD 1096# error Dynamic ftrace depends on MCOUNT_RECORD
1096#endif 1097#endif
1097 1098
1098static struct hlist_head ftrace_func_hash[FTRACE_FUNC_HASHSIZE] __read_mostly;
1099
1100struct ftrace_func_probe {
1101 struct hlist_node node;
1102 struct ftrace_probe_ops *ops;
1103 unsigned long flags;
1104 unsigned long ip;
1105 void *data;
1106 struct list_head free_list;
1107};
1108
1109struct ftrace_func_entry { 1099struct ftrace_func_entry {
1110 struct hlist_node hlist; 1100 struct hlist_node hlist;
1111 unsigned long ip; 1101 unsigned long ip;
1112}; 1102};
1113 1103
1104struct ftrace_func_probe {
1105 struct ftrace_probe_ops *probe_ops;
1106 struct ftrace_ops ops;
1107 struct trace_array *tr;
1108 struct list_head list;
1109 void *data;
1110 int ref;
1111};
1112
1114/* 1113/*
1115 * We make these constant because no one should touch them, 1114 * We make these constant because no one should touch them,
1116 * but they are used as the default "empty hash", to avoid allocating 1115 * but they are used as the default "empty hash", to avoid allocating
@@ -1271,7 +1270,7 @@ static void
1271remove_hash_entry(struct ftrace_hash *hash, 1270remove_hash_entry(struct ftrace_hash *hash,
1272 struct ftrace_func_entry *entry) 1271 struct ftrace_func_entry *entry)
1273{ 1272{
1274 hlist_del(&entry->hlist); 1273 hlist_del_rcu(&entry->hlist);
1275 hash->count--; 1274 hash->count--;
1276} 1275}
1277 1276
@@ -2807,18 +2806,28 @@ static int ftrace_shutdown(struct ftrace_ops *ops, int command)
2807 * callers are done before leaving this function. 2806 * callers are done before leaving this function.
2808 * The same goes for freeing the per_cpu data of the per_cpu 2807 * The same goes for freeing the per_cpu data of the per_cpu
2809 * ops. 2808 * ops.
2810 *
2811 * Again, normal synchronize_sched() is not good enough.
2812 * We need to do a hard force of sched synchronization.
2813 * This is because we use preempt_disable() to do RCU, but
2814 * the function tracers can be called where RCU is not watching
2815 * (like before user_exit()). We can not rely on the RCU
2816 * infrastructure to do the synchronization, thus we must do it
2817 * ourselves.
2818 */ 2809 */
2819 if (ops->flags & (FTRACE_OPS_FL_DYNAMIC | FTRACE_OPS_FL_PER_CPU)) { 2810 if (ops->flags & (FTRACE_OPS_FL_DYNAMIC | FTRACE_OPS_FL_PER_CPU)) {
2811 /*
2812 * We need to do a hard force of sched synchronization.
2813 * This is because we use preempt_disable() to do RCU, but
2814 * the function tracers can be called where RCU is not watching
2815 * (like before user_exit()). We can not rely on the RCU
2816 * infrastructure to do the synchronization, thus we must do it
2817 * ourselves.
2818 */
2820 schedule_on_each_cpu(ftrace_sync); 2819 schedule_on_each_cpu(ftrace_sync);
2821 2820
2821 /*
2822 * When the kernel is preeptive, tasks can be preempted
2823 * while on a ftrace trampoline. Just scheduling a task on
2824 * a CPU is not good enough to flush them. Calling
2825 * synchornize_rcu_tasks() will wait for those tasks to
2826 * execute and either schedule voluntarily or enter user space.
2827 */
2828 if (IS_ENABLED(CONFIG_PREEMPT))
2829 synchronize_rcu_tasks();
2830
2822 arch_ftrace_trampoline_free(ops); 2831 arch_ftrace_trampoline_free(ops);
2823 2832
2824 if (ops->flags & FTRACE_OPS_FL_PER_CPU) 2833 if (ops->flags & FTRACE_OPS_FL_PER_CPU)
@@ -3055,34 +3064,63 @@ struct ftrace_iterator {
3055 struct ftrace_page *pg; 3064 struct ftrace_page *pg;
3056 struct dyn_ftrace *func; 3065 struct dyn_ftrace *func;
3057 struct ftrace_func_probe *probe; 3066 struct ftrace_func_probe *probe;
3067 struct ftrace_func_entry *probe_entry;
3058 struct trace_parser parser; 3068 struct trace_parser parser;
3059 struct ftrace_hash *hash; 3069 struct ftrace_hash *hash;
3060 struct ftrace_ops *ops; 3070 struct ftrace_ops *ops;
3061 int hidx; 3071 int pidx;
3062 int idx; 3072 int idx;
3063 unsigned flags; 3073 unsigned flags;
3064}; 3074};
3065 3075
3066static void * 3076static void *
3067t_hash_next(struct seq_file *m, loff_t *pos) 3077t_probe_next(struct seq_file *m, loff_t *pos)
3068{ 3078{
3069 struct ftrace_iterator *iter = m->private; 3079 struct ftrace_iterator *iter = m->private;
3080 struct trace_array *tr = iter->ops->private;
3081 struct list_head *func_probes;
3082 struct ftrace_hash *hash;
3083 struct list_head *next;
3070 struct hlist_node *hnd = NULL; 3084 struct hlist_node *hnd = NULL;
3071 struct hlist_head *hhd; 3085 struct hlist_head *hhd;
3086 int size;
3072 3087
3073 (*pos)++; 3088 (*pos)++;
3074 iter->pos = *pos; 3089 iter->pos = *pos;
3075 3090
3076 if (iter->probe) 3091 if (!tr)
3077 hnd = &iter->probe->node;
3078 retry:
3079 if (iter->hidx >= FTRACE_FUNC_HASHSIZE)
3080 return NULL; 3092 return NULL;
3081 3093
3082 hhd = &ftrace_func_hash[iter->hidx]; 3094 func_probes = &tr->func_probes;
3095 if (list_empty(func_probes))
3096 return NULL;
3097
3098 if (!iter->probe) {
3099 next = func_probes->next;
3100 iter->probe = list_entry(next, struct ftrace_func_probe, list);
3101 }
3102
3103 if (iter->probe_entry)
3104 hnd = &iter->probe_entry->hlist;
3105
3106 hash = iter->probe->ops.func_hash->filter_hash;
3107 size = 1 << hash->size_bits;
3108
3109 retry:
3110 if (iter->pidx >= size) {
3111 if (iter->probe->list.next == func_probes)
3112 return NULL;
3113 next = iter->probe->list.next;
3114 iter->probe = list_entry(next, struct ftrace_func_probe, list);
3115 hash = iter->probe->ops.func_hash->filter_hash;
3116 size = 1 << hash->size_bits;
3117 iter->pidx = 0;
3118 }
3119
3120 hhd = &hash->buckets[iter->pidx];
3083 3121
3084 if (hlist_empty(hhd)) { 3122 if (hlist_empty(hhd)) {
3085 iter->hidx++; 3123 iter->pidx++;
3086 hnd = NULL; 3124 hnd = NULL;
3087 goto retry; 3125 goto retry;
3088 } 3126 }
@@ -3092,7 +3130,7 @@ t_hash_next(struct seq_file *m, loff_t *pos)
3092 else { 3130 else {
3093 hnd = hnd->next; 3131 hnd = hnd->next;
3094 if (!hnd) { 3132 if (!hnd) {
3095 iter->hidx++; 3133 iter->pidx++;
3096 goto retry; 3134 goto retry;
3097 } 3135 }
3098 } 3136 }
@@ -3100,26 +3138,28 @@ t_hash_next(struct seq_file *m, loff_t *pos)
3100 if (WARN_ON_ONCE(!hnd)) 3138 if (WARN_ON_ONCE(!hnd))
3101 return NULL; 3139 return NULL;
3102 3140
3103 iter->probe = hlist_entry(hnd, struct ftrace_func_probe, node); 3141 iter->probe_entry = hlist_entry(hnd, struct ftrace_func_entry, hlist);
3104 3142
3105 return iter; 3143 return iter;
3106} 3144}
3107 3145
3108static void *t_hash_start(struct seq_file *m, loff_t *pos) 3146static void *t_probe_start(struct seq_file *m, loff_t *pos)
3109{ 3147{
3110 struct ftrace_iterator *iter = m->private; 3148 struct ftrace_iterator *iter = m->private;
3111 void *p = NULL; 3149 void *p = NULL;
3112 loff_t l; 3150 loff_t l;
3113 3151
3114 if (!(iter->flags & FTRACE_ITER_DO_HASH)) 3152 if (!(iter->flags & FTRACE_ITER_DO_PROBES))
3115 return NULL; 3153 return NULL;
3116 3154
3117 if (iter->func_pos > *pos) 3155 if (iter->func_pos > *pos)
3118 return NULL; 3156 return NULL;
3119 3157
3120 iter->hidx = 0; 3158 iter->probe = NULL;
3159 iter->probe_entry = NULL;
3160 iter->pidx = 0;
3121 for (l = 0; l <= (*pos - iter->func_pos); ) { 3161 for (l = 0; l <= (*pos - iter->func_pos); ) {
3122 p = t_hash_next(m, &l); 3162 p = t_probe_next(m, &l);
3123 if (!p) 3163 if (!p)
3124 break; 3164 break;
3125 } 3165 }
@@ -3127,50 +3167,42 @@ static void *t_hash_start(struct seq_file *m, loff_t *pos)
3127 return NULL; 3167 return NULL;
3128 3168
3129 /* Only set this if we have an item */ 3169 /* Only set this if we have an item */
3130 iter->flags |= FTRACE_ITER_HASH; 3170 iter->flags |= FTRACE_ITER_PROBE;
3131 3171
3132 return iter; 3172 return iter;
3133} 3173}
3134 3174
3135static int 3175static int
3136t_hash_show(struct seq_file *m, struct ftrace_iterator *iter) 3176t_probe_show(struct seq_file *m, struct ftrace_iterator *iter)
3137{ 3177{
3138 struct ftrace_func_probe *rec; 3178 struct ftrace_func_entry *probe_entry;
3179 struct ftrace_probe_ops *probe_ops;
3180 struct ftrace_func_probe *probe;
3181
3182 probe = iter->probe;
3183 probe_entry = iter->probe_entry;
3139 3184
3140 rec = iter->probe; 3185 if (WARN_ON_ONCE(!probe || !probe_entry))
3141 if (WARN_ON_ONCE(!rec))
3142 return -EIO; 3186 return -EIO;
3143 3187
3144 if (rec->ops->print) 3188 probe_ops = probe->probe_ops;
3145 return rec->ops->print(m, rec->ip, rec->ops, rec->data);
3146 3189
3147 seq_printf(m, "%ps:%ps", (void *)rec->ip, (void *)rec->ops->func); 3190 if (probe_ops->print)
3191 return probe_ops->print(m, probe_entry->ip, probe_ops, probe->data);
3148 3192
3149 if (rec->data) 3193 seq_printf(m, "%ps:%ps\n", (void *)probe_entry->ip,
3150 seq_printf(m, ":%p", rec->data); 3194 (void *)probe_ops->func);
3151 seq_putc(m, '\n');
3152 3195
3153 return 0; 3196 return 0;
3154} 3197}
3155 3198
3156static void * 3199static void *
3157t_next(struct seq_file *m, void *v, loff_t *pos) 3200t_func_next(struct seq_file *m, loff_t *pos)
3158{ 3201{
3159 struct ftrace_iterator *iter = m->private; 3202 struct ftrace_iterator *iter = m->private;
3160 struct ftrace_ops *ops = iter->ops;
3161 struct dyn_ftrace *rec = NULL; 3203 struct dyn_ftrace *rec = NULL;
3162 3204
3163 if (unlikely(ftrace_disabled))
3164 return NULL;
3165
3166 if (iter->flags & FTRACE_ITER_HASH)
3167 return t_hash_next(m, pos);
3168
3169 (*pos)++; 3205 (*pos)++;
3170 iter->pos = iter->func_pos = *pos;
3171
3172 if (iter->flags & FTRACE_ITER_PRINTALL)
3173 return t_hash_start(m, pos);
3174 3206
3175 retry: 3207 retry:
3176 if (iter->idx >= iter->pg->index) { 3208 if (iter->idx >= iter->pg->index) {
@@ -3181,11 +3213,8 @@ t_next(struct seq_file *m, void *v, loff_t *pos)
3181 } 3213 }
3182 } else { 3214 } else {
3183 rec = &iter->pg->records[iter->idx++]; 3215 rec = &iter->pg->records[iter->idx++];
3184 if (((iter->flags & FTRACE_ITER_FILTER) && 3216 if (((iter->flags & (FTRACE_ITER_FILTER | FTRACE_ITER_NOTRACE)) &&
3185 !(ftrace_lookup_ip(ops->func_hash->filter_hash, rec->ip))) || 3217 !ftrace_lookup_ip(iter->hash, rec->ip)) ||
3186
3187 ((iter->flags & FTRACE_ITER_NOTRACE) &&
3188 !ftrace_lookup_ip(ops->func_hash->notrace_hash, rec->ip)) ||
3189 3218
3190 ((iter->flags & FTRACE_ITER_ENABLED) && 3219 ((iter->flags & FTRACE_ITER_ENABLED) &&
3191 !(rec->flags & FTRACE_FL_ENABLED))) { 3220 !(rec->flags & FTRACE_FL_ENABLED))) {
@@ -3196,24 +3225,51 @@ t_next(struct seq_file *m, void *v, loff_t *pos)
3196 } 3225 }
3197 3226
3198 if (!rec) 3227 if (!rec)
3199 return t_hash_start(m, pos); 3228 return NULL;
3200 3229
3230 iter->pos = iter->func_pos = *pos;
3201 iter->func = rec; 3231 iter->func = rec;
3202 3232
3203 return iter; 3233 return iter;
3204} 3234}
3205 3235
3236static void *
3237t_next(struct seq_file *m, void *v, loff_t *pos)
3238{
3239 struct ftrace_iterator *iter = m->private;
3240 loff_t l = *pos; /* t_hash_start() must use original pos */
3241 void *ret;
3242
3243 if (unlikely(ftrace_disabled))
3244 return NULL;
3245
3246 if (iter->flags & FTRACE_ITER_PROBE)
3247 return t_probe_next(m, pos);
3248
3249 if (iter->flags & FTRACE_ITER_PRINTALL) {
3250 /* next must increment pos, and t_probe_start does not */
3251 (*pos)++;
3252 return t_probe_start(m, &l);
3253 }
3254
3255 ret = t_func_next(m, pos);
3256
3257 if (!ret)
3258 return t_probe_start(m, &l);
3259
3260 return ret;
3261}
3262
3206static void reset_iter_read(struct ftrace_iterator *iter) 3263static void reset_iter_read(struct ftrace_iterator *iter)
3207{ 3264{
3208 iter->pos = 0; 3265 iter->pos = 0;
3209 iter->func_pos = 0; 3266 iter->func_pos = 0;
3210 iter->flags &= ~(FTRACE_ITER_PRINTALL | FTRACE_ITER_HASH); 3267 iter->flags &= ~(FTRACE_ITER_PRINTALL | FTRACE_ITER_PROBE);
3211} 3268}
3212 3269
3213static void *t_start(struct seq_file *m, loff_t *pos) 3270static void *t_start(struct seq_file *m, loff_t *pos)
3214{ 3271{
3215 struct ftrace_iterator *iter = m->private; 3272 struct ftrace_iterator *iter = m->private;
3216 struct ftrace_ops *ops = iter->ops;
3217 void *p = NULL; 3273 void *p = NULL;
3218 loff_t l; 3274 loff_t l;
3219 3275
@@ -3233,20 +3289,19 @@ static void *t_start(struct seq_file *m, loff_t *pos)
3233 * off, we can short cut and just print out that all 3289 * off, we can short cut and just print out that all
3234 * functions are enabled. 3290 * functions are enabled.
3235 */ 3291 */
3236 if ((iter->flags & FTRACE_ITER_FILTER && 3292 if ((iter->flags & (FTRACE_ITER_FILTER | FTRACE_ITER_NOTRACE)) &&
3237 ftrace_hash_empty(ops->func_hash->filter_hash)) || 3293 ftrace_hash_empty(iter->hash)) {
3238 (iter->flags & FTRACE_ITER_NOTRACE && 3294 iter->func_pos = 1; /* Account for the message */
3239 ftrace_hash_empty(ops->func_hash->notrace_hash))) {
3240 if (*pos > 0) 3295 if (*pos > 0)
3241 return t_hash_start(m, pos); 3296 return t_probe_start(m, pos);
3242 iter->flags |= FTRACE_ITER_PRINTALL; 3297 iter->flags |= FTRACE_ITER_PRINTALL;
3243 /* reset in case of seek/pread */ 3298 /* reset in case of seek/pread */
3244 iter->flags &= ~FTRACE_ITER_HASH; 3299 iter->flags &= ~FTRACE_ITER_PROBE;
3245 return iter; 3300 return iter;
3246 } 3301 }
3247 3302
3248 if (iter->flags & FTRACE_ITER_HASH) 3303 if (iter->flags & FTRACE_ITER_PROBE)
3249 return t_hash_start(m, pos); 3304 return t_probe_start(m, pos);
3250 3305
3251 /* 3306 /*
3252 * Unfortunately, we need to restart at ftrace_pages_start 3307 * Unfortunately, we need to restart at ftrace_pages_start
@@ -3256,13 +3311,13 @@ static void *t_start(struct seq_file *m, loff_t *pos)
3256 iter->pg = ftrace_pages_start; 3311 iter->pg = ftrace_pages_start;
3257 iter->idx = 0; 3312 iter->idx = 0;
3258 for (l = 0; l <= *pos; ) { 3313 for (l = 0; l <= *pos; ) {
3259 p = t_next(m, p, &l); 3314 p = t_func_next(m, &l);
3260 if (!p) 3315 if (!p)
3261 break; 3316 break;
3262 } 3317 }
3263 3318
3264 if (!p) 3319 if (!p)
3265 return t_hash_start(m, pos); 3320 return t_probe_start(m, pos);
3266 3321
3267 return iter; 3322 return iter;
3268} 3323}
@@ -3293,8 +3348,8 @@ static int t_show(struct seq_file *m, void *v)
3293 struct ftrace_iterator *iter = m->private; 3348 struct ftrace_iterator *iter = m->private;
3294 struct dyn_ftrace *rec; 3349 struct dyn_ftrace *rec;
3295 3350
3296 if (iter->flags & FTRACE_ITER_HASH) 3351 if (iter->flags & FTRACE_ITER_PROBE)
3297 return t_hash_show(m, iter); 3352 return t_probe_show(m, iter);
3298 3353
3299 if (iter->flags & FTRACE_ITER_PRINTALL) { 3354 if (iter->flags & FTRACE_ITER_PRINTALL) {
3300 if (iter->flags & FTRACE_ITER_NOTRACE) 3355 if (iter->flags & FTRACE_ITER_NOTRACE)
@@ -3355,12 +3410,13 @@ ftrace_avail_open(struct inode *inode, struct file *file)
3355 return -ENODEV; 3410 return -ENODEV;
3356 3411
3357 iter = __seq_open_private(file, &show_ftrace_seq_ops, sizeof(*iter)); 3412 iter = __seq_open_private(file, &show_ftrace_seq_ops, sizeof(*iter));
3358 if (iter) { 3413 if (!iter)
3359 iter->pg = ftrace_pages_start; 3414 return -ENOMEM;
3360 iter->ops = &global_ops;
3361 }
3362 3415
3363 return iter ? 0 : -ENOMEM; 3416 iter->pg = ftrace_pages_start;
3417 iter->ops = &global_ops;
3418
3419 return 0;
3364} 3420}
3365 3421
3366static int 3422static int
@@ -3369,13 +3425,14 @@ ftrace_enabled_open(struct inode *inode, struct file *file)
3369 struct ftrace_iterator *iter; 3425 struct ftrace_iterator *iter;
3370 3426
3371 iter = __seq_open_private(file, &show_ftrace_seq_ops, sizeof(*iter)); 3427 iter = __seq_open_private(file, &show_ftrace_seq_ops, sizeof(*iter));
3372 if (iter) { 3428 if (!iter)
3373 iter->pg = ftrace_pages_start; 3429 return -ENOMEM;
3374 iter->flags = FTRACE_ITER_ENABLED;
3375 iter->ops = &global_ops;
3376 }
3377 3430
3378 return iter ? 0 : -ENOMEM; 3431 iter->pg = ftrace_pages_start;
3432 iter->flags = FTRACE_ITER_ENABLED;
3433 iter->ops = &global_ops;
3434
3435 return 0;
3379} 3436}
3380 3437
3381/** 3438/**
@@ -3440,7 +3497,8 @@ ftrace_regex_open(struct ftrace_ops *ops, int flag,
3440 ret = -ENOMEM; 3497 ret = -ENOMEM;
3441 goto out_unlock; 3498 goto out_unlock;
3442 } 3499 }
3443 } 3500 } else
3501 iter->hash = hash;
3444 3502
3445 if (file->f_mode & FMODE_READ) { 3503 if (file->f_mode & FMODE_READ) {
3446 iter->pg = ftrace_pages_start; 3504 iter->pg = ftrace_pages_start;
@@ -3470,7 +3528,7 @@ ftrace_filter_open(struct inode *inode, struct file *file)
3470 struct ftrace_ops *ops = inode->i_private; 3528 struct ftrace_ops *ops = inode->i_private;
3471 3529
3472 return ftrace_regex_open(ops, 3530 return ftrace_regex_open(ops,
3473 FTRACE_ITER_FILTER | FTRACE_ITER_DO_HASH, 3531 FTRACE_ITER_FILTER | FTRACE_ITER_DO_PROBES,
3474 inode, file); 3532 inode, file);
3475} 3533}
3476 3534
@@ -3654,6 +3712,56 @@ ftrace_match_records(struct ftrace_hash *hash, char *buff, int len)
3654 return match_records(hash, buff, len, NULL); 3712 return match_records(hash, buff, len, NULL);
3655} 3713}
3656 3714
3715static void ftrace_ops_update_code(struct ftrace_ops *ops,
3716 struct ftrace_ops_hash *old_hash)
3717{
3718 struct ftrace_ops *op;
3719
3720 if (!ftrace_enabled)
3721 return;
3722
3723 if (ops->flags & FTRACE_OPS_FL_ENABLED) {
3724 ftrace_run_modify_code(ops, FTRACE_UPDATE_CALLS, old_hash);
3725 return;
3726 }
3727
3728 /*
3729 * If this is the shared global_ops filter, then we need to
3730 * check if there is another ops that shares it, is enabled.
3731 * If so, we still need to run the modify code.
3732 */
3733 if (ops->func_hash != &global_ops.local_hash)
3734 return;
3735
3736 do_for_each_ftrace_op(op, ftrace_ops_list) {
3737 if (op->func_hash == &global_ops.local_hash &&
3738 op->flags & FTRACE_OPS_FL_ENABLED) {
3739 ftrace_run_modify_code(op, FTRACE_UPDATE_CALLS, old_hash);
3740 /* Only need to do this once */
3741 return;
3742 }
3743 } while_for_each_ftrace_op(op);
3744}
3745
3746static int ftrace_hash_move_and_update_ops(struct ftrace_ops *ops,
3747 struct ftrace_hash **orig_hash,
3748 struct ftrace_hash *hash,
3749 int enable)
3750{
3751 struct ftrace_ops_hash old_hash_ops;
3752 struct ftrace_hash *old_hash;
3753 int ret;
3754
3755 old_hash = *orig_hash;
3756 old_hash_ops.filter_hash = ops->func_hash->filter_hash;
3757 old_hash_ops.notrace_hash = ops->func_hash->notrace_hash;
3758 ret = ftrace_hash_move(ops, enable, orig_hash, hash);
3759 if (!ret) {
3760 ftrace_ops_update_code(ops, &old_hash_ops);
3761 free_ftrace_hash_rcu(old_hash);
3762 }
3763 return ret;
3764}
3657 3765
3658/* 3766/*
3659 * We register the module command as a template to show others how 3767 * We register the module command as a template to show others how
@@ -3661,7 +3769,7 @@ ftrace_match_records(struct ftrace_hash *hash, char *buff, int len)
3661 */ 3769 */
3662 3770
3663static int 3771static int
3664ftrace_mod_callback(struct ftrace_hash *hash, 3772ftrace_mod_callback(struct trace_array *tr, struct ftrace_hash *hash,
3665 char *func, char *cmd, char *module, int enable) 3773 char *func, char *cmd, char *module, int enable)
3666{ 3774{
3667 int ret; 3775 int ret;
@@ -3695,16 +3803,11 @@ core_initcall(ftrace_mod_cmd_init);
3695static void function_trace_probe_call(unsigned long ip, unsigned long parent_ip, 3803static void function_trace_probe_call(unsigned long ip, unsigned long parent_ip,
3696 struct ftrace_ops *op, struct pt_regs *pt_regs) 3804 struct ftrace_ops *op, struct pt_regs *pt_regs)
3697{ 3805{
3698 struct ftrace_func_probe *entry; 3806 struct ftrace_probe_ops *probe_ops;
3699 struct hlist_head *hhd; 3807 struct ftrace_func_probe *probe;
3700 unsigned long key;
3701 3808
3702 key = hash_long(ip, FTRACE_HASH_BITS); 3809 probe = container_of(op, struct ftrace_func_probe, ops);
3703 3810 probe_ops = probe->probe_ops;
3704 hhd = &ftrace_func_hash[key];
3705
3706 if (hlist_empty(hhd))
3707 return;
3708 3811
3709 /* 3812 /*
3710 * Disable preemption for these calls to prevent a RCU grace 3813 * Disable preemption for these calls to prevent a RCU grace
@@ -3712,209 +3815,336 @@ static void function_trace_probe_call(unsigned long ip, unsigned long parent_ip,
3712 * on the hash. rcu_read_lock is too dangerous here. 3815 * on the hash. rcu_read_lock is too dangerous here.
3713 */ 3816 */
3714 preempt_disable_notrace(); 3817 preempt_disable_notrace();
3715 hlist_for_each_entry_rcu_notrace(entry, hhd, node) { 3818 probe_ops->func(ip, parent_ip, probe->tr, probe_ops, probe->data);
3716 if (entry->ip == ip)
3717 entry->ops->func(ip, parent_ip, &entry->data);
3718 }
3719 preempt_enable_notrace(); 3819 preempt_enable_notrace();
3720} 3820}
3721 3821
3722static struct ftrace_ops trace_probe_ops __read_mostly = 3822struct ftrace_func_map {
3723{ 3823 struct ftrace_func_entry entry;
3724 .func = function_trace_probe_call, 3824 void *data;
3725 .flags = FTRACE_OPS_FL_INITIALIZED,
3726 INIT_OPS_HASH(trace_probe_ops)
3727}; 3825};
3728 3826
3729static int ftrace_probe_registered; 3827struct ftrace_func_mapper {
3828 struct ftrace_hash hash;
3829};
3730 3830
3731static void __enable_ftrace_function_probe(struct ftrace_ops_hash *old_hash) 3831/**
3832 * allocate_ftrace_func_mapper - allocate a new ftrace_func_mapper
3833 *
3834 * Returns a ftrace_func_mapper descriptor that can be used to map ips to data.
3835 */
3836struct ftrace_func_mapper *allocate_ftrace_func_mapper(void)
3732{ 3837{
3733 int ret; 3838 struct ftrace_hash *hash;
3734 int i;
3735 3839
3736 if (ftrace_probe_registered) { 3840 /*
3737 /* still need to update the function call sites */ 3841 * The mapper is simply a ftrace_hash, but since the entries
3738 if (ftrace_enabled) 3842 * in the hash are not ftrace_func_entry type, we define it
3739 ftrace_run_modify_code(&trace_probe_ops, FTRACE_UPDATE_CALLS, 3843 * as a separate structure.
3740 old_hash); 3844 */
3741 return; 3845 hash = alloc_ftrace_hash(FTRACE_HASH_DEFAULT_BITS);
3742 } 3846 return (struct ftrace_func_mapper *)hash;
3847}
3743 3848
3744 for (i = 0; i < FTRACE_FUNC_HASHSIZE; i++) { 3849/**
3745 struct hlist_head *hhd = &ftrace_func_hash[i]; 3850 * ftrace_func_mapper_find_ip - Find some data mapped to an ip
3746 if (hhd->first) 3851 * @mapper: The mapper that has the ip maps
3747 break; 3852 * @ip: the instruction pointer to find the data for
3748 } 3853 *
3749 /* Nothing registered? */ 3854 * Returns the data mapped to @ip if found otherwise NULL. The return
3750 if (i == FTRACE_FUNC_HASHSIZE) 3855 * is actually the address of the mapper data pointer. The address is
3751 return; 3856 * returned for use cases where the data is no bigger than a long, and
3857 * the user can use the data pointer as its data instead of having to
3858 * allocate more memory for the reference.
3859 */
3860void **ftrace_func_mapper_find_ip(struct ftrace_func_mapper *mapper,
3861 unsigned long ip)
3862{
3863 struct ftrace_func_entry *entry;
3864 struct ftrace_func_map *map;
3752 3865
3753 ret = ftrace_startup(&trace_probe_ops, 0); 3866 entry = ftrace_lookup_ip(&mapper->hash, ip);
3867 if (!entry)
3868 return NULL;
3754 3869
3755 ftrace_probe_registered = 1; 3870 map = (struct ftrace_func_map *)entry;
3871 return &map->data;
3756} 3872}
3757 3873
3758static bool __disable_ftrace_function_probe(void) 3874/**
3875 * ftrace_func_mapper_add_ip - Map some data to an ip
3876 * @mapper: The mapper that has the ip maps
3877 * @ip: The instruction pointer address to map @data to
3878 * @data: The data to map to @ip
3879 *
3880 * Returns 0 on succes otherwise an error.
3881 */
3882int ftrace_func_mapper_add_ip(struct ftrace_func_mapper *mapper,
3883 unsigned long ip, void *data)
3759{ 3884{
3760 int i; 3885 struct ftrace_func_entry *entry;
3886 struct ftrace_func_map *map;
3761 3887
3762 if (!ftrace_probe_registered) 3888 entry = ftrace_lookup_ip(&mapper->hash, ip);
3763 return false; 3889 if (entry)
3890 return -EBUSY;
3764 3891
3765 for (i = 0; i < FTRACE_FUNC_HASHSIZE; i++) { 3892 map = kmalloc(sizeof(*map), GFP_KERNEL);
3766 struct hlist_head *hhd = &ftrace_func_hash[i]; 3893 if (!map)
3767 if (hhd->first) 3894 return -ENOMEM;
3768 return false;
3769 }
3770 3895
3771 /* no more funcs left */ 3896 map->entry.ip = ip;
3772 ftrace_shutdown(&trace_probe_ops, 0); 3897 map->data = data;
3773 3898
3774 ftrace_probe_registered = 0; 3899 __add_hash_entry(&mapper->hash, &map->entry);
3775 return true;
3776}
3777 3900
3901 return 0;
3902}
3778 3903
3779static void ftrace_free_entry(struct ftrace_func_probe *entry) 3904/**
3905 * ftrace_func_mapper_remove_ip - Remove an ip from the mapping
3906 * @mapper: The mapper that has the ip maps
3907 * @ip: The instruction pointer address to remove the data from
3908 *
3909 * Returns the data if it is found, otherwise NULL.
3910 * Note, if the data pointer is used as the data itself, (see
3911 * ftrace_func_mapper_find_ip(), then the return value may be meaningless,
3912 * if the data pointer was set to zero.
3913 */
3914void *ftrace_func_mapper_remove_ip(struct ftrace_func_mapper *mapper,
3915 unsigned long ip)
3780{ 3916{
3781 if (entry->ops->free) 3917 struct ftrace_func_entry *entry;
3782 entry->ops->free(entry->ops, entry->ip, &entry->data); 3918 struct ftrace_func_map *map;
3919 void *data;
3920
3921 entry = ftrace_lookup_ip(&mapper->hash, ip);
3922 if (!entry)
3923 return NULL;
3924
3925 map = (struct ftrace_func_map *)entry;
3926 data = map->data;
3927
3928 remove_hash_entry(&mapper->hash, entry);
3783 kfree(entry); 3929 kfree(entry);
3930
3931 return data;
3932}
3933
3934/**
3935 * free_ftrace_func_mapper - free a mapping of ips and data
3936 * @mapper: The mapper that has the ip maps
3937 * @free_func: A function to be called on each data item.
3938 *
3939 * This is used to free the function mapper. The @free_func is optional
3940 * and can be used if the data needs to be freed as well.
3941 */
3942void free_ftrace_func_mapper(struct ftrace_func_mapper *mapper,
3943 ftrace_mapper_func free_func)
3944{
3945 struct ftrace_func_entry *entry;
3946 struct ftrace_func_map *map;
3947 struct hlist_head *hhd;
3948 int size = 1 << mapper->hash.size_bits;
3949 int i;
3950
3951 if (free_func && mapper->hash.count) {
3952 for (i = 0; i < size; i++) {
3953 hhd = &mapper->hash.buckets[i];
3954 hlist_for_each_entry(entry, hhd, hlist) {
3955 map = (struct ftrace_func_map *)entry;
3956 free_func(map);
3957 }
3958 }
3959 }
3960 free_ftrace_hash(&mapper->hash);
3961}
3962
3963static void release_probe(struct ftrace_func_probe *probe)
3964{
3965 struct ftrace_probe_ops *probe_ops;
3966
3967 mutex_lock(&ftrace_lock);
3968
3969 WARN_ON(probe->ref <= 0);
3970
3971 /* Subtract the ref that was used to protect this instance */
3972 probe->ref--;
3973
3974 if (!probe->ref) {
3975 probe_ops = probe->probe_ops;
3976 /*
3977 * Sending zero as ip tells probe_ops to free
3978 * the probe->data itself
3979 */
3980 if (probe_ops->free)
3981 probe_ops->free(probe_ops, probe->tr, 0, probe->data);
3982 list_del(&probe->list);
3983 kfree(probe);
3984 }
3985 mutex_unlock(&ftrace_lock);
3986}
3987
3988static void acquire_probe_locked(struct ftrace_func_probe *probe)
3989{
3990 /*
3991 * Add one ref to keep it from being freed when releasing the
3992 * ftrace_lock mutex.
3993 */
3994 probe->ref++;
3784} 3995}
3785 3996
3786int 3997int
3787register_ftrace_function_probe(char *glob, struct ftrace_probe_ops *ops, 3998register_ftrace_function_probe(char *glob, struct trace_array *tr,
3788 void *data) 3999 struct ftrace_probe_ops *probe_ops,
4000 void *data)
3789{ 4001{
3790 struct ftrace_ops_hash old_hash_ops; 4002 struct ftrace_func_entry *entry;
3791 struct ftrace_func_probe *entry; 4003 struct ftrace_func_probe *probe;
3792 struct ftrace_glob func_g; 4004 struct ftrace_hash **orig_hash;
3793 struct ftrace_hash **orig_hash = &trace_probe_ops.func_hash->filter_hash; 4005 struct ftrace_hash *old_hash;
3794 struct ftrace_hash *old_hash = *orig_hash;
3795 struct ftrace_hash *hash; 4006 struct ftrace_hash *hash;
3796 struct ftrace_page *pg;
3797 struct dyn_ftrace *rec;
3798 int not;
3799 unsigned long key;
3800 int count = 0; 4007 int count = 0;
4008 int size;
3801 int ret; 4009 int ret;
4010 int i;
3802 4011
3803 func_g.type = filter_parse_regex(glob, strlen(glob), 4012 if (WARN_ON(!tr))
3804 &func_g.search, &not);
3805 func_g.len = strlen(func_g.search);
3806
3807 /* we do not support '!' for function probes */
3808 if (WARN_ON(not))
3809 return -EINVAL; 4013 return -EINVAL;
3810 4014
3811 mutex_lock(&trace_probe_ops.func_hash->regex_lock); 4015 /* We do not support '!' for function probes */
4016 if (WARN_ON(glob[0] == '!'))
4017 return -EINVAL;
3812 4018
3813 old_hash_ops.filter_hash = old_hash;
3814 /* Probes only have filters */
3815 old_hash_ops.notrace_hash = NULL;
3816 4019
3817 hash = alloc_and_copy_ftrace_hash(FTRACE_HASH_DEFAULT_BITS, old_hash); 4020 mutex_lock(&ftrace_lock);
3818 if (!hash) { 4021 /* Check if the probe_ops is already registered */
3819 count = -ENOMEM; 4022 list_for_each_entry(probe, &tr->func_probes, list) {
3820 goto out; 4023 if (probe->probe_ops == probe_ops)
4024 break;
3821 } 4025 }
3822 4026 if (&probe->list == &tr->func_probes) {
3823 if (unlikely(ftrace_disabled)) { 4027 probe = kzalloc(sizeof(*probe), GFP_KERNEL);
3824 count = -ENODEV; 4028 if (!probe) {
3825 goto out; 4029 mutex_unlock(&ftrace_lock);
4030 return -ENOMEM;
4031 }
4032 probe->probe_ops = probe_ops;
4033 probe->ops.func = function_trace_probe_call;
4034 probe->tr = tr;
4035 ftrace_ops_init(&probe->ops);
4036 list_add(&probe->list, &tr->func_probes);
3826 } 4037 }
3827 4038
3828 mutex_lock(&ftrace_lock); 4039 acquire_probe_locked(probe);
3829 4040
3830 do_for_each_ftrace_rec(pg, rec) { 4041 mutex_unlock(&ftrace_lock);
3831 4042
3832 if (rec->flags & FTRACE_FL_DISABLED) 4043 mutex_lock(&probe->ops.func_hash->regex_lock);
3833 continue;
3834 4044
3835 if (!ftrace_match_record(rec, &func_g, NULL, 0)) 4045 orig_hash = &probe->ops.func_hash->filter_hash;
3836 continue; 4046 old_hash = *orig_hash;
4047 hash = alloc_and_copy_ftrace_hash(FTRACE_HASH_DEFAULT_BITS, old_hash);
3837 4048
3838 entry = kmalloc(sizeof(*entry), GFP_KERNEL); 4049 ret = ftrace_match_records(hash, glob, strlen(glob));
3839 if (!entry) {
3840 /* If we did not process any, then return error */
3841 if (!count)
3842 count = -ENOMEM;
3843 goto out_unlock;
3844 }
3845 4050
3846 count++; 4051 /* Nothing found? */
4052 if (!ret)
4053 ret = -EINVAL;
3847 4054
3848 entry->data = data; 4055 if (ret < 0)
4056 goto out;
3849 4057
3850 /* 4058 size = 1 << hash->size_bits;
3851 * The caller might want to do something special 4059 for (i = 0; i < size; i++) {
3852 * for each function we find. We call the callback 4060 hlist_for_each_entry(entry, &hash->buckets[i], hlist) {
3853 * to give the caller an opportunity to do so. 4061 if (ftrace_lookup_ip(old_hash, entry->ip))
3854 */
3855 if (ops->init) {
3856 if (ops->init(ops, rec->ip, &entry->data) < 0) {
3857 /* caller does not like this func */
3858 kfree(entry);
3859 continue; 4062 continue;
4063 /*
4064 * The caller might want to do something special
4065 * for each function we find. We call the callback
4066 * to give the caller an opportunity to do so.
4067 */
4068 if (probe_ops->init) {
4069 ret = probe_ops->init(probe_ops, tr,
4070 entry->ip, data,
4071 &probe->data);
4072 if (ret < 0) {
4073 if (probe_ops->free && count)
4074 probe_ops->free(probe_ops, tr,
4075 0, probe->data);
4076 probe->data = NULL;
4077 goto out;
4078 }
3860 } 4079 }
4080 count++;
3861 } 4081 }
4082 }
3862 4083
3863 ret = enter_record(hash, rec, 0); 4084 mutex_lock(&ftrace_lock);
3864 if (ret < 0) {
3865 kfree(entry);
3866 count = ret;
3867 goto out_unlock;
3868 }
3869
3870 entry->ops = ops;
3871 entry->ip = rec->ip;
3872
3873 key = hash_long(entry->ip, FTRACE_HASH_BITS);
3874 hlist_add_head_rcu(&entry->node, &ftrace_func_hash[key]);
3875 4085
3876 } while_for_each_ftrace_rec(); 4086 if (!count) {
4087 /* Nothing was added? */
4088 ret = -EINVAL;
4089 goto out_unlock;
4090 }
3877 4091
3878 ret = ftrace_hash_move(&trace_probe_ops, 1, orig_hash, hash); 4092 ret = ftrace_hash_move_and_update_ops(&probe->ops, orig_hash,
4093 hash, 1);
4094 if (ret < 0)
4095 goto err_unlock;
3879 4096
3880 __enable_ftrace_function_probe(&old_hash_ops); 4097 /* One ref for each new function traced */
4098 probe->ref += count;
3881 4099
3882 if (!ret) 4100 if (!(probe->ops.flags & FTRACE_OPS_FL_ENABLED))
3883 free_ftrace_hash_rcu(old_hash); 4101 ret = ftrace_startup(&probe->ops, 0);
3884 else
3885 count = ret;
3886 4102
3887 out_unlock: 4103 out_unlock:
3888 mutex_unlock(&ftrace_lock); 4104 mutex_unlock(&ftrace_lock);
4105
4106 if (!ret)
4107 ret = count;
3889 out: 4108 out:
3890 mutex_unlock(&trace_probe_ops.func_hash->regex_lock); 4109 mutex_unlock(&probe->ops.func_hash->regex_lock);
3891 free_ftrace_hash(hash); 4110 free_ftrace_hash(hash);
3892 4111
3893 return count; 4112 release_probe(probe);
3894}
3895 4113
3896enum { 4114 return ret;
3897 PROBE_TEST_FUNC = 1,
3898 PROBE_TEST_DATA = 2
3899};
3900 4115
3901static void 4116 err_unlock:
3902__unregister_ftrace_function_probe(char *glob, struct ftrace_probe_ops *ops, 4117 if (!probe_ops->free || !count)
3903 void *data, int flags) 4118 goto out_unlock;
4119
4120 /* Failed to do the move, need to call the free functions */
4121 for (i = 0; i < size; i++) {
4122 hlist_for_each_entry(entry, &hash->buckets[i], hlist) {
4123 if (ftrace_lookup_ip(old_hash, entry->ip))
4124 continue;
4125 probe_ops->free(probe_ops, tr, entry->ip, probe->data);
4126 }
4127 }
4128 goto out_unlock;
4129}
4130
4131int
4132unregister_ftrace_function_probe_func(char *glob, struct trace_array *tr,
4133 struct ftrace_probe_ops *probe_ops)
3904{ 4134{
3905 struct ftrace_ops_hash old_hash_ops; 4135 struct ftrace_ops_hash old_hash_ops;
3906 struct ftrace_func_entry *rec_entry; 4136 struct ftrace_func_entry *entry;
3907 struct ftrace_func_probe *entry; 4137 struct ftrace_func_probe *probe;
3908 struct ftrace_func_probe *p;
3909 struct ftrace_glob func_g; 4138 struct ftrace_glob func_g;
3910 struct ftrace_hash **orig_hash = &trace_probe_ops.func_hash->filter_hash; 4139 struct ftrace_hash **orig_hash;
3911 struct ftrace_hash *old_hash = *orig_hash; 4140 struct ftrace_hash *old_hash;
3912 struct list_head free_list; 4141 struct ftrace_hash *hash = NULL;
3913 struct ftrace_hash *hash;
3914 struct hlist_node *tmp; 4142 struct hlist_node *tmp;
4143 struct hlist_head hhd;
3915 char str[KSYM_SYMBOL_LEN]; 4144 char str[KSYM_SYMBOL_LEN];
3916 int i, ret; 4145 int count = 0;
3917 bool disabled; 4146 int i, ret = -ENODEV;
4147 int size;
3918 4148
3919 if (glob && (strcmp(glob, "*") == 0 || !strlen(glob))) 4149 if (glob && (strcmp(glob, "*") == 0 || !strlen(glob)))
3920 func_g.search = NULL; 4150 func_g.search = NULL;
@@ -3928,95 +4158,104 @@ __unregister_ftrace_function_probe(char *glob, struct ftrace_probe_ops *ops,
3928 4158
3929 /* we do not support '!' for function probes */ 4159 /* we do not support '!' for function probes */
3930 if (WARN_ON(not)) 4160 if (WARN_ON(not))
3931 return; 4161 return -EINVAL;
4162 }
4163
4164 mutex_lock(&ftrace_lock);
4165 /* Check if the probe_ops is already registered */
4166 list_for_each_entry(probe, &tr->func_probes, list) {
4167 if (probe->probe_ops == probe_ops)
4168 break;
3932 } 4169 }
4170 if (&probe->list == &tr->func_probes)
4171 goto err_unlock_ftrace;
4172
4173 ret = -EINVAL;
4174 if (!(probe->ops.flags & FTRACE_OPS_FL_INITIALIZED))
4175 goto err_unlock_ftrace;
4176
4177 acquire_probe_locked(probe);
4178
4179 mutex_unlock(&ftrace_lock);
3933 4180
3934 mutex_lock(&trace_probe_ops.func_hash->regex_lock); 4181 mutex_lock(&probe->ops.func_hash->regex_lock);
4182
4183 orig_hash = &probe->ops.func_hash->filter_hash;
4184 old_hash = *orig_hash;
4185
4186 if (ftrace_hash_empty(old_hash))
4187 goto out_unlock;
3935 4188
3936 old_hash_ops.filter_hash = old_hash; 4189 old_hash_ops.filter_hash = old_hash;
3937 /* Probes only have filters */ 4190 /* Probes only have filters */
3938 old_hash_ops.notrace_hash = NULL; 4191 old_hash_ops.notrace_hash = NULL;
3939 4192
3940 hash = alloc_and_copy_ftrace_hash(FTRACE_HASH_DEFAULT_BITS, *orig_hash); 4193 ret = -ENOMEM;
4194 hash = alloc_and_copy_ftrace_hash(FTRACE_HASH_DEFAULT_BITS, old_hash);
3941 if (!hash) 4195 if (!hash)
3942 /* Hmm, should report this somehow */
3943 goto out_unlock; 4196 goto out_unlock;
3944 4197
3945 INIT_LIST_HEAD(&free_list); 4198 INIT_HLIST_HEAD(&hhd);
3946
3947 for (i = 0; i < FTRACE_FUNC_HASHSIZE; i++) {
3948 struct hlist_head *hhd = &ftrace_func_hash[i];
3949 4199
3950 hlist_for_each_entry_safe(entry, tmp, hhd, node) { 4200 size = 1 << hash->size_bits;
3951 4201 for (i = 0; i < size; i++) {
3952 /* break up if statements for readability */ 4202 hlist_for_each_entry_safe(entry, tmp, &hash->buckets[i], hlist) {
3953 if ((flags & PROBE_TEST_FUNC) && entry->ops != ops)
3954 continue;
3955
3956 if ((flags & PROBE_TEST_DATA) && entry->data != data)
3957 continue;
3958 4203
3959 /* do this last, since it is the most expensive */
3960 if (func_g.search) { 4204 if (func_g.search) {
3961 kallsyms_lookup(entry->ip, NULL, NULL, 4205 kallsyms_lookup(entry->ip, NULL, NULL,
3962 NULL, str); 4206 NULL, str);
3963 if (!ftrace_match(str, &func_g)) 4207 if (!ftrace_match(str, &func_g))
3964 continue; 4208 continue;
3965 } 4209 }
3966 4210 count++;
3967 rec_entry = ftrace_lookup_ip(hash, entry->ip); 4211 remove_hash_entry(hash, entry);
3968 /* It is possible more than one entry had this ip */ 4212 hlist_add_head(&entry->hlist, &hhd);
3969 if (rec_entry)
3970 free_hash_entry(hash, rec_entry);
3971
3972 hlist_del_rcu(&entry->node);
3973 list_add(&entry->free_list, &free_list);
3974 } 4213 }
3975 } 4214 }
4215
4216 /* Nothing found? */
4217 if (!count) {
4218 ret = -EINVAL;
4219 goto out_unlock;
4220 }
4221
3976 mutex_lock(&ftrace_lock); 4222 mutex_lock(&ftrace_lock);
3977 disabled = __disable_ftrace_function_probe(); 4223
3978 /* 4224 WARN_ON(probe->ref < count);
3979 * Remove after the disable is called. Otherwise, if the last 4225
3980 * probe is removed, a null hash means *all enabled*. 4226 probe->ref -= count;
3981 */ 4227
3982 ret = ftrace_hash_move(&trace_probe_ops, 1, orig_hash, hash); 4228 if (ftrace_hash_empty(hash))
4229 ftrace_shutdown(&probe->ops, 0);
4230
4231 ret = ftrace_hash_move_and_update_ops(&probe->ops, orig_hash,
4232 hash, 1);
3983 4233
3984 /* still need to update the function call sites */ 4234 /* still need to update the function call sites */
3985 if (ftrace_enabled && !disabled) 4235 if (ftrace_enabled && !ftrace_hash_empty(hash))
3986 ftrace_run_modify_code(&trace_probe_ops, FTRACE_UPDATE_CALLS, 4236 ftrace_run_modify_code(&probe->ops, FTRACE_UPDATE_CALLS,
3987 &old_hash_ops); 4237 &old_hash_ops);
3988 synchronize_sched(); 4238 synchronize_sched();
3989 if (!ret)
3990 free_ftrace_hash_rcu(old_hash);
3991 4239
3992 list_for_each_entry_safe(entry, p, &free_list, free_list) { 4240 hlist_for_each_entry_safe(entry, tmp, &hhd, hlist) {
3993 list_del(&entry->free_list); 4241 hlist_del(&entry->hlist);
3994 ftrace_free_entry(entry); 4242 if (probe_ops->free)
4243 probe_ops->free(probe_ops, tr, entry->ip, probe->data);
4244 kfree(entry);
3995 } 4245 }
3996 mutex_unlock(&ftrace_lock); 4246 mutex_unlock(&ftrace_lock);
3997 4247
3998 out_unlock: 4248 out_unlock:
3999 mutex_unlock(&trace_probe_ops.func_hash->regex_lock); 4249 mutex_unlock(&probe->ops.func_hash->regex_lock);
4000 free_ftrace_hash(hash); 4250 free_ftrace_hash(hash);
4001}
4002 4251
4003void 4252 release_probe(probe);
4004unregister_ftrace_function_probe(char *glob, struct ftrace_probe_ops *ops,
4005 void *data)
4006{
4007 __unregister_ftrace_function_probe(glob, ops, data,
4008 PROBE_TEST_FUNC | PROBE_TEST_DATA);
4009}
4010 4253
4011void 4254 return ret;
4012unregister_ftrace_function_probe_func(char *glob, struct ftrace_probe_ops *ops)
4013{
4014 __unregister_ftrace_function_probe(glob, ops, NULL, PROBE_TEST_FUNC);
4015}
4016 4255
4017void unregister_ftrace_function_probe_all(char *glob) 4256 err_unlock_ftrace:
4018{ 4257 mutex_unlock(&ftrace_lock);
4019 __unregister_ftrace_function_probe(glob, NULL, NULL, 0); 4258 return ret;
4020} 4259}
4021 4260
4022static LIST_HEAD(ftrace_commands); 4261static LIST_HEAD(ftrace_commands);
@@ -4068,9 +4307,11 @@ __init int unregister_ftrace_command(struct ftrace_func_command *cmd)
4068 return ret; 4307 return ret;
4069} 4308}
4070 4309
4071static int ftrace_process_regex(struct ftrace_hash *hash, 4310static int ftrace_process_regex(struct ftrace_iterator *iter,
4072 char *buff, int len, int enable) 4311 char *buff, int len, int enable)
4073{ 4312{
4313 struct ftrace_hash *hash = iter->hash;
4314 struct trace_array *tr = iter->ops->private;
4074 char *func, *command, *next = buff; 4315 char *func, *command, *next = buff;
4075 struct ftrace_func_command *p; 4316 struct ftrace_func_command *p;
4076 int ret = -EINVAL; 4317 int ret = -EINVAL;
@@ -4090,10 +4331,13 @@ static int ftrace_process_regex(struct ftrace_hash *hash,
4090 4331
4091 command = strsep(&next, ":"); 4332 command = strsep(&next, ":");
4092 4333
4334 if (WARN_ON_ONCE(!tr))
4335 return -EINVAL;
4336
4093 mutex_lock(&ftrace_cmd_mutex); 4337 mutex_lock(&ftrace_cmd_mutex);
4094 list_for_each_entry(p, &ftrace_commands, list) { 4338 list_for_each_entry(p, &ftrace_commands, list) {
4095 if (strcmp(p->name, command) == 0) { 4339 if (strcmp(p->name, command) == 0) {
4096 ret = p->func(hash, func, command, next, enable); 4340 ret = p->func(tr, hash, func, command, next, enable);
4097 goto out_unlock; 4341 goto out_unlock;
4098 } 4342 }
4099 } 4343 }
@@ -4130,7 +4374,7 @@ ftrace_regex_write(struct file *file, const char __user *ubuf,
4130 4374
4131 if (read >= 0 && trace_parser_loaded(parser) && 4375 if (read >= 0 && trace_parser_loaded(parser) &&
4132 !trace_parser_cont(parser)) { 4376 !trace_parser_cont(parser)) {
4133 ret = ftrace_process_regex(iter->hash, parser->buffer, 4377 ret = ftrace_process_regex(iter, parser->buffer,
4134 parser->idx, enable); 4378 parser->idx, enable);
4135 trace_parser_clear(parser); 4379 trace_parser_clear(parser);
4136 if (ret < 0) 4380 if (ret < 0)
@@ -4175,44 +4419,11 @@ ftrace_match_addr(struct ftrace_hash *hash, unsigned long ip, int remove)
4175 return add_hash_entry(hash, ip); 4419 return add_hash_entry(hash, ip);
4176} 4420}
4177 4421
4178static void ftrace_ops_update_code(struct ftrace_ops *ops,
4179 struct ftrace_ops_hash *old_hash)
4180{
4181 struct ftrace_ops *op;
4182
4183 if (!ftrace_enabled)
4184 return;
4185
4186 if (ops->flags & FTRACE_OPS_FL_ENABLED) {
4187 ftrace_run_modify_code(ops, FTRACE_UPDATE_CALLS, old_hash);
4188 return;
4189 }
4190
4191 /*
4192 * If this is the shared global_ops filter, then we need to
4193 * check if there is another ops that shares it, is enabled.
4194 * If so, we still need to run the modify code.
4195 */
4196 if (ops->func_hash != &global_ops.local_hash)
4197 return;
4198
4199 do_for_each_ftrace_op(op, ftrace_ops_list) {
4200 if (op->func_hash == &global_ops.local_hash &&
4201 op->flags & FTRACE_OPS_FL_ENABLED) {
4202 ftrace_run_modify_code(op, FTRACE_UPDATE_CALLS, old_hash);
4203 /* Only need to do this once */
4204 return;
4205 }
4206 } while_for_each_ftrace_op(op);
4207}
4208
4209static int 4422static int
4210ftrace_set_hash(struct ftrace_ops *ops, unsigned char *buf, int len, 4423ftrace_set_hash(struct ftrace_ops *ops, unsigned char *buf, int len,
4211 unsigned long ip, int remove, int reset, int enable) 4424 unsigned long ip, int remove, int reset, int enable)
4212{ 4425{
4213 struct ftrace_hash **orig_hash; 4426 struct ftrace_hash **orig_hash;
4214 struct ftrace_ops_hash old_hash_ops;
4215 struct ftrace_hash *old_hash;
4216 struct ftrace_hash *hash; 4427 struct ftrace_hash *hash;
4217 int ret; 4428 int ret;
4218 4429
@@ -4247,14 +4458,7 @@ ftrace_set_hash(struct ftrace_ops *ops, unsigned char *buf, int len,
4247 } 4458 }
4248 4459
4249 mutex_lock(&ftrace_lock); 4460 mutex_lock(&ftrace_lock);
4250 old_hash = *orig_hash; 4461 ret = ftrace_hash_move_and_update_ops(ops, orig_hash, hash, enable);
4251 old_hash_ops.filter_hash = ops->func_hash->filter_hash;
4252 old_hash_ops.notrace_hash = ops->func_hash->notrace_hash;
4253 ret = ftrace_hash_move(ops, enable, orig_hash, hash);
4254 if (!ret) {
4255 ftrace_ops_update_code(ops, &old_hash_ops);
4256 free_ftrace_hash_rcu(old_hash);
4257 }
4258 mutex_unlock(&ftrace_lock); 4462 mutex_unlock(&ftrace_lock);
4259 4463
4260 out_regex_unlock: 4464 out_regex_unlock:
@@ -4493,10 +4697,8 @@ static void __init set_ftrace_early_filters(void)
4493int ftrace_regex_release(struct inode *inode, struct file *file) 4697int ftrace_regex_release(struct inode *inode, struct file *file)
4494{ 4698{
4495 struct seq_file *m = (struct seq_file *)file->private_data; 4699 struct seq_file *m = (struct seq_file *)file->private_data;
4496 struct ftrace_ops_hash old_hash_ops;
4497 struct ftrace_iterator *iter; 4700 struct ftrace_iterator *iter;
4498 struct ftrace_hash **orig_hash; 4701 struct ftrace_hash **orig_hash;
4499 struct ftrace_hash *old_hash;
4500 struct trace_parser *parser; 4702 struct trace_parser *parser;
4501 int filter_hash; 4703 int filter_hash;
4502 int ret; 4704 int ret;
@@ -4526,16 +4728,12 @@ int ftrace_regex_release(struct inode *inode, struct file *file)
4526 orig_hash = &iter->ops->func_hash->notrace_hash; 4728 orig_hash = &iter->ops->func_hash->notrace_hash;
4527 4729
4528 mutex_lock(&ftrace_lock); 4730 mutex_lock(&ftrace_lock);
4529 old_hash = *orig_hash; 4731 ret = ftrace_hash_move_and_update_ops(iter->ops, orig_hash,
4530 old_hash_ops.filter_hash = iter->ops->func_hash->filter_hash; 4732 iter->hash, filter_hash);
4531 old_hash_ops.notrace_hash = iter->ops->func_hash->notrace_hash;
4532 ret = ftrace_hash_move(iter->ops, filter_hash,
4533 orig_hash, iter->hash);
4534 if (!ret) {
4535 ftrace_ops_update_code(iter->ops, &old_hash_ops);
4536 free_ftrace_hash_rcu(old_hash);
4537 }
4538 mutex_unlock(&ftrace_lock); 4733 mutex_unlock(&ftrace_lock);
4734 } else {
4735 /* For read only, the hash is the ops hash */
4736 iter->hash = NULL;
4539 } 4737 }
4540 4738
4541 mutex_unlock(&iter->ops->func_hash->regex_lock); 4739 mutex_unlock(&iter->ops->func_hash->regex_lock);
@@ -5274,6 +5472,50 @@ void ftrace_module_init(struct module *mod)
5274} 5472}
5275#endif /* CONFIG_MODULES */ 5473#endif /* CONFIG_MODULES */
5276 5474
5475void __init ftrace_free_init_mem(void)
5476{
5477 unsigned long start = (unsigned long)(&__init_begin);
5478 unsigned long end = (unsigned long)(&__init_end);
5479 struct ftrace_page **last_pg = &ftrace_pages_start;
5480 struct ftrace_page *pg;
5481 struct dyn_ftrace *rec;
5482 struct dyn_ftrace key;
5483 int order;
5484
5485 key.ip = start;
5486 key.flags = end; /* overload flags, as it is unsigned long */
5487
5488 mutex_lock(&ftrace_lock);
5489
5490 for (pg = ftrace_pages_start; pg; last_pg = &pg->next, pg = *last_pg) {
5491 if (end < pg->records[0].ip ||
5492 start >= (pg->records[pg->index - 1].ip + MCOUNT_INSN_SIZE))
5493 continue;
5494 again:
5495 rec = bsearch(&key, pg->records, pg->index,
5496 sizeof(struct dyn_ftrace),
5497 ftrace_cmp_recs);
5498 if (!rec)
5499 continue;
5500 pg->index--;
5501 if (!pg->index) {
5502 *last_pg = pg->next;
5503 order = get_count_order(pg->size / ENTRIES_PER_PAGE);
5504 free_pages((unsigned long)pg->records, order);
5505 kfree(pg);
5506 pg = container_of(last_pg, struct ftrace_page, next);
5507 if (!(*last_pg))
5508 ftrace_pages = pg;
5509 continue;
5510 }
5511 memmove(rec, rec + 1,
5512 (pg->index - (rec - pg->records)) * sizeof(*rec));
5513 /* More than one function may be in this block */
5514 goto again;
5515 }
5516 mutex_unlock(&ftrace_lock);
5517}
5518
5277void __init ftrace_init(void) 5519void __init ftrace_init(void)
5278{ 5520{
5279 extern unsigned long __start_mcount_loc[]; 5521 extern unsigned long __start_mcount_loc[];
@@ -5316,25 +5558,13 @@ void __weak arch_ftrace_update_trampoline(struct ftrace_ops *ops)
5316 5558
5317static void ftrace_update_trampoline(struct ftrace_ops *ops) 5559static void ftrace_update_trampoline(struct ftrace_ops *ops)
5318{ 5560{
5319
5320/*
5321 * Currently there's no safe way to free a trampoline when the kernel
5322 * is configured with PREEMPT. That is because a task could be preempted
5323 * when it jumped to the trampoline, it may be preempted for a long time
5324 * depending on the system load, and currently there's no way to know
5325 * when it will be off the trampoline. If the trampoline is freed
5326 * too early, when the task runs again, it will be executing on freed
5327 * memory and crash.
5328 */
5329#ifdef CONFIG_PREEMPT
5330 /* Currently, only non dynamic ops can have a trampoline */
5331 if (ops->flags & FTRACE_OPS_FL_DYNAMIC)
5332 return;
5333#endif
5334
5335 arch_ftrace_update_trampoline(ops); 5561 arch_ftrace_update_trampoline(ops);
5336} 5562}
5337 5563
5564void ftrace_init_trace_array(struct trace_array *tr)
5565{
5566 INIT_LIST_HEAD(&tr->func_probes);
5567}
5338#else 5568#else
5339 5569
5340static struct ftrace_ops global_ops = { 5570static struct ftrace_ops global_ops = {
@@ -5389,6 +5619,7 @@ __init void ftrace_init_global_array_ops(struct trace_array *tr)
5389{ 5619{
5390 tr->ops = &global_ops; 5620 tr->ops = &global_ops;
5391 tr->ops->private = tr; 5621 tr->ops->private = tr;
5622 ftrace_init_trace_array(tr);
5392} 5623}
5393 5624
5394void ftrace_init_array_ops(struct trace_array *tr, ftrace_func_t func) 5625void ftrace_init_array_ops(struct trace_array *tr, ftrace_func_t func)
@@ -5543,6 +5774,43 @@ ftrace_filter_pid_sched_switch_probe(void *data, bool preempt,
5543 trace_ignore_this_task(pid_list, next)); 5774 trace_ignore_this_task(pid_list, next));
5544} 5775}
5545 5776
5777static void
5778ftrace_pid_follow_sched_process_fork(void *data,
5779 struct task_struct *self,
5780 struct task_struct *task)
5781{
5782 struct trace_pid_list *pid_list;
5783 struct trace_array *tr = data;
5784
5785 pid_list = rcu_dereference_sched(tr->function_pids);
5786 trace_filter_add_remove_task(pid_list, self, task);
5787}
5788
5789static void
5790ftrace_pid_follow_sched_process_exit(void *data, struct task_struct *task)
5791{
5792 struct trace_pid_list *pid_list;
5793 struct trace_array *tr = data;
5794
5795 pid_list = rcu_dereference_sched(tr->function_pids);
5796 trace_filter_add_remove_task(pid_list, NULL, task);
5797}
5798
5799void ftrace_pid_follow_fork(struct trace_array *tr, bool enable)
5800{
5801 if (enable) {
5802 register_trace_sched_process_fork(ftrace_pid_follow_sched_process_fork,
5803 tr);
5804 register_trace_sched_process_exit(ftrace_pid_follow_sched_process_exit,
5805 tr);
5806 } else {
5807 unregister_trace_sched_process_fork(ftrace_pid_follow_sched_process_fork,
5808 tr);
5809 unregister_trace_sched_process_exit(ftrace_pid_follow_sched_process_exit,
5810 tr);
5811 }
5812}
5813
5546static void clear_ftrace_pids(struct trace_array *tr) 5814static void clear_ftrace_pids(struct trace_array *tr)
5547{ 5815{
5548 struct trace_pid_list *pid_list; 5816 struct trace_pid_list *pid_list;
diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c
index ca47a4fa2986..4ae268e687fe 100644
--- a/kernel/trace/ring_buffer.c
+++ b/kernel/trace/ring_buffer.c
@@ -438,6 +438,7 @@ struct ring_buffer_per_cpu {
438 raw_spinlock_t reader_lock; /* serialize readers */ 438 raw_spinlock_t reader_lock; /* serialize readers */
439 arch_spinlock_t lock; 439 arch_spinlock_t lock;
440 struct lock_class_key lock_key; 440 struct lock_class_key lock_key;
441 struct buffer_data_page *free_page;
441 unsigned long nr_pages; 442 unsigned long nr_pages;
442 unsigned int current_context; 443 unsigned int current_context;
443 struct list_head *pages; 444 struct list_head *pages;
@@ -4389,9 +4390,25 @@ EXPORT_SYMBOL_GPL(ring_buffer_swap_cpu);
4389 */ 4390 */
4390void *ring_buffer_alloc_read_page(struct ring_buffer *buffer, int cpu) 4391void *ring_buffer_alloc_read_page(struct ring_buffer *buffer, int cpu)
4391{ 4392{
4392 struct buffer_data_page *bpage; 4393 struct ring_buffer_per_cpu *cpu_buffer = buffer->buffers[cpu];
4394 struct buffer_data_page *bpage = NULL;
4395 unsigned long flags;
4393 struct page *page; 4396 struct page *page;
4394 4397
4398 local_irq_save(flags);
4399 arch_spin_lock(&cpu_buffer->lock);
4400
4401 if (cpu_buffer->free_page) {
4402 bpage = cpu_buffer->free_page;
4403 cpu_buffer->free_page = NULL;
4404 }
4405
4406 arch_spin_unlock(&cpu_buffer->lock);
4407 local_irq_restore(flags);
4408
4409 if (bpage)
4410 goto out;
4411
4395 page = alloc_pages_node(cpu_to_node(cpu), 4412 page = alloc_pages_node(cpu_to_node(cpu),
4396 GFP_KERNEL | __GFP_NORETRY, 0); 4413 GFP_KERNEL | __GFP_NORETRY, 0);
4397 if (!page) 4414 if (!page)
@@ -4399,6 +4416,7 @@ void *ring_buffer_alloc_read_page(struct ring_buffer *buffer, int cpu)
4399 4416
4400 bpage = page_address(page); 4417 bpage = page_address(page);
4401 4418
4419 out:
4402 rb_init_page(bpage); 4420 rb_init_page(bpage);
4403 4421
4404 return bpage; 4422 return bpage;
@@ -4408,13 +4426,29 @@ EXPORT_SYMBOL_GPL(ring_buffer_alloc_read_page);
4408/** 4426/**
4409 * ring_buffer_free_read_page - free an allocated read page 4427 * ring_buffer_free_read_page - free an allocated read page
4410 * @buffer: the buffer the page was allocate for 4428 * @buffer: the buffer the page was allocate for
4429 * @cpu: the cpu buffer the page came from
4411 * @data: the page to free 4430 * @data: the page to free
4412 * 4431 *
4413 * Free a page allocated from ring_buffer_alloc_read_page. 4432 * Free a page allocated from ring_buffer_alloc_read_page.
4414 */ 4433 */
4415void ring_buffer_free_read_page(struct ring_buffer *buffer, void *data) 4434void ring_buffer_free_read_page(struct ring_buffer *buffer, int cpu, void *data)
4416{ 4435{
4417 free_page((unsigned long)data); 4436 struct ring_buffer_per_cpu *cpu_buffer = buffer->buffers[cpu];
4437 struct buffer_data_page *bpage = data;
4438 unsigned long flags;
4439
4440 local_irq_save(flags);
4441 arch_spin_lock(&cpu_buffer->lock);
4442
4443 if (!cpu_buffer->free_page) {
4444 cpu_buffer->free_page = bpage;
4445 bpage = NULL;
4446 }
4447
4448 arch_spin_unlock(&cpu_buffer->lock);
4449 local_irq_restore(flags);
4450
4451 free_page((unsigned long)bpage);
4418} 4452}
4419EXPORT_SYMBOL_GPL(ring_buffer_free_read_page); 4453EXPORT_SYMBOL_GPL(ring_buffer_free_read_page);
4420 4454
diff --git a/kernel/trace/ring_buffer_benchmark.c b/kernel/trace/ring_buffer_benchmark.c
index c190a4d5013c..9fbcaf567886 100644
--- a/kernel/trace/ring_buffer_benchmark.c
+++ b/kernel/trace/ring_buffer_benchmark.c
@@ -171,7 +171,7 @@ static enum event_status read_page(int cpu)
171 } 171 }
172 } 172 }
173 } 173 }
174 ring_buffer_free_read_page(buffer, bpage); 174 ring_buffer_free_read_page(buffer, cpu, bpage);
175 175
176 if (ret < 0) 176 if (ret < 0)
177 return EVENT_DROPPED; 177 return EVENT_DROPPED;
diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
index 0029fe62b245..80eda7d254ed 100644
--- a/kernel/trace/trace.c
+++ b/kernel/trace/trace.c
@@ -257,7 +257,7 @@ unsigned long long ns2usecs(u64 nsec)
257 257
258/* trace_flags that are default zero for instances */ 258/* trace_flags that are default zero for instances */
259#define ZEROED_TRACE_FLAGS \ 259#define ZEROED_TRACE_FLAGS \
260 TRACE_ITER_EVENT_FORK 260 (TRACE_ITER_EVENT_FORK | TRACE_ITER_FUNC_FORK)
261 261
262/* 262/*
263 * The global_trace is the descriptor that holds the top-level tracing 263 * The global_trace is the descriptor that holds the top-level tracing
@@ -757,7 +757,7 @@ __trace_buffer_lock_reserve(struct ring_buffer *buffer,
757 return event; 757 return event;
758} 758}
759 759
760static void tracer_tracing_on(struct trace_array *tr) 760void tracer_tracing_on(struct trace_array *tr)
761{ 761{
762 if (tr->trace_buffer.buffer) 762 if (tr->trace_buffer.buffer)
763 ring_buffer_record_on(tr->trace_buffer.buffer); 763 ring_buffer_record_on(tr->trace_buffer.buffer);
@@ -894,23 +894,8 @@ int __trace_bputs(unsigned long ip, const char *str)
894EXPORT_SYMBOL_GPL(__trace_bputs); 894EXPORT_SYMBOL_GPL(__trace_bputs);
895 895
896#ifdef CONFIG_TRACER_SNAPSHOT 896#ifdef CONFIG_TRACER_SNAPSHOT
897/** 897static void tracing_snapshot_instance(struct trace_array *tr)
898 * trace_snapshot - take a snapshot of the current buffer.
899 *
900 * This causes a swap between the snapshot buffer and the current live
901 * tracing buffer. You can use this to take snapshots of the live
902 * trace when some condition is triggered, but continue to trace.
903 *
904 * Note, make sure to allocate the snapshot with either
905 * a tracing_snapshot_alloc(), or by doing it manually
906 * with: echo 1 > /sys/kernel/debug/tracing/snapshot
907 *
908 * If the snapshot buffer is not allocated, it will stop tracing.
909 * Basically making a permanent snapshot.
910 */
911void tracing_snapshot(void)
912{ 898{
913 struct trace_array *tr = &global_trace;
914 struct tracer *tracer = tr->current_trace; 899 struct tracer *tracer = tr->current_trace;
915 unsigned long flags; 900 unsigned long flags;
916 901
@@ -938,6 +923,27 @@ void tracing_snapshot(void)
938 update_max_tr(tr, current, smp_processor_id()); 923 update_max_tr(tr, current, smp_processor_id());
939 local_irq_restore(flags); 924 local_irq_restore(flags);
940} 925}
926
927/**
928 * trace_snapshot - take a snapshot of the current buffer.
929 *
930 * This causes a swap between the snapshot buffer and the current live
931 * tracing buffer. You can use this to take snapshots of the live
932 * trace when some condition is triggered, but continue to trace.
933 *
934 * Note, make sure to allocate the snapshot with either
935 * a tracing_snapshot_alloc(), or by doing it manually
936 * with: echo 1 > /sys/kernel/debug/tracing/snapshot
937 *
938 * If the snapshot buffer is not allocated, it will stop tracing.
939 * Basically making a permanent snapshot.
940 */
941void tracing_snapshot(void)
942{
943 struct trace_array *tr = &global_trace;
944
945 tracing_snapshot_instance(tr);
946}
941EXPORT_SYMBOL_GPL(tracing_snapshot); 947EXPORT_SYMBOL_GPL(tracing_snapshot);
942 948
943static int resize_buffer_duplicate_size(struct trace_buffer *trace_buf, 949static int resize_buffer_duplicate_size(struct trace_buffer *trace_buf,
@@ -1039,7 +1045,7 @@ void tracing_snapshot_alloc(void)
1039EXPORT_SYMBOL_GPL(tracing_snapshot_alloc); 1045EXPORT_SYMBOL_GPL(tracing_snapshot_alloc);
1040#endif /* CONFIG_TRACER_SNAPSHOT */ 1046#endif /* CONFIG_TRACER_SNAPSHOT */
1041 1047
1042static void tracer_tracing_off(struct trace_array *tr) 1048void tracer_tracing_off(struct trace_array *tr)
1043{ 1049{
1044 if (tr->trace_buffer.buffer) 1050 if (tr->trace_buffer.buffer)
1045 ring_buffer_record_off(tr->trace_buffer.buffer); 1051 ring_buffer_record_off(tr->trace_buffer.buffer);
@@ -1424,6 +1430,28 @@ static int wait_on_pipe(struct trace_iterator *iter, bool full)
1424} 1430}
1425 1431
1426#ifdef CONFIG_FTRACE_STARTUP_TEST 1432#ifdef CONFIG_FTRACE_STARTUP_TEST
1433static bool selftests_can_run;
1434
1435struct trace_selftests {
1436 struct list_head list;
1437 struct tracer *type;
1438};
1439
1440static LIST_HEAD(postponed_selftests);
1441
1442static int save_selftest(struct tracer *type)
1443{
1444 struct trace_selftests *selftest;
1445
1446 selftest = kmalloc(sizeof(*selftest), GFP_KERNEL);
1447 if (!selftest)
1448 return -ENOMEM;
1449
1450 selftest->type = type;
1451 list_add(&selftest->list, &postponed_selftests);
1452 return 0;
1453}
1454
1427static int run_tracer_selftest(struct tracer *type) 1455static int run_tracer_selftest(struct tracer *type)
1428{ 1456{
1429 struct trace_array *tr = &global_trace; 1457 struct trace_array *tr = &global_trace;
@@ -1434,6 +1462,14 @@ static int run_tracer_selftest(struct tracer *type)
1434 return 0; 1462 return 0;
1435 1463
1436 /* 1464 /*
1465 * If a tracer registers early in boot up (before scheduling is
1466 * initialized and such), then do not run its selftests yet.
1467 * Instead, run it a little later in the boot process.
1468 */
1469 if (!selftests_can_run)
1470 return save_selftest(type);
1471
1472 /*
1437 * Run a selftest on this tracer. 1473 * Run a selftest on this tracer.
1438 * Here we reset the trace buffer, and set the current 1474 * Here we reset the trace buffer, and set the current
1439 * tracer to be this tracer. The tracer can then run some 1475 * tracer to be this tracer. The tracer can then run some
@@ -1482,6 +1518,47 @@ static int run_tracer_selftest(struct tracer *type)
1482 printk(KERN_CONT "PASSED\n"); 1518 printk(KERN_CONT "PASSED\n");
1483 return 0; 1519 return 0;
1484} 1520}
1521
1522static __init int init_trace_selftests(void)
1523{
1524 struct trace_selftests *p, *n;
1525 struct tracer *t, **last;
1526 int ret;
1527
1528 selftests_can_run = true;
1529
1530 mutex_lock(&trace_types_lock);
1531
1532 if (list_empty(&postponed_selftests))
1533 goto out;
1534
1535 pr_info("Running postponed tracer tests:\n");
1536
1537 list_for_each_entry_safe(p, n, &postponed_selftests, list) {
1538 ret = run_tracer_selftest(p->type);
1539 /* If the test fails, then warn and remove from available_tracers */
1540 if (ret < 0) {
1541 WARN(1, "tracer: %s failed selftest, disabling\n",
1542 p->type->name);
1543 last = &trace_types;
1544 for (t = trace_types; t; t = t->next) {
1545 if (t == p->type) {
1546 *last = t->next;
1547 break;
1548 }
1549 last = &t->next;
1550 }
1551 }
1552 list_del(&p->list);
1553 kfree(p);
1554 }
1555
1556 out:
1557 mutex_unlock(&trace_types_lock);
1558
1559 return 0;
1560}
1561early_initcall(init_trace_selftests);
1485#else 1562#else
1486static inline int run_tracer_selftest(struct tracer *type) 1563static inline int run_tracer_selftest(struct tracer *type)
1487{ 1564{
@@ -1927,6 +2004,18 @@ void tracing_record_cmdline(struct task_struct *tsk)
1927 __this_cpu_write(trace_cmdline_save, false); 2004 __this_cpu_write(trace_cmdline_save, false);
1928} 2005}
1929 2006
2007/*
2008 * Several functions return TRACE_TYPE_PARTIAL_LINE if the trace_seq
2009 * overflowed, and TRACE_TYPE_HANDLED otherwise. This helper function
2010 * simplifies those functions and keeps them in sync.
2011 */
2012enum print_line_t trace_handle_return(struct trace_seq *s)
2013{
2014 return trace_seq_has_overflowed(s) ?
2015 TRACE_TYPE_PARTIAL_LINE : TRACE_TYPE_HANDLED;
2016}
2017EXPORT_SYMBOL_GPL(trace_handle_return);
2018
1930void 2019void
1931tracing_generic_entry_update(struct trace_entry *entry, unsigned long flags, 2020tracing_generic_entry_update(struct trace_entry *entry, unsigned long flags,
1932 int pc) 2021 int pc)
@@ -4122,6 +4211,9 @@ int set_tracer_flag(struct trace_array *tr, unsigned int mask, int enabled)
4122 if (mask == TRACE_ITER_EVENT_FORK) 4211 if (mask == TRACE_ITER_EVENT_FORK)
4123 trace_event_follow_fork(tr, enabled); 4212 trace_event_follow_fork(tr, enabled);
4124 4213
4214 if (mask == TRACE_ITER_FUNC_FORK)
4215 ftrace_pid_follow_fork(tr, enabled);
4216
4125 if (mask == TRACE_ITER_OVERWRITE) { 4217 if (mask == TRACE_ITER_OVERWRITE) {
4126 ring_buffer_change_overwrite(tr->trace_buffer.buffer, enabled); 4218 ring_buffer_change_overwrite(tr->trace_buffer.buffer, enabled);
4127#ifdef CONFIG_TRACER_MAX_TRACE 4219#ifdef CONFIG_TRACER_MAX_TRACE
@@ -5962,6 +6054,7 @@ static int tracing_clock_open(struct inode *inode, struct file *file)
5962struct ftrace_buffer_info { 6054struct ftrace_buffer_info {
5963 struct trace_iterator iter; 6055 struct trace_iterator iter;
5964 void *spare; 6056 void *spare;
6057 unsigned int spare_cpu;
5965 unsigned int read; 6058 unsigned int read;
5966}; 6059};
5967 6060
@@ -6291,9 +6384,11 @@ tracing_buffers_read(struct file *filp, char __user *ubuf,
6291 return -EBUSY; 6384 return -EBUSY;
6292#endif 6385#endif
6293 6386
6294 if (!info->spare) 6387 if (!info->spare) {
6295 info->spare = ring_buffer_alloc_read_page(iter->trace_buffer->buffer, 6388 info->spare = ring_buffer_alloc_read_page(iter->trace_buffer->buffer,
6296 iter->cpu_file); 6389 iter->cpu_file);
6390 info->spare_cpu = iter->cpu_file;
6391 }
6297 if (!info->spare) 6392 if (!info->spare)
6298 return -ENOMEM; 6393 return -ENOMEM;
6299 6394
@@ -6353,7 +6448,8 @@ static int tracing_buffers_release(struct inode *inode, struct file *file)
6353 __trace_array_put(iter->tr); 6448 __trace_array_put(iter->tr);
6354 6449
6355 if (info->spare) 6450 if (info->spare)
6356 ring_buffer_free_read_page(iter->trace_buffer->buffer, info->spare); 6451 ring_buffer_free_read_page(iter->trace_buffer->buffer,
6452 info->spare_cpu, info->spare);
6357 kfree(info); 6453 kfree(info);
6358 6454
6359 mutex_unlock(&trace_types_lock); 6455 mutex_unlock(&trace_types_lock);
@@ -6364,6 +6460,7 @@ static int tracing_buffers_release(struct inode *inode, struct file *file)
6364struct buffer_ref { 6460struct buffer_ref {
6365 struct ring_buffer *buffer; 6461 struct ring_buffer *buffer;
6366 void *page; 6462 void *page;
6463 int cpu;
6367 int ref; 6464 int ref;
6368}; 6465};
6369 6466
@@ -6375,7 +6472,7 @@ static void buffer_pipe_buf_release(struct pipe_inode_info *pipe,
6375 if (--ref->ref) 6472 if (--ref->ref)
6376 return; 6473 return;
6377 6474
6378 ring_buffer_free_read_page(ref->buffer, ref->page); 6475 ring_buffer_free_read_page(ref->buffer, ref->cpu, ref->page);
6379 kfree(ref); 6476 kfree(ref);
6380 buf->private = 0; 6477 buf->private = 0;
6381} 6478}
@@ -6409,7 +6506,7 @@ static void buffer_spd_release(struct splice_pipe_desc *spd, unsigned int i)
6409 if (--ref->ref) 6506 if (--ref->ref)
6410 return; 6507 return;
6411 6508
6412 ring_buffer_free_read_page(ref->buffer, ref->page); 6509 ring_buffer_free_read_page(ref->buffer, ref->cpu, ref->page);
6413 kfree(ref); 6510 kfree(ref);
6414 spd->partial[i].private = 0; 6511 spd->partial[i].private = 0;
6415} 6512}
@@ -6473,11 +6570,13 @@ tracing_buffers_splice_read(struct file *file, loff_t *ppos,
6473 kfree(ref); 6570 kfree(ref);
6474 break; 6571 break;
6475 } 6572 }
6573 ref->cpu = iter->cpu_file;
6476 6574
6477 r = ring_buffer_read_page(ref->buffer, &ref->page, 6575 r = ring_buffer_read_page(ref->buffer, &ref->page,
6478 len, iter->cpu_file, 1); 6576 len, iter->cpu_file, 1);
6479 if (r < 0) { 6577 if (r < 0) {
6480 ring_buffer_free_read_page(ref->buffer, ref->page); 6578 ring_buffer_free_read_page(ref->buffer, ref->cpu,
6579 ref->page);
6481 kfree(ref); 6580 kfree(ref);
6482 break; 6581 break;
6483 } 6582 }
@@ -6648,43 +6747,89 @@ static const struct file_operations tracing_dyn_info_fops = {
6648 6747
6649#if defined(CONFIG_TRACER_SNAPSHOT) && defined(CONFIG_DYNAMIC_FTRACE) 6748#if defined(CONFIG_TRACER_SNAPSHOT) && defined(CONFIG_DYNAMIC_FTRACE)
6650static void 6749static void
6651ftrace_snapshot(unsigned long ip, unsigned long parent_ip, void **data) 6750ftrace_snapshot(unsigned long ip, unsigned long parent_ip,
6751 struct trace_array *tr, struct ftrace_probe_ops *ops,
6752 void *data)
6652{ 6753{
6653 tracing_snapshot(); 6754 tracing_snapshot_instance(tr);
6654} 6755}
6655 6756
6656static void 6757static void
6657ftrace_count_snapshot(unsigned long ip, unsigned long parent_ip, void **data) 6758ftrace_count_snapshot(unsigned long ip, unsigned long parent_ip,
6759 struct trace_array *tr, struct ftrace_probe_ops *ops,
6760 void *data)
6658{ 6761{
6659 unsigned long *count = (long *)data; 6762 struct ftrace_func_mapper *mapper = data;
6763 long *count = NULL;
6660 6764
6661 if (!*count) 6765 if (mapper)
6662 return; 6766 count = (long *)ftrace_func_mapper_find_ip(mapper, ip);
6767
6768 if (count) {
6769
6770 if (*count <= 0)
6771 return;
6663 6772
6664 if (*count != -1)
6665 (*count)--; 6773 (*count)--;
6774 }
6666 6775
6667 tracing_snapshot(); 6776 tracing_snapshot_instance(tr);
6668} 6777}
6669 6778
6670static int 6779static int
6671ftrace_snapshot_print(struct seq_file *m, unsigned long ip, 6780ftrace_snapshot_print(struct seq_file *m, unsigned long ip,
6672 struct ftrace_probe_ops *ops, void *data) 6781 struct ftrace_probe_ops *ops, void *data)
6673{ 6782{
6674 long count = (long)data; 6783 struct ftrace_func_mapper *mapper = data;
6784 long *count = NULL;
6675 6785
6676 seq_printf(m, "%ps:", (void *)ip); 6786 seq_printf(m, "%ps:", (void *)ip);
6677 6787
6678 seq_puts(m, "snapshot"); 6788 seq_puts(m, "snapshot");
6679 6789
6680 if (count == -1) 6790 if (mapper)
6681 seq_puts(m, ":unlimited\n"); 6791 count = (long *)ftrace_func_mapper_find_ip(mapper, ip);
6792
6793 if (count)
6794 seq_printf(m, ":count=%ld\n", *count);
6682 else 6795 else
6683 seq_printf(m, ":count=%ld\n", count); 6796 seq_puts(m, ":unlimited\n");
6684 6797
6685 return 0; 6798 return 0;
6686} 6799}
6687 6800
6801static int
6802ftrace_snapshot_init(struct ftrace_probe_ops *ops, struct trace_array *tr,
6803 unsigned long ip, void *init_data, void **data)
6804{
6805 struct ftrace_func_mapper *mapper = *data;
6806
6807 if (!mapper) {
6808 mapper = allocate_ftrace_func_mapper();
6809 if (!mapper)
6810 return -ENOMEM;
6811 *data = mapper;
6812 }
6813
6814 return ftrace_func_mapper_add_ip(mapper, ip, init_data);
6815}
6816
6817static void
6818ftrace_snapshot_free(struct ftrace_probe_ops *ops, struct trace_array *tr,
6819 unsigned long ip, void *data)
6820{
6821 struct ftrace_func_mapper *mapper = data;
6822
6823 if (!ip) {
6824 if (!mapper)
6825 return;
6826 free_ftrace_func_mapper(mapper, NULL);
6827 return;
6828 }
6829
6830 ftrace_func_mapper_remove_ip(mapper, ip);
6831}
6832
6688static struct ftrace_probe_ops snapshot_probe_ops = { 6833static struct ftrace_probe_ops snapshot_probe_ops = {
6689 .func = ftrace_snapshot, 6834 .func = ftrace_snapshot,
6690 .print = ftrace_snapshot_print, 6835 .print = ftrace_snapshot_print,
@@ -6693,10 +6838,12 @@ static struct ftrace_probe_ops snapshot_probe_ops = {
6693static struct ftrace_probe_ops snapshot_count_probe_ops = { 6838static struct ftrace_probe_ops snapshot_count_probe_ops = {
6694 .func = ftrace_count_snapshot, 6839 .func = ftrace_count_snapshot,
6695 .print = ftrace_snapshot_print, 6840 .print = ftrace_snapshot_print,
6841 .init = ftrace_snapshot_init,
6842 .free = ftrace_snapshot_free,
6696}; 6843};
6697 6844
6698static int 6845static int
6699ftrace_trace_snapshot_callback(struct ftrace_hash *hash, 6846ftrace_trace_snapshot_callback(struct trace_array *tr, struct ftrace_hash *hash,
6700 char *glob, char *cmd, char *param, int enable) 6847 char *glob, char *cmd, char *param, int enable)
6701{ 6848{
6702 struct ftrace_probe_ops *ops; 6849 struct ftrace_probe_ops *ops;
@@ -6710,10 +6857,8 @@ ftrace_trace_snapshot_callback(struct ftrace_hash *hash,
6710 6857
6711 ops = param ? &snapshot_count_probe_ops : &snapshot_probe_ops; 6858 ops = param ? &snapshot_count_probe_ops : &snapshot_probe_ops;
6712 6859
6713 if (glob[0] == '!') { 6860 if (glob[0] == '!')
6714 unregister_ftrace_function_probe_func(glob+1, ops); 6861 return unregister_ftrace_function_probe_func(glob+1, tr, ops);
6715 return 0;
6716 }
6717 6862
6718 if (!param) 6863 if (!param)
6719 goto out_reg; 6864 goto out_reg;
@@ -6732,11 +6877,11 @@ ftrace_trace_snapshot_callback(struct ftrace_hash *hash,
6732 return ret; 6877 return ret;
6733 6878
6734 out_reg: 6879 out_reg:
6735 ret = alloc_snapshot(&global_trace); 6880 ret = alloc_snapshot(tr);
6736 if (ret < 0) 6881 if (ret < 0)
6737 goto out; 6882 goto out;
6738 6883
6739 ret = register_ftrace_function_probe(glob, ops, count); 6884 ret = register_ftrace_function_probe(glob, tr, ops, count);
6740 6885
6741 out: 6886 out:
6742 return ret < 0 ? ret : 0; 6887 return ret < 0 ? ret : 0;
@@ -7347,6 +7492,8 @@ static int instance_mkdir(const char *name)
7347 goto out_free_tr; 7492 goto out_free_tr;
7348 } 7493 }
7349 7494
7495 ftrace_init_trace_array(tr);
7496
7350 init_tracer_tracefs(tr, tr->dir); 7497 init_tracer_tracefs(tr, tr->dir);
7351 init_trace_flags_index(tr); 7498 init_trace_flags_index(tr);
7352 __update_tracer_options(tr); 7499 __update_tracer_options(tr);
@@ -7967,6 +8114,9 @@ __init static int tracer_alloc_buffers(void)
7967 8114
7968 register_tracer(&nop_trace); 8115 register_tracer(&nop_trace);
7969 8116
8117 /* Function tracing may start here (via kernel command line) */
8118 init_function_trace();
8119
7970 /* All seems OK, enable tracing */ 8120 /* All seems OK, enable tracing */
7971 tracing_disabled = 0; 8121 tracing_disabled = 0;
7972 8122
@@ -8001,7 +8151,7 @@ out:
8001 return ret; 8151 return ret;
8002} 8152}
8003 8153
8004void __init trace_init(void) 8154void __init early_trace_init(void)
8005{ 8155{
8006 if (tracepoint_printk) { 8156 if (tracepoint_printk) {
8007 tracepoint_print_iter = 8157 tracepoint_print_iter =
@@ -8012,6 +8162,10 @@ void __init trace_init(void)
8012 static_key_enable(&tracepoint_printk_key.key); 8162 static_key_enable(&tracepoint_printk_key.key);
8013 } 8163 }
8014 tracer_alloc_buffers(); 8164 tracer_alloc_buffers();
8165}
8166
8167void __init trace_init(void)
8168{
8015 trace_event_init(); 8169 trace_event_init();
8016} 8170}
8017 8171
diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h
index d19d52d600d6..291a1bca5748 100644
--- a/kernel/trace/trace.h
+++ b/kernel/trace/trace.h
@@ -262,6 +262,9 @@ struct trace_array {
262#ifdef CONFIG_FUNCTION_TRACER 262#ifdef CONFIG_FUNCTION_TRACER
263 struct ftrace_ops *ops; 263 struct ftrace_ops *ops;
264 struct trace_pid_list __rcu *function_pids; 264 struct trace_pid_list __rcu *function_pids;
265#ifdef CONFIG_DYNAMIC_FTRACE
266 struct list_head func_probes;
267#endif
265 /* function tracing enabled */ 268 /* function tracing enabled */
266 int function_enabled; 269 int function_enabled;
267#endif 270#endif
@@ -579,6 +582,8 @@ void tracing_reset_all_online_cpus(void);
579int tracing_open_generic(struct inode *inode, struct file *filp); 582int tracing_open_generic(struct inode *inode, struct file *filp);
580bool tracing_is_disabled(void); 583bool tracing_is_disabled(void);
581int tracer_tracing_is_on(struct trace_array *tr); 584int tracer_tracing_is_on(struct trace_array *tr);
585void tracer_tracing_on(struct trace_array *tr);
586void tracer_tracing_off(struct trace_array *tr);
582struct dentry *trace_create_file(const char *name, 587struct dentry *trace_create_file(const char *name,
583 umode_t mode, 588 umode_t mode,
584 struct dentry *parent, 589 struct dentry *parent,
@@ -696,6 +701,9 @@ extern void trace_event_follow_fork(struct trace_array *tr, bool enable);
696 701
697#ifdef CONFIG_DYNAMIC_FTRACE 702#ifdef CONFIG_DYNAMIC_FTRACE
698extern unsigned long ftrace_update_tot_cnt; 703extern unsigned long ftrace_update_tot_cnt;
704void ftrace_init_trace_array(struct trace_array *tr);
705#else
706static inline void ftrace_init_trace_array(struct trace_array *tr) { }
699#endif 707#endif
700#define DYN_FTRACE_TEST_NAME trace_selftest_dynamic_test_func 708#define DYN_FTRACE_TEST_NAME trace_selftest_dynamic_test_func
701extern int DYN_FTRACE_TEST_NAME(void); 709extern int DYN_FTRACE_TEST_NAME(void);
@@ -880,6 +888,14 @@ print_graph_function_flags(struct trace_iterator *iter, u32 flags)
880extern struct list_head ftrace_pids; 888extern struct list_head ftrace_pids;
881 889
882#ifdef CONFIG_FUNCTION_TRACER 890#ifdef CONFIG_FUNCTION_TRACER
891struct ftrace_func_command {
892 struct list_head list;
893 char *name;
894 int (*func)(struct trace_array *tr,
895 struct ftrace_hash *hash,
896 char *func, char *cmd,
897 char *params, int enable);
898};
883extern bool ftrace_filter_param __initdata; 899extern bool ftrace_filter_param __initdata;
884static inline int ftrace_trace_task(struct trace_array *tr) 900static inline int ftrace_trace_task(struct trace_array *tr)
885{ 901{
@@ -897,6 +913,8 @@ void ftrace_init_tracefs(struct trace_array *tr, struct dentry *d_tracer);
897void ftrace_init_tracefs_toplevel(struct trace_array *tr, 913void ftrace_init_tracefs_toplevel(struct trace_array *tr,
898 struct dentry *d_tracer); 914 struct dentry *d_tracer);
899void ftrace_clear_pids(struct trace_array *tr); 915void ftrace_clear_pids(struct trace_array *tr);
916int init_function_trace(void);
917void ftrace_pid_follow_fork(struct trace_array *tr, bool enable);
900#else 918#else
901static inline int ftrace_trace_task(struct trace_array *tr) 919static inline int ftrace_trace_task(struct trace_array *tr)
902{ 920{
@@ -916,15 +934,70 @@ static inline void ftrace_reset_array_ops(struct trace_array *tr) { }
916static inline void ftrace_init_tracefs(struct trace_array *tr, struct dentry *d) { } 934static inline void ftrace_init_tracefs(struct trace_array *tr, struct dentry *d) { }
917static inline void ftrace_init_tracefs_toplevel(struct trace_array *tr, struct dentry *d) { } 935static inline void ftrace_init_tracefs_toplevel(struct trace_array *tr, struct dentry *d) { }
918static inline void ftrace_clear_pids(struct trace_array *tr) { } 936static inline void ftrace_clear_pids(struct trace_array *tr) { }
937static inline int init_function_trace(void) { return 0; }
938static inline void ftrace_pid_follow_fork(struct trace_array *tr, bool enable) { }
919/* ftace_func_t type is not defined, use macro instead of static inline */ 939/* ftace_func_t type is not defined, use macro instead of static inline */
920#define ftrace_init_array_ops(tr, func) do { } while (0) 940#define ftrace_init_array_ops(tr, func) do { } while (0)
921#endif /* CONFIG_FUNCTION_TRACER */ 941#endif /* CONFIG_FUNCTION_TRACER */
922 942
923#if defined(CONFIG_FUNCTION_TRACER) && defined(CONFIG_DYNAMIC_FTRACE) 943#if defined(CONFIG_FUNCTION_TRACER) && defined(CONFIG_DYNAMIC_FTRACE)
944
945struct ftrace_probe_ops {
946 void (*func)(unsigned long ip,
947 unsigned long parent_ip,
948 struct trace_array *tr,
949 struct ftrace_probe_ops *ops,
950 void *data);
951 int (*init)(struct ftrace_probe_ops *ops,
952 struct trace_array *tr,
953 unsigned long ip, void *init_data,
954 void **data);
955 void (*free)(struct ftrace_probe_ops *ops,
956 struct trace_array *tr,
957 unsigned long ip, void *data);
958 int (*print)(struct seq_file *m,
959 unsigned long ip,
960 struct ftrace_probe_ops *ops,
961 void *data);
962};
963
964struct ftrace_func_mapper;
965typedef int (*ftrace_mapper_func)(void *data);
966
967struct ftrace_func_mapper *allocate_ftrace_func_mapper(void);
968void **ftrace_func_mapper_find_ip(struct ftrace_func_mapper *mapper,
969 unsigned long ip);
970int ftrace_func_mapper_add_ip(struct ftrace_func_mapper *mapper,
971 unsigned long ip, void *data);
972void *ftrace_func_mapper_remove_ip(struct ftrace_func_mapper *mapper,
973 unsigned long ip);
974void free_ftrace_func_mapper(struct ftrace_func_mapper *mapper,
975 ftrace_mapper_func free_func);
976
977extern int
978register_ftrace_function_probe(char *glob, struct trace_array *tr,
979 struct ftrace_probe_ops *ops, void *data);
980extern int
981unregister_ftrace_function_probe_func(char *glob, struct trace_array *tr,
982 struct ftrace_probe_ops *ops);
983
984int register_ftrace_command(struct ftrace_func_command *cmd);
985int unregister_ftrace_command(struct ftrace_func_command *cmd);
986
924void ftrace_create_filter_files(struct ftrace_ops *ops, 987void ftrace_create_filter_files(struct ftrace_ops *ops,
925 struct dentry *parent); 988 struct dentry *parent);
926void ftrace_destroy_filter_files(struct ftrace_ops *ops); 989void ftrace_destroy_filter_files(struct ftrace_ops *ops);
927#else 990#else
991struct ftrace_func_command;
992
993static inline __init int register_ftrace_command(struct ftrace_func_command *cmd)
994{
995 return -EINVAL;
996}
997static inline __init int unregister_ftrace_command(char *cmd_name)
998{
999 return -EINVAL;
1000}
928/* 1001/*
929 * The ops parameter passed in is usually undefined. 1002 * The ops parameter passed in is usually undefined.
930 * This must be a macro. 1003 * This must be a macro.
@@ -989,11 +1062,13 @@ extern int trace_get_user(struct trace_parser *parser, const char __user *ubuf,
989 1062
990#ifdef CONFIG_FUNCTION_TRACER 1063#ifdef CONFIG_FUNCTION_TRACER
991# define FUNCTION_FLAGS \ 1064# define FUNCTION_FLAGS \
992 C(FUNCTION, "function-trace"), 1065 C(FUNCTION, "function-trace"), \
1066 C(FUNC_FORK, "function-fork"),
993# define FUNCTION_DEFAULT_FLAGS TRACE_ITER_FUNCTION 1067# define FUNCTION_DEFAULT_FLAGS TRACE_ITER_FUNCTION
994#else 1068#else
995# define FUNCTION_FLAGS 1069# define FUNCTION_FLAGS
996# define FUNCTION_DEFAULT_FLAGS 0UL 1070# define FUNCTION_DEFAULT_FLAGS 0UL
1071# define TRACE_ITER_FUNC_FORK 0UL
997#endif 1072#endif
998 1073
999#ifdef CONFIG_STACKTRACE 1074#ifdef CONFIG_STACKTRACE
diff --git a/kernel/trace/trace_benchmark.c b/kernel/trace/trace_benchmark.c
index e49fbe901cfc..16a8cf02eee9 100644
--- a/kernel/trace/trace_benchmark.c
+++ b/kernel/trace/trace_benchmark.c
@@ -153,10 +153,18 @@ static int benchmark_event_kthread(void *arg)
153 trace_do_benchmark(); 153 trace_do_benchmark();
154 154
155 /* 155 /*
156 * We don't go to sleep, but let others 156 * We don't go to sleep, but let others run as well.
157 * run as well. 157 * This is bascially a "yield()" to let any task that
158 * wants to run, schedule in, but if the CPU is idle,
159 * we'll keep burning cycles.
160 *
161 * Note the _rcu_qs() version of cond_resched() will
162 * notify synchronize_rcu_tasks() that this thread has
163 * passed a quiescent state for rcu_tasks. Otherwise
164 * this thread will never voluntarily schedule which would
165 * block synchronize_rcu_tasks() indefinitely.
158 */ 166 */
159 cond_resched(); 167 cond_resched_rcu_qs();
160 } 168 }
161 169
162 return 0; 170 return 0;
diff --git a/kernel/trace/trace_events.c b/kernel/trace/trace_events.c
index 93116549a284..e7973e10398c 100644
--- a/kernel/trace/trace_events.c
+++ b/kernel/trace/trace_events.c
@@ -2460,15 +2460,8 @@ struct event_probe_data {
2460 bool enable; 2460 bool enable;
2461}; 2461};
2462 2462
2463static void 2463static void update_event_probe(struct event_probe_data *data)
2464event_enable_probe(unsigned long ip, unsigned long parent_ip, void **_data)
2465{ 2464{
2466 struct event_probe_data **pdata = (struct event_probe_data **)_data;
2467 struct event_probe_data *data = *pdata;
2468
2469 if (!data)
2470 return;
2471
2472 if (data->enable) 2465 if (data->enable)
2473 clear_bit(EVENT_FILE_FL_SOFT_DISABLED_BIT, &data->file->flags); 2466 clear_bit(EVENT_FILE_FL_SOFT_DISABLED_BIT, &data->file->flags);
2474 else 2467 else
@@ -2476,77 +2469,141 @@ event_enable_probe(unsigned long ip, unsigned long parent_ip, void **_data)
2476} 2469}
2477 2470
2478static void 2471static void
2479event_enable_count_probe(unsigned long ip, unsigned long parent_ip, void **_data) 2472event_enable_probe(unsigned long ip, unsigned long parent_ip,
2473 struct trace_array *tr, struct ftrace_probe_ops *ops,
2474 void *data)
2480{ 2475{
2481 struct event_probe_data **pdata = (struct event_probe_data **)_data; 2476 struct ftrace_func_mapper *mapper = data;
2482 struct event_probe_data *data = *pdata; 2477 struct event_probe_data *edata;
2478 void **pdata;
2483 2479
2484 if (!data) 2480 pdata = ftrace_func_mapper_find_ip(mapper, ip);
2481 if (!pdata || !*pdata)
2482 return;
2483
2484 edata = *pdata;
2485 update_event_probe(edata);
2486}
2487
2488static void
2489event_enable_count_probe(unsigned long ip, unsigned long parent_ip,
2490 struct trace_array *tr, struct ftrace_probe_ops *ops,
2491 void *data)
2492{
2493 struct ftrace_func_mapper *mapper = data;
2494 struct event_probe_data *edata;
2495 void **pdata;
2496
2497 pdata = ftrace_func_mapper_find_ip(mapper, ip);
2498 if (!pdata || !*pdata)
2485 return; 2499 return;
2486 2500
2487 if (!data->count) 2501 edata = *pdata;
2502
2503 if (!edata->count)
2488 return; 2504 return;
2489 2505
2490 /* Skip if the event is in a state we want to switch to */ 2506 /* Skip if the event is in a state we want to switch to */
2491 if (data->enable == !(data->file->flags & EVENT_FILE_FL_SOFT_DISABLED)) 2507 if (edata->enable == !(edata->file->flags & EVENT_FILE_FL_SOFT_DISABLED))
2492 return; 2508 return;
2493 2509
2494 if (data->count != -1) 2510 if (edata->count != -1)
2495 (data->count)--; 2511 (edata->count)--;
2496 2512
2497 event_enable_probe(ip, parent_ip, _data); 2513 update_event_probe(edata);
2498} 2514}
2499 2515
2500static int 2516static int
2501event_enable_print(struct seq_file *m, unsigned long ip, 2517event_enable_print(struct seq_file *m, unsigned long ip,
2502 struct ftrace_probe_ops *ops, void *_data) 2518 struct ftrace_probe_ops *ops, void *data)
2503{ 2519{
2504 struct event_probe_data *data = _data; 2520 struct ftrace_func_mapper *mapper = data;
2521 struct event_probe_data *edata;
2522 void **pdata;
2523
2524 pdata = ftrace_func_mapper_find_ip(mapper, ip);
2525
2526 if (WARN_ON_ONCE(!pdata || !*pdata))
2527 return 0;
2528
2529 edata = *pdata;
2505 2530
2506 seq_printf(m, "%ps:", (void *)ip); 2531 seq_printf(m, "%ps:", (void *)ip);
2507 2532
2508 seq_printf(m, "%s:%s:%s", 2533 seq_printf(m, "%s:%s:%s",
2509 data->enable ? ENABLE_EVENT_STR : DISABLE_EVENT_STR, 2534 edata->enable ? ENABLE_EVENT_STR : DISABLE_EVENT_STR,
2510 data->file->event_call->class->system, 2535 edata->file->event_call->class->system,
2511 trace_event_name(data->file->event_call)); 2536 trace_event_name(edata->file->event_call));
2512 2537
2513 if (data->count == -1) 2538 if (edata->count == -1)
2514 seq_puts(m, ":unlimited\n"); 2539 seq_puts(m, ":unlimited\n");
2515 else 2540 else
2516 seq_printf(m, ":count=%ld\n", data->count); 2541 seq_printf(m, ":count=%ld\n", edata->count);
2517 2542
2518 return 0; 2543 return 0;
2519} 2544}
2520 2545
2521static int 2546static int
2522event_enable_init(struct ftrace_probe_ops *ops, unsigned long ip, 2547event_enable_init(struct ftrace_probe_ops *ops, struct trace_array *tr,
2523 void **_data) 2548 unsigned long ip, void *init_data, void **data)
2524{ 2549{
2525 struct event_probe_data **pdata = (struct event_probe_data **)_data; 2550 struct ftrace_func_mapper *mapper = *data;
2526 struct event_probe_data *data = *pdata; 2551 struct event_probe_data *edata = init_data;
2552 int ret;
2553
2554 if (!mapper) {
2555 mapper = allocate_ftrace_func_mapper();
2556 if (!mapper)
2557 return -ENODEV;
2558 *data = mapper;
2559 }
2560
2561 ret = ftrace_func_mapper_add_ip(mapper, ip, edata);
2562 if (ret < 0)
2563 return ret;
2564
2565 edata->ref++;
2527 2566
2528 data->ref++; 2567 return 0;
2568}
2569
2570static int free_probe_data(void *data)
2571{
2572 struct event_probe_data *edata = data;
2573
2574 edata->ref--;
2575 if (!edata->ref) {
2576 /* Remove the SOFT_MODE flag */
2577 __ftrace_event_enable_disable(edata->file, 0, 1);
2578 module_put(edata->file->event_call->mod);
2579 kfree(edata);
2580 }
2529 return 0; 2581 return 0;
2530} 2582}
2531 2583
2532static void 2584static void
2533event_enable_free(struct ftrace_probe_ops *ops, unsigned long ip, 2585event_enable_free(struct ftrace_probe_ops *ops, struct trace_array *tr,
2534 void **_data) 2586 unsigned long ip, void *data)
2535{ 2587{
2536 struct event_probe_data **pdata = (struct event_probe_data **)_data; 2588 struct ftrace_func_mapper *mapper = data;
2537 struct event_probe_data *data = *pdata; 2589 struct event_probe_data *edata;
2538 2590
2539 if (WARN_ON_ONCE(data->ref <= 0)) 2591 if (!ip) {
2592 if (!mapper)
2593 return;
2594 free_ftrace_func_mapper(mapper, free_probe_data);
2540 return; 2595 return;
2541
2542 data->ref--;
2543 if (!data->ref) {
2544 /* Remove the SOFT_MODE flag */
2545 __ftrace_event_enable_disable(data->file, 0, 1);
2546 module_put(data->file->event_call->mod);
2547 kfree(data);
2548 } 2596 }
2549 *pdata = NULL; 2597
2598 edata = ftrace_func_mapper_remove_ip(mapper, ip);
2599
2600 if (WARN_ON_ONCE(!edata))
2601 return;
2602
2603 if (WARN_ON_ONCE(edata->ref <= 0))
2604 return;
2605
2606 free_probe_data(edata);
2550} 2607}
2551 2608
2552static struct ftrace_probe_ops event_enable_probe_ops = { 2609static struct ftrace_probe_ops event_enable_probe_ops = {
@@ -2578,10 +2635,9 @@ static struct ftrace_probe_ops event_disable_count_probe_ops = {
2578}; 2635};
2579 2636
2580static int 2637static int
2581event_enable_func(struct ftrace_hash *hash, 2638event_enable_func(struct trace_array *tr, struct ftrace_hash *hash,
2582 char *glob, char *cmd, char *param, int enabled) 2639 char *glob, char *cmd, char *param, int enabled)
2583{ 2640{
2584 struct trace_array *tr = top_trace_array();
2585 struct trace_event_file *file; 2641 struct trace_event_file *file;
2586 struct ftrace_probe_ops *ops; 2642 struct ftrace_probe_ops *ops;
2587 struct event_probe_data *data; 2643 struct event_probe_data *data;
@@ -2619,12 +2675,12 @@ event_enable_func(struct ftrace_hash *hash,
2619 ops = param ? &event_disable_count_probe_ops : &event_disable_probe_ops; 2675 ops = param ? &event_disable_count_probe_ops : &event_disable_probe_ops;
2620 2676
2621 if (glob[0] == '!') { 2677 if (glob[0] == '!') {
2622 unregister_ftrace_function_probe_func(glob+1, ops); 2678 ret = unregister_ftrace_function_probe_func(glob+1, tr, ops);
2623 ret = 0;
2624 goto out; 2679 goto out;
2625 } 2680 }
2626 2681
2627 ret = -ENOMEM; 2682 ret = -ENOMEM;
2683
2628 data = kzalloc(sizeof(*data), GFP_KERNEL); 2684 data = kzalloc(sizeof(*data), GFP_KERNEL);
2629 if (!data) 2685 if (!data)
2630 goto out; 2686 goto out;
@@ -2661,7 +2717,8 @@ event_enable_func(struct ftrace_hash *hash,
2661 ret = __ftrace_event_enable_disable(file, 1, 1); 2717 ret = __ftrace_event_enable_disable(file, 1, 1);
2662 if (ret < 0) 2718 if (ret < 0)
2663 goto out_put; 2719 goto out_put;
2664 ret = register_ftrace_function_probe(glob, ops, data); 2720
2721 ret = register_ftrace_function_probe(glob, tr, ops, data);
2665 /* 2722 /*
2666 * The above returns on success the # of functions enabled, 2723 * The above returns on success the # of functions enabled,
2667 * but if it didn't find any functions it returns zero. 2724 * but if it didn't find any functions it returns zero.
diff --git a/kernel/trace/trace_functions.c b/kernel/trace/trace_functions.c
index 0efa00d80623..a3bddbfd0874 100644
--- a/kernel/trace/trace_functions.c
+++ b/kernel/trace/trace_functions.c
@@ -267,10 +267,14 @@ static struct tracer function_trace __tracer_data =
267}; 267};
268 268
269#ifdef CONFIG_DYNAMIC_FTRACE 269#ifdef CONFIG_DYNAMIC_FTRACE
270static void update_traceon_count(void **data, bool on) 270static void update_traceon_count(struct ftrace_probe_ops *ops,
271 unsigned long ip,
272 struct trace_array *tr, bool on,
273 void *data)
271{ 274{
272 long *count = (long *)data; 275 struct ftrace_func_mapper *mapper = data;
273 long old_count = *count; 276 long *count;
277 long old_count;
274 278
275 /* 279 /*
276 * Tracing gets disabled (or enabled) once per count. 280 * Tracing gets disabled (or enabled) once per count.
@@ -301,23 +305,22 @@ static void update_traceon_count(void **data, bool on)
301 * setting the tracing_on file. But we currently don't care 305 * setting the tracing_on file. But we currently don't care
302 * about that. 306 * about that.
303 */ 307 */
304 if (!old_count) 308 count = (long *)ftrace_func_mapper_find_ip(mapper, ip);
309 old_count = *count;
310
311 if (old_count <= 0)
305 return; 312 return;
306 313
307 /* Make sure we see count before checking tracing state */ 314 /* Make sure we see count before checking tracing state */
308 smp_rmb(); 315 smp_rmb();
309 316
310 if (on == !!tracing_is_on()) 317 if (on == !!tracer_tracing_is_on(tr))
311 return; 318 return;
312 319
313 if (on) 320 if (on)
314 tracing_on(); 321 tracer_tracing_on(tr);
315 else 322 else
316 tracing_off(); 323 tracer_tracing_off(tr);
317
318 /* unlimited? */
319 if (old_count == -1)
320 return;
321 324
322 /* Make sure tracing state is visible before updating count */ 325 /* Make sure tracing state is visible before updating count */
323 smp_wmb(); 326 smp_wmb();
@@ -326,33 +329,41 @@ static void update_traceon_count(void **data, bool on)
326} 329}
327 330
328static void 331static void
329ftrace_traceon_count(unsigned long ip, unsigned long parent_ip, void **data) 332ftrace_traceon_count(unsigned long ip, unsigned long parent_ip,
333 struct trace_array *tr, struct ftrace_probe_ops *ops,
334 void *data)
330{ 335{
331 update_traceon_count(data, 1); 336 update_traceon_count(ops, ip, tr, 1, data);
332} 337}
333 338
334static void 339static void
335ftrace_traceoff_count(unsigned long ip, unsigned long parent_ip, void **data) 340ftrace_traceoff_count(unsigned long ip, unsigned long parent_ip,
341 struct trace_array *tr, struct ftrace_probe_ops *ops,
342 void *data)
336{ 343{
337 update_traceon_count(data, 0); 344 update_traceon_count(ops, ip, tr, 0, data);
338} 345}
339 346
340static void 347static void
341ftrace_traceon(unsigned long ip, unsigned long parent_ip, void **data) 348ftrace_traceon(unsigned long ip, unsigned long parent_ip,
349 struct trace_array *tr, struct ftrace_probe_ops *ops,
350 void *data)
342{ 351{
343 if (tracing_is_on()) 352 if (tracer_tracing_is_on(tr))
344 return; 353 return;
345 354
346 tracing_on(); 355 tracer_tracing_on(tr);
347} 356}
348 357
349static void 358static void
350ftrace_traceoff(unsigned long ip, unsigned long parent_ip, void **data) 359ftrace_traceoff(unsigned long ip, unsigned long parent_ip,
360 struct trace_array *tr, struct ftrace_probe_ops *ops,
361 void *data)
351{ 362{
352 if (!tracing_is_on()) 363 if (!tracer_tracing_is_on(tr))
353 return; 364 return;
354 365
355 tracing_off(); 366 tracer_tracing_off(tr);
356} 367}
357 368
358/* 369/*
@@ -364,144 +375,218 @@ ftrace_traceoff(unsigned long ip, unsigned long parent_ip, void **data)
364 */ 375 */
365#define STACK_SKIP 4 376#define STACK_SKIP 4
366 377
378static __always_inline void trace_stack(struct trace_array *tr)
379{
380 unsigned long flags;
381 int pc;
382
383 local_save_flags(flags);
384 pc = preempt_count();
385
386 __trace_stack(tr, flags, STACK_SKIP, pc);
387}
388
367static void 389static void
368ftrace_stacktrace(unsigned long ip, unsigned long parent_ip, void **data) 390ftrace_stacktrace(unsigned long ip, unsigned long parent_ip,
391 struct trace_array *tr, struct ftrace_probe_ops *ops,
392 void *data)
369{ 393{
370 trace_dump_stack(STACK_SKIP); 394 trace_stack(tr);
371} 395}
372 396
373static void 397static void
374ftrace_stacktrace_count(unsigned long ip, unsigned long parent_ip, void **data) 398ftrace_stacktrace_count(unsigned long ip, unsigned long parent_ip,
399 struct trace_array *tr, struct ftrace_probe_ops *ops,
400 void *data)
375{ 401{
376 long *count = (long *)data; 402 struct ftrace_func_mapper *mapper = data;
403 long *count;
377 long old_count; 404 long old_count;
378 long new_count; 405 long new_count;
379 406
407 if (!tracing_is_on())
408 return;
409
410 /* unlimited? */
411 if (!mapper) {
412 trace_stack(tr);
413 return;
414 }
415
416 count = (long *)ftrace_func_mapper_find_ip(mapper, ip);
417
380 /* 418 /*
381 * Stack traces should only execute the number of times the 419 * Stack traces should only execute the number of times the
382 * user specified in the counter. 420 * user specified in the counter.
383 */ 421 */
384 do { 422 do {
385
386 if (!tracing_is_on())
387 return;
388
389 old_count = *count; 423 old_count = *count;
390 424
391 if (!old_count) 425 if (!old_count)
392 return; 426 return;
393 427
394 /* unlimited? */
395 if (old_count == -1) {
396 trace_dump_stack(STACK_SKIP);
397 return;
398 }
399
400 new_count = old_count - 1; 428 new_count = old_count - 1;
401 new_count = cmpxchg(count, old_count, new_count); 429 new_count = cmpxchg(count, old_count, new_count);
402 if (new_count == old_count) 430 if (new_count == old_count)
403 trace_dump_stack(STACK_SKIP); 431 trace_stack(tr);
432
433 if (!tracing_is_on())
434 return;
404 435
405 } while (new_count != old_count); 436 } while (new_count != old_count);
406} 437}
407 438
408static int update_count(void **data) 439static int update_count(struct ftrace_probe_ops *ops, unsigned long ip,
440 void *data)
409{ 441{
410 unsigned long *count = (long *)data; 442 struct ftrace_func_mapper *mapper = data;
443 long *count = NULL;
411 444
412 if (!*count) 445 if (mapper)
413 return 0; 446 count = (long *)ftrace_func_mapper_find_ip(mapper, ip);
414 447
415 if (*count != -1) 448 if (count) {
449 if (*count <= 0)
450 return 0;
416 (*count)--; 451 (*count)--;
452 }
417 453
418 return 1; 454 return 1;
419} 455}
420 456
421static void 457static void
422ftrace_dump_probe(unsigned long ip, unsigned long parent_ip, void **data) 458ftrace_dump_probe(unsigned long ip, unsigned long parent_ip,
459 struct trace_array *tr, struct ftrace_probe_ops *ops,
460 void *data)
423{ 461{
424 if (update_count(data)) 462 if (update_count(ops, ip, data))
425 ftrace_dump(DUMP_ALL); 463 ftrace_dump(DUMP_ALL);
426} 464}
427 465
428/* Only dump the current CPU buffer. */ 466/* Only dump the current CPU buffer. */
429static void 467static void
430ftrace_cpudump_probe(unsigned long ip, unsigned long parent_ip, void **data) 468ftrace_cpudump_probe(unsigned long ip, unsigned long parent_ip,
469 struct trace_array *tr, struct ftrace_probe_ops *ops,
470 void *data)
431{ 471{
432 if (update_count(data)) 472 if (update_count(ops, ip, data))
433 ftrace_dump(DUMP_ORIG); 473 ftrace_dump(DUMP_ORIG);
434} 474}
435 475
436static int 476static int
437ftrace_probe_print(const char *name, struct seq_file *m, 477ftrace_probe_print(const char *name, struct seq_file *m,
438 unsigned long ip, void *data) 478 unsigned long ip, struct ftrace_probe_ops *ops,
479 void *data)
439{ 480{
440 long count = (long)data; 481 struct ftrace_func_mapper *mapper = data;
482 long *count = NULL;
441 483
442 seq_printf(m, "%ps:%s", (void *)ip, name); 484 seq_printf(m, "%ps:%s", (void *)ip, name);
443 485
444 if (count == -1) 486 if (mapper)
445 seq_puts(m, ":unlimited\n"); 487 count = (long *)ftrace_func_mapper_find_ip(mapper, ip);
488
489 if (count)
490 seq_printf(m, ":count=%ld\n", *count);
446 else 491 else
447 seq_printf(m, ":count=%ld\n", count); 492 seq_puts(m, ":unlimited\n");
448 493
449 return 0; 494 return 0;
450} 495}
451 496
452static int 497static int
453ftrace_traceon_print(struct seq_file *m, unsigned long ip, 498ftrace_traceon_print(struct seq_file *m, unsigned long ip,
454 struct ftrace_probe_ops *ops, void *data) 499 struct ftrace_probe_ops *ops,
500 void *data)
455{ 501{
456 return ftrace_probe_print("traceon", m, ip, data); 502 return ftrace_probe_print("traceon", m, ip, ops, data);
457} 503}
458 504
459static int 505static int
460ftrace_traceoff_print(struct seq_file *m, unsigned long ip, 506ftrace_traceoff_print(struct seq_file *m, unsigned long ip,
461 struct ftrace_probe_ops *ops, void *data) 507 struct ftrace_probe_ops *ops, void *data)
462{ 508{
463 return ftrace_probe_print("traceoff", m, ip, data); 509 return ftrace_probe_print("traceoff", m, ip, ops, data);
464} 510}
465 511
466static int 512static int
467ftrace_stacktrace_print(struct seq_file *m, unsigned long ip, 513ftrace_stacktrace_print(struct seq_file *m, unsigned long ip,
468 struct ftrace_probe_ops *ops, void *data) 514 struct ftrace_probe_ops *ops, void *data)
469{ 515{
470 return ftrace_probe_print("stacktrace", m, ip, data); 516 return ftrace_probe_print("stacktrace", m, ip, ops, data);
471} 517}
472 518
473static int 519static int
474ftrace_dump_print(struct seq_file *m, unsigned long ip, 520ftrace_dump_print(struct seq_file *m, unsigned long ip,
475 struct ftrace_probe_ops *ops, void *data) 521 struct ftrace_probe_ops *ops, void *data)
476{ 522{
477 return ftrace_probe_print("dump", m, ip, data); 523 return ftrace_probe_print("dump", m, ip, ops, data);
478} 524}
479 525
480static int 526static int
481ftrace_cpudump_print(struct seq_file *m, unsigned long ip, 527ftrace_cpudump_print(struct seq_file *m, unsigned long ip,
482 struct ftrace_probe_ops *ops, void *data) 528 struct ftrace_probe_ops *ops, void *data)
483{ 529{
484 return ftrace_probe_print("cpudump", m, ip, data); 530 return ftrace_probe_print("cpudump", m, ip, ops, data);
531}
532
533
534static int
535ftrace_count_init(struct ftrace_probe_ops *ops, struct trace_array *tr,
536 unsigned long ip, void *init_data, void **data)
537{
538 struct ftrace_func_mapper *mapper = *data;
539
540 if (!mapper) {
541 mapper = allocate_ftrace_func_mapper();
542 if (!mapper)
543 return -ENOMEM;
544 *data = mapper;
545 }
546
547 return ftrace_func_mapper_add_ip(mapper, ip, init_data);
548}
549
550static void
551ftrace_count_free(struct ftrace_probe_ops *ops, struct trace_array *tr,
552 unsigned long ip, void *data)
553{
554 struct ftrace_func_mapper *mapper = data;
555
556 if (!ip) {
557 free_ftrace_func_mapper(mapper, NULL);
558 return;
559 }
560
561 ftrace_func_mapper_remove_ip(mapper, ip);
485} 562}
486 563
487static struct ftrace_probe_ops traceon_count_probe_ops = { 564static struct ftrace_probe_ops traceon_count_probe_ops = {
488 .func = ftrace_traceon_count, 565 .func = ftrace_traceon_count,
489 .print = ftrace_traceon_print, 566 .print = ftrace_traceon_print,
567 .init = ftrace_count_init,
568 .free = ftrace_count_free,
490}; 569};
491 570
492static struct ftrace_probe_ops traceoff_count_probe_ops = { 571static struct ftrace_probe_ops traceoff_count_probe_ops = {
493 .func = ftrace_traceoff_count, 572 .func = ftrace_traceoff_count,
494 .print = ftrace_traceoff_print, 573 .print = ftrace_traceoff_print,
574 .init = ftrace_count_init,
575 .free = ftrace_count_free,
495}; 576};
496 577
497static struct ftrace_probe_ops stacktrace_count_probe_ops = { 578static struct ftrace_probe_ops stacktrace_count_probe_ops = {
498 .func = ftrace_stacktrace_count, 579 .func = ftrace_stacktrace_count,
499 .print = ftrace_stacktrace_print, 580 .print = ftrace_stacktrace_print,
581 .init = ftrace_count_init,
582 .free = ftrace_count_free,
500}; 583};
501 584
502static struct ftrace_probe_ops dump_probe_ops = { 585static struct ftrace_probe_ops dump_probe_ops = {
503 .func = ftrace_dump_probe, 586 .func = ftrace_dump_probe,
504 .print = ftrace_dump_print, 587 .print = ftrace_dump_print,
588 .init = ftrace_count_init,
589 .free = ftrace_count_free,
505}; 590};
506 591
507static struct ftrace_probe_ops cpudump_probe_ops = { 592static struct ftrace_probe_ops cpudump_probe_ops = {
@@ -525,7 +610,8 @@ static struct ftrace_probe_ops stacktrace_probe_ops = {
525}; 610};
526 611
527static int 612static int
528ftrace_trace_probe_callback(struct ftrace_probe_ops *ops, 613ftrace_trace_probe_callback(struct trace_array *tr,
614 struct ftrace_probe_ops *ops,
529 struct ftrace_hash *hash, char *glob, 615 struct ftrace_hash *hash, char *glob,
530 char *cmd, char *param, int enable) 616 char *cmd, char *param, int enable)
531{ 617{
@@ -537,10 +623,8 @@ ftrace_trace_probe_callback(struct ftrace_probe_ops *ops,
537 if (!enable) 623 if (!enable)
538 return -EINVAL; 624 return -EINVAL;
539 625
540 if (glob[0] == '!') { 626 if (glob[0] == '!')
541 unregister_ftrace_function_probe_func(glob+1, ops); 627 return unregister_ftrace_function_probe_func(glob+1, tr, ops);
542 return 0;
543 }
544 628
545 if (!param) 629 if (!param)
546 goto out_reg; 630 goto out_reg;
@@ -559,13 +643,13 @@ ftrace_trace_probe_callback(struct ftrace_probe_ops *ops,
559 return ret; 643 return ret;
560 644
561 out_reg: 645 out_reg:
562 ret = register_ftrace_function_probe(glob, ops, count); 646 ret = register_ftrace_function_probe(glob, tr, ops, count);
563 647
564 return ret < 0 ? ret : 0; 648 return ret < 0 ? ret : 0;
565} 649}
566 650
567static int 651static int
568ftrace_trace_onoff_callback(struct ftrace_hash *hash, 652ftrace_trace_onoff_callback(struct trace_array *tr, struct ftrace_hash *hash,
569 char *glob, char *cmd, char *param, int enable) 653 char *glob, char *cmd, char *param, int enable)
570{ 654{
571 struct ftrace_probe_ops *ops; 655 struct ftrace_probe_ops *ops;
@@ -576,24 +660,24 @@ ftrace_trace_onoff_callback(struct ftrace_hash *hash,
576 else 660 else
577 ops = param ? &traceoff_count_probe_ops : &traceoff_probe_ops; 661 ops = param ? &traceoff_count_probe_ops : &traceoff_probe_ops;
578 662
579 return ftrace_trace_probe_callback(ops, hash, glob, cmd, 663 return ftrace_trace_probe_callback(tr, ops, hash, glob, cmd,
580 param, enable); 664 param, enable);
581} 665}
582 666
583static int 667static int
584ftrace_stacktrace_callback(struct ftrace_hash *hash, 668ftrace_stacktrace_callback(struct trace_array *tr, struct ftrace_hash *hash,
585 char *glob, char *cmd, char *param, int enable) 669 char *glob, char *cmd, char *param, int enable)
586{ 670{
587 struct ftrace_probe_ops *ops; 671 struct ftrace_probe_ops *ops;
588 672
589 ops = param ? &stacktrace_count_probe_ops : &stacktrace_probe_ops; 673 ops = param ? &stacktrace_count_probe_ops : &stacktrace_probe_ops;
590 674
591 return ftrace_trace_probe_callback(ops, hash, glob, cmd, 675 return ftrace_trace_probe_callback(tr, ops, hash, glob, cmd,
592 param, enable); 676 param, enable);
593} 677}
594 678
595static int 679static int
596ftrace_dump_callback(struct ftrace_hash *hash, 680ftrace_dump_callback(struct trace_array *tr, struct ftrace_hash *hash,
597 char *glob, char *cmd, char *param, int enable) 681 char *glob, char *cmd, char *param, int enable)
598{ 682{
599 struct ftrace_probe_ops *ops; 683 struct ftrace_probe_ops *ops;
@@ -601,12 +685,12 @@ ftrace_dump_callback(struct ftrace_hash *hash,
601 ops = &dump_probe_ops; 685 ops = &dump_probe_ops;
602 686
603 /* Only dump once. */ 687 /* Only dump once. */
604 return ftrace_trace_probe_callback(ops, hash, glob, cmd, 688 return ftrace_trace_probe_callback(tr, ops, hash, glob, cmd,
605 "1", enable); 689 "1", enable);
606} 690}
607 691
608static int 692static int
609ftrace_cpudump_callback(struct ftrace_hash *hash, 693ftrace_cpudump_callback(struct trace_array *tr, struct ftrace_hash *hash,
610 char *glob, char *cmd, char *param, int enable) 694 char *glob, char *cmd, char *param, int enable)
611{ 695{
612 struct ftrace_probe_ops *ops; 696 struct ftrace_probe_ops *ops;
@@ -614,7 +698,7 @@ ftrace_cpudump_callback(struct ftrace_hash *hash,
614 ops = &cpudump_probe_ops; 698 ops = &cpudump_probe_ops;
615 699
616 /* Only dump once. */ 700 /* Only dump once. */
617 return ftrace_trace_probe_callback(ops, hash, glob, cmd, 701 return ftrace_trace_probe_callback(tr, ops, hash, glob, cmd,
618 "1", enable); 702 "1", enable);
619} 703}
620 704
@@ -687,9 +771,8 @@ static inline int init_func_cmd_traceon(void)
687} 771}
688#endif /* CONFIG_DYNAMIC_FTRACE */ 772#endif /* CONFIG_DYNAMIC_FTRACE */
689 773
690static __init int init_function_trace(void) 774__init int init_function_trace(void)
691{ 775{
692 init_func_cmd_traceon(); 776 init_func_cmd_traceon();
693 return register_tracer(&function_trace); 777 return register_tracer(&function_trace);
694} 778}
695core_initcall(init_function_trace);
diff --git a/kernel/trace/trace_kprobe.c b/kernel/trace/trace_kprobe.c
index 013f4e7146d4..8485f6738a87 100644
--- a/kernel/trace/trace_kprobe.c
+++ b/kernel/trace/trace_kprobe.c
@@ -25,6 +25,7 @@
25#include "trace_probe.h" 25#include "trace_probe.h"
26 26
27#define KPROBE_EVENT_SYSTEM "kprobes" 27#define KPROBE_EVENT_SYSTEM "kprobes"
28#define KRETPROBE_MAXACTIVE_MAX 4096
28 29
29/** 30/**
30 * Kprobe event core functions 31 * Kprobe event core functions
@@ -282,6 +283,7 @@ static struct trace_kprobe *alloc_trace_kprobe(const char *group,
282 void *addr, 283 void *addr,
283 const char *symbol, 284 const char *symbol,
284 unsigned long offs, 285 unsigned long offs,
286 int maxactive,
285 int nargs, bool is_return) 287 int nargs, bool is_return)
286{ 288{
287 struct trace_kprobe *tk; 289 struct trace_kprobe *tk;
@@ -309,6 +311,8 @@ static struct trace_kprobe *alloc_trace_kprobe(const char *group,
309 else 311 else
310 tk->rp.kp.pre_handler = kprobe_dispatcher; 312 tk->rp.kp.pre_handler = kprobe_dispatcher;
311 313
314 tk->rp.maxactive = maxactive;
315
312 if (!event || !is_good_name(event)) { 316 if (!event || !is_good_name(event)) {
313 ret = -EINVAL; 317 ret = -EINVAL;
314 goto error; 318 goto error;
@@ -598,8 +602,10 @@ static int create_trace_kprobe(int argc, char **argv)
598{ 602{
599 /* 603 /*
600 * Argument syntax: 604 * Argument syntax:
601 * - Add kprobe: p[:[GRP/]EVENT] [MOD:]KSYM[+OFFS]|KADDR [FETCHARGS] 605 * - Add kprobe:
602 * - Add kretprobe: r[:[GRP/]EVENT] [MOD:]KSYM[+0] [FETCHARGS] 606 * p[:[GRP/]EVENT] [MOD:]KSYM[+OFFS]|KADDR [FETCHARGS]
607 * - Add kretprobe:
608 * r[MAXACTIVE][:[GRP/]EVENT] [MOD:]KSYM[+0] [FETCHARGS]
603 * Fetch args: 609 * Fetch args:
604 * $retval : fetch return value 610 * $retval : fetch return value
605 * $stack : fetch stack address 611 * $stack : fetch stack address
@@ -619,6 +625,7 @@ static int create_trace_kprobe(int argc, char **argv)
619 int i, ret = 0; 625 int i, ret = 0;
620 bool is_return = false, is_delete = false; 626 bool is_return = false, is_delete = false;
621 char *symbol = NULL, *event = NULL, *group = NULL; 627 char *symbol = NULL, *event = NULL, *group = NULL;
628 int maxactive = 0;
622 char *arg; 629 char *arg;
623 unsigned long offset = 0; 630 unsigned long offset = 0;
624 void *addr = NULL; 631 void *addr = NULL;
@@ -637,8 +644,28 @@ static int create_trace_kprobe(int argc, char **argv)
637 return -EINVAL; 644 return -EINVAL;
638 } 645 }
639 646
640 if (argv[0][1] == ':') { 647 event = strchr(&argv[0][1], ':');
641 event = &argv[0][2]; 648 if (event) {
649 event[0] = '\0';
650 event++;
651 }
652 if (is_return && isdigit(argv[0][1])) {
653 ret = kstrtouint(&argv[0][1], 0, &maxactive);
654 if (ret) {
655 pr_info("Failed to parse maxactive.\n");
656 return ret;
657 }
658 /* kretprobes instances are iterated over via a list. The
659 * maximum should stay reasonable.
660 */
661 if (maxactive > KRETPROBE_MAXACTIVE_MAX) {
662 pr_info("Maxactive is too big (%d > %d).\n",
663 maxactive, KRETPROBE_MAXACTIVE_MAX);
664 return -E2BIG;
665 }
666 }
667
668 if (event) {
642 if (strchr(event, '/')) { 669 if (strchr(event, '/')) {
643 group = event; 670 group = event;
644 event = strchr(group, '/') + 1; 671 event = strchr(group, '/') + 1;
@@ -715,8 +742,8 @@ static int create_trace_kprobe(int argc, char **argv)
715 is_return ? 'r' : 'p', addr); 742 is_return ? 'r' : 'p', addr);
716 event = buf; 743 event = buf;
717 } 744 }
718 tk = alloc_trace_kprobe(group, event, addr, symbol, offset, argc, 745 tk = alloc_trace_kprobe(group, event, addr, symbol, offset, maxactive,
719 is_return); 746 argc, is_return);
720 if (IS_ERR(tk)) { 747 if (IS_ERR(tk)) {
721 pr_info("Failed to allocate trace_probe.(%d)\n", 748 pr_info("Failed to allocate trace_probe.(%d)\n",
722 (int)PTR_ERR(tk)); 749 (int)PTR_ERR(tk));
diff --git a/kernel/trace/trace_stack.c b/kernel/trace/trace_stack.c
index 5fb1f2c87e6b..76aa04d4c925 100644
--- a/kernel/trace/trace_stack.c
+++ b/kernel/trace/trace_stack.c
@@ -35,7 +35,7 @@ unsigned long stack_trace_max_size;
35arch_spinlock_t stack_trace_max_lock = 35arch_spinlock_t stack_trace_max_lock =
36 (arch_spinlock_t)__ARCH_SPIN_LOCK_UNLOCKED; 36 (arch_spinlock_t)__ARCH_SPIN_LOCK_UNLOCKED;
37 37
38static DEFINE_PER_CPU(int, trace_active); 38DEFINE_PER_CPU(int, disable_stack_tracer);
39static DEFINE_MUTEX(stack_sysctl_mutex); 39static DEFINE_MUTEX(stack_sysctl_mutex);
40 40
41int stack_tracer_enabled; 41int stack_tracer_enabled;
@@ -96,6 +96,14 @@ check_stack(unsigned long ip, unsigned long *stack)
96 if (in_nmi()) 96 if (in_nmi())
97 return; 97 return;
98 98
99 /*
100 * There's a slight chance that we are tracing inside the
101 * RCU infrastructure, and rcu_irq_enter() will not work
102 * as expected.
103 */
104 if (unlikely(rcu_irq_enter_disabled()))
105 return;
106
99 local_irq_save(flags); 107 local_irq_save(flags);
100 arch_spin_lock(&stack_trace_max_lock); 108 arch_spin_lock(&stack_trace_max_lock);
101 109
@@ -207,13 +215,12 @@ stack_trace_call(unsigned long ip, unsigned long parent_ip,
207 struct ftrace_ops *op, struct pt_regs *pt_regs) 215 struct ftrace_ops *op, struct pt_regs *pt_regs)
208{ 216{
209 unsigned long stack; 217 unsigned long stack;
210 int cpu;
211 218
212 preempt_disable_notrace(); 219 preempt_disable_notrace();
213 220
214 cpu = raw_smp_processor_id();
215 /* no atomic needed, we only modify this variable by this cpu */ 221 /* no atomic needed, we only modify this variable by this cpu */
216 if (per_cpu(trace_active, cpu)++ != 0) 222 __this_cpu_inc(disable_stack_tracer);
223 if (__this_cpu_read(disable_stack_tracer) != 1)
217 goto out; 224 goto out;
218 225
219 ip += MCOUNT_INSN_SIZE; 226 ip += MCOUNT_INSN_SIZE;
@@ -221,7 +228,7 @@ stack_trace_call(unsigned long ip, unsigned long parent_ip,
221 check_stack(ip, &stack); 228 check_stack(ip, &stack);
222 229
223 out: 230 out:
224 per_cpu(trace_active, cpu)--; 231 __this_cpu_dec(disable_stack_tracer);
225 /* prevent recursion in schedule */ 232 /* prevent recursion in schedule */
226 preempt_enable_notrace(); 233 preempt_enable_notrace();
227} 234}
@@ -253,7 +260,6 @@ stack_max_size_write(struct file *filp, const char __user *ubuf,
253 long *ptr = filp->private_data; 260 long *ptr = filp->private_data;
254 unsigned long val, flags; 261 unsigned long val, flags;
255 int ret; 262 int ret;
256 int cpu;
257 263
258 ret = kstrtoul_from_user(ubuf, count, 10, &val); 264 ret = kstrtoul_from_user(ubuf, count, 10, &val);
259 if (ret) 265 if (ret)
@@ -264,16 +270,15 @@ stack_max_size_write(struct file *filp, const char __user *ubuf,
264 /* 270 /*
265 * In case we trace inside arch_spin_lock() or after (NMI), 271 * In case we trace inside arch_spin_lock() or after (NMI),
266 * we will cause circular lock, so we also need to increase 272 * we will cause circular lock, so we also need to increase
267 * the percpu trace_active here. 273 * the percpu disable_stack_tracer here.
268 */ 274 */
269 cpu = smp_processor_id(); 275 __this_cpu_inc(disable_stack_tracer);
270 per_cpu(trace_active, cpu)++;
271 276
272 arch_spin_lock(&stack_trace_max_lock); 277 arch_spin_lock(&stack_trace_max_lock);
273 *ptr = val; 278 *ptr = val;
274 arch_spin_unlock(&stack_trace_max_lock); 279 arch_spin_unlock(&stack_trace_max_lock);
275 280
276 per_cpu(trace_active, cpu)--; 281 __this_cpu_dec(disable_stack_tracer);
277 local_irq_restore(flags); 282 local_irq_restore(flags);
278 283
279 return count; 284 return count;
@@ -307,12 +312,9 @@ t_next(struct seq_file *m, void *v, loff_t *pos)
307 312
308static void *t_start(struct seq_file *m, loff_t *pos) 313static void *t_start(struct seq_file *m, loff_t *pos)
309{ 314{
310 int cpu;
311
312 local_irq_disable(); 315 local_irq_disable();
313 316
314 cpu = smp_processor_id(); 317 __this_cpu_inc(disable_stack_tracer);
315 per_cpu(trace_active, cpu)++;
316 318
317 arch_spin_lock(&stack_trace_max_lock); 319 arch_spin_lock(&stack_trace_max_lock);
318 320
@@ -324,12 +326,9 @@ static void *t_start(struct seq_file *m, loff_t *pos)
324 326
325static void t_stop(struct seq_file *m, void *p) 327static void t_stop(struct seq_file *m, void *p)
326{ 328{
327 int cpu;
328
329 arch_spin_unlock(&stack_trace_max_lock); 329 arch_spin_unlock(&stack_trace_max_lock);
330 330
331 cpu = smp_processor_id(); 331 __this_cpu_dec(disable_stack_tracer);
332 per_cpu(trace_active, cpu)--;
333 332
334 local_irq_enable(); 333 local_irq_enable();
335} 334}
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 1e2af704938d..2c25de46c58f 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -65,6 +65,7 @@
65#include <linux/page_owner.h> 65#include <linux/page_owner.h>
66#include <linux/kthread.h> 66#include <linux/kthread.h>
67#include <linux/memcontrol.h> 67#include <linux/memcontrol.h>
68#include <linux/ftrace.h>
68 69
69#include <asm/sections.h> 70#include <asm/sections.h>
70#include <asm/tlbflush.h> 71#include <asm/tlbflush.h>
diff --git a/scripts/recordmcount.c b/scripts/recordmcount.c
index aeb34223167c..16e086dcc567 100644
--- a/scripts/recordmcount.c
+++ b/scripts/recordmcount.c
@@ -412,6 +412,7 @@ static int
412is_mcounted_section_name(char const *const txtname) 412is_mcounted_section_name(char const *const txtname)
413{ 413{
414 return strcmp(".text", txtname) == 0 || 414 return strcmp(".text", txtname) == 0 ||
415 strcmp(".init.text", txtname) == 0 ||
415 strcmp(".ref.text", txtname) == 0 || 416 strcmp(".ref.text", txtname) == 0 ||
416 strcmp(".sched.text", txtname) == 0 || 417 strcmp(".sched.text", txtname) == 0 ||
417 strcmp(".spinlock.text", txtname) == 0 || 418 strcmp(".spinlock.text", txtname) == 0 ||
diff --git a/scripts/recordmcount.pl b/scripts/recordmcount.pl
index 0b6002b36f20..1633c3e6c0b9 100755
--- a/scripts/recordmcount.pl
+++ b/scripts/recordmcount.pl
@@ -130,6 +130,7 @@ if ($inputfile =~ m,kernel/trace/ftrace\.o$,) {
130# Acceptable sections to record. 130# Acceptable sections to record.
131my %text_sections = ( 131my %text_sections = (
132 ".text" => 1, 132 ".text" => 1,
133 ".init.text" => 1,
133 ".ref.text" => 1, 134 ".ref.text" => 1,
134 ".sched.text" => 1, 135 ".sched.text" => 1,
135 ".spinlock.text" => 1, 136 ".spinlock.text" => 1,
diff --git a/tools/testing/selftests/ftrace/ftracetest b/tools/testing/selftests/ftrace/ftracetest
index 52e3c4df28d6..32e6211e1c6e 100755
--- a/tools/testing/selftests/ftrace/ftracetest
+++ b/tools/testing/selftests/ftrace/ftracetest
@@ -16,6 +16,7 @@ echo " -k|--keep Keep passed test logs"
16echo " -v|--verbose Increase verbosity of test messages" 16echo " -v|--verbose Increase verbosity of test messages"
17echo " -vv Alias of -v -v (Show all results in stdout)" 17echo " -vv Alias of -v -v (Show all results in stdout)"
18echo " -d|--debug Debug mode (trace all shell commands)" 18echo " -d|--debug Debug mode (trace all shell commands)"
19echo " -l|--logdir <dir> Save logs on the <dir>"
19exit $1 20exit $1
20} 21}
21 22
@@ -64,6 +65,10 @@ parse_opts() { # opts
64 DEBUG=1 65 DEBUG=1
65 shift 1 66 shift 1
66 ;; 67 ;;
68 --logdir|-l)
69 LOG_DIR=$2
70 shift 2
71 ;;
67 *.tc) 72 *.tc)
68 if [ -f "$1" ]; then 73 if [ -f "$1" ]; then
69 OPT_TEST_CASES="$OPT_TEST_CASES `abspath $1`" 74 OPT_TEST_CASES="$OPT_TEST_CASES `abspath $1`"
@@ -145,11 +150,16 @@ XFAILED_CASES=
145UNDEFINED_CASES= 150UNDEFINED_CASES=
146TOTAL_RESULT=0 151TOTAL_RESULT=0
147 152
153INSTANCE=
148CASENO=0 154CASENO=0
149testcase() { # testfile 155testcase() { # testfile
150 CASENO=$((CASENO+1)) 156 CASENO=$((CASENO+1))
151 desc=`grep "^#[ \t]*description:" $1 | cut -f2 -d:` 157 desc=`grep "^#[ \t]*description:" $1 | cut -f2 -d:`
152 prlog -n "[$CASENO]$desc" 158 prlog -n "[$CASENO]$INSTANCE$desc"
159}
160
161test_on_instance() { # testfile
162 grep -q "^#[ \t]*flags:.*instance" $1
153} 163}
154 164
155eval_result() { # sigval 165eval_result() { # sigval
@@ -266,6 +276,17 @@ for t in $TEST_CASES; do
266 run_test $t 276 run_test $t
267done 277done
268 278
279# Test on instance loop
280INSTANCE=" (instance) "
281for t in $TEST_CASES; do
282 test_on_instance $t || continue
283 SAVED_TRACING_DIR=$TRACING_DIR
284 export TRACING_DIR=`mktemp -d $TRACING_DIR/instances/ftracetest.XXXXXX`
285 run_test $t
286 rmdir $TRACING_DIR
287 TRACING_DIR=$SAVED_TRACING_DIR
288done
289
269prlog "" 290prlog ""
270prlog "# of passed: " `echo $PASSED_CASES | wc -w` 291prlog "# of passed: " `echo $PASSED_CASES | wc -w`
271prlog "# of failed: " `echo $FAILED_CASES | wc -w` 292prlog "# of failed: " `echo $FAILED_CASES | wc -w`
diff --git a/tools/testing/selftests/ftrace/test.d/00basic/basic2.tc b/tools/testing/selftests/ftrace/test.d/00basic/basic2.tc
index bf9a7b037924..ebfce83f35b4 100644
--- a/tools/testing/selftests/ftrace/test.d/00basic/basic2.tc
+++ b/tools/testing/selftests/ftrace/test.d/00basic/basic2.tc
@@ -1,5 +1,6 @@
1#!/bin/sh 1#!/bin/sh
2# description: Basic test for tracers 2# description: Basic test for tracers
3# flags: instance
3test -f available_tracers 4test -f available_tracers
4for t in `cat available_tracers`; do 5for t in `cat available_tracers`; do
5 echo $t > current_tracer 6 echo $t > current_tracer
diff --git a/tools/testing/selftests/ftrace/test.d/00basic/basic3.tc b/tools/testing/selftests/ftrace/test.d/00basic/basic3.tc
index bde6625d9785..9e33f841812f 100644
--- a/tools/testing/selftests/ftrace/test.d/00basic/basic3.tc
+++ b/tools/testing/selftests/ftrace/test.d/00basic/basic3.tc
@@ -1,5 +1,6 @@
1#!/bin/sh 1#!/bin/sh
2# description: Basic trace clock test 2# description: Basic trace clock test
3# flags: instance
3test -f trace_clock 4test -f trace_clock
4for c in `cat trace_clock | tr -d \[\]`; do 5for c in `cat trace_clock | tr -d \[\]`; do
5 echo $c > trace_clock 6 echo $c > trace_clock
diff --git a/tools/testing/selftests/ftrace/test.d/event/event-enable.tc b/tools/testing/selftests/ftrace/test.d/event/event-enable.tc
index 87eb9d6dd4ca..283b45ecb199 100644
--- a/tools/testing/selftests/ftrace/test.d/event/event-enable.tc
+++ b/tools/testing/selftests/ftrace/test.d/event/event-enable.tc
@@ -1,5 +1,6 @@
1#!/bin/sh 1#!/bin/sh
2# description: event tracing - enable/disable with event level files 2# description: event tracing - enable/disable with event level files
3# flags: instance
3 4
4do_reset() { 5do_reset() {
5 echo > set_event 6 echo > set_event
diff --git a/tools/testing/selftests/ftrace/test.d/event/event-pid.tc b/tools/testing/selftests/ftrace/test.d/event/event-pid.tc
index d4ab27b522f8..96c1a95be4f7 100644
--- a/tools/testing/selftests/ftrace/test.d/event/event-pid.tc
+++ b/tools/testing/selftests/ftrace/test.d/event/event-pid.tc
@@ -1,5 +1,6 @@
1#!/bin/sh 1#!/bin/sh
2# description: event tracing - restricts events based on pid 2# description: event tracing - restricts events based on pid
3# flags: instance
3 4
4do_reset() { 5do_reset() {
5 echo > set_event 6 echo > set_event
diff --git a/tools/testing/selftests/ftrace/test.d/event/subsystem-enable.tc b/tools/testing/selftests/ftrace/test.d/event/subsystem-enable.tc
index ced27ef0638f..b8fe2e5b9e67 100644
--- a/tools/testing/selftests/ftrace/test.d/event/subsystem-enable.tc
+++ b/tools/testing/selftests/ftrace/test.d/event/subsystem-enable.tc
@@ -1,5 +1,6 @@
1#!/bin/sh 1#!/bin/sh
2# description: event tracing - enable/disable with subsystem level files 2# description: event tracing - enable/disable with subsystem level files
3# flags: instance
3 4
4do_reset() { 5do_reset() {
5 echo > set_event 6 echo > set_event
diff --git a/tools/testing/selftests/ftrace/test.d/ftrace/func_event_triggers.tc b/tools/testing/selftests/ftrace/test.d/ftrace/func_event_triggers.tc
new file mode 100644
index 000000000000..07bb3e5930b4
--- /dev/null
+++ b/tools/testing/selftests/ftrace/test.d/ftrace/func_event_triggers.tc
@@ -0,0 +1,114 @@
1#!/bin/sh
2# description: ftrace - test for function event triggers
3# flags: instance
4#
5# Ftrace allows to add triggers to functions, such as enabling or disabling
6# tracing, enabling or disabling trace events, or recording a stack trace
7# within the ring buffer.
8#
9# This test is designed to test event triggers
10#
11
12# The triggers are set within the set_ftrace_filter file
13if [ ! -f set_ftrace_filter ]; then
14 echo "set_ftrace_filter not found? Is dynamic ftrace not set?"
15 exit_unsupported
16fi
17
18do_reset() {
19 reset_ftrace_filter
20 reset_tracer
21 disable_events
22 clear_trace
23 enable_tracing
24}
25
26fail() { # mesg
27 do_reset
28 echo $1
29 exit $FAIL
30}
31
32SLEEP_TIME=".1"
33
34do_reset
35
36echo "Testing function probes with events:"
37
38EVENT="sched:sched_switch"
39EVENT_ENABLE="events/sched/sched_switch/enable"
40
41cnt_trace() {
42 grep -v '^#' trace | wc -l
43}
44
45test_event_enabled() {
46 val=$1
47
48 e=`cat $EVENT_ENABLE`
49 if [ "$e" != $val ]; then
50 echo "Expected $val but found $e"
51 exit -1
52 fi
53}
54
55run_enable_disable() {
56 enable=$1 # enable
57 Enable=$2 # Enable
58 check_disable=$3 # 0
59 check_enable_star=$4 # 1*
60 check_disable_star=$5 # 0*
61
62 cnt=`cnt_trace`
63 if [ $cnt -ne 0 ]; then
64 fail "Found junk in trace file"
65 fi
66
67 echo "$Enable event all the time"
68
69 echo $check_disable > $EVENT_ENABLE
70 sleep $SLEEP_TIME
71
72 test_event_enabled $check_disable
73
74 echo "schedule:${enable}_event:$EVENT" > set_ftrace_filter
75
76 echo " make sure it works 5 times"
77
78 for i in `seq 5`; do
79 sleep $SLEEP_TIME
80 echo " test $i"
81 test_event_enabled $check_enable_star
82
83 echo $check_disable > $EVENT_ENABLE
84 done
85 sleep $SLEEP_TIME
86 echo " make sure it's still works"
87 test_event_enabled $check_enable_star
88
89 reset_ftrace_filter
90
91 echo " make sure it only works 3 times"
92
93 echo $check_disable > $EVENT_ENABLE
94 sleep $SLEEP_TIME
95
96 echo "schedule:${enable}_event:$EVENT:3" > set_ftrace_filter
97
98 for i in `seq 3`; do
99 sleep $SLEEP_TIME
100 echo " test $i"
101 test_event_enabled $check_enable_star
102
103 echo $check_disable > $EVENT_ENABLE
104 done
105
106 sleep $SLEEP_TIME
107 echo " make sure it stop working"
108 test_event_enabled $check_disable_star
109
110 do_reset
111}
112
113run_enable_disable enable Enable 0 "1*" "0*"
114run_enable_disable disable Disable 1 "0*" "1*"
diff --git a/tools/testing/selftests/ftrace/test.d/ftrace/func_set_ftrace_file.tc b/tools/testing/selftests/ftrace/test.d/ftrace/func_set_ftrace_file.tc
new file mode 100644
index 000000000000..113b4d9bc733
--- /dev/null
+++ b/tools/testing/selftests/ftrace/test.d/ftrace/func_set_ftrace_file.tc
@@ -0,0 +1,132 @@
1#!/bin/sh
2# description: ftrace - test reading of set_ftrace_filter
3#
4# The set_ftrace_filter file of ftrace is used to list functions as well as
5# triggers (probes) attached to functions. The code to read this file is not
6# straight forward and has had various bugs in the past. This test is designed
7# to add functions and triggers to that file in various ways and read that
8# file in various ways (cat vs dd).
9#
10
11# The triggers are set within the set_ftrace_filter file
12if [ ! -f set_ftrace_filter ]; then
13 echo "set_ftrace_filter not found? Is dynamic ftrace not set?"
14 exit_unsupported
15fi
16
17do_reset() {
18 reset_tracer
19 reset_ftrace_filter
20 disable_events
21 clear_trace
22 enable_tracing
23}
24
25fail() { # mesg
26 do_reset
27 echo $1
28 exit $FAIL
29}
30
31do_reset
32
33FILTER=set_ftrace_filter
34FUNC1="schedule"
35FUNC2="do_IRQ"
36
37ALL_FUNCS="#### all functions enabled ####"
38
39test_func() {
40 if ! echo "$1" | grep -q "^$2\$"; then
41 return 0
42 fi
43 echo "$1" | grep -v "^$2\$"
44 return 1
45}
46
47check_set_ftrace_filter() {
48 cat=`cat $FILTER`
49 dd1=`dd if=$FILTER bs=1 | grep -v -e 'records in' -e 'records out' -e 'bytes copied'`
50 dd100=`dd if=$FILTER bs=100 | grep -v -e 'records in' -e 'records out' -e 'bytes copied'`
51
52 echo "Testing '$@'"
53
54 while [ $# -gt 0 ]; do
55 echo "test $1"
56 if cat=`test_func "$cat" "$1"`; then
57 return 0
58 fi
59 if dd1=`test_func "$dd1" "$1"`; then
60 return 0
61 fi
62 if dd100=`test_func "$dd100" "$1"`; then
63 return 0
64 fi
65 shift
66 done
67
68 if [ -n "$cat" ]; then
69 return 0
70 fi
71 if [ -n "$dd1" ]; then
72 return 0
73 fi
74 if [ -n "$dd100" ]; then
75 return 0
76 fi
77 return 1;
78}
79
80if check_set_ftrace_filter "$ALL_FUNCS"; then
81 fail "Expected only $ALL_FUNCS"
82fi
83
84echo "$FUNC1:traceoff" > set_ftrace_filter
85if check_set_ftrace_filter "$ALL_FUNCS" "$FUNC1:traceoff:unlimited"; then
86 fail "Expected $ALL_FUNCS and $FUNC1:traceoff:unlimited"
87fi
88
89echo "$FUNC1" > set_ftrace_filter
90if check_set_ftrace_filter "$FUNC1" "$FUNC1:traceoff:unlimited"; then
91 fail "Expected $FUNC1 and $FUNC1:traceoff:unlimited"
92fi
93
94echo "$FUNC2" >> set_ftrace_filter
95if check_set_ftrace_filter "$FUNC1" "$FUNC2" "$FUNC1:traceoff:unlimited"; then
96 fail "Expected $FUNC1 $FUNC2 and $FUNC1:traceoff:unlimited"
97fi
98
99echo "$FUNC2:traceoff" >> set_ftrace_filter
100if check_set_ftrace_filter "$FUNC1" "$FUNC2" "$FUNC1:traceoff:unlimited" "$FUNC2:traceoff:unlimited"; then
101 fail "Expected $FUNC1 $FUNC2 $FUNC1:traceoff:unlimited and $FUNC2:traceoff:unlimited"
102fi
103
104echo "$FUNC1" > set_ftrace_filter
105if check_set_ftrace_filter "$FUNC1" "$FUNC1:traceoff:unlimited" "$FUNC2:traceoff:unlimited"; then
106 fail "Expected $FUNC1 $FUNC1:traceoff:unlimited and $FUNC2:traceoff:unlimited"
107fi
108
109echo > set_ftrace_filter
110if check_set_ftrace_filter "$ALL_FUNCS" "$FUNC1:traceoff:unlimited" "$FUNC2:traceoff:unlimited"; then
111 fail "Expected $ALL_FUNCS $FUNC1:traceoff:unlimited and $FUNC2:traceoff:unlimited"
112fi
113
114reset_ftrace_filter
115
116if check_set_ftrace_filter "$ALL_FUNCS"; then
117 fail "Expected $ALL_FUNCS"
118fi
119
120echo "$FUNC1" > set_ftrace_filter
121if check_set_ftrace_filter "$FUNC1" ; then
122 fail "Expected $FUNC1"
123fi
124
125echo "$FUNC2" >> set_ftrace_filter
126if check_set_ftrace_filter "$FUNC1" "$FUNC2" ; then
127 fail "Expected $FUNC1 and $FUNC2"
128fi
129
130do_reset
131
132exit 0
diff --git a/tools/testing/selftests/ftrace/test.d/ftrace/func_traceonoff_triggers.tc b/tools/testing/selftests/ftrace/test.d/ftrace/func_traceonoff_triggers.tc
new file mode 100644
index 000000000000..c8e02ec01eaf
--- /dev/null
+++ b/tools/testing/selftests/ftrace/test.d/ftrace/func_traceonoff_triggers.tc
@@ -0,0 +1,172 @@
1#!/bin/sh
2# description: ftrace - test for function traceon/off triggers
3# flags: instance
4#
5# Ftrace allows to add triggers to functions, such as enabling or disabling
6# tracing, enabling or disabling trace events, or recording a stack trace
7# within the ring buffer.
8#
9# This test is designed to test enabling and disabling tracing triggers
10#
11
12# The triggers are set within the set_ftrace_filter file
13if [ ! -f set_ftrace_filter ]; then
14 echo "set_ftrace_filter not found? Is dynamic ftrace not set?"
15 exit_unsupported
16fi
17
18do_reset() {
19 reset_ftrace_filter
20 reset_tracer
21 disable_events
22 clear_trace
23 enable_tracing
24}
25
26fail() { # mesg
27 do_reset
28 echo $1
29 exit $FAIL
30}
31
32SLEEP_TIME=".1"
33
34do_reset
35
36echo "Testing function probes with enabling disabling tracing:"
37
38cnt_trace() {
39 grep -v '^#' trace | wc -l
40}
41
42echo '** DISABLE TRACING'
43disable_tracing
44clear_trace
45
46cnt=`cnt_trace`
47if [ $cnt -ne 0 ]; then
48 fail "Found junk in trace"
49fi
50
51
52echo '** ENABLE EVENTS'
53
54echo 1 > events/enable
55
56echo '** ENABLE TRACING'
57enable_tracing
58
59cnt=`cnt_trace`
60if [ $cnt -eq 0 ]; then
61 fail "Nothing found in trace"
62fi
63
64# powerpc uses .schedule
65func="schedule"
66x=`grep '^\.schedule$' available_filter_functions | wc -l`
67if [ "$x" -eq 1 ]; then
68 func=".schedule"
69fi
70
71echo '** SET TRACEOFF'
72
73echo "$func:traceoff" > set_ftrace_filter
74
75cnt=`grep schedule set_ftrace_filter | wc -l`
76if [ $cnt -ne 1 ]; then
77 fail "Did not find traceoff trigger"
78fi
79
80cnt=`cnt_trace`
81sleep $SLEEP_TIME
82cnt2=`cnt_trace`
83
84if [ $cnt -ne $cnt2 ]; then
85 fail "Tracing is not stopped"
86fi
87
88on=`cat tracing_on`
89if [ $on != "0" ]; then
90 fail "Tracing is not off"
91fi
92
93line1=`cat trace | tail -1`
94sleep $SLEEP_TIME
95line2=`cat trace | tail -1`
96
97if [ "$line1" != "$line2" ]; then
98 fail "Tracing file is still changing"
99fi
100
101clear_trace
102
103cnt=`cnt_trace`
104if [ $cnt -ne 0 ]; then
105 fail "Tracing is still happeing"
106fi
107
108echo "!$func:traceoff" >> set_ftrace_filter
109
110cnt=`grep schedule set_ftrace_filter | wc -l`
111if [ $cnt -ne 0 ]; then
112 fail "traceoff trigger still exists"
113fi
114
115on=`cat tracing_on`
116if [ $on != "0" ]; then
117 fail "Tracing is started again"
118fi
119
120echo "$func:traceon" > set_ftrace_filter
121
122cnt=`grep schedule set_ftrace_filter | wc -l`
123if [ $cnt -ne 1 ]; then
124 fail "traceon trigger not found"
125fi
126
127cnt=`cnt_trace`
128if [ $cnt -eq 0 ]; then
129 fail "Tracing did not start"
130fi
131
132on=`cat tracing_on`
133if [ $on != "1" ]; then
134 fail "Tracing was not enabled"
135fi
136
137
138echo "!$func:traceon" >> set_ftrace_filter
139
140cnt=`grep schedule set_ftrace_filter | wc -l`
141if [ $cnt -ne 0 ]; then
142 fail "traceon trigger still exists"
143fi
144
145check_sleep() {
146 val=$1
147 sleep $SLEEP_TIME
148 cat set_ftrace_filter
149 on=`cat tracing_on`
150 if [ $on != "$val" ]; then
151 fail "Expected tracing_on to be $val, but it was $on"
152 fi
153}
154
155
156echo "$func:traceoff:3" > set_ftrace_filter
157check_sleep "0"
158echo 1 > tracing_on
159check_sleep "0"
160echo 1 > tracing_on
161check_sleep "0"
162echo 1 > tracing_on
163check_sleep "1"
164echo "!$func:traceoff:0" > set_ftrace_filter
165
166if grep -e traceon -e traceoff set_ftrace_filter; then
167 fail "Tracing on and off triggers still exist"
168fi
169
170disable_events
171
172exit 0
diff --git a/tools/testing/selftests/ftrace/test.d/functions b/tools/testing/selftests/ftrace/test.d/functions
index 91de1a8e4f19..9aec6fcb7729 100644
--- a/tools/testing/selftests/ftrace/test.d/functions
+++ b/tools/testing/selftests/ftrace/test.d/functions
@@ -30,6 +30,27 @@ reset_events_filter() { # reset all current setting filters
30 done 30 done
31} 31}
32 32
33reset_ftrace_filter() { # reset all triggers in set_ftrace_filter
34 echo > set_ftrace_filter
35 grep -v '^#' set_ftrace_filter | while read t; do
36 tr=`echo $t | cut -d: -f2`
37 if [ "$tr" == "" ]; then
38 continue
39 fi
40 if [ $tr == "enable_event" -o $tr == "disable_event" ]; then
41 tr=`echo $t | cut -d: -f1-4`
42 limit=`echo $t | cut -d: -f5`
43 else
44 tr=`echo $t | cut -d: -f1-2`
45 limit=`echo $t | cut -d: -f3`
46 fi
47 if [ "$limit" != "unlimited" ]; then
48 tr="$tr:$limit"
49 fi
50 echo "!$tr" > set_ftrace_filter
51 done
52}
53
33disable_events() { 54disable_events() {
34 echo 0 > events/enable 55 echo 0 > events/enable
35} 56}
diff --git a/tools/testing/selftests/ftrace/test.d/kprobe/kretprobe_maxactive.tc b/tools/testing/selftests/ftrace/test.d/kprobe/kretprobe_maxactive.tc
new file mode 100644
index 000000000000..57abdf1caabf
--- /dev/null
+++ b/tools/testing/selftests/ftrace/test.d/kprobe/kretprobe_maxactive.tc
@@ -0,0 +1,39 @@
1#!/bin/sh
2# description: Kretprobe dynamic event with maxactive
3
4[ -f kprobe_events ] || exit_unsupported # this is configurable
5
6echo > kprobe_events
7
8# Test if we successfully reject unknown messages
9if echo 'a:myprobeaccept inet_csk_accept' > kprobe_events; then false; else true; fi
10
11# Test if we successfully reject too big maxactive
12if echo 'r1000000:myprobeaccept inet_csk_accept' > kprobe_events; then false; else true; fi
13
14# Test if we successfully reject unparsable numbers for maxactive
15if echo 'r10fuzz:myprobeaccept inet_csk_accept' > kprobe_events; then false; else true; fi
16
17# Test for kretprobe with event name without maxactive
18echo 'r:myprobeaccept inet_csk_accept' > kprobe_events
19grep myprobeaccept kprobe_events
20test -d events/kprobes/myprobeaccept
21echo '-:myprobeaccept' >> kprobe_events
22
23# Test for kretprobe with event name with a small maxactive
24echo 'r10:myprobeaccept inet_csk_accept' > kprobe_events
25grep myprobeaccept kprobe_events
26test -d events/kprobes/myprobeaccept
27echo '-:myprobeaccept' >> kprobe_events
28
29# Test for kretprobe without event name without maxactive
30echo 'r inet_csk_accept' > kprobe_events
31grep inet_csk_accept kprobe_events
32echo > kprobe_events
33
34# Test for kretprobe without event name with a small maxactive
35echo 'r10 inet_csk_accept' > kprobe_events
36grep inet_csk_accept kprobe_events
37echo > kprobe_events
38
39clear_trace
diff --git a/tools/testing/selftests/ftrace/test.d/trigger/trigger-eventonoff.tc b/tools/testing/selftests/ftrace/test.d/trigger/trigger-eventonoff.tc
index 1a9445021bf1..c5435adfdd93 100644
--- a/tools/testing/selftests/ftrace/test.d/trigger/trigger-eventonoff.tc
+++ b/tools/testing/selftests/ftrace/test.d/trigger/trigger-eventonoff.tc
@@ -1,5 +1,6 @@
1#!/bin/sh 1#!/bin/sh
2# description: event trigger - test event enable/disable trigger 2# description: event trigger - test event enable/disable trigger
3# flags: instance
3 4
4do_reset() { 5do_reset() {
5 reset_trigger 6 reset_trigger
diff --git a/tools/testing/selftests/ftrace/test.d/trigger/trigger-filter.tc b/tools/testing/selftests/ftrace/test.d/trigger/trigger-filter.tc
index 514e466e198b..48849a8d577f 100644
--- a/tools/testing/selftests/ftrace/test.d/trigger/trigger-filter.tc
+++ b/tools/testing/selftests/ftrace/test.d/trigger/trigger-filter.tc
@@ -1,5 +1,6 @@
1#!/bin/sh 1#!/bin/sh
2# description: event trigger - test trigger filter 2# description: event trigger - test trigger filter
3# flags: instance
3 4
4do_reset() { 5do_reset() {
5 reset_trigger 6 reset_trigger
diff --git a/tools/testing/selftests/ftrace/test.d/trigger/trigger-hist-mod.tc b/tools/testing/selftests/ftrace/test.d/trigger/trigger-hist-mod.tc
index 400e98b64948..b7f86d10b549 100644
--- a/tools/testing/selftests/ftrace/test.d/trigger/trigger-hist-mod.tc
+++ b/tools/testing/selftests/ftrace/test.d/trigger/trigger-hist-mod.tc
@@ -1,5 +1,6 @@
1#!/bin/sh 1#!/bin/sh
2# description: event trigger - test histogram modifiers 2# description: event trigger - test histogram modifiers
3# flags: instance
3 4
4do_reset() { 5do_reset() {
5 reset_trigger 6 reset_trigger
diff --git a/tools/testing/selftests/ftrace/test.d/trigger/trigger-hist.tc b/tools/testing/selftests/ftrace/test.d/trigger/trigger-hist.tc
index a00184cd9c95..fb66f7d9339d 100644
--- a/tools/testing/selftests/ftrace/test.d/trigger/trigger-hist.tc
+++ b/tools/testing/selftests/ftrace/test.d/trigger/trigger-hist.tc
@@ -1,5 +1,6 @@
1#!/bin/sh 1#!/bin/sh
2# description: event trigger - test histogram trigger 2# description: event trigger - test histogram trigger
3# flags: instance
3 4
4do_reset() { 5do_reset() {
5 reset_trigger 6 reset_trigger
diff --git a/tools/testing/selftests/ftrace/test.d/trigger/trigger-multihist.tc b/tools/testing/selftests/ftrace/test.d/trigger/trigger-multihist.tc
index 3478b00ead57..f9153087dd7c 100644
--- a/tools/testing/selftests/ftrace/test.d/trigger/trigger-multihist.tc
+++ b/tools/testing/selftests/ftrace/test.d/trigger/trigger-multihist.tc
@@ -1,5 +1,6 @@
1#!/bin/sh 1#!/bin/sh
2# description: event trigger - test multiple histogram triggers 2# description: event trigger - test multiple histogram triggers
3# flags: instance
3 4
4do_reset() { 5do_reset() {
5 reset_trigger 6 reset_trigger
generated by cgit v1.2.2 (git 2.25.0) at 2025-06-01 12:19:26 -0400