1 // SPDX-License-Identifier: GPL-2.0
3 * ring buffer based function tracer
5 * Copyright (C) 2007-2012 Steven Rostedt <srostedt@redhat.com>
6 * Copyright (C) 2008 Ingo Molnar <mingo@redhat.com>
8 * Originally taken from the RT patch by:
9 * Arnaldo Carvalho de Melo <acme@redhat.com>
11 * Based on code from the latency_tracer, that is:
12 * Copyright (C) 2004-2006 Ingo Molnar
13 * Copyright (C) 2004 Nadia Yvette Chambers
15 #include <linux/ring_buffer.h>
16 #include <generated/utsrelease.h>
17 #include <linux/stacktrace.h>
18 #include <linux/writeback.h>
19 #include <linux/kallsyms.h>
20 #include <linux/security.h>
21 #include <linux/seq_file.h>
22 #include <linux/notifier.h>
23 #include <linux/irqflags.h>
24 #include <linux/debugfs.h>
25 #include <linux/tracefs.h>
26 #include <linux/pagemap.h>
27 #include <linux/hardirq.h>
28 #include <linux/linkage.h>
29 #include <linux/uaccess.h>
30 #include <linux/vmalloc.h>
31 #include <linux/ftrace.h>
32 #include <linux/module.h>
33 #include <linux/percpu.h>
34 #include <linux/splice.h>
35 #include <linux/kdebug.h>
36 #include <linux/string.h>
37 #include <linux/mount.h>
38 #include <linux/rwsem.h>
39 #include <linux/slab.h>
40 #include <linux/ctype.h>
41 #include <linux/init.h>
42 #include <linux/poll.h>
43 #include <linux/nmi.h>
45 #include <linux/trace.h>
46 #include <linux/sched/clock.h>
47 #include <linux/sched/rt.h>
48 #include <linux/fsnotify.h>
49 #include <linux/irq_work.h>
50 #include <linux/workqueue.h>
53 #include "trace_output.h"
56 * On boot up, the ring buffer is set to the minimum size, so that
57 * we do not waste memory on systems that are not using tracing.
59 bool ring_buffer_expanded
;
62 * We need to change this state when a selftest is running.
63 * A selftest will lurk into the ring-buffer to count the
64 * entries inserted during the selftest although some concurrent
65 * insertions into the ring-buffer such as trace_printk could occurred
66 * at the same time, giving false positive or negative results.
68 static bool __read_mostly tracing_selftest_running
;
71 * If a tracer is running, we do not want to run SELFTEST.
73 bool __read_mostly tracing_selftest_disabled
;
75 /* Pipe tracepoints to printk */
76 struct trace_iterator
*tracepoint_print_iter
;
77 int tracepoint_printk
;
78 static DEFINE_STATIC_KEY_FALSE(tracepoint_printk_key
);
80 /* For tracers that don't implement custom flags */
81 static struct tracer_opt dummy_tracer_opt
[] = {
86 dummy_set_flag(struct trace_array
*tr
, u32 old_flags
, u32 bit
, int set
)
92 * To prevent the comm cache from being overwritten when no
93 * tracing is active, only save the comm when a trace event
96 static DEFINE_PER_CPU(bool, trace_taskinfo_save
);
99 * Kill all tracing for good (never come back).
100 * It is initialized to 1 but will turn to zero if the initialization
101 * of the tracer is successful. But that is the only place that sets
104 static int tracing_disabled
= 1;
106 cpumask_var_t __read_mostly tracing_buffer_mask
;
109 * ftrace_dump_on_oops - variable to dump ftrace buffer on oops
111 * If there is an oops (or kernel panic) and the ftrace_dump_on_oops
112 * is set, then ftrace_dump is called. This will output the contents
113 * of the ftrace buffers to the console. This is very useful for
114 * capturing traces that lead to crashes and outputing it to a
117 * It is default off, but you can enable it with either specifying
118 * "ftrace_dump_on_oops" in the kernel command line, or setting
119 * /proc/sys/kernel/ftrace_dump_on_oops
120 * Set 1 if you want to dump buffers of all CPUs
121 * Set 2 if you want to dump the buffer of the CPU that triggered oops
124 enum ftrace_dump_mode ftrace_dump_on_oops
;
126 /* When set, tracing will stop when a WARN*() is hit */
127 int __disable_trace_on_warning
;
129 #ifdef CONFIG_TRACE_EVAL_MAP_FILE
130 /* Map of enums to their values, for "eval_map" file */
131 struct trace_eval_map_head
{
133 unsigned long length
;
136 union trace_eval_map_item
;
138 struct trace_eval_map_tail
{
140 * "end" is first and points to NULL as it must be different
141 * than "mod" or "eval_string"
143 union trace_eval_map_item
*next
;
144 const char *end
; /* points to NULL */
147 static DEFINE_MUTEX(trace_eval_mutex
);
150 * The trace_eval_maps are saved in an array with two extra elements,
151 * one at the beginning, and one at the end. The beginning item contains
152 * the count of the saved maps (head.length), and the module they
153 * belong to if not built in (head.mod). The ending item contains a
154 * pointer to the next array of saved eval_map items.
156 union trace_eval_map_item
{
157 struct trace_eval_map map
;
158 struct trace_eval_map_head head
;
159 struct trace_eval_map_tail tail
;
162 static union trace_eval_map_item
*trace_eval_maps
;
163 #endif /* CONFIG_TRACE_EVAL_MAP_FILE */
165 int tracing_set_tracer(struct trace_array
*tr
, const char *buf
);
166 static void ftrace_trace_userstack(struct trace_buffer
*buffer
,
167 unsigned long flags
, int pc
);
169 #define MAX_TRACER_SIZE 100
170 static char bootup_tracer_buf
[MAX_TRACER_SIZE
] __initdata
;
171 static char *default_bootup_tracer
;
173 static bool allocate_snapshot
;
175 static int __init
set_cmdline_ftrace(char *str
)
177 strlcpy(bootup_tracer_buf
, str
, MAX_TRACER_SIZE
);
178 default_bootup_tracer
= bootup_tracer_buf
;
179 /* We are using ftrace early, expand it */
180 ring_buffer_expanded
= true;
183 __setup("ftrace=", set_cmdline_ftrace
);
185 static int __init
set_ftrace_dump_on_oops(char *str
)
187 if (*str
++ != '=' || !*str
) {
188 ftrace_dump_on_oops
= DUMP_ALL
;
192 if (!strcmp("orig_cpu", str
)) {
193 ftrace_dump_on_oops
= DUMP_ORIG
;
199 __setup("ftrace_dump_on_oops", set_ftrace_dump_on_oops
);
201 static int __init
stop_trace_on_warning(char *str
)
203 if ((strcmp(str
, "=0") != 0 && strcmp(str
, "=off") != 0))
204 __disable_trace_on_warning
= 1;
207 __setup("traceoff_on_warning", stop_trace_on_warning
);
209 static int __init
boot_alloc_snapshot(char *str
)
211 allocate_snapshot
= true;
212 /* We also need the main ring buffer expanded */
213 ring_buffer_expanded
= true;
216 __setup("alloc_snapshot", boot_alloc_snapshot
);
219 static char trace_boot_options_buf
[MAX_TRACER_SIZE
] __initdata
;
221 static int __init
set_trace_boot_options(char *str
)
223 strlcpy(trace_boot_options_buf
, str
, MAX_TRACER_SIZE
);
226 __setup("trace_options=", set_trace_boot_options
);
228 static char trace_boot_clock_buf
[MAX_TRACER_SIZE
] __initdata
;
229 static char *trace_boot_clock __initdata
;
231 static int __init
set_trace_boot_clock(char *str
)
233 strlcpy(trace_boot_clock_buf
, str
, MAX_TRACER_SIZE
);
234 trace_boot_clock
= trace_boot_clock_buf
;
237 __setup("trace_clock=", set_trace_boot_clock
);
239 static int __init
set_tracepoint_printk(char *str
)
241 if ((strcmp(str
, "=0") != 0 && strcmp(str
, "=off") != 0))
242 tracepoint_printk
= 1;
245 __setup("tp_printk", set_tracepoint_printk
);
247 unsigned long long ns2usecs(u64 nsec
)
255 trace_process_export(struct trace_export
*export
,
256 struct ring_buffer_event
*event
, int flag
)
258 struct trace_entry
*entry
;
259 unsigned int size
= 0;
261 if (export
->flags
& flag
) {
262 entry
= ring_buffer_event_data(event
);
263 size
= ring_buffer_event_length(event
);
264 export
->write(export
, entry
, size
);
268 static DEFINE_MUTEX(ftrace_export_lock
);
270 static struct trace_export __rcu
*ftrace_exports_list __read_mostly
;
272 static DEFINE_STATIC_KEY_FALSE(trace_function_exports_enabled
);
273 static DEFINE_STATIC_KEY_FALSE(trace_event_exports_enabled
);
275 static inline void ftrace_exports_enable(struct trace_export
*export
)
277 if (export
->flags
& TRACE_EXPORT_FUNCTION
)
278 static_branch_inc(&trace_function_exports_enabled
);
280 if (export
->flags
& TRACE_EXPORT_EVENT
)
281 static_branch_inc(&trace_event_exports_enabled
);
284 static inline void ftrace_exports_disable(struct trace_export
*export
)
286 if (export
->flags
& TRACE_EXPORT_FUNCTION
)
287 static_branch_dec(&trace_function_exports_enabled
);
289 if (export
->flags
& TRACE_EXPORT_EVENT
)
290 static_branch_dec(&trace_event_exports_enabled
);
293 static void ftrace_exports(struct ring_buffer_event
*event
, int flag
)
295 struct trace_export
*export
;
297 preempt_disable_notrace();
299 export
= rcu_dereference_raw_check(ftrace_exports_list
);
301 trace_process_export(export
, event
, flag
);
302 export
= rcu_dereference_raw_check(export
->next
);
305 preempt_enable_notrace();
309 add_trace_export(struct trace_export
**list
, struct trace_export
*export
)
311 rcu_assign_pointer(export
->next
, *list
);
313 * We are entering export into the list but another
314 * CPU might be walking that list. We need to make sure
315 * the export->next pointer is valid before another CPU sees
316 * the export pointer included into the list.
318 rcu_assign_pointer(*list
, export
);
322 rm_trace_export(struct trace_export
**list
, struct trace_export
*export
)
324 struct trace_export
**p
;
326 for (p
= list
; *p
!= NULL
; p
= &(*p
)->next
)
333 rcu_assign_pointer(*p
, (*p
)->next
);
339 add_ftrace_export(struct trace_export
**list
, struct trace_export
*export
)
341 ftrace_exports_enable(export
);
343 add_trace_export(list
, export
);
347 rm_ftrace_export(struct trace_export
**list
, struct trace_export
*export
)
351 ret
= rm_trace_export(list
, export
);
352 ftrace_exports_disable(export
);
357 int register_ftrace_export(struct trace_export
*export
)
359 if (WARN_ON_ONCE(!export
->write
))
362 mutex_lock(&ftrace_export_lock
);
364 add_ftrace_export(&ftrace_exports_list
, export
);
366 mutex_unlock(&ftrace_export_lock
);
370 EXPORT_SYMBOL_GPL(register_ftrace_export
);
372 int unregister_ftrace_export(struct trace_export
*export
)
376 mutex_lock(&ftrace_export_lock
);
378 ret
= rm_ftrace_export(&ftrace_exports_list
, export
);
380 mutex_unlock(&ftrace_export_lock
);
384 EXPORT_SYMBOL_GPL(unregister_ftrace_export
);
386 /* trace_flags holds trace_options default values */
387 #define TRACE_DEFAULT_FLAGS \
388 (FUNCTION_DEFAULT_FLAGS | \
389 TRACE_ITER_PRINT_PARENT | TRACE_ITER_PRINTK | \
390 TRACE_ITER_ANNOTATE | TRACE_ITER_CONTEXT_INFO | \
391 TRACE_ITER_RECORD_CMD | TRACE_ITER_OVERWRITE | \
392 TRACE_ITER_IRQ_INFO | TRACE_ITER_MARKERS)
394 /* trace_options that are only supported by global_trace */
395 #define TOP_LEVEL_TRACE_FLAGS (TRACE_ITER_PRINTK | \
396 TRACE_ITER_PRINTK_MSGONLY | TRACE_ITER_RECORD_CMD)
398 /* trace_flags that are default zero for instances */
399 #define ZEROED_TRACE_FLAGS \
400 (TRACE_ITER_EVENT_FORK | TRACE_ITER_FUNC_FORK)
403 * The global_trace is the descriptor that holds the top-level tracing
404 * buffers for the live tracing.
406 static struct trace_array global_trace
= {
407 .trace_flags
= TRACE_DEFAULT_FLAGS
,
410 LIST_HEAD(ftrace_trace_arrays
);
412 int trace_array_get(struct trace_array
*this_tr
)
414 struct trace_array
*tr
;
417 mutex_lock(&trace_types_lock
);
418 list_for_each_entry(tr
, &ftrace_trace_arrays
, list
) {
425 mutex_unlock(&trace_types_lock
);
430 static void __trace_array_put(struct trace_array
*this_tr
)
432 WARN_ON(!this_tr
->ref
);
437 * trace_array_put - Decrement the reference counter for this trace array.
439 * NOTE: Use this when we no longer need the trace array returned by
440 * trace_array_get_by_name(). This ensures the trace array can be later
444 void trace_array_put(struct trace_array
*this_tr
)
449 mutex_lock(&trace_types_lock
);
450 __trace_array_put(this_tr
);
451 mutex_unlock(&trace_types_lock
);
453 EXPORT_SYMBOL_GPL(trace_array_put
);
455 int tracing_check_open_get_tr(struct trace_array
*tr
)
459 ret
= security_locked_down(LOCKDOWN_TRACEFS
);
463 if (tracing_disabled
)
466 if (tr
&& trace_array_get(tr
) < 0)
472 int call_filter_check_discard(struct trace_event_call
*call
, void *rec
,
473 struct trace_buffer
*buffer
,
474 struct ring_buffer_event
*event
)
476 if (unlikely(call
->flags
& TRACE_EVENT_FL_FILTERED
) &&
477 !filter_match_preds(call
->filter
, rec
)) {
478 __trace_event_discard_commit(buffer
, event
);
485 void trace_free_pid_list(struct trace_pid_list
*pid_list
)
487 vfree(pid_list
->pids
);
492 * trace_find_filtered_pid - check if a pid exists in a filtered_pid list
493 * @filtered_pids: The list of pids to check
494 * @search_pid: The PID to find in @filtered_pids
496 * Returns true if @search_pid is fonud in @filtered_pids, and false otherwis.
499 trace_find_filtered_pid(struct trace_pid_list
*filtered_pids
, pid_t search_pid
)
502 * If pid_max changed after filtered_pids was created, we
503 * by default ignore all pids greater than the previous pid_max.
505 if (search_pid
>= filtered_pids
->pid_max
)
508 return test_bit(search_pid
, filtered_pids
->pids
);
512 * trace_ignore_this_task - should a task be ignored for tracing
513 * @filtered_pids: The list of pids to check
514 * @task: The task that should be ignored if not filtered
516 * Checks if @task should be traced or not from @filtered_pids.
517 * Returns true if @task should *NOT* be traced.
518 * Returns false if @task should be traced.
521 trace_ignore_this_task(struct trace_pid_list
*filtered_pids
,
522 struct trace_pid_list
*filtered_no_pids
,
523 struct task_struct
*task
)
526 * If filterd_no_pids is not empty, and the task's pid is listed
527 * in filtered_no_pids, then return true.
528 * Otherwise, if filtered_pids is empty, that means we can
529 * trace all tasks. If it has content, then only trace pids
530 * within filtered_pids.
533 return (filtered_pids
&&
534 !trace_find_filtered_pid(filtered_pids
, task
->pid
)) ||
536 trace_find_filtered_pid(filtered_no_pids
, task
->pid
));
540 * trace_filter_add_remove_task - Add or remove a task from a pid_list
541 * @pid_list: The list to modify
542 * @self: The current task for fork or NULL for exit
543 * @task: The task to add or remove
545 * If adding a task, if @self is defined, the task is only added if @self
546 * is also included in @pid_list. This happens on fork and tasks should
547 * only be added when the parent is listed. If @self is NULL, then the
548 * @task pid will be removed from the list, which would happen on exit
551 void trace_filter_add_remove_task(struct trace_pid_list
*pid_list
,
552 struct task_struct
*self
,
553 struct task_struct
*task
)
558 /* For forks, we only add if the forking task is listed */
560 if (!trace_find_filtered_pid(pid_list
, self
->pid
))
564 /* Sorry, but we don't support pid_max changing after setting */
565 if (task
->pid
>= pid_list
->pid_max
)
568 /* "self" is set for forks, and NULL for exits */
570 set_bit(task
->pid
, pid_list
->pids
);
572 clear_bit(task
->pid
, pid_list
->pids
);
576 * trace_pid_next - Used for seq_file to get to the next pid of a pid_list
577 * @pid_list: The pid list to show
578 * @v: The last pid that was shown (+1 the actual pid to let zero be displayed)
579 * @pos: The position of the file
581 * This is used by the seq_file "next" operation to iterate the pids
582 * listed in a trace_pid_list structure.
584 * Returns the pid+1 as we want to display pid of zero, but NULL would
585 * stop the iteration.
587 void *trace_pid_next(struct trace_pid_list
*pid_list
, void *v
, loff_t
*pos
)
589 unsigned long pid
= (unsigned long)v
;
593 /* pid already is +1 of the actual prevous bit */
594 pid
= find_next_bit(pid_list
->pids
, pid_list
->pid_max
, pid
);
596 /* Return pid + 1 to allow zero to be represented */
597 if (pid
< pid_list
->pid_max
)
598 return (void *)(pid
+ 1);
604 * trace_pid_start - Used for seq_file to start reading pid lists
605 * @pid_list: The pid list to show
606 * @pos: The position of the file
608 * This is used by seq_file "start" operation to start the iteration
611 * Returns the pid+1 as we want to display pid of zero, but NULL would
612 * stop the iteration.
614 void *trace_pid_start(struct trace_pid_list
*pid_list
, loff_t
*pos
)
619 pid
= find_first_bit(pid_list
->pids
, pid_list
->pid_max
);
620 if (pid
>= pid_list
->pid_max
)
623 /* Return pid + 1 so that zero can be the exit value */
624 for (pid
++; pid
&& l
< *pos
;
625 pid
= (unsigned long)trace_pid_next(pid_list
, (void *)pid
, &l
))
631 * trace_pid_show - show the current pid in seq_file processing
632 * @m: The seq_file structure to write into
633 * @v: A void pointer of the pid (+1) value to display
635 * Can be directly used by seq_file operations to display the current
638 int trace_pid_show(struct seq_file
*m
, void *v
)
640 unsigned long pid
= (unsigned long)v
- 1;
642 seq_printf(m
, "%lu\n", pid
);
646 /* 128 should be much more than enough */
647 #define PID_BUF_SIZE 127
649 int trace_pid_write(struct trace_pid_list
*filtered_pids
,
650 struct trace_pid_list
**new_pid_list
,
651 const char __user
*ubuf
, size_t cnt
)
653 struct trace_pid_list
*pid_list
;
654 struct trace_parser parser
;
662 if (trace_parser_get_init(&parser
, PID_BUF_SIZE
+ 1))
666 * Always recreate a new array. The write is an all or nothing
667 * operation. Always create a new array when adding new pids by
668 * the user. If the operation fails, then the current list is
671 pid_list
= kmalloc(sizeof(*pid_list
), GFP_KERNEL
);
673 trace_parser_put(&parser
);
677 pid_list
->pid_max
= READ_ONCE(pid_max
);
679 /* Only truncating will shrink pid_max */
680 if (filtered_pids
&& filtered_pids
->pid_max
> pid_list
->pid_max
)
681 pid_list
->pid_max
= filtered_pids
->pid_max
;
683 pid_list
->pids
= vzalloc((pid_list
->pid_max
+ 7) >> 3);
684 if (!pid_list
->pids
) {
685 trace_parser_put(&parser
);
691 /* copy the current bits to the new max */
692 for_each_set_bit(pid
, filtered_pids
->pids
,
693 filtered_pids
->pid_max
) {
694 set_bit(pid
, pid_list
->pids
);
703 ret
= trace_get_user(&parser
, ubuf
, cnt
, &pos
);
704 if (ret
< 0 || !trace_parser_loaded(&parser
))
712 if (kstrtoul(parser
.buffer
, 0, &val
))
714 if (val
>= pid_list
->pid_max
)
719 set_bit(pid
, pid_list
->pids
);
722 trace_parser_clear(&parser
);
725 trace_parser_put(&parser
);
728 trace_free_pid_list(pid_list
);
733 /* Cleared the list of pids */
734 trace_free_pid_list(pid_list
);
739 *new_pid_list
= pid_list
;
744 static u64
buffer_ftrace_now(struct array_buffer
*buf
, int cpu
)
748 /* Early boot up does not have a buffer yet */
750 return trace_clock_local();
752 ts
= ring_buffer_time_stamp(buf
->buffer
, cpu
);
753 ring_buffer_normalize_time_stamp(buf
->buffer
, cpu
, &ts
);
758 u64
ftrace_now(int cpu
)
760 return buffer_ftrace_now(&global_trace
.array_buffer
, cpu
);
764 * tracing_is_enabled - Show if global_trace has been disabled
766 * Shows if the global trace has been enabled or not. It uses the
767 * mirror flag "buffer_disabled" to be used in fast paths such as for
768 * the irqsoff tracer. But it may be inaccurate due to races. If you
769 * need to know the accurate state, use tracing_is_on() which is a little
770 * slower, but accurate.
772 int tracing_is_enabled(void)
775 * For quick access (irqsoff uses this in fast path), just
776 * return the mirror variable of the state of the ring buffer.
777 * It's a little racy, but we don't really care.
780 return !global_trace
.buffer_disabled
;
784 * trace_buf_size is the size in bytes that is allocated
785 * for a buffer. Note, the number of bytes is always rounded
788 * This number is purposely set to a low number of 16384.
789 * If the dump on oops happens, it will be much appreciated
790 * to not have to wait for all that output. Anyway this can be
791 * boot time and run time configurable.
793 #define TRACE_BUF_SIZE_DEFAULT 1441792UL /* 16384 * 88 (sizeof(entry)) */
795 static unsigned long trace_buf_size
= TRACE_BUF_SIZE_DEFAULT
;
797 /* trace_types holds a link list of available tracers. */
798 static struct tracer
*trace_types __read_mostly
;
801 * trace_types_lock is used to protect the trace_types list.
803 DEFINE_MUTEX(trace_types_lock
);
806 * serialize the access of the ring buffer
808 * ring buffer serializes readers, but it is low level protection.
809 * The validity of the events (which returns by ring_buffer_peek() ..etc)
810 * are not protected by ring buffer.
812 * The content of events may become garbage if we allow other process consumes
813 * these events concurrently:
814 * A) the page of the consumed events may become a normal page
815 * (not reader page) in ring buffer, and this page will be rewrited
816 * by events producer.
817 * B) The page of the consumed events may become a page for splice_read,
818 * and this page will be returned to system.
820 * These primitives allow multi process access to different cpu ring buffer
823 * These primitives don't distinguish read-only and read-consume access.
824 * Multi read-only access are also serialized.
828 static DECLARE_RWSEM(all_cpu_access_lock
);
829 static DEFINE_PER_CPU(struct mutex
, cpu_access_lock
);
831 static inline void trace_access_lock(int cpu
)
833 if (cpu
== RING_BUFFER_ALL_CPUS
) {
834 /* gain it for accessing the whole ring buffer. */
835 down_write(&all_cpu_access_lock
);
837 /* gain it for accessing a cpu ring buffer. */
839 /* Firstly block other trace_access_lock(RING_BUFFER_ALL_CPUS). */
840 down_read(&all_cpu_access_lock
);
842 /* Secondly block other access to this @cpu ring buffer. */
843 mutex_lock(&per_cpu(cpu_access_lock
, cpu
));
847 static inline void trace_access_unlock(int cpu
)
849 if (cpu
== RING_BUFFER_ALL_CPUS
) {
850 up_write(&all_cpu_access_lock
);
852 mutex_unlock(&per_cpu(cpu_access_lock
, cpu
));
853 up_read(&all_cpu_access_lock
);
857 static inline void trace_access_lock_init(void)
861 for_each_possible_cpu(cpu
)
862 mutex_init(&per_cpu(cpu_access_lock
, cpu
));
867 static DEFINE_MUTEX(access_lock
);
869 static inline void trace_access_lock(int cpu
)
872 mutex_lock(&access_lock
);
875 static inline void trace_access_unlock(int cpu
)
878 mutex_unlock(&access_lock
);
881 static inline void trace_access_lock_init(void)
887 #ifdef CONFIG_STACKTRACE
888 static void __ftrace_trace_stack(struct trace_buffer
*buffer
,
890 int skip
, int pc
, struct pt_regs
*regs
);
891 static inline void ftrace_trace_stack(struct trace_array
*tr
,
892 struct trace_buffer
*buffer
,
894 int skip
, int pc
, struct pt_regs
*regs
);
897 static inline void __ftrace_trace_stack(struct trace_buffer
*buffer
,
899 int skip
, int pc
, struct pt_regs
*regs
)
902 static inline void ftrace_trace_stack(struct trace_array
*tr
,
903 struct trace_buffer
*buffer
,
905 int skip
, int pc
, struct pt_regs
*regs
)
911 static __always_inline
void
912 trace_event_setup(struct ring_buffer_event
*event
,
913 int type
, unsigned long flags
, int pc
)
915 struct trace_entry
*ent
= ring_buffer_event_data(event
);
917 tracing_generic_entry_update(ent
, type
, flags
, pc
);
920 static __always_inline
struct ring_buffer_event
*
921 __trace_buffer_lock_reserve(struct trace_buffer
*buffer
,
924 unsigned long flags
, int pc
)
926 struct ring_buffer_event
*event
;
928 event
= ring_buffer_lock_reserve(buffer
, len
);
930 trace_event_setup(event
, type
, flags
, pc
);
935 void tracer_tracing_on(struct trace_array
*tr
)
937 if (tr
->array_buffer
.buffer
)
938 ring_buffer_record_on(tr
->array_buffer
.buffer
);
940 * This flag is looked at when buffers haven't been allocated
941 * yet, or by some tracers (like irqsoff), that just want to
942 * know if the ring buffer has been disabled, but it can handle
943 * races of where it gets disabled but we still do a record.
944 * As the check is in the fast path of the tracers, it is more
945 * important to be fast than accurate.
947 tr
->buffer_disabled
= 0;
948 /* Make the flag seen by readers */
953 * tracing_on - enable tracing buffers
955 * This function enables tracing buffers that may have been
956 * disabled with tracing_off.
958 void tracing_on(void)
960 tracer_tracing_on(&global_trace
);
962 EXPORT_SYMBOL_GPL(tracing_on
);
965 static __always_inline
void
966 __buffer_unlock_commit(struct trace_buffer
*buffer
, struct ring_buffer_event
*event
)
968 __this_cpu_write(trace_taskinfo_save
, true);
970 /* If this is the temp buffer, we need to commit fully */
971 if (this_cpu_read(trace_buffered_event
) == event
) {
972 /* Length is in event->array[0] */
973 ring_buffer_write(buffer
, event
->array
[0], &event
->array
[1]);
974 /* Release the temp buffer */
975 this_cpu_dec(trace_buffered_event_cnt
);
977 ring_buffer_unlock_commit(buffer
, event
);
981 * __trace_puts - write a constant string into the trace buffer.
982 * @ip: The address of the caller
983 * @str: The constant string to write
984 * @size: The size of the string.
986 int __trace_puts(unsigned long ip
, const char *str
, int size
)
988 struct ring_buffer_event
*event
;
989 struct trace_buffer
*buffer
;
990 struct print_entry
*entry
;
991 unsigned long irq_flags
;
995 if (!(global_trace
.trace_flags
& TRACE_ITER_PRINTK
))
998 pc
= preempt_count();
1000 if (unlikely(tracing_selftest_running
|| tracing_disabled
))
1003 alloc
= sizeof(*entry
) + size
+ 2; /* possible \n added */
1005 local_save_flags(irq_flags
);
1006 buffer
= global_trace
.array_buffer
.buffer
;
1007 ring_buffer_nest_start(buffer
);
1008 event
= __trace_buffer_lock_reserve(buffer
, TRACE_PRINT
, alloc
,
1015 entry
= ring_buffer_event_data(event
);
1018 memcpy(&entry
->buf
, str
, size
);
1020 /* Add a newline if necessary */
1021 if (entry
->buf
[size
- 1] != '\n') {
1022 entry
->buf
[size
] = '\n';
1023 entry
->buf
[size
+ 1] = '\0';
1025 entry
->buf
[size
] = '\0';
1027 __buffer_unlock_commit(buffer
, event
);
1028 ftrace_trace_stack(&global_trace
, buffer
, irq_flags
, 4, pc
, NULL
);
1030 ring_buffer_nest_end(buffer
);
1033 EXPORT_SYMBOL_GPL(__trace_puts
);
1036 * __trace_bputs - write the pointer to a constant string into trace buffer
1037 * @ip: The address of the caller
1038 * @str: The constant string to write to the buffer to
1040 int __trace_bputs(unsigned long ip
, const char *str
)
1042 struct ring_buffer_event
*event
;
1043 struct trace_buffer
*buffer
;
1044 struct bputs_entry
*entry
;
1045 unsigned long irq_flags
;
1046 int size
= sizeof(struct bputs_entry
);
1050 if (!(global_trace
.trace_flags
& TRACE_ITER_PRINTK
))
1053 pc
= preempt_count();
1055 if (unlikely(tracing_selftest_running
|| tracing_disabled
))
1058 local_save_flags(irq_flags
);
1059 buffer
= global_trace
.array_buffer
.buffer
;
1061 ring_buffer_nest_start(buffer
);
1062 event
= __trace_buffer_lock_reserve(buffer
, TRACE_BPUTS
, size
,
1067 entry
= ring_buffer_event_data(event
);
1071 __buffer_unlock_commit(buffer
, event
);
1072 ftrace_trace_stack(&global_trace
, buffer
, irq_flags
, 4, pc
, NULL
);
1076 ring_buffer_nest_end(buffer
);
1079 EXPORT_SYMBOL_GPL(__trace_bputs
);
1081 #ifdef CONFIG_TRACER_SNAPSHOT
1082 static void tracing_snapshot_instance_cond(struct trace_array
*tr
,
1085 struct tracer
*tracer
= tr
->current_trace
;
1086 unsigned long flags
;
1089 internal_trace_puts("*** SNAPSHOT CALLED FROM NMI CONTEXT ***\n");
1090 internal_trace_puts("*** snapshot is being ignored ***\n");
1094 if (!tr
->allocated_snapshot
) {
1095 internal_trace_puts("*** SNAPSHOT NOT ALLOCATED ***\n");
1096 internal_trace_puts("*** stopping trace here! ***\n");
1101 /* Note, snapshot can not be used when the tracer uses it */
1102 if (tracer
->use_max_tr
) {
1103 internal_trace_puts("*** LATENCY TRACER ACTIVE ***\n");
1104 internal_trace_puts("*** Can not use snapshot (sorry) ***\n");
1108 local_irq_save(flags
);
1109 update_max_tr(tr
, current
, smp_processor_id(), cond_data
);
1110 local_irq_restore(flags
);
1113 void tracing_snapshot_instance(struct trace_array
*tr
)
1115 tracing_snapshot_instance_cond(tr
, NULL
);
1119 * tracing_snapshot - take a snapshot of the current buffer.
1121 * This causes a swap between the snapshot buffer and the current live
1122 * tracing buffer. You can use this to take snapshots of the live
1123 * trace when some condition is triggered, but continue to trace.
1125 * Note, make sure to allocate the snapshot with either
1126 * a tracing_snapshot_alloc(), or by doing it manually
1127 * with: echo 1 > /sys/kernel/debug/tracing/snapshot
1129 * If the snapshot buffer is not allocated, it will stop tracing.
1130 * Basically making a permanent snapshot.
1132 void tracing_snapshot(void)
1134 struct trace_array
*tr
= &global_trace
;
1136 tracing_snapshot_instance(tr
);
1138 EXPORT_SYMBOL_GPL(tracing_snapshot
);
1141 * tracing_snapshot_cond - conditionally take a snapshot of the current buffer.
1142 * @tr: The tracing instance to snapshot
1143 * @cond_data: The data to be tested conditionally, and possibly saved
1145 * This is the same as tracing_snapshot() except that the snapshot is
1146 * conditional - the snapshot will only happen if the
1147 * cond_snapshot.update() implementation receiving the cond_data
1148 * returns true, which means that the trace array's cond_snapshot
1149 * update() operation used the cond_data to determine whether the
1150 * snapshot should be taken, and if it was, presumably saved it along
1151 * with the snapshot.
1153 void tracing_snapshot_cond(struct trace_array
*tr
, void *cond_data
)
1155 tracing_snapshot_instance_cond(tr
, cond_data
);
1157 EXPORT_SYMBOL_GPL(tracing_snapshot_cond
);
1160 * tracing_snapshot_cond_data - get the user data associated with a snapshot
1161 * @tr: The tracing instance
1163 * When the user enables a conditional snapshot using
1164 * tracing_snapshot_cond_enable(), the user-defined cond_data is saved
1165 * with the snapshot. This accessor is used to retrieve it.
1167 * Should not be called from cond_snapshot.update(), since it takes
1168 * the tr->max_lock lock, which the code calling
1169 * cond_snapshot.update() has already done.
1171 * Returns the cond_data associated with the trace array's snapshot.
1173 void *tracing_cond_snapshot_data(struct trace_array
*tr
)
1175 void *cond_data
= NULL
;
1177 arch_spin_lock(&tr
->max_lock
);
1179 if (tr
->cond_snapshot
)
1180 cond_data
= tr
->cond_snapshot
->cond_data
;
1182 arch_spin_unlock(&tr
->max_lock
);
1186 EXPORT_SYMBOL_GPL(tracing_cond_snapshot_data
);
1188 static int resize_buffer_duplicate_size(struct array_buffer
*trace_buf
,
1189 struct array_buffer
*size_buf
, int cpu_id
);
1190 static void set_buffer_entries(struct array_buffer
*buf
, unsigned long val
);
1192 int tracing_alloc_snapshot_instance(struct trace_array
*tr
)
1196 if (!tr
->allocated_snapshot
) {
1198 /* allocate spare buffer */
1199 ret
= resize_buffer_duplicate_size(&tr
->max_buffer
,
1200 &tr
->array_buffer
, RING_BUFFER_ALL_CPUS
);
1204 tr
->allocated_snapshot
= true;
1210 static void free_snapshot(struct trace_array
*tr
)
1213 * We don't free the ring buffer. instead, resize it because
1214 * The max_tr ring buffer has some state (e.g. ring->clock) and
1215 * we want preserve it.
1217 ring_buffer_resize(tr
->max_buffer
.buffer
, 1, RING_BUFFER_ALL_CPUS
);
1218 set_buffer_entries(&tr
->max_buffer
, 1);
1219 tracing_reset_online_cpus(&tr
->max_buffer
);
1220 tr
->allocated_snapshot
= false;
1224 * tracing_alloc_snapshot - allocate snapshot buffer.
1226 * This only allocates the snapshot buffer if it isn't already
1227 * allocated - it doesn't also take a snapshot.
1229 * This is meant to be used in cases where the snapshot buffer needs
1230 * to be set up for events that can't sleep but need to be able to
1231 * trigger a snapshot.
1233 int tracing_alloc_snapshot(void)
1235 struct trace_array
*tr
= &global_trace
;
1238 ret
= tracing_alloc_snapshot_instance(tr
);
1243 EXPORT_SYMBOL_GPL(tracing_alloc_snapshot
);
1246 * tracing_snapshot_alloc - allocate and take a snapshot of the current buffer.
1248 * This is similar to tracing_snapshot(), but it will allocate the
1249 * snapshot buffer if it isn't already allocated. Use this only
1250 * where it is safe to sleep, as the allocation may sleep.
1252 * This causes a swap between the snapshot buffer and the current live
1253 * tracing buffer. You can use this to take snapshots of the live
1254 * trace when some condition is triggered, but continue to trace.
1256 void tracing_snapshot_alloc(void)
1260 ret
= tracing_alloc_snapshot();
1266 EXPORT_SYMBOL_GPL(tracing_snapshot_alloc
);
1269 * tracing_snapshot_cond_enable - enable conditional snapshot for an instance
1270 * @tr: The tracing instance
1271 * @cond_data: User data to associate with the snapshot
1272 * @update: Implementation of the cond_snapshot update function
1274 * Check whether the conditional snapshot for the given instance has
1275 * already been enabled, or if the current tracer is already using a
1276 * snapshot; if so, return -EBUSY, else create a cond_snapshot and
1277 * save the cond_data and update function inside.
1279 * Returns 0 if successful, error otherwise.
1281 int tracing_snapshot_cond_enable(struct trace_array
*tr
, void *cond_data
,
1282 cond_update_fn_t update
)
1284 struct cond_snapshot
*cond_snapshot
;
1287 cond_snapshot
= kzalloc(sizeof(*cond_snapshot
), GFP_KERNEL
);
1291 cond_snapshot
->cond_data
= cond_data
;
1292 cond_snapshot
->update
= update
;
1294 mutex_lock(&trace_types_lock
);
1296 ret
= tracing_alloc_snapshot_instance(tr
);
1300 if (tr
->current_trace
->use_max_tr
) {
1306 * The cond_snapshot can only change to NULL without the
1307 * trace_types_lock. We don't care if we race with it going
1308 * to NULL, but we want to make sure that it's not set to
1309 * something other than NULL when we get here, which we can
1310 * do safely with only holding the trace_types_lock and not
1311 * having to take the max_lock.
1313 if (tr
->cond_snapshot
) {
1318 arch_spin_lock(&tr
->max_lock
);
1319 tr
->cond_snapshot
= cond_snapshot
;
1320 arch_spin_unlock(&tr
->max_lock
);
1322 mutex_unlock(&trace_types_lock
);
1327 mutex_unlock(&trace_types_lock
);
1328 kfree(cond_snapshot
);
1331 EXPORT_SYMBOL_GPL(tracing_snapshot_cond_enable
);
1334 * tracing_snapshot_cond_disable - disable conditional snapshot for an instance
1335 * @tr: The tracing instance
1337 * Check whether the conditional snapshot for the given instance is
1338 * enabled; if so, free the cond_snapshot associated with it,
1339 * otherwise return -EINVAL.
1341 * Returns 0 if successful, error otherwise.
1343 int tracing_snapshot_cond_disable(struct trace_array
*tr
)
1347 arch_spin_lock(&tr
->max_lock
);
1349 if (!tr
->cond_snapshot
)
1352 kfree(tr
->cond_snapshot
);
1353 tr
->cond_snapshot
= NULL
;
1356 arch_spin_unlock(&tr
->max_lock
);
1360 EXPORT_SYMBOL_GPL(tracing_snapshot_cond_disable
);
1362 void tracing_snapshot(void)
1364 WARN_ONCE(1, "Snapshot feature not enabled, but internal snapshot used");
1366 EXPORT_SYMBOL_GPL(tracing_snapshot
);
1367 void tracing_snapshot_cond(struct trace_array
*tr
, void *cond_data
)
1369 WARN_ONCE(1, "Snapshot feature not enabled, but internal conditional snapshot used");
1371 EXPORT_SYMBOL_GPL(tracing_snapshot_cond
);
1372 int tracing_alloc_snapshot(void)
1374 WARN_ONCE(1, "Snapshot feature not enabled, but snapshot allocation used");
1377 EXPORT_SYMBOL_GPL(tracing_alloc_snapshot
);
1378 void tracing_snapshot_alloc(void)
1383 EXPORT_SYMBOL_GPL(tracing_snapshot_alloc
);
1384 void *tracing_cond_snapshot_data(struct trace_array
*tr
)
1388 EXPORT_SYMBOL_GPL(tracing_cond_snapshot_data
);
1389 int tracing_snapshot_cond_enable(struct trace_array
*tr
, void *cond_data
, cond_update_fn_t update
)
1393 EXPORT_SYMBOL_GPL(tracing_snapshot_cond_enable
);
1394 int tracing_snapshot_cond_disable(struct trace_array
*tr
)
1398 EXPORT_SYMBOL_GPL(tracing_snapshot_cond_disable
);
1399 #endif /* CONFIG_TRACER_SNAPSHOT */
1401 void tracer_tracing_off(struct trace_array
*tr
)
1403 if (tr
->array_buffer
.buffer
)
1404 ring_buffer_record_off(tr
->array_buffer
.buffer
);
1406 * This flag is looked at when buffers haven't been allocated
1407 * yet, or by some tracers (like irqsoff), that just want to
1408 * know if the ring buffer has been disabled, but it can handle
1409 * races of where it gets disabled but we still do a record.
1410 * As the check is in the fast path of the tracers, it is more
1411 * important to be fast than accurate.
1413 tr
->buffer_disabled
= 1;
1414 /* Make the flag seen by readers */
1419 * tracing_off - turn off tracing buffers
1421 * This function stops the tracing buffers from recording data.
1422 * It does not disable any overhead the tracers themselves may
1423 * be causing. This function simply causes all recording to
1424 * the ring buffers to fail.
1426 void tracing_off(void)
1428 tracer_tracing_off(&global_trace
);
1430 EXPORT_SYMBOL_GPL(tracing_off
);
1432 void disable_trace_on_warning(void)
1434 if (__disable_trace_on_warning
) {
1435 trace_array_printk_buf(global_trace
.array_buffer
.buffer
, _THIS_IP_
,
1436 "Disabling tracing due to warning\n");
1442 * tracer_tracing_is_on - show real state of ring buffer enabled
1443 * @tr : the trace array to know if ring buffer is enabled
1445 * Shows real state of the ring buffer if it is enabled or not.
1447 bool tracer_tracing_is_on(struct trace_array
*tr
)
1449 if (tr
->array_buffer
.buffer
)
1450 return ring_buffer_record_is_on(tr
->array_buffer
.buffer
);
1451 return !tr
->buffer_disabled
;
1455 * tracing_is_on - show state of ring buffers enabled
1457 int tracing_is_on(void)
1459 return tracer_tracing_is_on(&global_trace
);
1461 EXPORT_SYMBOL_GPL(tracing_is_on
);
1463 static int __init
set_buf_size(char *str
)
1465 unsigned long buf_size
;
1469 buf_size
= memparse(str
, &str
);
1470 /* nr_entries can not be zero */
1473 trace_buf_size
= buf_size
;
1476 __setup("trace_buf_size=", set_buf_size
);
1478 static int __init
set_tracing_thresh(char *str
)
1480 unsigned long threshold
;
1485 ret
= kstrtoul(str
, 0, &threshold
);
1488 tracing_thresh
= threshold
* 1000;
1491 __setup("tracing_thresh=", set_tracing_thresh
);
1493 unsigned long nsecs_to_usecs(unsigned long nsecs
)
1495 return nsecs
/ 1000;
1499 * TRACE_FLAGS is defined as a tuple matching bit masks with strings.
1500 * It uses C(a, b) where 'a' is the eval (enum) name and 'b' is the string that
1501 * matches it. By defining "C(a, b) b", TRACE_FLAGS becomes a list
1502 * of strings in the order that the evals (enum) were defined.
1507 /* These must match the bit postions in trace_iterator_flags */
1508 static const char *trace_options
[] = {
1516 int in_ns
; /* is this clock in nanoseconds? */
1517 } trace_clocks
[] = {
1518 { trace_clock_local
, "local", 1 },
1519 { trace_clock_global
, "global", 1 },
1520 { trace_clock_counter
, "counter", 0 },
1521 { trace_clock_jiffies
, "uptime", 0 },
1522 { trace_clock
, "perf", 1 },
1523 { ktime_get_mono_fast_ns
, "mono", 1 },
1524 { ktime_get_raw_fast_ns
, "mono_raw", 1 },
1525 { ktime_get_boot_fast_ns
, "boot", 1 },
1529 bool trace_clock_in_ns(struct trace_array
*tr
)
1531 if (trace_clocks
[tr
->clock_id
].in_ns
)
1538 * trace_parser_get_init - gets the buffer for trace parser
1540 int trace_parser_get_init(struct trace_parser
*parser
, int size
)
1542 memset(parser
, 0, sizeof(*parser
));
1544 parser
->buffer
= kmalloc(size
, GFP_KERNEL
);
1545 if (!parser
->buffer
)
1548 parser
->size
= size
;
1553 * trace_parser_put - frees the buffer for trace parser
1555 void trace_parser_put(struct trace_parser
*parser
)
1557 kfree(parser
->buffer
);
1558 parser
->buffer
= NULL
;
1562 * trace_get_user - reads the user input string separated by space
1563 * (matched by isspace(ch))
1565 * For each string found the 'struct trace_parser' is updated,
1566 * and the function returns.
1568 * Returns number of bytes read.
1570 * See kernel/trace/trace.h for 'struct trace_parser' details.
1572 int trace_get_user(struct trace_parser
*parser
, const char __user
*ubuf
,
1573 size_t cnt
, loff_t
*ppos
)
1580 trace_parser_clear(parser
);
1582 ret
= get_user(ch
, ubuf
++);
1590 * The parser is not finished with the last write,
1591 * continue reading the user input without skipping spaces.
1593 if (!parser
->cont
) {
1594 /* skip white space */
1595 while (cnt
&& isspace(ch
)) {
1596 ret
= get_user(ch
, ubuf
++);
1605 /* only spaces were written */
1606 if (isspace(ch
) || !ch
) {
1613 /* read the non-space input */
1614 while (cnt
&& !isspace(ch
) && ch
) {
1615 if (parser
->idx
< parser
->size
- 1)
1616 parser
->buffer
[parser
->idx
++] = ch
;
1621 ret
= get_user(ch
, ubuf
++);
1628 /* We either got finished input or we have to wait for another call. */
1629 if (isspace(ch
) || !ch
) {
1630 parser
->buffer
[parser
->idx
] = 0;
1631 parser
->cont
= false;
1632 } else if (parser
->idx
< parser
->size
- 1) {
1633 parser
->cont
= true;
1634 parser
->buffer
[parser
->idx
++] = ch
;
1635 /* Make sure the parsed string always terminates with '\0'. */
1636 parser
->buffer
[parser
->idx
] = 0;
1649 /* TODO add a seq_buf_to_buffer() */
1650 static ssize_t
trace_seq_to_buffer(struct trace_seq
*s
, void *buf
, size_t cnt
)
1654 if (trace_seq_used(s
) <= s
->seq
.readpos
)
1657 len
= trace_seq_used(s
) - s
->seq
.readpos
;
1660 memcpy(buf
, s
->buffer
+ s
->seq
.readpos
, cnt
);
1662 s
->seq
.readpos
+= cnt
;
1666 unsigned long __read_mostly tracing_thresh
;
1667 static const struct file_operations tracing_max_lat_fops
;
1669 #if (defined(CONFIG_TRACER_MAX_TRACE) || defined(CONFIG_HWLAT_TRACER)) && \
1670 defined(CONFIG_FSNOTIFY)
1672 static struct workqueue_struct
*fsnotify_wq
;
1674 static void latency_fsnotify_workfn(struct work_struct
*work
)
1676 struct trace_array
*tr
= container_of(work
, struct trace_array
,
1678 fsnotify_inode(tr
->d_max_latency
->d_inode
, FS_MODIFY
);
1681 static void latency_fsnotify_workfn_irq(struct irq_work
*iwork
)
1683 struct trace_array
*tr
= container_of(iwork
, struct trace_array
,
1685 queue_work(fsnotify_wq
, &tr
->fsnotify_work
);
1688 static void trace_create_maxlat_file(struct trace_array
*tr
,
1689 struct dentry
*d_tracer
)
1691 INIT_WORK(&tr
->fsnotify_work
, latency_fsnotify_workfn
);
1692 init_irq_work(&tr
->fsnotify_irqwork
, latency_fsnotify_workfn_irq
);
1693 tr
->d_max_latency
= trace_create_file("tracing_max_latency", 0644,
1694 d_tracer
, &tr
->max_latency
,
1695 &tracing_max_lat_fops
);
1698 __init
static int latency_fsnotify_init(void)
1700 fsnotify_wq
= alloc_workqueue("tr_max_lat_wq",
1701 WQ_UNBOUND
| WQ_HIGHPRI
, 0);
1703 pr_err("Unable to allocate tr_max_lat_wq\n");
1709 late_initcall_sync(latency_fsnotify_init
);
1711 void latency_fsnotify(struct trace_array
*tr
)
1716 * We cannot call queue_work(&tr->fsnotify_work) from here because it's
1717 * possible that we are called from __schedule() or do_idle(), which
1718 * could cause a deadlock.
1720 irq_work_queue(&tr
->fsnotify_irqwork
);
1724 * (defined(CONFIG_TRACER_MAX_TRACE) || defined(CONFIG_HWLAT_TRACER)) && \
1725 * defined(CONFIG_FSNOTIFY)
1729 #define trace_create_maxlat_file(tr, d_tracer) \
1730 trace_create_file("tracing_max_latency", 0644, d_tracer, \
1731 &tr->max_latency, &tracing_max_lat_fops)
1735 #ifdef CONFIG_TRACER_MAX_TRACE
1737 * Copy the new maximum trace into the separate maximum-trace
1738 * structure. (this way the maximum trace is permanently saved,
1739 * for later retrieval via /sys/kernel/tracing/tracing_max_latency)
1742 __update_max_tr(struct trace_array
*tr
, struct task_struct
*tsk
, int cpu
)
1744 struct array_buffer
*trace_buf
= &tr
->array_buffer
;
1745 struct array_buffer
*max_buf
= &tr
->max_buffer
;
1746 struct trace_array_cpu
*data
= per_cpu_ptr(trace_buf
->data
, cpu
);
1747 struct trace_array_cpu
*max_data
= per_cpu_ptr(max_buf
->data
, cpu
);
1750 max_buf
->time_start
= data
->preempt_timestamp
;
1752 max_data
->saved_latency
= tr
->max_latency
;
1753 max_data
->critical_start
= data
->critical_start
;
1754 max_data
->critical_end
= data
->critical_end
;
1756 strncpy(max_data
->comm
, tsk
->comm
, TASK_COMM_LEN
);
1757 max_data
->pid
= tsk
->pid
;
1759 * If tsk == current, then use current_uid(), as that does not use
1760 * RCU. The irq tracer can be called out of RCU scope.
1763 max_data
->uid
= current_uid();
1765 max_data
->uid
= task_uid(tsk
);
1767 max_data
->nice
= tsk
->static_prio
- 20 - MAX_RT_PRIO
;
1768 max_data
->policy
= tsk
->policy
;
1769 max_data
->rt_priority
= tsk
->rt_priority
;
1771 /* record this tasks comm */
1772 tracing_record_cmdline(tsk
);
1773 latency_fsnotify(tr
);
1777 * update_max_tr - snapshot all trace buffers from global_trace to max_tr
1779 * @tsk: the task with the latency
1780 * @cpu: The cpu that initiated the trace.
1781 * @cond_data: User data associated with a conditional snapshot
1783 * Flip the buffers between the @tr and the max_tr and record information
1784 * about which task was the cause of this latency.
1787 update_max_tr(struct trace_array
*tr
, struct task_struct
*tsk
, int cpu
,
1793 WARN_ON_ONCE(!irqs_disabled());
1795 if (!tr
->allocated_snapshot
) {
1796 /* Only the nop tracer should hit this when disabling */
1797 WARN_ON_ONCE(tr
->current_trace
!= &nop_trace
);
1801 arch_spin_lock(&tr
->max_lock
);
1803 /* Inherit the recordable setting from array_buffer */
1804 if (ring_buffer_record_is_set_on(tr
->array_buffer
.buffer
))
1805 ring_buffer_record_on(tr
->max_buffer
.buffer
);
1807 ring_buffer_record_off(tr
->max_buffer
.buffer
);
1809 #ifdef CONFIG_TRACER_SNAPSHOT
1810 if (tr
->cond_snapshot
&& !tr
->cond_snapshot
->update(tr
, cond_data
))
1813 swap(tr
->array_buffer
.buffer
, tr
->max_buffer
.buffer
);
1815 __update_max_tr(tr
, tsk
, cpu
);
1818 arch_spin_unlock(&tr
->max_lock
);
1822 * update_max_tr_single - only copy one trace over, and reset the rest
1824 * @tsk: task with the latency
1825 * @cpu: the cpu of the buffer to copy.
1827 * Flip the trace of a single CPU buffer between the @tr and the max_tr.
1830 update_max_tr_single(struct trace_array
*tr
, struct task_struct
*tsk
, int cpu
)
1837 WARN_ON_ONCE(!irqs_disabled());
1838 if (!tr
->allocated_snapshot
) {
1839 /* Only the nop tracer should hit this when disabling */
1840 WARN_ON_ONCE(tr
->current_trace
!= &nop_trace
);
1844 arch_spin_lock(&tr
->max_lock
);
1846 ret
= ring_buffer_swap_cpu(tr
->max_buffer
.buffer
, tr
->array_buffer
.buffer
, cpu
);
1848 if (ret
== -EBUSY
) {
1850 * We failed to swap the buffer due to a commit taking
1851 * place on this CPU. We fail to record, but we reset
1852 * the max trace buffer (no one writes directly to it)
1853 * and flag that it failed.
1855 trace_array_printk_buf(tr
->max_buffer
.buffer
, _THIS_IP_
,
1856 "Failed to swap buffers due to commit in progress\n");
1859 WARN_ON_ONCE(ret
&& ret
!= -EAGAIN
&& ret
!= -EBUSY
);
1861 __update_max_tr(tr
, tsk
, cpu
);
1862 arch_spin_unlock(&tr
->max_lock
);
1864 #endif /* CONFIG_TRACER_MAX_TRACE */
1866 static int wait_on_pipe(struct trace_iterator
*iter
, int full
)
1868 /* Iterators are static, they should be filled or empty */
1869 if (trace_buffer_iter(iter
, iter
->cpu_file
))
1872 return ring_buffer_wait(iter
->array_buffer
->buffer
, iter
->cpu_file
,
1876 #ifdef CONFIG_FTRACE_STARTUP_TEST
1877 static bool selftests_can_run
;
1879 struct trace_selftests
{
1880 struct list_head list
;
1881 struct tracer
*type
;
1884 static LIST_HEAD(postponed_selftests
);
1886 static int save_selftest(struct tracer
*type
)
1888 struct trace_selftests
*selftest
;
1890 selftest
= kmalloc(sizeof(*selftest
), GFP_KERNEL
);
1894 selftest
->type
= type
;
1895 list_add(&selftest
->list
, &postponed_selftests
);
1899 static int run_tracer_selftest(struct tracer
*type
)
1901 struct trace_array
*tr
= &global_trace
;
1902 struct tracer
*saved_tracer
= tr
->current_trace
;
1905 if (!type
->selftest
|| tracing_selftest_disabled
)
1909 * If a tracer registers early in boot up (before scheduling is
1910 * initialized and such), then do not run its selftests yet.
1911 * Instead, run it a little later in the boot process.
1913 if (!selftests_can_run
)
1914 return save_selftest(type
);
1917 * Run a selftest on this tracer.
1918 * Here we reset the trace buffer, and set the current
1919 * tracer to be this tracer. The tracer can then run some
1920 * internal tracing to verify that everything is in order.
1921 * If we fail, we do not register this tracer.
1923 tracing_reset_online_cpus(&tr
->array_buffer
);
1925 tr
->current_trace
= type
;
1927 #ifdef CONFIG_TRACER_MAX_TRACE
1928 if (type
->use_max_tr
) {
1929 /* If we expanded the buffers, make sure the max is expanded too */
1930 if (ring_buffer_expanded
)
1931 ring_buffer_resize(tr
->max_buffer
.buffer
, trace_buf_size
,
1932 RING_BUFFER_ALL_CPUS
);
1933 tr
->allocated_snapshot
= true;
1937 /* the test is responsible for initializing and enabling */
1938 pr_info("Testing tracer %s: ", type
->name
);
1939 ret
= type
->selftest(type
, tr
);
1940 /* the test is responsible for resetting too */
1941 tr
->current_trace
= saved_tracer
;
1943 printk(KERN_CONT
"FAILED!\n");
1944 /* Add the warning after printing 'FAILED' */
1948 /* Only reset on passing, to avoid touching corrupted buffers */
1949 tracing_reset_online_cpus(&tr
->array_buffer
);
1951 #ifdef CONFIG_TRACER_MAX_TRACE
1952 if (type
->use_max_tr
) {
1953 tr
->allocated_snapshot
= false;
1955 /* Shrink the max buffer again */
1956 if (ring_buffer_expanded
)
1957 ring_buffer_resize(tr
->max_buffer
.buffer
, 1,
1958 RING_BUFFER_ALL_CPUS
);
1962 printk(KERN_CONT
"PASSED\n");
1966 static __init
int init_trace_selftests(void)
1968 struct trace_selftests
*p
, *n
;
1969 struct tracer
*t
, **last
;
1972 selftests_can_run
= true;
1974 mutex_lock(&trace_types_lock
);
1976 if (list_empty(&postponed_selftests
))
1979 pr_info("Running postponed tracer tests:\n");
1981 tracing_selftest_running
= true;
1982 list_for_each_entry_safe(p
, n
, &postponed_selftests
, list
) {
1983 /* This loop can take minutes when sanitizers are enabled, so
1984 * lets make sure we allow RCU processing.
1987 ret
= run_tracer_selftest(p
->type
);
1988 /* If the test fails, then warn and remove from available_tracers */
1990 WARN(1, "tracer: %s failed selftest, disabling\n",
1992 last
= &trace_types
;
1993 for (t
= trace_types
; t
; t
= t
->next
) {
2004 tracing_selftest_running
= false;
2007 mutex_unlock(&trace_types_lock
);
2011 core_initcall(init_trace_selftests
);
2013 static inline int run_tracer_selftest(struct tracer
*type
)
2017 #endif /* CONFIG_FTRACE_STARTUP_TEST */
2019 static void add_tracer_options(struct trace_array
*tr
, struct tracer
*t
);
2021 static void __init
apply_trace_boot_options(void);
2024 * register_tracer - register a tracer with the ftrace system.
2025 * @type: the plugin for the tracer
2027 * Register a new plugin tracer.
2029 int __init
register_tracer(struct tracer
*type
)
2035 pr_info("Tracer must have a name\n");
2039 if (strlen(type
->name
) >= MAX_TRACER_SIZE
) {
2040 pr_info("Tracer has a name longer than %d\n", MAX_TRACER_SIZE
);
2044 if (security_locked_down(LOCKDOWN_TRACEFS
)) {
2045 pr_warn("Can not register tracer %s due to lockdown\n",
2050 mutex_lock(&trace_types_lock
);
2052 tracing_selftest_running
= true;
2054 for (t
= trace_types
; t
; t
= t
->next
) {
2055 if (strcmp(type
->name
, t
->name
) == 0) {
2057 pr_info("Tracer %s already registered\n",
2064 if (!type
->set_flag
)
2065 type
->set_flag
= &dummy_set_flag
;
2067 /*allocate a dummy tracer_flags*/
2068 type
->flags
= kmalloc(sizeof(*type
->flags
), GFP_KERNEL
);
2073 type
->flags
->val
= 0;
2074 type
->flags
->opts
= dummy_tracer_opt
;
2076 if (!type
->flags
->opts
)
2077 type
->flags
->opts
= dummy_tracer_opt
;
2079 /* store the tracer for __set_tracer_option */
2080 type
->flags
->trace
= type
;
2082 ret
= run_tracer_selftest(type
);
2086 type
->next
= trace_types
;
2088 add_tracer_options(&global_trace
, type
);
2091 tracing_selftest_running
= false;
2092 mutex_unlock(&trace_types_lock
);
2094 if (ret
|| !default_bootup_tracer
)
2097 if (strncmp(default_bootup_tracer
, type
->name
, MAX_TRACER_SIZE
))
2100 printk(KERN_INFO
"Starting tracer '%s'\n", type
->name
);
2101 /* Do we want this tracer to start on bootup? */
2102 tracing_set_tracer(&global_trace
, type
->name
);
2103 default_bootup_tracer
= NULL
;
2105 apply_trace_boot_options();
2107 /* disable other selftests, since this will break it. */
2108 tracing_selftest_disabled
= true;
2109 #ifdef CONFIG_FTRACE_STARTUP_TEST
2110 printk(KERN_INFO
"Disabling FTRACE selftests due to running tracer '%s'\n",
2118 static void tracing_reset_cpu(struct array_buffer
*buf
, int cpu
)
2120 struct trace_buffer
*buffer
= buf
->buffer
;
2125 ring_buffer_record_disable(buffer
);
2127 /* Make sure all commits have finished */
2129 ring_buffer_reset_cpu(buffer
, cpu
);
2131 ring_buffer_record_enable(buffer
);
2134 void tracing_reset_online_cpus(struct array_buffer
*buf
)
2136 struct trace_buffer
*buffer
= buf
->buffer
;
2141 ring_buffer_record_disable(buffer
);
2143 /* Make sure all commits have finished */
2146 buf
->time_start
= buffer_ftrace_now(buf
, buf
->cpu
);
2148 ring_buffer_reset_online_cpus(buffer
);
2150 ring_buffer_record_enable(buffer
);
2153 /* Must have trace_types_lock held */
2154 void tracing_reset_all_online_cpus(void)
2156 struct trace_array
*tr
;
2158 list_for_each_entry(tr
, &ftrace_trace_arrays
, list
) {
2159 if (!tr
->clear_trace
)
2161 tr
->clear_trace
= false;
2162 tracing_reset_online_cpus(&tr
->array_buffer
);
2163 #ifdef CONFIG_TRACER_MAX_TRACE
2164 tracing_reset_online_cpus(&tr
->max_buffer
);
2169 static int *tgid_map
;
2171 #define SAVED_CMDLINES_DEFAULT 128
2172 #define NO_CMDLINE_MAP UINT_MAX
2173 static arch_spinlock_t trace_cmdline_lock
= __ARCH_SPIN_LOCK_UNLOCKED
;
2174 struct saved_cmdlines_buffer
{
2175 unsigned map_pid_to_cmdline
[PID_MAX_DEFAULT
+1];
2176 unsigned *map_cmdline_to_pid
;
2177 unsigned cmdline_num
;
2179 char *saved_cmdlines
;
2181 static struct saved_cmdlines_buffer
*savedcmd
;
2183 /* temporary disable recording */
2184 static atomic_t trace_record_taskinfo_disabled __read_mostly
;
2186 static inline char *get_saved_cmdlines(int idx
)
2188 return &savedcmd
->saved_cmdlines
[idx
* TASK_COMM_LEN
];
2191 static inline void set_cmdline(int idx
, const char *cmdline
)
2193 strncpy(get_saved_cmdlines(idx
), cmdline
, TASK_COMM_LEN
);
2196 static int allocate_cmdlines_buffer(unsigned int val
,
2197 struct saved_cmdlines_buffer
*s
)
2199 s
->map_cmdline_to_pid
= kmalloc_array(val
,
2200 sizeof(*s
->map_cmdline_to_pid
),
2202 if (!s
->map_cmdline_to_pid
)
2205 s
->saved_cmdlines
= kmalloc_array(TASK_COMM_LEN
, val
, GFP_KERNEL
);
2206 if (!s
->saved_cmdlines
) {
2207 kfree(s
->map_cmdline_to_pid
);
2212 s
->cmdline_num
= val
;
2213 memset(&s
->map_pid_to_cmdline
, NO_CMDLINE_MAP
,
2214 sizeof(s
->map_pid_to_cmdline
));
2215 memset(s
->map_cmdline_to_pid
, NO_CMDLINE_MAP
,
2216 val
* sizeof(*s
->map_cmdline_to_pid
));
2221 static int trace_create_savedcmd(void)
2225 savedcmd
= kmalloc(sizeof(*savedcmd
), GFP_KERNEL
);
2229 ret
= allocate_cmdlines_buffer(SAVED_CMDLINES_DEFAULT
, savedcmd
);
2239 int is_tracing_stopped(void)
2241 return global_trace
.stop_count
;
2245 * tracing_start - quick start of the tracer
2247 * If tracing is enabled but was stopped by tracing_stop,
2248 * this will start the tracer back up.
2250 void tracing_start(void)
2252 struct trace_buffer
*buffer
;
2253 unsigned long flags
;
2255 if (tracing_disabled
)
2258 raw_spin_lock_irqsave(&global_trace
.start_lock
, flags
);
2259 if (--global_trace
.stop_count
) {
2260 if (global_trace
.stop_count
< 0) {
2261 /* Someone screwed up their debugging */
2263 global_trace
.stop_count
= 0;
2268 /* Prevent the buffers from switching */
2269 arch_spin_lock(&global_trace
.max_lock
);
2271 buffer
= global_trace
.array_buffer
.buffer
;
2273 ring_buffer_record_enable(buffer
);
2275 #ifdef CONFIG_TRACER_MAX_TRACE
2276 buffer
= global_trace
.max_buffer
.buffer
;
2278 ring_buffer_record_enable(buffer
);
2281 arch_spin_unlock(&global_trace
.max_lock
);
2284 raw_spin_unlock_irqrestore(&global_trace
.start_lock
, flags
);
2287 static void tracing_start_tr(struct trace_array
*tr
)
2289 struct trace_buffer
*buffer
;
2290 unsigned long flags
;
2292 if (tracing_disabled
)
2295 /* If global, we need to also start the max tracer */
2296 if (tr
->flags
& TRACE_ARRAY_FL_GLOBAL
)
2297 return tracing_start();
2299 raw_spin_lock_irqsave(&tr
->start_lock
, flags
);
2301 if (--tr
->stop_count
) {
2302 if (tr
->stop_count
< 0) {
2303 /* Someone screwed up their debugging */
2310 buffer
= tr
->array_buffer
.buffer
;
2312 ring_buffer_record_enable(buffer
);
2315 raw_spin_unlock_irqrestore(&tr
->start_lock
, flags
);
2319 * tracing_stop - quick stop of the tracer
2321 * Light weight way to stop tracing. Use in conjunction with
2324 void tracing_stop(void)
2326 struct trace_buffer
*buffer
;
2327 unsigned long flags
;
2329 raw_spin_lock_irqsave(&global_trace
.start_lock
, flags
);
2330 if (global_trace
.stop_count
++)
2333 /* Prevent the buffers from switching */
2334 arch_spin_lock(&global_trace
.max_lock
);
2336 buffer
= global_trace
.array_buffer
.buffer
;
2338 ring_buffer_record_disable(buffer
);
2340 #ifdef CONFIG_TRACER_MAX_TRACE
2341 buffer
= global_trace
.max_buffer
.buffer
;
2343 ring_buffer_record_disable(buffer
);
2346 arch_spin_unlock(&global_trace
.max_lock
);
2349 raw_spin_unlock_irqrestore(&global_trace
.start_lock
, flags
);
2352 static void tracing_stop_tr(struct trace_array
*tr
)
2354 struct trace_buffer
*buffer
;
2355 unsigned long flags
;
2357 /* If global, we need to also stop the max tracer */
2358 if (tr
->flags
& TRACE_ARRAY_FL_GLOBAL
)
2359 return tracing_stop();
2361 raw_spin_lock_irqsave(&tr
->start_lock
, flags
);
2362 if (tr
->stop_count
++)
2365 buffer
= tr
->array_buffer
.buffer
;
2367 ring_buffer_record_disable(buffer
);
2370 raw_spin_unlock_irqrestore(&tr
->start_lock
, flags
);
2373 static int trace_save_cmdline(struct task_struct
*tsk
)
2377 /* treat recording of idle task as a success */
2381 if (unlikely(tsk
->pid
> PID_MAX_DEFAULT
))
2385 * It's not the end of the world if we don't get
2386 * the lock, but we also don't want to spin
2387 * nor do we want to disable interrupts,
2388 * so if we miss here, then better luck next time.
2390 if (!arch_spin_trylock(&trace_cmdline_lock
))
2393 idx
= savedcmd
->map_pid_to_cmdline
[tsk
->pid
];
2394 if (idx
== NO_CMDLINE_MAP
) {
2395 idx
= (savedcmd
->cmdline_idx
+ 1) % savedcmd
->cmdline_num
;
2398 * Check whether the cmdline buffer at idx has a pid
2399 * mapped. We are going to overwrite that entry so we
2400 * need to clear the map_pid_to_cmdline. Otherwise we
2401 * would read the new comm for the old pid.
2403 pid
= savedcmd
->map_cmdline_to_pid
[idx
];
2404 if (pid
!= NO_CMDLINE_MAP
)
2405 savedcmd
->map_pid_to_cmdline
[pid
] = NO_CMDLINE_MAP
;
2407 savedcmd
->map_cmdline_to_pid
[idx
] = tsk
->pid
;
2408 savedcmd
->map_pid_to_cmdline
[tsk
->pid
] = idx
;
2410 savedcmd
->cmdline_idx
= idx
;
2413 set_cmdline(idx
, tsk
->comm
);
2415 arch_spin_unlock(&trace_cmdline_lock
);
2420 static void __trace_find_cmdline(int pid
, char comm
[])
2425 strcpy(comm
, "<idle>");
2429 if (WARN_ON_ONCE(pid
< 0)) {
2430 strcpy(comm
, "<XXX>");
2434 if (pid
> PID_MAX_DEFAULT
) {
2435 strcpy(comm
, "<...>");
2439 map
= savedcmd
->map_pid_to_cmdline
[pid
];
2440 if (map
!= NO_CMDLINE_MAP
)
2441 strlcpy(comm
, get_saved_cmdlines(map
), TASK_COMM_LEN
);
2443 strcpy(comm
, "<...>");
2446 void trace_find_cmdline(int pid
, char comm
[])
2449 arch_spin_lock(&trace_cmdline_lock
);
2451 __trace_find_cmdline(pid
, comm
);
2453 arch_spin_unlock(&trace_cmdline_lock
);
2457 int trace_find_tgid(int pid
)
2459 if (unlikely(!tgid_map
|| !pid
|| pid
> PID_MAX_DEFAULT
))
2462 return tgid_map
[pid
];
2465 static int trace_save_tgid(struct task_struct
*tsk
)
2467 /* treat recording of idle task as a success */
2471 if (unlikely(!tgid_map
|| tsk
->pid
> PID_MAX_DEFAULT
))
2474 tgid_map
[tsk
->pid
] = tsk
->tgid
;
2478 static bool tracing_record_taskinfo_skip(int flags
)
2480 if (unlikely(!(flags
& (TRACE_RECORD_CMDLINE
| TRACE_RECORD_TGID
))))
2482 if (atomic_read(&trace_record_taskinfo_disabled
) || !tracing_is_on())
2484 if (!__this_cpu_read(trace_taskinfo_save
))
2490 * tracing_record_taskinfo - record the task info of a task
2492 * @task: task to record
2493 * @flags: TRACE_RECORD_CMDLINE for recording comm
2494 * TRACE_RECORD_TGID for recording tgid
2496 void tracing_record_taskinfo(struct task_struct
*task
, int flags
)
2500 if (tracing_record_taskinfo_skip(flags
))
2504 * Record as much task information as possible. If some fail, continue
2505 * to try to record the others.
2507 done
= !(flags
& TRACE_RECORD_CMDLINE
) || trace_save_cmdline(task
);
2508 done
&= !(flags
& TRACE_RECORD_TGID
) || trace_save_tgid(task
);
2510 /* If recording any information failed, retry again soon. */
2514 __this_cpu_write(trace_taskinfo_save
, false);
2518 * tracing_record_taskinfo_sched_switch - record task info for sched_switch
2520 * @prev: previous task during sched_switch
2521 * @next: next task during sched_switch
2522 * @flags: TRACE_RECORD_CMDLINE for recording comm
2523 * TRACE_RECORD_TGID for recording tgid
2525 void tracing_record_taskinfo_sched_switch(struct task_struct
*prev
,
2526 struct task_struct
*next
, int flags
)
2530 if (tracing_record_taskinfo_skip(flags
))
2534 * Record as much task information as possible. If some fail, continue
2535 * to try to record the others.
2537 done
= !(flags
& TRACE_RECORD_CMDLINE
) || trace_save_cmdline(prev
);
2538 done
&= !(flags
& TRACE_RECORD_CMDLINE
) || trace_save_cmdline(next
);
2539 done
&= !(flags
& TRACE_RECORD_TGID
) || trace_save_tgid(prev
);
2540 done
&= !(flags
& TRACE_RECORD_TGID
) || trace_save_tgid(next
);
2542 /* If recording any information failed, retry again soon. */
2546 __this_cpu_write(trace_taskinfo_save
, false);
2549 /* Helpers to record a specific task information */
2550 void tracing_record_cmdline(struct task_struct
*task
)
2552 tracing_record_taskinfo(task
, TRACE_RECORD_CMDLINE
);
2555 void tracing_record_tgid(struct task_struct
*task
)
2557 tracing_record_taskinfo(task
, TRACE_RECORD_TGID
);
2561 * Several functions return TRACE_TYPE_PARTIAL_LINE if the trace_seq
2562 * overflowed, and TRACE_TYPE_HANDLED otherwise. This helper function
2563 * simplifies those functions and keeps them in sync.
2565 enum print_line_t
trace_handle_return(struct trace_seq
*s
)
2567 return trace_seq_has_overflowed(s
) ?
2568 TRACE_TYPE_PARTIAL_LINE
: TRACE_TYPE_HANDLED
;
2570 EXPORT_SYMBOL_GPL(trace_handle_return
);
2573 tracing_generic_entry_update(struct trace_entry
*entry
, unsigned short type
,
2574 unsigned long flags
, int pc
)
2576 struct task_struct
*tsk
= current
;
2578 entry
->preempt_count
= pc
& 0xff;
2579 entry
->pid
= (tsk
) ? tsk
->pid
: 0;
2582 #ifdef CONFIG_TRACE_IRQFLAGS_SUPPORT
2583 (irqs_disabled_flags(flags
) ? TRACE_FLAG_IRQS_OFF
: 0) |
2585 TRACE_FLAG_IRQS_NOSUPPORT
|
2587 ((pc
& NMI_MASK
) ? TRACE_FLAG_NMI
: 0) |
2588 ((pc
& HARDIRQ_MASK
) ? TRACE_FLAG_HARDIRQ
: 0) |
2589 ((pc
& SOFTIRQ_OFFSET
) ? TRACE_FLAG_SOFTIRQ
: 0) |
2590 (tif_need_resched() ? TRACE_FLAG_NEED_RESCHED
: 0) |
2591 (test_preempt_need_resched() ? TRACE_FLAG_PREEMPT_RESCHED
: 0);
2593 EXPORT_SYMBOL_GPL(tracing_generic_entry_update
);
2595 struct ring_buffer_event
*
2596 trace_buffer_lock_reserve(struct trace_buffer
*buffer
,
2599 unsigned long flags
, int pc
)
2601 return __trace_buffer_lock_reserve(buffer
, type
, len
, flags
, pc
);
2604 DEFINE_PER_CPU(struct ring_buffer_event
*, trace_buffered_event
);
2605 DEFINE_PER_CPU(int, trace_buffered_event_cnt
);
2606 static int trace_buffered_event_ref
;
2609 * trace_buffered_event_enable - enable buffering events
2611 * When events are being filtered, it is quicker to use a temporary
2612 * buffer to write the event data into if there's a likely chance
2613 * that it will not be committed. The discard of the ring buffer
2614 * is not as fast as committing, and is much slower than copying
2617 * When an event is to be filtered, allocate per cpu buffers to
2618 * write the event data into, and if the event is filtered and discarded
2619 * it is simply dropped, otherwise, the entire data is to be committed
2622 void trace_buffered_event_enable(void)
2624 struct ring_buffer_event
*event
;
2628 WARN_ON_ONCE(!mutex_is_locked(&event_mutex
));
2630 if (trace_buffered_event_ref
++)
2633 for_each_tracing_cpu(cpu
) {
2634 page
= alloc_pages_node(cpu_to_node(cpu
),
2635 GFP_KERNEL
| __GFP_NORETRY
, 0);
2639 event
= page_address(page
);
2640 memset(event
, 0, sizeof(*event
));
2642 per_cpu(trace_buffered_event
, cpu
) = event
;
2645 if (cpu
== smp_processor_id() &&
2646 this_cpu_read(trace_buffered_event
) !=
2647 per_cpu(trace_buffered_event
, cpu
))
2654 trace_buffered_event_disable();
2657 static void enable_trace_buffered_event(void *data
)
2659 /* Probably not needed, but do it anyway */
2661 this_cpu_dec(trace_buffered_event_cnt
);
2664 static void disable_trace_buffered_event(void *data
)
2666 this_cpu_inc(trace_buffered_event_cnt
);
2670 * trace_buffered_event_disable - disable buffering events
2672 * When a filter is removed, it is faster to not use the buffered
2673 * events, and to commit directly into the ring buffer. Free up
2674 * the temp buffers when there are no more users. This requires
2675 * special synchronization with current events.
2677 void trace_buffered_event_disable(void)
2681 WARN_ON_ONCE(!mutex_is_locked(&event_mutex
));
2683 if (WARN_ON_ONCE(!trace_buffered_event_ref
))
2686 if (--trace_buffered_event_ref
)
2690 /* For each CPU, set the buffer as used. */
2691 smp_call_function_many(tracing_buffer_mask
,
2692 disable_trace_buffered_event
, NULL
, 1);
2695 /* Wait for all current users to finish */
2698 for_each_tracing_cpu(cpu
) {
2699 free_page((unsigned long)per_cpu(trace_buffered_event
, cpu
));
2700 per_cpu(trace_buffered_event
, cpu
) = NULL
;
2703 * Make sure trace_buffered_event is NULL before clearing
2704 * trace_buffered_event_cnt.
2709 /* Do the work on each cpu */
2710 smp_call_function_many(tracing_buffer_mask
,
2711 enable_trace_buffered_event
, NULL
, 1);
2715 static struct trace_buffer
*temp_buffer
;
2717 struct ring_buffer_event
*
2718 trace_event_buffer_lock_reserve(struct trace_buffer
**current_rb
,
2719 struct trace_event_file
*trace_file
,
2720 int type
, unsigned long len
,
2721 unsigned long flags
, int pc
)
2723 struct ring_buffer_event
*entry
;
2726 *current_rb
= trace_file
->tr
->array_buffer
.buffer
;
2728 if (!ring_buffer_time_stamp_abs(*current_rb
) && (trace_file
->flags
&
2729 (EVENT_FILE_FL_SOFT_DISABLED
| EVENT_FILE_FL_FILTERED
)) &&
2730 (entry
= this_cpu_read(trace_buffered_event
))) {
2731 /* Try to use the per cpu buffer first */
2732 val
= this_cpu_inc_return(trace_buffered_event_cnt
);
2734 trace_event_setup(entry
, type
, flags
, pc
);
2735 entry
->array
[0] = len
;
2738 this_cpu_dec(trace_buffered_event_cnt
);
2741 entry
= __trace_buffer_lock_reserve(*current_rb
,
2742 type
, len
, flags
, pc
);
2744 * If tracing is off, but we have triggers enabled
2745 * we still need to look at the event data. Use the temp_buffer
2746 * to store the trace event for the tigger to use. It's recusive
2747 * safe and will not be recorded anywhere.
2749 if (!entry
&& trace_file
->flags
& EVENT_FILE_FL_TRIGGER_COND
) {
2750 *current_rb
= temp_buffer
;
2751 entry
= __trace_buffer_lock_reserve(*current_rb
,
2752 type
, len
, flags
, pc
);
2756 EXPORT_SYMBOL_GPL(trace_event_buffer_lock_reserve
);
2758 static DEFINE_SPINLOCK(tracepoint_iter_lock
);
2759 static DEFINE_MUTEX(tracepoint_printk_mutex
);
2761 static void output_printk(struct trace_event_buffer
*fbuffer
)
2763 struct trace_event_call
*event_call
;
2764 struct trace_event_file
*file
;
2765 struct trace_event
*event
;
2766 unsigned long flags
;
2767 struct trace_iterator
*iter
= tracepoint_print_iter
;
2769 /* We should never get here if iter is NULL */
2770 if (WARN_ON_ONCE(!iter
))
2773 event_call
= fbuffer
->trace_file
->event_call
;
2774 if (!event_call
|| !event_call
->event
.funcs
||
2775 !event_call
->event
.funcs
->trace
)
2778 file
= fbuffer
->trace_file
;
2779 if (test_bit(EVENT_FILE_FL_SOFT_DISABLED_BIT
, &file
->flags
) ||
2780 (unlikely(file
->flags
& EVENT_FILE_FL_FILTERED
) &&
2781 !filter_match_preds(file
->filter
, fbuffer
->entry
)))
2784 event
= &fbuffer
->trace_file
->event_call
->event
;
2786 spin_lock_irqsave(&tracepoint_iter_lock
, flags
);
2787 trace_seq_init(&iter
->seq
);
2788 iter
->ent
= fbuffer
->entry
;
2789 event_call
->event
.funcs
->trace(iter
, 0, event
);
2790 trace_seq_putc(&iter
->seq
, 0);
2791 printk("%s", iter
->seq
.buffer
);
2793 spin_unlock_irqrestore(&tracepoint_iter_lock
, flags
);
2796 int tracepoint_printk_sysctl(struct ctl_table
*table
, int write
,
2797 void *buffer
, size_t *lenp
,
2800 int save_tracepoint_printk
;
2803 mutex_lock(&tracepoint_printk_mutex
);
2804 save_tracepoint_printk
= tracepoint_printk
;
2806 ret
= proc_dointvec(table
, write
, buffer
, lenp
, ppos
);
2809 * This will force exiting early, as tracepoint_printk
2810 * is always zero when tracepoint_printk_iter is not allocated
2812 if (!tracepoint_print_iter
)
2813 tracepoint_printk
= 0;
2815 if (save_tracepoint_printk
== tracepoint_printk
)
2818 if (tracepoint_printk
)
2819 static_key_enable(&tracepoint_printk_key
.key
);
2821 static_key_disable(&tracepoint_printk_key
.key
);
2824 mutex_unlock(&tracepoint_printk_mutex
);
2829 void trace_event_buffer_commit(struct trace_event_buffer
*fbuffer
)
2831 if (static_key_false(&tracepoint_printk_key
.key
))
2832 output_printk(fbuffer
);
2834 if (static_branch_unlikely(&trace_event_exports_enabled
))
2835 ftrace_exports(fbuffer
->event
, TRACE_EXPORT_EVENT
);
2836 event_trigger_unlock_commit_regs(fbuffer
->trace_file
, fbuffer
->buffer
,
2837 fbuffer
->event
, fbuffer
->entry
,
2838 fbuffer
->flags
, fbuffer
->pc
, fbuffer
->regs
);
2840 EXPORT_SYMBOL_GPL(trace_event_buffer_commit
);
2845 * trace_buffer_unlock_commit_regs()
2846 * trace_event_buffer_commit()
2847 * trace_event_raw_event_xxx()
2849 # define STACK_SKIP 3
2851 void trace_buffer_unlock_commit_regs(struct trace_array
*tr
,
2852 struct trace_buffer
*buffer
,
2853 struct ring_buffer_event
*event
,
2854 unsigned long flags
, int pc
,
2855 struct pt_regs
*regs
)
2857 __buffer_unlock_commit(buffer
, event
);
2860 * If regs is not set, then skip the necessary functions.
2861 * Note, we can still get here via blktrace, wakeup tracer
2862 * and mmiotrace, but that's ok if they lose a function or
2863 * two. They are not that meaningful.
2865 ftrace_trace_stack(tr
, buffer
, flags
, regs
? 0 : STACK_SKIP
, pc
, regs
);
2866 ftrace_trace_userstack(buffer
, flags
, pc
);
2870 * Similar to trace_buffer_unlock_commit_regs() but do not dump stack.
2873 trace_buffer_unlock_commit_nostack(struct trace_buffer
*buffer
,
2874 struct ring_buffer_event
*event
)
2876 __buffer_unlock_commit(buffer
, event
);
2880 trace_function(struct trace_array
*tr
,
2881 unsigned long ip
, unsigned long parent_ip
, unsigned long flags
,
2884 struct trace_event_call
*call
= &event_function
;
2885 struct trace_buffer
*buffer
= tr
->array_buffer
.buffer
;
2886 struct ring_buffer_event
*event
;
2887 struct ftrace_entry
*entry
;
2889 event
= __trace_buffer_lock_reserve(buffer
, TRACE_FN
, sizeof(*entry
),
2893 entry
= ring_buffer_event_data(event
);
2895 entry
->parent_ip
= parent_ip
;
2897 if (!call_filter_check_discard(call
, entry
, buffer
, event
)) {
2898 if (static_branch_unlikely(&trace_function_exports_enabled
))
2899 ftrace_exports(event
, TRACE_EXPORT_FUNCTION
);
2900 __buffer_unlock_commit(buffer
, event
);
2904 #ifdef CONFIG_STACKTRACE
2906 /* Allow 4 levels of nesting: normal, softirq, irq, NMI */
2907 #define FTRACE_KSTACK_NESTING 4
2909 #define FTRACE_KSTACK_ENTRIES (PAGE_SIZE / FTRACE_KSTACK_NESTING)
2911 struct ftrace_stack
{
2912 unsigned long calls
[FTRACE_KSTACK_ENTRIES
];
2916 struct ftrace_stacks
{
2917 struct ftrace_stack stacks
[FTRACE_KSTACK_NESTING
];
2920 static DEFINE_PER_CPU(struct ftrace_stacks
, ftrace_stacks
);
2921 static DEFINE_PER_CPU(int, ftrace_stack_reserve
);
2923 static void __ftrace_trace_stack(struct trace_buffer
*buffer
,
2924 unsigned long flags
,
2925 int skip
, int pc
, struct pt_regs
*regs
)
2927 struct trace_event_call
*call
= &event_kernel_stack
;
2928 struct ring_buffer_event
*event
;
2929 unsigned int size
, nr_entries
;
2930 struct ftrace_stack
*fstack
;
2931 struct stack_entry
*entry
;
2935 * Add one, for this function and the call to save_stack_trace()
2936 * If regs is set, then these functions will not be in the way.
2938 #ifndef CONFIG_UNWINDER_ORC
2943 preempt_disable_notrace();
2945 stackidx
= __this_cpu_inc_return(ftrace_stack_reserve
) - 1;
2947 /* This should never happen. If it does, yell once and skip */
2948 if (WARN_ON_ONCE(stackidx
> FTRACE_KSTACK_NESTING
))
2952 * The above __this_cpu_inc_return() is 'atomic' cpu local. An
2953 * interrupt will either see the value pre increment or post
2954 * increment. If the interrupt happens pre increment it will have
2955 * restored the counter when it returns. We just need a barrier to
2956 * keep gcc from moving things around.
2960 fstack
= this_cpu_ptr(ftrace_stacks
.stacks
) + stackidx
;
2961 size
= ARRAY_SIZE(fstack
->calls
);
2964 nr_entries
= stack_trace_save_regs(regs
, fstack
->calls
,
2967 nr_entries
= stack_trace_save(fstack
->calls
, size
, skip
);
2970 size
= nr_entries
* sizeof(unsigned long);
2971 event
= __trace_buffer_lock_reserve(buffer
, TRACE_STACK
,
2972 sizeof(*entry
) + size
, flags
, pc
);
2975 entry
= ring_buffer_event_data(event
);
2977 memcpy(&entry
->caller
, fstack
->calls
, size
);
2978 entry
->size
= nr_entries
;
2980 if (!call_filter_check_discard(call
, entry
, buffer
, event
))
2981 __buffer_unlock_commit(buffer
, event
);
2984 /* Again, don't let gcc optimize things here */
2986 __this_cpu_dec(ftrace_stack_reserve
);
2987 preempt_enable_notrace();
2991 static inline void ftrace_trace_stack(struct trace_array
*tr
,
2992 struct trace_buffer
*buffer
,
2993 unsigned long flags
,
2994 int skip
, int pc
, struct pt_regs
*regs
)
2996 if (!(tr
->trace_flags
& TRACE_ITER_STACKTRACE
))
2999 __ftrace_trace_stack(buffer
, flags
, skip
, pc
, regs
);
3002 void __trace_stack(struct trace_array
*tr
, unsigned long flags
, int skip
,
3005 struct trace_buffer
*buffer
= tr
->array_buffer
.buffer
;
3007 if (rcu_is_watching()) {
3008 __ftrace_trace_stack(buffer
, flags
, skip
, pc
, NULL
);
3013 * When an NMI triggers, RCU is enabled via rcu_nmi_enter(),
3014 * but if the above rcu_is_watching() failed, then the NMI
3015 * triggered someplace critical, and rcu_irq_enter() should
3016 * not be called from NMI.
3018 if (unlikely(in_nmi()))
3021 rcu_irq_enter_irqson();
3022 __ftrace_trace_stack(buffer
, flags
, skip
, pc
, NULL
);
3023 rcu_irq_exit_irqson();
3027 * trace_dump_stack - record a stack back trace in the trace buffer
3028 * @skip: Number of functions to skip (helper handlers)
3030 void trace_dump_stack(int skip
)
3032 unsigned long flags
;
3034 if (tracing_disabled
|| tracing_selftest_running
)
3037 local_save_flags(flags
);
3039 #ifndef CONFIG_UNWINDER_ORC
3040 /* Skip 1 to skip this function. */
3043 __ftrace_trace_stack(global_trace
.array_buffer
.buffer
,
3044 flags
, skip
, preempt_count(), NULL
);
3046 EXPORT_SYMBOL_GPL(trace_dump_stack
);
3048 #ifdef CONFIG_USER_STACKTRACE_SUPPORT
3049 static DEFINE_PER_CPU(int, user_stack_count
);
3052 ftrace_trace_userstack(struct trace_buffer
*buffer
, unsigned long flags
, int pc
)
3054 struct trace_event_call
*call
= &event_user_stack
;
3055 struct ring_buffer_event
*event
;
3056 struct userstack_entry
*entry
;
3058 if (!(global_trace
.trace_flags
& TRACE_ITER_USERSTACKTRACE
))
3062 * NMIs can not handle page faults, even with fix ups.
3063 * The save user stack can (and often does) fault.
3065 if (unlikely(in_nmi()))
3069 * prevent recursion, since the user stack tracing may
3070 * trigger other kernel events.
3073 if (__this_cpu_read(user_stack_count
))
3076 __this_cpu_inc(user_stack_count
);
3078 event
= __trace_buffer_lock_reserve(buffer
, TRACE_USER_STACK
,
3079 sizeof(*entry
), flags
, pc
);
3081 goto out_drop_count
;
3082 entry
= ring_buffer_event_data(event
);
3084 entry
->tgid
= current
->tgid
;
3085 memset(&entry
->caller
, 0, sizeof(entry
->caller
));
3087 stack_trace_save_user(entry
->caller
, FTRACE_STACK_ENTRIES
);
3088 if (!call_filter_check_discard(call
, entry
, buffer
, event
))
3089 __buffer_unlock_commit(buffer
, event
);
3092 __this_cpu_dec(user_stack_count
);
3096 #else /* CONFIG_USER_STACKTRACE_SUPPORT */
3097 static void ftrace_trace_userstack(struct trace_buffer
*buffer
,
3098 unsigned long flags
, int pc
)
3101 #endif /* !CONFIG_USER_STACKTRACE_SUPPORT */
3103 #endif /* CONFIG_STACKTRACE */
3105 /* created for use with alloc_percpu */
3106 struct trace_buffer_struct
{
3108 char buffer
[4][TRACE_BUF_SIZE
];
3111 static struct trace_buffer_struct
*trace_percpu_buffer
;
3114 * Thise allows for lockless recording. If we're nested too deeply, then
3115 * this returns NULL.
3117 static char *get_trace_buf(void)
3119 struct trace_buffer_struct
*buffer
= this_cpu_ptr(trace_percpu_buffer
);
3121 if (!buffer
|| buffer
->nesting
>= 4)
3126 /* Interrupts must see nesting incremented before we use the buffer */
3128 return &buffer
->buffer
[buffer
->nesting
][0];
3131 static void put_trace_buf(void)
3133 /* Don't let the decrement of nesting leak before this */
3135 this_cpu_dec(trace_percpu_buffer
->nesting
);
3138 static int alloc_percpu_trace_buffer(void)
3140 struct trace_buffer_struct
*buffers
;
3142 if (trace_percpu_buffer
)
3145 buffers
= alloc_percpu(struct trace_buffer_struct
);
3146 if (MEM_FAIL(!buffers
, "Could not allocate percpu trace_printk buffer"))
3149 trace_percpu_buffer
= buffers
;
3153 static int buffers_allocated
;
3155 void trace_printk_init_buffers(void)
3157 if (buffers_allocated
)
3160 if (alloc_percpu_trace_buffer())
3163 /* trace_printk() is for debug use only. Don't use it in production. */
3166 pr_warn("**********************************************************\n");
3167 pr_warn("** NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE **\n");
3169 pr_warn("** trace_printk() being used. Allocating extra memory. **\n");
3171 pr_warn("** This means that this is a DEBUG kernel and it is **\n");
3172 pr_warn("** unsafe for production use. **\n");
3174 pr_warn("** If you see this message and you are not debugging **\n");
3175 pr_warn("** the kernel, report this immediately to your vendor! **\n");
3177 pr_warn("** NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE NOTICE **\n");
3178 pr_warn("**********************************************************\n");
3180 /* Expand the buffers to set size */
3181 tracing_update_buffers();
3183 buffers_allocated
= 1;
3186 * trace_printk_init_buffers() can be called by modules.
3187 * If that happens, then we need to start cmdline recording
3188 * directly here. If the global_trace.buffer is already
3189 * allocated here, then this was called by module code.
3191 if (global_trace
.array_buffer
.buffer
)
3192 tracing_start_cmdline_record();
3194 EXPORT_SYMBOL_GPL(trace_printk_init_buffers
);
3196 void trace_printk_start_comm(void)
3198 /* Start tracing comms if trace printk is set */
3199 if (!buffers_allocated
)
3201 tracing_start_cmdline_record();
3204 static void trace_printk_start_stop_comm(int enabled
)
3206 if (!buffers_allocated
)
3210 tracing_start_cmdline_record();
3212 tracing_stop_cmdline_record();
3216 * trace_vbprintk - write binary msg to tracing buffer
3217 * @ip: The address of the caller
3218 * @fmt: The string format to write to the buffer
3219 * @args: Arguments for @fmt
3221 int trace_vbprintk(unsigned long ip
, const char *fmt
, va_list args
)
3223 struct trace_event_call
*call
= &event_bprint
;
3224 struct ring_buffer_event
*event
;
3225 struct trace_buffer
*buffer
;
3226 struct trace_array
*tr
= &global_trace
;
3227 struct bprint_entry
*entry
;
3228 unsigned long flags
;
3230 int len
= 0, size
, pc
;
3232 if (unlikely(tracing_selftest_running
|| tracing_disabled
))
3235 /* Don't pollute graph traces with trace_vprintk internals */
3236 pause_graph_tracing();
3238 pc
= preempt_count();
3239 preempt_disable_notrace();
3241 tbuffer
= get_trace_buf();
3247 len
= vbin_printf((u32
*)tbuffer
, TRACE_BUF_SIZE
/sizeof(int), fmt
, args
);
3249 if (len
> TRACE_BUF_SIZE
/sizeof(int) || len
< 0)
3252 local_save_flags(flags
);
3253 size
= sizeof(*entry
) + sizeof(u32
) * len
;
3254 buffer
= tr
->array_buffer
.buffer
;
3255 ring_buffer_nest_start(buffer
);
3256 event
= __trace_buffer_lock_reserve(buffer
, TRACE_BPRINT
, size
,
3260 entry
= ring_buffer_event_data(event
);
3264 memcpy(entry
->buf
, tbuffer
, sizeof(u32
) * len
);
3265 if (!call_filter_check_discard(call
, entry
, buffer
, event
)) {
3266 __buffer_unlock_commit(buffer
, event
);
3267 ftrace_trace_stack(tr
, buffer
, flags
, 6, pc
, NULL
);
3271 ring_buffer_nest_end(buffer
);
3276 preempt_enable_notrace();
3277 unpause_graph_tracing();
3281 EXPORT_SYMBOL_GPL(trace_vbprintk
);
3285 __trace_array_vprintk(struct trace_buffer
*buffer
,
3286 unsigned long ip
, const char *fmt
, va_list args
)
3288 struct trace_event_call
*call
= &event_print
;
3289 struct ring_buffer_event
*event
;
3290 int len
= 0, size
, pc
;
3291 struct print_entry
*entry
;
3292 unsigned long flags
;
3295 if (tracing_disabled
|| tracing_selftest_running
)
3298 /* Don't pollute graph traces with trace_vprintk internals */
3299 pause_graph_tracing();
3301 pc
= preempt_count();
3302 preempt_disable_notrace();
3305 tbuffer
= get_trace_buf();
3311 len
= vscnprintf(tbuffer
, TRACE_BUF_SIZE
, fmt
, args
);
3313 local_save_flags(flags
);
3314 size
= sizeof(*entry
) + len
+ 1;
3315 ring_buffer_nest_start(buffer
);
3316 event
= __trace_buffer_lock_reserve(buffer
, TRACE_PRINT
, size
,
3320 entry
= ring_buffer_event_data(event
);
3323 memcpy(&entry
->buf
, tbuffer
, len
+ 1);
3324 if (!call_filter_check_discard(call
, entry
, buffer
, event
)) {
3325 __buffer_unlock_commit(buffer
, event
);
3326 ftrace_trace_stack(&global_trace
, buffer
, flags
, 6, pc
, NULL
);
3330 ring_buffer_nest_end(buffer
);
3334 preempt_enable_notrace();
3335 unpause_graph_tracing();
3341 int trace_array_vprintk(struct trace_array
*tr
,
3342 unsigned long ip
, const char *fmt
, va_list args
)
3344 return __trace_array_vprintk(tr
->array_buffer
.buffer
, ip
, fmt
, args
);
3348 * trace_array_printk - Print a message to a specific instance
3349 * @tr: The instance trace_array descriptor
3350 * @ip: The instruction pointer that this is called from.
3351 * @fmt: The format to print (printf format)
3353 * If a subsystem sets up its own instance, they have the right to
3354 * printk strings into their tracing instance buffer using this
3355 * function. Note, this function will not write into the top level
3356 * buffer (use trace_printk() for that), as writing into the top level
3357 * buffer should only have events that can be individually disabled.
3358 * trace_printk() is only used for debugging a kernel, and should not
3359 * be ever encorporated in normal use.
3361 * trace_array_printk() can be used, as it will not add noise to the
3362 * top level tracing buffer.
3364 * Note, trace_array_init_printk() must be called on @tr before this
3368 int trace_array_printk(struct trace_array
*tr
,
3369 unsigned long ip
, const char *fmt
, ...)
3377 /* This is only allowed for created instances */
3378 if (tr
== &global_trace
)
3381 if (!(tr
->trace_flags
& TRACE_ITER_PRINTK
))
3385 ret
= trace_array_vprintk(tr
, ip
, fmt
, ap
);
3389 EXPORT_SYMBOL_GPL(trace_array_printk
);
3392 * trace_array_init_printk - Initialize buffers for trace_array_printk()
3393 * @tr: The trace array to initialize the buffers for
3395 * As trace_array_printk() only writes into instances, they are OK to
3396 * have in the kernel (unlike trace_printk()). This needs to be called
3397 * before trace_array_printk() can be used on a trace_array.
3399 int trace_array_init_printk(struct trace_array
*tr
)
3404 /* This is only allowed for created instances */
3405 if (tr
== &global_trace
)
3408 return alloc_percpu_trace_buffer();
3410 EXPORT_SYMBOL_GPL(trace_array_init_printk
);
3413 int trace_array_printk_buf(struct trace_buffer
*buffer
,
3414 unsigned long ip
, const char *fmt
, ...)
3419 if (!(global_trace
.trace_flags
& TRACE_ITER_PRINTK
))
3423 ret
= __trace_array_vprintk(buffer
, ip
, fmt
, ap
);
3429 int trace_vprintk(unsigned long ip
, const char *fmt
, va_list args
)
3431 return trace_array_vprintk(&global_trace
, ip
, fmt
, args
);
3433 EXPORT_SYMBOL_GPL(trace_vprintk
);
3435 static void trace_iterator_increment(struct trace_iterator
*iter
)
3437 struct ring_buffer_iter
*buf_iter
= trace_buffer_iter(iter
, iter
->cpu
);
3441 ring_buffer_iter_advance(buf_iter
);
3444 static struct trace_entry
*
3445 peek_next_entry(struct trace_iterator
*iter
, int cpu
, u64
*ts
,
3446 unsigned long *lost_events
)
3448 struct ring_buffer_event
*event
;
3449 struct ring_buffer_iter
*buf_iter
= trace_buffer_iter(iter
, cpu
);
3452 event
= ring_buffer_iter_peek(buf_iter
, ts
);
3454 *lost_events
= ring_buffer_iter_dropped(buf_iter
) ?
3455 (unsigned long)-1 : 0;
3457 event
= ring_buffer_peek(iter
->array_buffer
->buffer
, cpu
, ts
,
3462 iter
->ent_size
= ring_buffer_event_length(event
);
3463 return ring_buffer_event_data(event
);
3469 static struct trace_entry
*
3470 __find_next_entry(struct trace_iterator
*iter
, int *ent_cpu
,
3471 unsigned long *missing_events
, u64
*ent_ts
)
3473 struct trace_buffer
*buffer
= iter
->array_buffer
->buffer
;
3474 struct trace_entry
*ent
, *next
= NULL
;
3475 unsigned long lost_events
= 0, next_lost
= 0;
3476 int cpu_file
= iter
->cpu_file
;
3477 u64 next_ts
= 0, ts
;
3483 * If we are in a per_cpu trace file, don't bother by iterating over
3484 * all cpu and peek directly.
3486 if (cpu_file
> RING_BUFFER_ALL_CPUS
) {
3487 if (ring_buffer_empty_cpu(buffer
, cpu_file
))
3489 ent
= peek_next_entry(iter
, cpu_file
, ent_ts
, missing_events
);
3491 *ent_cpu
= cpu_file
;
3496 for_each_tracing_cpu(cpu
) {
3498 if (ring_buffer_empty_cpu(buffer
, cpu
))
3501 ent
= peek_next_entry(iter
, cpu
, &ts
, &lost_events
);
3504 * Pick the entry with the smallest timestamp:
3506 if (ent
&& (!next
|| ts
< next_ts
)) {
3510 next_lost
= lost_events
;
3511 next_size
= iter
->ent_size
;
3515 iter
->ent_size
= next_size
;
3518 *ent_cpu
= next_cpu
;
3524 *missing_events
= next_lost
;
3529 #define STATIC_TEMP_BUF_SIZE 128
3530 static char static_temp_buf
[STATIC_TEMP_BUF_SIZE
];
3532 /* Find the next real entry, without updating the iterator itself */
3533 struct trace_entry
*trace_find_next_entry(struct trace_iterator
*iter
,
3534 int *ent_cpu
, u64
*ent_ts
)
3536 /* __find_next_entry will reset ent_size */
3537 int ent_size
= iter
->ent_size
;
3538 struct trace_entry
*entry
;
3541 * If called from ftrace_dump(), then the iter->temp buffer
3542 * will be the static_temp_buf and not created from kmalloc.
3543 * If the entry size is greater than the buffer, we can
3544 * not save it. Just return NULL in that case. This is only
3545 * used to add markers when two consecutive events' time
3546 * stamps have a large delta. See trace_print_lat_context()
3548 if (iter
->temp
== static_temp_buf
&&
3549 STATIC_TEMP_BUF_SIZE
< ent_size
)
3553 * The __find_next_entry() may call peek_next_entry(), which may
3554 * call ring_buffer_peek() that may make the contents of iter->ent
3555 * undefined. Need to copy iter->ent now.
3557 if (iter
->ent
&& iter
->ent
!= iter
->temp
) {
3558 if ((!iter
->temp
|| iter
->temp_size
< iter
->ent_size
) &&
3559 !WARN_ON_ONCE(iter
->temp
== static_temp_buf
)) {
3561 iter
->temp
= kmalloc(iter
->ent_size
, GFP_KERNEL
);
3565 memcpy(iter
->temp
, iter
->ent
, iter
->ent_size
);
3566 iter
->temp_size
= iter
->ent_size
;
3567 iter
->ent
= iter
->temp
;
3569 entry
= __find_next_entry(iter
, ent_cpu
, NULL
, ent_ts
);
3570 /* Put back the original ent_size */
3571 iter
->ent_size
= ent_size
;
3576 /* Find the next real entry, and increment the iterator to the next entry */
3577 void *trace_find_next_entry_inc(struct trace_iterator
*iter
)
3579 iter
->ent
= __find_next_entry(iter
, &iter
->cpu
,
3580 &iter
->lost_events
, &iter
->ts
);
3583 trace_iterator_increment(iter
);
3585 return iter
->ent
? iter
: NULL
;
3588 static void trace_consume(struct trace_iterator
*iter
)
3590 ring_buffer_consume(iter
->array_buffer
->buffer
, iter
->cpu
, &iter
->ts
,
3591 &iter
->lost_events
);
3594 static void *s_next(struct seq_file
*m
, void *v
, loff_t
*pos
)
3596 struct trace_iterator
*iter
= m
->private;
3600 WARN_ON_ONCE(iter
->leftover
);
3604 /* can't go backwards */
3609 ent
= trace_find_next_entry_inc(iter
);
3613 while (ent
&& iter
->idx
< i
)
3614 ent
= trace_find_next_entry_inc(iter
);
3621 void tracing_iter_reset(struct trace_iterator
*iter
, int cpu
)
3623 struct ring_buffer_iter
*buf_iter
;
3624 unsigned long entries
= 0;
3627 per_cpu_ptr(iter
->array_buffer
->data
, cpu
)->skipped_entries
= 0;
3629 buf_iter
= trace_buffer_iter(iter
, cpu
);
3633 ring_buffer_iter_reset(buf_iter
);
3636 * We could have the case with the max latency tracers
3637 * that a reset never took place on a cpu. This is evident
3638 * by the timestamp being before the start of the buffer.
3640 while (ring_buffer_iter_peek(buf_iter
, &ts
)) {
3641 if (ts
>= iter
->array_buffer
->time_start
)
3644 ring_buffer_iter_advance(buf_iter
);
3647 per_cpu_ptr(iter
->array_buffer
->data
, cpu
)->skipped_entries
= entries
;
3651 * The current tracer is copied to avoid a global locking
3654 static void *s_start(struct seq_file
*m
, loff_t
*pos
)
3656 struct trace_iterator
*iter
= m
->private;
3657 struct trace_array
*tr
= iter
->tr
;
3658 int cpu_file
= iter
->cpu_file
;
3664 * copy the tracer to avoid using a global lock all around.
3665 * iter->trace is a copy of current_trace, the pointer to the
3666 * name may be used instead of a strcmp(), as iter->trace->name
3667 * will point to the same string as current_trace->name.
3669 mutex_lock(&trace_types_lock
);
3670 if (unlikely(tr
->current_trace
&& iter
->trace
->name
!= tr
->current_trace
->name
))
3671 *iter
->trace
= *tr
->current_trace
;
3672 mutex_unlock(&trace_types_lock
);
3674 #ifdef CONFIG_TRACER_MAX_TRACE
3675 if (iter
->snapshot
&& iter
->trace
->use_max_tr
)
3676 return ERR_PTR(-EBUSY
);
3679 if (!iter
->snapshot
)
3680 atomic_inc(&trace_record_taskinfo_disabled
);
3682 if (*pos
!= iter
->pos
) {
3687 if (cpu_file
== RING_BUFFER_ALL_CPUS
) {
3688 for_each_tracing_cpu(cpu
)
3689 tracing_iter_reset(iter
, cpu
);
3691 tracing_iter_reset(iter
, cpu_file
);
3694 for (p
= iter
; p
&& l
< *pos
; p
= s_next(m
, p
, &l
))
3699 * If we overflowed the seq_file before, then we want
3700 * to just reuse the trace_seq buffer again.
3706 p
= s_next(m
, p
, &l
);
3710 trace_event_read_lock();
3711 trace_access_lock(cpu_file
);
3715 static void s_stop(struct seq_file
*m
, void *p
)
3717 struct trace_iterator
*iter
= m
->private;
3719 #ifdef CONFIG_TRACER_MAX_TRACE
3720 if (iter
->snapshot
&& iter
->trace
->use_max_tr
)
3724 if (!iter
->snapshot
)
3725 atomic_dec(&trace_record_taskinfo_disabled
);
3727 trace_access_unlock(iter
->cpu_file
);
3728 trace_event_read_unlock();
3732 get_total_entries_cpu(struct array_buffer
*buf
, unsigned long *total
,
3733 unsigned long *entries
, int cpu
)
3735 unsigned long count
;
3737 count
= ring_buffer_entries_cpu(buf
->buffer
, cpu
);
3739 * If this buffer has skipped entries, then we hold all
3740 * entries for the trace and we need to ignore the
3741 * ones before the time stamp.
3743 if (per_cpu_ptr(buf
->data
, cpu
)->skipped_entries
) {
3744 count
-= per_cpu_ptr(buf
->data
, cpu
)->skipped_entries
;
3745 /* total is the same as the entries */
3749 ring_buffer_overrun_cpu(buf
->buffer
, cpu
);
3754 get_total_entries(struct array_buffer
*buf
,
3755 unsigned long *total
, unsigned long *entries
)
3763 for_each_tracing_cpu(cpu
) {
3764 get_total_entries_cpu(buf
, &t
, &e
, cpu
);
3770 unsigned long trace_total_entries_cpu(struct trace_array
*tr
, int cpu
)
3772 unsigned long total
, entries
;
3777 get_total_entries_cpu(&tr
->array_buffer
, &total
, &entries
, cpu
);
3782 unsigned long trace_total_entries(struct trace_array
*tr
)
3784 unsigned long total
, entries
;
3789 get_total_entries(&tr
->array_buffer
, &total
, &entries
);
3794 static void print_lat_help_header(struct seq_file
*m
)
3796 seq_puts(m
, "# _------=> CPU# \n"
3797 "# / _-----=> irqs-off \n"
3798 "# | / _----=> need-resched \n"
3799 "# || / _---=> hardirq/softirq \n"
3800 "# ||| / _--=> preempt-depth \n"
3802 "# cmd pid ||||| time | caller \n"
3803 "# \\ / ||||| \\ | / \n");
3806 static void print_event_info(struct array_buffer
*buf
, struct seq_file
*m
)
3808 unsigned long total
;
3809 unsigned long entries
;
3811 get_total_entries(buf
, &total
, &entries
);
3812 seq_printf(m
, "# entries-in-buffer/entries-written: %lu/%lu #P:%d\n",
3813 entries
, total
, num_online_cpus());
3817 static void print_func_help_header(struct array_buffer
*buf
, struct seq_file
*m
,
3820 bool tgid
= flags
& TRACE_ITER_RECORD_TGID
;
3822 print_event_info(buf
, m
);
3824 seq_printf(m
, "# TASK-PID %s CPU# TIMESTAMP FUNCTION\n", tgid
? "TGID " : "");
3825 seq_printf(m
, "# | | %s | | |\n", tgid
? " | " : "");
3828 static void print_func_help_header_irq(struct array_buffer
*buf
, struct seq_file
*m
,
3831 bool tgid
= flags
& TRACE_ITER_RECORD_TGID
;
3832 const char *space
= " ";
3833 int prec
= tgid
? 10 : 2;
3835 print_event_info(buf
, m
);
3837 seq_printf(m
, "# %.*s _-----=> irqs-off\n", prec
, space
);
3838 seq_printf(m
, "# %.*s / _----=> need-resched\n", prec
, space
);
3839 seq_printf(m
, "# %.*s| / _---=> hardirq/softirq\n", prec
, space
);
3840 seq_printf(m
, "# %.*s|| / _--=> preempt-depth\n", prec
, space
);
3841 seq_printf(m
, "# %.*s||| / delay\n", prec
, space
);
3842 seq_printf(m
, "# TASK-PID %.*sCPU# |||| TIMESTAMP FUNCTION\n", prec
, " TGID ");
3843 seq_printf(m
, "# | | %.*s | |||| | |\n", prec
, " | ");
3847 print_trace_header(struct seq_file
*m
, struct trace_iterator
*iter
)
3849 unsigned long sym_flags
= (global_trace
.trace_flags
& TRACE_ITER_SYM_MASK
);
3850 struct array_buffer
*buf
= iter
->array_buffer
;
3851 struct trace_array_cpu
*data
= per_cpu_ptr(buf
->data
, buf
->cpu
);
3852 struct tracer
*type
= iter
->trace
;
3853 unsigned long entries
;
3854 unsigned long total
;
3855 const char *name
= "preemption";
3859 get_total_entries(buf
, &total
, &entries
);
3861 seq_printf(m
, "# %s latency trace v1.1.5 on %s\n",
3863 seq_puts(m
, "# -----------------------------------"
3864 "---------------------------------\n");
3865 seq_printf(m
, "# latency: %lu us, #%lu/%lu, CPU#%d |"
3866 " (M:%s VP:%d, KP:%d, SP:%d HP:%d",
3867 nsecs_to_usecs(data
->saved_latency
),
3871 #if defined(CONFIG_PREEMPT_NONE)
3873 #elif defined(CONFIG_PREEMPT_VOLUNTARY)
3875 #elif defined(CONFIG_PREEMPT)
3877 #elif defined(CONFIG_PREEMPT_RT)
3882 /* These are reserved for later use */
3885 seq_printf(m
, " #P:%d)\n", num_online_cpus());
3889 seq_puts(m
, "# -----------------\n");
3890 seq_printf(m
, "# | task: %.16s-%d "
3891 "(uid:%d nice:%ld policy:%ld rt_prio:%ld)\n",
3892 data
->comm
, data
->pid
,
3893 from_kuid_munged(seq_user_ns(m
), data
->uid
), data
->nice
,
3894 data
->policy
, data
->rt_priority
);
3895 seq_puts(m
, "# -----------------\n");
3897 if (data
->critical_start
) {
3898 seq_puts(m
, "# => started at: ");
3899 seq_print_ip_sym(&iter
->seq
, data
->critical_start
, sym_flags
);
3900 trace_print_seq(m
, &iter
->seq
);
3901 seq_puts(m
, "\n# => ended at: ");
3902 seq_print_ip_sym(&iter
->seq
, data
->critical_end
, sym_flags
);
3903 trace_print_seq(m
, &iter
->seq
);
3904 seq_puts(m
, "\n#\n");
3910 static void test_cpu_buff_start(struct trace_iterator
*iter
)
3912 struct trace_seq
*s
= &iter
->seq
;
3913 struct trace_array
*tr
= iter
->tr
;
3915 if (!(tr
->trace_flags
& TRACE_ITER_ANNOTATE
))
3918 if (!(iter
->iter_flags
& TRACE_FILE_ANNOTATE
))
3921 if (cpumask_available(iter
->started
) &&
3922 cpumask_test_cpu(iter
->cpu
, iter
->started
))
3925 if (per_cpu_ptr(iter
->array_buffer
->data
, iter
->cpu
)->skipped_entries
)
3928 if (cpumask_available(iter
->started
))
3929 cpumask_set_cpu(iter
->cpu
, iter
->started
);
3931 /* Don't print started cpu buffer for the first entry of the trace */
3933 trace_seq_printf(s
, "##### CPU %u buffer started ####\n",
3937 static enum print_line_t
print_trace_fmt(struct trace_iterator
*iter
)
3939 struct trace_array
*tr
= iter
->tr
;
3940 struct trace_seq
*s
= &iter
->seq
;
3941 unsigned long sym_flags
= (tr
->trace_flags
& TRACE_ITER_SYM_MASK
);
3942 struct trace_entry
*entry
;
3943 struct trace_event
*event
;
3947 test_cpu_buff_start(iter
);
3949 event
= ftrace_find_event(entry
->type
);
3951 if (tr
->trace_flags
& TRACE_ITER_CONTEXT_INFO
) {
3952 if (iter
->iter_flags
& TRACE_FILE_LAT_FMT
)
3953 trace_print_lat_context(iter
);
3955 trace_print_context(iter
);
3958 if (trace_seq_has_overflowed(s
))
3959 return TRACE_TYPE_PARTIAL_LINE
;
3962 return event
->funcs
->trace(iter
, sym_flags
, event
);
3964 trace_seq_printf(s
, "Unknown type %d\n", entry
->type
);
3966 return trace_handle_return(s
);
3969 static enum print_line_t
print_raw_fmt(struct trace_iterator
*iter
)
3971 struct trace_array
*tr
= iter
->tr
;
3972 struct trace_seq
*s
= &iter
->seq
;
3973 struct trace_entry
*entry
;
3974 struct trace_event
*event
;
3978 if (tr
->trace_flags
& TRACE_ITER_CONTEXT_INFO
)
3979 trace_seq_printf(s
, "%d %d %llu ",
3980 entry
->pid
, iter
->cpu
, iter
->ts
);
3982 if (trace_seq_has_overflowed(s
))
3983 return TRACE_TYPE_PARTIAL_LINE
;
3985 event
= ftrace_find_event(entry
->type
);
3987 return event
->funcs
->raw(iter
, 0, event
);
3989 trace_seq_printf(s
, "%d ?\n", entry
->type
);
3991 return trace_handle_return(s
);
3994 static enum print_line_t
print_hex_fmt(struct trace_iterator
*iter
)
3996 struct trace_array
*tr
= iter
->tr
;
3997 struct trace_seq
*s
= &iter
->seq
;
3998 unsigned char newline
= '\n';
3999 struct trace_entry
*entry
;
4000 struct trace_event
*event
;
4004 if (tr
->trace_flags
& TRACE_ITER_CONTEXT_INFO
) {
4005 SEQ_PUT_HEX_FIELD(s
, entry
->pid
);
4006 SEQ_PUT_HEX_FIELD(s
, iter
->cpu
);
4007 SEQ_PUT_HEX_FIELD(s
, iter
->ts
);
4008 if (trace_seq_has_overflowed(s
))
4009 return TRACE_TYPE_PARTIAL_LINE
;
4012 event
= ftrace_find_event(entry
->type
);
4014 enum print_line_t ret
= event
->funcs
->hex(iter
, 0, event
);
4015 if (ret
!= TRACE_TYPE_HANDLED
)
4019 SEQ_PUT_FIELD(s
, newline
);
4021 return trace_handle_return(s
);
4024 static enum print_line_t
print_bin_fmt(struct trace_iterator
*iter
)
4026 struct trace_array
*tr
= iter
->tr
;
4027 struct trace_seq
*s
= &iter
->seq
;
4028 struct trace_entry
*entry
;
4029 struct trace_event
*event
;
4033 if (tr
->trace_flags
& TRACE_ITER_CONTEXT_INFO
) {
4034 SEQ_PUT_FIELD(s
, entry
->pid
);
4035 SEQ_PUT_FIELD(s
, iter
->cpu
);
4036 SEQ_PUT_FIELD(s
, iter
->ts
);
4037 if (trace_seq_has_overflowed(s
))
4038 return TRACE_TYPE_PARTIAL_LINE
;
4041 event
= ftrace_find_event(entry
->type
);
4042 return event
? event
->funcs
->binary(iter
, 0, event
) :
4046 int trace_empty(struct trace_iterator
*iter
)
4048 struct ring_buffer_iter
*buf_iter
;
4051 /* If we are looking at one CPU buffer, only check that one */
4052 if (iter
->cpu_file
!= RING_BUFFER_ALL_CPUS
) {
4053 cpu
= iter
->cpu_file
;
4054 buf_iter
= trace_buffer_iter(iter
, cpu
);
4056 if (!ring_buffer_iter_empty(buf_iter
))
4059 if (!ring_buffer_empty_cpu(iter
->array_buffer
->buffer
, cpu
))
4065 for_each_tracing_cpu(cpu
) {
4066 buf_iter
= trace_buffer_iter(iter
, cpu
);
4068 if (!ring_buffer_iter_empty(buf_iter
))
4071 if (!ring_buffer_empty_cpu(iter
->array_buffer
->buffer
, cpu
))
4079 /* Called with trace_event_read_lock() held. */
4080 enum print_line_t
print_trace_line(struct trace_iterator
*iter
)
4082 struct trace_array
*tr
= iter
->tr
;
4083 unsigned long trace_flags
= tr
->trace_flags
;
4084 enum print_line_t ret
;
4086 if (iter
->lost_events
) {
4087 if (iter
->lost_events
== (unsigned long)-1)
4088 trace_seq_printf(&iter
->seq
, "CPU:%d [LOST EVENTS]\n",
4091 trace_seq_printf(&iter
->seq
, "CPU:%d [LOST %lu EVENTS]\n",
4092 iter
->cpu
, iter
->lost_events
);
4093 if (trace_seq_has_overflowed(&iter
->seq
))
4094 return TRACE_TYPE_PARTIAL_LINE
;
4097 if (iter
->trace
&& iter
->trace
->print_line
) {
4098 ret
= iter
->trace
->print_line(iter
);
4099 if (ret
!= TRACE_TYPE_UNHANDLED
)
4103 if (iter
->ent
->type
== TRACE_BPUTS
&&
4104 trace_flags
& TRACE_ITER_PRINTK
&&
4105 trace_flags
& TRACE_ITER_PRINTK_MSGONLY
)
4106 return trace_print_bputs_msg_only(iter
);
4108 if (iter
->ent
->type
== TRACE_BPRINT
&&
4109 trace_flags
& TRACE_ITER_PRINTK
&&
4110 trace_flags
& TRACE_ITER_PRINTK_MSGONLY
)
4111 return trace_print_bprintk_msg_only(iter
);
4113 if (iter
->ent
->type
== TRACE_PRINT
&&
4114 trace_flags
& TRACE_ITER_PRINTK
&&
4115 trace_flags
& TRACE_ITER_PRINTK_MSGONLY
)
4116 return trace_print_printk_msg_only(iter
);
4118 if (trace_flags
& TRACE_ITER_BIN
)
4119 return print_bin_fmt(iter
);
4121 if (trace_flags
& TRACE_ITER_HEX
)
4122 return print_hex_fmt(iter
);
4124 if (trace_flags
& TRACE_ITER_RAW
)
4125 return print_raw_fmt(iter
);
4127 return print_trace_fmt(iter
);
4130 void trace_latency_header(struct seq_file
*m
)
4132 struct trace_iterator
*iter
= m
->private;
4133 struct trace_array
*tr
= iter
->tr
;
4135 /* print nothing if the buffers are empty */
4136 if (trace_empty(iter
))
4139 if (iter
->iter_flags
& TRACE_FILE_LAT_FMT
)
4140 print_trace_header(m
, iter
);
4142 if (!(tr
->trace_flags
& TRACE_ITER_VERBOSE
))
4143 print_lat_help_header(m
);
4146 void trace_default_header(struct seq_file
*m
)
4148 struct trace_iterator
*iter
= m
->private;
4149 struct trace_array
*tr
= iter
->tr
;
4150 unsigned long trace_flags
= tr
->trace_flags
;
4152 if (!(trace_flags
& TRACE_ITER_CONTEXT_INFO
))
4155 if (iter
->iter_flags
& TRACE_FILE_LAT_FMT
) {
4156 /* print nothing if the buffers are empty */
4157 if (trace_empty(iter
))
4159 print_trace_header(m
, iter
);
4160 if (!(trace_flags
& TRACE_ITER_VERBOSE
))
4161 print_lat_help_header(m
);
4163 if (!(trace_flags
& TRACE_ITER_VERBOSE
)) {
4164 if (trace_flags
& TRACE_ITER_IRQ_INFO
)
4165 print_func_help_header_irq(iter
->array_buffer
,
4168 print_func_help_header(iter
->array_buffer
, m
,
4174 static void test_ftrace_alive(struct seq_file
*m
)
4176 if (!ftrace_is_dead())
4178 seq_puts(m
, "# WARNING: FUNCTION TRACING IS CORRUPTED\n"
4179 "# MAY BE MISSING FUNCTION EVENTS\n");
4182 #ifdef CONFIG_TRACER_MAX_TRACE
4183 static void show_snapshot_main_help(struct seq_file
*m
)
4185 seq_puts(m
, "# echo 0 > snapshot : Clears and frees snapshot buffer\n"
4186 "# echo 1 > snapshot : Allocates snapshot buffer, if not already allocated.\n"
4187 "# Takes a snapshot of the main buffer.\n"
4188 "# echo 2 > snapshot : Clears snapshot buffer (but does not allocate or free)\n"
4189 "# (Doesn't have to be '2' works with any number that\n"
4190 "# is not a '0' or '1')\n");
4193 static void show_snapshot_percpu_help(struct seq_file
*m
)
4195 seq_puts(m
, "# echo 0 > snapshot : Invalid for per_cpu snapshot file.\n");
4196 #ifdef CONFIG_RING_BUFFER_ALLOW_SWAP
4197 seq_puts(m
, "# echo 1 > snapshot : Allocates snapshot buffer, if not already allocated.\n"
4198 "# Takes a snapshot of the main buffer for this cpu.\n");
4200 seq_puts(m
, "# echo 1 > snapshot : Not supported with this kernel.\n"
4201 "# Must use main snapshot file to allocate.\n");
4203 seq_puts(m
, "# echo 2 > snapshot : Clears this cpu's snapshot buffer (but does not allocate)\n"
4204 "# (Doesn't have to be '2' works with any number that\n"
4205 "# is not a '0' or '1')\n");
4208 static void print_snapshot_help(struct seq_file
*m
, struct trace_iterator
*iter
)
4210 if (iter
->tr
->allocated_snapshot
)
4211 seq_puts(m
, "#\n# * Snapshot is allocated *\n#\n");
4213 seq_puts(m
, "#\n# * Snapshot is freed *\n#\n");
4215 seq_puts(m
, "# Snapshot commands:\n");
4216 if (iter
->cpu_file
== RING_BUFFER_ALL_CPUS
)
4217 show_snapshot_main_help(m
);
4219 show_snapshot_percpu_help(m
);
4222 /* Should never be called */
4223 static inline void print_snapshot_help(struct seq_file
*m
, struct trace_iterator
*iter
) { }
4226 static int s_show(struct seq_file
*m
, void *v
)
4228 struct trace_iterator
*iter
= v
;
4231 if (iter
->ent
== NULL
) {
4233 seq_printf(m
, "# tracer: %s\n", iter
->trace
->name
);
4235 test_ftrace_alive(m
);
4237 if (iter
->snapshot
&& trace_empty(iter
))
4238 print_snapshot_help(m
, iter
);
4239 else if (iter
->trace
&& iter
->trace
->print_header
)
4240 iter
->trace
->print_header(m
);
4242 trace_default_header(m
);
4244 } else if (iter
->leftover
) {
4246 * If we filled the seq_file buffer earlier, we
4247 * want to just show it now.
4249 ret
= trace_print_seq(m
, &iter
->seq
);
4251 /* ret should this time be zero, but you never know */
4252 iter
->leftover
= ret
;
4255 print_trace_line(iter
);
4256 ret
= trace_print_seq(m
, &iter
->seq
);
4258 * If we overflow the seq_file buffer, then it will
4259 * ask us for this data again at start up.
4261 * ret is 0 if seq_file write succeeded.
4264 iter
->leftover
= ret
;
4271 * Should be used after trace_array_get(), trace_types_lock
4272 * ensures that i_cdev was already initialized.
4274 static inline int tracing_get_cpu(struct inode
*inode
)
4276 if (inode
->i_cdev
) /* See trace_create_cpu_file() */
4277 return (long)inode
->i_cdev
- 1;
4278 return RING_BUFFER_ALL_CPUS
;
4281 static const struct seq_operations tracer_seq_ops
= {
4288 static struct trace_iterator
*
4289 __tracing_open(struct inode
*inode
, struct file
*file
, bool snapshot
)
4291 struct trace_array
*tr
= inode
->i_private
;
4292 struct trace_iterator
*iter
;
4295 if (tracing_disabled
)
4296 return ERR_PTR(-ENODEV
);
4298 iter
= __seq_open_private(file
, &tracer_seq_ops
, sizeof(*iter
));
4300 return ERR_PTR(-ENOMEM
);
4302 iter
->buffer_iter
= kcalloc(nr_cpu_ids
, sizeof(*iter
->buffer_iter
),
4304 if (!iter
->buffer_iter
)
4308 * trace_find_next_entry() may need to save off iter->ent.
4309 * It will place it into the iter->temp buffer. As most
4310 * events are less than 128, allocate a buffer of that size.
4311 * If one is greater, then trace_find_next_entry() will
4312 * allocate a new buffer to adjust for the bigger iter->ent.
4313 * It's not critical if it fails to get allocated here.
4315 iter
->temp
= kmalloc(128, GFP_KERNEL
);
4317 iter
->temp_size
= 128;
4320 * We make a copy of the current tracer to avoid concurrent
4321 * changes on it while we are reading.
4323 mutex_lock(&trace_types_lock
);
4324 iter
->trace
= kzalloc(sizeof(*iter
->trace
), GFP_KERNEL
);
4328 *iter
->trace
= *tr
->current_trace
;
4330 if (!zalloc_cpumask_var(&iter
->started
, GFP_KERNEL
))
4335 #ifdef CONFIG_TRACER_MAX_TRACE
4336 /* Currently only the top directory has a snapshot */
4337 if (tr
->current_trace
->print_max
|| snapshot
)
4338 iter
->array_buffer
= &tr
->max_buffer
;
4341 iter
->array_buffer
= &tr
->array_buffer
;
4342 iter
->snapshot
= snapshot
;
4344 iter
->cpu_file
= tracing_get_cpu(inode
);
4345 mutex_init(&iter
->mutex
);
4347 /* Notify the tracer early; before we stop tracing. */
4348 if (iter
->trace
->open
)
4349 iter
->trace
->open(iter
);
4351 /* Annotate start of buffers if we had overruns */
4352 if (ring_buffer_overruns(iter
->array_buffer
->buffer
))
4353 iter
->iter_flags
|= TRACE_FILE_ANNOTATE
;
4355 /* Output in nanoseconds only if we are using a clock in nanoseconds. */
4356 if (trace_clocks
[tr
->clock_id
].in_ns
)
4357 iter
->iter_flags
|= TRACE_FILE_TIME_IN_NS
;
4360 * If pause-on-trace is enabled, then stop the trace while
4361 * dumping, unless this is the "snapshot" file
4363 if (!iter
->snapshot
&& (tr
->trace_flags
& TRACE_ITER_PAUSE_ON_TRACE
))
4364 tracing_stop_tr(tr
);
4366 if (iter
->cpu_file
== RING_BUFFER_ALL_CPUS
) {
4367 for_each_tracing_cpu(cpu
) {
4368 iter
->buffer_iter
[cpu
] =
4369 ring_buffer_read_prepare(iter
->array_buffer
->buffer
,
4372 ring_buffer_read_prepare_sync();
4373 for_each_tracing_cpu(cpu
) {
4374 ring_buffer_read_start(iter
->buffer_iter
[cpu
]);
4375 tracing_iter_reset(iter
, cpu
);
4378 cpu
= iter
->cpu_file
;
4379 iter
->buffer_iter
[cpu
] =
4380 ring_buffer_read_prepare(iter
->array_buffer
->buffer
,
4382 ring_buffer_read_prepare_sync();
4383 ring_buffer_read_start(iter
->buffer_iter
[cpu
]);
4384 tracing_iter_reset(iter
, cpu
);
4387 mutex_unlock(&trace_types_lock
);
4392 mutex_unlock(&trace_types_lock
);
4395 kfree(iter
->buffer_iter
);
4397 seq_release_private(inode
, file
);
4398 return ERR_PTR(-ENOMEM
);
4401 int tracing_open_generic(struct inode
*inode
, struct file
*filp
)
4405 ret
= tracing_check_open_get_tr(NULL
);
4409 filp
->private_data
= inode
->i_private
;
4413 bool tracing_is_disabled(void)
4415 return (tracing_disabled
) ? true: false;
4419 * Open and update trace_array ref count.
4420 * Must have the current trace_array passed to it.
4422 int tracing_open_generic_tr(struct inode
*inode
, struct file
*filp
)
4424 struct trace_array
*tr
= inode
->i_private
;
4427 ret
= tracing_check_open_get_tr(tr
);
4431 filp
->private_data
= inode
->i_private
;
4436 static int tracing_release(struct inode
*inode
, struct file
*file
)
4438 struct trace_array
*tr
= inode
->i_private
;
4439 struct seq_file
*m
= file
->private_data
;
4440 struct trace_iterator
*iter
;
4443 if (!(file
->f_mode
& FMODE_READ
)) {
4444 trace_array_put(tr
);
4448 /* Writes do not use seq_file */
4450 mutex_lock(&trace_types_lock
);
4452 for_each_tracing_cpu(cpu
) {
4453 if (iter
->buffer_iter
[cpu
])
4454 ring_buffer_read_finish(iter
->buffer_iter
[cpu
]);
4457 if (iter
->trace
&& iter
->trace
->close
)
4458 iter
->trace
->close(iter
);
4460 if (!iter
->snapshot
&& tr
->stop_count
)
4461 /* reenable tracing if it was previously enabled */
4462 tracing_start_tr(tr
);
4464 __trace_array_put(tr
);
4466 mutex_unlock(&trace_types_lock
);
4468 mutex_destroy(&iter
->mutex
);
4469 free_cpumask_var(iter
->started
);
4472 kfree(iter
->buffer_iter
);
4473 seq_release_private(inode
, file
);
4478 static int tracing_release_generic_tr(struct inode
*inode
, struct file
*file
)
4480 struct trace_array
*tr
= inode
->i_private
;
4482 trace_array_put(tr
);
4486 static int tracing_single_release_tr(struct inode
*inode
, struct file
*file
)
4488 struct trace_array
*tr
= inode
->i_private
;
4490 trace_array_put(tr
);
4492 return single_release(inode
, file
);
4495 static int tracing_open(struct inode
*inode
, struct file
*file
)
4497 struct trace_array
*tr
= inode
->i_private
;
4498 struct trace_iterator
*iter
;
4501 ret
= tracing_check_open_get_tr(tr
);
4505 /* If this file was open for write, then erase contents */
4506 if ((file
->f_mode
& FMODE_WRITE
) && (file
->f_flags
& O_TRUNC
)) {
4507 int cpu
= tracing_get_cpu(inode
);
4508 struct array_buffer
*trace_buf
= &tr
->array_buffer
;
4510 #ifdef CONFIG_TRACER_MAX_TRACE
4511 if (tr
->current_trace
->print_max
)
4512 trace_buf
= &tr
->max_buffer
;
4515 if (cpu
== RING_BUFFER_ALL_CPUS
)
4516 tracing_reset_online_cpus(trace_buf
);
4518 tracing_reset_cpu(trace_buf
, cpu
);
4521 if (file
->f_mode
& FMODE_READ
) {
4522 iter
= __tracing_open(inode
, file
, false);
4524 ret
= PTR_ERR(iter
);
4525 else if (tr
->trace_flags
& TRACE_ITER_LATENCY_FMT
)
4526 iter
->iter_flags
|= TRACE_FILE_LAT_FMT
;
4530 trace_array_put(tr
);
4536 * Some tracers are not suitable for instance buffers.
4537 * A tracer is always available for the global array (toplevel)
4538 * or if it explicitly states that it is.
4541 trace_ok_for_array(struct tracer
*t
, struct trace_array
*tr
)
4543 return (tr
->flags
& TRACE_ARRAY_FL_GLOBAL
) || t
->allow_instances
;
4546 /* Find the next tracer that this trace array may use */
4547 static struct tracer
*
4548 get_tracer_for_array(struct trace_array
*tr
, struct tracer
*t
)
4550 while (t
&& !trace_ok_for_array(t
, tr
))
4557 t_next(struct seq_file
*m
, void *v
, loff_t
*pos
)
4559 struct trace_array
*tr
= m
->private;
4560 struct tracer
*t
= v
;
4565 t
= get_tracer_for_array(tr
, t
->next
);
4570 static void *t_start(struct seq_file
*m
, loff_t
*pos
)
4572 struct trace_array
*tr
= m
->private;
4576 mutex_lock(&trace_types_lock
);
4578 t
= get_tracer_for_array(tr
, trace_types
);
4579 for (; t
&& l
< *pos
; t
= t_next(m
, t
, &l
))
4585 static void t_stop(struct seq_file
*m
, void *p
)
4587 mutex_unlock(&trace_types_lock
);
4590 static int t_show(struct seq_file
*m
, void *v
)
4592 struct tracer
*t
= v
;
4597 seq_puts(m
, t
->name
);
4606 static const struct seq_operations show_traces_seq_ops
= {
4613 static int show_traces_open(struct inode
*inode
, struct file
*file
)
4615 struct trace_array
*tr
= inode
->i_private
;
4619 ret
= tracing_check_open_get_tr(tr
);
4623 ret
= seq_open(file
, &show_traces_seq_ops
);
4625 trace_array_put(tr
);
4629 m
= file
->private_data
;
4635 static int show_traces_release(struct inode
*inode
, struct file
*file
)
4637 struct trace_array
*tr
= inode
->i_private
;
4639 trace_array_put(tr
);
4640 return seq_release(inode
, file
);
4644 tracing_write_stub(struct file
*filp
, const char __user
*ubuf
,
4645 size_t count
, loff_t
*ppos
)
4650 loff_t
tracing_lseek(struct file
*file
, loff_t offset
, int whence
)
4654 if (file
->f_mode
& FMODE_READ
)
4655 ret
= seq_lseek(file
, offset
, whence
);
4657 file
->f_pos
= ret
= 0;
4662 static const struct file_operations tracing_fops
= {
4663 .open
= tracing_open
,
4665 .write
= tracing_write_stub
,
4666 .llseek
= tracing_lseek
,
4667 .release
= tracing_release
,
4670 static const struct file_operations show_traces_fops
= {
4671 .open
= show_traces_open
,
4673 .llseek
= seq_lseek
,
4674 .release
= show_traces_release
,
4678 tracing_cpumask_read(struct file
*filp
, char __user
*ubuf
,
4679 size_t count
, loff_t
*ppos
)
4681 struct trace_array
*tr
= file_inode(filp
)->i_private
;
4685 len
= snprintf(NULL
, 0, "%*pb\n",
4686 cpumask_pr_args(tr
->tracing_cpumask
)) + 1;
4687 mask_str
= kmalloc(len
, GFP_KERNEL
);
4691 len
= snprintf(mask_str
, len
, "%*pb\n",
4692 cpumask_pr_args(tr
->tracing_cpumask
));
4697 count
= simple_read_from_buffer(ubuf
, count
, ppos
, mask_str
, len
);
4705 int tracing_set_cpumask(struct trace_array
*tr
,
4706 cpumask_var_t tracing_cpumask_new
)
4713 local_irq_disable();
4714 arch_spin_lock(&tr
->max_lock
);
4715 for_each_tracing_cpu(cpu
) {
4717 * Increase/decrease the disabled counter if we are
4718 * about to flip a bit in the cpumask:
4720 if (cpumask_test_cpu(cpu
, tr
->tracing_cpumask
) &&
4721 !cpumask_test_cpu(cpu
, tracing_cpumask_new
)) {
4722 atomic_inc(&per_cpu_ptr(tr
->array_buffer
.data
, cpu
)->disabled
);
4723 ring_buffer_record_disable_cpu(tr
->array_buffer
.buffer
, cpu
);
4725 if (!cpumask_test_cpu(cpu
, tr
->tracing_cpumask
) &&
4726 cpumask_test_cpu(cpu
, tracing_cpumask_new
)) {
4727 atomic_dec(&per_cpu_ptr(tr
->array_buffer
.data
, cpu
)->disabled
);
4728 ring_buffer_record_enable_cpu(tr
->array_buffer
.buffer
, cpu
);
4731 arch_spin_unlock(&tr
->max_lock
);
4734 cpumask_copy(tr
->tracing_cpumask
, tracing_cpumask_new
);
4740 tracing_cpumask_write(struct file
*filp
, const char __user
*ubuf
,
4741 size_t count
, loff_t
*ppos
)
4743 struct trace_array
*tr
= file_inode(filp
)->i_private
;
4744 cpumask_var_t tracing_cpumask_new
;
4747 if (!alloc_cpumask_var(&tracing_cpumask_new
, GFP_KERNEL
))
4750 err
= cpumask_parse_user(ubuf
, count
, tracing_cpumask_new
);
4754 err
= tracing_set_cpumask(tr
, tracing_cpumask_new
);
4758 free_cpumask_var(tracing_cpumask_new
);
4763 free_cpumask_var(tracing_cpumask_new
);
4768 static const struct file_operations tracing_cpumask_fops
= {
4769 .open
= tracing_open_generic_tr
,
4770 .read
= tracing_cpumask_read
,
4771 .write
= tracing_cpumask_write
,
4772 .release
= tracing_release_generic_tr
,
4773 .llseek
= generic_file_llseek
,
4776 static int tracing_trace_options_show(struct seq_file
*m
, void *v
)
4778 struct tracer_opt
*trace_opts
;
4779 struct trace_array
*tr
= m
->private;
4783 mutex_lock(&trace_types_lock
);
4784 tracer_flags
= tr
->current_trace
->flags
->val
;
4785 trace_opts
= tr
->current_trace
->flags
->opts
;
4787 for (i
= 0; trace_options
[i
]; i
++) {
4788 if (tr
->trace_flags
& (1 << i
))
4789 seq_printf(m
, "%s\n", trace_options
[i
]);
4791 seq_printf(m
, "no%s\n", trace_options
[i
]);
4794 for (i
= 0; trace_opts
[i
].name
; i
++) {
4795 if (tracer_flags
& trace_opts
[i
].bit
)
4796 seq_printf(m
, "%s\n", trace_opts
[i
].name
);
4798 seq_printf(m
, "no%s\n", trace_opts
[i
].name
);
4800 mutex_unlock(&trace_types_lock
);
4805 static int __set_tracer_option(struct trace_array
*tr
,
4806 struct tracer_flags
*tracer_flags
,
4807 struct tracer_opt
*opts
, int neg
)
4809 struct tracer
*trace
= tracer_flags
->trace
;
4812 ret
= trace
->set_flag(tr
, tracer_flags
->val
, opts
->bit
, !neg
);
4817 tracer_flags
->val
&= ~opts
->bit
;
4819 tracer_flags
->val
|= opts
->bit
;
4823 /* Try to assign a tracer specific option */
4824 static int set_tracer_option(struct trace_array
*tr
, char *cmp
, int neg
)
4826 struct tracer
*trace
= tr
->current_trace
;
4827 struct tracer_flags
*tracer_flags
= trace
->flags
;
4828 struct tracer_opt
*opts
= NULL
;
4831 for (i
= 0; tracer_flags
->opts
[i
].name
; i
++) {
4832 opts
= &tracer_flags
->opts
[i
];
4834 if (strcmp(cmp
, opts
->name
) == 0)
4835 return __set_tracer_option(tr
, trace
->flags
, opts
, neg
);
4841 /* Some tracers require overwrite to stay enabled */
4842 int trace_keep_overwrite(struct tracer
*tracer
, u32 mask
, int set
)
4844 if (tracer
->enabled
&& (mask
& TRACE_ITER_OVERWRITE
) && !set
)
4850 int set_tracer_flag(struct trace_array
*tr
, unsigned int mask
, int enabled
)
4852 if ((mask
== TRACE_ITER_RECORD_TGID
) ||
4853 (mask
== TRACE_ITER_RECORD_CMD
))
4854 lockdep_assert_held(&event_mutex
);
4856 /* do nothing if flag is already set */
4857 if (!!(tr
->trace_flags
& mask
) == !!enabled
)
4860 /* Give the tracer a chance to approve the change */
4861 if (tr
->current_trace
->flag_changed
)
4862 if (tr
->current_trace
->flag_changed(tr
, mask
, !!enabled
))
4866 tr
->trace_flags
|= mask
;
4868 tr
->trace_flags
&= ~mask
;
4870 if (mask
== TRACE_ITER_RECORD_CMD
)
4871 trace_event_enable_cmd_record(enabled
);
4873 if (mask
== TRACE_ITER_RECORD_TGID
) {
4875 tgid_map
= kvcalloc(PID_MAX_DEFAULT
+ 1,
4879 tr
->trace_flags
&= ~TRACE_ITER_RECORD_TGID
;
4883 trace_event_enable_tgid_record(enabled
);
4886 if (mask
== TRACE_ITER_EVENT_FORK
)
4887 trace_event_follow_fork(tr
, enabled
);
4889 if (mask
== TRACE_ITER_FUNC_FORK
)
4890 ftrace_pid_follow_fork(tr
, enabled
);
4892 if (mask
== TRACE_ITER_OVERWRITE
) {
4893 ring_buffer_change_overwrite(tr
->array_buffer
.buffer
, enabled
);
4894 #ifdef CONFIG_TRACER_MAX_TRACE
4895 ring_buffer_change_overwrite(tr
->max_buffer
.buffer
, enabled
);
4899 if (mask
== TRACE_ITER_PRINTK
) {
4900 trace_printk_start_stop_comm(enabled
);
4901 trace_printk_control(enabled
);
4907 int trace_set_options(struct trace_array
*tr
, char *option
)
4912 size_t orig_len
= strlen(option
);
4915 cmp
= strstrip(option
);
4917 len
= str_has_prefix(cmp
, "no");
4923 mutex_lock(&event_mutex
);
4924 mutex_lock(&trace_types_lock
);
4926 ret
= match_string(trace_options
, -1, cmp
);
4927 /* If no option could be set, test the specific tracer options */
4929 ret
= set_tracer_option(tr
, cmp
, neg
);
4931 ret
= set_tracer_flag(tr
, 1 << ret
, !neg
);
4933 mutex_unlock(&trace_types_lock
);
4934 mutex_unlock(&event_mutex
);
4937 * If the first trailing whitespace is replaced with '\0' by strstrip,
4938 * turn it back into a space.
4940 if (orig_len
> strlen(option
))
4941 option
[strlen(option
)] = ' ';
4946 static void __init
apply_trace_boot_options(void)
4948 char *buf
= trace_boot_options_buf
;
4952 option
= strsep(&buf
, ",");
4958 trace_set_options(&global_trace
, option
);
4960 /* Put back the comma to allow this to be called again */
4967 tracing_trace_options_write(struct file
*filp
, const char __user
*ubuf
,
4968 size_t cnt
, loff_t
*ppos
)
4970 struct seq_file
*m
= filp
->private_data
;
4971 struct trace_array
*tr
= m
->private;
4975 if (cnt
>= sizeof(buf
))
4978 if (copy_from_user(buf
, ubuf
, cnt
))
4983 ret
= trace_set_options(tr
, buf
);
4992 static int tracing_trace_options_open(struct inode
*inode
, struct file
*file
)
4994 struct trace_array
*tr
= inode
->i_private
;
4997 ret
= tracing_check_open_get_tr(tr
);
5001 ret
= single_open(file
, tracing_trace_options_show
, inode
->i_private
);
5003 trace_array_put(tr
);
5008 static const struct file_operations tracing_iter_fops
= {
5009 .open
= tracing_trace_options_open
,
5011 .llseek
= seq_lseek
,
5012 .release
= tracing_single_release_tr
,
5013 .write
= tracing_trace_options_write
,
5016 static const char readme_msg
[] =
5017 "tracing mini-HOWTO:\n\n"
5018 "# echo 0 > tracing_on : quick way to disable tracing\n"
5019 "# echo 1 > tracing_on : quick way to re-enable tracing\n\n"
5020 " Important files:\n"
5021 " trace\t\t\t- The static contents of the buffer\n"
5022 "\t\t\t To clear the buffer write into this file: echo > trace\n"
5023 " trace_pipe\t\t- A consuming read to see the contents of the buffer\n"
5024 " current_tracer\t- function and latency tracers\n"
5025 " available_tracers\t- list of configured tracers for current_tracer\n"
5026 " error_log\t- error log for failed commands (that support it)\n"
5027 " buffer_size_kb\t- view and modify size of per cpu buffer\n"
5028 " buffer_total_size_kb - view total size of all cpu buffers\n\n"
5029 " trace_clock\t\t-change the clock used to order events\n"
5030 " local: Per cpu clock but may not be synced across CPUs\n"
5031 " global: Synced across CPUs but slows tracing down.\n"
5032 " counter: Not a clock, but just an increment\n"
5033 " uptime: Jiffy counter from time of boot\n"
5034 " perf: Same clock that perf events use\n"
5035 #ifdef CONFIG_X86_64
5036 " x86-tsc: TSC cycle counter\n"
5038 "\n timestamp_mode\t-view the mode used to timestamp events\n"
5039 " delta: Delta difference against a buffer-wide timestamp\n"
5040 " absolute: Absolute (standalone) timestamp\n"
5041 "\n trace_marker\t\t- Writes into this file writes into the kernel buffer\n"
5042 "\n trace_marker_raw\t\t- Writes into this file writes binary data into the kernel buffer\n"
5043 " tracing_cpumask\t- Limit which CPUs to trace\n"
5044 " instances\t\t- Make sub-buffers with: mkdir instances/foo\n"
5045 "\t\t\t Remove sub-buffer with rmdir\n"
5046 " trace_options\t\t- Set format or modify how tracing happens\n"
5047 "\t\t\t Disable an option by prefixing 'no' to the\n"
5048 "\t\t\t option name\n"
5049 " saved_cmdlines_size\t- echo command number in here to store comm-pid list\n"
5050 #ifdef CONFIG_DYNAMIC_FTRACE
5051 "\n available_filter_functions - list of functions that can be filtered on\n"
5052 " set_ftrace_filter\t- echo function name in here to only trace these\n"
5053 "\t\t\t functions\n"
5054 "\t accepts: func_full_name or glob-matching-pattern\n"
5055 "\t modules: Can select a group via module\n"
5056 "\t Format: :mod:<module-name>\n"
5057 "\t example: echo :mod:ext3 > set_ftrace_filter\n"
5058 "\t triggers: a command to perform when function is hit\n"
5059 "\t Format: <function>:<trigger>[:count]\n"
5060 "\t trigger: traceon, traceoff\n"
5061 "\t\t enable_event:<system>:<event>\n"
5062 "\t\t disable_event:<system>:<event>\n"
5063 #ifdef CONFIG_STACKTRACE
5066 #ifdef CONFIG_TRACER_SNAPSHOT
5071 "\t example: echo do_fault:traceoff > set_ftrace_filter\n"
5072 "\t echo do_trap:traceoff:3 > set_ftrace_filter\n"
5073 "\t The first one will disable tracing every time do_fault is hit\n"
5074 "\t The second will disable tracing at most 3 times when do_trap is hit\n"
5075 "\t The first time do trap is hit and it disables tracing, the\n"
5076 "\t counter will decrement to 2. If tracing is already disabled,\n"
5077 "\t the counter will not decrement. It only decrements when the\n"
5078 "\t trigger did work\n"
5079 "\t To remove trigger without count:\n"
5080 "\t echo '!<function>:<trigger> > set_ftrace_filter\n"
5081 "\t To remove trigger with a count:\n"
5082 "\t echo '!<function>:<trigger>:0 > set_ftrace_filter\n"
5083 " set_ftrace_notrace\t- echo function name in here to never trace.\n"
5084 "\t accepts: func_full_name, *func_end, func_begin*, *func_middle*\n"
5085 "\t modules: Can select a group via module command :mod:\n"
5086 "\t Does not accept triggers\n"
5087 #endif /* CONFIG_DYNAMIC_FTRACE */
5088 #ifdef CONFIG_FUNCTION_TRACER
5089 " set_ftrace_pid\t- Write pid(s) to only function trace those pids\n"
5091 " set_ftrace_notrace_pid\t- Write pid(s) to not function trace those pids\n"
5094 #ifdef CONFIG_FUNCTION_GRAPH_TRACER
5095 " set_graph_function\t- Trace the nested calls of a function (function_graph)\n"
5096 " set_graph_notrace\t- Do not trace the nested calls of a function (function_graph)\n"
5097 " max_graph_depth\t- Trace a limited depth of nested calls (0 is unlimited)\n"
5099 #ifdef CONFIG_TRACER_SNAPSHOT
5100 "\n snapshot\t\t- Like 'trace' but shows the content of the static\n"
5101 "\t\t\t snapshot buffer. Read the contents for more\n"
5102 "\t\t\t information\n"
5104 #ifdef CONFIG_STACK_TRACER
5105 " stack_trace\t\t- Shows the max stack trace when active\n"
5106 " stack_max_size\t- Shows current max stack size that was traced\n"
5107 "\t\t\t Write into this file to reset the max size (trigger a\n"
5108 "\t\t\t new trace)\n"
5109 #ifdef CONFIG_DYNAMIC_FTRACE
5110 " stack_trace_filter\t- Like set_ftrace_filter but limits what stack_trace\n"
5113 #endif /* CONFIG_STACK_TRACER */
5114 #ifdef CONFIG_DYNAMIC_EVENTS
5115 " dynamic_events\t\t- Create/append/remove/show the generic dynamic events\n"
5116 "\t\t\t Write into this file to define/undefine new trace events.\n"
5118 #ifdef CONFIG_KPROBE_EVENTS
5119 " kprobe_events\t\t- Create/append/remove/show the kernel dynamic events\n"
5120 "\t\t\t Write into this file to define/undefine new trace events.\n"
5122 #ifdef CONFIG_UPROBE_EVENTS
5123 " uprobe_events\t\t- Create/append/remove/show the userspace dynamic events\n"
5124 "\t\t\t Write into this file to define/undefine new trace events.\n"
5126 #if defined(CONFIG_KPROBE_EVENTS) || defined(CONFIG_UPROBE_EVENTS)
5127 "\t accepts: event-definitions (one definition per line)\n"
5128 "\t Format: p[:[<group>/]<event>] <place> [<args>]\n"
5129 "\t r[maxactive][:[<group>/]<event>] <place> [<args>]\n"
5130 #ifdef CONFIG_HIST_TRIGGERS
5131 "\t s:[synthetic/]<event> <field> [<field>]\n"
5133 "\t -:[<group>/]<event>\n"
5134 #ifdef CONFIG_KPROBE_EVENTS
5135 "\t place: [<module>:]<symbol>[+<offset>]|<memaddr>\n"
5136 "place (kretprobe): [<module>:]<symbol>[+<offset>]|<memaddr>\n"
5138 #ifdef CONFIG_UPROBE_EVENTS
5139 " place (uprobe): <path>:<offset>[(ref_ctr_offset)]\n"
5141 "\t args: <name>=fetcharg[:type]\n"
5142 "\t fetcharg: %<register>, @<address>, @<symbol>[+|-<offset>],\n"
5143 #ifdef CONFIG_HAVE_FUNCTION_ARG_ACCESS_API
5144 "\t $stack<index>, $stack, $retval, $comm, $arg<N>,\n"
5146 "\t $stack<index>, $stack, $retval, $comm,\n"
5148 "\t +|-[u]<offset>(<fetcharg>), \\imm-value, \\\"imm-string\"\n"
5149 "\t type: s8/16/32/64, u8/16/32/64, x8/16/32/64, string, symbol,\n"
5150 "\t b<bit-width>@<bit-offset>/<container-size>, ustring,\n"
5151 "\t <type>\\[<array-size>\\]\n"
5152 #ifdef CONFIG_HIST_TRIGGERS
5153 "\t field: <stype> <name>;\n"
5154 "\t stype: u8/u16/u32/u64, s8/s16/s32/s64, pid_t,\n"
5155 "\t [unsigned] char/int/long\n"
5158 " events/\t\t- Directory containing all trace event subsystems:\n"
5159 " enable\t\t- Write 0/1 to enable/disable tracing of all events\n"
5160 " events/<system>/\t- Directory containing all trace events for <system>:\n"
5161 " enable\t\t- Write 0/1 to enable/disable tracing of all <system>\n"
5163 " filter\t\t- If set, only events passing filter are traced\n"
5164 " events/<system>/<event>/\t- Directory containing control files for\n"
5166 " enable\t\t- Write 0/1 to enable/disable tracing of <event>\n"
5167 " filter\t\t- If set, only events passing filter are traced\n"
5168 " trigger\t\t- If set, a command to perform when event is hit\n"
5169 "\t Format: <trigger>[:count][if <filter>]\n"
5170 "\t trigger: traceon, traceoff\n"
5171 "\t enable_event:<system>:<event>\n"
5172 "\t disable_event:<system>:<event>\n"
5173 #ifdef CONFIG_HIST_TRIGGERS
5174 "\t enable_hist:<system>:<event>\n"
5175 "\t disable_hist:<system>:<event>\n"
5177 #ifdef CONFIG_STACKTRACE
5180 #ifdef CONFIG_TRACER_SNAPSHOT
5183 #ifdef CONFIG_HIST_TRIGGERS
5184 "\t\t hist (see below)\n"
5186 "\t example: echo traceoff > events/block/block_unplug/trigger\n"
5187 "\t echo traceoff:3 > events/block/block_unplug/trigger\n"
5188 "\t echo 'enable_event:kmem:kmalloc:3 if nr_rq > 1' > \\\n"
5189 "\t events/block/block_unplug/trigger\n"
5190 "\t The first disables tracing every time block_unplug is hit.\n"
5191 "\t The second disables tracing the first 3 times block_unplug is hit.\n"
5192 "\t The third enables the kmalloc event the first 3 times block_unplug\n"
5193 "\t is hit and has value of greater than 1 for the 'nr_rq' event field.\n"
5194 "\t Like function triggers, the counter is only decremented if it\n"
5195 "\t enabled or disabled tracing.\n"
5196 "\t To remove a trigger without a count:\n"
5197 "\t echo '!<trigger> > <system>/<event>/trigger\n"
5198 "\t To remove a trigger with a count:\n"
5199 "\t echo '!<trigger>:0 > <system>/<event>/trigger\n"
5200 "\t Filters can be ignored when removing a trigger.\n"
5201 #ifdef CONFIG_HIST_TRIGGERS
5202 " hist trigger\t- If set, event hits are aggregated into a hash table\n"
5203 "\t Format: hist:keys=<field1[,field2,...]>\n"
5204 "\t [:values=<field1[,field2,...]>]\n"
5205 "\t [:sort=<field1[,field2,...]>]\n"
5206 "\t [:size=#entries]\n"
5207 "\t [:pause][:continue][:clear]\n"
5208 "\t [:name=histname1]\n"
5209 "\t [:<handler>.<action>]\n"
5210 "\t [if <filter>]\n\n"
5211 "\t When a matching event is hit, an entry is added to a hash\n"
5212 "\t table using the key(s) and value(s) named, and the value of a\n"
5213 "\t sum called 'hitcount' is incremented. Keys and values\n"
5214 "\t correspond to fields in the event's format description. Keys\n"
5215 "\t can be any field, or the special string 'stacktrace'.\n"
5216 "\t Compound keys consisting of up to two fields can be specified\n"
5217 "\t by the 'keys' keyword. Values must correspond to numeric\n"
5218 "\t fields. Sort keys consisting of up to two fields can be\n"
5219 "\t specified using the 'sort' keyword. The sort direction can\n"
5220 "\t be modified by appending '.descending' or '.ascending' to a\n"
5221 "\t sort field. The 'size' parameter can be used to specify more\n"
5222 "\t or fewer than the default 2048 entries for the hashtable size.\n"
5223 "\t If a hist trigger is given a name using the 'name' parameter,\n"
5224 "\t its histogram data will be shared with other triggers of the\n"
5225 "\t same name, and trigger hits will update this common data.\n\n"
5226 "\t Reading the 'hist' file for the event will dump the hash\n"
5227 "\t table in its entirety to stdout. If there are multiple hist\n"
5228 "\t triggers attached to an event, there will be a table for each\n"
5229 "\t trigger in the output. The table displayed for a named\n"
5230 "\t trigger will be the same as any other instance having the\n"
5231 "\t same name. The default format used to display a given field\n"
5232 "\t can be modified by appending any of the following modifiers\n"
5233 "\t to the field name, as applicable:\n\n"
5234 "\t .hex display a number as a hex value\n"
5235 "\t .sym display an address as a symbol\n"
5236 "\t .sym-offset display an address as a symbol and offset\n"
5237 "\t .execname display a common_pid as a program name\n"
5238 "\t .syscall display a syscall id as a syscall name\n"
5239 "\t .log2 display log2 value rather than raw number\n"
5240 "\t .usecs display a common_timestamp in microseconds\n\n"
5241 "\t The 'pause' parameter can be used to pause an existing hist\n"
5242 "\t trigger or to start a hist trigger but not log any events\n"
5243 "\t until told to do so. 'continue' can be used to start or\n"
5244 "\t restart a paused hist trigger.\n\n"
5245 "\t The 'clear' parameter will clear the contents of a running\n"
5246 "\t hist trigger and leave its current paused/active state\n"
5248 "\t The enable_hist and disable_hist triggers can be used to\n"
5249 "\t have one event conditionally start and stop another event's\n"
5250 "\t already-attached hist trigger. The syntax is analogous to\n"
5251 "\t the enable_event and disable_event triggers.\n\n"
5252 "\t Hist trigger handlers and actions are executed whenever a\n"
5253 "\t a histogram entry is added or updated. They take the form:\n\n"
5254 "\t <handler>.<action>\n\n"
5255 "\t The available handlers are:\n\n"
5256 "\t onmatch(matching.event) - invoke on addition or update\n"
5257 "\t onmax(var) - invoke if var exceeds current max\n"
5258 "\t onchange(var) - invoke action if var changes\n\n"
5259 "\t The available actions are:\n\n"
5260 "\t trace(<synthetic_event>,param list) - generate synthetic event\n"
5261 "\t save(field,...) - save current event fields\n"
5262 #ifdef CONFIG_TRACER_SNAPSHOT
5263 "\t snapshot() - snapshot the trace buffer\n"
5269 tracing_readme_read(struct file
*filp
, char __user
*ubuf
,
5270 size_t cnt
, loff_t
*ppos
)
5272 return simple_read_from_buffer(ubuf
, cnt
, ppos
,
5273 readme_msg
, strlen(readme_msg
));
5276 static const struct file_operations tracing_readme_fops
= {
5277 .open
= tracing_open_generic
,
5278 .read
= tracing_readme_read
,
5279 .llseek
= generic_file_llseek
,
5282 static void *saved_tgids_next(struct seq_file
*m
, void *v
, loff_t
*pos
)
5286 if (*pos
|| m
->count
)
5291 for (; ptr
<= &tgid_map
[PID_MAX_DEFAULT
]; ptr
++) {
5292 if (trace_find_tgid(*ptr
))
5299 static void *saved_tgids_start(struct seq_file
*m
, loff_t
*pos
)
5309 v
= saved_tgids_next(m
, v
, &l
);
5317 static void saved_tgids_stop(struct seq_file
*m
, void *v
)
5321 static int saved_tgids_show(struct seq_file
*m
, void *v
)
5323 int pid
= (int *)v
- tgid_map
;
5325 seq_printf(m
, "%d %d\n", pid
, trace_find_tgid(pid
));
5329 static const struct seq_operations tracing_saved_tgids_seq_ops
= {
5330 .start
= saved_tgids_start
,
5331 .stop
= saved_tgids_stop
,
5332 .next
= saved_tgids_next
,
5333 .show
= saved_tgids_show
,
5336 static int tracing_saved_tgids_open(struct inode
*inode
, struct file
*filp
)
5340 ret
= tracing_check_open_get_tr(NULL
);
5344 return seq_open(filp
, &tracing_saved_tgids_seq_ops
);
5348 static const struct file_operations tracing_saved_tgids_fops
= {
5349 .open
= tracing_saved_tgids_open
,
5351 .llseek
= seq_lseek
,
5352 .release
= seq_release
,
5355 static void *saved_cmdlines_next(struct seq_file
*m
, void *v
, loff_t
*pos
)
5357 unsigned int *ptr
= v
;
5359 if (*pos
|| m
->count
)
5364 for (; ptr
< &savedcmd
->map_cmdline_to_pid
[savedcmd
->cmdline_num
];
5366 if (*ptr
== -1 || *ptr
== NO_CMDLINE_MAP
)
5375 static void *saved_cmdlines_start(struct seq_file
*m
, loff_t
*pos
)
5381 arch_spin_lock(&trace_cmdline_lock
);
5383 v
= &savedcmd
->map_cmdline_to_pid
[0];
5385 v
= saved_cmdlines_next(m
, v
, &l
);
5393 static void saved_cmdlines_stop(struct seq_file
*m
, void *v
)
5395 arch_spin_unlock(&trace_cmdline_lock
);
5399 static int saved_cmdlines_show(struct seq_file
*m
, void *v
)
5401 char buf
[TASK_COMM_LEN
];
5402 unsigned int *pid
= v
;
5404 __trace_find_cmdline(*pid
, buf
);
5405 seq_printf(m
, "%d %s\n", *pid
, buf
);
5409 static const struct seq_operations tracing_saved_cmdlines_seq_ops
= {
5410 .start
= saved_cmdlines_start
,
5411 .next
= saved_cmdlines_next
,
5412 .stop
= saved_cmdlines_stop
,
5413 .show
= saved_cmdlines_show
,
5416 static int tracing_saved_cmdlines_open(struct inode
*inode
, struct file
*filp
)
5420 ret
= tracing_check_open_get_tr(NULL
);
5424 return seq_open(filp
, &tracing_saved_cmdlines_seq_ops
);
5427 static const struct file_operations tracing_saved_cmdlines_fops
= {
5428 .open
= tracing_saved_cmdlines_open
,
5430 .llseek
= seq_lseek
,
5431 .release
= seq_release
,
5435 tracing_saved_cmdlines_size_read(struct file
*filp
, char __user
*ubuf
,
5436 size_t cnt
, loff_t
*ppos
)
5441 arch_spin_lock(&trace_cmdline_lock
);
5442 r
= scnprintf(buf
, sizeof(buf
), "%u\n", savedcmd
->cmdline_num
);
5443 arch_spin_unlock(&trace_cmdline_lock
);
5445 return simple_read_from_buffer(ubuf
, cnt
, ppos
, buf
, r
);
5448 static void free_saved_cmdlines_buffer(struct saved_cmdlines_buffer
*s
)
5450 kfree(s
->saved_cmdlines
);
5451 kfree(s
->map_cmdline_to_pid
);
5455 static int tracing_resize_saved_cmdlines(unsigned int val
)
5457 struct saved_cmdlines_buffer
*s
, *savedcmd_temp
;
5459 s
= kmalloc(sizeof(*s
), GFP_KERNEL
);
5463 if (allocate_cmdlines_buffer(val
, s
) < 0) {
5468 arch_spin_lock(&trace_cmdline_lock
);
5469 savedcmd_temp
= savedcmd
;
5471 arch_spin_unlock(&trace_cmdline_lock
);
5472 free_saved_cmdlines_buffer(savedcmd_temp
);
5478 tracing_saved_cmdlines_size_write(struct file
*filp
, const char __user
*ubuf
,
5479 size_t cnt
, loff_t
*ppos
)
5484 ret
= kstrtoul_from_user(ubuf
, cnt
, 10, &val
);
5488 /* must have at least 1 entry or less than PID_MAX_DEFAULT */
5489 if (!val
|| val
> PID_MAX_DEFAULT
)
5492 ret
= tracing_resize_saved_cmdlines((unsigned int)val
);
5501 static const struct file_operations tracing_saved_cmdlines_size_fops
= {
5502 .open
= tracing_open_generic
,
5503 .read
= tracing_saved_cmdlines_size_read
,
5504 .write
= tracing_saved_cmdlines_size_write
,
5507 #ifdef CONFIG_TRACE_EVAL_MAP_FILE
5508 static union trace_eval_map_item
*
5509 update_eval_map(union trace_eval_map_item
*ptr
)
5511 if (!ptr
->map
.eval_string
) {
5512 if (ptr
->tail
.next
) {
5513 ptr
= ptr
->tail
.next
;
5514 /* Set ptr to the next real item (skip head) */
5522 static void *eval_map_next(struct seq_file
*m
, void *v
, loff_t
*pos
)
5524 union trace_eval_map_item
*ptr
= v
;
5527 * Paranoid! If ptr points to end, we don't want to increment past it.
5528 * This really should never happen.
5531 ptr
= update_eval_map(ptr
);
5532 if (WARN_ON_ONCE(!ptr
))
5536 ptr
= update_eval_map(ptr
);
5541 static void *eval_map_start(struct seq_file
*m
, loff_t
*pos
)
5543 union trace_eval_map_item
*v
;
5546 mutex_lock(&trace_eval_mutex
);
5548 v
= trace_eval_maps
;
5552 while (v
&& l
< *pos
) {
5553 v
= eval_map_next(m
, v
, &l
);
5559 static void eval_map_stop(struct seq_file
*m
, void *v
)
5561 mutex_unlock(&trace_eval_mutex
);
5564 static int eval_map_show(struct seq_file
*m
, void *v
)
5566 union trace_eval_map_item
*ptr
= v
;
5568 seq_printf(m
, "%s %ld (%s)\n",
5569 ptr
->map
.eval_string
, ptr
->map
.eval_value
,
5575 static const struct seq_operations tracing_eval_map_seq_ops
= {
5576 .start
= eval_map_start
,
5577 .next
= eval_map_next
,
5578 .stop
= eval_map_stop
,
5579 .show
= eval_map_show
,
5582 static int tracing_eval_map_open(struct inode
*inode
, struct file
*filp
)
5586 ret
= tracing_check_open_get_tr(NULL
);
5590 return seq_open(filp
, &tracing_eval_map_seq_ops
);
5593 static const struct file_operations tracing_eval_map_fops
= {
5594 .open
= tracing_eval_map_open
,
5596 .llseek
= seq_lseek
,
5597 .release
= seq_release
,
5600 static inline union trace_eval_map_item
*
5601 trace_eval_jmp_to_tail(union trace_eval_map_item
*ptr
)
5603 /* Return tail of array given the head */
5604 return ptr
+ ptr
->head
.length
+ 1;
5608 trace_insert_eval_map_file(struct module
*mod
, struct trace_eval_map
**start
,
5611 struct trace_eval_map
**stop
;
5612 struct trace_eval_map
**map
;
5613 union trace_eval_map_item
*map_array
;
5614 union trace_eval_map_item
*ptr
;
5619 * The trace_eval_maps contains the map plus a head and tail item,
5620 * where the head holds the module and length of array, and the
5621 * tail holds a pointer to the next list.
5623 map_array
= kmalloc_array(len
+ 2, sizeof(*map_array
), GFP_KERNEL
);
5625 pr_warn("Unable to allocate trace eval mapping\n");
5629 mutex_lock(&trace_eval_mutex
);
5631 if (!trace_eval_maps
)
5632 trace_eval_maps
= map_array
;
5634 ptr
= trace_eval_maps
;
5636 ptr
= trace_eval_jmp_to_tail(ptr
);
5637 if (!ptr
->tail
.next
)
5639 ptr
= ptr
->tail
.next
;
5642 ptr
->tail
.next
= map_array
;
5644 map_array
->head
.mod
= mod
;
5645 map_array
->head
.length
= len
;
5648 for (map
= start
; (unsigned long)map
< (unsigned long)stop
; map
++) {
5649 map_array
->map
= **map
;
5652 memset(map_array
, 0, sizeof(*map_array
));
5654 mutex_unlock(&trace_eval_mutex
);
5657 static void trace_create_eval_file(struct dentry
*d_tracer
)
5659 trace_create_file("eval_map", 0444, d_tracer
,
5660 NULL
, &tracing_eval_map_fops
);
5663 #else /* CONFIG_TRACE_EVAL_MAP_FILE */
5664 static inline void trace_create_eval_file(struct dentry
*d_tracer
) { }
5665 static inline void trace_insert_eval_map_file(struct module
*mod
,
5666 struct trace_eval_map
**start
, int len
) { }
5667 #endif /* !CONFIG_TRACE_EVAL_MAP_FILE */
5669 static void trace_insert_eval_map(struct module
*mod
,
5670 struct trace_eval_map
**start
, int len
)
5672 struct trace_eval_map
**map
;
5679 trace_event_eval_update(map
, len
);
5681 trace_insert_eval_map_file(mod
, start
, len
);
5685 tracing_set_trace_read(struct file
*filp
, char __user
*ubuf
,
5686 size_t cnt
, loff_t
*ppos
)
5688 struct trace_array
*tr
= filp
->private_data
;
5689 char buf
[MAX_TRACER_SIZE
+2];
5692 mutex_lock(&trace_types_lock
);
5693 r
= sprintf(buf
, "%s\n", tr
->current_trace
->name
);
5694 mutex_unlock(&trace_types_lock
);
5696 return simple_read_from_buffer(ubuf
, cnt
, ppos
, buf
, r
);
5699 int tracer_init(struct tracer
*t
, struct trace_array
*tr
)
5701 tracing_reset_online_cpus(&tr
->array_buffer
);
5705 static void set_buffer_entries(struct array_buffer
*buf
, unsigned long val
)
5709 for_each_tracing_cpu(cpu
)
5710 per_cpu_ptr(buf
->data
, cpu
)->entries
= val
;
5713 #ifdef CONFIG_TRACER_MAX_TRACE
5714 /* resize @tr's buffer to the size of @size_tr's entries */
5715 static int resize_buffer_duplicate_size(struct array_buffer
*trace_buf
,
5716 struct array_buffer
*size_buf
, int cpu_id
)
5720 if (cpu_id
== RING_BUFFER_ALL_CPUS
) {
5721 for_each_tracing_cpu(cpu
) {
5722 ret
= ring_buffer_resize(trace_buf
->buffer
,
5723 per_cpu_ptr(size_buf
->data
, cpu
)->entries
, cpu
);
5726 per_cpu_ptr(trace_buf
->data
, cpu
)->entries
=
5727 per_cpu_ptr(size_buf
->data
, cpu
)->entries
;
5730 ret
= ring_buffer_resize(trace_buf
->buffer
,
5731 per_cpu_ptr(size_buf
->data
, cpu_id
)->entries
, cpu_id
);
5733 per_cpu_ptr(trace_buf
->data
, cpu_id
)->entries
=
5734 per_cpu_ptr(size_buf
->data
, cpu_id
)->entries
;
5739 #endif /* CONFIG_TRACER_MAX_TRACE */
5741 static int __tracing_resize_ring_buffer(struct trace_array
*tr
,
5742 unsigned long size
, int cpu
)
5747 * If kernel or user changes the size of the ring buffer
5748 * we use the size that was given, and we can forget about
5749 * expanding it later.
5751 ring_buffer_expanded
= true;
5753 /* May be called before buffers are initialized */
5754 if (!tr
->array_buffer
.buffer
)
5757 ret
= ring_buffer_resize(tr
->array_buffer
.buffer
, size
, cpu
);
5761 #ifdef CONFIG_TRACER_MAX_TRACE
5762 if (!(tr
->flags
& TRACE_ARRAY_FL_GLOBAL
) ||
5763 !tr
->current_trace
->use_max_tr
)
5766 ret
= ring_buffer_resize(tr
->max_buffer
.buffer
, size
, cpu
);
5768 int r
= resize_buffer_duplicate_size(&tr
->array_buffer
,
5769 &tr
->array_buffer
, cpu
);
5772 * AARGH! We are left with different
5773 * size max buffer!!!!
5774 * The max buffer is our "snapshot" buffer.
5775 * When a tracer needs a snapshot (one of the
5776 * latency tracers), it swaps the max buffer
5777 * with the saved snap shot. We succeeded to
5778 * update the size of the main buffer, but failed to
5779 * update the size of the max buffer. But when we tried
5780 * to reset the main buffer to the original size, we
5781 * failed there too. This is very unlikely to
5782 * happen, but if it does, warn and kill all
5786 tracing_disabled
= 1;
5791 if (cpu
== RING_BUFFER_ALL_CPUS
)
5792 set_buffer_entries(&tr
->max_buffer
, size
);
5794 per_cpu_ptr(tr
->max_buffer
.data
, cpu
)->entries
= size
;
5797 #endif /* CONFIG_TRACER_MAX_TRACE */
5799 if (cpu
== RING_BUFFER_ALL_CPUS
)
5800 set_buffer_entries(&tr
->array_buffer
, size
);
5802 per_cpu_ptr(tr
->array_buffer
.data
, cpu
)->entries
= size
;
5807 ssize_t
tracing_resize_ring_buffer(struct trace_array
*tr
,
5808 unsigned long size
, int cpu_id
)
5812 mutex_lock(&trace_types_lock
);
5814 if (cpu_id
!= RING_BUFFER_ALL_CPUS
) {
5815 /* make sure, this cpu is enabled in the mask */
5816 if (!cpumask_test_cpu(cpu_id
, tracing_buffer_mask
)) {
5822 ret
= __tracing_resize_ring_buffer(tr
, size
, cpu_id
);
5827 mutex_unlock(&trace_types_lock
);
5834 * tracing_update_buffers - used by tracing facility to expand ring buffers
5836 * To save on memory when the tracing is never used on a system with it
5837 * configured in. The ring buffers are set to a minimum size. But once
5838 * a user starts to use the tracing facility, then they need to grow
5839 * to their default size.
5841 * This function is to be called when a tracer is about to be used.
5843 int tracing_update_buffers(void)
5847 mutex_lock(&trace_types_lock
);
5848 if (!ring_buffer_expanded
)
5849 ret
= __tracing_resize_ring_buffer(&global_trace
, trace_buf_size
,
5850 RING_BUFFER_ALL_CPUS
);
5851 mutex_unlock(&trace_types_lock
);
5856 struct trace_option_dentry
;
5859 create_trace_option_files(struct trace_array
*tr
, struct tracer
*tracer
);
5862 * Used to clear out the tracer before deletion of an instance.
5863 * Must have trace_types_lock held.
5865 static void tracing_set_nop(struct trace_array
*tr
)
5867 if (tr
->current_trace
== &nop_trace
)
5870 tr
->current_trace
->enabled
--;
5872 if (tr
->current_trace
->reset
)
5873 tr
->current_trace
->reset(tr
);
5875 tr
->current_trace
= &nop_trace
;
5878 static void add_tracer_options(struct trace_array
*tr
, struct tracer
*t
)
5880 /* Only enable if the directory has been created already. */
5884 create_trace_option_files(tr
, t
);
5887 int tracing_set_tracer(struct trace_array
*tr
, const char *buf
)
5890 #ifdef CONFIG_TRACER_MAX_TRACE
5895 mutex_lock(&trace_types_lock
);
5897 if (!ring_buffer_expanded
) {
5898 ret
= __tracing_resize_ring_buffer(tr
, trace_buf_size
,
5899 RING_BUFFER_ALL_CPUS
);
5905 for (t
= trace_types
; t
; t
= t
->next
) {
5906 if (strcmp(t
->name
, buf
) == 0)
5913 if (t
== tr
->current_trace
)
5916 #ifdef CONFIG_TRACER_SNAPSHOT
5917 if (t
->use_max_tr
) {
5918 arch_spin_lock(&tr
->max_lock
);
5919 if (tr
->cond_snapshot
)
5921 arch_spin_unlock(&tr
->max_lock
);
5926 /* Some tracers won't work on kernel command line */
5927 if (system_state
< SYSTEM_RUNNING
&& t
->noboot
) {
5928 pr_warn("Tracer '%s' is not allowed on command line, ignored\n",
5933 /* Some tracers are only allowed for the top level buffer */
5934 if (!trace_ok_for_array(t
, tr
)) {
5939 /* If trace pipe files are being read, we can't change the tracer */
5940 if (tr
->trace_ref
) {
5945 trace_branch_disable();
5947 tr
->current_trace
->enabled
--;
5949 if (tr
->current_trace
->reset
)
5950 tr
->current_trace
->reset(tr
);
5952 /* Current trace needs to be nop_trace before synchronize_rcu */
5953 tr
->current_trace
= &nop_trace
;
5955 #ifdef CONFIG_TRACER_MAX_TRACE
5956 had_max_tr
= tr
->allocated_snapshot
;
5958 if (had_max_tr
&& !t
->use_max_tr
) {
5960 * We need to make sure that the update_max_tr sees that
5961 * current_trace changed to nop_trace to keep it from
5962 * swapping the buffers after we resize it.
5963 * The update_max_tr is called from interrupts disabled
5964 * so a synchronized_sched() is sufficient.
5971 #ifdef CONFIG_TRACER_MAX_TRACE
5972 if (t
->use_max_tr
&& !had_max_tr
) {
5973 ret
= tracing_alloc_snapshot_instance(tr
);
5980 ret
= tracer_init(t
, tr
);
5985 tr
->current_trace
= t
;
5986 tr
->current_trace
->enabled
++;
5987 trace_branch_enable(tr
);
5989 mutex_unlock(&trace_types_lock
);
5995 tracing_set_trace_write(struct file
*filp
, const char __user
*ubuf
,
5996 size_t cnt
, loff_t
*ppos
)
5998 struct trace_array
*tr
= filp
->private_data
;
5999 char buf
[MAX_TRACER_SIZE
+1];
6006 if (cnt
> MAX_TRACER_SIZE
)
6007 cnt
= MAX_TRACER_SIZE
;
6009 if (copy_from_user(buf
, ubuf
, cnt
))
6014 /* strip ending whitespace. */
6015 for (i
= cnt
- 1; i
> 0 && isspace(buf
[i
]); i
--)
6018 err
= tracing_set_tracer(tr
, buf
);
6028 tracing_nsecs_read(unsigned long *ptr
, char __user
*ubuf
,
6029 size_t cnt
, loff_t
*ppos
)
6034 r
= snprintf(buf
, sizeof(buf
), "%ld\n",
6035 *ptr
== (unsigned long)-1 ? -1 : nsecs_to_usecs(*ptr
));
6036 if (r
> sizeof(buf
))
6038 return simple_read_from_buffer(ubuf
, cnt
, ppos
, buf
, r
);
6042 tracing_nsecs_write(unsigned long *ptr
, const char __user
*ubuf
,
6043 size_t cnt
, loff_t
*ppos
)
6048 ret
= kstrtoul_from_user(ubuf
, cnt
, 10, &val
);
6058 tracing_thresh_read(struct file
*filp
, char __user
*ubuf
,
6059 size_t cnt
, loff_t
*ppos
)
6061 return tracing_nsecs_read(&tracing_thresh
, ubuf
, cnt
, ppos
);
6065 tracing_thresh_write(struct file
*filp
, const char __user
*ubuf
,
6066 size_t cnt
, loff_t
*ppos
)
6068 struct trace_array
*tr
= filp
->private_data
;
6071 mutex_lock(&trace_types_lock
);
6072 ret
= tracing_nsecs_write(&tracing_thresh
, ubuf
, cnt
, ppos
);
6076 if (tr
->current_trace
->update_thresh
) {
6077 ret
= tr
->current_trace
->update_thresh(tr
);
6084 mutex_unlock(&trace_types_lock
);
6089 #if defined(CONFIG_TRACER_MAX_TRACE) || defined(CONFIG_HWLAT_TRACER)
6092 tracing_max_lat_read(struct file
*filp
, char __user
*ubuf
,
6093 size_t cnt
, loff_t
*ppos
)
6095 return tracing_nsecs_read(filp
->private_data
, ubuf
, cnt
, ppos
);
6099 tracing_max_lat_write(struct file
*filp
, const char __user
*ubuf
,
6100 size_t cnt
, loff_t
*ppos
)
6102 return tracing_nsecs_write(filp
->private_data
, ubuf
, cnt
, ppos
);
6107 static int tracing_open_pipe(struct inode
*inode
, struct file
*filp
)
6109 struct trace_array
*tr
= inode
->i_private
;
6110 struct trace_iterator
*iter
;
6113 ret
= tracing_check_open_get_tr(tr
);
6117 mutex_lock(&trace_types_lock
);
6119 /* create a buffer to store the information to pass to userspace */
6120 iter
= kzalloc(sizeof(*iter
), GFP_KERNEL
);
6123 __trace_array_put(tr
);
6127 trace_seq_init(&iter
->seq
);
6128 iter
->trace
= tr
->current_trace
;
6130 if (!alloc_cpumask_var(&iter
->started
, GFP_KERNEL
)) {
6135 /* trace pipe does not show start of buffer */
6136 cpumask_setall(iter
->started
);
6138 if (tr
->trace_flags
& TRACE_ITER_LATENCY_FMT
)
6139 iter
->iter_flags
|= TRACE_FILE_LAT_FMT
;
6141 /* Output in nanoseconds only if we are using a clock in nanoseconds. */
6142 if (trace_clocks
[tr
->clock_id
].in_ns
)
6143 iter
->iter_flags
|= TRACE_FILE_TIME_IN_NS
;
6146 iter
->array_buffer
= &tr
->array_buffer
;
6147 iter
->cpu_file
= tracing_get_cpu(inode
);
6148 mutex_init(&iter
->mutex
);
6149 filp
->private_data
= iter
;
6151 if (iter
->trace
->pipe_open
)
6152 iter
->trace
->pipe_open(iter
);
6154 nonseekable_open(inode
, filp
);
6158 mutex_unlock(&trace_types_lock
);
6163 __trace_array_put(tr
);
6164 mutex_unlock(&trace_types_lock
);
6168 static int tracing_release_pipe(struct inode
*inode
, struct file
*file
)
6170 struct trace_iterator
*iter
= file
->private_data
;
6171 struct trace_array
*tr
= inode
->i_private
;
6173 mutex_lock(&trace_types_lock
);
6177 if (iter
->trace
->pipe_close
)
6178 iter
->trace
->pipe_close(iter
);
6180 mutex_unlock(&trace_types_lock
);
6182 free_cpumask_var(iter
->started
);
6183 mutex_destroy(&iter
->mutex
);
6186 trace_array_put(tr
);
6192 trace_poll(struct trace_iterator
*iter
, struct file
*filp
, poll_table
*poll_table
)
6194 struct trace_array
*tr
= iter
->tr
;
6196 /* Iterators are static, they should be filled or empty */
6197 if (trace_buffer_iter(iter
, iter
->cpu_file
))
6198 return EPOLLIN
| EPOLLRDNORM
;
6200 if (tr
->trace_flags
& TRACE_ITER_BLOCK
)
6202 * Always select as readable when in blocking mode
6204 return EPOLLIN
| EPOLLRDNORM
;
6206 return ring_buffer_poll_wait(iter
->array_buffer
->buffer
, iter
->cpu_file
,
6211 tracing_poll_pipe(struct file
*filp
, poll_table
*poll_table
)
6213 struct trace_iterator
*iter
= filp
->private_data
;
6215 return trace_poll(iter
, filp
, poll_table
);
6218 /* Must be called with iter->mutex held. */
6219 static int tracing_wait_pipe(struct file
*filp
)
6221 struct trace_iterator
*iter
= filp
->private_data
;
6224 while (trace_empty(iter
)) {
6226 if ((filp
->f_flags
& O_NONBLOCK
)) {
6231 * We block until we read something and tracing is disabled.
6232 * We still block if tracing is disabled, but we have never
6233 * read anything. This allows a user to cat this file, and
6234 * then enable tracing. But after we have read something,
6235 * we give an EOF when tracing is again disabled.
6237 * iter->pos will be 0 if we haven't read anything.
6239 if (!tracer_tracing_is_on(iter
->tr
) && iter
->pos
)
6242 mutex_unlock(&iter
->mutex
);
6244 ret
= wait_on_pipe(iter
, 0);
6246 mutex_lock(&iter
->mutex
);
6259 tracing_read_pipe(struct file
*filp
, char __user
*ubuf
,
6260 size_t cnt
, loff_t
*ppos
)
6262 struct trace_iterator
*iter
= filp
->private_data
;
6266 * Avoid more than one consumer on a single file descriptor
6267 * This is just a matter of traces coherency, the ring buffer itself
6270 mutex_lock(&iter
->mutex
);
6272 /* return any leftover data */
6273 sret
= trace_seq_to_user(&iter
->seq
, ubuf
, cnt
);
6277 trace_seq_init(&iter
->seq
);
6279 if (iter
->trace
->read
) {
6280 sret
= iter
->trace
->read(iter
, filp
, ubuf
, cnt
, ppos
);
6286 sret
= tracing_wait_pipe(filp
);
6290 /* stop when tracing is finished */
6291 if (trace_empty(iter
)) {
6296 if (cnt
>= PAGE_SIZE
)
6297 cnt
= PAGE_SIZE
- 1;
6299 /* reset all but tr, trace, and overruns */
6300 memset(&iter
->seq
, 0,
6301 sizeof(struct trace_iterator
) -
6302 offsetof(struct trace_iterator
, seq
));
6303 cpumask_clear(iter
->started
);
6304 trace_seq_init(&iter
->seq
);
6307 trace_event_read_lock();
6308 trace_access_lock(iter
->cpu_file
);
6309 while (trace_find_next_entry_inc(iter
) != NULL
) {
6310 enum print_line_t ret
;
6311 int save_len
= iter
->seq
.seq
.len
;
6313 ret
= print_trace_line(iter
);
6314 if (ret
== TRACE_TYPE_PARTIAL_LINE
) {
6315 /* don't print partial lines */
6316 iter
->seq
.seq
.len
= save_len
;
6319 if (ret
!= TRACE_TYPE_NO_CONSUME
)
6320 trace_consume(iter
);
6322 if (trace_seq_used(&iter
->seq
) >= cnt
)
6326 * Setting the full flag means we reached the trace_seq buffer
6327 * size and we should leave by partial output condition above.
6328 * One of the trace_seq_* functions is not used properly.
6330 WARN_ONCE(iter
->seq
.full
, "full flag set for trace type %d",
6333 trace_access_unlock(iter
->cpu_file
);
6334 trace_event_read_unlock();
6336 /* Now copy what we have to the user */
6337 sret
= trace_seq_to_user(&iter
->seq
, ubuf
, cnt
);
6338 if (iter
->seq
.seq
.readpos
>= trace_seq_used(&iter
->seq
))
6339 trace_seq_init(&iter
->seq
);
6342 * If there was nothing to send to user, in spite of consuming trace
6343 * entries, go back to wait for more entries.
6349 mutex_unlock(&iter
->mutex
);
6354 static void tracing_spd_release_pipe(struct splice_pipe_desc
*spd
,
6357 __free_page(spd
->pages
[idx
]);
6361 tracing_fill_pipe_page(size_t rem
, struct trace_iterator
*iter
)
6367 /* Seq buffer is page-sized, exactly what we need. */
6369 save_len
= iter
->seq
.seq
.len
;
6370 ret
= print_trace_line(iter
);
6372 if (trace_seq_has_overflowed(&iter
->seq
)) {
6373 iter
->seq
.seq
.len
= save_len
;
6378 * This should not be hit, because it should only
6379 * be set if the iter->seq overflowed. But check it
6380 * anyway to be safe.
6382 if (ret
== TRACE_TYPE_PARTIAL_LINE
) {
6383 iter
->seq
.seq
.len
= save_len
;
6387 count
= trace_seq_used(&iter
->seq
) - save_len
;
6390 iter
->seq
.seq
.len
= save_len
;
6394 if (ret
!= TRACE_TYPE_NO_CONSUME
)
6395 trace_consume(iter
);
6397 if (!trace_find_next_entry_inc(iter
)) {
6407 static ssize_t
tracing_splice_read_pipe(struct file
*filp
,
6409 struct pipe_inode_info
*pipe
,
6413 struct page
*pages_def
[PIPE_DEF_BUFFERS
];
6414 struct partial_page partial_def
[PIPE_DEF_BUFFERS
];
6415 struct trace_iterator
*iter
= filp
->private_data
;
6416 struct splice_pipe_desc spd
= {
6418 .partial
= partial_def
,
6419 .nr_pages
= 0, /* This gets updated below. */
6420 .nr_pages_max
= PIPE_DEF_BUFFERS
,
6421 .ops
= &default_pipe_buf_ops
,
6422 .spd_release
= tracing_spd_release_pipe
,
6428 if (splice_grow_spd(pipe
, &spd
))
6431 mutex_lock(&iter
->mutex
);
6433 if (iter
->trace
->splice_read
) {
6434 ret
= iter
->trace
->splice_read(iter
, filp
,
6435 ppos
, pipe
, len
, flags
);
6440 ret
= tracing_wait_pipe(filp
);
6444 if (!iter
->ent
&& !trace_find_next_entry_inc(iter
)) {
6449 trace_event_read_lock();
6450 trace_access_lock(iter
->cpu_file
);
6452 /* Fill as many pages as possible. */
6453 for (i
= 0, rem
= len
; i
< spd
.nr_pages_max
&& rem
; i
++) {
6454 spd
.pages
[i
] = alloc_page(GFP_KERNEL
);
6458 rem
= tracing_fill_pipe_page(rem
, iter
);
6460 /* Copy the data into the page, so we can start over. */
6461 ret
= trace_seq_to_buffer(&iter
->seq
,
6462 page_address(spd
.pages
[i
]),
6463 trace_seq_used(&iter
->seq
));
6465 __free_page(spd
.pages
[i
]);
6468 spd
.partial
[i
].offset
= 0;
6469 spd
.partial
[i
].len
= trace_seq_used(&iter
->seq
);
6471 trace_seq_init(&iter
->seq
);
6474 trace_access_unlock(iter
->cpu_file
);
6475 trace_event_read_unlock();
6476 mutex_unlock(&iter
->mutex
);
6481 ret
= splice_to_pipe(pipe
, &spd
);
6485 splice_shrink_spd(&spd
);
6489 mutex_unlock(&iter
->mutex
);
6494 tracing_entries_read(struct file
*filp
, char __user
*ubuf
,
6495 size_t cnt
, loff_t
*ppos
)
6497 struct inode
*inode
= file_inode(filp
);
6498 struct trace_array
*tr
= inode
->i_private
;
6499 int cpu
= tracing_get_cpu(inode
);
6504 mutex_lock(&trace_types_lock
);
6506 if (cpu
== RING_BUFFER_ALL_CPUS
) {
6507 int cpu
, buf_size_same
;
6512 /* check if all cpu sizes are same */
6513 for_each_tracing_cpu(cpu
) {
6514 /* fill in the size from first enabled cpu */
6516 size
= per_cpu_ptr(tr
->array_buffer
.data
, cpu
)->entries
;
6517 if (size
!= per_cpu_ptr(tr
->array_buffer
.data
, cpu
)->entries
) {
6523 if (buf_size_same
) {
6524 if (!ring_buffer_expanded
)
6525 r
= sprintf(buf
, "%lu (expanded: %lu)\n",
6527 trace_buf_size
>> 10);
6529 r
= sprintf(buf
, "%lu\n", size
>> 10);
6531 r
= sprintf(buf
, "X\n");
6533 r
= sprintf(buf
, "%lu\n", per_cpu_ptr(tr
->array_buffer
.data
, cpu
)->entries
>> 10);
6535 mutex_unlock(&trace_types_lock
);
6537 ret
= simple_read_from_buffer(ubuf
, cnt
, ppos
, buf
, r
);
6542 tracing_entries_write(struct file
*filp
, const char __user
*ubuf
,
6543 size_t cnt
, loff_t
*ppos
)
6545 struct inode
*inode
= file_inode(filp
);
6546 struct trace_array
*tr
= inode
->i_private
;
6550 ret
= kstrtoul_from_user(ubuf
, cnt
, 10, &val
);
6554 /* must have at least 1 entry */
6558 /* value is in KB */
6560 ret
= tracing_resize_ring_buffer(tr
, val
, tracing_get_cpu(inode
));
6570 tracing_total_entries_read(struct file
*filp
, char __user
*ubuf
,
6571 size_t cnt
, loff_t
*ppos
)
6573 struct trace_array
*tr
= filp
->private_data
;
6576 unsigned long size
= 0, expanded_size
= 0;
6578 mutex_lock(&trace_types_lock
);
6579 for_each_tracing_cpu(cpu
) {
6580 size
+= per_cpu_ptr(tr
->array_buffer
.data
, cpu
)->entries
>> 10;
6581 if (!ring_buffer_expanded
)
6582 expanded_size
+= trace_buf_size
>> 10;
6584 if (ring_buffer_expanded
)
6585 r
= sprintf(buf
, "%lu\n", size
);
6587 r
= sprintf(buf
, "%lu (expanded: %lu)\n", size
, expanded_size
);
6588 mutex_unlock(&trace_types_lock
);
6590 return simple_read_from_buffer(ubuf
, cnt
, ppos
, buf
, r
);
6594 tracing_free_buffer_write(struct file
*filp
, const char __user
*ubuf
,
6595 size_t cnt
, loff_t
*ppos
)
6598 * There is no need to read what the user has written, this function
6599 * is just to make sure that there is no error when "echo" is used
6608 tracing_free_buffer_release(struct inode
*inode
, struct file
*filp
)
6610 struct trace_array
*tr
= inode
->i_private
;
6612 /* disable tracing ? */
6613 if (tr
->trace_flags
& TRACE_ITER_STOP_ON_FREE
)
6614 tracer_tracing_off(tr
);
6615 /* resize the ring buffer to 0 */
6616 tracing_resize_ring_buffer(tr
, 0, RING_BUFFER_ALL_CPUS
);
6618 trace_array_put(tr
);
6624 tracing_mark_write(struct file
*filp
, const char __user
*ubuf
,
6625 size_t cnt
, loff_t
*fpos
)
6627 struct trace_array
*tr
= filp
->private_data
;
6628 struct ring_buffer_event
*event
;
6629 enum event_trigger_type tt
= ETT_NONE
;
6630 struct trace_buffer
*buffer
;
6631 struct print_entry
*entry
;
6632 unsigned long irq_flags
;
6637 /* Used in tracing_mark_raw_write() as well */
6638 #define FAULTED_STR "<faulted>"
6639 #define FAULTED_SIZE (sizeof(FAULTED_STR) - 1) /* '\0' is already accounted for */
6641 if (tracing_disabled
)
6644 if (!(tr
->trace_flags
& TRACE_ITER_MARKERS
))
6647 if (cnt
> TRACE_BUF_SIZE
)
6648 cnt
= TRACE_BUF_SIZE
;
6650 BUILD_BUG_ON(TRACE_BUF_SIZE
>= PAGE_SIZE
);
6652 local_save_flags(irq_flags
);
6653 size
= sizeof(*entry
) + cnt
+ 2; /* add '\0' and possible '\n' */
6655 /* If less than "<faulted>", then make sure we can still add that */
6656 if (cnt
< FAULTED_SIZE
)
6657 size
+= FAULTED_SIZE
- cnt
;
6659 buffer
= tr
->array_buffer
.buffer
;
6660 event
= __trace_buffer_lock_reserve(buffer
, TRACE_PRINT
, size
,
6661 irq_flags
, preempt_count());
6662 if (unlikely(!event
))
6663 /* Ring buffer disabled, return as if not open for write */
6666 entry
= ring_buffer_event_data(event
);
6667 entry
->ip
= _THIS_IP_
;
6669 len
= __copy_from_user_inatomic(&entry
->buf
, ubuf
, cnt
);
6671 memcpy(&entry
->buf
, FAULTED_STR
, FAULTED_SIZE
);
6678 if (tr
->trace_marker_file
&& !list_empty(&tr
->trace_marker_file
->triggers
)) {
6679 /* do not add \n before testing triggers, but add \0 */
6680 entry
->buf
[cnt
] = '\0';
6681 tt
= event_triggers_call(tr
->trace_marker_file
, entry
, event
);
6684 if (entry
->buf
[cnt
- 1] != '\n') {
6685 entry
->buf
[cnt
] = '\n';
6686 entry
->buf
[cnt
+ 1] = '\0';
6688 entry
->buf
[cnt
] = '\0';
6690 __buffer_unlock_commit(buffer
, event
);
6693 event_triggers_post_call(tr
->trace_marker_file
, tt
);
6701 /* Limit it for now to 3K (including tag) */
6702 #define RAW_DATA_MAX_SIZE (1024*3)
6705 tracing_mark_raw_write(struct file
*filp
, const char __user
*ubuf
,
6706 size_t cnt
, loff_t
*fpos
)
6708 struct trace_array
*tr
= filp
->private_data
;
6709 struct ring_buffer_event
*event
;
6710 struct trace_buffer
*buffer
;
6711 struct raw_data_entry
*entry
;
6712 unsigned long irq_flags
;
6717 #define FAULT_SIZE_ID (FAULTED_SIZE + sizeof(int))
6719 if (tracing_disabled
)
6722 if (!(tr
->trace_flags
& TRACE_ITER_MARKERS
))
6725 /* The marker must at least have a tag id */
6726 if (cnt
< sizeof(unsigned int) || cnt
> RAW_DATA_MAX_SIZE
)
6729 if (cnt
> TRACE_BUF_SIZE
)
6730 cnt
= TRACE_BUF_SIZE
;
6732 BUILD_BUG_ON(TRACE_BUF_SIZE
>= PAGE_SIZE
);
6734 local_save_flags(irq_flags
);
6735 size
= sizeof(*entry
) + cnt
;
6736 if (cnt
< FAULT_SIZE_ID
)
6737 size
+= FAULT_SIZE_ID
- cnt
;
6739 buffer
= tr
->array_buffer
.buffer
;
6740 event
= __trace_buffer_lock_reserve(buffer
, TRACE_RAW_DATA
, size
,
6741 irq_flags
, preempt_count());
6743 /* Ring buffer disabled, return as if not open for write */
6746 entry
= ring_buffer_event_data(event
);
6748 len
= __copy_from_user_inatomic(&entry
->id
, ubuf
, cnt
);
6751 memcpy(&entry
->buf
, FAULTED_STR
, FAULTED_SIZE
);
6756 __buffer_unlock_commit(buffer
, event
);
6764 static int tracing_clock_show(struct seq_file
*m
, void *v
)
6766 struct trace_array
*tr
= m
->private;
6769 for (i
= 0; i
< ARRAY_SIZE(trace_clocks
); i
++)
6771 "%s%s%s%s", i
? " " : "",
6772 i
== tr
->clock_id
? "[" : "", trace_clocks
[i
].name
,
6773 i
== tr
->clock_id
? "]" : "");
6779 int tracing_set_clock(struct trace_array
*tr
, const char *clockstr
)
6783 for (i
= 0; i
< ARRAY_SIZE(trace_clocks
); i
++) {
6784 if (strcmp(trace_clocks
[i
].name
, clockstr
) == 0)
6787 if (i
== ARRAY_SIZE(trace_clocks
))
6790 mutex_lock(&trace_types_lock
);
6794 ring_buffer_set_clock(tr
->array_buffer
.buffer
, trace_clocks
[i
].func
);
6797 * New clock may not be consistent with the previous clock.
6798 * Reset the buffer so that it doesn't have incomparable timestamps.
6800 tracing_reset_online_cpus(&tr
->array_buffer
);
6802 #ifdef CONFIG_TRACER_MAX_TRACE
6803 if (tr
->max_buffer
.buffer
)
6804 ring_buffer_set_clock(tr
->max_buffer
.buffer
, trace_clocks
[i
].func
);
6805 tracing_reset_online_cpus(&tr
->max_buffer
);
6808 mutex_unlock(&trace_types_lock
);
6813 static ssize_t
tracing_clock_write(struct file
*filp
, const char __user
*ubuf
,
6814 size_t cnt
, loff_t
*fpos
)
6816 struct seq_file
*m
= filp
->private_data
;
6817 struct trace_array
*tr
= m
->private;
6819 const char *clockstr
;
6822 if (cnt
>= sizeof(buf
))
6825 if (copy_from_user(buf
, ubuf
, cnt
))
6830 clockstr
= strstrip(buf
);
6832 ret
= tracing_set_clock(tr
, clockstr
);
6841 static int tracing_clock_open(struct inode
*inode
, struct file
*file
)
6843 struct trace_array
*tr
= inode
->i_private
;
6846 ret
= tracing_check_open_get_tr(tr
);
6850 ret
= single_open(file
, tracing_clock_show
, inode
->i_private
);
6852 trace_array_put(tr
);
6857 static int tracing_time_stamp_mode_show(struct seq_file
*m
, void *v
)
6859 struct trace_array
*tr
= m
->private;
6861 mutex_lock(&trace_types_lock
);
6863 if (ring_buffer_time_stamp_abs(tr
->array_buffer
.buffer
))
6864 seq_puts(m
, "delta [absolute]\n");
6866 seq_puts(m
, "[delta] absolute\n");
6868 mutex_unlock(&trace_types_lock
);
6873 static int tracing_time_stamp_mode_open(struct inode
*inode
, struct file
*file
)
6875 struct trace_array
*tr
= inode
->i_private
;
6878 ret
= tracing_check_open_get_tr(tr
);
6882 ret
= single_open(file
, tracing_time_stamp_mode_show
, inode
->i_private
);
6884 trace_array_put(tr
);
6889 int tracing_set_time_stamp_abs(struct trace_array
*tr
, bool abs
)
6893 mutex_lock(&trace_types_lock
);
6895 if (abs
&& tr
->time_stamp_abs_ref
++)
6899 if (WARN_ON_ONCE(!tr
->time_stamp_abs_ref
)) {
6904 if (--tr
->time_stamp_abs_ref
)
6908 ring_buffer_set_time_stamp_abs(tr
->array_buffer
.buffer
, abs
);
6910 #ifdef CONFIG_TRACER_MAX_TRACE
6911 if (tr
->max_buffer
.buffer
)
6912 ring_buffer_set_time_stamp_abs(tr
->max_buffer
.buffer
, abs
);
6915 mutex_unlock(&trace_types_lock
);
6920 struct ftrace_buffer_info
{
6921 struct trace_iterator iter
;
6923 unsigned int spare_cpu
;
6927 #ifdef CONFIG_TRACER_SNAPSHOT
6928 static int tracing_snapshot_open(struct inode
*inode
, struct file
*file
)
6930 struct trace_array
*tr
= inode
->i_private
;
6931 struct trace_iterator
*iter
;
6935 ret
= tracing_check_open_get_tr(tr
);
6939 if (file
->f_mode
& FMODE_READ
) {
6940 iter
= __tracing_open(inode
, file
, true);
6942 ret
= PTR_ERR(iter
);
6944 /* Writes still need the seq_file to hold the private data */
6946 m
= kzalloc(sizeof(*m
), GFP_KERNEL
);
6949 iter
= kzalloc(sizeof(*iter
), GFP_KERNEL
);
6957 iter
->array_buffer
= &tr
->max_buffer
;
6958 iter
->cpu_file
= tracing_get_cpu(inode
);
6960 file
->private_data
= m
;
6964 trace_array_put(tr
);
6970 tracing_snapshot_write(struct file
*filp
, const char __user
*ubuf
, size_t cnt
,
6973 struct seq_file
*m
= filp
->private_data
;
6974 struct trace_iterator
*iter
= m
->private;
6975 struct trace_array
*tr
= iter
->tr
;
6979 ret
= tracing_update_buffers();
6983 ret
= kstrtoul_from_user(ubuf
, cnt
, 10, &val
);
6987 mutex_lock(&trace_types_lock
);
6989 if (tr
->current_trace
->use_max_tr
) {
6994 arch_spin_lock(&tr
->max_lock
);
6995 if (tr
->cond_snapshot
)
6997 arch_spin_unlock(&tr
->max_lock
);
7003 if (iter
->cpu_file
!= RING_BUFFER_ALL_CPUS
) {
7007 if (tr
->allocated_snapshot
)
7011 /* Only allow per-cpu swap if the ring buffer supports it */
7012 #ifndef CONFIG_RING_BUFFER_ALLOW_SWAP
7013 if (iter
->cpu_file
!= RING_BUFFER_ALL_CPUS
) {
7018 if (tr
->allocated_snapshot
)
7019 ret
= resize_buffer_duplicate_size(&tr
->max_buffer
,
7020 &tr
->array_buffer
, iter
->cpu_file
);
7022 ret
= tracing_alloc_snapshot_instance(tr
);
7025 local_irq_disable();
7026 /* Now, we're going to swap */
7027 if (iter
->cpu_file
== RING_BUFFER_ALL_CPUS
)
7028 update_max_tr(tr
, current
, smp_processor_id(), NULL
);
7030 update_max_tr_single(tr
, current
, iter
->cpu_file
);
7034 if (tr
->allocated_snapshot
) {
7035 if (iter
->cpu_file
== RING_BUFFER_ALL_CPUS
)
7036 tracing_reset_online_cpus(&tr
->max_buffer
);
7038 tracing_reset_cpu(&tr
->max_buffer
, iter
->cpu_file
);
7048 mutex_unlock(&trace_types_lock
);
7052 static int tracing_snapshot_release(struct inode
*inode
, struct file
*file
)
7054 struct seq_file
*m
= file
->private_data
;
7057 ret
= tracing_release(inode
, file
);
7059 if (file
->f_mode
& FMODE_READ
)
7062 /* If write only, the seq_file is just a stub */
7070 static int tracing_buffers_open(struct inode
*inode
, struct file
*filp
);
7071 static ssize_t
tracing_buffers_read(struct file
*filp
, char __user
*ubuf
,
7072 size_t count
, loff_t
*ppos
);
7073 static int tracing_buffers_release(struct inode
*inode
, struct file
*file
);
7074 static ssize_t
tracing_buffers_splice_read(struct file
*file
, loff_t
*ppos
,
7075 struct pipe_inode_info
*pipe
, size_t len
, unsigned int flags
);
7077 static int snapshot_raw_open(struct inode
*inode
, struct file
*filp
)
7079 struct ftrace_buffer_info
*info
;
7082 /* The following checks for tracefs lockdown */
7083 ret
= tracing_buffers_open(inode
, filp
);
7087 info
= filp
->private_data
;
7089 if (info
->iter
.trace
->use_max_tr
) {
7090 tracing_buffers_release(inode
, filp
);
7094 info
->iter
.snapshot
= true;
7095 info
->iter
.array_buffer
= &info
->iter
.tr
->max_buffer
;
7100 #endif /* CONFIG_TRACER_SNAPSHOT */
7103 static const struct file_operations tracing_thresh_fops
= {
7104 .open
= tracing_open_generic
,
7105 .read
= tracing_thresh_read
,
7106 .write
= tracing_thresh_write
,
7107 .llseek
= generic_file_llseek
,
7110 #if defined(CONFIG_TRACER_MAX_TRACE) || defined(CONFIG_HWLAT_TRACER)
7111 static const struct file_operations tracing_max_lat_fops
= {
7112 .open
= tracing_open_generic
,
7113 .read
= tracing_max_lat_read
,
7114 .write
= tracing_max_lat_write
,
7115 .llseek
= generic_file_llseek
,
7119 static const struct file_operations set_tracer_fops
= {
7120 .open
= tracing_open_generic
,
7121 .read
= tracing_set_trace_read
,
7122 .write
= tracing_set_trace_write
,
7123 .llseek
= generic_file_llseek
,
7126 static const struct file_operations tracing_pipe_fops
= {
7127 .open
= tracing_open_pipe
,
7128 .poll
= tracing_poll_pipe
,
7129 .read
= tracing_read_pipe
,
7130 .splice_read
= tracing_splice_read_pipe
,
7131 .release
= tracing_release_pipe
,
7132 .llseek
= no_llseek
,
7135 static const struct file_operations tracing_entries_fops
= {
7136 .open
= tracing_open_generic_tr
,
7137 .read
= tracing_entries_read
,
7138 .write
= tracing_entries_write
,
7139 .llseek
= generic_file_llseek
,
7140 .release
= tracing_release_generic_tr
,
7143 static const struct file_operations tracing_total_entries_fops
= {
7144 .open
= tracing_open_generic_tr
,
7145 .read
= tracing_total_entries_read
,
7146 .llseek
= generic_file_llseek
,
7147 .release
= tracing_release_generic_tr
,
7150 static const struct file_operations tracing_free_buffer_fops
= {
7151 .open
= tracing_open_generic_tr
,
7152 .write
= tracing_free_buffer_write
,
7153 .release
= tracing_free_buffer_release
,
7156 static const struct file_operations tracing_mark_fops
= {
7157 .open
= tracing_open_generic_tr
,
7158 .write
= tracing_mark_write
,
7159 .llseek
= generic_file_llseek
,
7160 .release
= tracing_release_generic_tr
,
7163 static const struct file_operations tracing_mark_raw_fops
= {
7164 .open
= tracing_open_generic_tr
,
7165 .write
= tracing_mark_raw_write
,
7166 .llseek
= generic_file_llseek
,
7167 .release
= tracing_release_generic_tr
,
7170 static const struct file_operations trace_clock_fops
= {
7171 .open
= tracing_clock_open
,
7173 .llseek
= seq_lseek
,
7174 .release
= tracing_single_release_tr
,
7175 .write
= tracing_clock_write
,
7178 static const struct file_operations trace_time_stamp_mode_fops
= {
7179 .open
= tracing_time_stamp_mode_open
,
7181 .llseek
= seq_lseek
,
7182 .release
= tracing_single_release_tr
,
7185 #ifdef CONFIG_TRACER_SNAPSHOT
7186 static const struct file_operations snapshot_fops
= {
7187 .open
= tracing_snapshot_open
,
7189 .write
= tracing_snapshot_write
,
7190 .llseek
= tracing_lseek
,
7191 .release
= tracing_snapshot_release
,
7194 static const struct file_operations snapshot_raw_fops
= {
7195 .open
= snapshot_raw_open
,
7196 .read
= tracing_buffers_read
,
7197 .release
= tracing_buffers_release
,
7198 .splice_read
= tracing_buffers_splice_read
,
7199 .llseek
= no_llseek
,
7202 #endif /* CONFIG_TRACER_SNAPSHOT */
7204 #define TRACING_LOG_ERRS_MAX 8
7205 #define TRACING_LOG_LOC_MAX 128
7207 #define CMD_PREFIX " Command: "
7210 const char **errs
; /* ptr to loc-specific array of err strings */
7211 u8 type
; /* index into errs -> specific err string */
7212 u8 pos
; /* MAX_FILTER_STR_VAL = 256 */
7216 struct tracing_log_err
{
7217 struct list_head list
;
7218 struct err_info info
;
7219 char loc
[TRACING_LOG_LOC_MAX
]; /* err location */
7220 char cmd
[MAX_FILTER_STR_VAL
]; /* what caused err */
7223 static DEFINE_MUTEX(tracing_err_log_lock
);
7225 static struct tracing_log_err
*get_tracing_log_err(struct trace_array
*tr
)
7227 struct tracing_log_err
*err
;
7229 if (tr
->n_err_log_entries
< TRACING_LOG_ERRS_MAX
) {
7230 err
= kzalloc(sizeof(*err
), GFP_KERNEL
);
7232 err
= ERR_PTR(-ENOMEM
);
7233 tr
->n_err_log_entries
++;
7238 err
= list_first_entry(&tr
->err_log
, struct tracing_log_err
, list
);
7239 list_del(&err
->list
);
7245 * err_pos - find the position of a string within a command for error careting
7246 * @cmd: The tracing command that caused the error
7247 * @str: The string to position the caret at within @cmd
7249 * Finds the position of the first occurence of @str within @cmd. The
7250 * return value can be passed to tracing_log_err() for caret placement
7253 * Returns the index within @cmd of the first occurence of @str or 0
7254 * if @str was not found.
7256 unsigned int err_pos(char *cmd
, const char *str
)
7260 if (WARN_ON(!strlen(cmd
)))
7263 found
= strstr(cmd
, str
);
7271 * tracing_log_err - write an error to the tracing error log
7272 * @tr: The associated trace array for the error (NULL for top level array)
7273 * @loc: A string describing where the error occurred
7274 * @cmd: The tracing command that caused the error
7275 * @errs: The array of loc-specific static error strings
7276 * @type: The index into errs[], which produces the specific static err string
7277 * @pos: The position the caret should be placed in the cmd
7279 * Writes an error into tracing/error_log of the form:
7281 * <loc>: error: <text>
7285 * tracing/error_log is a small log file containing the last
7286 * TRACING_LOG_ERRS_MAX errors (8). Memory for errors isn't allocated
7287 * unless there has been a tracing error, and the error log can be
7288 * cleared and have its memory freed by writing the empty string in
7289 * truncation mode to it i.e. echo > tracing/error_log.
7291 * NOTE: the @errs array along with the @type param are used to
7292 * produce a static error string - this string is not copied and saved
7293 * when the error is logged - only a pointer to it is saved. See
7294 * existing callers for examples of how static strings are typically
7295 * defined for use with tracing_log_err().
7297 void tracing_log_err(struct trace_array
*tr
,
7298 const char *loc
, const char *cmd
,
7299 const char **errs
, u8 type
, u8 pos
)
7301 struct tracing_log_err
*err
;
7306 mutex_lock(&tracing_err_log_lock
);
7307 err
= get_tracing_log_err(tr
);
7308 if (PTR_ERR(err
) == -ENOMEM
) {
7309 mutex_unlock(&tracing_err_log_lock
);
7313 snprintf(err
->loc
, TRACING_LOG_LOC_MAX
, "%s: error: ", loc
);
7314 snprintf(err
->cmd
, MAX_FILTER_STR_VAL
,"\n" CMD_PREFIX
"%s\n", cmd
);
7316 err
->info
.errs
= errs
;
7317 err
->info
.type
= type
;
7318 err
->info
.pos
= pos
;
7319 err
->info
.ts
= local_clock();
7321 list_add_tail(&err
->list
, &tr
->err_log
);
7322 mutex_unlock(&tracing_err_log_lock
);
7325 static void clear_tracing_err_log(struct trace_array
*tr
)
7327 struct tracing_log_err
*err
, *next
;
7329 mutex_lock(&tracing_err_log_lock
);
7330 list_for_each_entry_safe(err
, next
, &tr
->err_log
, list
) {
7331 list_del(&err
->list
);
7335 tr
->n_err_log_entries
= 0;
7336 mutex_unlock(&tracing_err_log_lock
);
7339 static void *tracing_err_log_seq_start(struct seq_file
*m
, loff_t
*pos
)
7341 struct trace_array
*tr
= m
->private;
7343 mutex_lock(&tracing_err_log_lock
);
7345 return seq_list_start(&tr
->err_log
, *pos
);
7348 static void *tracing_err_log_seq_next(struct seq_file
*m
, void *v
, loff_t
*pos
)
7350 struct trace_array
*tr
= m
->private;
7352 return seq_list_next(v
, &tr
->err_log
, pos
);
7355 static void tracing_err_log_seq_stop(struct seq_file
*m
, void *v
)
7357 mutex_unlock(&tracing_err_log_lock
);
7360 static void tracing_err_log_show_pos(struct seq_file
*m
, u8 pos
)
7364 for (i
= 0; i
< sizeof(CMD_PREFIX
) - 1; i
++)
7366 for (i
= 0; i
< pos
; i
++)
7371 static int tracing_err_log_seq_show(struct seq_file
*m
, void *v
)
7373 struct tracing_log_err
*err
= v
;
7376 const char *err_text
= err
->info
.errs
[err
->info
.type
];
7377 u64 sec
= err
->info
.ts
;
7380 nsec
= do_div(sec
, NSEC_PER_SEC
);
7381 seq_printf(m
, "[%5llu.%06u] %s%s", sec
, nsec
/ 1000,
7382 err
->loc
, err_text
);
7383 seq_printf(m
, "%s", err
->cmd
);
7384 tracing_err_log_show_pos(m
, err
->info
.pos
);
7390 static const struct seq_operations tracing_err_log_seq_ops
= {
7391 .start
= tracing_err_log_seq_start
,
7392 .next
= tracing_err_log_seq_next
,
7393 .stop
= tracing_err_log_seq_stop
,
7394 .show
= tracing_err_log_seq_show
7397 static int tracing_err_log_open(struct inode
*inode
, struct file
*file
)
7399 struct trace_array
*tr
= inode
->i_private
;
7402 ret
= tracing_check_open_get_tr(tr
);
7406 /* If this file was opened for write, then erase contents */
7407 if ((file
->f_mode
& FMODE_WRITE
) && (file
->f_flags
& O_TRUNC
))
7408 clear_tracing_err_log(tr
);
7410 if (file
->f_mode
& FMODE_READ
) {
7411 ret
= seq_open(file
, &tracing_err_log_seq_ops
);
7413 struct seq_file
*m
= file
->private_data
;
7416 trace_array_put(tr
);
7422 static ssize_t
tracing_err_log_write(struct file
*file
,
7423 const char __user
*buffer
,
7424 size_t count
, loff_t
*ppos
)
7429 static int tracing_err_log_release(struct inode
*inode
, struct file
*file
)
7431 struct trace_array
*tr
= inode
->i_private
;
7433 trace_array_put(tr
);
7435 if (file
->f_mode
& FMODE_READ
)
7436 seq_release(inode
, file
);
7441 static const struct file_operations tracing_err_log_fops
= {
7442 .open
= tracing_err_log_open
,
7443 .write
= tracing_err_log_write
,
7445 .llseek
= seq_lseek
,
7446 .release
= tracing_err_log_release
,
7449 static int tracing_buffers_open(struct inode
*inode
, struct file
*filp
)
7451 struct trace_array
*tr
= inode
->i_private
;
7452 struct ftrace_buffer_info
*info
;
7455 ret
= tracing_check_open_get_tr(tr
);
7459 info
= kvzalloc(sizeof(*info
), GFP_KERNEL
);
7461 trace_array_put(tr
);
7465 mutex_lock(&trace_types_lock
);
7468 info
->iter
.cpu_file
= tracing_get_cpu(inode
);
7469 info
->iter
.trace
= tr
->current_trace
;
7470 info
->iter
.array_buffer
= &tr
->array_buffer
;
7472 /* Force reading ring buffer for first read */
7473 info
->read
= (unsigned int)-1;
7475 filp
->private_data
= info
;
7479 mutex_unlock(&trace_types_lock
);
7481 ret
= nonseekable_open(inode
, filp
);
7483 trace_array_put(tr
);
7489 tracing_buffers_poll(struct file
*filp
, poll_table
*poll_table
)
7491 struct ftrace_buffer_info
*info
= filp
->private_data
;
7492 struct trace_iterator
*iter
= &info
->iter
;
7494 return trace_poll(iter
, filp
, poll_table
);
7498 tracing_buffers_read(struct file
*filp
, char __user
*ubuf
,
7499 size_t count
, loff_t
*ppos
)
7501 struct ftrace_buffer_info
*info
= filp
->private_data
;
7502 struct trace_iterator
*iter
= &info
->iter
;
7509 #ifdef CONFIG_TRACER_MAX_TRACE
7510 if (iter
->snapshot
&& iter
->tr
->current_trace
->use_max_tr
)
7515 info
->spare
= ring_buffer_alloc_read_page(iter
->array_buffer
->buffer
,
7517 if (IS_ERR(info
->spare
)) {
7518 ret
= PTR_ERR(info
->spare
);
7521 info
->spare_cpu
= iter
->cpu_file
;
7527 /* Do we have previous read data to read? */
7528 if (info
->read
< PAGE_SIZE
)
7532 trace_access_lock(iter
->cpu_file
);
7533 ret
= ring_buffer_read_page(iter
->array_buffer
->buffer
,
7537 trace_access_unlock(iter
->cpu_file
);
7540 if (trace_empty(iter
)) {
7541 if ((filp
->f_flags
& O_NONBLOCK
))
7544 ret
= wait_on_pipe(iter
, 0);
7555 size
= PAGE_SIZE
- info
->read
;
7559 ret
= copy_to_user(ubuf
, info
->spare
+ info
->read
, size
);
7571 static int tracing_buffers_release(struct inode
*inode
, struct file
*file
)
7573 struct ftrace_buffer_info
*info
= file
->private_data
;
7574 struct trace_iterator
*iter
= &info
->iter
;
7576 mutex_lock(&trace_types_lock
);
7578 iter
->tr
->trace_ref
--;
7580 __trace_array_put(iter
->tr
);
7583 ring_buffer_free_read_page(iter
->array_buffer
->buffer
,
7584 info
->spare_cpu
, info
->spare
);
7587 mutex_unlock(&trace_types_lock
);
7593 struct trace_buffer
*buffer
;
7596 refcount_t refcount
;
7599 static void buffer_ref_release(struct buffer_ref
*ref
)
7601 if (!refcount_dec_and_test(&ref
->refcount
))
7603 ring_buffer_free_read_page(ref
->buffer
, ref
->cpu
, ref
->page
);
7607 static void buffer_pipe_buf_release(struct pipe_inode_info
*pipe
,
7608 struct pipe_buffer
*buf
)
7610 struct buffer_ref
*ref
= (struct buffer_ref
*)buf
->private;
7612 buffer_ref_release(ref
);
7616 static bool buffer_pipe_buf_get(struct pipe_inode_info
*pipe
,
7617 struct pipe_buffer
*buf
)
7619 struct buffer_ref
*ref
= (struct buffer_ref
*)buf
->private;
7621 if (refcount_read(&ref
->refcount
) > INT_MAX
/2)
7624 refcount_inc(&ref
->refcount
);
7628 /* Pipe buffer operations for a buffer. */
7629 static const struct pipe_buf_operations buffer_pipe_buf_ops
= {
7630 .release
= buffer_pipe_buf_release
,
7631 .get
= buffer_pipe_buf_get
,
7635 * Callback from splice_to_pipe(), if we need to release some pages
7636 * at the end of the spd in case we error'ed out in filling the pipe.
7638 static void buffer_spd_release(struct splice_pipe_desc
*spd
, unsigned int i
)
7640 struct buffer_ref
*ref
=
7641 (struct buffer_ref
*)spd
->partial
[i
].private;
7643 buffer_ref_release(ref
);
7644 spd
->partial
[i
].private = 0;
7648 tracing_buffers_splice_read(struct file
*file
, loff_t
*ppos
,
7649 struct pipe_inode_info
*pipe
, size_t len
,
7652 struct ftrace_buffer_info
*info
= file
->private_data
;
7653 struct trace_iterator
*iter
= &info
->iter
;
7654 struct partial_page partial_def
[PIPE_DEF_BUFFERS
];
7655 struct page
*pages_def
[PIPE_DEF_BUFFERS
];
7656 struct splice_pipe_desc spd
= {
7658 .partial
= partial_def
,
7659 .nr_pages_max
= PIPE_DEF_BUFFERS
,
7660 .ops
= &buffer_pipe_buf_ops
,
7661 .spd_release
= buffer_spd_release
,
7663 struct buffer_ref
*ref
;
7667 #ifdef CONFIG_TRACER_MAX_TRACE
7668 if (iter
->snapshot
&& iter
->tr
->current_trace
->use_max_tr
)
7672 if (*ppos
& (PAGE_SIZE
- 1))
7675 if (len
& (PAGE_SIZE
- 1)) {
7676 if (len
< PAGE_SIZE
)
7681 if (splice_grow_spd(pipe
, &spd
))
7685 trace_access_lock(iter
->cpu_file
);
7686 entries
= ring_buffer_entries_cpu(iter
->array_buffer
->buffer
, iter
->cpu_file
);
7688 for (i
= 0; i
< spd
.nr_pages_max
&& len
&& entries
; i
++, len
-= PAGE_SIZE
) {
7692 ref
= kzalloc(sizeof(*ref
), GFP_KERNEL
);
7698 refcount_set(&ref
->refcount
, 1);
7699 ref
->buffer
= iter
->array_buffer
->buffer
;
7700 ref
->page
= ring_buffer_alloc_read_page(ref
->buffer
, iter
->cpu_file
);
7701 if (IS_ERR(ref
->page
)) {
7702 ret
= PTR_ERR(ref
->page
);
7707 ref
->cpu
= iter
->cpu_file
;
7709 r
= ring_buffer_read_page(ref
->buffer
, &ref
->page
,
7710 len
, iter
->cpu_file
, 1);
7712 ring_buffer_free_read_page(ref
->buffer
, ref
->cpu
,
7718 page
= virt_to_page(ref
->page
);
7720 spd
.pages
[i
] = page
;
7721 spd
.partial
[i
].len
= PAGE_SIZE
;
7722 spd
.partial
[i
].offset
= 0;
7723 spd
.partial
[i
].private = (unsigned long)ref
;
7727 entries
= ring_buffer_entries_cpu(iter
->array_buffer
->buffer
, iter
->cpu_file
);
7730 trace_access_unlock(iter
->cpu_file
);
7733 /* did we read anything? */
7734 if (!spd
.nr_pages
) {
7739 if ((file
->f_flags
& O_NONBLOCK
) || (flags
& SPLICE_F_NONBLOCK
))
7742 ret
= wait_on_pipe(iter
, iter
->tr
->buffer_percent
);
7749 ret
= splice_to_pipe(pipe
, &spd
);
7751 splice_shrink_spd(&spd
);
7756 static const struct file_operations tracing_buffers_fops
= {
7757 .open
= tracing_buffers_open
,
7758 .read
= tracing_buffers_read
,
7759 .poll
= tracing_buffers_poll
,
7760 .release
= tracing_buffers_release
,
7761 .splice_read
= tracing_buffers_splice_read
,
7762 .llseek
= no_llseek
,
7766 tracing_stats_read(struct file
*filp
, char __user
*ubuf
,
7767 size_t count
, loff_t
*ppos
)
7769 struct inode
*inode
= file_inode(filp
);
7770 struct trace_array
*tr
= inode
->i_private
;
7771 struct array_buffer
*trace_buf
= &tr
->array_buffer
;
7772 int cpu
= tracing_get_cpu(inode
);
7773 struct trace_seq
*s
;
7775 unsigned long long t
;
7776 unsigned long usec_rem
;
7778 s
= kmalloc(sizeof(*s
), GFP_KERNEL
);
7784 cnt
= ring_buffer_entries_cpu(trace_buf
->buffer
, cpu
);
7785 trace_seq_printf(s
, "entries: %ld\n", cnt
);
7787 cnt
= ring_buffer_overrun_cpu(trace_buf
->buffer
, cpu
);
7788 trace_seq_printf(s
, "overrun: %ld\n", cnt
);
7790 cnt
= ring_buffer_commit_overrun_cpu(trace_buf
->buffer
, cpu
);
7791 trace_seq_printf(s
, "commit overrun: %ld\n", cnt
);
7793 cnt
= ring_buffer_bytes_cpu(trace_buf
->buffer
, cpu
);
7794 trace_seq_printf(s
, "bytes: %ld\n", cnt
);
7796 if (trace_clocks
[tr
->clock_id
].in_ns
) {
7797 /* local or global for trace_clock */
7798 t
= ns2usecs(ring_buffer_oldest_event_ts(trace_buf
->buffer
, cpu
));
7799 usec_rem
= do_div(t
, USEC_PER_SEC
);
7800 trace_seq_printf(s
, "oldest event ts: %5llu.%06lu\n",
7803 t
= ns2usecs(ring_buffer_time_stamp(trace_buf
->buffer
, cpu
));
7804 usec_rem
= do_div(t
, USEC_PER_SEC
);
7805 trace_seq_printf(s
, "now ts: %5llu.%06lu\n", t
, usec_rem
);
7807 /* counter or tsc mode for trace_clock */
7808 trace_seq_printf(s
, "oldest event ts: %llu\n",
7809 ring_buffer_oldest_event_ts(trace_buf
->buffer
, cpu
));
7811 trace_seq_printf(s
, "now ts: %llu\n",
7812 ring_buffer_time_stamp(trace_buf
->buffer
, cpu
));
7815 cnt
= ring_buffer_dropped_events_cpu(trace_buf
->buffer
, cpu
);
7816 trace_seq_printf(s
, "dropped events: %ld\n", cnt
);
7818 cnt
= ring_buffer_read_events_cpu(trace_buf
->buffer
, cpu
);
7819 trace_seq_printf(s
, "read events: %ld\n", cnt
);
7821 count
= simple_read_from_buffer(ubuf
, count
, ppos
,
7822 s
->buffer
, trace_seq_used(s
));
7829 static const struct file_operations tracing_stats_fops
= {
7830 .open
= tracing_open_generic_tr
,
7831 .read
= tracing_stats_read
,
7832 .llseek
= generic_file_llseek
,
7833 .release
= tracing_release_generic_tr
,
7836 #ifdef CONFIG_DYNAMIC_FTRACE
7839 tracing_read_dyn_info(struct file
*filp
, char __user
*ubuf
,
7840 size_t cnt
, loff_t
*ppos
)
7846 /* 256 should be plenty to hold the amount needed */
7847 buf
= kmalloc(256, GFP_KERNEL
);
7851 r
= scnprintf(buf
, 256, "%ld pages:%ld groups: %ld\n",
7852 ftrace_update_tot_cnt
,
7853 ftrace_number_of_pages
,
7854 ftrace_number_of_groups
);
7856 ret
= simple_read_from_buffer(ubuf
, cnt
, ppos
, buf
, r
);
7861 static const struct file_operations tracing_dyn_info_fops
= {
7862 .open
= tracing_open_generic
,
7863 .read
= tracing_read_dyn_info
,
7864 .llseek
= generic_file_llseek
,
7866 #endif /* CONFIG_DYNAMIC_FTRACE */
7868 #if defined(CONFIG_TRACER_SNAPSHOT) && defined(CONFIG_DYNAMIC_FTRACE)
7870 ftrace_snapshot(unsigned long ip
, unsigned long parent_ip
,
7871 struct trace_array
*tr
, struct ftrace_probe_ops
*ops
,
7874 tracing_snapshot_instance(tr
);
7878 ftrace_count_snapshot(unsigned long ip
, unsigned long parent_ip
,
7879 struct trace_array
*tr
, struct ftrace_probe_ops
*ops
,
7882 struct ftrace_func_mapper
*mapper
= data
;
7886 count
= (long *)ftrace_func_mapper_find_ip(mapper
, ip
);
7896 tracing_snapshot_instance(tr
);
7900 ftrace_snapshot_print(struct seq_file
*m
, unsigned long ip
,
7901 struct ftrace_probe_ops
*ops
, void *data
)
7903 struct ftrace_func_mapper
*mapper
= data
;
7906 seq_printf(m
, "%ps:", (void *)ip
);
7908 seq_puts(m
, "snapshot");
7911 count
= (long *)ftrace_func_mapper_find_ip(mapper
, ip
);
7914 seq_printf(m
, ":count=%ld\n", *count
);
7916 seq_puts(m
, ":unlimited\n");
7922 ftrace_snapshot_init(struct ftrace_probe_ops
*ops
, struct trace_array
*tr
,
7923 unsigned long ip
, void *init_data
, void **data
)
7925 struct ftrace_func_mapper
*mapper
= *data
;
7928 mapper
= allocate_ftrace_func_mapper();
7934 return ftrace_func_mapper_add_ip(mapper
, ip
, init_data
);
7938 ftrace_snapshot_free(struct ftrace_probe_ops
*ops
, struct trace_array
*tr
,
7939 unsigned long ip
, void *data
)
7941 struct ftrace_func_mapper
*mapper
= data
;
7946 free_ftrace_func_mapper(mapper
, NULL
);
7950 ftrace_func_mapper_remove_ip(mapper
, ip
);
7953 static struct ftrace_probe_ops snapshot_probe_ops
= {
7954 .func
= ftrace_snapshot
,
7955 .print
= ftrace_snapshot_print
,
7958 static struct ftrace_probe_ops snapshot_count_probe_ops
= {
7959 .func
= ftrace_count_snapshot
,
7960 .print
= ftrace_snapshot_print
,
7961 .init
= ftrace_snapshot_init
,
7962 .free
= ftrace_snapshot_free
,
7966 ftrace_trace_snapshot_callback(struct trace_array
*tr
, struct ftrace_hash
*hash
,
7967 char *glob
, char *cmd
, char *param
, int enable
)
7969 struct ftrace_probe_ops
*ops
;
7970 void *count
= (void *)-1;
7977 /* hash funcs only work with set_ftrace_filter */
7981 ops
= param
? &snapshot_count_probe_ops
: &snapshot_probe_ops
;
7984 return unregister_ftrace_function_probe_func(glob
+1, tr
, ops
);
7989 number
= strsep(¶m
, ":");
7991 if (!strlen(number
))
7995 * We use the callback data field (which is a pointer)
7998 ret
= kstrtoul(number
, 0, (unsigned long *)&count
);
8003 ret
= tracing_alloc_snapshot_instance(tr
);
8007 ret
= register_ftrace_function_probe(glob
, tr
, ops
, count
);
8010 return ret
< 0 ? ret
: 0;
8013 static struct ftrace_func_command ftrace_snapshot_cmd
= {
8015 .func
= ftrace_trace_snapshot_callback
,
8018 static __init
int register_snapshot_cmd(void)
8020 return register_ftrace_command(&ftrace_snapshot_cmd
);
8023 static inline __init
int register_snapshot_cmd(void) { return 0; }
8024 #endif /* defined(CONFIG_TRACER_SNAPSHOT) && defined(CONFIG_DYNAMIC_FTRACE) */
8026 static struct dentry
*tracing_get_dentry(struct trace_array
*tr
)
8028 if (WARN_ON(!tr
->dir
))
8029 return ERR_PTR(-ENODEV
);
8031 /* Top directory uses NULL as the parent */
8032 if (tr
->flags
& TRACE_ARRAY_FL_GLOBAL
)
8035 /* All sub buffers have a descriptor */
8039 static struct dentry
*tracing_dentry_percpu(struct trace_array
*tr
, int cpu
)
8041 struct dentry
*d_tracer
;
8044 return tr
->percpu_dir
;
8046 d_tracer
= tracing_get_dentry(tr
);
8047 if (IS_ERR(d_tracer
))
8050 tr
->percpu_dir
= tracefs_create_dir("per_cpu", d_tracer
);
8052 MEM_FAIL(!tr
->percpu_dir
,
8053 "Could not create tracefs directory 'per_cpu/%d'\n", cpu
);
8055 return tr
->percpu_dir
;
8058 static struct dentry
*
8059 trace_create_cpu_file(const char *name
, umode_t mode
, struct dentry
*parent
,
8060 void *data
, long cpu
, const struct file_operations
*fops
)
8062 struct dentry
*ret
= trace_create_file(name
, mode
, parent
, data
, fops
);
8064 if (ret
) /* See tracing_get_cpu() */
8065 d_inode(ret
)->i_cdev
= (void *)(cpu
+ 1);
8070 tracing_init_tracefs_percpu(struct trace_array
*tr
, long cpu
)
8072 struct dentry
*d_percpu
= tracing_dentry_percpu(tr
, cpu
);
8073 struct dentry
*d_cpu
;
8074 char cpu_dir
[30]; /* 30 characters should be more than enough */
8079 snprintf(cpu_dir
, 30, "cpu%ld", cpu
);
8080 d_cpu
= tracefs_create_dir(cpu_dir
, d_percpu
);
8082 pr_warn("Could not create tracefs '%s' entry\n", cpu_dir
);
8086 /* per cpu trace_pipe */
8087 trace_create_cpu_file("trace_pipe", 0444, d_cpu
,
8088 tr
, cpu
, &tracing_pipe_fops
);
8091 trace_create_cpu_file("trace", 0644, d_cpu
,
8092 tr
, cpu
, &tracing_fops
);
8094 trace_create_cpu_file("trace_pipe_raw", 0444, d_cpu
,
8095 tr
, cpu
, &tracing_buffers_fops
);
8097 trace_create_cpu_file("stats", 0444, d_cpu
,
8098 tr
, cpu
, &tracing_stats_fops
);
8100 trace_create_cpu_file("buffer_size_kb", 0444, d_cpu
,
8101 tr
, cpu
, &tracing_entries_fops
);
8103 #ifdef CONFIG_TRACER_SNAPSHOT
8104 trace_create_cpu_file("snapshot", 0644, d_cpu
,
8105 tr
, cpu
, &snapshot_fops
);
8107 trace_create_cpu_file("snapshot_raw", 0444, d_cpu
,
8108 tr
, cpu
, &snapshot_raw_fops
);
8112 #ifdef CONFIG_FTRACE_SELFTEST
8113 /* Let selftest have access to static functions in this file */
8114 #include "trace_selftest.c"
8118 trace_options_read(struct file
*filp
, char __user
*ubuf
, size_t cnt
,
8121 struct trace_option_dentry
*topt
= filp
->private_data
;
8124 if (topt
->flags
->val
& topt
->opt
->bit
)
8129 return simple_read_from_buffer(ubuf
, cnt
, ppos
, buf
, 2);
8133 trace_options_write(struct file
*filp
, const char __user
*ubuf
, size_t cnt
,
8136 struct trace_option_dentry
*topt
= filp
->private_data
;
8140 ret
= kstrtoul_from_user(ubuf
, cnt
, 10, &val
);
8144 if (val
!= 0 && val
!= 1)
8147 if (!!(topt
->flags
->val
& topt
->opt
->bit
) != val
) {
8148 mutex_lock(&trace_types_lock
);
8149 ret
= __set_tracer_option(topt
->tr
, topt
->flags
,
8151 mutex_unlock(&trace_types_lock
);
8162 static const struct file_operations trace_options_fops
= {
8163 .open
= tracing_open_generic
,
8164 .read
= trace_options_read
,
8165 .write
= trace_options_write
,
8166 .llseek
= generic_file_llseek
,
8170 * In order to pass in both the trace_array descriptor as well as the index
8171 * to the flag that the trace option file represents, the trace_array
8172 * has a character array of trace_flags_index[], which holds the index
8173 * of the bit for the flag it represents. index[0] == 0, index[1] == 1, etc.
8174 * The address of this character array is passed to the flag option file
8175 * read/write callbacks.
8177 * In order to extract both the index and the trace_array descriptor,
8178 * get_tr_index() uses the following algorithm.
8182 * As the pointer itself contains the address of the index (remember
8185 * Then to get the trace_array descriptor, by subtracting that index
8186 * from the ptr, we get to the start of the index itself.
8188 * ptr - idx == &index[0]
8190 * Then a simple container_of() from that pointer gets us to the
8191 * trace_array descriptor.
8193 static void get_tr_index(void *data
, struct trace_array
**ptr
,
8194 unsigned int *pindex
)
8196 *pindex
= *(unsigned char *)data
;
8198 *ptr
= container_of(data
- *pindex
, struct trace_array
,
8203 trace_options_core_read(struct file
*filp
, char __user
*ubuf
, size_t cnt
,
8206 void *tr_index
= filp
->private_data
;
8207 struct trace_array
*tr
;
8211 get_tr_index(tr_index
, &tr
, &index
);
8213 if (tr
->trace_flags
& (1 << index
))
8218 return simple_read_from_buffer(ubuf
, cnt
, ppos
, buf
, 2);
8222 trace_options_core_write(struct file
*filp
, const char __user
*ubuf
, size_t cnt
,
8225 void *tr_index
= filp
->private_data
;
8226 struct trace_array
*tr
;
8231 get_tr_index(tr_index
, &tr
, &index
);
8233 ret
= kstrtoul_from_user(ubuf
, cnt
, 10, &val
);
8237 if (val
!= 0 && val
!= 1)
8240 mutex_lock(&event_mutex
);
8241 mutex_lock(&trace_types_lock
);
8242 ret
= set_tracer_flag(tr
, 1 << index
, val
);
8243 mutex_unlock(&trace_types_lock
);
8244 mutex_unlock(&event_mutex
);
8254 static const struct file_operations trace_options_core_fops
= {
8255 .open
= tracing_open_generic
,
8256 .read
= trace_options_core_read
,
8257 .write
= trace_options_core_write
,
8258 .llseek
= generic_file_llseek
,
8261 struct dentry
*trace_create_file(const char *name
,
8263 struct dentry
*parent
,
8265 const struct file_operations
*fops
)
8269 ret
= tracefs_create_file(name
, mode
, parent
, data
, fops
);
8271 pr_warn("Could not create tracefs '%s' entry\n", name
);
8277 static struct dentry
*trace_options_init_dentry(struct trace_array
*tr
)
8279 struct dentry
*d_tracer
;
8284 d_tracer
= tracing_get_dentry(tr
);
8285 if (IS_ERR(d_tracer
))
8288 tr
->options
= tracefs_create_dir("options", d_tracer
);
8290 pr_warn("Could not create tracefs directory 'options'\n");
8298 create_trace_option_file(struct trace_array
*tr
,
8299 struct trace_option_dentry
*topt
,
8300 struct tracer_flags
*flags
,
8301 struct tracer_opt
*opt
)
8303 struct dentry
*t_options
;
8305 t_options
= trace_options_init_dentry(tr
);
8309 topt
->flags
= flags
;
8313 topt
->entry
= trace_create_file(opt
->name
, 0644, t_options
, topt
,
8314 &trace_options_fops
);
8319 create_trace_option_files(struct trace_array
*tr
, struct tracer
*tracer
)
8321 struct trace_option_dentry
*topts
;
8322 struct trace_options
*tr_topts
;
8323 struct tracer_flags
*flags
;
8324 struct tracer_opt
*opts
;
8331 flags
= tracer
->flags
;
8333 if (!flags
|| !flags
->opts
)
8337 * If this is an instance, only create flags for tracers
8338 * the instance may have.
8340 if (!trace_ok_for_array(tracer
, tr
))
8343 for (i
= 0; i
< tr
->nr_topts
; i
++) {
8344 /* Make sure there's no duplicate flags. */
8345 if (WARN_ON_ONCE(tr
->topts
[i
].tracer
->flags
== tracer
->flags
))
8351 for (cnt
= 0; opts
[cnt
].name
; cnt
++)
8354 topts
= kcalloc(cnt
+ 1, sizeof(*topts
), GFP_KERNEL
);
8358 tr_topts
= krealloc(tr
->topts
, sizeof(*tr
->topts
) * (tr
->nr_topts
+ 1),
8365 tr
->topts
= tr_topts
;
8366 tr
->topts
[tr
->nr_topts
].tracer
= tracer
;
8367 tr
->topts
[tr
->nr_topts
].topts
= topts
;
8370 for (cnt
= 0; opts
[cnt
].name
; cnt
++) {
8371 create_trace_option_file(tr
, &topts
[cnt
], flags
,
8373 MEM_FAIL(topts
[cnt
].entry
== NULL
,
8374 "Failed to create trace option: %s",
8379 static struct dentry
*
8380 create_trace_option_core_file(struct trace_array
*tr
,
8381 const char *option
, long index
)
8383 struct dentry
*t_options
;
8385 t_options
= trace_options_init_dentry(tr
);
8389 return trace_create_file(option
, 0644, t_options
,
8390 (void *)&tr
->trace_flags_index
[index
],
8391 &trace_options_core_fops
);
8394 static void create_trace_options_dir(struct trace_array
*tr
)
8396 struct dentry
*t_options
;
8397 bool top_level
= tr
== &global_trace
;
8400 t_options
= trace_options_init_dentry(tr
);
8404 for (i
= 0; trace_options
[i
]; i
++) {
8406 !((1 << i
) & TOP_LEVEL_TRACE_FLAGS
))
8407 create_trace_option_core_file(tr
, trace_options
[i
], i
);
8412 rb_simple_read(struct file
*filp
, char __user
*ubuf
,
8413 size_t cnt
, loff_t
*ppos
)
8415 struct trace_array
*tr
= filp
->private_data
;
8419 r
= tracer_tracing_is_on(tr
);
8420 r
= sprintf(buf
, "%d\n", r
);
8422 return simple_read_from_buffer(ubuf
, cnt
, ppos
, buf
, r
);
8426 rb_simple_write(struct file
*filp
, const char __user
*ubuf
,
8427 size_t cnt
, loff_t
*ppos
)
8429 struct trace_array
*tr
= filp
->private_data
;
8430 struct trace_buffer
*buffer
= tr
->array_buffer
.buffer
;
8434 ret
= kstrtoul_from_user(ubuf
, cnt
, 10, &val
);
8439 mutex_lock(&trace_types_lock
);
8440 if (!!val
== tracer_tracing_is_on(tr
)) {
8441 val
= 0; /* do nothing */
8443 tracer_tracing_on(tr
);
8444 if (tr
->current_trace
->start
)
8445 tr
->current_trace
->start(tr
);
8447 tracer_tracing_off(tr
);
8448 if (tr
->current_trace
->stop
)
8449 tr
->current_trace
->stop(tr
);
8451 mutex_unlock(&trace_types_lock
);
8459 static const struct file_operations rb_simple_fops
= {
8460 .open
= tracing_open_generic_tr
,
8461 .read
= rb_simple_read
,
8462 .write
= rb_simple_write
,
8463 .release
= tracing_release_generic_tr
,
8464 .llseek
= default_llseek
,
8468 buffer_percent_read(struct file
*filp
, char __user
*ubuf
,
8469 size_t cnt
, loff_t
*ppos
)
8471 struct trace_array
*tr
= filp
->private_data
;
8475 r
= tr
->buffer_percent
;
8476 r
= sprintf(buf
, "%d\n", r
);
8478 return simple_read_from_buffer(ubuf
, cnt
, ppos
, buf
, r
);
8482 buffer_percent_write(struct file
*filp
, const char __user
*ubuf
,
8483 size_t cnt
, loff_t
*ppos
)
8485 struct trace_array
*tr
= filp
->private_data
;
8489 ret
= kstrtoul_from_user(ubuf
, cnt
, 10, &val
);
8499 tr
->buffer_percent
= val
;
8506 static const struct file_operations buffer_percent_fops
= {
8507 .open
= tracing_open_generic_tr
,
8508 .read
= buffer_percent_read
,
8509 .write
= buffer_percent_write
,
8510 .release
= tracing_release_generic_tr
,
8511 .llseek
= default_llseek
,
8514 static struct dentry
*trace_instance_dir
;
8517 init_tracer_tracefs(struct trace_array
*tr
, struct dentry
*d_tracer
);
8520 allocate_trace_buffer(struct trace_array
*tr
, struct array_buffer
*buf
, int size
)
8522 enum ring_buffer_flags rb_flags
;
8524 rb_flags
= tr
->trace_flags
& TRACE_ITER_OVERWRITE
? RB_FL_OVERWRITE
: 0;
8528 buf
->buffer
= ring_buffer_alloc(size
, rb_flags
);
8532 buf
->data
= alloc_percpu(struct trace_array_cpu
);
8534 ring_buffer_free(buf
->buffer
);
8539 /* Allocate the first page for all buffers */
8540 set_buffer_entries(&tr
->array_buffer
,
8541 ring_buffer_size(tr
->array_buffer
.buffer
, 0));
8546 static int allocate_trace_buffers(struct trace_array
*tr
, int size
)
8550 ret
= allocate_trace_buffer(tr
, &tr
->array_buffer
, size
);
8554 #ifdef CONFIG_TRACER_MAX_TRACE
8555 ret
= allocate_trace_buffer(tr
, &tr
->max_buffer
,
8556 allocate_snapshot
? size
: 1);
8557 if (MEM_FAIL(ret
, "Failed to allocate trace buffer\n")) {
8558 ring_buffer_free(tr
->array_buffer
.buffer
);
8559 tr
->array_buffer
.buffer
= NULL
;
8560 free_percpu(tr
->array_buffer
.data
);
8561 tr
->array_buffer
.data
= NULL
;
8564 tr
->allocated_snapshot
= allocate_snapshot
;
8567 * Only the top level trace array gets its snapshot allocated
8568 * from the kernel command line.
8570 allocate_snapshot
= false;
8576 static void free_trace_buffer(struct array_buffer
*buf
)
8579 ring_buffer_free(buf
->buffer
);
8581 free_percpu(buf
->data
);
8586 static void free_trace_buffers(struct trace_array
*tr
)
8591 free_trace_buffer(&tr
->array_buffer
);
8593 #ifdef CONFIG_TRACER_MAX_TRACE
8594 free_trace_buffer(&tr
->max_buffer
);
8598 static void init_trace_flags_index(struct trace_array
*tr
)
8602 /* Used by the trace options files */
8603 for (i
= 0; i
< TRACE_FLAGS_MAX_SIZE
; i
++)
8604 tr
->trace_flags_index
[i
] = i
;
8607 static void __update_tracer_options(struct trace_array
*tr
)
8611 for (t
= trace_types
; t
; t
= t
->next
)
8612 add_tracer_options(tr
, t
);
8615 static void update_tracer_options(struct trace_array
*tr
)
8617 mutex_lock(&trace_types_lock
);
8618 __update_tracer_options(tr
);
8619 mutex_unlock(&trace_types_lock
);
8622 /* Must have trace_types_lock held */
8623 struct trace_array
*trace_array_find(const char *instance
)
8625 struct trace_array
*tr
, *found
= NULL
;
8627 list_for_each_entry(tr
, &ftrace_trace_arrays
, list
) {
8628 if (tr
->name
&& strcmp(tr
->name
, instance
) == 0) {
8637 struct trace_array
*trace_array_find_get(const char *instance
)
8639 struct trace_array
*tr
;
8641 mutex_lock(&trace_types_lock
);
8642 tr
= trace_array_find(instance
);
8645 mutex_unlock(&trace_types_lock
);
8650 static struct trace_array
*trace_array_create(const char *name
)
8652 struct trace_array
*tr
;
8656 tr
= kzalloc(sizeof(*tr
), GFP_KERNEL
);
8658 return ERR_PTR(ret
);
8660 tr
->name
= kstrdup(name
, GFP_KERNEL
);
8664 if (!alloc_cpumask_var(&tr
->tracing_cpumask
, GFP_KERNEL
))
8667 tr
->trace_flags
= global_trace
.trace_flags
& ~ZEROED_TRACE_FLAGS
;
8669 cpumask_copy(tr
->tracing_cpumask
, cpu_all_mask
);
8671 raw_spin_lock_init(&tr
->start_lock
);
8673 tr
->max_lock
= (arch_spinlock_t
)__ARCH_SPIN_LOCK_UNLOCKED
;
8675 tr
->current_trace
= &nop_trace
;
8677 INIT_LIST_HEAD(&tr
->systems
);
8678 INIT_LIST_HEAD(&tr
->events
);
8679 INIT_LIST_HEAD(&tr
->hist_vars
);
8680 INIT_LIST_HEAD(&tr
->err_log
);
8682 if (allocate_trace_buffers(tr
, trace_buf_size
) < 0)
8685 tr
->dir
= tracefs_create_dir(name
, trace_instance_dir
);
8689 ret
= event_trace_add_tracer(tr
->dir
, tr
);
8691 tracefs_remove(tr
->dir
);
8695 ftrace_init_trace_array(tr
);
8697 init_tracer_tracefs(tr
, tr
->dir
);
8698 init_trace_flags_index(tr
);
8699 __update_tracer_options(tr
);
8701 list_add(&tr
->list
, &ftrace_trace_arrays
);
8709 free_trace_buffers(tr
);
8710 free_cpumask_var(tr
->tracing_cpumask
);
8714 return ERR_PTR(ret
);
8717 static int instance_mkdir(const char *name
)
8719 struct trace_array
*tr
;
8722 mutex_lock(&event_mutex
);
8723 mutex_lock(&trace_types_lock
);
8726 if (trace_array_find(name
))
8729 tr
= trace_array_create(name
);
8731 ret
= PTR_ERR_OR_ZERO(tr
);
8734 mutex_unlock(&trace_types_lock
);
8735 mutex_unlock(&event_mutex
);
8740 * trace_array_get_by_name - Create/Lookup a trace array, given its name.
8741 * @name: The name of the trace array to be looked up/created.
8743 * Returns pointer to trace array with given name.
8744 * NULL, if it cannot be created.
8746 * NOTE: This function increments the reference counter associated with the
8747 * trace array returned. This makes sure it cannot be freed while in use.
8748 * Use trace_array_put() once the trace array is no longer needed.
8749 * If the trace_array is to be freed, trace_array_destroy() needs to
8750 * be called after the trace_array_put(), or simply let user space delete
8751 * it from the tracefs instances directory. But until the
8752 * trace_array_put() is called, user space can not delete it.
8755 struct trace_array
*trace_array_get_by_name(const char *name
)
8757 struct trace_array
*tr
;
8759 mutex_lock(&event_mutex
);
8760 mutex_lock(&trace_types_lock
);
8762 list_for_each_entry(tr
, &ftrace_trace_arrays
, list
) {
8763 if (tr
->name
&& strcmp(tr
->name
, name
) == 0)
8767 tr
= trace_array_create(name
);
8775 mutex_unlock(&trace_types_lock
);
8776 mutex_unlock(&event_mutex
);
8779 EXPORT_SYMBOL_GPL(trace_array_get_by_name
);
8781 static int __remove_instance(struct trace_array
*tr
)
8785 /* Reference counter for a newly created trace array = 1. */
8786 if (tr
->ref
> 1 || (tr
->current_trace
&& tr
->trace_ref
))
8789 list_del(&tr
->list
);
8791 /* Disable all the flags that were enabled coming in */
8792 for (i
= 0; i
< TRACE_FLAGS_MAX_SIZE
; i
++) {
8793 if ((1 << i
) & ZEROED_TRACE_FLAGS
)
8794 set_tracer_flag(tr
, 1 << i
, 0);
8797 tracing_set_nop(tr
);
8798 clear_ftrace_function_probes(tr
);
8799 event_trace_del_tracer(tr
);
8800 ftrace_clear_pids(tr
);
8801 ftrace_destroy_function_files(tr
);
8802 tracefs_remove(tr
->dir
);
8803 free_trace_buffers(tr
);
8805 for (i
= 0; i
< tr
->nr_topts
; i
++) {
8806 kfree(tr
->topts
[i
].topts
);
8810 free_cpumask_var(tr
->tracing_cpumask
);
8818 int trace_array_destroy(struct trace_array
*this_tr
)
8820 struct trace_array
*tr
;
8826 mutex_lock(&event_mutex
);
8827 mutex_lock(&trace_types_lock
);
8831 /* Making sure trace array exists before destroying it. */
8832 list_for_each_entry(tr
, &ftrace_trace_arrays
, list
) {
8833 if (tr
== this_tr
) {
8834 ret
= __remove_instance(tr
);
8839 mutex_unlock(&trace_types_lock
);
8840 mutex_unlock(&event_mutex
);
8844 EXPORT_SYMBOL_GPL(trace_array_destroy
);
8846 static int instance_rmdir(const char *name
)
8848 struct trace_array
*tr
;
8851 mutex_lock(&event_mutex
);
8852 mutex_lock(&trace_types_lock
);
8855 tr
= trace_array_find(name
);
8857 ret
= __remove_instance(tr
);
8859 mutex_unlock(&trace_types_lock
);
8860 mutex_unlock(&event_mutex
);
8865 static __init
void create_trace_instances(struct dentry
*d_tracer
)
8867 trace_instance_dir
= tracefs_create_instance_dir("instances", d_tracer
,
8870 if (MEM_FAIL(!trace_instance_dir
, "Failed to create instances directory\n"))
8875 init_tracer_tracefs(struct trace_array
*tr
, struct dentry
*d_tracer
)
8877 struct trace_event_file
*file
;
8880 trace_create_file("available_tracers", 0444, d_tracer
,
8881 tr
, &show_traces_fops
);
8883 trace_create_file("current_tracer", 0644, d_tracer
,
8884 tr
, &set_tracer_fops
);
8886 trace_create_file("tracing_cpumask", 0644, d_tracer
,
8887 tr
, &tracing_cpumask_fops
);
8889 trace_create_file("trace_options", 0644, d_tracer
,
8890 tr
, &tracing_iter_fops
);
8892 trace_create_file("trace", 0644, d_tracer
,
8895 trace_create_file("trace_pipe", 0444, d_tracer
,
8896 tr
, &tracing_pipe_fops
);
8898 trace_create_file("buffer_size_kb", 0644, d_tracer
,
8899 tr
, &tracing_entries_fops
);
8901 trace_create_file("buffer_total_size_kb", 0444, d_tracer
,
8902 tr
, &tracing_total_entries_fops
);
8904 trace_create_file("free_buffer", 0200, d_tracer
,
8905 tr
, &tracing_free_buffer_fops
);
8907 trace_create_file("trace_marker", 0220, d_tracer
,
8908 tr
, &tracing_mark_fops
);
8910 file
= __find_event_file(tr
, "ftrace", "print");
8911 if (file
&& file
->dir
)
8912 trace_create_file("trigger", 0644, file
->dir
, file
,
8913 &event_trigger_fops
);
8914 tr
->trace_marker_file
= file
;
8916 trace_create_file("trace_marker_raw", 0220, d_tracer
,
8917 tr
, &tracing_mark_raw_fops
);
8919 trace_create_file("trace_clock", 0644, d_tracer
, tr
,
8922 trace_create_file("tracing_on", 0644, d_tracer
,
8923 tr
, &rb_simple_fops
);
8925 trace_create_file("timestamp_mode", 0444, d_tracer
, tr
,
8926 &trace_time_stamp_mode_fops
);
8928 tr
->buffer_percent
= 50;
8930 trace_create_file("buffer_percent", 0444, d_tracer
,
8931 tr
, &buffer_percent_fops
);
8933 create_trace_options_dir(tr
);
8935 #if defined(CONFIG_TRACER_MAX_TRACE) || defined(CONFIG_HWLAT_TRACER)
8936 trace_create_maxlat_file(tr
, d_tracer
);
8939 if (ftrace_create_function_files(tr
, d_tracer
))
8940 MEM_FAIL(1, "Could not allocate function filter files");
8942 #ifdef CONFIG_TRACER_SNAPSHOT
8943 trace_create_file("snapshot", 0644, d_tracer
,
8944 tr
, &snapshot_fops
);
8947 trace_create_file("error_log", 0644, d_tracer
,
8948 tr
, &tracing_err_log_fops
);
8950 for_each_tracing_cpu(cpu
)
8951 tracing_init_tracefs_percpu(tr
, cpu
);
8953 ftrace_init_tracefs(tr
, d_tracer
);
8956 static struct vfsmount
*trace_automount(struct dentry
*mntpt
, void *ingore
)
8958 struct vfsmount
*mnt
;
8959 struct file_system_type
*type
;
8962 * To maintain backward compatibility for tools that mount
8963 * debugfs to get to the tracing facility, tracefs is automatically
8964 * mounted to the debugfs/tracing directory.
8966 type
= get_fs_type("tracefs");
8969 mnt
= vfs_submount(mntpt
, type
, "tracefs", NULL
);
8970 put_filesystem(type
);
8979 * tracing_init_dentry - initialize top level trace array
8981 * This is called when creating files or directories in the tracing
8982 * directory. It is called via fs_initcall() by any of the boot up code
8983 * and expects to return the dentry of the top level tracing directory.
8985 struct dentry
*tracing_init_dentry(void)
8987 struct trace_array
*tr
= &global_trace
;
8989 if (security_locked_down(LOCKDOWN_TRACEFS
)) {
8990 pr_warn("Tracing disabled due to lockdown\n");
8991 return ERR_PTR(-EPERM
);
8994 /* The top level trace array uses NULL as parent */
8998 if (WARN_ON(!tracefs_initialized()))
8999 return ERR_PTR(-ENODEV
);
9002 * As there may still be users that expect the tracing
9003 * files to exist in debugfs/tracing, we must automount
9004 * the tracefs file system there, so older tools still
9005 * work with the newer kerenl.
9007 tr
->dir
= debugfs_create_automount("tracing", NULL
,
9008 trace_automount
, NULL
);
9013 extern struct trace_eval_map
*__start_ftrace_eval_maps
[];
9014 extern struct trace_eval_map
*__stop_ftrace_eval_maps
[];
9016 static void __init
trace_eval_init(void)
9020 len
= __stop_ftrace_eval_maps
- __start_ftrace_eval_maps
;
9021 trace_insert_eval_map(NULL
, __start_ftrace_eval_maps
, len
);
9024 #ifdef CONFIG_MODULES
9025 static void trace_module_add_evals(struct module
*mod
)
9027 if (!mod
->num_trace_evals
)
9031 * Modules with bad taint do not have events created, do
9032 * not bother with enums either.
9034 if (trace_module_has_bad_taint(mod
))
9037 trace_insert_eval_map(mod
, mod
->trace_evals
, mod
->num_trace_evals
);
9040 #ifdef CONFIG_TRACE_EVAL_MAP_FILE
9041 static void trace_module_remove_evals(struct module
*mod
)
9043 union trace_eval_map_item
*map
;
9044 union trace_eval_map_item
**last
= &trace_eval_maps
;
9046 if (!mod
->num_trace_evals
)
9049 mutex_lock(&trace_eval_mutex
);
9051 map
= trace_eval_maps
;
9054 if (map
->head
.mod
== mod
)
9056 map
= trace_eval_jmp_to_tail(map
);
9057 last
= &map
->tail
.next
;
9058 map
= map
->tail
.next
;
9063 *last
= trace_eval_jmp_to_tail(map
)->tail
.next
;
9066 mutex_unlock(&trace_eval_mutex
);
9069 static inline void trace_module_remove_evals(struct module
*mod
) { }
9070 #endif /* CONFIG_TRACE_EVAL_MAP_FILE */
9072 static int trace_module_notify(struct notifier_block
*self
,
9073 unsigned long val
, void *data
)
9075 struct module
*mod
= data
;
9078 case MODULE_STATE_COMING
:
9079 trace_module_add_evals(mod
);
9081 case MODULE_STATE_GOING
:
9082 trace_module_remove_evals(mod
);
9089 static struct notifier_block trace_module_nb
= {
9090 .notifier_call
= trace_module_notify
,
9093 #endif /* CONFIG_MODULES */
9095 static __init
int tracer_init_tracefs(void)
9097 struct dentry
*d_tracer
;
9099 trace_access_lock_init();
9101 d_tracer
= tracing_init_dentry();
9102 if (IS_ERR(d_tracer
))
9107 init_tracer_tracefs(&global_trace
, d_tracer
);
9108 ftrace_init_tracefs_toplevel(&global_trace
, d_tracer
);
9110 trace_create_file("tracing_thresh", 0644, d_tracer
,
9111 &global_trace
, &tracing_thresh_fops
);
9113 trace_create_file("README", 0444, d_tracer
,
9114 NULL
, &tracing_readme_fops
);
9116 trace_create_file("saved_cmdlines", 0444, d_tracer
,
9117 NULL
, &tracing_saved_cmdlines_fops
);
9119 trace_create_file("saved_cmdlines_size", 0644, d_tracer
,
9120 NULL
, &tracing_saved_cmdlines_size_fops
);
9122 trace_create_file("saved_tgids", 0444, d_tracer
,
9123 NULL
, &tracing_saved_tgids_fops
);
9127 trace_create_eval_file(d_tracer
);
9129 #ifdef CONFIG_MODULES
9130 register_module_notifier(&trace_module_nb
);
9133 #ifdef CONFIG_DYNAMIC_FTRACE
9134 trace_create_file("dyn_ftrace_total_info", 0444, d_tracer
,
9135 NULL
, &tracing_dyn_info_fops
);
9138 create_trace_instances(d_tracer
);
9140 update_tracer_options(&global_trace
);
9145 static int trace_panic_handler(struct notifier_block
*this,
9146 unsigned long event
, void *unused
)
9148 if (ftrace_dump_on_oops
)
9149 ftrace_dump(ftrace_dump_on_oops
);
9153 static struct notifier_block trace_panic_notifier
= {
9154 .notifier_call
= trace_panic_handler
,
9156 .priority
= 150 /* priority: INT_MAX >= x >= 0 */
9159 static int trace_die_handler(struct notifier_block
*self
,
9165 if (ftrace_dump_on_oops
)
9166 ftrace_dump(ftrace_dump_on_oops
);
9174 static struct notifier_block trace_die_notifier
= {
9175 .notifier_call
= trace_die_handler
,
9180 * printk is set to max of 1024, we really don't need it that big.
9181 * Nothing should be printing 1000 characters anyway.
9183 #define TRACE_MAX_PRINT 1000
9186 * Define here KERN_TRACE so that we have one place to modify
9187 * it if we decide to change what log level the ftrace dump
9190 #define KERN_TRACE KERN_EMERG
9193 trace_printk_seq(struct trace_seq
*s
)
9195 /* Probably should print a warning here. */
9196 if (s
->seq
.len
>= TRACE_MAX_PRINT
)
9197 s
->seq
.len
= TRACE_MAX_PRINT
;
9200 * More paranoid code. Although the buffer size is set to
9201 * PAGE_SIZE, and TRACE_MAX_PRINT is 1000, this is just
9202 * an extra layer of protection.
9204 if (WARN_ON_ONCE(s
->seq
.len
>= s
->seq
.size
))
9205 s
->seq
.len
= s
->seq
.size
- 1;
9207 /* should be zero ended, but we are paranoid. */
9208 s
->buffer
[s
->seq
.len
] = 0;
9210 printk(KERN_TRACE
"%s", s
->buffer
);
9215 void trace_init_global_iter(struct trace_iterator
*iter
)
9217 iter
->tr
= &global_trace
;
9218 iter
->trace
= iter
->tr
->current_trace
;
9219 iter
->cpu_file
= RING_BUFFER_ALL_CPUS
;
9220 iter
->array_buffer
= &global_trace
.array_buffer
;
9222 if (iter
->trace
&& iter
->trace
->open
)
9223 iter
->trace
->open(iter
);
9225 /* Annotate start of buffers if we had overruns */
9226 if (ring_buffer_overruns(iter
->array_buffer
->buffer
))
9227 iter
->iter_flags
|= TRACE_FILE_ANNOTATE
;
9229 /* Output in nanoseconds only if we are using a clock in nanoseconds. */
9230 if (trace_clocks
[iter
->tr
->clock_id
].in_ns
)
9231 iter
->iter_flags
|= TRACE_FILE_TIME_IN_NS
;
9234 void ftrace_dump(enum ftrace_dump_mode oops_dump_mode
)
9236 /* use static because iter can be a bit big for the stack */
9237 static struct trace_iterator iter
;
9238 static atomic_t dump_running
;
9239 struct trace_array
*tr
= &global_trace
;
9240 unsigned int old_userobj
;
9241 unsigned long flags
;
9244 /* Only allow one dump user at a time. */
9245 if (atomic_inc_return(&dump_running
) != 1) {
9246 atomic_dec(&dump_running
);
9251 * Always turn off tracing when we dump.
9252 * We don't need to show trace output of what happens
9253 * between multiple crashes.
9255 * If the user does a sysrq-z, then they can re-enable
9256 * tracing with echo 1 > tracing_on.
9260 local_irq_save(flags
);
9261 printk_nmi_direct_enter();
9263 /* Simulate the iterator */
9264 trace_init_global_iter(&iter
);
9265 /* Can not use kmalloc for iter.temp */
9266 iter
.temp
= static_temp_buf
;
9267 iter
.temp_size
= STATIC_TEMP_BUF_SIZE
;
9269 for_each_tracing_cpu(cpu
) {
9270 atomic_inc(&per_cpu_ptr(iter
.array_buffer
->data
, cpu
)->disabled
);
9273 old_userobj
= tr
->trace_flags
& TRACE_ITER_SYM_USEROBJ
;
9275 /* don't look at user memory in panic mode */
9276 tr
->trace_flags
&= ~TRACE_ITER_SYM_USEROBJ
;
9278 switch (oops_dump_mode
) {
9280 iter
.cpu_file
= RING_BUFFER_ALL_CPUS
;
9283 iter
.cpu_file
= raw_smp_processor_id();
9288 printk(KERN_TRACE
"Bad dumping mode, switching to all CPUs dump\n");
9289 iter
.cpu_file
= RING_BUFFER_ALL_CPUS
;
9292 printk(KERN_TRACE
"Dumping ftrace buffer:\n");
9294 /* Did function tracer already get disabled? */
9295 if (ftrace_is_dead()) {
9296 printk("# WARNING: FUNCTION TRACING IS CORRUPTED\n");
9297 printk("# MAY BE MISSING FUNCTION EVENTS\n");
9301 * We need to stop all tracing on all CPUS to read the
9302 * the next buffer. This is a bit expensive, but is
9303 * not done often. We fill all what we can read,
9304 * and then release the locks again.
9307 while (!trace_empty(&iter
)) {
9310 printk(KERN_TRACE
"---------------------------------\n");
9314 trace_iterator_reset(&iter
);
9315 iter
.iter_flags
|= TRACE_FILE_LAT_FMT
;
9317 if (trace_find_next_entry_inc(&iter
) != NULL
) {
9320 ret
= print_trace_line(&iter
);
9321 if (ret
!= TRACE_TYPE_NO_CONSUME
)
9322 trace_consume(&iter
);
9324 touch_nmi_watchdog();
9326 trace_printk_seq(&iter
.seq
);
9330 printk(KERN_TRACE
" (ftrace buffer empty)\n");
9332 printk(KERN_TRACE
"---------------------------------\n");
9335 tr
->trace_flags
|= old_userobj
;
9337 for_each_tracing_cpu(cpu
) {
9338 atomic_dec(&per_cpu_ptr(iter
.array_buffer
->data
, cpu
)->disabled
);
9340 atomic_dec(&dump_running
);
9341 printk_nmi_direct_exit();
9342 local_irq_restore(flags
);
9344 EXPORT_SYMBOL_GPL(ftrace_dump
);
9346 int trace_run_command(const char *buf
, int (*createfn
)(int, char **))
9353 argv
= argv_split(GFP_KERNEL
, buf
, &argc
);
9358 ret
= createfn(argc
, argv
);
9365 #define WRITE_BUFSIZE 4096
9367 ssize_t
trace_parse_run_command(struct file
*file
, const char __user
*buffer
,
9368 size_t count
, loff_t
*ppos
,
9369 int (*createfn
)(int, char **))
9371 char *kbuf
, *buf
, *tmp
;
9376 kbuf
= kmalloc(WRITE_BUFSIZE
, GFP_KERNEL
);
9380 while (done
< count
) {
9381 size
= count
- done
;
9383 if (size
>= WRITE_BUFSIZE
)
9384 size
= WRITE_BUFSIZE
- 1;
9386 if (copy_from_user(kbuf
, buffer
+ done
, size
)) {
9393 tmp
= strchr(buf
, '\n');
9396 size
= tmp
- buf
+ 1;
9399 if (done
+ size
< count
) {
9402 /* This can accept WRITE_BUFSIZE - 2 ('\n' + '\0') */
9403 pr_warn("Line length is too long: Should be less than %d\n",
9411 /* Remove comments */
9412 tmp
= strchr(buf
, '#');
9417 ret
= trace_run_command(buf
, createfn
);
9422 } while (done
< count
);
9432 __init
static int tracer_alloc_buffers(void)
9438 if (security_locked_down(LOCKDOWN_TRACEFS
)) {
9439 pr_warn("Tracing disabled due to lockdown\n");
9444 * Make sure we don't accidently add more trace options
9445 * than we have bits for.
9447 BUILD_BUG_ON(TRACE_ITER_LAST_BIT
> TRACE_FLAGS_MAX_SIZE
);
9449 if (!alloc_cpumask_var(&tracing_buffer_mask
, GFP_KERNEL
))
9452 if (!alloc_cpumask_var(&global_trace
.tracing_cpumask
, GFP_KERNEL
))
9453 goto out_free_buffer_mask
;
9455 /* Only allocate trace_printk buffers if a trace_printk exists */
9456 if (&__stop___trace_bprintk_fmt
!= &__start___trace_bprintk_fmt
)
9457 /* Must be called before global_trace.buffer is allocated */
9458 trace_printk_init_buffers();
9460 /* To save memory, keep the ring buffer size to its minimum */
9461 if (ring_buffer_expanded
)
9462 ring_buf_size
= trace_buf_size
;
9466 cpumask_copy(tracing_buffer_mask
, cpu_possible_mask
);
9467 cpumask_copy(global_trace
.tracing_cpumask
, cpu_all_mask
);
9469 raw_spin_lock_init(&global_trace
.start_lock
);
9472 * The prepare callbacks allocates some memory for the ring buffer. We
9473 * don't free the buffer if the if the CPU goes down. If we were to free
9474 * the buffer, then the user would lose any trace that was in the
9475 * buffer. The memory will be removed once the "instance" is removed.
9477 ret
= cpuhp_setup_state_multi(CPUHP_TRACE_RB_PREPARE
,
9478 "trace/RB:preapre", trace_rb_cpu_prepare
,
9481 goto out_free_cpumask
;
9482 /* Used for event triggers */
9484 temp_buffer
= ring_buffer_alloc(PAGE_SIZE
, RB_FL_OVERWRITE
);
9486 goto out_rm_hp_state
;
9488 if (trace_create_savedcmd() < 0)
9489 goto out_free_temp_buffer
;
9491 /* TODO: make the number of buffers hot pluggable with CPUS */
9492 if (allocate_trace_buffers(&global_trace
, ring_buf_size
) < 0) {
9493 MEM_FAIL(1, "tracer: failed to allocate ring buffer!\n");
9494 goto out_free_savedcmd
;
9497 if (global_trace
.buffer_disabled
)
9500 if (trace_boot_clock
) {
9501 ret
= tracing_set_clock(&global_trace
, trace_boot_clock
);
9503 pr_warn("Trace clock %s not defined, going back to default\n",
9508 * register_tracer() might reference current_trace, so it
9509 * needs to be set before we register anything. This is
9510 * just a bootstrap of current_trace anyway.
9512 global_trace
.current_trace
= &nop_trace
;
9514 global_trace
.max_lock
= (arch_spinlock_t
)__ARCH_SPIN_LOCK_UNLOCKED
;
9516 ftrace_init_global_array_ops(&global_trace
);
9518 init_trace_flags_index(&global_trace
);
9520 register_tracer(&nop_trace
);
9522 /* Function tracing may start here (via kernel command line) */
9523 init_function_trace();
9525 /* All seems OK, enable tracing */
9526 tracing_disabled
= 0;
9528 atomic_notifier_chain_register(&panic_notifier_list
,
9529 &trace_panic_notifier
);
9531 register_die_notifier(&trace_die_notifier
);
9533 global_trace
.flags
= TRACE_ARRAY_FL_GLOBAL
;
9535 INIT_LIST_HEAD(&global_trace
.systems
);
9536 INIT_LIST_HEAD(&global_trace
.events
);
9537 INIT_LIST_HEAD(&global_trace
.hist_vars
);
9538 INIT_LIST_HEAD(&global_trace
.err_log
);
9539 list_add(&global_trace
.list
, &ftrace_trace_arrays
);
9541 apply_trace_boot_options();
9543 register_snapshot_cmd();
9548 free_saved_cmdlines_buffer(savedcmd
);
9549 out_free_temp_buffer
:
9550 ring_buffer_free(temp_buffer
);
9552 cpuhp_remove_multi_state(CPUHP_TRACE_RB_PREPARE
);
9554 free_cpumask_var(global_trace
.tracing_cpumask
);
9555 out_free_buffer_mask
:
9556 free_cpumask_var(tracing_buffer_mask
);
9561 void __init
early_trace_init(void)
9563 if (tracepoint_printk
) {
9564 tracepoint_print_iter
=
9565 kmalloc(sizeof(*tracepoint_print_iter
), GFP_KERNEL
);
9566 if (MEM_FAIL(!tracepoint_print_iter
,
9567 "Failed to allocate trace iterator\n"))
9568 tracepoint_printk
= 0;
9570 static_key_enable(&tracepoint_printk_key
.key
);
9572 tracer_alloc_buffers();
9575 void __init
trace_init(void)
9580 __init
static int clear_boot_tracer(void)
9583 * The default tracer at boot buffer is an init section.
9584 * This function is called in lateinit. If we did not
9585 * find the boot tracer, then clear it out, to prevent
9586 * later registration from accessing the buffer that is
9587 * about to be freed.
9589 if (!default_bootup_tracer
)
9592 printk(KERN_INFO
"ftrace bootup tracer '%s' not registered.\n",
9593 default_bootup_tracer
);
9594 default_bootup_tracer
= NULL
;
9599 fs_initcall(tracer_init_tracefs
);
9600 late_initcall_sync(clear_boot_tracer
);
9602 #ifdef CONFIG_HAVE_UNSTABLE_SCHED_CLOCK
9603 __init
static int tracing_set_default_clock(void)
9605 /* sched_clock_stable() is determined in late_initcall */
9606 if (!trace_boot_clock
&& !sched_clock_stable()) {
9607 if (security_locked_down(LOCKDOWN_TRACEFS
)) {
9608 pr_warn("Can not set tracing clock due to lockdown\n");
9613 "Unstable clock detected, switching default tracing clock to \"global\"\n"
9614 "If you want to keep using the local clock, then add:\n"
9615 " \"trace_clock=local\"\n"
9616 "on the kernel command line\n");
9617 tracing_set_clock(&global_trace
, "global");
9622 late_initcall_sync(tracing_set_default_clock
);