OpenCloudOS-Kernel/include/linux/ftrace_event.h

232 lines
6.3 KiB
C
Raw Normal View History

#ifndef _LINUX_FTRACE_EVENT_H
#define _LINUX_FTRACE_EVENT_H
#include <linux/ring_buffer.h>
#include <linux/trace_seq.h>
#include <linux/percpu.h>
#include <linux/hardirq.h>
#include <linux/perf_event.h>
struct trace_array;
struct tracer;
struct dentry;
DECLARE_PER_CPU(struct trace_seq, ftrace_event_seq);
struct trace_print_flags {
unsigned long mask;
const char *name;
};
const char *ftrace_print_flags_seq(struct trace_seq *p, const char *delim,
unsigned long flags,
const struct trace_print_flags *flag_array);
const char *ftrace_print_symbols_seq(struct trace_seq *p, unsigned long val,
const struct trace_print_flags *symbol_array);
/*
* The trace entry - the most basic unit of tracing. This is what
* is printed in the end as a single line in the trace output, such as:
*
* bash-15816 [01] 235.197585: idle_cpu <- irq_enter
*/
struct trace_entry {
unsigned short type;
unsigned char flags;
unsigned char preempt_count;
int pid;
int lock_depth;
};
#define FTRACE_MAX_EVENT \
((1 << (sizeof(((struct trace_entry *)0)->type) * 8)) - 1)
/*
* Trace iterator - used by printout routines who present trace
* results to users and which routines might sleep, etc:
*/
struct trace_iterator {
struct trace_array *tr;
struct tracer *trace;
void *private;
int cpu_file;
struct mutex mutex;
struct ring_buffer_iter *buffer_iter[NR_CPUS];
unsigned long iter_flags;
/* The below is zeroed out in pipe_read */
struct trace_seq seq;
struct trace_entry *ent;
unsigned long lost_events;
tracing: Buffer the output of seq_file in case of filled buffer If the seq_read fills the buffer it will call s_start again on the next itertation with the same position. This causes a problem with the function_graph tracer because it consumes the iteration in order to determine leaf functions. What happens is that the iterator stores the entry, and the function graph plugin will look at the next entry. If that next entry is a return of the same function and task, then the function is a leaf and the function_graph plugin calls ring_buffer_read which moves the ring buffer iterator forward (the trace iterator still points to the function start entry). The copying of the trace_seq to the seq_file buffer will fail if the seq_file buffer is full. The seq_read will not show this entry. The next read by userspace will cause seq_read to again call s_start which will reuse the trace iterator entry (the function start entry). But the function return entry was already consumed. The function graph plugin will think that this entry is a nested function and not a leaf. To solve this, the trace code now checks the return status of the seq_printf (trace_print_seq). If the writing to the seq_file buffer fails, we set a flag in the iterator (leftover) and we do not reset the trace_seq buffer. On the next call to s_start, we check the leftover flag, and if it is set, we just reuse the trace_seq buffer and do not call into the plugin print functions. Before this patch: 2) | fput() { 2) | __fput() { 2) 0.550 us | inotify_inode_queue_event(); 2) | __fsnotify_parent() { 2) 0.540 us | inotify_dentry_parent_queue_event(); After the patch: 2) | fput() { 2) | __fput() { 2) 0.550 us | inotify_inode_queue_event(); 2) 0.548 us | __fsnotify_parent(); 2) 0.540 us | inotify_dentry_parent_queue_event(); [ Updated the patch to fix a missing return 0 from the trace_print_seq() stub when CONFIG_TRACING is disabled. Reported-by: Ingo Molnar <mingo@elte.hu> ] Reported-by: Jiri Olsa <jolsa@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2009-12-07 22:11:39 +08:00
int leftover;
int cpu;
u64 ts;
loff_t pos;
long idx;
cpumask_var_t started;
};
typedef enum print_line_t (*trace_print_func)(struct trace_iterator *iter,
int flags);
struct trace_event {
struct hlist_node node;
struct list_head list;
int type;
trace_print_func trace;
trace_print_func raw;
trace_print_func hex;
trace_print_func binary;
};
extern int register_ftrace_event(struct trace_event *event);
extern int unregister_ftrace_event(struct trace_event *event);
/* Return values for print_line callback */
enum print_line_t {
TRACE_TYPE_PARTIAL_LINE = 0, /* Retry after flushing the seq */
TRACE_TYPE_HANDLED = 1,
TRACE_TYPE_UNHANDLED = 2, /* Relay to other output functions */
TRACE_TYPE_NO_CONSUME = 3 /* Handled but ask to not consume */
};
perf_counter: Fix/complete ftrace event records sampling This patch implements the kernel side support for ftrace event record sampling. A new counter sampling attribute is added: PERF_SAMPLE_TP_RECORD which requests ftrace events record sampling. In this case if a PERF_TYPE_TRACEPOINT counter is active and a tracepoint fires, we emit the tracepoint binary record to the perfcounter event buffer, as a sample. Result, after setting PERF_SAMPLE_TP_RECORD attribute from perf record: perf record -f -F 1 -a -e workqueue:workqueue_execution perf report -D 0x21e18 [0x48]: event: 9 . . ... raw event: size 72 bytes . 0000: 09 00 00 00 01 00 48 00 d0 c7 00 81 ff ff ff ff ......H........ . 0010: 0a 00 00 00 0a 00 00 00 21 00 00 00 00 00 00 00 ........!...... . 0020: 2b 00 01 02 0a 00 00 00 0a 00 00 00 65 76 65 6e +...........eve . 0030: 74 73 2f 31 00 00 00 00 00 00 00 00 0a 00 00 00 ts/1........... . 0040: e0 b1 31 81 ff ff ff ff ....... . 0x21e18 [0x48]: PERF_EVENT_SAMPLE (IP, 1): 10: 0xffffffff8100c7d0 period: 33 The raw ftrace binary record starts at offset 0020. Translation: struct trace_entry { type = 0x2b = 43; flags = 1; preempt_count = 2; pid = 0xa = 10; tgid = 0xa = 10; } thread_comm = "events/1" thread_pid = 0xa = 10; func = 0xffffffff8131b1e0 = flush_to_ldisc() What will come next? - Userspace support ('perf trace'), 'flight data recorder' mode for perf trace, etc. - The unconditional copy from the profiling callback brings some costs however if someone wants no such sampling to occur, and needs to be fixed in the future. For that we need to have an instant access to the perf counter attribute. This is a matter of a flag to add in the struct ftrace_event. - Take care of the events recursivity! Don't ever try to record a lock event for example, it seems some locking is used in the profiling fast path and lead to a tracing recursivity. That will be fixed using raw spinlock or recursivity protection. - [...] - Profit! :-) Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Li Zefan <lizf@cn.fujitsu.com> Cc: Tom Zanussi <tzanussi@gmail.com> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Cc: Mike Galbraith <efault@gmx.de> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Pekka Enberg <penberg@cs.helsinki.fi> Cc: Gabriel Munteanu <eduard.munteanu@linux360.ro> Cc: Lai Jiangshan <laijs@cn.fujitsu.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-08-07 07:25:54 +08:00
void tracing_generic_entry_update(struct trace_entry *entry,
unsigned long flags,
int pc);
struct ring_buffer_event *
trace_current_buffer_lock_reserve(struct ring_buffer **current_buffer,
int type, unsigned long len,
unsigned long flags, int pc);
void trace_current_buffer_unlock_commit(struct ring_buffer *buffer,
struct ring_buffer_event *event,
unsigned long flags, int pc);
void trace_nowake_buffer_unlock_commit(struct ring_buffer *buffer,
struct ring_buffer_event *event,
unsigned long flags, int pc);
void trace_current_buffer_discard_commit(struct ring_buffer *buffer,
struct ring_buffer_event *event);
void tracing_record_cmdline(struct task_struct *tsk);
struct event_filter;
tracing: Remove per event trace registering This patch removes the register functions of TRACE_EVENT() to enable and disable tracepoints. The registering of a event is now down directly in the trace_events.c file. The tracepoint_probe_register() is now called directly. The prototypes are no longer type checked, but this should not be an issue since the tracepoints are created automatically by the macros. If a prototype is incorrect in the TRACE_EVENT() macro, then other macros will catch it. The trace_event_class structure now holds the probes to be called by the callbacks. This removes needing to have each event have a separate pointer for the probe. To handle kprobes and syscalls, since they register probes in a different manner, a "reg" field is added to the ftrace_event_class structure. If the "reg" field is assigned, then it will be called for enabling and disabling of the probe for either ftrace or perf. To let the reg function know what is happening, a new enum (trace_reg) is created that has the type of control that is needed. With this new rework, the 82 kernel events and 618 syscall events has their footprint dramatically lowered: text data bss dec hex filename 4913961 1088356 861512 6863829 68bbd5 vmlinux.orig 4914025 1088868 861512 6864405 68be15 vmlinux.class 4918492 1084612 861512 6864616 68bee8 vmlinux.tracepoint 4900252 1057412 861512 6819176 680d68 vmlinux.regs The size went from 6863829 to 6819176, that's a total of 44K in savings. With tracepoints being continuously added, this is critical that the footprint becomes minimal. v5: Added #ifdef CONFIG_PERF_EVENTS around a reference to perf specific structure in trace_events.c. v4: Fixed trace self tests to check probe because regfunc no longer exists. v3: Updated to handle void *data in beginning of probe parameters. Also added the tracepoint: check_trace_callback_type_##call(). v2: Changed the callback probes to pass void * and typecast the value within the function. Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Acked-by: Masami Hiramatsu <mhiramat@redhat.com> Acked-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2010-04-22 00:27:06 +08:00
enum trace_reg {
TRACE_REG_REGISTER,
TRACE_REG_UNREGISTER,
TRACE_REG_PERF_REGISTER,
TRACE_REG_PERF_UNREGISTER,
};
struct ftrace_event_call;
struct ftrace_event_class {
char *system;
tracing: Remove per event trace registering This patch removes the register functions of TRACE_EVENT() to enable and disable tracepoints. The registering of a event is now down directly in the trace_events.c file. The tracepoint_probe_register() is now called directly. The prototypes are no longer type checked, but this should not be an issue since the tracepoints are created automatically by the macros. If a prototype is incorrect in the TRACE_EVENT() macro, then other macros will catch it. The trace_event_class structure now holds the probes to be called by the callbacks. This removes needing to have each event have a separate pointer for the probe. To handle kprobes and syscalls, since they register probes in a different manner, a "reg" field is added to the ftrace_event_class structure. If the "reg" field is assigned, then it will be called for enabling and disabling of the probe for either ftrace or perf. To let the reg function know what is happening, a new enum (trace_reg) is created that has the type of control that is needed. With this new rework, the 82 kernel events and 618 syscall events has their footprint dramatically lowered: text data bss dec hex filename 4913961 1088356 861512 6863829 68bbd5 vmlinux.orig 4914025 1088868 861512 6864405 68be15 vmlinux.class 4918492 1084612 861512 6864616 68bee8 vmlinux.tracepoint 4900252 1057412 861512 6819176 680d68 vmlinux.regs The size went from 6863829 to 6819176, that's a total of 44K in savings. With tracepoints being continuously added, this is critical that the footprint becomes minimal. v5: Added #ifdef CONFIG_PERF_EVENTS around a reference to perf specific structure in trace_events.c. v4: Fixed trace self tests to check probe because regfunc no longer exists. v3: Updated to handle void *data in beginning of probe parameters. Also added the tracepoint: check_trace_callback_type_##call(). v2: Changed the callback probes to pass void * and typecast the value within the function. Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Acked-by: Masami Hiramatsu <mhiramat@redhat.com> Acked-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2010-04-22 00:27:06 +08:00
void *probe;
#ifdef CONFIG_PERF_EVENTS
void *perf_probe;
#endif
int (*reg)(struct ftrace_event_call *event,
enum trace_reg type);
tracing: Move fields from event to class structure Move the defined fields from the event to the class structure. Since the fields of the event are defined by the class they belong to, it makes sense to have the class hold the information instead of the individual events. The events of the same class would just hold duplicate information. After this change the size of the kernel dropped another 3K: text data bss dec hex filename 4913961 1088356 861512 6863829 68bbd5 vmlinux.orig 4900252 1057412 861512 6819176 680d68 vmlinux.regs 4900375 1053380 861512 6815267 67fe23 vmlinux.fields Although the text increased, this was mainly due to the C files having to adapt to the change. This is a constant increase, where new tracepoints will not increase the Text. But the big drop is in the data size (as well as needed allocations to hold the fields). This will give even more savings as more tracepoints are created. Note, if just TRACE_EVENT()s are used and not DECLARE_EVENT_CLASS() with several DEFINE_EVENT()s, then the savings will be lost. But we are pushing developers to consolidate events with DEFINE_EVENT() so this should not be an issue. The kprobes define a unique class to every new event, but are dynamic so it should not be a issue. The syscalls however have a single class but the fields for the individual events are different. The syscalls use a metadata to define the fields. I moved the fields list from the event to the metadata and added a "get_fields()" function to the class. This function is used to find the fields. For normal events and kprobes, get_fields() just returns a pointer to the fields list_head in the class. For syscall events, it returns the fields list_head in the metadata for the event. v2: Fixed the syscall fields. The syscall metadata needs a list of fields for both enter and exit. Acked-by: Frederic Weisbecker <fweisbec@gmail.com> Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com> Acked-by: Masami Hiramatsu <mhiramat@redhat.com> Cc: Tom Zanussi <tzanussi@gmail.com> Cc: Peter Zijlstra <peterz@infradead.org> Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
2010-04-22 22:35:55 +08:00
int (*define_fields)(struct ftrace_event_call *);
struct list_head *(*get_fields)(struct ftrace_event_call *);
struct list_head fields;
int (*raw_init)(struct ftrace_event_call *);
};
struct ftrace_event_call {
struct list_head list;
struct ftrace_event_class *class;
char *name;
struct dentry *dir;
struct trace_event *event;
int enabled;
int id;
const char *print_fmt;
int filter_active;
struct event_filter *filter;
void *mod;
void *data;
int perf_refcount;
};
#define PERF_MAX_TRACE_SIZE 2048
tracing/filters: a better event parser Replace the current event parser hack with a better one. Filters are no longer specified predicate by predicate, but all at once and can use parens and any of the following operators: numeric fields: ==, !=, <, <=, >, >= string fields: ==, != predicates can be combined with the logical operators: &&, || examples: "common_preempt_count > 4" > filter "((sig >= 10 && sig < 15) || sig == 17) && comm != bash" > filter If there was an error, the erroneous string along with an error message can be seen by looking at the filter e.g.: ((sig >= 10 && sig < 15) || dsig == 17) && comm != bash ^ parse_error: Field not found Currently the caret for an error always appears at the beginning of the filter; a real position should be used, but the error message should be useful even without it. To clear a filter, '0' can be written to the filter file. Filters can also be set or cleared for a complete subsystem by writing the same filter as would be written to an individual event to the filter file at the root of the subsytem. Note however, that if any event in the subsystem lacks a field specified in the filter being set, the set will fail and all filters in the subsytem are automatically cleared. This change from the previous version was made because using only the fields that happen to exist for a given event would most likely result in a meaningless filter. Because the logical operators are now implemented as predicates, the maximum number of predicates in a filter was increased from 8 to 16. [ Impact: add new, extended trace-filter implementation ] Signed-off-by: Tom Zanussi <tzanussi@gmail.com> Acked-by: Steven Rostedt <rostedt@goodmis.org> Cc: fweisbec@gmail.com Cc: Li Zefan <lizf@cn.fujitsu.com> LKML-Reference: <1240905899.6416.121.camel@tropicana> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2009-04-28 16:04:59 +08:00
#define MAX_FILTER_PRED 32
#define MAX_FILTER_STR_VAL 256 /* Should handle KSYM_SYMBOL_LEN */
extern void destroy_preds(struct ftrace_event_call *call);
extern int filter_match_preds(struct event_filter *filter, void *rec);
extern int filter_current_check_discard(struct ring_buffer *buffer,
struct ftrace_event_call *call,
void *rec,
struct ring_buffer_event *event);
enum {
FILTER_OTHER = 0,
FILTER_STATIC_STRING,
FILTER_DYN_STRING,
FILTER_PTR_STRING,
};
extern int trace_event_raw_init(struct ftrace_event_call *call);
extern int trace_define_field(struct ftrace_event_call *call, const char *type,
const char *name, int offset, int size,
int is_signed, int filter_type);
tracing: Ftrace dynamic ftrace_event_call support Add dynamic ftrace_event_call support to ftrace. Trace engines can add new ftrace_event_call to ftrace on the fly. Each operator function of the call takes an ftrace_event_call data structure as an argument, because these functions may be shared among several ftrace_event_calls. Changes from v13: - Define remove_subsystem_dir() always (revirt a2ca5e03), because trace_remove_event_call() uses it. - Modify syscall tracer because of ftrace_event_call change. [fweisbec@gmail.com: Fixed conflict against latest tracing/core] Signed-off-by: Masami Hiramatsu <mhiramat@redhat.com> Cc: Ananth N Mavinakayanahalli <ananth@in.ibm.com> Cc: Avi Kivity <avi@redhat.com> Cc: Andi Kleen <ak@linux.intel.com> Cc: Christoph Hellwig <hch@infradead.org> Cc: Frank Ch. Eigler <fche@redhat.com> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Jason Baron <jbaron@redhat.com> Cc: Jim Keniston <jkenisto@us.ibm.com> Cc: K.Prasad <prasad@linux.vnet.ibm.com> Cc: Lai Jiangshan <laijs@cn.fujitsu.com> Cc: Li Zefan <lizf@cn.fujitsu.com> Cc: Przemysław Pawełczyk <przemyslaw@pawelczyk.it> Cc: Roland McGrath <roland@redhat.com> Cc: Sam Ravnborg <sam@ravnborg.org> Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Tom Zanussi <tzanussi@gmail.com> Cc: Vegard Nossum <vegard.nossum@gmail.com> LKML-Reference: <20090813203453.31965.71901.stgit@localhost.localdomain> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
2009-08-14 04:34:53 +08:00
extern int trace_add_event_call(struct ftrace_event_call *call);
extern void trace_remove_event_call(struct ftrace_event_call *call);
#define is_signed_type(type) (((type)(-1)) < 0)
int trace_set_clr_event(const char *system, const char *event, int set);
/*
* The double __builtin_constant_p is because gcc will give us an error
* if we try to allocate the static variable to fmt if it is not a
* constant. Even with the outer if statement optimizing out.
*/
#define event_trace_printk(ip, fmt, args...) \
do { \
__trace_printk_check_format(fmt, ##args); \
tracing_record_cmdline(current); \
if (__builtin_constant_p(fmt)) { \
static const char *trace_printk_fmt \
__attribute__((section("__trace_printk_fmt"))) = \
__builtin_constant_p(fmt) ? fmt : NULL; \
\
__trace_bprintk(ip, trace_printk_fmt, ##args); \
} else \
__trace_printk(ip, fmt, ##args); \
} while (0)
#ifdef CONFIG_PERF_EVENTS
struct perf_event;
perf: Take a hot regs snapshot for trace events We are taking a wrong regs snapshot when a trace event triggers. Either we use get_irq_regs(), which gives us the interrupted registers if we are in an interrupt, or we use task_pt_regs() which gives us the state before we entered the kernel, assuming we are lucky enough to be no kernel thread, in which case task_pt_regs() returns the initial set of regs when the kernel thread was started. What we want is different. We need a hot snapshot of the regs, so that we can get the instruction pointer to record in the sample, the frame pointer for the callchain, and some other things. Let's use the new perf_fetch_caller_regs() for that. Comparison with perf record -e lock: -R -a -f -g Before: perf [kernel] [k] __do_softirq | --- __do_softirq | |--55.16%-- __open | --44.84%-- __write_nocancel After: perf [kernel] [k] perf_tp_event | --- perf_tp_event | |--41.07%-- lock_acquire | | | |--39.36%-- _raw_spin_lock | | | | | |--7.81%-- hrtimer_interrupt | | | smp_apic_timer_interrupt | | | apic_timer_interrupt The old case was producing unreliable callchains. Now having right frame and instruction pointers, we have the trace we want. Also syscalls and kprobe events already have the right regs, let's use them instead of wasting a retrieval. v2: Follow the rename perf_save_regs() -> perf_fetch_caller_regs() Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Jason Baron <jbaron@redhat.com> Cc: Archs <linux-arch@vger.kernel.org>
2010-03-03 14:16:16 +08:00
DECLARE_PER_CPU(struct pt_regs, perf_trace_regs);
extern int perf_trace_enable(int event_id);
extern void perf_trace_disable(int event_id);
extern int ftrace_profile_set_filter(struct perf_event *event, int event_id,
char *filter_str);
extern void ftrace_profile_free_filter(struct perf_event *event);
extern void *
perf_trace_buf_prepare(int size, unsigned short type, int *rctxp,
unsigned long *irq_flags);
static inline void
perf_trace_buf_submit(void *raw_data, int size, int rctx, u64 addr,
perf: Take a hot regs snapshot for trace events We are taking a wrong regs snapshot when a trace event triggers. Either we use get_irq_regs(), which gives us the interrupted registers if we are in an interrupt, or we use task_pt_regs() which gives us the state before we entered the kernel, assuming we are lucky enough to be no kernel thread, in which case task_pt_regs() returns the initial set of regs when the kernel thread was started. What we want is different. We need a hot snapshot of the regs, so that we can get the instruction pointer to record in the sample, the frame pointer for the callchain, and some other things. Let's use the new perf_fetch_caller_regs() for that. Comparison with perf record -e lock: -R -a -f -g Before: perf [kernel] [k] __do_softirq | --- __do_softirq | |--55.16%-- __open | --44.84%-- __write_nocancel After: perf [kernel] [k] perf_tp_event | --- perf_tp_event | |--41.07%-- lock_acquire | | | |--39.36%-- _raw_spin_lock | | | | | |--7.81%-- hrtimer_interrupt | | | smp_apic_timer_interrupt | | | apic_timer_interrupt The old case was producing unreliable callchains. Now having right frame and instruction pointers, we have the trace we want. Also syscalls and kprobe events already have the right regs, let's use them instead of wasting a retrieval. v2: Follow the rename perf_save_regs() -> perf_fetch_caller_regs() Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Jason Baron <jbaron@redhat.com> Cc: Archs <linux-arch@vger.kernel.org>
2010-03-03 14:16:16 +08:00
u64 count, unsigned long irq_flags, struct pt_regs *regs)
{
struct trace_entry *entry = raw_data;
perf: Take a hot regs snapshot for trace events We are taking a wrong regs snapshot when a trace event triggers. Either we use get_irq_regs(), which gives us the interrupted registers if we are in an interrupt, or we use task_pt_regs() which gives us the state before we entered the kernel, assuming we are lucky enough to be no kernel thread, in which case task_pt_regs() returns the initial set of regs when the kernel thread was started. What we want is different. We need a hot snapshot of the regs, so that we can get the instruction pointer to record in the sample, the frame pointer for the callchain, and some other things. Let's use the new perf_fetch_caller_regs() for that. Comparison with perf record -e lock: -R -a -f -g Before: perf [kernel] [k] __do_softirq | --- __do_softirq | |--55.16%-- __open | --44.84%-- __write_nocancel After: perf [kernel] [k] perf_tp_event | --- perf_tp_event | |--41.07%-- lock_acquire | | | |--39.36%-- _raw_spin_lock | | | | | |--7.81%-- hrtimer_interrupt | | | smp_apic_timer_interrupt | | | apic_timer_interrupt The old case was producing unreliable callchains. Now having right frame and instruction pointers, we have the trace we want. Also syscalls and kprobe events already have the right regs, let's use them instead of wasting a retrieval. v2: Follow the rename perf_save_regs() -> perf_fetch_caller_regs() Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Ingo Molnar <mingo@elte.hu> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Steven Rostedt <rostedt@goodmis.org> Cc: Arnaldo Carvalho de Melo <acme@redhat.com> Cc: Masami Hiramatsu <mhiramat@redhat.com> Cc: Jason Baron <jbaron@redhat.com> Cc: Archs <linux-arch@vger.kernel.org>
2010-03-03 14:16:16 +08:00
perf_tp_event(entry->type, addr, count, raw_data, size, regs);
perf_swevent_put_recursion_context(rctx);
local_irq_restore(irq_flags);
}
#endif
#endif /* _LINUX_FTRACE_EVENT_H */