perf timechart: dynamically determine event fields offset
authorStanislav Fomichev <stfomichev@yandex-team.ru>
Wed, 27 Nov 2013 10:45:00 +0000 (14:45 +0400)
committerArnaldo Carvalho de Melo <acme@redhat.com>
Wed, 27 Nov 2013 18:10:11 +0000 (15:10 -0300)
Since b000c8065a92 "tracing: Remove the extra 4 bytes of padding in
events" removed padding bytes, perf timechart got out of sync with the
kernel's trace_entry structure.

Convert perf timechart to use dynamic fields offsets (via
perf_evsel__intval) not relying on a hardcoded copy of fields layout
from the kernel.

Signed-off-by: Stanislav Fomichev <stfomichev@yandex-team.ru>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Chia-I Wu <olvaffe@gmail.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Steven Rostedt <rostedt@goodmis.org>
Link: http://lkml.kernel.org/r/20131127104459.GB3309@stfomichev-desktop
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
tools/perf/builtin-timechart.c

index 491662bdfe0bb7a9ad97b59f2d26f20d273ed894..436cb5f9d751cd34111b3f759c849a3b9d61f850 100644 (file)
@@ -306,50 +306,10 @@ static int process_exit_event(struct perf_tool *tool __maybe_unused,
        return 0;
 }
 
-struct trace_entry {
-       unsigned short          type;
-       unsigned char           flags;
-       unsigned char           preempt_count;
-       int                     pid;
-       int                     lock_depth;
-};
-
 #ifdef SUPPORT_OLD_POWER_EVENTS
 static int use_old_power_events;
-struct power_entry_old {
-       struct trace_entry te;
-       u64     type;
-       u64     value;
-       u64     cpu_id;
-};
 #endif
 
-struct power_processor_entry {
-       struct trace_entry te;
-       u32     state;
-       u32     cpu_id;
-};
-
-#define TASK_COMM_LEN 16
-struct wakeup_entry {
-       struct trace_entry te;
-       char comm[TASK_COMM_LEN];
-       int   pid;
-       int   prio;
-       int   success;
-};
-
-struct sched_switch {
-       struct trace_entry te;
-       char prev_comm[TASK_COMM_LEN];
-       int  prev_pid;
-       int  prev_prio;
-       long prev_state; /* Arjan weeps. */
-       char next_comm[TASK_COMM_LEN];
-       int  next_pid;
-       int  next_prio;
-};
-
 static void c_state_start(int cpu, u64 timestamp, int state)
 {
        cpus_cstate_start_times[cpu] = timestamp;
@@ -409,25 +369,23 @@ static void p_state_change(int cpu, u64 timestamp, u64 new_freq)
                        turbo_frequency = max_freq;
 }
 
-static void
-sched_wakeup(int cpu, u64 timestamp, int pid, struct trace_entry *te,
-            const char *backtrace)
+static void sched_wakeup(int cpu, u64 timestamp, int waker, int wakee,
+                        u8 flags, const char *backtrace)
 {
        struct per_pid *p;
-       struct wakeup_entry *wake = (void *)te;
        struct wake_event *we = zalloc(sizeof(*we));
 
        if (!we)
                return;
 
        we->time = timestamp;
-       we->waker = pid;
+       we->waker = waker;
        we->backtrace = backtrace;
 
-       if ((te->flags & TRACE_FLAG_HARDIRQ) || (te->flags & TRACE_FLAG_SOFTIRQ))
+       if ((flags & TRACE_FLAG_HARDIRQ) || (flags & TRACE_FLAG_SOFTIRQ))
                we->waker = -1;
 
-       we->wakee = wake->pid;
+       we->wakee = wakee;
        we->next = wake_events;
        wake_events = we;
        p = find_create_pid(we->wakee);
@@ -444,24 +402,22 @@ sched_wakeup(int cpu, u64 timestamp, int pid, struct trace_entry *te,
        }
 }
 
-static void sched_switch(int cpu, u64 timestamp, struct trace_entry *te,
-                        const char *backtrace)
+static void sched_switch(int cpu, u64 timestamp, int prev_pid, int next_pid,
+                        u64 prev_state, const char *backtrace)
 {
        struct per_pid *p = NULL, *prev_p;
-       struct sched_switch *sw = (void *)te;
-
 
-       prev_p = find_create_pid(sw->prev_pid);
+       prev_p = find_create_pid(prev_pid);
 
-       p = find_create_pid(sw->next_pid);
+       p = find_create_pid(next_pid);
 
        if (prev_p->current && prev_p->current->state != TYPE_NONE)
-               pid_put_sample(sw->prev_pid, TYPE_RUNNING, cpu,
+               pid_put_sample(prev_pid, TYPE_RUNNING, cpu,
                               prev_p->current->state_since, timestamp,
                               backtrace);
        if (p && p->current) {
                if (p->current->state != TYPE_NONE)
-                       pid_put_sample(sw->next_pid, p->current->state, cpu,
+                       pid_put_sample(next_pid, p->current->state, cpu,
                                       p->current->state_since, timestamp,
                                       backtrace);
 
@@ -472,9 +428,9 @@ static void sched_switch(int cpu, u64 timestamp, struct trace_entry *te,
        if (prev_p->current) {
                prev_p->current->state = TYPE_NONE;
                prev_p->current->state_since = timestamp;
-               if (sw->prev_state & 2)
+               if (prev_state & 2)
                        prev_p->current->state = TYPE_BLOCKED;
-               if (sw->prev_state == 0)
+               if (prev_state == 0)
                        prev_p->current->state = TYPE_WAITING;
        }
 }
@@ -586,61 +542,69 @@ static int process_sample_event(struct perf_tool *tool __maybe_unused,
 }
 
 static int
-process_sample_cpu_idle(struct perf_evsel *evsel __maybe_unused,
+process_sample_cpu_idle(struct perf_evsel *evsel,
                        struct perf_sample *sample,
                        const char *backtrace __maybe_unused)
 {
-       struct power_processor_entry *ppe = sample->raw_data;
+       u32 state = perf_evsel__intval(evsel, sample, "state");
+       u32 cpu_id = perf_evsel__intval(evsel, sample, "cpu_id");
 
-       if (ppe->state == (u32) PWR_EVENT_EXIT)
-               c_state_end(ppe->cpu_id, sample->time);
+       if (state == (u32)PWR_EVENT_EXIT)
+               c_state_end(cpu_id, sample->time);
        else
-               c_state_start(ppe->cpu_id, sample->time, ppe->state);
+               c_state_start(cpu_id, sample->time, state);
        return 0;
 }
 
 static int
-process_sample_cpu_frequency(struct perf_evsel *evsel __maybe_unused,
+process_sample_cpu_frequency(struct perf_evsel *evsel,
                             struct perf_sample *sample,
                             const char *backtrace __maybe_unused)
 {
-       struct power_processor_entry *ppe = sample->raw_data;
+       u32 state = perf_evsel__intval(evsel, sample, "state");
+       u32 cpu_id = perf_evsel__intval(evsel, sample, "cpu_id");
 
-       p_state_change(ppe->cpu_id, sample->time, ppe->state);
+       p_state_change(cpu_id, sample->time, state);
        return 0;
 }
 
 static int
-process_sample_sched_wakeup(struct perf_evsel *evsel __maybe_unused,
+process_sample_sched_wakeup(struct perf_evsel *evsel,
                            struct perf_sample *sample,
                            const char *backtrace)
 {
-       struct trace_entry *te = sample->raw_data;
+       u8 flags = perf_evsel__intval(evsel, sample, "common_flags");
+       int waker = perf_evsel__intval(evsel, sample, "common_pid");
+       int wakee = perf_evsel__intval(evsel, sample, "pid");
 
-       sched_wakeup(sample->cpu, sample->time, sample->pid, te, backtrace);
+       sched_wakeup(sample->cpu, sample->time, waker, wakee, flags, backtrace);
        return 0;
 }
 
 static int
-process_sample_sched_switch(struct perf_evsel *evsel __maybe_unused,
+process_sample_sched_switch(struct perf_evsel *evsel,
                            struct perf_sample *sample,
                            const char *backtrace)
 {
-       struct trace_entry *te = sample->raw_data;
+       int prev_pid = perf_evsel__intval(evsel, sample, "prev_pid");
+       int next_pid = perf_evsel__intval(evsel, sample, "next_pid");
+       u64 prev_state = perf_evsel__intval(evsel, sample, "prev_state");
 
-       sched_switch(sample->cpu, sample->time, te, backtrace);
+       sched_switch(sample->cpu, sample->time, prev_pid, next_pid, prev_state,
+                    backtrace);
        return 0;
 }
 
 #ifdef SUPPORT_OLD_POWER_EVENTS
 static int
-process_sample_power_start(struct perf_evsel *evsel __maybe_unused,
+process_sample_power_start(struct perf_evsel *evsel,
                           struct perf_sample *sample,
                           const char *backtrace __maybe_unused)
 {
-       struct power_entry_old *peo = sample->raw_data;
+       u64 cpu_id = perf_evsel__intval(evsel, sample, "cpu_id");
+       u64 value = perf_evsel__intval(evsel, sample, "value");
 
-       c_state_start(peo->cpu_id, sample->time, peo->value);
+       c_state_start(cpu_id, sample->time, value);
        return 0;
 }
 
@@ -654,13 +618,14 @@ process_sample_power_end(struct perf_evsel *evsel __maybe_unused,
 }
 
 static int
-process_sample_power_frequency(struct perf_evsel *evsel __maybe_unused,
+process_sample_power_frequency(struct perf_evsel *evsel,
                               struct perf_sample *sample,
                               const char *backtrace __maybe_unused)
 {
-       struct power_entry_old *peo = sample->raw_data;
+       u64 cpu_id = perf_evsel__intval(evsel, sample, "cpu_id");
+       u64 value = perf_evsel__intval(evsel, sample, "value");
 
-       p_state_change(peo->cpu_id, sample->time, peo->value);
+       p_state_change(cpu_id, sample->time, value);
        return 0;
 }
 #endif /* SUPPORT_OLD_POWER_EVENTS */