perf evsel: Cache associated event_format

We already lookup the associated event_format when reading the perf.data
header, so that we can cache the tracepoint name in evsel->name, so do
it a little further and save the event_format itself, so that we can
avoid relookups in tools that need to access it.

Change the tools to take the most obvious advantage, when they were
using pevent_find_event directly. More work is needed for further
removing the need of a pointer to pevent, such as when asking for event
field values ("common_pid" and the other common fields and per
event_format fields).

This is something that was planned but only got actually done when
Andrey Wagin needed to do this lookup at perf_tool->sample() time, when
we don't have access to pevent (session->pevent) to use with
pevent_find_event().

Cc: Andrey Wagin <avagin@gmail.com>
Cc: David Ahern <dsahern@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Namhyung Kim <namhyung@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Link: http://lkml.kernel.org/n/tip-txkvew2ckko0b594ae8fbnyk@git.kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>

10 files changed, 70 insertions, 114 deletions

diff --git a/tools/perf/builtin-kmem.c b/tools/perf/builtin-kmem.c
index ce35015f2dc6..ffb93f424953 100644
--- a/tools/perf/builtin-kmem.c
+++ b/tools/perf/builtin-kmem.c
@@ -1,6 +1,7 @@
 #include "builtin.h"
 #include "perf.h"
 
+#include "util/evsel.h"
 #include "util/util.h"
 #include "util/cache.h"
 #include "util/symbol.h"
@@ -57,11 +58,6 @@ static unsigned long nr_allocs, nr_cross_allocs;
 
 #define PATH_SYS_NODE   "/sys/devices/system/node"
 
-struct perf_kmem {
-        struct perf_tool    tool;
-        struct perf_session *session;
-};
-
 static void init_cpunode_map(void)
 {
         FILE *fp;
@@ -283,16 +279,10 @@ static void process_free_event(void *data,
         s_alloc->alloc_cpu = -1;
 }
 
-static void process_raw_event(struct perf_tool *tool,
-                              union perf_event *raw_event __used, void *data,
+static void process_raw_event(struct perf_evsel *evsel, void *data,
                               int cpu, u64 timestamp, struct thread *thread)
 {
-        struct perf_kmem *kmem = container_of(tool, struct perf_kmem, tool);
-        struct event_format *event;
-        int type;
-
-        type = trace_parse_common_type(kmem->session->pevent, data);
-        event = pevent_find_event(kmem->session->pevent, type);
+        struct event_format *event = evsel->tp_format;
 
         if (!strcmp(event->name, "kmalloc") ||
             !strcmp(event->name, "kmem_cache_alloc")) {
@@ -313,10 +303,10 @@ static void process_raw_event(struct perf_tool *tool,
         }
 }
 
-static int process_sample_event(struct perf_tool *tool,
+static int process_sample_event(struct perf_tool *tool __used,
                                 union perf_event *event,
                                 struct perf_sample *sample,
-                                struct perf_evsel *evsel __used,
+                                struct perf_evsel *evsel,
                                 struct machine *machine)
 {
         struct thread *thread = machine__findnew_thread(machine, event->ip.pid);
@@ -329,18 +319,16 @@ static int process_sample_event(struct perf_tool *tool,
 
         dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid);
 
-        process_raw_event(tool, event, sample->raw_data, sample->cpu,
+        process_raw_event(evsel, sample->raw_data, sample->cpu,
                           sample->time, thread);
 
         return 0;
 }
 
-static struct perf_kmem perf_kmem = {
-        .tool = {
-                .sample                 = process_sample_event,
-                .comm                   = perf_event__process_comm,
-                .ordered_samples        = true,
-        },
+static struct perf_tool perf_kmem = {
+        .sample          = process_sample_event,
+        .comm            = perf_event__process_comm,
+        .ordered_samples = true,
 };
 
 static double fragmentation(unsigned long n_req, unsigned long n_alloc)
@@ -497,13 +485,10 @@ static int __cmd_kmem(void)
         int err = -EINVAL;
         struct perf_session *session;
 
-        session = perf_session__new(input_name, O_RDONLY, 0, false,
-                                    &perf_kmem.tool);
+        session = perf_session__new(input_name, O_RDONLY, 0, false, &perf_kmem);
         if (session == NULL)
                 return -ENOMEM;
 
-        perf_kmem.session = session;
-
         if (perf_session__create_kernel_maps(session) < 0)
                 goto out_delete;
 
@@ -511,7 +496,7 @@ static int __cmd_kmem(void)
                 goto out_delete;
 
         setup_pager();
-        err = perf_session__process_events(session, &perf_kmem.tool);
+        err = perf_session__process_events(session, &perf_kmem);
         if (err != 0)
                 goto out_delete;
         sort_result();
diff --git a/tools/perf/builtin-lock.c b/tools/perf/builtin-lock.c
index b3c428548868..142b3033e4be 100644
--- a/tools/perf/builtin-lock.c
+++ b/tools/perf/builtin-lock.c
@@ -1,6 +1,7 @@
 #include "builtin.h"
 #include "perf.h"
 
+#include "util/evsel.h"
 #include "util/util.h"
 #include "util/cache.h"
 #include "util/symbol.h"
@@ -718,14 +719,10 @@ process_lock_release_event(void *data,
                 trace_handler->release_event(&release_event, event, cpu, timestamp, thread);
 }
 
-static void
-process_raw_event(void *data, int cpu, u64 timestamp, struct thread *thread)
+static void process_raw_event(struct perf_evsel *evsel, void *data, int cpu,
+                              u64 timestamp, struct thread *thread)
 {
-        struct event_format *event;
-        int type;
-
-        type = trace_parse_common_type(session->pevent, data);
-        event = pevent_find_event(session->pevent, type);
+        struct event_format *event = evsel->tp_format;
 
         if (!strcmp(event->name, "lock_acquire"))
                 process_lock_acquire_event(data, event, cpu, timestamp, thread);
@@ -849,7 +846,7 @@ static void dump_info(void)
 static int process_sample_event(struct perf_tool *tool __used,
                                 union perf_event *event,
                                 struct perf_sample *sample,
-                                struct perf_evsel *evsel __used,
+                                struct perf_evsel *evsel,
                                 struct machine *machine)
 {
         struct thread *thread = machine__findnew_thread(machine, sample->tid);
@@ -860,7 +857,7 @@ static int process_sample_event(struct perf_tool *tool __used,
                 return -1;
         }
 
-        process_raw_event(sample->raw_data, sample->cpu, sample->time, thread);
+        process_raw_event(evsel, sample->raw_data, sample->cpu, sample->time, thread);
 
         return 0;