* [GIT PULL 0/4] perf/core improvements
@ 2010-12-02 19:39 Arnaldo Carvalho de Melo
2010-12-02 19:39 ` [PATCH 1/4] perf events: Separate the routines handling the PERF_SAMPLE_ identity fields Arnaldo Carvalho de Melo
` (5 more replies)
0 siblings, 6 replies; 15+ messages in thread
From: Arnaldo Carvalho de Melo @ 2010-12-02 19:39 UTC (permalink / raw)
To: Ingo Molnar
Cc: linux-kernel, Arnaldo Carvalho de Melo, Frederic Weisbecker,
Ian Munsie, Mike Galbraith, Paul Mackerras, Peter Zijlstra,
Stephane Eranian, Arnaldo Carvalho de Melo
Hi Ingo,
Please consider pulling after at least Ian and Peter acks it:
git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux-2.6 perf/core
Peter already acked a previous version of this, so I left his acked by
there, this new one is much simpler than the first one he acked, so I think
he'll be happy with this one as well, Peter?
I tested it using the perf tools built from what is in tip/perf/urgent
and it works as expected, i.e. the new fields are just discarded since they are
stashed as a trailer in the MMAP, FORK, etc events, so old tools process new
perf.data files just fine.
Also the new tools works on older kernels, also tested.
Ian, if you prefer, please test it before Ingo merges it.
Regards,
- Arnaldo
Arnaldo Carvalho de Melo (4):
perf events: Separate the routines handling the PERF_SAMPLE_ identity fields
perf events: Make sample_type identity fields available in all PERF_RECORD_ events
perf session: Parse sample earlier
perf tools: Ask for ID PERF_SAMPLE_ info on all PERF_RECORD_ events
include/linux/perf_event.h | 13 +-
kernel/perf_event.c | 207 +++++++++++++------
tools/perf/Documentation/perf-record.txt | 5 +
tools/perf/builtin-annotate.c | 6 +-
tools/perf/builtin-diff.c | 11 +-
tools/perf/builtin-inject.c | 39 ++--
tools/perf/builtin-kmem.c | 21 +--
tools/perf/builtin-lock.c | 12 +-
tools/perf/builtin-record.c | 23 ++
tools/perf/builtin-report.c | 15 +-
tools/perf/builtin-sched.c | 21 +--
tools/perf/builtin-script.c | 34 +--
tools/perf/builtin-timechart.c | 40 ++--
tools/perf/builtin-top.c | 12 +-
tools/perf/util/build-id.c | 7 +-
tools/perf/util/event.c | 341 +++++++++++++++++++----------
tools/perf/util/event.h | 27 ++-
tools/perf/util/header.c | 29 ++-
tools/perf/util/header.h | 1 +
tools/perf/util/hist.h | 2 +
tools/perf/util/session.c | 159 ++++++++++++---
tools/perf/util/session.h | 12 +-
22 files changed, 696 insertions(+), 341 deletions(-)
^ permalink raw reply [flat|nested] 15+ messages in thread
* [PATCH 1/4] perf events: Separate the routines handling the PERF_SAMPLE_ identity fields
2010-12-02 19:39 [GIT PULL 0/4] perf/core improvements Arnaldo Carvalho de Melo
@ 2010-12-02 19:39 ` Arnaldo Carvalho de Melo
2010-12-03 5:40 ` Ian Munsie
2010-12-02 19:39 ` [PATCH 2/4] perf events: Make sample_type identity fields available in all PERF_RECORD_ events Arnaldo Carvalho de Melo
` (4 subsequent siblings)
5 siblings, 1 reply; 15+ messages in thread
From: Arnaldo Carvalho de Melo @ 2010-12-02 19:39 UTC (permalink / raw)
To: Ingo Molnar
Cc: linux-kernel, Arnaldo Carvalho de Melo, Frederic Weisbecker,
Ian Munsie, Mike Galbraith, Peter Zijlstra, Paul Mackerras,
Stephane Eranian
From: Arnaldo Carvalho de Melo <acme@redhat.com>
Those will be made available in sample like events like MMAP, EXEC, etc in a
followup patch. So precalculate the extra id header space and have a separate
routine to fill them up.
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
include/linux/perf_event.h | 1 +
kernel/perf_event.c | 128 +++++++++++++++++++++++++------------------
2 files changed, 75 insertions(+), 54 deletions(-)
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index adf6d99..b9950b1 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -759,6 +759,7 @@ struct perf_event {
struct perf_event_attr attr;
u16 header_size;
+ u16 id_header_size;
u16 read_size;
struct hw_perf_event hw;
diff --git a/kernel/perf_event.c b/kernel/perf_event.c
index aede712..3b847b3 100644
--- a/kernel/perf_event.c
+++ b/kernel/perf_event.c
@@ -133,6 +133,28 @@ static void unclone_ctx(struct perf_event_context *ctx)
}
}
+static u32 perf_event_pid(struct perf_event *event, struct task_struct *p)
+{
+ /*
+ * only top level events have the pid namespace they were created in
+ */
+ if (event->parent)
+ event = event->parent;
+
+ return task_tgid_nr_ns(p, event->ns);
+}
+
+static u32 perf_event_tid(struct perf_event *event, struct task_struct *p)
+{
+ /*
+ * only top level events have the pid namespace they were created in
+ */
+ if (event->parent)
+ event = event->parent;
+
+ return task_pid_nr_ns(p, event->ns);
+}
+
/*
* If we inherit events we want to return the parent event id
* to userspace.
@@ -351,15 +373,30 @@ static void perf_event__header_size(struct perf_event *event)
if (sample_type & PERF_SAMPLE_IP)
size += sizeof(data->ip);
+ if (sample_type & PERF_SAMPLE_ADDR)
+ size += sizeof(data->addr);
+
+ if (sample_type & PERF_SAMPLE_PERIOD)
+ size += sizeof(data->period);
+
+ if (sample_type & PERF_SAMPLE_READ)
+ size += event->read_size;
+
+ event->header_size = size;
+}
+
+static void perf_event__id_header_size(struct perf_event *event)
+{
+ struct perf_sample_data *data;
+ u64 sample_type = event->attr.sample_type;
+ u16 size = 0;
+
if (sample_type & PERF_SAMPLE_TID)
size += sizeof(data->tid_entry);
if (sample_type & PERF_SAMPLE_TIME)
size += sizeof(data->time);
- if (sample_type & PERF_SAMPLE_ADDR)
- size += sizeof(data->addr);
-
if (sample_type & PERF_SAMPLE_ID)
size += sizeof(data->id);
@@ -369,13 +406,7 @@ static void perf_event__header_size(struct perf_event *event)
if (sample_type & PERF_SAMPLE_CPU)
size += sizeof(data->cpu_entry);
- if (sample_type & PERF_SAMPLE_PERIOD)
- size += sizeof(data->period);
-
- if (sample_type & PERF_SAMPLE_READ)
- size += event->read_size;
-
- event->header_size = size;
+ event->id_header_size = size;
}
static void perf_group_attach(struct perf_event *event)
@@ -3357,6 +3388,36 @@ __always_inline void perf_output_copy(struct perf_output_handle *handle,
} while (len);
}
+static void perf_event_header__init_id(struct perf_event_header *header,
+ struct perf_sample_data *data,
+ struct perf_event *event)
+{
+ u64 sample_type = event->attr.sample_type;
+
+ data->type = sample_type;
+ header->size += event->id_header_size;
+
+ if (sample_type & PERF_SAMPLE_TID) {
+ /* namespace issues */
+ data->tid_entry.pid = perf_event_pid(event, current);
+ data->tid_entry.tid = perf_event_tid(event, current);
+ }
+
+ if (sample_type & PERF_SAMPLE_TIME)
+ data->time = perf_clock();
+
+ if (sample_type & PERF_SAMPLE_ID)
+ data->id = primary_event_id(event);
+
+ if (sample_type & PERF_SAMPLE_STREAM_ID)
+ data->stream_id = event->id;
+
+ if (sample_type & PERF_SAMPLE_CPU) {
+ data->cpu_entry.cpu = raw_smp_processor_id();
+ data->cpu_entry.reserved = 0;
+ }
+}
+
int perf_output_begin(struct perf_output_handle *handle,
struct perf_event *event, unsigned int size,
int nmi, int sample)
@@ -3459,28 +3520,6 @@ void perf_output_end(struct perf_output_handle *handle)
rcu_read_unlock();
}
-static u32 perf_event_pid(struct perf_event *event, struct task_struct *p)
-{
- /*
- * only top level events have the pid namespace they were created in
- */
- if (event->parent)
- event = event->parent;
-
- return task_tgid_nr_ns(p, event->ns);
-}
-
-static u32 perf_event_tid(struct perf_event *event, struct task_struct *p)
-{
- /*
- * only top level events have the pid namespace they were created in
- */
- if (event->parent)
- event = event->parent;
-
- return task_pid_nr_ns(p, event->ns);
-}
-
static void perf_output_read_one(struct perf_output_handle *handle,
struct perf_event *event,
u64 enabled, u64 running)
@@ -3655,37 +3694,17 @@ void perf_prepare_sample(struct perf_event_header *header,
{
u64 sample_type = event->attr.sample_type;
- data->type = sample_type;
-
header->type = PERF_RECORD_SAMPLE;
header->size = sizeof(*header) + event->header_size;
header->misc = 0;
header->misc |= perf_misc_flags(regs);
+ perf_event_header__init_id(header, data, event);
+
if (sample_type & PERF_SAMPLE_IP)
data->ip = perf_instruction_pointer(regs);
- if (sample_type & PERF_SAMPLE_TID) {
- /* namespace issues */
- data->tid_entry.pid = perf_event_pid(event, current);
- data->tid_entry.tid = perf_event_tid(event, current);
- }
-
- if (sample_type & PERF_SAMPLE_TIME)
- data->time = perf_clock();
-
- if (sample_type & PERF_SAMPLE_ID)
- data->id = primary_event_id(event);
-
- if (sample_type & PERF_SAMPLE_STREAM_ID)
- data->stream_id = event->id;
-
- if (sample_type & PERF_SAMPLE_CPU) {
- data->cpu_entry.cpu = raw_smp_processor_id();
- data->cpu_entry.reserved = 0;
- }
-
if (sample_type & PERF_SAMPLE_CALLCHAIN) {
int size = 1;
@@ -5745,6 +5764,7 @@ SYSCALL_DEFINE5(perf_event_open,
* Precalculate sample_data sizes
*/
perf_event__header_size(event);
+ perf_event__id_header_size(event);
/*
* Drop the reference on the group_event after placing the
--
1.6.2.5
^ permalink raw reply related [flat|nested] 15+ messages in thread
* [PATCH 2/4] perf events: Make sample_type identity fields available in all PERF_RECORD_ events
2010-12-02 19:39 [GIT PULL 0/4] perf/core improvements Arnaldo Carvalho de Melo
2010-12-02 19:39 ` [PATCH 1/4] perf events: Separate the routines handling the PERF_SAMPLE_ identity fields Arnaldo Carvalho de Melo
@ 2010-12-02 19:39 ` Arnaldo Carvalho de Melo
2010-12-03 5:45 ` Ian Munsie
2010-12-02 19:39 ` [PATCH 3/4] perf session: Parse sample earlier Arnaldo Carvalho de Melo
` (3 subsequent siblings)
5 siblings, 1 reply; 15+ messages in thread
From: Arnaldo Carvalho de Melo @ 2010-12-02 19:39 UTC (permalink / raw)
To: Ingo Molnar
Cc: linux-kernel, Arnaldo Carvalho de Melo, Frederic Weisbecker,
Ian Munsie, Mike Galbraith, Peter Zijlstra, Paul Mackerras,
Stephane Eranian
From: Arnaldo Carvalho de Melo <acme@redhat.com>
If perf_event_attr.sample_id_all is set it will add the PERF_SAMPLE_ identity
info:
TID, TIME, ID, CPU, STREAM_ID
As a trailer, so that older perf tools can process new files, just ignoring the
extra payload.
With this its possible to do further analysis on problems in the event stream,
like detecting reordering of MMAP and FORK events, etc.
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
include/linux/perf_event.h | 12 ++++++-
kernel/perf_event.c | 79 ++++++++++++++++++++++++++++++++++++++++----
2 files changed, 83 insertions(+), 8 deletions(-)
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index b9950b1..2814ead 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -215,8 +215,9 @@ struct perf_event_attr {
*/
precise_ip : 2, /* skid constraint */
mmap_data : 1, /* non-exec mmap data */
+ sample_id_all : 1, /* sample_type all events */
- __reserved_1 : 46;
+ __reserved_1 : 45;
union {
__u32 wakeup_events; /* wakeup every n events */
@@ -327,6 +328,15 @@ struct perf_event_header {
enum perf_event_type {
/*
+ * If perf_event_attr.sample_id_all is set then all event types will
+ * have the sample_type selected fields related to where/when
+ * (identity) an event took place (TID, TIME, ID, CPU, STREAM_ID)
+ * described in PERF_RECORD_SAMPLE below, it will be stashed just after
+ * the perf_event_header and the fields already present for the existing
+ * fields, i.e. at the end of the payload. That way a newer perf.data
+ * file will be supported by older perf tools, with these new optional
+ * fields being ignored.
+ *
* The MMAP events record the PROT_EXEC mappings so that we can
* correlate userspace IPs to code. They have the following structure:
*
diff --git a/kernel/perf_event.c b/kernel/perf_event.c
index 3b847b3..bb77b8f 100644
--- a/kernel/perf_event.c
+++ b/kernel/perf_event.c
@@ -3418,6 +3418,35 @@ static void perf_event_header__init_id(struct perf_event_header *header,
}
}
+static void perf_output_id_sample(struct perf_output_handle *handle,
+ struct perf_sample_data *data)
+{
+ u64 sample_type = data->type;
+
+ if (sample_type & PERF_SAMPLE_TID)
+ perf_output_put(handle, data->tid_entry);
+
+ if (sample_type & PERF_SAMPLE_TIME)
+ perf_output_put(handle, data->time);
+
+ if (sample_type & PERF_SAMPLE_ID)
+ perf_output_put(handle, data->id);
+
+ if (sample_type & PERF_SAMPLE_STREAM_ID)
+ perf_output_put(handle, data->stream_id);
+
+ if (sample_type & PERF_SAMPLE_CPU)
+ perf_output_put(handle, data->cpu_entry);
+}
+
+static void perf_event__output_trailer(struct perf_event *event,
+ struct perf_output_handle *handle,
+ struct perf_sample_data *sample)
+{
+ if (event->attr.sample_id_all)
+ perf_output_id_sample(handle, sample);
+}
+
int perf_output_begin(struct perf_output_handle *handle,
struct perf_event *event, unsigned int size,
int nmi, int sample)
@@ -3425,6 +3454,7 @@ int perf_output_begin(struct perf_output_handle *handle,
struct perf_buffer *buffer;
unsigned long tail, offset, head;
int have_lost;
+ struct perf_sample_data sample_data;
struct {
struct perf_event_header header;
u64 id;
@@ -3451,8 +3481,13 @@ int perf_output_begin(struct perf_output_handle *handle,
goto out;
have_lost = local_read(&buffer->lost);
- if (have_lost)
- size += sizeof(lost_event);
+ if (have_lost) {
+ lost_event.header.size = sizeof(lost_event);
+ if (event->attr.sample_id_all)
+ perf_event_header__init_id(&lost_event.header,
+ &sample_data, event);
+ size += lost_event.header.size;
+ }
perf_output_get_handle(handle);
@@ -3483,11 +3518,11 @@ int perf_output_begin(struct perf_output_handle *handle,
if (have_lost) {
lost_event.header.type = PERF_RECORD_LOST;
lost_event.header.misc = 0;
- lost_event.header.size = sizeof(lost_event);
lost_event.id = event->id;
lost_event.lost = local_xchg(&buffer->lost, 0);
perf_output_put(handle, lost_event);
+ perf_event__output_trailer(event, handle, &sample_data);
}
return 0;
@@ -3768,6 +3803,7 @@ perf_event_read_event(struct perf_event *event,
struct task_struct *task)
{
struct perf_output_handle handle;
+ struct perf_sample_data sample;
struct perf_read_event read_event = {
.header = {
.type = PERF_RECORD_READ,
@@ -3779,12 +3815,15 @@ perf_event_read_event(struct perf_event *event,
};
int ret;
+ if (event->attr.sample_id_all)
+ perf_event_header__init_id(&read_event.header, &sample, event);
ret = perf_output_begin(&handle, event, read_event.header.size, 0, 0);
if (ret)
return;
perf_output_put(&handle, read_event);
perf_output_read(&handle, event);
+ perf_event__output_trailer(event, &handle, &sample);
perf_output_end(&handle);
}
@@ -3814,9 +3853,13 @@ static void perf_event_task_output(struct perf_event *event,
struct perf_task_event *task_event)
{
struct perf_output_handle handle;
+ struct perf_sample_data sample;
struct task_struct *task = task_event->task;
int size, ret;
+ if (event->attr.sample_id_all)
+ perf_event_header__init_id(&task_event->event_id.header,
+ &sample, event);
size = task_event->event_id.header.size;
ret = perf_output_begin(&handle, event, size, 0, 0);
@@ -3830,6 +3873,7 @@ static void perf_event_task_output(struct perf_event *event,
task_event->event_id.ptid = perf_event_tid(event, current);
perf_output_put(&handle, task_event->event_id);
+ perf_event__output_trailer(event, &handle, &sample);
perf_output_end(&handle);
}
@@ -3944,8 +3988,15 @@ static void perf_event_comm_output(struct perf_event *event,
struct perf_comm_event *comm_event)
{
struct perf_output_handle handle;
+ struct perf_sample_data sample;
int size = comm_event->event_id.header.size;
- int ret = perf_output_begin(&handle, event, size, 0, 0);
+ int ret;
+
+ if (event->attr.sample_id_all)
+ perf_event_header__init_id(&comm_event->event_id.header,
+ &sample, event);
+ size = comm_event->event_id.header.size;
+ ret = perf_output_begin(&handle, event, size, 0, 0);
if (ret)
return;
@@ -3956,6 +4007,7 @@ static void perf_event_comm_output(struct perf_event *event,
perf_output_put(&handle, comm_event->event_id);
perf_output_copy(&handle, comm_event->comm,
comm_event->comm_size);
+ perf_event__output_trailer(event, &handle, &sample);
perf_output_end(&handle);
}
@@ -4080,8 +4132,14 @@ static void perf_event_mmap_output(struct perf_event *event,
struct perf_mmap_event *mmap_event)
{
struct perf_output_handle handle;
- int size = mmap_event->event_id.header.size;
- int ret = perf_output_begin(&handle, event, size, 0, 0);
+ struct perf_sample_data sample;
+ int size, ret;
+
+ if (event->attr.sample_id_all)
+ perf_event_header__init_id(&mmap_event->event_id.header,
+ &sample, event);
+ size = mmap_event->event_id.header.size;
+ ret = perf_output_begin(&handle, event, size, 0, 0);
if (ret)
return;
@@ -4092,6 +4150,7 @@ static void perf_event_mmap_output(struct perf_event *event,
perf_output_put(&handle, mmap_event->event_id);
perf_output_copy(&handle, mmap_event->file_name,
mmap_event->file_size);
+ perf_event__output_trailer(event, &handle, &sample);
perf_output_end(&handle);
}
@@ -4245,6 +4304,7 @@ void perf_event_mmap(struct vm_area_struct *vma)
static void perf_log_throttle(struct perf_event *event, int enable)
{
struct perf_output_handle handle;
+ struct perf_sample_data sample;
int ret;
struct {
@@ -4266,11 +4326,16 @@ static void perf_log_throttle(struct perf_event *event, int enable)
if (enable)
throttle_event.header.type = PERF_RECORD_UNTHROTTLE;
- ret = perf_output_begin(&handle, event, sizeof(throttle_event), 1, 0);
+ if (event->attr.sample_id_all)
+ perf_event_header__init_id(&throttle_event.header,
+ &sample, event);
+ ret = perf_output_begin(&handle, event,
+ throttle_event.header.size, 1, 0);
if (ret)
return;
perf_output_put(&handle, throttle_event);
+ perf_event__output_trailer(event, &handle, &sample);
perf_output_end(&handle);
}
--
1.6.2.5
^ permalink raw reply related [flat|nested] 15+ messages in thread
* [PATCH 3/4] perf session: Parse sample earlier
2010-12-02 19:39 [GIT PULL 0/4] perf/core improvements Arnaldo Carvalho de Melo
2010-12-02 19:39 ` [PATCH 1/4] perf events: Separate the routines handling the PERF_SAMPLE_ identity fields Arnaldo Carvalho de Melo
2010-12-02 19:39 ` [PATCH 2/4] perf events: Make sample_type identity fields available in all PERF_RECORD_ events Arnaldo Carvalho de Melo
@ 2010-12-02 19:39 ` Arnaldo Carvalho de Melo
2010-12-03 5:46 ` Ian Munsie
2010-12-02 19:39 ` [PATCH 4/4] perf tools: Ask for ID PERF_SAMPLE_ info on all PERF_RECORD_ events Arnaldo Carvalho de Melo
` (2 subsequent siblings)
5 siblings, 1 reply; 15+ messages in thread
From: Arnaldo Carvalho de Melo @ 2010-12-02 19:39 UTC (permalink / raw)
To: Ingo Molnar
Cc: linux-kernel, Arnaldo Carvalho de Melo, Frederic Weisbecker,
Ian Munsie, Mike Galbraith, Peter Zijlstra, Paul Mackerras,
Stephane Eranian
From: Arnaldo Carvalho de Melo <acme@redhat.com>
At perf_session__process_event, so that we reduce the number of lines in eache
tool sample processing routine that now receives a sample_data pointer already
parsed.
This will also be useful in the next patch, where we'll allow sample the
identity fields in MMAP, FORK, EXIT, etc, when it will be possible to see (cpu,
timestamp) just after before every event.
Also validate callchains in perf_session__process_event, i.e. as early as
possible, and keep a counter of the number of events discarded due to invalid
callchains, warning the user about it if it happens.
There is an assumption that was kept that all events have the same sample_type,
that will be dealt with in the future, when this preexisting limitation will be
removed.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
tools/perf/builtin-annotate.c | 6 +-
tools/perf/builtin-diff.c | 11 ++--
tools/perf/builtin-inject.c | 39 +++++++++------
tools/perf/builtin-kmem.c | 21 ++------
tools/perf/builtin-lock.c | 12 ++---
tools/perf/builtin-record.c | 7 +++
tools/perf/builtin-report.c | 15 +++---
tools/perf/builtin-sched.c | 21 ++------
tools/perf/builtin-script.c | 34 +++++---------
tools/perf/builtin-timechart.c | 40 ++++++++--------
tools/perf/builtin-top.c | 4 +-
tools/perf/util/build-id.c | 7 ++-
tools/perf/util/event.c | 66 +++++++++++---------------
tools/perf/util/event.h | 20 +++++--
| 11 ++--
tools/perf/util/hist.h | 2 +
tools/perf/util/session.c | 103 ++++++++++++++++++++++++++++++----------
tools/perf/util/session.h | 9 ++-
18 files changed, 236 insertions(+), 192 deletions(-)
diff --git a/tools/perf/builtin-annotate.c b/tools/perf/builtin-annotate.c
index 6d5604d..569a276 100644
--- a/tools/perf/builtin-annotate.c
+++ b/tools/perf/builtin-annotate.c
@@ -58,12 +58,12 @@ static int hists__add_entry(struct hists *self, struct addr_location *al)
return hist_entry__inc_addr_samples(he, al->addr);
}
-static int process_sample_event(event_t *event, struct perf_session *session)
+static int process_sample_event(event_t *event, struct sample_data *sample,
+ struct perf_session *session)
{
struct addr_location al;
- struct sample_data data;
- if (event__preprocess_sample(event, session, &al, &data, NULL) < 0) {
+ if (event__preprocess_sample(event, session, &al, sample, NULL) < 0) {
pr_warning("problem processing %d event, skipping it.\n",
event->header.type);
return -1;
diff --git a/tools/perf/builtin-diff.c b/tools/perf/builtin-diff.c
index 221b823..5e1a043 100644
--- a/tools/perf/builtin-diff.c
+++ b/tools/perf/builtin-diff.c
@@ -30,12 +30,13 @@ static int hists__add_entry(struct hists *self,
return -ENOMEM;
}
-static int diff__process_sample_event(event_t *event, struct perf_session *session)
+static int diff__process_sample_event(event_t *event,
+ struct sample_data *sample,
+ struct perf_session *session)
{
struct addr_location al;
- struct sample_data data = { .period = 1, };
- if (event__preprocess_sample(event, session, &al, &data, NULL) < 0) {
+ if (event__preprocess_sample(event, session, &al, sample, NULL) < 0) {
pr_warning("problem processing %d event, skipping it.\n",
event->header.type);
return -1;
@@ -44,12 +45,12 @@ static int diff__process_sample_event(event_t *event, struct perf_session *sessi
if (al.filtered || al.sym == NULL)
return 0;
- if (hists__add_entry(&session->hists, &al, data.period)) {
+ if (hists__add_entry(&session->hists, &al, sample->period)) {
pr_warning("problem incrementing symbol period, skipping event\n");
return -1;
}
- session->hists.stats.total_period += data.period;
+ session->hists.stats.total_period += sample->period;
return 0;
}
diff --git a/tools/perf/builtin-inject.c b/tools/perf/builtin-inject.c
index 8e3e47b..4b66b85 100644
--- a/tools/perf/builtin-inject.c
+++ b/tools/perf/builtin-inject.c
@@ -16,8 +16,8 @@
static char const *input_name = "-";
static bool inject_build_ids;
-static int event__repipe(event_t *event __used,
- struct perf_session *session __used)
+static int event__repipe_synth(event_t *event,
+ struct perf_session *session __used)
{
uint32_t size;
void *buf = event;
@@ -36,22 +36,30 @@ static int event__repipe(event_t *event __used,
return 0;
}
-static int event__repipe_mmap(event_t *self, struct perf_session *session)
+static int event__repipe(event_t *event, struct sample_data *sample __used,
+ struct perf_session *session)
+{
+ return event__repipe_synth(event, session);
+}
+
+static int event__repipe_mmap(event_t *self, struct sample_data *sample,
+ struct perf_session *session)
{
int err;
- err = event__process_mmap(self, session);
- event__repipe(self, session);
+ err = event__process_mmap(self, sample, session);
+ event__repipe(self, sample, session);
return err;
}
-static int event__repipe_task(event_t *self, struct perf_session *session)
+static int event__repipe_task(event_t *self, struct sample_data *sample,
+ struct perf_session *session)
{
int err;
- err = event__process_task(self, session);
- event__repipe(self, session);
+ err = event__process_task(self, sample, session);
+ event__repipe(self, sample, session);
return err;
}
@@ -61,7 +69,7 @@ static int event__repipe_tracing_data(event_t *self,
{
int err;
- event__repipe(self, session);
+ event__repipe_synth(self, session);
err = event__process_tracing_data(self, session);
return err;
@@ -111,7 +119,8 @@ static int dso__inject_build_id(struct dso *self, struct perf_session *session)
return 0;
}
-static int event__inject_buildid(event_t *event, struct perf_session *session)
+static int event__inject_buildid(event_t *event, struct sample_data *sample,
+ struct perf_session *session)
{
struct addr_location al;
struct thread *thread;
@@ -146,7 +155,7 @@ static int event__inject_buildid(event_t *event, struct perf_session *session)
}
repipe:
- event__repipe(event, session);
+ event__repipe(event, sample, session);
return 0;
}
@@ -160,10 +169,10 @@ struct perf_event_ops inject_ops = {
.read = event__repipe,
.throttle = event__repipe,
.unthrottle = event__repipe,
- .attr = event__repipe,
- .event_type = event__repipe,
- .tracing_data = event__repipe,
- .build_id = event__repipe,
+ .attr = event__repipe_synth,
+ .event_type = event__repipe_synth,
+ .tracing_data = event__repipe_synth,
+ .build_id = event__repipe_synth,
};
extern volatile int session_done;
diff --git a/tools/perf/builtin-kmem.c b/tools/perf/builtin-kmem.c
index 31f60a2..d0a652e 100644
--- a/tools/perf/builtin-kmem.c
+++ b/tools/perf/builtin-kmem.c
@@ -304,22 +304,11 @@ process_raw_event(event_t *raw_event __used, void *data,
}
}
-static int process_sample_event(event_t *event, struct perf_session *session)
+static int process_sample_event(event_t *event, struct sample_data *sample,
+ struct perf_session *session)
{
- struct sample_data data;
- struct thread *thread;
+ struct thread *thread = perf_session__findnew(session, event->ip.pid);
- memset(&data, 0, sizeof(data));
- data.time = -1;
- data.cpu = -1;
- data.period = 1;
-
- event__parse_sample(event, session->sample_type, &data);
-
- dump_printf("(IP, %d): %d/%d: %#Lx period: %Ld\n", event->header.misc,
- data.pid, data.tid, data.ip, data.period);
-
- thread = perf_session__findnew(session, event->ip.pid);
if (thread == NULL) {
pr_debug("problem processing %d event, skipping it.\n",
event->header.type);
@@ -328,8 +317,8 @@ static int process_sample_event(event_t *event, struct perf_session *session)
dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid);
- process_raw_event(event, data.raw_data, data.cpu,
- data.time, thread);
+ process_raw_event(event, sample->raw_data, sample->cpu,
+ sample->time, thread);
return 0;
}
diff --git a/tools/perf/builtin-lock.c b/tools/perf/builtin-lock.c
index 8452a2a..92d3da5 100644
--- a/tools/perf/builtin-lock.c
+++ b/tools/perf/builtin-lock.c
@@ -834,22 +834,18 @@ static void dump_info(void)
die("Unknown type of information\n");
}
-static int process_sample_event(event_t *self, struct perf_session *s)
+static int process_sample_event(event_t *self, struct sample_data *sample,
+ struct perf_session *s)
{
- struct sample_data data;
- struct thread *thread;
+ struct thread *thread = perf_session__findnew(s, sample->tid);
- bzero(&data, sizeof(data));
- event__parse_sample(self, s->sample_type, &data);
-
- thread = perf_session__findnew(s, data.tid);
if (thread == NULL) {
pr_debug("problem processing %d event, skipping it.\n",
self->header.type);
return -1;
}
- process_raw_event(data.raw_data, data.cpu, data.time, thread);
+ process_raw_event(sample->raw_data, sample->cpu, sample->time, thread);
return 0;
}
diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index 024e144..b34de92 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -36,6 +36,7 @@ static int *fd[MAX_NR_CPUS][MAX_COUNTERS];
static u64 user_interval = ULLONG_MAX;
static u64 default_interval = 0;
+static u64 sample_type;
static int nr_cpus = 0;
static unsigned int page_size;
@@ -129,6 +130,7 @@ static void write_output(void *buf, size_t size)
}
static int process_synthesized_event(event_t *event,
+ struct sample_data *sample __used,
struct perf_session *self __used)
{
write_output(event, event->header.size);
@@ -287,6 +289,9 @@ static void create_counter(int counter, int cpu)
attr->sample_type |= PERF_SAMPLE_CPU;
}
+ if (!sample_type)
+ sample_type = attr->sample_type;
+
attr->mmap = track;
attr->comm = track;
attr->inherit = !no_inherit;
@@ -642,6 +647,8 @@ static int __cmd_record(int argc, const char **argv)
open_counters(cpumap[i]);
}
+ perf_session__set_sample_type(session, sample_type);
+
if (pipe_output) {
err = perf_header__write_pipe(output);
if (err < 0)
diff --git a/tools/perf/builtin-report.c b/tools/perf/builtin-report.c
index 5de405d..904519f 100644
--- a/tools/perf/builtin-report.c
+++ b/tools/perf/builtin-report.c
@@ -150,13 +150,13 @@ static int add_event_total(struct perf_session *session,
return 0;
}
-static int process_sample_event(event_t *event, struct perf_session *session)
+static int process_sample_event(event_t *event, struct sample_data *sample,
+ struct perf_session *session)
{
- struct sample_data data = { .period = 1, };
struct addr_location al;
struct perf_event_attr *attr;
- if (event__preprocess_sample(event, session, &al, &data, NULL) < 0) {
+ if (event__preprocess_sample(event, session, &al, sample, NULL) < 0) {
fprintf(stderr, "problem processing %d event, skipping it.\n",
event->header.type);
return -1;
@@ -165,14 +165,14 @@ static int process_sample_event(event_t *event, struct perf_session *session)
if (al.filtered || (hide_unresolved && al.sym == NULL))
return 0;
- if (perf_session__add_hist_entry(session, &al, &data)) {
+ if (perf_session__add_hist_entry(session, &al, sample)) {
pr_debug("problem incrementing symbol period, skipping event\n");
return -1;
}
- attr = perf_header__find_attr(data.id, &session->header);
+ attr = perf_header__find_attr(sample->id, &session->header);
- if (add_event_total(session, &data, attr)) {
+ if (add_event_total(session, sample, attr)) {
pr_debug("problem adding event period\n");
return -1;
}
@@ -180,7 +180,8 @@ static int process_sample_event(event_t *event, struct perf_session *session)
return 0;
}
-static int process_read_event(event_t *event, struct perf_session *session __used)
+static int process_read_event(event_t *event, struct sample_data *sample __used,
+ struct perf_session *session __used)
{
struct perf_event_attr *attr;
diff --git a/tools/perf/builtin-sched.c b/tools/perf/builtin-sched.c
index 26523c9..73d1e30 100644
--- a/tools/perf/builtin-sched.c
+++ b/tools/perf/builtin-sched.c
@@ -1606,25 +1606,15 @@ process_raw_event(event_t *raw_event __used, struct perf_session *session,
process_sched_migrate_task_event(data, session, event, cpu, timestamp, thread);
}
-static int process_sample_event(event_t *event, struct perf_session *session)
+static int process_sample_event(event_t *event, struct sample_data *sample,
+ struct perf_session *session)
{
- struct sample_data data;
struct thread *thread;
if (!(session->sample_type & PERF_SAMPLE_RAW))
return 0;
- memset(&data, 0, sizeof(data));
- data.time = -1;
- data.cpu = -1;
- data.period = -1;
-
- event__parse_sample(event, session->sample_type, &data);
-
- dump_printf("(IP, %d): %d/%d: %#Lx period: %Ld\n", event->header.misc,
- data.pid, data.tid, data.ip, data.period);
-
- thread = perf_session__findnew(session, data.pid);
+ thread = perf_session__findnew(session, sample->pid);
if (thread == NULL) {
pr_debug("problem processing %d event, skipping it.\n",
event->header.type);
@@ -1633,10 +1623,11 @@ static int process_sample_event(event_t *event, struct perf_session *session)
dump_printf(" ... thread: %s:%d\n", thread->comm, thread->pid);
- if (profile_cpu != -1 && profile_cpu != (int)data.cpu)
+ if (profile_cpu != -1 && profile_cpu != (int)sample->cpu)
return 0;
- process_raw_event(event, session, data.raw_data, data.cpu, data.time, thread);
+ process_raw_event(event, session, sample->raw_data, sample->cpu,
+ sample->time, thread);
return 0;
}
diff --git a/tools/perf/builtin-script.c b/tools/perf/builtin-script.c
index 4539551..683a305 100644
--- a/tools/perf/builtin-script.c
+++ b/tools/perf/builtin-script.c
@@ -63,22 +63,11 @@ static int cleanup_scripting(void)
static char const *input_name = "perf.data";
-static int process_sample_event(event_t *event, struct perf_session *session)
+static int process_sample_event(event_t *event, struct sample_data *sample,
+ struct perf_session *session)
{
- struct sample_data data;
- struct thread *thread;
+ struct thread *thread = perf_session__findnew(session, event->ip.pid);
- memset(&data, 0, sizeof(data));
- data.time = -1;
- data.cpu = -1;
- data.period = 1;
-
- event__parse_sample(event, session->sample_type, &data);
-
- dump_printf("(IP, %d): %d/%d: %#Lx period: %Ld\n", event->header.misc,
- data.pid, data.tid, data.ip, data.period);
-
- thread = perf_session__findnew(session, event->ip.pid);
if (thread == NULL) {
pr_debug("problem processing %d event, skipping it.\n",
event->header.type);
@@ -87,13 +76,13 @@ static int process_sample_event(event_t *event, struct perf_session *session)
if (session->sample_type & PERF_SAMPLE_RAW) {
if (debug_mode) {
- if (data.time < last_timestamp) {
+ if (sample->time < last_timestamp) {
pr_err("Samples misordered, previous: %llu "
"this: %llu\n", last_timestamp,
- data.time);
+ sample->time);
nr_unordered++;
}
- last_timestamp = data.time;
+ last_timestamp = sample->time;
return 0;
}
/*
@@ -101,18 +90,19 @@ static int process_sample_event(event_t *event, struct perf_session *session)
* field, although it should be the same than this perf
* event pid
*/
- scripting_ops->process_event(data.cpu, data.raw_data,
- data.raw_size,
- data.time, thread->comm);
+ scripting_ops->process_event(sample->cpu, sample->raw_data,
+ sample->raw_size,
+ sample->time, thread->comm);
}
- session->hists.stats.total_period += data.period;
+ session->hists.stats.total_period += sample->period;
return 0;
}
static u64 nr_lost;
-static int process_lost_event(event_t *event, struct perf_session *session __used)
+static int process_lost_event(event_t *event, struct sample_data *sample __used,
+ struct perf_session *session __used)
{
nr_lost += event->lost.lost;
diff --git a/tools/perf/builtin-timechart.c b/tools/perf/builtin-timechart.c
index 9bcc38f..1f158dc 100644
--- a/tools/perf/builtin-timechart.c
+++ b/tools/perf/builtin-timechart.c
@@ -272,19 +272,22 @@ static int cpus_cstate_state[MAX_CPUS];
static u64 cpus_pstate_start_times[MAX_CPUS];
static u64 cpus_pstate_state[MAX_CPUS];
-static int process_comm_event(event_t *event, struct perf_session *session __used)
+static int process_comm_event(event_t *event, struct sample_data *sample __used,
+ struct perf_session *session __used)
{
pid_set_comm(event->comm.tid, event->comm.comm);
return 0;
}
-static int process_fork_event(event_t *event, struct perf_session *session __used)
+static int process_fork_event(event_t *event, struct sample_data *sample __used,
+ struct perf_session *session __used)
{
pid_fork(event->fork.pid, event->fork.ppid, event->fork.time);
return 0;
}
-static int process_exit_event(event_t *event, struct perf_session *session __used)
+static int process_exit_event(event_t *event, struct sample_data *sample __used,
+ struct perf_session *session __used)
{
pid_exit(event->fork.pid, event->fork.time);
return 0;
@@ -470,24 +473,21 @@ static void sched_switch(int cpu, u64 timestamp, struct trace_entry *te)
}
-static int process_sample_event(event_t *event, struct perf_session *session)
+static int process_sample_event(event_t *event __used,
+ struct sample_data *sample,
+ struct perf_session *session)
{
- struct sample_data data;
struct trace_entry *te;
- memset(&data, 0, sizeof(data));
-
- event__parse_sample(event, session->sample_type, &data);
-
if (session->sample_type & PERF_SAMPLE_TIME) {
- if (!first_time || first_time > data.time)
- first_time = data.time;
- if (last_time < data.time)
- last_time = data.time;
+ if (!first_time || first_time > sample->time)
+ first_time = sample->time;
+ if (last_time < sample->time)
+ last_time = sample->time;
}
- te = (void *)data.raw_data;
- if (session->sample_type & PERF_SAMPLE_RAW && data.raw_size > 0) {
+ te = (void *)sample->raw_data;
+ if (session->sample_type & PERF_SAMPLE_RAW && sample->raw_size > 0) {
char *event_str;
struct power_entry *pe;
@@ -499,19 +499,19 @@ static int process_sample_event(event_t *event, struct perf_session *session)
return 0;
if (strcmp(event_str, "power:power_start") == 0)
- c_state_start(pe->cpu_id, data.time, pe->value);
+ c_state_start(pe->cpu_id, sample->time, pe->value);
if (strcmp(event_str, "power:power_end") == 0)
- c_state_end(pe->cpu_id, data.time);
+ c_state_end(pe->cpu_id, sample->time);
if (strcmp(event_str, "power:power_frequency") == 0)
- p_state_change(pe->cpu_id, data.time, pe->value);
+ p_state_change(pe->cpu_id, sample->time, pe->value);
if (strcmp(event_str, "sched:sched_wakeup") == 0)
- sched_wakeup(data.cpu, data.time, data.pid, te);
+ sched_wakeup(sample->cpu, sample->time, sample->pid, te);
if (strcmp(event_str, "sched:sched_switch") == 0)
- sched_switch(data.cpu, data.time, te);
+ sched_switch(sample->cpu, sample->time, te);
}
return 0;
}
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index 3d2b47d..5aa29e1 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -1025,6 +1025,8 @@ static void event__process_sample(const event_t *self,
if (self->header.misc & PERF_RECORD_MISC_EXACT_IP)
exact_samples++;
+ event__parse_sample(self, session->sample_type, &data);
+
if (event__preprocess_sample(self, session, &al, &data,
symbol_filter) < 0 ||
al.filtered)
@@ -1155,7 +1157,7 @@ static void perf_session__mmap_read_counter(struct perf_session *self,
if (event->header.type == PERF_RECORD_SAMPLE)
event__process_sample(event, self, md->counter);
else
- event__process(event, self);
+ event__process(event, NULL, self);
old += size;
}
diff --git a/tools/perf/util/build-id.c b/tools/perf/util/build-id.c
index e437edb..deffb8c 100644
--- a/tools/perf/util/build-id.c
+++ b/tools/perf/util/build-id.c
@@ -14,7 +14,9 @@
#include <linux/kernel.h>
#include "debug.h"
-static int build_id__mark_dso_hit(event_t *event, struct perf_session *session)
+static int build_id__mark_dso_hit(event_t *event,
+ struct sample_data *sample __used,
+ struct perf_session *session)
{
struct addr_location al;
u8 cpumode = event->header.misc & PERF_RECORD_MISC_CPUMODE_MASK;
@@ -35,7 +37,8 @@ static int build_id__mark_dso_hit(event_t *event, struct perf_session *session)
return 0;
}
-static int event__exit_del_thread(event_t *self, struct perf_session *session)
+static int event__exit_del_thread(event_t *self, struct sample_data *sample __used,
+ struct perf_session *session)
{
struct thread *thread = perf_session__findnew(session, self->fork.tid);
diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c
index 7260db7..34510f4 100644
--- a/tools/perf/util/event.c
+++ b/tools/perf/util/event.c
@@ -24,6 +24,15 @@ const char *event__name[] = {
[PERF_RECORD_HEADER_BUILD_ID] = "BUILD_ID",
};
+static struct sample_data synth_sample = {
+ .pid = -1,
+ .tid = -1,
+ .time = -1,
+ .stream_id = -1,
+ .cpu = -1,
+ .period = 1,
+};
+
static pid_t event__synthesize_comm(pid_t pid, int full,
event__handler_t process,
struct perf_session *session)
@@ -75,7 +84,7 @@ out_race:
if (!full) {
ev.comm.tid = pid;
- process(&ev, session);
+ process(&ev, &synth_sample, session);
goto out_fclose;
}
@@ -93,7 +102,7 @@ out_race:
ev.comm.tid = pid;
- process(&ev, session);
+ process(&ev, &synth_sample, session);
}
closedir(tasks);
@@ -173,7 +182,7 @@ static int event__synthesize_mmap_events(pid_t pid, pid_t tgid,
ev.mmap.pid = tgid;
ev.mmap.tid = pid;
- process(&ev, session);
+ process(&ev, &synth_sample, session);
}
}
@@ -219,7 +228,7 @@ int event__synthesize_modules(event__handler_t process,
memcpy(ev.mmap.filename, pos->dso->long_name,
pos->dso->long_name_len + 1);
- process(&ev, session);
+ process(&ev, &synth_sample, session);
}
return 0;
@@ -331,7 +340,7 @@ int event__synthesize_kernel_mmap(event__handler_t process,
ev.mmap.len = map->end - ev.mmap.start;
ev.mmap.pid = machine->pid;
- return process(&ev, session);
+ return process(&ev, &synth_sample, session);
}
static void thread__comm_adjust(struct thread *self, struct hists *hists)
@@ -361,7 +370,8 @@ static int thread__set_comm_adjust(struct thread *self, const char *comm,
return 0;
}
-int event__process_comm(event_t *self, struct perf_session *session)
+int event__process_comm(event_t *self, struct sample_data *sample __used,
+ struct perf_session *session)
{
struct thread *thread = perf_session__findnew(session, self->comm.tid);
@@ -376,7 +386,8 @@ int event__process_comm(event_t *self, struct perf_session *session)
return 0;
}
-int event__process_lost(event_t *self, struct perf_session *session)
+int event__process_lost(event_t *self, struct sample_data *sample __used,
+ struct perf_session *session)
{
dump_printf(": id:%Ld: lost:%Ld\n", self->lost.id, self->lost.lost);
session->hists.stats.total_lost += self->lost.lost;
@@ -485,7 +496,8 @@ out_problem:
return -1;
}
-int event__process_mmap(event_t *self, struct perf_session *session)
+int event__process_mmap(event_t *self, struct sample_data *sample __used,
+ struct perf_session *session)
{
struct machine *machine;
struct thread *thread;
@@ -526,7 +538,8 @@ out_problem:
return 0;
}
-int event__process_task(event_t *self, struct perf_session *session)
+int event__process_task(event_t *self, struct sample_data *sample __used,
+ struct perf_session *session)
{
struct thread *thread = perf_session__findnew(session, self->fork.tid);
struct thread *parent = perf_session__findnew(session, self->fork.ptid);
@@ -548,18 +561,19 @@ int event__process_task(event_t *self, struct perf_session *session)
return 0;
}
-int event__process(event_t *event, struct perf_session *session)
+int event__process(event_t *event, struct sample_data *sample,
+ struct perf_session *session)
{
switch (event->header.type) {
case PERF_RECORD_COMM:
- event__process_comm(event, session);
+ event__process_comm(event, sample, session);
break;
case PERF_RECORD_MMAP:
- event__process_mmap(event, session);
+ event__process_mmap(event, sample, session);
break;
case PERF_RECORD_FORK:
case PERF_RECORD_EXIT:
- event__process_task(event, session);
+ event__process_task(event, sample, session);
break;
default:
break;
@@ -674,32 +688,8 @@ int event__preprocess_sample(const event_t *self, struct perf_session *session,
symbol_filter_t filter)
{
u8 cpumode = self->header.misc & PERF_RECORD_MISC_CPUMODE_MASK;
- struct thread *thread;
-
- event__parse_sample(self, session->sample_type, data);
+ struct thread *thread = perf_session__findnew(session, self->ip.pid);
- dump_printf("(IP, %d): %d/%d: %#Lx period: %Ld cpu:%d\n",
- self->header.misc, data->pid, data->tid, data->ip,
- data->period, data->cpu);
-
- if (session->sample_type & PERF_SAMPLE_CALLCHAIN) {
- unsigned int i;
-
- dump_printf("... chain: nr:%Lu\n", data->callchain->nr);
-
- if (!ip_callchain__valid(data->callchain, self)) {
- pr_debug("call-chain problem with event, "
- "skipping it.\n");
- goto out_filtered;
- }
-
- if (dump_trace) {
- for (i = 0; i < data->callchain->nr; i++)
- dump_printf("..... %2d: %016Lx\n",
- i, data->callchain->ips[i]);
- }
- }
- thread = perf_session__findnew(session, self->ip.pid);
if (thread == NULL)
return -1;
diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h
index 8e790da..08c400b 100644
--- a/tools/perf/util/event.h
+++ b/tools/perf/util/event.h
@@ -135,7 +135,10 @@ void event__print_totals(void);
struct perf_session;
-typedef int (*event__handler_t)(event_t *event, struct perf_session *session);
+typedef int (*event__handler_synth_t)(event_t *event,
+ struct perf_session *session);
+typedef int (*event__handler_t)(event_t *event, struct sample_data *sample,
+ struct perf_session *session);
int event__synthesize_thread(pid_t pid, event__handler_t process,
struct perf_session *session);
@@ -150,11 +153,16 @@ int event__synthesize_modules(event__handler_t process,
struct perf_session *session,
struct machine *machine);
-int event__process_comm(event_t *self, struct perf_session *session);
-int event__process_lost(event_t *self, struct perf_session *session);
-int event__process_mmap(event_t *self, struct perf_session *session);
-int event__process_task(event_t *self, struct perf_session *session);
-int event__process(event_t *event, struct perf_session *session);
+int event__process_comm(event_t *self, struct sample_data *sample,
+ struct perf_session *session);
+int event__process_lost(event_t *self, struct sample_data *sample,
+ struct perf_session *session);
+int event__process_mmap(event_t *self, struct sample_data *sample,
+ struct perf_session *session);
+int event__process_task(event_t *self, struct sample_data *sample,
+ struct perf_session *session);
+int event__process(event_t *event, struct sample_data *sample,
+ struct perf_session *session);
struct addr_location;
int event__preprocess_sample(const event_t *self, struct perf_session *session,
--git a/tools/perf/util/header.c b/tools/perf/util/header.c
index f65d7dc..fe652f3 100644
--- a/tools/perf/util/header.c
+++ b/tools/perf/util/header.c
@@ -993,15 +993,14 @@ int event__synthesize_attr(struct perf_event_attr *attr, u16 ids, u64 *id,
ev->attr.header.type = PERF_RECORD_HEADER_ATTR;
ev->attr.header.size = size;
- err = process(ev, session);
+ err = process(ev, NULL, session);
free(ev);
return err;
}
-int event__synthesize_attrs(struct perf_header *self,
- event__handler_t process,
+int event__synthesize_attrs(struct perf_header *self, event__handler_t process,
struct perf_session *session)
{
struct perf_header_attr *attr;
@@ -1071,7 +1070,7 @@ int event__synthesize_event_type(u64 event_id, char *name,
ev.event_type.header.size = sizeof(ev.event_type) -
(sizeof(ev.event_type.event_type.name) - size);
- err = process(&ev, session);
+ err = process(&ev, NULL, session);
return err;
}
@@ -1126,7 +1125,7 @@ int event__synthesize_tracing_data(int fd, struct perf_event_attr *pattrs,
ev.tracing_data.header.size = sizeof(ev.tracing_data);
ev.tracing_data.size = aligned_size;
- process(&ev, session);
+ process(&ev, NULL, session);
err = read_tracing_data(fd, pattrs, nb_events);
write_padded(fd, NULL, 0, padding);
@@ -1186,7 +1185,7 @@ int event__synthesize_build_id(struct dso *pos, u16 misc,
ev.build_id.header.size = sizeof(ev.build_id) + len;
memcpy(&ev.build_id.filename, pos->long_name, pos->long_name_len);
- err = process(&ev, session);
+ err = process(&ev, NULL, session);
return err;
}
diff --git a/tools/perf/util/hist.h b/tools/perf/util/hist.h
index 587d375..ee78985 100644
--- a/tools/perf/util/hist.h
+++ b/tools/perf/util/hist.h
@@ -52,8 +52,10 @@ struct sym_priv {
struct events_stats {
u64 total_period;
u64 total_lost;
+ u64 total_invalid_chains;
u32 nr_events[PERF_RECORD_HEADER_MAX];
u32 nr_unknown_events;
+ u32 nr_invalid_chains;
};
enum hist_column {
diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c
index 52672da..08ec018 100644
--- a/tools/perf/util/session.c
+++ b/tools/perf/util/session.c
@@ -70,6 +70,11 @@ void perf_session__update_sample_type(struct perf_session *self)
self->sample_type = perf_header__sample_type(&self->header);
}
+void perf_session__set_sample_type(struct perf_session *session, u64 type)
+{
+ session->sample_type = type;
+}
+
int perf_session__create_kernel_maps(struct perf_session *self)
{
int ret = machine__create_kernel_maps(&self->host_machine);
@@ -240,7 +245,15 @@ struct map_symbol *perf_session__resolve_callchain(struct perf_session *self,
return syms;
}
+static int process_event_synth_stub(event_t *event __used,
+ struct perf_session *session __used)
+{
+ dump_printf(": unhandled!\n");
+ return 0;
+}
+
static int process_event_stub(event_t *event __used,
+ struct sample_data *sample __used,
struct perf_session *session __used)
{
dump_printf(": unhandled!\n");
@@ -280,13 +293,13 @@ static void perf_event_ops__fill_defaults(struct perf_event_ops *handler)
if (handler->unthrottle == NULL)
handler->unthrottle = process_event_stub;
if (handler->attr == NULL)
- handler->attr = process_event_stub;
+ handler->attr = process_event_synth_stub;
if (handler->event_type == NULL)
- handler->event_type = process_event_stub;
+ handler->event_type = process_event_synth_stub;
if (handler->tracing_data == NULL)
- handler->tracing_data = process_event_stub;
+ handler->tracing_data = process_event_synth_stub;
if (handler->build_id == NULL)
- handler->build_id = process_event_stub;
+ handler->build_id = process_event_synth_stub;
if (handler->finished_round == NULL) {
if (handler->ordered_samples)
handler->finished_round = process_finished_round;
@@ -419,6 +432,7 @@ static void flush_sample_queue(struct perf_session *s,
struct ordered_samples *os = &s->ordered_samples;
struct list_head *head = &os->samples;
struct sample_queue *tmp, *iter;
+ struct sample_data sample;
u64 limit = os->next_flush;
u64 last_ts = os->last_sample ? os->last_sample->timestamp : 0ULL;
@@ -429,7 +443,8 @@ static void flush_sample_queue(struct perf_session *s,
if (iter->timestamp > limit)
break;
- ops->sample(iter->event, s);
+ event__parse_sample(iter->event, s->sample_type, &sample);
+ ops->sample(iter->event, &sample, s);
os->last_flush = iter->timestamp;
list_del(&iter->list);
@@ -578,20 +593,29 @@ static int queue_sample_event(event_t *event, struct sample_data *data,
return 0;
}
-static int perf_session__process_sample(event_t *event, struct perf_session *s,
+static int perf_session__process_sample(event_t *event,
+ struct sample_data *sample,
+ struct perf_session *s,
struct perf_event_ops *ops)
{
- struct sample_data data;
-
if (!ops->ordered_samples)
- return ops->sample(event, s);
+ return ops->sample(event, sample, s);
+
+ queue_sample_event(event, sample, s);
+ return 0;
+}
- bzero(&data, sizeof(struct sample_data));
- event__parse_sample(event, s->sample_type, &data);
+static void callchain__dump(struct sample_data *sample)
+{
+ unsigned int i;
- queue_sample_event(event, &data, s);
+ if (!dump_trace)
+ return;
- return 0;
+ printf("... chain: nr:%Lu\n", sample->callchain->nr);
+
+ for (i = 0; i < sample->callchain->nr; i++)
+ printf("..... %2d: %016Lx\n", i, sample->callchain->ips[i]);
}
static int perf_session__process_event(struct perf_session *self,
@@ -599,8 +623,16 @@ static int perf_session__process_event(struct perf_session *self,
struct perf_event_ops *ops,
u64 file_offset)
{
+ struct sample_data sample;
+
trace_event(event);
+ if (self->header.needs_swap && event__swap_ops[event->header.type])
+ event__swap_ops[event->header.type](event);
+
+ if (event->header.type == PERF_RECORD_SAMPLE)
+ event__parse_sample(event, self->sample_type, &sample);
+
if (event->header.type < PERF_RECORD_HEADER_MAX) {
dump_printf("%#Lx [%#x]: PERF_RECORD_%s",
file_offset, event->header.size,
@@ -608,28 +640,41 @@ static int perf_session__process_event(struct perf_session *self,
hists__inc_nr_events(&self->hists, event->header.type);
}
- if (self->header.needs_swap && event__swap_ops[event->header.type])
- event__swap_ops[event->header.type](event);
-
switch (event->header.type) {
case PERF_RECORD_SAMPLE:
- return perf_session__process_sample(event, self, ops);
+ dump_printf("(IP, %d): %d/%d: %#Lx period: %Ld\n", event->header.misc,
+ sample.pid, sample.tid, sample.ip, sample.period);
+
+ if (self->sample_type & PERF_SAMPLE_CALLCHAIN) {
+ if (!ip_callchain__valid(sample.callchain, event)) {
+ pr_debug("call-chain problem with event, "
+ "skipping it.\n");
+ ++self->hists.stats.nr_invalid_chains;
+ self->hists.stats.total_invalid_chains += sample.period;
+ return 0;
+ }
+
+ callchain__dump(&sample);
+ }
+
+ return perf_session__process_sample(event, &sample, self, ops);
+
case PERF_RECORD_MMAP:
- return ops->mmap(event, self);
+ return ops->mmap(event, &sample, self);
case PERF_RECORD_COMM:
- return ops->comm(event, self);
+ return ops->comm(event, &sample, self);
case PERF_RECORD_FORK:
- return ops->fork(event, self);
+ return ops->fork(event, &sample, self);
case PERF_RECORD_EXIT:
- return ops->exit(event, self);
+ return ops->exit(event, &sample, self);
case PERF_RECORD_LOST:
- return ops->lost(event, self);
+ return ops->lost(event, &sample, self);
case PERF_RECORD_READ:
- return ops->read(event, self);
+ return ops->read(event, &sample, self);
case PERF_RECORD_THROTTLE:
- return ops->throttle(event, self);
+ return ops->throttle(event, &sample, self);
case PERF_RECORD_UNTHROTTLE:
- return ops->unthrottle(event, self);
+ return ops->unthrottle(event, &sample, self);
case PERF_RECORD_HEADER_ATTR:
return ops->attr(event, self);
case PERF_RECORD_HEADER_EVENT_TYPE:
@@ -894,6 +939,14 @@ out_err:
session->hists.stats.nr_unknown_events);
}
+ if (session->hists.stats.nr_invalid_chains != 0) {
+ ui__warning("Found invalid callchains!\n\n"
+ "%u out of %u events were discarded for this reason.\n\n"
+ "Consider reporting to linux-kernel@vger.kernel.org.\n\n",
+ session->hists.stats.nr_invalid_chains,
+ session->hists.stats.nr_events[PERF_RECORD_SAMPLE]);
+ }
+
perf_session_free_sample_buffers(session);
return err;
}
diff --git a/tools/perf/util/session.h b/tools/perf/util/session.h
index 5bf6efa..4578f86 100644
--- a/tools/perf/util/session.h
+++ b/tools/perf/util/session.h
@@ -54,7 +54,9 @@ struct perf_session {
struct perf_event_ops;
-typedef int (*event_op)(event_t *self, struct perf_session *session);
+typedef int (*event_op)(event_t *self, struct sample_data *sample,
+ struct perf_session *session);
+typedef int (*event_synth_op)(event_t *self, struct perf_session *session);
typedef int (*event_op2)(event_t *self, struct perf_session *session,
struct perf_event_ops *ops);
@@ -67,8 +69,8 @@ struct perf_event_ops {
lost,
read,
throttle,
- unthrottle,
- attr,
+ unthrottle;
+ event_synth_op attr,
event_type,
tracing_data,
build_id;
@@ -104,6 +106,7 @@ int perf_session__create_kernel_maps(struct perf_session *self);
int do_read(int fd, void *buf, size_t size);
void perf_session__update_sample_type(struct perf_session *self);
+void perf_session__set_sample_type(struct perf_session *session, u64 type);
void perf_session__remove_thread(struct perf_session *self, struct thread *th);
static inline
--
1.6.2.5
^ permalink raw reply related [flat|nested] 15+ messages in thread
* [PATCH 4/4] perf tools: Ask for ID PERF_SAMPLE_ info on all PERF_RECORD_ events
2010-12-02 19:39 [GIT PULL 0/4] perf/core improvements Arnaldo Carvalho de Melo
` (2 preceding siblings ...)
2010-12-02 19:39 ` [PATCH 3/4] perf session: Parse sample earlier Arnaldo Carvalho de Melo
@ 2010-12-02 19:39 ` Arnaldo Carvalho de Melo
2010-12-03 6:04 ` Ian Munsie
2010-12-02 23:00 ` [GIT PULL 0/4] perf/core improvements Thomas Gleixner
2010-12-03 6:10 ` [GIT PULL 0/4] perf/core improvements Ian Munsie
5 siblings, 1 reply; 15+ messages in thread
From: Arnaldo Carvalho de Melo @ 2010-12-02 19:39 UTC (permalink / raw)
To: Ingo Molnar
Cc: linux-kernel, Arnaldo Carvalho de Melo, Frederic Weisbecker,
Ian Munsie, Mike Galbraith, Peter Zijlstra, Paul Mackerras,
Stephane Eranian
From: Arnaldo Carvalho de Melo <acme@redhat.com>
So that we can use -T == --timestamp, asking for PERF_SAMPLE_TIME:
$ perf record -aT
$ perf report -D | grep PERF_RECORD_
<SNIP>
3 5951915425 0x47530 [0x58]: PERF_RECORD_SAMPLE(IP, 1): 16811/16811: 0xffffffff8138c1a2 period: 215979 cpu:3
3 5952026879 0x47588 [0x90]: PERF_RECORD_SAMPLE(IP, 1): 16811/16811: 0xffffffff810cb480 period: 215979 cpu:3
3 5952059959 0x47618 [0x38]: PERF_RECORD_FORK(6853:6853):(16811:16811)
3 5952138878 0x47650 [0x78]: PERF_RECORD_SAMPLE(IP, 1): 16811/16811: 0xffffffff811bac35 period: 431478 cpu:3
3 5952375068 0x476c8 [0x30]: PERF_RECORD_COMM: find:6853
3 5952395923 0x476f8 [0x50]: PERF_RECORD_MMAP 6853/6853: [0x400000(0x25000) @ 0]: /usr/bin/find
3 5952413756 0x47748 [0xa0]: PERF_RECORD_SAMPLE(IP, 1): 6853/6853: 0xffffffff810d080f period: 859332 cpu:3
3 5952419837 0x477e8 [0x58]: PERF_RECORD_MMAP 6853/6853: [0x3f44600000(0x21d000) @ 0]: /lib64/ld-2.5.so
3 5952437929 0x47840 [0x48]: PERF_RECORD_MMAP 6853/6853: [0x7fff7e1c9000(0x1000) @ 0x7fff7e1c9000]: [vdso]
3 5952570127 0x47888 [0x58]: PERF_RECORD_MMAP 6853/6853: [0x3f46200000(0x218000) @ 0]: /lib64/libselinux.so.1
3 5952623637 0x478e0 [0x58]: PERF_RECORD_MMAP 6853/6853: [0x3f44a00000(0x356000) @ 0]: /lib64/libc-2.5.so
3 5952675720 0x47938 [0x58]: PERF_RECORD_MMAP 6853/6853: [0x3f44e00000(0x204000) @ 0]: /lib64/libdl-2.5.so
3 5952710080 0x47990 [0x58]: PERF_RECORD_MMAP 6853/6853: [0x3f45a00000(0x246000) @ 0]: /lib64/libsepol.so.1
3 5952847802 0x479e8 [0x58]: PERF_RECORD_SAMPLE(IP, 1): 6853/6853: 0xffffffff813897f0 period: 1142536 cpu:3
<SNIP>
First column is the cpu and the second the timestamp.
That way we can investigate problems in the event stream.
If the new perf binary is run on an older kernel, it will disable this feature
automatically.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <new-submission>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
tools/perf/Documentation/perf-record.txt | 5 +
tools/perf/builtin-record.c | 16 ++
tools/perf/builtin-top.c | 14 +-
tools/perf/util/event.c | 285 +++++++++++++++++++++---------
tools/perf/util/event.h | 7 +-
| 18 ++
| 1 +
tools/perf/util/session.c | 66 +++++++-
tools/perf/util/session.h | 3 +
9 files changed, 315 insertions(+), 100 deletions(-)
diff --git a/tools/perf/Documentation/perf-record.txt b/tools/perf/Documentation/perf-record.txt
index 0ad1bc7..52462ae 100644
--- a/tools/perf/Documentation/perf-record.txt
+++ b/tools/perf/Documentation/perf-record.txt
@@ -108,6 +108,11 @@ OPTIONS
--data::
Sample addresses.
+-T::
+--timestamp::
+ Sample timestamps. Use it with 'perf report -D' to see the timestamps,
+ for instance.
+
-n::
--no-samples::
Don't sample.
diff --git a/tools/perf/builtin-record.c b/tools/perf/builtin-record.c
index b34de92..699dd21 100644
--- a/tools/perf/builtin-record.c
+++ b/tools/perf/builtin-record.c
@@ -49,6 +49,7 @@ static const char *output_name = "perf.data";
static int group = 0;
static int realtime_prio = 0;
static bool raw_samples = false;
+static bool sample_id_all_avail = true;
static bool system_wide = false;
static pid_t target_pid = -1;
static pid_t target_tid = -1;
@@ -61,6 +62,7 @@ static bool call_graph = false;
static bool inherit_stat = false;
static bool no_samples = false;
static bool sample_address = false;
+static bool sample_time = false;
static bool no_buildid = false;
static bool no_buildid_cache = false;
@@ -283,6 +285,9 @@ static void create_counter(int counter, int cpu)
if (system_wide)
attr->sample_type |= PERF_SAMPLE_CPU;
+ if (sample_time)
+ attr->sample_type |= PERF_SAMPLE_TIME;
+
if (raw_samples) {
attr->sample_type |= PERF_SAMPLE_TIME;
attr->sample_type |= PERF_SAMPLE_RAW;
@@ -299,6 +304,8 @@ static void create_counter(int counter, int cpu)
attr->disabled = 1;
attr->enable_on_exec = 1;
}
+retry_sample_id:
+ attr->sample_id_all = sample_id_all_avail ? 1 : 0;
for (thread_index = 0; thread_index < thread_num; thread_index++) {
try_again:
@@ -315,6 +322,12 @@ try_again:
else if (err == ENODEV && cpu_list) {
die("No such device - did you specify"
" an out-of-range profile CPU?\n");
+ } else if (err == EINVAL && sample_id_all_avail) {
+ /*
+ * Old kernel, no attr->sample_id_type_all field
+ */
+ sample_id_all_avail = false;
+ goto retry_sample_id;
}
/*
@@ -661,6 +674,8 @@ static int __cmd_record(int argc, const char **argv)
post_processing_offset = lseek(output, 0, SEEK_CUR);
+ perf_session__set_sample_id_all(session, sample_id_all_avail);
+
if (pipe_output) {
err = event__synthesize_attrs(&session->header,
process_synthesized_event,
@@ -841,6 +856,7 @@ const struct option record_options[] = {
"per thread counts"),
OPT_BOOLEAN('d', "data", &sample_address,
"Sample addresses"),
+ OPT_BOOLEAN('T', "timestamp", &sample_time, "Sample timestamps"),
OPT_BOOLEAN('n', "no-samples", &no_samples,
"don't sample"),
OPT_BOOLEAN('N', "no-buildid-cache", &no_buildid_cache,
diff --git a/tools/perf/builtin-top.c b/tools/perf/builtin-top.c
index 5aa29e1..0515ce9 100644
--- a/tools/perf/builtin-top.c
+++ b/tools/perf/builtin-top.c
@@ -977,12 +977,12 @@ static int symbol_filter(struct map *map, struct symbol *sym)
}
static void event__process_sample(const event_t *self,
- struct perf_session *session, int counter)
+ struct sample_data *sample,
+ struct perf_session *session, int counter)
{
u64 ip = self->ip.ip;
struct sym_entry *syme;
struct addr_location al;
- struct sample_data data;
struct machine *machine;
u8 origin = self->header.misc & PERF_RECORD_MISC_CPUMODE_MASK;
@@ -1025,9 +1025,7 @@ static void event__process_sample(const event_t *self,
if (self->header.misc & PERF_RECORD_MISC_EXACT_IP)
exact_samples++;
- event__parse_sample(self, session->sample_type, &data);
-
- if (event__preprocess_sample(self, session, &al, &data,
+ if (event__preprocess_sample(self, session, &al, sample,
symbol_filter) < 0 ||
al.filtered)
return;
@@ -1107,6 +1105,7 @@ static void perf_session__mmap_read_counter(struct perf_session *self,
unsigned int head = mmap_read_head(md);
unsigned int old = md->prev;
unsigned char *data = md->base + page_size;
+ struct sample_data sample;
int diff;
/*
@@ -1154,10 +1153,11 @@ static void perf_session__mmap_read_counter(struct perf_session *self,
event = &event_copy;
}
+ event__parse_sample(event, self, &sample);
if (event->header.type == PERF_RECORD_SAMPLE)
- event__process_sample(event, self, md->counter);
+ event__process_sample(event, &sample, self, md->counter);
else
- event__process(event, NULL, self);
+ event__process(event, &sample, self);
old += size;
}
diff --git a/tools/perf/util/event.c b/tools/perf/util/event.c
index 34510f4..e4cdc1e 100644
--- a/tools/perf/util/event.c
+++ b/tools/perf/util/event.c
@@ -33,11 +33,10 @@ static struct sample_data synth_sample = {
.period = 1,
};
-static pid_t event__synthesize_comm(pid_t pid, int full,
+static pid_t event__synthesize_comm(event_t *event, pid_t pid, int full,
event__handler_t process,
struct perf_session *session)
{
- event_t ev;
char filename[PATH_MAX];
char bf[BUFSIZ];
FILE *fp;
@@ -58,34 +57,39 @@ out_race:
return 0;
}
- memset(&ev.comm, 0, sizeof(ev.comm));
- while (!ev.comm.comm[0] || !ev.comm.pid) {
- if (fgets(bf, sizeof(bf), fp) == NULL)
- goto out_failure;
+ memset(&event->comm, 0, sizeof(event->comm));
+
+ while (!event->comm.comm[0] || !event->comm.pid) {
+ if (fgets(bf, sizeof(bf), fp) == NULL) {
+ pr_warning("couldn't get COMM and pgid, malformed %s\n", filename);
+ goto out;
+ }
if (memcmp(bf, "Name:", 5) == 0) {
char *name = bf + 5;
while (*name && isspace(*name))
++name;
size = strlen(name) - 1;
- memcpy(ev.comm.comm, name, size++);
+ memcpy(event->comm.comm, name, size++);
} else if (memcmp(bf, "Tgid:", 5) == 0) {
char *tgids = bf + 5;
while (*tgids && isspace(*tgids))
++tgids;
- tgid = ev.comm.pid = atoi(tgids);
+ tgid = event->comm.pid = atoi(tgids);
}
}
- ev.comm.header.type = PERF_RECORD_COMM;
+ event->comm.header.type = PERF_RECORD_COMM;
size = ALIGN(size, sizeof(u64));
- ev.comm.header.size = sizeof(ev.comm) - (sizeof(ev.comm.comm) - size);
-
+ memset(event->comm.comm + size, 0, session->id_hdr_size);
+ event->comm.header.size = (sizeof(event->comm) -
+ (sizeof(event->comm.comm) - size) +
+ session->id_hdr_size);
if (!full) {
- ev.comm.tid = pid;
+ event->comm.tid = pid;
- process(&ev, &synth_sample, session);
- goto out_fclose;
+ process(event, &synth_sample, session);
+ goto out;
}
snprintf(filename, sizeof(filename), "/proc/%d/task", pid);
@@ -100,22 +104,19 @@ out_race:
if (*end)
continue;
- ev.comm.tid = pid;
+ event->comm.tid = pid;
- process(&ev, &synth_sample, session);
+ process(event, &synth_sample, session);
}
- closedir(tasks);
-out_fclose:
+ closedir(tasks);
+out:
fclose(fp);
- return tgid;
-out_failure:
- pr_warning("couldn't get COMM and pgid, malformed %s\n", filename);
- return -1;
+ return tgid;
}
-static int event__synthesize_mmap_events(pid_t pid, pid_t tgid,
+static int event__synthesize_mmap_events(event_t *event, pid_t pid, pid_t tgid,
event__handler_t process,
struct perf_session *session)
{
@@ -133,29 +134,25 @@ static int event__synthesize_mmap_events(pid_t pid, pid_t tgid,
return -1;
}
+ event->header.type = PERF_RECORD_MMAP;
+ /*
+ * Just like the kernel, see __perf_event_mmap in kernel/perf_event.c
+ */
+ event->header.misc = PERF_RECORD_MISC_USER;
+
while (1) {
char bf[BUFSIZ], *pbf = bf;
- event_t ev = {
- .header = {
- .type = PERF_RECORD_MMAP,
- /*
- * Just like the kernel, see __perf_event_mmap
- * in kernel/perf_event.c
- */
- .misc = PERF_RECORD_MISC_USER,
- },
- };
int n;
size_t size;
if (fgets(bf, sizeof(bf), fp) == NULL)
break;
/* 00400000-0040c000 r-xp 00000000 fd:01 41038 /bin/cat */
- n = hex2u64(pbf, &ev.mmap.start);
+ n = hex2u64(pbf, &event->mmap.start);
if (n < 0)
continue;
pbf += n + 1;
- n = hex2u64(pbf, &ev.mmap.len);
+ n = hex2u64(pbf, &event->mmap.len);
if (n < 0)
continue;
pbf += n + 3;
@@ -170,19 +167,21 @@ static int event__synthesize_mmap_events(pid_t pid, pid_t tgid,
continue;
pbf += 3;
- n = hex2u64(pbf, &ev.mmap.pgoff);
+ n = hex2u64(pbf, &event->mmap.pgoff);
size = strlen(execname);
execname[size - 1] = '\0'; /* Remove \n */
- memcpy(ev.mmap.filename, execname, size);
+ memcpy(event->mmap.filename, execname, size);
size = ALIGN(size, sizeof(u64));
- ev.mmap.len -= ev.mmap.start;
- ev.mmap.header.size = (sizeof(ev.mmap) -
- (sizeof(ev.mmap.filename) - size));
- ev.mmap.pid = tgid;
- ev.mmap.tid = pid;
-
- process(&ev, &synth_sample, session);
+ event->mmap.len -= event->mmap.start;
+ event->mmap.header.size = (sizeof(event->mmap) -
+ (sizeof(event->mmap.filename) - size));
+ memset(event->mmap.filename + size, 0, session->id_hdr_size);
+ event->mmap.header.size += session->id_hdr_size;
+ event->mmap.pid = tgid;
+ event->mmap.tid = pid;
+
+ process(event, &synth_sample, session);
}
}
@@ -196,20 +195,27 @@ int event__synthesize_modules(event__handler_t process,
{
struct rb_node *nd;
struct map_groups *kmaps = &machine->kmaps;
- u16 misc;
+ event_t *event = zalloc(sizeof(event->mmap) + session->id_hdr_size);
+
+ if (event == NULL) {
+ pr_debug("Not enough memory synthesizing mmap event "
+ "for kernel modules\n");
+ return -1;
+ }
+
+ event->header.type = PERF_RECORD_MMAP;
/*
* kernel uses 0 for user space maps, see kernel/perf_event.c
* __perf_event_mmap
*/
if (machine__is_host(machine))
- misc = PERF_RECORD_MISC_KERNEL;
+ event->header.misc = PERF_RECORD_MISC_KERNEL;
else
- misc = PERF_RECORD_MISC_GUEST_KERNEL;
+ event->header.misc = PERF_RECORD_MISC_GUEST_KERNEL;
for (nd = rb_first(&kmaps->maps[MAP__FUNCTION]);
nd; nd = rb_next(nd)) {
- event_t ev;
size_t size;
struct map *pos = rb_entry(nd, struct map, rb_node);
@@ -217,39 +223,78 @@ int event__synthesize_modules(event__handler_t process,
continue;
size = ALIGN(pos->dso->long_name_len + 1, sizeof(u64));
- memset(&ev, 0, sizeof(ev));
- ev.mmap.header.misc = misc;
- ev.mmap.header.type = PERF_RECORD_MMAP;
- ev.mmap.header.size = (sizeof(ev.mmap) -
- (sizeof(ev.mmap.filename) - size));
- ev.mmap.start = pos->start;
- ev.mmap.len = pos->end - pos->start;
- ev.mmap.pid = machine->pid;
-
- memcpy(ev.mmap.filename, pos->dso->long_name,
+ event->mmap.header.type = PERF_RECORD_MMAP;
+ event->mmap.header.size = (sizeof(event->mmap) -
+ (sizeof(event->mmap.filename) - size));
+ memset(event->mmap.filename + size, 0, session->id_hdr_size);
+ event->mmap.header.size += session->id_hdr_size;
+ event->mmap.start = pos->start;
+ event->mmap.len = pos->end - pos->start;
+ event->mmap.pid = machine->pid;
+
+ memcpy(event->mmap.filename, pos->dso->long_name,
pos->dso->long_name_len + 1);
- process(&ev, &synth_sample, session);
+ process(event, &synth_sample, session);
}
+ free(event);
return 0;
}
-int event__synthesize_thread(pid_t pid, event__handler_t process,
- struct perf_session *session)
+static int __event__synthesize_thread(event_t *comm_event, event_t *mmap_event,
+ pid_t pid, event__handler_t process,
+ struct perf_session *session)
{
- pid_t tgid = event__synthesize_comm(pid, 1, process, session);
+ pid_t tgid = event__synthesize_comm(comm_event, pid, 1, process,
+ session);
if (tgid == -1)
return -1;
- return event__synthesize_mmap_events(pid, tgid, process, session);
+ return event__synthesize_mmap_events(mmap_event, pid, tgid,
+ process, session);
}
-void event__synthesize_threads(event__handler_t process,
- struct perf_session *session)
+int event__synthesize_thread(pid_t pid, event__handler_t process,
+ struct perf_session *session)
+{
+ event_t *comm_event, *mmap_event;
+ int err = -1;
+
+ comm_event = malloc(sizeof(comm_event->comm) + session->id_hdr_size);
+ if (comm_event == NULL)
+ goto out;
+
+ mmap_event = malloc(sizeof(mmap_event->mmap) + session->id_hdr_size);
+ if (mmap_event == NULL)
+ goto out_free_comm;
+
+ err = __event__synthesize_thread(comm_event, mmap_event, pid,
+ process, session);
+ free(mmap_event);
+out_free_comm:
+ free(comm_event);
+out:
+ return err;
+}
+
+int event__synthesize_threads(event__handler_t process,
+ struct perf_session *session)
{
DIR *proc;
struct dirent dirent, *next;
+ event_t *comm_event, *mmap_event;
+ int err = -1;
+
+ comm_event = malloc(sizeof(comm_event->comm) + session->id_hdr_size);
+ if (comm_event == NULL)
+ goto out;
+
+ mmap_event = malloc(sizeof(mmap_event->mmap) + session->id_hdr_size);
+ if (mmap_event == NULL)
+ goto out_free_comm;
proc = opendir("/proc");
+ if (proc == NULL)
+ goto out_free_mmap;
while (!readdir_r(proc, &dirent, &next) && next) {
char *end;
@@ -258,10 +303,18 @@ void event__synthesize_threads(event__handler_t process,
if (*end) /* only interested in proper numerical dirents */
continue;
- event__synthesize_thread(pid, process, session);
+ __event__synthesize_thread(comm_event, mmap_event, pid,
+ process, session);
}
closedir(proc);
+ err = 0;
+out_free_mmap:
+ free(mmap_event);
+out_free_comm:
+ free(comm_event);
+out:
+ return err;
}
struct process_symbol_args {
@@ -295,18 +348,20 @@ int event__synthesize_kernel_mmap(event__handler_t process,
char path[PATH_MAX];
char name_buff[PATH_MAX];
struct map *map;
-
- event_t ev = {
- .header = {
- .type = PERF_RECORD_MMAP,
- },
- };
+ int err;
/*
* We should get this from /sys/kernel/sections/.text, but till that is
* available use this, and after it is use this as a fallback for older
* kernels.
*/
struct process_symbol_args args = { .name = symbol_name, };
+ event_t *event = zalloc(sizeof(event->mmap) + session->id_hdr_size);
+
+ if (event == NULL) {
+ pr_debug("Not enough memory synthesizing mmap event "
+ "for kernel modules\n");
+ return -1;
+ }
mmap_name = machine__mmap_name(machine, name_buff, sizeof(name_buff));
if (machine__is_host(machine)) {
@@ -314,10 +369,10 @@ int event__synthesize_kernel_mmap(event__handler_t process,
* kernel uses PERF_RECORD_MISC_USER for user space maps,
* see kernel/perf_event.c __perf_event_mmap
*/
- ev.header.misc = PERF_RECORD_MISC_KERNEL;
+ event->header.misc = PERF_RECORD_MISC_KERNEL;
filename = "/proc/kallsyms";
} else {
- ev.header.misc = PERF_RECORD_MISC_GUEST_KERNEL;
+ event->header.misc = PERF_RECORD_MISC_GUEST_KERNEL;
if (machine__is_default_guest(machine))
filename = (char *) symbol_conf.default_guest_kallsyms;
else {
@@ -330,17 +385,21 @@ int event__synthesize_kernel_mmap(event__handler_t process,
return -ENOENT;
map = machine->vmlinux_maps[MAP__FUNCTION];
- size = snprintf(ev.mmap.filename, sizeof(ev.mmap.filename),
+ size = snprintf(event->mmap.filename, sizeof(event->mmap.filename),
"%s%s", mmap_name, symbol_name) + 1;
size = ALIGN(size, sizeof(u64));
- ev.mmap.header.size = (sizeof(ev.mmap) -
- (sizeof(ev.mmap.filename) - size));
- ev.mmap.pgoff = args.start;
- ev.mmap.start = map->start;
- ev.mmap.len = map->end - ev.mmap.start;
- ev.mmap.pid = machine->pid;
-
- return process(&ev, &synth_sample, session);
+ event->mmap.header.type = PERF_RECORD_MMAP;
+ event->mmap.header.size = (sizeof(event->mmap) -
+ (sizeof(event->mmap.filename) - size) + session->id_hdr_size);
+ event->mmap.pgoff = args.start;
+ event->mmap.start = map->start;
+ event->mmap.len = map->end - event->mmap.start;
+ event->mmap.pid = machine->pid;
+
+ err = process(event, &synth_sample, session);
+ free(event);
+
+ return err;
}
static void thread__comm_adjust(struct thread *self, struct hists *hists)
@@ -756,9 +815,65 @@ out_filtered:
return 0;
}
-int event__parse_sample(const event_t *event, u64 type, struct sample_data *data)
+static int event__parse_id_sample(const event_t *event,
+ struct perf_session *session,
+ struct sample_data *sample)
{
- const u64 *array = event->sample.array;
+ const u64 *array;
+ u64 type;
+
+ sample->cpu = sample->pid = sample->tid = -1;
+ sample->stream_id = sample->id = sample->time = -1ULL;
+
+ if (!session->sample_id_all)
+ return 0;
+
+ array = event->sample.array;
+ array += ((event->header.size -
+ sizeof(event->header)) / sizeof(u64)) - 1;
+ type = session->sample_type;
+
+ if (type & PERF_SAMPLE_CPU) {
+ u32 *p = (u32 *)array;
+ sample->cpu = *p;
+ array--;
+ }
+
+ if (type & PERF_SAMPLE_STREAM_ID) {
+ sample->stream_id = *array;
+ array--;
+ }
+
+ if (type & PERF_SAMPLE_ID) {
+ sample->id = *array;
+ array--;
+ }
+
+ if (type & PERF_SAMPLE_TIME) {
+ sample->time = *array;
+ array--;
+ }
+
+ if (type & PERF_SAMPLE_TID) {
+ u32 *p = (u32 *)array;
+ sample->pid = p[0];
+ sample->tid = p[1];
+ }
+
+ return 0;
+}
+
+int event__parse_sample(const event_t *event, struct perf_session *session,
+ struct sample_data *data)
+{
+ const u64 *array;
+ u64 type;
+
+ if (event->header.type != PERF_RECORD_SAMPLE)
+ return event__parse_id_sample(event, session, data);
+
+ array = event->sample.array;
+ type = session->sample_type;
if (type & PERF_SAMPLE_IP) {
data->ip = event->ip.ip;
diff --git a/tools/perf/util/event.h b/tools/perf/util/event.h
index 08c400b..a95ab18 100644
--- a/tools/perf/util/event.h
+++ b/tools/perf/util/event.h
@@ -142,8 +142,8 @@ typedef int (*event__handler_t)(event_t *event, struct sample_data *sample,
int event__synthesize_thread(pid_t pid, event__handler_t process,
struct perf_session *session);
-void event__synthesize_threads(event__handler_t process,
- struct perf_session *session);
+int event__synthesize_threads(event__handler_t process,
+ struct perf_session *session);
int event__synthesize_kernel_mmap(event__handler_t process,
struct perf_session *session,
struct machine *machine,
@@ -168,7 +168,8 @@ struct addr_location;
int event__preprocess_sample(const event_t *self, struct perf_session *session,
struct addr_location *al, struct sample_data *data,
symbol_filter_t filter);
-int event__parse_sample(const event_t *event, u64 type, struct sample_data *data);
+int event__parse_sample(const event_t *event, struct perf_session *session,
+ struct sample_data *sample);
extern const char *event__name[];
--git a/tools/perf/util/header.c b/tools/perf/util/header.c
index fe652f3..073f0e1 100644
--- a/tools/perf/util/header.c
+++ b/tools/perf/util/header.c
@@ -946,6 +946,24 @@ u64 perf_header__sample_type(struct perf_header *header)
return type;
}
+bool perf_header__sample_id_all(const struct perf_header *header)
+{
+ bool value = false, first = true;
+ int i;
+
+ for (i = 0; i < header->attrs; i++) {
+ struct perf_header_attr *attr = header->attr[i];
+
+ if (first) {
+ value = attr->attr.sample_id_all;
+ first = false;
+ } else if (value != attr->attr.sample_id_all)
+ die("non matching sample_id_all");
+ }
+
+ return value;
+}
+
struct perf_event_attr *
perf_header__find_attr(u64 id, struct perf_header *header)
{
--git a/tools/perf/util/header.h b/tools/perf/util/header.h
index ed550bf..6335965 100644
--- a/tools/perf/util/header.h
+++ b/tools/perf/util/header.h
@@ -81,6 +81,7 @@ void perf_header_attr__delete(struct perf_header_attr *self);
int perf_header_attr__add_id(struct perf_header_attr *self, u64 id);
u64 perf_header__sample_type(struct perf_header *header);
+bool perf_header__sample_id_all(const struct perf_header *header);
struct perf_event_attr *
perf_header__find_attr(u64 id, struct perf_header *header);
void perf_header__set_feat(struct perf_header *self, int feat);
diff --git a/tools/perf/util/session.c b/tools/perf/util/session.c
index 08ec018..5c75660 100644
--- a/tools/perf/util/session.c
+++ b/tools/perf/util/session.c
@@ -65,9 +65,37 @@ out_close:
return -1;
}
-void perf_session__update_sample_type(struct perf_session *self)
+static void perf_session__id_header_size(struct perf_session *session)
{
- self->sample_type = perf_header__sample_type(&self->header);
+ struct sample_data *data;
+ u64 sample_type = session->sample_type;
+ u16 size = 0;
+
+ if (!session->sample_id_all)
+ goto out;
+
+ if (sample_type & PERF_SAMPLE_TID)
+ size += sizeof(data->tid) * 2;
+
+ if (sample_type & PERF_SAMPLE_TIME)
+ size += sizeof(data->time);
+
+ if (sample_type & PERF_SAMPLE_ID)
+ size += sizeof(data->id);
+
+ if (sample_type & PERF_SAMPLE_STREAM_ID)
+ size += sizeof(data->stream_id);
+
+ if (sample_type & PERF_SAMPLE_CPU)
+ size += sizeof(data->cpu) * 2;
+out:
+ session->id_hdr_size = size;
+}
+
+void perf_session__set_sample_id_all(struct perf_session *session, bool value)
+{
+ session->sample_id_all = value;
+ perf_session__id_header_size(session);
}
void perf_session__set_sample_type(struct perf_session *session, u64 type)
@@ -75,6 +103,13 @@ void perf_session__set_sample_type(struct perf_session *session, u64 type)
session->sample_type = type;
}
+void perf_session__update_sample_type(struct perf_session *self)
+{
+ self->sample_type = perf_header__sample_type(&self->header);
+ self->sample_id_all = perf_header__sample_id_all(&self->header);
+ perf_session__id_header_size(self);
+}
+
int perf_session__create_kernel_maps(struct perf_session *self)
{
int ret = machine__create_kernel_maps(&self->host_machine);
@@ -443,7 +478,7 @@ static void flush_sample_queue(struct perf_session *s,
if (iter->timestamp > limit)
break;
- event__parse_sample(iter->event, s->sample_type, &sample);
+ event__parse_sample(iter->event, s, &sample);
ops->sample(iter->event, &sample, s);
os->last_flush = iter->timestamp;
@@ -618,6 +653,23 @@ static void callchain__dump(struct sample_data *sample)
printf("..... %2d: %016Lx\n", i, sample->callchain->ips[i]);
}
+static void perf_session__print_tstamp(struct perf_session *session,
+ event_t *event,
+ struct sample_data *sample)
+{
+ if (event->header.type != PERF_RECORD_SAMPLE &&
+ !session->sample_id_all) {
+ fputs("-1 -1 ", stdout);
+ return;
+ }
+
+ if ((session->sample_type & PERF_SAMPLE_CPU))
+ printf("%u ", sample->cpu);
+
+ if (session->sample_type & PERF_SAMPLE_TIME)
+ printf("%Lu ", sample->time);
+}
+
static int perf_session__process_event(struct perf_session *self,
event_t *event,
struct perf_event_ops *ops,
@@ -630,8 +682,12 @@ static int perf_session__process_event(struct perf_session *self,
if (self->header.needs_swap && event__swap_ops[event->header.type])
event__swap_ops[event->header.type](event);
- if (event->header.type == PERF_RECORD_SAMPLE)
- event__parse_sample(event, self->sample_type, &sample);
+ if (event->header.type >= PERF_RECORD_MMAP &&
+ event->header.type <= PERF_RECORD_SAMPLE) {
+ event__parse_sample(event, self, &sample);
+ if (dump_trace)
+ perf_session__print_tstamp(self, event, &sample);
+ }
if (event->header.type < PERF_RECORD_HEADER_MAX) {
dump_printf("%#Lx [%#x]: PERF_RECORD_%s",
diff --git a/tools/perf/util/session.h b/tools/perf/util/session.h
index 4578f86..ac36f99 100644
--- a/tools/perf/util/session.h
+++ b/tools/perf/util/session.h
@@ -46,6 +46,8 @@ struct perf_session {
int fd;
bool fd_pipe;
bool repipe;
+ bool sample_id_all;
+ u16 id_hdr_size;
int cwdlen;
char *cwd;
struct ordered_samples ordered_samples;
@@ -106,6 +108,7 @@ int perf_session__create_kernel_maps(struct perf_session *self);
int do_read(int fd, void *buf, size_t size);
void perf_session__update_sample_type(struct perf_session *self);
+void perf_session__set_sample_id_all(struct perf_session *session, bool value);
void perf_session__set_sample_type(struct perf_session *session, u64 type);
void perf_session__remove_thread(struct perf_session *self, struct thread *th);
--
1.6.2.5
^ permalink raw reply related [flat|nested] 15+ messages in thread
* Re: [GIT PULL 0/4] perf/core improvements
2010-12-02 19:39 [GIT PULL 0/4] perf/core improvements Arnaldo Carvalho de Melo
` (3 preceding siblings ...)
2010-12-02 19:39 ` [PATCH 4/4] perf tools: Ask for ID PERF_SAMPLE_ info on all PERF_RECORD_ events Arnaldo Carvalho de Melo
@ 2010-12-02 23:00 ` Thomas Gleixner
2010-12-03 0:44 ` Arnaldo Carvalho de Melo
2010-12-03 6:10 ` [GIT PULL 0/4] perf/core improvements Ian Munsie
5 siblings, 1 reply; 15+ messages in thread
From: Thomas Gleixner @ 2010-12-02 23:00 UTC (permalink / raw)
To: Arnaldo Carvalho de Melo
Cc: Ingo Molnar, linux-kernel, Frederic Weisbecker, Ian Munsie,
Mike Galbraith, Paul Mackerras, Peter Zijlstra, Stephane Eranian,
Arnaldo Carvalho de Melo
On Thu, 2 Dec 2010, Arnaldo Carvalho de Melo wrote:
> Hi Ingo,
>
> Please consider pulling after at least Ian and Peter acks it:
>
> git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux-2.6 perf/core
>
> Peter already acked a previous version of this, so I left his acked by
> there, this new one is much simpler than the first one he acked, so I think
> he'll be happy with this one as well, Peter?
>
> I tested it using the perf tools built from what is in tip/perf/urgent
> and it works as expected, i.e. the new fields are just discarded since they are
> stashed as a trailer in the MMAP, FORK, etc events, so old tools process new
> perf.data files just fine.
>
> Also the new tools works on older kernels, also tested.
>
> Ian, if you prefer, please test it before Ingo merges it.
Did so. Have some odd fallout:
tracepoint event: type 65535
data->tid 2621441 data->pid 9
Investigating. Please wait
tglx
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [GIT PULL 0/4] perf/core improvements
2010-12-02 23:00 ` [GIT PULL 0/4] perf/core improvements Thomas Gleixner
@ 2010-12-03 0:44 ` Arnaldo Carvalho de Melo
2010-12-03 11:51 ` Thomas Gleixner
0 siblings, 1 reply; 15+ messages in thread
From: Arnaldo Carvalho de Melo @ 2010-12-03 0:44 UTC (permalink / raw)
To: Thomas Gleixner
Cc: Ingo Molnar, linux-kernel, Frederic Weisbecker, Ian Munsie,
Mike Galbraith, Paul Mackerras, Peter Zijlstra, Stephane Eranian
Em Fri, Dec 03, 2010 at 12:00:09AM +0100, Thomas Gleixner escreveu:
> On Thu, 2 Dec 2010, Arnaldo Carvalho de Melo wrote:
>
> > Hi Ingo,
> >
> > Please consider pulling after at least Ian and Peter acks it:
> >
> > git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux-2.6 perf/core
> >
> > Peter already acked a previous version of this, so I left his acked by
> > there, this new one is much simpler than the first one he acked, so I think
> > he'll be happy with this one as well, Peter?
> >
> > I tested it using the perf tools built from what is in tip/perf/urgent
> > and it works as expected, i.e. the new fields are just discarded since they are
> > stashed as a trailer in the MMAP, FORK, etc events, so old tools process new
> > perf.data files just fine.
> >
> > Also the new tools works on older kernels, also tested.
> >
> > Ian, if you prefer, please test it before Ingo merges it.
>
> Did so. Have some odd fallout:
>
> tracepoint event: type 65535
>
> data->tid 2621441 data->pid 9
>
> Investigating. Please wait
Thanks, as I said on IRC, perhaps I neglected the non-perf bits of perf
:-\
- Arnaldo
Guess even for people understanding that "perf" is not just about
profiling that name gets in the way, "observe" is the only thing I think
I would think in a potential rename, but heck, it has 3+ letters :-\
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH 1/4] perf events: Separate the routines handling the PERF_SAMPLE_ identity fields
2010-12-02 19:39 ` [PATCH 1/4] perf events: Separate the routines handling the PERF_SAMPLE_ identity fields Arnaldo Carvalho de Melo
@ 2010-12-03 5:40 ` Ian Munsie
0 siblings, 0 replies; 15+ messages in thread
From: Ian Munsie @ 2010-12-03 5:40 UTC (permalink / raw)
To: Arnaldo Carvalho de Melo
Cc: Ingo Molnar, linux-kernel, Arnaldo Carvalho de Melo,
Frederic Weisbecker, Mike Galbraith, Peter Zijlstra,
Paul Mackerras, Stephane Eranian
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH 2/4] perf events: Make sample_type identity fields available in all PERF_RECORD_ events
2010-12-02 19:39 ` [PATCH 2/4] perf events: Make sample_type identity fields available in all PERF_RECORD_ events Arnaldo Carvalho de Melo
@ 2010-12-03 5:45 ` Ian Munsie
0 siblings, 0 replies; 15+ messages in thread
From: Ian Munsie @ 2010-12-03 5:45 UTC (permalink / raw)
To: Arnaldo Carvalho de Melo
Cc: Ingo Molnar, linux-kernel, Arnaldo Carvalho de Melo,
Frederic Weisbecker, Mike Galbraith, Peter Zijlstra,
Paul Mackerras, Stephane Eranian
> /*
> + * If perf_event_attr.sample_id_all is set then all event types will
> + * have the sample_type selected fields related to where/when
> + * (identity) an event took place (TID, TIME, ID, CPU, STREAM_ID)
> + * described in PERF_RECORD_SAMPLE below, it will be stashed just after
> + * the perf_event_header and the fields already present for the existing
> + * fields, i.e. at the end of the payload. That way a newer perf.data
> + * file will be supported by older perf tools, with these new optional
> + * fields being ignored.
I'd put a comment in each event to be clear, but this is fine too.
> @@ -3944,8 +3988,15 @@ static void perf_event_comm_output(struct perf_event *event,
> struct perf_comm_event *comm_event)
> {
> struct perf_output_handle handle;
> + struct perf_sample_data sample;
> int size = comm_event->event_id.header.size;
> - int ret = perf_output_begin(&handle, event, size, 0, 0);
> + int ret;
How about:
- int size = comm_event->event_id.header.size;
- int ret = perf_output_begin(&handle, event, size, 0, 0);
+ int size, ret;
Like you did for perf_event_mmap_output?
Other than that:
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH 3/4] perf session: Parse sample earlier
2010-12-02 19:39 ` [PATCH 3/4] perf session: Parse sample earlier Arnaldo Carvalho de Melo
@ 2010-12-03 5:46 ` Ian Munsie
0 siblings, 0 replies; 15+ messages in thread
From: Ian Munsie @ 2010-12-03 5:46 UTC (permalink / raw)
To: Arnaldo Carvalho de Melo
Cc: Ingo Molnar, linux-kernel, Arnaldo Carvalho de Melo,
Frederic Weisbecker, Mike Galbraith, Peter Zijlstra,
Paul Mackerras, Stephane Eranian
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [PATCH 4/4] perf tools: Ask for ID PERF_SAMPLE_ info on all PERF_RECORD_ events
2010-12-02 19:39 ` [PATCH 4/4] perf tools: Ask for ID PERF_SAMPLE_ info on all PERF_RECORD_ events Arnaldo Carvalho de Melo
@ 2010-12-03 6:04 ` Ian Munsie
0 siblings, 0 replies; 15+ messages in thread
From: Ian Munsie @ 2010-12-03 6:04 UTC (permalink / raw)
To: Arnaldo Carvalho de Melo
Cc: Ingo Molnar, linux-kernel, Arnaldo Carvalho de Melo,
Frederic Weisbecker, Mike Galbraith, Peter Zijlstra,
Paul Mackerras, Stephane Eranian
Excerpts from Arnaldo Carvalho de Melo's message of Fri Dec 03 06:39:32 +1100 2010:
> +-T::
> +--timestamp::
> + Sample timestamps. Use it with 'perf report -D' to see the timestamps,
> + for instance.
> +
I'm of two minds about allowing the user to request this. My patches
will enable this by default if sampling from multiple CPUs, which is the
norm unless no_inherit is specified, so this option will not have any
affect unless sampling from a single CPU as well.
> + } else if (err == EINVAL && sample_id_all_avail) {
> + /*
> + * Old kernel, no attr->sample_id_type_all field
> + */
> + sample_id_all_avail = false;
> + goto retry_sample_id;
Should we warn the user here? I guess the warning applies more to my
patches then these though, since I'm fixing a bug whereas you are just
adding a new feature, and samples will still have timestamps in your
patches either way.
> + printf("%Lu ", sample->time);
A little off topic, but I've noticed this in a few places throughout the
perf code - the L length modifier in printf is for a long double, not a
long long int (which is ll). The fact that this works with glibc is a
bug (and we've seen how little sympathy the glibc developers have for
programs not using glibc as per the standard). See the note in
CONFORMING TO in the printf man page.
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [GIT PULL 0/4] perf/core improvements
2010-12-02 19:39 [GIT PULL 0/4] perf/core improvements Arnaldo Carvalho de Melo
` (4 preceding siblings ...)
2010-12-02 23:00 ` [GIT PULL 0/4] perf/core improvements Thomas Gleixner
@ 2010-12-03 6:10 ` Ian Munsie
2010-12-03 13:05 ` Arnaldo Carvalho de Melo
5 siblings, 1 reply; 15+ messages in thread
From: Ian Munsie @ 2010-12-03 6:10 UTC (permalink / raw)
To: Arnaldo Carvalho de Melo
Cc: Ingo Molnar, linux-kernel, Frederic Weisbecker, Mike Galbraith,
Paul Mackerras, Peter Zijlstra, Stephane Eranian,
Arnaldo Carvalho de Melo
Excerpts from Arnaldo Carvalho de Melo's message of Fri Dec 03 06:39:28 +1100 2010:
> Ian, if you prefer, please test it before Ingo merges it.
Have done so and just about finished rebasing my patches on top of these
- I'll try to get them posted tonight. The bug I was working on is
fixed, so I'm happy :)
I haven't come across the bug that Thomas ran into, but then again I
haven't been testing the non-perf parts of perf for this either.
Which reminds me, does anyone have a test suite for perf other than
./perf test?
Cheers,
-Ian
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [GIT PULL 0/4] perf/core improvements
2010-12-03 0:44 ` Arnaldo Carvalho de Melo
@ 2010-12-03 11:51 ` Thomas Gleixner
2010-12-07 6:54 ` [tip:perf/core] perf events: Fix event inherit fallout of precalculated headers tip-bot for Thomas Gleixner
0 siblings, 1 reply; 15+ messages in thread
From: Thomas Gleixner @ 2010-12-03 11:51 UTC (permalink / raw)
To: Arnaldo Carvalho de Melo
Cc: Ingo Molnar, linux-kernel, Frederic Weisbecker, Ian Munsie,
Mike Galbraith, Paul Mackerras, Peter Zijlstra, Stephane Eranian
On Thu, 2 Dec 2010, Arnaldo Carvalho de Melo wrote:
> Em Fri, Dec 03, 2010 at 12:00:09AM +0100, Thomas Gleixner escreveu:
> > Investigating. Please wait
>
> Thanks, as I said on IRC, perhaps I neglected the non-perf bits of perf
It's a perf bit of perf. You simply broke the inheritance
code. Testing perf with something more complex than "sleep 1" might
have shown this :)
------------>
Subject: perf: Fix event inherit fallout of precalculated headers
From: Thomas Gleixner <tglx@linutronix.de>
Date: Fri, 03 Dec 2010 11:42:10 +0100
The precalculated header size is not updated when an event is inherited. That
results in bogus sample entries for all child events.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
---
kernel/perf_event.c | 6 ++++++
1 file changed, 6 insertions(+)
Index: linux-2.6-tip/kernel/perf_event.c
===================================================================
--- linux-2.6-tip.orig/kernel/perf_event.c
+++ linux-2.6-tip/kernel/perf_event.c
@@ -6184,6 +6184,12 @@ inherit_event(struct perf_event *parent_
child_event->overflow_handler = parent_event->overflow_handler;
/*
+ * Precalculate sample_data sizes
+ */
+ perf_event__header_size(child_event);
+ perf_event__id_header_size(child_event);
+
+ /*
* Link it up in the child's context:
*/
raw_spin_lock_irqsave(&child_ctx->lock, flags);
^ permalink raw reply [flat|nested] 15+ messages in thread
* Re: [GIT PULL 0/4] perf/core improvements
2010-12-03 6:10 ` [GIT PULL 0/4] perf/core improvements Ian Munsie
@ 2010-12-03 13:05 ` Arnaldo Carvalho de Melo
0 siblings, 0 replies; 15+ messages in thread
From: Arnaldo Carvalho de Melo @ 2010-12-03 13:05 UTC (permalink / raw)
To: Ian Munsie
Cc: Ingo Molnar, linux-kernel, Frederic Weisbecker, Mike Galbraith,
Paul Mackerras, Peter Zijlstra, Stephane Eranian
Em Fri, Dec 03, 2010 at 05:10:44PM +1100, Ian Munsie escreveu:
> Excerpts from Arnaldo Carvalho de Melo's message of Fri Dec 03 06:39:28 +1100 2010:
> > Ian, if you prefer, please test it before Ingo merges it.
>
> Have done so and just about finished rebasing my patches on top of these
> - I'll try to get them posted tonight. The bug I was working on is
> fixed, so I'm happy :)
Great!
> I haven't come across the bug that Thomas ran into, but then again I
> haven't been testing the non-perf parts of perf for this either.
I'll work on the problem related by Thomas today.
> Which reminds me, does anyone have a test suite for perf other than
> ./perf test?
Not that I know of, perf test is intended to be that, please consider
adding more tests :)
I want to do things like start a thread that will generate cache misses,
and then have a monitor thread measuring it, passing the test if it gets
into some range (its sampling after all), etc.
- Arnaldo
^ permalink raw reply [flat|nested] 15+ messages in thread
* [tip:perf/core] perf events: Fix event inherit fallout of precalculated headers
2010-12-03 11:51 ` Thomas Gleixner
@ 2010-12-07 6:54 ` tip-bot for Thomas Gleixner
0 siblings, 0 replies; 15+ messages in thread
From: tip-bot for Thomas Gleixner @ 2010-12-07 6:54 UTC (permalink / raw)
To: linux-tip-commits
Cc: acme, linux-kernel, eranian, paulus, hpa, mingo, a.p.zijlstra,
efault, imunsie, fweisbec, tglx, mingo
Commit-ID: 614b6780eb0c393d2fb49ff62d61f29b877bd07e
Gitweb: http://git.kernel.org/tip/614b6780eb0c393d2fb49ff62d61f29b877bd07e
Author: Thomas Gleixner <tglx@linutronix.de>
AuthorDate: Fri, 3 Dec 2010 16:24:32 -0200
Committer: Arnaldo Carvalho de Melo <acme@redhat.com>
CommitDate: Sat, 4 Dec 2010 22:56:11 -0200
perf events: Fix event inherit fallout of precalculated headers
The precalculated header size is not updated when an event is inherited. That
results in bogus sample entries for all child events. Bug introduced in c320c7b.
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Ian Munsie <imunsie@au1.ibm.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Stephane Eranian <eranian@google.com>
LKML-Reference: <alpine.LFD.2.00.1012031245220.2653@localhost6.localdomain6>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
kernel/perf_event.c | 5 +++++
1 files changed, 5 insertions(+), 0 deletions(-)
diff --git a/kernel/perf_event.c b/kernel/perf_event.c
index aede712..7961b27 100644
--- a/kernel/perf_event.c
+++ b/kernel/perf_event.c
@@ -6099,6 +6099,11 @@ inherit_event(struct perf_event *parent_event,
child_event->overflow_handler = parent_event->overflow_handler;
/*
+ * Precalculate sample_data sizes
+ */
+ perf_event__header_size(child_event);
+
+ /*
* Link it up in the child's context:
*/
raw_spin_lock_irqsave(&child_ctx->lock, flags);
^ permalink raw reply related [flat|nested] 15+ messages in thread
end of thread, other threads:[~2010-12-07 6:55 UTC | newest]
Thread overview: 15+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2010-12-02 19:39 [GIT PULL 0/4] perf/core improvements Arnaldo Carvalho de Melo
2010-12-02 19:39 ` [PATCH 1/4] perf events: Separate the routines handling the PERF_SAMPLE_ identity fields Arnaldo Carvalho de Melo
2010-12-03 5:40 ` Ian Munsie
2010-12-02 19:39 ` [PATCH 2/4] perf events: Make sample_type identity fields available in all PERF_RECORD_ events Arnaldo Carvalho de Melo
2010-12-03 5:45 ` Ian Munsie
2010-12-02 19:39 ` [PATCH 3/4] perf session: Parse sample earlier Arnaldo Carvalho de Melo
2010-12-03 5:46 ` Ian Munsie
2010-12-02 19:39 ` [PATCH 4/4] perf tools: Ask for ID PERF_SAMPLE_ info on all PERF_RECORD_ events Arnaldo Carvalho de Melo
2010-12-03 6:04 ` Ian Munsie
2010-12-02 23:00 ` [GIT PULL 0/4] perf/core improvements Thomas Gleixner
2010-12-03 0:44 ` Arnaldo Carvalho de Melo
2010-12-03 11:51 ` Thomas Gleixner
2010-12-07 6:54 ` [tip:perf/core] perf events: Fix event inherit fallout of precalculated headers tip-bot for Thomas Gleixner
2010-12-03 6:10 ` [GIT PULL 0/4] perf/core improvements Ian Munsie
2010-12-03 13:05 ` Arnaldo Carvalho de Melo
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.