From: Ajay Kaher <ajay.kaher@broadcom.com>
To: peterz@infradead.org, mingo@redhat.com, acme@kernel.org,
namhyung@kernel.org
Cc: mark.rutland@arm.com, rostedt@goodmis.org,
alexander.shishkin@linux.intel.com, jolsa@kernel.org,
irogers@google.com, adrian.hunter@intel.com,
kan.liang@linux.intel.com, yangjihong1@huawei.com,
zegao2021@gmail.com, leo.yan@linux.dev, asmadeus@codewreck.org,
siyanteng@loongson.cn, sunhaiyong@loongson.cn,
linux-perf-users@vger.kernel.org, linux-kernel@vger.kernel.org,
ajay.kaher@broadcom.com, alexey.makhalov@broadcom.com,
vasavi.sirnapalli@broadcom.com
Subject: [PATCH RFC 1/3] perf/core: add logic to collect off-cpu sample
Date: Thu, 11 Jul 2024 17:46:17 +0530 [thread overview]
Message-ID: <1720700179-22839-2-git-send-email-ajay.kaher@broadcom.com> (raw)
In-Reply-To: <1720700179-22839-1-git-send-email-ajay.kaher@broadcom.com>
following logics has been added to collect the off-cpu sample:
- 'task_pt_regs(current)' has been used to capture registers
status off-cpu sample.
- off-cpu time represent the time period for which the target
process not occupying the cpu cycles. And calculate as:
off-cpu time = swap-in time - swap-out time
Signed-off-by: Ajay Kaher <ajay.kaher@broadcom.com>
---
include/linux/perf_event.h | 16 ++++++++++++++++
include/uapi/linux/perf_event.h | 3 ++-
kernel/events/core.c | 27 ++++++++++++++++++++++-----
tools/include/uapi/linux/perf_event.h | 3 ++-
4 files changed, 42 insertions(+), 7 deletions(-)
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index a5304ae8c654..09dc3f695974 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -599,6 +599,11 @@ enum perf_event_state {
PERF_EVENT_STATE_ACTIVE = 1,
};
+enum perf_sample_cpu {
+ PERF_SAMPLE_ON_CPU,
+ PERF_SAMPLE_OFF_CPU,
+};
+
struct file;
struct perf_sample_data;
@@ -828,6 +833,7 @@ struct perf_event {
void *security;
#endif
struct list_head sb_list;
+ u64 stop_time;
/*
* Certain events gets forwarded to another pmu internally by over-
@@ -1301,6 +1307,16 @@ static inline u32 perf_sample_data_size(struct perf_sample_data *data,
return size;
}
+static inline void perf_sample_set_off_cpu(struct perf_sample_data *data)
+{
+ data->cpu_entry.reserved = PERF_SAMPLE_OFF_CPU;
+}
+
+static inline void perf_sample_set_on_cpu(struct perf_sample_data *data)
+{
+ data->cpu_entry.reserved = PERF_SAMPLE_ON_CPU;
+}
+
/*
* Clear all bitfields in the perf_branch_entry.
* The to and from fields are not cleared because they are
diff --git a/include/uapi/linux/perf_event.h b/include/uapi/linux/perf_event.h
index 3a64499b0f5d..9d0a23ab2549 100644
--- a/include/uapi/linux/perf_event.h
+++ b/include/uapi/linux/perf_event.h
@@ -460,7 +460,8 @@ struct perf_event_attr {
inherit_thread : 1, /* children only inherit if cloned with CLONE_THREAD */
remove_on_exec : 1, /* event is removed from task on exec */
sigtrap : 1, /* send synchronous SIGTRAP on event */
- __reserved_1 : 26;
+ off_cpu : 1, /* include off_cpu sample */
+ __reserved_1 : 25;
union {
__u32 wakeup_events; /* wakeup every n events */
diff --git a/kernel/events/core.c b/kernel/events/core.c
index 8f908f077935..e71eb7936134 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -7168,10 +7168,8 @@ static void __perf_event_header__init_id(struct perf_sample_data *data,
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_CPU)
+ data->cpu_entry.cpu = raw_smp_processor_id();
}
void perf_event_header__init_id(struct perf_event_header *header,
@@ -11083,8 +11081,8 @@ static enum hrtimer_restart perf_swevent_hrtimer(struct hrtimer *hrtimer)
return HRTIMER_NORESTART;
event->pmu->read(event);
-
perf_sample_data_init(&data, 0, event->hw.last_period);
+ perf_sample_set_on_cpu(&data);
regs = get_irq_regs();
if (regs && !perf_exclude_event(event, regs)) {
@@ -11099,6 +11097,18 @@ static enum hrtimer_restart perf_swevent_hrtimer(struct hrtimer *hrtimer)
return ret;
}
+static void perf_swevent_offCPUSample(struct perf_event *event, u64 period)
+{
+ struct perf_sample_data data;
+ struct pt_regs *regs;
+
+ event->pmu->read(event);
+ perf_sample_data_init(&data, 0, period);
+ perf_sample_set_off_cpu(&data);
+ regs = task_pt_regs(current);
+ __perf_event_overflow(event, 1, &data, regs);
+}
+
static void perf_swevent_start_hrtimer(struct perf_event *event)
{
struct hw_perf_event *hwc = &event->hw;
@@ -11107,6 +11117,11 @@ static void perf_swevent_start_hrtimer(struct perf_event *event)
if (!is_sampling_event(event))
return;
+ if (event->attr.off_cpu && event->stop_time && hwc->sample_period) {
+ perf_swevent_offCPUSample(event, perf_clock() - event->stop_time);
+ event->stop_time = 0;
+ }
+
period = local64_read(&hwc->period_left);
if (period) {
if (period < 0)
@@ -11128,6 +11143,7 @@ static void perf_swevent_cancel_hrtimer(struct perf_event *event)
ktime_t remaining = hrtimer_get_remaining(&hwc->hrtimer);
local64_set(&hwc->period_left, ktime_to_ns(remaining));
+ event->stop_time = perf_clock();
hrtimer_cancel(&hwc->hrtimer);
}
}
@@ -11139,6 +11155,7 @@ static void perf_swevent_init_hrtimer(struct perf_event *event)
if (!is_sampling_event(event))
return;
+ event->stop_time = 0;
hrtimer_init(&hwc->hrtimer, CLOCK_MONOTONIC, HRTIMER_MODE_REL_HARD);
hwc->hrtimer.function = perf_swevent_hrtimer;
diff --git a/tools/include/uapi/linux/perf_event.h b/tools/include/uapi/linux/perf_event.h
index 3a64499b0f5d..9d0a23ab2549 100644
--- a/tools/include/uapi/linux/perf_event.h
+++ b/tools/include/uapi/linux/perf_event.h
@@ -460,7 +460,8 @@ struct perf_event_attr {
inherit_thread : 1, /* children only inherit if cloned with CLONE_THREAD */
remove_on_exec : 1, /* event is removed from task on exec */
sigtrap : 1, /* send synchronous SIGTRAP on event */
- __reserved_1 : 26;
+ off_cpu : 1, /* include off_cpu sample */
+ __reserved_1 : 25;
union {
__u32 wakeup_events; /* wakeup every n events */
--
2.39.0
next prev parent reply other threads:[~2024-07-11 12:17 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-07-11 12:16 [PATCH RFC 0/3] perf: add logic to collect off-cpu samples Ajay Kaher
2024-07-11 12:16 ` Ajay Kaher [this message]
2024-07-11 21:49 ` [PATCH RFC 1/3] perf/core: add logic to collect off-cpu sample Peter Zijlstra
2024-07-14 16:23 ` Ajay Kaher
2024-07-15 11:48 ` Peter Zijlstra
2024-07-11 12:16 ` [PATCH RFC 2/3] perf/record: add options --off-cpu-kernel Ajay Kaher
2024-07-11 12:16 ` [PATCH RFC 3/3] perf/report: add off-cpu samples Ajay Kaher
2024-07-11 21:58 ` [PATCH RFC 0/3] perf: add logic to collect " Ian Rogers
2024-07-13 7:42 ` Ajay Kaher
2024-07-14 18:32 ` Ian Rogers
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1720700179-22839-2-git-send-email-ajay.kaher@broadcom.com \
--to=ajay.kaher@broadcom.com \
--cc=acme@kernel.org \
--cc=adrian.hunter@intel.com \
--cc=alexander.shishkin@linux.intel.com \
--cc=alexey.makhalov@broadcom.com \
--cc=asmadeus@codewreck.org \
--cc=irogers@google.com \
--cc=jolsa@kernel.org \
--cc=kan.liang@linux.intel.com \
--cc=leo.yan@linux.dev \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-perf-users@vger.kernel.org \
--cc=mark.rutland@arm.com \
--cc=mingo@redhat.com \
--cc=namhyung@kernel.org \
--cc=peterz@infradead.org \
--cc=rostedt@goodmis.org \
--cc=siyanteng@loongson.cn \
--cc=sunhaiyong@loongson.cn \
--cc=vasavi.sirnapalli@broadcom.com \
--cc=yangjihong1@huawei.com \
--cc=zegao2021@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).