From: Masami Hiramatsu <masami.hiramatsu@linaro.org>
To: Binoy Jayan <binoy.jayan@linaro.org>
Cc: "Steven Rostedt (Red Hat)" <rostedt@goodmis.org>,
Ingo Molnar <mingo@redhat.com>,
Daniel Wagner <daniel.wagner@bmw-carit.de>,
Arnd Bergmann <arnd@arndb.de>,
linux-kernel@vger.kernel.org
Subject: Re: [PATCH v4 3/3] tracing: Histogram for missed timer offsets
Date: Tue, 30 Aug 2016 19:50:34 +0900 [thread overview]
Message-ID: <CAA93ih3k+f-gSMqbmU_TPBs7rzwy9gRDaGNh_ZNofw1H63N83g@mail.gmail.com> (raw)
In-Reply-To: <1472552924-22297-4-git-send-email-binoy.jayan@linaro.org>
Hi Binoy,
2016-08-30 19:28 GMT+09:00 Binoy Jayan <binoy.jayan@linaro.org>:
> Latencies of missed timer offsets. Generate a histogram of missed
> timer offsets in microseconds. This will be a based along with irq
> and preemption latencies to calculate the effective process wakeup
> latencies.
>
> The following filter(s) may be used
>
> 'hist:key=common_pid.execname'
> 'hist:key=common_pid.execname:val=toffset,hitcount'
>
> Signed-off-by: Binoy Jayan <binoy.jayan@linaro.org>
> ---
> include/linux/hrtimer.h | 3 +++
> include/trace/events/latency.h | 29 +++++++++++++++++++++++++++++
> kernel/time/hrtimer.c | 40 ++++++++++++++++++++++++++++++++++++++++
> 3 files changed, 72 insertions(+)
>
> diff --git a/include/linux/hrtimer.h b/include/linux/hrtimer.h
> index 5e00f80..e09de14 100644
> --- a/include/linux/hrtimer.h
> +++ b/include/linux/hrtimer.h
> @@ -104,6 +104,9 @@ struct hrtimer {
> struct hrtimer_clock_base *base;
> u8 state;
> u8 is_rel;
> +#if defined(CONFIG_PREEMPT_TRACER) || defined(CONFIG_IRQSOFF_TRACER)
> + ktime_t praecox;
> +#endif
> #ifdef CONFIG_TIMER_STATS
> int start_pid;
> void *start_site;
> diff --git a/include/trace/events/latency.h b/include/trace/events/latency.h
> index e89be12..7fca7cd 100644
> --- a/include/trace/events/latency.h
> +++ b/include/trace/events/latency.h
> @@ -29,6 +29,35 @@ DEFINE_EVENT(latency_template, latency_preempt,
> TP_PROTO(int ltype, cycles_t latency),
> TP_ARGS(ltype, latency));
>
> +TRACE_EVENT(latency_hrtimer_interrupt,
> +
> + TP_PROTO(long long toffset, struct task_struct *curr,
> + struct task_struct *task),
> +
> + TP_ARGS(toffset, curr, task),
> +
> + TP_STRUCT__entry(
> + __field(long long, toffset)
> + __array(char, ccomm, TASK_COMM_LEN)
> + __field(int, cprio)
> + __array(char, tcomm, TASK_COMM_LEN)
> + __field(int, tprio)
> + ),
> +
> + TP_fast_assign(
> + __entry->toffset = toffset;
> + memcpy(__entry->ccomm, curr->comm, TASK_COMM_LEN);
> + __entry->cprio = curr->prio;
> + memcpy(__entry->tcomm, task != NULL ? task->comm : "<none>",
> + task != NULL ? TASK_COMM_LEN : 7);
> + __entry->tprio = task != NULL ? task->prio : -1;
> + ),
> +
> + TP_printk("toffset=%lld curr=%s[%d] thread=%s[%d]",
> + __entry->toffset, __entry->ccomm, __entry->cprio,
> + __entry->tcomm, __entry->tprio)
> +);
> +
> #endif /* _TRACE_HIST_H */
>
> /* This part must be outside protection */
> diff --git a/kernel/time/hrtimer.c b/kernel/time/hrtimer.c
> index 9ba7c82..f3e1d92 100644
> --- a/kernel/time/hrtimer.c
> +++ b/kernel/time/hrtimer.c
> @@ -53,9 +53,12 @@
> #include <asm/uaccess.h>
>
> #include <trace/events/timer.h>
> +#include <trace/events/latency.h>
>
> #include "tick-internal.h"
>
> +static enum hrtimer_restart hrtimer_wakeup(struct hrtimer *timer);
> +
> /*
> * The timer bases:
> *
> @@ -960,6 +963,39 @@ static inline ktime_t hrtimer_update_lowres(struct hrtimer *timer, ktime_t tim,
> return tim;
> }
>
> +static inline void trace_latency_hrtimer_mark_ts(struct hrtimer *timer,
> + struct hrtimer_clock_base *new_base,
> + ktime_t tim)
> +{
> +#if defined(CONFIG_PREEMPT_TRACER) || defined(CONFIG_IRQSOFF_TRACER)
> + if (trace_latency_hrtimer_interrupt_enabled()) {
You would better use unlikely() here.
> + ktime_t now = new_base->get_time();
> +
> + if (ktime_to_ns(tim) < ktime_to_ns(now))
Wouldn't we need to consider the case of wrap around?
> + timer->praecox = now;
> + else
> + timer->praecox = ktime_set(0, 0);
> + }
> +#endif
> +}
> +
> +static inline void trace_latency_missed_hrtimer(struct hrtimer *timer,
> + ktime_t basenow)
> +{
> +#if defined(CONFIG_PREEMPT_TRACER) || defined(CONFIG_IRQSOFF_TRACER)
> + if (trace_latency_hrtimer_interrupt_enabled())
Here, you'd better add unlikely() too.
> + trace_latency_hrtimer_interrupt(
> + ktime_to_ns(ktime_sub(ktime_to_ns(timer->praecox) ?
> + timer->praecox : hrtimer_get_expires(timer),
> + basenow)),
> + current,
> + timer->function == hrtimer_wakeup ?
> + container_of(timer, struct hrtimer_sleeper,
> + timer)->task : NULL);
> +#endif
> +
> +}
> +
> /**
> * hrtimer_start_range_ns - (re)start an hrtimer on the current CPU
> * @timer: the timer to be added
> @@ -992,6 +1028,8 @@ void hrtimer_start_range_ns(struct hrtimer *timer, ktime_t tim,
>
> timer_stats_hrtimer_set_start_info(timer);
>
> + trace_latency_hrtimer_mark_ts(timer, new_base, tim);
Hmm, since these calls are not actual tracepoints, it should not start
with "trace_".
I would like to ask you to use other function name, like
hrtimer_mark_start_timestamp()
> +
> leftmost = enqueue_hrtimer(timer, new_base);
> if (!leftmost)
> goto unlock;
> @@ -1284,6 +1322,8 @@ static void __hrtimer_run_queues(struct hrtimer_cpu_base *cpu_base, ktime_t now)
>
> timer = container_of(node, struct hrtimer, node);
>
> + trace_latency_missed_hrtimer(timer, basenow);
Here, too. hrtimer_expired_latency() etc.
BTW, I think "missed" is a bit misleadable, since the timer itself is
not missed,
it may be just not in time. :)
Thank you,
> +
> /*
> * The immediate goal for using the softexpires is
> * minimizing wakeups, not running timers at the
> --
> The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
> a Linux Foundation Collaborative Project
>
--
Masami Hiramatsu
next prev parent reply other threads:[~2016-08-30 10:50 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-08-30 10:28 [PATCH v4 0/3] *** Latency histograms - IRQSOFF,PREEMPTOFF *** Binoy Jayan
2016-08-30 10:28 ` [PATCH v4 1/3] tracing: Deference pointers without RCU checks Binoy Jayan
2016-08-30 10:28 ` [PATCH v4 2/3] tracing: Add trace_irqsoff tracepoints Binoy Jayan
2016-08-30 11:30 ` Daniel Wagner
2016-08-30 11:49 ` Binoy Jayan
2016-08-30 14:08 ` Steven Rostedt
2016-08-31 11:39 ` Binoy Jayan
2016-08-30 10:28 ` [PATCH v4 3/3] tracing: Histogram for missed timer offsets Binoy Jayan
2016-08-30 10:50 ` Masami Hiramatsu [this message]
2016-09-02 12:41 ` Binoy Jayan
2016-09-02 15:54 ` Masami Hiramatsu
2016-08-30 14:15 ` Steven Rostedt
2016-08-31 11:34 ` Binoy Jayan
2016-08-31 12:42 ` Steven Rostedt
2016-08-30 11:39 ` [PATCH v4 0/3] *** Latency histograms - IRQSOFF,PREEMPTOFF *** Daniel Wagner
2016-08-30 12:02 ` Masami Hiramatsu
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=CAA93ih3k+f-gSMqbmU_TPBs7rzwy9gRDaGNh_ZNofw1H63N83g@mail.gmail.com \
--to=masami.hiramatsu@linaro.org \
--cc=arnd@arndb.de \
--cc=binoy.jayan@linaro.org \
--cc=daniel.wagner@bmw-carit.de \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@redhat.com \
--cc=rostedt@goodmis.org \
/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).