From: Daniel Wagner <wagi@monom.org>
To: Binoy Jayan <binoy.jayan@linaro.org>
Cc: Arnd Bergmann <arnd@arndb.de>,
linaro-kernel@lists.linaro.org,
Daniel Wagner <daniel.wagner@bmw-carit.de>,
Carsten Emde <C.Emde@osadl.org>,
linux-kernel@vger.kernel.org,
"Steven Rostedt (Red Hat)" <rostedt@goodmis.org>
Subject: Re: [PATCH 2/3] tracing: Add trace_irqsoff tracepoints
Date: Thu, 18 Aug 2016 10:44:12 +0200 [thread overview]
Message-ID: <20160818084412.b24japksnn6hq45n@nobby.bmw-carit.de> (raw)
In-Reply-To: <1471504166-27665-3-git-send-email-binoy.jayan@linaro.org>
On Thu, Aug 18, 2016 at 12:39:25PM +0530, Binoy Jayan wrote:
> static struct trace_array *irqsoff_trace __read_mostly;
> static int tracer_enabled __read_mostly;
>
> static DEFINE_PER_CPU(int, tracing_cpu);
> -
> +static DEFINE_PER_CPU(cycle_t, ts_irqs);
> +static DEFINE_PER_CPU(cycle_t, ts_preempt);
> +static DEFINE_PER_CPU(cycle_t, ts_critical_timings);
> static DEFINE_RAW_SPINLOCK(max_trace_lock);
What about creating cycle_t array and introduce only one tracepoint
type which containts the type? I think it should be possible to get
the right plot via hist's filter option:
'hist:key=latency.bucket:val=hitcount:sort=latency if cpu==0 && type==0'
Obviously, there is a performance impact. It would be good to see some
numbers.
cheers,
daniel
Just as rough idea (not tested):
diff --git a/kernel/trace/trace_irqsoff.c b/kernel/trace/trace_irqsoff.c
index 03cdff8..012544c 100644
--- a/kernel/trace/trace_irqsoff.c
+++ b/kernel/trace/trace_irqsoff.c
@@ -19,7 +19,15 @@
static struct trace_array *irqsoff_trace __read_mostly;
static int tracer_enabled __read_mostly;
+enum latency_type {
+ LAT_IRQ,
+ LAT_PREEMPT,
+ LAT_CRITTIME,
+ LAT_MAX
+};
+
static DEFINE_PER_CPU(int, tracing_cpu);
+static DEFINE_PER_CPU(cycle_t[LAT_MAX], lat_data);
static DEFINE_RAW_SPINLOCK(max_trace_lock);
@@ -350,6 +358,19 @@ out:
__trace_function(tr, CALLER_ADDR0, parent_ip, flags, pc);
}
+static inline void lat_timestamp(enum latency_type type)
+{
+ int cpu = raw_smp_processor_id();
+ per_cpu(lat_data[type], cpu) = ftrace_now(cpu);
+}
+
+static inline void lat_trace(enum latency_type type)
+{
+ int cpu = raw_smp_processor_id();
+ trace_latency_preempt(type,
+ ftrace_now(cpu) - per_cpu(lat_data[type], cpu));
+}
+
static inline void
start_critical_timing(unsigned long ip, unsigned long parent_ip)
{
@@ -422,6 +443,9 @@ stop_critical_timing(unsigned long ip, unsigned long parent_ip)
/* start and stop critical timings used to for stoppage (in idle) */
void start_critical_timings(void)
{
+ if (trace_latency_preempt_enabled())
+ lat_timestamp(LAT_CRITTIME);
+
if (preempt_trace() || irq_trace())
start_critical_timing(CALLER_ADDR0, CALLER_ADDR1);
}
@@ -431,6 +455,9 @@ void stop_critical_timings(void)
{
if (preempt_trace() || irq_trace())
stop_critical_timing(CALLER_ADDR0, CALLER_ADDR1);
+
+ if (trace_latency_preempt_enabled())
+ lat_trace(LAT_CRITTIME);
}
EXPORT_SYMBOL_GPL(stop_critical_timings);
@@ -438,6 +465,9 @@ EXPORT_SYMBOL_GPL(stop_critical_timings);
#ifdef CONFIG_PROVE_LOCKING
void time_hardirqs_on(unsigned long a0, unsigned long a1)
{
+ if (trace_latency_preempt_enable())
+ lat_trace(LAT_CRITTIME);
+
if (!preempt_trace() && irq_trace())
stop_critical_timing(a0, a1);
}
@@ -446,6 +476,9 @@ void time_hardirqs_off(unsigned long a0, unsigned long a1)
{
if (!preempt_trace() && irq_trace())
start_critical_timing(a0, a1);
+
+ if (trace_latency_preempt_enabled())
+ lat_timestamp(LAT_IRQ);
}
#else /* !CONFIG_PROVE_LOCKING */
next prev parent reply other threads:[~2016-08-18 8:44 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-08-18 7:09 [PATCH 0/3] *** Latency trace events - irqs, preempt, critical timing *** Binoy Jayan
2016-08-18 7:09 ` [PATCH 1/3] tracing: Deference pointers without RCU checks Binoy Jayan
2016-08-18 7:37 ` Daniel Wagner
2016-08-18 7:09 ` [PATCH 2/3] tracing: Add trace_irqsoff tracepoints Binoy Jayan
2016-08-18 8:44 ` Daniel Wagner [this message]
2016-08-18 14:21 ` Steven Rostedt
2016-08-22 6:00 ` kbuild test robot
2016-08-22 7:37 ` kbuild test robot
2016-08-18 7:09 ` [PATCH 3/3] tracing: Histogram for missed timer offsets Binoy Jayan
2016-08-22 6:41 ` kbuild test robot
2016-08-22 8:44 ` kbuild test robot
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=20160818084412.b24japksnn6hq45n@nobby.bmw-carit.de \
--to=wagi@monom.org \
--cc=C.Emde@osadl.org \
--cc=arnd@arndb.de \
--cc=binoy.jayan@linaro.org \
--cc=daniel.wagner@bmw-carit.de \
--cc=linaro-kernel@lists.linaro.org \
--cc=linux-kernel@vger.kernel.org \
--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).