On 08/05/2016 04:52 PM, Steven Rostedt wrote:

quoted

quoted

--- a/kernel/trace/trace_hwlat.c
+++ b/kernel/trace/trace_hwlat.c

@@ -64,6 +64,15 @@ static struct dentry *hwlat_sample_window;	/* sample window us */

/* Save the previous tracing_thresh value */
static unsigned long save_tracing_thresh;

+/* NMI timestamp counters */
+static u64 nmi_ts_start;
+static u64 nmi_total_ts;
+static int nmi_count;
+static int nmi_cpu;  

and this is always limited to one CPU at a time?

Yes. Hence the "nmi_cpu".

I was just confused. So we check one CPU at a time. Okay.

quoted

quoted

@@ -125,6 +138,19 @@ static void trace_hwlat_sample(struct hwlat_sample *sample)

#define init_time(a, b)	(a = b)
#define time_u64(a)	a

+void trace_hwlat_callback(bool enter)
+{
+	if (smp_processor_id() != nmi_cpu)
+		return;
+
+	if (enter)
+		nmi_ts_start = time_get();  

but more interestingly: trace_clock_local() -> sched_clock()
and of kernel/time/sched_clock.c we do raw_read_seqcount(&cd.seq) which
means we are busted if the NMI triggers during update_clock_read_data().

Hmm, interesting. Because this is true for general tracing from an NMI.

/me looks at code.

Ah, this is when we have GENERIC_SCHED_CLOCK, which would break tracing
if any arch that has this also has NMIs. Probably need to look at arm64.

arm64 should use the generic code as they don't provide sched_clock()
(and I doubt they go for the weak jiffy version).

For x86, it has its own NMI safe sched_clock. I could make this "NMI"
code depend on:

#ifndef CONFIG_GENERIC_SCHED_CLOCK

that would be nice. That would be disable approx $(git grep
sched_clock_register | wc -l) users but better than a lock up I guess.

-- Steve

Sebastian