[PATCH] perf lock contention: Clear lock addr after use

[PATCH] perf lock contention: Clear lock addr after use

[Namhyung Kim]

It checks the current lock to calculated the delta of contention time.
The address is saved in the tstamp map which is allocated at begining of
contention and released at end of contention.

But it's possible for bpf_map_delete_elem() to fail.  In that case, the
element in the tstamp map kept for the current lock and it makes the
next contention for the same lock tracked incorrectly.  Specificially
the next contention begin will see the existing element for the task and
it'd just return.  Then the next contention end will see the element and
calculate the time using the timestamp for the previous begin.

This can result in a large value for two small contentions happened from
time to time.  Let's clear the lock address so that it can be updated
next time even if the bpf_map_delete_elem() failed.

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
---
 tools/perf/util/bpf_skel/lock_contention.bpf.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/tools/perf/util/bpf_skel/lock_contention.bpf.c b/tools/perf/util/bpf_skel/lock_contention.bpf.c
index 4900a5dfb4a4..b11179452e19 100644
--- a/tools/perf/util/bpf_skel/lock_contention.bpf.c
+++ b/tools/perf/util/bpf_skel/lock_contention.bpf.c
@@ -389,6 +389,7 @@ int contention_end(u64 *ctx)
 
 	duration = bpf_ktime_get_ns() - pelem->timestamp;
 	if ((__s64)duration < 0) {
+		pelem->lock = 0;
 		bpf_map_delete_elem(&tstamp, &pid);
 		__sync_fetch_and_add(&time_fail, 1);
 		return 0;
@@ -422,6 +423,7 @@ int contention_end(u64 *ctx)
 	data = bpf_map_lookup_elem(&lock_stat, &key);
 	if (!data) {
 		if (data_map_full) {
+			pelem->lock = 0;
 			bpf_map_delete_elem(&tstamp, &pid);
 			__sync_fetch_and_add(&data_fail, 1);
 			return 0;
@@ -445,6 +447,7 @@ int contention_end(u64 *ctx)
 				data_map_full = 1;
 			__sync_fetch_and_add(&data_fail, 1);
 		}
+		pelem->lock = 0;
 		bpf_map_delete_elem(&tstamp, &pid);
 		return 0;
 	}
@@ -458,6 +461,7 @@ int contention_end(u64 *ctx)
 	if (data->min_time > duration)
 		data->min_time = duration;
 
+	pelem->lock = 0;
 	bpf_map_delete_elem(&tstamp, &pid);
 	return 0;
 }
-- 
2.42.0.582.g8ccd20d70d-goog

Re: [PATCH] perf lock contention: Clear lock addr after use

[Arnaldo Carvalho de Melo]

Em Thu, Sep 28, 2023 at 04:50:18PM -0700, Namhyung Kim escreveu:
> It checks the current lock to calculated the delta of contention time.

> The address is saved in the tstamp map which is allocated at begining of
> contention and released at end of contention.
> 
> But it's possible for bpf_map_delete_elem() to fail.  In that case, the

How can it fail? 

You do:

        pelem = bpf_map_lookup_elem(&tstamp, &pid);
        if (!pelem || pelem->lock != ctx[0])
                return 0;

So it is there, why would the removal using the same key fail?

The patch should work as-is, I'm just curious about what would make
there removal of a map entry that was successfully looked up on the same
contention_end prog to fail when being removed...

- Arnaldo

> element in the tstamp map kept for the current lock and it makes the
> next contention for the same lock tracked incorrectly.  Specificially
> the next contention begin will see the existing element for the task and
> it'd just return.  Then the next contention end will see the element and
> calculate the time using the timestamp for the previous begin.
> 
> This can result in a large value for two small contentions happened from
> time to time.  Let's clear the lock address so that it can be updated
> next time even if the bpf_map_delete_elem() failed.
> 
> Signed-off-by: Namhyung Kim <namhyung@kernel.org>
> ---
>  tools/perf/util/bpf_skel/lock_contention.bpf.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/tools/perf/util/bpf_skel/lock_contention.bpf.c b/tools/perf/util/bpf_skel/lock_contention.bpf.c
> index 4900a5dfb4a4..b11179452e19 100644
> --- a/tools/perf/util/bpf_skel/lock_contention.bpf.c
> +++ b/tools/perf/util/bpf_skel/lock_contention.bpf.c
> @@ -389,6 +389,7 @@ int contention_end(u64 *ctx)
>  
>  	duration = bpf_ktime_get_ns() - pelem->timestamp;
>  	if ((__s64)duration < 0) {
> +		pelem->lock = 0;
>  		bpf_map_delete_elem(&tstamp, &pid);
>  		__sync_fetch_and_add(&time_fail, 1);
>  		return 0;
> @@ -422,6 +423,7 @@ int contention_end(u64 *ctx)
>  	data = bpf_map_lookup_elem(&lock_stat, &key);
>  	if (!data) {
>  		if (data_map_full) {
> +			pelem->lock = 0;
>  			bpf_map_delete_elem(&tstamp, &pid);
>  			__sync_fetch_and_add(&data_fail, 1);
>  			return 0;
> @@ -445,6 +447,7 @@ int contention_end(u64 *ctx)
>  				data_map_full = 1;
>  			__sync_fetch_and_add(&data_fail, 1);
>  		}
> +		pelem->lock = 0;
>  		bpf_map_delete_elem(&tstamp, &pid);
>  		return 0;
>  	}
> @@ -458,6 +461,7 @@ int contention_end(u64 *ctx)
>  	if (data->min_time > duration)
>  		data->min_time = duration;
>  
> +	pelem->lock = 0;
>  	bpf_map_delete_elem(&tstamp, &pid);
>  	return 0;
>  }
> -- 
> 2.42.0.582.g8ccd20d70d-goog
> 

-- 

- Arnaldo

Re: [PATCH] perf lock contention: Clear lock addr after use

[Namhyung Kim]

On Fri, Oct 13, 2023 at 4:00 PM Arnaldo Carvalho de Melo
<acme@kernel.org> wrote:
>
> Em Thu, Sep 28, 2023 at 04:50:18PM -0700, Namhyung Kim escreveu:
> > It checks the current lock to calculated the delta of contention time.
>
> > The address is saved in the tstamp map which is allocated at begining of
> > contention and released at end of contention.
> >
> > But it's possible for bpf_map_delete_elem() to fail.  In that case, the
>
> How can it fail?
>
> You do:
>
>         pelem = bpf_map_lookup_elem(&tstamp, &pid);
>         if (!pelem || pelem->lock != ctx[0])
>                 return 0;
>
> So it is there, why would the removal using the same key fail?

It can fail when it doesn't get a lock for the internal bucket.
See kernel/bpf/hashtab.c::htab_map_delete_elem().

But I'm not sure whether that's actually possible in this case.

>
> The patch should work as-is, I'm just curious about what would make
> there removal of a map entry that was successfully looked up on the same
> contention_end prog to fail when being removed...

Now I'm seeing some rare error cases like a spinlock wait
is longer than a minute.  I suspect a bug in this code and
try to be more defensive.

Thanks,
Namhyung