2021-05-19 20:22:38

by Aaron Tomlin

[permalink] [raw]
Subject: [PATCH v2] mm/page_alloc: bail out on fatal signal during reclaim/compaction retry attempt

It does not make sense to retry compaction when a fatal signal is
pending.

In the context of try_to_compact_pages(), indeed COMPACT_SKIPPED can be
returned; albeit, not every zone, on the zone list, would be considered
in the case a fatal signal is found to be pending.
Yet, in should_compact_retry(), given the last known compaction result,
each zone, on the zone list, can be considered/or checked
(see compaction_zonelist_suitable()). For example, if a zone was found
to succeed, then reclaim/compaction would be tried again
(notwithstanding the above).

This patch ensures that compaction is not needlessly retried
irrespective of the last known compaction result e.g. if it was skipped,
in the unlikely case a fatal signal is found pending.
So, OOM is at least attempted.

Signed-off-by: Aaron Tomlin <[email protected]>
---
mm/page_alloc.c | 3 +++
1 file changed, 3 insertions(+)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index aaa1655cf682..49f416ffb54f 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -4252,6 +4252,9 @@ should_compact_retry(struct alloc_context *ac, int order, int alloc_flags,
if (!order)
return false;

+ if (fatal_signal_pending(current))
+ goto out;
+
if (compaction_made_progress(compact_result))
(*compaction_retries)++;

--
2.26.3



2021-05-19 20:24:46

by Matthew Wilcox

[permalink] [raw]
Subject: Re: [PATCH v2] mm/page_alloc: bail out on fatal signal during reclaim/compaction retry attempt

On Wed, May 19, 2021 at 08:23:21PM +0100, Aaron Tomlin wrote:
> +++ b/mm/page_alloc.c
> @@ -4252,6 +4252,9 @@ should_compact_retry(struct alloc_context *ac, int order, int alloc_flags,
> if (!order)
> return false;
>
> + if (fatal_signal_pending(current))
> + goto out;

I think 'goto out' will be confusing. It'll output a tracepoint, which
isn't going to record that a fatal signal is pending, so it'll cause
some head scratching for someone looking through the traces. I
think we should just return false here and skip the tracepoint.

But I'd defer to someone like Vlastimil or Michal who know this code far
better than I do.