| Message ID | 20210520142901.3371299-1-atomlin@redhat.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [v4] mm/page_alloc: bail out on fatal signal during reclaim/compaction retry attempt | expand |
On 5/20/21 4:29 PM, Aaron Tomlin wrote: > A customer experienced a low-memory situation and decided to issue a > SIGKILL (i.e. a fatal signal). Instead of promptly terminating as one > would expect, the aforementioned task remained unresponsive. > > Further investigation indicated that the task was "stuck" in the > reclaim/compaction retry loop. Now, 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 <atomlin@redhat.com> Reviewed-by: Vlastimil Babka <vbabka@suse.cz> > --- > mm/page_alloc.c | 3 +++ > 1 file changed, 3 insertions(+) > > diff --git a/mm/page_alloc.c b/mm/page_alloc.c > index aaa1655cf682..b317057ac186 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)) > + return false; > + > if (compaction_made_progress(compact_result)) > (*compaction_retries)++; > >
On Thu 20-05-21 15:29:01, Aaron Tomlin wrote: > A customer experienced a low-memory situation and decided to issue a > SIGKILL (i.e. a fatal signal). Instead of promptly terminating as one > would expect, the aforementioned task remained unresponsive. > > Further investigation indicated that the task was "stuck" in the > reclaim/compaction retry loop. Now, it does not make sense to retry > compaction when a fatal signal is pending. Is this really true in general? The memory reclaim is retried even when fatal signals are pending. Why should be compaction different? I do agree that retrying way too much is bad but is there any reason why this special case doesn't follow the max retry logic?
On 5/31/21 1:33 PM, Michal Hocko wrote: > On Thu 20-05-21 15:29:01, Aaron Tomlin wrote: >> A customer experienced a low-memory situation and decided to issue a >> SIGKILL (i.e. a fatal signal). Instead of promptly terminating as one >> would expect, the aforementioned task remained unresponsive. >> >> Further investigation indicated that the task was "stuck" in the >> reclaim/compaction retry loop. Now, it does not make sense to retry >> compaction when a fatal signal is pending. > > Is this really true in general? The memory reclaim is retried even when > fatal signals are pending. Why should be compaction different? I do > agree that retrying way too much is bad but is there any reason why this > special case doesn't follow the max retry logic? Compaction doesn't do anything if fatal signal is pending, it bails out immediately and the checks are rather frequent. So why retry?
On Mon 31-05-21 13:35:31, Vlastimil Babka wrote: > On 5/31/21 1:33 PM, Michal Hocko wrote: > > On Thu 20-05-21 15:29:01, Aaron Tomlin wrote: > >> A customer experienced a low-memory situation and decided to issue a > >> SIGKILL (i.e. a fatal signal). Instead of promptly terminating as one > >> would expect, the aforementioned task remained unresponsive. > >> > >> Further investigation indicated that the task was "stuck" in the > >> reclaim/compaction retry loop. Now, it does not make sense to retry > >> compaction when a fatal signal is pending. > > > > Is this really true in general? The memory reclaim is retried even when > > fatal signals are pending. Why should be compaction different? I do > > agree that retrying way too much is bad but is there any reason why this > > special case doesn't follow the max retry logic? > > Compaction doesn't do anything if fatal signal is pending, it bails out > immediately and the checks are rather frequent. So why retry? OK, I was not aware of that and it would be helpful to have that mentioned in the changelog.
diff --git a/mm/page_alloc.c b/mm/page_alloc.c index aaa1655cf682..b317057ac186 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)) + return false; + if (compaction_made_progress(compact_result)) (*compaction_retries)++;
A customer experienced a low-memory situation and decided to issue a SIGKILL (i.e. a fatal signal). Instead of promptly terminating as one would expect, the aforementioned task remained unresponsive. Further investigation indicated that the task was "stuck" in the reclaim/compaction retry loop. Now, 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 <atomlin@redhat.com> --- mm/page_alloc.c | 3 +++ 1 file changed, 3 insertions(+)