Re: [PATCH 1/2] mm: thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings

[Vlastimil Babka <vbabka@suse.cz>]

On 10/10/18 12:51 AM, Andrea Arcangeli wrote:
> Yes there's one case where reclaim is "pointless", but it happens once
> and then COMPACT_DEFERRED is returned and __GFP_NORETRY will skip
> reclaim then.
> 
> So you're right when we hit fragmentation there's one and only one
> "pointless" reclaim invocation. And immediately after we also
> exponentially backoff on the compaction invocations with the
> compaction deferred logic.
> 
> We could try optimize away such "pointless" reclaim event for sure,
> but it's probably an optimization that may just get lost in the noise
> and may not be measurable, because it only happens once when the first
> full fragmentation is encountered.

Note there's a small catch in the above. defer_compaction() has always
only been called after a failure on higher priority than
COMPACT_PRIO_ASYNC, where it's assumed that async compaction can
terminate prematurely due to a number of reasons, so it doesn't mean
that the zone itself cannot be compacted.
And, for __GFP_NORETRY, if the initial compaction fails, we keep using
async compaction also for the second, after-reclaim attempt (which would
otherwise use SYNC_LIGHT):

        /*
         * Looks like reclaim/compaction is worth trying, but
         * sync compaction could be very expensive, so keep
         * using async compaction.
         */
        compact_priority = INIT_COMPACT_PRIORITY;

This doesn't affect current madvised THP allocation which doesn't use
__GFP_NORETRY, but could explain why you saw no benefit from changing it
to __GFP_NORETRY.