From: Michal Hocko <mhocko@suse.com>
To: Andrea Arcangeli <aarcange@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>,
linux-mm@kvack.org, Alex Williamson <alex.williamson@redhat.com>,
David Rientjes <rientjes@google.com>,
Vlastimil Babka <vbabka@suse.cz>
Subject: Re: [PATCH 2/2] mm: thp: fix transparent_hugepage/defrag = madvise || always
Date: Tue, 21 Aug 2018 13:50:57 +0200 [thread overview]
Message-ID: <20180821115057.GY29735@dhcp22.suse.cz> (raw)
In-Reply-To: <20180820032204.9591-3-aarcange@redhat.com>
On Sun 19-08-18 23:22:04, Andrea Arcangeli wrote:
> qemu uses MADV_HUGEPAGE which allows direct compaction (i.e.
> __GFP_DIRECT_RECLAIM is set).
>
> The problem is that direct compaction combined with the NUMA
> __GFP_THISNODE logic in mempolicy.c is telling reclaim to swap very
> hard the local node, instead of failing the allocation if there's no
> THP available in the local node.
>
> Such logic was ok until __GFP_THISNODE was added to the THP allocation
> path even with MPOL_DEFAULT.
>
> The idea behind the __GFP_THISNODE addition, is that it is better to
> provide local memory in PAGE_SIZE units than to use remote NUMA THP
> backed memory. That largely depends on the remote latency though, on
> threadrippers for example the overhead is relatively low in my
> experience.
>
> The combination of __GFP_THISNODE and __GFP_DIRECT_RECLAIM results in
> extremely slow qemu startup with vfio, if the VM is larger than the
> size of one host NUMA node. This is because it will try very hard to
> unsuccessfully swapout get_user_pages pinned pages as result of the
> __GFP_THISNODE being set, instead of falling back to PAGE_SIZE
> allocations and instead of trying to allocate THP on other nodes (it
> would be even worse without vfio type1 GUP pins of course, except it'd
> be swapping heavily instead).
>
> It's very easy to reproduce this by setting
> transparent_hugepage/defrag to "always", even with a simple memhog.
>
> 1) This can be fixed by retaining the __GFP_THISNODE logic also for
> __GFP_DIRECT_RELCAIM by allowing only one compaction run. Not even
> COMPACT_SKIPPED (i.e. compaction failing because not enough free
> memory in the zone) should be allowed to invoke reclaim.
>
> 2) An alternative is not use __GFP_THISNODE if __GFP_DIRECT_RELCAIM
> has been set by the caller (i.e. MADV_HUGEPAGE or
> defrag="always"). That would keep the NUMA locality restriction
> only when __GFP_DIRECT_RECLAIM is not set by the caller. So THP
> will be provided from remote nodes if available before falling back
> to PAGE_SIZE units in the local node, but an app using defrag =
> always (or madvise with MADV_HUGEPAGE) supposedly prefers that.
So does reverting 5265047ac301 ("mm, thp: really limit transparent
hugepage allocation to local node") help?
I really detest a new gfp flag for one time semantic that is muddy as
hell.
> + * __GFP_ONLY_COMPACT: Only invoke compaction. Do not try to succeed
> + * the allocation by freeing memory. Never risk to free any
> + * "PAGE_SIZE" memory unit even if compaction failed specifically
> + * because of not enough free pages in the zone. This only makes sense
> + * only in combination with __GFP_THISNODE (enforced with a
> + * VM_WARN_ON), to restrict the THP allocation in the local node that
> + * triggered the page fault and fallback into PAGE_SIZE allocations in
> + * the same node. We don't want to invoke reclaim because there may be
> + * plenty of free memory already in the local node. More importantly
> + * there may be even plenty of free THP available in remote nodes so
> + * we should allocate those if something instead of reclaiming any
> + * memory in the local node. Implementation detail: set ___GFP_NORETRY
> + * too so that ___GFP_ONLY_COMPACT only needs to be checked in a slow
> + * path.
This is simply incomprehensible. How can anybody who is not deeply
familiar with the allocator/reclaim internals know when to use it.
If this is really a regression then we should start by pinpointing the
real culprit and go from there. If this is really 5265047ac301 then just
start by reverting it. I strongly suspect there is some mismatch in
expectations here. What others consider acceptable seems to be a problem
for others. I believe that was one of the reasons why we have changed
the default THP direct compaction behavior, no?
--
Michal Hocko
SUSE Labs
next prev parent reply other threads:[~2018-08-21 11:51 UTC|newest]
Thread overview: 61+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-08-20 3:22 [PATCH 0/2] fix for "pathological THP behavior" Andrea Arcangeli
2018-08-20 3:22 ` [PATCH 1/2] mm: thp: consolidate policy_nodemask call Andrea Arcangeli
2018-08-20 3:22 ` [PATCH 2/2] mm: thp: fix transparent_hugepage/defrag = madvise || always Andrea Arcangeli
2018-08-20 3:26 ` [PATCH 0/1] fix for "pathological THP behavior" v2 Andrea Arcangeli
2018-08-20 3:26 ` [PATCH 1/1] mm: thp: fix transparent_hugepage/defrag = madvise || always Andrea Arcangeli
2018-08-20 12:35 ` [PATCH 2/2] " Zi Yan
2018-08-20 15:32 ` Andrea Arcangeli
2018-08-21 11:50 ` Michal Hocko [this message]
2018-08-21 21:40 ` Andrea Arcangeli
2018-08-22 9:02 ` Michal Hocko
2018-08-22 11:07 ` Michal Hocko
2018-08-22 14:24 ` Andrea Arcangeli
2018-08-22 14:45 ` Michal Hocko
2018-08-22 15:24 ` Andrea Arcangeli
2018-08-23 10:50 ` Michal Hocko
2018-08-22 15:52 ` Andrea Arcangeli
2018-08-23 10:52 ` Michal Hocko
2018-08-28 7:53 ` Michal Hocko
2018-08-28 8:18 ` Michal Hocko
2018-08-28 8:54 ` Stefan Priebe - Profihost AG
2018-08-29 11:11 ` Stefan Priebe - Profihost AG
[not found] ` <D5F4A33C-0A37-495C-9468-D6866A862097@cs.rutgers.edu>
2018-08-29 14:28 ` Michal Hocko
2018-08-29 14:35 ` Michal Hocko
2018-08-29 15:22 ` Zi Yan
2018-08-29 15:47 ` Michal Hocko
2018-08-29 16:06 ` Zi Yan
2018-08-29 16:25 ` Michal Hocko
2018-08-29 19:24 ` [PATCH] mm, thp: relax __GFP_THISNODE for MADV_HUGEPAGE mappings Michal Hocko
2018-08-29 22:54 ` Zi Yan
2018-08-30 7:00 ` Michal Hocko
2018-08-30 13:22 ` Zi Yan
2018-08-30 13:45 ` Michal Hocko
2018-08-30 14:02 ` Zi Yan
2018-08-30 16:19 ` Stefan Priebe - Profihost AG
2018-08-30 16:40 ` Michal Hocko
2018-09-05 3:44 ` Andrea Arcangeli
2018-09-05 7:08 ` Michal Hocko
2018-09-06 11:10 ` Vlastimil Babka
2018-09-06 11:16 ` Vlastimil Babka
2018-09-06 11:25 ` Michal Hocko
2018-09-06 12:35 ` Zi Yan
2018-09-06 10:59 ` Vlastimil Babka
2018-09-06 11:17 ` Zi Yan
2018-08-30 6:47 ` Michal Hocko
2018-09-06 11:18 ` Vlastimil Babka
2018-09-06 11:27 ` Michal Hocko
2018-09-12 17:29 ` Mel Gorman
2018-09-17 6:11 ` Michal Hocko
2018-09-17 7:04 ` Stefan Priebe - Profihost AG
2018-09-17 9:32 ` Stefan Priebe - Profihost AG
2018-09-17 11:27 ` Michal Hocko
2018-08-20 11:58 ` [PATCH 0/2] fix for "pathological THP behavior" Kirill A. Shutemov
2018-08-20 15:19 ` Andrea Arcangeli
2018-08-21 15:30 ` Vlastimil Babka
2018-08-21 17:26 ` David Rientjes
2018-08-21 22:18 ` Andrea Arcangeli
2018-08-21 22:05 ` Andrea Arcangeli
2018-08-22 9:24 ` Michal Hocko
2018-08-22 15:56 ` Andrea Arcangeli
2018-08-20 19:06 ` Yang Shi
2018-08-20 23:24 ` Andrea Arcangeli
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20180821115057.GY29735@dhcp22.suse.cz \
--to=mhocko@suse.com \
--cc=aarcange@redhat.com \
--cc=akpm@linux-foundation.org \
--cc=alex.williamson@redhat.com \
--cc=linux-mm@kvack.org \
--cc=rientjes@google.com \
--cc=vbabka@suse.cz \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).