linux-kernel - Re: mm: mkfs.ext4 invoked oom-killer on i386

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 17 Jun 2020 18:06:24 +0200
From:   Michal Hocko <mhocko@...nel.org>
To:     Naresh Kamboju <naresh.kamboju@...aro.org>
Cc:     Chris Down <chris@...isdown.name>,
        Yafang Shao <laoar.shao@...il.com>,
        Anders Roxell <anders.roxell@...aro.org>,
        "Linux F2FS DEV, Mailing List" 
        <linux-f2fs-devel@...ts.sourceforge.net>,
        linux-ext4 <linux-ext4@...r.kernel.org>,
        linux-block <linux-block@...r.kernel.org>,
        Andrew Morton <akpm@...ux-foundation.org>,
        open list <linux-kernel@...r.kernel.org>,
        Linux-Next Mailing List <linux-next@...r.kernel.org>,
        linux-mm <linux-mm@...ck.org>, Arnd Bergmann <arnd@...db.de>,
        Andreas Dilger <adilger.kernel@...ger.ca>,
        Jaegeuk Kim <jaegeuk@...nel.org>,
        Theodore Ts'o <tytso@....edu>, Chao Yu <chao@...nel.org>,
        Hugh Dickins <hughd@...gle.com>,
        Andrea Arcangeli <aarcange@...hat.com>,
        Matthew Wilcox <willy@...radead.org>,
        Chao Yu <yuchao0@...wei.com>, lkft-triage@...ts.linaro.org,
        Johannes Weiner <hannes@...xchg.org>,
        Roman Gushchin <guro@...com>, Cgroups <cgroups@...r.kernel.org>
Subject: Re: mm: mkfs.ext4 invoked oom-killer on i386 - pagecache_get_page

On Wed 17-06-20 21:23:05, Naresh Kamboju wrote:
> On Wed, 17 Jun 2020 at 19:41, Michal Hocko <mhocko@...nel.org> wrote:
> >
> > [Our emails have crossed]
> >
> > On Wed 17-06-20 14:57:58, Chris Down wrote:
> > > Naresh Kamboju writes:
> > > > mkfs -t ext4 /dev/disk/by-id/ata-TOSHIBA_MG04ACA100N_Y8RQK14KF6XF
> > > > mke2fs 1.43.8 (1-Jan-2018)
> > > > Creating filesystem with 244190646 4k blocks and 61054976 inodes
> > > > Filesystem UUID: 7c380766-0ed8-41ba-a0de-3c08e78f1891
> > > > Superblock backups stored on blocks:
> > > > 32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632, 2654208,
> > > > 4096000, 7962624, 11239424, 20480000, 23887872, 71663616, 78675968,
> > > > 102400000, 214990848
> > > > Allocating group tables:    0/7453 done
> > > > Writing inode tables:    0/7453 done
> > > > Creating journal (262144 blocks): [   51.544525] under min:0 emin:0
> > > > [   51.845304] under min:0 emin:0
> > > > [   51.848738] under min:0 emin:0
> > > > [   51.858147] under min:0 emin:0
> > > > [   51.861333] under min:0 emin:0
> > > > [   51.862034] under min:0 emin:0
> > > > [   51.862442] under min:0 emin:0
> > > > [   51.862763] under min:0 emin:0
> > >
> > > Thanks, this helps a lot. Somehow we're entering mem_cgroup_below_min even
> > > when min/emin is 0 (which should indeed be the case if you haven't set them
> > > in the hierarchy).
> > >
> > > My guess is that page_counter_read(&memcg->memory) is 0, which means
> > > mem_cgroup_below_min will return 1.
> >
> > Yes this is the case because this is likely the root memcg which skips
> > all charges.
> >
> > > However, I don't know for sure why that should then result in the OOM killer
> > > coming along. My guess is that since this memcg has 0 pages to scan anyway,
> > > we enter premature OOM under some conditions. I don't know why we wouldn't
> > > have hit that with the old version of mem_cgroup_protected that returned
> > > MEMCG_PROT_* members, though.
> >
> > Not really. There is likely no other memcg to reclaim from and assuming
> > min limit protection will result in no reclaimable memory and thus the
> > OOM killer.
> >
> > > Can you please try the patch with the `>=` checks in mem_cgroup_below_min
> > > and mem_cgroup_below_low changed to `>`? If that fixes it, then that gives a
> > > strong hint about what's going on here.
> >
> > This would work but I believe an explicit check for the root memcg would
> > be easier to spot the reasoning.
> 
> May I request you to send debugging or proposed fix patches here.
> I am happy to do more testing.

Sure, here is the diff to test.

diff --git a/include/linux/memcontrol.h b/include/linux/memcontrol.h
index c74a8f2323f1..6b5a31672fbe 100644
--- a/include/linux/memcontrol.h
+++ b/include/linux/memcontrol.h
@@ -392,6 +392,13 @@ static inline bool mem_cgroup_below_low(struct mem_cgroup *memcg)
 	if (mem_cgroup_disabled())
 		return false;
 
+	/*
+	 * Root memcg doesn't account charges and doesn't support
+	 * protection
+	 */
+	if (mem_cgroup_is_root(memcg))
+		return false;
+
 	return READ_ONCE(memcg->memory.elow) >=
 		page_counter_read(&memcg->memory);
 }
@@ -401,6 +408,13 @@ static inline bool mem_cgroup_below_min(struct mem_cgroup *memcg)
 	if (mem_cgroup_disabled())
 		return false;
 
+	/*
+	 * Root memcg doesn't account charges and doesn't support
+	 * protection
+	 */
+	if (mem_cgroup_is_root(memcg))
+		return false;
+
 	return READ_ONCE(memcg->memory.emin) >=
 		page_counter_read(&memcg->memory);
 }
-- 
Michal Hocko
SUSE Labs