linux-kernel - Re: Memory hotplug softlock issue

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]

Message-ID: <20181115143412.GS2653@MiWiFi-R3L-srv>
Date:   Thu, 15 Nov 2018 22:34:12 +0800
From:   Baoquan He <bhe@...hat.com>
To:     Michal Hocko <mhocko@...nel.org>
Cc:     David Hildenbrand <david@...hat.com>, linux-mm@...ck.org,
        pifang@...hat.com, linux-kernel@...r.kernel.org,
        akpm@...ux-foundation.org, aarcange@...hat.com
Subject: Re: Memory hotplug softlock issue

On 11/15/18 at 03:32pm, Michal Hocko wrote:
> On Thu 15-11-18 21:38:40, Baoquan He wrote:
> > On 11/15/18 at 02:19pm, Michal Hocko wrote:
> > > On Thu 15-11-18 21:12:11, Baoquan He wrote:
> > > > On 11/15/18 at 09:30am, Michal Hocko wrote:
> > > [...]
> > > > > It would be also good to find out whether this is fs specific. E.g. does
> > > > > it make any difference if you use a different one for your stress
> > > > > testing?
> > > > 
> > > > Created a ramdisk and put stress bin there, then run stress -m 200, now
> > > > seems it's stuck in libc-2.28.so migrating. And it's still xfs. So now xfs
> > > > is a big suspect. At bottom I paste numactl printing, you can see that it's
> > > > the last 4G.
> > > > 
> > > > Seems it's trying to migrate libc-2.28.so, but stress program keeps trying to
> > > > access and activate it.
> > > 
> > > Is this still with faultaround disabled? I have seen exactly same
> > > pattern in the bug I am working on. It was ext4 though.
> > 
> > After a long time struggling, the last 2nd block where libc-2.28.so is
> > located is reclaimed, now it comes to the last memory block, still
> > stress program itself. swap migration entry has been made and trying to
> > unmap, now it's looping there.
> > 
> > [  +0.004445] migrating pfn 190ff2bb0 failed 
> > [  +0.000013] page:ffffea643fcaec00 count:203 mapcount:201 mapping:ffff888dfb268f48 index:0x0
> > [  +0.012809] shmem_aops 
> > [  +0.000011] name:"stress" 
> > [  +0.002550] flags: 0x1dfffffc008004e(referenced|uptodate|dirty|workingset|swapbacked)
> > [  +0.010715] raw: 01dfffffc008004e ffffea643fcaec48 ffffea643fc714c8 ffff888dfb268f48
> > [  +0.007828] raw: 0000000000000000 0000000000000000 000000cb000000c8 ffff888e72e92000
> > [  +0.007810] page->mem_cgroup:ffff888e72e92000
> [...]
> > [  +0.004455] migrating pfn 190ff2bb0 failed 
> > [  +0.000018] page:ffffea643fcaec00 count:203 mapcount:201 mapping:ffff888dfb268f48 index:0x0
> > [  +0.014392] shmem_aops 
> > [  +0.000010] name:"stress" 
> > [  +0.002565] flags: 0x1dfffffc008004e(referenced|uptodate|dirty|workingset|swapbacked)
> > [  +0.010675] raw: 01dfffffc008004e ffffea643fcaec48 ffffea643fc714c8 ffff888dfb268f48
> > [  +0.007819] raw: 0000000000000000 0000000000000000 000000cb000000c8 ffff888e72e92000
> > [  +0.007808] page->mem_cgroup:ffff888e72e92000
> 
> OK, so this is tmpfs backed code of your stree test. This just tells us
> that this is not fs specific. Reference count is 2 more than the map
> count which is the expected state. So the reference count must have been
> elevated at the time when the migration was attempted. Shmem supports
> fault around so this might be still possible (assuming it is enabled).
> If not we really need to dig deeper. I will think of a debugging patch.

Yes, faultaround is enabled. Will disable it and test again. Will report
test result.