linux-ext4 - Re: [PATCH] ext4: Fix deadlock on page reclaim

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] [day] [month] [year] [list]

Date:   Wed, 31 Jul 2019 04:37:57 +0000
From:   Damien Le Moal <Damien.LeMoal@....com>
To:     Dave Chinner <david@...morbit.com>
CC:     Andreas Dilger <adilger@...ger.ca>,
        "Theodore Y. Ts'o" <tytso@....edu>,
        Christoph Hellwig <hch@...radead.org>,
        "linux-ext4@...r.kernel.org" <linux-ext4@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "linux-fsdevel@...r.kernel.org" <linux-fsdevel@...r.kernel.org>,
        Christoph Hellwig <hch@....de>,
        Johannes Thumshirn <jthumshirn@...e.de>,
        Naohiro Aota <Naohiro.Aota@....com>,
        Masato Suzuki <masato.suzuki@....com>
Subject: Re: [PATCH] ext4: Fix deadlock on page reclaim

Dave,

On 2019/07/31 8:48, Dave Chinner wrote:
> On Tue, Jul 30, 2019 at 02:06:33AM +0000, Damien Le Moal wrote:
>> If we had a pread_nofs()/pwrite_nofs(), that would work. Or we could define a
>> RWF_NORECLAIM flag for pwritev2()/preadv2(). This last one could actually be the
>> cleanest approach.
> 
> Clean, yes, but I'm not sure we want to expose kernel memory reclaim
> capabilities to userspace... It would be misleading, too, because we
> still want to allow reclaim to occur, just not have reclaim recurse
> into other filesystems....

When I wrote RWF_NORECLAIM, I was really thinking of RWF_NOFSRECLAIM. So
suppressing direct reclaim recursing into another FS rather than completely
disabling reclaim. Sorry for the confusion.

Would this be better ? This is still application controlled, so debatable if
control should be given. Most likely this would need to be limited to CAP_SYS
capable user processes (e.g. root processes).

I still need to check on FUSE if anything at all along these lines exists there.
I will dig.

Best regards.

-- 
Damien Le Moal
Western Digital Research