linux-ext4 - Re: [PATCH] jbd2: Fix dbench4 performance regression for 'nobarrier' mounts

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 for Android: free password hash cracker in your pocket

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

Message-ID: <20170428150324.GA20122@infradead.org>
Date:   Fri, 28 Apr 2017 08:03:24 -0700
From:   Christoph Hellwig <hch@...radead.org>
To:     Jan Kara <jack@...e.cz>
Cc:     Ted Tso <tytso@....edu>, linux-ext4@...r.kernel.org,
        Christoph Hellwig <hch@...radead.org>, stable@...r.kernel.org
Subject: Re: [PATCH] jbd2: Fix dbench4 performance regression for 'nobarrier'
 mounts

On Fri, Apr 28, 2017 at 11:59:34AM +0200, Jan Kara wrote:
> Fix the problem by making sure journal superblock writes are always
> treated as synchronous since they generally block progress of the
> journalling machinery and thus the whole filesystem.

The callchains leading down to jbd2_write_superblock looks a little
suspicious to me.  It seems like jbd2_journal_commit_transaction
will actually call without FUA in the JBD2_FLUSHED case. Is that
really intentional, and if yes should it be documented?

Except for that it would seem more useful to move to a "bool preflush"
argument passed down.

But I guess we'll need a quick fix first, for that:

Reviewed-by: Christoph Hellwig <hch@....de>