linux-ext4 - Re: [ext4 io hang] buffered write io hang in balance_dirty

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: <20230429044038.GA7561@lst.de>
Date:   Sat, 29 Apr 2023 06:40:38 +0200
From:   Christoph Hellwig <hch@....de>
To:     Ming Lei <ming.lei@...hat.com>
Cc:     Theodore Ts'o <tytso@....edu>, Baokun Li <libaokun1@...wei.com>,
        Matthew Wilcox <willy@...radead.org>,
        linux-ext4@...r.kernel.org,
        Andreas Dilger <adilger.kernel@...ger.ca>,
        linux-block@...r.kernel.org,
        Andrew Morton <akpm@...ux-foundation.org>,
        linux-fsdevel@...r.kernel.org, linux-mm@...ck.org,
        Dave Chinner <dchinner@...hat.com>,
        Eric Sandeen <sandeen@...hat.com>,
        Christoph Hellwig <hch@....de>, Zhang Yi <yi.zhang@...hat.com>,
        yangerkun <yangerkun@...wei.com>
Subject: Re: [ext4 io hang] buffered write io hang in balance_dirty_pages

On Sat, Apr 29, 2023 at 11:16:14AM +0800, Ming Lei wrote:
> OK, looks both Dave and you have same suggestion, and IMO, it isn't hard to
> add one interface for notifying FS, and it can be either one s_ops->shutdown()
> or shutdown_filesystem(struct super_block *sb).

It's not that simple.  You need to be able to do that for any device used
by a file system, not just s_bdev.  This means it needs go into ops
passed by the bdev owner, which is also needed to propagate this through
stackable devices.

I have some work on that, but the way how blkdev_get is called in the
generic mount helpers is a such a mess that I've not been happy with
the result yet.  Let me see if spending extra time with it will allow
me to come up with something that doesn't suck.

> But the main job should be how this interface is implemented in FS/VFS side,
> so it looks one more FS job, and block layer can call shutdown_filesystem()
> from del_gendisk() simply.

This needs to be called from blk_mark_disk_dead for drivers using that,
and from del_gendisk only if GD_DEAD isn't set yet.