linux-ext4 - Re: [PATCH v4 1/6] fs: enhance and rename shutdown() callback to remove

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: <5xno4s25lsd2sqq6judn7moorgy2h3konejgassnzlccfa6jsf@ez6ciofy3bwp>
Date: Thu, 10 Jul 2025 14:41:18 -0400
From: Kent Overstreet <kent.overstreet@...ux.dev>
To: Jan Kara <jack@...e.cz>
Cc: Dave Chinner <david@...morbit.com>, 
	Christian Brauner <brauner@...nel.org>, "Darrick J. Wong" <djwong@...nel.org>, 
	Qu Wenruo <quwenruo.btrfs@....com>, Qu Wenruo <wqu@...e.com>, linux-btrfs@...r.kernel.org, 
	linux-fsdevel@...r.kernel.org, viro@...iv.linux.org.uk, linux-ext4@...r.kernel.org, 
	linux-f2fs-devel@...ts.sourceforge.net, ntfs3@...ts.linux.dev, linux-xfs@...r.kernel.org, 
	linux-bcachefs@...r.kernel.org
Subject: Re: [PATCH v4 1/6] fs: enhance and rename shutdown() callback to
 remove_bdev()

On Thu, Jul 10, 2025 at 03:10:04PM +0200, Jan Kara wrote:
> On Wed 09-07-25 13:49:12, Kent Overstreet wrote:
> > On Wed, Jul 09, 2025 at 07:23:07PM +0200, Jan Kara wrote:
> > > > It also avoids the problem of ->mark_dead events being generated
> > > > from a context that holds filesystem/vfs locks and then deadlocking
> > > > waiting for those locks to be released.
> > > > 
> > > > IOWs, a multi-device filesystem should really be implementing
> > > > ->mark_dead itself, and should not be depending on being able to
> > > > lock the superblock to take an active reference to it.
> > > > 
> > > > It should be pretty clear that these are not issues that the generic
> > > > filesystem ->mark_dead implementation should be trying to
> > > > handle.....
> > > 
> > > Well, IMO every fs implementation needs to do the bdev -> sb transition and
> > > make sb somehow stable. It may be that grabbing s_umount and active sb
> > > reference is not what everybody wants but AFAIU btrfs as the second
> > > multi-device filesystem would be fine with that and for bcachefs this
> > > doesn't work only because they have special superblock instantiation
> > > behavior on mount for independent reasons (i.e., not because active ref
> > > + s_umount would be problematic for them) if I understand Kent right.
> > > So I'm still not fully convinced each multi-device filesystem should be
> > > shipping their special method to get from device to stable sb reference.
> > 
> > Honestly, the sync_filesystem() call seems bogus.
> > 
> > If the block device is truly dead, what's it going to accomplish?
> 
> Notice that fs_bdev_mark_dead() calls sync_filesystem() only in case
> 'surprise' argument is false - meaning this is actually a notification
> *before* the device is going away. I.e., graceful device hot unplug when
> you can access the device to clean up as much as possible.

That doesn't seem to be hooked up to anything?

blk_mark_disk_dead() -> blk_report_disk_dead(), surprise is always true

disk_force_media_change(), same

The only call where it's falso is in s390 code. If we know that a disk
is going away, that would be a userspace thing, and they can just
unmount.