Re: [syzbot] [f2fs?] WARNING in rcu_sync_dtor

[Dave Chinner <david@fromorbit.com>]

On Mon, Jul 29, 2024 at 03:27:21PM +0200, Jan Kara wrote:
> Also as the "filesystem shutdown" is spreading across multiple filesystems,
> I'm playing with the idea that maybe we could lift a flag like this to VFS
> so that we can check it in VFS paths and abort some operations early. 

I've been thinking the same thing since I saw what CIFS was doing a
couple of days ago with shutdowns. It's basically just stopping all
new incoming modification operations if the flag is set. i.e. it's
just a check in each filesystem method, and I suspect that many
other filesystems that support shutdown do the same thing.

It looks like exactly the same implementation as CIFS is about to be
added to exfat - stop all the incoming methods and check in the
writeback method - so having a generic superblock flag and generic
checks before calling into filesystem methods would make it real
easy for all filesystems to have basic ->shutdown support for when
block devices go away suddenly.

I also think that we should be lifting *IOC_SHUTDOWN to the VFS -
the same ioctl is now implemented in 4-5 filesystems and they
largely do the same thing - just set a bit in the internal
filesystem superblock flags. Yes, filesystems like XFS and ext4 do
special stuff with journals, but the generic VFS implemenation could
call the filesystem ->shutdown method to do that....

> But
> so far I'm not convinced the gain is worth the need to iron out various
> subtle semantical differences of "shutdown" among filesystems.

I don't think we need to change how any filesystem behaves when it
is shut down. As long as filesystems follow at least the "no new
modifications when shutdown" behaviour, filesystems can implement
internal shutdown however they want...

-Dave.
-- 
Dave Chinner
david@fromorbit.com