Re: [PATCH 07/12] zloop: take freeze protection when recording write pointers
From: Christoph Hellwig
Date: Fri Nov 20 2026 - 10:44:27 EST
On Mon, Nov 16, 2026 at 12:57:09PM +0100, Jan Kara wrote:
> On Mon 16-11-26 10:41:49, Christian Brauner wrote:
> > Fwiw, the check can't go into __vfs_setxattr(), __vfs_setxattr_noperm()
> > or __vfs_removexattr(). cap_inode_killpriv(), IMA and EVM call those
> > from ->write_iter() via file_remove_privs() and kiocb_start_write()
> > hides the freeze protection from lockdep for aio and io_uring. I have
> > not bothered to figure out why. Maybe you already know? In any case,
> > until that's fixed we can't place it there which means this doesn't
> > catch IMA/EVM at which point this patch becomes pointless.
>
> It's because of AIO and io_uring - if you don't hide the protection from
> lockdep, it will complain about it when returning to userspace while IO is
> still in flight (and thus protection is held). Sadly lockdep does not
> support releasing resources by anybody else than the process which acquired
> it so we fool it in this way...
Can we try to escalate this to the lockdep folks instead of piling hacks
deeper and deeper?
>
> Honza
> --
> Jan Kara <jack@xxxxxxxx>
> SUSE Labs, CR
---end quoted text---