linux-kernel - Re: [PATCH V2] loop: Add sanity check for read/write

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: <aAuOC8djgRrq-Gdj@infradead.org>
Date: Fri, 25 Apr 2025 06:28:43 -0700
From: Christoph Hellwig <hch@...radead.org>
To: Lizhi Xu <lizhi.xu@...driver.com>
Cc: ming.lei@...hat.com, axboe@...nel.dk, linux-block@...r.kernel.org,
	linux-kernel@...r.kernel.org,
	syzbot+6af973a3b8dfd2faefdc@...kaller.appspotmail.com,
	syzkaller-bugs@...glegroups.com
Subject: Re: [PATCH V2] loop: Add sanity check for read/write_iter

On Fri, Apr 25, 2025 at 01:38:03PM +0800, Lizhi Xu wrote:
> Some file systems do not support read_iter or write_iter, such as selinuxfs
> in this issue.
> So before calling them, first confirm that the interface is supported and
> then call it.

Nit: commit messages should not have lines longer than 73 characters.

Please also add a:

Fixes: f2fed441c69b ("loop: stop using vfs_iter__{read,write} for buffered I/O")

and maybe add a blurb that vfs_iter_read/write had this check.

Now the other interesting bit is why we did not hit this earlier with
direct I/O?  I guess it's because we basically have no instances
supporting direct I/O and not using the iter ops.

> @@ -603,6 +603,12 @@ static int loop_change_fd(struct loop_device *lo, struct block_device *bdev,
>  	if (!file)
>  		return -EBADF;
>  
> +	if (unlikely(!file->f_op->read_iter))
> +		return -EINVAL;
> +
> +	if (file->f_mode & FMODE_WRITE && unlikely(!file->f_op->write_iter))
> +		return -EINVAL;

Can we have a common helper for change_fd and configure, please?

Please also drop the unlikelys - this is not a fast path and we don't
need to micro-optimize.

A bit unrelated, but loop-configure actually checks for write_iter
and forces read-only for that.  Do we need the same kind of check in
change_fd?