Date:   Mon, 6 Feb 2023 19:33:27 +0100
From:   David Sterba <dsterba@suse.cz>
To:     Greg KH <gregkh@linuxfoundation.org>
Cc:     Kees Cook <keescook@chromium.org>, Chris Mason <clm@fb.com>,
        Josef Bacik <josef@toxicpanda.com>,
        David Sterba <dsterba@suse.com>, linux-btrfs@vger.kernel.org,
        linux-kernel@vger.kernel.org, linux-hardening@vger.kernel.org
Subject: Re: [PATCH] btrfs: sysfs: Handle NULL return values
Message-ID: <20230206183327.GC28288@twin.jikos.cz>
Reply-To: dsterba@suse.cz
References: <20230204183510.never.909-kees@kernel.org>
 <Y99XGrFvXBL32cOO@kroah.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <Y99XGrFvXBL32cOO@kroah.com>
User-Agent: Mutt/1.5.23.1-rc1 (2014-03-12)
Precedence: bulk

On Sun, Feb 05, 2023 at 08:13:30AM +0100, Greg KH wrote:
> On Sat, Feb 04, 2023 at 10:35:10AM -0800, Kees Cook wrote:
> > Each of to_fs_info(), discard_to_fs_info(), and to_space_info() can
> > return NULL values.
> 
> The code says it could, but I really do not think that is possible at
> all, especially based on the fact that there have never been any crashes
> reported here.
> 
> So the NULL returns should just be removed instead, right?

Yes. The access to any sysfs callback always implies an existing
fs_info, the sysfs files are destroyed before the instance of fs_info.
If the callback is slow or stuck for some reason, the removal will block
due to reference counts.

> Also, to_space_info() is a macro of container_of() which can not return
> NULL.
> 
> And get_btrfs_kobj() is just odd, that probably should be fixed up as
> well, that's an indication that something is wrong with the sysfs code
> if anyone has to attempt to walk the whole kobject parent path :(

The get_btrfs_kobj is relatively new and more generic than the other
helpers that basically hardcoded the kobj path relative to the main
sysfs object.

> So while this patch might fix up the compiler warning, it's logically
> not going to change anything, let's fix this properly.  I can look at it
> next week if someone doesn't beat me to it.

Let me have a look, I was meaning to clean up the code to use only
get_btrfs_kobj. So far there was no pressing reason for that but if gcc
13 is going to warn about the checks it needs to be fixed.