Received-SPF: pass (google.com: domain of linux-ext4-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Tue, 22 Nov 2022 03:40:31 +0000
From:   Matthew Wilcox <willy@infradead.org>
To:     Andrew Morton <akpm@linux-foundation.org>
Cc:     Alexander Potapenko <glider@google.com>,
        linux-kernel@vger.kernel.org, tytso@mit.edu,
        adilger.kernel@dilger.ca, jaegeuk@kernel.org, chao@kernel.org,
        linux-fsdevel@vger.kernel.org, linux-ext4@vger.kernel.org,
        linux-f2fs-devel@lists.sourceforge.net,
        Eric Biggers <ebiggers@kernel.org>,
        syzbot+9767be679ef5016b6082@syzkaller.appspotmail.com
Subject: Re: [PATCH 1/5] fs: ext4: initialize fsdata in pagecache_write()
Message-ID: <Y3xEr3hhbYfdei+k@casper.infradead.org>
References: <20221121112134.407362-1-glider@google.com>
 <20221121114840.c407626c13706ff993efabe3@linux-foundation.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20221121114840.c407626c13706ff993efabe3@linux-foundation.org>
Precedence: bulk

On Mon, Nov 21, 2022 at 11:48:40AM -0800, Andrew Morton wrote:
> On Mon, 21 Nov 2022 12:21:30 +0100 Alexander Potapenko <glider@google.com> wrote:
> 
> > When aops->write_begin() does not initialize fsdata, KMSAN reports
> > an error passing the latter to aops->write_end().
> > 
> > Fix this by unconditionally initializing fsdata.
> > 
> > ...
> >
> 
> I'm assuming that this is not-a-bug, and that these changes are purely
> workarounds for a KMSAN shortcoming?

It's a weird one.  It used to be not-a-bug.  Then we changed from
std=gnu99 to std=gnu11 or something.  And in the intervening years,
the C standards ctte decided that passing an uninitialised pointer to a
function was UB.  So we start by passing a pointer to the pointer to
->write_begin().  Some ->write_begin functions initialise that pointer;
others don't.  Then we pass the pointer directly to ->write_end.  If
->write_begin initialised the pointer, that's fine, and if not, it's UB.
Of course the ->write_end doesn't use it if the ->write_begin didn't
initialise it, but it's too late because merely calling the function
was UB.  Thanks, Itanium!