Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Wed, 31 Mar 2021 05:53:57 +0100
From:   Matthew Wilcox <willy@infradead.org>
To:     Andrew Morton <akpm@linux-foundation.org>
Cc:     "Gustavo A. R. Silva" <gustavoars@kernel.org>,
        linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org,
        linux-hardening@vger.kernel.org
Subject: Re: [PATCH][next] hfsplus: Fix out-of-bounds warnings in
 __hfsplus_setxattr
Message-ID: <20210331045357.GV351017@casper.infradead.org>
References: <20210330145226.GA207011@embeddedor>
 <20210330214320.93600506530f1ab18338b467@linux-foundation.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20210330214320.93600506530f1ab18338b467@linux-foundation.org>
Precedence: bulk

On Tue, Mar 30, 2021 at 09:43:20PM -0700, Andrew Morton wrote:
> On Tue, 30 Mar 2021 09:52:26 -0500 "Gustavo A. R. Silva" <gustavoars@kernel.org> wrote:
> 
> > Fix the following out-of-bounds warnings by enclosing
> > structure members file and finder into new struct info:
> > 
> > fs/hfsplus/xattr.c:300:5: warning: 'memcpy' offset [65, 80] from the object at 'entry' is out of the bounds of referenced subobject 'user_info' with type 'struct DInfo' at offset 48 [-Warray-bounds]
> > fs/hfsplus/xattr.c:313:5: warning: 'memcpy' offset [65, 80] from the object at 'entry' is out of the bounds of referenced subobject 'user_info' with type 'struct FInfo' at offset 48 [-Warray-bounds]
> > 
> > Refactor the code by making it more "structured."
> > 
> > Also, this helps with the ongoing efforts to enable -Warray-bounds and
> > makes the code clearer and avoid confusing the compiler.
> 
> Confused.  What was wrong with the old code?  Was this warning
> legitimate and if so, why?  Or is this patch a workaround for a
> compiler shortcoming?

The offending line is this:

-                               memcpy(&entry.file.user_info, value,
+                               memcpy(&entry.file.info, value,
                                                file_finderinfo_len);

what it's trying to do is copy two structs which are adjacent to each
other in a single call to memcpy().  gcc legitimately complains that the
memcpy to this struct overruns the bounds of the struct.  What Gustavo
has done here is introduce a new struct that contains the two structs,
and now gcc is happy that the memcpy doesn't overrun the length of this
containing struct.