From: Hannes Frederic Sowa Subject: [PATCH net-next] ipv6: revert misguided compiler warning fix on ipv6_exthdrs_len Date: Sun, 15 Dec 2013 18:18:06 +0100 Message-ID: <20131215171806.GA1715@order.stressinduktion.org> References: <1386824025-27413-1-git-send-email-hkchu@google.com> <20131214062929.GA23563@order.stressinduktion.org> <20131214.020211.969354536190248209.davem@davemloft.net> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: David Miller , Eric Dumazet , Herbert Xu , Or Gerlitz , Ben Hutchings , "netdev@vger.kernel.org" To: Jerry Chu Return-path: Received: from order.stressinduktion.org ([87.106.68.36]:57075 "EHLO order.stressinduktion.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751577Ab3LORSJ (ORCPT ); Sun, 15 Dec 2013 12:18:09 -0500 Content-Disposition: inline In-Reply-To: Sender: netdev-owner@vger.kernel.org List-ID: This essentlally reverts commit f52d81dc27c3456c702e83183035142c222acdc= 7 ("ipv6: fix compiler warning in ipv6_exthdrs_len") and silences the warning with uninitialized_var. Cc: Jerry Chu Signed-off-by: Hannes Frederic Sowa --- On Mon, Dec 16, 2013 at 12:20:22AM +0800, Jerry Chu wrote: > On Sat, Dec 14, 2013 at 3:02 PM, David Miller w= rote: > > From: Hannes Frederic Sowa > > Date: Sat, 14 Dec 2013 07:29:29 +0100 > > > >> Commit 299603e8370a93dd5d8e8d800f0dff1ce2c53d36 ("net-gro: Prepare= GRO > >> stack for the upcoming tunneling support") used an uninitialized v= ariable > >> which leads to the following compiler warning: > >> > >> net/ipv6/ip6_offload.c: In function =E2=80=98ipv6_gro_complete=E2=80= =99: > >> net/ipv6/ip6_offload.c:178:24: warning: =E2=80=98optlen=E2=80=99 m= ay be used uninitialized in this function [-Wmaybe-uninitialized] > >> opth =3D (void *)opth + optlen; > >> ^ > >> net/ipv6/ip6_offload.c:164:22: note: =E2=80=98optlen=E2=80=99 was = declared here > >> int len =3D 0, proto, optlen; > >> ^ > >> Fix it up. > >> > >> Cc: Jerry Chu > >> Signed-off-by: Hannes Frederic Sowa > > > > Applied. >=20 > Please back out the above "fix" - it actually breaks the original cod= e because > the fix will cause exthdrs_len to be one optlen short. > > There was no bug in the original code, just the compiler warning (and= I do not > get any warning, although I admit I did not run with the -W flag). Sorry, I see now, of course. Btw. I didn't use any special flags, just a normal kernel build with gcc 4.8.2. > In any case I will submit a revision shortly that will silent any > warning if exists. I hope you are ok with this. Sorry, Hannes net/ipv6/ip6_offload.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) diff --git a/net/ipv6/ip6_offload.c b/net/ipv6/ip6_offload.c index 08861f1..0a578fc 100644 --- a/net/ipv6/ip6_offload.c +++ b/net/ipv6/ip6_offload.c @@ -161,7 +161,7 @@ static int ipv6_exthdrs_len(struct ipv6hdr *iph, const struct net_offload **opps) { struct ipv6_opt_hdr *opth =3D NULL; - int len =3D 0, optlen =3D 0, proto; + int len =3D 0, proto, uninitialized_var(optlen); =20 proto =3D iph->nexthdr; for (;;) { @@ -172,12 +172,11 @@ static int ipv6_exthdrs_len(struct ipv6hdr *iph, if (!((*opps)->flags & INET6_PROTO_GSO_EXTHDR)) break; } - if (opth =3D=3D NULL) { + if (opth =3D=3D NULL) opth =3D (void *)(iph+1); - } else { - optlen =3D ipv6_optlen(opth); + else opth =3D (void *)opth + optlen; - } + optlen =3D ipv6_optlen(opth); len +=3D optlen; proto =3D opth->nexthdr; } --=20 1.8.3.1