From: Romain Francoise Subject: Re: Linux 3.6-rc5 Date: Mon, 10 Sep 2012 19:18:17 +0200 Message-ID: <878vchdcfa.fsf@silenus.orebokech.com> References: <87har7h84h.fsf@silenus.orebokech.com> <20120909191958.GB9782@gondor.apana.org.au> <20120909200910.GA11919@r00tworld.net> <20120909210000.GA10293@gondor.apana.org.au> Mime-Version: 1.0 Content-Type: text/plain Cc: Mathias Krause , Linus Torvalds , Linux Kernel Mailing List , Linux Crypto Mailing List To: Herbert Xu Return-path: Received: from stringer.orebokech.com ([88.190.240.207]:44576 "EHLO stringer.orebokech.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754626Ab2IJRST (ORCPT ); Mon, 10 Sep 2012 13:18:19 -0400 In-Reply-To: <20120909210000.GA10293@gondor.apana.org.au> (Herbert Xu's message of "Sun, 9 Sep 2012 14:00:00 -0700") Sender: linux-crypto-owner@vger.kernel.org List-ID: Herbert Xu writes: > Thanks! I think this patch should fix the problem. Can someone > please confirm this? Works for me as well, thanks! > crypto: authenc - Fix crash with zero-length assoc data > > The authenc code doesn't deal with zero-length associated data > correctly and ends up constructing a zero-length sg entry which > causes a crash when it's fed into the crypto system. > > This patch fixes this by avoiding the code-path that triggers > the SG construction if we have no associated data. > > This isn't the most optimal fix as it means that we'll end up > using the fallback code-path even when we could still execute > the digest function. However, this isn't a big deal as nobody > but the test path would supply zero-length associated data. > > Reported-by: Romain Francoise > Signed-off-by: Herbert Xu Feel free to change this to: Reported-and-tested-by: Romain Francoise