From mboxrd@z Thu Jan 1 00:00:00 1970 From: Pablo Neira Ayuso Subject: Re: [PATCH] mnl_attr_get_str: document that the result is not NUL-terminated Date: Tue, 14 Aug 2012 13:59:35 +0200 Message-ID: <20120814115935.GA28582@1984> References: <502A2624.6050701@redhat.com> <20120814111008.GA25235@1984> <502A3668.5010404@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: netfilter-devel@vger.kernel.org To: Florian Weimer Return-path: Received: from mail.us.es ([193.147.175.20]:57605 "EHLO mail.us.es" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751905Ab2HNL7j (ORCPT ); Tue, 14 Aug 2012 07:59:39 -0400 Content-Disposition: inline In-Reply-To: <502A3668.5010404@redhat.com> Sender: netfilter-devel-owner@vger.kernel.org List-ID: On Tue, Aug 14, 2012 at 01:28:40PM +0200, Florian Weimer wrote: > On 08/14/2012 01:10 PM, Pablo Neira Ayuso wrote: > >On Tue, Aug 14, 2012 at 12:19:16PM +0200, Florian Weimer wrote: > >>Hi, > >> > >>while reviewing from libmnl-using code, I discovered that the result > >>of mnl_attr_get_str was used as a NUL-terminated string, although in > >>reality the string wasn't. I think this should be mentioned in the > >>documentation. > > > >Looking at the kernel: > > > >static inline int nla_put_string(struct sk_buff *skb, int attrtype, > > const char *str) > >{ > > return nla_put(skb, attrtype, strlen(str) + 1, str); > >} > > > >It seems it always returns the string plus one byte (that is assumed > >to be NULL). > > > >Are you looking at any netlink subsystem in particular? What are you > >noticing? > > Here is what happens: The kernel sends an attribute of type > NLA_NUL_STRING. The application checks it with > mnl_attr_validate(attr, MNL_TYPE_STRING). The application then > proceeds to call mnl_attr_get_str(attr) and uses the result as if it > were NUL-terminated. So if the Netlink packet actually comes from > the kernel, the application code is correct in the sense that it > works with current libmnl. If the packet does not come from the > kernel, the application is incorrect. > > Based on your comments, I think the application should use > mnl_attr_validate(attr, MNL_TYPE_NUL_STRING) instead. In this > light, the documentation for mnl_attr_get_str should probably > suggest to validate with MNL_TYPE_NUL_STRING before calling > mnl_attr_get_str, and my original patch is wrong. That makes sense. MNL_TYPE_NUL_STRING should be used if possible. Still, IIRC, old kernels used to send strings without guaranteeing NULL-termination of strings. So validating this with MNL_TYPE_NUL_STRING may result in problems if your application needs to work with an old kernel.