netdev - Re: [PATCH net v4] xen-netback: fix fragment detection in checksum setup

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20131203151328.GB728@zion.uk.xensource.com>
Date:	Tue, 3 Dec 2013 15:13:29 +0000
From:	Wei Liu <wei.liu2@...rix.com>
To:	Paul Durrant <Paul.Durrant@...rix.com>
CC:	Wei Liu <wei.liu2@...rix.com>,
	"xen-devel@...ts.xen.org" <xen-devel@...ts.xen.org>,
	"netdev@...r.kernel.org" <netdev@...r.kernel.org>,
	Zoltan Kiss <zoltan.kiss@...rix.com>,
	Ian Campbell <Ian.Campbell@...rix.com>,
	David Vrabel <david.vrabel@...rix.com>,
	David Miller <davem@...emloft.net>
Subject: Re: [PATCH net v4] xen-netback: fix fragment detection in checksum
 setup

On Tue, Dec 03, 2013 at 03:05:40PM +0000, Paul Durrant wrote:
> > -----Original Message-----
> > From: Wei Liu [mailto:wei.liu2@...rix.com]
> > Sent: 03 December 2013 14:58
> > To: Paul Durrant
> > Cc: Wei Liu; xen-devel@...ts.xen.org; netdev@...r.kernel.org; Zoltan Kiss;
> > Ian Campbell; David Vrabel; David Miller
> > Subject: Re: [PATCH net v4] xen-netback: fix fragment detection in checksum
> > setup
> > 
> > On Tue, Dec 03, 2013 at 02:34:56PM +0000, Paul Durrant wrote:
> > > > -----Original Message-----
> > > > From: Wei Liu [mailto:wei.liu2@...rix.com]
> > > > Sent: 03 December 2013 14:29
> > > > To: Paul Durrant
> > > > Cc: Wei Liu; xen-devel@...ts.xen.org; netdev@...r.kernel.org; Zoltan
> > Kiss;
> > > > Ian Campbell; David Vrabel; David Miller
> > > > Subject: Re: [PATCH net v4] xen-netback: fix fragment detection in
> > checksum
> > > > setup
> > > >
> > > > On Tue, Dec 03, 2013 at 02:05:17PM +0000, Paul Durrant wrote:
> > > > [...]
> > > > > > >
> > > > > > > -	header_size = skb->network_header + off +
> > MAX_IPOPTLEN;
> > > > > > > -	maybe_pull_tail(skb, header_size);
> > > > > > > +	if (!maybe_pull_tail(skb, sizeof(struct iphdr),
> > MAX_IP_HDR_LEN))
> > > > > > > +		goto out;
> > > > > > > +
> > > > > >
> > > > > > I think you need to correctly update err to reflect this failure.
> > > > > > Using -EPROTO will wrongly blame frontend while it is backend that's
> > > > > > failing to process the packet.
> > > > > >
> > > > >
> > > > > But a failure should only occur if the packet is malformed, so that would
> > be
> > > > a frontend error wouldn't it?
> > > > >
> > > >
> > > > __pskb_pull_tail may fail due to malloc failure.
> > > >
> > > > However the return value of __pskb_pull_tail cannot reflect the wether
> > > > the failure is due to malformed packet or OOM. Not sure what's the best
> > > > solution here. What's the malformed packet you were talking about?
> > > >
> > >
> > > For example, the pull would fail if the packet had an either_type of
> > > IP but didn't contain an IP header, or perhaps an IPv6 packet that had
> > > an incomplete option header sequence. I would have thought such a
> > > packet was a more likely cause of failure than OOM, so -EPROTO seems a
> > > reasonable best guess.
> > 
> > How? __pskb_pull_tail doesn't seem to care about upper layer protocols.
> > And maybe_pull_tail has already done some lenght comparisions.
> > 
> 
> No, __pskb_pull_tail() doesn't care but the final check in
> maybe_pull_tail() means it will return false if skb_headlen() is not
> at least as big as what it was asked for. So if we try to pull up an
> IP header and there's fewer bytes than that available then we hit the
> error condition. Or maybe I'm missing something.

OK, we'er still on the same boat here. ;-)

Would it make sense to make maybe_pull_tail to return int to reflect
__pskb_pull_fail? In that case we can distinguish backend failure and
frontend failure.

I pay extra attention to this as we often have no access to frontend and
we probably don't want to blame frontend for non-existent misbehavior.

Wei.

> 
>   Paul
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html