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

[Wei Liu <wei.liu2@xxxxxxxxxx>]

On Tue, Dec 03, 2013 at 03:05:40PM +0000, Paul Durrant wrote:
> > -----Original Message-----
> > From: Wei Liu [mailto:wei.liu2@xxxxxxxxxx]
> > Sent: 03 December 2013 14:58
> > To: Paul Durrant
> > Cc: Wei Liu; xen-devel@xxxxxxxxxxxxx; netdev@xxxxxxxxxxxxxxx; 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@xxxxxxxxxx]
> > > > Sent: 03 December 2013 14:29
> > > > To: Paul Durrant
> > > > Cc: Wei Liu; xen-devel@xxxxxxxxxxxxx; netdev@xxxxxxxxxxxxxxx; 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

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel