[Intel-wired-lan] [PATCH V3] ixgbe: Handle extended IPv6 headers in tx path

[Rustad, Mark D]

Alexander Duyck <alexander.duyck at gmail.com> wrote:

>> +                       rc = ipv6_skip_exthdr(skb, network_hdr.raw - skb->data +
>> +                                                  sizeof(struct ipv6hdr),
>> +                                             &l4_hdr, &frag_off);
>> +                       if (unlikely(rc < 0 || frag_off)) {
>> +                               if (net_ratelimit()) {
>> +                                       dev_warn(tx_ring->dev,
>> +                                                "ipv6_skip_exthdr returns %d\n",
>> +                                                rc);
> 
> Now that I am seeing this it seems like this is getting kind of
> clunky.  Perhaps we should look at coalescing all of these warning
> messages into one message at the end of the l4_hdr section.
> 
> Also it seems like you could ignore the rc value and just focus on the
> frag_off value.

Almost. If rc < 0, then frag_off will not have been assigned to. If I initialize frag_off to 0, then rc can be completely eliminated as you suggest.

> If the function fails the l4_hdr will remain the
> extended header value and fall into the error path in the l4_hdr
> switch section below.  You could probably use NEXTHDR_FRAGMENT as the
> l4_hdr value if frag_off is non-zero.

Clever. That does get a bit of information out and coalesce the error cases nicely. I'll try that out.

--
Mark Rustad, Networking Division, Intel Corporation
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 841 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://lists.osuosl.org/pipermail/intel-wired-lan/attachments/20151118/7776b41b/attachment-0001.asc>