So people don't think that I'm doing nothing but sleeping today :-) I gave a shot yesterday and today to kill this off, but it really can't be done easily in one fell swoop, and there are real issues to resolve.
So I'll work on this in stages. First I'll get rid of the most trivial uses of skb->h (those where skb->data is equally correct to use). Then I'll take care of the less trivial cases (such as TCP, which simply wants the FIN/SYN flags of received packets, which we can store in the TCP_SKB_CB() as we do for the output path). Then we'll need to discuss the checksum usage. Unlike Alexey seemed to indicate, non-ipv4 checksum offload via CHECKSUM_HW really does get used. In particular, sungem and sunhme support this perfectly, and the s2io stuff will likely do so as well. So we have to provide some way to let the card know the offset into the packet at which to start the checksum. I think we can stick the necessary pointer into skb_shared_info() just like we do for the TSO offload state, but that's just one idea. Perhaps there is some more clever way to replace the existing usage without adding anything to skb_shared_info(), suggestions welcome :-) More tomorrow, hopefully... I need to look at some long standing sparc64 SMP problems as well some time this week. - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
