On Fri, Dec 18, 2015 at 2:41 AM, Edward Cree <ec...@solarflare.com> wrote: > On 17/12/15 18:06, Tom Herbert wrote: >> I'm not sure that we need bits in VXLAN or any other encapsulation. It >> should be sufficient in udp_set_csum that if we already have >> CHECKSUM_PARTIAL that can always be used to do local checksum offload. > My understandingis that otherwise iptunnel_handle_offloads() will do the > inner checksum in sw, because csum_help will be passed as true. It will > call skb_checksum_help(). >> This is also should be independent as to whether the device does >> NETIF_F_HW_CSUM or can offload NETIF_F_IP[V6]_CSUM for encapsulated >> packets. > I was wary of drivers that declare NETIF_F_IP[V6]_CSUM but don't cope with > encapsulated packets. Would they do the right thing if the inner_csum bool > in patch 2 just tested for NETIF_F_CSUM_MASK, or do I need to test things > like NETIF_F_GSO_UDP_TUNNEL_CSUM? I'm afraid I don't entirely understand > the infrastructure here, so I just did the minimal thing I was sure worked, > i.e. testing for NETIF_F_HW_CSUM.
Drivers indicate that can do NETIF_F_IP[V6]_CSUM for encapsulation by setting enc_features. This is checked in validate_xmit_skb so that if drive can't handle encapsulated checksum skb_checksum_help is called there. >> It would be nice to have a more formal documentation also. This is a >> very powerful mechanism but the math behind it and requirements are >> subtle. >> >> Tom > What would be a good place to put such documentation? In > Documentation/networking, or as part of the big checksums comment at the > top of skbuff.h? > I don't think this right for skbuff.h that should just describe the interface. LCO has no interface like checksum-unnecessary conversion. Checksumming, encapsulation, segmentation offloads are complex enough now there should really be a Linux doc on this maybe modeled after scaling.txt. That's probably more than you bargained for in this patch, but if someone wants to learn how this infrastructure really works, writing a doc is a good way! :-) Tom -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
