On Wed, Mar 7, 2012 at 2:35 PM, Ben Pfaff <b...@nicira.com> wrote: > On Wed, Mar 07, 2012 at 02:26:46PM -0800, Jesse Gross wrote: >> On Wed, Mar 7, 2012 at 10:46 AM, Ben Pfaff <b...@nicira.com> wrote: >> > On Tue, Mar 06, 2012 at 03:40:51PM -0800, Jesse Gross wrote: >> >> When modifying IP addresses or ports on a UDP packet we don't >> >> correctly follow the rules for unchecksummed packets. This meant >> >> that packets without a checksum can be given a incorrect new checksum >> >> and packets with a checksum can become marked as being unchecksummed. >> >> This fixes it to handle those requirements. >> >> >> >> Bug #8937 >> >> >> >> Signed-off-by: Jesse Gross <je...@nicira.com> >> >> --- >> >> v2: CSUM_MANGLED_0 was not introduced until 2.6.20 >> > >> > It's not obvious to me why the UDP cases in this file need a check for >> > OVS_CSUM_PARTIAL but the TCP cases do not. >> >> There are two things: >> * inet_proto_csum_replaceX includes a check for CHECKSUM_PARTIAL and >> it is no-op if it's an offloaded checksum and the modified part is not >> in the pseudoheader, so both TCP and UDP do handle this (and did >> before this patch). >> * When using CHECKSUM_PARTIAL with UDP the checksum of the >> pseudoheader is stored in the checksum field without using >> CSUM_MANGLED_0 so it's possible for it to have the value 0 and be >> interpreted as an unchecksummed packet. Actual unchecksummed packets >> won't use CHECKSUM_PARTIAL so we know the we always need to update >> these. > > Thank you for the explanations. After staring at the code and the > changes for a while longer, it all makes sense. > > Acked-by: Ben Pfaff <b...@nicira.com> > (again)
Thanks, pushed to master and branches 1.5 and 1.6. _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
