On Wed, Jun 26, 2013 at 9:53 PM, Simon Horman <ho...@verge.net.au> wrote: > On Mon, Jun 24, 2013 at 12:29:55PM -0700, Jesse Gross wrote: >> On Sun, Jun 23, 2013 at 7:29 PM, Simon Horman <ho...@verge.net.au> wrote: >> > It appears that in the case where vs_match_init() is called from >> > ovs_flow_metadata_from_nlattrs() it is undesirable to set the flow key as >> > some of its values are set earlier on in ovs_flow_metadata_from_nlattrs(). >> > Furthermore in the case where ovs_flow_metadata_from_nlattrs() is called >> > from ovs_packet_cmd_execute() key has been partially initialised by >> > ovs_flow_extract(). >> > >> > This manifests in a problem when executing actions as elements of key are >> > used when verifying some actions. For example a dec_ttl action verifies the >> > proto of the flow. An example of a flow that fails as a result of this >> > problem is: >> > >> > ovs-ofctl add-flow br0 "ip actions=dec_ttl,normal" >> > >> > This patch resolves this problem by not clearing key in ovs_match_init() >> > and instead clearing it before calling ovs_match_init() when it is called >> > other than by ovs_flow_metadata_from_nlattrs(). >> > >> > This appears to be a regression added by "datapath: Mega flow >> > implementation", a1c564be1e2ffc31f8da09ab654c8ed987907fe5. >> > >> > Cc: Andy Zhou <az...@nicira.com> >> > Signed-off-by: Simon Horman <ho...@verge.net.au> >> > >> > --- >> > >> > I am unsure of the exact intention of clearing the key and thus the >> > correctness of this patch. However, it does appear to resolve >> > the problem that I describe above. >> >> Thanks, I think this can simplified a little bit. I applied the following >> patch: > > Thanks, I have confirmed that resolves the problem that I observed.
Thanks and sorry about not looking at your patches yet - I've been trying to help stabilize master first (it's getting better but there's still one that's eluding me at the moment). X-CudaMail-Whitelist-To: dev@openvswitch.org _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
