On Sun, Mar 25, 2012 at 11:20:40PM -0700, Justin Pettit wrote:
> Signed-off-by: Justin Pettit <jpet...@nicira.com>
I don't really understand this change.
First, my understanding was that on modify and delete operations,
"cookie_mask" is supposed to be a mask for matching an existing
cookie. But the new code in ofputil_flow_mod_usable_protocols() seems
to regard it as a mask for bits that should be set or updated when a
flow is added or modified.
I'd be inclined to change struct ofputil_flow_mod so that there are
three fields, like this:
ovs_be64 cookie; /* Cookie bits to match. */
ovs_be64 cookie_mask; /* 1-bit in each 'cookie' bit to match. */
ovs_be64 new_cookie; /* New cookie to install or UINT64_MAX. */
which would make it clear what is intended although I haven't
carefully though it through.
Second, the comment here:
# In NXM, the modification of an existing flow should not update the
# cookie. For OpenFlow, the behavior is different depending on the
# version. See below.
seems to inaccurately describe what OVS actually implements. If this
is correct, then the following:
ovs-ofctl -F nxm add-flow br0 cookie=0x1,in_port=1,actions=0
ovs-ofctl -F nxm mod-flows br0 cookie=0x4321/1,in_port=1,actions=1
should yield a flow table like this, with the cookie unchanged:
cookie=0x1 in_port=1 actions=output:1
but in fact in my testing it changes the cookie:
cookie=0x4321 in_port=1 actions=output:1
What am I missing?
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
