[ovs-dev] [PATCH] ofproto: Properly refresh rule modified time when nothing else changes.
Ben Pfaff
blp at nicira.com
Sat Jan 26 01:24:00 UTC 2013
Thanks, I applied this to master. I know we talked about applying it
to earlier branches too, but I didn't do that yet.
On Fri, Jan 25, 2013 at 04:51:51PM -0800, Justin Pettit wrote:
> Looks good.
>
> --Justin
>
>
> On Jan 25, 2013, at 3:00 PM, Ben Pfaff <blp at nicira.com> wrote:
>
> > In Open vSwitch, a "modify" or "modify_strict" flow_mod is supposed to
> > refresh the flow's last-modified time even if nothing else changes, because
> > this interpretation makes the "learn" action more useful. As commit
> > 308881afb (ofproto: Reinterpret meaning of OpenFlow hard timeouts with
> > OFPFC_MODIFY.) notes:
> >
> > I finally found a good use for hard timeouts in OpenFlow, but they
> > require a slight reinterpretation of the meaning of hard timeouts.
> > Until now, a hard timeout meant that a flow would be removed the
> > specified number of seconds after a flow was created. Intervening
> > modifications with OFPFC_MODIFY(_STRICT) had no effect on the hard
> > timeout; the flow would still be deleted the specified number of
> > seconds after its original creation.
> >
> > This commit changes the effect of OFPFC_MODIFY(_STRICT). Now,
> > modifying a flow resets its hard timeout counter. A flow will time out
> > the specified number of seconds after creation or after the last time
> > it is modified, whichever comes later.
> >
> > However, commit 080437614b (ofproto: Represent flow cookie changes as
> > operations too.) broke this behavior because it incorrectly optimized out
> > "modify" operations that didn't change the flow's actions or flow cookie.
> > This commit fixes the problem, and adds a test to prevent future
> > regression.
> >
> > Thanks to Amar Padmanabhan <amar at nicira.com> for helping to track this
> > down.
> >
> > Bug #14538.
> > Reported-by: Hiroshi Tanaka <htanaka at vmware.com>
> > CC: Amar Padmanabhan <amar at nicira.com>
> > Signed-off-by: Ben Pfaff <blp at nicira.com>
> > ---
> > ofproto/ofproto.c | 18 +++++++++---
> > tests/learn.at | 75 +++++++++++++++++++++++++++++++++++++++++++++++++++++
> > 2 files changed, 88 insertions(+), 5 deletions(-)
> >
> > diff --git a/ofproto/ofproto.c b/ofproto/ofproto.c
> > index bc27079..249f2d5 100644
> > --- a/ofproto/ofproto.c
> > +++ b/ofproto/ofproto.c
> > @@ -3261,10 +3261,6 @@ modify_flows__(struct ofproto *ofproto, struct ofconn *ofconn,
> > new_cookie = (fm->new_cookie != htonll(UINT64_MAX)
> > ? fm->new_cookie
> > : rule->flow_cookie);
> > - if (!actions_changed && new_cookie == rule->flow_cookie) {
> > - /* No change at all. */
> > - continue;
> > - }
> >
> > op = ofoperation_create(group, rule, OFOPERATION_MODIFY, 0);
> > rule->flow_cookie = new_cookie;
> > @@ -4247,7 +4243,19 @@ ofopgroup_complete(struct ofopgroup *group)
> > LIST_FOR_EACH_SAFE (op, next_op, group_node, &group->ops) {
> > struct rule *rule = op->rule;
> >
> > - if (!op->error && !ofproto_rule_is_hidden(rule)) {
> > + /* We generally want to report the change to active OpenFlow flow
> > + monitors (e.g. NXST_FLOW_MONITOR). There are three exceptions:
> > +
> > + - The operation failed.
> > +
> > + - The affected rule is not visible to controllers.
> > +
> > + - The operation's only effect was to update rule->modified. */
> > + if (!(op->error
> > + || ofproto_rule_is_hidden(rule)
> > + || (op->type == OFOPERATION_MODIFY
> > + && op->ofpacts
> > + && rule->flow_cookie == op->flow_cookie))) {
> > /* Check that we can just cast from ofoperation_type to
> > * nx_flow_update_event. */
> > BUILD_ASSERT_DECL((enum nx_flow_update_event) OFOPERATION_ADD
> > diff --git a/tests/learn.at b/tests/learn.at
> > index 47c1d32..800dc14 100644
> > --- a/tests/learn.at
> > +++ b/tests/learn.at
> > @@ -122,6 +122,81 @@ NXST_FLOW reply:
> > OVS_VSWITCHD_STOP
> > AT_CLEANUP
> >
> > +dnl This test checks that repeated uses of a "learn" action cause the
> > +dnl modified time of the learned flow to advance. Otherwise, the
> > +dnl learned flow will expire after its hard timeout even though it's
> > +dnl supposed to be refreshed. (The expiration can be hard to see since
> > +dnl it gets re-learned again the next time a packet appears, but
> > +dnl sometimes the expiration can cause temporary flooding etc.)
> > +AT_SETUP([learning action - learn refreshes hard_age])
> > +OVS_VSWITCHD_START(
> > + [add-port br0 p1 -- set Interface p1 type=dummy ofport_request=1 -- \
> > + add-port br0 p2 -- set Interface p2 type=dummy ofport_request=2 -- \
> > + add-port br0 p3 -- set Interface p3 type=dummy ofport_request=3])
> > +
> > +ovs-appctl time/stop
> > +
> > +# Set up flow table for MAC learning.
> > +AT_DATA([flows.txt], [[
> > +table=0 actions=learn(table=1, hard_timeout=10, NXM_OF_ETH_DST[]=NXM_OF_ETH_SRC[], output:NXM_OF_IN_PORT[]), resubmit(,1)
> > +table=1 priority=0 actions=flood
> > +]])
> > +AT_CHECK([ovs-ofctl add-flows br0 flows.txt])
> > +
> > +# Trace an ICMP packet arriving on port 3, to create a MAC learning entry.
> > +flow="in_port(3),eth(src=50:54:00:00:00:07,dst=50:54:00:00:00:05),eth_type(0x0800),ipv4(src=192.168.0.2,dst=192.168.0.1,proto=1,tos=0,ttl=64,frag=no),icmp(type=0,code=0)"
> > +AT_CHECK([ovs-appctl ofproto/trace br0 "$flow" -generate], [0], [stdout])
> > +actual=`tail -1 stdout | sed 's/Datapath actions: //'`
> > +
> > +expected="1,2,100"
> > +AT_CHECK([ovs-dpctl normalize-actions "$flow" "$expected"], [0], [stdout])
> > +mv stdout expout
> > +AT_CHECK([ovs-dpctl normalize-actions "$flow" "$actual"], [0], [expout])
> > +
> > +# Check that the MAC learning entry appeared.
> > +AT_CHECK([ovs-ofctl dump-flows br0 table=1 | ofctl_strip | sort], [0], [dnl
> > + table=1, hard_timeout=10, dl_dst=50:54:00:00:00:07 actions=output:3
> > + table=1, priority=0 actions=FLOOD
> > +NXST_FLOW reply:
> > +])
> > +
> > +# For 25 seconds, make sure that the MAC learning entry doesn't
> > +# disappear as long as we refresh it every second.
> > +for i in 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25; do
> > + ovs-appctl time/warp 1000
> > + AT_CHECK([ovs-appctl ofproto/trace br0 "$flow" -generate], [0], [stdout])
> > +
> > + # Check that the entry is there.
> > + AT_CHECK([ovs-ofctl dump-flows br0 table=1], [0], [stdout])
> > + AT_CHECK([ofctl_strip < stdout | sort], [0], [dnl
> > + table=1, hard_timeout=10, dl_dst=50:54:00:00:00:07 actions=output:3
> > + table=1, priority=0 actions=FLOOD
> > +NXST_FLOW reply:
> > +])
> > +
> > + if test $i != 1; then
> > + # Check that hard_age has appeared. We need to do this separately
> > + # from the above check because ofctl_strip removes it. dump-flows
> > + # only prints hard_age when it is different from the flow's duration
> > + # (that is, the number of seconds from the time it was created),
> > + # so we only check for it after we've refreshed the flow once.
> > + AT_CHECK([grep dl_dst=50:54:00:00:00:07 stdout | grep -c hard_age],
> > + [0], [1
> > +])
> > + fi
> > +done
> > +
> > +# Make sure that 15 seconds without refreshing makes the flow time out.
> > +ovs-appctl time/warp 5000
> > +ovs-appctl time/warp 5000
> > +ovs-appctl time/warp 5000
> > + AT_CHECK([ovs-ofctl dump-flows br0 table=1 | ofctl_strip | sort], [0], [dnl
> > + table=1, priority=0 actions=FLOOD
> > +NXST_FLOW reply:
> > +])
> > +OVS_VSWITCHD_STOP
> > +AT_CLEANUP
> > +
> > AT_SETUP([learning action - TCPv4 port learning])
> > OVS_VSWITCHD_START(
> > [add-port br0 p1 -- set Interface p1 type=dummy -- \
> > --
> > 1.7.2.5
> >
> > _______________________________________________
> > dev mailing list
> > dev at openvswitch.org
> > http://openvswitch.org/mailman/listinfo/dev
>
More information about the dev
mailing list