On Thu, Apr 05, 2012 at 05:57:57PM -0700, Mehak Mahajan wrote:
> The CFM packets that are out of sequence or contain invalid cfm_interval were
> previously not ignored. The behavior is changed with this patch to not
> process those CFM frames.
>
> Signed-off-by: Mehak Mahajan
Please put a space
Thanks, this looks good to me.
> + if (rmp->mpid == cfm->mpid) {
> + cfm_fault |= CFM_FAULT_LOOPBACK;
> + VLOG_WARN_RL(&rl,"%s: received CCM with local MPID"
> + " %"PRIu64, cfm->name, rmp->mpid);
> + }
The indentatio
The CFM packets that are out of sequence or contain invalid cfm_interval were
previously not ignored. The behavior is changed with this patch to not
process those CFM frames.
Signed-off-by: Mehak Mahajan
---
lib/cfm.c| 61 ++---
lib/cfm.h
Hey,
Thank Ben and Ethan for your comments.
I have taken care of them.
We are now setting all possible faults when we get the heartbeat message.
thanx!
mehak
On Thu, Apr 5, 2012 at 4:00 PM, Ethan Jackson wrote:
> > -rmp->recv = true;
> > -if (!fault) {
> > +
> - rmp->recv = true;
> - if (!fault) {
> + if (!cfm_fault && !ccm_rdi) {
> rmp->num_health_ccm++;
> }
> + if (!cfm_fault) {
> + rmp->recv = true;
> + }
> + cfm->recv_fault |= cfm_fault;
I
On Thu, Apr 05, 2012 at 03:41:45PM -0700, Ethan Jackson wrote:
> > That sounds even better, because the CFM object would signal RDI
> > immediately without waiting for the next fault check.
>
> No actually this data only changes on fault checks.
Ah, you're right, of course.
__
> No actually this data only changes on fault checks.
i.e. it's only written to the database on fault checks.
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
> That sounds even better, because the CFM object would signal RDI
> immediately without waiting for the next fault check.
No actually this data only changes on fault checks.
___
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinf
On Thu, Apr 05, 2012 at 03:32:50PM -0700, Ethan Jackson wrote:
> > It looks like rdi gets updated on every CCM packet receive, so that if
> > we receive a CCM without RDI after one with it between fault checks,
> > cfm_run() won't report the RDI. I wonder whether we should actually
> > only reset
> It looks like rdi gets updated on every CCM packet receive, so that if
> we receive a CCM without RDI after one with it between fault checks,
> cfm_run() won't report the RDI. I wonder whether we should actually
> only reset rdi to false in cfm_run() and only set it to true, never to
> false, in
On Thu, Apr 05, 2012 at 03:12:21PM -0700, Mehak Mahajan wrote:
> The CFM packets that are out of sequence or contain invalid cfm_interval were
> previously not ignored. The behavior is changed with this patch to not
> process those CFM frames.
>
> Signed-off-by: Mehak Mahajan
I guess Ethan ought
The CFM packets that are out of sequence or contain invalid cfm_interval were
previously not ignored. The behavior is changed with this patch to not
process those CFM frames.
Signed-off-by: Mehak Mahajan
---
lib/cfm.c| 21 ++---
lib/cfm.h|4 +++-
vsw
Thanks for doing this. It basically looks like what we want. One question,
shouldn't we calculate the new cfm_fault flag and then set then OR it into the
old recv fault flag? It looks like with this code the new fault reasons won't
cause the fault flag to be triggered.
Ethan
On Apr 4, 2
The CFM packets that are out of sequence or contain invalid cfm_interval were
previously not ignored. The behavior is changed with this patch to not
process those CFM frames.
Signed-off-by: Mehak Mahajan
---
lib/cfm.c|9 -
lib/cfm.h|4 +++-
vswitchd/vswitc
14 matches
Mail list logo
