[ovs-dev] [PATCH 7/7] datapath-windows: Fixes in HvCreatePort() to re-add a port

Ankur Sharma ankursharma at vmware.com
Wed Nov 19 00:57:13 UTC 2014


Acked-by: Ankur Sharma <ankursharma at vmware.com>
________________________________________
From: dev <dev-bounces at openvswitch.org> on behalf of Nithin Raju <nithin at vmware.com>
Sent: Wednesday, November 12, 2014 10:56 AM
To: Eitan Eliahu
Cc: dev at openvswitch.org
Subject: Re: [ovs-dev] [PATCH 7/7] datapath-windows: Fixes in HvCreatePort() to re-add a port

On Nov 12, 2014, at 9:30 AM, Eitan Eliahu <eliahue at vmware.com> wrote:

> Hi Nithin,
> For all port add/delete functionality we need to validate in three major port states:
> [1] The port exists  on the Hyper-V Virtual Switch but there is not corresponding port in OVS.
> [2] The port exists in OVS user-mode but there is no corresponding VSwitch port.
> [3] The port is on the VSwitch and there is a corresponding OVS port.
> [4] The port is not connected to the switch (VM is down).
> [5] Special handling for Internal and External ports.

#1, #3 are the same.

#4 is not practically important since from the OVS extension perspective, we can't even know if a VM is powered on or not. The only thing we care about is if a VM's NIC gets connected/disconnected.

As mentioned in the cover letter for the patch series #5 is not tested, and I know it won't work. I know the fix though, and will make it in a subsequent patch. This series is already complicated enough, and I want to make progress for VIF ports. I can assure you that the change for the #5 is not big.

> Also, for memory leak verification uninstallation should be performed after there were packets were transferred over this port.
> We need to perform these test with the Verifier enabled.

Running with the verifer is a separate activity, and I have not exercised it. I did do uninstall though, and uninstall did not hang, nor did I hit the following ASSERTs()
ASSERT(switchContext->numHvPorts == 0);
ASSERT(switchContext->numNonHvPorts == 0);

So, I am satisfied that no vports are being leaked.

> Can you exercise these scenarios ?

Yes, with the caveats I have mentioned for #5, I have exercised the scenarios mentioned.

thanks,
-- Nithin
_______________________________________________
dev mailing list
dev at openvswitch.org
https://urldefense.proofpoint.com/v2/url?u=http-3A__openvswitch.org_mailman_listinfo_dev&d=AAIGaQ&c=Sqcl0Ez6M0X8aeM67LKIiDJAXVeAw-YihVMNtXt-uEs&r=RDZsrBxhAiOewDSD-jiF-W03FqHevF2o7haW6eQzmtA&m=ffPIGgyXtwNQITHY3DoBSIfnVFaFpu7-UbU3_knTwkg&s=LIjIK45QUcdNPlTYVI93F2dk2x8Hid0ooVwUp60OyRU&e=



More information about the dev mailing list