On Mon, Nov 21, 2011 at 10:29 AM, Ben Pfaff <b...@nicira.com> wrote:
> On Mon, Nov 21, 2011 at 10:16:01AM -0800, Jesse Gross wrote:
>> On Mon, Nov 21, 2011 at 10:09 AM, Ben Pfaff <b...@nicira.com> wrote:
>> > On Sat, Nov 19, 2011 at 02:40:13PM -0800, Jesse Gross wrote:
>> >> We currently notify for port deletions in the namespace of the device
>> >> that was deleted. ??In general this should be initial namespace because
>> >> that's the only place where we look but it's possible that the device
>> >> was moved after being attached. ??However, it's not semantically correct
>> >> because we really care about the namespace of the userspace process, not
>> >> that of the device. ??This switches to genl_multicast() which always uses
>> >> the initial namespace and seems more appropriate anyways.
>> >>
>> >> Signed-off-by: Jesse Gross <je...@nicira.com>
>> >
>> > I don't understand the primary explanation of this issue. ??What
>> > userspace process's namespace do we care about?
>>
>> ovs-vswitchd, which will only be listening in the initial namespace
>> (at the moment but still a single namespace in the future).  The
>> kernel should send all notifications to that namespace and it is not
>> related to where the devices are attached.
>
> OK.
>
> My guess is that if we get improved namespace support then a datapath
> will belong to a namespace and then the notification will go to that
> namespace?

Yes, I think that's right.  Right now all datapaths de facto belong to
the initial namespace, which is why I wanted to make this change to
always notify in that namespace.

Thanks for the reviews, I've pushed all 3 series.
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev

Reply via email to