On Aug 13, 2014, at 7:25 AM, Samuel Ghinet <sghi...@cloudbasesolutions.com>
wrote:
> Create a NBL for each NB when required
>
> ovs/ovs-issues#15
>
> All NET_BUFFERs of a NET_BUFFER_LIST must go through
> the pipeline: extract, find flow, execute. Previously,
> only the first NET_BUFFER of a NET_BUFFER_LIST was going
> through this pipeline, which was erroneous.
>
> OvsPartialCopyToMultipleNBLs is used to make each NET_BUFFER
> have its own NET_BUFFER_LIST.
>
> Some functions that used to take as argument a NET_BUFFER_LIST
> now take as argument a NET_BUFFER. I have also added a few
> ASSERTs where the NET_BUFFER_LIST is expected to have
> only one NET_BUFFER."
Samuel,
This is an important change. Thanks for taking a stab at it.
As you know, in the existing code, we block processing of the multi-NB NBL here
in OvsPacketIO.c:
OvsStartNBLIngress():
/* Ethernet Header is a guaranteed safe access. */
curNb = NET_BUFFER_LIST_FIRST_NB(curNbl);
if (curNb->Next != NULL) {
/* XXX: This case is not handled yet. */
ASSERT(FALSE);
} else {
I see that your patch does not make any changes here. So, are the changes you
sent out infrastructure changes to buffer management? How were you able to
validate these changes?
I'm looking at your changes and will provide input. But, clarity about what is
implemented and what is missing would help.
thanks,
Nithin
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
