[ovs-dev] [PATCH v5 0/6] netdev-dpdk: Fix management of pre-existing mempools.
antonio.fischetti at intel.com
Fri Oct 13 17:55:28 UTC 2017
Thanks Mark for your review, some comment inline.
> -----Original Message-----
> From: Kavanagh, Mark B
> Sent: Friday, October 13, 2017 3:46 PM
> To: Fischetti, Antonio <antonio.fischetti at intel.com>; dev at openvswitch.org
> Subject: RE: [ovs-dev] [PATCH v5 0/6] netdev-dpdk: Fix management of pre-
> existing mempools.
> >From: ovs-dev-bounces at openvswitch.org [mailto:ovs-dev-bounces at openvswitch.org]
> >On Behalf Of antonio.fischetti at intel.com
> >Sent: Wednesday, October 11, 2017 5:01 PM
> >To: dev at openvswitch.org
> >Subject: [ovs-dev] [PATCH v5 0/6] netdev-dpdk: Fix management of pre-existing
> >List of versions:
> > - v5: manage new MTU value when a pre-existing mempool is returned.
> > - v4: fix NUMA awareness usecase
> > - v3: avoid deletion of pre-existing mempools
> > - v2: rework to accomodate code changes for dpdk ports too
> > - v1: 1st implementation.
> >Fischetti, Antonio (6):
> > netdev-dpdk: fix management of pre-existing mempools.
> > netdev-dpdk: Fix mempool names to reflect socket id.
> > netdev-dpdk: skip init for existing mempools.
> > netdev-dpdk: assert mempool names.
> > netdev-dpdk: Reword mp_size as n_mbufs.
> > netdev-dpdk: Rename dpdk_mp_put as dpdk_mp_free.
> Hi Antonio,
> Some general comments on this patchset:
> - patches 1 and 3 of the series should be squashed into one.
The fixes are for two different issues:
- Issue #1: detecting when previous mempools must be released
- Issue #2: mempool name generation for NUMA-awareness test case
The 1st issue came earlier from a PVP test case and while its fix
was being tested, the 2nd issue for the NUMA-awareness usecase
came later on. I kept these two fixes into two separate
patches because the vHost Zero-copy development was blocked
by the issue #1, so keeping two independent patches would speed
up upstreaming patch #1 - and unblocking vHost Zero-copy
> - patches 3-6 are purely cosmetic, and do not contribute to fixing the mempool
> reconfiguration mechanism;
[Antonio] Agree, the name of the series says "Fix management.."
but the real 'fixes' are in patches #1 and #2.
> as such, they should be part of a separate patchset.
> - initial testing has been successful; I'll send on final results as soon as
> they are available.
> > lib/netdev-dpdk.c | 88 +++++++++++++++++++++++++++++++-----------------------
> > 1 file changed, 50 insertions(+), 38 deletions(-)
> >dev mailing list
> >dev at openvswitch.org
More information about the dev