mbox series

[ovs-dev,0/2] Enable support for non-contiguous NUMA nodes

Message ID 20210512162755.2176656-1-drc@linux.vnet.ibm.com
Headers show
Series Enable support for non-contiguous NUMA nodes | expand

Message

David Christensen May 12, 2021, 4:27 p.m. UTC
Systems such as the IBM POWER9 do not allocate contiguous NUMA nodes,
nor does the DPDK framework require that they be contiguous.  This
patchset enables OVS support for systems with non-contiguous NUMA nodes
and adds additional tests using the "--dummy-numa" parameter to verify
the functionality.

David Christensen (2):
  dpdk: support non-contiguous NUMA nodes for IBM POWER systems
  dpdk: add non-contiguous NUMA node support to auto tests

 lib/dpdk.c            |  27 ++++++---
 lib/ovs-numa.c        |  12 +---
 lib/ovs-numa.h        |   1 +
 tests/dpif-netdev.at  |  64 +++++++++++---------
 tests/ofproto-dpif.at | 100 ++++++++++++++++++--------------
 tests/pmd.at          | 132 ++++++++++++++++++++++--------------------
 6 files changed, 186 insertions(+), 150 deletions(-)

Comments

Ilya Maximets May 13, 2021, 12:03 p.m. UTC | #1
On 5/12/21 6:27 PM, David Christensen wrote:
> Systems such as the IBM POWER9 do not allocate contiguous NUMA nodes,
> nor does the DPDK framework require that they be contiguous.  This
> patchset enables OVS support for systems with non-contiguous NUMA nodes
> and adds additional tests using the "--dummy-numa" parameter to verify
> the functionality.
> 
> David Christensen (2):
>   dpdk: support non-contiguous NUMA nodes for IBM POWER systems
>   dpdk: add non-contiguous NUMA node support to auto tests
> 
>  lib/dpdk.c            |  27 ++++++---
>  lib/ovs-numa.c        |  12 +---
>  lib/ovs-numa.h        |   1 +
>  tests/dpif-netdev.at  |  64 +++++++++++---------
>  tests/ofproto-dpif.at | 100 ++++++++++++++++++--------------
>  tests/pmd.at          | 132 ++++++++++++++++++++++--------------------
>  6 files changed, 186 insertions(+), 150 deletions(-)
> 

Hi, David Christensen.  Thanks for working on this, but there is already
almost exactly the same patch-set from David Wilder here:
  https://patchwork.ozlabs.org/project/openvswitch/list/?series=157389

It didn't get enough attention review-wise, so it would be great if you
can review it.  That patch set also additionally accounts for offline
cores.  It needs a slight rebase, though.  I was looking at oldest patches
in our patchwork recently and had an intention to accept these, as they
are in a relatively good shape.

Best regards, Ilya Maximets.