On Fri, May 01, 2020 at 07:28:14PM +0000, Saeed Mahameed wrote:

On Fri, 2020-05-01 at 11:44 -0300, Jason Gunthorpe wrote:

quoted

On Thu, Apr 30, 2020 at 10:21:30PM +0300, Maor Gottlieb wrote:

quoted

Hi Dave,

This series is a combination of netdev and RDMA, so in order to
avoid
conflicts, we would like to ask you to route this series through
mlx5-next shared branch. It is based on v5.7-rc2 tag.


The following series adds support to get the LAG master xmit slave
by
introducing new .ndo - ndo_get_xmit_slave. Every LAG module can
implement it and it first implemented in the bond driver. 
This is follow-up to the RFC discussion [1].

The main motivation for doing this is for drivers that offload part
of the LAG functionality. For example, Mellanox Connect-X hardware
implements RoCE LAG which selects the TX affinity when the
resources
are created and port is remapped when it goes down.

The first part of this patchset introduces the new .ndo and add the
support to the bonding module.

The second part adds support to get the RoCE LAG xmit slave by
building
skb of the RoCE packet based on the AH attributes and call to the
new
.ndo.

The third part change the mlx5 driver driver to set the QP's
affinity
port according to the slave which found by the .ndo.

Thanks

[1]
https://lore.kernel.org/netdev/20200126132126.9981-1-maorg@xxxxxxxxxxxx/

where did these xxxxx's come from?

quoted

Change log:
v8: Fix bad numbering of v7. 
v7: Change only in RDMA part:
	- return slave and as output
	- Don't hold lock while allocating skb.
    In addition, reorder patches, so mlx5 patches are before RDMA.
v6: patch 1 - Fix commit message and add function description. 
    patch 10 - Keep udata as function argument.
v5: patch 1 - Remove rcu lock.
    patch 10 - Refactor patch that group the AH attributes in
struct.
    patch 11 - call the ndo while holding the rcu and initialize
xmit_slave.
    patch 12 - Store the xmit slave in rdma_ah_init_attr and
qp_attr.

v4: 1. Rename master_get_xmit_slave to netdev_get_xmit_slave and
move
the implementation to dev.c 
    2. Remove unnecessary check of NULL pointer.
    3. Fix typo.
v3: 1. Move master_get_xmit_slave to netdevice.h and change the
flags
arg.
to bool.
    2. Split helper functions commit to multiple commits for each
bond
mode.
    3. Extract refcotring changes to seperate commits.
v2: The first patch wasn't sent in v1.
v1:
https://lore.kernel.org/netdev/ac373456-b838-29cf-645f-b1ea1a93e3b0@xxxxxxxxx/T/#t 

Maor Gottlieb (16):
  net/core: Introduce netdev_get_xmit_slave
  bonding: Export skip slave logic to function
  bonding: Rename slave_arr to usable_slaves
  bonding/alb: Add helper functions to get the xmit slave
  bonding: Add helper function to get the xmit slave based on hash
  bonding: Add helper function to get the xmit slave in rr mode
  bonding: Add function to get the xmit slave in active-backup mode
  bonding: Add array of all slaves
  bonding: Implement ndo_get_xmit_slave
  net/mlx5: Change lag mutex lock to spin lock
  net/mlx5: Add support to get lag physical port
  RDMA: Group create AH arguments in struct
  RDMA/core: Add LAG functionality
  RDMA/core: Get xmit slave for LAG
  RDMA/mlx5: Refactor affinity related code
  RDMA/mlx5: Set lag tx affinity according to slave

It seems fine to me too, Saeed, can you apply the net parts to the
mlx5 shared branch with DaveM's ack? Thanks

Done:

https://git.kernel.org/pub/scm/linux/kernel/git/mellanox/linux.git/log/?h=mlx5-next

Okay, applied thanks

Jason