Re: [PATCH net-next v5 00/10] add sriov support for wangxun NICs

[Simon Horman <horms@kernel.org>]

On Sun, Aug 04, 2024 at 08:48:31PM +0800, Mengyuan Lou wrote:
> Add sriov_configure for ngbe and txgbe drivers.
> Reallocate queue and irq resources when sriov is enabled.
> Add wx_msg_task in interrupts handler, which is used to process the
> configuration sent by vfs.
> Add ping_vf for wx_pf to tell vfs about pf link change.
> Make devlink allocation function generic to use it for PF and for VF.
> Add PF/VF devlink port creation. It will be used to set/get VFs.

I think it would be good to summarise the overall status of SR-IOV support
with this patch, and what follow-up work is planned. As Jakub mentioned [1]
this does not seem complete as is.

[1] https://lore.kernel.org/netdev/988BFB51-32C8-499C-837D-91CC1C0FFE42@net-swift.com/

I mean, I understand the NDOs were removed from the patchset (see more on
that below) but there needs to be a plan to support users of this device
in a meaningful way.

> 
> v5:
> - Add devlink allocation which will be used to add uAPI.
> - Remove unused EXPORT_SYMBOL.
> - Unify some functions return styles in patch 1 and patch 4.
> - Make the code line less than 80 columns.
> v4:
> https://lore.kernel.org/netdev/3601E5DE87D2BC4F+20240604155850.51983-1-mengyuanlou@net-swift.com/
> - Move wx_ping_vf to patch 6.
> - Modify return section format in Kernel docs.
> v3:
> https://lore.kernel.org/netdev/587FAB7876D85676+20240415110225.75132-1-mengyuanlou@net-swift.com/
> - Do not accept any new implementations of the old SR-IOV API.
> - So remove ndo_vf_xxx in these patches. Switch mode ops will be added
> - in vf driver which will be submitted later.

FYI, this policy was recently significantly relaxed [2]:

[2] https://lore.kernel.org/netdev/20240620002741.1029936-1-kuba@kernel.org/

> v2:
> https://lore.kernel.org/netdev/EF19E603F7CCA7B9+20240403092714.3027-1-mengyuanlou@net-swift.com/
> - Fix some used uninitialised.
> - Use poll + yield with delay instead of busy poll of 10 times in
>  mbx_lock obtain.
> - Split msg_task and flow into separate patches.
> v1:
> https://lore.kernel.org/netdev/DA3033FE3CCBBB84+20240307095755.7130-1-mengyuanlou@net-swift.com/

...