Re: [PATCH RFC for-next 2/6] RDMA/mlx5: remove deliver net device event

[Leon Romanovsky <leon@kernel.org>]

On Mon, Jan 20, 2020 at 03:31:14PM +0800, Lang Cheng wrote:
>
> On 2020/1/16 19:41, Leon Romanovsky wrote:
> > On Thu, Jan 16, 2020 at 12:10:43PM +0800, Weihang Li wrote:
> > > From: Lang Cheng <chenglang@huawei.com>
> > >
> > > The code that handles the link event of the net device has been moved
> > > into the core, and the related processing should been removed from the
> > > provider's driver.
> > I have serious doubts that this patch broke mlx5 LAG functionality.
>
> All vendor drivers need to remove port link event code,
> and query slave info(only if support bonding) in ops.query_port callback.
> Here is about 4 function:
>
> mlx5_netdev_event(): remove all port link event code after ib core supports
> sending them,
>
> mlx5_get_rep_roce(): Only mlx5_netdev_event() ever called it, and now no
> one, so remove it.
>
> get_port_state():just move public operation to ib core.
>
> mlx5_query_port_roce():	query more info, no impact on existing code.
>
>
> Is there any hidden relationship that I didn't notice?

You didn't missed the functions which are relevant to bond, but from
what I saw you implemented wrongly events handling related to mlx5 bond.

I didn't look very deeply yet because the series is far from
completion and maybe I'm mistaken and bond works perfectly.

Thanks

>
> Thanks.
>
> >
> > Thanks
> > _______________________________________________
> > Linuxarm mailing list
> > Linuxarm@huawei.com
> > http://hulk.huawei.com/mailman/listinfo/linuxarm
> >