[PATCH mlx5-next] RDMA/mlx5: Enable multiplane mode only when it is supported

[PATCH mlx5-next] RDMA/mlx5: Enable multiplane mode only when it is supported

[Leon Romanovsky]

From: Mark Zhang <markzhang@nvidia.com>

Driver queries vport_cxt.num_plane and enables multiplane when it is
greater then 0, but some old FWs (versions from x.40.1000 till x.42.1000),
report vport_cxt.num_plane = 1 unexpectedly.

Fix it by querying num_plane only when HCA_CAP2.multiplane bit is set.

Fixes: 2a5db20fa532 ("RDMA/mlx5: Add support to multi-plane device and port")
Cc: stable@vger.kernel.org
Reported-by: Francesco Poli <invernomuto@paranoici.org>
Closes: https://lore.kernel.org/all/nvs4i2v7o6vn6zhmtq4sgazy2hu5kiulukxcntdelggmznnl7h@so3oul6uwgbl/
Signed-off-by: Mark Zhang <markzhang@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
---
 drivers/infiniband/hw/mlx5/main.c | 2 +-
 include/linux/mlx5/mlx5_ifc.h     | 4 +++-
 2 files changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index c2314797afc9..f5b59d02f4d3 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -2839,7 +2839,7 @@ static int mlx5_ib_get_plane_num(struct mlx5_core_dev *mdev, u8 *num_plane)
 	int err;
 
 	*num_plane = 0;
-	if (!MLX5_CAP_GEN(mdev, ib_virt))
+	if (!MLX5_CAP_GEN(mdev, ib_virt) || !MLX5_CAP_GEN_2(mdev, multiplane))
 		return 0;
 
 	err = mlx5_query_hca_vport_context(mdev, 0, 1, 0, &vport_ctx);
diff --git a/include/linux/mlx5/mlx5_ifc.h b/include/linux/mlx5/mlx5_ifc.h
index 4fbbcf35498b..48d47181c7cd 100644
--- a/include/linux/mlx5/mlx5_ifc.h
+++ b/include/linux/mlx5/mlx5_ifc.h
@@ -2119,7 +2119,9 @@ struct mlx5_ifc_cmd_hca_cap_2_bits {
 	u8	   migration_in_chunks[0x1];
 	u8	   reserved_at_d1[0x1];
 	u8	   sf_eq_usage[0x1];
-	u8	   reserved_at_d3[0xd];
+	u8	   reserved_at_d3[0x5];
+	u8	   multiplane[0x1];
+	u8	   reserved_at_d9[0x7];
 
 	u8	   cross_vhca_object_to_object_supported[0x20];
 
-- 
2.47.0

Re: [PATCH mlx5-next] RDMA/mlx5: Enable multiplane mode only when it is supported

[Michal Swiatkowski]

On Thu, Dec 19, 2024 at 02:23:36PM +0200, Leon Romanovsky wrote:
> From: Mark Zhang <markzhang@nvidia.com>
> 
> Driver queries vport_cxt.num_plane and enables multiplane when it is
> greater then 0, but some old FWs (versions from x.40.1000 till x.42.1000),
> report vport_cxt.num_plane = 1 unexpectedly.
> 
> Fix it by querying num_plane only when HCA_CAP2.multiplane bit is set.
> 
> Fixes: 2a5db20fa532 ("RDMA/mlx5: Add support to multi-plane device and port")
> Cc: stable@vger.kernel.org
> Reported-by: Francesco Poli <invernomuto@paranoici.org>
> Closes: https://lore.kernel.org/all/nvs4i2v7o6vn6zhmtq4sgazy2hu5kiulukxcntdelggmznnl7h@so3oul6uwgbl/
> Signed-off-by: Mark Zhang <markzhang@nvidia.com>
> Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
> ---
>  drivers/infiniband/hw/mlx5/main.c | 2 +-
>  include/linux/mlx5/mlx5_ifc.h     | 4 +++-
>  2 files changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
> index c2314797afc9..f5b59d02f4d3 100644
> --- a/drivers/infiniband/hw/mlx5/main.c
> +++ b/drivers/infiniband/hw/mlx5/main.c
> @@ -2839,7 +2839,7 @@ static int mlx5_ib_get_plane_num(struct mlx5_core_dev *mdev, u8 *num_plane)
>  	int err;
>  
>  	*num_plane = 0;
> -	if (!MLX5_CAP_GEN(mdev, ib_virt))
> +	if (!MLX5_CAP_GEN(mdev, ib_virt) || !MLX5_CAP_GEN_2(mdev, multiplane))
>  		return 0;
>  
>  	err = mlx5_query_hca_vport_context(mdev, 0, 1, 0, &vport_ctx);
> diff --git a/include/linux/mlx5/mlx5_ifc.h b/include/linux/mlx5/mlx5_ifc.h
> index 4fbbcf35498b..48d47181c7cd 100644
> --- a/include/linux/mlx5/mlx5_ifc.h
> +++ b/include/linux/mlx5/mlx5_ifc.h
> @@ -2119,7 +2119,9 @@ struct mlx5_ifc_cmd_hca_cap_2_bits {
>  	u8	   migration_in_chunks[0x1];
>  	u8	   reserved_at_d1[0x1];
>  	u8	   sf_eq_usage[0x1];
> -	u8	   reserved_at_d3[0xd];
> +	u8	   reserved_at_d3[0x5];
> +	u8	   multiplane[0x1];
> +	u8	   reserved_at_d9[0x7];
>  
>  	u8	   cross_vhca_object_to_object_supported[0x20];

Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>

Just out of curiosity, don't you have mlx5-net or sth like that for
fixes?

>  
> -- 
> 2.47.0

Re: [PATCH mlx5-next] RDMA/mlx5: Enable multiplane mode only when it is supported

[Leon Romanovsky]

On Thu, Dec 19, 2024 at 02:23:04PM +0100, Michal Swiatkowski wrote:
> On Thu, Dec 19, 2024 at 02:23:36PM +0200, Leon Romanovsky wrote:
> > From: Mark Zhang <markzhang@nvidia.com>
> > 
> > Driver queries vport_cxt.num_plane and enables multiplane when it is
> > greater then 0, but some old FWs (versions from x.40.1000 till x.42.1000),
> > report vport_cxt.num_plane = 1 unexpectedly.
> > 
> > Fix it by querying num_plane only when HCA_CAP2.multiplane bit is set.
> > 
> > Fixes: 2a5db20fa532 ("RDMA/mlx5: Add support to multi-plane device and port")
> > Cc: stable@vger.kernel.org
> > Reported-by: Francesco Poli <invernomuto@paranoici.org>
> > Closes: https://lore.kernel.org/all/nvs4i2v7o6vn6zhmtq4sgazy2hu5kiulukxcntdelggmznnl7h@so3oul6uwgbl/
> > Signed-off-by: Mark Zhang <markzhang@nvidia.com>
> > Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
> > ---
> >  drivers/infiniband/hw/mlx5/main.c | 2 +-
> >  include/linux/mlx5/mlx5_ifc.h     | 4 +++-
> >  2 files changed, 4 insertions(+), 2 deletions(-)
> > 
> > diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
> > index c2314797afc9..f5b59d02f4d3 100644
> > --- a/drivers/infiniband/hw/mlx5/main.c
> > +++ b/drivers/infiniband/hw/mlx5/main.c
> > @@ -2839,7 +2839,7 @@ static int mlx5_ib_get_plane_num(struct mlx5_core_dev *mdev, u8 *num_plane)
> >  	int err;
> >  
> >  	*num_plane = 0;
> > -	if (!MLX5_CAP_GEN(mdev, ib_virt))
> > +	if (!MLX5_CAP_GEN(mdev, ib_virt) || !MLX5_CAP_GEN_2(mdev, multiplane))
> >  		return 0;
> >  
> >  	err = mlx5_query_hca_vport_context(mdev, 0, 1, 0, &vport_ctx);
> > diff --git a/include/linux/mlx5/mlx5_ifc.h b/include/linux/mlx5/mlx5_ifc.h
> > index 4fbbcf35498b..48d47181c7cd 100644
> > --- a/include/linux/mlx5/mlx5_ifc.h
> > +++ b/include/linux/mlx5/mlx5_ifc.h
> > @@ -2119,7 +2119,9 @@ struct mlx5_ifc_cmd_hca_cap_2_bits {
> >  	u8	   migration_in_chunks[0x1];
> >  	u8	   reserved_at_d1[0x1];
> >  	u8	   sf_eq_usage[0x1];
> > -	u8	   reserved_at_d3[0xd];
> > +	u8	   reserved_at_d3[0x5];
> > +	u8	   multiplane[0x1];
> > +	u8	   reserved_at_d9[0x7];
> >  
> >  	u8	   cross_vhca_object_to_object_supported[0x20];
> 
> Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
> 
> Just out of curiosity, don't you have mlx5-net or sth like that for
> fixes?

No, we don't have such as it is so rare situation that we have fix for
shared branch. I wrote here mlx5-next target because this patch changes
the shared mlx5_ifc.h file and for the visibility, but it doesn't affect
mlx5 eth devices.

Most likely, we will end taking this patch directly to rdma-rc and sending
as part of usual PR to Linus.

Thanks

> 
> >  
> > -- 
> > 2.47.0

Re: [PATCH mlx5-next] RDMA/mlx5: Enable multiplane mode only when it is supported

[Jason Gunthorpe]

On Thu, Dec 19, 2024 at 02:23:36PM +0200, Leon Romanovsky wrote:
> From: Mark Zhang <markzhang@nvidia.com>
> 
> Driver queries vport_cxt.num_plane and enables multiplane when it is
> greater then 0, but some old FWs (versions from x.40.1000 till x.42.1000),
> report vport_cxt.num_plane = 1 unexpectedly.
> 
> Fix it by querying num_plane only when HCA_CAP2.multiplane bit is set.
> 
> Fixes: 2a5db20fa532 ("RDMA/mlx5: Add support to multi-plane device and port")
> Cc: stable@vger.kernel.org
> Reported-by: Francesco Poli <invernomuto@paranoici.org>
> Closes: https://lore.kernel.org/all/nvs4i2v7o6vn6zhmtq4sgazy2hu5kiulukxcntdelggmznnl7h@so3oul6uwgbl/
> Signed-off-by: Mark Zhang <markzhang@nvidia.com>
> Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
> Reviewed-by: Michal Swiatkowski <michal.swiatkowski@linux.intel.com>
> ---
>  drivers/infiniband/hw/mlx5/main.c | 2 +-
>  include/linux/mlx5/mlx5_ifc.h     | 4 +++-
>  2 files changed, 4 insertions(+), 2 deletions(-)

Applied to for-rc thanks

Jason