On 10/19/2021 2:16 PM, Max Gurtovoy wrote:

On 10/19/2021 1:58 PM, Yishai Hadas wrote:

quoted

Use mlx5_vf_get_core_dev() to get PF device instead of accessing
directly the PF data structure from the VF one.

The mlx5_vf_get_core_dev() API in its turn uses the generic PCI API
(i.e. pci_iov_get_pf_drvdata) to get it.

Signed-off-by: Yishai Hadas <yishaih@nvidia.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
---
  drivers/vdpa/mlx5/net/mlx5_vnet.c | 27 +++++++++++++++++++++------
  1 file changed, 21 insertions(+), 6 deletions(-)

diff --git a/drivers/vdpa/mlx5/net/mlx5_vnet.c 

b/drivers/vdpa/mlx5/net/mlx5_vnet.c
index 5c7d2a953dbd..97b8917bc34d 100644

--- a/drivers/vdpa/mlx5/net/mlx5_vnet.c
+++ b/drivers/vdpa/mlx5/net/mlx5_vnet.c

@@ -1445,7 +1445,10 @@ static virtio_net_ctrl_ack 

handle_ctrl_mac(struct mlx5_vdpa_dev *mvdev, u8 cmd)
      size_t read;
      u8 mac[ETH_ALEN];
  -    pfmdev = pci_get_drvdata(pci_physfn(mvdev->mdev->pdev));
+    pfmdev = mlx5_vf_get_core_dev(mvdev->mdev->pdev);
+    if (!pfmdev)
+        return status;
+
      switch (cmd) {

Yishai/Jason,

I think this patch breaks VPDA over SF.

Did you verify it ?

It seems that you are right, the SF use case requires some different 
handling/path compared to VF.

As this patch is orthogonal to what we try to achieve in this series, we 
may apparently drop it from V3 and let vdpa/mlx5 people come with a 
complete tested patch for both cases on top.

Thanks,
Yishai