Re: [PATCH v3] vdpa/mlx5: should not activate virtq object when suspended

[Si-Wei Liu <si-wei.liu@oracle.com>]

On 2/15/2023 8:48 PM, Jason Wang wrote:
> On Wed, Feb 15, 2023 at 9:31 AM Si-Wei Liu <si-wei.liu@oracle.com> wrote:
>> Otherwise the virtqueue object to instate could point to invalid address
>> that was unmapped from the MTT:
>>
>>    mlx5_core 0000:41:04.2: mlx5_cmd_out_err:782:(pid 8321):
>>    CREATE_GENERAL_OBJECT(0xa00) op_mod(0xd) failed, status
>>    bad parameter(0x3), syndrome (0x5fa1c), err(-22)
>>
>> Fixes: cae15c2ed8e6 ("vdpa/mlx5: Implement susupend virtqueue callback")
>> Cc: Eli Cohen <elic@nvidia.com>
>> Signed-off-by: Si-Wei Liu <si-wei.liu@oracle.com>
>> Reviewed-by: Eli Cohen <elic@nvidia.com>
>>
>> ---
>> v3: move suspended to struct mlx5_vdpa_dev
>> v2: removed the change for improving warning message
>> ---
>>   drivers/vdpa/mlx5/core/mlx5_vdpa.h | 1 +
>>   drivers/vdpa/mlx5/net/mlx5_vnet.c  | 6 +++++-
>>   2 files changed, 6 insertions(+), 1 deletion(-)
>>
>> diff --git a/drivers/vdpa/mlx5/core/mlx5_vdpa.h b/drivers/vdpa/mlx5/core/mlx5_vdpa.h
>> index 058fbe2..25fc412 100644
>> --- a/drivers/vdpa/mlx5/core/mlx5_vdpa.h
>> +++ b/drivers/vdpa/mlx5/core/mlx5_vdpa.h
>> @@ -96,6 +96,7 @@ struct mlx5_vdpa_dev {
>>          struct mlx5_control_vq cvq;
>>          struct workqueue_struct *wq;
>>          unsigned int group2asid[MLX5_VDPA_NUMVQ_GROUPS];
>> +       bool suspended;
>>   };
>>
>>   int mlx5_vdpa_alloc_pd(struct mlx5_vdpa_dev *dev, u32 *pdn, u16 uid);
>> diff --git a/drivers/vdpa/mlx5/net/mlx5_vnet.c b/drivers/vdpa/mlx5/net/mlx5_vnet.c
>> index 3a6dbbc6..daac3ab 100644
>> --- a/drivers/vdpa/mlx5/net/mlx5_vnet.c
>> +++ b/drivers/vdpa/mlx5/net/mlx5_vnet.c
>> @@ -2411,7 +2411,7 @@ static int mlx5_vdpa_change_map(struct mlx5_vdpa_dev *mvdev,
>>          if (err)
>>                  goto err_mr;
>>
>> -       if (!(mvdev->status & VIRTIO_CONFIG_S_DRIVER_OK))
>> +       if (!(mvdev->status & VIRTIO_CONFIG_S_DRIVER_OK) || mvdev->suspended)
> One more thought,
>
> Does this mean set_map() is forbidden during suspending?
No, it will not. Instead it now allows set_map() to proceed even if 
mapping is shrinking while device is suspended. The "goto err_mr" below 
actually returns 0 when it leaves mlx5_vdpa_change_map().

-Siwei

>   I'm not sure
> this is correct or at least we need restrict in in the vDPA core.
>
> Thanks
>
>>                  goto err_mr;
>>
>>          restore_channels_info(ndev);
>> @@ -2579,6 +2579,7 @@ static int mlx5_vdpa_reset(struct vdpa_device *vdev)
>>          clear_vqs_ready(ndev);
>>          mlx5_vdpa_destroy_mr(&ndev->mvdev);
>>          ndev->mvdev.status = 0;
>> +       ndev->mvdev.suspended = false;
>>          ndev->cur_num_vqs = 0;
>>          ndev->mvdev.cvq.received_desc = 0;
>>          ndev->mvdev.cvq.completed_desc = 0;
>> @@ -2815,6 +2816,8 @@ static int mlx5_vdpa_suspend(struct vdpa_device *vdev)
>>          struct mlx5_vdpa_virtqueue *mvq;
>>          int i;
>>
>> +       mlx5_vdpa_info(mvdev, "suspending device\n");
>> +
>>          down_write(&ndev->reslock);
>>          ndev->nb_registered = false;
>>          mlx5_notifier_unregister(mvdev->mdev, &ndev->nb);
>> @@ -2824,6 +2827,7 @@ static int mlx5_vdpa_suspend(struct vdpa_device *vdev)
>>                  suspend_vq(ndev, mvq);
>>          }
>>          mlx5_vdpa_cvq_suspend(mvdev);
>> +       mvdev->suspended = true;
>>          up_write(&ndev->reslock);
>>          return 0;
>>   }
>> --
>> 1.8.3.1
>>