Re: [PATCH net V4 3/4] net/mlx5e: SD, Fix missing cleanup on probe error

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] NULL Pointer Dereference in _mlx5e_remove() during suspend/resume error or hot-unplug of a secondary SD device.
--

commit 6088c89e67042c08a966bd363281ba624421bac1
Author: Shay Drory <shayd@nvidia.com>

net/mlx5e: SD, Fix missing cleanup on probe error

This commit adds a missing mlx5_sd_cleanup() call to the probe error path.
The commit message notes a similar gap exists on the resume path leaving
sd == NULL on failure, but defers the fix to a follow-up series.

> diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
> index 5a46870c4b74..e21affd0ffc4 100644
> --- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
> +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c
[ ... ]
> @@ -6912,9 +6912,16 @@ static int mlx5e_probe(struct auxiliary_device *adev,
>  		return err;
>  
>  	actual_adev = mlx5_sd_get_adev(mdev, adev, edev->idx);
> -	if (actual_adev)
> -		return _mlx5e_probe(actual_adev);
> +	if (actual_adev) {
> +		err = _mlx5e_probe(actual_adev);
> +		if (err)
> +			goto sd_cleanup;
> +	}
>  	return 0;
> +
> +sd_cleanup:
> +	mlx5_sd_cleanup(mdev);
> +	return err;
>  }
>  
>  static void _mlx5e_remove(struct auxiliary_device *adev)

Will leaving the resume gap unfixed as mentioned in the commit message cause
a NULL pointer dereference when a secondary device is later removed?

If a secondary device fails resume, sd becomes NULL. Later during removal,
mlx5_sd_get_adev() will check !sd and return the secondary adev.

This causes _mlx5e_remove() to be executed on the secondary adev:

_mlx5e_remove()
{
        struct mlx5_adev *edev = container_of(adev, struct mlx5_adev, adev);
        struct mlx5e_dev *mlx5e_dev = auxiliary_get_drvdata(adev);
        struct net_device *netdev = mlx5e_dev->netdev;
        ...
}

Since _mlx5e_probe() is never called on secondary devices,
auxiliary_get_drvdata(adev) returns NULL. Does this mean accessing
mlx5e_dev->netdev will crash?

Would it be safer to add a NULL check for mlx5e_dev in _mlx5e_remove()
until the proper PM lifecycle fixes are implemented?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260428060111.221086-1-tariqt@nvidia.com?part=3