Re: [net-next 03/14] net/mlx5e: rep, store send to vport rules per peer

[Simon Horman <simon.horman@corigine.com>]

On Wed, May 31, 2023 at 11:01:07PM -0700, Saeed Mahameed wrote:
> From: Mark Bloch <mbloch@nvidia.com>
> 
> Each representor, for each send queue, is holding a
> send_to_vport rule for the peer eswitch.
> 
> In order to support more than one peer, and to map between the peer
> rules and peer eswitches, refactor representor to hold both the peer
> rules and pointer to the peer eswitches.
> This enables mlx5 to store send_to_vport rules per peer, where each
> peer have dedicate index via mlx5_get_dev_index().
> 
> Signed-off-by: Mark Bloch <mbloch@nvidia.com>
> Signed-off-by: Shay Drory <shayd@nvidia.com>
> Reviewed-by: Roi Dayan <roid@nvidia.com>
> Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

...

> @@ -426,15 +437,24 @@ static int mlx5e_sqs2vport_start(struct mlx5_eswitch *esw,
>  		rep_sq->sqn = sqns_array[i];
>  
>  		if (peer_esw) {
> +			int peer_rule_idx = mlx5_get_dev_index(peer_esw->dev);
> +
> +			sq_peer = kzalloc(sizeof(*sq_peer), GFP_KERNEL);
> +			if (!sq_peer)
> +				goto out_sq_peer_err;

Hi Mark and Saeed,

Jumping to out_sq_peer_err will return err.
But err seems to be uninitialised here.

> +
>  			flow_rule = mlx5_eswitch_add_send_to_vport_rule(peer_esw, esw,
>  									rep, sqns_array[i]);
>  			if (IS_ERR(flow_rule)) {
>  				err = PTR_ERR(flow_rule);
> -				mlx5_eswitch_del_send_to_vport_rule(rep_sq->send_to_vport_rule);
> -				kfree(rep_sq);
> -				goto out_err;
> +				goto out_flow_rule_err;
>  			}
> -			rep_sq->send_to_vport_rule_peer = flow_rule;
> +
> +			sq_peer->rule = flow_rule;
> +			sq_peer->peer = peer_esw;
> +			err = xa_insert(&rep_sq->sq_peer, peer_rule_idx, sq_peer, GFP_KERNEL);
> +			if (err)
> +				goto out_xa_err;
>  		}
>  
>  		list_add(&rep_sq->list, &rpriv->vport_sqs_list);
> @@ -445,6 +465,14 @@ static int mlx5e_sqs2vport_start(struct mlx5_eswitch *esw,
>  
>  	return 0;
>  
> +out_xa_err:
> +	mlx5_eswitch_del_send_to_vport_rule(flow_rule);
> +out_flow_rule_err:
> +	kfree(sq_peer);
> +out_sq_peer_err:
> +	mlx5_eswitch_del_send_to_vport_rule(rep_sq->send_to_vport_rule);
> +	xa_destroy(&rep_sq->sq_peer);
> +	kfree(rep_sq);
>  out_err:
>  	mlx5e_sqs2vport_stop(esw, rep);
>  

...