Re: [PATCH net 1/1] net: fec: using the standard return codes when xdp xmit errors

[Simon Horman]

On Tue, May 09, 2023 at 02:38:45PM -0500, Shenwei Wang wrote:
> The existing logic did not properly handle unsuccessful xdp xmit frames,
> this patch revises the logic to return standard error codes (-EBUSY or
> -ENOMEM) on unsuccessful transmit.

Hi Shenwei,

I'm not sure that changing the calling convention is the fix here.
The caller seems to get it right both before and after this part
of this patch change. So this seems to be a cleanup rather than a fix.

> Start the xmit of the frame immediately right after configuring the
> tx descriptors.

This is the money part of this change. I think it warrants more
explanation.

If there are two changes to be made, they probably belong in two
patches.

> 
> Fixes: e8a17397180f ("net: fec: correct the counting of XDP sent frames")

I think this tag should be:

Fixes: 26312c685ae0 ("net: fec: correct the counting of XDP sent frames")

> Signed-off-by: Shenwei Wang <shenwei.wang@xxxxxxx>
> ---
>  drivers/net/ethernet/freescale/fec_main.c | 25 ++++++++++++++---------
>  1 file changed, 15 insertions(+), 10 deletions(-)
> 
> diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c
> index 42ec6ca3bf03..438fc1c3aea2 100644
> --- a/drivers/net/ethernet/freescale/fec_main.c
> +++ b/drivers/net/ethernet/freescale/fec_main.c
> @@ -3798,8 +3798,7 @@ static int fec_enet_txq_xmit_frame(struct fec_enet_private *fep,
>  	entries_free = fec_enet_get_free_txdesc_num(txq);
>  	if (entries_free < MAX_SKB_FRAGS + 1) {
>  		netdev_err(fep->netdev, "NOT enough BD for SG!\n");
> -		xdp_return_frame(frame);
> -		return NETDEV_TX_BUSY;
> +		return -EBUSY;
>  	}
>  
>  	/* Fill in a Tx ring entry */
> @@ -3813,7 +3812,7 @@ static int fec_enet_txq_xmit_frame(struct fec_enet_private *fep,
>  	dma_addr = dma_map_single(&fep->pdev->dev, frame->data,
>  				  frame->len, DMA_TO_DEVICE);
>  	if (dma_mapping_error(&fep->pdev->dev, dma_addr))
> -		return FEC_ENET_XDP_CONSUMED;
> +		return -ENOMEM;
>  
>  	status |= (BD_ENET_TX_INTR | BD_ENET_TX_LAST);
>  	if (fep->bufdesc_ex)
> @@ -3835,6 +3834,11 @@ static int fec_enet_txq_xmit_frame(struct fec_enet_private *fep,
>  	index = fec_enet_get_bd_index(last_bdp, &txq->bd);
>  	txq->tx_skbuff[index] = NULL;
>  
> +	/* Make sure the updates to rest of the descriptor are performed before
> +	 * transferring ownership.
> +	 */
> +	wmb();
> +
>  	/* Send it on its way.  Tell FEC it's ready, interrupt when done,
>  	 * it's the last BD of the frame, and to put the CRC on the end.
>  	 */
> @@ -3844,8 +3848,15 @@ static int fec_enet_txq_xmit_frame(struct fec_enet_private *fep,
>  	/* If this was the last BD in the ring, start at the beginning again. */
>  	bdp = fec_enet_get_nextdesc(last_bdp, &txq->bd);
>  
> +	/* Make sure the update to bdp and tx_skbuff are performed before
> +	 * txq->bd.cur.
> +	 */
> +	wmb();
>  	txq->bd.cur = bdp;
>  
> +	/* Trigger transmission start */
> +	writel(0, txq->bd.reg_desc_active);
> +
>  	return 0;
>  }
>  
> @@ -3869,17 +3880,11 @@ static int fec_enet_xdp_xmit(struct net_device *dev,
>  	__netif_tx_lock(nq, cpu);
>  
>  	for (i = 0; i < num_frames; i++) {
> -		if (fec_enet_txq_xmit_frame(fep, txq, frames[i]) != 0)
> +		if (fec_enet_txq_xmit_frame(fep, txq, frames[i]) < 0)
>  			break;
>  		sent_frames++;
>  	}
>  
> -	/* Make sure the update to bdp and tx_skbuff are performed. */
> -	wmb();
> -
> -	/* Trigger transmission start */
> -	writel(0, txq->bd.reg_desc_active);
> -
>  	__netif_tx_unlock(nq);
>  
>  	return sent_frames;
> -- 
> 2.34.1
> 
>