Re: [PATCH v2 2/2] net: Add fsl_enetc network driver support

[Ahmad Fatoum <a.fatoum@pengutronix.de>]

Hello Sascha,

On 04.01.24 13:18, Sascha Hauer wrote:
> On Thu, Jan 04, 2024 at 12:22:41PM +0100, Sascha Hauer wrote:
>> This adds support for the fsl_enetc network controller found on several
>> Layerscape SoCs. The code is originally from U-Boot-2023.10-rc1.
>>
>> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
>> ---
>>
>> Notes:
>>     Changes since v1:
>>     
>>     - Use get_unaligned to access unaligned pointers
>>     - use physical address returned by dma_alloc_coherent()
>>     - implement struct pci_driver::remove hook
>>     - do not printf in poller function
>>     - use write[lwq] to access rx/tx rings to avoid dmb()
> 
> Actually this last part is missing, here it is:
> 
> ---------------------------------8<---------------------------
> 
> From 3b312b7d501be3b0049de27582a26b19fdaf84e1 Mon Sep 17 00:00:00 2001
> From: Sascha Hauer <s.hauer@pengutronix.de>
> Date: Wed, 3 Jan 2024 20:55:39 +0100
> Subject: [PATCH] fixup! net: Add fsl_enetc network driver support
> 
> ---
>  drivers/net/fsl_enetc.c | 19 +++++++------------
>  1 file changed, 7 insertions(+), 12 deletions(-)
> 
> diff --git a/drivers/net/fsl_enetc.c b/drivers/net/fsl_enetc.c
> index eb90c526bb..07e9a6b7c8 100644
> --- a/drivers/net/fsl_enetc.c
> +++ b/drivers/net/fsl_enetc.c
> @@ -454,13 +454,10 @@ static int enetc_send(struct eth_device *edev, void *packet, int length)
>  	dma = dma_map_single(priv->dev, packet, length, DMA_TO_DEVICE);
>  
>  	/* prepare Tx BD */
> -	memset(&priv->enetc_txbd[pi], 0x0, sizeof(struct enetc_tx_bd));
> -	priv->enetc_txbd[pi].addr = cpu_to_le64(dma);
> -	priv->enetc_txbd[pi].buf_len = cpu_to_le16(length);
> -	priv->enetc_txbd[pi].frm_len = cpu_to_le16(length);
> -	priv->enetc_txbd[pi].flags = cpu_to_le16(ENETC_TXBD_FLAGS_F);
> -
> -	dmb();
> +	writeq(cpu_to_le64(dma), &priv->enetc_txbd[pi].addr);
> +	writew(cpu_to_le16(length), &priv->enetc_txbd[pi].buf_len);
> +	writew(cpu_to_le16(length), &priv->enetc_txbd[pi].frm_len);
> +	writew(cpu_to_le16(ENETC_TXBD_FLAGS_F), &priv->enetc_txbd[pi].flags);

The cpu_to_le* can be dropped because writew/writeq always do conversion
to little endian.

>  
>  	/* send frame: increment producer index */
>  	pi = (pi + 1) % txr->bd_count;
> @@ -508,9 +505,7 @@ static int enetc_recv(struct eth_device *edev)
>  	if (!ENETC_RXBD_STATUS_R(status))
>  		return 0;
>  
> -	dmb();
> -
> -	len = le16_to_cpu(priv->enetc_rxbd[pi].r.buf_len);
> +	len = le16_to_cpu(readw(&priv->enetc_rxbd[pi].r.buf_len));

Same.

>  
>  	dev_dbg(&edev->dev, "RxBD[%d]: len=%d err=%d pkt=0x%p\n", pi, len,
>  		  ENETC_RXBD_STATUS_ERRORS(status), pkg);
> @@ -520,8 +515,8 @@ static int enetc_recv(struct eth_device *edev)
>  	dma_sync_single_for_device(priv->dev, priv->rx_pkg_phys[pi], PKTSIZE, DMA_FROM_DEVICE);
>  
>  	/* BD clean up and advance to next in ring */
> -	memset(&priv->enetc_rxbd[pi], 0, sizeof(union enetc_rx_bd));
> -	priv->enetc_rxbd[pi].w.addr = priv->rx_pkg_phys[pi];
> +	memset_io(&priv->enetc_rxbd[pi], 0, sizeof(union enetc_rx_bd));
> +	writeq(priv->rx_pkg_phys[pi], &priv->enetc_rxbd[pi].w.addr);
>  	rxr->next_prod_idx = (pi + 1) % rxr->bd_count;
>  	ci = (ci + 1) % rxr->bd_count;
>  	rxr->next_cons_idx = ci;
LGTM with changes applied:

Reviewed-by: Ahmad Fatoum <a.fatoum@pengutronix.de>

-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |