Re: [PATCH net-next 3/3] net: ethernet: am65-cpsw: Error out if Enable TX/RX channel fails

[Vladimir Oltean]

On Mon, Nov 13, 2023 at 01:07:08PM +0200, Roger Quadros wrote:
> k3_udma_glue_enable_rx/tx_chn returns error code on failure.
> Bail out on error while enabling TX/RX channel.
> 
> Signed-off-by: Roger Quadros <rogerq@xxxxxxxxxx>
> ---
>  drivers/net/ethernet/ti/am65-cpsw-nuss.c | 33 +++++++++++++++++++-----
>  1 file changed, 26 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/net/ethernet/ti/am65-cpsw-nuss.c b/drivers/net/ethernet/ti/am65-cpsw-nuss.c
> index 7c440899c93c..340f25bf33b1 100644
> --- a/drivers/net/ethernet/ti/am65-cpsw-nuss.c
> +++ b/drivers/net/ethernet/ti/am65-cpsw-nuss.c
> @@ -372,7 +372,7 @@ static void am65_cpsw_init_port_emac_ale(struct am65_cpsw_port *port);
>  static int am65_cpsw_nuss_common_open(struct am65_cpsw_common *common)
>  {
>  	struct am65_cpsw_host *host_p = am65_common_get_host(common);
> -	int port_idx, i, ret;
> +	int port_idx, i, ret, tx;
>  	struct sk_buff *skb;
>  	u32 val, port_mask;
>  
> @@ -453,13 +453,22 @@ static int am65_cpsw_nuss_common_open(struct am65_cpsw_common *common)
>  		}
>  		kmemleak_not_leak(skb);
>  	}
> -	k3_udma_glue_enable_rx_chn(common->rx_chns.rx_chn);
>  
> -	for (i = 0; i < common->tx_ch_num; i++) {
> -		ret = k3_udma_glue_enable_tx_chn(common->tx_chns[i].tx_chn);
> -		if (ret)
> -			return ret;

Can you comment on the kmemleak_not_leak(skb) call above, and its
relationship to the pre-existing error handling path in am65_cpsw_nuss_common_open()?
I see that the dev_kfree_skb_any() call is being made from am65_cpsw_nuss_rx_cleanup(),
which is only called from am65_cpsw_nuss_common_stop().

So if there are errors during am65_cpsw_nuss_common_open() and
descriptors have already been added to the RX DMA channel, they will not
be removed either from hardware or from software. How does that work?