Re: [PATCH v6 RESEND net-next 4/5] net: enetc: add LSO support for i.MX95 ENETC PF

[Simon Horman <horms@kernel.org>]

On Wed, Dec 04, 2024 at 01:29:31PM +0800, Wei Fang wrote:
> ENETC rev 4.1 supports large send offload (LSO), segmenting large TCP
> and UDP transmit units into multiple Ethernet frames. To support LSO,
> software needs to fill some auxiliary information in Tx BD, such as LSO
> header length, frame length, LSO maximum segment size, etc.
> 
> At 1Gbps link rate, TCP segmentation was tested using iperf3, and the
> CPU performance before and after applying the patch was compared through
> the top command. It can be seen that LSO saves a significant amount of
> CPU cycles compared to software TSO.
> 
> Before applying the patch:
> %Cpu(s):  0.1 us,  4.1 sy,  0.0 ni, 85.7 id,  0.0 wa,  0.5 hi,  9.7 si
> 
> After applying the patch:
> %Cpu(s):  0.1 us,  2.3 sy,  0.0 ni, 94.5 id,  0.0 wa,  0.4 hi,  2.6 si
> 
> Signed-off-by: Wei Fang <wei.fang@nxp.com>
> Reviewed-by: Frank Li <Frank.Li@nxp.com>
> Reviewed-by: Claudiu Manoil <claudiu.manoil@nxp.com>
> ---
> v2: no changes
> v3: use enetc_skb_is_ipv6() helper fucntion which is added in patch 2
> v4: fix a typo
> v5: no changes
> v6: remove error logs from the datapath
> ---
>  drivers/net/ethernet/freescale/enetc/enetc.c  | 259 +++++++++++++++++-
>  drivers/net/ethernet/freescale/enetc/enetc.h  |  15 +
>  .../net/ethernet/freescale/enetc/enetc4_hw.h  |  22 ++
>  .../net/ethernet/freescale/enetc/enetc_hw.h   |  15 +-
>  .../freescale/enetc/enetc_pf_common.c         |   3 +
>  5 files changed, 304 insertions(+), 10 deletions(-)
> 
> diff --git a/drivers/net/ethernet/freescale/enetc/enetc.c b/drivers/net/ethernet/freescale/enetc/enetc.c
> index dafe7aeac26b..82a7932725f9 100644
> --- a/drivers/net/ethernet/freescale/enetc/enetc.c
> +++ b/drivers/net/ethernet/freescale/enetc/enetc.c
> @@ -523,6 +523,226 @@ static void enetc_tso_complete_csum(struct enetc_bdr *tx_ring, struct tso_t *tso
>  	}
>  }
>  
> +static inline int enetc_lso_count_descs(const struct sk_buff *skb)
> +{
> +	/* 4 BDs: 1 BD for LSO header + 1 BD for extended BD + 1 BD
> +	 * for linear area data but not include LSO header, namely
> +	 * skb_headlen(skb) - lso_hdr_len. And 1 BD for gap.
> +	 */
> +	return skb_shinfo(skb)->nr_frags + 4;
> +}
> +
> +static int enetc_lso_get_hdr_len(const struct sk_buff *skb)
> +{
> +	int hdr_len, tlen;
> +
> +	tlen = skb_is_gso_tcp(skb) ? tcp_hdrlen(skb) : sizeof(struct udphdr);
> +	hdr_len = skb_transport_offset(skb) + tlen;

Hi Wei,

I am wondering if packets that are neither TCP nor UDP can be process
by the LSO code added by this patch, and if so, what the implications are.

> +
> +	return hdr_len;
> +}

...