Re: [PATCH v4 net-next 1/9] i40e: don't reserve excessive XDP_PACKET_HEADROOM on XSK Rx to skb

On 09/12/2021 18.33, Alexander Lobakin wrote:

From: Jesper Dangaard Brouer <redacted>
Date: Thu, 9 Dec 2021 09:19:46 +0100

quoted

On 08/12/2021 15.06, Alexander Lobakin wrote:

quoted

{__,}napi_alloc_skb() allocates and reserves additional NET_SKB_PAD
+ NET_IP_ALIGN for any skb.
OTOH, i40e_construct_skb_zc() currently allocates and reserves
additional `xdp->data - xdp->data_hard_start`, which is
XDP_PACKET_HEADROOM for XSK frames.
There's no need for that at all as the frame is post-XDP and will
go only to the networking stack core.

I disagree with this assumption, that headroom is not needed by netstack.
Why "no need for that at all" for netstack?

napi_alloc_skb() in our particular case will reserve 64 bytes, it is
sufficient for {TCP,UDP,SCTP,...}/IPv{4,6} etc.

My bad, I misunderstood you. I now see (looking at code) that (as you say) 64 bytes of headroom *is* reserved (in bottom of __napi_alloc_skb).
Thus, the SKB *do* have headroom, so this patch should be fine.

Acked-by: Jesper Dangaard Brouer <redacted>

Do watch out that 64 bytes is not always enough. Notice the define LL_MAX_HEADER and MAX_HEADER in include/linux/netdevice.h (that tries to determine worst-case header length) which is above 64 bytes. It is also affected by HyperV and WiFi configs.