> Thank you for your contribution! Sashiko AI review found more severe > issue(s) in other patches of this series. This mail is only an FYI, in case > you want to address the nit picks below when you respin. > > Low priority / nit picks for your consideration: > > - [Low] The kerneldoc for the newly exported __xdp_build_skb_from_buff() > (net/core/xdp.c:622-633) says only "Protocol parsing is left to the… > > -- > > net: xdp: add __xdp_build_skb_from_buff() helper > > This patch splits xdp_build_skb_from_buff() in two. The new exported > __xdp_build_skb_from_buff() builds the skb without calling > eth_type_trans(). A static inline wrapper in include/net/xdp.h keeps the > old behaviour and still calls eth_type_trans(). The goal is to let stmmac > process the L2 header (software VLAN) before protocol parsing. > > > diff --git a/net/core/xdp.c b/net/core/xdp.c > > index 1d679e8fd6493..470bf90d93ef2 100644 > > --- a/net/core/xdp.c > > +++ b/net/core/xdp.c > > @@ -620,17 +620,18 @@ void xdp_warn(const char *msg, const char *func, const int line) > > EXPORT_SYMBOL_GPL(xdp_warn); > > > > /** > > - * xdp_build_skb_from_buff - create an skb from &xdp_buff > > + * __xdp_build_skb_from_buff - create an skb from &xdp_buff > > * @xdp: &xdp_buff to convert to an skb > > * > > * Perform common operations to create a new skb to pass up the stack from > > * &xdp_buff: allocate an skb head from the NAPI percpu cache, initialize > > * skb data pointers and offsets, set the recycle bit if the buff is > > - * PP-backed, Rx queue index, protocol and update frags info. > > + * PP-backed, record the Rx queue index and update frags info. Protocol > > + * parsing is left to the caller. > > [Severity: Low] > Does "Protocol parsing is left to the caller" fully describe what the > caller now has to do? > > Moving eth_type_trans() out of __xdp_build_skb_from_buff() also moves out > the other work it does: > > net/ethernet/eth.c:eth_type_trans() { > skb->dev = dev; > skb_reset_mac_header(skb); > > eth = eth_skb_pull_mac(skb); > eth_skb_pkt_type(skb, dev); > ... > } > > So the skb returned by __xdp_build_skb_from_buff() has skb->dev == NULL > and no mac header, and protocol and pkt_type are not set. This is true > even though rxq->dev is available inside the helper. > > The peer helper __xdp_build_skb_from_frame() still sets these: > > /* Essential SKB info: protocol and skb->dev */ > skb->protocol = eth_type_trans(skb, dev); > > As a result, the two __xdp_build_skb_* helpers now return different > minimal skb state. > > Nothing breaks today. The only in-tree user of the underscore variant > later in the series is stmmac_rx(), and it calls > eth_type_trans(skb, priv->dev) before napi_gro_receive(). A future caller > that does its own L2 handling and reads "protocol parsing" as just > skb->protocol could pass an skb with no dev to napi_gro_receive() or > netif_receive_skb(). > > Could the kerneldoc say that the caller must also set skb->dev and reset > the mac header, for example by calling eth_type_trans()? > > A smaller point: the public xdp_build_skb_from_buff() wrapper in > include/net/xdp.h no longer has any kerneldoc. The kerneldoc for > xdp_build_skb_from_zc() still refers readers to it: > > * Similar to xdp_build_skb_from_buff(), but for XSk frames. Allocate an skb > > [ ... ] I will fix it in v11. Regards, Lorenzo > > -- > Sashiko AI review · https://netdev-ai.bots.linux.dev/sashiko/#/patchset/20261112-stmmac-rx-mb-v10-0-d3d04b65c6bc%40oss.qualcomm.com