RE: [PATCH net v6 1/3] igb: do not let XDP frames span more than one Rx buffer

From: Tjerk Kusters

Date: Fri Nov 20 2026 - 06:08:49 EST


> - [Medium] igb_xdp_setup() now calls netdev_update_features(dev) without
> first checking that the device is running or awake.

Right. In v7, igb_xdp_setup() resumes a runtime-suspended device before
it turns rx-all back on, as ethtool does before it changes features, and
leaves the features as they are if the resume fails. Tested on an I350:
with the interface down and runtime suspended, removing XDP with rx-all
requested resumes the device and turns rx-all on, and the link comes up
normally afterwards.

> - [Medium] igb_xdp_setup() ignores the return value of igb_open() and
> then calls the new netdev_update_features(dev) unconditionally.

Right. In v7 rx-all is not turned back on when igb_open() failed. The
return value of igb_xdp_setup() stays 0 as before: the program is
already removed at that point, and returning an error would make the
core's XDP state disagree with the driver's: dev_xdp_install() would
keep the old program in its state, and dev_xdp_detach_link() and
dev_xdp_uninstall() warn on an error.

> - [High] This problem wasn't introduced by this patch, but
> igb_set_rx_mode() runs without netif_addr_lock_bh() from several paths…

That is older than this series. This patch only stops igb_set_rx_mode()
from writing the receive length limit. I listed it in the v7 cover
letter as an older issue.

> - [High] This problem wasn't introduced by this patch, but
> igb_set_vf_rlpml() takes the VF-supplied msgbuf[1] (u32) into 'int…

Fixed separately:
https://lore.kernel.org/all/20261104-igb-vf-rlpml-v1-1-729621b60eee@xxxxxxxx/

> - [High] This problem wasn't introduced by this patch, but on 2 KiB
> build_skb Rx rings (PAGE_SIZE < 8192, legacy-rx off,…

That is the first older issue in the cover letter. It dates from the
XDP support and needs a change of the build_skb buffer layout, which
affects all igb users, so I left it out of this series.

pw-bot: cr