Re: [PATCH bpf 0/2] Fix some corner cases in xskxceiver

[Larysa Zaremba]

On Fri, Feb 06, 2026 at 09:39:16AM -0800, Alexei Starovoitov wrote:
> On Tue, Feb 3, 2026 at 8:18 AM Larysa Zaremba <larysa.zaremba@xxxxxxxxx> wrote:
> >
> > While working on XDP and AF_XDP support for ixgbevf driver,
> > I came across two distinct problems that caused tests to fail
> > when they shouldn't have.
> 
> Applied.
> 
> What is the status of kernel side fixes?
> 
> Running test_xsk.sh I see plenty of
> 
> [   93.609026] BUG: Bad page state in process xskxceiver  pfn:115920
> ...
> [   93.633927]  bad_page.cold+0xd0/0x140
> [   93.633929]  __free_frozen_pages+0x85a/0x1150
> [   93.633935]  bpf_xdp_adjust_tail+0x808/0x1120


I believe the call trace that you see when running test_xsk.sh is due to test 
24: XDP_ADJUST_TAIL_SHRINK_MULTI_BUFF. I have this call trace too, when running 
automated veth test and when running xskxceiver in skb mode on a page-pool-using 
driver (WiP ixgbevf). The problem is specifically in the skb mode.

The patchset I have recently submitted fixes tail-growing in multi-buffer 
packets, veth is not affected. Those issues are weirdly not related in any way.

Maciej has submitted a patchset that fixes the skb mode some time ago [0], I can 
confirm it gets rid of the call traces.

[0] 
https://lore.kernel.org/bpf/20251029221315.2694841-1-maciej.fijalkowski@xxxxxxxxx/#t