Re: [PATCH bpf 1/1] selftests/bpf: Enable BLK_DEV_NBD for raw_tp_writable_reject_nbd_invalid

[sun jian]

On Tue, Jul 14, 2026 at 11:32 AM Shung-Hsi Yu <shung-hsi.yu@xxxxxxxx> wrote:
>
> On Mon, Jul 13, 2026 at 01:47:18PM -0700, Eduard Zingerman wrote:
> > On Mon, 2026-07-13 at 14:35 +0800, Shung-Hsi Yu wrote:
> > > The raw_tp_writable_reject_nbd_invalid test relies on availability of the
> > > nbd_send_request tracepoint, which is only present if the selftest kernel is
> > > built with CONFIG_BLK_DEV_NBD=y and the kernel built from current BPF selftests
> > > config lacks.
> > >
> > > Without it, the bpf_raw_tracepoint_open() call always returns with -2, leaving
> > > raw_tp_writable_reject_nbd_invalid test always passing without exercising the
> > > checks bpf_probe_register().
> > >
> > > Cc: <stable@xxxxxxxxxxxxxxx> # 5.2.0
> > > Link: https://lore.kernel.org/bpf/alRtilWhKw4zzMkI@u94a
> > > Signed-off-by: Shung-Hsi Yu <shung-hsi.yu@xxxxxxxx>
> > > ---
> > > Not sure if fixes tag is the right thing to use here, so use the cc
> > > stable tag instead to get this config change propogated to other stable
> > > branches to make stable BPF CI's job easier.
> >
> > Shung-Hsi,
> >
> > Thank you for figuring this out.
> > I'd suggest we switch to bpf_testmod_test_writable_bare_tp() [1]
> > from the test module to avoid the config dependency and let
> > Sun pack all of this as a single patch-set to simplify backports
> > (if such are necessary). Wdyt?
>
> Make sense, that's probably for the better.
>

Hi Shung-Hsi and Eduard,

Thanks for tracking down the CONFIG_BLK_DEV_NBD false
pass and for the test-module suggestion.

I'll fold everything into the v5 series:

1. Keep only the necessary effective-start check in the verifier fix,
dropping the BPF_MAX_VAR_OFF and size < 0 checks as Eduard suggested.
2. Move the attach tests to bpf_testmod_test_writable_bare_tp(), use
subtests, and remove the NBD configuration dependency.
3. Add the PTR_TO_BUF negative-offset verifier case requested by Eduard.

> pw-bot: changes-requested
>
> @Sun can you make sure to include the follow tag for [1] when you send
> it? Should give a better guarantee that is will be picked up by AUTOSEL.
> Thanks!
>
>   Cc: <stable@xxxxxxxxxxxxxxx> # 5.2.0

Will do.

Best Regards,
Sun Jian