Re: [PATCH bpf-next] selftests/bpf: Fix htab_update/reenter_update selftest failure

[Alexei Starovoitov]

On Wed, Nov 12, 2025 at 7:08 AM Saket Kumar Bhaskar <skb99@xxxxxxxxxxxxx> wrote:
>
> On Tue, Nov 11, 2025 at 10:35:39AM -0800, Alexei Starovoitov wrote:
> > On Tue, Nov 11, 2025 at 6:33 AM Saket Kumar Bhaskar <skb99@xxxxxxxxxxxxx> wrote:
> > >
> > > On Thu, Nov 06, 2025 at 09:15:39AM -0800, Alexei Starovoitov wrote:
> > > > On Wed, Nov 5, 2025 at 9:26 PM Saket Kumar Bhaskar <skb99@xxxxxxxxxxxxx> wrote:
> > > > >
> > > > > Since commit 31158ad02ddb ("rqspinlock: Add deadlock detection and recovery")
> > > > > the updated path on re-entrancy now reports deadlock via
> > > > > -EDEADLK instead of the previous -EBUSY.
> > > > >
> > > > > The selftest is updated to align with expected errno
> > > > > with the kernel’s current behavior.
> > > > >
> > > > > Signed-off-by: Saket Kumar Bhaskar <skb99@xxxxxxxxxxxxx>
> > > > > ---
> > > > >  tools/testing/selftests/bpf/prog_tests/htab_update.c | 2 +-
> > > > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > > >
> > > > > diff --git a/tools/testing/selftests/bpf/prog_tests/htab_update.c b/tools/testing/selftests/bpf/prog_tests/htab_update.c
> > > > > index 2bc85f4814f4..98d52bb1446f 100644
> > > > > --- a/tools/testing/selftests/bpf/prog_tests/htab_update.c
> > > > > +++ b/tools/testing/selftests/bpf/prog_tests/htab_update.c
> > > > > @@ -40,7 +40,7 @@ static void test_reenter_update(void)
> > > > >         if (!ASSERT_OK(err, "add element"))
> > > > >                 goto out;
> > > > >
> > > > > -       ASSERT_EQ(skel->bss->update_err, -EBUSY, "no reentrancy");
> > > > > +       ASSERT_EQ(skel->bss->update_err, -EDEADLK, "no reentrancy");
> > > >
> > > > Makes sense, but looks like the test was broken for quite some time.
> > > > It fails with
> > > >         /* lookup_elem_raw() may be inlined and find_kernel_btf_id()
> > > > will return -ESRCH */
> > > >         bpf_program__set_autoload(skel->progs.lookup_elem_raw, true);
> > > >         err = htab_update__load(skel);
> > > >         if (!ASSERT_TRUE(!err || err == -ESRCH, "htab_update__load") || err)
> > > >
> > > > before reaching deadlk check.
> > > > Pls make it more robust.
> > > > __pcpu_freelist_pop() might be better alternative then lookup_elem_raw().
> > > >
> > > > pw-bot: cr
> > >
> > > Hi Alexei,
> > >
> > > I tried for __pcpu_freelist_pop, looks like it is not good candidate to
> > > attach fentry for, as it is non traceable:
> > >
> > > trace_kprobe: Could not probe notrace function __pcpu_freelist_pop
> > >
> > > I wasn't able to find any other function for this.
> >
> > alloc_htab_elem() is not inlined for me.
> > bpf_obj_free_fields() would be another option.
> Since alloc_htab_elem() is a static function, wouldn’t its
> inlining behavior be compiler-dependent?

of course. Just like lookup_elem_raw(), but alloc is much bigger
and less likely to be inlined.

> static struct htab_elem *alloc_htab_elem(struct bpf_htab *htab, void *key,
>                                          void *value, u32 key_size, u32 hash,
>                                          bool percpu, bool onallcpus,
>                                          struct htab_elem *old_elem)
>
> When the fentry program is instead attached to bpf_obj_free_fields(),
> the bpf_map_update_elem() call returns 0 rather than -EDEADLK,
> because bpf_obj_free_fields() is not invoked in the bpf_map_update_elem()
> re-entrancy path:

Then make it so. Think what you need to do to make
check_and_free_fields() call it.