Re: [PATCH bpf v3 2/2] selftests/bpf: Add test to ensure kprobe_multi is not sleepable

[Jiri Olsa]

On Thu, Apr 02, 2026 at 12:50:10AM +0200, Kumar Kartikeya Dwivedi wrote:
> On Wed, 1 Apr 2026 at 21:11, Varun R Mallya <varunrmallya@xxxxxxxxx> wrote:
> >
> > Add a selftest to ensure that kprobe_multi programs cannot be attached
> > using the BPF_F_SLEEPABLE flag. This test succeeds when the kernel
> > rejects attachment of kprobe_multi when the BPF_F_SLEEPABLE flag is set.
> >
> > Signed-off-by: Varun R Mallya <varunrmallya@xxxxxxxxx>
> > ---
> 
> Acked-by: Kumar Kartikeya Dwivedi <memxor@xxxxxxxxx>
> 
> >  .../bpf/prog_tests/kprobe_multi_test.c        | 41 +++++++++++++++++++
> >  .../bpf/progs/kprobe_multi_sleepable.c        | 13 ++++++
> >  2 files changed, 54 insertions(+)
> >  create mode 100644 tools/testing/selftests/bpf/progs/kprobe_multi_sleepable.c
> >
> > diff --git a/tools/testing/selftests/bpf/prog_tests/kprobe_multi_test.c b/tools/testing/selftests/bpf/prog_tests/kprobe_multi_test.c
> > index 78c974d4ea33..f02fec2b6fda 100644
> > --- a/tools/testing/selftests/bpf/prog_tests/kprobe_multi_test.c
> > +++ b/tools/testing/selftests/bpf/prog_tests/kprobe_multi_test.c
> > @@ -10,6 +10,7 @@
> >  #include "kprobe_multi_session_cookie.skel.h"
> >  #include "kprobe_multi_verifier.skel.h"
> >  #include "kprobe_write_ctx.skel.h"
> > +#include "kprobe_multi_sleepable.skel.h"
> >  #include "bpf/libbpf_internal.h"
> >  #include "bpf/hashmap.h"
> >
> > @@ -633,6 +634,44 @@ static void test_attach_write_ctx(void)
> >  }
> >  #endif
> >
> > +static void test_attach_multi_sleepable(void)
> > +{
> > +       struct kprobe_multi_sleepable *skel;
> > +       int err;
> > +
> > +       skel = kprobe_multi_sleepable__open();
> > +       if (!ASSERT_OK_PTR(skel, "kprobe_multi_sleepable__open"))
> > +               return;
> > +
> > +       err = bpf_program__set_flags(skel->progs.handle_kprobe_multi_sleepable,
> > +                                    BPF_F_SLEEPABLE);
> > +       if (!ASSERT_OK(err, "bpf_program__set_flags"))
> > +               goto cleanup;
> > +
> > +       /* Load should succeed even with BPF_F_SLEEPABLE for KPROBE types */
> > +       err = kprobe_multi_sleepable__load(skel);
> > +       if (!ASSERT_OK(err, "kprobe_multi_sleepable__load"))
> > +               goto cleanup;
> > +
> > +       /* Attachment must fail for kprobe.multi + BPF_F_SLEEPABLE.
> > +        * Also chosen a stable symbol to send into opts
> > +        */
> > +       LIBBPF_OPTS(bpf_kprobe_multi_opts, opts);
> > +       const char *sym = "vfs_read";
> > +
> > +       opts.syms = &sym;
> > +       opts.cnt = 1;
> > +
> > +       skel->links.handle_kprobe_multi_sleepable =
> > +               bpf_program__attach_kprobe_multi_opts(skel->progs.handle_kprobe_multi_sleepable,
> > +                                                     NULL, &opts);
> > +       ASSERT_ERR_PTR(skel->links.handle_kprobe_multi_sleepable,
> > +                      "bpf_program__attach_kprobe_multi_opts");
> 
> Nit: While vfs_read will likely remain stable, the check could
> probably be stronger to distinguish an attach error from -EINVAL?
> I added a typo to vfs_read and it still passed, because it failed to
> attach instead of getting rejected on unfixed kernel.
> May not be a big deal since vfs_read is unlikely to break.
> I verified it works by adding bpf_copy_from_user to the program and
> attaching to SYS_PREFIX sys_getpid and invoking the splat though, so
> LGTM otherwise.

why not use bpf_fentry_test2 ? you could also put it in pattern argument
and bypass opts completely (up to you)

also there's test_attach_api_fails test, please move it over there

thanks,
jirka