[RFC,bpf-next,0/4] bpf_panic() helper

Message

Artem Savkov July 11, 2022, 8:32 a.m. UTC

eBPF is often used for kernel debugging, and one of the widely used and
powerful debugging techniques is post-mortem debugging with a full memory dump.
Triggering a panic at exactly the right moment allows the user to get such a
dump and thus a better view at the system's state. This patchset adds
bpf_panic() helper to do exactly that.

I realize that even though there are multiple guards present, a helper like
this is contrary to BPF being "safe", so this is sent as RFC to have a
discussion on whether adding destructive capabilities is deemed acceptable.

Artem Savkov (4):
  bpf: add a sysctl to enable destructive bpf helpers
  bpf: add BPF_F_DESTRUCTIVE flag for BPF_PROG_LOAD
  bpf: add bpf_panic() helper
  selftests/bpf: bpf_panic selftest

 include/linux/bpf.h                           |   8 +
 include/uapi/linux/bpf.h                      |  13 ++
 kernel/bpf/core.c                             |   1 +
 kernel/bpf/helpers.c                          |  13 ++
 kernel/bpf/syscall.c                          |  33 +++-
 kernel/bpf/verifier.c                         |   7 +
 kernel/trace/bpf_trace.c                      |   2 +
 tools/include/uapi/linux/bpf.h                |  13 ++
 .../selftests/bpf/prog_tests/bpf_panic.c      | 144 ++++++++++++++++++
 9 files changed, 233 insertions(+), 1 deletion(-)
 create mode 100644 tools/testing/selftests/bpf/prog_tests/bpf_panic.c

Comments

Jiri Olsa July 11, 2022, 10:51 a.m. UTC | #1

On Mon, Jul 11, 2022 at 10:32:16AM +0200, Artem Savkov wrote:
> eBPF is often used for kernel debugging, and one of the widely used and
> powerful debugging techniques is post-mortem debugging with a full memory dump.
> Triggering a panic at exactly the right moment allows the user to get such a
> dump and thus a better view at the system's state. This patchset adds
> bpf_panic() helper to do exactly that.

FWIW I was asked for such helper some time ago from Daniel Vacek, cc-ed

jirka

> 
> I realize that even though there are multiple guards present, a helper like
> this is contrary to BPF being "safe", so this is sent as RFC to have a
> discussion on whether adding destructive capabilities is deemed acceptable.
> 
> Artem Savkov (4):
>   bpf: add a sysctl to enable destructive bpf helpers
>   bpf: add BPF_F_DESTRUCTIVE flag for BPF_PROG_LOAD
>   bpf: add bpf_panic() helper
>   selftests/bpf: bpf_panic selftest
> 
>  include/linux/bpf.h                           |   8 +
>  include/uapi/linux/bpf.h                      |  13 ++
>  kernel/bpf/core.c                             |   1 +
>  kernel/bpf/helpers.c                          |  13 ++
>  kernel/bpf/syscall.c                          |  33 +++-
>  kernel/bpf/verifier.c                         |   7 +
>  kernel/trace/bpf_trace.c                      |   2 +
>  tools/include/uapi/linux/bpf.h                |  13 ++
>  .../selftests/bpf/prog_tests/bpf_panic.c      | 144 ++++++++++++++++++
>  9 files changed, 233 insertions(+), 1 deletion(-)
>  create mode 100644 tools/testing/selftests/bpf/prog_tests/bpf_panic.c
> 
> -- 
> 2.35.3
>

Daniel Vacek Aug. 1, 2022, 1:58 p.m. UTC | #2

On Mon, Jul 11, 2022 at 12:51 PM Jiri Olsa <[e-mail]> wrote:
>
> On Mon, Jul 11, 2022 at 10:32:16AM +0200, Artem Savkov wrote:
> > eBPF is often used for kernel debugging, and one of the widely used and
> > powerful debugging techniques is post-mortem debugging with a full memory dump.
> > Triggering a panic at exactly the right moment allows the user to get such a
> > dump and thus a better view at the system's state. This patchset adds
> > bpf_panic() helper to do exactly that.
>
> FWIW I was asked for such helper some time ago from Daniel Vacek, cc-ed

Nice :-)
This is totally welcome. Though, IIRC, I was asking if I could do a
NULL pointer dereference within perf probe (or ftrace) back then.
Still, the outcome is similar. So kudos to Artem.

--nX

> jirka
>
> >
> > I realize that even though there are multiple guards present, a helper like
> > this is contrary to BPF being "safe", so this is sent as RFC to have a
> > discussion on whether adding destructive capabilities is deemed acceptable.
> >
> > Artem Savkov (4):
> >   bpf: add a sysctl to enable destructive bpf helpers
> >   bpf: add BPF_F_DESTRUCTIVE flag for BPF_PROG_LOAD
> >   bpf: add bpf_panic() helper
> >   selftests/bpf: bpf_panic selftest
> >
> >  include/linux/bpf.h                           |   8 +
> >  include/uapi/linux/bpf.h                      |  13 ++
> >  kernel/bpf/core.c                             |   1 +
> >  kernel/bpf/helpers.c                          |  13 ++
> >  kernel/bpf/syscall.c                          |  33 +++-
> >  kernel/bpf/verifier.c                         |   7 +
> >  kernel/trace/bpf_trace.c                      |   2 +
> >  tools/include/uapi/linux/bpf.h                |  13 ++
> >  .../selftests/bpf/prog_tests/bpf_panic.c      | 144 ++++++++++++++++++
> >  9 files changed, 233 insertions(+), 1 deletion(-)
> >  create mode 100644 tools/testing/selftests/bpf/prog_tests/bpf_panic.c
> >
> > --
> > 2.35.3
> >
>