Re: [PATCHv2 bpf 0/2] bpf: Fix prog_array_map_poke_run map poke update

[Ilya Leoshkevich <iii@linux.ibm.com>]

On Fri, 2023-12-01 at 14:13 +0100, Jiri Olsa wrote:
> On Wed, Nov 29, 2023 at 02:23:04PM +0100, Jiri Olsa wrote:
> > On Tue, Nov 28, 2023 at 11:44:33PM +0100, Ilya Leoshkevich wrote:
> > > On Tue, 2023-11-28 at 10:28 +0100, Jiri Olsa wrote:
> > > > hi,
> > > > this patchset fixes the issue reported in [0].
> > > > 
> > > > For the actual fix in patch 2 I'm changing bpf_arch_text_poke
> > > > to
> > > > allow to skip
> > > > ip address check in patch 1. I considered adding separate
> > > > function
> > > > for that,
> > > > but because each arch implementation is bit different, adding
> > > > extra
> > > > arg seemed
> > > > like better option.
> > > > 
> > > > v2 changes:
> > > >   - make it work for other archs
> > > > 
> > > > thanks,
> > > > jirka
> > > > 
> > > > 
> > > > [0]
> > > > https://syzkaller.appspot.com/bug?extid=97a4fe20470e9bc30810
> > > > ---
> > > > Jiri Olsa (2):
> > > >       bpf: Add checkip argument to bpf_arch_text_poke
> > > >       bpf, x64: Fix prog_array_map_poke_run map poke update
> > > > 
> > > >  arch/arm64/net/bpf_jit_comp.c   |  3 ++-
> > > >  arch/riscv/net/bpf_jit_comp64.c |  5 +++--
> > > >  arch/s390/net/bpf_jit_comp.c    |  3 ++-
> > > >  arch/x86/net/bpf_jit_comp.c     | 24 +++++++++++++-----------
> > > >  include/linux/bpf.h             |  2 +-
> > > >  kernel/bpf/arraymap.c           | 31 +++++++++++--------------
> > > > ------
> > > >  kernel/bpf/core.c               |  2 +-
> > > >  kernel/bpf/trampoline.c         | 12 ++++++------
> > > >  8 files changed, 39 insertions(+), 43 deletions(-)
> > > 
> > > Would it be possible to add a minimized version of the reproducer
> > > as a
> > > testcase?
> > 
> > there's reproducer I used in here:
> >   https://syzkaller.appspot.com/text?tag=ReproC&x=1397180f680000
> > 
> > I can try, but not sure I'll be able to come up with something that
> > would fit as testcase.. I'll check
> 
> the test below reproduces it for me.. the only tricky part is that
> I need to repeat the loop 10 times to trigger that on my setup..
> which is not terrible, but not great for a test I think
> 
> jirka

The test looks useful to me. I think having magic repetition counts
like this 10 here is almost inevitable when trying to reproduce race
conditions. The test also runs quickly for me. You can have my

Acked-by: Ilya Leoshkevich <iii@linux.ibm.com>

in case you decide to make a formal patch.