Re: [PATCH] seccomp: allow BPF_MOD ALU instructions

[Kees Cook]

On Sun, Aug 11, 2019 at 10:58:33AM +0200, Paul Chaignon wrote:
> On Fri, Aug 9, 2019 at 8:26 PM Paul Chaignon <paul.chaignon@xxxxxxxxxx> wrote:
> >
> > We need BPF_MOD to match system calls against whitelists encoded as 32-bit
> > bit arrays.  The selection of the syscall's bit in the appropriate bit
> > array requires a modulo operation such that X = 1 << nr % 32.
> 
> Of course, X = 1 << nr & 0x1F, and we can do without BPF_MOD in our case.
> I'll put that on a lack of sleep...

No worries! Changing the dialect of seccomp BPF isn't something I'd like
to do without really good reason since it creates a split in the filter
correctness from userspace (i.e. a filter using BPF_MOD on an older
kernel will fail). So there would need to be a distinct flag set
somewhere, etc. So, if you do end up discovering later you really want
BPF_MOD, we can figure that out, but for now if you can get by with "&",
that would be best. :)

Thanks!

-Kees

> 
> >
> > Signed-off-by: Paul Chaignon <paul.chaignon@xxxxxxxxxx>
> > ---
> >  kernel/seccomp.c | 2 ++
> >  1 file changed, 2 insertions(+)
> >
> > diff --git a/kernel/seccomp.c b/kernel/seccomp.c
> > index 811b4a86cdf6..87de6532ff6d 100644
> > --- a/kernel/seccomp.c
> > +++ b/kernel/seccomp.c
> > @@ -205,6 +205,8 @@ static int seccomp_check_filter(struct sock_filter *filter, unsigned int flen)
> >                 case BPF_ALU | BPF_MUL | BPF_X:
> >                 case BPF_ALU | BPF_DIV | BPF_K:
> >                 case BPF_ALU | BPF_DIV | BPF_X:
> > +               case BPF_ALU | BPF_MOD | BPF_K:
> > +               case BPF_ALU | BPF_MOD | BPF_X:
> >                 case BPF_ALU | BPF_AND | BPF_K:
> >                 case BPF_ALU | BPF_AND | BPF_X:
> >                 case BPF_ALU | BPF_OR | BPF_K:
> > --
> > 2.17.1

-- 
Kees Cook