Re: [PATCH bpf-next 1/2] bpf: Add bpf_lsm_set_bprm_opts helper

From: KP Singh
Date: Mon Nov 16 2020 - 17:53:20 EST


On Mon, Nov 16, 2020 at 11:48 PM KP Singh <kpsingh@xxxxxxxxxxxx> wrote:
>
> [...]
>
> > >
> > > +BPF_CALL_2(bpf_lsm_set_bprm_opts, struct linux_binprm *, bprm, u64, flags)
> > > +{
> >
> > This should also reject invalid flags. I'd rather change this helper from RET_VOID
> > to RET_INTEGER and throw -EINVAL for everything other than BPF_LSM_F_BPRM_SECUREEXEC
> > passed in here including zero so it can be extended in future.
>
> Sounds good, I added:
>
> enum {
> BPF_LSM_F_BPRM_SECUREEXEC = (1ULL << 0),
> + /* Mask for all the currently supported BPRM options */
> + BPF_LSM_F_BRPM_OPTS_MASK = 0x1ULL,
> };
>
> changed the return type to RET_INTEGER as suggested checking for
> invalid flags as:
>
> BPF_CALL_2(bpf_lsm_set_bprm_opts, struct linux_binprm *, bprm, u64, flags)
> {
> +
> + if (flags & !BPF_LSM_F_BRPM_OPTS_MASK)
> + return -EINVAL;
>
> Do let me know if this is okay and I can spin up a v2 with these changes.

Oops this should have been:

if (flags & ~BPF_LSM_F_BRPM_OPTS_MASK)
return -EINVAL;

>
> - KP
>
> >
> > > + bprm->secureexec = (flags & BPF_LSM_F_BPRM_SECUREEXEC);
> > > + return 0;
> > > +}
> > > +
> > > +BTF_ID_LIST_SINGLE(bpf_lsm_set_bprm_opts_btf_ids, struct, linux_binprm)
> > > +
> > > +const static struct bpf_func_proto bpf_lsm_set_bprm_opts_proto = {
> > > + .func = bpf_lsm_set_bprm_opts,
> > > + .gpl_only = false,
> > > + .ret_type = RET_VOID,
> > > + .arg1_type = ARG_PTR_TO_BTF_ID,
> > > + .arg1_btf_id = &bpf_lsm_set_bprm_opts_btf_ids[0],
> > > + .arg2_type = ARG_ANYTHING,
> > > +};
> > > +