Re: [PATCH 5/6] seccomp: add a way to attach a filter via eBPF fd

[Tycho Andersen <tycho.andersen@canonical.com>]

On Wed, Sep 09, 2015 at 08:14:04AM -0700, Alexei Starovoitov wrote:
> On Wed, Sep 09, 2015 at 08:47:24AM -0600, Tycho Andersen wrote:
> > On Tue, Sep 08, 2015 at 05:07:03PM -0700, Kees Cook wrote:
> > >
> > > Yeah, bpf's union looks good. Let's add a "command" flag, though:
> > > 
> > > seccomp(SECCOMP_MODE_FILTER_EBPF, int cmd, union, size);
> > > 
> > > And this cmd could be ADD_FD or something?
> > > 
> > > How's that look?
> > 
> > I think we can drop the size (using the same strategy as bpf() and
> > checking for zeroes at the end), and keep the same signature for
> > seccomp(); so:
> > 
> > seccomp(SECCOMP_MODE_FILTER_EBPF, SECCOMP_ADD_BPF_FD, &union)
> > 
> > Yes, I'll use this in the next version.
> 
> actually bpf() has size as the last argument:
> SYSCALL_DEFINE3(bpf, int, cmd, union bpf_attr __user *, uattr, unsigned int, size)
> perf_event_open() doesn't and size is embedded as one of the fields.
> Both approaches are equivally powerfull from extensitiblity
> point of view. My preference was to keep size as an explicit
> argument.

Yep, sorry that was poorly written. I meant keeping the size as a
member of the struct as Michael originally suggested, mostly to avoid
having to change the signature of seccomp().

Tycho