Re: [RFC,PATCH 1/2] seccomp_filters: system call filtering using BPF

[Will Drewry <wad@chromium.org>]

On Mon, Jan 16, 2012 at 12:37 PM, Oleg Nesterov <oleg@redhat.com> wrote:
> On 01/13, Will Drewry wrote:
>>
>> On Fri, Jan 13, 2012 at 1:01 PM, Will Drewry <wad@chromium.org> wrote:
>> > On Fri, Jan 13, 2012 at 11:31 AM, Oleg Nesterov <oleg@redhat.com> wrote:
>> >>
>> >> Me too. I see no point in using user_regs_struct.
>> >
>> > I'll rev the change to use pt_regs and drop all the helper code.  If
>> > no one says otherwise, that certainly seems ideal from a performance
>> > perspective, and I see pt_regs exported to userland along with ptrace
>> > abi register offset macros.
>>
>> On second thought, pt_regs is scary :)
>>
>> From looking at
>>   http://lxr.linux.no/linux+v3.2.1/arch/x86/include/asm/syscall.h#L97
>> and ia32syscall enty code, it appears that for x86, at least, the
>> pt_regs for compat processes will be 8 bytes wide per register on the
>> stack.  This means if a self-filtering 32-bit program runs on a 64-bit host in
>> IA32_EMU, its filters will always index into pt_regs incorrectly.
>
> Yes, thanks, I forgot about compat tasks again. But this is easy, just
> we need regs_64_to_32().

Yup - we could make the assumption that is_compat_task is always
32-bit and the pt_regs is always 64-bit, then copy_and_truncate with
regs_64_to_32.  Seems kinda wonky though :/

> Doesn't matter. I think Indan has a better suggestion.

I disagree, but perhaps I'm not fully understanding!

Thanks!
will