Re: [PATCH v5 1/3] seccomp: kill the seccomp_t typedef

[Will Drewry <wad@chromium.org>]

On Sat, Jan 28, 2012 at 5:13 AM, Cong Wang <amwang@redhat.com> wrote:
> On Fri, 2012-01-27 at 17:47 -0600, Will Drewry wrote:
>> On Fri, Jan 27, 2012 at 5:41 PM, Greg KH <gregkh@suse.de> wrote:
>> > On Fri, Jan 27, 2012 at 05:24:39PM -0600, Will Drewry wrote:
>> >> Replaces the seccomp_t typedef with seccomp_struct to match modern
>> >> kernel style.
>> >>
>> >> Signed-off-by: Will Drewry <wad@chromium.org>
>> >> ---
>> >>  include/linux/sched.h   |    2 +-
>> >>  include/linux/seccomp.h |   10 ++++++----
>> >>  2 files changed, 7 insertions(+), 5 deletions(-)
>> >>
>> >> diff --git a/include/linux/sched.h b/include/linux/sched.h
>> >> index 4032ec1..288b5cb 100644
>> >> --- a/include/linux/sched.h
>> >> +++ b/include/linux/sched.h
>> >> @@ -1418,7 +1418,7 @@ struct task_struct {
>> >>       uid_t loginuid;
>> >>       unsigned int sessionid;
>> >>  #endif
>> >> -     seccomp_t seccomp;
>> >> +     struct seccomp_struct seccomp;
>> >
>> > Isn't 'struct seccomp_struct' a bit redundant?
>> >
>> > How about a simple 'struct seccomp' instead?
>>
>> Works for me - I can't recall why that seemed to make sense (other
>> than the user of similar redundant names elsewhere).
>
> seccomp_struct for a type is okay, but you also have:

Thanks. Looking again, I see the main benefit of seccomp_struct over
seccomp being that it help avoids error.  Ignoring task_struct :),
signal_struct and sighand_structs are a good example.  Since the
member name is "signal" and "sighand", I think this bit of redundancy
helps disambiuguate, but perhaps I'm just being overly conservative.

I'll leave it as is unless Greg or someone else feels strongly :)

> +#define seccomp_struct_init_task(_seccomp) do { } while (0);
> +#define seccomp_struct_fork(_tsk, _orig) do { } while (0);
> +#define seccomp_struct_free_task(_seccomp) do { } while (0);
>
> in patch 2/3, "struct" in these function/macro names is redundant.

Fixed - thanks!