Re: [PATCH v2 1/3] syscalls,x86 Expose arch_prctl on x86-32.

[Kyle Huey]

On Wed, Sep 14, 2016 at 2:29 PM, Dave Hansen
<dave.hansen@xxxxxxxxxxxxxxx> wrote:
> On 09/14/2016 02:01 PM, Kyle Huey wrote:
>> Signed-off-by: Kyle Huey <khuey@xxxxxxxxxxxx>
>> ---
>>  arch/x86/entry/syscalls/syscall_32.tbl |  1 +
>>  arch/x86/kernel/process.c              | 80 ++++++++++++++++++++++++++++++++++
>>  arch/x86/kernel/process_64.c           | 66 ----------------------------
>>  3 files changed, 81 insertions(+), 66 deletions(-)
>
> Could you explain a bit about what is going on here?  Is it just a plain
> old code move, _why_ you had to do it this way, etc...?

Sure.  In the subsequent patches in this series I add an arch_prctl
that is available for both 64 and 32 bit programs/kernels.  Since
process_64.c is only built for 64 bit kernels, this syscall can't stay
there anymore.

It's not quite a plain move.  To leave the existing arch_prctls only
accessible to 64 bit callers, I added the is_32 bit and the four early
returns for each existing ARCH_BLAH.  These cases are now
conditionally compiled out in a 32 bit kernel, so we only have to
handle the 32 bit process on a 64 bit kernel case at runtime.

I considered doing this instead with a compat wrapper for the syscall
on 32 bit systems that would filter these arch_prctls before getting
to do_arch_prctl. I didn't see any prior art for it, so decided not to
proceed that way.

- Kyle