[MIPS] Fix syscall clone implementation

[MIPS] Fix syscall clone implementation

After compiling lss with Clang/LLVM, two issues came up:

a) clang complains about named register __r7 being listed both as input
   argument and input-output argument, and
b) clang does weird things if input argument is not actually initialized
   before asm block.

This change fixes those two issues by removing __r7 from the input argument
list and by moving __v0 initialization out of asm block.

[petarj, 2015-02-10 15:20:11]

petarj@mips.com changed reviewers:
+ vapier@chromium.org

[petarj, 2015-02-12 13:28:46]

ping.

[petarj, 2015-02-12 13:29:19]

petarj@mips.com changed reviewers:
+ markus@chromium.org

[petarj, 2015-02-18 17:17:08]

On 2015/02/12 13:28:46, petarj wrote:
> ping.

ping.

[Mark Seaborn, 2015-02-18 17:27:40]

On 2015/02/18 17:17:08, petarj wrote:
> On 2015/02/12 13:28:46, petarj wrote:
> > ping.
> 
> ping.

Who are you pinging?  Are you expecting vapier@ to review this?  Do you want me
to review it if vapier@ is not responding?  (When you ping, it's useful to add
information, like what your expectations are...)

[petarj, 2015-02-18 18:02:28]

On 2015/02/18 17:27:40, Mark Seaborn wrote:
> On 2015/02/18 17:17:08, petarj wrote:
> > On 2015/02/12 13:28:46, petarj wrote:
> > > ping.
> > 
> > ping.
> 
> Who are you pinging?  Are you expecting vapier@ to review this?  Do you want
me
> to review it if vapier@ is not responding?  (When you ping, it's useful to add
> information, like what your expectations are...)


I have added Mike (vapier) since he was involved in recent clang/lss issues on
MIPS.
If you have time to review it, I would be more than grateful.
Thanks.

[vapier, 2015-02-20 20:44:09]

it still builds with clang, and the changes look pretty mechanical.  plus the
patch is coming from someone @mips.com ;).

this would be an example where:
- unittests sure would be nice
- using named asm constraints would prevent code shuffling

but so it goes

lgtm

[vapier, 2015-02-20 20:46:49]

committed in: https://code.google.com/p/linux-syscall-support/source/detail?r=34

[vapier, 2015-02-20 20:49:28]

btw, in exchange, could you review
https://code.google.com/p/linux-syscall-support/issues/detail?id=5 ? ;)

[petarj, 2015-02-24 00:21:03]

On 2015/02/20 20:46:49, vapier wrote:
> committed in:
https://code.google.com/p/linux-syscall-support/source/detail?r=34

Thanks. You did not really need to commit it, since I have commit rights as
well. :)