Re: [RFC] de-asmify the x86-64 system call slowpath

["H. Peter Anvin" <hpa@zytor.com>]

On 01/26/2014 08:32 PM, Linus Torvalds wrote:
> On Sun, Jan 26, 2014 at 4:22 PM, Al Viro <viro@zeniv.linux.org.uk> wrote:
>>
>> Umm...  Can't uprobe_notify_resume() modify regs as well?
> 
> Probably.
> 
> .. and on the other hand, we should actually be able to use 'sysret'
> for signal handling on x86-64, because while sysret destroys %rcx and
> doesn't allow for returning to odd modes, for calling a signal handler
> I don't think we really care..
> 

Yes, it is the fourth argument register, but we only have three
arguments to a signal handler.  I had to think about that one.

>> While we
>> are at it, when we start using the same thing on 32bit kernels, we'll
>> need to watch out for execve() - the reason why start_thread() sets
>> TIF_NOTIFY_RESUME is to force us away from sysexit path.  IIRC, vm86
>> is another thing to watch out for (same reasons).
> 
> Yes, the 32-bit code I didn't want to touch, partly because I no
> longer have a test-case. And it does end up having some more
> interesting cases.

That is one way to put it.  However, this code is incredibly ugly and
getting it cleaned up would really, really help, of course.

	-hpa