Re: linux-next: manual merge of the akpm-current tree with the tip tree

[Stephen Rothwell <sfr@canb.auug.org.au>]

Hi Andrea,

On Wed, 29 Jul 2015 19:12:56 +0200 Andrea Arcangeli <aarcange@redhat.com> wrote:
>
> On Tue, Jul 28, 2015 at 04:00:15PM +1000, Stephen Rothwell wrote:
> >  -359	i386	userfaultfd		sys_userfaultfd
> > ++374	i386	userfaultfd		sys_userfaultfd
> 
> Do I understand correctly the syscall number of userfaultfd for x86
> 32bit has just changed from 359 to 374? Appreciated that you CCed me
> on such a relevant change to be sure I didn't miss it.
> 
> Then the below is needed as well.

I have added the below patch to linux-next from today.

> One related question: is it ok to ship kernels in production right now
> with the userfaultfd syscall number 374 for x86 32bit ABI (after the
> above change) and 323 for x86-64 64bit ABI, with these syscalls number
> registered in linux-next or it may keep changing like it has just
> happened? I refer only to userfaultfd syscalls of x86 32bit and x86-64
> 64bit, not all other syscalls in linux-next.

These numbers are certainly not in any way official, they are just the
result of my merge conflict fixup.  So, yes, they could change again if
someone adds another new syscall to any tree but Andrew's.

> Of course, I know full well that the standard answer is no, and in
> fact the above is an expected and fine change. In other words what I'm
> really asking is if I wonder if I could get an agreement here that
> from now on, the syscall number of userfaultfd for x86 32bit and
> x86-64 64bit won't change anymore in linux-next and it's already
> reserved just like if it was already upstream.

Like Thomas said, send a patch to the x86 maintainers.  I suspect (if
the rest of the implementation needs to stay in Andrew's tree) that it
could be a simple as a patch to the syscall tables using ni_syscall and
a comment.  Thomas?

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au