Re: [PATCH 15/15] x86: use non-set_fs based maccess routines

[Christoph Hellwig]

On Wed, May 06, 2020 at 10:51:51AM -0700, Linus Torvalds wrote:
> My private tree no longer has those __get/put_user_size() things,
> because "unsafe_get/put_user()" is the only thing that remains with my
> conversion to asm goto.
> 
> And we're actively trying to get rid of the whole __get_user() mess.
> Admittedly "__get_user_size()" is just the internal helper that
> doesn't have the problem, but it really is an internal helper for a
> legacy operation, and the new op that uses it is that
> "unsafe_get_user()".
> 
> Also, because you use __get_user_size(), you then have to duplicate
> the error handling logic that we already have in unsafe_get_user().
> 
> IOW - is there some reason why you didn't just make these use
> "unsafe_get/put_user()" directly, and avoid both of those issues?

That was the first prototype, and or x86 it works great, just the
__user cases in maccess.c are a little ugly.  And they point to
the real problem - for architectures like sparc and s390 that use
an entirely separate address space for the kernel vs userspace
I dont think just use unsafe_{get,put}_user will work, as they need
different instructions.

Btw, where is you magic private tree and what is the plan for it?