Re: [RFC PATCH v1 7/9] x86: Add unsafe_copy_from_user()

[Yury Norov <ynorov@nvidia.com>]

On Mon, Apr 27, 2026 at 08:20:38PM +0200, Christophe Leroy (CS GROUP) wrote:
> 
> 
> Le 27/04/2026 à 19:58, Yury Norov a écrit :
> > On Mon, Apr 27, 2026 at 07:13:48PM +0200, Christophe Leroy (CS GROUP) wrote:
> > > At the time being, x86 and arm64 are missing unsafe_copy_from_user().
> > 
> > No, they don't. They (should) rely on a generic implementation from
> > linux/uaccess.h, like every other arch, except for  PPC and RISCV.
> > 
> > But they #define arch_unsafe_get_user, and the unsafe_copy_from_user()
> > becomes undefined conditionally on that.
> > 
> > So please, fix that bug instead of introducing another arch flavor.
> > We'd always choose generic version, unless there's strong evidence
> > that arch one is better.
> 
> But they both implement the exact same unsafe_copy_to_user(). What is the
> difference here ?
>
> Should that function become generic too ?

This is what Linus said when added x86 implementation for copy_from_user()
in c512c69187197:

  Note that it only does this [arch version] for the copying _to_ user space,
  and we still don't have a unsafe version of copy_from_user().
  
  That's partly because we have no current users of it, but also partly
  because the copy_from_user() case is slightly different and cannot
  efficiently be implemented in terms of a unsafe_get_user() loop (because
  gcc can't do asm goto with outputs).

In the unsafe_copy_to_user case, arch versions were justified. Just as
said, I'm not against arch version for unsafe_copy_from_user(), but it
should be explained very well.