[PATCH] x86/mm: Clean up types in xlate_dev_mem_ptr() some more

[Ingo Molnar]

* Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:

> Ugh, I pulled, but:
> 
> On Wed, May 6, 2015 at 5:58 AM, Ingo Molnar <mingo2.kernel.org@xxxxxxxxx> wrote:
> >
> > Ingo Molnar (1):
> >       x86/mm: Clean up types in xlate_dev_mem_ptr()
> >
> > diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c
> > index fdf617c00e2f..4bf037b20f47 100644
> > --- a/arch/x86/mm/ioremap.c
> > +++ b/arch/x86/mm/ioremap.c
> > @@ -332,18 +332,20 @@ EXPORT_SYMBOL(iounmap);
> >   */
> >  void *xlate_dev_mem_ptr(phys_addr_t phys)
> >  {
> > +       unsigned long start  = phys &  PAGE_MASK;
> > +       unsigned long offset = phys & ~PAGE_MASK;
> > +       unsigned long vaddr;
> 
> That "unsigned long vaddr" is just stupid and not a cleanup.
> 
> It causes two pointless casts:
> 
> > +       vaddr = (unsigned long)ioremap_cache(start, PAGE_SIZE);
> > +       /* Only add the offset on success and return NULL if the ioremap() failed: */
> > +       if (vaddr)
> > +               vaddr += offset;
> >
> > +       return (void *)vaddr;
> 
> neither of which is helpful in the least. And the "vaddr += offset" 
> would work equally well in "void *", gcc is perfectly happy to treat 
> "void *" arithmetic as byte offsets, it's both documented and 
> already extensively used in the kernel.

Yeah, not sure why I didn't notice that, I love void * arithmetics.

> So the cleanup to use "start/offset" is a good cleanup, but you 
> should have kept "addr" as a pointer.

Something like the patch below?

Thanks,

	Ingo

===================>