Re: [PATCH 1/6] parisc: add support for patching multiple words

[Sven Schnelle <svens@stackframe.org>]

On Wed, May 29, 2019 at 07:58:57PM +0200, Rolf Eike Beer wrote:
> Sven Schnelle wrote:
> > On Tue, May 28, 2019 at 10:19:11AM +0200, Rolf Eike Beer wrote:
> > > Another nitpick: the "+4" and "-sizeof(u32)" are just the same at the end,
> > > but why do they use entirely different wording? What do we need "addr" for
> > > anyway, one could just look at "p" which would cross a page boundary at
> > > the
> > > same time, no?
> > 
> > You can't, because of the patch_map() call below which updates the fixed
> > mapping. That call needs the real virtual address, while *p holds the
> > virtual address of the fixed mapping for patching.
> 
> Can that remapping really place it at a non-zero offset regarding to the 
> underlying page? That it moves the page descriptor around is normal, but it 
> will keep the low order bits intact, so the page boundary crossing will be 
> still the same, no?

For the page crossing check it doesn't make a difference whether you check p or
addr, but for the patch_map() you can only use addr. So we have to update both
variables.

Regards
Sven