Re: RFC: being more anal about iospace accesses..

[Matthew Wilcox <willy@debian.org>]

On Thu, Sep 16, 2004 at 03:42:34PM -0700, Linus Torvalds wrote:
> On Thu, 16 Sep 2004, Matthew Wilcox wrote:
> > 2004/09/14 torvalds   |  * They do _not_ update the port address. If you
> > 2004/09/14 torvalds   |  * want MMIO that copies stuff laid out in MMIO
> > 2004/09/14 torvalds   |  * memory across multiple ports, use "memcpy_toio()"
> > 2004/09/14 torvalds   |  * and friends.
> > 
> > That implies you can pass an iomap cookie to memcpy_toio()
> 
> Actually, no it does not. It very much says "if you want __MMIO__ that 
> copies stuff".
> 
> You can pass cookies the other way:  if you create a cookie with the old
> "ioremap()" interface (which means MMIO only, of course), I intended for
> that to work well with the new interfaces (ioread/iowrite). But a
> new-style IO-map cookie will _not_ work with the old interfaces. 

OK, you're thinking in x86-specific terms and I'm thinking in PA-specific
terms ;-)  I don't particularly want to be able to memcpy_fooio() on
ioport space.  But I do want to be able to memcpy_fooio() on big-endian
mmio space and I do want to design the iomap cookies such that you can't
pass the old ioremap cookies to the new interface.

-- 
"Next the statesmen will invent cheap lies, putting the blame upon 
the nation that is attacked, and every man will be glad of those
conscience-soothing falsities, and will diligently study them, and refuse
to examine any refutations of them; and thus he will by and by convince 
himself that the war is just, and will thank God for the better sleep 
he enjoys after this process of grotesque self-deception." -- Mark Twain