Re: [parisc-linux] Re: RFC: mmap patch

[Matthew Wilcox <willy@debian.org>]

On Sat, Mar 08, 2003 at 03:00:23PM -0800, David S. Miller wrote:
> You need merely 8MB of address space (2 * 4MB) if you implement
> my {copy,clear}_user_page() dynamic mapping hack, that will be
> tons more cheaper than any kmap based scheme and also be nicer
> on the TLB as there will be zero TLB changes occurring around
> the copy/clear.
> 
> People can continue to talk about all their bright new idea, which
> is fine, but it feels like my known-working ideas are being ignored.

Sorry, I think you've misunderstood the problem.  This is write() vs mmap().
We call flush_dcache_page() after the write() -- but PA's flush_dcache_page() only flushes the kernel's mapping of that page, not the user's.  The question is how to fix that.

Randolph published a patch which flushes two pages -- one which is the
kernel's view of the page, and one which is the user's view of the page.
John disagreed that was necessary, and I'm proposing a way (with kmap)
to avoid it.

Nothing to do with copy/clear_user_page().

-- 
"It's not Hollywood.  War is real, war is primarily not about defeat or
victory, it is about death.  I've seen thousands and thousands of dead bodies.
Do you think I want to have an academic debate on this subject?" -- Robert Fisk