Re: [BUGFIX]{PATCH] flush icache on ia64 take2

[Zoltan Menyhart <Zoltan.Menyhart@bull.net>]

KAMEZAWA Hiroyuki wrote:

> A bit new idea.  How about this ?
> =
> - Set PG_arch_1 if  "icache is *not* coherent"

page-flags.h:
 * PG_arch_1 is an architecture specific page state bit.  The generic code
 * guarantees that this bit is cleared for a page when it first is entered into
 * the page cache.

I do not think you can easily change it.
I can agree, making nfs_readpage() call an architecture dependent service
is not an easy stuff either. :-)

> - make flush_dcache_page() to be empty func.
> - For Montecito, add kmap_atomic(). This function just set PG_arch1.

kmap_atomic() is used at several places. Do you want to set
PG_arch1, everywhere kmap_atomic() is called?

>   Then, "the page which is copied by the kernel" is marked as "not icache coherent page"
> - icache_flush_page() just flushes a page which has PG_arch_1.
> - Anonymous page is always has PG_arch_1. Tkae care of Copy-On-Write.

You can allocate (even in user mode) an anonymous page, hand-create or
read() in some code from a file, and mprotect(...., EXEC)-it. The page has
to become I-cache coherent.

I am not sure I can really understand your proposal.
I cannot see how the compatibility to the existing code is made sure.

Thanks,

Zoltan