Re: [PATCH] M68k cache

[Roman Zippel <zippel@linux-m68k.org>]

Hi,

On Fri, 29 Aug 2003, Geert Uytterhoeven wrote:

> M68k icache flush fixes from Roman Zippel:
>   - uninline flush_icache_range() and rename it to flush_icache_user_range()
>   - add virt_to_phys_slow() which handles vmalloc()ed space
>   - add flush_icache_range and flush_icache_user_page
> 
> [...]
> 
> --- linux-2.4.23-pre1/kernel/ptrace.c	Wed May 28 13:11:52 2003
> +++ linux-m68k-2.4.23-pre1/kernel/ptrace.c	Fri Jul 25 20:02:39 2003
> @@ -165,7 +165,7 @@
>  		if (write) {
>  			memcpy(maddr + offset, buf, bytes);
>  			flush_page_to_ram(page);
> -			flush_icache_user_range(vma, page, addr, len);
> +			flush_icache_user_page(vma, page, addr, len);
>  			set_page_dirty(page);
>  		} else {
>  			memcpy(buf, maddr + offset, bytes);

Geert, did you intend to include this part?
This part needs an update of all archs includes (to define 
flush_icache_user_page at least as flush_icache_user_range) and the 
changes to binfmt_{elf,aout}.c are part of this patch.
Maybe I should explain the reason behind this patch:
The actual problem is the usage of flush_icache_range(). In 
kernel/module.c it's used to flush data from the kernel cache, in 
binfmt_{elf,aout}.c it's used to flush data from the user cache and in 
both situations it's called with a user space context, so that 
flush_icache_range() doesn't know which cache to flush and I'd really 
like to avoid having to flush both caches.
So the full patch renames flush_icache_range() in binfmt_{elf,aout}.c into 
flush_icache_user_range(), but which already exists, so I renamed this 
into flush_icache_user_page().

bye, Roman