Re: Make 2.5.17 TLB even more friendlier

[Linus Torvalds]

On Tue, 21 May 2002, Paul Mackerras wrote:
> 
> I have a bit in the PTE that tells me if there is an MMU hash table
> entry for the virtual address represented by the PTE.

This is exactly why 2.5.17 has the "tlb_remove_pte_entry()", and why it is 
passed down the pte that we just cleared out - so that architectures can 
hide details like this in the page tables (the other use is to hide things 
like "iTBL vs dTLB" etc).

Sp why don't you just make "tlb_remove_pte_entry()" look at your bit, and 
then if that bit is set you try to remove the entry from the hash chains 
at that point?

You _have_ to do it this way, in fact, since reserved pages and other 
"VM-invisible" pages aren't even passed down to "tlb_remove_page()" 
(because they have no freeing logic, and they have no impact on RSS).

		Linus