Re: zap_page_range(): TLB flush race

From: Manfred Spraul (manfreds@colorfullife.com)
Date: Tue Apr 11 2000 - 09:41:38 EST

"Stephen C. Tweedie" wrote:
>
>
> OK, I'm sure there are optimisation issues, but I was worried about
> correctness problems from what Alan said.
>

They are correctness problems:
[I'm only reading the asm source, I might be wrong]

* s390 has hardware support for tlb flush ipis.
* it doesn't support hardware contexts (address space numbers, region
id,...)
* They need the old pte value and the virtual address for their flush
ipi.

-->
        set_pte()
        flush_tlb_page()
doesn't work.

Obviously their work-around
        flush_tlb_page()
        set_pte()
is wrong as well, and it breaks all other architectures :-/

They need an atomic set_and_flush_pte() macro.
OTHO on i386 a tlb flush during set_pte() would cause a dramatic
slowdown. 

--
	Manfred

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/

This archive was generated by hypermail 2b29 : Sat Apr 15 2000 - 21:00:16 EST