On 10 Jan 2003, Juan Quintela wrote:
> The only thing that could be controversial is the _l1() thing, and as
> current thing is broken, I vote for insclusion.
>
> maciej> diff -up --recursive --new-file
> linux-mips-2.4.20-pre6-20030107.macro/arch/mips64/mm/c-r4k.c
> linux-mips-2.4.20-pre6-20030107/arch/mips64/mm/c-r4k.c
> maciej> --- linux-mips-2.4.20-pre6-20030107.macro/arch/mips64/mm/c-r4k.c
> 2002-12-20 03:56:52.000000000 +0000
> maciej> +++ linux-mips-2.4.20-pre6-20030107/arch/mips64/mm/c-r4k.c
> 2003-01-09 23:21:39.000000000 +0000
> @@ -979,7 +980,7 @@ static void r4k_dma_cache_wback_inv_sc(u
> unsigned long end, a;
>
> if (size >= scache_size) {
> - flush_cache_l1();
> + flush_cache_all();
> return;
> }
>
> This one is fixing a bug, we are talking about a chip with Secondary
> cache and don't touch the secondary cache at all :(
That bug is inactive -- both function pointers are defined to the same
value as surprisinly enough "l1" means "both caches" for the R4k. Anyway,
I for removing flush_cache_l1() altogether in the next step.
--
+ Maciej W. Rozycki, Technical University of Gdansk, Poland +
+--------------------------------------------------------------+
+ e-mail: macro@ds2.pg.gda.pl, PGP key available +
|