next branch update

[Sean MacLennan]

On Thu, 11 Jun 2009 10:50:52 +1000
Benjamin Herrenschmidt <benh at kernel.crashing.org> wrote:

> n Wed, 2009-06-10 at 17:56 -0400, Sean MacLennan wrote:
> > What ever happened to this patch?  
> 
> Dunno... It should have been in patchwork. I remember the patch
> in fact and I intended to merge it... Can you re-submit with
> appropriate cset comment, signed-off-by etc... ?
> 
> (Mistakes -do- happen, which is why we can be flexible with the rules,
> btw, but heh, 33 new patches the day the merge window opens is a bit
> over the top).

I was expecting an rc9 and another week. Probably others where caught
off guard. Patch below.

Cheers,
   Sean

If no device is passed to __dma_alloc_coherent, it defaults to using ISA_DMA_THRESHOLD
for the mask. This patch provides a reasonable default rather than 0.

Signed-off-by: Sean MacLennan <smaclennan at pikatech.com>
---
diff --git a/arch/powerpc/platforms/44x/warp.c b/arch/powerpc/platforms/44x/warp.c
index 21ff868..7f3c1c7 100644
--- a/arch/powerpc/platforms/44x/warp.c
+++ b/arch/powerpc/platforms/44x/warp.c
@@ -43,7 +43,13 @@ static int __init warp_probe(void)
 {
 	unsigned long root = of_get_flat_dt_root();
 
-	return of_flat_dt_is_compatible(root, "pika,warp");
+	if (!of_flat_dt_is_compatible(root, "pika,warp"))
+		return 0;
+
+	/* For __dma_alloc_coherent */
+	ISA_DMA_THRESHOLD = ~0L;
+
+	return 1;
 }
 
 define_machine(warp) {