[PATCH] zone table removal miss merge

[PATCH] zone table removal miss merge

[Andy Whitcroft]

As suspected this is not related to SPARSEMEM configuration at all.
But relates to the case where the node,zone size is zero.  Here we
then are trying to shift (sizeof(int) - 0) which is illegal.

We should be defining ZONEID_SHIFT in terms of ZONE_PGSHIFT not
ZONE_PGOFF.  As this was correct in the orginal patch I assume this
was somehow damaged during merge.

The below should fix it.

-apw
=== 8< ===
zone table removal miss-merge

It looks very much like zone table removal v2 suffered during merge
into -mm.  This patch is needed to get rid of the following errors
on arm (and I suspect other platforms):

  include/linux/mm.h: In function `page_zone_id':
  include/linux/mm.h:450: warning: right shift count >= width of type

Signed-off-by: Andy Whitcroft <apw@shadowen.org>
---
diff --git a/include/linux/mm.h b/include/linux/mm.h
index a7997d9..2eb64fa 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -421,7 +421,7 @@ #define ZONEID_SHIFT		(SECTIONS_SHIFT + 
 #else
 #define ZONEID_SHIFT		(NODES_SHIFT + ZONES_SHIFT)
 #endif
-#define ZONEID_PGSHIFT		ZONES_PGOFF
+#define ZONEID_PGSHIFT		ZONES_PGSHIFT
 
 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED
 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > FLAGS_RESERVED

Re: [PATCH] zone table removal miss merge

[Christoph Lameter]

On Wed, 27 Sep 2006, Andy Whitcroft wrote:

> The below should fix it.

Acked-by: Christoph Lameter <clameter@sgi.com>


Note that if ZONE_DMA is off then ZONES_WIDTH may become
0 and therefore also ZONES_PGSHIFT is zero.

If you then do

#define ZONEID_PGSHIFT ZONES_PGSHIFT

then ZONEID_PGSHIFT will be 0!

So this could be an issue for the optional ZONE_DMA patch.

Could you also make sure that ZONEID_PGSHIFT is set correctly even if 
ZONES_WIDTH is zero?

This affects the optional ZONE_DMA patch. zone table removal should be 
fine with just the above patch.