Re: [Q] mm/memblock.c: cast truncates bits from RED_INACTIVE

[Andrew Morton <akpm@linux-foundation.org>]

On Tue, 14 Jun 2011 19:47:19 -0500
H Hartley Sweeten <hartleys@visionengravers.com> wrote:

> Hello all,
> 
> Sparse is reporting a couple warnings in mm/memblock.c:
> 
> 	warning: cast truncates bits from constant value (9f911029d74e35b becomes 9d74e35b)
> 
> The warnings are due to the cast of RED_INACTIVE in memblock_analyze():
> 
> 	/* Check marker in the unused last array entry */
> 	WARN_ON(memblock_memory_init_regions[INIT_MEMBLOCK_REGIONS].base
> 		!= (phys_addr_t)RED_INACTIVE);
> 	WARN_ON(memblock_reserved_init_regions[INIT_MEMBLOCK_REGIONS].base
> 		!= (phys_addr_t)RED_INACTIVE);
> 
> And in memblock_init():
> 
> 	/* Write a marker in the unused last array entry */
> 	memblock.memory.regions[INIT_MEMBLOCK_REGIONS].base = (phys_addr_t)RED_INACTIVE;
> 	memblock.reserved.regions[INIT_MEMBLOCK_REGIONS].base = (phys_addr_t)RED_INACTIVE;
> 
> Could this cause any problems?  If not, is there anyway to quiet the sparse noise?
> 

It's all just a debugging check and that check will continue to work OK
despite this bug.

But yes, it's ugly and should be fixed.

I don't think that mm/memblock.c should have reused RED_INACTIVE. 
That's a slab thing and wedging it into a phys_addr_t was
inappropriate.

In fact I don't think RED_INACTIVE should exist.  It's just inviting
other subsystems to (ab)use it.  It should be replaced by a
slab-specific SLAB_RED_INACTIVE, as slub did with SLUB_RED_INACTIVE.


I'd suggest something like the below, which I didn't test.  Feel free to
send it back at me, or ignore it ;)


diff -puN include/linux/poison.h~a include/linux/poison.h
--- a/include/linux/poison.h~a
+++ a/include/linux/poison.h
@@ -40,6 +40,12 @@
 #define	RED_INACTIVE	0x09F911029D74E35BULL	/* when obj is inactive */
 #define	RED_ACTIVE	0xD84156C5635688C0ULL	/* when obj is active */
 
+#ifdef CONFIG_PHYS_ADDR_T_64BIT
+#define MEMBLOCK_INACTIVE	0x3a84fb0144c9e71bULL
+#else
+#define MEMBLOCK_INACTIVE	0x44c9e71bUL
+#endif
+
 #define SLUB_RED_INACTIVE	0xbb
 #define SLUB_RED_ACTIVE		0xcc
 
diff -puN mm/memblock.c~a mm/memblock.c
--- a/mm/memblock.c~a
+++ a/mm/memblock.c
@@ -758,9 +758,9 @@ void __init memblock_analyze(void)
 
 	/* Check marker in the unused last array entry */
 	WARN_ON(memblock_memory_init_regions[INIT_MEMBLOCK_REGIONS].base
-		!= (phys_addr_t)RED_INACTIVE);
+		!= MEMBLOCK_INACTIVE);
 	WARN_ON(memblock_reserved_init_regions[INIT_MEMBLOCK_REGIONS].base
-		!= (phys_addr_t)RED_INACTIVE);
+		!= MEMBLOCK_INACTIVE);
 
 	memblock.memory_size = 0;
 
@@ -786,8 +786,8 @@ void __init memblock_init(void)
 	memblock.reserved.max	= INIT_MEMBLOCK_REGIONS;
 
 	/* Write a marker in the unused last array entry */
-	memblock.memory.regions[INIT_MEMBLOCK_REGIONS].base = (phys_addr_t)RED_INACTIVE;
-	memblock.reserved.regions[INIT_MEMBLOCK_REGIONS].base = (phys_addr_t)RED_INACTIVE;
+	memblock.memory.regions[INIT_MEMBLOCK_REGIONS].base = MEMBLOCK_INACTIVE;
+	memblock.reserved.regions[INIT_MEMBLOCK_REGIONS].base = MEMBLOCK_INACTIVE;
 
 	/* Create a dummy zero size MEMBLOCK which will get coalesced away later.
 	 * This simplifies the memblock_add() code below...
_