Broken CLEAR_BITMAP() macro

Hello all,
It appears that the CLEAR_BITMAP() macro in include/linux/types.h is
broken.

(Examples done with BITS_PER_LONG == 32, but would work with == 64, or anything but 8 for that matter! :)

If, for example, we DECLARE_BITMAP(foo, 64), we're going to get:

	unsigned long foo[((64)+32-1)/32] =>
	unsigned long foo[95/32] =>
	unsigned long foo[2]

Now, that's all well and good (and correct! ;) But, look at what happens if we do a CLEAR_BITMAP(foo, 64):

	memset(foo, 0, ((64)+32-1)/8) =>
	memset(foo, 0, 95/8) =>
	memset(foo, 0, 11)

So the memset is going to try and clear 11 bytes starting at foo, which will overflow the foo array by 3 bytes. This is bad.

What CLEAR_BITMAP wants to be doing is this:
#define CLEAR_BITMAP(name,bits) \
memset(name, 0,
(((bits)+BITS_PER_LONG-1)/BITS_PER_LONG)*sizeof(unsigned long))

Attatched is a patch that creates a macro BITS_TO_LONGS that just rounds a number of bits up to the closest number of unsigned longs. This makes the DECLARE & CLEAR _BITMAP macros more readable. I also modify the CLEAR_BITMAP macro to work correctly.

Cheers!

-Matt

diff -Nur --exclude-from=/usr/src/.dontdiff linux-2.5.59-vanilla/include/linux/types.h linux-2.5.59-bitmap_fix/include/linux/types.h
@@ -4,10 +4,12 @@
 #ifdef	__KERNEL__
 #include <linux/config.h>
 
+#define BITS_TO_LONGS(bits) \
+	(((bits)+BITS_PER_LONG-1)/BITS_PER_LONG)
 #define DECLARE_BITMAP(name,bits) \
-	unsigned long name[((bits)+BITS_PER_LONG-1)/BITS_PER_LONG]
+	unsigned long name[BITS_TO_LONGS(bits)]
 #define CLEAR_BITMAP(name,bits) \
-	memset(name, 0, ((bits)+BITS_PER_LONG-1)/8)
+	memset(name, 0, BITS_TO_LONGS(bits)*sizeof(unsigned long))
 #endif
 
 #include <linux/posix_types.h>