Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
x86: fix power-of-2 round_up/round_down macros
These macros had two bugs:
- the type of the mask was not correctly expanded to the full size of
the argument being expanded, resulting in possible loss of high bits
when mixing types.
- the alignment argument was evaluated twice, despite the macro looking
like a fancy function (but it really does need to be a macro, since
it works on arbitrary integer types)
Noticed by Peter Anvin, and with a fix that is a modification of his
suggestion (bug noticed by Yinghai Lu).
Cc: Peter Anvin <hpa@zytor.com>
Cc: Yinghai Lu <yinghai@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
+9
-2
+9
-2
arch/x86/include/asm/proto.h
+9
-2
arch/x86/include/asm/proto.h
···
22
22
23
23
long do_arch_prctl(struct task_struct *task, int code, unsigned long addr);
24
24
25
-
#define round_up(x, y) (((x) + (y) - 1) & ~((y) - 1))
26
-
#define round_down(x, y) ((x) & ~((y) - 1))
25
+
/*
26
+
* This looks more complex than it should be. But we need to
27
+
* get the type for the ~ right in round_down (it needs to be
28
+
* as wide as the result!), and we want to evaluate the macro
29
+
* arguments just once each.
30
+
*/
31
+
#define __round_mask(x,y) ((__typeof__(x))((y)-1))
32
+
#define round_up(x,y) ((((x)-1) | __round_mask(x,y))+1)
33
+
#define round_down(x,y) ((x) & ~__round_mask(x,y))
27
34
28
35
#endif /* _ASM_X86_PROTO_H */