[PATCH] Fix MMC warnings

[PATCH] Fix MMC warnings

[Pierre Ossman]

[-- Attachment #1: Type: text/plain, Size: 58 bytes --]

Here's a patch that fixes the compiler warnings in mmc.c.

[-- Attachment #2: mmc-warning.patch --]
[-- Type: text/x-patch, Size: 714 bytes --]

Index: linux-wbsd/drivers/mmc/mmc.c
===================================================================
--- linux-wbsd/drivers/mmc/mmc.c	(revision 117)
+++ linux-wbsd/drivers/mmc/mmc.c	(working copy)
@@ -428,14 +428,14 @@
 
 #define UNSTUFF_BITS(resp,start,size)					\
 	({								\
-		const u32 __mask = (1 << (size)) - 1;			\
+		const u32 __mask = ((size >= 32)?0:(1 << (size))) - 1;	\
 		const int __off = 3 - ((start) / 32);			\
 		const int __shft = (start) & 31;			\
 		u32 __res;						\
 									\
 		__res = resp[__off] >> __shft;				\
 		if ((size) + __shft >= 32)				\
-			__res |= resp[__off-1] << (32 - __shft);	\
+			__res |= resp[__off-1] << ((32 - __shft) % 32);	\
 		__res & __mask;						\
 	})
 

Re: [PATCH] Fix MMC warnings

[Russell King]

On Sun, Jan 02, 2005 at 01:07:31AM +0100, Pierre Ossman wrote:
> Here's a patch that fixes the compiler warnings in mmc.c.

I'd rather the compiler was fixed so that it does proper analysis of
the code, rather than blatently issuing warnings for code which is
unreachable.

-- 
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:  2.6 PCMCIA      - http://pcmcia.arm.linux.org.uk/
                 2.6 Serial core

Re: [PATCH] Fix MMC warnings

[Russell King]

On Sun, Jan 02, 2005 at 08:44:09AM +0000, Russell King wrote:
> On Sun, Jan 02, 2005 at 01:07:31AM +0100, Pierre Ossman wrote:
> > Here's a patch that fixes the compiler warnings in mmc.c.
> 
> I'd rather the compiler was fixed so that it does proper analysis of
> the code, rather than blatently issuing warnings for code which is
> unreachable.

Actually, you're quite correct about the first instance - 'mask' may
be generated from 1 << 32 - 1, which is definitely reachable, and gcc
will warn about.  There's also another bug - an out by one on whether
to use the second word to construct the value.

However, the other 5 warnings are due to (corrected version):

	if (16 + 0 > 32)
		__res |= resp[__off-1] << (32 - 0);

Obviously, this code is not reachable since the if condition is false.
Therefore, gcc's whinging about the shift being >= 32 is wrong.

That said, the fix doesn't change gcc's code generation, so I think I'll
apply the changes.  Thanks.

-- 
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:  2.6 PCMCIA      - http://pcmcia.arm.linux.org.uk/
                 2.6 Serial core