lkml.org 
[lkml]   [2014]   [Jan]   [12]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
Date
From
SubjectRe: [PATCH] mm: nobootmem: avoid type warning about alignment value
On Mon, Dec 09, 2013 at 08:02:30PM -0500, Santosh Shilimkar wrote:
> On Monday 09 December 2013 07:54 PM, Russell King - ARM Linux wrote:
> > The underlying reason is that - as I've already explained - ARM's __ffs()
> > differs from other architectures in that it ends up being an int, whereas
> > almost everyone else is unsigned long.
> >
> > The fix is to fix ARMs __ffs() to conform to other architectures.
> >
> I was just about to cross-post your reply here. Obviously I didn't think
> this far when I made $subject fix.
>
> So lets ignore the $subject patch which is not correct. Sorry for noise

Well, here we are, a month on, and this still remains unfixed despite
my comments pointing to what the problem is. So, here's a patch to fix
this problem the correct way. I took the time to add some comments to
these functions as I find that I wonder about their return values, and
these comments make the patch a little larger than it otherwise would be.

This patch makes their types match exactly with x86's definitions of
the same, which is the basic problem: on ARM, they all took "int" values
and returned "int"s, which leads to min() in nobootmem.c complaining.

arch/arm/include/asm/bitops.h | 54 +++++++++++++++++++++++++++++++++++--------
1 file changed, 44 insertions(+), 10 deletions(-)

diff --git a/arch/arm/include/asm/bitops.h b/arch/arm/include/asm/bitops.h
index e691ec91e4d3..b2e298a90d76 100644
--- a/arch/arm/include/asm/bitops.h
+++ b/arch/arm/include/asm/bitops.h
@@ -254,25 +254,59 @@ static inline int constant_fls(int x)
}

/*
- * On ARMv5 and above those functions can be implemented around
- * the clz instruction for much better code efficiency.
+ * On ARMv5 and above those functions can be implemented around the
+ * clz instruction for much better code efficiency. __clz returns
+ * the number of leading zeros, zero input will return 32, and
+ * 0x80000000 will return 0.
*/
+static inline unsigned int __clz(unsigned int x)
+{
+ unsigned int ret;
+
+ asm("clz\t%0, %1" : "=r" (ret) : "r" (x));

+ return ret;
+}
+
+/*
+ * fls() returns zero if the input is zero, otherwise returns the bit
+ * position of the last set bit, where the LSB is 1 and MSB is 32.
+ */
static inline int fls(int x)
{
- int ret;
-
if (__builtin_constant_p(x))
return constant_fls(x);

- asm("clz\t%0, %1" : "=r" (ret) : "r" (x));
- ret = 32 - ret;
- return ret;
+ return 32 - __clz(x);
+}
+
+/*
+ * __fls() returns the bit position of the last bit set, where the
+ * LSB is 0 and MSB is 31. Zero input is undefined.
+ */
+static inline unsigned long __fls(unsigned long x)
+{
+ return fls(x) - 1;
+}
+
+/*
+ * ffs() returns zero if the input was zero, otherwise returns the bit
+ * position of the first set bit, where the LSB is 1 and MSB is 32.
+ */
+static inline int ffs(int x)
+{
+ return fls(x & -x);
+}
+
+/*
+ * __ffs() returns the bit position of the first bit set, where the
+ * LSB is 0 and MSB is 31. Zero input is undefined.
+ */
+static inline unsigned long __ffs(unsigned long x)
+{
+ return ffs(x) - 1;
}

-#define __fls(x) (fls(x) - 1)
-#define ffs(x) ({ unsigned long __t = (x); fls(__t & -__t); })
-#define __ffs(x) (ffs(x) - 1)
#define ffz(x) __ffs( ~(x) )

#endif

--
FTTC broadband for 0.8mile line: 5.8Mbps down 500kbps up. Estimation
in database were 13.1 to 19Mbit for a good line, about 7.5+ for a bad.
Estimate before purchase was "up to 13.2Mbit".


\
 
 \ /
  Last update: 2014-01-12 12:21    [W:0.107 / U:0.392 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site