[PATCH 7/8] fs: use get_unaligned_* helpers

[PATCH 7/8] fs: use get_unaligned_* helpers

[Harvey Harrison]

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
---
 fs/partitions/ldm.h |    6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/fs/partitions/ldm.h b/fs/partitions/ldm.h
index 80f63b5..8c341f3 100644
--- a/fs/partitions/ldm.h
+++ b/fs/partitions/ldm.h
@@ -99,9 +99,9 @@ struct parsed_partitions;
 #define TOC_BITMAP2		"log"		/* bitmaps in the TOCBLOCK. */
 
 /* Most numbers we deal with are big-endian and won't be aligned. */
-#define BE16(x)			((u16)be16_to_cpu(get_unaligned((__be16*)(x))))
-#define BE32(x)			((u32)be32_to_cpu(get_unaligned((__be32*)(x))))
-#define BE64(x)			((u64)be64_to_cpu(get_unaligned((__be64*)(x))))
+#define BE16(x)		get_unaligned_be16((x))
+#define BE32(x)		get_unaligned_be32((x))
+#define BE64(x)		get_unaligned_be64((x))
 
 /* Borrowed from msdos.c */
 #define SYS_IND(p)		(get_unaligned(&(p)->sys_ind))
-- 
1.5.5.1.350.gbbbf

Re: [PATCH 7/8] fs: use get_unaligned_* helpers

[Stefan Richter]

Harvey Harrison wrote:
>  fs/partitions/ldm.h |    6 +++---
>  1 files changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/fs/partitions/ldm.h b/fs/partitions/ldm.h
...
>  /* Most numbers we deal with are big-endian and won't be aligned. */
> -#define BE16(x)			((u16)be16_to_cpu(get_unaligned((__be16*)(x))))
> -#define BE32(x)			((u32)be32_to_cpu(get_unaligned((__be32*)(x))))
> -#define BE64(x)			((u64)be64_to_cpu(get_unaligned((__be64*)(x))))
> +#define BE16(x)		get_unaligned_be16((x))
> +#define BE32(x)		get_unaligned_be32((x))
> +#define BE64(x)		get_unaligned_be64((x))

Why the double parentheses?

More importantly, I would say ldm.c should rather use
get_unaligned_be*() directly.  No need to hide what these macros are
about by using a local alias.

After that, ldm.c should be inspected for whether really all of the
accesses are unaligned.  If there are any which are guaranteed to be
aligned, they should of course be turned into be*_to_cpu().  But
separately from this conversion to get_unaligned_be*().
-- 
Stefan Richter
-=====-==--- -=-= ---=-
http://arcgraph.de/sr/

Re: [PATCH 7/8] fs: use get_unaligned_* helpers

[Harvey Harrison]

On Fri, 2008-05-02 at 13:40 +0200, Stefan Richter wrote:
> Harvey Harrison wrote:
> >  fs/partitions/ldm.h |    6 +++---
> >  1 files changed, 3 insertions(+), 3 deletions(-)
> > 
> > diff --git a/fs/partitions/ldm.h b/fs/partitions/ldm.h
> ...
> >  /* Most numbers we deal with are big-endian and won't be aligned. */
> > -#define BE16(x)			((u16)be16_to_cpu(get_unaligned((__be16*)(x))))
> > -#define BE32(x)			((u32)be32_to_cpu(get_unaligned((__be32*)(x))))
> > -#define BE64(x)			((u64)be64_to_cpu(get_unaligned((__be64*)(x))))
> > +#define BE16(x)		get_unaligned_be16((x))
> > +#define BE32(x)		get_unaligned_be32((x))
> > +#define BE64(x)		get_unaligned_be64((x))
> 
> Why the double parentheses?

No reason.

> 
> More importantly, I would say ldm.c should rather use
> get_unaligned_be*() directly.  No need to hide what these macros are
> about by using a local alias.

Done, will send out a revised set.

> 
> After that, ldm.c should be inspected for whether really all of the
> accesses are unaligned.  If there are any which are guaranteed to be
> aligned, they should of course be turned into be*_to_cpu().  But
> separately from this conversion to get_unaligned_be*().

It looks like they are all needed from a quick inspection, the comments
seem to suggest that lots of the accesses will be unaligned.

Cheers,

Harvey