| To: | Alex Elder <aelder@xxxxxxx> |
|---|---|
| Subject: | Re: [PATCH] xfsprogs: fix unaligned access in libxfs |
| From: | Christoph Hellwig <hch@xxxxxxxxxxxxx> |
| Date: | Fri, 28 Aug 2009 18:15:30 -0400 |
| Cc: | Christoph Hellwig <hch@xxxxxxxxxxxxx>, Gabriel Vlasiu <gabrielvlasiu@xxxxxxxxx>, xfs@xxxxxxxxxxx |
| In-reply-to: | <1AB9A794DBDDF54A8A81BE2296F7BDFE83ABEF@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> |
| References: | <20090828153718.GA26409@xxxxxxxxxxxxx> <1AB9A794DBDDF54A8A81BE2296F7BDFE83ABEF@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> |
| User-agent: | Mutt/1.5.19 (2009-01-05) |
On Fri, Aug 28, 2009 at 04:39:49PM -0500, Alex Elder wrote: > > +static inline __uint32_t __get_unaligned_be32(const __uint8_t *p) > > > It would be nice for this interface (and others) to take a (void *) > argument, but that would also mean immediately casting it to do > pointer arithmetic portably. Not a big deal. the __ routines are internal implementation details, the kernel has non-__ prefixed versions that take void pointers, but I didn't bother to add them to xfsprogs as we only do unaligned access to 64bit values. That beeing said I dislike the interface taking void pointers as it removes important error checking. The actually visible interface should take __be16/32/64 types to enforce they are used on the right type. Once I get some time I will push for that in the kernel again, but for xfsprogs I really do want to stick to the kernel interfaces so that we can share the code unmodified. |
| <Prev in Thread] | Current Thread | [Next in Thread> |
|---|---|---|
| ||
| Previous by Date: | [PATCH] xfsprogs: update CHANGES, Christoph Hellwig |
|---|---|
| Next by Date: | RE: [PATCH] xfsprogs: fix unaligned access in libxfs, Alex Elder |
| Previous by Thread: | RE: [PATCH] xfsprogs: fix unaligned access in libxfs, Alex Elder |
| Next by Thread: | RE: [PATCH] xfsprogs: fix unaligned access in libxfs, Alex Elder |
| Indexes: | [Date] [Thread] [Top] [All Lists] |