I don't see an API change:
the 2.2.17 "struct nfs_fs" and the 2.2.18 "struct nfs2_fh" are identical.
Ok, I see the problem:
struct nfs_fh {
unsigned short size;
unsigned char data[NFS_MAXFHSIZE];
}
The compiler thinks that data is a character array, thus no padding is inserted.
nlm_lookup_file casts f->data to "struct knfs_fs", a structure with pointers and
u32. --> unaligned u32 read --> boom.
Is that correct?
Is &(((struct nfs_fh*)0)->data) 2 or 4?
ARM isn't the only cpu that can't handle unaligned memory reads, why doesn't the
code fail on Alpha/Sparc? Does gcc insert padding on these cpus?
-- Manfred - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org Please read the FAQ at http://www.tux.org/lkml/