On Wednesday September 28, agk@xxxxxxxxxx wrote:
> On Wed, Sep 14, 2005 at 02:32:59PM +1000, Neil Brown wrote:
> > It is worth noting that the code uses the same bitsets for in-memory
> > and on-disk logs. As these bitsets are host-endian and host-sized,
> > this means that they cannot safely be moved between computers with
> > different architectures. I don't know if the dm doco makes this
> > clear...
>
> Are you able to try out the (untested) patch from Patrick below?
> Since the current implementation is broken for 64-bit BE we might
> as well fix both problems at the same time.
>
I've ask the person who reported the problem to test this patch. I'll
let you know what eventuates.
... it's a pity about the 'ext2' appearing in there instead of a more
sensible 'set_le_bit', isn't it :-(
Thanks,
NeilBrown
> Alasdair
>
>
> From: Patrick Caulfield <pcaulfie@xxxxxxxxxx>
>
> The solution seems to be to use to set_le_bit test_le_bit functions that ext2
> uses. That makes the BE machines keep the bitmap internally in LE order - it
> does mean you can't use any other type of operations on the bitmap words but
> that looks to be OK in this instance. The efficiency tradeoff is very minimal as
> you would expect for something that ext2 uses.
>
> The nice thing about this way is that there is no on-disk format change for
> 32bit LE boxes at least, and even nicer is that it removes code and data
> structures because log->disk_bits disappears.
>
>
> Index: linux-2.6.14-rc2/drivers/md/dm-log.c
> ===================================================================
> --- linux-2.6.14-rc2.orig/drivers/md/dm-log.c 2005-09-28 18:32:53.000000000 +0100
> +++ linux-2.6.14-rc2/drivers/md/dm-log.c 2005-09-28 18:33:30.000000000 +0100
> @@ -157,7 +157,6 @@
> struct log_header *disk_header;
>
> struct io_region bits_location;
> - uint32_t *disk_bits;
> };
>
> /*
> @@ -166,20 +165,20 @@
> */
> static inline int log_test_bit(uint32_t *bs, unsigned bit)
> {
> - return test_bit(bit, (unsigned long *) bs) ? 1 : 0;
> + return ext2_test_bit(bit, (unsigned long *) bs) ? 1 : 0;
> }
--
dm-devel@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/dm-devel
