Hi Finn,
[resent after it bounced from vger ...]
On 15/06/23 12:13, Finn Thain wrote:
On Wed, 14 Jun 2023, Michael Schmitz wrote:
The reason why only rdb_SummedLongs, rdb_BlockBytes and
rdb_PartitionList are explicitly declared big endian is probably quite
simple - nothing else was used by the Linux RDB parser. My patch adds
rdb_CylBlocks to that list, so that ought to be changed to big endian,
too.
affs_hardblocks.h is a UAPI header - what are the rules and
ramifications around changes to those? Might not be worth the hassle in
the end.
I think it's safe to fix the UAPI header if we are talking about
endianness annotations that affect static checking and not code
generation. The existing annotations in that struct would appear to
support that notion, if indeed they were put there for the benefit of the
kernel.
The annotations pre-date mainstream git history - looking at Thomas
Gleixner's full history git
(https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git/), the
annotations were added by Al Viro in 2004 (commit
380768f5a50524ff7eab00d03d82a56cf2fdfe7b
<https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git/commit/include/linux/affs_hardblocks.h?id=380768f5a50524ff7eab00d03d82a56cf2fdfe7b>
there).
Al said at the time that he only annotated only those fields that are
used by the kernel, and left everything else alone even though it all
ought to be big endian.
Reason enough to now change the rdb_CylBlocks annotation.
Thanks,
Michael
