On Tue, 24 Feb 2004, Alexander Fisher wrote:
> Unfortunately backporting the fix in any sensible time
> period is beyond me. At least one of the files being
> patched doesn't even exist in the 1.3.1 release.
page_buf.[ch] turned into xfs_buf.[ch], and moved
from fs/xfs/pagebuf to fs/xfs/linux-2.[46]/
With that, it will be much closer to applying.
> If I
> try to remedy this by applying other updates to the
> xfs code first I'm going to get into the situation
> where other core kernel updates are also required.
> These will too have their dependencies. I think it
> really needs somebody who knows the xfs code and
> understands the problem to produce a 1.3.1 fix.
We can't be in the business of applying all new bugfixes
to all past releases, we'd soon run out of time
to do real work, I'm afraid.
> You said that one of my workarounds would be to
> downgrade xfstools (or at least mkfs.xfs). Wouldn't
> it be better if the latest released version of
> xfstools worked correctly with the latest released
> version of xfs?
It does; XFS in 2.4.25 + xfsprogs 2.6.3 does not
exhibit this problem.
> Perhaps its possible for mkfs.xfs to
> detect the use of xfs1.3.1 (and earlier?) and not
> create large allocation groups under those conditions.
mkfs is a purely userspace app, it has no knowledge of
(and no need for) any kernelspace code, so that's
not a workable solution.
I'd suggest that you use 2.4.25, downgrade xfsprogs,
specify 4g AG size with later xfsprogs, or work just
a bit more with the patch to get it to apply to
1.3.1 - at least you have a few options. :)
Thanks,
-Eric
|