On Sat, Aug 11, 2012 at 12:50:45PM +0200, Michael Monnerie wrote:
> Am Freitag, 10. August 2012, 08:15:18 schrieb Dave Chinner:
> > The freeze/thaw deadlock is not fixed in 3.5 and i think it missed
> > 3.6 as well so you're going to be waiting until 3.7 for this to be
> > fixed, I suspect....
> OK, at least you guys have a plan for a fix. I'm happy to get that fixed
> soon. Thank you.
It got included late in the 3.6-rc1 cycle, so it will be in 3.6.
> But when I think about it: Isn't that worth a backport fix? VMware isn't
> a seldom used piece of software, and we use NetApp Storage with
> snapshots, which is also used a lot worldwide. I can imagine others have
> the same problem, or is it just me?
It's a problem that has been around for, well, since freeze was
first introduced many, many years ago. And the complexity of the fix
(20+ patch series, IIRC) makes it a difficult and time consuming fix
that requires a lot of testing - it touches the VFS and has changes
all through different filesystems, including XFS. I just can't see
anyone other than enterprise distro vendors spending that much
effort to back port such an intrusive fix....