[dm-devel] Re: Was: [patch 1/1] md/dm: Reduce stack usage with stacked block devices

[Molle Bestefich <molle bestefich gmail com>]

Alasdair G Kergon wrote:
> There are several races.  The code makes some ordering assumptions
> about the I/O but does nothing to enforce those assumptions.
> So if you're unlucky and I/O happens to complete in the wrong order
> or too soon you can see transient corruption in the snapshot or
> reference memory that has been freed and perhaps reallocated to
> something else.

Thanks a lot for the status report.

> Some references:
[snip]

Thanks.  I opened a bug report myself at one point:
https://bugzilla.redhat.com/bugzilla/attachment.cgi?id=120937

> Not actively developed, no.  Long-term we want to replace it with
> something better.

That's really too bad :-(.
There's a lot of people over in the Xen community that could *really*
use something like dm-snapshot if it was stable and production mature.

> But at the moment I'm reviewing the existing code
> and fixing as many problems as I can find.

Sounds great.  Shout out if you need help :-).
(With testing or other monkey work, not C coding.  I'm not a good C coder.)

Thanks!