Ext4 Developer Interlock Call: 9/20/06 Minutes
Attendees: Shaggy (Dave Kleikamp), Jean-Pierre Dion, Alexandre Ratchov,
Val?rie Cl?ment, Eric Sandeen, Avantika Mathur
- Shaggy has ported the patch set to 2.6.18-rc7-mm1, and will post the
patches today. Some of the patches are redundant and should be merged
together, Shaggy and Alexandre will be working on this. These patches
have only been compile tested, and need more thorough testing.
Alexandre has been running ext4 on a 20 TB sparse disk for a month and
has not encountered any problems.
- Shaggy and Alexandre will also look back at all of the comments on the
original post, to verify they have been addressed in the updated series.
- Now that 2.6.18 is up, We assume Andrew Morton will be pushing current
ext3 patches in the -mm tree to 2.6.19-rc1. At that point we will try
to merge ext4 patches.
- Alexandre is waiting for feedback from Ted on e2fsprogs patches. Plan
to discuss and review through the mailing list.
- A bug in the JBD patches was found last week; JBD and JBD2 use the
same slab. The kernel crashes if ext3 and ext4 are both loaded. This
problem is probably still present in current set. Easily reproduced by
loading ext3 and ext4 modules, the second one will crash because of
duplicate slab name.
On Wed, Sep 20, 2006 at 10:03:51AM -0700, Avantika Mathur LTC wrote:
> - Alexandre is waiting for feedback from Ted on e2fsprogs patches. Plan
> to discuss and review through the mailing list.
My posting from earlier this week was a result of reviewing the
clusterfs patches, althoguh perhaps I should have been more explicit
about my complaints:
*) In appropriate namespace leakage (all new functions that are
visible should have the ext2fs_ prefix)
*) Interfaces that expose the on-disk extent format, and will
therefore require changing (and thus breaking) the API and ABI when we
add support for 48-bit or 64-bit extent patches.
I was trying to be a bit more constructive, though, and have been in
the midst of reimplementing a better interface. Although if Alexandre
has time in the next week or two, things might go faster if I can just
explain what I want the new interface to look like and he can
implement it. :-)