Ext4 Developer Interlock Call: 03-04-07 Minutes
Attendees: Mingming Cao, Dave Kleikamp, Jean-Pierre Dion, Jean Noel Cordenner, Akira Fujita, Jose Santos, Badari Pulavarty, Ted Ts'o, Andreas Dilger, Avantika Mathur
Minutes can be accessed at: http://ext4.wiki.kernel.org/index.php/Ext4_Developer%27s_Conference_Call
Ext4 Patch Queue Status:
- The first four bug-fix patches in the ext4 git tree have been accepted into Linus' tree. Once 2.6.22-rc4 is out, the ext4 tree will be updated with these patches removed.
Inode Versioning Patches:
- There was a lot of discussion about the different requirements and plans for this feature.
- NFSv4 requires a 64-bit counter which is updated on every file change including metadata updates. unsure sure if it should be updated on atime updates. The counter is only checked for inequality. The current patches by Jean Noel fulfill these requirements.
- Andreas explained Lustre's requirements which are a superset of what is required by NFSv4. Every time an operation is executed on a file, the counter is updated by a specific value. The counter is globally unique on the entire filesystem. Andreas said they can use the current patches, and build upon them with their own patches for their purposes.
- The current patches use the i_version field in the vfs inode for the counter, and all updates to the field are implemented in ext4 code. Jean Noel will ask Trond if these patches are ok, or if he would prefer the the updates to be implemented at the vfs layer (if it is possible). This may make it easier for other filesystems to start using the feature.
- Ted suggested having a flag in the superblock for i_version, with an additional option of having a global counter for Lustre's requirements. i_version would be on by default, but give users who do not use the feature the option to turn it off and avoid any performance impact.
- Andreas will ask Kalpak to separate the i_version specific changes from his expand-inode-size patch, and merge these changes with Jean Noel's patches.
- The i_version is being reset in the ext4_quota_write() function. Jean Noel does not think that this needs to be changed, Andreas and Mingming felt that resetting the counter at this place did not make sense. We will keep the patch unchanged, unless there are any
- Once Kalpak's patches are updated and the patches are ok for NFSv4, the plan is to post the patches and if there are no issues, push them to Linus' tree.
- Ted plans to push e2fsprogs-1.40 this week, so any bug fixes will be available to distros. Ted would like to get the extents capability into mainline soon as well.
- Currently there are many patches to e2fsprogs, Ted hopes to start merging these to mainline.
- Valerie is working on patches to e2fsprogs for supporting 64-bit block numbers. Ted is very interested in seeing the direction of these patches and strongly encourages Valerie to post her work in progress patches soon.
Uninitialized Block Groups:
- Avantika started porting Andreas' uninitialized block group kernel and e2fsprogs patches, but ran into problems in testing, with e2fsprogs. She will post the patches to the list for further review.