Here are the topics that are cooking in e2fsprogs. Commits prefixed
with '-' are only in 'pu' while commits prefixed with '+' are in
'next'. Commits prefixed with '.' are not in either branch for now.
* tt/64bit-bitmaps (Sun Oct 14 22:51:51 2007 -0400) 1 commit
- Initial design for 64-bit bitmaps
This a sketch for how to do 64-bit bitmaps; it's mostly comments and
function signatures for now.
* js/uninit (Thu Oct 11 14:17:19 2007 -0500) 16 commits
. Add m_uninit test case.
. Add new mm_lazy test case.
. Fix test cases.
. Update uninit block group documetation for some of the utilities.
. Make e2fsck uninit block group aware.
. Make debugfs uninit block group aware.
. Make resize2fs uninit block group aware.
. Make dumpe2fs uninit block group aware.
. Make tune2fs uninit block group aware.
. Add support for creating filesystems using uninit block group.
. Rename feature name from gdt_checksum to uninit_groups.
. Add uninit block group support on libe2fs.
. Add initial checksum support.
This doesn't compile on my Ubuntu system, due to the use of
linux/types.h. Jose is reworking this patch seris, so I've left it out
of pu for now.
* cl/remove-masix (Thu Aug 23 15:09:03 2007 +0800) 8 commits
- mke2fs: remove masix support
- e2fsck: remove masix support
- libext2fs: remove masix support
- document: remove masix from data structure
- debugfs: remove masix support
I've done a quick review, and this will probably going in soon.
* tt/extents (Mon Aug 20 21:31:11 2007 -0400) 5 commits
- e2fsck: Add support for extents
- Add support for extents to libext2fs
- e2fsck: factor out code to clear an inode into
e2fsck_clear_inode()
- Don't byte swap extents information in the inode
- Allow debugfs to be extended for use by test programs
This will correctly check a valid filesystem, but it may crash or loop
forever on an invalid filesystem. More work needs to be done on the
e2fsck patch, but the core extent support in libext2fs is mostly there.
* js/flex-bg (Mon Aug 13 23:33:14 2007 -0500) 7 commits
- New bitmap and inode table allocation for FLEX_BG
- Relax group descriptor checking.
- Allow FLEX_BG to be use as a feature option at mke2fs time.
- Reserve the INCOMPAT feature number for FLEX_BG.
Relatively straight forward, so we should be able to merge this soon.
* ak/undo-mgr (Mon Aug 13 15:56:26 2007 +0530) 9 commits
- e2fsprogs: Add test case for undoe2fs
- e2fsprogs: Fix the resize inode test case
- e2fsprogs: Support for large inode migration.
- e2fsprogs: Make mke2fs use undo I/O manager.
- e2fsprogs: Add undoe2fs
- e2fsprogs: Add undo I/O manager
The undo manager is mostly OK, but the large inode migration is
extremely unsatisfactory, given that it halves the number of inodes and
only works if the no more than 50% of the inodes in every single block
group is used. As a standalone kludge it may be OK, but the right thing
to do is to allocate more blocks for the inode table.
* ad/extents-testcases (Thu Jul 12 11:20:14 2007 -0400) 19 commits
. Add extent test: f_extents_shrd_blk
. Add extent test: f_extents_unsorted
. Add extent test: f_extents_res_blk
. Add extent test: f_extents_overlap
. Add extent test: f_extents_orphan_blks
. Add extent test: f_extents_inrlevel-incons
. Add extent test: f_extents_imbalanced_tree
. Add extent test: f_extents_ei_leaf
. Add extent test: f_extents_ei_block
. Add extent test: f_extents_eh_max
. Add extent test: f_extents_eh_magic
. Add extent test: f_extents_eh_entries
. Add extent test: f_extents_eh_depth
. Add extent test: f_extents_ee_start
. Add extent test: f_extents_ee_len
. Add extent test: f_extents_ee_block
. Add extent test: f_extents_bad_blk
. Add extent test: f_extents
. Add extent test: f_bad_disconnected_inode
These are the extent test cases from Clusterfs. They will be pulled in
to the extents branch once they don't cause e2fsck to loop forever.
- Ted
Theodore Ts'o wrote:
>
> * ak/undo-mgr (Mon Aug 13 15:56:26 2007 +0530) 9 commits
> - e2fsprogs: Add test case for undoe2fs
> - e2fsprogs: Fix the resize inode test case
> - e2fsprogs: Support for large inode migration.
> - e2fsprogs: Make mke2fs use undo I/O manager.
> - e2fsprogs: Add undoe2fs
> - e2fsprogs: Add undo I/O manager
>
> The undo manager is mostly OK, but the large inode migration is
> extremely unsatisfactory, given that it halves the number of inodes and
> only works if the no more than 50% of the inodes in every single block
> group is used. As a standalone kludge it may be OK, but the right thing
> to do is to allocate more blocks for the inode table.
>
>
We don't do that.
git show 9ff4bdac259a942b6e1afc49d868075d8b19463a
...
As a part of increasing the inode size we increase the
inode table size. We also move the used data blocks around
and update the respective inodes to point to the new block
....
We are right now increasing the inode table size moving the data blocks around
and fixing the inodes to point to right data blocks.
-aneesh
On Tue, 16 Oct 2007 00:48:21 -0400
"Theodore Ts'o" <[email protected]> wrote:
> * js/flex-bg (Mon Aug 13 23:33:14 2007 -0500) 7 commits
> - New bitmap and inode table allocation for FLEX_BG
This one was more of an RFC to let folks play with different meta data
layouts. Should probably stay outside the tree for now. The other
three patches are whats needed to get e2fsprog in sync with the kernel.
> - Relax group descriptor checking.
This one needs updating to be more in line with what we define FLEX_BG
to be in the kernel. I thing I also encounter conflicts when pulling
from the git tree a while back, so the one in the mailing list will not
apply cleanly.
> - Allow FLEX_BG to be use as a feature option at mke2fs time.
> - Reserve the INCOMPAT feature number for FLEX_BG.
>
> Relatively straight forward, so we should be able to merge this soon.
The last two are fine but its still missing documentation. Skip this
series for now. I will update and resend.
-JRS