Hi Ted,
Here is the new version of our e2fsprogs patches to add the 64-bit
support in the e2fsprogs tools.
First, I introduced a new compilation option to build e2fsprogs
utilities for ext4 filesytems and not break backwards compatiblity with
ext2/3 filesystems.
So most of the new code for ext4 is included under #ifdef _EXT4FS_.
The major change is that when compiling with _EXT4FS_ option, blk_t is
64 bits.
I tried to create new interfaces for ext4 which use the new type blk64_t
you defined, so most of the code in the patches 03 and 05 are simply
duplicated from the 32-bit version.
This patchset is also not complete:
TODO:
- support of 48-bit in extents to complete,
- support of 48-bit block number for ACL to do,
- change the format string %u for a block number to %llu in case of
64 bits (many occurences in the code),
- update code under ext2ed/,
- and surely other utilities/interfaces I don't use or know to port.
Code compiled, booted and briefly tested on a 20TB device (mkfs,
debugfs, e2fsck, dumpe2fs, filefrag tested)
Some non-regression tests also done on ext3 filesystems when compiling
code without _EXT4FS_ option.
I splitted the changes into 12 patches to facilitate the review of the
patches. I hope this work will be able to make progress on the code for
the 64-bit support.
Val?rie
On Jun 11, 2007 18:41 +0200, Valerie Clement wrote:
> This patchset is also not complete:
> TODO:
> - update code under ext2ed/,
This code should just be removed entirely. It is dangerous to use, and
by including it in e2fsprogs it lends a false sense of credibility to
the code. Hopefully most of the functionality here has been incorporated
into debugfs?
> Code compiled, booted and briefly tested on a 20TB device (mkfs,
> debugfs, e2fsck, dumpe2fs, filefrag tested)
> Some non-regression tests also done on ext3 filesystems when compiling
> code without _EXT4FS_ option.
Did you run the "make check" tests? We've found these invaluable for
catching regressions. Ideally there would also be some test cases
for this support using "lazy_bg", a script to generate the filesystem,
and conditional upon being able to create sparse files that big.
Cheers, Andreas
--
Andreas Dilger
Principal Software Engineer
Cluster File Systems, Inc.
Andreas Dilger wrote:
> On Jun 11, 2007 18:41 +0200, Valerie Clement wrote:
>> This patchset is also not complete:
>> TODO:
>> - update code under ext2ed/,
>
> This code should just be removed entirely. It is dangerous to use, and
> by including it in e2fsprogs it lends a false sense of credibility to
> the code. Hopefully most of the functionality here has been incorporated
> into debugfs?
>
OK.
>> Code compiled, booted and briefly tested on a 20TB device (mkfs,
>> debugfs, e2fsck, dumpe2fs, filefrag tested)
>> Some non-regression tests also done on ext3 filesystems when compiling
>> code without _EXT4FS_ option.
>
> Did you run the "make check" tests? We've found these invaluable for
> catching regressions. Ideally there would also be some test cases
> for this support using "lazy_bg", a script to generate the filesystem,
> and conditional upon being able to create sparse files that big.
>
I missed to mention that some changes have to be done under tests/ too.
Without changes, "make check" fails.
Val?rie