[Ocfs2-tools-devel] [PATCH]: Add removing slots support for ocfs2-tools

[tao.ma]

Sunil Mushran wrote:
> For testing, I was thinking of adding a "preferred slot" mount option.
> As in, mount the volume using that slot, if possible. This will make
> it easier for us to test "remove slots".
>
> Well, not thinking. I am adding it. But let me know if you have a
> better idea.
That would be really cool and very helpful.
But please note that currently ocfs2 kernel only use extent_alloc:0000 
to alloc extent blocks, so even if you can let mount use different 
inode_alloc, all the extent block will be in extent_alloc:0000. So you 
have to define "OCFS2_USE_ALL_METADATA_SUBALLOCATORS" and build a 
non-standard kernel.

So my current solution is a little tricky:
1. mount the volume and create test files using inode_alloc:0000 and 
extent_alloc:0000, umount.
2. Exchange the inode information between some specified inodes. e.g, if 
I want to let inode_alloc:0007 have some information, I will exchange 
the inode_alloc:0000 and inode_alloc:0007. Just swap the inode and don't 
touch the group descriptor and the corresponding "Sub Alloc Slot".
3. Run "fsck.ocfs2 -fy", this tool will modify the group descriptor and 
all the inode allocated information.
The same process goes with extent alloc.

I am still implementing the test script to test whether tunefs.ocfs2 can 
works OK with multiple chains and groups.
I also wrote a patch for adding some panic condition for tunefs.ocfs2, 
so that the tunefs.ocfs2 can panic at the specified place to check 
whether fsck.ocfs2 can works OK with the corruption.