Re: [patch 12/16] fix race between writeback and unlink

[Andrew Morton <akpm@zip.com.au>]

Linus Torvalds wrote:
> 
> ...
> I'd still love to see a "fast and slightly stupid" allocator for both
> blocks and inodes, and have some infrastructure to do run-time defragging
> in the background.
> 

I think runtime defrag could yield really good benefits.  In
particular it would allow us to find_group_dir(), and always
put directories in the same blockgroup as their parent (big
speedups for the `untar-a-kernel-tree' workload).

There's a patch at
http://www.zip.com.au/~akpm/linux/patches/2.4/2.4.19-pre10/ext3-reloc-page.patch
which provides a simple `relocate page' ioctl for ext3 files.  It
relocates a page's blocks.  The operation is fully journalled and
pagecache-coherent.  So you can turn off the power in the middle
of a defrag operation and the fs will come back just fine.  It doesn't
make any attempt to relocate inodes.  If the page relocation attempt fails
then it just returns -EAGAIN and userspace gets to worry about what
to do.

I simply have not had the time to do anything about the userspace
program which drives that ioctl.  So if there's anyone out there
who has a little time on their hands...

-