Re: [PATCH 1/9] writeback: move dirty inodes from super_block to backing_dev_info

[Christoph Hellwig <hch@infradead.org>]

On Wed, Aug 12, 2009 at 06:12:50PM +0200, Jens Axboe wrote:
> OK, so you'd prefer leaving the super block lists in place and rather
> have the super blocks hanging off the bdi?

That would solve the above problem.  It would also implicitly provide
increased locality by always writing batches of dirty inodes per fs.

> What about file systems that
> support more than one block device per mount, like btrfs?

Or XFS :)

> Can we assume
> that they will forever provide a single bdi backing? btrfs currently has
> this, just wondering about future implications.

I don't see any point to assume things are forever.  For making progress
on this and getting it merged in .32 making that assumption is a good
one IMHO.

Now the question about that to do with a filesystem on multiple actual
backing device is an interesting one.  What about the case of having
btrfs just one half of two disks each?  Or same with a "normal" fs
ontop of LVM/MD?  Maybe in the end one thread(-pool) per filesystem
and not just per backing dev is the way forward, with the threads
schedule so that they don't interfer if they operate on the same
backing dev?


> 
> -- 
> Jens Axboe
> 
---end quoted text---