Re: [Jfs-discussion] direct-to-BIO pagecache IO

[Christoph H. <hc...@in...>]

On Mon, Sep 02, 2002 at 12:55:51AM -0700, Andrew Morton wrote:
> It may be beneficial to implement a smarter jfs_get_blocks() in
> that case.  The one I put there is just a degenerate version to
> keep direct-IO happy.  But if you change it to map more disk,
> it will be called less often during direct IO.
> 
> It's pretty simple - it's passed the number of blocks to attempt
> to map, and it returns the amount of contiguous disk which it _did_
> map in bh->b_size.

It's already in the tree.  Btw, is there any chance to make ->readpages
and ->writepages take a get_blocks callback?  This would help extent
based filesystems.