On Fri, Oct 06, 2006 at 11:05:13AM +0900, YAMAMOTO Takashi wrote:
> 
> i withdraw this because filesystems are free to read-in pages
> without explicit requests from upper layers.
> eg. when aligning i/o range to block boundaries.
>  

OK - so, we may need to read with spinlocks held though if I can get
this working via uvm_aio_iodone() it still may not be necessary to do
this dance still.

-- 
Brett Lymn