On Sun, Mar 06, 2005 at 05:38:09PM -0800, Chuck Silvers wrote:
>
>  - the read code should make sure that medium size (more than a page, smaller
>    than MAXPHYS) reads use the minimum number of disk I/Os (ie. 1) to fetch
>    the data.

This is actually a symptom of our discarding read-size information in
the middle of the VM system, no?  So, basically, every read looks like
a single-page fault (which is magically turned into two pages) but
readahead sometimes "rescues" us by causing a 64k read instead, which
is often what we want.

I have always thought that this was a pretty bogus way to do things!
The underlying interfaces can get multiple pages at once -- but between
read() and the filesystem, we "forget" that we wanted them.  Yuck.

-- 
 Thor Lancelot Simon	                                      tls@rek.tjls.com

"The inconsistency is startling, though admittedly, if consistency is to be
 abandoned or transcended, there is no problem."		- Noam Chomsky