On Tue, Sep 28, 2004 at 12:42:26PM -0700, Jason Thorpe wrote:
> 
> On Sep 28, 2004, at 12:38 PM, Manuel Bouyer wrote:
> 
> >The check against 0xffffff was added specifically so that with drives
> >supporting LBA48, we could still access the first 128GB if the 
> >controller
> >didn't support LBA48 (like some promise controllers).
> 
> Oh, I didn't realize that the controller has to also support LBA48 
> operation (silly me for thinking that they just passed all those 
> register accesses through directly to the drive...).  "Ok, nevermind!" 
> :-)

You're perfectly correct, lba48 limitation is mostly artificial made by
vendors, e.g. promise cards Manuel has mentioned can be fixed with bios
upgrade.

btw, on my dumb sectors reading perfomance tests there's some benefit in
not using lba48 for short transfers since lba48 commands require 2 times
more io registers writes.

> 
>         -- Jason R. Thorpe <thorpej@shagadelic.org>
> 



-- 
   Alexander Yurchenko (aka grange)