Subject: Re: MAXPHYS
To: None <wojtek@wojtek.from.pl>
From: Thor Lancelot Simon <tls@rek.tjls.com>
List: port-i386
Date: 02/04/2001 17:40:10
On Sun, Feb 04, 2001 at 10:36:40PM +0100, wojtek@wojtek.from.pl wrote:
> > becomes a problem.
> > 
> > Finally, things like RAIDframe *really* want the old MAXPHYS limit to go
> > away so that they don't ultimately send 8k transfers to the disks.
> couldn't be #define MAXPHYS changed to int tell_maxphys(device_major)???
> 
> pciide will tell 64kB unless modified, ccd depending of
> stripe size and value from disks, raidframe a_very_big_value etc., scsi
> drivers depending of implement etc..

It's not that simple.  The maximum transfer size may be limited not by
the device you're talking to but by, for example, the bus you're talking
to it across.  But in general, yes, I think that we should try to migrate
to this kind of approach.

Note that it was essentially pointless to do so before the advent of UBC
because increasing MAXBSIZE beyond 64K would have been horrifically
inefficient.  At the moment, there seem to be some performance problems
with UBC in which it doesn't cluster up to MAXPHYS anyway, even with a
64k MAXPHYS, but those will surely be fixed soon; so it would seem to me
that replacing the constant MAXPHYS with a scheme that propagates this
kind of information through the device tree would be a good idea.