Mr Hawkinson,

Thank you for your bug report and analysis.

From: John Hawkinson <jhawk@MIT.EDU>
Subject: Re: kern/10318 -- CVS commit: syssrc
Date: Thu, 8 Jun 2000 16:33:41 -0400 (EDT)
Message-ID: <200006082033.QAA22210@contents-vnder-pressvre.mit.edu>

 > | Module Name:	syssrc
 > | Committed By:	haya
 > | Date:		Tue Jun  6 08:17:38 UTC 2000
 > | 
 > | Modified Files:
 > | 	  syssrc/sys/arch/i386/i386: rbus_machdep.c
 > | 
 > | Log Message:
 > | Use entire extent region when PCIBIOS_ADDR_FIXUP is set.  PCIBIOS
 > | provides PCI-dedicated memory and io extent.
 > | 
 > | To generate a diff of this commit:
 > | cvs rdiff -r1.5 -r1.6 syssrc/sys/arch/i386/i386/rbus_machdep.c
 > |
 > Hayakawa-san:
 > 
 > This breaks cardbus on my Sony VAIO (in weird ways), where
 > PCIBIOS_ADDR_FIXUP is set. Please see PR 10318 (copy sent to you under
 > seperate cover).

I saw it.  It looks some address space confliction occurs.
I met the same situation with another machine (Toshiba
Dynabook SS 3020).  Occasionally, there are some hidden
devices (I'm not sure about it) just on system memory area.


 > What's the right way to investigate this and should we back it
 > out in the meantime?

Right.  We must avoid using memory space on the system
memory.


 > Somewhat unrelatedly, I think that pcibios(4) needs to be a lot more
 > clear about whether users should turn on the *FIXUPs by default, or
 > whether they should only do so in special circumstances.

Exactly.
-- 
HAYAKAWA, Koichi