Subject: Re: xinit causes kernel panic
To: None <port-sparc@netbsd.org>
From: Masahiro =Rhythm Drive= Ishiyama <masahiro@isl.rdc.toshiba.co.jp>
List: port-sparc
Date: 06/22/1999 17:35:53
>>>>> On Fri, 18 Jun 1999 00:24:33 -0700, Jason Thorpe <thorpej@nas.nasa.gov> said:

 > I'm wondering if the prestoserve is getting counted as RAM, and that
 > is having some sort of Bad Effect when accessed as normal RAM.

	What do you mean by "normal RAM"? You mean framebuffer?
	Is there any way to check it?
	
	I found that kernel panics on dereferencing softc structure in
	cgsixioctl():

        case FBIOGATTR:
#define fba ((struct fbgattr *)data)
                fba->real_type = sc->sc_fb.fb_type.fb_type;  <= here

	because sc points a bad address. I tryed to observe sc's value.  
	It is 0xf0536000 on cgsixattach(), and it is changed to 0x536000 on
	cgsixioctl() if SS20 has a prestoserve.
	I'm wandering that somebody writes first byte of sc, 
	i.e. cgsix_cd.cd_devs[minor(dev)]

	According to a dmesg, prestserve is not configured, and it is probed
	after cgsix.  Is there any possibility that probing prestserve
	has a bad side effect?

  --------------------------------------------------------------------------
    When you hear the music, after over it, | Masahiro Ishiyama
    it's gone into the air.                 | Communication Platform
    You can never capture it again.         | Laboratory, R&D Center, 
                          --- Eric Dolphy   | TOSHIBA CORPORATION
  --------------------------------------------------------------------------