Subject: Re: Testers needed for upcoming 3.1 release
To: Joachim Thiemann <joachim.thiemann@gmail.com>
From: Michael Lorenz <macallan@netbsd.org>
List: port-macppc
Date: 09/17/2006 16:09:51
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

On Sep 17, 2006, at 11:08, Joachim Thiemann wrote:

> My B&W G3 just had a kernel panic dropping into db this morning, with
> the 3.1_RC2 I installed yesterday.
>
> I was just starting a cvs update of my pkgsrc tree (cvs -q update -P
> -d as user "source") when:
>
> trap: kernel write DSI trap @ 0xdd1b5318 by 0x3606fc (DSISR 
> 0x42000000, err=14)
> panic: trap
> Stopped in pid 10291.1 (cvs) at netbsd:cpu_Debugger+0x10: lwz r0, r1, 0
>
> Note that this _could_ be hardware related: I have the rev.1 B&W G3,
> and thus the IDE chip that is a bit brain-damaged.  Usually my custom
> kernel restricts IDE transfer mode to (IIRC) multibyte DMA, that is,
> disallow UDMA.  This usually affects the slave drive only, on which
> the pkgsrc tree is located.  The system and swap are on the master
> drive, which should(?) be OK.

The IDE chip doesn't support UDMA anyway and allowing DMA on my beige 
G3 leads to severe data corruption on the disk and in kernel memory 
which ultimately ends in a DSI or a MCHK trap. Please disable DMA for 
now and see if it remains stable. I put a workaround into -current 
which simply disables DMA on heathrow-ata, if the problem goes away 
with DMA disabled I'll have it pulled into 3.1.

have fun
Michael
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (Darwin)

iQEVAwUBRQ2rkMpnzkX8Yg2nAQIrSAgApVlhxPplMOEj34n3DwolzZFCBqWTcUUm
mDC3KcErRI/2ZLLE2/Pn9+5no1YrdT9RPLtPRcG74FG4qJe86h6DxicmhjFm8A8P
8P2XbSup8YeyaGngk44VKiTVCReMO49TRWYvgX1nEz3gGEXCF8Nuvji7w1eJ6yeC
yTGzCPPicIV2yPusNzhVQo1WWC4qrx1nD/jC5s0PcBYPaU6CBon8mEgsbyzTGkPV
XZpU+1+p1UpQIaxyoSfOKd2H7eh1wGFhdZMuGNtvZTTSjy9bvi6gL2RKUajWF/Cr
trDI6E6Luht7TbLYFyHfe+irRbgRBe8noQrXHdfj4SmQfsJvtLQmrg==
=oxj2
-----END PGP SIGNATURE-----