Hauke Fath | 1 Dec 21:48 2010
Picon

Serial console unusable in -current?

All,

current kernels booted to serial console invariably drop into the kernel
debugger after having probed nubus:

Bootstrapping NetBSD/mac68k.
Getting mapping from MMU.
Loaded at 0x0
System RAM: 71303168 bytes in 17408 pages.
     Low = 0x0, high = 0x4400000
On-board video at addr 0x0xf9001000 (phys 0x0xf9001000), len 0xff000.
Done.
Bootstrapping the pmap system.
Pmap bootstrapped.
Moving ROMBase from 0x40800000 to 0x4ee000.
Video address 0x0xf9001000 -> 0x0x6ee000.
Loaded initial symtab at 0x355878, strtab at 0x3ac738, # entries 21436
Copyright (c) 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003, 2004, 2005,
    2006, 2007, 2008, 2009, 2010
    The NetBSD Foundation, Inc.  All rights reserved.
Copyright (c) 1982, 1986, 1989, 1991, 1993
    The Regents of the University of California.  All rights reserved.

NetBSD 5.99.40 (DEBUG) #0: Wed Nov 24 18:21:20 CET 2010

hf <at> Hochstuhl:/var/obj/netbsd-builds/developer/mac68k/sys/arch/mac68k/compile/DEB
UG
Apple Macintosh Quadra 700  (68040)
cpu: delay factor 800
fpu: mc68040
(Continue reading)

Hauke Fath | 2 Dec 22:25 2010
Picon

Quiescing Nubus DMA?

All,

given the size of the audience, this question should probably go to
tech-kern, and may eventually. Still...

I have (once again) taken up toying with an FWB Jackhammer Nubus SCSI card,
which features a 53c720 controller. It looks like the 53c720 is active even
after the kernel has taken over, punching random holes into memory early
during the boot and way before Nubus is probed. I've looked at the sn(4)
DMA code, but we have other means of learning early up that the
corresponding DMA hardware is present, and silence it, so that does not
really help.

What is the general NetBSD take on this, short of hacking the booter into
shutting down the 53c720 DMA before handing control to the kernel?
Unfortunately, mac68k hardware does not go through a full reset sequence
before booting into NetBSD.

Puzzled,
	hauke

--
"It's never straight up and down"     (DEVO)

Donald Lee | 3 Dec 01:08 2010

Re: Quiescing Nubus DMA?

It's hard to imagine any OS running reliably if it does not "take charge"
of such a peripheral card at boot time, to reset and initialize it.

How does MacOS do this?  I would think it would *have* to be done
in the ROM, since nothing else would be safe from the DMA engine.

I would expect that NuBus would have some sort of reset sequence, and
that would *have* to "park" the card.

No help, but curious. ;->

-dgl-

>All,
>
>given the size of the audience, this question should probably go to
>tech-kern, and may eventually. Still...
>
>I have (once again) taken up toying with an FWB Jackhammer Nubus SCSI card,
>which features a 53c720 controller. It looks like the 53c720 is active even
>after the kernel has taken over, punching random holes into memory early
>during the boot and way before Nubus is probed. I've looked at the sn(4)
>DMA code, but we have other means of learning early up that the
>corresponding DMA hardware is present, and silence it, so that does not
>really help.
>
>What is the general NetBSD take on this, short of hacking the booter into
>shutting down the 53c720 DMA before handing control to the kernel?
>Unfortunately, mac68k hardware does not go through a full reset sequence
>before booting into NetBSD.
(Continue reading)

Hauke Fath | 3 Dec 11:42 2010
Picon

Re: Quiescing Nubus DMA?

At 18:08 Uhr -0600 02.12.2010, Donald Lee wrote:
>It's hard to imagine any OS running reliably if it does not "take charge"
>of such a peripheral card at boot time, to reset and initialize it.

Well, NetBSD/mac68k does that, to the extent that it knows the hardware.

>How does MacOS do this?  I would think it would *have* to be done
>in the ROM, since nothing else would be safe from the DMA engine.

The Nubus is sent a power-on reset signal, and the ROM startup routine
sends another two reset pulses, to be sure. Since the Mac expects to be
able to boot from the Jackhammer, it sets up the card pretty early on.

>I would expect that NuBus would have some sort of reset sequence, and
>that would *have* to "park" the card.

It does, and that would. According to DC&D, the Nubus reset line can be
toggled by software, so that would be a quick and dirty way to discipline
the Jackhammer.

The main problem with applying this sledgehammer is that we lose access to
any Nubus graphics cards, which only MacOS (rather: the board's ROM driver,
on behalf of the Macintosh ROMs) knows how to set up. AFAICS, that is the
main reason why native boots (i.e. bypassing the Macintosh operating
system) never got any traction on !OF Macintoshes.

	hauke

--

-- 
Hauke Fath                        <hauke <at> Espresso.Rhein-Neckar.DE>
(Continue reading)

Hauke Fath | 3 Dec 12:09 2010
Picon

[Link] TI Nubus Interface Products

All,

I just came across

<http://www.bitsavers.org/pdf/ti/2242825-0001_NuBus_Spec1983.pdf>

and especially

<http://computer-refuge.org/bitsavers/pdf/ti/_dataBooks/Nubus_Interface_Products_1991.pdf>

which might be interesting - I have data sheets for some of TI's Nubus
interface chips (used on e.g. the Jackhammer), but I have not seen a copy
of the complete data book online before, which appears to have more Nubus
specifics, and application notes.

	hauke

--

-- 
Hauke Fath                        <hauke <at> Espresso.Rhein-Neckar.DE>
Friedrich-Ebert-Straße 70
64347 Griesheim
Germany

Hauke Fath | 3 Dec 13:21 2010
Picon

[Link] CERN - Developing for the Macintosh Nubus

The CERN document at

<http://cdsweb.cern.ch/record/200836/files/CM-P00062891.pdf>

is an interesting hands-on summary of hardware oriented programming on the Mac.

	hauke

--

-- 
Hauke Fath                        <hauke <at> Espresso.Rhein-Neckar.DE>
Friedrich-Ebert-Straße 70
64347 Griesheim
Germany

Donald Lee | 3 Dec 18:08 2010

Re: Quiescing Nubus DMA?

>The main problem with applying this sledgehammer is that we lose access to
>any Nubus graphics cards, which only MacOS (rather: the board's ROM driver,
>on behalf of the Macintosh ROMs) knows how to set up. AFAICS, that is the
>main reason why native boots (i.e. bypassing the Macintosh operating
>system) never got any traction on !OF Macintoshes.

This stuff is never as simple as it seems.

I guess that's why, when you confuse a sledgehammer with a
jackhammer, you break stuff. ;->

-dgl-

Hauke Fath | 4 Dec 22:18 2010
Picon

Re: Serial console unusable in -current?

At 21:48 Uhr +0100 1.12.2010, Hauke Fath wrote:
>current kernels booted to serial console invariably drop into the kernel
>debugger after having probed nubus:

Scratch that - the 5_99_40 serial console works fine with a printer cable
and ZTerm, so it's comms/conserver8 and the sparc acting up.

	hauke

--
"It's never straight up and down"     (DEVO)

carlidbifriends | 7 Dec 03:02 2010
Picon

You have received an important Message!

You have received an important Message! Check your message here:
http://gabyone.zoomshare.com/files/hottie.htm

carlidbifriends | 8 Dec 12:40 2010
Picon

I have added you to my friends network today!

I created this cool friends network and added you to my friends network. Hit-up now:
http://girlaadvance.zoomshare.com/files/photos.htm


Gmane