Adaptec 4 port ethernet not working

Hi,

I'm trying to get a 4 port Adaptec AN-6944A/TX ethernet
card working on a 
macppc machine (which is a Motorola Starmax). In dmesg I
see:

ppb0 at pci0 dev 13 function 0: Digital Equipment DC21152
PCI-PCI Bridge (rev. 0x03)
pci1 at ppb0 bus 1
pci1: i/o space, memory space enabled
tlp0 at pci1 dev 4 function 0: DECchip 21140A Ethernet, pass
2.2
tlp0: unable to map device registers
tlp1 at pci1 dev 5 function 0: DECchip 21140A Ethernet, pass
2.2
tlp1: broken MicroWire interface detected; setting SROM size
to 1Kb
tlp1: sorry, unable to handle your board
tlp2 at pci1 dev 6 function 0: DECchip 21140A Ethernet, pass
2.2
tlp2: broken MicroWire interface detected; setting SROM size
to 1Kb
tlp2: sorry, unable to handle your board
tlp3 at pci1 dev 7 function 0: DECchip 21140A Ethernet, pass
2.2
tlp3: broken MicroWire interface detected; setting SROM size
to 1Kb
tlp3: sorry, unable to handle your board

Apparently, this board works in other machines running
NetBSD going back 
to some version of 1.5, but this is happening on both
NetBSD-3 and 
NetBSD-current.

Any ideas why this isn't working?

Thanks,
John Klos
-- 
Today is the day that my destiny calls me!
   Meowth

On Wed, Jul 05, 2006 at 04:30:00PM -0700, John Klos wrote:
> Hi,
> 
> I'm trying to get a 4 port Adaptec AN-6944A/TX
ethernet card working on a 
> macppc machine (which is a Motorola Starmax). In dmesg
I see:
> 
> ppb0 at pci0 dev 13 function 0: Digital Equipment
DC21152 PCI-PCI Bridge 
> (rev. 0x03)
> pci1 at ppb0 bus 1
> pci1: i/o space, memory space enabled
> tlp0 at pci1 dev 4 function 0: DECchip 21140A Ethernet,
pass 2.2
> tlp0: unable to map device registers

Hmm, interesting, perhaps OFW is not setting things behind
bridges up for us?
But in short, pci_mapreg_map(9) is failing somehow.

> tlp1 at pci1 dev 5 function 0: DECchip 21140A Ethernet,
pass 2.2
> tlp1: broken MicroWire interface detected; setting SROM
size to 1Kb
> tlp1: sorry, unable to handle your board
> tlp2 at pci1 dev 6 function 0: DECchip 21140A Ethernet,
pass 2.2
> tlp2: broken MicroWire interface detected; setting SROM
size to 1Kb
> tlp2: sorry, unable to handle your board
> tlp3 at pci1 dev 7 function 0: DECchip 21140A Ethernet,
pass 2.2
> tlp3: broken MicroWire interface detected; setting SROM
size to 1Kb
> tlp3: sorry, unable to handle your board

I'd like to see the output of the debugging code near here.
tlp_srom_debug
needs to be non-zero. (short pause while I go off to find
the email
that explained how) Backup your kernel, then:

gdb --write $TEST_KERNEL
set tlp_srom_debug=1
quit

and then boot $TEST_KERNEL.

> 
> Apparently, this board works in other machines running
NetBSD going back 
> to some version of 1.5, but this is happening on both
NetBSD-3 and 
> NetBSD-current.
> 
> Any ideas why this isn't working?

Something (hardware, firmware, kernel) is not handling the
PCI bus
properly.

And why don't we see identical errors for all the tlp(4)s
behind the
ppb(4)?

	Jonathan Kollasch

> Hmm, interesting, perhaps OFW is not setting things
behind bridges up for us?
> But in short, pci_mapreg_map(9) is failing somehow.

I douubt that OFW knows about bridges other than the ones
that shipped in 
Macs...

> I'd like to see the output of the debugging code near
here. tlp_srom_debug
> needs to be non-zero. (short pause while I go off to
find the email
> that explained how) Backup your kernel, then:
>
> gdb --write $TEST_KERNEL
> set tlp_srom_debug=1
> quit
>
> and then boot $TEST_KERNEL.

Thanks. Good flag to know about. Here're my dmesg snippets:

ppb0 at pci0 dev 13 function 0: Digital Equipment DC21152
PCI-PCI Bridge 
(rev. 0x03)
pci1 at ppb0 bus 1
pci1: i/o space, memory space enabled
tlp0 at pci1 dev 4 function 0: DECchip 21140A Ethernet, pass
2.2
tlp0: unable to map device registers
tlp1 at pci1 dev 5 function 0: DECchip 21140A Ethernet, pass
2.2
tlp1: broken MicroWire interface detected; setting SROM size
to 1Kb
SROM CONTENTS:
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
tlp1: sorry, unable to handle your board
tlp2 at pci1 dev 6 function 0: DECchip 21140A Ethernet, pass
2.2
tlp2: broken MicroWire interface detected; setting SROM size
to 1Kb
SROM CONTENTS:
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
tlp2: sorry, unable to handle your board
tlp3 at pci1 dev 7 function 0: DECchip 21140A Ethernet, pass
2.2
tlp3: broken MicroWire interface detected; setting SROM size
to 1Kb
SROM CONTENTS:
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
         0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff
tlp3: sorry, unable to handle your board


> Something (hardware, firmware, kernel) is not handling
the PCI bus
> properly.

Ya, something's amiss...

> And why don't we see identical errors for all the
tlp(4)s behind the
> ppb(4)?

Good question...

Thanks,
John
-- 
Today is the day that my destiny calls me!
   Meowth

John Klos wrote:
> 
> I douubt that OFW knows about bridges other than the
ones that shipped 
> in Macs...

Not necessarily true. I have 2 4 port Adaptec Starfire
cards. They have 
a PCI bridge that sets itself up correctly and 4 sf
interfaces, but it 
won't send any packets and anything that it receives is
garbage and 
mostly 0xDEADBEEF.

At 19:51 -0700 07/07/2006, John Klos wrote:

>I douubt that OFW knows about bridges other than the
ones that 
>shipped in Macs...

PPBs are a part of the PCI specification.  OFW should know
about 
bridges and should properly setup a device tree even when
bridges are 
involved.   IIRC, something like 256 PPBs should be
supported for up 
to 1024 PCI slots.   However, most firmware implementations
are 
unlikely to support that many layers of PPBs.  Still, any
firmware 
that meets the PCI specification should properly support
some layers 
of PPBs.

Jeff Walther

At 20:34 -0500 07/07/2006, Jonathan A. Kollasch wrote:
>On Wed, Jul 05, 2006 at 04:30:00PM -0700, John Klos
wrote:

>>  I'm trying to get a 4 port Adaptec AN-6944A/TX
ethernet card working on a
>>  macppc machine (which is a Motorola Starmax). In
dmesg I see:

>Something (hardware, firmware, kernel) is not handling
the PCI bus
>properly.

The x500 series of Macintoshes, which are different models
but same 
generation as the StarMax, have a definite firmware bug with
regard 
to properly handling PCI-PCI Bridges.

That's the 7500 through the 9600 and related clones.   The
symptom in 
those cases is that PPBs more than one layer deep cause
hangs during 
initialization.  There are several details and exceptions,
but that's 
the gist.   It seems likely that the Starmax would have
related 
bug(s).

How do I know it isn't hardware?   Well, the problem could
be in 
Bandit (CPU bus--PCI bridge), I suppose (but PPB hierarchies
should 
be handled in firmware or software,  not in hardware).  
However, it 
definitely isn't the PPB.   I  know this because I replaced
the DEC 
21052 in the Umax S900 (Mac Clone) with four different pin
compatible 
PPB chips (desoldered and soldered-in 160 pin QFPs) without
the 
symptoms changing.

It wasn't an OS issue because the hang occurs before
anything has 
loaded from disk.   It's purely a firmware driven issue.

So I would wager that the issue in this same-generation
machine is a 
firmware issue.

The ROM revision in the x500 machines and clones is
$77D.28F2.

Jeff Walther

On Fri, Jul 07, 2006 at 08:34:00PM -0500, Jonathan A.
Kollasch wrote:
> On Wed, Jul 05, 2006 at 04:30:00PM -0700, John Klos
wrote:
> > Hi,
> > 
> > I'm trying to get a 4 port Adaptec AN-6944A/TX
ethernet card working on a 
> > macppc machine (which is a Motorola Starmax). In
dmesg I see:
> > 
> > ppb0 at pci0 dev 13 function 0: Digital Equipment
DC21152 PCI-PCI Bridge 
> > (rev. 0x03)
> > pci1 at ppb0 bus 1
> > pci1: i/o space, memory space enabled
> > tlp0 at pci1 dev 4 function 0: DECchip 21140A
Ethernet, pass 2.2
> > tlp0: unable to map device registers
> 
> Hmm, interesting, perhaps OFW is not setting things
behind bridges up for us?
> But in short, pci_mapreg_map(9) is failing somehow.

That's been my experience.

What you need to do is somehow fix up the bridge.

Take care,

Bill

On Jul 18, 2006, at 3:01 PM, Bill Studenmund wrote:

> On Fri, Jul 07, 2006 at 08:34:00PM -0500, Jonathan A.
Kollasch wrote:
>> On Wed, Jul 05, 2006 at 04:30:00PM -0700, John Klos
wrote:
>>> Hi,
>>>
>>> I'm trying to get a 4 port Adaptec AN-6944A/TX
ethernet card  
>>> working on a
>>> macppc machine (which is a Motorola Starmax).
In dmesg I see:
>>>
>>> ppb0 at pci0 dev 13 function 0: Digital
Equipment DC21152 PCI-PCI  
>>> Bridge
>>> (rev. 0x03)
>>> pci1 at ppb0 bus 1
>>> pci1: i/o space, memory space enabled
>>> tlp0 at pci1 dev 4 function 0: DECchip 21140A
Ethernet, pass 2.2
>>> tlp0: unable to map device registers
>>
>> Hmm, interesting, perhaps OFW is not setting things
behind bridges  
>> up for us?
>> But in short, pci_mapreg_map(9) is failing somehow.
>
> That's been my experience.
>
> What you need to do is somehow fix up the bridge.

Here's what I had the last time I tried a similar card:

http://mail-index.netbsd.org/port-macppc/2004/12/13
/0048.html