7025-F40 Support

With the valued help of Mauricio Lima Pilla, who provided me
direct console
access to an F40 (and lots of manual power cycles), I have
been able to fix the
final bug that was preventing NetBSD from booting on the
7025-F40.  I have also
asked that this path be pulled up to NetBSD-4.0, so when the
release is finally
available, F40 support should be present.

One caveat though, The F40 has a secondary PCI bus, which we
do not currently
support.  This PCI bus covers two of the PCI slots on the
board.. (Possibly
slot 1 and 2).  I will need to write specialized support for
this bus, as it is
unlike anything else in prep.  This support is not likely to
make it into 4.0.

When the pullup is made, I will let the list know which
automated build will
contain support for it in 4.0_BETA.  If you are missing a
device with a card in
the slot, you might be on one of those special slots, and
you just need to move
the card.  I apologize for the trouble.

Also, at this time, support is only for uni-processor.  Dual
CPU will also not
likely make it for 4.0, but I will work on it in -current.

---
Tim Rightnour <rootgarbled.net>
NetBSD: Free multi-architecture OS http://www.netbsd.org/
Genecys: Open Source 3D MMORPG: http://www.genecys.org/

Great work on this - I have tried the 20060910 snapshot on
my F40
and it works great (except for slots 1&2 as you mention
below).

On Wed, Sep 06, 2006 at 11:21:39PM -0700, Tim Rightnour
wrote:
> One caveat though, The F40 has a secondary PCI bus,
which we do not currently
> support.  This PCI bus covers two of the PCI slots on
the board.. (Possibly
> slot 1 and 2).  I will need to write specialized
support for this bus, as it is
> unlike anything else in prep.  This support is not
likely to make it into 4.0.

I don't think slots 1&2 are on a secondary bus. 
According to the
IBM docs, slots 1-3 are in a primary bus and 4-9 are on a
secondary
bus (i.e. behind a ppb).  The 7043-140 is similar, with
slots 1&2
primary and 3-5 secondary.  What makes slots 1&2 unique
on F40's is
that they are 64-bit slots - do these need to be treated
differently
by NetBSD?

Here are boot messages from my F40, with three SCSI cards in
slots
1-3, and two Ethernet cards in slots 4&5.  The esiop0,
esiop1, and
pcn0 devices on pci0 are on-board devices.  The esiop2
device is the
PCI SCSI card in slot 3, and pcn1 and pcn2 are the PCI
Ethernet cards
in slots 4&5.  The PCI SCSI cards in slots 1&2 are
not detected.

NetBSD 4.0_BETA (GENERIC) #0: Mon Sep 11 05:19:42 UTC 2006
        buildspb:/home/builds/ab/netbsd-4/prep/200609100000Z-obj/ho
me/builds/ab/netbsd-4/src/sys/arch/prep/compile/GENERIC
Model: IBM PPS Model 7025 (ED)
total memory = 1024 MB
avail memory = 992 MB
OpenPIC Version 1.1: Supports 4 CPUs and 16 interrupt
sources.
timecounter: Timecounters tick every 10.000 msec
mainbus0 (root)
cpu0 at mainbus0: 604e (Revision 2.3), ID 0 (primary)
cpu0: HID0 c084<ICE,DCE,SGE,BHT>, powersave: 1
cpu0: 233.34 MHz
pnpbus0 at mainbus0
mcclock0 at pnpbus0: port 0x70-0x71, irq 8: mc146818
compatible time-of-day clock
nvram0 at pnpbus0: port 0x74-0x75 0x76
pci0 at mainbus0 bus 0: indirect configuration space access
pci0: i/o space, memory space enabled
pchb0 at pci0 dev 0 function 0
pchb0: IBM product 0x003a (rev. 0x00)
esiop0 at pci0 dev 1 function 0: Symbios Logic 53c825a (fast
wide scsi)
esiop0: using on-board RAM
esiop0: interrupting at irq 17
scsibus0 at esiop0: 16 targets, 8 luns per target
esiop1 at pci0 dev 2 function 0: Symbios Logic 53c825a (fast
wide scsi)
esiop1: using on-board RAM
esiop1: interrupting at irq 18
scsibus1 at esiop1: 16 targets, 8 luns per target
pcn0 at pci0 dev 3 function 0: AMD PCnet-PCI Ethernet
pcn0: Am79c970A PCnet-PCI II rev 6, Ethernet address
08:00:5a:ba:a9:fd
pcn0: interrupting at irq 19
pcn0: 10base5, 10base5-FDX, 10baseT, 10baseT-FDX, auto,
auto-FDX

ppb0 at pci0 dev 4 function 0: IBM 82351 PCI-PCI Bridge
(rev. 0x01)
pci1 at ppb0 bus 1
pci1: i/o space, memory space enabled
pcn1 at pci1 dev 0 function 0: AMD PCnet-PCI Ethernet
pcn1: Am79c971 PCnet-FAST rev 5, Ethernet address
00:20:35:12:64:e6
pcn1: interrupting at irq 25
nsphy0 at pcn1 phy 1: DP83840 10/100 media interface, rev. 1
nsphy0: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, auto
ukphy0 at pcn1 phy 31: Generic IEEE 802.3u media interface
ukphy0: OUI 0x000058, model 0x0001, rev. 1
ukphy0: 10baseT, 10baseT-FDX, auto
pcn2 at pci1 dev 1 function 0: AMD PCnet-PCI Ethernet
pcn2: Am79c971 PCnet-FAST rev 6, Ethernet address
00:06:29:6c:a3:73
pcn2: interrupting at irq 26
nsphy1 at pcn2 phy 1: DP83840 10/100 media interface, rev. 1
nsphy1: 10baseT, 10baseT-FDX, 100baseTX, 100baseTX-FDX, auto
ukphy1 at pcn2 phy 31: Generic IEEE 802.3u media interface
ukphy1: OUI 0x000058, model 0x0001, rev. 1
ukphy1: 10baseT, 10baseT-FDX, auto
esiop2 at pci0 dev 5 function 0: Symbios Logic 53c825a (fast
wide scsi)
esiop2: using on-board RAM
esiop2: interrupting at irq 20
scsibus2 at esiop2: 16 targets, 8 luns per target
pcib0 at pci0 dev 6 function 0: IBM ISA Bridge w/PnP (rev.
0x03)
pcib0: PIRQ[0-3] not used
IBM MPIC (undefined subclass 0x00) at pci0 dev 7 function 0
not configured
isa0 at pcib0
lpt0 at isa0 port 0x3bc-0x3bf irq 7
com0 at isa0 port 0x3f8-0x3ff irq 4: ns16550a, working fifo
com0: console
com1 at isa0 port 0x2f8-0x2ff irq 3: ns16550a, working fifo
pckbc0 at isa0 port 0x60-0x64
attimer0 at isa0 port 0x40-0x43: AT Timer
pcppi0 at isa0 port 0x61
isabeep0 at pcppi0
pcppi0: attached to attimer0
biomask 160000 netmask 61e0000 ttymask 61e0080
timecounter: Timecounter "clockinterrupt"
frequency 100 Hz quality 0
timecounter: selected timecounter
"clockinterrupt" frequency 100 Hz quality 0
timecounter: Timecounter "prep_mftb" frequency
16662777 Hz quality 0
timecounter: selected timecounter "prep_mftb"
frequency 16662777 Hz quality 0
Kernelized RAIDframe activated
scsibus0: waiting 2 seconds for devices to settle...
scsibus1: waiting 2 seconds for devices to settle...
scsibus2: waiting 2 seconds for devices to settle...


> Also, at this time, support is only for uni-processor. 
Dual CPU will also not
> likely make it for 4.0, but I will work on it in
-current.

Hey - you've got to walk before you can run 

On 11-Sep-2006 Cory Bajus wrote:
> I don't think slots 1&2 are on a secondary bus. 
According to the
> IBM docs, slots 1-3 are in a primary bus and 4-9 are on
a secondary
> bus (i.e. behind a ppb).  The 7043-140 is similar, with
slots 1&2
> primary and 3-5 secondary.  What makes slots 1&2
unique on F40's is
> that they are 64-bit slots - do these need to be
treated differently
> by NetBSD?

According to the residual data: 
http://www.garbled.net/tmp/residual/residual.7025-F40

There are three complete PCI bridges on the machine.  When
scanning the primary
host bridge, you see the pci-pci bridge on there.  However,
even scanning both
bridges, we never see the third bridge.  Given the fact that
they number the
bridge 128, and give a parent locator of 0 (supposedly
meaning it's
cpu-attached) I suspect this third bridge is actually a
second host bridge.  So
what I think this machine needs is:

pci0 at mainbus
ppb0 at pci0
pci1 at ppb0
pci2 at mainbus

Complications arise here, because the second primary bus
seems to have a
different base address than the primary bus, (which makes
sense actually).  So
I need to rewrite some of the attachment code in mainbus I
believe.

Note that in the residual, "Pci Slot 3" is noted
as being under the primary
bus, (DEVICE 2) and slots 1 and 2 are under this third,
wierd bus.  The bridge
is listed as having slots 4-9.

Given that you say they are 64 bit slots.. it wouldn't
surprise me at all if
IBM put a second primary bus in the machine to make them run
extra-fast.

I'd like to see the IBM docs you spoke of..  do you have a
part number for the
book, or have it in PDF form?

---
Tim Rightnour <rootgarbled.net>
NetBSD: Free multi-architecture OS http://www.netbsd.org/
Genecys: Open Source 3D MMORPG: http://www.genecys.org/

On Mon, Sep 11, 2006 at 02:44:27PM -0700, Tim Rightnour
wrote:
> On 11-Sep-2006 Cory Bajus wrote:
> > I don't think slots 1&2 are on a secondary
bus.  According to the
> > IBM docs, slots 1-3 are in a primary bus and 4-9
are on a secondary
> > bus (i.e. behind a ppb).  The 7043-140 is similar,
with slots 1&2
> > primary and 3-5 secondary.  What makes slots
1&2 unique on F40's is
> > that they are 64-bit slots - do these need to be
treated differently
> > by NetBSD?
> 
> According to the residual data: 
> http://www.garbled.net/tmp/residual/residual.7025-F40
> 
> There are three complete PCI bridges on the machine. 
When scanning the primary
> host bridge, you see the pci-pci bridge on there. 
However, even scanning both
> bridges, we never see the third bridge.  Given the fact
that they number the
> bridge 128, and give a parent locator of 0 (supposedly
meaning it's
> cpu-attached) I suspect this third bridge is actually a
second host bridge.  So
> what I think this machine needs is:
> 
> pci0 at mainbus
> ppb0 at pci0
> pci1 at ppb0
> pci2 at mainbus

OK, this makes sense.  I was confused by the description of
the missing
bus as a secondary bus - in IBM-speak a 'secondary' bus is
always behind
a PCI-PCI bridge.

> Complications arise here, because the second primary
bus seems to have a
> different base address than the primary bus, (which
makes sense actually).  So
> I need to rewrite some of the attachment code in
mainbus I believe.
> 
> Note that in the residual, "Pci Slot 3" is
noted as being under the primary
> bus, (DEVICE 2) and slots 1 and 2 are under this third,
wierd bus.  The bridge
> is listed as having slots 4-9.
> 
> Given that you say they are 64 bit slots.. it wouldn't
surprise me at all if
> IBM put a second primary bus in the machine to make
them run extra-fast.

That looks to be the case.  Multiple primary buses are very
common in the
newer CHRP systems as well (although I didn't realize the
F40 had two
primary buses).

> I'd like to see the IBM docs you spoke of..  do you
have a part number for the
> book, or have it in PDF form?

The IBM doc is the old RS/6000 'PCI Adapter Placement
Guide'.  It doesn't
have any useful technical info for probing the buses; it
just shows the
slot layout (primary/secondary, width, speed, voltage, etc.)
for the various
PCI-based RS/6000 models.