Re: HPPA support for IGNITE-UX install discs

Re: HPPA support for IGNITE-UX install discs

[Helge Deller]

On 21.09.20 05:41, Philippe Mathieu-Daudé wrote:
> Cc'ing the HPPA team.

CC'ing parisc linux development list.

> On 9/20/20 9:42 PM, Keith Monahan via wrote:
>> HI there,
>>
>> In QEMU 5.1.50 (also tried 4.2.1) built from source, I have a particular
>> IGNITE-UX disk associated with an HP Logic Analyzer 16700A that I'm
>> trying to boot and install. It's based on HP-UX 10.20. The regular LA
>> hardware shows up via uname as "HP-UX no_hostn B.10.20 C 9000/779
>> 2008038539 32-user license" I've seen reference to 9000/779/rock. This
>> should be fairly similar to the B132L/B160L.

The 9000/779 seems to be like an RDI PrecisionBook:
https://www.openpa.net/systems/rdi_precisionbook.html
which is quite similiar to a B160L.

There is even an entry for your logic analyzer:
https://www.openpa.net/systems/hp_16600_16700_logic_agilent.html

>> While SeaBIOS identifies the attached virtual drive, and the specified
>> CDROM boots, the install init script fails to find any IO devices, and
>> so no installation can occur.
>>
>> Other HP-UX 10.20 Ignite-UX disks work without issue in the same qemu
>> environment with the same command line options.
>>
>> A successful run has the "Scanning system for IO devices..." take a few
>> seconds, followed by a "Querying disk device" message for each of the
>> devices it sees. That whole process takes 30 seconds or so in emulation.
>>
>> A bad run immediately goes from "Scanning system..." to "There were no
>> disk devices found during the scan."
>>
>> I have noticed that the "install/init" Revision on one that works is a
>> 10.3, but the failing one is 10.124. Not sure if this telltale or not.
>>
>> Anyone super familiar with this type of install disk? Or how the boot
>> process works with HP-UX CDs in general? What gets loaded first? How can
>> I find this script and/or executable that's evaluating the readiness of
>> the drives? My thoughts are that I could "update" via editing the ISO,
>> which both mount easily in Linux.

In qemu the B160L emulation isn't complete yet.
The SCSI disc is emulated by a "virtual built-in" PCI SCSI card, while
the original B160L has a NCR 53C710 SCSI chip internal in the LASI controller.
Additionally the LASI emulation isn't complete either.
My assumption is, that your Install CD for the analyzer will only detect
real LASI/53C710 chips and ignore the virtual PCI card. That's why it doesn't
detect any discs.

>> But I'm open to thoughts, including redirection to a different more
>> appropriate mailing list! :)
>>
>> Thanks for any general advice.
>> Keith
>>
>>
>> I'm invoking like this:
>>
>> ./qemu-system-hppa -boot d -serial mon:stdio -drive
>> if=scsi,bus=0,index=6,file=./hpux.img,format=raw -D /tmp/foo -nographic
>> -m 512 -d nochain -cdrom ./HP16700_cdrom.iso -net nic,model=tulip  -net tap

You could drop the "nochain", it slows things down.
Without "nographic" the STI framebuffer might work.

>> and see this:
>>
>>   Primary boot path:    FWSCSI.6.0
>>   Alternate boot path:  FWSCSI.2.0

both are on the virtual PCI card...

>>   Console path:         SERIAL_1.9600.8.none
>>   Keyboard path:        PS2
>>
>> Available boot devices:
>> 1. DVD/CD [lsi 00:00.0 2:0 Drive QEMU QEMU CD-ROM 2.5+]
>> 2. lsi 00:00.0 6:0 Drive QEMU QEMU HARDDISK 2.5+
>>
>> Booting from DVD/CD [lsi 00:00.0 2:0 Drive QEMU QEMU CD-ROM 2.5+]
>>
>> Booting...
>> Boot IO Dependent Code (IODC) revision 153
>>
>> HARD Booted.
>>
>> ISL Revision A.00.38  OCT 26, 1994
>>
>> ISL booting  hpux (;0):INSTALL
>>
>> Boot
>> : disc(8/0/0/0.2.0;0):INSTALL
>> 4999156 + 446464 + 361504 start 0x184268
>> No BTLB entries found for processor 0
>> vuseg=1a5a000
>>
>>     System Console is on the Built-In Serial Interface
>> Networking memory for fragment reassembly is restricted to 49061888 bytes
>>     Swap device table:  (start & size given in 512-byte blocks)
>>         entry 0 - auto-configured on root device; ignored - no room
>> WARNING: No dump device configured.  Dump is disabled.
>> Starting the STREAMS daemons.
>>     9245XB HP-UX (B.10.20) #1: Sun Jun  9 06:31:19 PDT 1996
>>
>> Memory Information:
>>     physical page size = 4096 bytes, logical page size = 4096 bytes
>>     Physical: 524288 Kbytes, lockable: 404768 Kbytes, available: 465588
>> Kbytes
>>
>> =======  09/19/20 16:01:14 EDT  HP-UX Installation Initialization. (Sat
>> Sep 19
>>          16:01:14 EDT 2020)
>>          @(#) Ignite-UX Revision 1.32
>>          @(#) install/init (opt) $Revision: 10.124 $
>>        * EISA configuration has completed.  Following the completion of a
>>          successful HP-UX installation, please check the
>> "/etc/eisa/config.err"
>>          file for any EISA configuration messages.
>>        * Scanning system for IO devices...
>> NOTE:    There were no disk devices found during the scan.  Make sure
>> that the
>>          destination disks are connected and powered on.  You may choose to
>>          scan for more disk drives from next menu on the console.
>> WARNING: Could not verify access to LAN interface: 8/0/1/0
>> ERROR:   could not find source device 8/0/0/0 in ioscan output

That's strange.... I think 8/... might be LASI.

>> Would you like to switch source to network? (y/[n]): y
>>
>>
>> Hardware Summary:         System Model: 9000/778/B160L
>>   +---------------------+----------------+-------------------+ [ Scan
>> Again  ]
>>   | Disks: 0  (  0.0GB) |  Floppies: 0   | LAN cards:   1    |
>>   | CDs:   0            |  Tapes:    0   | Memory:    512Mb  |
>>   | Graphics Ports: 0   |  IO Buses: 1   |                   | [ H/W
>> Details ]
>>   +---------------------+----------------+-------------------+

I'm not an HP-UX expert.
Maybe it's possible to install HP-UX from another install CD and
install the Analyzer packages afterwards?

Helge

Re: HPPA support for IGNITE-UX install discs

[Keith Monahan]

Helge,

Thanks for looking a this!

Responses in-line


On 9/21/2020 10:46 AM, Helge Deller wrote:

 > The 9000/779 seems to be like an RDI PrecisionBook:
 > https://www.openpa.net/systems/rdi_precisionbook.html
 > which is quite similiar to a B160L.
 >
 > There is even an entry for your logic analyzer:
 > https://www.openpa.net/systems/hp_16600_16700_logic_agilent.html

I haven't heard of those precisionbooks! Portable HP-UX. Pretty cool!

My conversations with Paul W over there at the beginning of the year led 
to the creation of that page. He did all the hard work, I provided a 
minor correction here or there.

 > In qemu the B160L emulation isn't complete yet.
 > The SCSI disc is emulated by a "virtual built-in" PCI SCSI card, while
 > the original B160L has a NCR 53C710 SCSI chip internal in the LASI 
controller.
 > Additionally the LASI emulation isn't complete either.

Thanks for this.

I have another project I'm working on that is semi-related. The 
NCR53C710 megacell as implemented in LASI is limited to 5MB/s. This 
logic analyzer has no (physical) PCI slots, but has TE/AMP connectors 
that are electrically PCI. I'm attempting to make a small adapter that 
will convert that AMP to a standard 32-bit white PCI slot. The first 
thing I'd like to do is try a 53C875E card.....to increase disk speeds.

There's a card that sits on that same bus which is wired directly to a 
chip's PCI interface.....so I'm pretty sure this is the case.

I'm also pretty sure it's DINO that's bridging the processor/GSC bus to 
PCI and exposing it on these connectors.

I've created a breakout board that takes the TE/AMP connector and 
provides 20-pin connectors for a termination adapter to attach, which 
then goes to a(nother) logic analyzer.

 > My assumption is, that your Install CD for the analyzer will only detect
 > real LASI/53C710 chips and ignore the virtual PCI card. That's why it 
doesn't
 > detect any discs.

Right. I wonder if it's simply the matter of drivers. Do you know if 
this scan is actually "ioscan -C disk" ?

Here's the ioscan from the actual hardware

no_hostname:/> ioscan
H/W Path    Class                   Description
===============================================
             bc
8           bc                      Pseudo Bus Converter
8/0             ba                  PCI Bus Bridge
8/0/15.0              instrument    PCI(103c1650)
8/16            ba                  Core I/O Adapter
8/16/0             ext_bus          Built-in Parallel Interface
8/16/1             audio            Built-in Audio
8/16/4             tty              Built-in RS-232C
8/16/5             ext_bus          Built-in SCSI
8/16/5.0              target
8/16/5.0.0               disk        codesrc         SCSI2SD
8/16/5.6              target
8/16/5.6.0               disk        codesrc         SCSI2SD
8/16/5.7              target
8/16/5.7.0               ctl        Initiator
8/16/6             lan              Built-in LAN
8/16/7             ps2              Built-in Keyboard/Mouse
8/16/10            pc               Built-in Floppy Drive
8/16/10.1             floppy        HP_PC_FDC_FLOPPY
8/24            graphics            Graphics
8/27            unknown
62          processor               Processor
63          memory                  Memory


 >>>         * Scanning system for IO devices...
 >>> NOTE:    There were no disk devices found during the scan.  Make sure
 >>> that the
 >>>           destination disks are connected and powered on.  You may 
choose to
 >>>           scan for more disk drives from next menu on the console.
 >>> WARNING: Could not verify access to LAN interface: 8/0/1/0
 >>> ERROR:   could not find source device 8/0/0/0 in ioscan output
 >
 > That's strange.... I think 8/... might be LASI.
 >
 > Maybe it's possible to install HP-UX from another install CD and
 > install the Analyzer packages afterwards?
 >
 > Helge
 >

Well it likely is. What I'm trying to do, however, have a sandbox that 
very closely matches the LA that I can play around with, reset, break, 
try again, and so on. There's other ways of accomplishing this, but most 
involve physical hardware, re-burning SD cards, time, presence, and so on.

For my purposes, I don't know how much the LA's HP-UX installation 
matches a standard one. And having the LA software might be nice as well.

For what it's worth, and I'm guessing you guys also have the bug --- 
there's just something about taking a 25-year old CD, popping it in 
software on a modern computer, and duplicating the experience. The mix 
of old and new just does it for me! Hard to explain, but it's just so cool!

Thanks
Keith

Re: HPPA support for IGNITE-UX install discs

[Helge Deller]

Somewhat late reply...

On 9/22/20 2:28 AM, Keith Monahan wrote:
> Here's the ioscan from the actual hardware
> 
> no_hostname:/> ioscan
> H/W Path    Class                   Description
> ===============================================
>             bc
> 8           bc                      Pseudo Bus Converter
> 8/0             ba                  PCI Bus Bridge
> 8/0/15.0              instrument    PCI(103c1650)

That seems to be a specific HP PCI card.
Maybe it's possible to pass-through it at some point in an emulation ?

> 8/16            ba                  Core I/O Adapter
> 8/16/0             ext_bus          Built-in Parallel Interface
> 8/16/1             audio            Built-in Audio
> 8/16/4             tty              Built-in RS-232C
> 8/16/5             ext_bus          Built-in SCSI

^ this one isn't implemented yet in qemu.
As I said in another mail, we currently emulate a PCI SCSI card instead.
Maybe emulating the original SCSI controller isn't hard, but I don't know
and I'm not a SCSI expert.

> 8/16/5.0              target
> 8/16/5.0.0               disk        codesrc         SCSI2SD
> 8/16/5.6              target
> 8/16/5.6.0               disk        codesrc         SCSI2SD
> 8/16/5.7              target
> 8/16/5.7.0               ctl        Initiator
> 8/16/6             lan              Built-in LAN

^For that LASI LAN I do have code, but it doesn't work yet.
In HP-UX it fails the self-tests, but in Linux it works.
Still need to bring it upstream at some point.


> 8/16/7             ps2              Built-in Keyboard/Mouse
> 8/16/10            pc               Built-in Floppy Drive
> 8/16/10.1             floppy        HP_PC_FDC_FLOPPY
> 8/24            graphics            Graphics
> 8/27            unknown
> 62          processor               Processor
> 63          memory                  Memory
> 
> 
>>>>         * Scanning system for IO devices...
>>>> NOTE:    There were no disk devices found during the scan.  Make sure
>>>> that the
>>>>           destination disks are connected and powered on.  You may choose to
>>>>           scan for more disk drives from next menu on the console.
>>>> WARNING: Could not verify access to LAN interface: 8/0/1/0
>>>> ERROR:   could not find source device 8/0/0/0 in ioscan output

I think the SCSI controller is the biggest issue for now...
If that works, the installations should continue.

Helge

Re: HPPA support for IGNITE-UX install discs

[Keith Monahan]

Helge,

Thanks for your reply!


On October 26, 2020 4:01:52 PM Helge Deller <deller@gmx.de> wrote:

> Somewhat late reply...
>
> On 9/22/20 2:28 AM, Keith Monahan wrote:
>> Here's the ioscan from the actual hardware
>>
>> no_hostname:/> ioscan
>> H/W Path    Class Description
>> ===============================================
>>             bc
>> 8           bc                      Pseudo Bus Converter
>> 8/0             ba                  PCI Bus Bridge
>> 8/0/15.0              instrument PCI(103c1650)
>
> That seems to be a specific HP PCI card.
> Maybe it's possible to pass-through it at some point in an emulation ?

Not super concerned with emulating that hw there. That card is used to 
connect the PCI bus (in turn, connected to GSC via DINO) to the logic 
analyzer backplane.

>
>> 8/16            ba                  Core I/O Adapter
>> 8/16/0             ext_bus          Built-in Parallel Interface
>> 8/16/1             audio            Built-in Audio
>> 8/16/4             tty              Built-in RS-232C
>> 8/16/5             ext_bus          Built-in SCSI
>
> ^ this one isn't implemented yet in qemu.
> As I said in another mail, we currently emulate a PCI SCSI card instead.
> Maybe emulating the original SCSI controller isn't hard, but I don't know
> and I'm not a SCSI expert.

The NCR 53C710 SCSI that's present inside LASI was pretty common. One of 
the uses includes the Commodore A4091, a SCSI controller sold for the 
Amiga 4000. The A4091 is emulated on WinUAE, which is open source. To 
make my story go full circle, looking at that source, located below, is 
based on QEMU source!!

https://github.com/tonioni/WinUAE/blob/master/qemuvga/lsi53c710.cpp

which is based on

qemu/hw/scsi/lsi53c895a.c

I don't know the significance/complexity of the difference between 
emulating a PCI card vs accessing things via LASI, which would sit on 
the GSC bus. Wishful thinking is that once we get there, that we've got 
existing (albeit different system emulation platform) working code that 
emulates those chip functions.

so then maybe there's still hope for this cause eventually! :)

>
> I think the SCSI controller is the biggest issue for now...
> If that works, the installations should continue.
>
> Helge

Thanks,
Keith

Re: HPPA support for IGNITE-UX install discs

[Helge Deller]

On 10/28/20 1:29 PM, Keith Monahan wrote:
> On October 26, 2020 4:01:52 PM Helge Deller <deller@gmx.de> wrote:
>> On 9/22/20 2:28 AM, Keith Monahan wrote:
>>> Here's the ioscan from the actual hardware
>>>
>>> no_hostname:/> ioscan
>>> H/W Path    Class Description
>>> ===============================================
>>>             bc
>>> 8           bc                      Pseudo Bus Converter
>>> 8/0             ba                  PCI Bus Bridge
>>> 8/0/15.0              instrument PCI(103c1650)
>>
>> That seems to be a specific HP PCI card.
>> Maybe it's possible to pass-through it at some point in an emulation ?
>
> Not super concerned with emulating that hw there. That card is used to connect the PCI bus (in turn, connected to GSC via DINO) to the logic analyzer backplane.
>
>>
>>> 8/16            ba                  Core I/O Adapter
>>> 8/16/0             ext_bus          Built-in Parallel Interface
>>> 8/16/1             audio            Built-in Audio
>>> 8/16/4             tty              Built-in RS-232C
>>> 8/16/5             ext_bus          Built-in SCSI
>>
>> ^ this one isn't implemented yet in qemu.
>> As I said in another mail, we currently emulate a PCI SCSI card instead.
>> Maybe emulating the original SCSI controller isn't hard, but I don't know
>> and I'm not a SCSI expert.
>
> The NCR 53C710 SCSI that's present inside LASI was pretty common. One
> of the uses includes the Commodore A4091, a SCSI controller sold for
> the Amiga 4000. The A4091 is emulated on WinUAE, which is open
> source. To make my story go full circle, looking at that source,
> located below, is based on QEMU source!!
>
> https://github.com/tonioni/WinUAE/blob/master/qemuvga/lsi53c710.cpp
>
> which is based on
>
> qemu/hw/scsi/lsi53c895a.c

Cool, so someone already modified the qemu lsi53c895a.c driver
to work as a 53C710 controller.
And it probably works, which is good.

> I don't know the significance/complexity of the difference between
> emulating a PCI card vs accessing things via LASI, which would sit on
> the GSC bus. Wishful thinking is that once we get there, that we've
> got existing (albeit different system emulation platform) working
> code that emulates those chip functions.

I think to get it glued up with the GSC bus isn't hard.
The LASI SCSI documentation is here:
https://parisc.wiki.kernel.org/images-parisc/7/79/Lasi_ers.pdf
page 8 of 114 (page 20 in pdf)
The SCSI controller ports are mapped at offset 0x100 of the SCSI
part of LASI.
So, if someone has enough time to add a NC53C710 code to qemu, it
should be possible.
I think I could do that, but as soon as something doesn't work
I'm not sure I would be able to debug it - I'm simply missing
the SCSI knowledge.

Helge

Re: HPPA support for IGNITE-UX install discs

[Helge Deller]

On 11/16/20 5:35 PM, rs.rdp@zoho.eu wrote:
> In qemu the B160L emulation isn't complete yet.
> The SCSI disc is emulated by a "virtual built-in" PCI SCSI card, while
> the original B160L has a NCR 53C710 SCSI chip internal in the LASI
> controller.
> Additionally the LASI emulation isn't complete either.
> My assumption is, that your Install CD for the analyzer will only detect
> real LASI/53C710 chips and ignore the virtual PCI card. That's why it
> doesn't
> detect any discs.
> _________________________________________________________
>
> Many years ago I had to install HP-UX 10.20 on a 712/9000 machine.
> At first I tried with a third-party SCSI hard disk, unsuccessfully.
> Then I replaced it with an HP hard disk and could install the os successfully.
> So I would confirm that the 10.20 installation process looks for HP physical drives and don't accept anything else.
> Any chance to deceive it?

The IGNITE-UX install disc doesn't detect any *SCSI controllers*,
so I can not say if you need real HP hard discs in this case.

> It seems it's not so for the 11.11 installation: I'm just installing it on Qemu.

Yes, there the emulated LSI 53c895 SCSI controller is supported & detected.

Helge