Folks, as usual with IBM, we are left with suspicions and suggestive passages.
This morning, I fed my Brittanies around 0510, brewed up a pot of French Press coffee. Then the utterances of Saint Katherine of Rohl came to me, unbidden.
The Corvette lacks a SCSI BIOS and the SCSI firmware has a "stub". What
this stub does, or what procedure it calls, or what address it points
to, I dunno.
Since the Corvette is an RS/6000 adapter, there is no announcement
letter for it, instead, it is part of a system announcement letter, fer instance, for the 580. The POWER2 systems seem to be the first ones to support the Corvette's need for the SCSI BIOS internalized in IML code.
If we look at the "Enhanced Complex BIOS" or "Dual Boot" upgrade complex BIOS for the T1 -AND- T2, both are from Mar '92. The upgrade needs the
T1 ver 1.31 refdisk or the T2 ver 1.21 refdisk are both from Nov '92.
The RETAIN tip about the Corvette in a 8595 says there are improvements
in SCSI stuff, but no mention of the F/W.
The earliest POWER2 systems that offered the F/W are in mid-'94. But
that doesn't mean the Corvette couldn't be ordered separately from the system. RPQ or special bid...
My SWAG is the Korvetten-class were under development in '92, and IBM
most likely planned for introduction on the high-end 90/95 and on POWER2
and better RS/6000.
Another thought, offering a high-end SCSI adapter that requires an
upper-end product seems like pure IBM marketing.
On 5/20/26 20:51, Louis Ohland wrote:--- Synchronet 3.22a-Linux NewsLink 1.2
Folks, as usual with IBM, we are left with suspicions and suggestive
passages.
This morning, I fed my Brittanies around 0510, brewed up a pot of
French Press coffee. Then the utterances of Saint Katherine of Rohl
came to me, unbidden.
The Corvette lacks a SCSI BIOS and the SCSI firmware has a "stub".
What this stub does, or what procedure it calls, or what address it
points to, I dunno.
Since the Corvette is an RS/6000 adapter, there is no announcement
letter for it, instead, it is part of a system announcement letter,
fer instance, for the 580. The POWER2 systems seem to be the first
ones to support the Corvette's need for the SCSI BIOS internalized in
IML code.
Is it? I know the 4-C Corvette Turbo is but I think the standard one
was always intended for dual use.
If we look at the "Enhanced Complex BIOS" or "Dual Boot" upgrade
complex BIOS for the T1 -AND- T2, both are from Mar '92. The upgrade
needs the T1 ver 1.31 refdisk or the T2 ver 1.21 refdisk are both from
Nov '92.
The RETAIN tip about the Corvette in a 8595 says there are
improvements in SCSI stuff, but no mention of the F/W.
The earliest POWER2 systems that offered the F/W are in mid-'94. But
that doesn't mean the Corvette couldn't be ordered separately from the
system. RPQ or special bid...
https://www.ardent-tool.com/IBM_SCSI/SCSI-FW.html#RS6000_Boot_Support
Note that the way these (early) RS/6000s boot is opposite of a PC. The
ROS code enumerates the devices and selects a boot record from a default search order or a persisted nvram order. So it needs to know enough
about the device to do that. That's why it can't boot the older systems.
My SWAG is the Korvetten-class were under development in '92, and IBM
most likely planned for introduction on the high-end 90/95 and on
POWER2 and better RS/6000.
Another thought, offering a high-end SCSI adapter that requires an
upper-end product seems like pure IBM marketing.
SCSI-2 was formally ratified in 1994 although products were shipping earlier. It took a bit more horse power for everything (HBA, drive
ASICs) to run F/W and the cabling was once more complicated. Until
drives started to push the speeds, there wasn't yet a big need for a PC class system to worry about it. The 0664 HDD was contemporary, offered
in F/W, and might peak at 5MB/s. So you'd need a few simultaneously accessed for F/W to show a major improvement.
Note that the way these (early) RS/6000s boot is opposite of a PC. The--- Synchronet 3.22a-Linux NewsLink 1.2
ROS code enumerates the devices and selects a boot record from a default search order or a persisted nvram order. So it needs to know enough
about the device to do that. That's why it can't boot the older systems.
I'm not worthy.https://ps-2.kev009.com/rs6000/docs/SA23-2647-00_RS6000_Hardware_Technical_Reference_Micro_Channel_Architecture_1990.pdf
How is a PS/2 adapter's ADF converted into AIX friendly ODM entry? Does
that count as a "persistent nvram order" ? Just looking to chisel an exception from the generic rules...
Kevin Bowling wrote:
Note that the way these (early) RS/6000s boot is opposite of a PC.
The ROS code enumerates the devices and selects a boot record from a
default search order or a persisted nvram order. So it needs to know
enough about the device to do that. That's why it can't boot the
older systems.
On 5/21/26 06:50, Louis Ohland wrote:
I'm not worthy.https://ps-2.kev009.com/rs6000/docs/SA23-2647-00_RS6000_Hardware_Technical_Reference_Micro_Channel_Architecture_1990.pdf
How is a PS/2 adapter's ADF converted into AIX friendly ODM entry?
Does that count as a "persistent nvram order" ? Just looking to chisel
an exception from the generic rules...
Pg. 1-73+
https://ps-2.kev009.com/rs6000/docs/POWERstation%20Hardware%20Technical%20Information%20-%20General%20Architectures%20-%20cleaned%20OCR.pdf
Ch 2,3,4 are more illustrative but complicated. There is enough data in the VPD to bring up any card directly in basic I/O mode (but who does
it, ROS or IPL ROM?).
Once the AIX kernel is up, it rescans the bus and has its own
configuration routines. The equivalent of a PS/2 .ADF file is the per-adapter ODM "add" file (e.g. adapter.mca.8efc.add for the SCSI-2
Diff Fast/Wide card). A PdDv stanza plus PdAt/PdCn stanzas listing the same resource menu (DMA window, I/O address, IRQ, DMA level, SCSI
IDs...) that an ADF would present. These load into the system-wide Predefined ODM at install time.
Kevin Bowling wrote:
Note that the way these (early) RS/6000s boot is opposite of a PC.
The ROS code enumerates the devices and selects a boot record from a
default search order or a persisted nvram order. So it needs to know
enough about the device to do that. That's why it can't boot the
older systems.
So, does this mean that an adapter NOT in the ROS can be bootable?
Kevin Bowling wrote:
On 5/21/26 06:50, Louis Ohland wrote:
I'm not worthy.https://ps-2.kev009.com/rs6000/docs/
How is a PS/2 adapter's ADF converted into AIX friendly ODM entry?
Does that count as a "persistent nvram order" ? Just looking to
chisel an exception from the generic rules...
SA23-2647-00_RS6000_Hardware_Technical_Reference_Micro_Channel_Architecture_1990.pdf
Pg. 1-73+
https://ps-2.kev009.com/rs6000/docs/
POWERstation%20Hardware%20Technical%20Information%20-
%20General%20Architectures%20-%20cleaned%20OCR.pdf
Ch 2,3,4 are more illustrative but complicated. There is enough data
in the VPD to bring up any card directly in basic I/O mode (but who
does it, ROS or IPL ROM?).
Once the AIX kernel is up, it rescans the bus and has its own
configuration routines. The equivalent of a PS/2 .ADF file is the
per-adapter ODM "add" file (e.g. adapter.mca.8efc.add for the SCSI-2
Diff Fast/Wide card). A PdDv stanza plus PdAt/PdCn stanzas listing
the same resource menu (DMA window, I/O address, IRQ, DMA level, SCSI
IDs...) that an ADF would present. These load into the system-wide
Predefined ODM at install time.
Kevin Bowling wrote:
Note that the way these (early) RS/6000s boot is opposite of a PC.
The ROS code enumerates the devices and selects a boot record from a
default search order or a persisted nvram order. So it needs to
know enough about the device to do that. That's why it can't boot
the older systems.
Another thought, offering a high-end SCSI adapter that requires an
upper-end product seems like pure IBM marketing.
Kevin, there is no available detailed documentation on the Korvetten-
class production or functions. We have documented the Tribble, Spock,
and DFW variants. There is no simular RS/6000 SCSI adapter collection :(
I used that RS/6000 boot support list as a guide to identifying RS/6K systems that came with / supported the F/W. Tis true that's pretty
darned paltry.
It is interesting that the T1 / T2 enhanced complex BIOSes were
available in '92, and the requisite refdisks were also fielded in '92.
That does NOT prove that the F/W was available then, but suggests that
IBM had engineered the SCSI functionality to support providing the SCSI
BIOS code in IML at that time.
One RETAIN tip states that ver 1.3x / 1.2 Refdisks are version 05, this
must be a typo. SC.EXE never got above ver 3.1
http://ps-2.kev009.com/ohland/Setup/SC_EXE_Versions.html
The Type 4 SurePath BIOS must be 05+
Your thoughts on ROS and the SCSI-2 standard are spot-on, but do not
address the actual functioning of the Korvette's SCSI Firmware "stub"
and the complex IML code [aka BIOS].
Unit 5 needs more data!
Kevin Bowling wrote:
On 5/20/26 20:51, Louis Ohland wrote:
Folks, as usual with IBM, we are left with suspicions and suggestive
passages.
This morning, I fed my Brittanies around 0510, brewed up a pot of
French Press coffee. Then the utterances of Saint Katherine of Rohl
came to me, unbidden.
The Corvette lacks a SCSI BIOS and the SCSI firmware has a "stub".
What this stub does, or what procedure it calls, or what address it
points to, I dunno.
Since the Corvette is an RS/6000 adapter, there is no announcement
letter for it, instead, it is part of a system announcement letter,
fer instance, for the 580. The POWER2 systems seem to be the first
ones to support the Corvette's need for the SCSI BIOS internalized in
IML code.
Is it? I know the 4-C Corvette Turbo is but I think the standard one
was always intended for dual use.
If we look at the "Enhanced Complex BIOS" or "Dual Boot" upgrade
complex BIOS for the T1 -AND- T2, both are from Mar '92. The upgrade
needs the T1 ver 1.31 refdisk or the T2 ver 1.21 refdisk are both
from Nov '92.
The RETAIN tip about the Corvette in a 8595 says there are
improvements in SCSI stuff, but no mention of the F/W.
The earliest POWER2 systems that offered the F/W are in mid-'94. But
that doesn't mean the Corvette couldn't be ordered separately from
the system. RPQ or special bid...
https://www.ardent-tool.com/IBM_SCSI/SCSI-FW.html#RS6000_Boot_Support
Note that the way these (early) RS/6000s boot is opposite of a PC.
The ROS code enumerates the devices and selects a boot record from a
default search order or a persisted nvram order. So it needs to know
enough about the device to do that. That's why it can't boot the
older systems.
My SWAG is the Korvetten-class were under development in '92, and IBM
most likely planned for introduction on the high-end 90/95 and on
POWER2 and better RS/6000.
Another thought, offering a high-end SCSI adapter that requires an
upper-end product seems like pure IBM marketing.
SCSI-2 was formally ratified in 1994 although products were shipping
earlier. It took a bit more horse power for everything (HBA, drive
ASICs) to run F/W and the cabling was once more complicated. Until
drives started to push the speeds, there wasn't yet a big need for a
PC class system to worry about it. The 0664 HDD was contemporary,
offered in F/W, and might peak at 5MB/s. So you'd need a few
simultaneously accessed for F/W to show a major improvement.
So does it carry a BIOS hook and "install" itself on Intel systems? That--- Synchronet 3.22a-Linux NewsLink 1.2
was "common" on PC hardware later on. I am not familiar, someone with
more BIOS expertise will have to weigh in.
Kevin, in my years of fame 'n glory, I foundt that you had to know
enough about the problem to ask an intelligent question. So it sux if
one can only detect a problem, but cannot describe it with clinical precision...
Kevin Bowling wrote:
So does it carry a BIOS hook and "install" itself on Intel systems?
That was "common" on PC hardware later on. I am not familiar, someone
with more BIOS expertise will have to weigh in.
| Sysop: | DaiTengu |
|---|---|
| Location: | Appleton, WI |
| Users: | 1,118 |
| Nodes: | 10 (0 / 10) |
| Uptime: | 39:22:36 |
| Calls: | 14,340 |
| Files: | 186,357 |
| D/L today: |
23,668 files (7,691M bytes) |
| Messages: | 2,532,986 |