OBP flash failed, serial terminal, Fcode.
It's about an Ultra1e 170 I'm afraid I have killed.
I will though appreciate some post-mortrn analyse from gurus, to learn more from that experience.
I need also advice: must I swap the whole systemboard (I think it's what is needed) or only the NVRAM.
So the story:
an U1e170 happily running BSDs, Solaris, Linux.
I decided to flash the OBP to the latest 3.35.(didn't trigger the POST flashing)
The flash-file loaded and when it was in the middle of the reprogramming I realised I FORGOT TO WRITE-ENABLE the PROM jumper J2003 !!! Damn...
When the flashing printed "Programming done", the system went down to reset.And then no more graphics and no keyboard.
I opened the box to write enable J2003 and hooked a serial terminal.
The OBP prompt was alive:
Fast Data Access MMU Miss
ok .version
Release 3.35 Version 0 created 2004/04/19 12:17
OBP 3.35.0 2004/04/19 12:17
POST 3.11.4 1997/05/27 02:26
ok banner
Sun Ultra 1 UPA/SBus (UltraSPARC 167 Mhz), No keyboard
OpenBoot 3.35 384 MB memory installed, Serial# 9535409
Ethernet address 8:0:20:91:7f:b1 Host ID: 80917fb1
Diag and so did show no knowledge of the OBP current state for the FAS scsi driver, nor the HME ethernet....
The point was to reload the flash program
boot net or boot cdrom weren't possible.
The floppy driver was still alive, so I tested a dd of the flash-update (less than 1 mb) to a floppy:
ok boot floppy
Initializing memory...
Boot device: /sbus/SUNW,fdtwo
Bad magic number in disk label
Can't open disk label package
Can't open boot device
when trying with
ok boot floppy nolabel, same error
I tried then to test the floppy boot ability by using a know working boot disk,
as OpenBSD 3.6 and Linux/Debian-stable.
The openbsd boot loader died shortly:
"OpenBSD IEEE 1275 bootblock 1.1
...program terminated"
Here, I was considering if some forth programming would help to debug further the floppy problem, or if typing in a driver for the hme was possible.
The only available doc on line is the "Writing Fcode for OpenBoot 3.x".
I wasn't willing to buy the forth programming book, learn the language, get the specs for the hme and try that.
But is it a reasonable option for someone comfortable with forth?
On that topic I found almost nothing on the Sun sites, nor much relevant by an hour of web combing....(nothing to easily apply to Sun prom's).
Any advices ?
I maybe better had to wait and follow the Fcode path...
But as the serial console was alive, there was too the possibility to flash the board that way.
Here again,I found no docs.Does somebody have links ?
Issues: is the flash-upgrade file usable with a serial transfer.Which kind of serial protocol, if so: xmodem, zmodem, ymodem,...?
Which forth word to trigger in the OBP ?
Related words in the OBP dictionary: flash-upgrade, flash-erase, get-flash-filename.
Then these ones which seems (from "Writing Fcode...") related to interpreting forth ascii only: dload, dloader, dlbin...
Help about more docs appreciated.
From that point, without being sure about it, I opened a minicom session in a xterm and launched "flash-update" (which requires a filename as argument), and on another xterm i triggered "lsx flashfile" (XMODEM sending).
While the flash file was being sent a flow of "characters" was displaying on the OBP console.After few minutes, the console remained quiet.Serial transfers are slows and 900k is quite much.As all that was taking time, I decided to stop the xmodem and the console which was so quiet.
I got bak the ok prompt.Did a reset and...tada:
RED State Exception
TL=0000.0000..... TT=..... TPC=... TnPC=.... TSTATE=....
that all the console spit out now, in a continous flow.
So I believe I have really fucked the systemboard.
Or it is only the NVRAM ?
What had helped is more docs about Fcode drivers, and about serial console repair (I guess rescue was possible by doing the right thing).
Last detail: I didn't move the J2204 to switch the half-boot positions.
So any advices and pointers to documentation very much appreciated.
Thanks.
