Coming rather late to this party, though have been playing with RISCiX on and off for over decade.
To add my experience:
Hard discs of 1GB+ working with RISCiX - SCSI card is AKA32 with SCSIFS/CDFS ROM (except * which are fitted in an R140 with AKA31, with "Mk.III ROM")
Quantum Fireball 1.2S aka 1280S, 1.2GB
IBM DPES-31080, 1GB
Fujitsu 1606SAU, 1GB (needs a scsidm 'probe' to kickstart before booting from RO supervisor, fine from desktop)
*Seagate ST51080N, 1GB
An IBM DCAS-32160 failed because its capacity is slightly more than 2GB.
disktab entries found include NEC 1700S, Nomai 540MB (removable cartridge).
When using scsidm to prepare the disc, the RISCOS section can be up to 512MB for Filecore, with RISCiX partitions sitting in the space beyond that. Brian Brunswick wrote a utility allowing a partition in the RISCiX section to be accessible to SCSIFS and used as an additional Filecore drive.
Somewhat confusingly, disktab (as created from the iX version of scsidm - much easier than doing it by hand, once you've made the necessary /dev/entries) reports a cylinder count covering the RISCiX section only but the sectors count for the 'h' partition represents the whole disc.
scsidm sees the whole physical disk, so it seems possible to rewrite the partition table dusplicating the 4corn Root/Boot (as an aside why that and not Root/Swap?) partition values and appending a third (fourth, ... seventh) partition to make use of spare space when the 270MB image has been dd'ed to a larger HD. Oddity: the Root partition begins at sector 123001 but when starting a new table the starting sector offered by scsidm is 277489 - I suspect scsidm sees the 1.2GB HD and calculates the minimum RISCOS section, which is larger than the actual RISCOS section (60MB, circa 22%) on the original 270MB HD.
RISCiXFS configure values
I have four RiX partitions on a single HD:
0 Root the 4corn installation
1 Swap swap space
2 nz Alan Williams' installation downloaded from Drobe
3 hopside the hopside_root installation from jfc.org.uk (which includes an 8MB swapfile)
In /dev, these correspond to sd2a, sd2b, sd2c and sd2d.
Configuring 'partition 0' boots 4corn as Root no problem; 'partition 1' fails (understandably); 'partition 2' which should boot from 'nz' yields a kernel panic and reports 'root fstype 4.3 name /dev/sd2b' (which is swap, so the panic is understandable - but then how did the kernel load to get this far?).
'partition 3' boots with root name /dev/sd2c and comes up single-user but seems to be the hopside_root version, 'partition 4' immediately "fails to find object".
Either booting from the fourth partition of the third HD is overly ambitious or I need to populate /dev with sd2a, sd2b etc nodes (and maybe use sd2S for the swap space - I can see st0S there and the swap space seems to be automagically named sd2S by the system, so perhaps using sd2b as a nodename for the swap partition has been my mistake). Time to get a 'master' /dev set up and cpio it onto other partitions. Or dedicate a different HD to each installation, so always booting from partition 0/sdNa
I will experiment further!
UPDATE: with some additional nodes in /dev, it started working as expected for single-user booting, though for proper (multi-user) booting one would have to edit the fstab references to correspond to the relevant node names.
Cloning
For cloning an existing setup, use the guide at jfc.org.uk but I would suggest mounting the 'slave' at /0 and omitting the "grep -v /mnt" part of the tar command because (1) /mnt is not empty and (2) /0 will be seen so early in the ls -A1 process that recursion is not an issue.
Could one use dump and restore as an alternative, perhaps more efficiently and sidestepping the /dev cpio point? See
http://www.futuretech.blinkenlights.nl/ ... html#CLONE for comparable process on SGI IRIX (xfs filesystem).
Ethernet
/etc/rc.net contains a conditional check on the EtherI/II driver, so better to leave "ETHERNET" blank near the top of the file and (given reports of Ether3 working) add an "ea0" section - though if Ether3 is ok "out of the box" then perhaps the en driver is sufficient for EtherII and Ether3. I have EtherI but can try EtherH later (with no real hope...).
Presumably those without econet can comment out the `econetup` command!
X session
Cannot get more colours, even with a -c8 or -c4 flag on startx, using monitortype 3 or 4. But having access to xedit in monochrome is a blessing!
Kernels
Details of the 4corn kernel are higher up this thread - thought it might be useful to list what else I have seen:
hopside_root contains two kernels
vmunix.orig 1.21b, 13/5/1992, 874,357 bytes (presumably the stock 1.21b kernel)
vmunix "test kwelton #161 special", 23/3/1993 (presumably built by Kevin Welton at Acorn)
Alan Williams' RISCiX.tar contains two kernels:
vmunix "1.21c test kernel root #1 special", dated 8/12/1993, 910,136 bytes
vmunix- 1.21c, no label, 7/9/1993, 909,763 bytes (same size as the "NSICT colonel #1 special" in the 4corn setup, but as it has no label I wonder if this is a stock 1.21c kernel)
I already had a couple of other kernels from an R260 donated to Bletchley Park, and thought it may be down to kernel differences that prevented booting with the SCSI card in slot 1 rather than 0 in my past experience. However, experiments this morning show that slot 1 seems fine, and better for keeping the SCSI card cool (I can see why the A680 reportedly had problems with is onboard SCSI overheating, and now regret not having kept a short ribbon cable). Slot 0 perhaps mandatory for RISCiX 1.xx in the R140, per notes in the old boot file).
vmunix 1.21, no label so assumed to be stock kernel, dated 30/11/1990, 838,925 bytes - won't boot with more than 4MB RAM (gets to "init: single user boot" but no prompt appears; with 4MB it works but shows what a speed difference the extra RAM makes!).
vmunix.NSICT-8Mb - 1.21c, label "colonel admin #2 special", dated 17/01/1995, 909,763 bytes