Re: NetBSD on DECsystem 5000/240
"Maciej W. Rozycki" <[email protected]> Sun, 12 May 2024 01:12:03 +0100 (BST)
| Newsgroups | gmane.os.netbsd.ports.mips.devel |
|---|---|
| Message-ID | <[email protected]> |
On Tue, 20 Feb 2024, Jonathan Stone wrote: > Kernel support appears present. I've also got a PMAGD-BA and -CA, but > I've read those need more Turbochannel-slot address space than the > boot-ROM on a /240 will give you, or than the 4MB HW limit in a > 5000/200. [Found this old thread in old mailing list traffic review.] With the R3400 CPU you need to use the KSEG2 mapped address space and go via the TLB (i.e. have page tables and TLB exception handlers set up for virtual addressing) to access the extended TC slot space in the /240 for options that require more than 8MiB. That is surely (and sadly) beyond what the REX firmware spec envisaged. With the /260 and R4400SC CPU you could use the XKPHYS unmapped address space instead, however that in turn requires 64-bit addressing and is beyond what the REX environment can handle too. So oddly enough only the lesser 3MIN and MAXINE systems (and their R4k siblings) are ones in the PMAX family you can make the proper use of the PMAGD-BA and -CA framebuffers with. Improperly, they can be persuaded to work with the /240 or the /260, but that's a hack really and a proof of concept only. I have verified it can work though and extended TC slot space decoding does work as hinted in documentation. Kudos to the hw designers for the foresight and a shame that the WSE project got cancelled early, before the software side had a chance to evolve. FWIW, Maciej