QSIC update and request
jnc at mercury.lcs.mit.edu
Tue Jan 2 08:01:53 CST 2018
> From: Mark J. Blair
> I wonder if it might also be useful in any of the QBUS MicroVAXen?
Hardwarewise, it should be fine. Softwarewise... well...
The issue is that we're currently only planning to emulate the RK11 and RP11,
because we're not up for the hassle involved in emulating more recent
controllers. (That's not an issue for the systems we want to run.)
We looked at the RP04, and _full_ emulation even of that one is a significant
amount of work. (I stress the 'full' because a partial, simple emulation
might not be too bad, but since we have no idea what various OS's will expect
to be there, it's not clear how much use such a partial emulation would be.)
However, that presents a problem. There are no 22-bit versions of either (in
fact, there's no QBUS RP11 at all; and the QBUS RK11 is restricted to Q16,
for reasons that surpass me). 22-bit operation is needed to make them really
useable as mass storage under Unix, for swapping (because all file system
access is buffered through low memory, purely file system use would be OK
without Q22 support) - at least on -11's with more than 256KB of memory.
(Probably DEC OS's too, but I know nothing of them.)
So, in addition to the dead-stock emulations of the two, we will also support
slightly 'adjusted' versions of the controllers, to have an 'address
extension' register (in exactly the same way the RLV12 has an extra register
over the RLV11).
(And we're also going to have an adjusted version of the RP11, which extends
the size of the disk, using unused bits/values in the disk address registers,
to allow up to 1TB on an 'RP11-D', as we're calling it. Hey, if you're going
to change the controller _at all_... But I digress.)
Anyway, you can see where this is going. For people who can tweak their
drivers, no biggie. The changes aren't major - a line or two. For people who
want to run stock software...
I don't know enough about how the QBUS VAX systems use their disks. Will uVAX
OS's run with only an RKV11-D for mass storage? Somehow I doubt it..
I assume on the later ones (the ones with the private memory bus so they can
have more than 4MB of memory) there's some sort of QBUS map, to map from the
QBUS' 22-bit address space, to the full memory. But does the hardware and
software expect to use the entire 22-bit address space, or are they prepared
to limit it (e.g. for working with an RKV11-D), or what?
I suppose we could add the RLV12 to the list of things we emulate; that's not
_that_ complicated a controller. The problem is that RL's aren't that big
(10MB), and that gets to be an issue with later OS's. And even then VAX OS's
might not run off an RLV12 - I just don't know.
Getting around this is all, of course, a SMOP (Small Matter of Programming),
since a new FPGA load, with support for more emulations, can be installed on
an existing QSIC at any time.
Now, whether Dave would be interested in supporting later devices, or whether
someone else could be convinced to emulate something more modern ... who knows?
More information about the cctalk