Re: New pmap module for m68k

[Jason Thorpe <thorpej%me.com@localhost>]

> On Nov 9, 2025, at 11:10 PM, Romain Dolbeau <romain%dolbeau.org@localhost> wrote:
> 
> s it possible to have more than one 32-bits space in the qemu model?
> That would be a way to create something equivalent to a 34-bits area.
> I am wondering if it would be possible to have UPA[1:0] (stored in
> bits [9:8] of descriptors, and the treated same as address bits all
> the way to the pins) used to discriminate between 4 32-bits area, that
> would give the virtual model e.g. 4 GiB of I/O space and up to 12 GiB
> of memory. The process virtual space would still be limited to <4 GiB,
> but several of them could be resident simultaneously. Completely
> useless on a real '040 (too much data, not enough speed), but might be
> handy on a build host.

I keep forgetting about the UPA bits (partially because the Hibler pmap uses one of them as a software-defined bit, which we thankfully got away with for all these years).

Yes, it would be possible to do that with UPA, but it would need special support in the pmap module (and any other place that has reason to consult a PTE value).

That said… I tried booting Qemu with 3.5GB and it errored out… Qemu itself does impose a limit, just above 3GB.

-- thorpej