Quoting "de Dinechin, Christophe (Integrity VM)"
> Another peculiar thing about HPUX, which probably has some impact, is that
> its performance is tuned and dependent on variable page sizes. So you should
> expect for example user-space mappings to contain, say, 4K, 16K, 1M and 4M
> pages for the same process. The reason it may matter for Xen is if you do
> physical address translation (as I assume you do). In that case, it's
> possible that the guest requests 4M of what it thinks is physically
> contiguous memory, but the corresponding host memory is not contiguous. In
> that case, you need to break up the large page into smaller pages.
We do exactly that: break up large pages into smaller pages. Not really good
> Stupid question from a newcomer: what does "metaphysical" mean? Funny name
Indeed. It is the mode when the guest is in physical (ie MMU disabled) mode.
Invented by Dan, from HP.
> > I think the support is complete enough
> > for VMS although we don't yet support protection keys. Does HP-UX
> > use them ?
> Yes, pretty extensively.
So we will have to add support for them...
> Well, we used to boot with IDE a long time ago, but we have dropped IDE
> support for so long that I cannot really tell if a modern HPUX can still boot
> that way.
Can you boot from CD ?
Which disk controller does HP support and which one does the HP VM emulate ?
(If you are allowed to answer!)
> > Our disk controller and VGA adapter are really basic.
> HPUX can boot headless, so no problem even if there is no VGA at all.
> > This is definitely not a difficult problem, but may prevent HP-UX
> > from going far.
> I don't know enough about how Xen emulates instructions. If there is any
> binary translation of privileged operations (I think that's not the case,
> just checking), then the generated code accesses data in guest endianness,
> and must be ready to deal with (unprivileged) endianness changes.
There are no translation, just emulation.
Xen-ia64-devel mailing list