Hi Alex,
Here is a string of patches which fix most of the many problems with
system reset on the pseriss machine. They apply on top of my other
string of pseries patches which you already merged. They apply before
Li Zhang's usb and vga patches, since it looks like those will go
another iteration, I can easily rebase after those if that would be
more convenient.
1/7 is a generic patch which I have already sent to Anthony, but it
hasn't gone into mainline yet, the rest of the series is dependent on
it, though, so it's included here. It's also dependent on newer
kernel headers than are in mainline, but which I think you already
have in your tree
This does quite a pit of rework to the pseries reset sequence, with
some influence on the ppc target at large. It fixes both general and
kvm specific bugs, although a number of the general bugs were very
difficult to actually trigger without kvm anyway (because full emu SMP
is so achingly slow, and I think has some other bugs I haven't had
time to investigate yet). There are some known reset problems still
remaining, specifically:
* We need to reset the VPA registration in KVM as well. I
have a tentative patch for that, but I'm waiting for Paul to send the
necessary KVM bits upstream.
* We should reset the TCE table on the emulated PCI host
bridge as well as VIO devices. I just haven't yet had a chance to
figure out the right place to wire in that reset yet.
