On Mon, Nov 21, 2016 at 03:58:31PM +1100, David Gibson wrote: > Unfortunately, a bug already slipped into the qemu-2.7 release which > breaks migration of the pseries-2.6 machine type from qemu v2.6.0 to > qemu v2.7.0. That bug also breaks migration from v2.6.0 to current > master. In addition there have been several more changes which have > inadvertently broken migration from v2.7.0 to current master for both > pseries-2.6 and pseries-2.7 machine types. > > One of those was fixed - in a way - by 9b54ca0, however it was done in > such a way as to break backwards migration from current master to > v2.7.0. It turns out that supporting backwards migration is more > important than I realized at the time. > > So, this series contains several fixes for migration. Between them, > they allow migration both forwards and backwards between qemu v2.6.0 > and current master (and therefore, I hope, v2.8.0). > > Unfortunately, because of the already released bug, there's no > reasonable way to fix migration from current master back to v2.7.0 as > well, and, obviously, no way at all to fix the existing breakage > between v2.6.0 and v2.7.0. However, I'm also hoping to get patch 1/5 > of this series applied to the 2.7 stable branch. If that happens, it > should be possible to freely migrate between v2.6.0, v2.7.1 and > current master / 2.8, which is about the best we can hope for. > > I've done basic testing of these with all the combinations of source > and destination being either v2.6.0, v2.7.0 + patch 1/5 and > ppc-for-2.8 + this series, and with all the applicable machine types > for each of those combinations. Each test was with just a basic > default hardware setup, running a RHEL7.3 guest, no fancier > combinations. > > Dave, could you please sanity check my compatibility sheanningans and > send an R-b (and an acked-by for patch 2/5, which isn't strictly ppc > related). > > Michael, if you could test this on any combinations that occur to you, > that would be most helpful. > > I'm hoping to get some reviews and squeeze this into v2.8.0, which > will mean a pull request within a couple of days.
Bother, just realised I left a chunk of debug code in there. Will
resend momentarily.
>
> David Gibson (5): target-ppc: Fix CPU migration from qemu-2.6 <->
> later versions migration: Add VMSTATE_UINTTL_TEST() target-ppc:
> Allow eventual removal of old migration mistakes Revert "spapr: Fix
> migration of PCI host bridges from qemu-2.7" spapr: Fix 2.7<->2.8
> migration of PCI host bridge
>
> hw/ppc/spapr.c | 10 +++++++++
> hw/ppc/spapr_pci.c | 35 +++++++++++++++++++++--------
> include/hw/pci-host/spapr.h | 6 +++++
> include/migration/cpu.h | 4 ++++
> target-ppc/cpu.h | 7 ++++++
> target-ppc/machine.c | 55
> +++++++++++++++++++++++++++++++++++++++++----
> target-ppc/translate_init.c | 6 +++++
> 7 files changed, 110 insertions(+), 13 deletions(-)
>
--
David Gibson | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_
| _way_ _around_!
http://www.ozlabs.org/~dgibson
signature.asc
Description: PGP signature
