At the moment the ITS is not properly reset. On System reset or reboot, previous ITS register values and caches are left unchanged. Some of the registers might point to some guest RAM tables which are not provisionned. This leads to state inconsistencies that are detected by the kernel save/restore code. And eventually this may cause qemu abort on source or destination.
The 1st patch, suggested to be cc'ed stable proposes to remove the abort in case of table save/restore failure. This is definitively not ideal but looks the most reasonable until we get a proper way to reset the ITS. Still a message is emitted to report the save/restore did not happen correctly. Subsequent patches add the support of explicit reset using a new kvm device group/attribute combo. The associated kernel series is not upstream [1], hence the RFC. ITS specification is not very clear about reset. There is no reset wire. Some register fields are documented to have architecturally defined reset values and we use those here: Most importantly the Valid bit of GITS_CBASER and GITS_BASER are cleared and the GITS_CTLR.Enabled bit is cleared as well. Best Regards Eric Host Kernel dependencies: - [1] [PATCH 0/10 v2] vITS Migration fixes and reset The series is available at: https://github.com/eauger/qemu/tree/v2.10-its-reset-v1 Eric Auger (3): hw/intc/arm_gicv3_its: Don't abort on table save/restore linux-headers: Partial header update for ITS reset hw/intc/arm_gicv3_its: Implement reset hw/intc/arm_gicv3_its_common.c | 5 ++--- hw/intc/arm_gicv3_its_kvm.c | 37 ++++++++++++++++++++++++---------- include/hw/intc/arm_gicv3_its_common.h | 1 + linux-headers/asm-arm/kvm.h | 1 + linux-headers/asm-arm64/kvm.h | 1 + 5 files changed, 31 insertions(+), 14 deletions(-) -- 2.5.5