The EEH (Enhanced Error Handling) is one of RAS features on IBM Power machines. In order to support EEH, the VFIO needs some modification as the patchset addresses. Firstly, the address (domain:bus:slot:function) of passed PCI devices looks quite different from host and guest perspectives. So we have to mantain the address mapping in host so that the EEH could direct the EEH errors from guest to proper PCI device. Unfortunately, it seems that the VFIO implementation doesn't include the mechanism yet. On the other hand, it's totally business of individual platforms. So I introduced some weak functions in VFIO driver and individual platforms can override that to figure out more information that platform needs. Apart from that, the last patch [3/3] is changing the current behavior of accessing uncoverred config space for specific PCI device. The patchset is expected to be applied after Alexy's patchset (supporting VFIO on PowerNV platform). Besides, there're patchset based on it queued in my personal tree for EEH core to support PowerKVM guest. With all of them (Alexy's patchset, this patchset, EEH core patchset), I can sucessfully pass PCI device to guest and recover it from EEH errors. drivers/vfio/pci/vfio_pci.c | 42 ++++++++++++++++++++++++++++++------ drivers/vfio/pci/vfio_pci_config.c | 31 +++++++++++++++++--------- include/linux/vfio.h | 7 +++++- include/uapi/linux/vfio.h | 16 +++++++++++++ 4 files changed, 77 insertions(+), 19 deletions(-) Thanks, Gavin -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
