Created attachment 920987 [details] hotplug-log.txt Description of problem: Libguestfs has a comprehensive test suite which covers hotplugging. It uses virtio-scsi as a the disk interface to test this. Hot plug of virtio-scsi doesn't appear to work on aarch64. ------------------- First question: On aarch64 we use virtio-mmio, not virtio-pci. Is virtio-mmio even supposed to support hotplugging? My internet searches are inconclusive. If virtio-mmio just doesn't support hotplugging, then feel free to move the component to `libguestfs' so I can disable the feature. However if it's supposed to work then don't do this, because I'd like to test it. ------------------- When you try to hot add a disk, you get no error. The disk just doesn't show up after 30 seconds (which is the timeout in this test). However there are virtio-scsi errors that happen during boot which might be related. The full log of the test is attached. Version-Release number of selected component (if applicable): qemu from git @ ab6d3749c4915cd569 kernel 3.16.0-0.rc6.git1.1.efirtcfix1.fc22.aarch64 libvirt-daemon-1.2.6-1.fc21.aarch64 How reproducible: 100% Steps to Reproduce: 1. make -C tests/hotplug check Additional info: See attached log
My understanding is that this is not expected to work, virtio-mmio is not an actual bus and has no concept of hotplug. Maybe it should fail in a better way, but this is all better suited for the upstream list where the arm experts are actually watching.
Hotplugging a scsi-disk onto an existing virtio-scsi device should work, though.
This bug appears to have been reported against 'rawhide' during the Fedora 22 development cycle. Changing version to '22'. More information and reason for this action is here: https://fedoraproject.org/wiki/Fedora_Program_Management/HouseKeeping/Fedora22
I verified that f23 x86 host + f23 aarch64 guest works fine for hotplugging a virtio-scsi disk
Can confirm this *is* fixed ... which surprised me. I thought that virtio-mmio would never have working hotplug. Thanks Cole.
(In reply to Richard W.M. Jones from comment #5) > Can confirm this *is* fixed ... which surprised me. I thought > that virtio-mmio would never have working hotplug. Thanks Cole. I'm guessing virtio-mmio isn't really a blocker here, since as paolo says in comment #2 this is just hotplugging a scsi-disk to a scsi controller, but the controller happens to be virtio