errors from /var/log/xcp-xapi.log: 652Z|debug||1227 UNIX /var/lib/xcp/xapi|VM.start R:0b98978366e4|xapi] creating device emulator 652Z|debug||1227 UNIX /var/lib/xcp/xapi|VM.start R:0b98978366e4|xenops] Device.Dm.start domid=10 661Z|debug||1227 UNIX /var/lib/xcp/xapi|VM.start R:0b98978366e4|xenops] qemu-dm: should be running in the background (stdout redirected to syslog) 661Z|debug||1227 UNIX /var/lib/xcp/xapi|VM.start R:0b98978366e4|xenops] qemu-dm: pid = 9169. Waiting for /local/domain/0/device-model/10/state 661Z|debug||1227 UNIX /var/lib/xcp/xapi|VM.start R:0b98978366e4|xenops] watch: watching xenstore paths: [ /local/domain/0/device-model/10/state ] with timeout 5.000000 seconds 667Z|error||1227 UNIX /var/lib/xcp/xapi|VM.start R:0b98978366e4|xenops] watch: timeout while watching xenstore after 5.000000 seconds 667Z|error||1227 UNIX /var/lib/xcp/xapi|VM.start R:0b98978366e4|xenops] qemu-dm: unexpected exit with code: 1 667Z|debug||1227 UNIX /var/lib/xcp/xapi|VM.start R:0b98978366e4|backtrace] Raised at device.ml:1606.12-57 -> vmops.ml:627.2-32 -> vmops.ml:1123.8-104 667Z|error||1227 UNIX /var/lib/xcp/xapi|VM.start R:0b98978366e4|xapi] Vmops.start_paused caught: INTERNAL_ERROR: [ Device.Ioemu_failed("qemu-dm exitted unexpectedly") ] and more from /var/log/daemon.log: xcp-fe: qemu-dm-10[9169]: domid: 10 xcp-fe: qemu-dm-10[9169]: -c config qemu network with xen bridge for xcp-fe: qemu-dm-10[9169]: tap10.0 xapi1 xcp-fe: qemu-dm-10[9169]: can't add tap10.0 to bridge xapi1: Operation not supported xcp-fe: qemu-dm-10[9169]: /etc/xen/scripts/qemu-ifup: could not launch network script xcp-fe: qemu-dm-10[9169]: Could not initialize device 'tap' The message about "can't add tap10.0 to bridge xapi1: Operation not supported" seems to be the point where it fails. xapi1 does exist and two PV domUs are connected to it already: # brctl show xapi1 bridge name bridge id STP enabled interfaces xapi1 0000.de699bd4dc4d no vif1.2 vif3.1 It is only the HVM domU that has this error. Why is it talking about tap10.0 and not vif10.0 in the qemu-dm log message? Do HVM vifs use the tapX naming convention? Or is this the cause of the problem? dmesg output shows that a vif10.0 did exist briefly