]> git.ipfire.org Git - thirdparty/libvirt.git/commit
qemu: Don't set address type too early during virtio disk hotplug
authorGuido Günther <agx@sigxcpu.org>
Fri, 22 Mar 2013 09:25:42 +0000 (10:25 +0100)
committerGuido Günther <agx@sigxcpu.org>
Tue, 26 Mar 2013 17:54:41 +0000 (18:54 +0100)
commitea2e31fa5b2d4fe5c3dada3617042ecbf28e6faa
treee13359b63641c1986b0752e77a149562f4d9c77d
parent41db895f9e9efd51452568678759d6241e3ed84c
qemu: Don't set address type too early during virtio disk hotplug

f946462e14ac036357b7c11ce5c23f94a3ee4e49 changed behavior by settings
VIR_DOMAIN_DEVICE_ADDRESS_TYPE_PCI upfront. If we do so before invoking
qemuDomainPCIAddressEnsureAddr we merely try to set the PCI slot via
qemuDomainPCIAddressReserveSlot instead reserving a new address via
qemuDomainPCIAddressSetNextAddr which fails with

$ ~/run-tck-test domain/200-disk-hotplug.t
./scripts/domain/200-disk-hotplug.t .. # Creating a new transient domain
./scripts/domain/200-disk-hotplug.t .. 1/5 # Attaching the new disk /var/lib/jenkins/jobs/libvirt-tck-build/workspace/scratchdir/200-disk-hotplug/extra.img

 #   Failed test 'disk has been attached'
 #   at ./scripts/domain/200-disk-hotplug.t line 67.
 # died: Sys::Virt::Error (libvirt error code: 1, message: internal error unable to reserve PCI address 0:0:0.0
 # )
src/qemu/qemu_hotplug.c