One problem with qemuProcessStop() is that it may run from "qemu-event" thread. This thread processes events from all guests, meaning there's just one thread that serves all running guests. But as such has no virIdentity set, which looks correct on the first glance - there's no identity to associate it with anyway. But, the problem arises when we need to connect to other daemons (e.g. virtnetworkd) to call APIs (e.g. delete a port). Here we want the thread to have the identity that of the guest. But since we haven't set any identity, opening connection to secondary drivers fails (because virIdentityGetCurrent() returns NULL which is treated by virGetConnectGeneric() as an error). As a terrible hack we can set the system identity. Resolves: https://bugzilla.redhat.com/show_bug.cgi?id=2013573 Signed-off-by: Michal Privoznik <mprivozn@xxxxxxxxxx> --- src/qemu/qemu_process.c | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/src/qemu/qemu_process.c b/src/qemu/qemu_process.c index d5f8a47ac2..7ae1974710 100644 --- a/src/qemu/qemu_process.c +++ b/src/qemu/qemu_process.c @@ -7888,6 +7888,7 @@ void qemuProcessStop(virQEMUDriver *driver, g_autofree char *timestamp = NULL; g_autoptr(virQEMUDriverConfig) cfg = virQEMUDriverGetConfig(driver); g_autoptr(virConnect) conn = NULL; + g_autoptr(virIdentity) ident = NULL; VIR_DEBUG("Shutting down vm=%p name=%s id=%d pid=%lld, " "reason=%s, asyncJob=%s, flags=0x%x", @@ -7901,6 +7902,9 @@ void qemuProcessStop(virQEMUDriver *driver, * reporting so we don't squash a legit error. */ virErrorPreserveLast(&orig_err); + ident = virIdentityGetSystem(); + virIdentitySetCurrent(ident); + if (asyncJob != QEMU_ASYNC_JOB_NONE) { if (qemuDomainObjBeginNestedJob(driver, vm, asyncJob) < 0) goto cleanup; -- 2.32.0
