Re: [libvirt] [PATCH 04/10] qemuSecurityRestoreAllLabel: Don't use transactions

On Fri, Jan 20, 2017 at 10:42:44AM +0100, Michal Privoznik wrote: > Because of the nature of security driver transactions, it is > impossible to use them properly. The thing is, transactions enter > the domain namespace and commit all the seclabel changes. > However, in RestoreAllLabel() this is impossible - the qemu > process, the only process running in the namespace, is gone. And > thus is the namespace. Therefore we shouldn't use the transactions > as there is no namespace to enter. > > Signed-off-by: Michal Privoznik <mprivozn(a)redhat.com&gt; > --- > src/qemu/qemu_security.c | 25 +++++++++---------------- > 1 file changed, 9 insertions(+), 16 deletions(-) > > diff --git a/src/qemu/qemu_security.c b/src/qemu/qemu_security.c > index 544feeb4a..13d99cdbd 100644 > --- a/src/qemu/qemu_security.c > +++ b/src/qemu/qemu_security.c > @@ -73,22 +73,15 @@ qemuSecurityRestoreAllLabel(virQEMUDriverPtr driver, > virDomainObjPtr vm, > bool migrated) > { > - if (qemuDomainNamespaceEnabled(vm, QEMU_DOMAIN_NS_MOUNT) && > - virSecurityManagerTransactionStart(driver->securityManager) < 0) > - goto cleanup; > - > - if (virSecurityManagerRestoreAllLabel(driver->securityManager, > - vm->def, > - migrated) < 0) > - goto cleanup; > - > - if (qemuDomainNamespaceEnabled(vm, QEMU_DOMAIN_NS_MOUNT) && > - virSecurityManagerTransactionCommit(driver->securityManager, > - vm->pid) < 0) > - goto cleanup; > - > - cleanup: > - virSecurityManagerTransactionAbort(driver->securityManager); > + /* In contrast to qemuSecuritySetAllLabel, do not use > + * secdriver transactions here. This function is called from > + * qemuProcessStop() which is meant to do cleanup after qemu > + * process died. If it did do, the namespace is gone as qemu > + * was the only process running there. We would not succeed > + * in entering the namespace then. */ > + virSecurityManagerRestoreAllLabel(driver->securityManager, > + vm->def, > + migrated); This means we'll be running restore on /dev/BLAH in the host namespace, which is a file we didn't set a label/permissions on originally. I think this is probably safe though, as we'd just be resetting it to a label that it should already have and thus be a no-op. I'm a little more concerned about file permiissions though, as we might end up changing group from "disk" to "root" for example.

It feels like we need to explicitly skip restore for block devices if we had namespaces in use previously.