Re: [libvirt] [PATCH 2/9] qemu: process: Don't try to redetect missing qemuCaps on reconnect

On Mon, Nov 25, 2019 at 02:18:54PM +0100, Peter Krempa wrote: > On Mon, Nov 25, 2019 at 13:11:10 +0000, Daniel Berrange wrote: > > On Mon, Nov 25, 2019 at 02:07:13PM +0100, Peter Krempa wrote: > > > On Mon, Nov 25, 2019 at 12:59:00 +0000, Daniel Berrange wrote: > > > > On Mon, Nov 25, 2019 at 01:53:47PM +0100, Peter Krempa wrote: > > > > > The redetection was originally added in 43c01d3838 as a way to recover > > > > > from libvirtd upgrade from the time when we didn't persist the qemu > > > > > capabilities in the status XML. Also this the oldest supported qemu by > > > > > more than two years. > > > > > > > > > > Even if somebody would have a running VM running at least qemu 1.5 with > > > > > such an old libvirt we certainly wouldn't do the right thing by > > > > > redetecting the capabilities and then trying to communicate with qemu. > > > > > > > > > > For now it will be the best to just stop considering this scenario any > > > > > more. > > > > > > > > > > Signed-off-by: Peter Krempa <pkrempa(a)redhat.com&gt; > > > > > --- > > > > > src/qemu/qemu_process.c | 7 ------- > > > > > 1 file changed, 7 deletions(-) > > > > > > > > > > diff --git a/src/qemu/qemu_process.c b/src/qemu/qemu_process.c > > > > > index 1b88c471f4..a76a8da841 100644 > > > > > --- a/src/qemu/qemu_process.c > > > > > +++ b/src/qemu/qemu_process.c > > > > > @@ -8085,13 +8085,6 @@ qemuProcessReconnect(void *opaque) > > > > > goto error; > > > > > } > > > > > > > > > > - /* If upgrading from old libvirtd we won't have found any > > > > > - * caps in the domain status, so re-query them > > > > > - */ > > > > > - if (!priv->qemuCaps && > > > > > - (qemuDomainUpdateQEMUCaps(obj, driver->qemuCapsCache) < 0)) > > > > > - goto error; > > > > > > > > Shouldn't we be making missing qemuCaps into a fatal error > > > > when loading that VM. > > > > > > I thought about it for a bit. If we make it an error the VM will vanish > > > after upgrade (and continue running). If we don't make an error it will > > > not vanish and you'll be able to at least destroy it. > > > > > > Given that such situation is super-unlikely, I opted to remove it. > > > Nobody will probably run into this situation and if yes then the VM will > > > not vanish at least. > > > > > > I can add an error though, but to me it seems it's not worth. > > > > My concern is that pretty much all our later code will assume that > > priv->qemuCaps is non-NULL. So by ignoring the error, the VM might > > not vanish initially, but libvirtd may well crash shortly after. > > Good point. > > The caps are parsed in qemuDomainObjPrivateXMLParse > > How about we either change the condition to allocate priv->qemuCaps even > if 0 elements are found in the XML so that priv->qemuCaps is not NULL > (but empty). This is fine, especially if we mark the VM tainted when we reconnect.