Re: [libvirt] [PATCH v2] qemu: Introduce caching whether /dev/kvm is accessible

On Tue, Oct 30, 2018 at 03:07:59PM +0000, Daniel P. Berrangé wrote: > On Tue, Oct 30, 2018 at 03:45:36PM +0100, Michal Privoznik wrote: > > On 10/30/2018 02:46 PM, Michal Privoznik wrote: > > > > > > It is kernel problem. In my testing, the moment I call: > > > > > > ioctl(kvm, KVM_CREATE_VM, 0); > > > > Okay, I have to retract this claim. 'udevadm monitor' shows some events: > > > > KERNEL[3631.129645] change /devices/virtual/misc/kvm (misc) > > UDEV [3631.130816] change /devices/virtual/misc/kvm (misc) > > > > and stopping udevd leaves all three times untouched. So it is udev after > > all. I just don't know how to find the rule that is causing the issue. > > Anyway, as for this patch, I think we can merge it in the end, can't we? > > Not really. Udev is in use everywhere, so this behaviour makes the > patch useless in practice, even though it is technically right in > theory :-( > Does it? With this behaviour we still do the "expensive" work after any machine has started. But for one machine starting it still has the effect of running it only once. And we *need* to run it for each machine unless we also cache the result per (at least) user:group of that machine as every machine can run under different user:group.