Can you please try that a few more times? If I script that, I find that I get your results if I don't put a sleep before the udevadm trigger, but the right results if I do. I.e.:
cat > kvmudevtest << EOF mv /lib/udev/rules.d/40-qemu-system.rules{,.disabled} mv /lib/udev/udev-acl{,.disabled} rmmod kvm_intel modprobe kvm_intel cp /lib/udev/rules.d/40-qemu-system.rules{.disabled,} sleep 2 udevadm trigger --sysname-match=kvm --verbose ls -l /dev/kvm EOF
bash kvmudevtest
with the sleep, it works. Without the sleep, /dev/kvm does not get chowned. (obviously just a race with inotify and udev's handling thereof)
So disabling udev-acl, on my raring test system, gives just the right result.
Can you please try that a few more times? If I script that, I find that
I get your results if I don't put a sleep before the udevadm trigger,
but the right results if I do. I.e.:
cat > kvmudevtest << EOF rules.d/ 40-qemu- system. rules{, .disabled} udev-acl{ ,.disabled} rules.d/ 40-qemu- system. rules{. disabled, }
mv /lib/udev/
mv /lib/udev/
rmmod kvm_intel
modprobe kvm_intel
cp /lib/udev/
sleep 2
udevadm trigger --sysname-match=kvm --verbose
ls -l /dev/kvm
EOF
bash kvmudevtest
with the sleep, it works. Without the sleep, /dev/kvm does not get
chowned. (obviously just a race with inotify and udev's handling
thereof)
So disabling udev-acl, on my raring test system, gives just the
right result.