** Description changed: + [Impact] + + * Libvirt will no more be able to start guests with newer kernels + (>=4.18) + + * We brought a fix upstream that we want to backport to potentially + affected releases (B+C) + + [Test Case] + + Note: This can but does not have to be tested in nested virtualization (Since we need to change the kernel it is probably the preferred way to do it in nested, so this is what I describe) + 1. Get a KVM guest with the Release to test (Cosmic as an example) and give it a size to be able to host a nested guest + $ sudo apt install uvtool-libvirt + $ uvt-simplestreams-libvirt --verbose sync --source http://cloud-images.ubuntu.com/daily arch=amd64 release=cosmic label=daily + $ uvt-kvm create --cpu 8 --memory 4096 --password ubuntu c-nested-kvm-4.18 arch=amd64 release=cosmic label=daily + $ uvt-kvm ssh c-nested-kvm-4.18 + 2. In the guest set up another guest + $ uvt-kvm ssh c-nested-kvm-4.18 + $ uvt-simplestreams-libvirt --verbose sync --source http://cloud-images.ubuntu.com/daily arch=amd64 release=cosmic label=daily + $ uvt-kvm create --password ubuntu c-nested-guest arch=amd64 release=cosmic label=daily + $ sudo apt install uvtool-libvirt + 3. The above will work, now get a kernel >=4.18 e.g. the mainline kernels at + https://wiki.ubuntu.com/Kernel/MainlineBuilds + $ wget (link of the day) + $ sudo dpkg -i *.deb + 4. reboot into the 4.18 kernel + $ sudo reboot + 5. Start the guest again + $ virsh start c-nested-guest + + Without fix this will fail like: + error: Failed to start domain c-nested-guest + error: internal error: child reported: Kernel does not provide mount namespace: Permission denied + + And an associated Dmesg/Apparmor entry: + [ 519.529760] audit: type=1400 audit(1535364820.976:27): apparmor="DENIED" operation="ptrace" profile="/usr/sbin/libvirtd" pid=1692 comm="libvirtd" requested_mask="read" denied_mask="read" peer="libvirt-4ae5dbbe-641c-4934-ae7c-956bcc8b9e2f" + + 6. In the guest update from PPA/Proposed that you want to test to bring in the libvirt package with the updated rules (Example with the PPA). + $ sudo add-apt-repository ppa:ci-train-ppa-service/3381 + $ sudo apt upgrade + + 7. Start the guest again and it will now work + $ virsh start c-nested-guest + + [Regression Potential] + + * The change is "only" opening up the apparmor profile for libvirt a + little bit for the access now being reported as read in addition. + That should in itself be rather regression free, but in addition it has + an ack of Jamie Strandboge to be sure from Securities POV (you can + follow the upstreaming thread at to check: + https://www.redhat.com/archives/libvir-list/2018-August/msg01597.html + + + [Other Info] + + * This is no issue "yet" but would be one when the Cosmic Kernel becomes available as HWE kernel. We'd want to have the apparmor profile in libvirt ready at the time so that we never have users see this error in the Field on a Released Ubuntu version. + + --- + $ virsh start cosmic-i386 error: Failed to start domain cosmic-i386 error: internal error: child reported: Kernel does not provide mount namespace: Permission denied - Happens with 4.18 only, 4.17 is fine. 27013 openat(AT_FDCWD, "/proc/27012/ns/mnt", O_RDONLY) = -1 EACCES (Permission denied) Opening a self or parent mount namespace works fine as tested with cat /proc/self/ns/mnt and cat /proc/self/$$/mnt. Still investigating.
-- You received this bug notification because you are a member of Kernel Packages, which is subscribed to linux in Ubuntu. https://bugs.launchpad.net/bugs/1788603 Title: libvirt fails with failure to open mount namespace Status in libvirt package in Ubuntu: In Progress Status in linux package in Ubuntu: In Progress Status in libvirt source package in Bionic: Confirmed Status in linux source package in Bionic: New Status in libvirt source package in Cosmic: In Progress Status in linux source package in Cosmic: In Progress Bug description: [Impact] * Libvirt will no more be able to start guests with newer kernels (>=4.18) * We brought a fix upstream that we want to backport to potentially affected releases (B+C) [Test Case] Note: This can but does not have to be tested in nested virtualization (Since we need to change the kernel it is probably the preferred way to do it in nested, so this is what I describe) 1. Get a KVM guest with the Release to test (Cosmic as an example) and give it a size to be able to host a nested guest $ sudo apt install uvtool-libvirt $ uvt-simplestreams-libvirt --verbose sync --source http://cloud-images.ubuntu.com/daily arch=amd64 release=cosmic label=daily $ uvt-kvm create --cpu 8 --memory 4096 --password ubuntu c-nested-kvm-4.18 arch=amd64 release=cosmic label=daily $ uvt-kvm ssh c-nested-kvm-4.18 2. In the guest set up another guest $ uvt-kvm ssh c-nested-kvm-4.18 $ uvt-simplestreams-libvirt --verbose sync --source http://cloud-images.ubuntu.com/daily arch=amd64 release=cosmic label=daily $ uvt-kvm create --password ubuntu c-nested-guest arch=amd64 release=cosmic label=daily $ sudo apt install uvtool-libvirt 3. The above will work, now get a kernel >=4.18 e.g. the mainline kernels at https://wiki.ubuntu.com/Kernel/MainlineBuilds $ wget (link of the day) $ sudo dpkg -i *.deb 4. reboot into the 4.18 kernel $ sudo reboot 5. Start the guest again $ virsh start c-nested-guest Without fix this will fail like: error: Failed to start domain c-nested-guest error: internal error: child reported: Kernel does not provide mount namespace: Permission denied And an associated Dmesg/Apparmor entry: [ 519.529760] audit: type=1400 audit(1535364820.976:27): apparmor="DENIED" operation="ptrace" profile="/usr/sbin/libvirtd" pid=1692 comm="libvirtd" requested_mask="read" denied_mask="read" peer="libvirt-4ae5dbbe-641c-4934-ae7c-956bcc8b9e2f" 6. In the guest update from PPA/Proposed that you want to test to bring in the libvirt package with the updated rules (Example with the PPA). $ sudo add-apt-repository ppa:ci-train-ppa-service/3381 $ sudo apt upgrade 7. Start the guest again and it will now work $ virsh start c-nested-guest [Regression Potential] * The change is "only" opening up the apparmor profile for libvirt a little bit for the access now being reported as read in addition. That should in itself be rather regression free, but in addition it has an ack of Jamie Strandboge to be sure from Securities POV (you can follow the upstreaming thread at to check: https://www.redhat.com/archives/libvir-list/2018-August/msg01597.html [Other Info] * This is no issue "yet" but would be one when the Cosmic Kernel becomes available as HWE kernel. We'd want to have the apparmor profile in libvirt ready at the time so that we never have users see this error in the Field on a Released Ubuntu version. --- $ virsh start cosmic-i386 error: Failed to start domain cosmic-i386 error: internal error: child reported: Kernel does not provide mount namespace: Permission denied Happens with 4.18 only, 4.17 is fine. 27013 openat(AT_FDCWD, "/proc/27012/ns/mnt", O_RDONLY) = -1 EACCES (Permission denied) Opening a self or parent mount namespace works fine as tested with cat /proc/self/ns/mnt and cat /proc/self/$$/mnt. Still investigating. To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/libvirt/+bug/1788603/+subscriptions -- Mailing list: https://launchpad.net/~kernel-packages Post to : kernel-packages@lists.launchpad.net Unsubscribe : https://launchpad.net/~kernel-packages More help : https://help.launchpad.net/ListHelp
