Thanks John,

it has been confirmed that

1ea37b26d720 UBUNTU: SAUCE: apparmor4.0.0 [73/76]: userns - allow
restricting unprivileged change_profile

is causing the issue. It has a sysctl to disable its behavior, but the sysctl 
can't be defaulted to off in the kernel. So to disable the sysctl, either
1. lxd needs to do it dynamically like it is doing for some other sysctls
2. we need the disable it at the system level
3. we revert the patch

For the time frame we are looking at, I recommend reverting the patch.
Doing so will not materially affect the userns mediation feature. This
patch is about closing off a confinement escape.

-- 
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/2038567

Title:
  Mantic 6.5.0-7 kernel causes regression in LXD container usage

Status in Release Notes for Ubuntu:
  New
Status in apparmor package in Ubuntu:
  New
Status in linux package in Ubuntu:
  Incomplete
Status in lxd package in Ubuntu:
  New
Status in snapd package in Ubuntu:
  New

Bug description:
  Following upgrade to 6.5.0-7 kernel in mantic cloud images we are
  seeing a regression in our cloud image tests. The test runs the
  following:

  ```
  lxd init --auto --storage-backend dir
  lxc launch ubuntu-daily:mantic mantic
  lxc info mantic
  lxc exec mantic -- cloud-init status --wait
  ```

  The `lxc exec mantic -- cloud-init status --wait` times out after 240s
  and will fail our test as a result.

  I have been able to replicate in a local VM

  ```
  wget 
http://cloud-images.ubuntu.com/mantic/20231005/mantic-server-cloudimg-amd64.img 
  wget --output-document=launch-qcow2-image-qemu.sh 
https://gist.githubusercontent.com/philroche/14c241c086a5730481e24178b654268f/raw/7af95cd4dfc8e1d0600e6118803d2c866765714e/gistfile1.txt
 
  chmod +x launch-qcow2-image-qemu.sh 

  ./launch-qcow2-image-qemu.sh --password passw0rd --image 
./mantic-server-cloudimg-amd64.img 
  cat <<EOF > "./reproducer.sh"
  #!/bin/bash -eux
  lxd init --auto --storage-backend dir
  lxc launch ubuntu-daily:mantic mantic
  lxc info mantic
  lxc exec mantic -- cloud-init status --wait
  EOF
  chmod +x ./reproducer.sh
  sshpass -p passw0rd scp -o UserKnownHostsFile=/dev/null -o CheckHostIP=no -o 
StrictHostKeyChecking=no -P 2222 ./reproducer.sh ubuntu@127.0.0.1:~/
  sshpass -p passw0rd ssh -o UserKnownHostsFile=/dev/null -o CheckHostIP=no -o 
StrictHostKeyChecking=no -p 2222 ubuntu@127.0.0.1 sudo apt-get update
  sshpass -p passw0rd ssh -o UserKnownHostsFile=/dev/null -o CheckHostIP=no -o 
StrictHostKeyChecking=no -p 2222 ubuntu@127.0.0.1 sudo apt-get upgrade 
--assume-yes
  sshpass -p passw0rd ssh -o UserKnownHostsFile=/dev/null -o CheckHostIP=no -o 
StrictHostKeyChecking=no -p 2222 ubuntu@127.0.0.1 ./reproducer.sh
  ```

  The issue is not present with the 6.5.0-5 kernel and the issue is
  present regardless of the container launched. I tried the jammy
  container to test this.

  From my test VM

  ```
  ubuntu@cloudimg:~$ uname --all
  Linux cloudimg 6.5.0-7-generic #7-Ubuntu SMP PREEMPT_DYNAMIC Fri Sep 29 
09:14:56 UTC 2023 x86_64 x86_64 x86_64 GNU/Linux
  ubuntu@cloudimg:~$ uname --kernel-release
  6.5.0-7-generic
  ```

  This is a regression in our test that will block 23.10 cloud image
  release next week.

To manage notifications about this bug go to:
https://bugs.launchpad.net/ubuntu-release-notes/+bug/2038567/+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

Reply via email to