ruri v3.9.5 Release

It's another huge release since v3.9.3. So v3.9.5 is now ready, all caught up :>

• Add --freeze/--thaw subcommands to pause and resume a container via cgroup freezer (NOTE: needs cgroup freezer support, both cgroup v2 built-in freezer and v1 /sys/fs/cgroup/freezer or Android /dev/freezer are detected).
• Add --set-flag for developers to enable feature flags.
• Add a .ruri_umounted file to indicate that the container has been unmounted.
• Fix re-enter unshare container failed in v3.9.4.
• Fix pipe-sync issue in unshare container.
• Fix losetup failed when udev didn't create /dev/loopx device before ruri tries to mount it.
• Fix losetup failed on Android due to SELinux issue.
• Always convert mount source to absolute path.
• Add loop partition and btrfs mount support.
• Support tmpfs as rootfs.
• Support BIND: prefix to force bind-mounting a regular file instead of trying to mount it as an image.
• Fix subarch for seccomp profile.
• Support new feature flags:
  • ban_futex_pi: Ban futex_pi syscalls, for GhostLock mitigation.
  • wait_before_exec: Wait for SIGUSR1 signal before exec() in the container.
  • allow_personality: Allow personality() syscall, for compatibility with some software like debian reprotest, box86/wine, etc.
  • force_panic: The internal implementation of --strict-mode, will force ruri to panic on any error.
  • no_time_ns: disable time namespace.
  • no_uts_ns: disable UTS namespace.
  • no_ipc_ns: disable IPC namespace.
  • no_pid_ns: disable PID namespace.
  • no_cgroup_ns: disable cgroup namespace.
  • fork_as_init: The internal implementation of --fork-as-init, will make ruri fork() before exec() to be the init process in the container.
  • disable_warnings: The internal implementation of --no-warnings, will disable all warnings.
  • auto_umount: The internal implementation of --auto-umount, will automatically umount the container when it exits.
  • auto_umount_on_panic: The internal implementation of --umount-on-panic, will automatically umount the container when it panics.
  • is_health_check: The internal implementation of --health-check, will run as health check process in the container.
  • systemd_init: The internal implementation of --systemd, will enable systemd init support in the container.
  • enable_tty_signals: The internal implementation of --enable-tty-signals, will not mask SIGTTIN and SIGTTOU signals in the container.
  • skip_setgroups: The internal implementation of --skip-setgroups, will skip setgroups() call when changing the user in the container.
  • make_kvm_node: Will be converted to dev_nodes=+kvm.
  • empty_net_ns: The internal implementation of --no-network, will disable network in the container.
  • create_gunyah_node: Will be converted to dev_nodes=+gunyah.
  • create_geniezone_node: Will be converted to dev_nodes=+gzvm.
  • no_reset_pidfile: will keep writing to the pidfile without cleaning it. For debugging.
  • no_logs: ruri will auto convert ruri_no_logs env to this flag, and will disable all logs. For debugging.
  • wait_pidfile_lock: As pidfile is updated asynchronously, this flag will make sure the pidfile is updated before exiting.
  • no_seccomp: Disable all seccomp-based features.
  • no_rurienv: The internal implementation of --no-rurienv.
  • no_cgroup: Disable all cgroup-based features.
  • no_pidfile_daemon: Disable the pidfile daemon, --auto-umount and --umount-on-panic will also be disabled.
  • no_drop_caps: Do not really call cap_drop_bound(), only for debugging.
  • no_memory_cgroup: Disable all memory cgroup based features.
  • no_cpuset_cgroup: Disable all cpuset cgroup based features.
  • no_cpupercent_cgroup: Disable all cpupercent cgroup based features.
  • no_pids_cgroup: Disable all pids cgroup based features.
  • no_io_cgroup: Disable all io cgroup based features.
  • no_freezer_cgroup: Disable all freezer cgroup based features.
  • no_pidfd: Disable all pidfd based features, for debugging.
  • dev_nodes: A comma-separated list to override default device nodes in the container. For example, dev_nodes=+kvm,-full means create /dev/kvm but disable /dev/full in the container.
  • just_chroot: The internal implementation of --just-chroot, will just chroot into the container without creating runtime directories.
  • ruri_dbg: Enable ruri debug mode, will print logs and do other debug stuff. For debugging.
  • use_host_runtime: The internal implementation of --host-runtime, will bind-mount /dev/, /sys/, and /proc/ from host.
  • no_mask_paths: The internal implementation of --unmask-dirs, will not mask sensitive paths in /proc and /sys.
  • read_only_rootfs: The internal implementation of --read-only, will mount / as read-only.
  • no_new_privs: The internal implementation of --no-new-privs, will set NO_NEW_PRIVS flag.
  • rlimits: A comma-separated list to set rlimits in the container. For example, rlimits=nproc:16:32,core:1 means set RLIMIT_NPROC to 16 (soft) and 32 (hard), and set RLIMIT_CORE to 1 (soft) and 1 (hard).
  • outside_rurienv: Use outside .rurienv file instead of the one in the container. For example, outside_rurienv=/tmp/rurienv means use /tmp/rurienv instead of /.rurienv in the container.
  • rw_rurienv: make .rurienv rw, will not set immutable flag and ro bind-mount on it.
  • ruri_perf: enable profiling log, only for debugging.
  • is_termux: if we are running in termux.
  • img_sectx: SELinux context for image file, to fix loop-mount on android.
  • new_tty: create a new pty in container.
  • create_ntsync_node: Will be converted to dev_nodes=+ntsync.
  • ban_setuid: A comma-separated list to disallow setuid() to specified uid, for example ban_setuid=5,2,1 will diasllow setuid to uid 5 or 2 or 1.
  • no_subarch: Do not add subarch to seccomp profile.
  • ban_sctp: Ban IPPROTO_SCTP for socket().