github.com/opencontainers/runc

There's a frood who really knows where his towel is.

runc now requires a minimum of Go 1.20 to compile.

NOTE: runc currently will not work properly when compiled with Go 1.22 or newer. This is due to some unfortunate glibc behaviour that Go 1.22 exacerbates in a way that results in containers not being able to start on some systems. See this issue for more information.

Breaking

• Several aspects of how mount options work has been adjusted in a way that could theoretically break users that have very strange mount option strings. This was necessary to fix glaring issues in how mount options were being treated. The key changes are:

  • Mount options on bind-mounts that clear a mount flag are now always applied. Previously, if a user requested a bind-mount with only clearing options (such as rw,exec,dev) the options would be ignored and the original bind-mount options would be set. Unfortunately this also means that container configurations which specified only clearing mount options will now actually get what they asked for, which could break existing containers (though it seems unlikely that a user who requested a specific mount option would consider it "broken" to get the mount options they asked foruser who requested a specific mount option would consider it "broken" to get the mount options they asked for). This also allows us to silently add locked mount flags the user did not explicitly request to be cleared in rootless mode, allowing for easier use of bind-mounts for rootless containers. (#3967)

  • Container configurations using bind-mounts with superblock mount flags (i.e. filesystem-specific mount flags, referred to as "data" in mount(2), as opposed to VFS generic mount flags like MS_NODEV) will now return an error. This is because superblock mount flags will also affect the host mount (as the superblock is shared when bind-mounting), which is obviously not acceptable. Previously, these flags were silently ignored so this change simply tells users that runc cannot fulfil their request rather than just ignoring it. (#3990)

  If any of these changes cause problems in real-world workloads, please open an issue so we can adjust the behaviour to avoid compatibility issues.

Added

• runc has been updated to OCI runtime-spec 1.2.0, and supports all Linux features with a few minor exceptions. See docs/spec-conformance.md for more details.
• runc now supports id-mapped mounts for bind-mounts (with no restrictions on the mapping used for each mount). Other mount types are not currently supported. This feature requires MOUNT_ATTR_IDMAP kernel support (Linux 5.12 or newer) as well as kernel support for the underlying filesystem used for the bind-mount. See mount_setattr(2) for a list of supported filesystems and other restrictions. (#3717, #3985, #3993)
• Two new mechanisms for reducing the memory usage of our protections against CVE-2019-5736 have been introduced:
  • runc-dmz is a minimal binary (~8K) which acts as an additional execve stage, allowing us to only need to protect the smaller binary. It should be noted that there have been several compatibility issues reported with the usage of runc-dmz (namely related to capabilities and SELinux). As such, this mechanism is opt-in and can be enabled by running runc with the environment variable RUNC_DMZ=true (setting this environment variable in config.json will have no effect). This feature can be disabled at build time using the runc_nodmz build tag. (#3983, #3987)
  • contrib/memfd-bind is a helper daemon which will bind-mount a memfd copy of /usr/bin/runc on top of /usr/bin/runc. This entirely eliminates per-container copies of the binary, but requires care to ensure that upgrades to runc are handled properly, and requires a long-running daemon (unfortunately memfds cannot be bind-mounted directly and thus require a daemon to keep them alive). (#3987)
• runc will now use cgroup.kill if available to kill all processes in a container (such as when doing runc kill). (#3135, #3825)
• Add support for setting the umask for runc exec. (#3661)
• libct/cg: support SCHED_IDLE for runc cgroupfs. (#3377)
• checkpoint/restore: implement --manage-cgroups-mode=ignore. (#3546)
• seccomp: refactor flags support; add flags to features, set SPEC_ALLOW by default. (#3588)
• libct/cg/sd: use systemd v240+ new MAJOR:* syntax. (#3843)
• Support CFS bandwidth burst for CPU. (#3749, #3145)
• Support time namespaces. (#3876)
• Reduce the runc binary size by ~11% by updating github.com/checkpoint-restore/go-criu. (#3652)
• Add --pidfd-socket to runc run and runc exec to allow for management processes to receive a pidfd for the new process, allowing them to avoid pid reuse attacks. (#4045)

Deprecated

• runc option --criu is now ignored (with a warning), and the option will be removed entirely in a future release. Users who need a non-standard criu binary should rely on the standard way of looking up binaries in $PATH. (#3316)
• runc kill option -a is now deprecated. Previously, it had to be specified to kill a container (with SIGKILL) which does not have its own private PID namespace (so that runc would send SIGKILL to all processes). Now, this is done automatically. (#3864, #3825)
• github.com/opencontainers/runc/libcontainer/user is now deprecated, please use github.com/moby/sys/user instead. It will be removed in a future release. (#4017)

Changed

• When Intel RDT feature is not available, its initialization is skipped, resulting in slightly faster runc exec and runc run. (#3306)
• runc features is no longer experimental. (#3861)
• libcontainer users that create and kill containers from a daemon process (so that the container init is a child of that process) must now implement a proper child reaper in case a container does not have its own private PID namespace, as documented in container.Signal. (#3825)
• Sum anon and file from memory.stat for cgroupv2 root usage, as the root does not have memory.current for cgroupv2. This aligns cgroupv2 root usage more closely with cgroupv1 reporting. Additionally, report root swap usage as sum of swap and memory usage, aligned with v1 and existing non-root v2 reporting. (#3933)
• Add swapOnlyUsage in MemoryStats. This field reports swap-only usage. For cgroupv1, Usage and Failcnt are set by subtracting memory usage from memory+swap usage. For cgroupv2, Usage, Limit, and MaxUsage are set. (#4010)
• libcontainer users that create and kill containers from a daemon process (so that the container init is a child of that process) must now implement a proper child reaper in case a container does not have its own private PID namespace, as documented in container.Signal. (#3825)
• libcontainer: container.Signal no longer takes an all argument. Whether or not it is necessary to kill all processes in the container individually is now determined automatically. (#3825, #3885)
• seccomp: enable seccomp binary tree optimization. (#3405)
• runc run/runc exec: ignore SIGURG. (#3368)
• Remove tun/tap from the default device allowlist. (#3468)
• runc --root non-existent-dir list now reports an error for non-existent root directory. (#3374)

Fixed

• In case the runc binary resides on tmpfs, runc init no longer re-execs itself twice. (#3342)
• Our seccomp -ENOSYS stub now correctly handles multiplexed syscalls on s390 and s390x. This solves the issue where syscalls the host kernel did not support would return -EPERM despite the existence of the -ENOSYS stub code (this was due to how s390x does syscall multiplexing). (#3474)
• Remove tun/tap from the default device rules. (#3468)
• specconv: avoid mapping "acl" to MS_POSIXACL. (#3739)
• libcontainer: fix private PID namespace detection when killing the container. (#3866, #3825)
• systemd socket notification: fix race where runc exited before systemd properly handled the READY notification. (#3291, #3293)
• The -ENOSYS seccomp stub is now always generated for the native architecture that runc is running on. This is needed to work around some arguably specification-incompliant behaviour from Docker on architectures such as ppc64le, where the allowed architecture list is set to null. This ensures that we always generate at least one -ENOSYS stub for the native architecture even with these weird configs. (#4219)

Removed

• In order to fix performance issues in the "lightweight" bindfd protection against CVE-2019-5736, the temporary ro bind-mount of /proc/self/exe has been removed. runc now creates a binary copy in all cases. See the above notes about memfd-bind and runc-dmz as well as contrib/cmd/memfd-bind/README.md for more information about how this (minor) change in memory usage can be further reduced. (#3987, #3599, #2532, #3931)
• libct/cg: Remove EnterPid (a function with no users). (#3797)
• libcontainer: Remove {Pre,Post}MountCmds which were never used and are obsoleted by more generic container hooks. (#3350)

How, dear sir, did you cross the flood? By not stopping, friend, and by not straining I crossed the flood.

Fixed

• The -ENOSYS seccomp stub is now always generated for the native architecture that runc is running on. This is needed to work around some arguably specification-incompliant behaviour from Docker on architectures such as ppc64le, where the allowed architecture list is set to null. This ensures that we always generate at least one -ENOSYS stub for the native architecture even with these weird configs. (#4391)
• On a system with older kernel, reading /proc/self/mountinfo may skip some entries, as a consequence runc may not properly set mount propagation, causing container mounts leak onto the host mount namespace. (#2404, #4425)

Removed

• In order to fix performance issues in the "lightweight" bindfd protection against CVE-2019-5736, the temporary ro bind-mount of /proc/self/exe has been removed. runc now creates a binary copy in all cases. (#4392, #2532)

年を取っていいことは、驚かなくなることね。

Security

• Fix CVE-2024-45310, a low-severity attack that allowed maliciously configured containers to create empty files and directories on the host.

Added

• Add support for Go 1.23. (#4360, #4372)

Fixed

• Revert "allow overriding VERSION value in Makefile" and add EXTRA_VERSION. (#4370, #4382)
• rootfs: consolidate mountpoint creation logic. (#4359)

There is no certainty in the world. This is the only certainty I have.

Important Notes

• If building with Go 1.22.x, make sure to use 1.22.4 or a later version. (see #4233 for more details)

Fixed

• Support go 1.22.4+. (#4313)
• runc list: fix race with runc delete. (#4231)
• Fix set nofile rlimit error. (#4277, #4299)
• libct/cg/fs: fix setting rt_period vs rt_runtime. (#4284)
• Fix a debug msg for user ns in nsexec. (#4315)
• script/*: fix gpg usage wrt keyboxd. (#4316)
• CI fixes and misc backports. (#4241)
• Fix codespell warnings. (#4300)

Changed

• Silence security false positives from golang/net. (#4244)
• libcontainer: allow containers to make apps think fips is enabled/disabled for testing. (#4257)
• allow overriding VERSION value in Makefile. (#4270)
• Vagrantfile.fedora: bump Fedora to 39. (#4261)
• ci/cirrus: rm centos stream 8. (#4305, #4308)

Now you're thinking with Portals™!

Security

• Fix CVE-2024-21626, a container breakout attack that took advantage of a file descriptor that was leaked internally within runc (but never leaked to the container process). In addition to fixing the leak, several strict hardening measures were added to ensure that future internal leaks could not be used to break out in this manner again. Based on our research, while no other container runtime had a similar leak, none had any of the hardening steps we've introduced (and some runtimes would not check for any file descriptors that a calling process may have leaked to them, allowing for container breakouts due to basic user error).

Happy New Year!

Fixed

• Fix several issues with userns path handling. (#4122, #4124, #4134, #4144)

Changed

• Support memory.peak and memory.swap.peak in cgroups v2. Add swapOnlyUsage in MemoryStats. This field reports swap-only usage. For cgroupv1, Usage and Failcnt are set by subtracting memory usage from memory+swap usage. For cgroupv2, Usage, Limit, and MaxUsage are set. (#4000, #4010, #4131)
• build(deps): bump github.com/cyphar/filepath-securejoin. (#4140)

Śruba, przykręcona we śnie, nie zmieni sytuacji, jaka panuje na jawie.

Added

• Support for hugetlb.<pagesize>.rsvd limiting and accounting. Fixes the issue of postres failing when hugepage limits are set. (#3859, #4077)

Fixed

• Fixed permissions of a newly created directories to not depend on the value of umask in tmpcopyup feature implementation. (#3991, #4060)
• libcontainer: cgroup v1 GetStats now ignores missing kmem.limit_in_bytes (fixes the compatibility with Linux kernel 6.1+). (#4028)
• Fix a semi-arbitrary cgroup write bug when given a malicious hugetlb configuration. This issue is not a security issue because it requires a malicious config.json, which is outside of our threat model. (#4103)
• Various CI fixes. (#4081, #4055)

There is a crack in everything. That's how the light gets in.

Added

• Added go 1.21 to the CI matrix; other CI updates. (#3976, #3958)

Fixed

• Fixed losing sticky bit on tmpfs (a regression in 1.1.8). (#3952, #3961)
• intelrdt: fixed ignoring ClosID on some systems. (#3550, #3978)

Changed

• Sum anon and file from memory.stat for cgroupv2 root usage, as the root does not have memory.current for cgroupv2. This aligns cgroupv2 root usage more closely with cgroupv1 reporting. Additionally, report root swap usage as sum of swap and memory usage, aligned with v1 and existing non-root v2 reporting. (#3933)

海纳百川 有容乃大

Added

• Support riscv64. (#3905)

Fixed

• init: do not print environment variable value. (#3879)
• libct: fix a race with systemd removal. (#3877)
• tests/int: increase num retries for oom tests. (#3891)
• man/runc: fixes. (#3892)
• Fix tmpfs mode opts when dir already exists. (#3916)
• docs/systemd: fix a broken link. (#3917)
• ci/cirrus: enable some rootless tests on cs9. (#3918)
• runc delete: call systemd's reset-failed. (#3932)
• libct/cg/sd/v1: do not update non-frozen cgroup after frozen failed. (#3921)

Changed

• CI: bump Fedora, Vagrant, bats. (#3878)
• .codespellrc: update for 2.2.5. (#3909)

Ночевала тучка золотая на груди утеса-великана.

Fixed

• When used with systemd v240+, systemd cgroup drivers no longer skip DeviceAllow rules if the device does not exist (a regression introduced in runc 1.1.3). This fix also reverts the workaround added in runc 1.1.5, removing an extra warning emitted by runc run/start. (#3845, #3708, #3671)

Added

• The source code now has a new file, runc.keyring, which contains the keys used to sign runc releases. (#3838)

In this world nothing is certain but death and taxes.

Compatibility

• This release can no longer be built from sources using Go 1.16. Using a latest maintained Go 1.20.x or Go 1.19.x release is recommended. Go 1.17 can still be used.

Fixed

• systemd cgroup v1 and v2 drivers were deliberately ignoring UnitExist error from systemd while trying to create a systemd unit, which in some scenarios may result in a container not being added to the proper systemd unit and cgroup. (#3780, #3806)
• systemd cgroup v2 driver was incorrectly translating cpuset range from spec's resources.cpu.cpus to systemd unit property (AllowedCPUs) in case of more than 8 CPUs, resulting in the wrong AllowedCPUs setting. (#3808)
• systemd cgroup v1 driver was prefixing container's cgroup path with the path of PID 1 cgroup, resulting in inability to place PID 1 in a non-root cgroup. (#3811)
• runc run/start may return "permission denied" error when starting a rootless container when the file to be executed does not have executable bit set for the user, not taking the CAP_DAC_OVERRIDE capability into account. This is a regression in runc 1.1.4, as well as in Go 1.20 and 1.20.1 (#3715, #3817)
• cgroup v1 drivers are now aware of misc controller. (#3823)
• Various CI fixes and improvements, mostly to ensure Go 1.19.x and Go 1.20.x compatibility.

囚われた屈辱は 反撃の嚆矢だ

Security

The following CVEs were fixed in this release:

• CVE-2023-25809 is a vulnerability involving rootless containers where (under specific configurations), the container would have write access to the /sys/fs/cgroup/user.slice/... cgroup hierarchy. No other hierarchies on the host were affected. This vulnerability was discovered by Akihiro Suda.

• CVE-2023-27561 was a regression in our protections against tricky /proc and /sys configurations (where the container mountpoint is a symlink) causing us to be tricked into incorrectly configuring the container, which effectively re-introduced CVE-2019-19921. This regression was present from v1.0.0-rc95 to v1.1.4 and was discovered by @Beuc. (#3785)

• CVE-2023-28642 is a different attack vector using the same regression as in CVE-2023-27561. This was reported by Lei Wang.

Fixed

• Fix the inability to use /dev/null when inside a container. (#3620)
• Fix changing the ownership of host's /dev/null caused by fd redirection (a regression in 1.1.1). (#3674, #3731)
• Fix rare runc exec/enter unshare error on older kernels, including CentOS < 7.7. (#3776)
• nsexec: Check for errors in write_log(). (#3721)
• Various CI fixes and updates. (#3618, #3630, #3640, #3729)

If you look for perfection, you'll never be content.

Fixed

• Fix mounting via wrong proc fd. When the user and mount namespaces are used, and the bind mount is followed by the cgroup mount in the spec, the cgroup was mounted using the bind mount's mount fd. (#3511)
• Switch kill() in libcontainer/nsenter to sane_kill(). (#3536)
• Fix "permission denied" error from runc run on noexec fs. (#3541)
• Fix failed exec after systemctl daemon-reload. Due to a regression in v1.1.3, the DeviceAllow=char-pts rwm rule was no longer added and was causing an error open /dev/pts/0: operation not permitted: unknown when systemd was reloaded. (#3554)
• Various CI fixes. (#3538, #3558, #3562)