Because the target of a mount is inside a container (which may be a
volume that is shared with another container), there exists a race
condition where the target of the mount may change to a path containing
a symlink after we have sanitised the path -- resulting in us
inadvertently mounting the path outside of the container.
This is not immediately useful because we are in a mount namespace with
MS_SLAVE mount propagation applied to "/", so we cannot mount on top of
host paths in the host namespace. However, if any subsequent mountpoints
in the configuration use a subdirectory of that host path as a source,
those subsequent mounts will use an attacker-controlled source path
(resolved within the host rootfs) -- allowing the bind-mounting of "/"
into the container.
While arguably configuration issues like this are not entirely within
runc's threat model, within the context of Kubernetes (and possibly
other container managers that provide semi-arbitrary container creation
privileges to untrusted users) this is a legitimate issue. Since we
cannot block mounting from the host into the container, we need to block
the first stage of this attack (mounting onto a path outside the
container).
The long-term plan to solve this would be to migrate to libpathrs, but
as a stop-gap we implement libpathrs-like path verification through
readlink(/proc/self/fd/$n) and then do mount operations through the
procfd once it's been verified to be inside the container. The target
could move after we've checked it, but if it is inside the container
then we can assume that it is safe for the same reason that libpathrs
operations would be safe.
A slight wrinkle is the "copyup" functionality we provide for tmpfs,
which is the only case where we want to do a mount on the host
filesystem. To facilitate this, I split out the copy-up functionality
entirely so that the logic isn't interspersed with the regular tmpfs
logic. In addition, all dependencies on m.Destination being overwritten
have been removed since that pattern was just begging to be a source of
more mount-target bugs (we do still have to modify m.Destination for
tmpfs-copyup but we only do it temporarily).
Fixes: CVE-2021-30465
Reported-by: Etienne Champetier <champetier.etienne@gmail.com>
Co-authored-by: Noah Meyerhans <nmeyerha@amazon.com>
Reviewed-by: Samuel Karp <skarp@amazon.com>
Reviewed-by: Kir Kolyshkin <kolyshkin@gmail.com> (@kolyshkin)
Reviewed-by: Akihiro Suda <akihiro.suda.cz@hco.ntt.co.jp>
Signed-off-by: Aleksa Sarai <cyphar@cyphar.com>
Dockerfile: switch from go 1.15 to go 1.16.x.
Vagrantfile.centos7: switch from go 1.15 to go 1.16.4.
Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
Kir Kolyshkin (8):
tests/int/helpers: rm old code
libct/cg/sd/v2: call initPath from Path
tests/int/update.bats: don't set cpuset in setup
tests/int/helpers: generalize require cgroups_freezer
tests/int: enable/use requires cgroups_<ctrl>
tests/int/cgroups: don't check for hugetlb
tests/int: run rootless_cgroup tests for v2+systemd
Vagrantfile.fedora: set Delegate=yes
LGTMs: AkihiroSuda cyphar
Closes#2944
Instead of listing all individual controllers we want to be delegated,
just say "yes" which means "all the controllers that are available".
Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
Before this commit, "require rootless_cgroup" feature check required "cgroup"
to be present in ROOTLESS_FEATURES. The idea of the requirement, though, is
to ensure that rootless runc can manage cgroups.
In case of systemd + cgroup v2, rootless runc can manage cgroups,
thanks to systemd delegation, so modify the feature check accordingly.
Next, convert (simplify) some of the existing users to the modified check.
Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
Systemd is not able to delegate hugetlb controller, and it is needed for
cgroup v2 + systemd + rootless case (which is currently skipped because
of "requires rootless_cgroup", but will be enabled by a later commit).
The failure being fixed looks like this:
> not ok 4 runc create (limits + cgrouppath + permission on the cgroup dir) succeeds
> # (from function `check_cgroup_value' in file /vagrant/tests/integration/helpers.bash, line 188,
> # in test file /vagrant/tests/integration/cgroups.bats, line 53)
> # `check_cgroup_value "cgroup.controllers" "$(cat /sys/fs/cgroup/user.slice/user-$(id -u).slice/cgroup.controllers)"' failed
> # <....>
> # /sys/fs/cgroup/user.slice/user-2000.slice/user@2000.service/machine.slice/runc-cgroups-integration-test-20341.scope/cgroup.controllers
> # current cpuset cpu io memory pids !? cpuset cpu io memory hugetlb pids
Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
1. In case of cgroup v2 + systemd + rootless, get the list of controllers from
the own cgroup, rather than from the root one (as systemd cgroup delegation
is required in this case, and it might not be set or fully working).
2. Use "requires cgroups_<controller>" in tests that need those
controllers.
Tested on Fedora 31 (which does not support delegation of cpuset controller).
Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
Setting cpuset.cpus requires cpuset controller, which might
not be available in case of cgroup v2 + systemd < 244.
Rather than require cpuset support from every test case, do not set the initial
cpuset value from setup(), and do not check it.
Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
Sometimes Path() is called before m.path is initialized (in particular,
this happens from (*linuxContainer).newInitConfig), so we do need to
make sure to call initPath.
This fixes the following integration tests (for cgroup v2 + systemd case,
currently not enabled -- to be enabled by further commits):
* runc run (blkio weight)
* runc run (cgroupv2 mount inside container)
Fixes: ff692f289b ("Fix cgroup2 mount for rootless case")
Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
This was added by commit ca4f427af, together with set_resources_limit,
and was needed for the latter, since sed was used to update resources.
Now when we switched to jq in commit 79fe41d3c1, this kludge is no
longer needed.
Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
Add a test to check that container.Run do not leak file descriptors.
Before the previous commit, it fails like this:
exec_test.go:2030: extra fd 8 -> socket:[659703]
exec_test.go:2030: extra fd 11 -> socket:[658715]
exec_test.go:2033: found 2 extra fds after container.Run
Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
Using per cgroup manager dbus connection instances means
that every cgroup manager instance gets a new connection,
and those connections are never closed, ultimately resulting
in file descriptors limit being hit.
Revert back to using a single global dbus connection for everything,
without changing the callers.
NOTE that it is assumed a runtime can't use both root and rootless
dbus at the same time. If this happens, we panic.
Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
Commit 47ef9a104f forgot to wrap GetManagerProperty("ControlGroup")
into retryOnDisconnect. Since there's one other user of
GetManagerProperty, add getManagerProperty wrapper and use it.
Fixes: 47ef9a104f
Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
os.FindProcess never returns an error on Unix/Linux.
Use kill(0) to actually check if the process exists.
Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
Instead of manually figuring out the file and line number of the caller,
use t.Helper() so t.Fatal prints the correct one.
Before:
> utils_test.go:85: exec_test.go:536: unexpected error: container_linux.go:380: starting container process caused: exec: "catt": executable file not found in $PATH
After:
> exec_test.go:536: unexpected error: container_linux.go:380: starting container process caused: exec: "catt": executable file not found in $PATH
(the error is introduced by s/cat/catt/)
Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
I hate to keep adding those kludges, but lately TestFreeze (and
TestSystemdFreeze) from libcontainer/integration fails a lot. The
failure comes and goes, and is probably this is caused by a slow host
allocated for the test, and a slow VM on top of it.
To remediate, add a small sleep on every 25th iteration in between
asking the kernel to freeze and checking its status.
In the worst case scenario (failure to freeze), this adds about 0.4 ms
(40 x 10 us) to the duration of the call.
It is hard to measure how this affects CI as GHA plays a roulette when
allocating a node to run the test on, but it seems to help. With
additional debug info, I saw somewhat frequent "frozen after 24 retries"
or "frozen after 49 retries", meaning it succeeded right after the added
sleep.
While at it, rewrite/improve the comments.
Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
This is required for environments/build systems where a specific
go version / command needs to be used.
Signed-off-by: Enrico Weigelt, metux IT consult <info@metux.net>
https://golang.org/cmd/go/#hdr-Build_and_test_caching says:
> If you have made changes to the C libraries on your system, you will
> need to clean the cache explicitly or else use the -a build flag
> (see 'go help build') to force rebuilding of packages that depend
> on the updated C libraries.
This means that:
1. We need to either 'go clean -cache' or 'go build -a' when building
the release binary. Adding '-a' seems less intrusive / more focused.
2. The check for existing libseccomp.a (added by commit d748280aa)
is no longer needed.
Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
This package was recently updated to add support for Linux on
32-bit PowerPC (ppc), implemented by gccgo.
Signed-off-by: Ben Hutchings <ben.hutchings@essensium.com>
Akihiro Suda (4):
Revert "Makefile: rm go 1.13 workaround"
libcontainer: avoid using t.Cleanup
go.mod: demote to Go 1.13
CI: enable Go 1.13 again
LGTMs: kolyskhin cyphar
Closes#2925
t.Cleanup is not present in Go 1.13.
Dockre/Moby still builds runc with Go 1.13, so we should still support
Go 1.13.
Signed-off-by: Akihiro Suda <akihiro.suda.cz@hco.ntt.co.jp>
Aleksa Sarai