AF_ALG (address family 38) exposes the Linux kernel crypto API to
userspace via socket(2). Containers have no legitimate need for this
interface under the default profile, and leaving it accessible widens
the kernel attack surface unnecessarily (see https://copy.fail/).
Signed-off-by: Paweł Gronowski <pawel.gronowski@docker.com>
Add a comment explaining the purpose of the socket rules and noting that
on 32-bit x86, socket() goes through socketcall(2) which is allowed
unconditionally, so these arg filters only apply to the direct socket
syscall.
Signed-off-by: Paweł Gronowski <pawel.gronowski@docker.com>
Tests in pkg/cri/[sb]server/container_create_linux_test.go depends on go:noinline
since Go 1.21.
e.g.,
> ```
> === FAIL: pkg/cri/sbserver TestGenerateSeccompSecurityProfileSpecOpts/should_set_default_seccomp_when_seccomp_is_runtime/default (0.00s)
> container_create_linux_test.go:1013:
> Error Trace: /home/runner/work/containerd/containerd/pkg/cri/sbserver/container_create_linux_test.go:1013
> Error: Not equal:
> expected: 0x263d880
> actual : 0x263cbc0
> Test: TestGenerateSeccompSecurityProfileSpecOpts/should_set_default_seccomp_when_seccomp_is_runtime/default
> ```
See comments in PR 8957.
Thanks to Wei Fu for analyzing this.
Co-authored-by: Wei Fu <fuweid89@gmail.com>
Signed-off-by: Akihiro Suda <akihiro.suda.cz@hco.ntt.co.jp>
Follow up to 94faa70df4. The commit referenced allowed `ptrace` calls in the default seccomp profile following the usual tracing security checks in for Kernels newer than 4.8. Kernels prior to this version are susceptible to [CVE-2019-2054](https://github.com/advisories/GHSA-qgfr-27qf-f323). Moby's default had allowed for `ptrace` for kernels newer than 4.8 at the time the commit was created. The current [seccomp default](https://github.com/moby/moby/blob/master/profiles/seccomp/default_linux.go#L405-L417) has been updated to include `process_vm_read` and `process_vm_write`. Mirror that policy to complete the classic ptrace set of APIs.
Signed-off-by: Juan Hoyos <juan.s.hoyos@outlook.com>
Add pkey_alloc(2), pkey_free(2) and pkey_mprotect(2) in seccomp default profile.
pkey_alloc(2), pkey_free(2) and pkey_mprotect(2) can only configure
the calling process's own memory, so they are existing "safe for everyone" syscalls.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Allow the following syscalls by default:
- `landlock_add_rule`
- `landlock_create_ruleset`
- `landlock_restrict_self`
See https://landlock.io/
Signed-off-by: Akihiro Suda <akihiro.suda.cz@hco.ntt.co.jp>
This system call is only available on 32- and 64-bit PowerPC, it is used
by modern programming language implementations to implement coroutine
features through userspace context switches.
moby [1] and systemd nspawn [2] already whitelist this system call so it
makes sense to whitelist it in containerd as well.
[1]: https://github.com/moby/moby/pull/43092
[2]: https://github.com/systemd/systemd/pull/9487
Signed-off-by: Sören Tempel <soeren+git@soeren-tempel.net>
The io/ioutil package has been deprecated as of Go 1.16, see
https://golang.org/doc/go1.16#ioutil. This commit replaces the existing
io/ioutil functions with their new definitions in io and os packages.
Signed-off-by: Eng Zer Jun <engzerjun@gmail.com>
clone3 is explicitly requested to give ENOSYS instead of the default EPERM, when CAP_SYS_ADMIN is unset.
See moby/moby PR 42681 (thanks to berrange).
Without this commit, rawhide image does not work:
```console
$ sudo ctr run --rm --net-host --seccomp registry.fedoraproject.org/fedora:rawhide foo /usr/bin/curl google.com
curl: (6) getaddrinfo() thread failed to start
```
Signed-off-by: Akihiro Suda <akihiro.suda.cz@hco.ntt.co.jp>
These syscalls (some of which have been in Linux for a while but were
missing from the profile) fall into a few buckets:
* close_range(2), epoll_wait2(2) are just extensions of existing "safe
for everyone" syscalls.
* The mountv2 API syscalls (fs*(2), move_mount(2), open_tree(2)) are
all equivalent to aspects of mount(2) and thus go into the
CAP_SYS_ADMIN category.
* process_madvise(2) is similar to the other process_*(2) syscalls and
thus goes in the CAP_SYS_PTRACE category.
Co-authored-by: Aleksa Sarai <asarai@suse.de>
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Adds the io-uring related system call introduced in kernel 5.1 to the
seccomp whitelist. With older kernels or older versions of libseccomp,
this configure will be omitted.
Note that io_uring will grow support for more syscalls in the future
so we should keep an eye on this.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
On a ppc64le host, running postgres (tried with 9.4 to 9.6) gives the following
warning when trying to flush data to disks (which happens very frequently):
WARNING: could not flush dirty data: Operation not permitted.
A quick dig in postgres source code indicate it uses sync_file_range(2) to
flush data; which on ppe64le and arm64 is translated to sync_file_range2(2)
for alignements reasons.
The profile did not allow sync_file_range2(2), making postgres sad because
it can not flush its buffers. arm_sync_file_range(2) is an ancient alias to
sync_file_range2(2), the syscall was renamed in Linux 2.6.22 when the same
syscall was added for PowerPC.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
From personality(2):
Have uname(2) report a 2.6.40+ version number rather than a 3.x version
number. Added as a stopgap measure to support broken applications that
could not handle the kernel version-numbering switch from 2.6.x to 3.x.
This allows both "UNAME26|PER_LINUX" and "UNAME26|PER_LINUX32".
Fixes: "setarch broken in docker packages from Debian stretch"
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Add the membarrier syscall to the default seccomp profile.
It is for example used in the implementation of dlopen() in
the musl libc of Alpine images.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Enabled adjtimex in the default profile without requiring CAP_SYS_TIME privilege.
The kernel will check CAP_SYS_TIME and won't allow setting the time.
Fixes: Getting the system time with ntptime returns an error in an unprivileged
container
To verify, inside a CentOS 7 container:
yum install -y ntp
ntptime
# ntp_gettime() returns code 0 (OK)
ntpdate -v time.nist.gov
# ntpdate[84]: Can't adjust the time of day: Operation not permitted
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
This call is what is used to implement `dmesg` to get kernel messages
about the host. This can leak substantial information about the host.
It is normally available to unprivileged users on the host, unless
the sysctl `kernel.dmesg_restrict = 1` is set, but this is not set
by standard on the majority of distributions. Blocking this to restrict
leaks about the configuration seems correct.
Relates to docker/docker#37897 "docker exposes dmesg to containers by default"
See also https://googleprojectzero.blogspot.com/2018/09/a-cache-invalidation-bug-in-linux.html
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Paweł Gronowski