Rachid Koucha [Sat, 15 Jun 2019 13:17:50 +0000 (15:17 +0200)]
Fixed file descriptor leak for network namespace
In privileged mode, the container startup looses a file descriptor for "handler->nsfd[LX_NS_NET]". At line 1782, we preserve the namespaces file descriptor (in privileged mode, the network namespace is also preserved) :
for (i = 0; i < LXC_NS_MAX; i++)
if (handler->ns_on_clone_flags & ns_info[i].clone_flag)
INFO("Cloned %s", ns_info[i].flag_name);
if (!lxc_try_preserve_namespaces(handler, handler->ns_on_clone_flags, handler->pid)) {
ERROR("Failed to preserve cloned namespaces for lxc.hook.stop");
goto out_delete_net;
}
Then at line 1830, we preserve one more time the network namespace :
ret = lxc_try_preserve_ns(handler->pid, "net");
if (ret < 0) {
if (ret != -EOPNOTSUPP) {
SYSERROR("Failed to preserve net namespace");
goto out_delete_net;
}
The latter overwrites the file descriptor already stored in handler->nsfd[LXC_NS_NET] at line 1786.
So, this fix checks that the entry is not already filled.
seccomp: do not set SECCOMP_FILTER_FLAG_NEW_LISTENER
Do not set SECCOMP_FILTER_FLAG_NEW_LISTENER as seccomp attribute.
Prior to libseccomp merging support for SECCOMP_RET_USER_NOTIF there was a
libseccomp specific attribute that needed to be set before
SECCOMP_RET_USER_NOTIF could be used. This has been removed.
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
BugLink: https://bugs.launchpad.net/bugs/1831258 Cc: Dimitri John Ledkov <xnox@ubuntu.com> Cc: Scott Moser <smoser@ubuntu.com> Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
In addition to isolated cpus we also need to account for offline cpus when our
ancestor cgroup is the root cgroup and we have not been initialized yet.
Closes #2953.
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
Thomas Parrott [Wed, 15 May 2019 14:54:12 +0000 (15:54 +0100)]
network: move phys netdevs back to monitor's net ns rather than pid 1's
Updates lxc_restore_phys_nics_to_netns() to move phys netdevs back to the monitor's network namespace rather than the previously hardcoded PID 1 net ns.
This is to fix instances where LXC is started inside a net ns different from PID 1 and physical devices are moved back to a different net ns when the container is shutdown than the net ns than where the container was started from.
Signed-off-by: Thomas Parrott <thomas.parrott@canonical.com>
Tycho Andersen [Thu, 9 May 2019 18:18:10 +0000 (14:18 -0400)]
lxc_clone: get rid of some indirection
We have a do_clone(), which just calls a void f(void *) that it gets
passed. We build up a struct consisting of two args that are just the
actual arg and actual function. Let's just have the syscall do this for us.
Tycho Andersen [Thu, 9 May 2019 17:52:30 +0000 (13:52 -0400)]
lxc_clone: pass non-stack allocated stack to clone
There are two problems with this code:
1. The math is wrong. We allocate a char *foo[__LXC_STACK_SIZE]; which
means it's really sizeof(char *) * __LXC_STACK_SIZE, instead of just
__LXC_STACK SIZE.
2. We can't actually allocate it on our stack. When we use CLONE_VM (which
we do in the shared ns case) that means that the new thread is just
running one page lower on the stack, but anything that allocates a page
on the stack may clobber data. This is a pretty short race window since
we just do the shared ns stuff and then do a clone without CLONE_VM.
However, it does point out an interesting possible privilege escalation if
things aren't configured correctly: do_share_ns() sets up namespaces while
it shares the address space of the task that spawned it; once it enters the
pid ns of the thing it's sharing with, the thing it's sharing with can
ptrace it and write stuff into the host's address space. Since the function
that does the clone() is lxc_spawn(), it has a struct cgroup_ops* on the
stack, which itself has function pointers called later in the function, so
it's possible to allocate shellcode in the address space of the host and
run it fairly easily.
ASLR doesn't mitigate this since we know exactly the stack offsets; however
this patch has the kernel allocate a new stack, which will help. Of course,
the attacker could just check /proc/pid/maps to find the location of the
stack, but they'd still have to guess where to write stuff in.
The thing that does prevent this is the default configuration of apparmor.
Since the apparmor profile is set in the second clone, and apparmor
prevents ptracing things under a different profile, attackers confined by
apparmor can't do this. However, if users are using a custom configuration
with shared namespaces, care must be taken to avoid this race.
Shared namespaces aren't widely used now, so perhaps this isn't a problem,
but with the advent of crio-lxc for k8s, this functionality will be used
more.
Rachid Koucha [Mon, 13 May 2019 11:13:18 +0000 (13:13 +0200)]
Config: check for %m availability
GLIBC supports %m to avoid calling strerror(). Using it saves some code space.
==> This check will define HAVE_M_FORMAT to be use wherever possible (e.g. log.h)
Rachid Koucha [Fri, 10 May 2019 16:56:12 +0000 (18:56 +0200)]
lxccontainer: do not display if missing privileges
lxc-ls without root privileges on privileged containers should not display
information. In lxc_container_new(), ongoing_create()'s result is not checked
for all possible returned values. Hence, an unprivileged user can send command
messages to the container's monitor. For example:
$ lxc-ls -P /.../tests -f
NAME STATE AUTOSTART GROUPS IPV4 IPV6 UNPRIVILEGED
ctr - 0 - - - false
$ sudo lxc-ls -P /.../tests -f
NAME STATE AUTOSTART GROUPS IPV4 IPV6 UNPRIVILEGED
ctr RUNNING 0 - 10.0.3.51 - false
After this change:
$ lxc-ls -P /.../tests -f <-------- No more display without root privileges
$ sudo lxc-ls -P /.../tests -f
NAME STATE AUTOSTART GROUPS IPV4 IPV6 UNPRIVILEGED
ctr RUNNING 0 - 10.0.3.37 - false
$
Signed-off-by: Rachid Koucha <rachid.koucha@gmail.com> Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
Rachid Koucha [Fri, 10 May 2019 15:01:13 +0000 (17:01 +0200)]
New --bbpath option and unecessary --rootfs checks
. Add the "--bbpath" option to pass an alternate busybox pathname instead of the one found from ${PATH}.
. Take this opportunity to add some formatting in the usage display
. As a try is done to pick rootfs from the config file and set it to ${path}/rootfs, it is unnecessary to make it mandatory
Note the clone() syscall ignores unknown flags which is usually a design
mistake. However, for us this bug is a feature since we can just pass the flag
along and see whether the kernel has given us a pidfd.
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>