In the Linux kernel, the following vulnerability has been resolved:
nstree: check listing permission before taking a namespace reference
legitimize_ns() takes a reference on the candidate namespace before
may_list_ns() has decided whether the caller may see it. The
__free(ns_put) cleanup on the denied path can drop the last reference to a
mount namespace while we still hold the rcu read lock, and put_mnt_ns()
may sleep there. This is the same problem commit 2ec2aff3c8e2 ("ns: make
sure reference are dropped outside of rcu lock") fixed for the put_user()
path. Neither ns_requested() nor may_list_ns() needs a reference, both
only look at the namespace type and at the caller's own namespaces, so do
the checks first and take the reference last.
Splat:
Voluntary context switch within RCU read-side critical section!
WARNING: kernel/rcu/tree_plugin.h:332 at rcu_note_context_switch+0x238/0x2a0, CPU#5: a/3442
CPU: 5 UID: 1000 PID: 3442 Comm: a Not tainted 7.0.0-30-generic #30-Ubuntu PREEMPT(lazy)
RIP: 0010:rcu_note_context_switch+0x238/0x2a0
Call Trace:
<TASK>
__schedule+0xcf/0x650
schedule+0x27/0x90
schedule_preempt_disabled+0x15/0x30
__mutex_lock.constprop.0+0x550/0xaf0
__mutex_lock_slowpath+0x13/0x20
mutex_lock+0x3b/0x50
exp_funnel_lock+0xb2/0x260
synchronize_rcu_expedited+0xe7/0x220
namespace_unlock+0x26a/0x320
put_mnt_ns+0xd3/0x120
mntns_put+0xe/0x20
do_listns+0x13e/0x560
__do_sys_listns+0x126/0x2d0
__x64_sys_listns+0x20/0x30
x64_sys_call+0x2366/0x2390
do_syscall_64+0x105/0x5a0
entry_SYSCALL_64_after_hwframe+0x76/0x7e
</TASK>
CVSS Vector: CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H
CVSS Score: 7.8
AV:L - The bug is reached only through the local listns() syscall (SYSCALL_DEFINE4(listns) -> do_listns()/do_listns_userns() -> legitimize_ns()). No remote protocol supplies any input.
AC:L - The attacker controls both sides of the race. One process loops listns() while another creates and tears down its own mount namespaces with unshare(CLONE_NEWUSER|CLONE_NEWNS), so legitimize_ns() ends up holding the last reference on the may_list_ns() denied path.
PR:L - listns() has no capability gate, and may_list_ns() only filters what is returned after the reference is taken. The commit's splat shows the trigger from UID 1000, and the victim mount namespaces can be created without privilege through user namespaces.
UI:N - The attacker's own syscalls produce both the listing and the namespace teardown. No other user takes part.
S:U - This is corruption and a hang inside the kernel's own namespace and RCU handling. No hypervisor, IOMMU or other separately governed boundary is crossed.
C:H - put_mnt_ns() queues call_rcu(mnt_ns_release_rcu) and then sleeps in synchronize_rcu_expedited() inside do_listns()'s rcu_read_lock. On non-preemptible-RCU kernels that lets the mnt_namespace be kfree'd, and next_ns_common(ns) then walks list pointers read from freed memory.
I:H - After the use-after-free, the iterator hands a freed, reallocatable object to legitimize_ns(), which increments its refcount through ns_get_unless_inactive() and calls ns->ops->put(). That gives a write primitive and an indirect call through attacker-sprayable memory.
A:H - On PREEMPT_RCU kernels, synchronize_rcu_expedited() called from within the caller's own RCU read-side section waits on itself and hangs the task. Elsewhere the result is a sleeping-in-atomic BUG or a use-after-free crash, and an unprivileged user can repeat it at will.
| Attack Vector |
Local |
Scope |
Unchanged |
| Attack Complexity |
Low |
Confidentiality Impact |
High |
| Privileges Required |
Low |
Integrity Impact |
High |
| User Interaction |
None |
Availability Impact |
High |
AV:L - The bug is reached only through the local listns() syscall (SYSCALL_DEFINE4(listns) -> do_listns()/do_listns_userns() -> legitimize_ns()). No remote protocol supplies any input.
AC:L - The attacker controls both sides of the race. One process loops listns() while another creates and tears down its own mount namespaces with unshare(CLONE_NEWUSER|CLONE_NEWNS), so legitimize_ns() ends up holding the last reference on the may_list_ns() denied path.
PR:L - listns() has no capability gate, and may_list_ns() only filters what is returned after the reference is taken. The commit's splat shows the trigger from UID 1000, and the victim mount namespaces can be created without privilege through user namespaces.
UI:N - The attacker's own syscalls produce both the listing and the namespace teardown. No other user takes part.
S:U - This is corruption and a hang inside the kernel's own namespace and RCU handling. No hypervisor, IOMMU or other separately governed boundary is crossed.
C:H - put_mnt_ns() queues call_rcu(mnt_ns_release_rcu) and then sleeps in synchronize_rcu_expedited() inside do_listns()'s rcu_read_lock. On non-preemptible-RCU kernels that lets the mnt_namespace be kfree'd, and next_ns_common(ns) then walks list pointers read from freed memory.
I:H - After the use-after-free, the iterator hands a freed, reallocatable object to legitimize_ns(), which increments its refcount through ns_get_unless_inactive() and calls ns->ops->put(). That gives a write primitive and an indirect call through attacker-sprayable memory.
A:H - On PREEMPT_RCU kernels, synchronize_rcu_expedited() called from within the caller's own RCU read-side section waits on itself and hangs the task. Elsewhere the result is a sleeping-in-atomic BUG or a use-after-free crash, and an unprivileged user can repeat it at will.
CVSS 3.1