In the Linux kernel, the following vulnerability has been resolved:
ceph: bound MDSCapAuth path and fs_name decode in handle_session()
handle_session() decodes the MDSCapAuth records carried by a
CEPH_SESSION_OPEN message (msg_version >= 6). For each record the
match.path and match.fs_name byte strings are read by first decoding a
32-bit length and then copying that many bytes with the bare
ceph_decode_copy(). Unlike the surrounding fields, which all use the
_safe decode variants, these two copies are not preceded by a
ceph_decode_need() bounds check, and the enclosing MDSCapAuth and
MDSCapMatch struct_len fields are skipped rather than enforced as an
upper bound. A length larger than the bytes remaining in the message
front makes ceph_decode_copy() read past the end of the front buffer.
The message front is a dedicated allocation (ceph_msg_new2() ->
kvmalloc), so the over-read runs off that object. A malicious or
compromised MDS can trigger this with the first post-connect message on
mount, with no client-side user interaction; under KASAN it is reported
as a slab-out-of-bounds read in handle_session().
Impact: a malicious MDS can force the kernel client to read up to 4 GiB
past the message front allocation during session setup, crashing the
client (out-of-bounds read).
Switch both copies to ceph_decode_copy_safe(), which performs the
ceph_decode_need() bounds check before the copy and branches to the
existing bad label, matching the rest of the decoder and the error path
that frees the partially decoded cap_auths array.
CVSS Vector: CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H
CVSS Score: 9.8
AV:N - CEPH_MSG_CLIENT_SESSION arrives over the kernel Ceph messenger TCP session from an MDS peer; mds_dispatch() hands it to handle_session(), so a malicious or compromised MDS (or on-path attacker on an unsigned msgr session) reaches the unbounded decode with network data alone.
AC:L - The attacker fully controls msg->hdr.version and the MDSCapAuth path/fs_name u32 lengths; setting version>=6 and a length larger than remaining front bytes (or UINT_MAX) deterministically hits the unchecked ceph_decode_copy() with no race or layout dependency.
PR:N - Exploitation requires only sending a crafted CEPH_SESSION_OPEN as the remote MDS peer on an established session; the attacker needs no account, capability, or mount privilege on the victim beyond the client already being connected to that cluster.
UI:N - Once CephFS is mounted (standard in Rook/Kubernetes/OpenStack), the MDS emits CEPH_SESSION_OPEN as the first post-connect message on mount or reconnect, processed automatically in the ceph-msgr kworker with no further victim action at exploit time.
S:U - The out-of-bounds read and any resulting kernel memory corruption stay inside the victim host's CephFS client kernel; they do not cross a VM, sandbox, or IOMMU security boundary.
C:H - Unchecked ceph_decode_copy() of an attacker-chosen u32 length reads past the kvmalloc'd message front into adjacent kernel memory and stores those bytes in match.path/match.fs_name; unbounded out-of-bounds reads are High per kernel guidance.
I:H - Attacker-controlled _len is both the memcpy size and kcalloc(_len+1) count, so UINT_MAX wraps the allocation to ZERO_SIZE_PTR and the copy writes from address 16; over-read bytes are also installed as cap-auth path/fs_name used by ceph_mds_check_access, a memory-corruption primitive.
A:H - The over-read runs off the dedicated message-front object and can fault on unmapped pages or vmalloc guard pages, oopsing the ceph-msgr worker as KASAN reported; a malicious MDS can repeat this on every session open.
| Attack Vector |
Network |
Scope |
Unchanged |
| Attack Complexity |
Low |
Confidentiality Impact |
High |
| Privileges Required |
None |
Integrity Impact |
High |
| User Interaction |
None |
Availability Impact |
High |
AV:N - CEPH_MSG_CLIENT_SESSION arrives over the kernel Ceph messenger TCP session from an MDS peer; mds_dispatch() hands it to handle_session(), so a malicious or compromised MDS (or on-path attacker on an unsigned msgr session) reaches the unbounded decode with network data alone.
AC:L - The attacker fully controls msg->hdr.version and the MDSCapAuth path/fs_name u32 lengths; setting version>=6 and a length larger than remaining front bytes (or UINT_MAX) deterministically hits the unchecked ceph_decode_copy() with no race or layout dependency.
PR:N - Exploitation requires only sending a crafted CEPH_SESSION_OPEN as the remote MDS peer on an established session; the attacker needs no account, capability, or mount privilege on the victim beyond the client already being connected to that cluster.
UI:N - Once CephFS is mounted (standard in Rook/Kubernetes/OpenStack), the MDS emits CEPH_SESSION_OPEN as the first post-connect message on mount or reconnect, processed automatically in the ceph-msgr kworker with no further victim action at exploit time.
S:U - The out-of-bounds read and any resulting kernel memory corruption stay inside the victim host's CephFS client kernel; they do not cross a VM, sandbox, or IOMMU security boundary.
C:H - Unchecked ceph_decode_copy() of an attacker-chosen u32 length reads past the kvmalloc'd message front into adjacent kernel memory and stores those bytes in match.path/match.fs_name; unbounded out-of-bounds reads are High per kernel guidance.
I:H - Attacker-controlled _len is both the memcpy size and kcalloc(_len+1) count, so UINT_MAX wraps the allocation to ZERO_SIZE_PTR and the copy writes from address 16; over-read bytes are also installed as cap-auth path/fs_name used by ceph_mds_check_access, a memory-corruption primitive.
A:H - The over-read runs off the dedicated message-front object and can fault on unmapped pages or vmalloc guard pages, oopsing the ceph-msgr worker as KASAN reported; a malicious MDS can repeat this on every session open.
CVSS 3.1