In the Linux kernel, the following vulnerability has been resolved:
geneve: gate GRO hint in geneve_gro_complete() on gs->gro_hint
geneve_gro_receive() reads the GRO hint through geneve_sk_gro_hint_off(),
which honours it only when the socket enabled IFLA_GENEVE_GRO_HINT
(gs->gro_hint). geneve_gro_complete() instead calls the low-level
geneve_opt_gro_hint_off() and acts on the hint unconditionally.
On a tunnel without the hint, receive aggregates the frames as plain
ETH_P_TEB while complete still honours an attacker-supplied hint option: it
inflates gh_len by gro_hint->nested_hdr_len (u8) and redirects the dispatch
type, so the inner gro_complete handler runs at nhoff + gh_len, an offset
receive never pulled nor validated, reading out of bounds of the skb head:
BUG: KASAN: slab-out-of-bounds in ipv6_gro_complete (net/ipv6/ip6_offload.c:196)
Read of size 1 at addr ffff88800fe91980 by task exploit/153
ipv6_gro_complete (net/ipv6/ip6_offload.c:196)
geneve_gro_complete (drivers/net/geneve.c:965)
udp_gro_complete (net/ipv4/udp_offload.c:940)
inet_gro_complete (net/ipv4/af_inet.c:1621)
__gro_flush (net/core/gro.c:306)
Gate the complete path on gs->gro_hint too via geneve_sk_gro_hint_off(), so
both paths agree. Tunnels that enable the hint are unaffected.
CVSS Vector: CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H
CVSS Score: 10
AV:N - The bug is triggered by remotely delivered UDP/GENEVE packets processed in the NAPI GRO completion path (udp_gro_complete → geneve_gro_complete → __gro_flush), not via local syscalls or physical interfaces.
AC:L - The attacker controls all packet fields (GRO hint option, nested_hdr_len, inner_proto_id) and can send multiple matching datagrams to force GRO aggregation/flush; no uncontrollable race or memory layout is required.
PR:N - No authentication or local privileges are needed; any remote host that can send UDP to a victim's GENEVE tunnel port triggers the flaw on default tunnels without IFLA_GENEVE_GRO_HINT enabled.
UI:N - Exploitation requires only transmission of crafted network packets; no victim interaction such as opening files, mounting filesystems, or clicking links is needed.
S:C - In cloud/Kubernetes overlay deployments using GENEVE, a malicious tenant VM or pod can attack the host/hypervisor kernel over the underlay network, crossing the tenant-to-host security boundary.
C:H - geneve_gro_complete() honors attacker-supplied GRO hint options that gro_receive() ignored, inflating gh_len and dispatching inner gro_complete past validated skb data, causing slab out-of-bounds reads and enabling arbitrary kernel memory disclosure.
I:H - Misdirected gro_complete handlers (e.g., ipv6_gro_complete) read and rewrite protocol header fields at attacker-chosen offsets beyond the skb head, providing out-of-bounds write primitives exploitable for control-flow hijacking.
A:H - The vulnerability provokes KASAN slab-out-of-bounds faults and kernel oops during GRO flush; repeated remote triggering can panic or hang the host, denying availability to all workloads.
| Attack Vector |
Network |
Scope |
Changed |
| Attack Complexity |
Low |
Confidentiality Impact |
High |
| Privileges Required |
None |
Integrity Impact |
High |
| User Interaction |
None |
Availability Impact |
High |
AV:N - The bug is triggered by remotely delivered UDP/GENEVE packets processed in the NAPI GRO completion path (udp_gro_complete → geneve_gro_complete → __gro_flush), not via local syscalls or physical interfaces.
AC:L - The attacker controls all packet fields (GRO hint option, nested_hdr_len, inner_proto_id) and can send multiple matching datagrams to force GRO aggregation/flush; no uncontrollable race or memory layout is required.
PR:N - No authentication or local privileges are needed; any remote host that can send UDP to a victim's GENEVE tunnel port triggers the flaw on default tunnels without IFLA_GENEVE_GRO_HINT enabled.
UI:N - Exploitation requires only transmission of crafted network packets; no victim interaction such as opening files, mounting filesystems, or clicking links is needed.
S:C - In cloud/Kubernetes overlay deployments using GENEVE, a malicious tenant VM or pod can attack the host/hypervisor kernel over the underlay network, crossing the tenant-to-host security boundary.
C:H - geneve_gro_complete() honors attacker-supplied GRO hint options that gro_receive() ignored, inflating gh_len and dispatching inner gro_complete past validated skb data, causing slab out-of-bounds reads and enabling arbitrary kernel memory disclosure.
I:H - Misdirected gro_complete handlers (e.g., ipv6_gro_complete) read and rewrite protocol header fields at attacker-chosen offsets beyond the skb head, providing out-of-bounds write primitives exploitable for control-flow hijacking.
A:H - The vulnerability provokes KASAN slab-out-of-bounds faults and kernel oops during GRO flush; repeated remote triggering can panic or hang the host, denying availability to all workloads.
CVSS 3.1