In the Linux kernel, the following vulnerability has been resolved:
geneve: validate inner network offset in geneve_gro_complete()
Even with both paths gated on gs->gro_hint, geneve_gro_complete()
re-derives the inner dispatch type and length from the packet and the
current gs->gro_hint, independently of geneve_gro_receive(). The two can
disagree if gs->gro_hint flips under a concurrent geneve_quiesce()/
geneve_unquiesce() (sk_user_data is NULL across a synchronize_net()), or if
the re-read option bytes differ from the ones receive parsed.
geneve_gro_receive() already records the inner network header position in
NAPI_GRO_CB()->inner_network_offset. Have geneve_gro_complete() compute the
offset it is about to dispatch at, adding ETH_HLEN in the ETH_P_TEB case
where eth_gro_complete() steps over the inner MAC header, and bail out if
it lands past inner_network_offset.
Use a lower bound rather than exact equality: between gh_len and the inner
L3 header, geneve_gro_receive() may also have pulled an inner VLAN tag
(vlan_gro_receive() advances the recorded offset past it), which only moves
inner_network_offset further out. A valid frame therefore always satisfies
inner_nh <= inner_network_offset, while a gh_len inflated by a hint
gro_receive() did not honour dispatches past the validated inner header,
i.e. the out-of-bounds completion. Only the latter is rejected.
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 in geneve_gro_complete() on the NAPI GRO receive path for UDP-encapsulated GENEVE frames; a remote peer can deliver crafted packets to a listening GENEVE UDP port on cloud/overlay hosts without local access.
AC:L - An attacker can reliably trigger the mismatch by flooding GRO-eligible GENEVE frames with crafted GRO-hint options and timing them against geneve_quiesce()/geneve_unquiesce() during changelink, or self-trigger that race with CAP_NET_ADMIN in a user namespace.
PR:N - Exploitation requires no credentials or capabilities on the victim; the attacker only needs network reachability to a host already running a gro_hint-enabled GENEVE tunnel, which is common on overlay networks.
UI:N - Exploitation is fully automated through malicious network traffic and does not require any victim user action such as mounting filesystems or opening files.
S:C - On cloud/overlay deployments a tenant VM or remote overlay peer can corrupt the hypervisor/host kernel through underlay GENEVE GRO processing, crossing the guest/host security boundary.
C:H - When gro_complete() dispatches past inner_network_offset, eth_gro_complete() and inet_gro_complete() read headers from attacker-influenced out-of-bounds skb memory, enabling arbitrary kernel memory disclosure.
I:H - The same out-of-bounds gro_complete dispatch writes header fields (e.g., iph->tot_len and iph->check in inet_gro_complete()), causing memory corruption that can be leveraged for arbitrary write or code execution.
A:H - Out-of-bounds GRO completion can cause kernel oops/panic or forced skb drop under error paths, enabling repeatable remote denial of service against the host.
| 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 in geneve_gro_complete() on the NAPI GRO receive path for UDP-encapsulated GENEVE frames; a remote peer can deliver crafted packets to a listening GENEVE UDP port on cloud/overlay hosts without local access.
AC:L - An attacker can reliably trigger the mismatch by flooding GRO-eligible GENEVE frames with crafted GRO-hint options and timing them against geneve_quiesce()/geneve_unquiesce() during changelink, or self-trigger that race with CAP_NET_ADMIN in a user namespace.
PR:N - Exploitation requires no credentials or capabilities on the victim; the attacker only needs network reachability to a host already running a gro_hint-enabled GENEVE tunnel, which is common on overlay networks.
UI:N - Exploitation is fully automated through malicious network traffic and does not require any victim user action such as mounting filesystems or opening files.
S:C - On cloud/overlay deployments a tenant VM or remote overlay peer can corrupt the hypervisor/host kernel through underlay GENEVE GRO processing, crossing the guest/host security boundary.
C:H - When gro_complete() dispatches past inner_network_offset, eth_gro_complete() and inet_gro_complete() read headers from attacker-influenced out-of-bounds skb memory, enabling arbitrary kernel memory disclosure.
I:H - The same out-of-bounds gro_complete dispatch writes header fields (e.g., iph->tot_len and iph->check in inet_gro_complete()), causing memory corruption that can be leveraged for arbitrary write or code execution.
A:H - Out-of-bounds GRO completion can cause kernel oops/panic or forced skb drop under error paths, enabling repeatable remote denial of service against the host.
CVSS 3.1