In the Linux kernel, the following vulnerability has been resolved:
xdp: fix zero-copy frame layout
xdp_convert_zc_to_xdp_frame() clones an XSK packet into an order-0 page
and advertises PAGE_SIZE as its frame size. It allows the copied frame
to occupy the page tail needed by skb_shared_info and records zero
headroom even when metadata separates the frame header from packet data.
An AF_XDP zero-copy packet redirected through cpumap can therefore make
the skb overlap skb_shared_info or place it beyond the allocated page.
Limit the copied layout to SKB_WITH_OVERHEAD(PAGE_SIZE) and include the
metadata length in frame headroom. Redirect callers already handle a
NULL conversion result.
BUG: KASAN: slab-out-of-bounds in skb_gro_receive
Write of size 4 at addr ffff88800cf37004 by task cpumap/1/map:1/146
Call Trace:
skb_gro_receive (net/core/gro.c:174)
udp_gro_receive (net/ipv4/udp_offload.c:812)
inet_gro_receive (net/ipv4/af_inet.c:1539)
dev_gro_receive (net/core/gro.c:515)
gro_receive_skb (net/core/gro.c:633)
cpu_map_kthread_run (kernel/bpf/cpumap.c:395)
kthread (kernel/kthread.c:436)
ret_from_fork (arch/x86/kernel/process.c:164)
ret_from_fork_asm (arch/x86/entry/entry_64.S:255)
Kernel panic - not syncing: KASAN: panic_on_warn set ...
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 - Received AF_XDP zero-copy packets reach xdp_convert_zc_to_xdp_frame() on XDP_REDIRECT into cpumap or a veth. On an internet-facing NIC with that hybrid XDP/AF_XDP setup, unauthenticated remote frames drive the cloned page into skb/GRO; AF_XDP bind is a victim-side prerequisite, not a local-only syscall bug.
AC:L - The sender chooses packet length, so a 4K UMEM/jumbo frame makes the clone occupy the skb_shared_info tail, and normal cpumap/veth metadata (bpf_xdp_adjust_meta) misplaces sinfo past the page. Either layout is attacker-reproducible with no race or uninfluenced memory-layout dependency.
PR:N - The remote sender needs no account or capability. CAP_NET_RAW, CAP_NET_ADMIN, and CAP_BPF are required only by the victim process that binds the XSK queue and loads the XDP redirect program, matching other AF_XDP RX-path kernel scores.
UI:N - After the interface is configured, inbound packets are converted and passed to GRO automatically in NAPI or the cpumap kthread; no victim mount, click, or other per-packet action is required.
S:U - The out-of-bounds skb_shared_info write corrupts host kernel heap/page memory in the same kernel authority. It is ordinary kernel memory corruption, not a VM, IOMMU, or sandbox boundary crossing.
C:H - skb_shared_info is placed past the order-0 page or overlapped with the copied frame, so GRO and build_skb write into adjacent objects that hold kernel pointers; that corruption can be turned into an arbitrary kernel read, so confidentiality is High.
I:H - KASAN records a 4-byte out-of-bounds write in skb_gro_receive while building the skb around the mis-sized ZC clone. An OOB write of skb_shared_info (including pointer fields) is an arbitrary-write/control-flow primitive, so integrity is High.
A:H - The same GRO path panics under KASAN and oopses without it when skb_shared_info is written off the allocated page, so a remote packet can crash the kernel.
| 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 - Received AF_XDP zero-copy packets reach xdp_convert_zc_to_xdp_frame() on XDP_REDIRECT into cpumap or a veth. On an internet-facing NIC with that hybrid XDP/AF_XDP setup, unauthenticated remote frames drive the cloned page into skb/GRO; AF_XDP bind is a victim-side prerequisite, not a local-only syscall bug.
AC:L - The sender chooses packet length, so a 4K UMEM/jumbo frame makes the clone occupy the skb_shared_info tail, and normal cpumap/veth metadata (bpf_xdp_adjust_meta) misplaces sinfo past the page. Either layout is attacker-reproducible with no race or uninfluenced memory-layout dependency.
PR:N - The remote sender needs no account or capability. CAP_NET_RAW, CAP_NET_ADMIN, and CAP_BPF are required only by the victim process that binds the XSK queue and loads the XDP redirect program, matching other AF_XDP RX-path kernel scores.
UI:N - After the interface is configured, inbound packets are converted and passed to GRO automatically in NAPI or the cpumap kthread; no victim mount, click, or other per-packet action is required.
S:U - The out-of-bounds skb_shared_info write corrupts host kernel heap/page memory in the same kernel authority. It is ordinary kernel memory corruption, not a VM, IOMMU, or sandbox boundary crossing.
C:H - skb_shared_info is placed past the order-0 page or overlapped with the copied frame, so GRO and build_skb write into adjacent objects that hold kernel pointers; that corruption can be turned into an arbitrary kernel read, so confidentiality is High.
I:H - KASAN records a 4-byte out-of-bounds write in skb_gro_receive while building the skb around the mis-sized ZC clone. An OOB write of skb_shared_info (including pointer fields) is an arbitrary-write/control-flow primitive, so integrity is High.
A:H - The same GRO path panics under KASAN and oopses without it when skb_shared_info is written off the allocated page, so a remote packet can crash the kernel.
CVSS 3.1