In the Linux kernel, the following vulnerability has been resolved:
net/mlx5e: Prevent stale XSK buffer release on refill retry
When an XDP redirect to an AF_XDP socket fails because its RX ring is
full, the XSK core frees the buffer. During the subsequent batched refill
of a legacy cyclic RQ, mlx5e also releases the WQE's XSK buffer before
allocating a replacement. If that refill succeeds only partially, a WQE
left without a replacement retains its old buffer pointer.
The buffer can meanwhile be allocated to another WQE. A later refill
retry can then free the live buffer through the stale pointer and publish
the same UMEM frame twice.
Mark the WQE as released immediately after the driver-side free. The flag
is already cleared when a replacement buffer is assigned, so refill
retries no longer release stale pointers.
The failure is silent and produces no kernel warning or splat. A
standalone legacy cyclic-RQ zero-copy libxsk reproducer, using 64-byte UDP
traffic offered at 12 Mpps, detected it: stock stopped after 2,854,914
packets in 4.094 seconds, with 4,542 xdp_rx_ring_full events and 64
ownership/double-publication errors. With this change it processed
356,904,225 packets in 30 seconds despite 571,405 xdp_rx_ring_full events,
with no ownership or data errors.
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 - UDP datagrams reaching mlx5 drive mlx5e_completion_event() -> mlx5e_napi_poll() -> mlx5e_poll_rx_cq()/XDP redirect, then mlx5e_napi_xsk_post() -> mlx5e_post_rx_wqes() -> patched mlx5e_xsk_free_rx_wqe(); sustained routable traffic fills the XSK RX ring and triggers stale release.
AC:L - A sender can sustain small-packet traffic until __xsk_rcv_zc() reports RX-ring-full and mlx5e_xsk_alloc_rx_wqes_batched() refills only part of the missing WQEs; retries are deterministic and require no uncontrolled race or memory layout.
PR:N - The NIC-CQE/NAPI receive path contains no capability or authentication check; xsk_create()'s CAP_NET_RAW check applies to victim-side AF_XDP provisioning, while the remote UDP sender needs no privileges.
UI:N - Once the zero-copy XSK pool and XDP redirect are active, xsk_rcv_zc() failure, mlx5e_napi_xsk_post(), and refill retries occur automatically as packets arrive, without further victim action.
S:U - mlx5e_xsk_free_rx_wqe() corrupts the host kernel's xsk_buff_pool and mlx5 receive state without crossing a VM, IOMMU, or other security-authority boundary.
C:H - Duplicate xdp_buff_xsk ownership lets repeated successful __xsk_rcv_zc() calls invoke unbounded xp_release(); in unaligned-UMEM mode, free_heads_cnt can write kernel pointers past the flexible free_heads[] array, yielding heap corruption exploitable for arbitrary disclosure.
I:H - The duplicate WQEs can make xp_release() write beyond free_heads[], while xp_get_xskb() later trusts pointers popped through the corrupted count; this supplies out-of-bounds and corrupted-pointer write primitives capable of control-flow hijacking.
A:H - The free_heads[] overflow or a bogus xskb consumed by xp_get_xskb() can oops or panic; additionally, the dma_dev_need_sync refill path can store NULL on allocation failure and have the stale retry pass it to xsk_buff_free().
| 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 - UDP datagrams reaching mlx5 drive mlx5e_completion_event() -> mlx5e_napi_poll() -> mlx5e_poll_rx_cq()/XDP redirect, then mlx5e_napi_xsk_post() -> mlx5e_post_rx_wqes() -> patched mlx5e_xsk_free_rx_wqe(); sustained routable traffic fills the XSK RX ring and triggers stale release.
AC:L - A sender can sustain small-packet traffic until __xsk_rcv_zc() reports RX-ring-full and mlx5e_xsk_alloc_rx_wqes_batched() refills only part of the missing WQEs; retries are deterministic and require no uncontrolled race or memory layout.
PR:N - The NIC-CQE/NAPI receive path contains no capability or authentication check; xsk_create()'s CAP_NET_RAW check applies to victim-side AF_XDP provisioning, while the remote UDP sender needs no privileges.
UI:N - Once the zero-copy XSK pool and XDP redirect are active, xsk_rcv_zc() failure, mlx5e_napi_xsk_post(), and refill retries occur automatically as packets arrive, without further victim action.
S:U - mlx5e_xsk_free_rx_wqe() corrupts the host kernel's xsk_buff_pool and mlx5 receive state without crossing a VM, IOMMU, or other security-authority boundary.
C:H - Duplicate xdp_buff_xsk ownership lets repeated successful __xsk_rcv_zc() calls invoke unbounded xp_release(); in unaligned-UMEM mode, free_heads_cnt can write kernel pointers past the flexible free_heads[] array, yielding heap corruption exploitable for arbitrary disclosure.
I:H - The duplicate WQEs can make xp_release() write beyond free_heads[], while xp_get_xskb() later trusts pointers popped through the corrupted count; this supplies out-of-bounds and corrupted-pointer write primitives capable of control-flow hijacking.
A:H - The free_heads[] overflow or a bogus xskb consumed by xp_get_xskb() can oops or panic; additionally, the dma_dev_need_sync refill path can store NULL on allocation failure and have the stale retry pass it to xsk_buff_free().
CVSS 3.1