PatchSiren cyber security CVE debrief
CVE-2026-43464 Linux CVE debrief
CVE-2026-43464 is a Linux kernel vulnerability in the mlx5e receive path that affects XDP multi-buffer handling on legacy RQ. The bug can cause fragment accounting to go out of sync after XDP actions that change buffer layout, which may lead to a negative page-pool reference count and a kernel warning during page release. The issue is documented by NVD as a high-severity availability problem, and the official kernel fixes are available in the linked stable patches.
- Vendor
- Linux
- Product
- Unknown
- CVSS
- HIGH 7.5
- CISA KEV
- Not listed in stored evidence
- Original CVE published
- 2026-05-08
- Original CVE updated
- 2026-05-20
- Advisory published
- 2026-05-08
- Advisory updated
- 2026-05-20
Who should care
Kernel and platform teams running Linux systems with the mlx5 driver, especially hosts using XDP multi-buffer programs, legacy RQ, or high-throughput networking on NVIDIA/Mellanox hardware. Cloud operators, network appliance maintainers, and anyone backporting mlx5e/XDP fixes to supported kernel branches should treat this as relevant.
Technical summary
The mlx5e RX path assumed XDP buffer layout would remain stable during program execution, but XDP multi-buffer programs can alter layout via bpf_xdp_pull_data() or bpf_xdp_adjust_tail(). The reported problem is that the driver did not count dropped original fragments for all relevant actions (XDP_TX, XDP_REDIRECT, and XDP_PASS) after an earlier fix changed the accounting logic. According to the supplied description, this can leave the driver with incorrect fragment tracking and a negative pp_ref_count when pages are released. The fix restores page-fragment counting for the original XDP buffer fragments and adjusts how nr_frags is computed for xdp_update_skb_frags_info() when frag_page still points to the original tail.
Defensive priority
High for systems using mlx5e with XDP multi-buffer workloads. The impact is availability-focused, but the kernel warning and refcount corruption indicate a correctness issue that can affect stability. Prioritize if you run exposed networking workloads or have observability of intermittent page-pool warnings.
Recommended defensive actions
- Check whether your kernels fall within the vulnerable NVD ranges: 6.6.115 through before 6.7, 6.12.56 through before 6.13, 6.17.6 through before 6.18, 6.18.1 through before 6.18.19, or 6.19 through before 6.19.9.
- Apply the relevant upstream stable patch(es) linked in the official kernel references or ensure your vendor kernel has backported the fix.
- If you rely on mlx5e and XDP multi-buffer programs, prioritize validation on representative traffic and test for page-pool or refcount warnings after updating.
- Monitor for kernel warnings involving page_pool helpers or mlx5e_page_release_fragmented, as the supplied report shows this is a visible failure mode.
- Track vendor advisories and backport status for your supported kernel branches before scheduling production rollout.
Evidence notes
The vulnerability description states that mlx5e RX legacy RQ fragment counting was incorrect for XDP multi-buffer programs after bpf_xdp_pull_data() or bpf_xdp_adjust_tail(), and that the resulting mismatch can produce a negative pp_ref_count and a kernel warning. NVD marks the CVE as analyzed, severity HIGH, and lists the vulnerable version ranges. The official reference links are kernel stable patch commits, which support the remediation guidance. Published and modified timestamps in the supplied record were used as the CVE timing context.
Official resources
-
CVE-2026-43464 CVE record
CVE.org
-
CVE-2026-43464 NVD detail
NVD
-
Source item URL
nvd_modified
-
Mitigation or vendor reference
416baaa9-dc9f-4396-8d5f-8c081fb06d67 - Patch
-
Mitigation or vendor reference
416baaa9-dc9f-4396-8d5f-8c081fb06d67 - Patch
-
Mitigation or vendor reference
416baaa9-dc9f-4396-8d5f-8c081fb06d67 - Patch
Publicly disclosed in the supplied record on 2026-05-08T15:16:59.283Z, with a later metadata update on 2026-05-20T18:39:13.850Z. The CVE record and NVD entry are the authoritative sources for timing context.