New from Wiz Research: Critical NGINX flaw. Understand the risk and impact.

CVE-2026-99302: 
Linux Kernel vulnerability analysis and mitigation

Overview

CVE-2026-99302 is a buffer over-read vulnerability in the Linux kernel's NVMe RDMA target (nvmet-rdma) driver. The flaw exists in nvmet_rdma_parse_cm_connect_req(), which accepts RDMA-CM connection requests with arbitrarily short private data lengths without validating that the full nvme_rdma_cm_req structure is present. This allows a network-accessible attacker to trigger an out-of-bounds read of kernel memory. Affected kernel versions begin at 4.8; fixed versions include 6.12.117, 6.18.63, 7.3.4, and 7.4-rc1. The vulnerability was published on November 16, 2026, with a CVSS category estimate of Medium (Github Advisory).

Technical details

The root cause is insufficient input validation (CWE class: buffer over-read) in nvmet_rdma_parse_cm_connect_req() within drivers/nvme/target/rdma.c. The iWARP path allocates a buffer of exactly the peer-provided private_data_len bytes; when an attacker sends a six-byte MPA connection request, the parser proceeds to read the hsqsize field, which is two bytes past the end of the allocated region, triggering a KASAN-detected slab-out-of-bounds read. The fix requires that the full sizeof(struct nvme_rdma_cm_req) be present before any fields are accessed. The KASAN report traces the read through nvmet_rdma_cm_handler → cma_cm_event_handler → iw_conn_req_handler → cm_work_handler (Github Advisory).

Impact

An unauthenticated attacker with network access to a system running the NVMe RDMA target can send a crafted short RDMA connection request to read kernel memory beyond the allocated buffer, potentially disclosing sensitive kernel data. The primary impact is confidentiality — kernel memory contents may be leaked — with no direct integrity or availability impact reported. Systems exposing NVMe-oF RDMA targets (particularly over iWARP) to untrusted networks are most at risk (Github Advisory).

Exploitability

There is no known public proof-of-concept exploit and no evidence of in-the-wild exploitation as of the disclosure date. The EPSS score is 0.00209 (approximately 0.2%), indicating a low probability of exploitation in the near term. The vulnerability is not listed in the CISA Known Exploited Vulnerabilities catalog. Exploitation requires network access to an exposed NVMe RDMA target endpoint but does not require authentication (Github Advisory).

Exploitation steps

  1. Reconnaissance: Identify hosts exposing NVMe-oF RDMA targets over iWARP (TCP port 4420 by default) using network scanning tools such as Nmap.
  2. Craft malformed MPA request: Construct an RDMA-CM connection request with a private_data_len of 6 bytes (or any value smaller than sizeof(struct nvme_rdma_cm_req)), which causes the iWARP layer to allocate a buffer of exactly that size.
  3. Send the request: Initiate an iWARP connection to the target NVMe RDMA endpoint, supplying the undersized private data payload.
  4. Trigger out-of-bounds read: The kernel's nvmet_rdma_parse_cm_connect_req() reads the hsqsize field two bytes past the end of the allocated buffer, potentially disclosing adjacent kernel heap memory contents to the attacker via error paths or side channels (Github Advisory).

Indicators of compromise

  • Logs: KASAN kernel log messages containing BUG: KASAN: slab-out-of-bounds in nvmet_rdma_cm_handler in /var/log/kern.log or dmesg output.
  • Logs: Kernel stack traces referencing drivers/nvme/target/rdma.c lines 1427, 1475, 1614, or 1814 in system logs.
  • Network: Unusual or repeated short RDMA-CM connection attempts (MPA requests with very small private data) to NVMe-oF RDMA target ports (default TCP 4420) from unexpected source addresses.
  • Process: Workqueue entries for iw_cm_wq/cm_work_handler appearing at anomalous rates in kernel task listings (Github Advisory).

Mitigation and workarounds

Apply the upstream kernel patches that add validation requiring the complete nvme_rdma_cm_req structure before parsing any fields. Fixed commits are available for stable branches: 22d0eb8d7aa4 (6.12.x), 36c3af0e94b6 (6.18.x), c17d443d1e98 (7.3.x), and 635e760169a8 (mainline). Fixed stable releases are Linux 6.12.117, 6.18.63, 7.3.4, and 7.4-rc1. As a workaround, restrict network access to NVMe RDMA target endpoints to trusted hosts and networks using firewall rules or RDMA fabric access controls (Github Advisory).

Additional resources

Linux Distribution fix status

Fix availability across major Linux distributions and their releases.

Debian

Fixed

bookworm

linux

Affected

sid

linux: 7.3.4-1

Fixed

trixie

linux: 6.12.117-1

Fixed

Ubuntu

Affected

bionic (esm-infra)

linux

Affected

bionic (fips-updates)

linux-fips

Affected

bionic (fips)

linux-fips

Affected

devel

linux

Affected

focal

linux-azure-fde-5.15

Not Affected

focal (esm-infra)

linux

Affected

focal (fips-updates)

linux-fips

Affected

focal (fips)

linux-fips

Affected

Source: This report was generated using AI

Related Linux Kernel vulnerabilities:

CVE ID

Severity

Score

Technologies

Component name

CISA KEV exploit

Has fix

Published date

CVE-2026-99310HIGH7.5
  • Linux Kernel logoLinux Kernel
  • linux
NoYesNov 16, 2026
CVE-2026-99308MEDIUM6.8
  • Linux Kernel logoLinux Kernel
  • linux-gcp-4.15
NoYesNov 16, 2026
CVE-2026-99299MEDIUM4.6
  • Linux Kernel logoLinux Kernel
  • linux-gcp
NoYesNov 16, 2026
CVE-2026-99304NONEN/A
  • Linux Kernel logoLinux Kernel
  • kernel-64k-debug-modules-extra
NoYesNov 16, 2026
CVE-2026-99302NONEN/A
  • Linux Kernel logoLinux Kernel
  • libbpf
NoYesNov 16, 2026

Free Vulnerability Assessment

Benchmark your Cloud Security Posture

Evaluate your cloud security practices across 9 security domains to benchmark your risk level and identify gaps in your defenses.

Request assessment

Additional Wiz resources

Get a personalized demo

Ready to see Wiz in action?

"Best User Experience I have ever seen, provides full visibility to cloud workloads."
David EstlickCISO
"Wiz provides a single pane of glass to see what is going on in our cloud environments."
Adam FletcherChief Security Officer
"We know that if Wiz identifies something as critical, it actually is."
Greg PoniatowskiHead of Threat and Vulnerability Management