From mboxrd@z Thu Jan 1 00:00:00 1970 From: AL-KERNEL To: kernel-cve@kernelcve.org Subject: [CVE-2026-93107][MODERATE 7.0] RDMA/rxe: Avoid reprocessing the current packet after the QP enters the error state [ Upstream Date: Fri, 18 Sep 2026 05:42:05 -0400 Message-ID: MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit X-AL-KERNEL-CVE: CVE-2026-93107 X-AL-KERNEL-Priority: MODERATE 7.0 X-AL-KERNEL-Severity: MODERATE 7.0 X-AL-KERNEL-Base-Severity: MODERATE X-AL-KERNEL-KPANIC: YES X-AL-KERNEL-ActionableScore: 5 X-AL-KERNEL-ActionableScore-Lower: 4 X-AL-KERNEL-Commit: c319b986a88cedf600c4497cd7814a891b2f9e7c List-Id: CVE: CVE-2026-93107 Priority: MODERATE 7.0 AL-KERNEL base severity: MODERATE KPANIC flag: YES Patch: RDMA/rxe: Avoid reprocessing the current packet after the QP enters the error state [ Upstream Commit: c319b986a88cedf600c4497cd7814a891b2f9e7c Upstream patch: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=c319b986a88cedf600c4497cd7814a891b2f9e7c Original CVE announcement: https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-93107 Analysis date: Fri, 18 Sep 2026 05:42:05 -0400 ActionableScore: 5 ActionableScore lower bound: 4 Actionable bucket: Actionable Moderate Manual review required: YES Summary: RDMA RXE responder processing can reprocess an already completed packet during a QP error-state race, causing duplicate successful receives, upper-layer data corruption, and eventually a NULL pointer dereference kernel crash from a reachable RDMA peer. ====================================================================== ABOUT THIS REPORT ====================================================================== The original Linux kernel CVE announcement for CVE-2026-93107 is available here: https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-93107 The original announcement does not normally provide a security severity estimate, CVSS assessment, or enough information to determine whether the reported kernel bug represents a practically relevant security issue. This report was generated by AL-KERNEL, an AI-assisted Linux kernel vulnerability analysis system developed by Alexander Larkin. It combines an autonomous classifier with LLM-assisted technical analysis and a separate ActionableScore mechanism. The purpose of this report is to prioritize Linux kernel CVEs before manual review, identify cases that require prompt investigation, and support automatic closure of issues that are unlikely to have meaningful security impact. Published priority for this report: MODERATE 7.0 Manual review required: YES A detailed explanation of the methodology and priority rules is included at the end of this message. ====================================================================== AL-KERNEL CLASSIFICATION RESULT ====================================================================== CVE-2026-93107 MODERATE https://www.kernelcve.org/list/?q=CVE-2026-93107 CHECK WITH IMPACT FROM ORIG NN MODERATE Maybe valid. Check manually. Hints by AL-KERNEL: The best (paranoid) CVSS is 'AV:A/AC:H/PR:N/UI:N/S:U/C:L/I:H/A:H';CWE-362;**CWE-476;*CWE-664;Other CVSS 'AV:A/AC:H/PR:N/UI:N/S:U/C:N/I:H/A:H';BEST CVSS score: '7.1';DESCR 'RDMA RXE responder processing can reprocess an already completed packet when a QP enters the error state during receive completion. A remote RDMA peer may trigger this through a disconnect race while packets are being completed, causing the same payload to be copied into multiple receive WQEs and reported as multiple successful completions. This can corrupt upper layer protocol streams that rely on exactly once RC delivery, and after the receive queue is exhausted the path can hit a NULL pointer dereference and crash the kernel. For the CVSS the PR:N is used because the attacker does not need a local account on the victim if the RDMA RXE endpoint is reachable as a peer. The issue is network reachable within RDMA or storage fabrics, but these fabrics are typically isolated data center networks and are usually not exposed to the public Internet. Impact is integrity loss through duplicate delivery and denial of service via kernel crash.';YES REQUIRES MANUAL CHECK; ,and ActionableScore result is Actionable Moderate (with actual score 5) YES DANGER NULLPTR INIT NETWORK KERNEL_PANIC_PLUS_UAF LINUS TEST KPANIC PACKET YES NO checked ====================================================================== ACTIONABLESCORE ANALYSIS ====================================================================== ActionableScore=5 ActionableScoreLower=4 PrimitiveClass=CLASS_C RealUAF=NO HostAdminRequired=NO NamespaceScopedPrivilege=NO ImportantCandidate=NO ManualReview=REQUIRED AutoDowngradeAllowed=NO ## 1. ActionableScore * Raw conservative score: 4 * Conservative score: 4 * Raw paranoid score: 5 * Paranoid score: 5 * Final recommended bucket: **Actionable Moderate**:: No floor or cap was applied. The paranoid score is higher because the RDS stress-test evidence shows a practical upper-layer integrity consequence, not only a theoretical duplicate-completion issue. ## 2. Signal breakdown Conservative signals: * +2 Remote or network-adjacent reachable path. A remote RDMA/RXE peer can plausibly reach responder packet processing when the RXE endpoint is exposed. * +1 Reliable kernel crash / strong DoS. The commit shows a NULL pointer dereference in `send_data_in()` / `copy_data()` after the receive queue is exhausted. * +1 Integrity plausible and demonstrated. The same SEND payload can be copied into multiple receive WQEs and reported as multiple `IB_WC_SUCCESS` completions, corrupting ULP semantics that rely on exactly-once RC delivery. * -1 Timing/race requirement. The described trigger requires QP transition to error state while packet completion is in progress, e.g. disconnect racing receive processing. Paranoid additional signal: * +1 Practical consequence demonstrated. The RDS stress test accepted duplicate completions as new messages and delivered the same datagram hundreds of times to user space, showing practical stream corruption beyond a bare crash. Material non-awarded signals: * Real UAF not awarded. The patch describes reprocessing an already completed but still held packet, not use of memory after it was freed. * Concrete repeated release/refcount not awarded. No second `put`, `kref_put`, `free`, or equivalent lifetime mutation on the same dangling reference is established. * Controlled reclaim not demonstrated. * Arbitrary or directed kernel write not demonstrated. * Confidentiality not awarded. No disclosure path is shown. * Strong LPE / strong exploitation leverage not awarded. There is no evidence of controlled object replacement, callback influence, credential impact, or arbitrary kernel R/W. * Extra availability point not awarded. The NULL dereference crash is already counted as the DoS signal. ## 3. Reachability analysis The evaluated attacker is a remote RDMA peer, not a local host user. If an RXE/RDMA endpoint is reachable, the peer can send traffic and can participate in connection teardown or disconnect behavior that races with receive completion. No host-admin-equivalent privilege on the victim is established as necessary for the peer-side trigger, so `HostAdminRequired=NO`. This assumes the RDMA/RXE service or ULP is already deployed and reachable. That prior deployment is a normal service exposure condition, not attacker host privilege. This is not a general Internet-facing networking core bug. RDMA/RXE and RDS-style deployments are typically used inside storage, cluster, or data-center fabrics. Public Internet exposure is uncommon, but internal network reachability is enough for practical triage concern. Namespace-scoped privilege is not relevant from the supplied evidence. No complete namespace-local trigger chain is established. The race is real but not a one-shot impossible condition. The commit reports many duplicate completions in stress conditions, so the race reduces conservative confidence but does not justify dismissing the issue. ## 4. Primitive and severity interpretation The exact primitive is a responder state-machine error: after a QP enters error state during completion, the already completed packet can be sent back through the request-processing chain instead of being consumed. This causes the same payload to be treated as fresh input for multiple receive WQEs. This is not an active-object UAF and not fixed stale storage in the memory-lifetime sense. It is a protocol/state-machine stale-packet reuse leading to duplicate successful completions, upper-layer integrity corruption, and then NULL pointer dereference when `qp->resp.wqe` becomes NULL. Primitive class is `CLASS_C`: demonstrated invalid-state behavior with DoS and data-integrity consequences, but without an independently useful kernel memory corruption primitive. Demonstrated consequences: * duplicate successful RDMA receive completions; * upper-layer stream/datagram corruption; * kernel NULL pointer dereference and crash. Speculative consequences: * broader application-level integrity impact in ULPs relying on exactly-once RC delivery. Not established: * privilege escalation; * arbitrary kernel read/write; * use-after-free; * controlled reclaim; * information disclosure. `ImportantCandidate=NO` because the evidence does not establish Class A or unresolved Class B memory/lifetime corruption leverage. `AutoDowngradeAllowed=NO` because this is network-reachable within RDMA fabrics and has confirmed integrity impact plus kernel crash, so it should not be automatically closed as an ordinary low-risk DoS. Operational severity is Actionable Moderate. It is important enough for manual triage in environments using RXE/RDMA/RDS, but not a strong Important candidate without additional evidence of memory corruption or privilege escalation. ## 5. One-sentence report phrase RDMA RXE responder processing can reprocess an already completed packet during a QP error-state race, causing duplicate successful receives, upper-layer data corruption, and eventually a NULL pointer dereference kernel crash from a reachable RDMA peer. ## 6. Manual review recommendation MANUAL CHECK REQUIRED Reason: the issue is network-reachable within RDMA/RXE fabrics and has confirmed integrity impact in addition to kernel crash, even though no UAF or strong memory-corruption primitive is established. ## 7. Call-site confidence Call-site confidence: high ====================================================================== UPSTREAM PATCH SUMMARY ====================================================================== Patch: RDMA/rxe: Avoid reprocessing the current packet after the QP enters the error state [ Upstream Commit: c319b986a88cedf600c4497cd7814a891b2f9e7c Upstream URL: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=c319b986a88cedf600c4497cd7814a891b2f9e7c Changed files: drivers/infiniband/sw/rxe/rxe_resp.c Diff excerpt: Not included in this email. See the upstream URL for the full patch. Full patch: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=c319b986a88cedf600c4497cd7814a891b2f9e7c ====================================================================== DETAILED REPORT METHODOLOGY ====================================================================== The original Linux kernel CVE announcement for CVE-2026-93107 can be found here: https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-93107 The original CVE announcement normally does not include a security-level estimate. In particular, it may not contain a CVSS assessment, an impact level, or enough information to determine whether the reported bug is a practically relevant security issue. One purpose of this parallel CVE list is to provide that missing technical and prioritization information. The original goal of the AL-KERNEL project was to prioritize Linux kernel CVE analysis automatically before manual review. The system can also help identify non-security issues that may be suitable for automatic closure. This report was generated by AL-KERNEL, an AI-assisted Linux kernel vulnerability analysis system developed by Alexander Larkin. The first analysis stage combines an autonomous classifier with additional LLM-based analysis. The autonomous classifier runs locally on a CPU and is based on a backpropagation neural network. Together, these mechanisms produce a technical vulnerability description, identify likely weakness types, estimate CVSS severity, and provide input for ActionableScore. Two CVSS estimates are retained because incomplete kernel vulnerability information often permits more than one defensible interpretation: Conservative CVSS vector: AV:A/AC:H/PR:N/UI:N/S:U/C:N/I:H/A:H The Best / paranoid CVSS vector: AV:A/AC:H/PR:N/UI:N/S:U/C:L/I:H/A:H The Best / paranoid CVSS score: 7.1 The conservative vector represents a lower-impact interpretation. The Best/paranoid vector intentionally represents a plausible upper-bound interpretation and should not automatically be treated as demonstrated real-world impact. CVSS may also need to be adjusted for a particular Linux deployment, because actual reachability, privileges, enabled kernel configuration, hardware, namespaces, exposed device nodes, and other environmental conditions can differ significantly between systems. A separate ActionableScore mechanism evaluates practical remediation urgency. Its analysis may include reachability, attack prerequisites, subsystem exposure, memory-corruption characteristics, denial-of-service reliability, and possible confidentiality, integrity, or privilege-escalation impact. Conservative ActionableScore: 4 Paranoid ActionableScore: 5 The final base severity is taken directly from the second tab-separated field of the AL-KERNEL classification result. ActionableScore does not replace or independently override that final AL-KERNEL decision, and if ActionableScore adjusted impact level of ALKERNEL, then you would see self-readable flags above like INCREASED_TO_HIGH_BASED_ON_ACTIONABLESCOREHIGHEREQTHAN7. For an AL-KERNEL result of MODERATE, this report uses the following additional presentation split: ActionableScore below 5 -> MODERATE REGULAR ActionableScore 5 or more -> MODERATE 7.0 The distinction between MODERATE REGULAR and MODERATE 7.0 makes it possible to identify Moderate issues that should receive manual analysis and fixes before lower-priority MODERATE REGULAR issues. In many cases, MODERATE REGULAR fixes may wait for a later rebase or routine update. There is one override in which MODERATE REGULAR becomes MODERATE 7.0 even when the ActionableScore is below 5. When the AL-KERNEL result contains the KPANIC flag, a MODERATE result is always presented as MODERATE 7.0. The KPANIC flag selected with few regexps without usage of AI at all, so it helps to detect cases when Kernel Crash happens and similar (to filter False-Negative results from the LLM usage). KPANIC indicates that a reliable kernel crash, kernel panic, or similarly serious kernel availability impact was identified by the classification workflow. AL-KERNEL base severity for this report: MODERATE KPANIC detected for this report: YES Published priority for this report (same as in Subject): MODERATE 7.0 These results are intended to support engineering triage. They are machine-generated estimates, and cases marked for manual review should be validated by a human security engineer before final disposition. For more info read docs linked from here: https://kernelcve.org/ (and you can submit you own patch there to generate such a report for non-existant CVE-id yet). Note that in many cases this AI tool selects higher severity, than real is (means you can expect Importants instead of Moderate 7.0 or Moderates 7.0 instead of regular Moderates). If you see such cases, please use reply email interface to add additional manual analyses info to this particular CVE. And please, please, let me know when you see Lows instead of Importants or Important instead of Low (because particular for such cases I need to tune this AI tool to make it better for this one and next similar). My contact email for such notifications is alexanjelausa@gmail.com (and both send reply to CVE record itself too and see "reply" button below for howto reply).