From mboxrd@z Thu Jan 1 00:00:00 1970 From: AL-KERNEL To: kernel-cve@kernelcve.org Subject: [CVE-2026-97495][MODERATE 7.0] drm/amdkfd: Check bounds on allocate_doorbell [ Upstream Date: Thu, 24 Sep 2026 14:37:20 -0400 Message-ID: MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit X-AL-KERNEL-CVE: CVE-2026-97495 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: unknown X-AL-KERNEL-Commit: e0cecd04a4fc4c75e375b91cf53eab5ba2e4ab7f List-Id: CVE: CVE-2026-97495 Priority: MODERATE 7.0 AL-KERNEL base severity: MODERATE KPANIC flag: YES Patch: drm/amdkfd: Check bounds on allocate_doorbell [ Upstream Commit: e0cecd04a4fc4c75e375b91cf53eab5ba2e4ab7f Upstream patch: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=e0cecd04a4fc4c75e375b91cf53eab5ba2e4ab7f Original CVE announcement: https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-97495 Analysis date: Thu, 24 Sep 2026 14:37:20 -0400 ActionableScore: 5 ActionableScore lower bound: unknown Actionable bucket: Actionable Moderate / Strong Important candidate requiring manual review Manual review required: YES Summary: No one-sentence report phrase was found. ====================================================================== ABOUT THIS REPORT ====================================================================== The original Linux kernel CVE announcement for CVE-2026-97495 is available here: https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-97495 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-2026-97495 MODERATE https://www.kernelcve.org/list/?q=CVE-2026-97495 CHECK WITH IMPACT FROM ORIG NN LOW Maybe valid. Check manually. Hints by AL-KERNEL: The best (paranoid) CVSS is 'AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H';*CWE-787;CWE-129;*CWE-20;Other CVSS 'AV:L/AC:L/PR:L/UI:N/S:U/C:L/I:L/A:H';BEST CVSS score: '7.8';DESCR 'An unchecked doorbell ID in the AMD KFD queue restore path is passed to __test_and_set_bit as an index into qpd doorbell_bitmap. A value at or above KFD_MAX_NUM_OF_QUEUES_PER_PROCESS can make the atomic bitmap operation access and modify a machine word outside the allocated bitmap, creating a limited out of bounds kernel write. A local process may reach this path while restoring GPU queue state through CRIU or a related KFD interface. For the CVSS the PR:L value is selected because exploitation requires a local process with access to the KFD device and GPU queue restore interface, but this access can be delegated to an unprivileged compute user, container, or service account and does not necessarily require full administrative control. The issue is not network reachable. Impact is at least denial of service through kernel memory corruption and may include confidentiality or integrity impact or local privilege escalation if the affected adjacent memory can be controlled. The exact exploitability depends on bitmap allocation layout, reachable restore ID values, and the kernel object located after the bitmap.';YES REQUIRES MANUAL CHECK; ,and ActionableScore result is Actionable Moderate / Strong Important candidate requiring manual review (with actual score 5) SKIP CVE-2026-97495 UNKNOWN SKIP No affected files built, so skip this CVE NO - - unknown MAYBE OOB HARDWARE KPANIC KPANIC INCREASED_TO_HIGH_BASED_ON_GUESSCVSS DECREASED_TO_MODERATE7_BASED_ON_FALSEPOSCHECKOFKP - - checked ====================================================================== ACTIONABLESCORE ANALYSIS ====================================================================== ActionableScore=5 PrimitiveClass=CLASS_A RealUAF=NO HostAdminRequired=UNKNOWN NamespaceScopedPrivilege=UNKNOWN ImportantCandidate=YES ManualReview=REQUIRED AutoDowngradeAllowed=NO 1. ActionableScore * Raw conservative score: 3 * Conservative score: 5 * Raw paranoid score: 3 * Paranoid score: 5 * Applied conservative safety floor: 3 -> 5 because the patch establishes an attacker-influenced out-of-bounds atomic kernel write. * Applied paranoid safety floor: 3 -> 5 for the same concrete corruption primitive. No speculative exploitation points were added. * Final recommended bucket: **Actionable Moderate / Strong Important candidate requiring manual review**:: 2. Signal breakdown * +2 Strong corruption primitive. An unchecked `restore_id` is used as the bit index for `__test_and_set_bit()` on `qpd->doorbell_bitmap`. An out-of-range value causes an atomic read-modify-write outside the intended bitmap. * +1 Reliable kernel crash or strong DoS. A sufficiently invalid index can access unrelated or unmapped kernel memory, while a smaller over-index can corrupt an adjacent allocation. * +0 Local unprivileged trigger. Access through `/dev/kfd` is plausible, but the supplied evidence does not establish that the complete CRIU queue-restore operation is available to an ordinary user. * +0 Strong LPE leverage. The attacker selects the bit index, but the operation can only set one bit relative to a fixed bitmap base. The reachable range, allocation layout, and adjacent target objects are not supplied. * +0 Confidentiality. No disclosure or kernel read-back path is established. * +0 separate Integrity point. The OOB modification is already represented by the strong corruption primitive, and no security-sensitive target is identified. * +0 additional Availability. The same potential kernel crash is not counted twice. * No privilege penalty was applied because host-admin-equivalent privilege is not positively established. * No hardware penalty was applied. AMD GPU support is hardware-specific, but the supplied evidence does not show rare or physical-only hardware. * No race penalty was applied because this is direct missing bounds validation, not a timing race. * Controlled reclaim, arbitrary kernel read/write, callback control, credential corruption, and demonstrated privilege escalation are not established. * No release, refcount, or object-lifetime sequence exists in the patch. The repeated-release rules and UAF floors do not apply. 3. Reachability analysis The vulnerable value is supplied through the AMD KFD queue restoration path used by CRIU and reaches `allocate_doorbell()` as `restore_id`. The caller can therefore request a specific doorbell bitmap index. A local process must have access to the relevant KFD device and queue-restore interface. Ordinary ROCm workloads can use `/dev/kfd` when device permissions are granted, but the supplied patch does not show the complete ioctl path or whether this particular restoration operation additionally requires `CAP_CHECKPOINT_RESTORE`, `CAP_SYS_ADMIN`, ptrace-style authorization, or another privileged CRIU context. Therefore, below-host-admin reachability is not positively established and `HostAdminRequired` remains `UNKNOWN`. No user-namespace capability checks are visible, and there is no evidence that the complete restore sequence can be performed in an attacker-created namespace. `NamespaceScopedPrivilege` is therefore `UNKNOWN`. Prior availability of a compatible AMD GPU, KFD, and the SOC15 CP queue path is required. These are deployment prerequisites rather than evidence of rare hardware. There is no network trigger and no race requirement. Trigger repeatability depends on whether a caller can repeatedly submit restoration requests after an invalid request corrupts state or terminates the process. 4. Primitive and severity interpretation The exact primitive is an attacker-influenced out-of-bounds atomic bit-set operation. `__test_and_set_bit()` reads and modifies the machine word containing the selected bit. The missing upper-bound check permits that word to lie beyond `qpd->doorbell_bitmap`. This is an OOB kernel write rather than a validation-only issue, passive metadata mismatch, read-only over-index, or UAF. The destination is relative to a fixed bitmap base, and the payload is constrained to setting one selected bit. Nevertheless, the index controls the offset, so the primitive is stronger than an ordinary invalid-pointer crash and is classified as `CLASS_A`. The demonstrated practical consequences are kernel memory corruption and potential denial of service. Local privilege escalation is technically plausible only if the reachable index range and allocator layout permit modification of an exploitation-relevant adjacent object. That stronger chain is not demonstrated. `ImportantCandidate=YES` because a concrete attacker-indexed kernel write must not be automatically reduced to a validation bug merely because the affected adjacent allocation is unknown. `AutoDowngradeAllowed=NO` because automatic closure would discard unresolved exploitation leverage from the established OOB write. Operationally, this is at least Actionable Moderate. Important severity depends on confirming below-host-admin reachability and identifying useful adjacent targets or repeatable corruption behavior. 5. One-sentence report phrase An unchecked AMD KFD doorbell restore ID reaches an atomic bitmap operation as an attacker-selected index, enabling an out-of-bounds kernel bit write that can crash the host and may provide stronger corruption leverage depending on the reachable offset and adjacent allocation layout. 6. Manual review recommendation MANUAL CHECK REQUIRED Trace the complete KFD CRIU ioctl path, determine the actual privilege checks and `restore_id` type, inspect the allocation size and placement of `doorbell_bitmap`, and identify which kernel objects can be reached by practical out-of-range indices. 7. Call-site confidence Call-site confidence: medium ====================================================================== UPSTREAM PATCH SUMMARY ====================================================================== Patch: drm/amdkfd: Check bounds on allocate_doorbell [ Upstream Commit: e0cecd04a4fc4c75e375b91cf53eab5ba2e4ab7f Upstream URL: https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=e0cecd04a4fc4c75e375b91cf53eab5ba2e4ab7f Changed files: drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.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=e0cecd04a4fc4c75e375b91cf53eab5ba2e4ab7f ====================================================================== DETAILED REPORT METHODOLOGY ====================================================================== The original Linux kernel CVE announcement for CVE-2026-97495 can be found here: https://lore.kernel.org/linux-cve-announce/?q=CVE-2026-97495 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:L/AC:L/PR:L/UI:N/S:U/C:L/I:L/A:H The Best / paranoid CVSS vector: AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H The Best / paranoid CVSS score: 7.8 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: unknown 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).