Skip to main content

Linux Kernel CVE-2026-74721

| EUVDEUVD-2026-64408 HIGH
2026-08-22 Linux GHSA-grmx-x9wr-gmvh
7.8
CVSS 3.1 · Vendor: Linux
Share

Severity by source

Vendor (Linux) PRIMARY
7.8 HIGH
AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H
vuln.today AI
7.8 HIGH

AV:L confirmed as mmap requires local kernel access; PR:L for any user with device node access; S:U as kernel impact is contained within the local system.

3.1 AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H
4.0 AV:L/AC:L/AT:N/PR:L/UI:N/VC:H/VI:H/VA:H/SC:H/SI:H/SA:H

Primary rating from Vendor (Linux).

CVSS VectorVendor: Linux

CVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H
Attack Vector
Local
Attack Complexity
Low
Privileges Required
Low
User Interaction
None
Scope
Unchanged
Confidentiality
High
Integrity
High
Availability
High

Lifecycle Timeline

5
Analysis Generated
Aug 25, 2026 - 07:31 vuln.today
CVSS changed
Aug 25, 2026 - 06:22 NVD
7.8 (HIGH)
Patch available
Aug 22, 2026 - 16:01 EUVD
CVE Published
Aug 22, 2026 - 15:33 cve.org
UNKNOWN (no severity yet)
CVE Published
Aug 22, 2026 - 15:33 cve.org
HIGH 7.8

DescriptionCVE.org

In the Linux kernel, the following vulnerability has been resolved:

accel/amxdna: Fix page-insertion errors in amdxdna_insert_pages()

Two error paths in amdxdna_insert_pages() called vma->vm_ops->close(vma) before returning an error code to the caller. This is incorrect: amdxdna_gem_obj_mmap() registers an HMM interval notifier before calling amdxdna_insert_pages(), and on a hard error it jumps to hmm_unreg to undo that registration. Calling vm_ops->close() manually - which drops the shmem pages_pin_count and the GEM object reference that backs the VMA - before the mmap syscall has even returned causes those resources to be released while the VMA is still alive. The kernel VMA teardown will call vm_ops->close() a second time when the process later unmaps the range, producing a reference count underflow.

Replace both hard-error returns with a deferred-fault approach that keeps the VMA alive and retries page insertion through the HMM range-fault path.

AnalysisAI

Reference count underflow in the Linux kernel amdxdna (AMD XDNA NPU accelerator) driver allows a local low-privileged user to corrupt kernel memory via a double invocation of vm_ops->close() triggered by failed mmap page insertions. Two error paths in amdxdna_insert_pages() incorrectly release the shmem pages_pin_count and backing GEM object reference while the VMA remains live; the kernel's subsequent VMA teardown fires vm_ops->close() a second time, producing a reference count underflow with potential use-after-free consequences and high impact across confidentiality, integrity, and availability. …

Unlock full vulnerability intelligence

  • Risk assessment & exploitation conditions
  • Attack chain visualization
  • Remediation with exact patch versions
  • Threat intelligence from 22 sources
  • Personal watchlist & email alerts

Free forever · No credit card required

Attack ChainAIDerived

Hypothetical attack flow derived from CVE metadata

Recon
Gain local user access on Ryzen AI system
Delivery
Open amdxdna device node
Exploit
Call mmap on NPU memory region
Install
Induce page insertion error in amdxdna_insert_pages()
C2
Premature vm_ops->close() drops GEM object reference
Execute
VMA teardown fires second close() on unmap
Impact
Reference count underflow enables kernel memory corruption and privilege escalation

Vulnerability AssessmentAI

Exploitation Exploitation requires local access to a Linux system equipped with AMD XDNA NPU hardware (Ryzen AI series SoC), a local user account with permission to open the amdxdna device node and invoke mmap (any local user if device node permissions are default), and the ability to induce an error condition within amdxdna_insert_pages() - such as a page insertion failure driven by memory exhaustion or malformed parameters. … Additional conditions and limiting factors are described in the full assessment.
Risk Assessment Despite the CVSS 3.1 score of 7.8 (High) with AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H, real-world risk is substantially constrained by hardware specificity: exploitation requires a system equipped with an AMD XDNA NPU (Ryzen AI series), limiting the affected population to a narrow set of consumer and enterprise devices. … Full risk analysis with EPSS, KEV, and SSVC signal comparison available after sign-in.
Exploit Scenario A local user on a Ryzen AI-equipped system opens the amdxdna device node and issues a mmap syscall designed to trigger an error inside amdxdna_insert_pages() - for example, by exhausting available memory or supplying parameters that cause page insertion to fail. The premature vm_ops->close() call releases the GEM-backed shmem reference while the VMA remains live; when the attacker later unmaps the region or the process exits, the kernel's VMA teardown fires vm_ops->close() a second time, causing a reference count underflow that corrupts kernel memory and can be leveraged for kernel information disclosure or local privilege escalation to root. …
Remediation The primary remediation is upgrading to Linux kernel 7.1.9 or 7.2, which contain the upstream fix commits 8d51e0fd3e698919d2adeff71936377f0c0d4aa0 and 1501e4d07c6fee0d50531a0d1cb2be01a63e6e75 (available at https://git.kernel.org/stable/c/8d51e0fd3e698919d2adeff71936377f0c0d4aa0 and https://git.kernel.org/stable/c/1501e4d07c6fee0d50531a0d1cb2be01a63e6e75). … Detailed patch versions, workarounds, and compensating controls in full report.

Recommended ActionAI

Within 24 hours, audit your infrastructure to identify all systems running AMD Ryzen AI hardware with Linux kernels. …

Sign in for detailed remediation steps and compensating controls.

Threat intelligence, references, and detailed analysis are available after sign-in.

Share

CVE-2026-74721 vulnerability details – vuln.today

This site uses cookies essential for authentication and security. No tracking or analytics cookies are used. Privacy Policy