Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB
CVE-2026-98522 Overview
CVE-2026-98522 is a Linux kernel vulnerability in the drm/imagination (PowerVR) GPU driver. The pvr_mmu_map_sgl() function failed to correctly propagate map errors back to its caller pvr_mmu_map(). As a result, pvr_mmu_map() continued execution after a failed mapping, leading to a null pointer dereference in pvr_mmu_op_context_unmap_curr_page(). The flaw triggers a kernel oops and system instability when exercised by a local user-space process issuing GPU memory mapping ioctls through the DRM subsystem.
Critical Impact
A local, unprivileged user with access to the PowerVR DRM device can trigger a kernel null pointer dereference, producing a denial-of-service condition on affected ARM64 systems using Imagination PowerVR GPUs.
Affected Products
- Linux kernel builds including the powervr (drm/imagination) driver prior to the fix commits
- ARM64 platforms using Imagination PowerVR GPUs (observed on Texas Instruments AM625 SK)
- Distributions shipping Linux 6.17.0 and earlier releases that contain the vulnerable pvr_mmu_map_sgl() path
Discovery Timeline
- 2026-10-18 - CVE-2026-98522 published to NVD
- 2026-10-19 - Last updated in NVD database
Technical Details for CVE-2026-98522
Vulnerability Analysis
The defect is a missing error propagation in the PowerVR GPU MMU mapping code path. When pvr_mmu_map_sgl() encounters a mapping failure while walking a scatter-gather list, it returns without surfacing the failure to the caller. The upstream caller pvr_mmu_map() then proceeds under the assumption that the mapping succeeded.
Subsequent cleanup paths dereference partially initialized page-table state inside pvr_mmu_op_context_unmap_curr_page(). The kernel trace shows a data abort at virtual address 0x8 on ARM64, with pgd=0 indicating no valid translation. The oops is reproducible via the deqp-vk Vulkan conformance test suite calling the DRM_IOCTL_PVR_VM_MAP ioctl.
Root Cause
The root cause is improper error handling [CWE-252, Unchecked Return Value class] inside pvr_mmu_map_sgl(). The function did not return a proper negative error code when the underlying map operation failed. Without a propagated error, pvr_mmu_map() skipped its failure branch and advanced state tracking as though the mapping had been installed. This left MMU operation context pointers in an inconsistent state, which the unmap cleanup routine then dereferenced.
Attack Vector
Triggering the condition requires local access to the DRM device node (typically /dev/dri/card0) and the ability to issue pvr_ioctl_vm_map calls. An attacker invokes the ioctl with inputs that induce a failure inside pvr_mmu_map_sgl(), for example by supplying a scatter-gather configuration the driver cannot map. The resulting null pointer dereference crashes the kernel task and can be used to deny service to the host. The crash occurs in kernel context and taints the kernel with the D (die) flag.
The vulnerability mechanism is a kernel-level null pointer dereference triggered through DRM ioctl misuse. See the upstream fix commits linked below for the exact code change.
Detection Methods for CVE-2026-98522
Indicators of Compromise
- Kernel oops messages referencing pvr_mmu_op_context_unmap_curr_page+0x6c/0x134 [powervr] in dmesg or /var/log/kern.log
- Unable to handle kernel NULL pointer dereference at virtual address 0000000000000008 entries originating from the powervr module
- Repeated process crashes from GPU workloads such as deqp-vk, Vulkan test harnesses, or Mesa-based clients on PowerVR hardware
- Kernel taint flag D appearing after GPU-related ioctl activity
Detection Strategies
- Monitor kernel ring buffer for stack traces that include pvr_mmu_map, pvr_vm_gpuva_map, and drm_gpuvm_sm_map symbol frames
- Correlate DRM_IOCTL_PVR_VM_MAP ioctl invocations from untrusted processes with subsequent kernel oops events
- Audit loaded kernel modules for the powervr driver on ARM64 fleet systems to scope exposure
Monitoring Recommendations
- Forward kern.log and journald kernel-facility entries to a centralized log platform for pattern matching on powervr crash signatures
- Alert on any kernel oops or panic events on production hosts, particularly those running GPU compute workloads
- Track uptime regressions on AM625-class and other PowerVR-equipped edge devices as an indirect indicator of exploitation attempts
How to Mitigate CVE-2026-98522
Immediate Actions Required
- Apply the upstream Linux kernel patches that fix pvr_mmu_map_sgl() error propagation and rebuild or update affected kernels
- Restrict access to /dev/dri/* device nodes to trusted users and GPU-rendering groups only
- Prioritize patching ARM64 edge and embedded devices running the powervr driver, including Texas Instruments AM62x platforms
Patch Information
The fix returns a proper error code from pvr_mmu_map_sgl() so that pvr_mmu_map() aborts cleanly on mapping failure. The upstream fix is distributed across the following stable commits:
- Kernel Git Commit 13c7ae1
- Kernel Git Commit 7b824c2
- Kernel Git Commit afe5b4d
- Kernel Git Commit d30ea27
Workarounds
- Unload the powervr kernel module on systems that do not require GPU acceleration using modprobe -r powervr
- Blacklist the powervr module via /etc/modprobe.d/ on hosts where GPU workloads are not needed
- Limit DRM device node access using filesystem permissions and udev rules until a patched kernel is deployed
# Configuration example: restrict DRM device access and blacklist vulnerable module
echo 'blacklist powervr' | sudo tee /etc/modprobe.d/disable-powervr.conf
sudo chmod 0660 /dev/dri/card0
sudo chown root:video /dev/dri/card0
Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB

