CVE-2026-68273
High · 7.8CVSS
7.8
EPSS
0.1%
Published
Aug 10, 2026
CISA due date
—
Cloudanix analysis
A bug in the Linux graphics driver for AMD video cards could let someone already on the machine cause crashes or gain higher access rights.
- Exploitation
- Local Privilege
- Attacker needs
- Local User
- Complexity
- Moderate
- Impact
- Full Compromise
- Component
- OS kernel
- Class
- Use After Free
- Remediation
- Upgrade
- Effort
- Medium
- Patch risk
- Low
What it actually does
The vulnerability permits use-after-free and NULL pointer dereference conditions in the AMD GPU driver's context pstate handling because amdgpu_ctx_init() omits the required adev->pm.stable_pstate_ctx_lock, while context fini incorrectly restores stale pstate values set via sysfs.
Why it matters to the business
Systems using AMD GPUs may experience kernel instability or local privilege escalation; low exploit prevalence and declining EPSS trend indicate limited immediate operational risk, but kernel-level flaws still warrant prompt patching to avoid potential downtime or compromise.
How an attack unfolds
A local attacker triggers repeated GPU context creation and destruction while racing against sysfs pstate modifications; the missing lock allows the stored context pointer to become invalid, leading to a use-after-free dereference during initialization or exit.
How to fix it
Apply the upstream fix from commit 1b5e413713c0a93bc1818394d0ce49aaad21bd27 or later equivalent kernel patches that enforce lock-protected pstate ownership transitions and save/restore state only at override points.
How to confirm the fix worked
Confirm the patched kernel source contains the simplified locking logic around stable_pstate_ctx_lock; test by exercising sysfs pstate overrides concurrently with context creation and verify absence of crashes or incorrect state restoration.
Typical attack path
- Obtain local unprivileged access on an AMD GPU system
- Rapidly allocate and free GPU contexts via DRM ioctls
- Concurrently modify pstate via exposed sysfs interfaces
- Trigger the unlocked context initialization path to dereference freed memory
- Escalate privileges or cause kernel panic from the resulting UAF
How to detect it
Review kernel logs for amdgpu-related oops, NULL dereference, or use-after-free warnings involving ctx_init or stable_pstate_ctx_lock; correlate with unusual sysfs pstate activity preceding crashes.
What your security team should do
- Prioritize kernel updates on all AMD GPU hosts within standard change windows given the low EPSS and absence from KEV
- Restrict unprivileged local access where feasible until patches are deployed
- Include amdgpu context and pstate sysfs paths in kernel crash monitoring
The record
In the Linux kernel, the following vulnerability has been resolved: drm/amdgpu: Fix context pstate override handling There are several problems in the context pstate handling code. The most serious ones are potential use-after-free and NULL pointer dereferences at context initialization time. Both are due amdgpu_ctx_init() not holding the adev->pm.stable_pstate_ctx_lock, which is otherwise used from both sysfs and the context code itself for modifying and clearing the stored context pointer. Second issue is that context fini can trample over the pstate configuration set via sysfs. This is due the restore state (ctx->stable_pstate) being saved at context init time, and not if, or when the context actually changes the pstate. As the context exits it will therefore incorrectly restore to what was set before the sysfs override was requested. The simplest fix is to drastically simplify how the state is tracked, by clearly defining the points at which pstate ownership is taken and released, and to handle all transitions under the correct lock. Instead of at context init time, the previous state is saved only at the point the context overrides the current state, and is restored on context exit only if the context is still the owner of the current override state. (cherry picked from commit 1b5e413713c0a93bc1818394d0ce49aaad21bd27)
Affected packages
| Ecosystem | Package | Introduced | Fixed |
|---|
Related vulnerabilities
- CVE-2026-121230 Race condition in ANGLE in Google Chrome prior to 157.0.8103.38 allowed a remote attacker… CWE-362
- CVE-2026-121079 Race condition in V8 in Google Chrome prior to 157.0.8103.38 allowed a remote attacker to… CWE-362
- CVE-2026-121628 Bagisto before 2.3.16 and 2.4.x before 2.4.1 contains a race condition vulnerability in… CWE-362
- CVE-2026-121408 FreeRDP before 3.32.2 contains a use-after-free in the urbdrc USB redirection channel… CWE-362
- CVE-2026-58926 A Use-After-Free (UAF) vulnerability exists in the EVS HAL, allowing a privileged client… CWE-362
- CVE-2026-59042 In __pkvm_pages_to_ppages of mmu.c, there is a possible out-of-bounds write due to a race… CWE-362
- CVE-2026-59043 In arm_lpae_split_blk_unmap of io-pgtable-arm-common.c, there is a possible memory… CWE-362
- CVE-2026-59045 In multiple locations, there is a possible memory corruption due to a race condition. CWE-362