CVE-2026-99467 in Linuxinfo

Summary

by MITRE • 11/19/2026

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

media: marvell: mcam: stop DMA and cancel s_bh_work before freeing DMA buffers

In vmalloc mode the frame-completion IRQ queues mcam_frame_work() on cam->s_bh_work, which memcpy()s from the dma_bufs[] that
mcam_free_dma_bufs() frees with dma_free_coherent(). mccic_shutdown() frees those buffers without stopping the controller, so while streaming a late frame IRQ can re-arm the work after the buffers are gone, causing a use-after-free.

Stop the controller in mccic_shutdown() when streaming is still active (gated on an open fd, which holds a runtime-PM reference, so the device is powered), cancel s_bh_work in mcam_free_dma_bufs(), and move INIT_WORK() into mccic_register() before the device can be published.

If you want to get best quality of vulnerability data, you may have to visit VulDB.

Analysis

by VulDB Data Team • 11/19/2026

The vulnerability identified within the Linux kernel media subsystem specifically affects the Marvell camera sensor driver (mcam). This issue represents a classic use-after-free condition arising from improper synchronization between hardware state management, interrupt handling, and memory lifecycle operations. The root cause lies in the sequence of events during device shutdown or buffer deallocation when operating in vmalloc mode. In this configuration, frame-completion interrupts trigger an action handler that queues work to be executed in a bottom-half context via cam->s_bh_work. This queued work item is responsible for copying data from DMA buffers into user-space accessible memory structures using memcpy operations on the dma_bufs array.

The critical flaw occurs because mccic_shutdown() proceeds to free these DMA buffers by calling dma_free_coherent without first ensuring that the hardware controller has been stopped or that pending interrupt handlers have completed their execution. If a frame completion interrupt arrives after the shutdown process begins but before it completes, the system may re-arm the s_bh_work item even though the underlying memory resources (dma_bufs) have already been released back to the kernel allocator. Consequently, when the scheduled work eventually executes in the bottom-half context, it attempts to access freed memory through memcpy, leading to a use-after-free vulnerability. This can result in data corruption, kernel panics, or potentially arbitrary code execution if an attacker can influence the contents of the reclaimed memory and trigger the specific race condition.

From a technical perspective, this flaw is categorized under CWE-416: Use After Free. The operational impact extends beyond simple stability issues; it compromises the integrity of the media subsystem and exposes the system to potential privilege escalation vectors if exploited in conjunction with other vulnerabilities or through careful manipulation of memory allocation patterns. The vulnerability highlights a failure in enforcing strict ordering between resource deallocation and interrupt handler cleanup, which is essential for maintaining data consistency in asynchronous hardware-driven environments.

To mitigate this risk, several corrective measures have been implemented within the driver codebase. First, mccic_shutdown() has been modified to stop the controller when streaming is still active. This gating mechanism relies on checking for an open file descriptor, which holds a runtime-power management reference, ensuring that the device remains powered and responsive during shutdown but preventing new operations from starting once the process begins. Second, s_bh_work is explicitly cancelled in mcam_free_dma_bufs() before any memory is freed. This ensures that no pending work items will execute after the associated buffers are released. Finally, INIT_WORK has been moved to mccic_register(), which occurs early during device initialization and before the device can be published or accessed by user-space applications. This structural change guarantees that the work queue infrastructure is properly initialized in a controlled context prior to any potential race conditions arising from hardware interrupts.

These changes align with best practices for kernel driver development, particularly regarding resource management and interrupt handling safety. By ensuring that all pending asynchronous tasks are flushed or cancelled before memory release, and by halting hardware activity during shutdown sequences, the driver now adheres to safer concurrency models. This approach minimizes the window of vulnerability where race conditions could lead to invalid memory accesses. The fix reinforces the principle that device state must be fully quiesced before resources are reclaimed, thereby preventing use-after-free scenarios in high-throughput media processing pipelines.

Responsible

Linux

Reservation

09/25/2026

Disclosure

11/19/2026

Moderation

accepted

CPE

ready

EPSS

0.00168

KEV

no

Activities

very low

Sources

Do you want to use VulDB in your project?

Use the official API to access entries easily!