Improper Handling of Scheduler Work Queue in Linux Kernel Affects AMD GPUs

CVE-2023-53351

What is CVE-2023-53351?

This vulnerability in the Linux kernel arises due to improper management of the scheduler work queue, which can lead to GPU faults during reset operations on AMD GPUs. When the GPU undergoes a reset test, it sometimes triggers an unexpected fault that can result in a NULL pointer dereference. This situation occurs if the interrupt service routine is executed before the GPU's scheduler is fully initialized, leading to potential system instability during graphics operations. The recommended solution is to utilize the timeout_wq field to ensure that uninitialized schedulers do not cause faults, thereby maintaining system reliability.

References