CVE-2025-37854

In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> drm/amdkfd: Fix mode1 reset crash issue<br /> <br /> If HW scheduler hangs and mode1 reset is used to recover GPU, KFD signal<br /> user space to abort the processes. After process abort exit, user queues<br /> still use the GPU to access system memory before h/w is reset while KFD<br /> cleanup worker free system memory and free VRAM.<br /> <br /> There is use-after-free race bug that KFD allocate and reuse the freed<br /> system memory, and user queue write to the same system memory to corrupt<br /> the data structure and cause driver crash.<br /> <br /> To fix this race, KFD cleanup worker terminate user queues, then flush<br /> reset_domain wq to wait for any GPU ongoing reset complete, and then<br /> free outstanding BOs.