perf: Fix perf_pending_task() UaF

Vulnerability Description

In the Linux kernel, the following vulnerability has been resolved: perf: Fix perf_pending_task() UaF Per syzbot it is possible for perf_pending_task() to run after the event is free()'d. There are two related but distinct cases: - the task_work was already queued before destroying the event; - destroying the event itself queues the task_work. The first cannot be solved using task_work_cancel() since perf_release() itself might be called from a task_work (____fput), which means the current->task_works list is already empty and task_work_cancel() won't be able to find the perf_pending_task() entry. The simplest alternative is extending the perf_event lifetime to cover the task_work. The second is just silly, queueing a task_work while you know the event is going away makes no sense and is easily avoided by re-arranging how the event is marked STATE_DEAD and ensuring it goes through STATE_OFF on the way down.

References

• https://git.kernel.org/stable/c/8bffa95ac19ff27c8261904f89d36c7fcf215d59
• https://git.kernel.org/stable/c/78e1317a174edbfd1182599bf76c092a2877672c
• https://git.kernel.org/stable/c/517e6a301f34613bff24a8e35b5455884f2d83d8