Waiman Long

A issue in the Linux kernel has been identified where the `clocksource verify choose cpus()` function is invoked with preemption disabled, leading to the acquisition of sleeping locks in atomic context. This occurs because `get random u32()` is called to obtain random numbers, which can acquire the `batched entropy 32` local lock and/or the `base crng.lock` spinlock. In PREEMPT RT kernels, these locks are sleeping locks and cannot be acquired in atomic context, resulting in a bug report.

A divide-by-0 panic can occur in the `padata mt helper()` function when `ps->chunk size` is 0. This happens when the `min chunk` in the passed-in `padata mt job` structure is 0. The issue is resolved by ensuring that `chunk size` is at least 1, regardless of the input parameters.

The issue is related to a NULL pointer dereference in the `pfn section valid()` function. This occurs due to a race condition with `section deactivate()` where `ms->usage` can be cleared. The addition of a `READ ONCE()` call around `ms->usage` is not sufficient to prevent the NULL pointer dereference, and a check of its value before dereferencing is necessary. There is no information provided about the estimated number of potentially affected devices or real-world incidents where this issue was exploited.