On 12/22/22 14:34, Peter Zijlstra wrote:
On Mon, Dec 05, 2022 at 11:48:32AM -0500, Waiman Long wrote:
Since commit 07ec77a1d4e8 ("sched: Allow task CPU affinity to be restricted on asymmetric systems"), the setting and clearing of user_cpus_ptr are done under pi_lock for arm64 architecture. However, dup_user_cpus_ptr() accesses user_cpus_ptr without any lock protection. When racing with the clearing of user_cpus_ptr in __set_cpus_allowed_ptr_locked(), it can lead to user-after-free and double-free in arm64 kernel.
How? the task cannot be in migrate_enable() and fork() at the same time, no?
I believe a task A can call sched_setaffinity() to modify the cpu affinity of a different task, say B, which can be under fork() at the same time. So we need to use the pi_lock to synchronize the access of user_cpus_ptr to avoid the kind of race that can cause double-free.
Cheers, Longman