On Mon, 7 Feb 2022 11:31:38 -0800 Suren Baghdasaryan wrote: > > Sorry for the delay. Are you concerned about the use of spin_lock() > inside percpu_rwsem_async_destroy() which would become a sleeping lock > in case of PREEMPT_RT? If so, we can use raw_spin_lock() when locking > destroy_list_lock. Please confirm. Thanks! Yes please replace spin lock with the raw version which can fit in more scenarios. Hillf
