On Fri, Jun 17, 2022 at 07:14:45AM +0000, chenjun (AM) wrote: > 在 2022/6/17 13:59, Sebastian Andrzej Siewior 写道: > > On 2022-06-16 12:41:27 [-0700], Andrew Morton wrote: > >> ------------------------------------------------------ > >> From: Chen Jun <chenjun102@xxxxxxxxxx> > >> Subject: lockdep: make PROVE_RAW_LOCK_NESTING depend on PREEMPT_RT > >> Date: Thu, 16 Jun 2022 02:29:11 +0000 > >> > >> CONFIG_PROVE_RAW_LOCK_NESTING is used to check two things. > >> 1. calling spin_lock in hardirq. > >> 2. raw_spinlock - spinlock nesting. > >> The error messages like below: > >> [ 13.485711] [ BUG: Invalid wait context ] > >> > >> However there is no problem on non-PREEMPT_RT kernel. The messages may be > >> confusing and disturbing. > >> > >> Therefore, do not check raw_spinlock - spinlock nesting > >> on non-PREEMPT_RT kernel. > > > > This is not correct. The annotation is used to detect problems on > > non-PREEMPT_RT kernels which occur on PREEMPT_RT. This is documented in > > Kconfig. Letting this option depend on PREEMPT_RT renders it pointless > > because warning are spit on PREEMPT_RT even without that option enabled. > > > > Please drop that patch. > > Maybe we can add some obvious description that the error messages have > no effect on non-PREEMPT_RT kernel? They still need to be fixed though....
