On Fri, Jun 21, 2019 at 06:38:21PM +0200, Sebastian Andrzej Siewior wrote: > On 2019-06-20 14:18:26 [-0700], Paul E. McKenney wrote: > > > Example #1: > > > > > > 1. preempt_disable() > > > 2. local_bh_disable() > > > 3. preempt_enable() > > > 4. local_bh_enable() > > > > > > Example #2: > > > > > > 1. rcu_read_lock() > > > 2. local_irq_disable() > > > 3. rcu_read_unlock() > > > 4. local_irq_enable() > > > > > > Example #3: > > > > > > 1. preempt_disable() > > > 2. local_irq_disable() > > > 3. preempt_enable() > > > 4. local_irq_enable() > > > > OK for -rt, but as long as people can code those sequences without getting > > their wrists slapped, RCU needs to deal with it. So I cannot accept > > this in mainline at the current time. Yes, I will know when it is safe > > to accept it when rcutorture's virtual wrist gets slapped in mainline. > > All three examples are not symmetrical so if people use this mainline > then they should get their wrists slapped. Since RT trips over each one > of those I try to get rid of them if I notice something like that. > > In example #3 you would lose a scheduling event if TIF_NEED_RESCHED gets > set between step 1 and 2 (as local schedule requirement) because the > preempt_enable() would trigger schedule() which does not happen due to > IRQ-off. I have no objection to the outlawing of a number of these sequences in mainline, but am rather pointing out that until they really are outlawed and eliminated, rcutorture must continue to test them in mainline. Of course, an rcutorture running in -rt should avoid testing things that break -rt, including these sequences. Thanx, Paul
