sched/core: Validate rq_clock*() serialization (cebde6d6) · Commits · 戴 / test

rq->clock{,_task} are serialized by rq->lock, verify this. One immediate fail is the usage in scale_rt_capability, so 'annotate' that for now, there's more 'funny' there. Maybe change rq->lock into a raw_seqlock_t? (Only 32-bit is affected) Signed-off-by:

Peter Zijlstra (Intel) <peterz@infradead.org> Link: http://lkml.kernel.org/r/20150105103554.361872747@infradead.org Cc: Linus Torvalds <torvalds@linux-foundation.org> Cc: umgwanakikbuti@gmail.com Signed-off-by:

Ingo Molnar <mingo@kernel.org>

kernel/sched/fair.c

+1 −1

Original line number	Diff line number	Diff line
		@@ -5948,8 +5948,8 @@ static unsigned long scale_rt_capacity(int cpu)
		*/
		age_stamp = ACCESS_ONCE(rq->age_stamp);
		avg = ACCESS_ONCE(rq->rt_avg);
		delta = __rq_clock_broken(rq) - age_stamp;

		delta = rq_clock(rq) - age_stamp;
		if (unlikely(delta < 0))
		delta = 0;

kernel/sched/sched.h

+7 −0

Original line number	Diff line number	Diff line
		@@ -687,13 +687,20 @@ DECLARE_PER_CPU_SHARED_ALIGNED(struct rq, runqueues);
		#define cpu_curr(cpu) (cpu_rq(cpu)->curr)
		#define raw_rq() raw_cpu_ptr(&runqueues)

		static inline u64 __rq_clock_broken(struct rq *rq)
		{
		return ACCESS_ONCE(rq->clock);
		}

		static inline u64 rq_clock(struct rq *rq)
		{
		lockdep_assert_held(&rq->lock);
		return rq->clock;
		}

		static inline u64 rq_clock_task(struct rq *rq)
		{
		lockdep_assert_held(&rq->lock);
		return rq->clock_task;
		}

Admin message