Re: [PATCH 3/3] rcu/trace: Add name of the source for gp_seq to prevent confusion

2020-06-20 Thread Paul E. McKenney
On Sat, Jun 20, 2020 at 01:36:08AM -0400, Joel Fernandes wrote: > On Fri, Jun 19, 2020 at 10:40:01AM -0700, Paul E. McKenney wrote: > > On Thu, Jun 18, 2020 at 10:07:18PM -0400, Joel Fernandes wrote: > > > On Thu, Jun 18, 2020 at 09:36:41PM -0400, Joel Fernandes (Google) wrote: > > > [...] > > > >

Re: [PATCH 3/3] rcu/trace: Add name of the source for gp_seq to prevent confusion

2020-06-19 Thread Joel Fernandes
On Fri, Jun 19, 2020 at 10:40:01AM -0700, Paul E. McKenney wrote: > On Thu, Jun 18, 2020 at 10:07:18PM -0400, Joel Fernandes wrote: > > On Thu, Jun 18, 2020 at 09:36:41PM -0400, Joel Fernandes (Google) wrote: > > [...] > > > @@ -2019,7 +2019,7 @@ static int __noreturn rcu_gp_kthread(void *unused) >

Re: [PATCH 3/3] rcu/trace: Add name of the source for gp_seq to prevent confusion

2020-06-19 Thread Paul E. McKenney
On Thu, Jun 18, 2020 at 10:07:18PM -0400, Joel Fernandes wrote: > On Thu, Jun 18, 2020 at 09:36:41PM -0400, Joel Fernandes (Google) wrote: > [...] > > @@ -2019,7 +2019,7 @@ static int __noreturn rcu_gp_kthread(void *unused) > > cond_resched_tasks_rcu_qs(); > >

Re: [PATCH 3/3] rcu/trace: Add name of the source for gp_seq to prevent confusion

2020-06-18 Thread Joel Fernandes
On Thu, Jun 18, 2020 at 09:36:41PM -0400, Joel Fernandes (Google) wrote: [...] > @@ -2019,7 +2019,7 @@ static int __noreturn rcu_gp_kthread(void *unused) > cond_resched_tasks_rcu_qs(); > WRITE_ONCE(rcu_state.gp_activity, jiffies); >

[PATCH 3/3] rcu/trace: Add name of the source for gp_seq to prevent confusion

2020-06-18 Thread Joel Fernandes (Google)
The gp_seq value can come from either of rdp, rsp or rnp. This makes the rcu_grace_period tracepoint extremely confusing since once if left constantly applying effort to reason about what a gp_seq means. Only the rsp is the global source of truth (most accurate GP info). The rnp can be off by ~1 a