On Thu, Jun 18, 2020 at 03:27:21PM -0700, Paul E. McKenney wrote:
> On Thu, Jun 18, 2020 at 04:29:53PM -0400, Joel Fernandes (Google) wrote:
> > During acceleration of CB, the rsp's gp_seq is rcu_seq_snap'd. This is
> > the value used for acceleration - it is the value of gp_seq at which it
> > is safe the execute all callbacks in the callback list.
> >
> > The rdp's gp_seq is not very useful for this scenario. Make
> > rcu_grace_period report the rsp's gp_seq instead as it allows one to
> > reason about how the acceleration works.
>
> Good catch, but please instead trace the gp_seq_req local variable.
Oh sure, I will do that. Actually I was planning to, but was wondering if it
would be inconsistent with other trace_rcu_grace_period() invocations which
pass a real gp_seq (one that is as old as rcu_state.gp_seq). But agreed, this
is better since the acceleration uses it.
thanks,
- Joel
> Thanx, Paul
>
> > Signed-off-by: Joel Fernandes (Google) <joel@xxxxxxxxxxxxxxxxx>
> > ---
> > kernel/rcu/tree.c | 4 ++--
> > 1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> > index 81df1b837dd9d..c3bae7a83d792 100644
> > --- a/kernel/rcu/tree.c
> > +++ b/kernel/rcu/tree.c
> > @@ -1437,9 +1437,9 @@ static bool rcu_accelerate_cbs(struct rcu_node *rnp, struct rcu_data *rdp)
> >
> > /* Trace depending on how much we were able to accelerate. */
> > if (rcu_segcblist_restempty(&rdp->cblist, RCU_WAIT_TAIL))
> > - trace_rcu_grace_period(rcu_state.name, "rdp", rdp->gp_seq, TPS("AccWaitCB"));
> > + trace_rcu_grace_period(rcu_state.name, "rsp", rcu_state.gp_seq, TPS("AccWaitCB"));
> > else
> > - trace_rcu_grace_period(rcu_state.name, "rdp", rdp->gp_seq, TPS("AccReadyCB"));
> > + trace_rcu_grace_period(rcu_state.name, "rsp", rcu_state.gp_seq, TPS("AccReadyCB"));
> >
> > /* Count CBs for tracing. */
> > rcu_segcblist_countseq(&rdp->cblist, cbs, gps);
> > --
> > 2.27.0.111.gc72c7da667-goog
> >
