On Wed, Sep 29, 2021 at 05:17:31PM +0200, Peter Zijlstra wrote:
> XXX I'm pretty sure I broke task-trace-rcu.
> -static noinstr void rcu_dynticks_eqs_enter(void)
> -{
> - int seq;
> -
> - /*
> - * CPUs seeing atomic_add_return() must see prior RCU read-side
> - * critical sections, and we also must force ordering with the
> - * next idle sojourn.
> - */
> - rcu_dynticks_task_trace_enter(); // Before ->dynticks update!
> - seq = rcu_dynticks_inc(1);
> - // RCU is no longer watching. Better be in extended quiescent state!
> - WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) && (seq & 0x1));
> -}
> -static noinstr void rcu_dynticks_eqs_exit(void)
> -{
> - int seq;
> -
> - /*
> - * CPUs seeing atomic_add_return() must see prior idle sojourns,
> - * and we also must force ordering with the next RCU read-side
> - * critical section.
> - */
> - seq = rcu_dynticks_inc(1);
> - // RCU is now watching. Better not be in an extended quiescent state!
> - rcu_dynticks_task_trace_exit(); // After ->dynticks update!
> - WARN_ON_ONCE(IS_ENABLED(CONFIG_RCU_EQS_DEBUG) && !(seq & 0x1));
> -}
So specifically rcu_dynticks_task_trace_{enter,exit}() are now orphaned.
After this patch, nothing calls them.
However, looking at this again, we've got:
__context_tracking_enter()
rcu_user_enter()
rcu_eqs_enter()
rcu_dynticks_eqs_enter()
rcu_dynticks_task_trace_enter()
rcu_dynticks_inc();
rcu_dynticks_task_enter();
ct_seq_user_enter()
atomic_add_return()
and on the other end:
__context_tracking_exit()
ct_seq_user_exit()
atomic_add_return()
rcu_user_exit()
rcu_esq_exit()
rcu_dynticks_task_exit()
rcu_dynticks_eqs_exit()
rcu_dynticks_inc()
rcu_dynticks_task_trace_exit()
And since we want to replace dynticks_inc() with ct_seq_*() the
rcu_dynticks_task_{enter,exit}() ought to be pulled before that..
Instead I orphaned rcu_dynticks_task_trace_{enter,exit}() which should
more or less stay where they are.
I seems to have confused the two :/
