On Mon, Jan 09, 2017 at 03:32:04PM -0800, Paul E. McKenney wrote:
> We could move rcu_scheduler_starting() later, as long as there
> is no chance of preemption or context switch before it is invoked.
> Would that help in this case, or are we already context switching before
> acpi_os_map_cleanup() is invoked? (If we are already context switching,
> short-circuiting synchronize_rcu_expedited() would be a bug.)
Hmm, how about the below?
It would still happen before
/*
* The boot idle thread must execute schedule()
* at least once to get things moving:
*/
init_idle_bootup_task(current);
schedule_preempt_disabled();
in rest_init() and right after native_smp_prepare_cpus() which is where
we're splatting.
Lemme run it.
Even if it works, we would have to stress-test this seriously...
---
diff --git a/init/main.c b/init/main.c
index b0c9d6facef9..9be221cc87c3 100644
--- a/init/main.c
+++ b/init/main.c
@@ -385,7 +385,6 @@ static noinline void __ref rest_init(void)
{
int pid;
- rcu_scheduler_starting();
/*
* We need to spawn init first so that it obtains pid 1, however
* the init task will end up wanting to create kthreads, which, if
@@ -1019,6 +1018,8 @@ static noinline void __init kernel_init_freeable(void)
smp_prepare_cpus(setup_max_cpus);
+ rcu_scheduler_starting();
+
workqueue_init();
do_pre_smp_initcalls();
--
Regards/Gruss,
Boris.
Good mailing practices for 400: avoid top-posting and trim the reply.
