On Thu, Aug 30, 2018 at 10:35:09AM -0700, Paul E. McKenney wrote:
> On Thu, Aug 30, 2018 at 12:44:44PM -0400, Steven Rostedt wrote:
> > On Wed, 29 Aug 2018 20:23:15 -0700
> > "Paul E. McKenney" <paul...@linux.vnet.ibm.com> wrote:
> >
> > > > > Glad you like it! Does it actually work for you? ;-)
> > > >
> > > > Oh, you want me to actually test it too? ;-)
> > >
> > > ;-) ;-) ;-)
> > >
> > > > I'll try to add that in my todo list tomorrow.
> > >
> > > Much appreciated!
> >
> > I reverted the change that prevents calling call_srcu() early:
> >
> > I checked out v4.19-rc1 and applied these three patches, then did:
> >
> > git show f8a79d5c7ef47c62d97a30e16064caf2ef91f648 | patch -p1 -R
> >
> > But still triggered the following:
> >
> > WARNING: CPU: 0 PID: 0 at
> > /work/git/linux-trace.git/kernel/rcu/srcutree.c:242
> > check_init_srcu_struct+0x85/0x90
>
> Gah!!! I needed to have removed that WARN_ON_ONCE(), didn't I?
> In fact, I should have removed that once I started using workqueues,
> quite some time back.
As in the below patch.
Thanx, Paul
------------------------------------------------------------------------
diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
index 75571ff09c62..a8846ed7f352 100644
--- a/kernel/rcu/srcutree.c
+++ b/kernel/rcu/srcutree.c
@@ -239,7 +239,6 @@ static void check_init_srcu_struct(struct srcu_struct *sp)
{
unsigned long flags;
- WARN_ON_ONCE(rcu_scheduler_active == RCU_SCHEDULER_INIT);
/* The smp_load_acquire() pairs with the smp_store_release(). */
if (!rcu_seq_state(smp_load_acquire(&sp->srcu_gp_seq_needed))) /*^^^*/
return; /* Already initialized. */
