On Thu, Nov 17, 2022 at 03:20:25PM +0100, Frederic Weisbecker wrote:
> On Thu, Nov 17, 2022 at 07:20:50PM +0800, Zqiang wrote:
> > Before SRCU_SIZE_WAIT_CALL srcu_size_state is reached, the srcu
> > callback only insert to boot-CPU srcu_data structure's->srcu_cblist
> > and other CPU srcu_data structure's->srcu_cblist is always empty. so
> > before SRCU_SIZE_WAIT_CALL is reached, need to correctly check boot-CPU
> > pend cbs in srcu_might_be_idle().
> >
> > Signed-off-by: Zqiang <qiang1.zhang@xxxxxxxxx>
> > ---
> > kernel/rcu/srcutree.c | 7 +++++--
> > 1 file changed, 5 insertions(+), 2 deletions(-)
> >
> > diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
> > index 6af031200580..6d9af9901765 100644
> > --- a/kernel/rcu/srcutree.c
> > +++ b/kernel/rcu/srcutree.c
> > @@ -1098,8 +1098,11 @@ static bool srcu_might_be_idle(struct srcu_struct *ssp)
> > unsigned long tlast;
> >
> > check_init_srcu_struct(ssp);
> > - /* If the local srcu_data structure has callbacks, not idle. */
> > - sdp = raw_cpu_ptr(ssp->sda);
> > + /* If the boot CPU or local srcu_data structure has callbacks, not idle. */
> > + if (smp_load_acquire(&ssp->srcu_size_state) < SRCU_SIZE_WAIT_CALL)
> > + sdp = per_cpu_ptr(ssp->sda, get_boot_cpu_id());
> > + else
> > + sdp = raw_cpu_ptr(ssp->sda);
>
> While at it if someone is interested in documenting/commenting on the meaning of
> all these SRCU_SIZE_* things, it would be much appreciated!
>
Due to a conflict understanding to the code, once hesitate to jump in.
But anyway, just bold to move ahead.
I have send a series "[PATCH 0/3] srcu: shrink the num of
srcu_size_state", which opens the discussion. (Have Cced you and Zqiang)
Please have a look.
Thanks,
Pingfan
> Thanks.
>
> > spin_lock_irqsave_rcu_node(sdp, flags);
> > if (rcu_segcblist_pend_cbs(&sdp->srcu_cblist)) {
> > spin_unlock_irqrestore_rcu_node(sdp, flags);
> > --
> > 2.25.1
> >
