On Mon, Sep 08, 2025 at 01:19:44PM +0000, Zqiang wrote:
> >
> > On Mon, Sep 08, 2025 at 08:31:55AM +0800, Zqiang wrote:
> >
> > >
> > > Currently, the srcu_gp_start_if_needed() is always be invoked in
> > > preempt disable's critical section, this commit therefore remove
> > > redundant preempt_disable/enable() in srcu_gp_start_if_needed().
> > >
> > > Fixes: 65b4a59557f6 ("srcu: Make Tiny SRCU explicitly disable
> > > preemption")
> > > Signed-off-by: Zqiang <[email protected]>
> > >
> > Looks good, but what would be a good way to make this code defend itself
> > against being invoked from someplace else that did have preemption
> > enabled? Especially given that the Tree SRCU version of this function
> > does get invoked with preemption enabled?
>
> ok, maybe we can add lockdep_assert_preemption_disabled() in
> the srcu_gp_start_if_needed() ?
That sounds like a good idea to me!
Thanx, Paul
> Thanks
> Zqiang
>
>
> >
> > Thanx, Paul
> >
> > >
> > > ---
> > > kernel/rcu/srcutiny.c | 3 ---
> > > 1 file changed, 3 deletions(-)
> > >
> > > diff --git a/kernel/rcu/srcutiny.c b/kernel/rcu/srcutiny.c
> > > index b52ec45698e8..417bd0e4457c 100644
> > > --- a/kernel/rcu/srcutiny.c
> > > +++ b/kernel/rcu/srcutiny.c
> > > @@ -181,10 +181,8 @@ static void srcu_gp_start_if_needed(struct
> > > srcu_struct *ssp)
> > > {
> > > unsigned long cookie;
> > >
> > > - preempt_disable(); // Needed for PREEMPT_LAZY
> > > cookie = get_state_synchronize_srcu(ssp);
> > > if (ULONG_CMP_GE(READ_ONCE(ssp->srcu_idx_max), cookie)) {
> > > - preempt_enable();
> > > return;
> > > }
> > > WRITE_ONCE(ssp->srcu_idx_max, cookie);
> > > @@ -194,7 +192,6 @@ static void srcu_gp_start_if_needed(struct
> > > srcu_struct *ssp)
> > > else if (list_empty(&ssp->srcu_work.entry))
> > > list_add(&ssp->srcu_work.entry, &srcu_boot_list);
> > > }
> > > - preempt_enable();
> > > }
> > >
> > > /*
> > > --
> > > 2.48.1
> > >
> >
