On Wed, Feb 12, 2025 at 8:52 AM Josh Poimboeuf <jpoimboe@xxxxxxxxxx> wrote:
>
> On Tue, Feb 11, 2025 at 02:24:37PM +0800, Yafang Shao wrote:
> > +++ b/kernel/livepatch/transition.c
> > @@ -491,9 +491,18 @@ void klp_try_complete_transition(void)
> > complete = false;
> > break;
> > }
> > +
> > + /* Avoid potential RCU stalls */
> > + if (need_resched()) {
> > + complete = false;
> > + break;
> > + }
> > }
> > read_unlock(&tasklist_lock);
> >
> > + /* The above operation might be expensive. */
> > + cond_resched();
> > +
>
> This is also nasty, yet another reason to use rcu_read_lock() if we can.
The RCU stalls still happen even if we use rcu_read_lock() as it is
still in the RCU read-side critical section.
>
> Also, with the new lazy preemption model, I believe cond_resched() is
> pretty much deprecated.
I'm not familiar with the newly introduced PREEMPT_LAZY, but it
appears to be a configuration option. Therefore, we still need this
cond_resched() for users who don't have PREEMPT_LAZY set as the
default.
>
> Also, for future patch sets, please also add lkml to cc. Thanks.
Sure.
--
Regards
Yafang
