Re: [PATCH v2] psi: fix race between psi_trigger_create and psimon

[Suren Baghdasaryan]

On Mon, May 17, 2021 at 7:16 PM Zhaoyang Huang <huangzhaoyang@xxxxxxxxx> wrote:
>
> resend for adding other maintainers in cc
>
> On Tue, May 18, 2021 at 10:11 AM Huangzhaoyang <huangzhaoyang@xxxxxxxxx> wrote:
> >
> > From: Zhaoyang Huang <zhaoyang.huang@xxxxxxxxxx>
> >
> > Race detected between psimon_new and psimon_old as shown below, which
> > cause panic by accessing invalid psi_system->poll_wait->wait_queue_entry
> > and psi_system->poll_timer->entry->next. It is not necessary to reinit
> > resource of psi_system when psi_trigger_create.

Please clarify that this fixes a race between psi_trigger_create and
psi_trigger_destroy, when psi_trigger_destroy is destroying the last
trigger in a psi group while racing with psi_trigger_create.

> >
> > psi_trigger_create      psimon_new     psimon_old
> >  init_waitqueue_head                    finish_wait
> >                                           spin_lock(lock_old)
> >         spin_lock_init(lock_new)
> >  wake_up_process(psimon_new)
> >
> >                         finish_wait
> >                           spin_lock(lock_new)
> >                             list_del       list_del

Add:
Fixes: 461daba06bdc ("psi: eliminate kthread_worker from psi trigger
scheduling mechanism")

> >
> > Signed-off-by: ziwei.dai <ziwei.dai@xxxxxxxxxx>
> > Signed-off-by: ke.wang <ke.wang@xxxxxxxxxx>
> > Signed-off-by: Zhaoyang Huang <zhaoyang.huang@xxxxxxxxxx>
> > +++
> > v2: change del_timer_sync to del_timer in psi_trigger_destroy

Please see my reply to your original RFC:
https://lore.kernel.org/patchwork/patch/1429498/#1627196
You are still missing:
- removal of timer_setup() in psi_trigger_create()
- removal of an obsolete comment in psi_trigger_destroy()
- you are calling del_timer() from outside of trigger_lock protected
section, which leaves the following race:

psi_trigger_destroy
    mutex_lock(trigger_lock)
    mutex_unlock(trigger_lock)
<preempted>
                       psi_trigger_create
                       psi_group_change
                           psi_schedule_poll_work
                               mod_timer(poll_timer)
<resumes>
    del_timer(poll_timer)

This would lead to deletion of the new poll_timer created by
psi_trigger_create() and would stop periodic psi_poll_work(). This
results in psi trigger not firing when it should. In my suggestion
del_timer() was called under trigger_lock protection which would
prevent psi_trigger_create() from starting a new timer before
psi_trigger_destroy() can call del_timer().

> > +++
> > ---
> >  kernel/sched/psi.c | 7 +++++--
> >  1 file changed, 5 insertions(+), 2 deletions(-)
> >
> > diff --git a/kernel/sched/psi.c b/kernel/sched/psi.c
> > index cc25a3c..fe29022 100644
> > --- a/kernel/sched/psi.c
> > +++ b/kernel/sched/psi.c
> > @@ -182,6 +182,8 @@ struct psi_group psi_system = {
> >
> >  static void psi_avgs_work(struct work_struct *work);
> >
> > +static void poll_timer_fn(struct timer_list *t);
> > +
> >  static void group_init(struct psi_group *group)
> >  {
> >         int cpu;
> > @@ -201,6 +203,8 @@ static void group_init(struct psi_group *group)
> >         memset(group->polling_total, 0, sizeof(group->polling_total));
> >         group->polling_next_update = ULLONG_MAX;
> >         group->polling_until = 0;
> > +       init_waitqueue_head(&group->poll_wait);
> > +       timer_setup(&group->poll_timer, poll_timer_fn, 0);
> >         rcu_assign_pointer(group->poll_task, NULL);
> >  }
> >
> > @@ -1157,7 +1161,6 @@ struct psi_trigger *psi_trigger_create(struct psi_group *group,
> >                         return ERR_CAST(task);
> >                 }
> >                 atomic_set(&group->poll_wakeup, 0);
> > -               init_waitqueue_head(&group->poll_wait);
> >                 wake_up_process(task);
> >                 timer_setup(&group->poll_timer, poll_timer_fn, 0);
> >                 rcu_assign_pointer(group->poll_task, task);
> > @@ -1233,7 +1236,7 @@ static void psi_trigger_destroy(struct kref *ref)
> >                  * But it might have been already scheduled before
> >                  * that - deschedule it cleanly before destroying it.
> >                  */
> > -               del_timer_sync(&group->poll_timer);
> > +               del_timer(&group->poll_timer);
> >                 kthread_stop(task_to_destroy);
> >         }
> >         kfree(t);
> > --
> > 1.9.1
> >