Re: [PATCH v6] psi: fix race between psi_trigger_create/destroy

[Peter Zijlstra]

On Thu, Jun 10, 2021 at 09:31:33AM +0200, Peter Zijlstra wrote:
> On Fri, May 21, 2021 at 10:18:53AM -0400, Johannes Weiner wrote:
> > On Fri, May 21, 2021 at 10:05:54AM +0800, Huangzhaoyang wrote:
> > > From: Zhaoyang Huang <zhaoyang.huang@xxxxxxxxxx>
> > > 
> > > Race detected between psi_trigger_destroy/create as shown below, which
> > > cause panic by accessing invalid psi_system->poll_wait->wait_queue_entry
> > > and psi_system->poll_timer->entry->next. Under this modification, the
> > > race window is removed by initialising poll_wait and poll_timer in
> > > group_init which are executed only once at beginning.
> > > 
> > > psi_trigger_destroy                      psi_trigger_create
> > > mutex_lock(trigger_lock);
> > > rcu_assign_pointer(poll_task, NULL);
> > > mutex_unlock(trigger_lock);
> > > 					mutex_lock(trigger_lock);
> > > 					if (!rcu_access_pointer(group->poll_task)) {
> > > 
> > > 						timer_setup(poll_timer, poll_timer_fn, 0);
> > > 
> > > 						rcu_assign_pointer(poll_task, task);
> > > 					}
> > > 					mutex_unlock(trigger_lock);
> > > 
> > > synchronize_rcu();
> > > del_timer_sync(poll_timer); <-- poll_timer has been reinitialized by
> > > psi_trigger_create
> > > 
> > > So, trigger_lock/RCU correctly protects destruction of group->poll_task but
> > > misses this race affecting poll_timer and poll_wait.
> > > 
> > > 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>

Argh noticed the SoB chain is invalid. Please fix.

> > 
> > Acked-by: Johannes Weiner <hannes@xxxxxxxxxxx>
> 
> Thanks!