Re: [PATCH v2] workqueue: Fix UAF report by KASAN in pwq_release_workfn()

[Tejun Heo]

On Tue, Sep 19, 2023 at 07:42:39AM -1000, Tejun Heo wrote:
> Hello,
> 
> The patch looks fine but
> 
> On Tue, Sep 19, 2023 at 04:19:58PM +0800, Zqiang wrote:
> > @@ -4743,6 +4743,8 @@ struct workqueue_struct *alloc_workqueue(const char *fmt,
> >  	wq_free_lockdep(wq);
> >  err_free_wq:
> >  	free_workqueue_attrs(wq->unbound_attrs);
> 
> Can you please add a comment here explaining why the flushing is necessary?
> 
> > +	if (wq->flags & WQ_UNBOUND)
> > +		kthread_flush_worker(pwq_release_worker);

Hmm... also, wouldn't it be clearer to flush from alloc_and_link_pwqs()
right after apply_workqueue_attrs() failed? That's the only case that
requires flushing, right?

Thanks.

-- 
tejun