Hello, Joseph. On Sat, Feb 24, 2018 at 09:45:49AM +0800, Joseph Qi wrote: > > IIRC, as long as the blkcg and the device are there, the blkgs aren't > > gonna be destroyed. So, if you have a ref to the blkcg through > > tryget, the blkg shouldn't go away. > > > > Maybe we have misunderstanding here. > > In this case, blkg doesn't go away as we have rcu protect, but > blkg_destroy() can be called, in which blkg_put() will put the last > refcnt and then schedule __blkg_release_rcu(). > > css refcnt can't prevent blkcg css from offlining, instead it is css > online_cnt. > > css_tryget() will only get a refcnt of blkcg css, but can't be > guaranteed to fail when css is confirmed to kill. Ah, you're right. I was thinking we only destroy blkgs from blkcg release path. Given that we primarily use blkcg refcnting to pin them, I believe that's what we should do - ie. only call pd_offline_fn() from blkcg_css_offline() path and do the rest of destruction from blkcg_css_free(). What do you think? Thanks. -- tejun -- To unsubscribe from this list: send the line "unsubscribe cgroups" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
