On Fri, 4 Jul 2008 16:48:28 +0900 (JST), yamamoto@xxxxxxxxxxxxx (YAMAMOTO Takashi) wrote: > > Hi, Yamamoto-san. > > > > Thank you for your comment. > > > > On Fri, 4 Jul 2008 15:54:31 +0900 (JST), yamamoto@xxxxxxxxxxxxx (YAMAMOTO Takashi) wrote: > > > hi, > > > > > > > +/* > > > > + * uncharge all the entries that are charged to the group. > > > > + */ > > > > +void __swap_cgroup_force_empty(struct mem_cgroup *mem) > > > > +{ > > > > + struct swap_info_struct *p; > > > > + int type; > > > > + > > > > + spin_lock(&swap_lock); > > > > + for (type = swap_list.head; type >= 0; type = swap_info[type].next) { > > > > + p = swap_info + type; > > > > + > > > > + if ((p->flags & SWP_ACTIVE) == SWP_ACTIVE) { > > > > + unsigned int i = 0; > > > > + > > > > + spin_unlock(&swap_lock); > > > > > > what prevents the device from being swapoff'ed while you drop swap_lock? > > > > > Nothing. > > > > After searching the entry to be uncharged(find_next_to_unuse below), > > I recheck under swap_lock whether the entry is charged to the group. > > Even if the device is swapoff'ed, swap_off must have uncharged the entry, > > so I don't think it's needed anyway. > > > > > YAMAMOTO Takashi > > > > > > > + while ((i = find_next_to_unuse(p, i, mem)) != 0) { > > > > + spin_lock(&swap_lock); > > > > + if (p->swap_map[i] && p->memcg[i] == mem) > > Ah, I think it should be added !p->swap_map to check the device has not > > been swapoff'ed. > > find_next_to_unuse seems to have fragile assumptions and > can dereference p->swap_map as well. > You're right. Thank you for pointing it out! I'll consider more. Thanks, Daisuke Nishimura. _______________________________________________ Containers mailing list Containers@xxxxxxxxxxxxxxxxxxxxxxxxxx https://lists.linux-foundation.org/mailman/listinfo/containers