>> struct mem_cgroup *memcg = mem_cgroup_from_cont(cgrp); >> + int ret = -EAGAIN; >> >> if (val >= (1 << NR_MOVE_TYPE)) >> return -EINVAL; >> @@ -4123,10 +4130,13 @@ static int mem_cgroup_move_charge_write(struct cgroup *cgrp, >> * inconsistent. >> */ >> cgroup_lock(); >> + if (memcg->attach_in_progress) >> + goto out; >> memcg->move_charge_at_immigrate = val; >> + ret = 0; >> +out: >> cgroup_unlock(); >> - >> - return 0; >> + return ret; > > Unsure whether this is a good behavior. to be honest, I am not so sure myself. I kinda leaned towards this after some consideration, because the other solution I saw was basically busy waiting... > It's a bit nasty to fail for > internal temporary reasons like this. If it ever causes a problem, > the occurrences are likely to be far and between making it difficult > to debug. Can't you determine to immigrate or not in ->can_attach(), > record whether to do that or not on the css, and finish it in > ->attach() according to that. There's no need to consult the config > multiple times. > Well, yeah... that is an option too, and indeed better. Good call. I will send it again soon -- 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
