On Thu, May 12, 2011 at 10:53 AM, KAMEZAWA Hiroyuki <kamezawa.hiroyu@xxxxxxxxxxxxxx> wrote: > On Thu, 12 May 2011 10:30:45 +0900 > Minchan Kim <minchan.kim@xxxxxxxxx> wrote: > >> Hi Kame, >> >> On Thu, May 12, 2011 at 9:52 AM, KAMEZAWA Hiroyuki >> <kamezawa.hiroyu@xxxxxxxxxxxxxx> wrote: >> > On Tue, 10 May 2011 17:15:01 +0900 (JST) >> > KOSAKI Motohiro <kosaki.motohiro@xxxxxxxxxxxxxx> wrote: >> > >> >> This patch introduces do_each_thread_reverse() and >> >> select_bad_process() uses it. The benefits are two, >> >> 1) oom-killer can kill younger process than older if >> >> they have a same oom score. Usually younger process >> >> is less important. 2) younger task often have PF_EXITING >> >> because shell script makes a lot of short lived processes. >> >> Reverse order search can detect it faster. >> >> >> >> Reported-by: CAI Qian <caiqian@xxxxxxxxxx> >> >> Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@xxxxxxxxxxxxxx> >> > >> > IIUC, for_each_thread() can be called under rcu_read_lock() but >> > for_each_thread_reverse() must be under tasklist_lock. >> >> Just out of curiosity. >> You mentioned it when I sent forkbomb killer patch. :) >> From at that time, I can't understand why we need holding >> tasklist_lock not rcu_read_lock. Sorry for the dumb question. >> >> At present, it seems that someone uses tasklist_lock and others uses >> rcu_read_lock. But I can't find any rule for that. >> > > for_each_list_rcu() makes use of RCU list's characteristics and allows > walk a list under rcu_read_lock() without taking any atomic locks. > > list_del() of RCU list works as folllowing. > > == > Â1) assume ÂA, B, C, are linked in the list. > Â Â Â Â(head)<->(A) <-> (B) Â<-> (C) > > Â2) remove B. > Â Â Â Â(head)<->(A) <-> (C) > Â Â Â Â Â Â Â Â Â Â Â Â/ > Â Â Â Â Â Â Â Â Â Â (B) > > ÂBecause (B)'s next points to (C) even after (B) is removed, (B)->next > Âpoints to the alive object. Even if (C) is removed at the same time, > Â(C) is not freed until rcu glace period and (C)'s next points to (head) > > Then, for_each_list_rcu() can work well under rcu_read_lock(), it will visit > only alive objects (but may not be valid.) > > == > > please see include/linux/rculist.h and check list_add_rcu() ;) > > As above implies, (B)->prev pointer is invalid pointer after list_del(). > So, there will be race with list modification and for_each_list_reverse under > rcu_read__lock() > > So, when you need to take atomic lock (as tasklist lock is) is... > > Â1) You can't check 'entry' is valid or not... > Â ÂIn above for_each_list_rcu(), you may visit an object which is under removing. > Â ÂYou need some flag or check to see the object is valid or not. > > Â2) you want to use list_for_each_safe(). > Â ÂYou can't do list_del() an object which is under removing... > > Â3) You want to walk the list in reverse. > > Â3) Some other reasons. For example, you'll access an object pointed by the > Â Â'entry' and the object is not rcu safe. > > make sense ? Yes. Thanks, Kame. It seems It is caused by prev poisoning of list_del_rcu. If we remove it, isn't it possible to traverse reverse without atomic lock? > > Thanks, > -Kame > > >> Could you elaborate it, please? >> Doesn't it need document about it? >> >> -- >> Kind regards, >> Minchan Kim >> >> -- >> To unsubscribe, send a message with 'unsubscribe linux-mm' in >> the body to majordomo@xxxxxxxxxx ÂFor more info on Linux MM, >> see: http://www.linux-mm.org/ . >> Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/ >> Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a> >> > > -- Kind regards, Minchan Kim -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxxx For more info on Linux MM, see: http://www.linux-mm.org/ . Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/ Don't email: <a href