>>> But memcg_update_cache_sizes calls memcg_kmem_clear_activated on the >>> error path. >>> >> >> But memcg_kmem_mark_dead() checks the ACCOUNT flag not the ACCOUNTED flag. >> Am I missing something? >> > > Dang. You are right! Glauber, is there any reason why > memcg_kmem_mark_dead checks only KMEM_ACCOUNTED_ACTIVE rather than > KMEM_ACCOUNTED_MASK? > > This all is very confusing to say the least. > > Anyway, this all means that Li's first patch is correct. I am not sure I > like it though. I think that the refcount cleanup should be done as > close to where it has been taken as possible otherwise we will end up in > this "chase the nasty details" again and again. There are definitely two > bugs here. The one introduced by e4715f01 and the other one introduced > even earlier (I haven't checked that history yet). I think we should do > something like the 2 follow up patches but if you guys think that the smaller > patch from Li is more appropriate then I will not block it. > Or we can queue my patch for 3.9, and then see if we want to change the tear down process, and if yes we make the change for 3.10. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>