Hi Brian, Thanks for the report. I managed to reproduce the bug on the latest mmotm kernel using the script you attached, so it isn't ubuntu-specific: : kernel BUG at mm/memcontrol.c:2929! : invalid opcode: 0000 [#1] SMP : CPU: 0 PID: 4441 Comm: kworker/0:2 Not tainted 4.4.0-mm1+ #256 : Workqueue: cgroup_destroy css_killed_work_fn : task: ffff8800aaddd880 ti: ffff8800369b0000 task.ti: ffff8800369b0000 : RIP: 0010:[<ffffffff81220551>] [<ffffffff81220551>] memcg_offline_kmem+0xd1/0xe0 : RSP: 0018:ffff8800369b3b08 EFLAGS: 00010293 : RAX: ffff8800a9cba800 RBX: ffff8800ab1c7000 RCX: 0000000000000003 : RDX: ffff8800a9cba850 RSI: ffff8800ab1c7060 RDI: ffff8800ab1c1000 : RBP: ffff8800369b3b28 R08: 0000000000000001 R09: 0000000000000000 : R10: 0000000000000000 R11: 0000000000000000 R12: ffff8800ab1c5000 : R13: 0000000000000000 R14: ffff8800ab1c7650 R15: ffff8800ab1c7640 : FS: 0000000000000000(0000) GS:ffff88014ae00000(0000) knlGS:0000000000000000 : CS: 0010 DS: 0000 ES: 0000 CR0: 000000008005003b : CR2: 00007f3ad9d3b090 CR3: 0000000148d61000 CR4: 00000000000006f0 : Stack: : ffff8800369b3b28 ffff8800ab1c7640 ffff8800ab1c7640 ffff8800ab1c7000 : ffff8800369b3b88 ffffffff81220601 0000000000000001 0000000000000000 : 0000000000000002 ffff8800aaddd880 ffff8800369b3b88 ffff8800ab1c7090 : Call Trace: : [<ffffffff81220601>] mem_cgroup_css_offline+0xa1/0xc0 : [<ffffffff81124b5c>] css_killed_work_fn+0x5c/0x170 : [<ffffffff8109ea30>] process_one_work+0x200/0x560 : [<ffffffff8109e99f>] ? process_one_work+0x16f/0x560 : [<ffffffff810d2bb2>] ? __lock_acquire+0x1a2/0x440 : [<ffffffff8109f6d4>] ? worker_thread+0x204/0x530 : [<ffffffff8109f6c7>] ? worker_thread+0x1f7/0x530 : [<ffffffff8109f63e>] worker_thread+0x16e/0x530 : [<ffffffff816f2eb4>] ? __schedule+0x354/0x900 : [<ffffffff810b2e22>] ? default_wake_function+0x12/0x20 : [<ffffffff810ca356>] ? __wake_up_common+0x56/0x90 : [<ffffffff8109f4d0>] ? maybe_create_worker+0x110/0x110 : [<ffffffff816f3567>] ? schedule+0x47/0xc0 : [<ffffffff8109f4d0>] ? maybe_create_worker+0x110/0x110 : [<ffffffff810a4ac9>] kthread+0xe9/0x110 : [<ffffffff810af22e>] ? schedule_tail+0x1e/0xd0 : [<ffffffff810a49e0>] ? __init_kthread_worker+0x70/0x70 : [<ffffffff816f92cf>] ret_from_fork+0x3f/0x70 : [<ffffffff810a49e0>] ? __init_kthread_worker+0x70/0x70 >From first glance, it looks like the bug was triggered, because mem_cgroup_css_offline was run for a child cgroup earlier than for its parent. This couldn't happen for sure before the cgroup was switched to percpu_ref, because cgroup_destroy_wq has always had max_active == 1. Now, however, it looks like this is perfectly possible for css_killed_ref_fn is called from an rcu callback - see kill_css -> percpu_ref_kill_and_confirm. This breaks kmemcg assumptions. I'll take a look what can be done about that. Thanks, Vladimir -- 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>