warning in mem_cgroup_lruvec()

warning in mem_cgroup_lruvec()

[Matthew Wilcox]

Running xfstests against next-20201124 + the THP patchset
https://git.infradead.org/users/willy/pagecache.git/shortlog/refs/heads/next

4005 ------------[ cut here ]------------
4006 WARNING: CPU: 2 PID: 45 at include/linux/memcontrol.h:621 do_try_to_free_pages+0x40b/0x460
4006 Modules linked in: crc_t10dif crct10dif_generic crct10dif_common [last unloaded: scsi_debug]
4006 CPU: 2 PID: 45 Comm: khugepaged Kdump: loaded Tainted: G        W         5.10.0-rc5-next-20201124-00062-gfedc7682a240 #990
4006 Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.14.0-1 04/01/2014
4006 RIP: 0010:do_try_to_free_pages+0x40b/0x460
4006 Code: 00 00 48 89 c3 e9 31 ff ff ff 48 3b 89 e8 20 00 00 48 8d 91 60 20 00 00 0f 84 f0 fe ff ff 48 89 8a 88 00 00 00 e9 e4 fe ff ff <0f> 0b 48 8b 15 ec 90 05 01 e9 5b fe ff ff 0f 0b 0f b6 5d cb 83 e0
4006 RSP: 0018:ffff888003053b50 EFLAGS: 00010246
4006 RAX: ffffffff821878c0 RBX: ffffffff82189800 RCX: 0000000000000004
4006 RDX: 0000000000000000 RSI: ffffffff82187f00 RDI: ffffffff82189820
4006 RBP: ffff888003053b90 R08: 00000000000003e3 R09: 0000000000004d64
4006 R10: 0000000000000009 R11: 0000000000000000 R12: 0000000000000000
4006 R13: ffffffff821878c0 R14: 0000000000000000 R15: ffff888003053ba0
4006 FS:  0000000000000000(0000) GS:ffff88807d880000(0000) knlGS:0000000000000000
4006 CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
4006 CR2: 000055e23c731368 CR3: 0000000053cfe002 CR4: 0000000000370ea0
4006 Call Trace:
4006  try_to_free_pages+0xdc/0x1b0
4006  __alloc_pages_slowpath.constprop.0+0x2d2/0xd00
4006  __alloc_pages_nodemask+0x249/0x290
4006  khugepaged+0x2cd/0x28e0
4006  ? add_wait_queue_exclusive+0x80/0x80
4006  kthread+0x135/0x150
4006  ? collapse_pte_mapped_thp+0x4a0/0x4a0
4006  ? kthread_create_worker_on_cpu+0x40/0x40
4006  ret_from_fork+0x1f/0x30
4006 ---[ end trace 4da165ba10d2d88a ]---

I see a few of these with different callchains.  eg:

3749 RIP: 0010:prepare_kswapd_sleep+0x66/0xc0
3749  kswapd+0xe2/0x350
3749  ? add_wait_queue_exclusive+0x80/0x80
3749  kthread+0x135/0x150
3749  ? balance_pgdat+0x600/0x600
3749  ? kthread_create_worker_on_cpu+0x40/0x40
3749  ret_from_fork+0x1f/0x30

3749 RIP: 0010:balance_pgdat+0x46f/0x600
3749 RIP: 0010:workingset_eviction+0x184/0x250

but they're all:
        VM_WARN_ON_ONCE(!memcg);

which is clearly not actually warning once at all.  not even once per
inlined site.