xfs ロックアップが原因でシステムがハングアップする

Issue

• システムがハングアップした状態になりましたが、システムログにはメッセージが出力されません。

• vmcore では、D 状態の複数のプロセスについて、以下のログが確認できます。

  [179039.139457] INFO: task varnishd:2941 blocked for more than 120 seconds.
  [179039.139477] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
  [179039.139500] varnishd        D 0000000000000000     0  2941   2586 0x00000080
  [179039.139505]  ffff880078177b58 0000000000000086 ffff880078163ec0 ffff880078177fd8
  [179039.139506]  ffff880078177fd8 ffff880078177fd8 ffff880078163ec0 ffff880079f2dc00
  [179039.139507]  ffff880053ffeac8 ffff880079f2ddc0 00000000000009cc 0000000000000000
  [179039.139508] Call Trace:
  [179039.139510]  [<ffffffff8168b879>] schedule+0x29/0x70
  [179039.139522]  [<ffffffffa027d672>] xlog_grant_head_wait+0xb2/0x190 [xfs]
  [179039.139532]  [<ffffffffa027d7ee>] xlog_grant_head_check+0x9e/0x110 [xfs]
  [179039.139542]  [<ffffffffa02812e9>] xfs_log_reserve+0xd9/0x1d0 [xfs]
  [179039.139551]  [<ffffffffa027b84d>] xfs_trans_reserve+0x18d/0x200 [xfs]
  [179039.139562]  [<ffffffffa026c226>] xfs_vn_update_time+0x56/0x190 [xfs]
  [179039.139563]  [<ffffffff81219a75>] update_time+0x25/0xd0
  [179039.139564]  [<ffffffff81200d38>] ?__sb_start_write+0x58/0x110
  [179039.139565]  [<ffffffff81219d20>] file_update_time+0xa0/0xf0
  [179039.139575]  [<ffffffffa026150e>] xfs_filemap_page_mkwrite+0x4e/0x120 [xfs]
  [179039.139577]  [<ffffffff811ac3b4>] do_page_mkwrite+0x54/0xa0
  [179039.139578]  [<ffffffff811aef36>] do_wp_page+0x326/0x530
  [179039.139579]  [<ffffffff811b0f8c>] handle_mm_fault+0x72c/0xfe0
  [179039.139581]  [<ffffffff81691c94>] __do_page_fault+0x154/0x450
  [179039.139583]  [<ffffffff81691fc5>] do_page_fault+0x35/0x90
  [179039.139584]  [<ffffffff8168e288>] page_fault+0x28/0x30
  [187467.575335] audit: audit_backlog=321 > audit_backlog_limit=320
  [187467.575339] audit: audit_lost=1 audit_rate_limit=0 audit_backlog_limit=320
  [187467.575340] audit: backlog limit exceeded
  [187467.575439] audit: audit_backlog=321 > audit_backlog_limit=320
  [187467.575440] audit: audit_lost=2 audit_rate_limit=0 audit_backlog_limit=320
  [187467.575441] audit: backlog limit exceeded