[LU-3653] exclusive open lock reference leak

• Type: Technical task
• Resolution: Fixed
• Priority: Critical
• Fix Version/s: Lustre 2.5.0
• Assignee: Jinshan Xiong (Inactive)
• Reporter: Oleg Drokin

Description

Doing unrelated testing with --enable-lu_ref I hit this failure twice already:

<4>[ 8263.047598] Lustre: DEBUG MARKER: == sanity test 208: Exclusive open == 14:
28:11 (1375036091)
<3>[ 8263.163127] LustreError: 15854:0:(lu_ref.c:105:lu_ref_print()) lu_ref: ffff
88005dcfff98 2 0 ldlm_lock_new:505
<3>[ 8263.164133] LustreError: 15854:0:(lu_ref.c:107:lu_ref_print())      link: h
andle ffff880096d0c340
<0>[ 8263.167064] LustreError: 15854:0:(lu_ref.c:149:lu_ref_fini()) ASSERTION( 0 
) failed: 
<0>[ 8263.167560] LustreError: 15854:0:(lu_ref.c:149:lu_ref_fini()) LBUG
<4>[ 8263.167850] Pid: 15854, comm: multiop
<4>[ 8263.168105] 
<4>[ 8263.168105] Call Trace:
<4>[ 8263.168500]  [<ffffffffa0e768a5>] libcfs_debug_dumpstack+0x55/0x80 [libcfs]
<4>[ 8263.168817]  [<ffffffffa0e76ea7>] lbug_with_loc+0x47/0xb0 [libcfs]
<4>[ 8263.169107]  [<ffffffffa0f62280>] lu_ref_fini+0xa0/0xf0 [obdclass]
<4>[ 8263.169394]  [<ffffffffa04a483a>] ldlm_lock_put+0x19a/0x560 [ptlrpc]
<4>[ 8263.169736]  [<ffffffffa04c33e2>] ldlm_cli_cancel_list+0xf2/0x3e0 [ptlrpc]
<4>[ 8263.170030]  [<ffffffffa04c511e>] ldlm_cli_cancel+0x14e/0x380 [ptlrpc]
<4>[ 8263.170314]  [<ffffffffa0c2a1d4>] ll_lease_close+0x1a4/0x210 [lustre]
<4>[ 8263.170611]  [<ffffffffa0c36115>] ll_file_ioctl+0x1dc5/0x3080 [lustre]
<4>[ 8263.170881]  [<ffffffff810464a4>] ? __do_page_fault+0x204/0x490
<4>[ 8263.171134]  [<ffffffff81195512>] vfs_ioctl+0x22/0xa0

Crashdump and modules: /exports/crashdumps/192.168.10.219-2013-07-28-14\:28\:15/
another crashdump: /export/crashdumps/192.168.10.224-2013-07-28-16\:00\:43
Source tag in my tree: master-20130728