[LU-19202] nfsd processes stuck in trying to double lock the inode

• Type: Bug
• Resolution: Unresolved
• Priority: Medium
• Severity: 3
• Assignee: Zhenyu Xu
• Reporter: Zhenyu Xu

Description

nfsd mounted on Lustre could hang with following processes.

PID: 164009   TASK: ff4d87028f53c000  CPU: 45   COMMAND: "nfsd"
 #0 [ff6ac3f222de75c0] __schedule at ffffffffb97eb171
 #1 [ff6ac3f222de7618] schedule at ffffffffb97eb735
 #2 [ff6ac3f222de7628] rwsem_down_write_slowpath at ffffffffb8f551ed
 #3 [ff6ac3f222de76c0] vvp_inode_ops at ffffffffc19a6725 [lustre]
 #4 [ff6ac3f222de76d8] cl_object_inode_ops at ffffffffc137ab86 [obdclass]
 #5 [ff6ac3f222de7708] lov_conf_set at ffffffffc18b1b36 [lov]
 #6 [ff6ac3f222de7788] cl_conf_set at ffffffffc137a6d9 [obdclass]
 #7 [ff6ac3f222de77b0] cl_file_inode_init at ffffffffc199c7f8 [lustre]
 #8 [ff6ac3f222de7820] ll_update_inode at ffffffffc196c46f [lustre]
 #9 [ff6ac3f222de7858] ll_prep_inode at ffffffffc196eedf [lustre]
#10 [ff6ac3f222de7920] ll_revalidate_it_finish at ffffffffc193545a [lustre]
#11 [ff6ac3f222de7968] ll_inode_revalidate.constprop.67 at ffffffffc194b3e5 [lustre]
#12 [ff6ac3f222de79e8] ll_getattr_dentry at ffffffffc1958739 [lustre]
#13 [ff6ac3f222de7aa0] fill_pre_wcc at ffffffffc1a589f7 [nfsd]
#14 [ff6ac3f222de7b68] nfsd_setattr at ffffffffc1a502a9 [nfsd]
#15 [ff6ac3f222de7c08] nfs4_get_vfs_file at ffffffffc1a693f8 [nfsd]
#16 [ff6ac3f222de7cc8] nfsd4_process_open2 at ffffffffc1a6e26f [nfsd]
#17 [ff6ac3f222de7db8] nfsd4_open at ffffffffc1a5d102 [nfsd]
#18 [ff6ac3f222de7df8] nfsd4_proc_compound at ffffffffc1a5d4de [nfsd]
#19 [ff6ac3f222de7e58] nfsd_dispatch at ffffffffc1a48c5e [nfsd]
#20 [ff6ac3f222de7e80] svc_process_common at ffffffffc0c78e10 [sunrpc]
#21 [ff6ac3f222de7ed8] svc_process at ffffffffc0c79127 [sunrpc]
#22 [ff6ac3f222de7ef0] nfsd at ffffffffc1a48603 [nfsd]
#23 [ff6ac3f222de7f10] kthread at ffffffffb8f1808b
#24 [ff6ac3f222de7f50] ret_from_fork at ffffffffb980024f

PID: 164010   TASK: ff4d87028f64c000  CPU: 20   COMMAND: "nfsd"
 #0 [ff6ac3f222def6f0] __schedule at ffffffffb97eb171
 #1 [ff6ac3f222def748] schedule at ffffffffb97eb735
 #2 [ff6ac3f222def758] rwsem_down_write_slowpath at ffffffffb8f551ed
 #3 [ff6ac3f222def7f0] vvp_inode_ops at ffffffffc19a6725 [lustre]
 #4 [ff6ac3f222def808] cl_object_inode_ops at ffffffffc137ab86 [obdclass]
 #5 [ff6ac3f222def838] lov_conf_set at ffffffffc18b1b36 [lov]
 #6 [ff6ac3f222def8b8] cl_conf_set at ffffffffc137a6d9 [obdclass]
 #7 [ff6ac3f222def8e0] cl_file_inode_init at ffffffffc199c7f8 [lustre]
 #8 [ff6ac3f222def950] ll_update_inode at ffffffffc196c46f [lustre]
 #9 [ff6ac3f222def988] ll_prep_inode at ffffffffc196eedf [lustre]
#10 [ff6ac3f222defa50] ll_intent_file_open at ffffffffc1951a80 [lustre]
#11 [ff6ac3f222defac8] ll_file_open at ffffffffc1951df3 [lustre]
#12 [ff6ac3f222defb78] do_dentry_open at ffffffffb9157ac3
#13 [ff6ac3f222defba8] dentry_open at ffffffffb9157dec
#14 [ff6ac3f222defbc8] nfsd_open at ffffffffc1a4de68 [nfsd]
#15 [ff6ac3f222defc08] nfs4_get_vfs_file at ffffffffc1a6943d [nfsd]
#16 [ff6ac3f222defcc8] nfsd4_process_open2 at ffffffffc1a6e26f [nfsd]
#17 [ff6ac3f222defdb8] nfsd4_open at ffffffffc1a5d102 [nfsd]
#18 [ff6ac3f222defdf8] nfsd4_proc_compound at ffffffffc1a5d4de [nfsd]
#19 [ff6ac3f222defe58] nfsd_dispatch at ffffffffc1a48c5e [nfsd]
#20 [ff6ac3f222defe80] svc_process_common at ffffffffc0c78e10 [sunrpc]
#21 [ff6ac3f222defed8] svc_process at ffffffffc0c79127 [sunrpc]
#22 [ff6ac3f222defef0] nfsd at ffffffffc1a48603 [nfsd]
#23 [ff6ac3f222deff10] kthread at ffffffffb8f1808b
#24 [ff6ac3f222deff50] ret_from_fork at ffffffffb980024f

The nfsd could lock the inode (by fh_lock()/fh_lock_nested()) before calling vfs interface so that Lustre is unaware and tries to lock the inode again.