[LU-15832] race between dt_try_as_dir() and lod_striping_free_nolock()

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.16.0
• Severity: 3
• Assignee: Lai Siyao
• Reporter: Lai Siyao

Description

In lod_striping_free_nolock(), if object is directory, its striping is freed, but .ldo_index_ops is not cleared, if a subsequent dt_try_as_dir() is called on this object, the .ldo_index_ops will be called, while its striping is not initialized yet, which may cause crash.

MDT0000 crashed with the following LBUG:

LustreError: 18825:0:(mdt_xattr.c:421:mdt_dir_layout_shrink()) fs-MDT0000: [0x20002c3a3:0x47:0x0] migrate mdt index mismatch 0 != 4294967295
LustreError: 18825:0:(mdt_xattr.c:421:mdt_dir_layout_shrink()) fs-MDT0000: [0x20002c3a3:0x69:0x0] migrate mdt index mismatch 0 != 4294967295
LustreError: 18825:0:(mdt_xattr.c:421:mdt_dir_layout_shrink()) Skipped 26 previous similar messages
LustreError: 18832:0:(mdt_xattr.c:421:mdt_dir_layout_shrink()) fs-MDT0000: [0x20002c3a3:0xb3:0x0] migrate mdt index mismatch 0 != 4294967295
LustreError: 18832:0:(mdt_xattr.c:421:mdt_dir_layout_shrink()) Skipped 47 previous similar messages
LustreError: 18839:0:(mdd_dir.c:4521:mdd_migrate()) fs-MDD0000: 'chrne' migration was interrupted, run 'lfs migrate -m 0 -c 3 -H 2 j4675' to finish migration.
LustreError: 18828:0:(lod_object.c:383:lod_striped_lookup()) ASSERTION( lo->ldo_dir_stripe_count > 0 ) failed:
LustreError: 18828:0:(lod_object.c:383:lod_striped_lookup()) LBUG
Kernel panic - not syncing: LBUG
 CPU: 5 PID: 18828 Comm: mdt01_007  3.10.0-1160.49.1.el7_lustre.x86_64 #1
 Call Trace:
 dump_stack+0x19/0x1b
 panic+0xe8/0x21f
 lbug_with_loc+0x9b/0xa0 [libcfs]
 lod_striped_lookup+0x2ba/0x2f0 [lod]
 __mdd_lookup.isra.17+0x298/0x440 [mdd]
 mdd_lookup+0xaf/0x170 [mdd]
 mdt_getattr_name_lock+0x1353/0x21e0 [mdt]
 mdt_getattr_name+0xc4/0x2b0 [mdt]
 tgt_request_handle+0xaee/0x15f0 [ptlrpc]
 ptlrpc_server_handle_request+0x24b/0xab0 [ptlrpc]
 ptlrpc_main+0xb34/0x1470 [ptlrpc]
 kthread+0xd1/0xe0