[LU-2895] recovery-small 24a: osp_precreate_get_fid(): ASSERTION( lu_fid_diff(&d->opd_pre_used_fid, &d->opd_pre_last_created_fid) < 0 ) failed: next fid [0x0:0x1:0x0] last created fid [0x0:0x1:0x0]

• Type: Bug
• Resolution: Fixed
• Priority: Blocker
• Fix Version/s: Lustre 2.4.0
• Affects Version/s: Lustre 2.4.0
• Labels: LB
• Severity: 3
• Assignee: Niu Yawei (Inactive)
• Reporter: Maloo

Description

This issue was created by maloo for Li Wei <[e-mail]>

This issue relates to the following test suite run: https://maloo.whamcloud.com/test_sets/0c78ac7a-81ad-11e2-9f6b-52540035b04c.

The sub-test test_24a failed with the following error:

test failed to respond and timed out

Info required for matching: recovery-small 24a

First, the test failed with return code 1. Second, the MDS hit an assertion failure afterward:

12:39:11:Lustre: DEBUG MARKER: == recovery-small test 24a: fsync error (should return error) == 12:39:10 (1361997550)
12:39:36:Lustre: lustre-OST0000-osc-MDT0000: slow creates, last=[0x0:0x1:0x0], next=[0x0:0x1:0x0], reserved=0, syn_changes=0, syn_rpc_in_progress=0, status=0
12:39:36:LustreError: 18701:0:(osp_precreate.c:1164:osp_precreate_get_fid()) ASSERTION( lu_fid_diff(&d->opd_pre_used_fid, &d->opd_pre_last_created_fid) < 0 ) failed: next fid [0x0:0x1:0x0] last created fid [0x0:0x1:0x0]
12:39:36:LustreError: 18701:0:(osp_precreate.c:1164:osp_precreate_get_fid()) LBUG
12:39:36:Pid: 18701, comm: mdt00_000
12:39:36:
12:39:37:Call Trace:
12:39:37: [<ffffffffa06f1895>] libcfs_debug_dumpstack+0x55/0x80 [libcfs]
12:39:37: [<ffffffffa06f1e97>] lbug_with_loc+0x47/0xb0 [libcfs]
12:39:37: [<ffffffffa0ba4f63>] osp_precreate_get_fid+0x253/0x310 [osp]
12:39:37: [<ffffffffa07022e1>] ? libcfs_debug_msg+0x41/0x50 [libcfs]
12:39:37: [<ffffffffa0ba28cb>] osp_object_assign_fid+0x7b/0x160 [osp]
12:39:37: [<ffffffffa0ba30b6>] osp_object_create+0x406/0x890 [osp]
12:39:38: [<ffffffffa07022e1>] ? libcfs_debug_msg+0x41/0x50 [libcfs]
12:39:38: [<ffffffffa0b5e683>] lod_striping_create+0x193/0x320 [lod]
12:39:38: [<ffffffffa0b5e983>] lod_object_create+0x173/0x260 [lod]
12:39:38: [<ffffffffa0a1a580>] mdd_object_create_internal+0xe0/0x220 [mdd]
12:39:38: [<ffffffffa0a2d22c>] mdd_create+0x77c/0x11f0 [mdd]
12:39:38: [<ffffffffa07022e1>] ? libcfs_debug_msg+0x41/0x50 [libcfs]
12:39:38: [<ffffffffa0ac0ef9>] mdt_reint_open+0x12b9/0x1ab0 [mdt]
12:39:39: [<ffffffffa07022e1>] ? libcfs_debug_msg+0x41/0x50 [libcfs]
12:39:39: [<ffffffffa0aab7f1>] mdt_reint_rec+0x41/0xe0 [mdt]
12:39:39: [<ffffffffa0aa4e53>] mdt_reint_internal+0x4e3/0x7d0 [mdt]
12:39:39: [<ffffffffa0aa540d>] mdt_intent_reint+0x1ed/0x4f0 [mdt]
12:39:39: [<ffffffffa0aa0fde>] mdt_intent_policy+0x3ae/0x750 [mdt]
12:39:39: [<ffffffffa0fb0351>] ldlm_lock_enqueue+0x361/0x8d0 [ptlrpc]
12:39:39: [<ffffffffa0fd6447>] ldlm_handle_enqueue0+0x4f7/0x1080 [ptlrpc]
12:39:39: [<ffffffffa0aa14b6>] mdt_enqueue+0x46/0x110 [mdt]
12:39:39: [<ffffffffa0a96078>] mdt_handle_common+0x628/0x1620 [mdt]
12:39:39: [<ffffffffa0acd845>] mds_regular_handle+0x15/0x20 [mdt]
12:39:39: [<ffffffffa100802c>] ptlrpc_server_handle_request+0x41c/0xdf0 [ptlrpc]
12:39:39: [<ffffffffa06f25de>] ? cfs_timer_arm+0xe/0x10 [libcfs]
12:39:39: [<ffffffffa0fff759>] ? ptlrpc_wait_event+0xa9/0x290 [ptlrpc]
12:39:39: [<ffffffff8105fa40>] ? default_wake_function+0x0/0x20
12:39:39: [<ffffffffa1009576>] ptlrpc_main+0xb76/0x1870 [ptlrpc]
12:39:39: [<ffffffffa1008a00>] ? ptlrpc_main+0x0/0x1870 [ptlrpc]
12:39:39: [<ffffffff8100c0ca>] child_rip+0xa/0x20
12:39:39: [<ffffffffa1008a00>] ? ptlrpc_main+0x0/0x1870 [ptlrpc]
12:39:40: [<ffffffffa1008a00>] ? ptlrpc_main+0x0/0x1870 [ptlrpc]
12:39:40: [<ffffffff8100c0c0>] ? child_rip+0x0/0x20
12:39:40:
12:39:40:Kernel panic - not syncing: LBUG

Comments

[Peter Jones]

Landed for 2.4

[Alex Zhuravlev]

inability to setup FLD client or cleanup orphans should block any activity on this OSP, I'd think.

[Niu Yawei (Inactive)]

Alex, the opd_pre_status was intialized as -EAGAIN, and it could be cleared to 0 in osp_precreate_thread() -> osp_statfs_update(), and the following osp_precreate_cleanup_orphans() wasn't success somehow, then we'll end up in above state: uninitialized opd_pre_last_created_fid & opd_pre_used_fid, opd_pre_status == 0, and osp_precreate_reserve() is possbile timedout with rc == 0.

[Alex Zhuravlev]

why rc taken from is not enough ?

[Niu Yawei (Inactive)]

fix the LASSERT problem: http://review.whamcloud.com/5683

[Niu Yawei (Inactive)]

Looks the osp_precreate_reserve() should timedout:

12:39:36:Lustre: lustre-OST0000-osc-MDT0000: slow creates, last=[0x0:0x1:0x0], next=[0x0:0x1:0x0], reserved=0, syn_changes=0, syn_rpc_in_progress=0, status=0

but the osp_precreate_reserve() didn't return proper error for such case:

                if (cfs_time_aftereq(cfs_time_current(), expire))
                        break;

The mds debug log is missed, so I'm not sure the exact reason of precreation isn't done (even the seq hasn't been initialized), but at least, we should return proper error in such case.

[Peter Jones]

Niu

Could you please look into this one?

Thanks

Peter