[LU-4445] lu_object.c:1199:lu_device_fini()) ASSERTION( t->ldt_device_nr > 0 )

• Type: Bug
• Resolution: Fixed
• Priority: Critical
• Fix Version/s: Lustre 2.6.0

Environment

I hit the following ASSERT() several times on client-side during LFSCK performance testing:

{code}
LustreError: 16887:0:(lu_object.c:1199:lu_device_fini()) ASSERTION( t->ldt_device_nr > 0 ) failed:
LustreError: 16887:0:(lu_object.c:1199:lu_device_fini()) LBUG
Pid: 16887, comm: umount

Call Trace:
 [<ffffffffa03fc895>] libcfs_debug_dumpstack+0x55/0x80 [libcfs]
 [<ffffffffa03fce97>] lbug_with_loc+0x47/0xb0 [libcfs]
 [<ffffffffa0589434>] lu_device_fini+0x84/0xc0 [obdclass]
 [<ffffffffa0b8e778>] lov_device_free+0x38/0x250 [lov]
 [<ffffffffa058d1fe>] lu_stack_fini+0x7e/0xc0 [obdclass]
 [<ffffffffa0593b8e>] cl_stack_fini+0xe/0x10 [obdclass]
 [<ffffffffa117ce8d>] cl_sb_fini+0x6d/0x190 [lustre]
 [<ffffffffa1140364>] client_common_put_super+0x54/0x9f0 [lustre]
 [<ffffffffa1140e16>] ll_put_super+0x116/0x500 [lustre]
 [<ffffffffa116e6e3>] ? ll_destroy_inode+0xc3/0x100 [lustre]
 [<ffffffff8119d03f>] ? destroy_inode+0x2f/0x60
 [<ffffffff8119d50c>] ? dispose_list+0xfc/0x120
 [<ffffffff8119d906>] ? invalidate_inodes+0xf6/0x190
 [<ffffffff8118366b>] generic_shutdown_super+0x5b/0xe0
 [<ffffffff81183756>] kill_anon_super+0x16/0x60
 [<ffffffffa057f2fa>] lustre_kill_super+0x4a/0x60 [obdclass]
 [<ffffffff81183ef7>] deactivate_super+0x57/0x80
 [<ffffffff811a21ef>] mntput_no_expire+0xbf/0x110
 [<ffffffff811a2c5b>] sys_umount+0x7b/0x3a0
 [<ffffffff8100b072>] system_call_fastpath+0x16/0x1b

Kernel panic - not syncing: LBUG
{code}

During the test, I mount multiple logic clients on the same node, and create in parallel, and then umount all of them after the create. LFSCK patches do not touch client-side code. So I do not know it is LFSCK patches caused the issue, but it is blocking the LFSCK performance test.

Jay, would you please give some look when you have time? Thanks!

• Severity: 3
• Assignee: nasf (Inactive)
• Reporter: nasf (Inactive)

Issue Links

• is related to: LU-6053 Review #12531 (LU-4604) failed to compile in lustre 2.4.3

Comments

[nasf (Inactive)]

http://review.whamcloud.com/#/c/8750/ has been merged into 8694, and has been landed to master.

[nasf (Inactive)]

Here is the patch:

http://review.whamcloud.com/#/c/8750/

[nasf (Inactive)]

I think that we lack the protection on the lu_device_type::ldt_device_nr. I will work on that.