[LU-1045] kernel BUG at fs/jbd2/transaction.c:1033!

• Type: Bug
• Resolution: Duplicate
• Priority: Critical
• Environment: lustre 2.1.0-17chaos, ldiskfs-4.0.6 (LLNL's external ldiskfs build, based on ldiskfs from 2.1.0-17chaos. Source at http://github.com/chaos/ldiskfs
• Severity: 2
• Assignee: Andreas Dilger
• Reporter: Christopher Morrone (Inactive)

Description

We hit an assertion in jbd2 on our MDS less than 24 hours after upgrading the 1.8 filesystem to 2.1 (clients are still 1.8).

It looks like the assertion in question is in fucntion jbd2_journal_dirty_metadata(), here:

	if (jh->b_modified == 0) {
		/*
		 * This buffer's got modified and becoming part
		 * of the transaction. This needs to be done
		 * once a transaction -bzzz
		 */
		jh->b_modified = 1;
		J_ASSERT_JH(jh, handle->h_buffer_credits > 0);
		handle->h_buffer_credits--;
	}

So Alex probably needs to look at this. Here is our backtrace:

crash> bt
PID: 9121   TASK: ffff8807db9a8b40  CPU: 2   COMMAND: "mdt_159"
 #0 [ffff8807db9af280] machine_kexec at ffffffff81031fcb
 #1 [ffff8807db9af2e0] crash_kexec at ffffffff810b8d92
 #2 [ffff8807db9af3b0] oops_end at ffffffff814f1f40
 #3 [ffff8807db9af3e0] die at ffffffff8100f26b
 #4 [ffff8807db9af410] do_trap at ffffffff814f1834
 #5 [ffff8807db9af470] do_invalid_op at ffffffff8100ce35
 #6 [ffff8807db9af510] invalid_op at ffffffff8100bedb
    [exception RIP: jbd2_journal_dirty_metadata+269]
    RIP: ffffffffa0ab476d  RSP: ffff8807db9af5c0  RFLAGS: 00010246
    RAX: ffff8801a9fc8c80  RBX: ffff8801a919aa98  RCX: ffff880832eeaf68
    RDX: 0000000000000000  RSI: ffff880832eeaf68  RDI: 0000000000000000
    RBP: ffff8807db9af5e0   R8: d010000000000000   R9: f7af135d53a3da02
    R10: 0000000000000001  R11: 0000000000000000  R12: ffff8801b3030358
    R13: ffff880832eeaf68  R14: ffff880631a7a800  R15: 0000000000002000
    ORIG_RAX: ffffffffffffffff  CS: 0010  SS: 0018
 #7 [ffff8807db9af5e8] __ldiskfs_handle_dirty_metadata at ffffffffa0acdf3b [ldiskfs]
 #8 [ffff8807db9af628] fsfilt_ldiskfs_write_handle at ffffffffa0b88758 [fsfilt_ldiskfs]
 #9 [ffff8807db9af6b8] fsfilt_ldiskfs_write_record at ffffffffa0b889dc [fsfilt_ldiskfs]
#10 [ffff8807db9af708] llog_lvfs_write_blob at ffffffffa05b215b [obdclass]
#11 [ffff8807db9af778] llog_lvfs_write_rec at ffffffffa05b327a [obdclass]
#12 [ffff8807db9af818] llog_cat_add_rec at ffffffffa05b1189 [obdclass]
#13 [ffff8807db9af868] llog_obd_origin_add at ffffffffa05b5906 [obdclass]
#14 [ffff8807db9af898] llog_add at ffffffffa05b5a61 [obdclass]
#15 [ffff8807db9af8d8] lov_llog_origin_add at ffffffffa09330c4 [lov]
#16 [ffff8807db9af958] llog_add at ffffffffa05b5a61 [obdclass]
#17 [ffff8807db9af998] mds_llog_origin_add at ffffffffa0ba8193 [mds]
#18 [ffff8807db9af9e8] llog_add at ffffffffa05b5a61 [obdclass]
#19 [ffff8807db9afa28] mdd_log_op_setattr at ffffffffa0bdd3d6 [mdd]
#20 [ffff8807db9afa88] mdd_setattr_log at ffffffffa0bddc78 [mdd]
#21 [ffff8807db9afad8] mdd_attr_set at ffffffffa0bd62fc [mdd]
#22 [ffff8807db9afbc8] cml_attr_set at ffffffffa0cfab6d [cmm]
#23 [ffff8807db9afc08] mdt_attr_set at ffffffffa0c5266b [mdt]
#24 [ffff8807db9afc68] mdt_reint_setattr at ffffffffa0c52c45 [mdt]
#25 [ffff8807db9afce8] mdt_reint_rec at ffffffffa0c4f110 [mdt]
#26 [ffff8807db9afd18] mdt_reint_internal at ffffffffa0c49f78 [mdt]
#27 [ffff8807db9afd68] mdt_reint at ffffffffa0c4a234 [mdt]
#28 [ffff8807db9afd98] mdt_handle_common at ffffffffa0c3e74d [mdt]
#29 [ffff8807db9afde8] mdt_regular_handle at ffffffffa0c3f395 [mdt]
#30 [ffff8807db9afdf8] ptlrpc_main at ffffffffa0728181 [ptlrpc]
#31 [ffff8807db9aff48] kernel_thread at ffffffff8100c14a

Issue Links

• duplicates: LU-601 kernel BUG at fs/jbd2/transaction.c:1030