[LU-11156] scrub treat project quota inode as IGIF last_id

• Type: Bug
• Resolution: Fixed
• Priority: Critical
• Fix Version/s: Lustre 2.12.0, Lustre 2.10.7
• Affects Version/s: Lustre 2.11.0
• Labels: patch
• Environment: enabled project quota
• Severity: 3
• Assignee: Alexander Boyko
• Reporter: Alexander Boyko

Description

If project quota is enabled, scrub treat project quota inode as IGIF last id, cause it has no lma fid and version is 0. And try to create O/inum/LAST_ID O/inum/d{0..31}

And it fails with 

LDISKFS-fs: ldiskfs_getblk:901: aborting transaction: error 28 in __ldiskfs_handle_dirty_metadata
[783792.593392] Call Trace:
 [783792.597289]  [<ffffffff816ae7c8>] dump_stack+0x19/0x1b
 [783792.603819]  [<ffffffff8108ae58>] __warn+0xd8/0x100
 [783792.610095]  [<ffffffff8108af9d>] warn_slowpath_null+0x1d/0x20
 [783792.617369]  [<ffffffffc1564bd2>] __ldiskfs_handle_dirty_metadata+0x1c2/0x220 [ldiskfs]
 [783792.626835]  [<ffffffffc153c841>] ldiskfs_getblk+0x131/0x200 [ldiskfs]
 [783792.634665]  [<ffffffffc153c937>] ldiskfs_bread+0x27/0xc0 [ldiskfs]
 [783792.642167]  [<ffffffffc152e241>] ldiskfs_append+0x81/0x150 [ldiskfs]
 [783792.649826]  [<ffffffffc153524f>] ldiskfs_init_new_dir+0xcf/0x230 [ldiskfs]
 [783792.658049]  [<ffffffffc15355cb>] ldiskfs_mkdir+0x18b/0x290 [ldiskfs]
 [783792.665711]  [<ffffffff81211cdd>] vfs_mkdir+0xbd/0x170
 [783792.672121]  [<ffffffffc1608e61>] simple_mkdir+0x571/0x6b0 [osd_ldiskfs]
 [783792.680600]  [<ffffffffc161b875>] osd_seq_load_locked.isra.21+0x3b5/0x5c3 [osd_ldiskfs]
 [783792.689844]  [<ffffffff81003564>] ? perf_assign_events+0x234/0x2e0
 [783792.697293]  [<ffffffffc16092fc>] osd_seq_load+0x35c/0x520 [osd_ldiskfs]
 [783792.705209]  [<ffffffffc1609501>] osd_object_spec_find+0x41/0x170 [osd_ldiskfs]
 [783792.713706]  [<ffffffffc160c6fc>] osd_obj_spec_insert+0x5c/0x150 [osd_ldiskfs]
 [783792.722131]  [<ffffffffc15f9714>] osd_oi_insert+0xb4/0x4d0 [osd_ldiskfs]
 [783792.729992]  [<ffffffffc160d8ba>] ? osd_scrub_refresh_mapping+0x6a/0x440 [osd_ldiskfs]
 [783792.739074]  [<ffffffffc1564539>] ? __ldiskfs_journal_start_sb+0x69/0xe0 [ldiskfs]
 [783792.747775]  [<ffffffffc160da74>] osd_scrub_refresh_mapping+0x224/0x440 [osd_ldiskfs]
 [783792.756877]  [<ffffffffc15edd2d>] ? osd_ea_fid_set+0xdd/0x420 [osd_ldiskfs]
 [783792.764934]  [<ffffffffc160fcd0>] osd_scrub_check_update+0x280/0x10e0 [osd_ldiskfs]
 [783792.773674]  [<ffffffffc1610b95>] osd_scrub_exec+0x65/0x4f0 [osd_ldiskfs]
 [783792.781529]  [<ffffffffc156b006>] ? ldiskfs_read_inode_bitmap+0x176/0x5c0 [ldiskfs]
 [783792.790229]  [<ffffffffc16114c8>] osd_inode_iteration+0x4a8/0xcf0 [osd_ldiskfs]
 [783792.798834]  [<ffffffffc1610b30>] ? osd_scrub_check_update+0x10e0/0x10e0 [osd_ldiskfs]
 [783792.807873]  [<ffffffffc160e060>] ? osd_preload_next+0xa0/0xa0 [osd_ldiskfs]
 [783792.816258]  [<ffffffffc16125b1>] osd_scrub_main+0x8a1/0xe40 [osd_ldiskfs]
 [783792.824341]  [<ffffffff810c7c80>] ? wake_up_state+0x20/0x20
 [783792.830944]  [<ffffffffc1611d10>] ? osd_inode_iteration+0xcf0/0xcf0 [osd_ldiskfs]
 [783792.839510]  [<ffffffff810b4031>] kthread+0xd1/0xe0
 [783792.845447]  [<ffffffff810b3f60>] ? insert_kthread_work+0x40/0x40
 [783792.852490]  [<ffffffff816c155d>] ret_from_fork+0x5d/0xb0
 [783792.858832]  [<ffffffff810b3f60>] ? insert_kthread_work+0x40/0x40
 [783792.865823] ---[ end trace 6993707a81974a09 ]---
[783792.871346] LDISKFS-fs: ldiskfs_getblk:901: aborting transaction: error 28 in __ldiskfs_handle_dirty_metadata
[783792.882087] LDISKFS-fs error (device md65): ldiskfs_getblk:901: inode #579862668: block 362545323: comm OI_scrub: journal_dirty_metadata failed: handle type 0 started at line 84, credits 16/0, errcode -28
 [783792.917540] Aborting journal on device md65-8.

The fail relates to LU-10598, IGIF with no credits, but the root cause is different - scrub processing the project quota inode.

Issue Links

• is related to: LU-10598 Ignore IGIF formatted last_id