Re: [PATCH 2/2] xfs: fuzz every field of every structure and test kernel crashes

[Eryu Guan <guaneryu@gmail.com>]

On Tue, Jul 03, 2018 at 09:50:37PM -0700, Darrick J. Wong wrote:
> From: Darrick J. Wong <darrick.wong@oracle.com>
> 
> Fuzz every field of every structure and then try to write the
> filesystem, to see how many of these writes can crash the kernel.
> 
> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>

The "re-repair" failures are gone, but I still see some test failures
like (xfs/1398 for example)

+re-mount failed (32) with magic = zeroes.
+re-mount failed (32) with magic = ones.
...

Looks like the re-mount is expected to fail as we skipped all the repair work.

Also, there're _check_dmesg failures too (they were buried among other
failures so I didn't notice them in last review), like this "Internal
error" from xfs/1397

[1513573.879719] [U] ++ Try to write filesystem again
[1513574.092652] XFS (dm-1): Internal error XFS_WANT_CORRUPTED_GOTO at line 756 of file fs/xfs/libxfs/xfs_rmap.c.  Caller xfs_rmap_finish_one+0x206/0x2b0 [xfs]
[1513574.094001] CPU: 1 PID: 7087 Comm: kworker/u4:2 Tainted: G        W  OE     4.18.0-rc1 #1
[1513574.094839] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.11.0-2.fc28 04/01/2014
[1513574.095650] Workqueue: writeback wb_workfn (flush-253:1)
[1513574.096145] Call Trace:
[1513574.096390]  dump_stack+0x5c/0x80
[1513574.096719]  xfs_rmap_map+0x18c/0x8d0 [xfs]
[1513574.097138]  ? xfs_free_extent_fix_freelist+0x7d/0xb0 [xfs]
[1513574.097662]  ? _cond_resched+0x15/0x30
[1513574.098021]  ? kmem_cache_alloc+0x16a/0x1d0
[1513574.098435]  ? kmem_zone_alloc+0x61/0xe0 [xfs]
[1513574.098877]  xfs_rmap_finish_one+0x206/0x2b0 [xfs]
[1513574.099355]  ? xfs_trans_free+0x55/0xc0 [xfs]
[1513574.099788]  xfs_trans_log_finish_rmap_update+0x2f/0x40 [xfs]
[1513574.100346]  xfs_rmap_update_finish_item+0x2d/0x40 [xfs]
[1513574.100865]  xfs_defer_finish+0x164/0x470 [xfs]
[1513574.101318]  ? xfs_rmap_update_cancel_item+0x10/0x10 [xfs]
[1513574.101852]  xfs_iomap_write_allocate+0x182/0x370 [xfs]
[1513574.102371]  xfs_map_blocks+0x209/0x290 [xfs]
[1513574.102819]  xfs_do_writepage+0x147/0x690 [xfs]
[1513574.103265]  ? clear_page_dirty_for_io+0x224/0x290
[1513574.103718]  write_cache_pages+0x1dc/0x450
[1513574.104141]  ? xfs_vm_readpage+0x70/0x70 [xfs]
[1513574.104594]  ? btrfs_wq_submit_bio+0xc9/0xf0 [btrfs]
[1513574.105098]  xfs_vm_writepages+0x59/0x90 [xfs]
[1513574.105534]  do_writepages+0x41/0xd0
[1513574.105886]  ? __switch_to_asm+0x40/0x70
[1513574.106281]  ? __switch_to_asm+0x34/0x70
[1513574.106673]  ? __switch_to_asm+0x40/0x70
[1513574.107067]  ? __switch_to_asm+0x34/0x70
[1513574.107453]  ? __switch_to_asm+0x40/0x70
[1513574.107843]  ? __switch_to_asm+0x34/0x70
[1513574.108235]  ? __switch_to_asm+0x40/0x70
[1513574.108623]  ? __switch_to_asm+0x34/0x70
[1513574.109016]  ? __switch_to_asm+0x40/0x70
[1513574.109406]  ? __switch_to_asm+0x40/0x70
[1513574.109790]  __writeback_single_inode+0x3d/0x350
[1513574.110247]  writeback_sb_inodes+0x1d0/0x460
[1513574.110669]  __writeback_inodes_wb+0x5d/0xb0
[1513574.111172]  wb_writeback+0x255/0x2f0
[1513574.111535]  ? get_nr_inodes+0x35/0x50
[1513574.111904]  ? cpumask_next+0x16/0x20
[1513574.112273]  wb_workfn+0x186/0x400
[1513574.112608]  ? sched_clock+0x5/0x10
[1513574.112955]  process_one_work+0x1a1/0x350
[1513574.113343]  worker_thread+0x30/0x380
[1513574.113702]  ? wq_update_unbound_numa+0x1a0/0x1a0
[1513574.114158]  kthread+0x112/0x130
[1513574.114484]  ? kthread_create_worker_on_cpu+0x70/0x70
[1513574.114980]  ret_from_fork+0x35/0x40
[1513574.115352] XFS (dm-1): xfs_do_force_shutdown(0x8) called from line 222 of file fs/xfs/libxfs/xfs_defer.c.  Return address = 0000000000b9898a
[1513574.309527] XFS (dm-1): Corruption of in-memory data detected.  Shutting down filesystem
[1513574.313154] XFS (dm-1): Please umount the filesystem and rectify the problem(s)

Should the dmesg check be disabled as well?

Thanks,
Eryu

P.S.
BTW, patch 1/1 looks fine, I'll take it for this week's update.