[Ocfs2-devel] ocfs2 crashes on xfstest 013

[Tristan Ye <tristan.ye@oracle.com>]

Hi Goldwyn,
	The root cause was due to we didn't skip an existing hole somewhere when
punching a hole, to speak more accurately, the bug occurs at the time when we
removing a partial extent record(which also means the last extent record to
remove), as a result, the hole-punching code attempts to remove more than the
length of whole extent record, which leads to the failure of following
assert(fs/ocfs2/alloc.c):

5507         BUG_ON(cpos < le32_to_cpu(rec->e_cpos) || trunc_range > rec_range);

	Could you please make a verification against my coming patch, I passed xfs's
013 testcase with the patch;-)


Tristan

Goldwyn Rodrigues wrote:
> Hi,
> 
> While running xfstest number 013 on an ocfs2 partition, it crashes with
> the following message:
> 
> [  463.874451] (fsstress,2615,0):ocfs2_unlink:953 ERROR: status = -39
> [  463.876197] (fsstress,2615,0):ocfs2_unlink:953 ERROR: status = -39
> [  463.983814] (fsstress,2619,1):ocfs2_unlink:953 ERROR: status = -39
> [  464.025495] (fsstress,2630,0):ocfs2_unlink:953 ERROR: status = -39
> [  464.401278] ------------[ cut here ]------------
> [  464.401291] kernel BUG at fs/ocfs2/alloc.c:5507!
> [  464.401304] invalid opcode: 0000 [#1] PREEMPT SMP
> [  464.401317] last sysfs file:
> /sys/devices/pci0000:00/0000:00:0e.0/host0/target0:0:0/0:0:0:0/block/sda/dev
> [  464.401332] Modules linked in: ocfs2 ocfs2_nodemanager configfs
> ocfs2_stackglue quota_tree snd_pcm_oss snd_mixer_oss snd_seq
> snd_seq_device edd microcode dm_mod snd_hda_codec_realtek snd_hda_intel
> snd_hda_codec snd_hwdep snd_pcm snd_timer sr_mod sg cdrom i2c_nforce2
> pcspkr serio_raw snd wmi r8169 soundcore snd_page_alloc ext4 jbd2 crc16
> nouveau ttm drm_kms_helper drm i2c_algo_bit video button fan processor
> ata_generic pata_amd thermal thermal_sys
> [  464.401403]
> [  464.401408] Pid: 2626, comm: fsstress Not tainted 2.6.39-rc7-up+ #1
> eMachines ET1810/EMCP73VT-PM
> [  464.401419] EIP: 0060:[<f7ec8930>] EFLAGS: 00210293 CPU: 0
> [  464.401463] EIP is at ocfs2_remove_extent+0xba0/0xbb0 [ocfs2]
> [  464.401470] EAX: 00000043 EBX: 00000129 ECX: 00000000 EDX: edff30c0
> [  464.401476] ESI: 00000043 EDI: 00000000 EBP: f2321ce0 ESP: f2321c4c
> [  464.401482]  DS: 007b ES: 007b FS: 00d8 GS: 0033 SS: 0068
> [  464.401488] Process fsstress (pid: 2626, ti=f2320000 task=f231ede0
> task.ti=f2320000)
> [  464.401495] Stack:
> [  464.401498]  c02674b0 f2321c50 f2321c50 00200082 efccfc88 efd00be4
> efccfc88 f2321c70
> [  464.401510]  f7ef357b f2321c78 f7f43920 00000043 efd00be4 00000001
> 00000000 000ee000
> [  464.401522]  00000129 0000002b edff30c0 000ee000 f1ea02c0 f2321de0
> 0000003b c03673eb
> [  464.401533] Call Trace:
> [  464.401547]  [<c02674b0>] ? autoremove_wake_function+0x40/0x40
> [  464.401577]  [<f7ef357b>] ? ocfs2_inode_cache_io_unlock+0xb/0x10
> [ocfs2]
> [  464.401615]  [<f7f43920>] ? ocfs2_metadata_cache_io_unlock+0x10/0x20
> [ocfs2]
> [  464.401627]  [<c03673eb>] ? __dquot_free_space+0x3b/0x2f0
> [  464.401654]  [<f7efa2ca>] ? ocfs2_journal_access_di+0x1a/0x20 [ocfs2]
> [  464.401679]  [<f7ec9028>] ocfs2_remove_btree_range+0x6e8/0xa20
> [ocfs2]
> [  464.401690]  [<c0311546>] ? kmem_cache_alloc_trace+0x156/0x160
> [  464.401716]  [<f7efa050>] ? ocfs2_commit_trans+0xf0/0x110 [ocfs2]
> [  464.401744]  [<f7eee449>] __ocfs2_change_file_space+0x1199/0x19f0
> [ocfs2]
> [  464.401753]  [<c032a941>] ? acl_permission_check+0x21/0xb0
> [  464.401780]  [<f7efa2b0>] ? ocfs2_extend_trans+0x240/0x240 [ocfs2]
> [  464.401808]  [<f7ef1962>] ocfs2_change_file_space+0xb2/0xc0 [ocfs2]
> [  464.401836]  [<f7ef8169>] ocfs2_ioctl+0x319/0x6f0 [ocfs2]
> [  464.401844]  [<c032ab75>] ? path_put+0x15/0x20
> [  464.401868]  [<f7ee3efa>] ? ocfs2_inode_unlock+0x4a/0xa0 [ocfs2]
> [  464.401893]  [<f7ee3efa>] ? ocfs2_inode_unlock+0x4a/0xa0 [ocfs2]
> [  464.401904]  [<c040e92e>] ? copy_to_user+0x2e/0x50
> [  464.401911]  [<c0325a5c>] ? cp_new_stat64+0xdc/0xf0
> [  464.401937]  [<f7ef7e50>] ? ocfs2_info_handle+0x80/0x80 [ocfs2]
> [  464.401945]  [<c03315bd>] do_vfs_ioctl+0x43d/0x5b0
> [  464.401952]  [<c03261f2>] ? sys_fstat64+0x22/0x30
> [  464.401958]  [<c03317af>] sys_ioctl+0x7f/0x90
> [  464.401967]  [<c0683998>] sysenter_do_call+0x12/0x28
> [  464.401973] Code: 0c bb e2 ff ff ff 89 7c 24 14 89 74 24 10 c7 44 24
> 08 30 b6 f6 f7 c7 44 24 04 03 03 f6 f7 89 04 24 e8 45 9e 07 00 e9 00 f8
> ff ff <0f> 0b 8d b4 26 00 00 00 00 8d bc 27 00 00 00 00 55 89 e5 57 56
> [  464.402024] EIP: [<f7ec8930>] ocfs2_remove_extent+0xba0/0xbb0 [ocfs2]
> SS:ESP 0068:f2321c4c
> [  464.402436] ---[ end trace 51ce626abe09c6ef ]---
> [  464.435066] (fsstress,2625,0):ocfs2_unlink:953 ERROR: status = -39
> [  464.435162] (fsstress,2631,1):ocfs2_unlink:953 ERROR: status = -39
> [  464.435316] (fsstress,2621,1):ocfs2_unlink:953 ERROR: status = -39
> [  464.436405] (fsstress,2634,0):ocfs2_unlink:953 ERROR: status = -39
> [  464.436443] (fsstress,2634,0):ocfs2_unlink:953 ERROR: status = -39
> 
> This is reproducible always, and even on a local mount.
> The kernel I used is the Linus's latest git tree, or version 2.6.39-rc7
> 
> Let me know if you need more information.
>