2.6.0-test1 ext3 slab/fs corruption

I woke up this morning and discovered that a compile had died due to I/O errors. I found the following in my logs (the rest of the relevant bits from the log is attached -- 5k bz2, 146k uncompressed). Preempt is on, htree is in use, e2fsck worked fine.

Jul 21 19:25:32 entropy kernel: EXT3-fs warning (device hda1):
ext3_unlink: Deleting nonexistent file (5632341), 0
Jul 21 19:28:31 entropy kernel: Slab corruption: start=c2a4ba30,
expend=c2a4bbff, problemat=c2a4baa8
Jul 21 19:28:31 entropy kernel: Last user:
[<c584f877>](ext3_destroy_inode+0x17/0x20 [ext3])
Jul 21 19:28:31 entropy kernel: Data:
************************************************************************************************************************6C 13 EC C4 ***************************************************************************************************************************************************************************************************************************************************************************************************************************************************A5
Jul 21 19:28:31 entropy kernel: Next: 71 F0 2C .77 F8 84 C5 A5 C2 0F 17
14 02 C8 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
Jul 21 19:28:31 entropy kernel: slab error in check_poison_obj(): cache
`ext3_inode_cache': object was modified after freeing
Jul 21 19:28:31 entropy kernel: Call Trace:
Jul 21 19:28:31 entropy kernel:  [<c013b33f>]
check_poison_obj+0x13f/0x190
Jul 21 19:28:31 entropy kernel:  [<c013ca67>]
kmem_cache_alloc+0x117/0x160
Jul 21 19:28:31 entropy kernel:  [<c584f830>] ext3_alloc_inode+0x10/0x40
[ext3]
Jul 21 19:28:31 entropy kernel:  [<c584f830>] ext3_alloc_inode+0x10/0x40
[ext3]
Jul 21 19:28:31 entropy kernel:  [<c0168909>] alloc_inode+0x19/0x150
Jul 21 19:28:32 entropy kernel:  [<c0169336>] new_inode+0x16/0x90
Jul 21 19:28:32 entropy kernel:  [<c5847792>] ext3_new_inode+0x42/0x750
[ext3]
Jul 21 19:28:32 entropy kernel:  [<c013ca72>]
kmem_cache_alloc+0x122/0x160
Jul 21 19:28:32 entropy kernel:  [<c5814442>] new_handle+0x12/0x50 [jbd]
Jul 21 19:28:32 entropy kernel:  [<c5814509>] journal_start+0x89/0xb0
[jbd]
Jul 21 19:28:32 entropy kernel:  [<c584de0f>] ext3_create+0x3f/0x90
[ext3]
Jul 21 19:28:32 entropy kernel:  [<c015edec>] vfs_create+0x7c/0xe0
Jul 21 19:28:32 entropy kernel:  [<c015f43f>] open_namei+0x3ef/0x450
Jul 21 19:28:32 entropy kernel:  [<c014fc55>] filp_open+0x35/0x60
Jul 21 19:28:32 entropy kernel:  [<c015011f>] sys_open+0x3f/0x70
Jul 21 19:28:32 entropy kernel:  [<c010aec7>] syscall_call+0x7/0xb
Jul 21 19:28:32 entropy kernel:

please, show your .config

NM> I woke up this morning and discovered that a compile had died due to I/O NM> errors. I found the following in my logs (the rest of the relevant bits NM> from the log is attached -- 5k bz2, 146k uncompressed). Preempt is on, NM> htree is in use, e2fsck worked fine.

NM> Jul 21 19:25:32 entropy kernel: EXT3-fs warning (device hda1):
NM> ext3_unlink: Deleting nonexistent file (5632341), 0
NM> Jul 21 19:28:31 entropy kernel: Slab corruption: start=c2a4ba30,
NM> expend=c2a4bbff, problemat=c2a4baa8
NM> Jul 21 19:28:31 entropy kernel: Last user:
NM> [<c584f877>](ext3_destroy_inode+0x17/0x20 [ext3])
NM> Jul 21 19:28:31 entropy kernel: Data:
NM> ************************************************************************************************************************6C 13 EC C4 ***************************************************************************************************************************************************************************************************************************************************************************************************************************************************A5 NM> Jul 21 19:28:31 entropy kernel: Next: 71 F0 2C .77 F8 84 C5 A5 C2 0F 17
NM> 14 02 C8 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
NM> Jul 21 19:28:31 entropy kernel: slab error in check_poison_obj(): cache
NM> `ext3_inode_cache': object was modified after freeing
NM> Jul 21 19:28:31 entropy kernel: Call Trace:
NM> Jul 21 19:28:31 entropy kernel: [<c013b33f>]
NM> check_poison_obj+0x13f/0x190
NM> Jul 21 19:28:31 entropy kernel: [<c013ca67>]
NM> kmem_cache_alloc+0x117/0x160
NM> Jul 21 19:28:31 entropy kernel: [<c584f830>] ext3_alloc_inode+0x10/0x40
NM> [ext3]
NM> Jul 21 19:28:31 entropy kernel: [<c584f830>] ext3_alloc_inode+0x10/0x40
NM> [ext3]

 NM> Jul 21 19:28:31 entropy kernel:  [<c0168909>] alloc_inode+0x19/0x150
 NM> Jul 21 19:28:32 entropy kernel:  [<c0169336>] new_inode+0x16/0x90
 NM> Jul 21 19:28:32 entropy kernel:  [<c5847792>] ext3_new_inode+0x42/0x750
 NM> [ext3]
 NM> Jul 21 19:28:32 entropy kernel:  [<c013ca72>]
 NM> kmem_cache_alloc+0x122/0x160
 NM> Jul 21 19:28:32 entropy kernel:  [<c5814442>] new_handle+0x12/0x50 [jbd]
 NM> Jul 21 19:28:32 entropy kernel:  [<c5814509>] journal_start+0x89/0xb0
 NM> [jbd]
 NM> Jul 21 19:28:32 entropy kernel:  [<c584de0f>] ext3_create+0x3f/0x90
 NM> [ext3]
 NM> Jul 21 19:28:32 entropy kernel:  [<c015edec>] vfs_create+0x7c/0xe0
 NM> Jul 21 19:28:32 entropy kernel:  [<c015f43f>] open_namei+0x3ef/0x450
 NM> Jul 21 19:28:32 entropy kernel:  [<c014fc55>] filp_open+0x35/0x60
 NM> Jul 21 19:28:32 entropy kernel:  [<c015011f>] sys_open+0x3f/0x70
 NM> Jul 21 19:28:32 entropy kernel:  [<c010aec7>] syscall_call+0x7/0xb

NM> Jul 21 19:28:32 entropy kernel:

hmm. it really looks like memory problem. have you tried to test it? all the corruptions happened at the same address.

10 hours of memtest86 found nothing, and my config is attached.