Re: PROBLEM: doing a dd to a dbfile

[Jeff Mahoney <jeffm@suse.com>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Ing. Raúl Alvarez Aguileta wrote:
> Hi
> 
> I've a corruption problem with one innodb file, when i'm trying to
> recover a corrupted file (it's filled with no data at byte 524288+, but
> it have valid info later [aprox 4mb ahead] ) with:
> 
> # dd if=/dev/zero of=ibdata1 bs=1 seek=524288 count=4194304 conv=notrunc
> 
> the command ends dumping the folllowing message:
> 
> output from /var/log/messages
> ---->
> Jun  1 18:43:07 g4nd4lf Kernel BUG at file:534
> Jun  1 18:43:07 g4nd4lf invalid operand: 0000 [1] SMP
> Jun  1 18:43:07 g4nd4lf CPU 0
> Jun  1 18:43:07 g4nd4lf Modules linked in: ipt_LOG ipt_state
> iptable_filter ip_tables ip_conntrack_tftp ip_conntrack_irc
> ip_conntrack_ftp ip_conntrack_amanda ip_conntrack amd74xx sd_mod aic79xx
> mptctl mptscsih mptbase sata_sx4 sata_promise libata usb_storage
> ide_core scsi_mod ohci_hcd uhci_hcd ehci_hcd usbcore
> Jun  1 18:43:07 g4nd4lf Pid: 8993, comm: dd Not tainted 2.6.11-gentoo-r9
> Jun  1 18:43:07 g4nd4lf RIP: 0010:[<ffffffff801a9372>]
> <ffffffff801a9372>{reiserfs_file_write+5298}
> Jun  1 18:43:07 g4nd4lf RSP: 0018:ffff81007c273ba8  EFLAGS: 00010246
> Jun  1 18:43:08 g4nd4lf RAX: 0000000001000020 RBX: 0000000000000000 RCX:
> 0000000000000001
> Jun  1 18:43:08 g4nd4lf RDX: ffff810002b04038 RSI: ffff81007c26e670 RDI:
> ffff81007c273d98
> Jun  1 18:43:08 g4nd4lf RBP: ffff81007e8bbdf0 R08: 0000000000000000 R09:
> ffff810002c13c80
> Jun  1 18:43:08 g4nd4lf R10: ffff81007fdd8ec0 R11: ffff81007e8bbdf0 R12:
> 0000000000000001
> Jun  1 18:43:08 g4nd4lf R13: 0000000000000000 R14: ffff81007c26e670 R15:
> ffff81007e8c0678
> Jun  1 18:43:08 g4nd4lf FS:  00002aaaaade7b00(0000)
> GS:ffffffff803ed880(0000) knlGS:0000000000000000
> Jun  1 18:43:08 g4nd4lf CS:  0010 DS: 0000 ES: 0000 CR0:
> 000000008005003b
> Jun  1 18:43:08 g4nd4lf CR2: 00002aaaaac4d3e0 CR3: 000000007c267000 CR4:
> 00000000000006e0
> Jun  1 18:43:08 g4nd4lf Process dd (pid: 8993, threadinfo
> ffff81007c272000, task ffff81007f6fa1b0)
> Jun  1 18:43:08 g4nd4lf Stack: 0000000000000286 ffff81000000d680
> 0000000000000000 ffff81007c26e730
> Jun  1 18:43:08 g4nd4lf 0000000100000000 000000000008cfff
> 00000000000080d2 0000000000000000
> Jun  1 18:43:08 g4nd4lf 0000000000000000 0000000000000001
> Jun  1 18:43:08 g4nd4lf Call Trace:<ffffffff80145f9e>{buffered_rmqueue
> +654} <ffffffff801527ab>{do_no_page+939}
> Jun  1 18:43:08 g4nd4lf <ffffffff8010de4d>{error_exit+0}
> <ffffffff8018b88f>{inotify_inode_queue_event+47}
> Jun  1 18:43:08 g4nd4lf <ffffffff801f1160>{read_zero+480}
> <ffffffff80164669>{vfs_write+233}
> Jun  1 18:43:08 g4nd4lf <ffffffff801647f3>{sys_write+83}
> <ffffffff8010d44a>{system_call+126}
> Jun  1 18:43:08 g4nd4lf
> Jun  1 18:43:08 g4nd4lf
> Jun  1 18:43:08 g4nd4lf Code: 0f 0b 8b f1 2d 80 ff ff ff ff 16 02 8b 02
> 4c 8b 42 10 f6 c4
> Jun  1 18:43:08 g4nd4lf RIP <ffffffff801a9372>{reiserfs_file_write+5298}
> RSP <ffff81007c273ba8>
> <----
> 
> # sh /usr/src/linux/scripts/ver_linux
> ---->
> If some fields are empty or look unusual you may have an old version.
> Compare to the current minimal requirements in Documentation/Changes.
> 
> Linux g4nd4lf 2.6.11-gentoo-r9 #1 SMP Wed May 18 07:53:38 CDT 2005
> x86_64 AMD Opteron(tm) Processor 250 AuthenticAMD GNU/Linux

Ok, this trace maps to page_buffers() in
reiserfs_allocate_blocks_for_region(), specifically the BUG_ON
(!PagePrivate(page)) check inside it. Somehow, the buffers associated
with that page are getting freed, despite it being locked.

Just to double check that the page it's checking is valid - could you
post your vmlinux somewhere or send it to me off list? I'd like to check
the Oops against the disassembly.

- -Jeff

- --
Jeff Mahoney
SuSE Labs
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.0 (GNU/Linux)

iD8DBQFCn1u2LPWxlyuTD7IRAtipAJ0dnkRURMCvDVi7BsqCLB3PKTRlXACfXXip
nREAY313itZta85HPacCSKc=
=4U9c
-----END PGP SIGNATURE-----