mirrored lv with 6TB

Hi,

*i am currently using the kernel 2.6.9-42.0.3.ELsmp and lvm2-2.02.06-6.0.RHEL4 on RHEL4.4 (32bitCPU) and i have three disks:* /dev/sdb: 6497.8 GB
/dev/sdc: 6497.8 GB
/dev/sdd: 283 MB

*i want to create a volumegroup with a mirrored lv. but this is only possible up to 4TB. if i want to create a mirrored lv with 6TB, i got kernel panic with following in the /var/log/messages:*

kernel: ------------[ cut here ]------------
kernel: kernel BUG at drivers/md/dm-io.c:148!
kernel: invalid operand: 0000 [#1]
kernel: SMP
kernel: Modules linked in: dlm(U) cman(U) ipv6 sunrpc crc32c libcrc32c md5 iscsi_sfnet scsi_transport_iscsi dm_mirror dm_mod button bat tery ac ohci_hcd tg3 bonding(U) floppy sg ext3 jbd mptscsih mptsas mptspi mptfc mptscsi mptbase sd_mod scsi_mod

kernel: CPU:    3
kernel: EIP:    0060:[<f892e424>]    Not tainted VLI
kernel: EFLAGS: 00010206   (2.6.9-42.0.3.ELsmp)
kernel: EIP is at bio_set_alloc+0x95/0x11c [dm_mod]
kernel: eax: f532c700   ebx: 0000081f   ecx: f7073980   edx: f531a000
kernel: esi: 00000000   edi: 00000105   ebp: f532c700   esp: f5b85e20
kernel: ds: 007b   es: 007b   ss: 0068
kernel: Process vgchange (pid: 3266, threadinfo=f5b85000 task=f54ca0b0)
kernel: Stack: 00000004 00000000 00000010 f8937bc0 0000081f 00000000 00000000 00000000
kernel:        f892e810 c2b3b164 00000000 00000010 00000019 c283dde0 f5b539b0 f5b85ee8
kernel:        00000010 00000000 00000000 f892e957 f5b85ee8 f5b85eb0 c2b3b164 00000001
kernel: Call Trace:
kernel:  [<f892e810>] do_region+0x40/0x104 [dm_mod]
kernel:  [<f892e957>] dispatch_io+0x83/0xa4 [dm_mod]
kernel:  [<f892e773>] vm_get_page+0x0/0x2a [dm_mod]
kernel:  [<f892e79d>] vm_next_page+0x0/0x17 [dm_mod]
kernel:  [<f892e9f3>] sync_io+0x7b/0xdf [dm_mod]
kernel:  [<f892eb85>] dm_io_sync_vm+0x2a/0x32 [dm_mod]
kernel:  [<f892e773>] vm_get_page+0x0/0x2a [dm_mod]
kernel:  [<f892e79d>] vm_next_page+0x0/0x17 [dm_mod]
kernel:  [<f88b929b>] read_bits+0x16/0x1a [dm_mirror]
kernel:  [<f88b9760>] disk_resume+0x2c/0x184 [dm_mirror]
kernel:  [<f88b9734>] disk_resume+0x0/0x184 [dm_mirror]
kernel:  [<f88bbe12>] mirror_resume+0x14/0x2c [dm_mirror]
kernel:  [<f892b6e5>] dm_table_resume_targets+0x27/0x37 [dm_mod]
kernel:  [<f892a538>] dm_resume+0x68/0x102 [dm_mod]
kernel:  [<f892cf8e>] do_resume+0x158/0x178 [dm_mod]
kernel:  [<f892df16>] ctl_ioctl+0xd9/0x13c [dm_mod]
kernel:  [<f892cfae>] dev_suspend+0x0/0x10 [dm_mod]
kernel:  [<c016add6>] sys_ioctl+0x227/0x269
kernel:  [<c02d47cb>] syscall_call+0x7/0xb
kernel: Code: 04 24 00 00 00 00 4f 74 31 eb 20 c7 04 24 04 00 00 00 81 ff 80 00 00 00 76 20 81 ff 00 01 00 00 77 09 c7 04 24 05 00 00 0

0 eb 0f <0f> 0b 94 00 59 04 93 f8 c7 04 24 00 00 00 00 8b 04 24 8b 74 24
kernel: <0>Fatal exception: panic in 5 seconds

*is there any patch outside, which i can use to create bigger volumes? *
michael

...

Bioset allocation code in device mapper is different in RHEL4.5, please test it with RHEL4.5 update.

If you hit this bug again, please use Red Hat support or bugzilla to report this problem, thanks.

Milan
--
[email protected]

Hi,

It looks that dm-log requested too big write to dm-io and dm-io couldn't handle it.

You should be able to work around it using larger region size (-R option of lvcreate).

For example, the following command creates a LV with 4MB region size.

# lvcreate -L 6T -R 4 ...

Thanks,

Jun'ichi Nomura, NEC Corporation of America