[linux-lvm] Problems with LVM2 pvmove again

[Jan Niehusmann]

Hi!

I see strange failures with devmapper, again.

First some information about my current setup: linux-2.4.22, device-mapper
from current cvs, otherwise it's the same as last time I had problems
with pvmove...

Again, problems were caused by a pvmove /dev/hda4. While the pvmove is
running, reading /dev/vgraid/vmware returned pure nonsense starting at
some point in the middle of the device.

# dmsetup table vgraid-vmware      
0 7471104 linear 009:002 65995136
7471104 4194304 linear 254:029 0

The first 7471104 blocks are ok, but then the mapping fails somehow:

# dd if=/dev/mapper/vgraid-vmware bs=512 skip=7471103
[...]
c017c590 T pci_write_config_word
c017c5d0 T pci_write_config_dword
c017c610 T pci_set_master
c017c660 T pci_set_dma_mask
c017c690 T pci_dac_set_dma_maskNÝÀu©n1JãrqIIè^?ÿ×ú<95>-®aÝ1ž8.^K&y#°&²nn^Ud^?8.p
H\åq<9E>¿ZÆ<9B>ÄB^F<93>c        ^U²<91><9D>À<9C>ç<81><9A>æîo®g<9C>³^_õø=I ú^QØ}+
úV0<94>µgá<8B>O<81><9D>¢§<9F>óFû>@ÊBä(Ç@+jÊ9£<8B>c;HØßç<81><8D>Ã=*-2^X<85><9C>RÊ
ãnÅF^D<80>rG&ºKW<86>^T^AB:^F^M<8D>Ã<91>ô¬e;.TkfÛM<8B>k+È6H^]UHl<91><8E>=« µ{v!]€

The ASCII parts seems to be the correct content, and the binary stuff 
starting at block 7471104 is wrong. 
What is 254:029 in the dmsetup output above? Let's check dmsetup ls:
[...]
vgraid-pvmove0  (254, 29)
[...]

Well, as expected, it's the pvmove device. Let's get the table:

# dmsetup table vgraid-pvmove0
0 4194304 mirror core 1 1024 2 003:004 64288496 009:002 73466240 
4194304 1245184 linear 003:004 87619312
5439488 98304 linear 003:004 91420400
5537792 2097152 linear 003:004 78640880
7634944 229376 linear 003:004 7370480
7864320 622592 linear 003:004 110524144
8486912 327680 linear 003:004 138508016
8814592 32768 linear 003:004 106002160
[...]

And what's in there?

# dd if=/dev/mapper/vgraid-pvmove0 | head

c017c6b0 t pci_size
c017c6d0 t pci_read_bases
c017c8f0 t pci_read_irq
c017c930 T pci_setup_device
c017caa0 t pci_pm_save_state_device
c017cad0 t pci_pm_suspend_device
c017cb00 t pci_pm_resume_device
c017cb20 t pci_pm_save_state_bus
c017cb90 t pci_pm_suspend_bus
[...]

So this looks like the stuff that should be in block 7471104 of
vgraid-vmware. Why does vgraid-vmware show something else?
(BTW, I checked, both /dev/hda4 and /dev/md2 contain the same content as 
vgraid-pvmove0, ie. the correct content)

Jan