[nycbug-talk] FreeBSD PseudoRAID RAID0 array broken on atapci1: <Intel ICH5 SATA150 controller>

[Yarema]

Hi, I need some help recovering from this.  First some back story.  Running 
6.2-STABLE i386 from Sep 17, 2007.  My /home slice is mounted from 
/dev/ar0s1e where the relevant kernel messages look like so when all is 
good:

atapci1: <Intel ICH5 SATA150 controller>
ata2: <ATA channel 0> on atapci1
ata3: <ATA channel 1> on atapci1
ad4: 381554MB <WDC WD4000YR-01PLB0 01.06A01> at ata2-master SATA150
ad6: 381554MB <WDC WD4000YR-01PLB0 01.06A01> at ata3-master SATA150
ar0: 763108MB <FreeBSD PseudoRAID RAID0 (stripe 256 KB)> status: READY
ar0: disk0 READY using ad4 at ata2-master
ar0: disk1 READY using ad6 at ata3-master

Today this server crashed with the following loggeed:

ad4: TIMEOUT - WRITE_DMA retrying (1 retry left) LBA=144888320
ad4: TIMEOUT - READ_DMA retrying (1 retry left) LBA=143390319
ad4: FAILURE - device detached
ar0: FAILURE - RAID0 array broken
subdisk4: detached
ad4: detached
g_vfs_done():ar0s1e[WRITE(offset=146002964480, length=2048)]error = 5
initiate_write_filepage: already started
g_vfs_done():ar0s1e[WRITE(offset=146002964480, length=2048)]error = 5
g_vfs_done():ar0s1e[WRITE(offset=6144000, length=16384)]error = 5
g_vfs_done():ar0s1e[WRITE(offset=6160384, length=16384)]error = 5
g_vfs_done():ar0s1e[WRITE(offset=6176768, length=16384)]error = 5
g_vfs_done():ar0s1e[WRITE(offset=6193152, length=16384)]error = 5
g_vfs_done():ar0s1e[WRITE(offset=6209536, length=2048)]error = 5
g_vfs_done():ar0s1e[WRITE(offset=65536, length=2048)]error = 5
g_vfs_done():ar0s1e[WRITE(offset=147801325568, length=12288)]error = 5
g_vfs_done():ar0s1e[WRITE(offset=147142686720, length=2048)]error = 5
g_vfs_done():ar0s1e[WRITE(offset=65536, length=2048)]error = 5
g_vfs_done():ar0s1e[WRITE(offset=6144000, length=16384)]error = 5
g_vfs_done():ar0s1e[WRITE(offset=6160384, length=16384)]error = 5
g_vfs_done():ar0s1e[WRITE(offset=6176768, length=16384)]error = 5
g_vfs_done():ar0s1e[WRITE(offset=6193152, length=16384)]error = 5
g_vfs_done():ar0s1e[WRITE(offset=6209536, length=2048)]error = 5
g_vfs_done():ar0s1e[WRITE(offset=146831867904, length=16384)]error = 5
g_vfs_done():ar0s1e[WRITE(offset=147024330752, length=16384)]error = 5
initiate_write_filepage: already started
g_vfs_done():ar0s1e[WRITE(offset=146002964480, length=2048)]error = 5
initiate_write_filepage: already started
g_vfs_done():ar0s1e[WRITE(offset=146002964480, length=2048)]error = 5
initiate_write_filepage: already started
g_vfs_done():ar0s1e[WRITE(offset=147801325568, length=12288)]error = 5
initiate_write_filepage: already started
g_vfs_done():ar0s1e[WRITE(offset=147142686720, length=2048)]error = 5

Now the kernel messages read:

ar0: FAILURE - RAID0 array broken
ar0: 763108MB <FreeBSD PseudoRAID RAID0 (stripe 256 KB)> status: BROKEN
ar0: disk0 READY using ad4 at ata2-master
ar0: disk1 DOWN no device found for this subdisk
ar1: 763108MB <FreeBSD PseudoRAID RAID0 (stripe 256 KB)> status: BROKEN
ar1: disk0 DOWN no device found for this subdisk
ar1: disk1 READY using ad6 at ata3-master

For some reason the second disk in the array shows up as ar1 instead of 
being part of ar0.  I suspect there's gotta be some way to force the two 
drives to show up as part of the same array by perhaps editing the 
PseudoRAID metadata on disk without putting any of the UFS2 data in 
"jeopardy".  Any pointers on where to start poking around for the relevant 
metadata structures on disk or what to search for?  I figure if I can dd 
the metadata off the disks, tweak a field or two and then dd the whole mess 
back I stand a chance of either hosing the array irrevocably or getting it 
all back. ;)  Or maybe atacontrol could be used to re-create the metadata 
without destroying the UFS2 on the array?  I have a coredump of the kernel 
from this crash if that helps analyze things any.

-- 
Yarema