I am new to Linux, so forgive any obvious missteps or use of incorrect
terms. With that :-), here is my problem:
- HP DL580 G8, four quad-core X7350s, 128GB of ram
- RHEL 4 update 4, 64-bit, installed yesterday
- EMC Clariion cx3-40, f/w revision 3.26.40.5.010, PROM revision
3.58.0
- 16 RAID 0 LUNs, 8 assigned to each controller
- QLA2432 (2462) 4Gb FC card;
- qla2xxx-v8.01.07.15-2 driver, compiled on the system

I cannot access the LUNs on controller B. In dmesg, I see the
following for the controller A LUNs:

Attached scsi disk sdk at scsi0, channel 0, id 5, lun 6
Vendor: DGC Model: RAID 0 Rev: 0326
Type: Direct-Access ANSI SCSI revision: 04
qla2400 0000:1c:00.0: scsi(0:0:5:7): Enabled tagged queuing, queue
depth 32.
SCSI device sdl: 786432000 512-byte hdwr sectors (402653 MB)
SCSI device sdl: drive cache: write through
SCSI device sdl: 786432000 512-byte hdwr sectors (402653 MB)
SCSI device sdl: drive cache: write through
sdl: sdl1 sdl2 sdl3 sdl4 < sdl5 sdl6 sdl7 sdl8 sdl9 sdl10 sdl11
sdl12 sdl13 sdl14 >
Attached scsi disk sdl at scsi0, channel 0, id 5, lun 7


But for the next LUN, on controller B, I see:

Vendor: DGC Model: RAID 0 Rev: 0326
Type: Direct-Access ANSI SCSI revision: 04
qla2400 0000:1c:00.0: scsi(0:0:5:8): Enabled tagged queuing, queue
depth 32.
SCSI device sdm: 838860800 512-byte hdwr sectors (429497 MB)
sdm: asking for cache data failed
sdm: assuming drive cache: write through
SCSI device sdm: 838860800 512-byte hdwr sectors (429497 MB)
sdm: asking for cache data failed
sdm: assuming drive cache: write through
sdm:<6>Device sdm not ready.
end_request: I/O error, dev sdm, sector 0
Buffer I/O error on device sdm, logical block 0
Device sdm not ready.
end_request: I/O error, dev sdm, sector 0
Buffer I/O error on device sdm, logical block 0
Device sdm not ready.
end_request: I/O error, dev sdm, sector 0
Buffer I/O error on device sdm, logical block 0
unable to read partition table
Attached scsi disk sdm at scsi0, channel 0, id 5, lun 8

There is an entry in /proc/partitions for /dev/sdm, and it correctly
responds to scsi_id, but no sign of its partitions in /proc/
partitions, and I cannot dd from it, etc. If I trespass the LUNs from
controller B and move them all to controller A, all is well.

I installed the multipath package device-mapper-
multipath-0.4.5-27.el4_6.3.x86_64.rpm, and that didn't help. I found
that no matter what I did, the system never discovered two paths for
any of the LUNs for me to worry about picking the right one.

What am I doing wrong?

Thanks,
Reza Taheri