Re: [SCSI REGRESSION] 3.10.2 or 3.10.3: arcmsr failure at bootup / early userspace transition

[Nix <nix@esperi.org.uk>]

On 29 Jul 2013, Bernd Schubert spake thusly:
> Could you try to run these commands with 3.10.1?
>
> # # check if reporting opcodes works
> # sg_opcodes -v  -n /dev/sdX

spindle:/boot# sg_opcodes -v -n /dev/sda
    inquiry cdb: 12 00 00 00 24 00
    Report Supported Operation Codes cmd: a3 0c 00 00 00 00 00 00 20 00 00 00
Report Supported Operation Codes:  Fixed format, current;  Sense key: Illegal Request
 Additional sense: Invalid command operation code
  Info fld=0x0 [0]
  Sense Key Specific: Error in Command byte 3840
Report supported operation codes: operation not supported

(sdb is the same, obviously, since they are both separate RAID volumes
controlled by the same controller.)

> # check ata information page
> # sg_vpd --page=0x89 /dev/sdX

spindle:/boot# sg_vpd --page=0x89 /dev/sda
ATA information VPD page:
fetching VPD page failed

Not very helpful, I know :(

I'll try rebooting into a kernel with that commit reverted next.

Areca controllers appear to be a bit weird: e.g. they needed special
support in smartctl...

>> No changes to arcmsr between those versions... I suspect I'll have to
>> bisect, which will be a complete pig because every failure means a hard
>> powerdown of this box. Always-on servers rarely appreciate hard
>> powerdowns :(
>
> Maybe just revert this commit? Helpful would be some scsi logging to
> see which command actually fails. I guess you don't have a serial
> console?

I could set one up, in theory, but the problem is that all my machines
are rather dependent on my NFS-mounted $HOME. Guess where it's mounted
from... in any case, the machine has no serial port, so it would have to
be a usb-serial console, and we know exactly how reliable those are :/

-- 
NULL && (void)