[OmniOS-discuss] smartmontools on omnios

Narayan Desai narayan.desai at gmail.com
Thu Sep 12 15:10:42 UTC 2013


I'm trying to figure out how to probe smart statistics reliably for SATA
drives behind a SAS expander. I've installed the omni-ms version of
smartmontools, and it appears to work, with a few caveats.

If I run with -d sat,12 (as esproul suggested in IRC), it looks like I get
full information back from the drive, but the soft errors and recoverable
errors counts get incremented. I also get this message in the ring buffer:
Sep 11 23:35:15 v3 scsi: [ID 107833 kern.warning] WARNING:
/scsi_vhci/disk at g5000cca369e606ce (sd79):
Sep 11 23:35:15 v3      Error for Command: <undecoded cmd 0xa1>    Error
Level: Recovered
Sep 11 23:35:15 v3 scsi: [ID 107833 kern.notice]        Requested Block: 0
                        Error Block: 0
Sep 11 23:35:15 v3 scsi: [ID 107833 kern.notice]        Vendor: ATA
                       Serial Number:       MN3220
Sep 11 23:35:15 v3 scsi: [ID 107833 kern.notice]        Sense Key:
Soft_Error
Sep 11 23:35:15 v3 scsi: [ID 107833 kern.notice]        ASC: 0x0 (<vendor
unique code 0x0>), ASCQ: 0x1d, FRU: 0x0

If I run with -d scsi, a subset of the smart information is reported, and
the illegal request counter gets incremented.

It looks like the scsi_vhci driver gets upset in two different ways,
depending on how smartctl is called. In the former case it gets an
unexpected message, and in the latter, it thinks the smart request is an
illegal request.

So, from cursory digging into the code, it appears that the common scsi
layer doesn't know how to decode this command, and that is what is causing
the ruckus. It the solution just a matter of adding:
0xa1 0x0 "SAT SMART Data"
to extended_sense_list[] in usr/src/uts/common/io/scsi/impl/scsi_subr.c ?

I'm not sure if I'm reading the error message correctly or not. Is this
basically the right thing to do?
thanks.
 -nld
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://omniosce.org/ml-archive/attachments/20130912/f6f02f3c/attachment.html>


More information about the OmniOS-discuss mailing list