Smartctl 5.43 testing issue in Suse 10 (LSI 1064E)

[jackson4015]

Server+ LSI RAID card 1064E+ Seagate new Gen of 10K HDD:ST 900MM0006

RAID 1 seeting on 2 x SAS HDD;
Use smartctlv5.43 to do test DST, skip;
report error: not supported

But for centos/federo, no such issue;

[Christian Franke]

Please provide smartctl debug (-r ioctl,2) outputs as attachments(!) to this ticket.

[jackson4015]

Debug data for Suse 10

[jackson4015]

Replying to chrfranke:

Please provide smartctl debug (-r ioctl,2) outputs as attachments(!) to this ticket.

Hello,Chrfranke.

Thx for your advice.

As V6.0 has been released, so we try it,still fail again,same as V5.43,
we provide the logs of V6.0,also you could get the Linux ver(suse/10sp3/64bit)in the
attach.And only RAID1 could observe such failure.
PLS note it.

[Christian Franke]

Replying to jackson4015:

And only RAID1 could observe such failure.

Please retry with correct -d option and device name, see man page and Wiki.

[jackson4015]

debug -d data

[jackson4015]

And only RAID1 could observe such failure.

Please retry with correct -d option and device name, see man page and Wiki.

Hello,chrfranke.

As the raid card is neither LSI's 3ware nor Megaraid card,but 1064E, so we use:

smartctl -d scsi /dev/sgx

to retry.Observerd the the diagnostic CMD for SDST was sent : 1d 20 00 00 00 00
Errno is set to EPERM:"SG_IO ioctl failed".

[Christian Franke]

Adding -d scsi has no effect because it is the default.

The outputs suggest that the (mptlinux?) driver does not properly implement the SG_IO I/O-control. If this is the case, we cannot fix this.

Does it work with same controller under other distributions?

(BTW: Please provide future text attachments as plain .txt, not as .rtf)

[jackson4015]

Replying to chrfranke:

Does it work with same controller under other distributions?

Yes. Same system/config. but different distribut:
Use Fedora17 and CentOS to do short DST test on SAS HDD:ST900MM0006,it works OK;
Pls see the attach#1;

Btw,

In order to exclude HDD reason, use another SAS Drv:ST9146853SS under Suse 10,
smartctl 5.38, short DST passed, but switch to smart5.43/6.0, failure took place.
That means only smartctl Ver. is different,it seems that the bug follows the tool because of "SG_IO ioctl failed",PLS see attach#2.
We are eager to know why it happened.If any evidence proved that it is related to
Suse 10, we would report that bug to Novell.
Thx!

[jackson4015]

Attach#1

[jackson4015]

Attach#2

[Christian Franke]

Please report a bug to Novell and provide the link to the ticket here. Make sure to provide detailed info about controller, driver, kernel and smartmontools version. The smartmontools package may contain Suse-specific patches.

Testing with 5.38, 5.39, ..., 5.43 (each build from src tarball) would possibly help to track down the change which introduced the problem.

[jackson4015]

OK, thx,would do it right now.
But could you pls ask the smartmontools developers to review the code of V5.43,
Novell may challenge why 5.38 works OK on same OS,hope get more clue of
"SG_IO ioctl failed".

Btw,
Testing with 5.38, 5.39, ..., 5.43 (each build from src tarball)
Yes, is there more debugging method in smartctl helping to track down that change?
At present,only debug (-r ioctl,2) is available.

[Christian Franke]

Test which smartmontools versions (rebuild from tarballs) in the mentioned interval work as expected or not. This info is needed to review the code changes.

[Alex Samorukov]

6 months without feedback, please reopen if needed.