Tape Tools (hp_ltt) Support For C1537A DDS-3 Drive

[ltt.team]

The drive is responding to TUR's however the Inquiry command is failing with 'Invalid field in SCSI CDB'. For this reason, LTT is unable to detect the drive attached to the host.

Is this a virtual environment?

- Satya

[ANDREW PIZIALI]

Satya, you wrote:

> The drive is responding to TUR's however the Inquiry command is failing with 'Invalid field in SCSI CDB'. For this reason, LTT is unable to detect the drive attached to the host.

> Is this a virtual environment?

No. Fedora Core 8 is running the Linux 2.6.26.5 kernel on a Pentium 4 processor in a socket 478 Gigabyte P4 Titan motherboard.

[ltt.team]

hmmm... can you email the output of '/var/log/messages' & EventLog.ltt as an email attachment to ltt.team@hp.com

I am not sure who (device communication path) is not supporting generic SCSI commands. The logs might help us identify it.

Thanks,
Satya

[ANDREW PIZIALI]

Satya, I mailed a new EventLog.ltt and my current /var/log/messages to ltt.team@hp.com with the subject line "Re: Tape Tools (hp_ltt) Support For C1537A DDS-3 Drive."

[ANDREW PIZIALI]

Satya, you wrote:

> Thanks for the logs.

> Sadly, the messages output didn???t contain the errors we were seeking as the output was captured before L&TT was initialized, based on the timestamps.

This cannot be because the sequence I followed was (1) run hp_ltt, (2) capture EventLog.ltt, (3) capture /var/log/messages. The last two lines recorded in messages:

..... Oct 20 10:45:57 piziali ntpd: synchronized to 66.7.96.1, stratum 2
..... Oct 20 10:45:57 piziali ntpd: kernel time sync status change 4001

were written AFTER hp_ltt was run. Hence, hp_ltt apparently did not trigger any messages to be written to this file. Shall I run the experiment again, recording the two files?

> We were looking for the error being logged (& by whom) when LTT communicates to the tape drive. There is a sense condition 5, 0x2400 in LTT???s event log when it sends an inquiry command to the tape drive.

Attached are the EventLog.ltt and /var/log/messages files captured after hp_ltt was run (concatenated together), wherein I executed the command sequence:

........ 1. Initialize LTT
........ 2. Select 'Hardware scan'
........ 3. Enter command> continue
........ 4. Enter command> preference
........ 5. Enter command> 4 (add IO history to Event log)
........ 6. Enter command> up
........ 7. Enter command> rescan
........ 8. Enter command> exit

> Alternatively, we could update the LSI adapter driver and verify if it resolves the issue.

As you noted in the messages log, I am using the Fusion MPT base driver, version 3.04.06. Are you aware of a newer version compatible with the Linux 2.6.26.5 kernel?

Thanks.

[ANDREW PIZIALI]

Satya, you wrote:

> Earlier, I wasnâ t sure if the timestamps correlated as the EventLog.ltt start time was 10:46. Based on the current results, I believe the errors were available earlier also as the log contains the timestamp when it was created. I am sorry for requesting the logs again.

> From the logs, I noticed the following events are recorded when LTT issues the inquiry command to the drive.

> Oct 21 06:59:16 piziali kernel: mptbase: ioc0: Recovered from IOC FAULT
> Oct 21 06:59:16 piziali kernel: ioc0: LSI53C1030 C0: Capabilities={Initiator}
> Oct 21 06:59:16 piziali kernel: mptbase: ioc0: pci-resume: success
> Oct 21 06:59:16 piziali kernel: target2:0:0: Beginning Domain Validation
> Oct 21 06:59:16 piziali kernel: eth0: link up, 100Mbps, full-duplex, lpa 0x45E1
> Oct 21 06:59:16 piziali kernel: NVRM: RmPowerManagement: 4
> Oct 21 06:59:16 piziali kernel: target2:0:0: Domain Validation skipping write tests
> Oct 21 06:59:16 piziali kernel: target2:0:0: Ending Domain Validation
> Oct 21 06:59:16 piziali kernel: target2:0:0: FAST-10 SCSI 10.0 MB/s ST (100 ns, offset 32)

This sequence of log messages is in response to an ACPI resume, not an inquiry from LTT. The LTT inquiry was issued about "Oct 21 07:45," where no messages are logged. I have seen the "Recovered from IOC FAULT" message ever since I installed the LSI Logic LSI20320-R host bus adapter, both during boot as well as resume.

> It correlates with the sense condition we are seeing and the behavior of any Linux driver.

We'll need to find a record of error elsewhere since it is not recorded in /var/log/messages. Perhaps dmesg(1)?

> Based on the available information, it looks like the current driver doesnâ t support our SCSI commands.

What SCSI commands is LTT issuing?

> I am not sure of the latest drivers compatible with your kernel.

I will track down the latest version of the Fusion MPT base driver.

Thanks.

[ANDREW PIZIALI]

Satya,

It occurred to me we might get more insight into why the LTT SCSI commands are failing by examining a system call trace. I ran hp_ltt under strace(1) and recorded its output in the "hp_ltt.err" section of the attached file, hp_ltt.log. The section "hp_ltt_sg3.err" contains only those calls from the trace that reference the tape drive generic device /dev/sg3. Note that none of the ioctl calls failed (i.e. they each returned zero).

The "typescript" section of the attached file is the command line log of my interaction with LTT.

If this does not offer sufficient visibility into the SCSI command failures, please forward me a copy of hp_ltt without the symbol table stripped. I will run it under gdb and forward you those results.

Thank you.

[ANDREW PIZIALI]

Satya, as we discussed hp_ltt does not recognize a tape drive if any one of several inquiries to the drive fail. In this case, the inquiry for a serial number from the C1537A drive failed so hp_ltt did not find the drive, preventing its use by hp_ltt. At your suggestion, I replaced LTT 4.6 with version 4.2 and this solved the problem. hp_ltt 4.2 recognizes the drive and allows me to perform diagnostics.

Thank you!