Skip to contentSkip to conversation

SCSI Tape Device FATAL error on 2.4.10

DISCUSSION Jeff V. Merkey 2 participants 3 messages Oct 25, 2001 – Oct 26, 2001

lore
Share & tools
lore
#0 lore raw

On a ServerWorks HE Chipset system with an Exabyte EXB-480 Robotics Tape library we are seeing a fatal SCSI IO problem that results in a SCSI bus hang on the system. This error is very fatal, and requires that the machine be rebooted
to recover. Following this error, the Linux
Operating System is still running OK, but the affected SCSI bus does not respond to any commands nor do any devices attached to this bus.

The Tape Drive is an Exabyte SCSI Tape. The error occurs when the device reaches end of tape (EOT) during a write operation while writing to the tape.

With tape programming, there really is no good way to know where the end of tape is while archiving data real time, so this error is pretty much fatal. We are using tape partitioning, which we have noticed not many applications in Linux use at present, so these code paths may be related to the problem. I have reviewed st.c but it is not readily apparent where the problem may be in this code, which is leading me to suspect it's related to some interaction between st.c and the drivers with regard to multiple seeks and writes between tape partitions.

Configuration is:

2.4.10
AIC7XXX
ST
EXSCH (Exabyte Robotics Library for Linux 2.2.X/2.4.X) We wrote this.

The EXB-480 has 4 Exabyte tape drives and 30 150GB Tape slots with an import/export port.

THe system we are using is the SuperMicro Serverworks HE Motherboard with 2GB of memory and 3Ware IDE Controllers. We are not using the 3Ware controllers for any of the tape support.

Please advise,

Jeff Merkey
TRG

#1 lore raw

"Jeff V. Merkey" <[email protected]> wrote in message news:[email protected]...

Quoted 15 lines

On a ServerWorks HE Chipset system with an Exabyte EXB-480 Robotics Tape library we are seeing a fatal SCSI IO problem that results in a SCSI bus hang on the system. This error is very fatal, and requires that the machine be rebooted
to recover. Following this error, the Linux
Operating System is still running OK, but the affected SCSI bus does not respond to any commands nor do any devices attached to this bus.

The Tape Drive is an Exabyte SCSI Tape. The error occurs when the device reaches end of tape (EOT) during a write operation while writing to the tape.

Quoted 10 lines

With tape programming, there really is no good way to know where the end of tape is while archiving data real time, so this error is pretty much fatal. We are using tape partitioning, which we have noticed not many applications in Linux use at present, so these code paths may be related to the problem. I have reviewed st.c but it is not readily apparent where the problem may be in this code, which is leading me to suspect it's related to some interaction between st.c and the drivers with regard to multiple seeks and writes between tape partitions.

Jeff,

Logical end-of-tape handling is clearly defined in Exabyte's SCSI reference manual which you can download from their web page. There's nothing fatal about it, your application should handle it. The SCSI Write command that reaches logical end-of-tape (or end-of-partition) will end with a Check condition, Sense key 0h, ASC=00h, ASCQ=00h, EOM bit = 1. Your data will be written to tape just fine. There is plenty of tape left so you can gracefully write a delimiting filemark or whatever you need to close the current data set. All write commands after reaching LEOT will also result in a Check Condition with the above sense information (until you really run out of tape). It's a Logical end-of-tape you deal with, not a Physical one.

As a side note, make sure you have the latest firmware loaded in your M2 drives.

Rob

#2 lore raw
In reply to Jeff V. Merkey
Quoted 41 lines · Rob Turk

"Jeff V. Merkey" <[email protected]> wrote in message news:[email protected]...

Quoted 15 lines

On a ServerWorks HE Chipset system with an Exabyte EXB-480 Robotics Tape library we are seeing a fatal SCSI IO problem that results in a SCSI bus hang on the system. This error is very fatal, and requires that the machine be rebooted
to recover. Following this error, the Linux
Operating System is still running OK, but the affected SCSI bus does not respond to any commands nor do any devices attached to this bus.

The Tape Drive is an Exabyte SCSI Tape. The error occurs when the device reaches end of tape (EOT) during a write operation while writing to the tape.

Quoted 10 lines

With tape programming, there really is no good way to know where the end of tape is while archiving data real time, so this error is pretty much fatal. We are using tape partitioning, which we have noticed not many applications in Linux use at present, so these code paths may be related to the problem. I have reviewed st.c but it is not readily apparent where the problem may be in this code, which is leading me to suspect it's related to some interaction between st.c and the drivers with regard to multiple seeks and writes between tape partitions.

Jeff,

Logical end-of-tape handling is clearly defined in Exabyte's SCSI reference manual which you can download from their web page. There's nothing fatal about it, your application should handle it. The SCSI Write command that reaches logical end-of-tape (or end-of-partition) will end with a Check condition, Sense key 0h, ASC=00h, ASCQ=00h, EOM bit = 1. Your data will be written to tape just fine. There is plenty of tape left so you can gracefully write a delimiting filemark or whatever you need to close the current data set. All write commands after reaching LEOT will also result in a Check Condition with the above sense information (until you really run out of tape). It's a Logical end-of-tape you deal with, not a Physical one.

Quoted 5 lines

As a side note, make sure you have the latest firmware loaded in your M2 drives.

Rob

I am using st.o and it is not handling this condition correctly under Linux. It has nothing to do with your firmware. I am programming the robot directly over SCSI, but I am going through the Linux tape support module st.c for tape drive support. When this error occurs, the
entire SCSI bus gets hosed in some of the failure cases. Sounds like
I need to submit a patch to st.c

:-)

Jeff

Quoted 8 lines

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/