Complex Legato Networker Problem with AIX4.3.3 and L700e

Hi,

I've discovered a serious problem with Legato Networker v6.1.3 under
AIX 4.3.3 and the FC-native connected robotic controller of a STK
L700e to an IBM 6228 FC-Adapter:

Everytime the customer used the import/export slot (cap) during
library moves, Legato Networker reports I/O errors. Due to the fact
that everything is well configured an on currrent revision levels, I
did several fibre channel traces with a GTX Protocol Analyzer of that
problem. Every trace shows the same what is:

Initiator (IBM Server) did a Mode Sense, Page 0x1d;
Target (L700e LibController) respond Check Condition, Unit Attention
(0x06), Import or Export Element Accessed (ASC+ASCQ is 0x28+0x01);
Initiator did a Request Sense (0x03 0x00 0x00 0x20 0x00 0x00);
Target respond Check Condition, Illegal Request (0x05), Invalid Field
in CDB
(0x24+0x00) Sense-Key Specific=0xc00003) witch point out Byte 3 in the
Request Sense Command.

The SCSI-specifications tells me that Byte 3 in request sense should
be 0x00, not 0x20, as requested from the initiator.

My problem is now, to decide if that illegal request command comes
from Legatos lus driver or from IBMs FC-driver?

I know that this is a very complex problem, so thank you very much for
any hints.

Rene

"
wrote in message
om...
Hi,

I've discovered a serious problem with Legato Networker v6.1.3 under
AIX 4.3.3 and the FC-native connected robotic controller of a STK
L700e to an IBM 6228 FC-Adapter:

Everytime the customer used the import/export slot (cap) during
library moves, Legato Networker reports I/O errors. Due to the fact
that everything is well configured an on currrent revision levels, I
did several fibre channel traces with a GTX Protocol Analyzer of that
problem. Every trace shows the same what is:

Initiator (IBM Server) did a Mode Sense, Page 0x1d;
Target (L700e LibController) respond Check Condition, Unit Attention
(0x06), Import or Export Element Accessed (ASC+ASCQ is 0x28+0x01);
Initiator did a Request Sense (0x03 0x00 0x00 0x20 0x00 0x00);
Target respond Check Condition, Illegal Request (0x05), Invalid Field
in CDB
(0x24+0x00) Sense-Key Specific=0xc00003) witch point out Byte 3 in the
Request Sense Command.

The SCSI-specifications tells me that Byte 3 in request sense should
be 0x00, not 0x20, as requested from the initiator.

My problem is now, to decide if that illegal request command comes
from Legatos lus driver or from IBMs FC-driver?

I know that this is a very complex problem, so thank you very much for
any hints.

Rene

Do you have any tools other than Legato to address the library? Did Legato
install a custom library driver or are they using SCSI passthrough? Maybe
you can try to generate a check condition with some other tool and see if
the same request sense command follows.

From experience I know that most OSes do an auto-sense on check conditions.
AIX is an exception to the rule in some cases. All recovery including
auto-sense is switched off when using SCSI passthrough. The application is
at that point responsible for retrieving the sense data.

I would not be surprised if this is a Legato issue. Byte 4 is 0x00, which is
unusual as well. It should contain the sense allocation length. I guess the
developer mistyped sense_cdb[3]=0x20; instead of sense_cdb[4]=0x20;.

Rob

Hi,

thank you very much for your response Rob. I also would not be
surprised if this is a Legato issue. Only a simple bug in sense_cdb[3]
and sense_cdb[4]. Well indeed, sense allocation length 0x00 is unusual
but it is allowed and no target should generate an illegal request for
that. Anyway, due to the fact that this L700e library is very critical
for my customer, I only can escalate this problem to Legato, hoping
for a fast solution.
Thank you again for your time to write a well thought out response,
all of your points were excellent and very much appreciated.

René

"Rob Turk" wrote in message ll.nl...
"
wrote in message
om...
Hi,

I've discovered a serious problem with Legato Networker v6.1.3 under
AIX 4.3.3 and the FC-native connected robotic controller of a STK
L700e to an IBM 6228 FC-Adapter:

Everytime the customer used the import/export slot (cap) during
library moves, Legato Networker reports I/O errors. Due to the fact
that everything is well configured an on currrent revision levels, I
did several fibre channel traces with a GTX Protocol Analyzer of that
problem. Every trace shows the same what is:

Initiator (IBM Server) did a Mode Sense, Page 0x1d;
Target (L700e LibController) respond Check Condition, Unit Attention
(0x06), Import or Export Element Accessed (ASC+ASCQ is 0x28+0x01);
Initiator did a Request Sense (0x03 0x00 0x00 0x20 0x00 0x00);
Target respond Check Condition, Illegal Request (0x05), Invalid Field
in CDB
(0x24+0x00) Sense-Key Specific=0xc00003) witch point out Byte 3 in the
Request Sense Command.

The SCSI-specifications tells me that Byte 3 in request sense should
be 0x00, not 0x20, as requested from the initiator.

My problem is now, to decide if that illegal request command comes
from Legatos lus driver or from IBMs FC-driver?

I know that this is a very complex problem, so thank you very much for
any hints.

Rene

Do you have any tools other than Legato to address the library? Did Legato
install a custom library driver or are they using SCSI passthrough? Maybe
you can try to generate a check condition with some other tool and see if
the same request sense command follows.

From experience I know that most OSes do an auto-sense on check conditions.
AIX is an exception to the rule in some cases. All recovery including
auto-sense is switched off when using SCSI passthrough. The application is
at that point responsible for retrieving the sense data.

I would not be surprised if this is a Legato issue. Byte 4 is 0x00, which is
unusual as well. It should contain the sense allocation length. I guess the
developer mistyped sense_cdb[3]=0x20; instead of sense_cdb[4]=0x20;.

Rob