RE: Timeout values for SCSI requests

[julian_satran@il.ibm.com]

Usually timeouts are set and enforced by the (class) driver at initiator.

Julo

"Lakshmi Ramasubramanian" <nramas@windows.microsoft.com> on 28-06-2001
21:13:13

Please respond to "Lakshmi Ramasubramanian" <nramas@windows.microsoft.com>

To:   Julian Satran/Haifa/IBM@IBMIL
cc:
Subject:  RE: Timeout values for SCSI requests




Julian,

I agree transport timeouts should be fixed based on round trip delays,
etc.

But I was refering to the timeout value to be set by the target driver
when
it sends the scsi command to the device - I mean, the timer starts
ticking
after the scsi layer on the target has sent the command to the device.
As you
know this'd depend on the device type, command, etc.

Thanks!
 -lakshmi

-----Original Message-----
From: julian_satran@il.ibm.com [mailto:julian_satran@il.ibm.com]
Sent: Thursday, June 28, 2001 6:06 AM
To: Lakshmi Ramasubramanian
Subject: Re: Timeout values for SCSI requests




Lakshmi,

It was felt that execution timeouts are beyond the scope of iSCSI and
transport timeouts should be determined dynamically based on the
measured round trip delays.


Julo

"Lakshmi Ramasubramanian" <nramas@windows.microsoft.com> on 27-06-2001
20:32:05

Please respond to "Lakshmi Ramasubramanian"
<nramas@windows.microsoft.com>

To:   Julian Satran/Haifa/IBM@IBMIL, ips@ece.cmu.edu
cc:
Subject:  Timeout values for SCSI requests




iSCSI structures for SCSI commands do not have fields to indicate
timeout values for various commands. Only CDBs can be sent across. Do we
expect the target driver to determine the required timeout value based
on the CDB Opcode? Or, is there a way for the initiator to explicitly
indicate the timeout values?

Thanks!
 -lakshmi