On Sun, 3 Aug 2014, Brian Buhrow wrote:
I'm wondering if you have a particular drive that our driver doesn't like? The message youre seeing suggests the LSI controller is receiving more data than it expects from a particular SCSI target. Unfortunately, the driver isn't telling you which target it is. I'm not sure if that's because the driver doesn't have that data from the LSI controller or because we just didn't write the driver to present it. Having spent a lot of time working on this driver, however, I'm pretty sure the problem you're seeing is between a particular drive and the controller itself. If it works on Linux, it probably means we're setting up the controller differently and it's a matter of figuring out what, exactly is different.
I think you are right. I found some information about scsi problems between EuroStore raids and LSI controllers. The vendor writes "If 16 bytes CDB is to be used, user must make sure U320 IU mode is used." The raid is using CDB16.
The tested Linux has IU active and the raid works well. So I think, that could really be the problem. In the netbsd documentation I found the following:
Not every device obeys the SCSI specification as faithfully as it should. As such devices are discovered by the NetBSD Project, their names are added to a quirk list compiled into the scsi driver along a list of flags indicating which particular bad behaviors the device exhibits (and that the driver should be prepared to work around).
Can I force 16 byte CDB with U320 IU by using the quirk list? Can anyone create a patch?
Thank you for your efforts Regards Uwe