Salyzyn, Mark wrote: > Hannes, I am working the firmware and management software folks to find out how > to mine for this information. In your patch, the controller serial number may be > helpful, but not necessarily compliant? > Well, the information in page 0x80 is 'vendor specific'. So anything we put in there is compliant :-) > Besides persistent device id, what else is gained? The controller ensures persistent > device id for the arrays through the meta-data, only migrating if it happens to be a > foreign (from another controller) array that collides with a native array at the same > ID. Plug the constituent drives into totally different IDs in the SAS infrastructure, > the ID remains the same as presented by the controller. Plug those drives into another > controller, and if not taken by an existing array, the ID will again remain the same. > Yes, that's what I thought would happen. Main reason for adding page 0x80 support is to get it to work with the current persistent device ID setup from udev. There we're using the scsi_id program which doesn't return a serial number if neither page 0x83 nor page 0x80 are supported. So just hacking them into the driver was the easiest way to get it fixed. And there is actually a second reason for adding page 0x80 support: If there were something like a build uuid for the array (either a proper uuid or maybe even the build date) we would be able to detect an array re-initialisation. Currently we won't be able to deal with this as the array will have the same ID, so we would try to mount a drive with a totally different content. That was actually what I was aiming for with that patch, but after two days of hacking I gave up and settled for the easy way out. I know that at least the array build date is somewhere, as the management software displays them :-). So it should be relatively easy to use that as the serial number. > If this is all of your concerns, adding the baggage to the driver > will be redundant. > Actually, not quite. In not supporting EVPD page any compliant software must assume that there is not way of reliably distinguish between two drives from the same vendor/model. And, incidentally, it should be possible to hook in two controllers, and creating two arrays with the same name, right? So we _don't_ have a reliable way for doing so. And this is where EVPD page 0x80 support comes in. Cheers, Hannes -- Dr. Hannes Reinecke hare@xxxxxxx SuSE Linux Products GmbH S390 & zSeries Maxfeldstraße 5 +49 911 74053 688 90409 Nürnberg http://www.suse.de - To unsubscribe from this list: send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
