This note is just to let PSARC-ext know that I have talked with Jerry and we have concluded that this fast track will have no adverse affect on Solaris Cluster. From a cluster perspective, I have no objection to this case. I am adding my interpretations of the answers to the questions below.
Charles DeBardeleben wrote: > Two more questions, and a comment. > > The Solaris Cluster did driver is both a consumer and calls > ddi_devid_register. > The device id that we register is related to the device id we consume. > > Question 1. Does this mean that we need the new property? The intent > is for > an instance of did to always have the same device id. No need to add the property. If you do not provide the property in driver.conf, then this case will not change device ID cache behavior. You get the old behavior. > > Question 2. If the disk under us moves, how do we discover this so > that we can fix up our > layer open to get the correct disk under us. Solaris Cluster currently has manual disk management procedures that have been tested. These remain the same with this change unless Solaris Cluster adds ddi-devid-registrant to our device driver. If Solaris Cluster adds this property to the device driver, we may need to review our procedures. -Charles > > Comment on virtual disk driver. Solaris cluster requires that anything > that looks > like a disk or tape driver to have a device id. We require that vdc > provide a device id. > Another thing, we require that the device id not change if the disk or > virtual disk is still > the same. So if the device id presented is a fabricated device id, > then it should be > stored persistently so that it is the same after reboot. > > -Charles > > Jerry Gilliam wrote: >> >>>> Immediate consumers (sd, ssd, cmdk, and dad target drivers) are all >>>> in ON. There are no contracts needed at this time. I am seeking >>>> patch binding. >>>> >>> There are two other drivers that use ddi_devid_register: >>> sata and vdc (ldom's virtual disk client) - should they be listed >>> here ? >>> >>> >> >> It wasn't apparent to me that vdc needs to provide this property >> for its encapsulated devid. This is either an existing bug or >> there is no requirement. I'll follow up on this. >> >> And with Prasad's work put back now, yes sata is now a devid >> registrant. >> >> >> >>> What happens if a driver uses ddi_devid_register but the corresponding >>> driver.conf is not updated to have this flag ? >>> >>> >> >> If the cache were to become damaged or the disk moved, the cache >> auto-rebuild mechanism would not include trying to configure that >> driver to re-discover the disk. >> >> >> -jg >> >> >
