On Fri, Jun 06, 2003 at 11:11:49AM -0500, Kevin Corry wrote: > Where is your dev tree located? people.sistina.com/~thornber/internal-patches/raid1/ > So are you saying it would be alright to remove a suspended device that has no > pending I/O or isn't open? If so, the current code (in 2.5.70) doesn't seem > to coordinate the removal of such a device with another process trying to > open it or submit new I/O. Some new locking of the device would be necessary > to prevent a device which is being removed from being opened at the same > time. I'll have another look at the code, but I'm pretty certain that the reference counting of the mapped_device object takes care of all this. - Joe