On Mon, 2004-11-01 at 16:20 -0700, David Brownell wrote: > Hmm, I'll have to let that grow on me. The notion seems maybe > specific to one style of queue management, and I'm not sure it > makes sense outside of the block layer. Which is close to beeing the only one with a notion of request queues anyway :) This is done that way because we can't afford losing requests on the block layer (discarding them). USB is different, you probably jus want to discard URBs sent to a sleeping HCD there. usb-storage itself probably doesn't care neither, since the whole procedure I described in another mail would ideally be dealt with entirely within the SCSI layer when "sd" gets the suspend call. But yes, to reply to your sentence, this mecanism is specific to the block layer (and the PM request is defined generically in the block layer). Ben.
