On Monday 01 November 2004 13:48, Benjamin Herrenschmidt wrote: > On Mon, 2004-11-01 at 10:18 -0700, David Brownell wrote: > > > But this also highlights one of the reasons for that last sentence. > > The usb-storage driver is literally a bridge between two different > > frameworks. It can know things about how that bridging works, > > things that no generic PM layer can reasonably know. In this > > case, that there's a low power mode for the bridge that doesn't > > need to be visible to SCSI. > > It should just do what IDE does and what I didn't have time to do for > SCSI yet, that is for system PM have a PM request go down the queue > acting as a barrier, ensuring completion of previous pending requests > and blocking any further from coming down. The wakeup is done by > injecting a wakeup request at the head of the queue. 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. - Dave
