On Sat, 2004-11-06 at 15:12 -0500, Alan Stern wrote: > This would have to be done extremely carefully. The model used in the USB > subsystem is that when a device is disconnected it goes away completely. > Even if it reconnects, the code doesn't recognize it as the same device -- > it's treated as a totally new device. This is particularly relevant for > USB mass storage devices; a disconnect followed by reconnect would break > mounted filesystems, for example. Ok. On a totally separate issue, aren't there some kind of unique IDs or serial number on mass storage to avoid just that ? I know ieee1394 does, though I don't think linux can deal with that properly (MacOS can afaik). > Unfortunately UHCI controllers aren't very smart. They're either running > or stopped, and whenever they run they do DMA. If some devices disconnect > when they find the bus is suspended, then they'll be gone when we resume > from the FREEZE. Nothing we can do about it. Well, that isn't too bad, they wouldn't have supported suspend anyway, and it's hopeful that devices that are neccessary for wakeup, that is that do wakeup events, are also those who can suspend ;) Ben.
