Thanks Oliver Neukum!You solved lots of puzzle in my mind!! Some further questions: Q5: I still can't fully understand how autosuspension saving power. If a device driver only saves the device state and do not fill any device's physical registering to communicate with device, (hi, please sleep), then suspend() return ok . I guess the HW device should continue to consume power as normarlly, since HW device knows nothing about the suspension request from host system. Is it mean there are some physical power supply change in USB bus for autosuspension callback? Or autosuspend means a special USB package broadcast in USB bus after suspend() callback return, USB device who received it must do some power saving work per USB SPEC? Q6: Does it mean the most power-optimal usb system is: everything marked it can be autosuspend, including hubs and usb devices? there isn't any squeeze potential for usb power any more? Q7: A more specific question is: If a device running on Linux is already enabled autosuspend and idle now, while it is physically consumes more power than running on Windows when idle. Is there anything known issue can cause this power difference? Is there any know advantage for USB windows power management, e.g for more mechnism than autosuspend? Thanks again! Samuel 2010/6/30 Oliver Neukum <oneukum@xxxxxxx>: > Am Mittwoch, 30. Juni 2010, 11:04:58 schrieb Samuel Xu: >> Appreciate any USB power expert can help me on those Linux USB power >> management puzzles: >> >> Q1: I found some USB material mentioned : Relationship between ACPI >> Dx(e.g. D0/D1/D3) states and USB PM states (active/suspended) is >> orthogonal. > > Almost. The root hub must be in D0 unless it is suspended. But the root > hub can be suspended only if all devices on the bus are suspended. > >> Suspend/resume might not effect device Dx state. Is it a correct >> statement for general usb device? > > Yes. > >> Q2: How to tell USB device’s ACPI Dx state. It seems lsusb can’t tell >> us those info. (lspci works for PCI device’s Dx state) > > USB devices don't have Dx states. Therefore you can't query them. > >> Q3: How to tell USB device’s suspension state? > > You don't. You can just query whether the device should autosuspend itself > in /sys/bus/usb/devices/$DEVICE/power > >> will any query via urb >> will cause resume of USB device? > > Yes. > >> Q4: should USB device driver response to do some device-specific power >> action (e.g. device register writing) to put a specific USB device >> into low power state when responding to suspend action? > > No. A driver should implement suspend()/resume() hooks that > a) save device state if necessary > b) make it safe to go into a suspended state > c) cease communication > (in suspend() ) > > Please clarify the question. > >> Q5: If Q4 is Yes, should device vendor respond for those >> device-specific code? will Linux kernel code typically incorporate >> those device-specific code? > > There is nothing necessarily device specific in suspend()/resume() > Please clarify the question. > > HTH > Oliver > -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
