On Tue, 12 Jan 2010, Oliver Neukum wrote: > Am Montag, 11. Januar 2010 20:37:18 schrieb Alan Stern: > > More to the point, usb_autopm_put_interface() can cause the usage count > > to drop to 0. If the last_busy value was too far in the past, the > > device would be autosuspended immediately. That's not what we want; > > the autosuspend shouldn't occur until after the appropriate delay. And > > the delay should begin at the time the device is no longer being used, > > which is the time when usb_autopm_put_interface() is called. > > But is this the time the device is used for the last time? > It seems to me that the last time a device is used is the last time > IO is done to it. In most cases usb_autopm_put_interface() is called > in close(), which can be a long time after IO was done for the last time. What's wrong with that? If the device has been prevented from autosuspending for "a long time", I don't see any problem with adding on a few extra seconds for the autosuspend delay to expire. We can't depend on the driver to update last_busy when I/O actually finishes, anyway. Or do you think usb_hcd_giveback_urb() should update udev->last_busy every time an URB completes with 0 status? Alan Stern -- 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
