On Fri, 11 Feb 2011, Oliver Neukum wrote: > Hi, > > I am looking at IO queued to be executed upon resumption failing. Do you mean that the I/O fails or that the resume fails? I'll assume the former. > However usb_autopm_get_interface_async() was already called for > it. How can I balance the PM counters from within resume()? Let's see. I/O was submitted while the device was suspended, and usb_autopm_get_interface_async() was called. The resume routine starts the I/O and encounters a failure. Since the I/O will never complete, you want the device to be suspended again as soon as possible. Right? The correct approach is to call usb_autopm_put_interface_no_suspend(). The resume will complete normally, and then the PM core will see that the usage_count is 0, so it will send an asynchronous idle notification which will start an autosuspend. usb_autopm_put_interface() and usb_autopm_put_interface_async() should work too, but they will expend some futile extra effort (they will try to send an immediate idle notification, which will fail because the interface is in the wrong state). 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
