From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757721AbZGQDIJ (ORCPT ); Thu, 16 Jul 2009 23:08:09 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1757711AbZGQDII (ORCPT ); Thu, 16 Jul 2009 23:08:08 -0400 Received: from mga09.intel.com ([134.134.136.24]:40980 "EHLO mga09.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751979AbZGQDIH (ORCPT ); Thu, 16 Jul 2009 23:08:07 -0400 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="4.42,415,1243839600"; d="scan'208";a="430960473" Subject: Re: [PATCH 8/8] i8042: support device async suspend & shutdown From: Zhang Rui To: "Rafael J. Wysocki" Cc: Pavel Machek , Linux Kernel Mailing List , linux-pm , linux-acpi , Len Brown , Arjan van de Ven In-Reply-To: <200907170440.16678.rjw@sisk.pl> References: <1247643522.26272.83.camel@rzhang-dt> <20090715003716.GE2360@ucw.cz> <1247795488.26272.227.camel@rzhang-dt> <200907170440.16678.rjw@sisk.pl> Content-Type: text/plain Date: Fri, 17 Jul 2009 11:08:57 +0800 Message-Id: <1247800137.26272.297.camel@rzhang-dt> Mime-Version: 1.0 X-Mailer: Evolution 2.22.1 (2.22.1-2.fc9) Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Fri, 2009-07-17 at 10:40 +0800, Rafael J. Wysocki wrote: > On Friday 17 July 2009, Zhang Rui wrote: > > On Wed, 2009-07-15 at 08:37 +0800, Pavel Machek wrote: > > > On Wed 2009-07-15 15:38:42, Zhang Rui wrote: > > > > > > > > i8042 controller support device async actions. > > > > > > > > If boot option "dev_async_action" is added, > > > > i8042 controller and its child devices can be > > > > suspended/resumed/shutdown asynchronously. > > > > > > > > Signed-off-by: Zhang Rui > > > > --- > > > > drivers/input/serio/i8042.c | 10 +++++++++- > > > > 1 file changed, 9 insertions(+), 1 deletion(-) > > > > > > > > Index: linux-2.6/drivers/input/serio/i8042.c > > > > =================================================================== > > > > --- linux-2.6.orig/drivers/input/serio/i8042.c > > > > +++ linux-2.6/drivers/input/serio/i8042.c > > > > @@ -1284,14 +1284,21 @@ static int __init i8042_init(void) > > > > goto err_unregister_driver; > > > > } > > > > > > > > - err = platform_device_add(i8042_platform_device); > > > > + err = dev_async_register(&i8042_platform_device->dev, > > > > + DEV_ASYNC_SUSPEND | DEV_ASYNC_SHUTDOWN); > > > > if (err) > > > > goto err_free_device; > > > > > > > > + err = platform_device_add(i8042_platform_device); > > > > + if (err) > > > > + goto err_dev_async_unregister; > > > > + > > > > panic_blink = i8042_panic_blink; > > > > > > > > return 0; > > > > > > > > + err_dev_async_unregister: > > > > + dev_async_unregister(&i8042_platform_device->dev); > > > > > > > > > > err_free_device: > > > > platform_device_put(i8042_platform_device); > > > > err_unregister_driver: > > > > @@ -1304,6 +1311,7 @@ static int __init i8042_init(void) > > > > > > > > static void __exit i8042_exit(void) > > > > { > > > > + dev_async_unregister(&i8042_platform_device->dev); > > > > platform_device_unregister(i8042_platform_device); > > > > platform_driver_unregister(&i8042_driver); > > > > i8042_platform_exit(); > > > > > > Could we get the core to do async_unregister during device_unregister? > > > > > yes, we can. > > I think it's a little strange that drivers need to invoke > > dev_async_register to register an async device group, but don't need to > > call dev_async_unregister to unregister it. > > > > > Or maybe better, just add flags during device_add? > > > > > Many devices don't know if they support async actions or not during > > device initialization time, like PCI and ACPI devices. > > If we want to create an async group for a PCI device, we have to call > > dev_async_register in the specified PCI device driver, rather than in > > the PCI bus scan stage. > > That is quite a big design issue, IMO. > > Did you consider any other approach to this? > not yet. Just like the async boot approach, this async action mechanism is not designed for all the devices. Plus different device actions should be handled differently. E.g device suspend and shutdown are more complicate than resume case. Because in resume case, child devices are ALWAYS resumed after its parent device, so it's easy to register an DEV_ASYNC_RESUME group. But if we want to register an async device group that supports DEV_ASYNC_SUSPEND/DEV_ASYNC_SHUTDOWN, we must make sure that the device groups is something like a Closure, i.e. they don't depend on ANY OTHER devices. Anyway, IMO, if we want to register a new async device group, we should do it case by case. thanks, rui