From: "Rafael J. Wysocki" Subject: Re: [RFC][PATCH update] ACPI / PM: Allow PCI root bridges to wake up the system Date: Tue, 1 Sep 2009 21:03:19 +0200 Message-ID: <200909012103.19761.rjw@sisk.pl> References: <200908300041.14541.rjw@sisk.pl> <200908312324.20850.rjw@sisk.pl> <20090901132547.GA13084@srcf.ucam.org> Mime-Version: 1.0 Content-Type: Text/Plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20090901132547.GA13084@srcf.ucam.org> Sender: linux-pci-owner@vger.kernel.org To: Matthew Garrett Cc: linux-pm@lists.linux-foundation.org, Linux PCI , Henrique de Moraes Holschuh , ACPI Devel Maling List , Jesse Barnes List-Id: linux-acpi@vger.kernel.org On Tuesday 01 September 2009, Matthew Garrett wrote: > On Mon, Aug 31, 2009 at 11:24:20PM +0200, Rafael J. Wysocki wrote: > > > Updated patch is appended, but there's a problem with it. Namely, when the > > root bridge GPE is shared with other devices, those devices are also enabled > > to wake up and they may be kind of "interesting" (on one of my test boxes they > > include the ethernet and audio adapters). I'm not sure what to do about that. > > Perhaps only enable it if a child device is enabled? Well, the problem is that we have to enable the devices that share the GPE to wake up together, but the child device we _really_ want to wake-up need not be any of them. Also, the child device may be enabled to wake up after we've run the "glue" code. Well, perhaps it's better to rework pci_enable_wake() so that it, if the device doesn't have an ACPI handle and is not PCIe, it will go and enable the root bridge to wake up as well? I wonder what about the bridges between the desired wake-up device and the root bridge. I guess they also should be enabled to wake-up, so that they can forward the PME# in case the system is designed this way. Thoughts? > This is difficult with the current code - the refcounted GPE stuff I'm > working on ought to make it easier. Ah, I think that's going to be useful. Thanks, Rafael