Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1758303AbYCYUfm (ORCPT ); Tue, 25 Mar 2008 16:35:42 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1755003AbYCYUfe (ORCPT ); Tue, 25 Mar 2008 16:35:34 -0400 Received: from ogre.sisk.pl ([217.79.144.158]:34632 "EHLO ogre.sisk.pl" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754171AbYCYUfb (ORCPT ); Tue, 25 Mar 2008 16:35:31 -0400 From: "Rafael J. Wysocki" To: Alan Stern Subject: Re: [RFC][PATCH] PM: Introduce new top level suspend and hibernation callbacks (rev. 3) Date: Tue, 25 Mar 2008 21:35:02 +0100 User-Agent: KMail/1.9.6 (enterprise 20070904.708012) Cc: pm list , ACPI Devel Maling List , Greg KH , Len Brown , LKML , Alexey Starikovskiy , David Brownell , Pavel Machek , Benjamin Herrenschmidt References: In-Reply-To: MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200803252135.04247.rjw@sisk.pl> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 3582 Lines: 99 On Tuesday, 25 of March 2008, Alan Stern wrote: > On Mon, 24 Mar 2008, Rafael J. Wysocki wrote: > > > > Can we also have a DPM_PREPARING state, set when ->prepare() is about > > > to be called? The PM core wouldn't make use of it but some drivers > > > would. (I can't think of any use at all for the analogous > > > DPM_COMPLETING state, however.) > > > > Hmm. dev->power.status is protected by dpm_list_mtx. Do you think it would be > > useful to have an accessor function for reading it under the lock? > > I don't think so. What I have in mind is situations where there > accessed has already been synchronized by other means, while the > prepare() method is running. For example: > > Task 0 Task 1 > ------ ------ > ->prepare() is called > Waits for currently-running > registration in task 1 > to finish > Does other stuff > Receives a request to register > a new child under dev > Sees that dev->power.state is > still DPM_ON, so goes ahead > with the child's registration > ->prepare() returns > dev->power.state is set to > DPM_SUSPENDING > device_pm_add() checks > dev->power.state and fails > the registration > > If dev->power.state had been set to DPM_PREPARING before ->prepare() > was called, then task 1 would have avoided trying to register the > child. > > > > > + dev->power.status = DPM_RESUMING; > > > > + get_device(dev); > > > > + mutex_unlock(&dpm_list_mtx); > > > > + > > > > + resume_device(dev, state); > > > > + > > > > + mutex_lock(&dpm_list_mtx); > > > > + put_device(dev); > > > > + } > > > > + if (!list_empty(&dev->power.entry)) > > > > + list_move_tail(&dev->power.entry, &list); > > > > > > A little problem here: You refer to dev after calling put_device(). > > > > The device can't be removed at this point, because we hold dpm_list_mtx, which > > is needed by device_del(). > > True, it can't be removed at this point. But it _can_ be removed > between the calls to resume_device() and mutex_lock(). > > > > > } > > > > - if (!error) > > > > - all_sleeping = true; > > > > + list_splice(&list, &dpm_list); > > > > > > Instead you could eliminate the list_splice_init() above and put here: > > > > > > list_splice(&list, dpm_list->prev); > > > > > > This will move the entries from list to the end of dpm_list. > > > > dpm_list may be empty at this point. Wouldn't that cause any trouble? > > It will still work correctly. If dpm_list is empty then dpm_list->prev > is equal to &dpm_list, so it will do the same thing as your current > code does. > > > I just thought of another problem. At the point where > local_irq_disable() is called, in between device_suspend() and > device_power_down(), it is possible in a preemptible kernel that > another task is holding dpm_list_mtx and is in the middle of updating > the list pointers. This would mess up the traversal in > device_power_down(). > > I'm not sure about the best way to prevent this. Is it legal to call > unlock_mutex() while interrupts or preemption are disabled? Well, I think it is, but I'm not sure how that can help. To prevent the race from happening, we can lock dpm_list_mtx before switching interrupts off in kernel/power/main.c:suspend_enter() and analogously in kernel/power/disk.c . Thanks, Rafael -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/