Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S934948AbYCFVaX (ORCPT ); Thu, 6 Mar 2008 16:30:23 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1756663AbYCFVaH (ORCPT ); Thu, 6 Mar 2008 16:30:07 -0500 Received: from ogre.sisk.pl ([217.79.144.158]:38486 "EHLO ogre.sisk.pl" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756303AbYCFVaF (ORCPT ); Thu, 6 Mar 2008 16:30:05 -0500 From: "Rafael J. Wysocki" To: Alan Stern Subject: Re: [RFC][PATCH] PM: Make PM core handle device registrations concurrent with suspend/hibernation Date: Thu, 6 Mar 2008 22:28:27 +0100 User-Agent: KMail/1.9.6 (enterprise 20070904.708012) Cc: pm list , Alexey Starikovskiy , Pavel Machek , LKML 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: <200803062228.28104.rjw@sisk.pl> Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 9462 Lines: 295 On Thursday, 6 of March 2008, Alan Stern wrote: > On Thu, 6 Mar 2008, Rafael J. Wysocki wrote: > > > On Thursday, 6 of March 2008, Alan Stern wrote: > > > On Thu, 6 Mar 2008, Rafael J. Wysocki wrote: > > > > > > > > I thought of one more thing you might want to add: device_pm_add() > > > > > doesn't handle the case where dev->parent is NULL. > > > > > > > > I'm not sure what you mean. > > > > > > > > If dev->parent is NULL, we get into the "successful" branch where the device is > > > > added to dpm_active. Do you think we should add any extra handling of this > > > > case? > > > > > > If a device is registered after dpm_suspend() has returned, the > > > device won't be suspended properly before the system goes to sleep. > > > > > > If the device has a parent then you're okay, because the parent must > > > already be suspended and so device_pm_add() will fail. But if the > > > device doesn't have a parent then device_pm_add() will succeed, which > > > you don't want. > > > > Well, can it happen in practice? If it can, then what way can it happen? > > Yes, it can happen in practice when a new module is loaded. In some > ways, modules' init routines are like probe methods. Hmm, I'm not sure if it really is possible to load a module when all devices have been suspended. Never mind, though. > Maybe it can happen some other ways too, but I don't know of any. > > > It seems to me that we're discussing purely academic stuff here. If it turns > > out to be of any practical relevance, it will be trivial to add the handling of > > it in the future. > > To be safe, I think we should make system sleep mutually exclusive with > module loading. Okay, is the (yet another) version of the patch below fine by you? Rafael --- From: Rafael J. Wysocki Modify the PM core to protect its data structures, specifically the dpm_active list, from being corrupted if a child of the currently suspending device is registered concurrently with its ->suspend() callback. In that case, since the new device (the child) is added to dpm_active after its parent, the PM core will attempt to suspend it after the parent, which is wrong. Introduce a new member of struct dev_pm_info, called 'sleeping', and use it to check if the parent of the device being added to dpm_active has been suspended, in which case the device registration fails. Also, use 'sleeping' for checking if the ordering of devices on dpm_active is correct. Remove pm_sleep_rwsem which is not necessary any more. Special thanks to Alan Stern for discussions and suggestions that lead to the creation of this patch. Signed-off-by: Rafael J. Wysocki --- Documentation/power/devices.txt | 5 +++ drivers/base/core.c | 6 +++- drivers/base/power/main.c | 57 ++++++++++++++++++---------------------- drivers/base/power/power.h | 23 ++-------------- include/linux/pm.h | 1 5 files changed, 40 insertions(+), 52 deletions(-) Index: linux-2.6/include/linux/pm.h =================================================================== --- linux-2.6.orig/include/linux/pm.h +++ linux-2.6/include/linux/pm.h @@ -183,6 +183,7 @@ typedef struct pm_message { struct dev_pm_info { pm_message_t power_state; unsigned can_wakeup:1; + bool sleeping:1; /* Owned by the PM core */ #ifdef CONFIG_PM_SLEEP unsigned should_wakeup:1; struct list_head entry; Index: linux-2.6/drivers/base/power/main.c =================================================================== --- linux-2.6.orig/drivers/base/power/main.c +++ linux-2.6/drivers/base/power/main.c @@ -54,7 +54,8 @@ static LIST_HEAD(dpm_destroy); static DEFINE_MUTEX(dpm_list_mtx); -static DECLARE_RWSEM(pm_sleep_rwsem); +/* 'true' if all devices have been suspended, protected by dpm_list_mtx */ +static bool all_sleeping; int (*platform_enable_wakeup)(struct device *dev, int is_on); @@ -62,14 +63,28 @@ int (*platform_enable_wakeup)(struct dev * device_pm_add - add a device to the list of active devices * @dev: Device to be added to the list */ -void device_pm_add(struct device *dev) +int device_pm_add(struct device *dev) { + int error = 0; + pr_debug("PM: Adding info for %s:%s\n", dev->bus ? dev->bus->name : "No Bus", kobject_name(&dev->kobj)); mutex_lock(&dpm_list_mtx); - list_add_tail(&dev->power.entry, &dpm_active); + if ((dev->parent && dev->parent->power.sleeping) || all_sleeping) { + if (dev->parent->power.sleeping) + dev_warn(dev, + "parent %s is sleeping, will not add\n", + dev->parent->bus_id); + else + dev_warn(dev, "devices are sleeping, will not add\n"); + WARN_ON(true); + error = -EBUSY; + } else { + list_add_tail(&dev->power.entry, &dpm_active); + } mutex_unlock(&dpm_list_mtx); + return error; } /** @@ -107,32 +122,6 @@ void device_pm_schedule_removal(struct d } EXPORT_SYMBOL_GPL(device_pm_schedule_removal); -/** - * pm_sleep_lock - mutual exclusion for registration and suspend - * - * Returns 0 if no suspend is underway and device registration - * may proceed, otherwise -EBUSY. - */ -int pm_sleep_lock(void) -{ - if (down_read_trylock(&pm_sleep_rwsem)) - return 0; - - return -EBUSY; -} - -/** - * pm_sleep_unlock - mutual exclusion for registration and suspend - * - * This routine undoes the effect of device_pm_add_lock - * when a device's registration is complete. - */ -void pm_sleep_unlock(void) -{ - up_read(&pm_sleep_rwsem); -} - - /*------------------------- Resume routines -------------------------*/ /** @@ -242,11 +231,13 @@ static int resume_device(struct device * static void dpm_resume(void) { mutex_lock(&dpm_list_mtx); + all_sleeping = false; while(!list_empty(&dpm_off)) { struct list_head *entry = dpm_off.next; struct device *dev = to_device(entry); list_move_tail(entry, &dpm_active); + dev->power.sleeping = false; mutex_unlock(&dpm_list_mtx); resume_device(dev); mutex_lock(&dpm_list_mtx); @@ -285,7 +276,6 @@ void device_resume(void) might_sleep(); dpm_resume(); unregister_dropped_devices(); - up_write(&pm_sleep_rwsem); } EXPORT_SYMBOL_GPL(device_resume); @@ -426,6 +416,9 @@ static int dpm_suspend(pm_message_t stat struct list_head *entry = dpm_active.prev; struct device *dev = to_device(entry); + WARN_ON(dev->parent && dev->parent->power.sleeping); + + dev->power.sleeping = true; mutex_unlock(&dpm_list_mtx); error = suspend_device(dev, state); mutex_lock(&dpm_list_mtx); @@ -437,11 +430,14 @@ static int dpm_suspend(pm_message_t stat (error == -EAGAIN ? " (please convert to suspend_late)" : "")); + dev->power.sleeping = false; break; } if (!list_empty(&dev->power.entry)) list_move(&dev->power.entry, &dpm_off); } + if (!error) + all_sleeping = true; mutex_unlock(&dpm_list_mtx); return error; @@ -459,7 +455,6 @@ int device_suspend(pm_message_t state) int error; might_sleep(); - down_write(&pm_sleep_rwsem); error = dpm_suspend(state); if (error) device_resume(); Index: linux-2.6/drivers/base/power/power.h =================================================================== --- linux-2.6.orig/drivers/base/power/power.h +++ linux-2.6/drivers/base/power/power.h @@ -11,30 +11,13 @@ static inline struct device *to_device(s return container_of(entry, struct device, power.entry); } -extern void device_pm_add(struct device *); +extern int device_pm_add(struct device *); extern void device_pm_remove(struct device *); -extern int pm_sleep_lock(void); -extern void pm_sleep_unlock(void); #else /* CONFIG_PM_SLEEP */ - -static inline void device_pm_add(struct device *dev) -{ -} - -static inline void device_pm_remove(struct device *dev) -{ -} - -static inline int pm_sleep_lock(void) -{ - return 0; -} - -static inline void pm_sleep_unlock(void) -{ -} +static inline int device_pm_add(struct device *dev) { return 0; } +static inline void device_pm_remove(struct device *dev) {} #endif Index: linux-2.6/drivers/base/core.c =================================================================== --- linux-2.6.orig/drivers/base/core.c +++ linux-2.6/drivers/base/core.c @@ -815,7 +815,11 @@ int device_add(struct device *dev) error = dpm_sysfs_add(dev); if (error) goto PMError; - device_pm_add(dev); + error = device_pm_add(dev); + if (error) { + dpm_sysfs_remove(dev); + goto PMError; + } error = bus_add_device(dev); if (error) goto BusError; Index: linux-2.6/Documentation/power/devices.txt =================================================================== --- linux-2.6.orig/Documentation/power/devices.txt +++ linux-2.6/Documentation/power/devices.txt @@ -196,6 +196,11 @@ its parent; and can't be removed or susp The policy is that the device tree should match hardware bus topology. (Or at least the control bus, for devices which use multiple busses.) +In particular, this means that a device registration may fail if the parent of +the device is suspending (ie. has been chosen by the PM core as the next +device to suspend) or has already suspended, as well as after all of the other +devices have been suspended. Device drivers must be prepared to cope with such +situations. Suspending Devices -- 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/