| Message ID | 1562298.ZjRvhqQzT7@vostro.rjw.lan (mailing list archive) |
|---|---|
| State | Accepted, archived |
| Headers | show |
On Thu, 2013-08-29 at 23:18 +0200, Rafael J. Wysocki wrote: > From: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > > Since all of the memory hotplug operations have to be carried out > under device_hotplug_lock, they won't need to acquire pm_mutex if > device_hotplug_lock is held around hibernation. > > For this reason, make the hibernation code acquire > device_hotplug_lock after freezing user space processes and > release it before thawing them. At the same tim drop the > lock_system_sleep() and unlock_system_sleep() calls from > lock_memory_hotplug() and unlock_memory_hotplug(), respectively. > > Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > --- > kernel/power/hibernate.c | 4 ++++ > kernel/power/user.c | 2 ++ > mm/memory_hotplug.c | 4 ---- > 3 files changed, 6 insertions(+), 4 deletions(-) > > Index: linux-pm/kernel/power/hibernate.c > =================================================================== > --- linux-pm.orig/kernel/power/hibernate.c > +++ linux-pm/kernel/power/hibernate.c > @@ -652,6 +652,7 @@ int hibernate(void) > if (error) > goto Exit; > > + lock_device_hotplug(); Since hibernate() can be called from sysfs, do you think the tool may see this as a circular dependency with p_active again? This shouldn't be a problem in practice, though. Thanks, -Toshi -- To unsubscribe from this list: send the line "unsubscribe linux-pm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Sat, 2013-08-31 at 02:39 +0200, Rafael J. Wysocki wrote: > On Friday, August 30, 2013 06:23:19 PM Toshi Kani wrote: > > On Thu, 2013-08-29 at 23:18 +0200, Rafael J. Wysocki wrote: > > > From: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > > > > > > Since all of the memory hotplug operations have to be carried out > > > under device_hotplug_lock, they won't need to acquire pm_mutex if > > > device_hotplug_lock is held around hibernation. > > > > > > For this reason, make the hibernation code acquire > > > device_hotplug_lock after freezing user space processes and > > > release it before thawing them. At the same tim drop the > > > lock_system_sleep() and unlock_system_sleep() calls from > > > lock_memory_hotplug() and unlock_memory_hotplug(), respectively. > > > > > > Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > > > --- > > > kernel/power/hibernate.c | 4 ++++ > > > kernel/power/user.c | 2 ++ > > > mm/memory_hotplug.c | 4 ---- > > > 3 files changed, 6 insertions(+), 4 deletions(-) > > > > > > Index: linux-pm/kernel/power/hibernate.c > > > =================================================================== > > > --- linux-pm.orig/kernel/power/hibernate.c > > > +++ linux-pm/kernel/power/hibernate.c > > > @@ -652,6 +652,7 @@ int hibernate(void) > > > if (error) > > > goto Exit; > > > > > > + lock_device_hotplug(); > > > > Since hibernate() can be called from sysfs, do you think the tool may > > see this as a circular dependency with p_active again? This shouldn't > > be a problem in practice, though. > > /sys/power/state isn't a device attribute even and is never removed, so it > would be very sad and disappointing if lockdep reported that as a circular > dependency. The deadlock is surely not possible here anyway. Agreed. The code looks good otherwise, and this is a nice cleanup. If it is OK to ignore the possible warning from the tool (which I do not know the rule here), feel free to add my ack to patch 2/3 and 3/3 as well. Thanks, -Toshi -- To unsubscribe from this list: send the line "unsubscribe linux-pm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Friday, August 30, 2013 06:23:19 PM Toshi Kani wrote: > On Thu, 2013-08-29 at 23:18 +0200, Rafael J. Wysocki wrote: > > From: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > > > > Since all of the memory hotplug operations have to be carried out > > under device_hotplug_lock, they won't need to acquire pm_mutex if > > device_hotplug_lock is held around hibernation. > > > > For this reason, make the hibernation code acquire > > device_hotplug_lock after freezing user space processes and > > release it before thawing them. At the same tim drop the > > lock_system_sleep() and unlock_system_sleep() calls from > > lock_memory_hotplug() and unlock_memory_hotplug(), respectively. > > > > Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > > --- > > kernel/power/hibernate.c | 4 ++++ > > kernel/power/user.c | 2 ++ > > mm/memory_hotplug.c | 4 ---- > > 3 files changed, 6 insertions(+), 4 deletions(-) > > > > Index: linux-pm/kernel/power/hibernate.c > > =================================================================== > > --- linux-pm.orig/kernel/power/hibernate.c > > +++ linux-pm/kernel/power/hibernate.c > > @@ -652,6 +652,7 @@ int hibernate(void) > > if (error) > > goto Exit; > > > > + lock_device_hotplug(); > > Since hibernate() can be called from sysfs, do you think the tool may > see this as a circular dependency with p_active again? This shouldn't > be a problem in practice, though. /sys/power/state isn't a device attribute even and is never removed, so it would be very sad and disappointing if lockdep reported that as a circular dependency. The deadlock is surely not possible here anyway. Thanks, Rafael -- To unsubscribe from this list: send the line "unsubscribe linux-pm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Friday, August 30, 2013 06:35:12 PM Toshi Kani wrote: > On Sat, 2013-08-31 at 02:39 +0200, Rafael J. Wysocki wrote: > > On Friday, August 30, 2013 06:23:19 PM Toshi Kani wrote: > > > On Thu, 2013-08-29 at 23:18 +0200, Rafael J. Wysocki wrote: > > > > From: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > > > > > > > > Since all of the memory hotplug operations have to be carried out > > > > under device_hotplug_lock, they won't need to acquire pm_mutex if > > > > device_hotplug_lock is held around hibernation. > > > > > > > > For this reason, make the hibernation code acquire > > > > device_hotplug_lock after freezing user space processes and > > > > release it before thawing them. At the same tim drop the > > > > lock_system_sleep() and unlock_system_sleep() calls from > > > > lock_memory_hotplug() and unlock_memory_hotplug(), respectively. > > > > > > > > Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> > > > > --- > > > > kernel/power/hibernate.c | 4 ++++ > > > > kernel/power/user.c | 2 ++ > > > > mm/memory_hotplug.c | 4 ---- > > > > 3 files changed, 6 insertions(+), 4 deletions(-) > > > > > > > > Index: linux-pm/kernel/power/hibernate.c > > > > =================================================================== > > > > --- linux-pm.orig/kernel/power/hibernate.c > > > > +++ linux-pm/kernel/power/hibernate.c > > > > @@ -652,6 +652,7 @@ int hibernate(void) > > > > if (error) > > > > goto Exit; > > > > > > > > + lock_device_hotplug(); > > > > > > Since hibernate() can be called from sysfs, do you think the tool may > > > see this as a circular dependency with p_active again? This shouldn't > > > be a problem in practice, though. > > > > /sys/power/state isn't a device attribute even and is never removed, so it > > would be very sad and disappointing if lockdep reported that as a circular > > dependency. The deadlock is surely not possible here anyway. > > Agreed. The code looks good otherwise, and this is a nice cleanup. If > it is OK to ignore the possible warning from the tool (which I do not > know the rule here), Well, if it complains, we'll just need to add some annotations to this. The code is correct I believe. > feel free to add my ack to patch 2/3 and 3/3 as well. I will, thanks!
Index: linux-pm/kernel/power/hibernate.c =================================================================== --- linux-pm.orig/kernel/power/hibernate.c +++ linux-pm/kernel/power/hibernate.c @@ -652,6 +652,7 @@ int hibernate(void) if (error) goto Exit; + lock_device_hotplug(); /* Allocate memory management structures */ error = create_basic_memory_bitmaps(); if (error) @@ -685,6 +686,7 @@ int hibernate(void) Free_bitmaps: free_basic_memory_bitmaps(); Thaw: + unlock_device_hotplug(); thaw_processes(); /* Don't bother checking whether freezer_test_done is true */ @@ -814,6 +816,7 @@ static int software_resume(void) pr_debug("PM: Loading hibernation image.\n"); + lock_device_hotplug(); error = create_basic_memory_bitmaps(); if (error) goto Thaw; @@ -827,6 +830,7 @@ static int software_resume(void) swsusp_free(); free_basic_memory_bitmaps(); Thaw: + unlock_device_hotplug(); thaw_processes(); Finish: pm_notifier_call_chain(PM_POST_RESTORE); Index: linux-pm/kernel/power/user.c =================================================================== --- linux-pm.orig/kernel/power/user.c +++ linux-pm/kernel/power/user.c @@ -201,6 +201,7 @@ static long snapshot_ioctl(struct file * if (!mutex_trylock(&pm_mutex)) return -EBUSY; + lock_device_hotplug(); data = filp->private_data; switch (cmd) { @@ -373,6 +374,7 @@ static long snapshot_ioctl(struct file * } + unlock_device_hotplug(); mutex_unlock(&pm_mutex); return error; Index: linux-pm/mm/memory_hotplug.c =================================================================== --- linux-pm.orig/mm/memory_hotplug.c +++ linux-pm/mm/memory_hotplug.c @@ -51,14 +51,10 @@ DEFINE_MUTEX(mem_hotplug_mutex); void lock_memory_hotplug(void) { mutex_lock(&mem_hotplug_mutex); - - /* for exclusive hibernation if CONFIG_HIBERNATION=y */ - lock_system_sleep(); } void unlock_memory_hotplug(void) { - unlock_system_sleep(); mutex_unlock(&mem_hotplug_mutex); }