[PATCH 0/2] Protect cpu map updates from ACPI processor

[PATCH 0/2] Protect cpu map updates from ACPI processor

[Toshi Kani]

CPU system maps are protected with reader/writer locks.  The reader
lock, put_online_cpus(), assures that the maps are not updated while
holding the lock.  The writer lock, cpu_hotplug_begin(), is used to
udpate the cpu maps along with cpu_maps_update_begin().

However, the ACPI processor handler updates the cpu maps without
holding the the writer lock.  This patchset fixes this problem.

---
The patchset is based on linux-next of the pm tree.

---
Toshi Kani (2):
 CPU hotplug: Export cpu_hotplug_begin() interface
 ACPI: Acquire writer lock to update cpu maps

---
 drivers/acpi/acpi_processor.c | 21 ++++++++++++++++-----
 include/linux/cpu.h           |  2 ++
 kernel/cpu.c                  |  4 ++--
 3 files changed, 20 insertions(+), 7 deletions(-)

[PATCH 1/2] CPU hotplug: Export cpu_hotplug_begin() interface

[Toshi Kani]

cpu_hotplug_begin() and cpu_hotplug_done() are the interfaces to
acquire the writer lock of the cpu_hotplug.lock.  The interfaces are
used along with cpu_maps_update_begin() / cpu_maps_update_done() to
serialize the updates to the system cpu maps, such as cpu_online_mask
and cpu_present_mask.

cpu_present_mask is updated by the ACPI processor handler.  Therefore,
this patch changes the interfaces to global, and exports them in cpu.h.

Signed-off-by: Toshi Kani <toshi.kani@hp.com>
---
 include/linux/cpu.h |    4 ++++
 kernel/cpu.c        |    9 +++------
 2 files changed, 7 insertions(+), 6 deletions(-)

diff --git a/include/linux/cpu.h b/include/linux/cpu.h
index ab0eade..956c0a1 100644
--- a/include/linux/cpu.h
+++ b/include/linux/cpu.h
@@ -172,6 +172,8 @@ extern struct bus_type cpu_subsys;
 #ifdef CONFIG_HOTPLUG_CPU
 /* Stop CPUs going up and down. */
 
+extern void cpu_hotplug_begin(void);
+extern void cpu_hotplug_done(void);
 extern void get_online_cpus(void);
 extern void put_online_cpus(void);
 extern void cpu_hotplug_disable(void);
@@ -197,6 +199,8 @@ static inline void cpu_hotplug_driver_unlock(void)
 
 #else		/* CONFIG_HOTPLUG_CPU */
 
+static inline void cpu_hotplug_begin(void) {}
+static inline void cpu_hotplug_done(void) {}
 #define get_online_cpus()	do { } while (0)
 #define put_online_cpus()	do { } while (0)
 #define cpu_hotplug_disable()	do { } while (0)
diff --git a/kernel/cpu.c b/kernel/cpu.c
index b2b227b..d7f07a2 100644
--- a/kernel/cpu.c
+++ b/kernel/cpu.c
@@ -113,7 +113,7 @@ EXPORT_SYMBOL_GPL(put_online_cpus);
  * get_online_cpus() not an api which is called all that often.
  *
  */
-static void cpu_hotplug_begin(void)
+void cpu_hotplug_begin(void)
 {
 	cpu_hotplug.active_writer = current;
 
@@ -127,7 +127,7 @@ static void cpu_hotplug_begin(void)
 	}
 }
 
-static void cpu_hotplug_done(void)
+void cpu_hotplug_done(void)
 {
 	cpu_hotplug.active_writer = NULL;
 	mutex_unlock(&cpu_hotplug.lock);
@@ -154,10 +154,7 @@ void cpu_hotplug_enable(void)
 	cpu_maps_update_done();
 }
 
-#else /* #if CONFIG_HOTPLUG_CPU */
-static void cpu_hotplug_begin(void) {}
-static void cpu_hotplug_done(void) {}
-#endif	/* #else #if CONFIG_HOTPLUG_CPU */
+#endif	/* CONFIG_HOTPLUG_CPU */
 
 /* Need to know about CPUs going up/down? */
 int __ref register_cpu_notifier(struct notifier_block *nb)

[PATCH 2/2] ACPI: Acquire writer lock to update cpu maps

[Toshi Kani]

acpi_map_lsapic() is called from acpi_processor_hotadd_init() to update
cpu_possible_mask and cpu_present_mask.  acpi_unmap_lsapic() is called
from acpi_processor_remove() to update cpu_possible_mask.  Currently,
they are either unprotected or protected with the reader lock, which is
not correct.

This patch protects them with the writer lock with cpu_hotplug_begin()
along with cpu_maps_update_begin(), which must be held before calling
cpu_hotplug_begin().  It also protects arch_register_cpu() /
arch_unregister_cpu(), which creates / deletes a sysfs cpu device
interface.

Signed-off-by: Toshi Kani <toshi.kani@hp.com>
---
 drivers/acpi/acpi_processor.c |   21 ++++++++++++++++-----
 1 file changed, 16 insertions(+), 5 deletions(-)

diff --git a/drivers/acpi/acpi_processor.c b/drivers/acpi/acpi_processor.c
index 5a74a9c..f29e06e 100644
--- a/drivers/acpi/acpi_processor.c
+++ b/drivers/acpi/acpi_processor.c
@@ -178,14 +178,17 @@ static int acpi_processor_hotadd_init(struct acpi_processor *pr)
 	if (ACPI_FAILURE(status) || !(sta & ACPI_STA_DEVICE_PRESENT))
 		return -ENODEV;
 
+	cpu_maps_update_begin();
+	cpu_hotplug_begin();
+
 	ret = acpi_map_lsapic(pr->handle, &pr->id);
 	if (ret)
-		return ret;
+		goto out;
 
 	ret = arch_register_cpu(pr->id);
 	if (ret) {
 		acpi_unmap_lsapic(pr->id);
-		return ret;
+		goto out;
 	}
 
 	/*
@@ -195,7 +198,11 @@ static int acpi_processor_hotadd_init(struct acpi_processor *pr)
 	 */
 	pr_info("CPU%d has been hot-added\n", pr->id);
 	pr->flags.need_hotplug_init = 1;
-	return 0;
+
+out:
+	cpu_hotplug_done();
+	cpu_maps_update_done();
+	return ret;
 }
 #else
 static inline int acpi_processor_hotadd_init(struct acpi_processor *pr)
@@ -452,11 +459,15 @@ static void acpi_processor_remove(struct acpi_device *device)
 	per_cpu(processor_device_array, pr->id) = NULL;
 	per_cpu(processors, pr->id) = NULL;
 
+	cpu_maps_update_begin();
+	cpu_hotplug_begin();
+
 	/* Remove the CPU. */
-	get_online_cpus();
 	arch_unregister_cpu(pr->id);
 	acpi_unmap_lsapic(pr->id);
-	put_online_cpus();
+
+	cpu_hotplug_done();
+	cpu_maps_update_done();
 
 	try_offline_node(cpu_to_node(pr->id));
 

Re: [PATCH 0/2] Protect cpu map updates from ACPI processor

[Rafael J. Wysocki]

On Monday, August 12, 2013 09:45:52 AM Toshi Kani wrote:
> CPU system maps are protected with reader/writer locks.  The reader
> lock, put_online_cpus(), assures that the maps are not updated while
> holding the lock.  The writer lock, cpu_hotplug_begin(), is used to
> udpate the cpu maps along with cpu_maps_update_begin().
> 
> However, the ACPI processor handler updates the cpu maps without
> holding the the writer lock.  This patchset fixes this problem.
> 
> ---
> The patchset is based on linux-next of the pm tree.

Basically looks OK to me, but I'd just merge the patches together,
because the [2/2] is the very reason for the change made by [1/2].

Also it would be good to mention the actual scenario that may lead
to problems with that.

Thanks,
Rafael


> ---
> Toshi Kani (2):
>  CPU hotplug: Export cpu_hotplug_begin() interface
>  ACPI: Acquire writer lock to update cpu maps
> 
> ---
>  drivers/acpi/acpi_processor.c | 21 ++++++++++++++++-----
>  include/linux/cpu.h           |  2 ++
>  kernel/cpu.c                  |  4 ++--
>  3 files changed, 20 insertions(+), 7 deletions(-)
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
-- 
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.

Re: [PATCH 0/2] Protect cpu map updates from ACPI processor

[Toshi Kani]

On Tue, 2013-08-13 at 02:43 +0200, Rafael J. Wysocki wrote:
> On Monday, August 12, 2013 09:45:52 AM Toshi Kani wrote:
> > CPU system maps are protected with reader/writer locks.  The reader
> > lock, put_online_cpus(), assures that the maps are not updated while
> > holding the lock.  The writer lock, cpu_hotplug_begin(), is used to
> > udpate the cpu maps along with cpu_maps_update_begin().
> > 
> > However, the ACPI processor handler updates the cpu maps without
> > holding the the writer lock.  This patchset fixes this problem.
> > 
> > ---
> > The patchset is based on linux-next of the pm tree.
> 
> Basically looks OK to me, but I'd just merge the patches together,
> because the [2/2] is the very reason for the change made by [1/2].

I separated for reviewing purpose, but yes, they can be together.  Let
me know if you need me to re-send it together.

> Also it would be good to mention the actual scenario that may lead
> to problems with that.

For example, the get_online_cpus() below is supposed to assure that
cpu_possible_mask is not changed while the code is iterating with
for_each_possible_cpu().

        get_online_cpus();
        for_each_possible_cpu(cpu) {
		:
        }
        put_online_cpus();

However, this lock has no protection with CPU hotplug since the ACPI
processor handler does not use the writer lock when it updates
cpu_possible_mask.  The reader lock does not serialize within the
readers.

Thanks,
-Toshi

Re: [PATCH 0/2] Protect cpu map updates from ACPI processor

[Rafael J. Wysocki]

On Monday, August 12, 2013 06:55:03 PM Toshi Kani wrote:
> On Tue, 2013-08-13 at 02:43 +0200, Rafael J. Wysocki wrote:
> > On Monday, August 12, 2013 09:45:52 AM Toshi Kani wrote:
> > > CPU system maps are protected with reader/writer locks.  The reader
> > > lock, put_online_cpus(), assures that the maps are not updated while
> > > holding the lock.  The writer lock, cpu_hotplug_begin(), is used to
> > > udpate the cpu maps along with cpu_maps_update_begin().
> > > 
> > > However, the ACPI processor handler updates the cpu maps without
> > > holding the the writer lock.  This patchset fixes this problem.
> > > 
> > > ---
> > > The patchset is based on linux-next of the pm tree.
> > 
> > Basically looks OK to me, but I'd just merge the patches together,
> > because the [2/2] is the very reason for the change made by [1/2].
> 
> I separated for reviewing purpose, but yes, they can be together.  Let
> me know if you need me to re-send it together.

I combined them and added a changelog built out of the messages you sent in
this thread.  Please check:

http://git.kernel.org/cgit/linux/kernel/git/rafael/linux-pm.git/commit/?h=bleeding-edge&id=b9d10be7a8e88fdcb12540387c219cdde87b0795

Thanks,
Rafael

Re: [PATCH 0/2] Protect cpu map updates from ACPI processor

[Toshi Kani]

On Tue, 2013-08-13 at 13:47 +0200, Rafael J. Wysocki wrote:
> On Monday, August 12, 2013 06:55:03 PM Toshi Kani wrote:
> > On Tue, 2013-08-13 at 02:43 +0200, Rafael J. Wysocki wrote:
> > > On Monday, August 12, 2013 09:45:52 AM Toshi Kani wrote:
> > > > CPU system maps are protected with reader/writer locks.  The reader
> > > > lock, put_online_cpus(), assures that the maps are not updated while
> > > > holding the lock.  The writer lock, cpu_hotplug_begin(), is used to
> > > > udpate the cpu maps along with cpu_maps_update_begin().
> > > > 
> > > > However, the ACPI processor handler updates the cpu maps without
> > > > holding the the writer lock.  This patchset fixes this problem.
> > > > 
> > > > ---
> > > > The patchset is based on linux-next of the pm tree.
> > > 
> > > Basically looks OK to me, but I'd just merge the patches together,
> > > because the [2/2] is the very reason for the change made by [1/2].
> > 
> > I separated for reviewing purpose, but yes, they can be together.  Let
> > me know if you need me to re-send it together.
> 
> I combined them and added a changelog built out of the messages you sent in
> this thread.  Please check:
> 
> http://git.kernel.org/cgit/linux/kernel/git/rafael/linux-pm.git/commit/?h=bleeding-edge&id=b9d10be7a8e88fdcb12540387c219cdde87b0795
> 

Looks good!
Thanks,
-Toshi