This is a note to let you know that I've just added the patch titled
PM: runtime: Add safety net to supplier device release
to the 5.10-stable tree which can be found at:
http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary
The filename of the patch is:
pm-runtime-add-safety-net-to-supplier-device-release.patch
and it can be found in the queue-5.10 subdirectory.
If you, or anyone else, feels it should not be added to the stable tree,
please let <stable@xxxxxxxxxxxxxxx> know about it.
commit 078902dbd0754664e17a69576275d08602a9e720
Author: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx>
Date: Fri Dec 10 17:10:13 2021 +0100
PM: runtime: Add safety net to supplier device release
[ Upstream commit d1579e61192e0e686faa4208500ef4c3b529b16c ]
Because refcount_dec_not_one() returns true if the target refcount
becomes saturated, it is generally unsafe to use its return value as
a loop termination condition, but that is what happens when a device
link's supplier device is released during runtime PM suspend
operations and on device link removal.
To address this, introduce pm_runtime_release_supplier() to be used
in the above cases which will check the supplier device's runtime
PM usage counter in addition to the refcount_dec_not_one() return
value, so the loop can be terminated in case the rpm_active refcount
value becomes invalid, and update the code in question to use it as
appropriate.
This change is not expected to have any visible functional impact.
Reported-by: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx>
Acked-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
Acked-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
diff --git a/drivers/base/core.c b/drivers/base/core.c
index 389d13616d1df..c0566aff53551 100644
--- a/drivers/base/core.c
+++ b/drivers/base/core.c
@@ -348,8 +348,7 @@ static void device_link_release_fn(struct work_struct *work)
/* Ensure that all references to the link object have been dropped. */
device_link_synchronize_removal();
- while (refcount_dec_not_one(&link->rpm_active))
- pm_runtime_put(link->supplier);
+ pm_runtime_release_supplier(link, true);
put_device(link->consumer);
put_device(link->supplier);
diff --git a/drivers/base/power/runtime.c b/drivers/base/power/runtime.c
index bc649da4899a0..1573319404888 100644
--- a/drivers/base/power/runtime.c
+++ b/drivers/base/power/runtime.c
@@ -305,19 +305,40 @@ static int rpm_get_suppliers(struct device *dev)
return 0;
}
+/**
+ * pm_runtime_release_supplier - Drop references to device link's supplier.
+ * @link: Target device link.
+ * @check_idle: Whether or not to check if the supplier device is idle.
+ *
+ * Drop all runtime PM references associated with @link to its supplier device
+ * and if @check_idle is set, check if that device is idle (and so it can be
+ * suspended).
+ */
+void pm_runtime_release_supplier(struct device_link *link, bool check_idle)
+{
+ struct device *supplier = link->supplier;
+
+ /*
+ * The additional power.usage_count check is a safety net in case
+ * the rpm_active refcount becomes saturated, in which case
+ * refcount_dec_not_one() would return true forever, but it is not
+ * strictly necessary.
+ */
+ while (refcount_dec_not_one(&link->rpm_active) &&
+ atomic_read(&supplier->power.usage_count) > 0)
+ pm_runtime_put_noidle(supplier);
+
+ if (check_idle)
+ pm_request_idle(supplier);
+}
+
static void __rpm_put_suppliers(struct device *dev, bool try_to_suspend)
{
struct device_link *link;
list_for_each_entry_rcu(link, &dev->links.suppliers, c_node,
- device_links_read_lock_held()) {
-
- while (refcount_dec_not_one(&link->rpm_active))
- pm_runtime_put_noidle(link->supplier);
-
- if (try_to_suspend)
- pm_request_idle(link->supplier);
- }
+ device_links_read_lock_held())
+ pm_runtime_release_supplier(link, try_to_suspend);
}
static void rpm_put_suppliers(struct device *dev)
@@ -1755,9 +1776,7 @@ void pm_runtime_drop_link(struct device_link *link)
return;
pm_runtime_drop_link_count(link->consumer);
-
- while (refcount_dec_not_one(&link->rpm_active))
- pm_runtime_put(link->supplier);
+ pm_runtime_release_supplier(link, true);
}
static bool pm_runtime_need_not_resume(struct device *dev)
diff --git a/include/linux/pm_runtime.h b/include/linux/pm_runtime.h
index 161acd4ede448..30091ab5de287 100644
--- a/include/linux/pm_runtime.h
+++ b/include/linux/pm_runtime.h
@@ -58,6 +58,7 @@ extern void pm_runtime_get_suppliers(struct device *dev);
extern void pm_runtime_put_suppliers(struct device *dev);
extern void pm_runtime_new_link(struct device *dev);
extern void pm_runtime_drop_link(struct device_link *link);
+extern void pm_runtime_release_supplier(struct device_link *link, bool check_idle);
/**
* pm_runtime_get_if_in_use - Conditionally bump up runtime PM usage counter.
@@ -279,6 +280,8 @@ static inline void pm_runtime_get_suppliers(struct device *dev) {}
static inline void pm_runtime_put_suppliers(struct device *dev) {}
static inline void pm_runtime_new_link(struct device *dev) {}
static inline void pm_runtime_drop_link(struct device_link *link) {}
+static inline void pm_runtime_release_supplier(struct device_link *link,
+ bool check_idle) {}
#endif /* !CONFIG_PM */
