ACPI / PM: Walk physical_node_list under physical_node_lock
The list of physical devices corresponding to an ACPI device
object is walked by acpi_system_wakeup_device_seq_show() and
physical_device_enable_wakeup() without taking that object's
physical_node_lock mutex. Since each of those functions may be
run at any time as a result of a user space action, the lack of
appropriate locking in them may lead to a kernel crash if that
happens during device hot-add or hot-remove involving the device
object in question.
Fix the issue by modifying acpi_system_wakeup_device_seq_show() and
physical_device_enable_wakeup() to use physical_node_lock as
appropriate.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Cc: All <stable@vger.kernel.org>
diff --git a/drivers/acpi/proc.c b/drivers/acpi/proc.c
index aa1227a..04a1378 100644
--- a/drivers/acpi/proc.c
+++ b/drivers/acpi/proc.c
@@ -311,6 +311,8 @@
dev->pnp.bus_id,
(u32) dev->wakeup.sleep_state);
+ mutex_lock(&dev->physical_node_lock);
+
if (!dev->physical_node_count) {
seq_printf(seq, "%c%-8s\n",
dev->wakeup.flags.run_wake ? '*' : ' ',
@@ -338,6 +340,8 @@
put_device(ldev);
}
}
+
+ mutex_unlock(&dev->physical_node_lock);
}
mutex_unlock(&acpi_device_lock);
return 0;
@@ -347,12 +351,16 @@
{
struct acpi_device_physical_node *entry;
+ mutex_lock(&adev->physical_node_lock);
+
list_for_each_entry(entry,
&adev->physical_node_list, node)
if (entry->dev && device_can_wakeup(entry->dev)) {
bool enable = !device_may_wakeup(entry->dev);
device_set_wakeup_enable(entry->dev, enable);
}
+
+ mutex_unlock(&adev->physical_node_lock);
}
static ssize_t