Skip to contentSkip to conversation

ACPI: bus: Avoid non-ACPI device objects in walks over children

PATCH In review Rafael J. Wysocki 3 participants 3 messages Apr 22, 2022

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ACPI: bus: Avoid non-ACPI device objects in walks over children

From: Rafael J. Wysocki <[email protected]>

When walking the children of an ACPI device, take extra care to avoid using to_acpi_device() on the ones that are not ACPI devices, because that may lead to out-of-bounds access and memory corruption.

While at it, make the function passed to acpi_dev_for_each_child() take a struct acpi_device pointer argument (instead of a struct device one), so it is more straightforward to use.

Fixes b7dd6298db81 ("ACPI: PM: Introduce acpi_dev_power_up_children_with_adr()")

---

The commit being fixed is present in linux-next.

---
 drivers/acpi/bus.c       |   24 ++++++++++++++++++++++--
 drivers/acpi/device_pm.c |    5 +----
 include/acpi/acpi_bus.h  |    2 +-
 3 files changed, 24 insertions(+), 7 deletions(-)
drivers/acpi/bus.c +22 -2
Index: linux-pm/drivers/acpi/bus.c
===================================================================
@@ -1070,10 +1070,30 @@ int acpi_bus_for_each_dev(int (*fn)(stru
 }
 EXPORT_SYMBOL_GPL(acpi_bus_for_each_dev);
 
+struct acpi_dev_walk_context {
+	int (*fn)(struct acpi_device *, void *);
+	void *data;
+};
+
+static int acpi_dev_for_one_check(struct device *dev, void *context)
+{
+	struct acpi_dev_walk_context *adwc = context;
+
+	if (dev->bus != &acpi_bus_type)
+		return 0;
+
+	return adwc->fn(to_acpi_device(dev), adwc->data);
+}
+
 int acpi_dev_for_each_child(struct acpi_device *adev,
-			    int (*fn)(struct device *, void *), void *data)
+			    int (*fn)(struct acpi_device *, void *), void *data)
 {
-	return device_for_each_child(&adev->dev, data, fn);
+	struct acpi_dev_walk_context adwc = {
+		.fn = fn,
+		.data = data,
+	};
+
+	return device_for_each_child(&adev->dev, &adwc, acpi_dev_for_one_check);
 }
 
 /* --------------------------------------------------------------------------
include/acpi/acpi_bus.h +1 -1
Index: linux-pm/include/acpi/acpi_bus.h
===================================================================
@@ -482,7 +482,7 @@ extern struct bus_type acpi_bus_type;
 
 int acpi_bus_for_each_dev(int (*fn)(struct device *, void *), void *data);
 int acpi_dev_for_each_child(struct acpi_device *adev,
-			    int (*fn)(struct device *, void *), void *data);
+			    int (*fn)(struct acpi_device *, void *), void *data);
 
 /*
  * Events
drivers/acpi/device_pm.c +1 -4
Index: linux-pm/drivers/acpi/device_pm.c
===================================================================
@@ -429,11 +429,8 @@ bool acpi_bus_power_manageable(acpi_hand
 }
 EXPORT_SYMBOL(acpi_bus_power_manageable);
 
-static int acpi_power_up_if_adr_present(struct device *dev, void *not_used)
+static int acpi_power_up_if_adr_present(struct acpi_device *adev, void *not_used)
 {
-	struct acpi_device *adev;
-
-	adev = to_acpi_device(dev);
 	if (!(adev->flags.power_manageable && adev->pnp.type.bus_address))
 		return 0;
 
Reviewed-by on v1 1/1
#1 lore raw

Hi Rafael,

Quoted 33 lines · Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

When walking the children of an ACPI device, take extra care to avoid using to_acpi_device() on the ones that are not ACPI devices, because that may lead to out-of-bounds access and memory corruption.

While at it, make the function passed to acpi_dev_for_each_child() take a struct acpi_device pointer argument (instead of a struct device one), so it is more straightforward to use.

Fixes b7dd6298db81 ("ACPI: PM: Introduce acpi_dev_power_up_children_with_adr()")

---

The commit being fixed is present in linux-next.

---
 drivers/acpi/bus.c       |   24 ++++++++++++++++++++++--
 drivers/acpi/device_pm.c |    5 +----
 include/acpi/acpi_bus.h  |    2 +-
 3 files changed, 24 insertions(+), 7 deletions(-)

Index: linux-pm/drivers/acpi/bus.c
===================================================================

drivers/acpi/bus.c +1 -0
@@ -1070,10 +1070,30 @@ int acpi_bus_for_each_dev(int (*fn)(stru
 }
 EXPORT_SYMBOL_GPL(acpi_bus_for_each_dev);
 
+struct acpi_dev_walk_context {

drivers/acpi/bus.c int acpi_bus_for_each_dev(int (*fn)(stru · line 1083

Quoted 10 lines
drivers/acpi/bus.c +10 -0
+	int (*fn)(struct acpi_device *, void *);
+	void *data;
+};
+
+static int acpi_dev_for_one_check(struct device *dev, void *context)
+{
+	struct acpi_dev_walk_context *adwc = context;
+
+	if (dev->bus != &acpi_bus_type)
+		return 0;

I wonder if it make sense to add dev_is_acpi() that does the above analoguos to dev_is_pci()?

Regardless of that,

#2 lore raw
Quoted 36 lines · Mika Westerberg

Hi Rafael,

Quoted 32 lines · Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

When walking the children of an ACPI device, take extra care to avoid using to_acpi_device() on the ones that are not ACPI devices, because that may lead to out-of-bounds access and memory corruption.

While at it, make the function passed to acpi_dev_for_each_child() take a struct acpi_device pointer argument (instead of a struct device one), so it is more straightforward to use.

Fixes b7dd6298db81 ("ACPI: PM: Introduce acpi_dev_power_up_children_with_adr()")

---

The commit being fixed is present in linux-next.

---
 drivers/acpi/bus.c       |   24 ++++++++++++++++++++++--
 drivers/acpi/device_pm.c |    5 +----
 include/acpi/acpi_bus.h  |    2 +-
 3 files changed, 24 insertions(+), 7 deletions(-)

Index: linux-pm/drivers/acpi/bus.c
===================================================================

drivers/acpi/bus.c +0 -0
@@ -1070,10 +1070,30 @@ int acpi_bus_for_each_dev(int (*fn)(stru
 }
 EXPORT_SYMBOL_GPL(acpi_bus_for_each_dev);
 
Quoted 14 lines

drivers/acpi/bus.c int acpi_bus_for_each_dev(int (*fn)(stru · line 1083

Quoted 11 lines
drivers/acpi/bus.c +11 -0
+struct acpi_dev_walk_context {
+     int (*fn)(struct acpi_device *, void *);
+     void *data;
+};
+
+static int acpi_dev_for_one_check(struct device *dev, void *context)
+{
+     struct acpi_dev_walk_context *adwc = context;
+
+     if (dev->bus != &acpi_bus_type)
+             return 0;

I wonder if it make sense to add dev_is_acpi() that does the above analoguos to dev_is_pci()?

I thought about that, but this is the only place where it would be needed ATM, so for now this isn't necessary.

Quoted 3 lines

Regardless of that,

Thanks!