device property: Track owner device of device property
Subsystems (4)
PATCH 1/1 device property: Track owner device of device property
Deletion of subdevice will remove device properties associated to parent when they share the same firmware node after commit 478573c93abd ("driver
core: Don't leak secondary fwnode on device removal"). This was observed with a driver adding subdevice that driver wasn't able to read device properties after rmmod/modprobe cycle.
Consider the lifecycle of it:
parent device registration
ACPI_COMPANION_SET() device_add_properties() pset_copy_set() set_secondary_fwnode(dev, &p->fwnode) device_add()
parent probe
read device properties ACPI_COMPANION_SET(subdevice, ACPI_COMPANION(parent)) device_add(subdevice)
parent remove
device_del(subdevice) device_remove_properties() set_secondary_fwnode(dev, NULL); pset_free()
Parent device will have its primary firmware node pointing to an ACPI node and secondary firmware node point to device properties. ACPI_COMPANION_SET() call in parent probe will set the subdevice's firmware node to point to the same 'struct fwnode_handle' and the associated secondary firmware node, i.e. the device properties as the parent.
When subdevice is deleted in parent remove that will remove those device properties and attempt to read device properties in next parent probe call will fail.
Fix this by tracking the owner device of device properties and delete them only when owner device is being deleted.
--- drivers/base/property.c | 15 +++++++++------ 1 file changed, 9 insertions(+), 6 deletions(-)
drivers/base/property.c +9 -6
@@ -21,6 +21,7 @@#include <linux/phy.h> struct property_set { +struct device *dev;struct fwnode_handle fwnode;const struct property_entry *properties; };
@@ -891,6 +892,7 @@ static struct property_set *pset_copy_set(const struct property_set *pset)void device_remove_properties (struct device *dev) {struct fwnode_handle *fwnode; +struct property_set *pset; fwnode =dev_fwnode (dev);if (!fwnode)
@@ -900,16 +902,16 @@ void device_remove_properties(struct device *dev)* the pset. If there is no real firmware node (ACPI/DT) primary * will hold the pset. */ -if (is_pset_node (fwnode)) { + pset =to_pset_node (fwnode); +if (pset) {set_primary_fwnode (dev,NULL ); -pset_free_set (to_pset_node (fwnode)); }else { - fwnode = fwnode->secondary; -if (!IS_ERR (fwnode) &&is_pset_node (fwnode)) { + pset =to_pset_node (fwnode->secondary); +if (pset && dev == pset->dev)set_secondary_fwnode (dev,NULL ); -pset_free_set (to_pset_node (fwnode)); - } } +if (pset && dev == pset->dev) +pset_free_set (pset); }EXPORT_SYMBOL_GPL (device_remove_properties);
@@ -938,6 +940,7 @@ int device_add_properties(struct device *dev, p->fwnode.ops = &pset_fwnode_ops;set_secondary_fwnode (dev, &p->fwnode); + p->dev = dev;return 0 ; }EXPORT_SYMBOL_GPL (device_add_properties);
Signature · 1 line
2.14.2
Greg Kroah-Hartman
maintainerQuoted 90 lines · Jarkko Nikula
Deletion of subdevice will remove device properties associated to parent when they share the same firmware node after commit
478573c93abd("driver
core: Don't leak secondary fwnode on device removal"). This was observed with a driver adding subdevice that driver wasn't able to read device properties after rmmod/modprobe cycle.Consider the lifecycle of it:
parent device registration
ACPI_COMPANION_SET() device_add_properties() pset_copy_set() set_secondary_fwnode(dev, &p->fwnode) device_add()parent probe
read device propertiesACPI_COMPANION_SET(subdevice, ACPI_COMPANION(parent))device_add(subdevice)parent remove
device_del(subdevice) device_remove_properties() set_secondary_fwnode(dev, NULL); pset_free()Parent device will have its primary firmware node pointing to an ACPI node and secondary firmware node point to device properties.
ACPI_COMPANION_SET()call in parent probe will set the subdevice's firmware node to point to the same 'structfwnode_handle' and the associated secondary firmware node, i.e. the device properties as the parent.When subdevice is deleted in parent remove that will remove those device properties and attempt to read device properties in next parent probe call will fail.
Fix this by tracking the owner device of device properties and delete them only when owner device is being deleted.
Fixes 478573c93abd ("driver core: Don't leak secondary fwnode on device removal")Cc <[email protected]> # v4.9+Signed-off-by Jarkko Nikula <[email protected]>--- drivers/base/property.c | 15 +++++++++------ 1 file changed, 9 insertions(+), 6 deletions(-)drivers/base/property.c +8
-6@@ -21,6 +21,7 @@#include <linux/phy.h> struct property_set { +struct device *dev;struct fwnode_handle fwnode;const struct property_entry *properties; };@@ -891,6 +892,7 @@ static struct property_set *pset_copy_set(const struct property_set *pset)void device_remove_properties (struct device *dev) {struct fwnode_handle *fwnode; +struct property_set *pset; fwnode =dev_fwnode (dev);if (!fwnode)@@ -900,16 +902,16 @@ void device_remove_properties(struct device *dev)* the pset. If there is no real firmware node (ACPI/DT) primary * will hold the pset. */ -if (is_pset_node (fwnode)) { + pset =to_pset_node (fwnode); +if (pset) {set_primary_fwnode (dev,NULL ); -pset_free_set (to_pset_node (fwnode)); }else { - fwnode = fwnode->secondary; -if (!IS_ERR (fwnode) &&is_pset_node (fwnode)) { + pset =to_pset_node (fwnode->secondary); +if (pset && dev == pset->dev)set_secondary_fwnode (dev,NULL ); -pset_free_set (to_pset_node (fwnode)); - } } +if (pset && dev == pset->dev) +pset_free_set (pset); }EXPORT_SYMBOL_GPL (device_remove_properties);
drivers/base/property.c int device_add_properties(struct device *dev, · line 943
Quoted 4 lines
drivers/base/property.c +1
-0@@ -938,6 +940,7 @@ int device_add_properties(struct device *dev, p->fwnode.ops = &pset_fwnode_ops;set_secondary_fwnode (dev, &p->fwnode); + p->dev = dev;
Don't you also need to increment the reference counter here? Or how is it assured that it will not go away?
thanks,
greg k-h
Quoted 9 lines · Greg Kroah-Hartman
drivers/base/property.c int device_add_properties(struct device *dev, · line 943
Quoted 5 lines
drivers/base/property.c +1
-0@@ -938,6 +940,7 @@ int device_add_properties(struct device *dev, p->fwnode.ops = &pset_fwnode_ops;set_secondary_fwnode (dev, &p->fwnode); + p->dev = dev;Don't you also need to increment the reference counter here? Or how is it assured that it will not go away?
I need to scratch my head on this. It sounds more robust to track references and remove properties when last reference is dropped in device_remove_properties(). What I don't know are properties be expected to be copied and be usable for another device for instance by ACPI_COMPANION_SET().
I can figure out case where properties are added to one device, another device gets reference to them via ACPI_COMPANION_SET() and properties get freed when the first device is removed. But should the second device be able to use those properties at first place?
Rafael: What's you opinion: should there be reference counting for device properties? Initial increment in device_add_properties(), other around ACPI_COMPANION_SET()/set_primary_fwnode() and decrementing in device_remove_properties().
Signature · 1 line
Jarkko