Skip to contentSkip to conversation

software node: Introduce software_node_find_by_name()

PATCH In review Heikki Krogerus 5 participants 10 messages Aug 19, 2019 – Sep 4, 2019

Subsystems (5)
lore
Share & tools
lore

Heikki Krogerus

author reviewer
#0 lore raw

v3 cover software node: Introduce software_node_find_by_name()

Hi,

There was still one bug spotted by Andy in v2. The role switch node was not removed in case of an error (patch 2/3). It is now fixed.

The cover letter from v2:

This is the second version of this series where I'm introducing that helper.

Hans and Andy! Because of the changes I made to patch 2/3, I'm not carrying your reviewed-by tags in it. I would appreciate if you could take another look at that patch.

I added a note to the kernel-doc comment in patch 1/3 that the caller of the helper function needs to release the ref count after use as proposed by Andy.

In patch 2/3, since we have to now modify the role switch descriptor, I'm filling the structure in stack memory and removing the constant static version. The content of the descriptor is copied during switch registration in any case, so we don't need to store it in the driver.

I also noticed a bug in 2/3. I never properly destroyed the software node when the mux was removed. That leak is now also fixed.

thanks,

Heikki Krogerus (3):
software node: Add software_node_find_by_name()
usb: roles: intel_xhci: Supplying software node for the role mux platform/x86: intel_cht_int33fe: Use new API to gain access to the
role switch

 drivers/base/swnode.c                         | 37 ++++++++++++
 drivers/platform/x86/intel_cht_int33fe.c      | 57 ++++---------------
 .../usb/roles/intel-xhci-usb-role-switch.c    | 27 ++++++---
 include/linux/property.h                      |  4 ++
 4 files changed, 71 insertions(+), 54 deletions(-)
Signature · 1 line
2.23.0.rc1
author reviewer
#1 lore raw

PATCH v3 1/3 software node: Add software_node_find_by_name()

Function that searches software nodes by node name.

---
 drivers/base/swnode.c    | 37 +++++++++++++++++++++++++++++++++++++
 include/linux/property.h |  4 ++++
 2 files changed, 41 insertions(+)
drivers/base/swnode.c +37 -0
index e7b3aa3bd55a..ee2a405cca9a 100644
@@ -620,6 +620,43 @@ static const struct fwnode_operations software_node_ops = {
 
 /* -------------------------------------------------------------------------- */
 
+/**
+ * software_node_find_by_name - Find software node by name
+ * @parent: Parent of the software node
+ * @name: Name of the software node
+ *
+ * The function will find a node that is child of @parent and that is named
+ * @name. If no node is found, the function returns NULL.
+ *
+ * NOTE: you will need to drop the reference with fwnode_handle_put() after use.
+ */
+const struct software_node *
+software_node_find_by_name(const struct software_node *parent, const char *name)
+{
+	struct swnode *swnode;
+	struct kobject *k;
+
+	if (!name)
+		return NULL;
+
+	spin_lock(&swnode_kset->list_lock);
+
+	list_for_each_entry(k, &swnode_kset->list, entry) {
+		swnode = kobj_to_swnode(k);
+		if (parent == swnode->node->parent && swnode->node->name &&
+		    !strcmp(name, swnode->node->name)) {
+			kobject_get(&swnode->kobj);
+			break;
+		}
+		swnode = NULL;
+	}
+
+	spin_unlock(&swnode_kset->list_lock);
+
+	return swnode ? swnode->node : NULL;
+}
+EXPORT_SYMBOL_GPL(software_node_find_by_name);
+
 static int
 software_node_register_properties(struct software_node *node,
 				  const struct property_entry *properties)
include/linux/property.h +4 -0
index 5a910ad79591..9b3d4ca3a73a 100644
@@ -421,6 +421,10 @@ bool is_software_node(const struct fwnode_handle *fwnode);
 const struct software_node *to_software_node(struct fwnode_handle *fwnode);
 struct fwnode_handle *software_node_fwnode(const struct software_node *node);
 
+const struct software_node *
+software_node_find_by_name(const struct software_node *parent,
+			   const char *name);
+
 int software_node_register_nodes(const struct software_node *nodes);
 void software_node_unregister_nodes(const struct software_node *nodes);
 
Signature · 1 line
2.23.0.rc1
author reviewer
#3 lore raw

PATCH v3 2/3 usb: roles: intel_xhci: Supplying software node for the role mux

In reply to Heikki Krogerus

The primary purpose for this node will be to allow linking the users of the switch to it. The users will be for example USB Type-C connectors. By supplying a reference to this node in the software nodes representing the USB Type-C controllers or connectors, the drivers for those devices can access the switch.

---
 .../usb/roles/intel-xhci-usb-role-switch.c    | 27 ++++++++++++++-----
 1 file changed, 20 insertions(+), 7 deletions(-)
drivers/usb/roles/intel-xhci-usb-role-switch.c +20 -7
index 277de96181f9..7325a84dd1c8 100644
@@ -39,6 +39,10 @@ struct intel_xhci_usb_data {
 	void __iomem *base;
 };
 
+static const struct software_node intel_xhci_usb_node = {
+	"intel-xhci-usb-sw",
+};
+
 static int intel_xhci_usb_set_role(struct device *dev, enum usb_role role)
 {
 	struct intel_xhci_usb_data *data = dev_get_drvdata(dev);
@@ -122,17 +126,13 @@ static enum usb_role intel_xhci_usb_get_role(struct device *dev)
 	return role;
 }
 
-static const struct usb_role_switch_desc sw_desc = {
-	.set = intel_xhci_usb_set_role,
-	.get = intel_xhci_usb_get_role,
-	.allow_userspace_control = true,
-};
-
 static int intel_xhci_usb_probe(struct platform_device *pdev)
 {
+	struct usb_role_switch_desc sw_desc = { };
 	struct device *dev = &pdev->dev;
 	struct intel_xhci_usb_data *data;
 	struct resource *res;
+	int ret;
 
 	data = devm_kzalloc(dev, sizeof(*data), GFP_KERNEL);
 	if (!data)
@@ -147,9 +147,20 @@ static int intel_xhci_usb_probe(struct platform_device *pdev)
 
 	platform_set_drvdata(pdev, data);
 
+	ret = software_node_register(&intel_xhci_usb_node);
+	if (ret)
+		return ret;
+
+	sw_desc.set = intel_xhci_usb_set_role,
+	sw_desc.get = intel_xhci_usb_get_role,
+	sw_desc.allow_userspace_control = true,
+	sw_desc.fwnode = software_node_fwnode(&intel_xhci_usb_node);
+
 	data->role_sw = usb_role_switch_register(dev, &sw_desc);
-	if (IS_ERR(data->role_sw))
+	if (IS_ERR(data->role_sw)) {
+		fwnode_handle_put(sw_desc.fwnode);
 		return PTR_ERR(data->role_sw);
+	}
 
 	pm_runtime_set_active(dev);
 	pm_runtime_enable(dev);
@@ -164,6 +175,8 @@ static int intel_xhci_usb_remove(struct platform_device *pdev)
 	pm_runtime_disable(&pdev->dev);
 
 	usb_role_switch_unregister(data->role_sw);
+	fwnode_handle_put(software_node_fwnode(&intel_xhci_usb_node));
+
 	return 0;
 }
 
Signature · 1 line
2.23.0.rc1
Reviewed-by on v3 2/3
#4 lore raw

Re: [PATCH v3 2/3] usb: roles: intel_xhci: Supplying software node for the role mux

Quoted 8 lines · Heikki Krogerus

The primary purpose for this node will be to allow linking the users of the switch to it. The users will be for example USB Type-C connectors. By supplying a reference to this node in the software nodes representing the USB Type-C controllers or connectors, the drivers for those devices can access the switch.

FWIW,

Quoted 74 lines
---
 .../usb/roles/intel-xhci-usb-role-switch.c    | 27 ++++++++++++++-----
 1 file changed, 20 insertions(+), 7 deletions(-)
drivers/usb/roles/intel-xhci-usb-role-switch.c +20 -8
index 277de96181f9..7325a84dd1c8 100644
@@ -39,6 +39,10 @@ struct intel_xhci_usb_data {
        void __iomem *base;
 };
 
+static const struct software_node intel_xhci_usb_node = {
+       "intel-xhci-usb-sw",
+};
+
 static int intel_xhci_usb_set_role(struct device *dev, enum usb_role role)
 {
        struct intel_xhci_usb_data *data = dev_get_drvdata(dev);
@@ -122,17 +126,13 @@ static enum usb_role intel_xhci_usb_get_role(struct device *dev)
        return role;
 }
 
-static const struct usb_role_switch_desc sw_desc = {
-       .set = intel_xhci_usb_set_role,
-       .get = intel_xhci_usb_get_role,
-       .allow_userspace_control = true,
-};
-
 static int intel_xhci_usb_probe(struct platform_device *pdev)
 {
+       struct usb_role_switch_desc sw_desc = { };
        struct device *dev = &pdev->dev;
        struct intel_xhci_usb_data *data;
        struct resource *res;
+       int ret;
 
        data = devm_kzalloc(dev, sizeof(*data), GFP_KERNEL);
        if (!data)
@@ -147,9 +147,20 @@ static int intel_xhci_usb_probe(struct platform_device *pdev)
 
        platform_set_drvdata(pdev, data);
 
+       ret = software_node_register(&intel_xhci_usb_node);
+       if (ret)
+               return ret;
+
+       sw_desc.set = intel_xhci_usb_set_role,
+       sw_desc.get = intel_xhci_usb_get_role,
+       sw_desc.allow_userspace_control = true,
+       sw_desc.fwnode = software_node_fwnode(&intel_xhci_usb_node);
+
        data->role_sw = usb_role_switch_register(dev, &sw_desc);
-       if (IS_ERR(data->role_sw))
+       if (IS_ERR(data->role_sw)) {
+               fwnode_handle_put(sw_desc.fwnode);
                return PTR_ERR(data->role_sw);
+       }
 
        pm_runtime_set_active(dev);
        pm_runtime_enable(dev);
@@ -164,6 +175,8 @@ static int intel_xhci_usb_remove(struct platform_device *pdev)
        pm_runtime_disable(&pdev->dev);
 
        usb_role_switch_unregister(data->role_sw);
+       fwnode_handle_put(software_node_fwnode(&intel_xhci_usb_node));
+
        return 0;
 }
 
--

2.23.0.rc1

Signature · 2 lines
With Best Regards,
Andy Shevchenko
maintainer Reviewed-by on v3 2/3
#5 lore raw

Re: [PATCH v3 2/3] usb: roles: intel_xhci: Supplying software node for the role mux

In reply to Heikki Krogerus
Quoted 8 lines · Heikki Krogerus

The primary purpose for this node will be to allow linking the users of the switch to it. The users will be for example USB Type-C connectors. By supplying a reference to this node in the software nodes representing the USB Type-C controllers or connectors, the drivers for those devices can access the switch.

author reviewer
#6 lore raw

PATCH v3 3/3 platform/x86: intel_cht_int33fe: Use new API to gain access to the role switch

In reply to Heikki Krogerus

The driver for the Intel USB role mux now always supplies software node for the role switch, so no longer checking that, and never creating separate node for the role switch. From now on using software_node_find_by_name() function to get the handle to the USB role switch.

---
 drivers/platform/x86/intel_cht_int33fe.c | 57 +++++-------------------
 1 file changed, 10 insertions(+), 47 deletions(-)
drivers/platform/x86/intel_cht_int33fe.c +10 -47
index 4fbdff48a4b5..1d5d877b9582 100644
@@ -34,7 +34,6 @@ enum {
 	INT33FE_NODE_MAX17047,
 	INT33FE_NODE_PI3USB30532,
 	INT33FE_NODE_DISPLAYPORT,
-	INT33FE_NODE_ROLE_SWITCH,
 	INT33FE_NODE_USB_CONNECTOR,
 	INT33FE_NODE_MAX,
 };
@@ -45,7 +44,6 @@ struct cht_int33fe_data {
 	struct i2c_client *pi3usb30532;
 
 	struct fwnode_handle *dp;
-	struct fwnode_handle *mux;
 };
 
 static const struct software_node nodes[];
@@ -139,46 +137,10 @@ static const struct software_node nodes[] = {
 	{ "max17047", NULL, max17047_props },
 	{ "pi3usb30532" },
 	{ "displayport" },
-	{ "usb-role-switch" },
 	{ "connector", &nodes[0], usb_connector_props, usb_connector_refs },
 	{ }
 };
 
-static int cht_int33fe_setup_mux(struct cht_int33fe_data *data)
-{
-	struct fwnode_handle *fwnode;
-	struct device *dev;
-	struct device *p;
-
-	fwnode = software_node_fwnode(&nodes[INT33FE_NODE_ROLE_SWITCH]);
-	if (!fwnode)
-		return -ENODEV;
-
-	/* First finding the platform device */
-	p = bus_find_device_by_name(&platform_bus_type, NULL,
-				    "intel_xhci_usb_sw");
-	if (!p)
-		return -EPROBE_DEFER;
-
-	/* Then the mux child device */
-	dev = device_find_child_by_name(p, "intel_xhci_usb_sw-role-switch");
-	put_device(p);
-	if (!dev)
-		return -EPROBE_DEFER;
-
-	/* If there already is a node for the mux, using that one. */
-	if (dev->fwnode)
-		fwnode_remove_software_node(fwnode);
-	else
-		dev->fwnode = fwnode;
-
-	data->mux = fwnode_handle_get(dev->fwnode);
-	put_device(dev);
-	mux_ref.node = to_software_node(data->mux);
-
-	return 0;
-}
-
 static int cht_int33fe_setup_dp(struct cht_int33fe_data *data)
 {
 	struct fwnode_handle *fwnode;
@@ -211,10 +173,9 @@ static void cht_int33fe_remove_nodes(struct cht_int33fe_data *data)
 {
 	software_node_unregister_nodes(nodes);
 
-	if (data->mux) {
-		fwnode_handle_put(data->mux);
+	if (mux_ref.node) {
+		fwnode_handle_put(software_node_fwnode(mux_ref.node));
 		mux_ref.node = NULL;
-		data->mux = NULL;
 	}
 
 	if (data->dp) {
@@ -235,14 +196,16 @@ static int cht_int33fe_add_nodes(struct cht_int33fe_data *data)
 	/* The devices that are not created in this driver need extra steps. */
 
 	/*
-	 * There is no ACPI device node for the USB role mux, so we need to find
-	 * the mux device and assign our node directly to it. That means we
-	 * depend on the mux driver. This function will return -PROBE_DEFER
-	 * until the mux device is registered.
+	 * There is no ACPI device node for the USB role mux, so we need to wait
+	 * until the mux driver has created software node for the mux device.
+	 * It means we depend on the mux driver. This function will return
+	 * -EPROBE_DEFER until the mux device is registered.
 	 */
-	ret = cht_int33fe_setup_mux(data);
-	if (ret)
+	mux_ref.node = software_node_find_by_name(NULL, "intel-xhci-usb-sw");
+	if (!mux_ref.node) {
+		ret = -EPROBE_DEFER;
 		goto err_remove_nodes;
+	}
 
 	/*
 	 * The DP connector does have ACPI device node. In this case we can just
Signature · 1 line
2.23.0.rc1
#7 lore raw
In reply to Heikki Krogerus
Quoted 5 lines · Heikki Krogerus

Hi,

There was still one bug spotted by Andy in v2. The role switch node was not removed in case of an error (patch 2/3). It is now fixed.

It would be nice to have immutable branch for these changes. There is at least some other activity regard to intel_cht_int33fe driver.

Quoted 38 lines

The cover letter from v2:

This is the second version of this series where I'm introducing that helper.

Hans and Andy! Because of the changes I made to patch 2/3, I'm not carrying your reviewed-by tags in it. I would appreciate if you could take another look at that patch.

I added a note to the kernel-doc comment in patch 1/3 that the caller of the helper function needs to release the ref count after use as proposed by Andy.

In patch 2/3, since we have to now modify the role switch descriptor, I'm filling the structure in stack memory and removing the constant static version. The content of the descriptor is copied during switch registration in any case, so we don't need to store it in the driver.

I also noticed a bug in 2/3. I never properly destroyed the software node when the mux was removed. That leak is now also fixed.

thanks,

Heikki Krogerus (3):
software node: Add software_node_find_by_name()
usb: roles: intel_xhci: Supplying software node for the role mux platform/x86: intel_cht_int33fe: Use new API to gain access to the
role switch

 drivers/base/swnode.c                         | 37 ++++++++++++
 drivers/platform/x86/intel_cht_int33fe.c      | 57 ++++---------------
 .../usb/roles/intel-xhci-usb-role-switch.c    | 27 ++++++---
 include/linux/property.h                      |  4 ++
 4 files changed, 71 insertions(+), 54 deletions(-)

--
2.23.0.rc1

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#8 lore raw
Quoted 11 lines · Andy Shevchenko
Quoted 5 lines · Heikki Krogerus

Hi,

There was still one bug spotted by Andy in v2. The role switch node was not removed in case of an error (patch 2/3). It is now fixed.

It would be nice to have immutable branch for these changes. There is at least some other activity regard to intel_cht_int33fe driver.

When they land in my devprop branch, that will be immutable.

#9 lore raw
In reply to Heikki Krogerus

Hi,

Quoted 19 lines · Heikki Krogerus

Hi,

There was still one bug spotted by Andy in v2. The role switch node was not removed in case of an error (patch 2/3). It is now fixed.

The cover letter from v2:

This is the second version of this series where I'm introducing that helper.

Hans and Andy! Because of the changes I made to patch 2/3, I'm not carrying your reviewed-by tags in it. I would appreciate if you could take another look at that patch.

I added a note to the kernel-doc comment in patch 1/3 that the caller of the helper function needs to release the ref count after use as proposed by Andy.

Quoted 7 lines

In patch 2/3, since we have to now modify the role switch descriptor, I'm filling the structure in stack memory and removing the constant static version. The content of the descriptor is copied during switch registration in any case, so we don't need to store it in the driver.

I also noticed a bug in 2/3. I never properly destroyed the software node when the mux was removed. That leak is now also fixed.

I've just given this series a try on a device with an ACPI INT33FE node describing its TypeC connector and the mux is still properly found and controlled.

So from my pov this series looks ready for merging.

Regards,

Hans