[PATCH net-next v2 1/3] rtnetlink: add alloc() method to rtnl_link_ops

[PATCH net-next v2 1/3] rtnetlink: add alloc() method to rtnl_link_ops

[Loic Poulain]

From: Johannes Berg <johannes.berg@intel.com>

In order to make rtnetlink ops that can create different
kinds of devices, like what we want to add to the WWAN
framework, the priv_size and setup parameters aren't quite
sufficient. Make this easier to manage by allowing ops to
allocate their own netdev via an @alloc method that gets
the tb netlink data.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
---
 v2: no change

 include/net/rtnetlink.h |  8 ++++++++
 net/core/rtnetlink.c    | 19 ++++++++++++++-----
 2 files changed, 22 insertions(+), 5 deletions(-)

diff --git a/include/net/rtnetlink.h b/include/net/rtnetlink.h
index 479f60e..384e800 100644
--- a/include/net/rtnetlink.h
+++ b/include/net/rtnetlink.h
@@ -37,6 +37,9 @@ static inline int rtnl_msg_family(const struct nlmsghdr *nlh)
  *	@maxtype: Highest device specific netlink attribute number
  *	@policy: Netlink policy for device specific attribute validation
  *	@validate: Optional validation function for netlink/changelink parameters
+ *	@alloc: netdev allocation function, can be %NULL and is then used
+ *		in place of alloc_netdev_mqs(), in this case @priv_size
+ *		and @setup are unused. Returns a netdev or ERR_PTR().
  *	@priv_size: sizeof net_device private space
  *	@setup: net_device setup function
  *	@newlink: Function for configuring and registering a new device
@@ -63,6 +66,11 @@ struct rtnl_link_ops {
 	const char		*kind;
 
 	size_t			priv_size;
+	struct net_device	*(*alloc)(struct nlattr *tb[],
+					  const char *ifname,
+					  unsigned char name_assign_type,
+					  unsigned int num_tx_queues,
+					  unsigned int num_rx_queues);
 	void			(*setup)(struct net_device *dev);
 
 	bool			netns_refund;
diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
index cd87c76..92c3e43 100644
--- a/net/core/rtnetlink.c
+++ b/net/core/rtnetlink.c
@@ -376,12 +376,12 @@ int __rtnl_link_register(struct rtnl_link_ops *ops)
 	if (rtnl_link_ops_get(ops->kind))
 		return -EEXIST;
 
-	/* The check for setup is here because if ops
+	/* The check for alloc/setup is here because if ops
 	 * does not have that filled up, it is not possible
 	 * to use the ops for creating device. So do not
 	 * fill up dellink as well. That disables rtnl_dellink.
 	 */
-	if (ops->setup && !ops->dellink)
+	if ((ops->alloc || ops->setup) && !ops->dellink)
 		ops->dellink = unregister_netdevice_queue;
 
 	list_add_tail(&ops->list, &link_ops);
@@ -3165,8 +3165,17 @@ struct net_device *rtnl_create_link(struct net *net, const char *ifname,
 		return ERR_PTR(-EINVAL);
 	}
 
-	dev = alloc_netdev_mqs(ops->priv_size, ifname, name_assign_type,
-			       ops->setup, num_tx_queues, num_rx_queues);
+	if (ops->alloc) {
+		dev = ops->alloc(tb, ifname, name_assign_type,
+				 num_tx_queues, num_rx_queues);
+		if (IS_ERR(dev))
+			return dev;
+	} else {
+		dev = alloc_netdev_mqs(ops->priv_size, ifname,
+				       name_assign_type, ops->setup,
+				       num_tx_queues, num_rx_queues);
+	}
+
 	if (!dev)
 		return ERR_PTR(-ENOMEM);
 
@@ -3399,7 +3408,7 @@ static int __rtnl_newlink(struct sk_buff *skb, struct nlmsghdr *nlh,
 		return -EOPNOTSUPP;
 	}
 
-	if (!ops->setup)
+	if (!ops->alloc && !ops->setup)
 		return -EOPNOTSUPP;
 
 	if (!ifname[0]) {
-- 
2.7.4

[PATCH net-next v2 2/3] rtnetlink: add IFLA_PARENT_[DEV|DEV_BUS]_NAME

[Loic Poulain]

From: Johannes Berg <johannes.berg@intel.com>

In some cases, for example in the upcoming WWAN framework changes,
there's no natural "parent netdev", so sometimes dummy netdevs are
created or similar. IFLA_PARENT_DEV_NAME is a new attribute intended to
contain a device (sysfs, struct device) name that can be used instead
when creating a new netdev, if the rtnetlink family implements it.

As suggested by Parav Pandit, we also introduce IFLA_PARENT_DEV_BUS_NAME
attribute in order to uniquely identify a device on the system (with
bus/name pair).

ip-link(8) support for the generic parent device attributes will help
us avoid code duplication, so no other link type will require a custom
code to handle the parent name attribute. E.g. the WWAN interface
creation command will looks like this:

$ ip link add wwan0-1 parent-dev wwan0 type wwan channel-id 1

So, some future subsystem (or driver) FOO will have an interface
creation command that looks like this:

$ ip link add foo1-3 parent-dev foo1 type foo bar-id 3 baz-type Y

Below is an example of dumping link info of a random device with these
new attributes:

$ ip --details link show wlp0s20f3
  4: wlp0s20f3: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue
     state UP mode DORMANT group default qlen 1000
     ...
     parent_devname 0000:00:14.3 parent_busname pci

Co-developed-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Signed-off-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Co-developed-by: Loic Poulain <loic.poulain@linaro.org>
Signed-off-by: Loic Poulain <loic.poulain@linaro.org>
Suggested-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
---
 v2: - Squashed Johannes and Sergey changes
     - Added IFLA_PARENT_DEV_BUS_NAME attribute
     - reworded commit message + introduce Sergey's comment

 include/uapi/linux/if_link.h |  7 +++++++
 net/core/rtnetlink.c         | 12 ++++++++++++
 2 files changed, 19 insertions(+)

diff --git a/include/uapi/linux/if_link.h b/include/uapi/linux/if_link.h
index a5a7f0e..4882e81 100644
--- a/include/uapi/linux/if_link.h
+++ b/include/uapi/linux/if_link.h
@@ -341,6 +341,13 @@ enum {
 	IFLA_ALT_IFNAME, /* Alternative ifname */
 	IFLA_PERM_ADDRESS,
 	IFLA_PROTO_DOWN_REASON,
+
+	/* device (sysfs) name as parent, used instead
+	 * of IFLA_LINK where there's no parent netdev
+	 */
+	IFLA_PARENT_DEV_NAME,
+	IFLA_PARENT_DEV_BUS_NAME,
+
 	__IFLA_MAX
 };
 
diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
index 92c3e43..32599f3 100644
--- a/net/core/rtnetlink.c
+++ b/net/core/rtnetlink.c
@@ -1821,6 +1821,16 @@ static int rtnl_fill_ifinfo(struct sk_buff *skb,
 	if (rtnl_fill_prop_list(skb, dev))
 		goto nla_put_failure;
 
+	if (dev->dev.parent &&
+	    nla_put_string(skb, IFLA_PARENT_DEV_NAME,
+			   dev_name(dev->dev.parent)))
+		goto nla_put_failure;
+
+	if (dev->dev.parent && dev->dev.parent->bus &&
+	    nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
+			   dev->dev.parent->bus->name))
+		goto nla_put_failure;
+
 	nlmsg_end(skb, nlh);
 	return 0;
 
@@ -1880,6 +1890,8 @@ static const struct nla_policy ifla_policy[IFLA_MAX+1] = {
 	[IFLA_PERM_ADDRESS]	= { .type = NLA_REJECT },
 	[IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
 	[IFLA_NEW_IFINDEX]	= NLA_POLICY_MIN(NLA_S32, 1),
+	[IFLA_PARENT_DEV_NAME]	= { .type = NLA_NUL_STRING },
+	[IFLA_PARENT_DEV_BUS_NAME] = { .type = NLA_NUL_STRING },
 };
 
 static const struct nla_policy ifla_info_policy[IFLA_INFO_MAX+1] = {
-- 
2.7.4

[PATCH net-next v2 3/3] wwan: add interface creation support

[Loic Poulain]

From: Johannes Berg <johannes.berg@intel.com>

Add support to create (and destroy) interfaces via a new
rtnetlink kind "wwan". The responsible driver has to use
the new wwan_register_ops() to make this possible.

Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
Signed-off-by: Loic Poulain <loic.poulain@linaro.org>
---
 v2: no change

 drivers/net/wwan/wwan_core.c | 231 +++++++++++++++++++++++++++++++++++++++++--
 include/linux/wwan.h         |  38 +++++++
 include/uapi/linux/wwan.h    |  16 +++
 3 files changed, 278 insertions(+), 7 deletions(-)
 create mode 100644 include/uapi/linux/wwan.h

diff --git a/drivers/net/wwan/wwan_core.c b/drivers/net/wwan/wwan_core.c
index 45a41ae..125426d 100644
--- a/drivers/net/wwan/wwan_core.c
+++ b/drivers/net/wwan/wwan_core.c
@@ -14,6 +14,8 @@
 #include <linux/types.h>
 #include <linux/termios.h>
 #include <linux/wwan.h>
+#include <net/rtnetlink.h>
+#include <uapi/linux/wwan.h>
 
 /* Maximum number of minors in use */
 #define WWAN_MAX_MINORS		(1 << MINORBITS)
@@ -35,10 +37,16 @@ static int wwan_major;
  *
  * @id: WWAN device unique ID.
  * @dev: Underlying device.
+ * @port_id: Current available port ID to pick.
+ * @ops: wwan device ops
+ * @ops_ctxt: context to pass to ops
  */
 struct wwan_device {
 	unsigned int id;
 	struct device dev;
+	atomic_t port_id;
+	const struct wwan_ops *ops;
+	void *ops_ctxt;
 };
 
 /**
@@ -102,7 +110,8 @@ static const struct device_type wwan_dev_type = {
 
 static int wwan_dev_parent_match(struct device *dev, const void *parent)
 {
-	return (dev->type == &wwan_dev_type && dev->parent == parent);
+	return (dev->type == &wwan_dev_type &&
+		(dev->parent == parent || dev == parent));
 }
 
 static struct wwan_device *wwan_dev_get_by_parent(struct device *parent)
@@ -116,6 +125,23 @@ static struct wwan_device *wwan_dev_get_by_parent(struct device *parent)
 	return to_wwan_dev(dev);
 }
 
+static int wwan_dev_name_match(struct device *dev, const void *name)
+{
+	return dev->type == &wwan_dev_type &&
+	       strcmp(dev_name(dev), name) == 0;
+}
+
+static struct wwan_device *wwan_dev_get_by_name(const char *name)
+{
+	struct device *dev;
+
+	dev = class_find_device(wwan_class, NULL, name, wwan_dev_name_match);
+	if (!dev)
+		return ERR_PTR(-ENODEV);
+
+	return to_wwan_dev(dev);
+}
+
 /* This function allocates and registers a new WWAN device OR if a WWAN device
  * already exist for the given parent, it gets a reference and return it.
  * This function is not exported (for now), it is called indirectly via
@@ -180,9 +206,14 @@ static void wwan_remove_dev(struct wwan_device *wwandev)
 	/* WWAN device is created and registered (get+add) along with its first
 	 * child port, and subsequent port registrations only grab a reference
 	 * (get). The WWAN device must then be unregistered (del+put) along with
-	 * its latest port, and reference simply dropped (put) otherwise.
+	 * its last port, and reference simply dropped (put) otherwise. In the
+	 * same fashion, we must not unregister it when the ops are still there.
 	 */
-	ret = device_for_each_child(&wwandev->dev, NULL, is_wwan_child);
+	if (wwandev->ops)
+		ret = 1;
+	else
+		ret = device_for_each_child(&wwandev->dev, NULL, is_wwan_child);
+
 	if (!ret)
 		device_unregister(&wwandev->dev);
 	else
@@ -750,26 +781,212 @@ static const struct file_operations wwan_port_fops = {
 	.llseek = noop_llseek,
 };
 
+int wwan_register_ops(struct device *parent, const struct wwan_ops *ops,
+		      void *ctxt)
+{
+	struct wwan_device *wwandev;
+
+	if (WARN_ON(!parent || !ops))
+		return -EINVAL;
+
+	wwandev = wwan_create_dev(parent);
+	if (!wwandev)
+		return -ENOMEM;
+
+	if (WARN_ON(wwandev->ops)) {
+		wwan_remove_dev(wwandev);
+		return -EBUSY;
+	}
+
+	if (!try_module_get(ops->owner)) {
+		wwan_remove_dev(wwandev);
+		return -ENODEV;
+	}
+
+	wwandev->ops = ops;
+	wwandev->ops_ctxt = ctxt;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(wwan_register_ops);
+
+void wwan_unregister_ops(struct device *parent)
+{
+	struct wwan_device *wwandev = wwan_dev_get_by_parent(parent);
+	bool has_ops;
+
+	if (WARN_ON(IS_ERR(wwandev)))
+		return;
+
+	has_ops = wwandev->ops;
+
+	/* put the reference obtained by wwan_dev_get_by_parent(),
+	 * we should still have one (that the owner is giving back
+	 * now) due to the ops being assigned, check that below
+	 * and return if not.
+	 */
+	put_device(&wwandev->dev);
+
+	if (WARN_ON(!has_ops))
+		return;
+
+	module_put(wwandev->ops->owner);
+
+	wwandev->ops = NULL;
+	wwandev->ops_ctxt = NULL;
+	wwan_remove_dev(wwandev);
+}
+EXPORT_SYMBOL_GPL(wwan_unregister_ops);
+
+static int wwan_rtnl_validate(struct nlattr *tb[], struct nlattr *data[],
+			      struct netlink_ext_ack *extack)
+{
+	if (!data)
+		return -EINVAL;
+
+	if (!tb[IFLA_PARENT_DEV_NAME])
+		return -EINVAL;
+
+	if (!data[IFLA_WWAN_LINK_ID])
+		return -EINVAL;
+
+	return 0;
+}
+
+static struct device_type wwan_type = { .name = "wwan" };
+
+static struct net_device *wwan_rtnl_alloc(struct nlattr *tb[],
+					  const char *ifname,
+					  unsigned char name_assign_type,
+					  unsigned int num_tx_queues,
+					  unsigned int num_rx_queues)
+{
+	const char *devname = nla_data(tb[IFLA_PARENT_DEV_NAME]);
+	struct wwan_device *wwandev = wwan_dev_get_by_name(devname);
+	struct net_device *dev;
+
+	if (IS_ERR(wwandev))
+		return ERR_CAST(wwandev);
+
+	/* only supported if ops were registered (not just ports) */
+	if (!wwandev->ops) {
+		dev = ERR_PTR(-EOPNOTSUPP);
+		goto out;
+	}
+
+	dev = alloc_netdev_mqs(wwandev->ops->priv_size, ifname, name_assign_type,
+			       wwandev->ops->setup, num_tx_queues, num_rx_queues);
+
+	if (dev) {
+		SET_NETDEV_DEV(dev, &wwandev->dev);
+		SET_NETDEV_DEVTYPE(dev, &wwan_type);
+	}
+
+out:
+	/* release the reference */
+	put_device(&wwandev->dev);
+	return dev;
+}
+
+static int wwan_rtnl_newlink(struct net *src_net, struct net_device *dev,
+			     struct nlattr *tb[], struct nlattr *data[],
+			     struct netlink_ext_ack *extack)
+{
+	struct wwan_device *wwandev = wwan_dev_get_by_parent(dev->dev.parent);
+	u32 link_id = nla_get_u32(data[IFLA_WWAN_LINK_ID]);
+	int ret;
+
+	if (IS_ERR(wwandev))
+		return PTR_ERR(wwandev);
+
+	/* shouldn't have a netdev (left) with us as parent so WARN */
+	if (WARN_ON(!wwandev->ops)) {
+		ret = -EOPNOTSUPP;
+		goto out;
+	}
+
+	if (wwandev->ops->newlink)
+		ret = wwandev->ops->newlink(wwandev->ops_ctxt, dev,
+					    link_id, extack);
+	else
+		ret = register_netdevice(dev);
+
+out:
+	/* release the reference */
+	put_device(&wwandev->dev);
+	return ret;
+}
+
+static void wwan_rtnl_dellink(struct net_device *dev, struct list_head *head)
+{
+	struct wwan_device *wwandev = wwan_dev_get_by_parent(dev->dev.parent);
+
+	if (IS_ERR(wwandev))
+		return;
+
+	/* shouldn't have a netdev (left) with us as parent so WARN */
+	if (WARN_ON(!wwandev->ops))
+		goto out;
+
+	if (wwandev->ops->dellink)
+		wwandev->ops->dellink(wwandev->ops_ctxt, dev, head);
+	else
+		unregister_netdevice(dev);
+
+out:
+	/* release the reference */
+	put_device(&wwandev->dev);
+}
+
+static const struct nla_policy wwan_rtnl_policy[IFLA_WWAN_MAX + 1] = {
+	[IFLA_WWAN_LINK_ID] = { .type = NLA_U32 },
+};
+
+static struct rtnl_link_ops wwan_rtnl_link_ops __read_mostly = {
+	.kind = "wwan",
+	.maxtype = __IFLA_WWAN_MAX,
+	.alloc = wwan_rtnl_alloc,
+	.validate = wwan_rtnl_validate,
+	.newlink = wwan_rtnl_newlink,
+	.dellink = wwan_rtnl_dellink,
+	.policy = wwan_rtnl_policy,
+};
+
 static int __init wwan_init(void)
 {
+	int err;
+
+	err = rtnl_link_register(&wwan_rtnl_link_ops);
+	if (err)
+		return err;
+
 	wwan_class = class_create(THIS_MODULE, "wwan");
-	if (IS_ERR(wwan_class))
-		return PTR_ERR(wwan_class);
+	if (IS_ERR(wwan_class)) {
+		err = PTR_ERR(wwan_class);
+		goto unregister;
+	}
 
 	/* chrdev used for wwan ports */
 	wwan_major = __register_chrdev(0, 0, WWAN_MAX_MINORS, "wwan_port",
 				       &wwan_port_fops);
 	if (wwan_major < 0) {
-		class_destroy(wwan_class);
-		return wwan_major;
+		err = wwan_major;
+		goto destroy;
 	}
 
 	return 0;
+
+destroy:
+	class_destroy(wwan_class);
+unregister:
+	rtnl_link_unregister(&wwan_rtnl_link_ops);
+	return err;
 }
 
 static void __exit wwan_exit(void)
 {
 	__unregister_chrdev(wwan_major, 0, WWAN_MAX_MINORS, "wwan_port");
+	rtnl_link_unregister(&wwan_rtnl_link_ops);
 	class_destroy(wwan_class);
 }
 
diff --git a/include/linux/wwan.h b/include/linux/wwan.h
index fa33cc1..7071f95 100644
--- a/include/linux/wwan.h
+++ b/include/linux/wwan.h
@@ -7,6 +7,7 @@
 #include <linux/device.h>
 #include <linux/kernel.h>
 #include <linux/skbuff.h>
+#include <linux/netlink.h>
 
 /**
  * enum wwan_port_type - WWAN port types
@@ -116,4 +117,41 @@ void wwan_port_txon(struct wwan_port *port);
  */
 void *wwan_port_get_drvdata(struct wwan_port *port);
 
+/**
+ * struct wwan_ops - WWAN device ops
+ * @owner: module owner of the WWAN ops
+ * @priv_size: size of private netdev data area
+ * @setup: set up a new netdev
+ * @newlink: register the new netdev
+ * @dellink: remove the given netdev
+ */
+struct wwan_ops {
+	struct module *owner;
+	unsigned int priv_size;
+	void (*setup)(struct net_device *dev);
+	int (*newlink)(void *ctxt, struct net_device *dev,
+		       u32 if_id, struct netlink_ext_ack *extack);
+	void (*dellink)(void *ctxt, struct net_device *dev,
+			struct list_head *head);
+};
+
+/**
+ * wwan_register_ops - register WWAN device ops
+ * @parent: Device to use as parent and shared by all WWAN ports and
+ *	created netdevs
+ * @ops: operations to register
+ * @ctxt: context to pass to operations
+ *
+ * Returns: 0 on success, a negative error code on failure
+ */
+int wwan_register_ops(struct device *parent, const struct wwan_ops *ops,
+		      void *ctxt);
+
+/**
+ * wwan_unregister_ops - remove WWAN device ops
+ * @parent: Device to use as parent and shared by all WWAN ports and
+ *	created netdevs
+ */
+void wwan_unregister_ops(struct device *parent);
+
 #endif /* __WWAN_H */
diff --git a/include/uapi/linux/wwan.h b/include/uapi/linux/wwan.h
new file mode 100644
index 0000000..32a2720
--- /dev/null
+++ b/include/uapi/linux/wwan.h
@@ -0,0 +1,16 @@
+/* SPDX-License-Identifier: GPL-2.0-only WITH Linux-syscall-note */
+/*
+ * Copyright (C) 2021 Intel Corporation.
+ */
+#ifndef _UAPI_WWAN_H_
+#define _UAPI_WWAN_H_
+
+enum {
+	IFLA_WWAN_UNSPEC,
+	IFLA_WWAN_LINK_ID, /* u32 */
+
+	__IFLA_WWAN_MAX
+};
+#define IFLA_WWAN_MAX (__IFLA_WWAN_MAX - 1)
+
+#endif /* _UAPI_WWAN_H_ */
-- 
2.7.4

RE: [PATCH net-next v2 3/3] wwan: add interface creation support

[Parav Pandit]

> From: Loic Poulain <loic.poulain@linaro.org>
> Sent: Thursday, June 10, 2021 11:15 PM
> 
> From: Johannes Berg <johannes.berg@intel.com>
> 
> Add support to create (and destroy) interfaces via a new rtnetlink kind
> "wwan". The responsible driver has to use the new wwan_register_ops() to
> make this possible.
The responsible driver must be in same series.

> 
> +int wwan_register_ops(struct device *parent, const struct wwan_ops *ops,
> +		      void *ctxt)
> +{
> +	struct wwan_device *wwandev;
> +
> +	if (WARN_ON(!parent || !ops))
> +		return -EINVAL;
> +
> +	wwandev = wwan_create_dev(parent);
> +	if (!wwandev)
> +		return -ENOMEM;
> +
> +	if (WARN_ON(wwandev->ops)) {
> +		wwan_remove_dev(wwandev);
> +		return -EBUSY;
> +	}
> +
> +	if (!try_module_get(ops->owner)) {
> +		wwan_remove_dev(wwandev);
> +		return -ENODEV;
> +	}
> +
> +	wwandev->ops = ops;
> +	wwandev->ops_ctxt = ctxt;
> +
> +	return 0;
> +}
> +EXPORT_SYMBOL_GPL(wwan_register_ops);
> +
I am not familiar with wwan so cannot comment much on code correctness.
But I know for sure that this series is in complete.
Above wwan_register_ops() and below symbols are exported and never used in this kernel.
This practice is no longer in use to have dead code.
Please remove this dead code or have the user of these APIs in same series along with cover letter.

> +void wwan_unregister_ops(struct device *parent) {
> +	struct wwan_device *wwandev =
> wwan_dev_get_by_parent(parent);
> +	bool has_ops;
> +
> +	if (WARN_ON(IS_ERR(wwandev)))
> +		return;
> +
> +	has_ops = wwandev->ops;
> +
> +	/* put the reference obtained by wwan_dev_get_by_parent(),
> +	 * we should still have one (that the owner is giving back
> +	 * now) due to the ops being assigned, check that below
> +	 * and return if not.
> +	 */
> +	put_device(&wwandev->dev);
> +
> +	if (WARN_ON(!has_ops))
> +		return;
> +
> +	module_put(wwandev->ops->owner);
> +
> +	wwandev->ops = NULL;
> +	wwandev->ops_ctxt = NULL;
> +	wwan_remove_dev(wwandev);
> +}
> +EXPORT_SYMBOL_GPL(wwan_unregister_ops);
> +
> +static int wwan_rtnl_validate(struct nlattr *tb[], struct nlattr *data[],
> +			      struct netlink_ext_ack *extack) {
> +	if (!data)
> +		return -EINVAL;
> +
> +	if (!tb[IFLA_PARENT_DEV_NAME])
> +		return -EINVAL;
> +
> +	if (!data[IFLA_WWAN_LINK_ID])
> +		return -EINVAL;
> +
> +	return 0;
> +}
> +
> +static struct device_type wwan_type = { .name = "wwan" };
> +
> +static struct net_device *wwan_rtnl_alloc(struct nlattr *tb[],
> +					  const char *ifname,
> +					  unsigned char name_assign_type,
> +					  unsigned int num_tx_queues,
> +					  unsigned int num_rx_queues)
> +{
> +	const char *devname = nla_data(tb[IFLA_PARENT_DEV_NAME]);
> +	struct wwan_device *wwandev =
> wwan_dev_get_by_name(devname);
> +	struct net_device *dev;
> +
> +	if (IS_ERR(wwandev))
> +		return ERR_CAST(wwandev);
> +
> +	/* only supported if ops were registered (not just ports) */
> +	if (!wwandev->ops) {
> +		dev = ERR_PTR(-EOPNOTSUPP);
> +		goto out;
> +	}
> +
> +	dev = alloc_netdev_mqs(wwandev->ops->priv_size, ifname,
> name_assign_type,
> +			       wwandev->ops->setup, num_tx_queues,
> num_rx_queues);
> +
> +	if (dev) {
> +		SET_NETDEV_DEV(dev, &wwandev->dev);
> +		SET_NETDEV_DEVTYPE(dev, &wwan_type);
> +	}
> +
> +out:
> +	/* release the reference */
> +	put_device(&wwandev->dev);
> +	return dev;
> +}
> +
> +static int wwan_rtnl_newlink(struct net *src_net, struct net_device *dev,
> +			     struct nlattr *tb[], struct nlattr *data[],
> +			     struct netlink_ext_ack *extack) {
> +	struct wwan_device *wwandev = wwan_dev_get_by_parent(dev-
> >dev.parent);
> +	u32 link_id = nla_get_u32(data[IFLA_WWAN_LINK_ID]);
> +	int ret;
> +
> +	if (IS_ERR(wwandev))
> +		return PTR_ERR(wwandev);
> +
> +	/* shouldn't have a netdev (left) with us as parent so WARN */
> +	if (WARN_ON(!wwandev->ops)) {
> +		ret = -EOPNOTSUPP;
> +		goto out;
> +	}
> +
> +	if (wwandev->ops->newlink)
> +		ret = wwandev->ops->newlink(wwandev->ops_ctxt, dev,
> +					    link_id, extack);
> +	else
> +		ret = register_netdevice(dev);
> +
> +out:
> +	/* release the reference */
> +	put_device(&wwandev->dev);
> +	return ret;
> +}
> +
> +static void wwan_rtnl_dellink(struct net_device *dev, struct list_head
> +*head) {
> +	struct wwan_device *wwandev = wwan_dev_get_by_parent(dev-
> >dev.parent);
> +
> +	if (IS_ERR(wwandev))
> +		return;
> +
> +	/* shouldn't have a netdev (left) with us as parent so WARN */
> +	if (WARN_ON(!wwandev->ops))
> +		goto out;
> +
> +	if (wwandev->ops->dellink)
> +		wwandev->ops->dellink(wwandev->ops_ctxt, dev, head);
> +	else
> +		unregister_netdevice(dev);
> +
> +out:
> +	/* release the reference */
> +	put_device(&wwandev->dev);
> +}
> +
> +static const struct nla_policy wwan_rtnl_policy[IFLA_WWAN_MAX + 1] = {
> +	[IFLA_WWAN_LINK_ID] = { .type = NLA_U32 }, };
> +
> +static struct rtnl_link_ops wwan_rtnl_link_ops __read_mostly = {
> +	.kind = "wwan",
> +	.maxtype = __IFLA_WWAN_MAX,
> +	.alloc = wwan_rtnl_alloc,
> +	.validate = wwan_rtnl_validate,
> +	.newlink = wwan_rtnl_newlink,
> +	.dellink = wwan_rtnl_dellink,
> +	.policy = wwan_rtnl_policy,
> +};
> +
>  static int __init wwan_init(void)
>  {
> +	int err;
> +
> +	err = rtnl_link_register(&wwan_rtnl_link_ops);
> +	if (err)
> +		return err;
> +
>  	wwan_class = class_create(THIS_MODULE, "wwan");
> -	if (IS_ERR(wwan_class))
> -		return PTR_ERR(wwan_class);
> +	if (IS_ERR(wwan_class)) {
> +		err = PTR_ERR(wwan_class);
> +		goto unregister;
> +	}
> 
>  	/* chrdev used for wwan ports */
>  	wwan_major = __register_chrdev(0, 0, WWAN_MAX_MINORS,
> "wwan_port",
>  				       &wwan_port_fops);
>  	if (wwan_major < 0) {
> -		class_destroy(wwan_class);
> -		return wwan_major;
> +		err = wwan_major;
> +		goto destroy;
>  	}
> 
>  	return 0;
> +
> +destroy:
> +	class_destroy(wwan_class);
> +unregister:
> +	rtnl_link_unregister(&wwan_rtnl_link_ops);
> +	return err;
>  }
> 
>  static void __exit wwan_exit(void)
>  {
>  	__unregister_chrdev(wwan_major, 0, WWAN_MAX_MINORS,
> "wwan_port");
> +	rtnl_link_unregister(&wwan_rtnl_link_ops);
>  	class_destroy(wwan_class);
>  }
> 
> diff --git a/include/linux/wwan.h b/include/linux/wwan.h index
> fa33cc1..7071f95 100644
> --- a/include/linux/wwan.h
> +++ b/include/linux/wwan.h
> @@ -7,6 +7,7 @@
>  #include <linux/device.h>
>  #include <linux/kernel.h>
>  #include <linux/skbuff.h>
> +#include <linux/netlink.h>
> 
>  /**
>   * enum wwan_port_type - WWAN port types @@ -116,4 +117,41 @@ void
> wwan_port_txon(struct wwan_port *port);
>   */
>  void *wwan_port_get_drvdata(struct wwan_port *port);
> 
> +/**
> + * struct wwan_ops - WWAN device ops
> + * @owner: module owner of the WWAN ops
> + * @priv_size: size of private netdev data area
> + * @setup: set up a new netdev
> + * @newlink: register the new netdev
> + * @dellink: remove the given netdev
> + */
> +struct wwan_ops {
> +	struct module *owner;
> +	unsigned int priv_size;
> +	void (*setup)(struct net_device *dev);
> +	int (*newlink)(void *ctxt, struct net_device *dev,
> +		       u32 if_id, struct netlink_ext_ack *extack);
> +	void (*dellink)(void *ctxt, struct net_device *dev,
> +			struct list_head *head);
> +};
> +
> +/**
> + * wwan_register_ops - register WWAN device ops
> + * @parent: Device to use as parent and shared by all WWAN ports and
> + *	created netdevs
> + * @ops: operations to register
> + * @ctxt: context to pass to operations
> + *
> + * Returns: 0 on success, a negative error code on failure  */ int
Kdoc comment section goes in the .c file.

RE: [PATCH net-next v2 2/3] rtnetlink: add IFLA_PARENT_[DEV|DEV_BUS]_NAME

[Parav Pandit]

> From: Loic Poulain <loic.poulain@linaro.org>
> Sent: Thursday, June 10, 2021 11:15 PM
> 
> From: Johannes Berg <johannes.berg@intel.com>
> 
> In some cases, for example in the upcoming WWAN framework changes,
> there's no natural "parent netdev", so sometimes dummy netdevs are
> created or similar. IFLA_PARENT_DEV_NAME is a new attribute intended to
> contain a device (sysfs, struct device) name that can be used instead when
> creating a new netdev, if the rtnetlink family implements it.
> 
> As suggested by Parav Pandit, we also introduce
> IFLA_PARENT_DEV_BUS_NAME attribute in order to uniquely identify a
> device on the system (with bus/name pair).
> 
> ip-link(8) support for the generic parent device attributes will help us avoid
> code duplication, so no other link type will require a custom code to handle
> the parent name attribute. E.g. the WWAN interface creation command will
> looks like this:
> 
> $ ip link add wwan0-1 parent-dev wwan0 type wwan channel-id 1
> 
> So, some future subsystem (or driver) FOO will have an interface creation
> command that looks like this:
> 
> $ ip link add foo1-3 parent-dev foo1 type foo bar-id 3 baz-type Y
> 
> Below is an example of dumping link info of a random device with these new
> attributes:
> 
> $ ip --details link show wlp0s20f3
>   4: wlp0s20f3: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc
> noqueue
>      state UP mode DORMANT group default qlen 1000
>      ...
>      parent_devname 0000:00:14.3 parent_busname pci

Showing bus first followed device is more preferred approach to see hierarchy.
Please change their sequence.

You should drop "name" suffix.
"parent_bus" and "parent_dev" are just fine.

> 
> Co-developed-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
> Signed-off-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
> Co-developed-by: Loic Poulain <loic.poulain@linaro.org>
> Signed-off-by: Loic Poulain <loic.poulain@linaro.org>
> Suggested-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
> Signed-off-by: Johannes Berg <johannes.berg@intel.com>
> ---
>  v2: - Squashed Johannes and Sergey changes
>      - Added IFLA_PARENT_DEV_BUS_NAME attribute
>      - reworded commit message + introduce Sergey's comment
> 
>  include/uapi/linux/if_link.h |  7 +++++++
>  net/core/rtnetlink.c         | 12 ++++++++++++
>  2 files changed, 19 insertions(+)
> 
> diff --git a/include/uapi/linux/if_link.h b/include/uapi/linux/if_link.h index
> a5a7f0e..4882e81 100644
> --- a/include/uapi/linux/if_link.h
> +++ b/include/uapi/linux/if_link.h
> @@ -341,6 +341,13 @@ enum {
>  	IFLA_ALT_IFNAME, /* Alternative ifname */
>  	IFLA_PERM_ADDRESS,
>  	IFLA_PROTO_DOWN_REASON,
> +
> +	/* device (sysfs) name as parent, used instead
> +	 * of IFLA_LINK where there's no parent netdev
> +	 */
> +	IFLA_PARENT_DEV_NAME,
> +	IFLA_PARENT_DEV_BUS_NAME,
> +
>  	__IFLA_MAX
>  };
> 
> diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c index 92c3e43..32599f3
> 100644
> --- a/net/core/rtnetlink.c
> +++ b/net/core/rtnetlink.c
> @@ -1821,6 +1821,16 @@ static int rtnl_fill_ifinfo(struct sk_buff *skb,
>  	if (rtnl_fill_prop_list(skb, dev))
>  		goto nla_put_failure;
> 
> +	if (dev->dev.parent &&
> +	    nla_put_string(skb, IFLA_PARENT_DEV_NAME,
> +			   dev_name(dev->dev.parent)))
> +		goto nla_put_failure;
> +
> +	if (dev->dev.parent && dev->dev.parent->bus &&
> +	    nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
> +			   dev->dev.parent->bus->name))
> +		goto nla_put_failure;
> +
>  	nlmsg_end(skb, nlh);
>  	return 0;
> 
> @@ -1880,6 +1890,8 @@ static const struct nla_policy
> ifla_policy[IFLA_MAX+1] = {
>  	[IFLA_PERM_ADDRESS]	= { .type = NLA_REJECT },
>  	[IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
>  	[IFLA_NEW_IFINDEX]	= NLA_POLICY_MIN(NLA_S32, 1),
> +	[IFLA_PARENT_DEV_NAME]	= { .type = NLA_NUL_STRING },
> +	[IFLA_PARENT_DEV_BUS_NAME] = { .type = NLA_NUL_STRING },
>  };
> 
This hunk should go in the patch that enables users to use these fields to specify it for new link creation.

Re: [PATCH net-next v2 2/3] rtnetlink: add IFLA_PARENT_[DEV|DEV_BUS]_NAME

[Loic Poulain]

Hi Parav,

On Fri, 11 Jun 2021 at 15:01, Parav Pandit <parav@nvidia.com> wrote:
>
>
>
> > From: Loic Poulain <loic.poulain@linaro.org>
> > Sent: Thursday, June 10, 2021 11:15 PM
> >
> > From: Johannes Berg <johannes.berg@intel.com>
> >
> > In some cases, for example in the upcoming WWAN framework changes,
> > there's no natural "parent netdev", so sometimes dummy netdevs are
> > created or similar. IFLA_PARENT_DEV_NAME is a new attribute intended to
> > contain a device (sysfs, struct device) name that can be used instead when
> > creating a new netdev, if the rtnetlink family implements it.
> >
> > As suggested by Parav Pandit, we also introduce
> > IFLA_PARENT_DEV_BUS_NAME attribute in order to uniquely identify a
> > device on the system (with bus/name pair).
> >
> > ip-link(8) support for the generic parent device attributes will help us avoid
> > code duplication, so no other link type will require a custom code to handle
> > the parent name attribute. E.g. the WWAN interface creation command will
> > looks like this:
> >
> > $ ip link add wwan0-1 parent-dev wwan0 type wwan channel-id 1
> >
> > So, some future subsystem (or driver) FOO will have an interface creation
> > command that looks like this:
> >
> > $ ip link add foo1-3 parent-dev foo1 type foo bar-id 3 baz-type Y
> >
> > Below is an example of dumping link info of a random device with these new
> > attributes:
> >
> > $ ip --details link show wlp0s20f3
> >   4: wlp0s20f3: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc
> > noqueue
> >      state UP mode DORMANT group default qlen 1000
> >      ...
> >      parent_devname 0000:00:14.3 parent_busname pci
>
> Showing bus first followed device is more preferred approach to see hierarchy.
> Please change their sequence.
>
> You should drop "name" suffix.
> "parent_bus" and "parent_dev" are just fine.

Do you think it can also be dropped for the IFLA symbol
(IFLA_PARENT_DEV && IFLA_PARENT_BUS_NAME).

>
> >
> > Co-developed-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
> > Signed-off-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
> > Co-developed-by: Loic Poulain <loic.poulain@linaro.org>
> > Signed-off-by: Loic Poulain <loic.poulain@linaro.org>
> > Suggested-by: Sergey Ryazanov <ryazanov.s.a@gmail.com>
> > Signed-off-by: Johannes Berg <johannes.berg@intel.com>
> > ---
> >  v2: - Squashed Johannes and Sergey changes
> >      - Added IFLA_PARENT_DEV_BUS_NAME attribute
> >      - reworded commit message + introduce Sergey's comment
> >
> >  include/uapi/linux/if_link.h |  7 +++++++
> >  net/core/rtnetlink.c         | 12 ++++++++++++
> >  2 files changed, 19 insertions(+)
> >
> > diff --git a/include/uapi/linux/if_link.h b/include/uapi/linux/if_link.h index
> > a5a7f0e..4882e81 100644
> > --- a/include/uapi/linux/if_link.h
> > +++ b/include/uapi/linux/if_link.h
> > @@ -341,6 +341,13 @@ enum {
> >       IFLA_ALT_IFNAME, /* Alternative ifname */
> >       IFLA_PERM_ADDRESS,
> >       IFLA_PROTO_DOWN_REASON,
> > +
> > +     /* device (sysfs) name as parent, used instead
> > +      * of IFLA_LINK where there's no parent netdev
> > +      */
> > +     IFLA_PARENT_DEV_NAME,
> > +     IFLA_PARENT_DEV_BUS_NAME,
> > +
> >       __IFLA_MAX
> >  };
> >
> > diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c index 92c3e43..32599f3
> > 100644
> > --- a/net/core/rtnetlink.c
> > +++ b/net/core/rtnetlink.c
> > @@ -1821,6 +1821,16 @@ static int rtnl_fill_ifinfo(struct sk_buff *skb,
> >       if (rtnl_fill_prop_list(skb, dev))
> >               goto nla_put_failure;
> >
> > +     if (dev->dev.parent &&
> > +         nla_put_string(skb, IFLA_PARENT_DEV_NAME,
> > +                        dev_name(dev->dev.parent)))
> > +             goto nla_put_failure;
> > +
> > +     if (dev->dev.parent && dev->dev.parent->bus &&
> > +         nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
> > +                        dev->dev.parent->bus->name))
> > +             goto nla_put_failure;
> > +
> >       nlmsg_end(skb, nlh);
> >       return 0;
> >
> > @@ -1880,6 +1890,8 @@ static const struct nla_policy
> > ifla_policy[IFLA_MAX+1] = {
> >       [IFLA_PERM_ADDRESS]     = { .type = NLA_REJECT },
> >       [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
> >       [IFLA_NEW_IFINDEX]      = NLA_POLICY_MIN(NLA_S32, 1),
> > +     [IFLA_PARENT_DEV_NAME]  = { .type = NLA_NUL_STRING },
> > +     [IFLA_PARENT_DEV_BUS_NAME] = { .type = NLA_NUL_STRING },
> >  };
> >
> This hunk should go in the patch that enables users to use these fields to specify it for new link creation.

ack.

Thx,
Loic

RE: [PATCH net-next v2 2/3] rtnetlink: add IFLA_PARENT_[DEV|DEV_BUS]_NAME

[Parav Pandit]

> From: Loic Poulain <loic.poulain@linaro.org>
> Sent: Friday, June 11, 2021 8:37 PM
> 
> Hi Parav,
> 
> On Fri, 11 Jun 2021 at 15:01, Parav Pandit <parav@nvidia.com> wrote:
> >
> >
> >
> > > From: Loic Poulain <loic.poulain@linaro.org>
> > > Sent: Thursday, June 10, 2021 11:15 PM
> > >
> > > From: Johannes Berg <johannes.berg@intel.com>
> > >
> > > In some cases, for example in the upcoming WWAN framework changes,
> > > there's no natural "parent netdev", so sometimes dummy netdevs are
> > > created or similar. IFLA_PARENT_DEV_NAME is a new attribute intended
> > > to contain a device (sysfs, struct device) name that can be used
> > > instead when creating a new netdev, if the rtnetlink family implements it.
> > >
> > > As suggested by Parav Pandit, we also introduce
> > > IFLA_PARENT_DEV_BUS_NAME attribute in order to uniquely identify a
> > > device on the system (with bus/name pair).
> > >
> > > ip-link(8) support for the generic parent device attributes will
> > > help us avoid code duplication, so no other link type will require a
> > > custom code to handle the parent name attribute. E.g. the WWAN
> > > interface creation command will looks like this:
> > >
> > > $ ip link add wwan0-1 parent-dev wwan0 type wwan channel-id 1
> > >
> > > So, some future subsystem (or driver) FOO will have an interface
> > > creation command that looks like this:
> > >
> > > $ ip link add foo1-3 parent-dev foo1 type foo bar-id 3 baz-type Y
> > >
> > > Below is an example of dumping link info of a random device with
> > > these new
> > > attributes:
> > >
> > > $ ip --details link show wlp0s20f3
> > >   4: wlp0s20f3: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc
> > > noqueue
> > >      state UP mode DORMANT group default qlen 1000
> > >      ...
> > >      parent_devname 0000:00:14.3 parent_busname pci
> >
> > Showing bus first followed device is more preferred approach to see
> hierarchy.
> > Please change their sequence.
> >
> > You should drop "name" suffix.
> > "parent_bus" and "parent_dev" are just fine.
> 
> Do you think it can also be dropped for the IFLA symbol (IFLA_PARENT_DEV
> && IFLA_PARENT_BUS_NAME).
IFLA symbol with _NAME is fine as it covers what the field is.

Re: [PATCH net-next v2 2/3] rtnetlink: add IFLA_PARENT_[DEV|DEV_BUS]_NAME

[Loic Poulain]

Hi Parav,

On Fri, 11 Jun 2021 at 15:01, Parav Pandit <parav@nvidia.com> wrote:
>
>
>
> > From: Loic Poulain <loic.poulain@linaro.org>
> > Sent: Thursday, June 10, 2021 11:15 PM
> >
> > From: Johannes Berg <johannes.berg@intel.com>
> >
> > In some cases, for example in the upcoming WWAN framework changes,
> > there's no natural "parent netdev", so sometimes dummy netdevs are
> > created or similar. IFLA_PARENT_DEV_NAME is a new attribute intended to
> > contain a device (sysfs, struct device) name that can be used instead when
> > creating a new netdev, if the rtnetlink family implements it.
> >
> > As suggested by Parav Pandit, we also introduce
> > IFLA_PARENT_DEV_BUS_NAME attribute in order to uniquely identify a
> > device on the system (with bus/name pair).

[...]

> > diff --git a/include/uapi/linux/if_link.h b/include/uapi/linux/if_link.h index
> > a5a7f0e..4882e81 100644
> > --- a/include/uapi/linux/if_link.h
> > +++ b/include/uapi/linux/if_link.h
> > @@ -341,6 +341,13 @@ enum {
> >       IFLA_ALT_IFNAME, /* Alternative ifname */
> >       IFLA_PERM_ADDRESS,
> >       IFLA_PROTO_DOWN_REASON,
> > +
> > +     /* device (sysfs) name as parent, used instead
> > +      * of IFLA_LINK where there's no parent netdev
> > +      */
> > +     IFLA_PARENT_DEV_NAME,
> > +     IFLA_PARENT_DEV_BUS_NAME,
> > +
> >       __IFLA_MAX
> >  };
> >
> > diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c index 92c3e43..32599f3
> > 100644
> > --- a/net/core/rtnetlink.c
> > +++ b/net/core/rtnetlink.c
> > @@ -1821,6 +1821,16 @@ static int rtnl_fill_ifinfo(struct sk_buff *skb,
> >       if (rtnl_fill_prop_list(skb, dev))
> >               goto nla_put_failure;
> >
> > +     if (dev->dev.parent &&
> > +         nla_put_string(skb, IFLA_PARENT_DEV_NAME,
> > +                        dev_name(dev->dev.parent)))
> > +             goto nla_put_failure;
> > +
> > +     if (dev->dev.parent && dev->dev.parent->bus &&
> > +         nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
> > +                        dev->dev.parent->bus->name))
> > +             goto nla_put_failure;
> > +
> >       nlmsg_end(skb, nlh);
> >       return 0;
> >
> > @@ -1880,6 +1890,8 @@ static const struct nla_policy
> > ifla_policy[IFLA_MAX+1] = {
> >       [IFLA_PERM_ADDRESS]     = { .type = NLA_REJECT },
> >       [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
> >       [IFLA_NEW_IFINDEX]      = NLA_POLICY_MIN(NLA_S32, 1),
> > +     [IFLA_PARENT_DEV_NAME]  = { .type = NLA_NUL_STRING },
> > +     [IFLA_PARENT_DEV_BUS_NAME] = { .type = NLA_NUL_STRING },
> >  };
> >
> This hunk should go in the patch that enables users to use these fields to specify it for new link creation.

Don't get it, the previous changes I see in the tree change both
if_link.h and rtnetlink.c for new atributes (e.g. f74877a5457d). Can
you elaborate on what you expect here?

Thanks,
Loic

RE: [PATCH net-next v2 2/3] rtnetlink: add IFLA_PARENT_[DEV|DEV_BUS]_NAME

[Parav Pandit]

> From: Loic Poulain <loic.poulain@linaro.org>
> Sent: Friday, June 11, 2021 9:16 PM
> 
> Hi Parav,
> 
> On Fri, 11 Jun 2021 at 15:01, Parav Pandit <parav@nvidia.com> wrote:
> >
> >
> >
> > > From: Loic Poulain <loic.poulain@linaro.org>
> > > Sent: Thursday, June 10, 2021 11:15 PM
> > >
> > > From: Johannes Berg <johannes.berg@intel.com>
> > >
> > > In some cases, for example in the upcoming WWAN framework changes,
> > > there's no natural "parent netdev", so sometimes dummy netdevs are
> > > created or similar. IFLA_PARENT_DEV_NAME is a new attribute intended
> > > to contain a device (sysfs, struct device) name that can be used
> > > instead when creating a new netdev, if the rtnetlink family implements it.
> > >
> > > As suggested by Parav Pandit, we also introduce
> > > IFLA_PARENT_DEV_BUS_NAME attribute in order to uniquely identify a
> > > device on the system (with bus/name pair).
> 
> [...]
> 
> > > diff --git a/include/uapi/linux/if_link.h
> > > b/include/uapi/linux/if_link.h index
> > > a5a7f0e..4882e81 100644
> > > --- a/include/uapi/linux/if_link.h
> > > +++ b/include/uapi/linux/if_link.h
> > > @@ -341,6 +341,13 @@ enum {
> > >       IFLA_ALT_IFNAME, /* Alternative ifname */
> > >       IFLA_PERM_ADDRESS,
> > >       IFLA_PROTO_DOWN_REASON,
> > > +
> > > +     /* device (sysfs) name as parent, used instead
> > > +      * of IFLA_LINK where there's no parent netdev
> > > +      */
> > > +     IFLA_PARENT_DEV_NAME,
> > > +     IFLA_PARENT_DEV_BUS_NAME,
> > > +
> > >       __IFLA_MAX
> > >  };
> > >
> > > diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c index
> > > 92c3e43..32599f3
> > > 100644
> > > --- a/net/core/rtnetlink.c
> > > +++ b/net/core/rtnetlink.c
> > > @@ -1821,6 +1821,16 @@ static int rtnl_fill_ifinfo(struct sk_buff *skb,
> > >       if (rtnl_fill_prop_list(skb, dev))
> > >               goto nla_put_failure;
> > >
> > > +     if (dev->dev.parent &&
> > > +         nla_put_string(skb, IFLA_PARENT_DEV_NAME,
> > > +                        dev_name(dev->dev.parent)))
> > > +             goto nla_put_failure;
> > > +
> > > +     if (dev->dev.parent && dev->dev.parent->bus &&
> > > +         nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
> > > +                        dev->dev.parent->bus->name))
> > > +             goto nla_put_failure;
> > > +
> > >       nlmsg_end(skb, nlh);
> > >       return 0;
> > >
> > > @@ -1880,6 +1890,8 @@ static const struct nla_policy
> > > ifla_policy[IFLA_MAX+1] = {
> > >       [IFLA_PERM_ADDRESS]     = { .type = NLA_REJECT },
> > >       [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
> > >       [IFLA_NEW_IFINDEX]      = NLA_POLICY_MIN(NLA_S32, 1),
> > > +     [IFLA_PARENT_DEV_NAME]  = { .type = NLA_NUL_STRING },
> > > +     [IFLA_PARENT_DEV_BUS_NAME] = { .type = NLA_NUL_STRING },
> > >  };
> > >
> > This hunk should go in the patch that enables users to use these fields to
> specify it for new link creation.
> 
> Don't get it, the previous changes I see in the tree change both if_link.h and
> rtnetlink.c for new atributes (e.g. f74877a5457d). Can you elaborate on what
> you expect here?
> 
Commit f74877a5457d did in same patch because one of the objective of commit f74877a5457d is to also block PERM_ADDR using policy NLA_REJECT.
This patch-2 is not enabling user to pass these params via new_link command.
It is done in a later patch. So ifla_policy doesn’t need to have these fields in this patch.

Re: [PATCH net-next v2 2/3] rtnetlink: add IFLA_PARENT_[DEV|DEV_BUS]_NAME

[Loic Poulain]

On Fri, 11 Jun 2021 at 18:00, Parav Pandit <parav@nvidia.com> wrote:
>
>
>
> > From: Loic Poulain <loic.poulain@linaro.org>
> > Sent: Friday, June 11, 2021 9:16 PM
> >
> > Hi Parav,
> >
> > On Fri, 11 Jun 2021 at 15:01, Parav Pandit <parav@nvidia.com> wrote:
> > >
> > >
> > >
> > > > From: Loic Poulain <loic.poulain@linaro.org>
> > > > Sent: Thursday, June 10, 2021 11:15 PM
> > > >
> > > > From: Johannes Berg <johannes.berg@intel.com>
> > > >
> > > > In some cases, for example in the upcoming WWAN framework changes,
> > > > there's no natural "parent netdev", so sometimes dummy netdevs are
> > > > created or similar. IFLA_PARENT_DEV_NAME is a new attribute intended
> > > > to contain a device (sysfs, struct device) name that can be used
> > > > instead when creating a new netdev, if the rtnetlink family implements it.
> > > >
> > > > As suggested by Parav Pandit, we also introduce
> > > > IFLA_PARENT_DEV_BUS_NAME attribute in order to uniquely identify a
> > > > device on the system (with bus/name pair).
> >
> > [...]
> >
> > > > diff --git a/include/uapi/linux/if_link.h
> > > > b/include/uapi/linux/if_link.h index
> > > > a5a7f0e..4882e81 100644
> > > > --- a/include/uapi/linux/if_link.h
> > > > +++ b/include/uapi/linux/if_link.h
> > > > @@ -341,6 +341,13 @@ enum {
> > > >       IFLA_ALT_IFNAME, /* Alternative ifname */
> > > >       IFLA_PERM_ADDRESS,
> > > >       IFLA_PROTO_DOWN_REASON,
> > > > +
> > > > +     /* device (sysfs) name as parent, used instead
> > > > +      * of IFLA_LINK where there's no parent netdev
> > > > +      */
> > > > +     IFLA_PARENT_DEV_NAME,
> > > > +     IFLA_PARENT_DEV_BUS_NAME,
> > > > +
> > > >       __IFLA_MAX
> > > >  };
> > > >
> > > > diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c index
> > > > 92c3e43..32599f3
> > > > 100644
> > > > --- a/net/core/rtnetlink.c
> > > > +++ b/net/core/rtnetlink.c
> > > > @@ -1821,6 +1821,16 @@ static int rtnl_fill_ifinfo(struct sk_buff *skb,
> > > >       if (rtnl_fill_prop_list(skb, dev))
> > > >               goto nla_put_failure;
> > > >
> > > > +     if (dev->dev.parent &&
> > > > +         nla_put_string(skb, IFLA_PARENT_DEV_NAME,
> > > > +                        dev_name(dev->dev.parent)))
> > > > +             goto nla_put_failure;
> > > > +
> > > > +     if (dev->dev.parent && dev->dev.parent->bus &&
> > > > +         nla_put_string(skb, IFLA_PARENT_DEV_BUS_NAME,
> > > > +                        dev->dev.parent->bus->name))
> > > > +             goto nla_put_failure;
> > > > +
> > > >       nlmsg_end(skb, nlh);
> > > >       return 0;
> > > >
> > > > @@ -1880,6 +1890,8 @@ static const struct nla_policy
> > > > ifla_policy[IFLA_MAX+1] = {
> > > >       [IFLA_PERM_ADDRESS]     = { .type = NLA_REJECT },
> > > >       [IFLA_PROTO_DOWN_REASON] = { .type = NLA_NESTED },
> > > >       [IFLA_NEW_IFINDEX]      = NLA_POLICY_MIN(NLA_S32, 1),
> > > > +     [IFLA_PARENT_DEV_NAME]  = { .type = NLA_NUL_STRING },
> > > > +     [IFLA_PARENT_DEV_BUS_NAME] = { .type = NLA_NUL_STRING },
> > > >  };
> > > >
> > > This hunk should go in the patch that enables users to use these fields to
> > specify it for new link creation.
> >
> > Don't get it, the previous changes I see in the tree change both if_link.h and
> > rtnetlink.c for new atributes (e.g. f74877a5457d). Can you elaborate on what
> > you expect here?
> >
> Commit f74877a5457d did in same patch because one of the objective of commit f74877a5457d is to also block PERM_ADDR using policy NLA_REJECT.
> This patch-2 is not enabling user to pass these params via new_link command.
> It is done in a later patch. So ifla_policy doesn’t need to have these fields in this patch.

Understood, thanks.
Loic