1 | // SPDX-License-Identifier: GPL-2.0-only |
2 | #include <linux/etherdevice.h> |
3 | #include "ipvlan.h" |
4 | #include <linux/if_vlan.h> |
5 | #include <linux/if_tap.h> |
6 | #include <linux/interrupt.h> |
7 | #include <linux/nsproxy.h> |
8 | #include <linux/compat.h> |
9 | #include <linux/if_tun.h> |
10 | #include <linux/module.h> |
11 | #include <linux/skbuff.h> |
12 | #include <linux/cache.h> |
13 | #include <linux/sched.h> |
14 | #include <linux/types.h> |
15 | #include <linux/slab.h> |
16 | #include <linux/wait.h> |
17 | #include <linux/cdev.h> |
18 | #include <linux/idr.h> |
19 | #include <linux/fs.h> |
20 | #include <linux/uio.h> |
21 | |
22 | #include <net/net_namespace.h> |
23 | #include <net/rtnetlink.h> |
24 | #include <net/sock.h> |
25 | #include <linux/virtio_net.h> |
26 | |
27 | #define TUN_OFFLOADS (NETIF_F_HW_CSUM | NETIF_F_TSO_ECN | NETIF_F_TSO | \ |
28 | NETIF_F_TSO6) |
29 | |
30 | static dev_t ipvtap_major; |
31 | static struct cdev ipvtap_cdev; |
32 | |
33 | static const void *ipvtap_net_namespace(const struct device *d) |
34 | { |
35 | const struct net_device *dev = to_net_dev(d->parent); |
36 | return dev_net(dev); |
37 | } |
38 | |
39 | static struct class ipvtap_class = { |
40 | .name = "ipvtap" , |
41 | .ns_type = &net_ns_type_operations, |
42 | .namespace = ipvtap_net_namespace, |
43 | }; |
44 | |
45 | struct ipvtap_dev { |
46 | struct ipvl_dev vlan; |
47 | struct tap_dev tap; |
48 | }; |
49 | |
50 | static void ipvtap_count_tx_dropped(struct tap_dev *tap) |
51 | { |
52 | struct ipvtap_dev *vlantap = container_of(tap, struct ipvtap_dev, tap); |
53 | struct ipvl_dev *vlan = &vlantap->vlan; |
54 | |
55 | this_cpu_inc(vlan->pcpu_stats->tx_drps); |
56 | } |
57 | |
58 | static void ipvtap_count_rx_dropped(struct tap_dev *tap) |
59 | { |
60 | struct ipvtap_dev *vlantap = container_of(tap, struct ipvtap_dev, tap); |
61 | struct ipvl_dev *vlan = &vlantap->vlan; |
62 | |
63 | ipvlan_count_rx(ipvlan: vlan, len: 0, success: 0, mcast: 0); |
64 | } |
65 | |
66 | static void ipvtap_update_features(struct tap_dev *tap, |
67 | netdev_features_t features) |
68 | { |
69 | struct ipvtap_dev *vlantap = container_of(tap, struct ipvtap_dev, tap); |
70 | struct ipvl_dev *vlan = &vlantap->vlan; |
71 | |
72 | vlan->sfeatures = features; |
73 | netdev_update_features(dev: vlan->dev); |
74 | } |
75 | |
76 | static int ipvtap_newlink(struct net *src_net, struct net_device *dev, |
77 | struct nlattr *tb[], struct nlattr *data[], |
78 | struct netlink_ext_ack *extack) |
79 | { |
80 | struct ipvtap_dev *vlantap = netdev_priv(dev); |
81 | int err; |
82 | |
83 | INIT_LIST_HEAD(list: &vlantap->tap.queue_list); |
84 | |
85 | /* Since macvlan supports all offloads by default, make |
86 | * tap support all offloads also. |
87 | */ |
88 | vlantap->tap.tap_features = TUN_OFFLOADS; |
89 | vlantap->tap.count_tx_dropped = ipvtap_count_tx_dropped; |
90 | vlantap->tap.update_features = ipvtap_update_features; |
91 | vlantap->tap.count_rx_dropped = ipvtap_count_rx_dropped; |
92 | |
93 | err = netdev_rx_handler_register(dev, rx_handler: tap_handle_frame, rx_handler_data: &vlantap->tap); |
94 | if (err) |
95 | return err; |
96 | |
97 | /* Don't put anything that may fail after macvlan_common_newlink |
98 | * because we can't undo what it does. |
99 | */ |
100 | err = ipvlan_link_new(src_net, dev, tb, data, extack); |
101 | if (err) { |
102 | netdev_rx_handler_unregister(dev); |
103 | return err; |
104 | } |
105 | |
106 | vlantap->tap.dev = vlantap->vlan.dev; |
107 | |
108 | return err; |
109 | } |
110 | |
111 | static void ipvtap_dellink(struct net_device *dev, |
112 | struct list_head *head) |
113 | { |
114 | struct ipvtap_dev *vlan = netdev_priv(dev); |
115 | |
116 | netdev_rx_handler_unregister(dev); |
117 | tap_del_queues(tap: &vlan->tap); |
118 | ipvlan_link_delete(dev, head); |
119 | } |
120 | |
121 | static void ipvtap_setup(struct net_device *dev) |
122 | { |
123 | ipvlan_link_setup(dev); |
124 | dev->tx_queue_len = TUN_READQ_SIZE; |
125 | dev->priv_flags &= ~IFF_NO_QUEUE; |
126 | } |
127 | |
128 | static struct rtnl_link_ops ipvtap_link_ops __read_mostly = { |
129 | .kind = "ipvtap" , |
130 | .setup = ipvtap_setup, |
131 | .newlink = ipvtap_newlink, |
132 | .dellink = ipvtap_dellink, |
133 | .priv_size = sizeof(struct ipvtap_dev), |
134 | }; |
135 | |
136 | static int ipvtap_device_event(struct notifier_block *unused, |
137 | unsigned long event, void *ptr) |
138 | { |
139 | struct net_device *dev = netdev_notifier_info_to_dev(info: ptr); |
140 | struct ipvtap_dev *vlantap; |
141 | struct device *classdev; |
142 | dev_t devt; |
143 | int err; |
144 | char tap_name[IFNAMSIZ]; |
145 | |
146 | if (dev->rtnl_link_ops != &ipvtap_link_ops) |
147 | return NOTIFY_DONE; |
148 | |
149 | snprintf(buf: tap_name, IFNAMSIZ, fmt: "tap%d" , dev->ifindex); |
150 | vlantap = netdev_priv(dev); |
151 | |
152 | switch (event) { |
153 | case NETDEV_REGISTER: |
154 | /* Create the device node here after the network device has |
155 | * been registered but before register_netdevice has |
156 | * finished running. |
157 | */ |
158 | err = tap_get_minor(major: ipvtap_major, tap: &vlantap->tap); |
159 | if (err) |
160 | return notifier_from_errno(err); |
161 | |
162 | devt = MKDEV(MAJOR(ipvtap_major), vlantap->tap.minor); |
163 | classdev = device_create(cls: &ipvtap_class, parent: &dev->dev, devt, |
164 | drvdata: dev, fmt: "%s" , tap_name); |
165 | if (IS_ERR(ptr: classdev)) { |
166 | tap_free_minor(major: ipvtap_major, tap: &vlantap->tap); |
167 | return notifier_from_errno(err: PTR_ERR(ptr: classdev)); |
168 | } |
169 | err = sysfs_create_link(kobj: &dev->dev.kobj, target: &classdev->kobj, |
170 | name: tap_name); |
171 | if (err) |
172 | return notifier_from_errno(err); |
173 | break; |
174 | case NETDEV_UNREGISTER: |
175 | /* vlan->minor == 0 if NETDEV_REGISTER above failed */ |
176 | if (vlantap->tap.minor == 0) |
177 | break; |
178 | sysfs_remove_link(kobj: &dev->dev.kobj, name: tap_name); |
179 | devt = MKDEV(MAJOR(ipvtap_major), vlantap->tap.minor); |
180 | device_destroy(cls: &ipvtap_class, devt); |
181 | tap_free_minor(major: ipvtap_major, tap: &vlantap->tap); |
182 | break; |
183 | case NETDEV_CHANGE_TX_QUEUE_LEN: |
184 | if (tap_queue_resize(tap: &vlantap->tap)) |
185 | return NOTIFY_BAD; |
186 | break; |
187 | } |
188 | |
189 | return NOTIFY_DONE; |
190 | } |
191 | |
192 | static struct notifier_block ipvtap_notifier_block __read_mostly = { |
193 | .notifier_call = ipvtap_device_event, |
194 | }; |
195 | |
196 | static int __init ipvtap_init(void) |
197 | { |
198 | int err; |
199 | |
200 | err = tap_create_cdev(tap_cdev: &ipvtap_cdev, tap_major: &ipvtap_major, device_name: "ipvtap" , |
201 | THIS_MODULE); |
202 | if (err) |
203 | goto out1; |
204 | |
205 | err = class_register(class: &ipvtap_class); |
206 | if (err) |
207 | goto out2; |
208 | |
209 | err = register_netdevice_notifier(nb: &ipvtap_notifier_block); |
210 | if (err) |
211 | goto out3; |
212 | |
213 | err = ipvlan_link_register(ops: &ipvtap_link_ops); |
214 | if (err) |
215 | goto out4; |
216 | |
217 | return 0; |
218 | |
219 | out4: |
220 | unregister_netdevice_notifier(nb: &ipvtap_notifier_block); |
221 | out3: |
222 | class_unregister(class: &ipvtap_class); |
223 | out2: |
224 | tap_destroy_cdev(major: ipvtap_major, tap_cdev: &ipvtap_cdev); |
225 | out1: |
226 | return err; |
227 | } |
228 | module_init(ipvtap_init); |
229 | |
230 | static void __exit ipvtap_exit(void) |
231 | { |
232 | rtnl_link_unregister(ops: &ipvtap_link_ops); |
233 | unregister_netdevice_notifier(nb: &ipvtap_notifier_block); |
234 | class_unregister(class: &ipvtap_class); |
235 | tap_destroy_cdev(major: ipvtap_major, tap_cdev: &ipvtap_cdev); |
236 | } |
237 | module_exit(ipvtap_exit); |
238 | MODULE_ALIAS_RTNL_LINK("ipvtap" ); |
239 | MODULE_AUTHOR("Sainath Grandhi <sainath.grandhi@intel.com>" ); |
240 | MODULE_LICENSE("GPL" ); |
241 | |