net/switchdev/switchdev.c

   1 /*
   2  * net/switchdev/switchdev.c - Switch device API
   3  * Copyright (c) 2014 Jiri Pirko <jiri@resnulli.us>
   4  *
   5  * This program is free software; you can redistribute it and/or modify
   6  * it under the terms of the GNU General Public License as published by
   7  * the Free Software Foundation; either version 2 of the License, or
   8  * (at your option) any later version.
   9  */
  10
  11 #include <linux/kernel.h>
  12 #include <linux/types.h>
  13 #include <linux/init.h>
  14 #include <linux/mutex.h>
  15 #include <linux/notifier.h>
  16 #include <linux/netdevice.h>
  17 #include <net/ip_fib.h>
  18 #include <net/switchdev.h>
  19
  20 /**
  21  *      netdev_switch_parent_id_get - Get ID of a switch
  22  *      @dev: port device
  23  *      @psid: switch ID
  24  *
  25  *      Get ID of a switch this port is part of.
  26  */
  27 int netdev_switch_parent_id_get(struct net_device *dev,
  28                                 struct netdev_phys_item_id *psid)
  29 {
  30         const struct net_device_ops *ops = dev->netdev_ops;
  31
  32         if (!ops->ndo_switch_parent_id_get)
  33                 return -EOPNOTSUPP;
  34         return ops->ndo_switch_parent_id_get(dev, psid);
  35 }
  36 EXPORT_SYMBOL(netdev_switch_parent_id_get);
  37
  38 /**
  39  *      netdev_switch_port_stp_update - Notify switch device port of STP
  40  *                                      state change
  41  *      @dev: port device
  42  *      @state: port STP state
  43  *
  44  *      Notify switch device port of bridge port STP state change.
  45  */
  46 int netdev_switch_port_stp_update(struct net_device *dev, u8 state)
  47 {
  48         const struct net_device_ops *ops = dev->netdev_ops;
  49
  50         if (!ops->ndo_switch_port_stp_update)
  51                 return -EOPNOTSUPP;
  52         WARN_ON(!ops->ndo_switch_parent_id_get);
  53         return ops->ndo_switch_port_stp_update(dev, state);
  54 }
  55 EXPORT_SYMBOL(netdev_switch_port_stp_update);
  56
  57 static DEFINE_MUTEX(netdev_switch_mutex);
  58 static RAW_NOTIFIER_HEAD(netdev_switch_notif_chain);
  59
  60 /**
  61  *      register_netdev_switch_notifier - Register nofifier
  62  *      @nb: notifier_block
  63  *
  64  *      Register switch device notifier. This should be used by code
  65  *      which needs to monitor events happening in particular device.
  66  *      Return values are same as for atomic_notifier_chain_register().
  67  */
  68 int register_netdev_switch_notifier(struct notifier_block *nb)
  69 {
  70         int err;
  71
  72         mutex_lock(&netdev_switch_mutex);
  73         err = raw_notifier_chain_register(&netdev_switch_notif_chain, nb);
  74         mutex_unlock(&netdev_switch_mutex);
  75         return err;
  76 }
  77 EXPORT_SYMBOL(register_netdev_switch_notifier);
  78
  79 /**
  80  *      unregister_netdev_switch_notifier - Unregister nofifier
  81  *      @nb: notifier_block
  82  *
  83  *      Unregister switch device notifier.
  84  *      Return values are same as for atomic_notifier_chain_unregister().
  85  */
  86 int unregister_netdev_switch_notifier(struct notifier_block *nb)
  87 {
  88         int err;
  89
  90         mutex_lock(&netdev_switch_mutex);
  91         err = raw_notifier_chain_unregister(&netdev_switch_notif_chain, nb);
  92         mutex_unlock(&netdev_switch_mutex);
  93         return err;
  94 }
  95 EXPORT_SYMBOL(unregister_netdev_switch_notifier);
  96
  97 /**
  98  *      call_netdev_switch_notifiers - Call nofifiers
  99  *      @val: value passed unmodified to notifier function
 100  *      @dev: port device
 101  *      @info: notifier information data
 102  *
 103  *      Call all network notifier blocks. This should be called by driver
 104  *      when it needs to propagate hardware event.
 105  *      Return values are same as for atomic_notifier_call_chain().
 106  */
 107 int call_netdev_switch_notifiers(unsigned long val, struct net_device *dev,
 108                                  struct netdev_switch_notifier_info *info)
 109 {
 110         int err;
 111
 112         info->dev = dev;
 113         mutex_lock(&netdev_switch_mutex);
 114         err = raw_notifier_call_chain(&netdev_switch_notif_chain, val, info);
 115         mutex_unlock(&netdev_switch_mutex);
 116         return err;
 117 }
 118 EXPORT_SYMBOL(call_netdev_switch_notifiers);
 119
 120 /**
 121  *      netdev_switch_port_bridge_setlink - Notify switch device port of bridge
 122  *      port attributes
 123  *
 124  *      @dev: port device
 125  *      @nlh: netlink msg with bridge port attributes
 126  *      @flags: bridge setlink flags
 127  *
 128  *      Notify switch device port of bridge port attributes
 129  */
 130 int netdev_switch_port_bridge_setlink(struct net_device *dev,
 131                                       struct nlmsghdr *nlh, u16 flags)
 132 {
 133         const struct net_device_ops *ops = dev->netdev_ops;
 134
 135         if (!(dev->features & NETIF_F_HW_SWITCH_OFFLOAD))
 136                 return 0;
 137
 138         if (!ops->ndo_bridge_setlink)
 139                 return -EOPNOTSUPP;
 140
 141         return ops->ndo_bridge_setlink(dev, nlh, flags);
 142 }
 143 EXPORT_SYMBOL(netdev_switch_port_bridge_setlink);
 144
 145 /**
 146  *      netdev_switch_port_bridge_dellink - Notify switch device port of bridge
 147  *      port attribute delete
 148  *
 149  *      @dev: port device
 150  *      @nlh: netlink msg with bridge port attributes
 151  *      @flags: bridge setlink flags
 152  *
 153  *      Notify switch device port of bridge port attribute delete
 154  */
 155 int netdev_switch_port_bridge_dellink(struct net_device *dev,
 156                                       struct nlmsghdr *nlh, u16 flags)
 157 {
 158         const struct net_device_ops *ops = dev->netdev_ops;
 159
 160         if (!(dev->features & NETIF_F_HW_SWITCH_OFFLOAD))
 161                 return 0;
 162
 163         if (!ops->ndo_bridge_dellink)
 164                 return -EOPNOTSUPP;
 165
 166         return ops->ndo_bridge_dellink(dev, nlh, flags);
 167 }
 168 EXPORT_SYMBOL(netdev_switch_port_bridge_dellink);
 169
 170 /**
 171  *      ndo_dflt_netdev_switch_port_bridge_setlink - default ndo bridge setlink
 172  *                                                   op for master devices
 173  *
 174  *      @dev: port device
 175  *      @nlh: netlink msg with bridge port attributes
 176  *      @flags: bridge setlink flags
 177  *
 178  *      Notify master device slaves of bridge port attributes
 179  */
 180 int ndo_dflt_netdev_switch_port_bridge_setlink(struct net_device *dev,
 181                                                struct nlmsghdr *nlh, u16 flags)
 182 {
 183         struct net_device *lower_dev;
 184         struct list_head *iter;
 185         int ret = 0, err = 0;
 186
 187         if (!(dev->features & NETIF_F_HW_SWITCH_OFFLOAD))
 188                 return ret;
 189
 190         netdev_for_each_lower_dev(dev, lower_dev, iter) {
 191                 err = netdev_switch_port_bridge_setlink(lower_dev, nlh, flags);
 192                 if (err && err != -EOPNOTSUPP)
 193                         ret = err;
 194         }
 195
 196         return ret;
 197 }
 198 EXPORT_SYMBOL(ndo_dflt_netdev_switch_port_bridge_setlink);
 199
 200 /**
 201  *      ndo_dflt_netdev_switch_port_bridge_dellink - default ndo bridge dellink
 202  *                                                   op for master devices
 203  *
 204  *      @dev: port device
 205  *      @nlh: netlink msg with bridge port attributes
 206  *      @flags: bridge dellink flags
 207  *
 208  *      Notify master device slaves of bridge port attribute deletes
 209  */
 210 int ndo_dflt_netdev_switch_port_bridge_dellink(struct net_device *dev,
 211                                                struct nlmsghdr *nlh, u16 flags)
 212 {
 213         struct net_device *lower_dev;
 214         struct list_head *iter;
 215         int ret = 0, err = 0;
 216
 217         if (!(dev->features & NETIF_F_HW_SWITCH_OFFLOAD))
 218                 return ret;
 219
 220         netdev_for_each_lower_dev(dev, lower_dev, iter) {
 221                 err = netdev_switch_port_bridge_dellink(lower_dev, nlh, flags);
 222                 if (err && err != -EOPNOTSUPP)
 223                         ret = err;
 224         }
 225
 226         return ret;
 227 }
 228 EXPORT_SYMBOL(ndo_dflt_netdev_switch_port_bridge_dellink);
 229
 230 static struct net_device *netdev_switch_get_lowest_dev(struct net_device *dev)
 231 {
 232         const struct net_device_ops *ops = dev->netdev_ops;
 233         struct net_device *lower_dev;
 234         struct net_device *port_dev;
 235         struct list_head *iter;
 236
 237         /* Recusively search down until we find a sw port dev.
 238          * (A sw port dev supports ndo_switch_parent_id_get).
 239          */
 240
 241         if (dev->features & NETIF_F_HW_SWITCH_OFFLOAD &&
 242             ops->ndo_switch_parent_id_get)
 243                 return dev;
 244
 245         netdev_for_each_lower_dev(dev, lower_dev, iter) {
 246                 port_dev = netdev_switch_get_lowest_dev(lower_dev);
 247                 if (port_dev)
 248                         return port_dev;
 249         }
 250
 251         return NULL;
 252 }
 253
 254 static struct net_device *netdev_switch_get_dev_by_nhs(struct fib_info *fi)
 255 {
 256         struct netdev_phys_item_id psid;
 257         struct netdev_phys_item_id prev_psid;
 258         struct net_device *dev = NULL;
 259         int nhsel;
 260
 261         /* For this route, all nexthop devs must be on the same switch. */
 262
 263         for (nhsel = 0; nhsel < fi->fib_nhs; nhsel++) {
 264                 const struct fib_nh *nh = &fi->fib_nh[nhsel];
 265
 266                 if (!nh->nh_dev)
 267                         return NULL;
 268
 269                 dev = netdev_switch_get_lowest_dev(nh->nh_dev);
 270                 if (!dev)
 271                         return NULL;
 272
 273                 if (netdev_switch_parent_id_get(dev, &psid))
 274                         return NULL;
 275
 276                 if (nhsel > 0) {
 277                         if (prev_psid.id_len != psid.id_len)
 278                                 return NULL;
 279                         if (memcmp(prev_psid.id, psid.id, psid.id_len))
 280                                 return NULL;
 281                 }
 282
 283                 prev_psid = psid;
 284         }
 285
 286         return dev;
 287 }
 288
 289 /**
 290  *      netdev_switch_fib_ipv4_add - Add IPv4 route entry to switch
 291  *
 292  *      @dst: route's IPv4 destination address
 293  *      @dst_len: destination address length (prefix length)
 294  *      @fi: route FIB info structure
 295  *      @tos: route TOS
 296  *      @type: route type
 297  *      @tb_id: route table ID
 298  *
 299  *      Add IPv4 route entry to switch device.
 300  */
 301 int netdev_switch_fib_ipv4_add(u32 dst, int dst_len, struct fib_info *fi,
 302                                u8 tos, u8 type, u32 tb_id)
 303 {
 304         struct net_device *dev;
 305         const struct net_device_ops *ops;
 306         int err = 0;
 307
 308         /* Don't offload route if using custom ip rules or if
 309          * IPv4 FIB offloading has been disabled completely.
 310          */
 311
 312 #ifdef CONFIG_IP_MULTIPLE_TABLES
 313         if (fi->fib_net->ipv4.fib_has_custom_rules)
 314                 return 0;
 315 #endif
 316
 317         if (fi->fib_net->ipv4.fib_offload_disabled)
 318                 return 0;
 319
 320         dev = netdev_switch_get_dev_by_nhs(fi);
 321         if (!dev)
 322                 return 0;
 323         ops = dev->netdev_ops;
 324
 325         if (ops->ndo_switch_fib_ipv4_add) {
 326                 err = ops->ndo_switch_fib_ipv4_add(dev, htonl(dst), dst_len,
 327                                                    fi, tos, type, tb_id);
 328                 if (!err)
 329                         fi->fib_flags |= RTNH_F_EXTERNAL;
 330         }
 331
 332         return err;
 333 }
 334 EXPORT_SYMBOL(netdev_switch_fib_ipv4_add);
 335
 336 /**
 337  *      netdev_switch_fib_ipv4_del - Delete IPv4 route entry from switch
 338  *
 339  *      @dst: route's IPv4 destination address
 340  *      @dst_len: destination address length (prefix length)
 341  *      @fi: route FIB info structure
 342  *      @tos: route TOS
 343  *      @type: route type
 344  *      @tb_id: route table ID
 345  *
 346  *      Delete IPv4 route entry from switch device.
 347  */
 348 int netdev_switch_fib_ipv4_del(u32 dst, int dst_len, struct fib_info *fi,
 349                                u8 tos, u8 type, u32 tb_id)
 350 {
 351         struct net_device *dev;
 352         const struct net_device_ops *ops;
 353         int err = 0;
 354
 355         if (!(fi->fib_flags & RTNH_F_EXTERNAL))
 356                 return 0;
 357
 358         dev = netdev_switch_get_dev_by_nhs(fi);
 359         if (!dev)
 360                 return 0;
 361         ops = dev->netdev_ops;
 362
 363         if (ops->ndo_switch_fib_ipv4_del) {
 364                 err = ops->ndo_switch_fib_ipv4_del(dev, htonl(dst), dst_len,
 365                                                    fi, tos, type, tb_id);
 366                 if (!err)
 367                         fi->fib_flags &= ~RTNH_F_EXTERNAL;
 368         }
 369
 370         return err;
 371 }
 372 EXPORT_SYMBOL(netdev_switch_fib_ipv4_del);
 373
 374 /**
 375  *      netdev_switch_fib_ipv4_abort - Abort an IPv4 FIB operation
 376  *
 377  *      @fi: route FIB info structure
 378  */
 379 void netdev_switch_fib_ipv4_abort(struct fib_info *fi)
 380 {
 381         /* There was a problem installing this route to the offload
 382          * device.  For now, until we come up with more refined
 383          * policy handling, abruptly end IPv4 fib offloading for
 384          * for entire net by flushing offload device(s) of all
 385          * IPv4 routes, and mark IPv4 fib offloading broken from
 386          * this point forward.
 387          */
 388
 389         fib_flush_external(fi->fib_net);
 390         fi->fib_net->ipv4.fib_offload_disabled = true;
 391 }
 392 EXPORT_SYMBOL(netdev_switch_fib_ipv4_abort);