net/switchdev/switchdev.c

   1 /*
   2  * net/switchdev/switchdev.c - Switch device API
   3  * Copyright (c) 2014 Jiri Pirko <jiri@resnulli.us>
   4  * Copyright (c) 2014-2015 Scott Feldman <sfeldma@gmail.com>
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License as published by
   8  * the Free Software Foundation; either version 2 of the License, or
   9  * (at your option) any later version.
  10  */
  11
  12 #include <linux/kernel.h>
  13 #include <linux/types.h>
  14 #include <linux/init.h>
  15 #include <linux/mutex.h>
  16 #include <linux/notifier.h>
  17 #include <linux/netdevice.h>
  18 #include <net/ip_fib.h>
  19 #include <net/switchdev.h>
  20
  21 /**
  22  *      netdev_switch_parent_id_get - Get ID of a switch
  23  *      @dev: port device
  24  *      @psid: switch ID
  25  *
  26  *      Get ID of a switch this port is part of.
  27  */
  28 int netdev_switch_parent_id_get(struct net_device *dev,
  29                                 struct netdev_phys_item_id *psid)
  30 {
  31         const struct swdev_ops *ops = dev->swdev_ops;
  32
  33         if (!ops || !ops->swdev_parent_id_get)
  34                 return -EOPNOTSUPP;
  35         return ops->swdev_parent_id_get(dev, psid);
  36 }
  37 EXPORT_SYMBOL_GPL(netdev_switch_parent_id_get);
  38
  39 /**
  40  *      netdev_switch_port_stp_update - Notify switch device port of STP
  41  *                                      state change
  42  *      @dev: port device
  43  *      @state: port STP state
  44  *
  45  *      Notify switch device port of bridge port STP state change.
  46  */
  47 int netdev_switch_port_stp_update(struct net_device *dev, u8 state)
  48 {
  49         const struct swdev_ops *ops = dev->swdev_ops;
  50         struct net_device *lower_dev;
  51         struct list_head *iter;
  52         int err = -EOPNOTSUPP;
  53
  54         if (ops && ops->swdev_port_stp_update)
  55                 return ops->swdev_port_stp_update(dev, state);
  56
  57         netdev_for_each_lower_dev(dev, lower_dev, iter) {
  58                 err = netdev_switch_port_stp_update(lower_dev, state);
  59                 if (err && err != -EOPNOTSUPP)
  60                         return err;
  61         }
  62
  63         return err;
  64 }
  65 EXPORT_SYMBOL_GPL(netdev_switch_port_stp_update);
  66
  67 static DEFINE_MUTEX(netdev_switch_mutex);
  68 static RAW_NOTIFIER_HEAD(netdev_switch_notif_chain);
  69
  70 /**
  71  *      register_netdev_switch_notifier - Register notifier
  72  *      @nb: notifier_block
  73  *
  74  *      Register switch device notifier. This should be used by code
  75  *      which needs to monitor events happening in particular device.
  76  *      Return values are same as for atomic_notifier_chain_register().
  77  */
  78 int register_netdev_switch_notifier(struct notifier_block *nb)
  79 {
  80         int err;
  81
  82         mutex_lock(&netdev_switch_mutex);
  83         err = raw_notifier_chain_register(&netdev_switch_notif_chain, nb);
  84         mutex_unlock(&netdev_switch_mutex);
  85         return err;
  86 }
  87 EXPORT_SYMBOL_GPL(register_netdev_switch_notifier);
  88
  89 /**
  90  *      unregister_netdev_switch_notifier - Unregister notifier
  91  *      @nb: notifier_block
  92  *
  93  *      Unregister switch device notifier.
  94  *      Return values are same as for atomic_notifier_chain_unregister().
  95  */
  96 int unregister_netdev_switch_notifier(struct notifier_block *nb)
  97 {
  98         int err;
  99
 100         mutex_lock(&netdev_switch_mutex);
 101         err = raw_notifier_chain_unregister(&netdev_switch_notif_chain, nb);
 102         mutex_unlock(&netdev_switch_mutex);
 103         return err;
 104 }
 105 EXPORT_SYMBOL_GPL(unregister_netdev_switch_notifier);
 106
 107 /**
 108  *      call_netdev_switch_notifiers - Call notifiers
 109  *      @val: value passed unmodified to notifier function
 110  *      @dev: port device
 111  *      @info: notifier information data
 112  *
 113  *      Call all network notifier blocks. This should be called by driver
 114  *      when it needs to propagate hardware event.
 115  *      Return values are same as for atomic_notifier_call_chain().
 116  */
 117 int call_netdev_switch_notifiers(unsigned long val, struct net_device *dev,
 118                                  struct netdev_switch_notifier_info *info)
 119 {
 120         int err;
 121
 122         info->dev = dev;
 123         mutex_lock(&netdev_switch_mutex);
 124         err = raw_notifier_call_chain(&netdev_switch_notif_chain, val, info);
 125         mutex_unlock(&netdev_switch_mutex);
 126         return err;
 127 }
 128 EXPORT_SYMBOL_GPL(call_netdev_switch_notifiers);
 129
 130 /**
 131  *      netdev_switch_port_bridge_setlink - Notify switch device port of bridge
 132  *      port attributes
 133  *
 134  *      @dev: port device
 135  *      @nlh: netlink msg with bridge port attributes
 136  *      @flags: bridge setlink flags
 137  *
 138  *      Notify switch device port of bridge port attributes
 139  */
 140 int netdev_switch_port_bridge_setlink(struct net_device *dev,
 141                                       struct nlmsghdr *nlh, u16 flags)
 142 {
 143         const struct net_device_ops *ops = dev->netdev_ops;
 144
 145         if (!(dev->features & NETIF_F_HW_SWITCH_OFFLOAD))
 146                 return 0;
 147
 148         if (!ops->ndo_bridge_setlink)
 149                 return -EOPNOTSUPP;
 150
 151         return ops->ndo_bridge_setlink(dev, nlh, flags);
 152 }
 153 EXPORT_SYMBOL_GPL(netdev_switch_port_bridge_setlink);
 154
 155 /**
 156  *      netdev_switch_port_bridge_dellink - Notify switch device port of bridge
 157  *      port attribute delete
 158  *
 159  *      @dev: port device
 160  *      @nlh: netlink msg with bridge port attributes
 161  *      @flags: bridge setlink flags
 162  *
 163  *      Notify switch device port of bridge port attribute delete
 164  */
 165 int netdev_switch_port_bridge_dellink(struct net_device *dev,
 166                                       struct nlmsghdr *nlh, u16 flags)
 167 {
 168         const struct net_device_ops *ops = dev->netdev_ops;
 169
 170         if (!(dev->features & NETIF_F_HW_SWITCH_OFFLOAD))
 171                 return 0;
 172
 173         if (!ops->ndo_bridge_dellink)
 174                 return -EOPNOTSUPP;
 175
 176         return ops->ndo_bridge_dellink(dev, nlh, flags);
 177 }
 178 EXPORT_SYMBOL_GPL(netdev_switch_port_bridge_dellink);
 179
 180 /**
 181  *      ndo_dflt_netdev_switch_port_bridge_setlink - default ndo bridge setlink
 182  *                                                   op for master devices
 183  *
 184  *      @dev: port device
 185  *      @nlh: netlink msg with bridge port attributes
 186  *      @flags: bridge setlink flags
 187  *
 188  *      Notify master device slaves of bridge port attributes
 189  */
 190 int ndo_dflt_netdev_switch_port_bridge_setlink(struct net_device *dev,
 191                                                struct nlmsghdr *nlh, u16 flags)
 192 {
 193         struct net_device *lower_dev;
 194         struct list_head *iter;
 195         int ret = 0, err = 0;
 196
 197         if (!(dev->features & NETIF_F_HW_SWITCH_OFFLOAD))
 198                 return ret;
 199
 200         netdev_for_each_lower_dev(dev, lower_dev, iter) {
 201                 err = netdev_switch_port_bridge_setlink(lower_dev, nlh, flags);
 202                 if (err && err != -EOPNOTSUPP)
 203                         ret = err;
 204         }
 205
 206         return ret;
 207 }
 208 EXPORT_SYMBOL_GPL(ndo_dflt_netdev_switch_port_bridge_setlink);
 209
 210 /**
 211  *      ndo_dflt_netdev_switch_port_bridge_dellink - default ndo bridge dellink
 212  *                                                   op for master devices
 213  *
 214  *      @dev: port device
 215  *      @nlh: netlink msg with bridge port attributes
 216  *      @flags: bridge dellink flags
 217  *
 218  *      Notify master device slaves of bridge port attribute deletes
 219  */
 220 int ndo_dflt_netdev_switch_port_bridge_dellink(struct net_device *dev,
 221                                                struct nlmsghdr *nlh, u16 flags)
 222 {
 223         struct net_device *lower_dev;
 224         struct list_head *iter;
 225         int ret = 0, err = 0;
 226
 227         if (!(dev->features & NETIF_F_HW_SWITCH_OFFLOAD))
 228                 return ret;
 229
 230         netdev_for_each_lower_dev(dev, lower_dev, iter) {
 231                 err = netdev_switch_port_bridge_dellink(lower_dev, nlh, flags);
 232                 if (err && err != -EOPNOTSUPP)
 233                         ret = err;
 234         }
 235
 236         return ret;
 237 }
 238 EXPORT_SYMBOL_GPL(ndo_dflt_netdev_switch_port_bridge_dellink);
 239
 240 static struct net_device *netdev_switch_get_lowest_dev(struct net_device *dev)
 241 {
 242         const struct swdev_ops *ops = dev->swdev_ops;
 243         struct net_device *lower_dev;
 244         struct net_device *port_dev;
 245         struct list_head *iter;
 246
 247         /* Recusively search down until we find a sw port dev.
 248          * (A sw port dev supports swdev_parent_id_get).
 249          */
 250
 251         if (dev->features & NETIF_F_HW_SWITCH_OFFLOAD &&
 252             ops && ops->swdev_parent_id_get)
 253                 return dev;
 254
 255         netdev_for_each_lower_dev(dev, lower_dev, iter) {
 256                 port_dev = netdev_switch_get_lowest_dev(lower_dev);
 257                 if (port_dev)
 258                         return port_dev;
 259         }
 260
 261         return NULL;
 262 }
 263
 264 static struct net_device *netdev_switch_get_dev_by_nhs(struct fib_info *fi)
 265 {
 266         struct netdev_phys_item_id psid;
 267         struct netdev_phys_item_id prev_psid;
 268         struct net_device *dev = NULL;
 269         int nhsel;
 270
 271         /* For this route, all nexthop devs must be on the same switch. */
 272
 273         for (nhsel = 0; nhsel < fi->fib_nhs; nhsel++) {
 274                 const struct fib_nh *nh = &fi->fib_nh[nhsel];
 275
 276                 if (!nh->nh_dev)
 277                         return NULL;
 278
 279                 dev = netdev_switch_get_lowest_dev(nh->nh_dev);
 280                 if (!dev)
 281                         return NULL;
 282
 283                 if (netdev_switch_parent_id_get(dev, &psid))
 284                         return NULL;
 285
 286                 if (nhsel > 0) {
 287                         if (prev_psid.id_len != psid.id_len)
 288                                 return NULL;
 289                         if (memcmp(prev_psid.id, psid.id, psid.id_len))
 290                                 return NULL;
 291                 }
 292
 293                 prev_psid = psid;
 294         }
 295
 296         return dev;
 297 }
 298
 299 /**
 300  *      netdev_switch_fib_ipv4_add - Add IPv4 route entry to switch
 301  *
 302  *      @dst: route's IPv4 destination address
 303  *      @dst_len: destination address length (prefix length)
 304  *      @fi: route FIB info structure
 305  *      @tos: route TOS
 306  *      @type: route type
 307  *      @nlflags: netlink flags passed in (NLM_F_*)
 308  *      @tb_id: route table ID
 309  *
 310  *      Add IPv4 route entry to switch device.
 311  */
 312 int netdev_switch_fib_ipv4_add(u32 dst, int dst_len, struct fib_info *fi,
 313                                u8 tos, u8 type, u32 nlflags, u32 tb_id)
 314 {
 315         struct net_device *dev;
 316         const struct swdev_ops *ops;
 317         int err = 0;
 318
 319         /* Don't offload route if using custom ip rules or if
 320          * IPv4 FIB offloading has been disabled completely.
 321          */
 322
 323 #ifdef CONFIG_IP_MULTIPLE_TABLES
 324         if (fi->fib_net->ipv4.fib_has_custom_rules)
 325                 return 0;
 326 #endif
 327
 328         if (fi->fib_net->ipv4.fib_offload_disabled)
 329                 return 0;
 330
 331         dev = netdev_switch_get_dev_by_nhs(fi);
 332         if (!dev)
 333                 return 0;
 334         ops = dev->swdev_ops;
 335
 336         if (ops->swdev_fib_ipv4_add) {
 337                 err = ops->swdev_fib_ipv4_add(dev, htonl(dst), dst_len,
 338                                               fi, tos, type, nlflags,
 339                                               tb_id);
 340                 if (!err)
 341                         fi->fib_flags |= RTNH_F_OFFLOAD;
 342         }
 343
 344         return err;
 345 }
 346 EXPORT_SYMBOL_GPL(netdev_switch_fib_ipv4_add);
 347
 348 /**
 349  *      netdev_switch_fib_ipv4_del - Delete IPv4 route entry from switch
 350  *
 351  *      @dst: route's IPv4 destination address
 352  *      @dst_len: destination address length (prefix length)
 353  *      @fi: route FIB info structure
 354  *      @tos: route TOS
 355  *      @type: route type
 356  *      @tb_id: route table ID
 357  *
 358  *      Delete IPv4 route entry from switch device.
 359  */
 360 int netdev_switch_fib_ipv4_del(u32 dst, int dst_len, struct fib_info *fi,
 361                                u8 tos, u8 type, u32 tb_id)
 362 {
 363         struct net_device *dev;
 364         const struct swdev_ops *ops;
 365         int err = 0;
 366
 367         if (!(fi->fib_flags & RTNH_F_OFFLOAD))
 368                 return 0;
 369
 370         dev = netdev_switch_get_dev_by_nhs(fi);
 371         if (!dev)
 372                 return 0;
 373         ops = dev->swdev_ops;
 374
 375         if (ops->swdev_fib_ipv4_del) {
 376                 err = ops->swdev_fib_ipv4_del(dev, htonl(dst), dst_len,
 377                                               fi, tos, type, tb_id);
 378                 if (!err)
 379                         fi->fib_flags &= ~RTNH_F_OFFLOAD;
 380         }
 381
 382         return err;
 383 }
 384 EXPORT_SYMBOL_GPL(netdev_switch_fib_ipv4_del);
 385
 386 /**
 387  *      netdev_switch_fib_ipv4_abort - Abort an IPv4 FIB operation
 388  *
 389  *      @fi: route FIB info structure
 390  */
 391 void netdev_switch_fib_ipv4_abort(struct fib_info *fi)
 392 {
 393         /* There was a problem installing this route to the offload
 394          * device.  For now, until we come up with more refined
 395          * policy handling, abruptly end IPv4 fib offloading for
 396          * for entire net by flushing offload device(s) of all
 397          * IPv4 routes, and mark IPv4 fib offloading broken from
 398          * this point forward.
 399          */
 400
 401         fib_flush_external(fi->fib_net);
 402         fi->fib_net->ipv4.fib_offload_disabled = true;
 403 }
 404 EXPORT_SYMBOL_GPL(netdev_switch_fib_ipv4_abort);