Version:  2.0.40 2.2.26 2.4.37 3.9 3.10 3.11 3.12 3.13 3.14 3.15 3.16 3.17 3.18 3.19 4.0 4.1 4.2 4.3 4.4 4.5 4.6

Linux/net/bridge/br_if.c

  1 /*
  2  *      Userspace interface
  3  *      Linux ethernet bridge
  4  *
  5  *      Authors:
  6  *      Lennert Buytenhek               <buytenh@gnu.org>
  7  *
  8  *      This program is free software; you can redistribute it and/or
  9  *      modify it under the terms of the GNU General Public License
 10  *      as published by the Free Software Foundation; either version
 11  *      2 of the License, or (at your option) any later version.
 12  */
 13 
 14 #include <linux/kernel.h>
 15 #include <linux/netdevice.h>
 16 #include <linux/etherdevice.h>
 17 #include <linux/netpoll.h>
 18 #include <linux/ethtool.h>
 19 #include <linux/if_arp.h>
 20 #include <linux/module.h>
 21 #include <linux/init.h>
 22 #include <linux/rtnetlink.h>
 23 #include <linux/if_ether.h>
 24 #include <linux/slab.h>
 25 #include <net/sock.h>
 26 #include <linux/if_vlan.h>
 27 #include <net/switchdev.h>
 28 
 29 #include "br_private.h"
 30 
 31 /*
 32  * Determine initial path cost based on speed.
 33  * using recommendations from 802.1d standard
 34  *
 35  * Since driver might sleep need to not be holding any locks.
 36  */
 37 static int port_cost(struct net_device *dev)
 38 {
 39         struct ethtool_link_ksettings ecmd;
 40 
 41         if (!__ethtool_get_link_ksettings(dev, &ecmd)) {
 42                 switch (ecmd.base.speed) {
 43                 case SPEED_10000:
 44                         return 2;
 45                 case SPEED_1000:
 46                         return 4;
 47                 case SPEED_100:
 48                         return 19;
 49                 case SPEED_10:
 50                         return 100;
 51                 }
 52         }
 53 
 54         /* Old silly heuristics based on name */
 55         if (!strncmp(dev->name, "lec", 3))
 56                 return 7;
 57 
 58         if (!strncmp(dev->name, "plip", 4))
 59                 return 2500;
 60 
 61         return 100;     /* assume old 10Mbps */
 62 }
 63 
 64 
 65 /* Check for port carrier transitions. */
 66 void br_port_carrier_check(struct net_bridge_port *p)
 67 {
 68         struct net_device *dev = p->dev;
 69         struct net_bridge *br = p->br;
 70 
 71         if (!(p->flags & BR_ADMIN_COST) &&
 72             netif_running(dev) && netif_oper_up(dev))
 73                 p->path_cost = port_cost(dev);
 74 
 75         if (!netif_running(br->dev))
 76                 return;
 77 
 78         spin_lock_bh(&br->lock);
 79         if (netif_running(dev) && netif_oper_up(dev)) {
 80                 if (p->state == BR_STATE_DISABLED)
 81                         br_stp_enable_port(p);
 82         } else {
 83                 if (p->state != BR_STATE_DISABLED)
 84                         br_stp_disable_port(p);
 85         }
 86         spin_unlock_bh(&br->lock);
 87 }
 88 
 89 static void br_port_set_promisc(struct net_bridge_port *p)
 90 {
 91         int err = 0;
 92 
 93         if (br_promisc_port(p))
 94                 return;
 95 
 96         err = dev_set_promiscuity(p->dev, 1);
 97         if (err)
 98                 return;
 99 
100         br_fdb_unsync_static(p->br, p);
101         p->flags |= BR_PROMISC;
102 }
103 
104 static void br_port_clear_promisc(struct net_bridge_port *p)
105 {
106         int err;
107 
108         /* Check if the port is already non-promisc or if it doesn't
109          * support UNICAST filtering.  Without unicast filtering support
110          * we'll end up re-enabling promisc mode anyway, so just check for
111          * it here.
112          */
113         if (!br_promisc_port(p) || !(p->dev->priv_flags & IFF_UNICAST_FLT))
114                 return;
115 
116         /* Since we'll be clearing the promisc mode, program the port
117          * first so that we don't have interruption in traffic.
118          */
119         err = br_fdb_sync_static(p->br, p);
120         if (err)
121                 return;
122 
123         dev_set_promiscuity(p->dev, -1);
124         p->flags &= ~BR_PROMISC;
125 }
126 
127 /* When a port is added or removed or when certain port flags
128  * change, this function is called to automatically manage
129  * promiscuity setting of all the bridge ports.  We are always called
130  * under RTNL so can skip using rcu primitives.
131  */
132 void br_manage_promisc(struct net_bridge *br)
133 {
134         struct net_bridge_port *p;
135         bool set_all = false;
136 
137         /* If vlan filtering is disabled or bridge interface is placed
138          * into promiscuous mode, place all ports in promiscuous mode.
139          */
140         if ((br->dev->flags & IFF_PROMISC) || !br_vlan_enabled(br))
141                 set_all = true;
142 
143         list_for_each_entry(p, &br->port_list, list) {
144                 if (set_all) {
145                         br_port_set_promisc(p);
146                 } else {
147                         /* If the number of auto-ports is <= 1, then all other
148                          * ports will have their output configuration
149                          * statically specified through fdbs.  Since ingress
150                          * on the auto-port becomes forwarding/egress to other
151                          * ports and egress configuration is statically known,
152                          * we can say that ingress configuration of the
153                          * auto-port is also statically known.
154                          * This lets us disable promiscuous mode and write
155                          * this config to hw.
156                          */
157                         if (br->auto_cnt == 0 ||
158                             (br->auto_cnt == 1 && br_auto_port(p)))
159                                 br_port_clear_promisc(p);
160                         else
161                                 br_port_set_promisc(p);
162                 }
163         }
164 }
165 
166 static void nbp_update_port_count(struct net_bridge *br)
167 {
168         struct net_bridge_port *p;
169         u32 cnt = 0;
170 
171         list_for_each_entry(p, &br->port_list, list) {
172                 if (br_auto_port(p))
173                         cnt++;
174         }
175         if (br->auto_cnt != cnt) {
176                 br->auto_cnt = cnt;
177                 br_manage_promisc(br);
178         }
179 }
180 
181 static void nbp_delete_promisc(struct net_bridge_port *p)
182 {
183         /* If port is currently promiscuous, unset promiscuity.
184          * Otherwise, it is a static port so remove all addresses
185          * from it.
186          */
187         dev_set_allmulti(p->dev, -1);
188         if (br_promisc_port(p))
189                 dev_set_promiscuity(p->dev, -1);
190         else
191                 br_fdb_unsync_static(p->br, p);
192 }
193 
194 static void release_nbp(struct kobject *kobj)
195 {
196         struct net_bridge_port *p
197                 = container_of(kobj, struct net_bridge_port, kobj);
198         kfree(p);
199 }
200 
201 static struct kobj_type brport_ktype = {
202 #ifdef CONFIG_SYSFS
203         .sysfs_ops = &brport_sysfs_ops,
204 #endif
205         .release = release_nbp,
206 };
207 
208 static void destroy_nbp(struct net_bridge_port *p)
209 {
210         struct net_device *dev = p->dev;
211 
212         p->br = NULL;
213         p->dev = NULL;
214         dev_put(dev);
215 
216         kobject_put(&p->kobj);
217 }
218 
219 static void destroy_nbp_rcu(struct rcu_head *head)
220 {
221         struct net_bridge_port *p =
222                         container_of(head, struct net_bridge_port, rcu);
223         destroy_nbp(p);
224 }
225 
226 static unsigned get_max_headroom(struct net_bridge *br)
227 {
228         unsigned max_headroom = 0;
229         struct net_bridge_port *p;
230 
231         list_for_each_entry(p, &br->port_list, list) {
232                 unsigned dev_headroom = netdev_get_fwd_headroom(p->dev);
233 
234                 if (dev_headroom > max_headroom)
235                         max_headroom = dev_headroom;
236         }
237 
238         return max_headroom;
239 }
240 
241 static void update_headroom(struct net_bridge *br, int new_hr)
242 {
243         struct net_bridge_port *p;
244 
245         list_for_each_entry(p, &br->port_list, list)
246                 netdev_set_rx_headroom(p->dev, new_hr);
247 
248         br->dev->needed_headroom = new_hr;
249 }
250 
251 /* Delete port(interface) from bridge is done in two steps.
252  * via RCU. First step, marks device as down. That deletes
253  * all the timers and stops new packets from flowing through.
254  *
255  * Final cleanup doesn't occur until after all CPU's finished
256  * processing packets.
257  *
258  * Protected from multiple admin operations by RTNL mutex
259  */
260 static void del_nbp(struct net_bridge_port *p)
261 {
262         struct net_bridge *br = p->br;
263         struct net_device *dev = p->dev;
264 
265         sysfs_remove_link(br->ifobj, p->dev->name);
266 
267         nbp_delete_promisc(p);
268 
269         spin_lock_bh(&br->lock);
270         br_stp_disable_port(p);
271         spin_unlock_bh(&br->lock);
272 
273         br_ifinfo_notify(RTM_DELLINK, p);
274 
275         list_del_rcu(&p->list);
276         if (netdev_get_fwd_headroom(dev) == br->dev->needed_headroom)
277                 update_headroom(br, get_max_headroom(br));
278         netdev_reset_rx_headroom(dev);
279 
280         nbp_vlan_flush(p);
281         br_fdb_delete_by_port(br, p, 0, 1);
282         switchdev_deferred_process();
283 
284         nbp_update_port_count(br);
285 
286         netdev_upper_dev_unlink(dev, br->dev);
287 
288         dev->priv_flags &= ~IFF_BRIDGE_PORT;
289 
290         netdev_rx_handler_unregister(dev);
291 
292         br_multicast_del_port(p);
293 
294         kobject_uevent(&p->kobj, KOBJ_REMOVE);
295         kobject_del(&p->kobj);
296 
297         br_netpoll_disable(p);
298 
299         call_rcu(&p->rcu, destroy_nbp_rcu);
300 }
301 
302 /* Delete bridge device */
303 void br_dev_delete(struct net_device *dev, struct list_head *head)
304 {
305         struct net_bridge *br = netdev_priv(dev);
306         struct net_bridge_port *p, *n;
307 
308         list_for_each_entry_safe(p, n, &br->port_list, list) {
309                 del_nbp(p);
310         }
311 
312         br_fdb_delete_by_port(br, NULL, 0, 1);
313 
314         br_vlan_flush(br);
315         br_multicast_dev_del(br);
316         del_timer_sync(&br->gc_timer);
317 
318         br_sysfs_delbr(br->dev);
319         unregister_netdevice_queue(br->dev, head);
320 }
321 
322 /* find an available port number */
323 static int find_portno(struct net_bridge *br)
324 {
325         int index;
326         struct net_bridge_port *p;
327         unsigned long *inuse;
328 
329         inuse = kcalloc(BITS_TO_LONGS(BR_MAX_PORTS), sizeof(unsigned long),
330                         GFP_KERNEL);
331         if (!inuse)
332                 return -ENOMEM;
333 
334         set_bit(0, inuse);      /* zero is reserved */
335         list_for_each_entry(p, &br->port_list, list) {
336                 set_bit(p->port_no, inuse);
337         }
338         index = find_first_zero_bit(inuse, BR_MAX_PORTS);
339         kfree(inuse);
340 
341         return (index >= BR_MAX_PORTS) ? -EXFULL : index;
342 }
343 
344 /* called with RTNL but without bridge lock */
345 static struct net_bridge_port *new_nbp(struct net_bridge *br,
346                                        struct net_device *dev)
347 {
348         int index;
349         struct net_bridge_port *p;
350 
351         index = find_portno(br);
352         if (index < 0)
353                 return ERR_PTR(index);
354 
355         p = kzalloc(sizeof(*p), GFP_KERNEL);
356         if (p == NULL)
357                 return ERR_PTR(-ENOMEM);
358 
359         p->br = br;
360         dev_hold(dev);
361         p->dev = dev;
362         p->path_cost = port_cost(dev);
363         p->priority = 0x8000 >> BR_PORT_BITS;
364         p->port_no = index;
365         p->flags = BR_LEARNING | BR_FLOOD;
366         br_init_port(p);
367         br_set_state(p, BR_STATE_DISABLED);
368         br_stp_port_timer_init(p);
369         br_multicast_add_port(p);
370 
371         return p;
372 }
373 
374 int br_add_bridge(struct net *net, const char *name)
375 {
376         struct net_device *dev;
377         int res;
378 
379         dev = alloc_netdev(sizeof(struct net_bridge), name, NET_NAME_UNKNOWN,
380                            br_dev_setup);
381 
382         if (!dev)
383                 return -ENOMEM;
384 
385         dev_net_set(dev, net);
386         dev->rtnl_link_ops = &br_link_ops;
387 
388         res = register_netdev(dev);
389         if (res)
390                 free_netdev(dev);
391         return res;
392 }
393 
394 int br_del_bridge(struct net *net, const char *name)
395 {
396         struct net_device *dev;
397         int ret = 0;
398 
399         rtnl_lock();
400         dev = __dev_get_by_name(net, name);
401         if (dev == NULL)
402                 ret =  -ENXIO;  /* Could not find device */
403 
404         else if (!(dev->priv_flags & IFF_EBRIDGE)) {
405                 /* Attempt to delete non bridge device! */
406                 ret = -EPERM;
407         }
408 
409         else if (dev->flags & IFF_UP) {
410                 /* Not shutdown yet. */
411                 ret = -EBUSY;
412         }
413 
414         else
415                 br_dev_delete(dev, NULL);
416 
417         rtnl_unlock();
418         return ret;
419 }
420 
421 /* MTU of the bridge pseudo-device: ETH_DATA_LEN or the minimum of the ports */
422 int br_min_mtu(const struct net_bridge *br)
423 {
424         const struct net_bridge_port *p;
425         int mtu = 0;
426 
427         ASSERT_RTNL();
428 
429         if (list_empty(&br->port_list))
430                 mtu = ETH_DATA_LEN;
431         else {
432                 list_for_each_entry(p, &br->port_list, list) {
433                         if (!mtu  || p->dev->mtu < mtu)
434                                 mtu = p->dev->mtu;
435                 }
436         }
437         return mtu;
438 }
439 
440 static void br_set_gso_limits(struct net_bridge *br)
441 {
442         unsigned int gso_max_size = GSO_MAX_SIZE;
443         u16 gso_max_segs = GSO_MAX_SEGS;
444         const struct net_bridge_port *p;
445 
446         list_for_each_entry(p, &br->port_list, list) {
447                 gso_max_size = min(gso_max_size, p->dev->gso_max_size);
448                 gso_max_segs = min(gso_max_segs, p->dev->gso_max_segs);
449         }
450         br->dev->gso_max_size = gso_max_size;
451         br->dev->gso_max_segs = gso_max_segs;
452 }
453 
454 /*
455  * Recomputes features using slave's features
456  */
457 netdev_features_t br_features_recompute(struct net_bridge *br,
458         netdev_features_t features)
459 {
460         struct net_bridge_port *p;
461         netdev_features_t mask;
462 
463         if (list_empty(&br->port_list))
464                 return features;
465 
466         mask = features;
467         features &= ~NETIF_F_ONE_FOR_ALL;
468 
469         list_for_each_entry(p, &br->port_list, list) {
470                 features = netdev_increment_features(features,
471                                                      p->dev->features, mask);
472         }
473         features = netdev_add_tso_features(features, mask);
474 
475         return features;
476 }
477 
478 /* called with RTNL */
479 int br_add_if(struct net_bridge *br, struct net_device *dev)
480 {
481         struct net_bridge_port *p;
482         int err = 0;
483         unsigned br_hr, dev_hr;
484         bool changed_addr;
485 
486         /* Don't allow bridging non-ethernet like devices, or DSA-enabled
487          * master network devices since the bridge layer rx_handler prevents
488          * the DSA fake ethertype handler to be invoked, so we do not strip off
489          * the DSA switch tag protocol header and the bridge layer just return
490          * RX_HANDLER_CONSUMED, stopping RX processing for these frames.
491          */
492         if ((dev->flags & IFF_LOOPBACK) ||
493             dev->type != ARPHRD_ETHER || dev->addr_len != ETH_ALEN ||
494             !is_valid_ether_addr(dev->dev_addr) ||
495             netdev_uses_dsa(dev))
496                 return -EINVAL;
497 
498         /* No bridging of bridges */
499         if (dev->netdev_ops->ndo_start_xmit == br_dev_xmit)
500                 return -ELOOP;
501 
502         /* Device is already being bridged */
503         if (br_port_exists(dev))
504                 return -EBUSY;
505 
506         /* No bridging devices that dislike that (e.g. wireless) */
507         if (dev->priv_flags & IFF_DONT_BRIDGE)
508                 return -EOPNOTSUPP;
509 
510         p = new_nbp(br, dev);
511         if (IS_ERR(p))
512                 return PTR_ERR(p);
513 
514         call_netdevice_notifiers(NETDEV_JOIN, dev);
515 
516         err = dev_set_allmulti(dev, 1);
517         if (err)
518                 goto put_back;
519 
520         err = kobject_init_and_add(&p->kobj, &brport_ktype, &(dev->dev.kobj),
521                                    SYSFS_BRIDGE_PORT_ATTR);
522         if (err)
523                 goto err1;
524 
525         err = br_sysfs_addif(p);
526         if (err)
527                 goto err2;
528 
529         err = br_netpoll_enable(p);
530         if (err)
531                 goto err3;
532 
533         err = netdev_rx_handler_register(dev, br_handle_frame, p);
534         if (err)
535                 goto err4;
536 
537         dev->priv_flags |= IFF_BRIDGE_PORT;
538 
539         err = netdev_master_upper_dev_link(dev, br->dev, NULL, NULL);
540         if (err)
541                 goto err5;
542 
543         dev_disable_lro(dev);
544 
545         list_add_rcu(&p->list, &br->port_list);
546 
547         nbp_update_port_count(br);
548 
549         netdev_update_features(br->dev);
550 
551         br_hr = br->dev->needed_headroom;
552         dev_hr = netdev_get_fwd_headroom(dev);
553         if (br_hr < dev_hr)
554                 update_headroom(br, dev_hr);
555         else
556                 netdev_set_rx_headroom(dev, br_hr);
557 
558         if (br_fdb_insert(br, p, dev->dev_addr, 0))
559                 netdev_err(dev, "failed insert local address bridge forwarding table\n");
560 
561         err = nbp_vlan_init(p);
562         if (err) {
563                 netdev_err(dev, "failed to initialize vlan filtering on this port\n");
564                 goto err6;
565         }
566 
567         spin_lock_bh(&br->lock);
568         changed_addr = br_stp_recalculate_bridge_id(br);
569 
570         if (netif_running(dev) && netif_oper_up(dev) &&
571             (br->dev->flags & IFF_UP))
572                 br_stp_enable_port(p);
573         spin_unlock_bh(&br->lock);
574 
575         br_ifinfo_notify(RTM_NEWLINK, p);
576 
577         if (changed_addr)
578                 call_netdevice_notifiers(NETDEV_CHANGEADDR, br->dev);
579 
580         dev_set_mtu(br->dev, br_min_mtu(br));
581         br_set_gso_limits(br);
582 
583         kobject_uevent(&p->kobj, KOBJ_ADD);
584 
585         return 0;
586 
587 err6:
588         list_del_rcu(&p->list);
589         br_fdb_delete_by_port(br, p, 0, 1);
590         nbp_update_port_count(br);
591         netdev_upper_dev_unlink(dev, br->dev);
592 
593 err5:
594         dev->priv_flags &= ~IFF_BRIDGE_PORT;
595         netdev_rx_handler_unregister(dev);
596 err4:
597         br_netpoll_disable(p);
598 err3:
599         sysfs_remove_link(br->ifobj, p->dev->name);
600 err2:
601         kobject_put(&p->kobj);
602         p = NULL; /* kobject_put frees */
603 err1:
604         dev_set_allmulti(dev, -1);
605 put_back:
606         dev_put(dev);
607         kfree(p);
608         return err;
609 }
610 
611 /* called with RTNL */
612 int br_del_if(struct net_bridge *br, struct net_device *dev)
613 {
614         struct net_bridge_port *p;
615         bool changed_addr;
616 
617         p = br_port_get_rtnl(dev);
618         if (!p || p->br != br)
619                 return -EINVAL;
620 
621         /* Since more than one interface can be attached to a bridge,
622          * there still maybe an alternate path for netconsole to use;
623          * therefore there is no reason for a NETDEV_RELEASE event.
624          */
625         del_nbp(p);
626 
627         dev_set_mtu(br->dev, br_min_mtu(br));
628         br_set_gso_limits(br);
629 
630         spin_lock_bh(&br->lock);
631         changed_addr = br_stp_recalculate_bridge_id(br);
632         spin_unlock_bh(&br->lock);
633 
634         if (changed_addr)
635                 call_netdevice_notifiers(NETDEV_CHANGEADDR, br->dev);
636 
637         netdev_update_features(br->dev);
638 
639         return 0;
640 }
641 
642 void br_port_flags_change(struct net_bridge_port *p, unsigned long mask)
643 {
644         struct net_bridge *br = p->br;
645 
646         if (mask & BR_AUTO_MASK)
647                 nbp_update_port_count(br);
648 }
649 

This page was automatically generated by LXR 0.3.1 (source).  •  Linux is a registered trademark of Linus Torvalds  •  Contact us