datapath/vport.c

   1 /*
   2  * Copyright (c) 2007-2011 Nicira Networks.
   3  *
   4  * This program is free software; you can redistribute it and/or
   5  * modify it under the terms of version 2 of the GNU General Public
   6  * License as published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it will be useful, but
   9  * WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  11  * General Public License for more details.
  12  *
  13  * You should have received a copy of the GNU General Public License
  14  * along with this program; if not, write to the Free Software
  15  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
  16  * 02110-1301, USA
  17  */
  18
  19 #include <linux/dcache.h>
  20 #include <linux/etherdevice.h>
  21 #include <linux/if.h>
  22 #include <linux/if_vlan.h>
  23 #include <linux/kernel.h>
  24 #include <linux/list.h>
  25 #include <linux/mutex.h>
  26 #include <linux/percpu.h>
  27 #include <linux/rcupdate.h>
  28 #include <linux/rtnetlink.h>
  29 #include <linux/compat.h>
  30 #include <linux/version.h>
  31
  32 #include "vport.h"
  33 #include "vport-internal_dev.h"
  34
  35 /* List of statically compiled vport implementations.  Don't forget to also
  36  * add yours to the list at the bottom of vport.h. */
  37 static const struct vport_ops *base_vport_ops_list[] = {
  38         &netdev_vport_ops,
  39         &internal_vport_ops,
  40         &patch_vport_ops,
  41         &gre_vport_ops,
  42 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,26)
  43         &capwap_vport_ops,
  44 #endif
  45 };
  46
  47 static const struct vport_ops **vport_ops_list;
  48 static int n_vport_types;
  49
  50 /* Protected by RCU read lock for reading, RTNL lock for writing. */
  51 static struct hlist_head *dev_table;
  52 #define VPORT_HASH_BUCKETS 1024
  53
  54 /**
  55  *      vport_init - initialize vport subsystem
  56  *
  57  * Called at module load time to initialize the vport subsystem and any
  58  * compiled in vport types.
  59  */
  60 int vport_init(void)
  61 {
  62         int err;
  63         int i;
  64
  65         dev_table = kzalloc(VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
  66                             GFP_KERNEL);
  67         if (!dev_table) {
  68                 err = -ENOMEM;
  69                 goto error;
  70         }
  71
  72         vport_ops_list = kmalloc(ARRAY_SIZE(base_vport_ops_list) *
  73                                  sizeof(struct vport_ops *), GFP_KERNEL);
  74         if (!vport_ops_list) {
  75                 err = -ENOMEM;
  76                 goto error_dev_table;
  77         }
  78
  79         for (i = 0; i < ARRAY_SIZE(base_vport_ops_list); i++) {
  80                 const struct vport_ops *new_ops = base_vport_ops_list[i];
  81
  82                 if (new_ops->init)
  83                         err = new_ops->init();
  84                 else
  85                         err = 0;
  86
  87                 if (!err)
  88                         vport_ops_list[n_vport_types++] = new_ops;
  89                 else if (new_ops->flags & VPORT_F_REQUIRED) {
  90                         vport_exit();
  91                         goto error;
  92                 }
  93         }
  94
  95         return 0;
  96
  97 error_dev_table:
  98         kfree(dev_table);
  99 error:
 100         return err;
 101 }
 102
 103 /**
 104  *      vport_exit - shutdown vport subsystem
 105  *
 106  * Called at module exit time to shutdown the vport subsystem and any
 107  * initialized vport types.
 108  */
 109 void vport_exit(void)
 110 {
 111         int i;
 112
 113         for (i = 0; i < n_vport_types; i++) {
 114                 if (vport_ops_list[i]->exit)
 115                         vport_ops_list[i]->exit();
 116         }
 117
 118         kfree(vport_ops_list);
 119         kfree(dev_table);
 120 }
 121
 122 static struct hlist_head *hash_bucket(const char *name)
 123 {
 124         unsigned int hash = full_name_hash(name, strlen(name));
 125         return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
 126 }
 127
 128 /**
 129  *      vport_locate - find a port that has already been created
 130  *
 131  * @name: name of port to find
 132  *
 133  * Must be called with RTNL or RCU read lock.
 134  */
 135 struct vport *vport_locate(const char *name)
 136 {
 137         struct hlist_head *bucket = hash_bucket(name);
 138         struct vport *vport;
 139         struct hlist_node *node;
 140
 141         hlist_for_each_entry_rcu(vport, node, bucket, hash_node)
 142                 if (!strcmp(name, vport->ops->get_name(vport)))
 143                         return vport;
 144
 145         return NULL;
 146 }
 147
 148 static void release_vport(struct kobject *kobj)
 149 {
 150         struct vport *p = container_of(kobj, struct vport, kobj);
 151         kfree(p);
 152 }
 153
 154 static struct kobj_type brport_ktype = {
 155 #ifdef CONFIG_SYSFS
 156         .sysfs_ops = &brport_sysfs_ops,
 157 #endif
 158         .release = release_vport
 159 };
 160
 161 /**
 162  *      vport_alloc - allocate and initialize new vport
 163  *
 164  * @priv_size: Size of private data area to allocate.
 165  * @ops: vport device ops
 166  *
 167  * Allocate and initialize a new vport defined by @ops.  The vport will contain
 168  * a private data area of size @priv_size that can be accessed using
 169  * vport_priv().  vports that are no longer needed should be released with
 170  * vport_free().
 171  */
 172 struct vport *vport_alloc(int priv_size, const struct vport_ops *ops,
 173                           const struct vport_parms *parms)
 174 {
 175         struct vport *vport;
 176         size_t alloc_size;
 177
 178         alloc_size = sizeof(struct vport);
 179         if (priv_size) {
 180                 alloc_size = ALIGN(alloc_size, VPORT_ALIGN);
 181                 alloc_size += priv_size;
 182         }
 183
 184         vport = kzalloc(alloc_size, GFP_KERNEL);
 185         if (!vport)
 186                 return ERR_PTR(-ENOMEM);
 187
 188         vport->dp = parms->dp;
 189         vport->port_no = parms->port_no;
 190         vport->upcall_pid = parms->upcall_pid;
 191         vport->ops = ops;
 192
 193         /* Initialize kobject for bridge.  This will be added as
 194          * /sys/class/net/<devname>/brport later, if sysfs is enabled. */
 195         vport->kobj.kset = NULL;
 196         kobject_init(&vport->kobj, &brport_ktype);
 197
 198         vport->percpu_stats = alloc_percpu(struct vport_percpu_stats);
 199         if (!vport->percpu_stats)
 200                 return ERR_PTR(-ENOMEM);
 201
 202         spin_lock_init(&vport->stats_lock);
 203
 204         return vport;
 205 }
 206
 207 /**
 208  *      vport_free - uninitialize and free vport
 209  *
 210  * @vport: vport to free
 211  *
 212  * Frees a vport allocated with vport_alloc() when it is no longer needed.
 213  *
 214  * The caller must ensure that an RCU grace period has passed since the last
 215  * time @vport was in a datapath.
 216  */
 217 void vport_free(struct vport *vport)
 218 {
 219         free_percpu(vport->percpu_stats);
 220
 221         kobject_put(&vport->kobj);
 222 }
 223
 224 /**
 225  *      vport_add - add vport device (for kernel callers)
 226  *
 227  * @parms: Information about new vport.
 228  *
 229  * Creates a new vport with the specified configuration (which is dependent on
 230  * device type).  RTNL lock must be held.
 231  */
 232 struct vport *vport_add(const struct vport_parms *parms)
 233 {
 234         struct vport *vport;
 235         int err = 0;
 236         int i;
 237
 238         ASSERT_RTNL();
 239
 240         for (i = 0; i < n_vport_types; i++) {
 241                 if (vport_ops_list[i]->type == parms->type) {
 242                         vport = vport_ops_list[i]->create(parms);
 243                         if (IS_ERR(vport)) {
 244                                 err = PTR_ERR(vport);
 245                                 goto out;
 246                         }
 247
 248                         hlist_add_head_rcu(&vport->hash_node,
 249                                            hash_bucket(vport->ops->get_name(vport)));
 250                         return vport;
 251                 }
 252         }
 253
 254         err = -EAFNOSUPPORT;
 255
 256 out:
 257         return ERR_PTR(err);
 258 }
 259
 260 /**
 261  *      vport_set_options - modify existing vport device (for kernel callers)
 262  *
 263  * @vport: vport to modify.
 264  * @port: New configuration.
 265  *
 266  * Modifies an existing device with the specified configuration (which is
 267  * dependent on device type).  RTNL lock must be held.
 268  */
 269 int vport_set_options(struct vport *vport, struct nlattr *options)
 270 {
 271         ASSERT_RTNL();
 272
 273         if (!vport->ops->set_options)
 274                 return -EOPNOTSUPP;
 275         return vport->ops->set_options(vport, options);
 276 }
 277
 278 /**
 279  *      vport_del - delete existing vport device
 280  *
 281  * @vport: vport to delete.
 282  *
 283  * Detaches @vport from its datapath and destroys it.  It is possible to fail
 284  * for reasons such as lack of memory.  RTNL lock must be held.
 285  */
 286 void vport_del(struct vport *vport)
 287 {
 288         ASSERT_RTNL();
 289
 290         hlist_del_rcu(&vport->hash_node);
 291
 292         vport->ops->destroy(vport);
 293 }
 294
 295 /**
 296  *      vport_set_addr - set device Ethernet address (for kernel callers)
 297  *
 298  * @vport: vport on which to set Ethernet address.
 299  * @addr: New address.
 300  *
 301  * Sets the Ethernet address of the given device.  Some devices may not support
 302  * setting the Ethernet address, in which case the result will always be
 303  * -EOPNOTSUPP.  RTNL lock must be held.
 304  */
 305 int vport_set_addr(struct vport *vport, const unsigned char *addr)
 306 {
 307         ASSERT_RTNL();
 308
 309         if (!is_valid_ether_addr(addr))
 310                 return -EADDRNOTAVAIL;
 311
 312         if (vport->ops->set_addr)
 313                 return vport->ops->set_addr(vport, addr);
 314         else
 315                 return -EOPNOTSUPP;
 316 }
 317
 318 /**
 319  *      vport_set_stats - sets offset device stats
 320  *
 321  * @vport: vport on which to set stats
 322  * @stats: stats to set
 323  *
 324  * Provides a set of transmit, receive, and error stats to be added as an
 325  * offset to the collect data when stats are retreived.  Some devices may not
 326  * support setting the stats, in which case the result will always be
 327  * -EOPNOTSUPP.
 328  *
 329  * Must be called with RTNL lock.
 330  */
 331 void vport_set_stats(struct vport *vport, struct ovs_vport_stats *stats)
 332 {
 333         ASSERT_RTNL();
 334
 335         spin_lock_bh(&vport->stats_lock);
 336         vport->offset_stats = *stats;
 337         spin_unlock_bh(&vport->stats_lock);
 338 }
 339
 340 /**
 341  *      vport_get_stats - retrieve device stats
 342  *
 343  * @vport: vport from which to retrieve the stats
 344  * @stats: location to store stats
 345  *
 346  * Retrieves transmit, receive, and error stats for the given device.
 347  *
 348  * Must be called with RTNL lock or rcu_read_lock.
 349  */
 350 void vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
 351 {
 352         int i;
 353
 354         /* We potentially have 3 sources of stats that need to be
 355          * combined: those we have collected (split into err_stats and
 356          * percpu_stats), offset_stats from set_stats(), and device
 357          * error stats from netdev->get_stats() (for errors that happen
 358          * downstream and therefore aren't reported through our
 359          * vport_record_error() function).
 360          * Stats from first two sources are merged and reported by ovs over
 361          * OVS_VPORT_ATTR_STATS.
 362          * netdev-stats can be directly read over netlink-ioctl.
 363          */
 364
 365         spin_lock_bh(&vport->stats_lock);
 366
 367         *stats = vport->offset_stats;
 368
 369         stats->rx_errors        += vport->err_stats.rx_errors;
 370         stats->tx_errors        += vport->err_stats.tx_errors;
 371         stats->tx_dropped       += vport->err_stats.tx_dropped;
 372         stats->rx_dropped       += vport->err_stats.rx_dropped;
 373
 374         spin_unlock_bh(&vport->stats_lock);
 375
 376         for_each_possible_cpu(i) {
 377                 const struct vport_percpu_stats *percpu_stats;
 378                 struct vport_percpu_stats local_stats;
 379                 unsigned int start;
 380
 381                 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
 382
 383                 do {
 384                         start = u64_stats_fetch_begin_bh(&percpu_stats->sync);
 385                         local_stats = *percpu_stats;
 386                 } while (u64_stats_fetch_retry_bh(&percpu_stats->sync, start));
 387
 388                 stats->rx_bytes         += local_stats.rx_bytes;
 389                 stats->rx_packets       += local_stats.rx_packets;
 390                 stats->tx_bytes         += local_stats.tx_bytes;
 391                 stats->tx_packets       += local_stats.tx_packets;
 392         }
 393 }
 394
 395 /**
 396  *      vport_get_options - retrieve device options
 397  *
 398  * @vport: vport from which to retrieve the options.
 399  * @skb: sk_buff where options should be appended.
 400  *
 401  * Retrieves the configuration of the given device, appending an
 402  * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
 403  * vport-specific attributes to @skb.
 404  *
 405  * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
 406  * negative error code if a real error occurred.  If an error occurs, @skb is
 407  * left unmodified.
 408  *
 409  * Must be called with RTNL lock or rcu_read_lock.
 410  */
 411 int vport_get_options(const struct vport *vport, struct sk_buff *skb)
 412 {
 413         struct nlattr *nla;
 414
 415         nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
 416         if (!nla)
 417                 return -EMSGSIZE;
 418
 419         if (vport->ops->get_options) {
 420                 int err = vport->ops->get_options(vport, skb);
 421                 if (err) {
 422                         nla_nest_cancel(skb, nla);
 423                         return err;
 424                 }
 425         }
 426
 427         nla_nest_end(skb, nla);
 428         return 0;
 429 }
 430
 431 /**
 432  *      vport_receive - pass up received packet to the datapath for processing
 433  *
 434  * @vport: vport that received the packet
 435  * @skb: skb that was received
 436  *
 437  * Must be called with rcu_read_lock.  The packet cannot be shared and
 438  * skb->data should point to the Ethernet header.  The caller must have already
 439  * called compute_ip_summed() to initialize the checksumming fields.
 440  */
 441 void vport_receive(struct vport *vport, struct sk_buff *skb)
 442 {
 443         struct vport_percpu_stats *stats;
 444
 445         stats = per_cpu_ptr(vport->percpu_stats, smp_processor_id());
 446
 447         u64_stats_update_begin(&stats->sync);
 448         stats->rx_packets++;
 449         stats->rx_bytes += skb->len;
 450         u64_stats_update_end(&stats->sync);
 451
 452         if (!(vport->ops->flags & VPORT_F_FLOW))
 453                 OVS_CB(skb)->flow = NULL;
 454
 455         if (!(vport->ops->flags & VPORT_F_TUN_ID))
 456                 OVS_CB(skb)->tun_id = 0;
 457
 458         dp_process_received_packet(vport, skb);
 459 }
 460
 461 /**
 462  *      vport_send - send a packet on a device
 463  *
 464  * @vport: vport on which to send the packet
 465  * @skb: skb to send
 466  *
 467  * Sends the given packet and returns the length of data sent.  Either RTNL
 468  * lock or rcu_read_lock must be held.
 469  */
 470 int vport_send(struct vport *vport, struct sk_buff *skb)
 471 {
 472         int sent = vport->ops->send(vport, skb);
 473
 474         if (likely(sent)) {
 475                 struct vport_percpu_stats *stats;
 476
 477                 stats = per_cpu_ptr(vport->percpu_stats, smp_processor_id());
 478
 479                 u64_stats_update_begin(&stats->sync);
 480                 stats->tx_packets++;
 481                 stats->tx_bytes += sent;
 482                 u64_stats_update_end(&stats->sync);
 483         }
 484         return sent;
 485 }
 486
 487 /**
 488  *      vport_record_error - indicate device error to generic stats layer
 489  *
 490  * @vport: vport that encountered the error
 491  * @err_type: one of enum vport_err_type types to indicate the error type
 492  *
 493  * If using the vport generic stats layer indicate that an error of the given
 494  * type has occured.
 495  */
 496 void vport_record_error(struct vport *vport, enum vport_err_type err_type)
 497 {
 498         spin_lock(&vport->stats_lock);
 499
 500         switch (err_type) {
 501         case VPORT_E_RX_DROPPED:
 502                 vport->err_stats.rx_dropped++;
 503                 break;
 504
 505         case VPORT_E_RX_ERROR:
 506                 vport->err_stats.rx_errors++;
 507                 break;
 508
 509         case VPORT_E_TX_DROPPED:
 510                 vport->err_stats.tx_dropped++;
 511                 break;
 512
 513         case VPORT_E_TX_ERROR:
 514                 vport->err_stats.tx_errors++;
 515                 break;
 516         };
 517
 518         spin_unlock(&vport->stats_lock);
 519 }