datapath/vport.c

   1 /*
   2  * Copyright (c) 2007-2012 Nicira, Inc.
   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/etherdevice.h>
  20 #include <linux/if.h>
  21 #include <linux/if_vlan.h>
  22 #include <linux/jhash.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 #include <net/net_namespace.h>
  32
  33 #include "datapath.h"
  34 #include "vport.h"
  35 #include "vport-internal_dev.h"
  36
  37 /* List of statically compiled vport implementations.  Don't forget to also
  38  * add yours to the list at the bottom of vport.h. */
  39 static const struct vport_ops *base_vport_ops_list[] = {
  40         &ovs_netdev_vport_ops,
  41         &ovs_internal_vport_ops,
  42         &ovs_gre_vport_ops,
  43         &ovs_gre64_vport_ops,
  44 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,26)
  45         &ovs_vxlan_vport_ops,
  46         &ovs_lisp_vport_ops,
  47 #endif
  48 };
  49
  50 static const struct vport_ops **vport_ops_list;
  51 static int n_vport_types;
  52
  53 /* Protected by RCU read lock for reading, ovs_mutex for writing. */
  54 static struct hlist_head *dev_table;
  55 #define VPORT_HASH_BUCKETS 1024
  56
  57 /**
  58  *      ovs_vport_init - initialize vport subsystem
  59  *
  60  * Called at module load time to initialize the vport subsystem and any
  61  * compiled in vport types.
  62  */
  63 int ovs_vport_init(void)
  64 {
  65         int err;
  66         int i;
  67
  68         dev_table = kzalloc(VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
  69                             GFP_KERNEL);
  70         if (!dev_table) {
  71                 err = -ENOMEM;
  72                 goto error;
  73         }
  74
  75         vport_ops_list = kmalloc(ARRAY_SIZE(base_vport_ops_list) *
  76                                  sizeof(struct vport_ops *), GFP_KERNEL);
  77         if (!vport_ops_list) {
  78                 err = -ENOMEM;
  79                 goto error_dev_table;
  80         }
  81
  82         for (i = 0; i < ARRAY_SIZE(base_vport_ops_list); i++) {
  83                 const struct vport_ops *new_ops = base_vport_ops_list[i];
  84
  85                 if (new_ops->init)
  86                         err = new_ops->init();
  87                 else
  88                         err = 0;
  89
  90                 if (!err)
  91                         vport_ops_list[n_vport_types++] = new_ops;
  92                 else if (new_ops->flags & VPORT_F_REQUIRED) {
  93                         ovs_vport_exit();
  94                         goto error;
  95                 }
  96         }
  97
  98         return 0;
  99
 100 error_dev_table:
 101         kfree(dev_table);
 102 error:
 103         return err;
 104 }
 105
 106 /**
 107  *      ovs_vport_exit - shutdown vport subsystem
 108  *
 109  * Called at module exit time to shutdown the vport subsystem and any
 110  * initialized vport types.
 111  */
 112 void ovs_vport_exit(void)
 113 {
 114         int i;
 115
 116         for (i = 0; i < n_vport_types; i++) {
 117                 if (vport_ops_list[i]->exit)
 118                         vport_ops_list[i]->exit();
 119         }
 120
 121         kfree(vport_ops_list);
 122         kfree(dev_table);
 123 }
 124
 125 static struct hlist_head *hash_bucket(struct net *net, const char *name)
 126 {
 127         unsigned int hash = jhash(name, strlen(name), (unsigned long) net);
 128         return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
 129 }
 130
 131 /**
 132  *      ovs_vport_locate - find a port that has already been created
 133  *
 134  * @name: name of port to find
 135  *
 136  * Must be called with ovs or RCU read lock.
 137  */
 138 struct vport *ovs_vport_locate(struct net *net, const char *name)
 139 {
 140         struct hlist_head *bucket = hash_bucket(net, name);
 141         struct vport *vport;
 142
 143         hlist_for_each_entry_rcu(vport, bucket, hash_node)
 144                 if (!strcmp(name, vport->ops->get_name(vport)) &&
 145                     net_eq(ovs_dp_get_net(vport->dp), net))
 146                         return vport;
 147
 148         return NULL;
 149 }
 150
 151 /**
 152  *      ovs_vport_alloc - allocate and initialize new vport
 153  *
 154  * @priv_size: Size of private data area to allocate.
 155  * @ops: vport device ops
 156  *
 157  * Allocate and initialize a new vport defined by @ops.  The vport will contain
 158  * a private data area of size @priv_size that can be accessed using
 159  * vport_priv().  vports that are no longer needed should be released with
 160  * ovs_vport_free().
 161  */
 162 struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
 163                               const struct vport_parms *parms)
 164 {
 165         struct vport *vport;
 166         size_t alloc_size;
 167
 168         alloc_size = sizeof(struct vport);
 169         if (priv_size) {
 170                 alloc_size = ALIGN(alloc_size, VPORT_ALIGN);
 171                 alloc_size += priv_size;
 172         }
 173
 174         vport = kzalloc(alloc_size, GFP_KERNEL);
 175         if (!vport)
 176                 return ERR_PTR(-ENOMEM);
 177
 178         vport->dp = parms->dp;
 179         vport->port_no = parms->port_no;
 180         vport->upcall_portid = parms->upcall_portid;
 181         vport->ops = ops;
 182         INIT_HLIST_NODE(&vport->dp_hash_node);
 183
 184         vport->percpu_stats = alloc_percpu(struct vport_percpu_stats);
 185         if (!vport->percpu_stats) {
 186                 kfree(vport);
 187                 return ERR_PTR(-ENOMEM);
 188         }
 189
 190         spin_lock_init(&vport->stats_lock);
 191
 192         return vport;
 193 }
 194
 195 /**
 196  *      ovs_vport_free - uninitialize and free vport
 197  *
 198  * @vport: vport to free
 199  *
 200  * Frees a vport allocated with ovs_vport_alloc() when it is no longer needed.
 201  *
 202  * The caller must ensure that an RCU grace period has passed since the last
 203  * time @vport was in a datapath.
 204  */
 205 void ovs_vport_free(struct vport *vport)
 206 {
 207         free_percpu(vport->percpu_stats);
 208         kfree(vport);
 209 }
 210
 211 /**
 212  *      ovs_vport_add - add vport device (for kernel callers)
 213  *
 214  * @parms: Information about new vport.
 215  *
 216  * Creates a new vport with the specified configuration (which is dependent on
 217  * device type).  ovs_mutex must be held.
 218  */
 219 struct vport *ovs_vport_add(const struct vport_parms *parms)
 220 {
 221         struct vport *vport;
 222         int err = 0;
 223         int i;
 224
 225         for (i = 0; i < n_vport_types; i++) {
 226                 if (vport_ops_list[i]->type == parms->type) {
 227                         struct hlist_head *bucket;
 228
 229                         vport = vport_ops_list[i]->create(parms);
 230                         if (IS_ERR(vport)) {
 231                                 err = PTR_ERR(vport);
 232                                 goto out;
 233                         }
 234
 235                         bucket = hash_bucket(ovs_dp_get_net(vport->dp),
 236                                              vport->ops->get_name(vport));
 237                         hlist_add_head_rcu(&vport->hash_node, bucket);
 238                         return vport;
 239                 }
 240         }
 241
 242         err = -EAFNOSUPPORT;
 243
 244 out:
 245         return ERR_PTR(err);
 246 }
 247
 248 /**
 249  *      ovs_vport_set_options - modify existing vport device (for kernel callers)
 250  *
 251  * @vport: vport to modify.
 252  * @port: New configuration.
 253  *
 254  * Modifies an existing device with the specified configuration (which is
 255  * dependent on device type).  ovs_mutex must be held.
 256  */
 257 int ovs_vport_set_options(struct vport *vport, struct nlattr *options)
 258 {
 259         if (!vport->ops->set_options)
 260                 return -EOPNOTSUPP;
 261         return vport->ops->set_options(vport, options);
 262 }
 263
 264 /**
 265  *      ovs_vport_del - delete existing vport device
 266  *
 267  * @vport: vport to delete.
 268  *
 269  * Detaches @vport from its datapath and destroys it.  It is possible to fail
 270  * for reasons such as lack of memory.  ovs_mutex must be held.
 271  */
 272 void ovs_vport_del(struct vport *vport)
 273 {
 274         ASSERT_OVSL();
 275
 276         hlist_del_rcu(&vport->hash_node);
 277         vport->ops->destroy(vport);
 278 }
 279
 280 /**
 281  *      ovs_vport_set_stats - sets offset device stats
 282  *
 283  * @vport: vport on which to set stats
 284  * @stats: stats to set
 285  *
 286  * Provides a set of transmit, receive, and error stats to be added as an
 287  * offset to the collect data when stats are retreived.  Some devices may not
 288  * support setting the stats, in which case the result will always be
 289  * -EOPNOTSUPP.
 290  *
 291  * Must be called with ovs_mutex.
 292  */
 293 void ovs_vport_set_stats(struct vport *vport, struct ovs_vport_stats *stats)
 294 {
 295         spin_lock_bh(&vport->stats_lock);
 296         vport->offset_stats = *stats;
 297         spin_unlock_bh(&vport->stats_lock);
 298 }
 299
 300 /**
 301  *      ovs_vport_get_stats - retrieve device stats
 302  *
 303  * @vport: vport from which to retrieve the stats
 304  * @stats: location to store stats
 305  *
 306  * Retrieves transmit, receive, and error stats for the given device.
 307  *
 308  * Must be called with ovs_mutex or rcu_read_lock.
 309  */
 310 void ovs_vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
 311 {
 312         int i;
 313
 314         /* We potentially have 3 sources of stats that need to be
 315          * combined: those we have collected (split into err_stats and
 316          * percpu_stats), offset_stats from set_stats(), and device
 317          * error stats from netdev->get_stats() (for errors that happen
 318          * downstream and therefore aren't reported through our
 319          * vport_record_error() function).
 320          * Stats from first two sources are merged and reported by ovs over
 321          * OVS_VPORT_ATTR_STATS.
 322          * netdev-stats can be directly read over netlink-ioctl.
 323          */
 324
 325         spin_lock_bh(&vport->stats_lock);
 326
 327         *stats = vport->offset_stats;
 328
 329         stats->rx_errors        += vport->err_stats.rx_errors;
 330         stats->tx_errors        += vport->err_stats.tx_errors;
 331         stats->tx_dropped       += vport->err_stats.tx_dropped;
 332         stats->rx_dropped       += vport->err_stats.rx_dropped;
 333
 334         spin_unlock_bh(&vport->stats_lock);
 335
 336         for_each_possible_cpu(i) {
 337                 const struct vport_percpu_stats *percpu_stats;
 338                 struct vport_percpu_stats local_stats;
 339                 unsigned int start;
 340
 341                 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
 342
 343                 do {
 344                         start = u64_stats_fetch_begin_bh(&percpu_stats->sync);
 345                         local_stats = *percpu_stats;
 346                 } while (u64_stats_fetch_retry_bh(&percpu_stats->sync, start));
 347
 348                 stats->rx_bytes         += local_stats.rx_bytes;
 349                 stats->rx_packets       += local_stats.rx_packets;
 350                 stats->tx_bytes         += local_stats.tx_bytes;
 351                 stats->tx_packets       += local_stats.tx_packets;
 352         }
 353 }
 354
 355 /**
 356  *      ovs_vport_get_options - retrieve device options
 357  *
 358  * @vport: vport from which to retrieve the options.
 359  * @skb: sk_buff where options should be appended.
 360  *
 361  * Retrieves the configuration of the given device, appending an
 362  * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
 363  * vport-specific attributes to @skb.
 364  *
 365  * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
 366  * negative error code if a real error occurred.  If an error occurs, @skb is
 367  * left unmodified.
 368  *
 369  * Must be called with ovs_mutex or rcu_read_lock.
 370  */
 371 int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
 372 {
 373         struct nlattr *nla;
 374         int err;
 375
 376         if (!vport->ops->get_options)
 377                 return 0;
 378
 379         nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
 380         if (!nla)
 381                 return -EMSGSIZE;
 382
 383         err = vport->ops->get_options(vport, skb);
 384         if (err) {
 385                 nla_nest_cancel(skb, nla);
 386                 return err;
 387         }
 388
 389         nla_nest_end(skb, nla);
 390         return 0;
 391 }
 392
 393 /**
 394  *      ovs_vport_receive - pass up received packet to the datapath for processing
 395  *
 396  * @vport: vport that received the packet
 397  * @skb: skb that was received
 398  *
 399  * Must be called with rcu_read_lock.  The packet cannot be shared and
 400  * skb->data should point to the Ethernet header.  The caller must have already
 401  * called compute_ip_summed() to initialize the checksumming fields.
 402  */
 403 void ovs_vport_receive(struct vport *vport, struct sk_buff *skb)
 404 {
 405         struct vport_percpu_stats *stats;
 406
 407         stats = this_cpu_ptr(vport->percpu_stats);
 408         u64_stats_update_begin(&stats->sync);
 409         stats->rx_packets++;
 410         stats->rx_bytes += skb->len;
 411         u64_stats_update_end(&stats->sync);
 412
 413         if (!(vport->ops->flags & VPORT_F_TUN_ID))
 414                 OVS_CB(skb)->tun_key = NULL;
 415
 416         ovs_dp_process_received_packet(vport, skb);
 417 }
 418
 419 /**
 420  *      ovs_vport_send - send a packet on a device
 421  *
 422  * @vport: vport on which to send the packet
 423  * @skb: skb to send
 424  *
 425  * Sends the given packet and returns the length of data sent.  Either ovs
 426  * lock or rcu_read_lock must be held.
 427  */
 428 int ovs_vport_send(struct vport *vport, struct sk_buff *skb)
 429 {
 430         int sent = vport->ops->send(vport, skb);
 431
 432         if (likely(sent)) {
 433                 struct vport_percpu_stats *stats;
 434
 435                 stats = this_cpu_ptr(vport->percpu_stats);
 436
 437                 u64_stats_update_begin(&stats->sync);
 438                 stats->tx_packets++;
 439                 stats->tx_bytes += sent;
 440                 u64_stats_update_end(&stats->sync);
 441         }
 442         return sent;
 443 }
 444
 445 /**
 446  *      ovs_vport_record_error - indicate device error to generic stats layer
 447  *
 448  * @vport: vport that encountered the error
 449  * @err_type: one of enum vport_err_type types to indicate the error type
 450  *
 451  * If using the vport generic stats layer indicate that an error of the given
 452  * type has occured.
 453  */
 454 void ovs_vport_record_error(struct vport *vport, enum vport_err_type err_type)
 455 {
 456         spin_lock(&vport->stats_lock);
 457
 458         switch (err_type) {
 459         case VPORT_E_RX_DROPPED:
 460                 vport->err_stats.rx_dropped++;
 461                 break;
 462
 463         case VPORT_E_RX_ERROR:
 464                 vport->err_stats.rx_errors++;
 465                 break;
 466
 467         case VPORT_E_TX_DROPPED:
 468                 vport->err_stats.tx_dropped++;
 469                 break;
 470
 471         case VPORT_E_TX_ERROR:
 472                 vport->err_stats.tx_errors++;
 473                 break;
 474         }
 475
 476         spin_unlock(&vport->stats_lock);
 477 }