net/openvswitch/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 <net/net_namespace.h>
  31
  32 #include "datapath.h"
  33 #include "vport.h"
  34 #include "vport-internal_dev.h"
  35
  36 /* List of statically compiled vport implementations.  Don't forget to also
  37  * add yours to the list at the bottom of vport.h. */
  38 static const struct vport_ops *vport_ops_list[] = {
  39         &ovs_netdev_vport_ops,
  40         &ovs_internal_vport_ops,
  41
  42 #ifdef CONFIG_OPENVSWITCH_GRE
  43         &ovs_gre_vport_ops,
  44 #endif
  45 };
  46
  47 /* Protected by RCU read lock for reading, ovs_mutex for writing. */
  48 static struct hlist_head *dev_table;
  49 #define VPORT_HASH_BUCKETS 1024
  50
  51 /**
  52  *      ovs_vport_init - initialize vport subsystem
  53  *
  54  * Called at module load time to initialize the vport subsystem.
  55  */
  56 int ovs_vport_init(void)
  57 {
  58         dev_table = kzalloc(VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
  59                             GFP_KERNEL);
  60         if (!dev_table)
  61                 return -ENOMEM;
  62
  63         return 0;
  64 }
  65
  66 /**
  67  *      ovs_vport_exit - shutdown vport subsystem
  68  *
  69  * Called at module exit time to shutdown the vport subsystem.
  70  */
  71 void ovs_vport_exit(void)
  72 {
  73         kfree(dev_table);
  74 }
  75
  76 static struct hlist_head *hash_bucket(struct net *net, const char *name)
  77 {
  78         unsigned int hash = jhash(name, strlen(name), (unsigned long) net);
  79         return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
  80 }
  81
  82 /**
  83  *      ovs_vport_locate - find a port that has already been created
  84  *
  85  * @name: name of port to find
  86  *
  87  * Must be called with ovs or RCU read lock.
  88  */
  89 struct vport *ovs_vport_locate(struct net *net, const char *name)
  90 {
  91         struct hlist_head *bucket = hash_bucket(net, name);
  92         struct vport *vport;
  93
  94         hlist_for_each_entry_rcu(vport, bucket, hash_node)
  95                 if (!strcmp(name, vport->ops->get_name(vport)) &&
  96                     net_eq(ovs_dp_get_net(vport->dp), net))
  97                         return vport;
  98
  99         return NULL;
 100 }
 101
 102 /**
 103  *      ovs_vport_alloc - allocate and initialize new vport
 104  *
 105  * @priv_size: Size of private data area to allocate.
 106  * @ops: vport device ops
 107  *
 108  * Allocate and initialize a new vport defined by @ops.  The vport will contain
 109  * a private data area of size @priv_size that can be accessed using
 110  * vport_priv().  vports that are no longer needed should be released with
 111  * vport_free().
 112  */
 113 struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
 114                           const struct vport_parms *parms)
 115 {
 116         struct vport *vport;
 117         size_t alloc_size;
 118
 119         alloc_size = sizeof(struct vport);
 120         if (priv_size) {
 121                 alloc_size = ALIGN(alloc_size, VPORT_ALIGN);
 122                 alloc_size += priv_size;
 123         }
 124
 125         vport = kzalloc(alloc_size, GFP_KERNEL);
 126         if (!vport)
 127                 return ERR_PTR(-ENOMEM);
 128
 129         vport->dp = parms->dp;
 130         vport->port_no = parms->port_no;
 131         vport->upcall_portid = parms->upcall_portid;
 132         vport->ops = ops;
 133         INIT_HLIST_NODE(&vport->dp_hash_node);
 134
 135         vport->percpu_stats = alloc_percpu(struct pcpu_tstats);
 136         if (!vport->percpu_stats) {
 137                 kfree(vport);
 138                 return ERR_PTR(-ENOMEM);
 139         }
 140
 141         spin_lock_init(&vport->stats_lock);
 142
 143         return vport;
 144 }
 145
 146 /**
 147  *      ovs_vport_free - uninitialize and free vport
 148  *
 149  * @vport: vport to free
 150  *
 151  * Frees a vport allocated with vport_alloc() when it is no longer needed.
 152  *
 153  * The caller must ensure that an RCU grace period has passed since the last
 154  * time @vport was in a datapath.
 155  */
 156 void ovs_vport_free(struct vport *vport)
 157 {
 158         free_percpu(vport->percpu_stats);
 159         kfree(vport);
 160 }
 161
 162 /**
 163  *      ovs_vport_add - add vport device (for kernel callers)
 164  *
 165  * @parms: Information about new vport.
 166  *
 167  * Creates a new vport with the specified configuration (which is dependent on
 168  * device type).  ovs_mutex must be held.
 169  */
 170 struct vport *ovs_vport_add(const struct vport_parms *parms)
 171 {
 172         struct vport *vport;
 173         int err = 0;
 174         int i;
 175
 176         for (i = 0; i < ARRAY_SIZE(vport_ops_list); i++) {
 177                 if (vport_ops_list[i]->type == parms->type) {
 178                         struct hlist_head *bucket;
 179
 180                         vport = vport_ops_list[i]->create(parms);
 181                         if (IS_ERR(vport)) {
 182                                 err = PTR_ERR(vport);
 183                                 goto out;
 184                         }
 185
 186                         bucket = hash_bucket(ovs_dp_get_net(vport->dp),
 187                                              vport->ops->get_name(vport));
 188                         hlist_add_head_rcu(&vport->hash_node, bucket);
 189                         return vport;
 190                 }
 191         }
 192
 193         err = -EAFNOSUPPORT;
 194
 195 out:
 196         return ERR_PTR(err);
 197 }
 198
 199 /**
 200  *      ovs_vport_set_options - modify existing vport device (for kernel callers)
 201  *
 202  * @vport: vport to modify.
 203  * @port: New configuration.
 204  *
 205  * Modifies an existing device with the specified configuration (which is
 206  * dependent on device type).  ovs_mutex must be held.
 207  */
 208 int ovs_vport_set_options(struct vport *vport, struct nlattr *options)
 209 {
 210         if (!vport->ops->set_options)
 211                 return -EOPNOTSUPP;
 212         return vport->ops->set_options(vport, options);
 213 }
 214
 215 /**
 216  *      ovs_vport_del - delete existing vport device
 217  *
 218  * @vport: vport to delete.
 219  *
 220  * Detaches @vport from its datapath and destroys it.  It is possible to fail
 221  * for reasons such as lack of memory.  ovs_mutex must be held.
 222  */
 223 void ovs_vport_del(struct vport *vport)
 224 {
 225         ASSERT_OVSL();
 226
 227         hlist_del_rcu(&vport->hash_node);
 228
 229         vport->ops->destroy(vport);
 230 }
 231
 232 /**
 233  *      ovs_vport_get_stats - retrieve device stats
 234  *
 235  * @vport: vport from which to retrieve the stats
 236  * @stats: location to store stats
 237  *
 238  * Retrieves transmit, receive, and error stats for the given device.
 239  *
 240  * Must be called with ovs_mutex or rcu_read_lock.
 241  */
 242 void ovs_vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
 243 {
 244         int i;
 245
 246         memset(stats, 0, sizeof(*stats));
 247
 248         /* We potentially have 2 sources of stats that need to be combined:
 249          * those we have collected (split into err_stats and percpu_stats) from
 250          * set_stats() and device error stats from netdev->get_stats() (for
 251          * errors that happen  downstream and therefore aren't reported through
 252          * our vport_record_error() function).
 253          * Stats from first source are reported by ovs (OVS_VPORT_ATTR_STATS).
 254          * netdev-stats can be directly read over netlink-ioctl.
 255          */
 256
 257         spin_lock_bh(&vport->stats_lock);
 258
 259         stats->rx_errors        = vport->err_stats.rx_errors;
 260         stats->tx_errors        = vport->err_stats.tx_errors;
 261         stats->tx_dropped       = vport->err_stats.tx_dropped;
 262         stats->rx_dropped       = vport->err_stats.rx_dropped;
 263
 264         spin_unlock_bh(&vport->stats_lock);
 265
 266         for_each_possible_cpu(i) {
 267                 const struct pcpu_tstats *percpu_stats;
 268                 struct pcpu_tstats local_stats;
 269                 unsigned int start;
 270
 271                 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
 272
 273                 do {
 274                         start = u64_stats_fetch_begin_bh(&percpu_stats->syncp);
 275                         local_stats = *percpu_stats;
 276                 } while (u64_stats_fetch_retry_bh(&percpu_stats->syncp, start));
 277
 278                 stats->rx_bytes         += local_stats.rx_bytes;
 279                 stats->rx_packets       += local_stats.rx_packets;
 280                 stats->tx_bytes         += local_stats.tx_bytes;
 281                 stats->tx_packets       += local_stats.tx_packets;
 282         }
 283 }
 284
 285 /**
 286  *      ovs_vport_get_options - retrieve device options
 287  *
 288  * @vport: vport from which to retrieve the options.
 289  * @skb: sk_buff where options should be appended.
 290  *
 291  * Retrieves the configuration of the given device, appending an
 292  * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
 293  * vport-specific attributes to @skb.
 294  *
 295  * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
 296  * negative error code if a real error occurred.  If an error occurs, @skb is
 297  * left unmodified.
 298  *
 299  * Must be called with ovs_mutex or rcu_read_lock.
 300  */
 301 int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
 302 {
 303         struct nlattr *nla;
 304         int err;
 305
 306         if (!vport->ops->get_options)
 307                 return 0;
 308
 309         nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
 310         if (!nla)
 311                 return -EMSGSIZE;
 312
 313         err = vport->ops->get_options(vport, skb);
 314         if (err) {
 315                 nla_nest_cancel(skb, nla);
 316                 return err;
 317         }
 318
 319         nla_nest_end(skb, nla);
 320         return 0;
 321 }
 322
 323 /**
 324  *      ovs_vport_receive - pass up received packet to the datapath for processing
 325  *
 326  * @vport: vport that received the packet
 327  * @skb: skb that was received
 328  *
 329  * Must be called with rcu_read_lock.  The packet cannot be shared and
 330  * skb->data should point to the Ethernet header.
 331  */
 332 void ovs_vport_receive(struct vport *vport, struct sk_buff *skb,
 333                        struct ovs_key_ipv4_tunnel *tun_key)
 334 {
 335         struct pcpu_tstats *stats;
 336
 337         stats = this_cpu_ptr(vport->percpu_stats);
 338         u64_stats_update_begin(&stats->syncp);
 339         stats->rx_packets++;
 340         stats->rx_bytes += skb->len;
 341         u64_stats_update_end(&stats->syncp);
 342
 343         OVS_CB(skb)->tun_key = tun_key;
 344         ovs_dp_process_received_packet(vport, skb);
 345 }
 346
 347 /**
 348  *      ovs_vport_send - send a packet on a device
 349  *
 350  * @vport: vport on which to send the packet
 351  * @skb: skb to send
 352  *
 353  * Sends the given packet and returns the length of data sent.  Either ovs
 354  * lock or rcu_read_lock must be held.
 355  */
 356 int ovs_vport_send(struct vport *vport, struct sk_buff *skb)
 357 {
 358         int sent = vport->ops->send(vport, skb);
 359
 360         if (likely(sent > 0)) {
 361                 struct pcpu_tstats *stats;
 362
 363                 stats = this_cpu_ptr(vport->percpu_stats);
 364
 365                 u64_stats_update_begin(&stats->syncp);
 366                 stats->tx_packets++;
 367                 stats->tx_bytes += sent;
 368                 u64_stats_update_end(&stats->syncp);
 369         } else if (sent < 0) {
 370                 ovs_vport_record_error(vport, VPORT_E_TX_ERROR);
 371                 kfree_skb(skb);
 372         } else
 373                 ovs_vport_record_error(vport, VPORT_E_TX_DROPPED);
 374
 375         return sent;
 376 }
 377
 378 /**
 379  *      ovs_vport_record_error - indicate device error to generic stats layer
 380  *
 381  * @vport: vport that encountered the error
 382  * @err_type: one of enum vport_err_type types to indicate the error type
 383  *
 384  * If using the vport generic stats layer indicate that an error of the given
 385  * type has occurred.
 386  */
 387 void ovs_vport_record_error(struct vport *vport, enum vport_err_type err_type)
 388 {
 389         spin_lock(&vport->stats_lock);
 390
 391         switch (err_type) {
 392         case VPORT_E_RX_DROPPED:
 393                 vport->err_stats.rx_dropped++;
 394                 break;
 395
 396         case VPORT_E_RX_ERROR:
 397                 vport->err_stats.rx_errors++;
 398                 break;
 399
 400         case VPORT_E_TX_DROPPED:
 401                 vport->err_stats.tx_dropped++;
 402                 break;
 403
 404         case VPORT_E_TX_ERROR:
 405                 vport->err_stats.tx_errors++;
 406                 break;
 407         }
 408
 409         spin_unlock(&vport->stats_lock);
 410 }
 411
 412 static void free_vport_rcu(struct rcu_head *rcu)
 413 {
 414         struct vport *vport = container_of(rcu, struct vport, rcu);
 415
 416         ovs_vport_free(vport);
 417 }
 418
 419 void ovs_vport_deferred_free(struct vport *vport)
 420 {
 421         if (!vport)
 422                 return;
 423
 424         call_rcu(&vport->rcu, free_vport_rcu);
 425 }