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 #ifdef CONFIG_OPENVSWITCH_VXLAN
  46         &ovs_vxlan_vport_ops,
  47 #endif
  48 };
  49
  50 /* Protected by RCU read lock for reading, ovs_mutex for writing. */
  51 static struct hlist_head *dev_table;
  52 #define VPORT_HASH_BUCKETS 1024
  53
  54 /**
  55  *      ovs_vport_init - initialize vport subsystem
  56  *
  57  * Called at module load time to initialize the vport subsystem.
  58  */
  59 int ovs_vport_init(void)
  60 {
  61         dev_table = kzalloc(VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
  62                             GFP_KERNEL);
  63         if (!dev_table)
  64                 return -ENOMEM;
  65
  66         return 0;
  67 }
  68
  69 /**
  70  *      ovs_vport_exit - shutdown vport subsystem
  71  *
  72  * Called at module exit time to shutdown the vport subsystem.
  73  */
  74 void ovs_vport_exit(void)
  75 {
  76         kfree(dev_table);
  77 }
  78
  79 static struct hlist_head *hash_bucket(struct net *net, const char *name)
  80 {
  81         unsigned int hash = jhash(name, strlen(name), (unsigned long) net);
  82         return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
  83 }
  84
  85 /**
  86  *      ovs_vport_locate - find a port that has already been created
  87  *
  88  * @name: name of port to find
  89  *
  90  * Must be called with ovs or RCU read lock.
  91  */
  92 struct vport *ovs_vport_locate(struct net *net, const char *name)
  93 {
  94         struct hlist_head *bucket = hash_bucket(net, name);
  95         struct vport *vport;
  96
  97         hlist_for_each_entry_rcu(vport, bucket, hash_node)
  98                 if (!strcmp(name, vport->ops->get_name(vport)) &&
  99                     net_eq(ovs_dp_get_net(vport->dp), net))
 100                         return vport;
 101
 102         return NULL;
 103 }
 104
 105 /**
 106  *      ovs_vport_alloc - allocate and initialize new vport
 107  *
 108  * @priv_size: Size of private data area to allocate.
 109  * @ops: vport device ops
 110  *
 111  * Allocate and initialize a new vport defined by @ops.  The vport will contain
 112  * a private data area of size @priv_size that can be accessed using
 113  * vport_priv().  vports that are no longer needed should be released with
 114  * vport_free().
 115  */
 116 struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
 117                           const struct vport_parms *parms)
 118 {
 119         struct vport *vport;
 120         size_t alloc_size;
 121         int i;
 122
 123         alloc_size = sizeof(struct vport);
 124         if (priv_size) {
 125                 alloc_size = ALIGN(alloc_size, VPORT_ALIGN);
 126                 alloc_size += priv_size;
 127         }
 128
 129         vport = kzalloc(alloc_size, GFP_KERNEL);
 130         if (!vport)
 131                 return ERR_PTR(-ENOMEM);
 132
 133         vport->dp = parms->dp;
 134         vport->port_no = parms->port_no;
 135         vport->upcall_portid = parms->upcall_portid;
 136         vport->ops = ops;
 137         INIT_HLIST_NODE(&vport->dp_hash_node);
 138
 139         vport->percpu_stats = alloc_percpu(struct pcpu_tstats);
 140         if (!vport->percpu_stats) {
 141                 kfree(vport);
 142                 return ERR_PTR(-ENOMEM);
 143         }
 144
 145         for_each_possible_cpu(i) {
 146                 struct pcpu_tstats *vport_stats;
 147                 vport_stats = per_cpu_ptr(vport->percpu_stats, i);
 148                 u64_stats_init(&vport_stats->syncp);
 149         }
 150
 151
 152         spin_lock_init(&vport->stats_lock);
 153
 154         return vport;
 155 }
 156
 157 /**
 158  *      ovs_vport_free - uninitialize and free vport
 159  *
 160  * @vport: vport to free
 161  *
 162  * Frees a vport allocated with vport_alloc() when it is no longer needed.
 163  *
 164  * The caller must ensure that an RCU grace period has passed since the last
 165  * time @vport was in a datapath.
 166  */
 167 void ovs_vport_free(struct vport *vport)
 168 {
 169         free_percpu(vport->percpu_stats);
 170         kfree(vport);
 171 }
 172
 173 /**
 174  *      ovs_vport_add - add vport device (for kernel callers)
 175  *
 176  * @parms: Information about new vport.
 177  *
 178  * Creates a new vport with the specified configuration (which is dependent on
 179  * device type).  ovs_mutex must be held.
 180  */
 181 struct vport *ovs_vport_add(const struct vport_parms *parms)
 182 {
 183         struct vport *vport;
 184         int err = 0;
 185         int i;
 186
 187         for (i = 0; i < ARRAY_SIZE(vport_ops_list); i++) {
 188                 if (vport_ops_list[i]->type == parms->type) {
 189                         struct hlist_head *bucket;
 190
 191                         vport = vport_ops_list[i]->create(parms);
 192                         if (IS_ERR(vport)) {
 193                                 err = PTR_ERR(vport);
 194                                 goto out;
 195                         }
 196
 197                         bucket = hash_bucket(ovs_dp_get_net(vport->dp),
 198                                              vport->ops->get_name(vport));
 199                         hlist_add_head_rcu(&vport->hash_node, bucket);
 200                         return vport;
 201                 }
 202         }
 203
 204         err = -EAFNOSUPPORT;
 205
 206 out:
 207         return ERR_PTR(err);
 208 }
 209
 210 /**
 211  *      ovs_vport_set_options - modify existing vport device (for kernel callers)
 212  *
 213  * @vport: vport to modify.
 214  * @options: New configuration.
 215  *
 216  * Modifies an existing device with the specified configuration (which is
 217  * dependent on device type).  ovs_mutex must be held.
 218  */
 219 int ovs_vport_set_options(struct vport *vport, struct nlattr *options)
 220 {
 221         if (!vport->ops->set_options)
 222                 return -EOPNOTSUPP;
 223         return vport->ops->set_options(vport, options);
 224 }
 225
 226 /**
 227  *      ovs_vport_del - delete existing vport device
 228  *
 229  * @vport: vport to delete.
 230  *
 231  * Detaches @vport from its datapath and destroys it.  It is possible to fail
 232  * for reasons such as lack of memory.  ovs_mutex must be held.
 233  */
 234 void ovs_vport_del(struct vport *vport)
 235 {
 236         ASSERT_OVSL();
 237
 238         hlist_del_rcu(&vport->hash_node);
 239
 240         vport->ops->destroy(vport);
 241 }
 242
 243 /**
 244  *      ovs_vport_get_stats - retrieve device stats
 245  *
 246  * @vport: vport from which to retrieve the stats
 247  * @stats: location to store stats
 248  *
 249  * Retrieves transmit, receive, and error stats for the given device.
 250  *
 251  * Must be called with ovs_mutex or rcu_read_lock.
 252  */
 253 void ovs_vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
 254 {
 255         int i;
 256
 257         memset(stats, 0, sizeof(*stats));
 258
 259         /* We potentially have 2 sources of stats that need to be combined:
 260          * those we have collected (split into err_stats and percpu_stats) from
 261          * set_stats() and device error stats from netdev->get_stats() (for
 262          * errors that happen  downstream and therefore aren't reported through
 263          * our vport_record_error() function).
 264          * Stats from first source are reported by ovs (OVS_VPORT_ATTR_STATS).
 265          * netdev-stats can be directly read over netlink-ioctl.
 266          */
 267
 268         spin_lock_bh(&vport->stats_lock);
 269
 270         stats->rx_errors        = vport->err_stats.rx_errors;
 271         stats->tx_errors        = vport->err_stats.tx_errors;
 272         stats->tx_dropped       = vport->err_stats.tx_dropped;
 273         stats->rx_dropped       = vport->err_stats.rx_dropped;
 274
 275         spin_unlock_bh(&vport->stats_lock);
 276
 277         for_each_possible_cpu(i) {
 278                 const struct pcpu_tstats *percpu_stats;
 279                 struct pcpu_tstats local_stats;
 280                 unsigned int start;
 281
 282                 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
 283
 284                 do {
 285                         start = u64_stats_fetch_begin_bh(&percpu_stats->syncp);
 286                         local_stats = *percpu_stats;
 287                 } while (u64_stats_fetch_retry_bh(&percpu_stats->syncp, start));
 288
 289                 stats->rx_bytes         += local_stats.rx_bytes;
 290                 stats->rx_packets       += local_stats.rx_packets;
 291                 stats->tx_bytes         += local_stats.tx_bytes;
 292                 stats->tx_packets       += local_stats.tx_packets;
 293         }
 294 }
 295
 296 /**
 297  *      ovs_vport_get_options - retrieve device options
 298  *
 299  * @vport: vport from which to retrieve the options.
 300  * @skb: sk_buff where options should be appended.
 301  *
 302  * Retrieves the configuration of the given device, appending an
 303  * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
 304  * vport-specific attributes to @skb.
 305  *
 306  * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
 307  * negative error code if a real error occurred.  If an error occurs, @skb is
 308  * left unmodified.
 309  *
 310  * Must be called with ovs_mutex or rcu_read_lock.
 311  */
 312 int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
 313 {
 314         struct nlattr *nla;
 315         int err;
 316
 317         if (!vport->ops->get_options)
 318                 return 0;
 319
 320         nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
 321         if (!nla)
 322                 return -EMSGSIZE;
 323
 324         err = vport->ops->get_options(vport, skb);
 325         if (err) {
 326                 nla_nest_cancel(skb, nla);
 327                 return err;
 328         }
 329
 330         nla_nest_end(skb, nla);
 331         return 0;
 332 }
 333
 334 /**
 335  *      ovs_vport_receive - pass up received packet to the datapath for processing
 336  *
 337  * @vport: vport that received the packet
 338  * @skb: skb that was received
 339  * @tun_key: tunnel (if any) that carried packet
 340  *
 341  * Must be called with rcu_read_lock.  The packet cannot be shared and
 342  * skb->data should point to the Ethernet header.
 343  */
 344 void ovs_vport_receive(struct vport *vport, struct sk_buff *skb,
 345                        struct ovs_key_ipv4_tunnel *tun_key)
 346 {
 347         struct pcpu_tstats *stats;
 348
 349         stats = this_cpu_ptr(vport->percpu_stats);
 350         u64_stats_update_begin(&stats->syncp);
 351         stats->rx_packets++;
 352         stats->rx_bytes += skb->len;
 353         u64_stats_update_end(&stats->syncp);
 354
 355         OVS_CB(skb)->tun_key = tun_key;
 356         ovs_dp_process_received_packet(vport, skb);
 357 }
 358
 359 /**
 360  *      ovs_vport_send - send a packet on a device
 361  *
 362  * @vport: vport on which to send the packet
 363  * @skb: skb to send
 364  *
 365  * Sends the given packet and returns the length of data sent.  Either ovs
 366  * lock or rcu_read_lock must be held.
 367  */
 368 int ovs_vport_send(struct vport *vport, struct sk_buff *skb)
 369 {
 370         int sent = vport->ops->send(vport, skb);
 371
 372         if (likely(sent > 0)) {
 373                 struct pcpu_tstats *stats;
 374
 375                 stats = this_cpu_ptr(vport->percpu_stats);
 376
 377                 u64_stats_update_begin(&stats->syncp);
 378                 stats->tx_packets++;
 379                 stats->tx_bytes += sent;
 380                 u64_stats_update_end(&stats->syncp);
 381         } else if (sent < 0) {
 382                 ovs_vport_record_error(vport, VPORT_E_TX_ERROR);
 383                 kfree_skb(skb);
 384         } else
 385                 ovs_vport_record_error(vport, VPORT_E_TX_DROPPED);
 386
 387         return sent;
 388 }
 389
 390 /**
 391  *      ovs_vport_record_error - indicate device error to generic stats layer
 392  *
 393  * @vport: vport that encountered the error
 394  * @err_type: one of enum vport_err_type types to indicate the error type
 395  *
 396  * If using the vport generic stats layer indicate that an error of the given
 397  * type has occurred.
 398  */
 399 void ovs_vport_record_error(struct vport *vport, enum vport_err_type err_type)
 400 {
 401         spin_lock(&vport->stats_lock);
 402
 403         switch (err_type) {
 404         case VPORT_E_RX_DROPPED:
 405                 vport->err_stats.rx_dropped++;
 406                 break;
 407
 408         case VPORT_E_RX_ERROR:
 409                 vport->err_stats.rx_errors++;
 410                 break;
 411
 412         case VPORT_E_TX_DROPPED:
 413                 vport->err_stats.tx_dropped++;
 414                 break;
 415
 416         case VPORT_E_TX_ERROR:
 417                 vport->err_stats.tx_errors++;
 418                 break;
 419         }
 420
 421         spin_unlock(&vport->stats_lock);
 422 }
 423
 424 static void free_vport_rcu(struct rcu_head *rcu)
 425 {
 426         struct vport *vport = container_of(rcu, struct vport, rcu);
 427
 428         ovs_vport_free(vport);
 429 }
 430
 431 void ovs_vport_deferred_free(struct vport *vport)
 432 {
 433         if (!vport)
 434                 return;
 435
 436         call_rcu(&vport->rcu, free_vport_rcu);
 437 }