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/kconfig.h>
  24 #include <linux/kernel.h>
  25 #include <linux/list.h>
  26 #include <linux/mutex.h>
  27 #include <linux/percpu.h>
  28 #include <linux/rcupdate.h>
  29 #include <linux/rtnetlink.h>
  30 #include <linux/compat.h>
  31 #include <linux/version.h>
  32 #include <net/net_namespace.h>
  33
  34 #include "datapath.h"
  35 #include "vport.h"
  36 #include "vport-internal_dev.h"
  37
  38 /* List of statically compiled vport implementations.  Don't forget to also
  39  * add yours to the list at the bottom of vport.h. */
  40 static const struct vport_ops *vport_ops_list[] = {
  41         &ovs_netdev_vport_ops,
  42         &ovs_internal_vport_ops,
  43 #if IS_ENABLED(CONFIG_NET_IPGRE_DEMUX)
  44         &ovs_gre_vport_ops,
  45         &ovs_gre64_vport_ops,
  46 #endif
  47         &ovs_vxlan_vport_ops,
  48         &ovs_lisp_vport_ops,
  49 };
  50
  51 /* Protected by RCU read lock for reading, ovs_mutex for writing. */
  52 static struct hlist_head *dev_table;
  53 #define VPORT_HASH_BUCKETS 1024
  54
  55 /**
  56  *      ovs_vport_init - initialize vport subsystem
  57  *
  58  * Called at module load time to initialize the vport subsystem.
  59  */
  60 int ovs_vport_init(void)
  61 {
  62         dev_table = kzalloc(VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
  63                             GFP_KERNEL);
  64         if (!dev_table)
  65                 return -ENOMEM;
  66
  67         return 0;
  68 }
  69
  70 /**
  71  *      ovs_vport_exit - shutdown vport subsystem
  72  *
  73  * Called at module exit time to shutdown the vport subsystem.
  74  */
  75 void ovs_vport_exit(void)
  76 {
  77         kfree(dev_table);
  78 }
  79
  80 static struct hlist_head *hash_bucket(struct net *net, const char *name)
  81 {
  82         unsigned int hash = jhash(name, strlen(name), (unsigned long) net);
  83         return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
  84 }
  85
  86 /**
  87  *      ovs_vport_locate - find a port that has already been created
  88  *
  89  * @name: name of port to find
  90  *
  91  * Must be called with ovs or RCU read lock.
  92  */
  93 struct vport *ovs_vport_locate(struct net *net, const char *name)
  94 {
  95         struct hlist_head *bucket = hash_bucket(net, name);
  96         struct vport *vport;
  97
  98         hlist_for_each_entry_rcu(vport, bucket, hash_node)
  99                 if (!strcmp(name, vport->ops->get_name(vport)) &&
 100                     net_eq(ovs_dp_get_net(vport->dp), net))
 101                         return vport;
 102
 103         return NULL;
 104 }
 105
 106 /**
 107  *      ovs_vport_alloc - allocate and initialize new vport
 108  *
 109  * @priv_size: Size of private data area to allocate.
 110  * @ops: vport device ops
 111  *
 112  * Allocate and initialize a new vport defined by @ops.  The vport will contain
 113  * a private data area of size @priv_size that can be accessed using
 114  * vport_priv().  vports that are no longer needed should be released with
 115  * ovs_vport_free().
 116  */
 117 struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
 118                               const struct vport_parms *parms)
 119 {
 120         struct vport *vport;
 121         size_t alloc_size;
 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         spin_lock_init(&vport->stats_lock);
 146
 147         return vport;
 148 }
 149
 150 /**
 151  *      ovs_vport_free - uninitialize and free vport
 152  *
 153  * @vport: vport to free
 154  *
 155  * Frees a vport allocated with ovs_vport_alloc() when it is no longer needed.
 156  *
 157  * The caller must ensure that an RCU grace period has passed since the last
 158  * time @vport was in a datapath.
 159  */
 160 void ovs_vport_free(struct vport *vport)
 161 {
 162         free_percpu(vport->percpu_stats);
 163         kfree(vport);
 164 }
 165
 166 /**
 167  *      ovs_vport_add - add vport device (for kernel callers)
 168  *
 169  * @parms: Information about new vport.
 170  *
 171  * Creates a new vport with the specified configuration (which is dependent on
 172  * device type).  ovs_mutex must be held.
 173  */
 174 struct vport *ovs_vport_add(const struct vport_parms *parms)
 175 {
 176         struct vport *vport;
 177         int err = 0;
 178         int i;
 179
 180         for (i = 0; i < ARRAY_SIZE(vport_ops_list); i++) {
 181                 if (vport_ops_list[i]->type == parms->type) {
 182                         struct hlist_head *bucket;
 183
 184                         vport = vport_ops_list[i]->create(parms);
 185                         if (IS_ERR(vport)) {
 186                                 err = PTR_ERR(vport);
 187                                 goto out;
 188                         }
 189
 190                         bucket = hash_bucket(ovs_dp_get_net(vport->dp),
 191                                              vport->ops->get_name(vport));
 192                         hlist_add_head_rcu(&vport->hash_node, bucket);
 193                         return vport;
 194                 }
 195         }
 196
 197         err = -EAFNOSUPPORT;
 198
 199 out:
 200         return ERR_PTR(err);
 201 }
 202
 203 /**
 204  *      ovs_vport_set_options - modify existing vport device (for kernel callers)
 205  *
 206  * @vport: vport to modify.
 207  * @options: New configuration.
 208  *
 209  * Modifies an existing device with the specified configuration (which is
 210  * dependent on device type).  ovs_mutex must be held.
 211  */
 212 int ovs_vport_set_options(struct vport *vport, struct nlattr *options)
 213 {
 214         if (!vport->ops->set_options)
 215                 return -EOPNOTSUPP;
 216         return vport->ops->set_options(vport, options);
 217 }
 218
 219 /**
 220  *      ovs_vport_del - delete existing vport device
 221  *
 222  * @vport: vport to delete.
 223  *
 224  * Detaches @vport from its datapath and destroys it.  It is possible to fail
 225  * for reasons such as lack of memory.  ovs_mutex must be held.
 226  */
 227 void ovs_vport_del(struct vport *vport)
 228 {
 229         ASSERT_OVSL();
 230
 231         hlist_del_rcu(&vport->hash_node);
 232         vport->ops->destroy(vport);
 233 }
 234
 235 /**
 236  *      ovs_vport_set_stats - sets offset device stats
 237  *
 238  * @vport: vport on which to set stats
 239  * @stats: stats to set
 240  *
 241  * Provides a set of transmit, receive, and error stats to be added as an
 242  * offset to the collected data when stats are retrieved.  Some devices may not
 243  * support setting the stats, in which case the result will always be
 244  * -EOPNOTSUPP.
 245  *
 246  * Must be called with ovs_mutex.
 247  */
 248 void ovs_vport_set_stats(struct vport *vport, struct ovs_vport_stats *stats)
 249 {
 250         spin_lock_bh(&vport->stats_lock);
 251         vport->offset_stats = *stats;
 252         spin_unlock_bh(&vport->stats_lock);
 253 }
 254
 255 /**
 256  *      ovs_vport_get_stats - retrieve device stats
 257  *
 258  * @vport: vport from which to retrieve the stats
 259  * @stats: location to store stats
 260  *
 261  * Retrieves transmit, receive, and error stats for the given device.
 262  *
 263  * Must be called with ovs_mutex or rcu_read_lock.
 264  */
 265 void ovs_vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
 266 {
 267         int i;
 268
 269         /* We potentially have 3 sources of stats that need to be
 270          * combined: those we have collected (split into err_stats and
 271          * percpu_stats), offset_stats from set_stats(), and device
 272          * error stats from netdev->get_stats() (for errors that happen
 273          * downstream and therefore aren't reported through our
 274          * vport_record_error() function).
 275          * Stats from first two sources are merged and reported by ovs over
 276          * OVS_VPORT_ATTR_STATS.
 277          * netdev-stats can be directly read over netlink-ioctl.
 278          */
 279
 280         spin_lock_bh(&vport->stats_lock);
 281
 282         *stats = vport->offset_stats;
 283
 284         stats->rx_errors        += vport->err_stats.rx_errors;
 285         stats->tx_errors        += vport->err_stats.tx_errors;
 286         stats->tx_dropped       += vport->err_stats.tx_dropped;
 287         stats->rx_dropped       += vport->err_stats.rx_dropped;
 288
 289         spin_unlock_bh(&vport->stats_lock);
 290
 291         for_each_possible_cpu(i) {
 292                 const struct pcpu_tstats *percpu_stats;
 293                 struct pcpu_tstats local_stats;
 294                 unsigned int start;
 295
 296                 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
 297
 298                 do {
 299                         start = u64_stats_fetch_begin_bh(&percpu_stats->syncp);
 300                         local_stats = *percpu_stats;
 301                 } while (u64_stats_fetch_retry_bh(&percpu_stats->syncp, start));
 302
 303                 stats->rx_bytes         += local_stats.rx_bytes;
 304                 stats->rx_packets       += local_stats.rx_packets;
 305                 stats->tx_bytes         += local_stats.tx_bytes;
 306                 stats->tx_packets       += local_stats.tx_packets;
 307         }
 308 }
 309
 310 /**
 311  *      ovs_vport_get_options - retrieve device options
 312  *
 313  * @vport: vport from which to retrieve the options.
 314  * @skb: sk_buff where options should be appended.
 315  *
 316  * Retrieves the configuration of the given device, appending an
 317  * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
 318  * vport-specific attributes to @skb.
 319  *
 320  * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
 321  * negative error code if a real error occurred.  If an error occurs, @skb is
 322  * left unmodified.
 323  *
 324  * Must be called with ovs_mutex or rcu_read_lock.
 325  */
 326 int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
 327 {
 328         struct nlattr *nla;
 329         int err;
 330
 331         if (!vport->ops->get_options)
 332                 return 0;
 333
 334         nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
 335         if (!nla)
 336                 return -EMSGSIZE;
 337
 338         err = vport->ops->get_options(vport, skb);
 339         if (err) {
 340                 nla_nest_cancel(skb, nla);
 341                 return err;
 342         }
 343
 344         nla_nest_end(skb, nla);
 345         return 0;
 346 }
 347
 348 /**
 349  *      ovs_vport_receive - pass up received packet to the datapath for processing
 350  *
 351  * @vport: vport that received the packet
 352  * @skb: skb that was received
 353  * @tun_key: tunnel (if any) that carried packet
 354  *
 355  * Must be called with rcu_read_lock.  The packet cannot be shared and
 356  * skb->data should point to the Ethernet header.  The caller must have already
 357  * called compute_ip_summed() to initialize the checksumming fields.
 358  */
 359 void ovs_vport_receive(struct vport *vport, struct sk_buff *skb,
 360                        struct ovs_key_ipv4_tunnel *tun_key)
 361 {
 362         struct pcpu_tstats *stats;
 363
 364         stats = this_cpu_ptr(vport->percpu_stats);
 365         u64_stats_update_begin(&stats->syncp);
 366         stats->rx_packets++;
 367         stats->rx_bytes += skb->len;
 368         u64_stats_update_end(&stats->syncp);
 369
 370         OVS_CB(skb)->tun_key = tun_key;
 371         ovs_dp_process_received_packet(vport, skb);
 372 }
 373
 374 /**
 375  *      ovs_vport_send - send a packet on a device
 376  *
 377  * @vport: vport on which to send the packet
 378  * @skb: skb to send
 379  *
 380  * Sends the given packet and returns the length of data sent.  Either ovs
 381  * lock or rcu_read_lock must be held.
 382  */
 383 int ovs_vport_send(struct vport *vport, struct sk_buff *skb)
 384 {
 385         int sent = vport->ops->send(vport, skb);
 386
 387         if (likely(sent > 0)) {
 388                 struct pcpu_tstats *stats;
 389
 390                 stats = this_cpu_ptr(vport->percpu_stats);
 391
 392                 u64_stats_update_begin(&stats->syncp);
 393                 stats->tx_packets++;
 394                 stats->tx_bytes += sent;
 395                 u64_stats_update_end(&stats->syncp);
 396         } else if (sent < 0) {
 397                 ovs_vport_record_error(vport, VPORT_E_TX_ERROR);
 398                 kfree_skb(skb);
 399         } else
 400                 ovs_vport_record_error(vport, VPORT_E_TX_DROPPED);
 401
 402         return sent;
 403 }
 404
 405 /**
 406  *      ovs_vport_record_error - indicate device error to generic stats layer
 407  *
 408  * @vport: vport that encountered the error
 409  * @err_type: one of enum vport_err_type types to indicate the error type
 410  *
 411  * If using the vport generic stats layer indicate that an error of the given
 412  * type has occurred.
 413  */
 414 void ovs_vport_record_error(struct vport *vport, enum vport_err_type err_type)
 415 {
 416         spin_lock(&vport->stats_lock);
 417
 418         switch (err_type) {
 419         case VPORT_E_RX_DROPPED:
 420                 vport->err_stats.rx_dropped++;
 421                 break;
 422
 423         case VPORT_E_RX_ERROR:
 424                 vport->err_stats.rx_errors++;
 425                 break;
 426
 427         case VPORT_E_TX_DROPPED:
 428                 vport->err_stats.tx_dropped++;
 429                 break;
 430
 431         case VPORT_E_TX_ERROR:
 432                 vport->err_stats.tx_errors++;
 433                 break;
 434         }
 435
 436         spin_unlock(&vport->stats_lock);
 437 }
 438
 439 static void free_vport_rcu(struct rcu_head *rcu)
 440 {
 441         struct vport *vport = container_of(rcu, struct vport, rcu);
 442
 443         ovs_vport_free(vport);
 444 }
 445
 446 void ovs_vport_deferred_free(struct vport *vport)
 447 {
 448         if (!vport)
 449                 return;
 450
 451         call_rcu(&vport->rcu, free_vport_rcu);
 452 }