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
  43 /* Protected by RCU read lock for reading, RTNL lock for writing. */
  44 static struct hlist_head *dev_table;
  45 #define VPORT_HASH_BUCKETS 1024
  46
  47 /**
  48  *      ovs_vport_init - initialize vport subsystem
  49  *
  50  * Called at module load time to initialize the vport subsystem.
  51  */
  52 int ovs_vport_init(void)
  53 {
  54         dev_table = kzalloc(VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
  55                             GFP_KERNEL);
  56         if (!dev_table)
  57                 return -ENOMEM;
  58
  59         return 0;
  60 }
  61
  62 /**
  63  *      ovs_vport_exit - shutdown vport subsystem
  64  *
  65  * Called at module exit time to shutdown the vport subsystem.
  66  */
  67 void ovs_vport_exit(void)
  68 {
  69         kfree(dev_table);
  70 }
  71
  72 static struct hlist_head *hash_bucket(struct net *net, const char *name)
  73 {
  74         unsigned int hash = jhash(name, strlen(name), (unsigned long) net);
  75         return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
  76 }
  77
  78 /**
  79  *      ovs_vport_locate - find a port that has already been created
  80  *
  81  * @name: name of port to find
  82  *
  83  * Must be called with RTNL or RCU read lock.
  84  */
  85 struct vport *ovs_vport_locate(struct net *net, const char *name)
  86 {
  87         struct hlist_head *bucket = hash_bucket(net, name);
  88         struct vport *vport;
  89
  90         hlist_for_each_entry_rcu(vport, bucket, hash_node)
  91                 if (!strcmp(name, vport->ops->get_name(vport)) &&
  92                     net_eq(ovs_dp_get_net(vport->dp), net))
  93                         return vport;
  94
  95         return NULL;
  96 }
  97
  98 /**
  99  *      ovs_vport_alloc - allocate and initialize new vport
 100  *
 101  * @priv_size: Size of private data area to allocate.
 102  * @ops: vport device ops
 103  *
 104  * Allocate and initialize a new vport defined by @ops.  The vport will contain
 105  * a private data area of size @priv_size that can be accessed using
 106  * vport_priv().  vports that are no longer needed should be released with
 107  * vport_free().
 108  */
 109 struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
 110                           const struct vport_parms *parms)
 111 {
 112         struct vport *vport;
 113         size_t alloc_size;
 114
 115         alloc_size = sizeof(struct vport);
 116         if (priv_size) {
 117                 alloc_size = ALIGN(alloc_size, VPORT_ALIGN);
 118                 alloc_size += priv_size;
 119         }
 120
 121         vport = kzalloc(alloc_size, GFP_KERNEL);
 122         if (!vport)
 123                 return ERR_PTR(-ENOMEM);
 124
 125         vport->dp = parms->dp;
 126         vport->port_no = parms->port_no;
 127         vport->upcall_portid = parms->upcall_portid;
 128         vport->ops = ops;
 129         INIT_HLIST_NODE(&vport->dp_hash_node);
 130
 131         vport->percpu_stats = alloc_percpu(struct vport_percpu_stats);
 132         if (!vport->percpu_stats) {
 133                 kfree(vport);
 134                 return ERR_PTR(-ENOMEM);
 135         }
 136
 137         spin_lock_init(&vport->stats_lock);
 138
 139         return vport;
 140 }
 141
 142 /**
 143  *      ovs_vport_free - uninitialize and free vport
 144  *
 145  * @vport: vport to free
 146  *
 147  * Frees a vport allocated with vport_alloc() when it is no longer needed.
 148  *
 149  * The caller must ensure that an RCU grace period has passed since the last
 150  * time @vport was in a datapath.
 151  */
 152 void ovs_vport_free(struct vport *vport)
 153 {
 154         free_percpu(vport->percpu_stats);
 155         kfree(vport);
 156 }
 157
 158 /**
 159  *      ovs_vport_add - add vport device (for kernel callers)
 160  *
 161  * @parms: Information about new vport.
 162  *
 163  * Creates a new vport with the specified configuration (which is dependent on
 164  * device type).  RTNL lock must be held.
 165  */
 166 struct vport *ovs_vport_add(const struct vport_parms *parms)
 167 {
 168         struct vport *vport;
 169         int err = 0;
 170         int i;
 171
 172         ASSERT_RTNL();
 173
 174         for (i = 0; i < ARRAY_SIZE(vport_ops_list); i++) {
 175                 if (vport_ops_list[i]->type == parms->type) {
 176                         struct hlist_head *bucket;
 177
 178                         vport = vport_ops_list[i]->create(parms);
 179                         if (IS_ERR(vport)) {
 180                                 err = PTR_ERR(vport);
 181                                 goto out;
 182                         }
 183
 184                         bucket = hash_bucket(ovs_dp_get_net(vport->dp),
 185                                              vport->ops->get_name(vport));
 186                         hlist_add_head_rcu(&vport->hash_node, bucket);
 187                         return vport;
 188                 }
 189         }
 190
 191         err = -EAFNOSUPPORT;
 192
 193 out:
 194         return ERR_PTR(err);
 195 }
 196
 197 /**
 198  *      ovs_vport_set_options - modify existing vport device (for kernel callers)
 199  *
 200  * @vport: vport to modify.
 201  * @port: New configuration.
 202  *
 203  * Modifies an existing device with the specified configuration (which is
 204  * dependent on device type).  RTNL lock must be held.
 205  */
 206 int ovs_vport_set_options(struct vport *vport, struct nlattr *options)
 207 {
 208         ASSERT_RTNL();
 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.  RTNL lock must be held.
 222  */
 223 void ovs_vport_del(struct vport *vport)
 224 {
 225         ASSERT_RTNL();
 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 RTNL lock 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 vport_percpu_stats *percpu_stats;
 268                 struct vport_percpu_stats 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->sync);
 275                         local_stats = *percpu_stats;
 276                 } while (u64_stats_fetch_retry_bh(&percpu_stats->sync, 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 RTNL lock 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
 305         nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
 306         if (!nla)
 307                 return -EMSGSIZE;
 308
 309         if (vport->ops->get_options) {
 310                 int err = vport->ops->get_options(vport, skb);
 311                 if (err) {
 312                         nla_nest_cancel(skb, nla);
 313                         return err;
 314                 }
 315         }
 316
 317         nla_nest_end(skb, nla);
 318         return 0;
 319 }
 320
 321 /**
 322  *      ovs_vport_receive - pass up received packet to the datapath for processing
 323  *
 324  * @vport: vport that received the packet
 325  * @skb: skb that was received
 326  *
 327  * Must be called with rcu_read_lock.  The packet cannot be shared and
 328  * skb->data should point to the Ethernet header.
 329  */
 330 void ovs_vport_receive(struct vport *vport, struct sk_buff *skb)
 331 {
 332         struct vport_percpu_stats *stats;
 333
 334         stats = this_cpu_ptr(vport->percpu_stats);
 335         u64_stats_update_begin(&stats->sync);
 336         stats->rx_packets++;
 337         stats->rx_bytes += skb->len;
 338         u64_stats_update_end(&stats->sync);
 339
 340         ovs_dp_process_received_packet(vport, skb);
 341 }
 342
 343 /**
 344  *      ovs_vport_send - send a packet on a device
 345  *
 346  * @vport: vport on which to send the packet
 347  * @skb: skb to send
 348  *
 349  * Sends the given packet and returns the length of data sent.  Either RTNL
 350  * lock or rcu_read_lock must be held.
 351  */
 352 int ovs_vport_send(struct vport *vport, struct sk_buff *skb)
 353 {
 354         int sent = vport->ops->send(vport, skb);
 355
 356         if (likely(sent)) {
 357                 struct vport_percpu_stats *stats;
 358
 359                 stats = this_cpu_ptr(vport->percpu_stats);
 360
 361                 u64_stats_update_begin(&stats->sync);
 362                 stats->tx_packets++;
 363                 stats->tx_bytes += sent;
 364                 u64_stats_update_end(&stats->sync);
 365         }
 366         return sent;
 367 }
 368
 369 /**
 370  *      ovs_vport_record_error - indicate device error to generic stats layer
 371  *
 372  * @vport: vport that encountered the error
 373  * @err_type: one of enum vport_err_type types to indicate the error type
 374  *
 375  * If using the vport generic stats layer indicate that an error of the given
 376  * type has occured.
 377  */
 378 void ovs_vport_record_error(struct vport *vport, enum vport_err_type err_type)
 379 {
 380         spin_lock(&vport->stats_lock);
 381
 382         switch (err_type) {
 383         case VPORT_E_RX_DROPPED:
 384                 vport->err_stats.rx_dropped++;
 385                 break;
 386
 387         case VPORT_E_RX_ERROR:
 388                 vport->err_stats.rx_errors++;
 389                 break;
 390
 391         case VPORT_E_TX_DROPPED:
 392                 vport->err_stats.tx_dropped++;
 393                 break;
 394
 395         case VPORT_E_TX_ERROR:
 396                 vport->err_stats.tx_errors++;
 397                 break;
 398         }
 399
 400         spin_unlock(&vport->stats_lock);
 401 }