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, ovs_mutex 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 ovs 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 pcpu_tstats);
 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).  ovs_mutex 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         for (i = 0; i < ARRAY_SIZE(vport_ops_list); i++) {
 173                 if (vport_ops_list[i]->type == parms->type) {
 174                         struct hlist_head *bucket;
 175
 176                         vport = vport_ops_list[i]->create(parms);
 177                         if (IS_ERR(vport)) {
 178                                 err = PTR_ERR(vport);
 179                                 goto out;
 180                         }
 181
 182                         bucket = hash_bucket(ovs_dp_get_net(vport->dp),
 183                                              vport->ops->get_name(vport));
 184                         hlist_add_head_rcu(&vport->hash_node, bucket);
 185                         return vport;
 186                 }
 187         }
 188
 189         err = -EAFNOSUPPORT;
 190
 191 out:
 192         return ERR_PTR(err);
 193 }
 194
 195 /**
 196  *      ovs_vport_set_options - modify existing vport device (for kernel callers)
 197  *
 198  * @vport: vport to modify.
 199  * @port: New configuration.
 200  *
 201  * Modifies an existing device with the specified configuration (which is
 202  * dependent on device type).  ovs_mutex must be held.
 203  */
 204 int ovs_vport_set_options(struct vport *vport, struct nlattr *options)
 205 {
 206         if (!vport->ops->set_options)
 207                 return -EOPNOTSUPP;
 208         return vport->ops->set_options(vport, options);
 209 }
 210
 211 /**
 212  *      ovs_vport_del - delete existing vport device
 213  *
 214  * @vport: vport to delete.
 215  *
 216  * Detaches @vport from its datapath and destroys it.  It is possible to fail
 217  * for reasons such as lack of memory.  ovs_mutex must be held.
 218  */
 219 void ovs_vport_del(struct vport *vport)
 220 {
 221         ASSERT_OVSL();
 222
 223         hlist_del_rcu(&vport->hash_node);
 224
 225         vport->ops->destroy(vport);
 226 }
 227
 228 /**
 229  *      ovs_vport_get_stats - retrieve device stats
 230  *
 231  * @vport: vport from which to retrieve the stats
 232  * @stats: location to store stats
 233  *
 234  * Retrieves transmit, receive, and error stats for the given device.
 235  *
 236  * Must be called with ovs_mutex or rcu_read_lock.
 237  */
 238 void ovs_vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
 239 {
 240         int i;
 241
 242         memset(stats, 0, sizeof(*stats));
 243
 244         /* We potentially have 2 sources of stats that need to be combined:
 245          * those we have collected (split into err_stats and percpu_stats) from
 246          * set_stats() and device error stats from netdev->get_stats() (for
 247          * errors that happen  downstream and therefore aren't reported through
 248          * our vport_record_error() function).
 249          * Stats from first source are reported by ovs (OVS_VPORT_ATTR_STATS).
 250          * netdev-stats can be directly read over netlink-ioctl.
 251          */
 252
 253         spin_lock_bh(&vport->stats_lock);
 254
 255         stats->rx_errors        = vport->err_stats.rx_errors;
 256         stats->tx_errors        = vport->err_stats.tx_errors;
 257         stats->tx_dropped       = vport->err_stats.tx_dropped;
 258         stats->rx_dropped       = vport->err_stats.rx_dropped;
 259
 260         spin_unlock_bh(&vport->stats_lock);
 261
 262         for_each_possible_cpu(i) {
 263                 const struct pcpu_tstats *percpu_stats;
 264                 struct pcpu_tstats local_stats;
 265                 unsigned int start;
 266
 267                 percpu_stats = per_cpu_ptr(vport->percpu_stats, i);
 268
 269                 do {
 270                         start = u64_stats_fetch_begin_bh(&percpu_stats->syncp);
 271                         local_stats = *percpu_stats;
 272                 } while (u64_stats_fetch_retry_bh(&percpu_stats->syncp, start));
 273
 274                 stats->rx_bytes         += local_stats.rx_bytes;
 275                 stats->rx_packets       += local_stats.rx_packets;
 276                 stats->tx_bytes         += local_stats.tx_bytes;
 277                 stats->tx_packets       += local_stats.tx_packets;
 278         }
 279 }
 280
 281 /**
 282  *      ovs_vport_get_options - retrieve device options
 283  *
 284  * @vport: vport from which to retrieve the options.
 285  * @skb: sk_buff where options should be appended.
 286  *
 287  * Retrieves the configuration of the given device, appending an
 288  * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
 289  * vport-specific attributes to @skb.
 290  *
 291  * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
 292  * negative error code if a real error occurred.  If an error occurs, @skb is
 293  * left unmodified.
 294  *
 295  * Must be called with ovs_mutex or rcu_read_lock.
 296  */
 297 int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
 298 {
 299         struct nlattr *nla;
 300         int err;
 301
 302         if (!vport->ops->get_options)
 303                 return 0;
 304
 305         nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
 306         if (!nla)
 307                 return -EMSGSIZE;
 308
 309         err = vport->ops->get_options(vport, skb);
 310         if (err) {
 311                 nla_nest_cancel(skb, nla);
 312                 return err;
 313         }
 314
 315         nla_nest_end(skb, nla);
 316         return 0;
 317 }
 318
 319 /**
 320  *      ovs_vport_receive - pass up received packet to the datapath for processing
 321  *
 322  * @vport: vport that received the packet
 323  * @skb: skb that was received
 324  *
 325  * Must be called with rcu_read_lock.  The packet cannot be shared and
 326  * skb->data should point to the Ethernet header.
 327  */
 328 void ovs_vport_receive(struct vport *vport, struct sk_buff *skb)
 329 {
 330         struct pcpu_tstats *stats;
 331
 332         stats = this_cpu_ptr(vport->percpu_stats);
 333         u64_stats_update_begin(&stats->syncp);
 334         stats->rx_packets++;
 335         stats->rx_bytes += skb->len;
 336         u64_stats_update_end(&stats->syncp);
 337
 338         ovs_dp_process_received_packet(vport, skb);
 339 }
 340
 341 /**
 342  *      ovs_vport_send - send a packet on a device
 343  *
 344  * @vport: vport on which to send the packet
 345  * @skb: skb to send
 346  *
 347  * Sends the given packet and returns the length of data sent.  Either ovs
 348  * lock or rcu_read_lock must be held.
 349  */
 350 int ovs_vport_send(struct vport *vport, struct sk_buff *skb)
 351 {
 352         int sent = vport->ops->send(vport, skb);
 353
 354         if (likely(sent)) {
 355                 struct pcpu_tstats *stats;
 356
 357                 stats = this_cpu_ptr(vport->percpu_stats);
 358
 359                 u64_stats_update_begin(&stats->syncp);
 360                 stats->tx_packets++;
 361                 stats->tx_bytes += sent;
 362                 u64_stats_update_end(&stats->syncp);
 363         }
 364         return sent;
 365 }
 366
 367 /**
 368  *      ovs_vport_record_error - indicate device error to generic stats layer
 369  *
 370  * @vport: vport that encountered the error
 371  * @err_type: one of enum vport_err_type types to indicate the error type
 372  *
 373  * If using the vport generic stats layer indicate that an error of the given
 374  * type has occured.
 375  */
 376 void ovs_vport_record_error(struct vport *vport, enum vport_err_type err_type)
 377 {
 378         spin_lock(&vport->stats_lock);
 379
 380         switch (err_type) {
 381         case VPORT_E_RX_DROPPED:
 382                 vport->err_stats.rx_dropped++;
 383                 break;
 384
 385         case VPORT_E_RX_ERROR:
 386                 vport->err_stats.rx_errors++;
 387                 break;
 388
 389         case VPORT_E_TX_DROPPED:
 390                 vport->err_stats.tx_dropped++;
 391                 break;
 392
 393         case VPORT_E_TX_ERROR:
 394                 vport->err_stats.tx_errors++;
 395                 break;
 396         }
 397
 398         spin_unlock(&vport->stats_lock);
 399 }