2 * Copyright (c) 2007-2012 Nicira, Inc.
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.
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.
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
19 #include <linux/etherdevice.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 <linux/version.h>
31 #include <net/net_namespace.h>
35 #include "vport-internal_dev.h"
37 /* List of statically compiled vport implementations. Don't forget to also
38 * add yours to the list at the bottom of vport.h. */
39 static const struct vport_ops
*base_vport_ops_list
[] = {
40 &ovs_netdev_vport_ops
,
41 &ovs_internal_vport_ops
,
45 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,26)
46 &ovs_capwap_vport_ops
,
50 static const struct vport_ops
**vport_ops_list
;
51 static int n_vport_types
;
53 /* Protected by RCU read lock for reading, RTNL lock for writing. */
54 static struct hlist_head
*dev_table
;
55 #define VPORT_HASH_BUCKETS 1024
58 * ovs_vport_init - initialize vport subsystem
60 * Called at module load time to initialize the vport subsystem and any
61 * compiled in vport types.
63 int ovs_vport_init(void)
68 dev_table
= kzalloc(VPORT_HASH_BUCKETS
* sizeof(struct hlist_head
),
75 vport_ops_list
= kmalloc(ARRAY_SIZE(base_vport_ops_list
) *
76 sizeof(struct vport_ops
*), GFP_KERNEL
);
77 if (!vport_ops_list
) {
82 for (i
= 0; i
< ARRAY_SIZE(base_vport_ops_list
); i
++) {
83 const struct vport_ops
*new_ops
= base_vport_ops_list
[i
];
86 err
= new_ops
->init();
91 vport_ops_list
[n_vport_types
++] = new_ops
;
92 else if (new_ops
->flags
& VPORT_F_REQUIRED
) {
107 * ovs_vport_exit - shutdown vport subsystem
109 * Called at module exit time to shutdown the vport subsystem and any
110 * initialized vport types.
112 void ovs_vport_exit(void)
116 for (i
= 0; i
< n_vport_types
; i
++) {
117 if (vport_ops_list
[i
]->exit
)
118 vport_ops_list
[i
]->exit();
121 kfree(vport_ops_list
);
125 static struct hlist_head
*hash_bucket(struct net
*net
, const char *name
)
127 unsigned int hash
= jhash(name
, strlen(name
), (unsigned long) net
);
128 return &dev_table
[hash
& (VPORT_HASH_BUCKETS
- 1)];
132 * ovs_vport_locate - find a port that has already been created
134 * @name: name of port to find
136 * Must be called with RTNL or RCU read lock.
138 struct vport
*ovs_vport_locate(struct net
*net
, const char *name
)
140 struct hlist_head
*bucket
= hash_bucket(net
, name
);
142 struct hlist_node
*node
;
144 hlist_for_each_entry_rcu(vport
, node
, bucket
, hash_node
)
145 if (!strcmp(name
, vport
->ops
->get_name(vport
)) &&
146 net_eq(ovs_dp_get_net(vport
->dp
), net
))
152 static void release_vport(struct kobject
*kobj
)
154 struct vport
*p
= container_of(kobj
, struct vport
, kobj
);
158 static struct kobj_type brport_ktype
= {
160 .sysfs_ops
= &ovs_brport_sysfs_ops
,
162 .release
= release_vport
166 * ovs_vport_alloc - allocate and initialize new vport
168 * @priv_size: Size of private data area to allocate.
169 * @ops: vport device ops
171 * Allocate and initialize a new vport defined by @ops. The vport will contain
172 * a private data area of size @priv_size that can be accessed using
173 * vport_priv(). vports that are no longer needed should be released with
176 struct vport
*ovs_vport_alloc(int priv_size
, const struct vport_ops
*ops
,
177 const struct vport_parms
*parms
)
182 alloc_size
= sizeof(struct vport
);
184 alloc_size
= ALIGN(alloc_size
, VPORT_ALIGN
);
185 alloc_size
+= priv_size
;
188 vport
= kzalloc(alloc_size
, GFP_KERNEL
);
190 return ERR_PTR(-ENOMEM
);
192 vport
->dp
= parms
->dp
;
193 vport
->port_no
= parms
->port_no
;
194 vport
->upcall_pid
= parms
->upcall_pid
;
196 INIT_HLIST_NODE(&vport
->dp_hash_node
);
198 /* Initialize kobject for bridge. This will be added as
199 * /sys/class/net/<devname>/brport later, if sysfs is enabled. */
200 vport
->kobj
.kset
= NULL
;
201 kobject_init(&vport
->kobj
, &brport_ktype
);
203 vport
->percpu_stats
= alloc_percpu(struct vport_percpu_stats
);
204 if (!vport
->percpu_stats
) {
206 return ERR_PTR(-ENOMEM
);
209 spin_lock_init(&vport
->stats_lock
);
215 * ovs_vport_free - uninitialize and free vport
217 * @vport: vport to free
219 * Frees a vport allocated with ovs_vport_alloc() when it is no longer needed.
221 * The caller must ensure that an RCU grace period has passed since the last
222 * time @vport was in a datapath.
224 void ovs_vport_free(struct vport
*vport
)
226 free_percpu(vport
->percpu_stats
);
228 kobject_put(&vport
->kobj
);
232 * ovs_vport_add - add vport device (for kernel callers)
234 * @parms: Information about new vport.
236 * Creates a new vport with the specified configuration (which is dependent on
237 * device type). RTNL lock must be held.
239 struct vport
*ovs_vport_add(const struct vport_parms
*parms
)
247 for (i
= 0; i
< n_vport_types
; i
++) {
248 if (vport_ops_list
[i
]->type
== parms
->type
) {
249 struct hlist_head
*bucket
;
251 vport
= vport_ops_list
[i
]->create(parms
);
253 err
= PTR_ERR(vport
);
257 bucket
= hash_bucket(ovs_dp_get_net(vport
->dp
),
258 vport
->ops
->get_name(vport
));
259 hlist_add_head_rcu(&vport
->hash_node
, bucket
);
271 * ovs_vport_set_options - modify existing vport device (for kernel callers)
273 * @vport: vport to modify.
274 * @port: New configuration.
276 * Modifies an existing device with the specified configuration (which is
277 * dependent on device type). RTNL lock must be held.
279 int ovs_vport_set_options(struct vport
*vport
, struct nlattr
*options
)
283 if (!vport
->ops
->set_options
)
285 return vport
->ops
->set_options(vport
, options
);
289 * ovs_vport_del - delete existing vport device
291 * @vport: vport to delete.
293 * Detaches @vport from its datapath and destroys it. It is possible to fail
294 * for reasons such as lack of memory. RTNL lock must be held.
296 void ovs_vport_del(struct vport
*vport
)
300 hlist_del_rcu(&vport
->hash_node
);
302 vport
->ops
->destroy(vport
);
306 * ovs_vport_set_addr - set device Ethernet address (for kernel callers)
308 * @vport: vport on which to set Ethernet address.
309 * @addr: New address.
311 * Sets the Ethernet address of the given device. Some devices may not support
312 * setting the Ethernet address, in which case the result will always be
313 * -EOPNOTSUPP. RTNL lock must be held.
315 int ovs_vport_set_addr(struct vport
*vport
, const unsigned char *addr
)
319 if (!is_valid_ether_addr(addr
))
320 return -EADDRNOTAVAIL
;
322 if (vport
->ops
->set_addr
)
323 return vport
->ops
->set_addr(vport
, addr
);
329 * ovs_vport_set_stats - sets offset device stats
331 * @vport: vport on which to set stats
332 * @stats: stats to set
334 * Provides a set of transmit, receive, and error stats to be added as an
335 * offset to the collect data when stats are retreived. Some devices may not
336 * support setting the stats, in which case the result will always be
339 * Must be called with RTNL lock.
341 void ovs_vport_set_stats(struct vport
*vport
, struct ovs_vport_stats
*stats
)
345 spin_lock_bh(&vport
->stats_lock
);
346 vport
->offset_stats
= *stats
;
347 spin_unlock_bh(&vport
->stats_lock
);
351 * ovs_vport_get_stats - retrieve device stats
353 * @vport: vport from which to retrieve the stats
354 * @stats: location to store stats
356 * Retrieves transmit, receive, and error stats for the given device.
358 * Must be called with RTNL lock or rcu_read_lock.
360 void ovs_vport_get_stats(struct vport
*vport
, struct ovs_vport_stats
*stats
)
364 /* We potentially have 3 sources of stats that need to be
365 * combined: those we have collected (split into err_stats and
366 * percpu_stats), offset_stats from set_stats(), and device
367 * error stats from netdev->get_stats() (for errors that happen
368 * downstream and therefore aren't reported through our
369 * vport_record_error() function).
370 * Stats from first two sources are merged and reported by ovs over
371 * OVS_VPORT_ATTR_STATS.
372 * netdev-stats can be directly read over netlink-ioctl.
375 spin_lock_bh(&vport
->stats_lock
);
377 *stats
= vport
->offset_stats
;
379 stats
->rx_errors
+= vport
->err_stats
.rx_errors
;
380 stats
->tx_errors
+= vport
->err_stats
.tx_errors
;
381 stats
->tx_dropped
+= vport
->err_stats
.tx_dropped
;
382 stats
->rx_dropped
+= vport
->err_stats
.rx_dropped
;
384 spin_unlock_bh(&vport
->stats_lock
);
386 for_each_possible_cpu(i
) {
387 const struct vport_percpu_stats
*percpu_stats
;
388 struct vport_percpu_stats local_stats
;
391 percpu_stats
= per_cpu_ptr(vport
->percpu_stats
, i
);
394 start
= u64_stats_fetch_begin_bh(&percpu_stats
->sync
);
395 local_stats
= *percpu_stats
;
396 } while (u64_stats_fetch_retry_bh(&percpu_stats
->sync
, start
));
398 stats
->rx_bytes
+= local_stats
.rx_bytes
;
399 stats
->rx_packets
+= local_stats
.rx_packets
;
400 stats
->tx_bytes
+= local_stats
.tx_bytes
;
401 stats
->tx_packets
+= local_stats
.tx_packets
;
406 * ovs_vport_get_options - retrieve device options
408 * @vport: vport from which to retrieve the options.
409 * @skb: sk_buff where options should be appended.
411 * Retrieves the configuration of the given device, appending an
412 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
413 * vport-specific attributes to @skb.
415 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
416 * negative error code if a real error occurred. If an error occurs, @skb is
419 * Must be called with RTNL lock or rcu_read_lock.
421 int ovs_vport_get_options(const struct vport
*vport
, struct sk_buff
*skb
)
425 nla
= nla_nest_start(skb
, OVS_VPORT_ATTR_OPTIONS
);
429 if (vport
->ops
->get_options
) {
430 int err
= vport
->ops
->get_options(vport
, skb
);
432 nla_nest_cancel(skb
, nla
);
437 nla_nest_end(skb
, nla
);
442 * ovs_vport_receive - pass up received packet to the datapath for processing
444 * @vport: vport that received the packet
445 * @skb: skb that was received
447 * Must be called with rcu_read_lock. The packet cannot be shared and
448 * skb->data should point to the Ethernet header. The caller must have already
449 * called compute_ip_summed() to initialize the checksumming fields.
451 void ovs_vport_receive(struct vport
*vport
, struct sk_buff
*skb
)
453 struct vport_percpu_stats
*stats
;
455 stats
= per_cpu_ptr(vport
->percpu_stats
, smp_processor_id());
457 u64_stats_update_begin(&stats
->sync
);
459 stats
->rx_bytes
+= skb
->len
;
460 u64_stats_update_end(&stats
->sync
);
462 if (!(vport
->ops
->flags
& VPORT_F_FLOW
))
463 OVS_CB(skb
)->flow
= NULL
;
465 if (!(vport
->ops
->flags
& VPORT_F_TUN_ID
))
466 OVS_CB(skb
)->tun_id
= 0;
468 ovs_dp_process_received_packet(vport
, skb
);
472 * ovs_vport_send - send a packet on a device
474 * @vport: vport on which to send the packet
477 * Sends the given packet and returns the length of data sent. Either RTNL
478 * lock or rcu_read_lock must be held.
480 int ovs_vport_send(struct vport
*vport
, struct sk_buff
*skb
)
482 int sent
= vport
->ops
->send(vport
, skb
);
485 struct vport_percpu_stats
*stats
;
487 stats
= per_cpu_ptr(vport
->percpu_stats
, smp_processor_id());
489 u64_stats_update_begin(&stats
->sync
);
491 stats
->tx_bytes
+= sent
;
492 u64_stats_update_end(&stats
->sync
);
498 * ovs_vport_record_error - indicate device error to generic stats layer
500 * @vport: vport that encountered the error
501 * @err_type: one of enum vport_err_type types to indicate the error type
503 * If using the vport generic stats layer indicate that an error of the given
506 void ovs_vport_record_error(struct vport
*vport
, enum vport_err_type err_type
)
508 spin_lock(&vport
->stats_lock
);
511 case VPORT_E_RX_DROPPED
:
512 vport
->err_stats
.rx_dropped
++;
515 case VPORT_E_RX_ERROR
:
516 vport
->err_stats
.rx_errors
++;
519 case VPORT_E_TX_DROPPED
:
520 vport
->err_stats
.tx_dropped
++;
523 case VPORT_E_TX_ERROR
:
524 vport
->err_stats
.tx_errors
++;
528 spin_unlock(&vport
->stats_lock
);