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/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>
36 #include "vport-internal_dev.h"
38 static void ovs_vport_record_error(struct vport
*,
39 enum vport_err_type err_type
);
41 /* List of statically compiled vport implementations. Don't forget to also
42 * add yours to the list at the bottom of vport.h. */
43 static const struct vport_ops
*vport_ops_list
[] = {
44 &ovs_netdev_vport_ops
,
45 &ovs_internal_vport_ops
,
46 #if IS_ENABLED(CONFIG_NET_IPGRE_DEMUX)
54 /* Protected by RCU read lock for reading, ovs_mutex for writing. */
55 static struct hlist_head
*dev_table
;
56 #define VPORT_HASH_BUCKETS 1024
59 * ovs_vport_init - initialize vport subsystem
61 * Called at module load time to initialize the vport subsystem.
63 int ovs_vport_init(void)
65 dev_table
= kzalloc(VPORT_HASH_BUCKETS
* sizeof(struct hlist_head
),
74 * ovs_vport_exit - shutdown vport subsystem
76 * Called at module exit time to shutdown the vport subsystem.
78 void ovs_vport_exit(void)
83 static struct hlist_head
*hash_bucket(struct net
*net
, const char *name
)
85 unsigned int hash
= jhash(name
, strlen(name
), (unsigned long) net
);
86 return &dev_table
[hash
& (VPORT_HASH_BUCKETS
- 1)];
90 * ovs_vport_locate - find a port that has already been created
92 * @name: name of port to find
94 * Must be called with ovs or RCU read lock.
96 struct vport
*ovs_vport_locate(struct net
*net
, const char *name
)
98 struct hlist_head
*bucket
= hash_bucket(net
, name
);
101 hlist_for_each_entry_rcu(vport
, bucket
, hash_node
)
102 if (!strcmp(name
, vport
->ops
->get_name(vport
)) &&
103 net_eq(ovs_dp_get_net(vport
->dp
), net
))
110 * ovs_vport_alloc - allocate and initialize new vport
112 * @priv_size: Size of private data area to allocate.
113 * @ops: vport device ops
115 * Allocate and initialize a new vport defined by @ops. The vport will contain
116 * a private data area of size @priv_size that can be accessed using
117 * vport_priv(). vports that are no longer needed should be released with
120 struct vport
*ovs_vport_alloc(int priv_size
, const struct vport_ops
*ops
,
121 const struct vport_parms
*parms
)
127 alloc_size
= sizeof(struct vport
);
129 alloc_size
= ALIGN(alloc_size
, VPORT_ALIGN
);
130 alloc_size
+= priv_size
;
133 vport
= kzalloc(alloc_size
, GFP_KERNEL
);
135 return ERR_PTR(-ENOMEM
);
137 vport
->dp
= parms
->dp
;
138 vport
->port_no
= parms
->port_no
;
140 INIT_HLIST_NODE(&vport
->dp_hash_node
);
142 if (ovs_vport_set_upcall_portids(vport
, parms
->upcall_portids
))
143 return ERR_PTR(-EINVAL
);
145 vport
->percpu_stats
= alloc_percpu(struct pcpu_sw_netstats
);
146 if (!vport
->percpu_stats
) {
148 return ERR_PTR(-ENOMEM
);
151 for_each_possible_cpu(i
) {
152 struct pcpu_sw_netstats
*vport_stats
;
153 vport_stats
= per_cpu_ptr(vport
->percpu_stats
, i
);
154 u64_stats_init(&vport_stats
->syncp
);
157 spin_lock_init(&vport
->stats_lock
);
163 * ovs_vport_free - uninitialize and free vport
165 * @vport: vport to free
167 * Frees a vport allocated with ovs_vport_alloc() when it is no longer needed.
169 * The caller must ensure that an RCU grace period has passed since the last
170 * time @vport was in a datapath.
172 void ovs_vport_free(struct vport
*vport
)
174 kfree((struct vport_portids __force
*)vport
->upcall_portids
);
175 free_percpu(vport
->percpu_stats
);
180 * ovs_vport_add - add vport device (for kernel callers)
182 * @parms: Information about new vport.
184 * Creates a new vport with the specified configuration (which is dependent on
185 * device type). ovs_mutex must be held.
187 struct vport
*ovs_vport_add(const struct vport_parms
*parms
)
193 for (i
= 0; i
< ARRAY_SIZE(vport_ops_list
); i
++) {
194 if (vport_ops_list
[i
]->type
== parms
->type
) {
195 struct hlist_head
*bucket
;
197 vport
= vport_ops_list
[i
]->create(parms
);
199 err
= PTR_ERR(vport
);
203 bucket
= hash_bucket(ovs_dp_get_net(vport
->dp
),
204 vport
->ops
->get_name(vport
));
205 hlist_add_head_rcu(&vport
->hash_node
, bucket
);
217 * ovs_vport_set_options - modify existing vport device (for kernel callers)
219 * @vport: vport to modify.
220 * @options: New configuration.
222 * Modifies an existing device with the specified configuration (which is
223 * dependent on device type). ovs_mutex must be held.
225 int ovs_vport_set_options(struct vport
*vport
, struct nlattr
*options
)
227 if (!vport
->ops
->set_options
)
229 return vport
->ops
->set_options(vport
, options
);
233 * ovs_vport_del - delete existing vport device
235 * @vport: vport to delete.
237 * Detaches @vport from its datapath and destroys it. It is possible to fail
238 * for reasons such as lack of memory. ovs_mutex must be held.
240 void ovs_vport_del(struct vport
*vport
)
244 hlist_del_rcu(&vport
->hash_node
);
245 vport
->ops
->destroy(vport
);
249 * ovs_vport_set_stats - sets offset device stats
251 * @vport: vport on which to set stats
252 * @stats: stats to set
254 * Provides a set of transmit, receive, and error stats to be added as an
255 * offset to the collected data when stats are retrieved. Some devices may not
256 * support setting the stats, in which case the result will always be
259 * Must be called with ovs_mutex.
261 void ovs_vport_set_stats(struct vport
*vport
, struct ovs_vport_stats
*stats
)
263 spin_lock_bh(&vport
->stats_lock
);
264 vport
->offset_stats
= *stats
;
265 spin_unlock_bh(&vport
->stats_lock
);
269 * ovs_vport_get_stats - retrieve device stats
271 * @vport: vport from which to retrieve the stats
272 * @stats: location to store stats
274 * Retrieves transmit, receive, and error stats for the given device.
276 * Must be called with ovs_mutex or rcu_read_lock.
278 void ovs_vport_get_stats(struct vport
*vport
, struct ovs_vport_stats
*stats
)
282 /* We potentially have 3 sources of stats that need to be
283 * combined: those we have collected (split into err_stats and
284 * percpu_stats), offset_stats from set_stats(), and device
285 * error stats from netdev->get_stats() (for errors that happen
286 * downstream and therefore aren't reported through our
287 * vport_record_error() function).
288 * Stats from first two sources are merged and reported by ovs over
289 * OVS_VPORT_ATTR_STATS.
290 * netdev-stats can be directly read over netlink-ioctl.
293 spin_lock_bh(&vport
->stats_lock
);
295 *stats
= vport
->offset_stats
;
297 stats
->rx_errors
+= vport
->err_stats
.rx_errors
;
298 stats
->tx_errors
+= vport
->err_stats
.tx_errors
;
299 stats
->tx_dropped
+= vport
->err_stats
.tx_dropped
;
300 stats
->rx_dropped
+= vport
->err_stats
.rx_dropped
;
302 spin_unlock_bh(&vport
->stats_lock
);
304 for_each_possible_cpu(i
) {
305 const struct pcpu_sw_netstats
*percpu_stats
;
306 struct pcpu_sw_netstats local_stats
;
309 percpu_stats
= per_cpu_ptr(vport
->percpu_stats
, i
);
312 start
= u64_stats_fetch_begin_bh(&percpu_stats
->syncp
);
313 local_stats
= *percpu_stats
;
314 } while (u64_stats_fetch_retry_bh(&percpu_stats
->syncp
, start
));
316 stats
->rx_bytes
+= local_stats
.rx_bytes
;
317 stats
->rx_packets
+= local_stats
.rx_packets
;
318 stats
->tx_bytes
+= local_stats
.tx_bytes
;
319 stats
->tx_packets
+= local_stats
.tx_packets
;
324 * ovs_vport_get_options - retrieve device options
326 * @vport: vport from which to retrieve the options.
327 * @skb: sk_buff where options should be appended.
329 * Retrieves the configuration of the given device, appending an
330 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
331 * vport-specific attributes to @skb.
333 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
334 * negative error code if a real error occurred. If an error occurs, @skb is
337 * Must be called with ovs_mutex or rcu_read_lock.
339 int ovs_vport_get_options(const struct vport
*vport
, struct sk_buff
*skb
)
344 if (!vport
->ops
->get_options
)
347 nla
= nla_nest_start(skb
, OVS_VPORT_ATTR_OPTIONS
);
351 err
= vport
->ops
->get_options(vport
, skb
);
353 nla_nest_cancel(skb
, nla
);
357 nla_nest_end(skb
, nla
);
361 static void vport_portids_destroy_rcu_cb(struct rcu_head
*rcu
)
363 struct vport_portids
*ids
= container_of(rcu
, struct vport_portids
,
370 * ovs_vport_set_upcall_portids - set upcall portids of @vport.
372 * @vport: vport to modify.
373 * @ids: new configuration, an array of port ids.
375 * Sets the vport's upcall_portids to @ids.
377 * Returns 0 if successful, -EINVAL if @ids is zero length or cannot be parsed
378 * as an array of U32.
380 * Must be called with ovs_mutex.
382 int ovs_vport_set_upcall_portids(struct vport
*vport
, struct nlattr
*ids
)
384 struct vport_portids
*old
, *vport_portids
;
386 if (!nla_len(ids
) || nla_len(ids
) % sizeof(u32
))
389 old
= ovsl_dereference(vport
->upcall_portids
);
391 vport_portids
= kmalloc(sizeof *vport_portids
+ nla_len(ids
),
393 vport_portids
->n_ids
= nla_len(ids
) / sizeof(u32
);
394 vport_portids
->rn_ids
= reciprocal_value(vport_portids
->n_ids
);
395 nla_memcpy(vport_portids
->ids
, ids
, nla_len(ids
));
397 rcu_assign_pointer(vport
->upcall_portids
, vport_portids
);
400 call_rcu(&old
->rcu
, vport_portids_destroy_rcu_cb
);
406 * ovs_vport_get_upcall_portids - get the upcall_portids of @vport.
408 * @vport: vport from which to retrieve the portids.
409 * @skb: sk_buff where portids should be appended.
411 * Retrieves the configuration of the given vport, appending the
412 * %OVS_VPORT_ATTR_UPCALL_PID attribute which is the array of upcall
415 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room.
416 * If an error occurs, @skb is left unmodified. Must be called with
417 * ovs_mutex or rcu_read_lock.
419 int ovs_vport_get_upcall_portids(const struct vport
*vport
,
422 struct vport_portids
*ids
;
424 ids
= rcu_dereference_ovsl(vport
->upcall_portids
);
426 if (vport
->dp
->user_features
& OVS_DP_F_VPORT_PIDS
)
427 return nla_put(skb
, OVS_VPORT_ATTR_UPCALL_PID
,
428 ids
->n_ids
* sizeof(u32
), (void *) ids
->ids
);
430 return nla_put_u32(skb
, OVS_VPORT_ATTR_UPCALL_PID
, ids
->ids
[0]);
434 * ovs_vport_find_upcall_portid - find the upcall portid to send upcall.
436 * @vport: vport from which the missed packet is received.
437 * @skb: skb that the missed packet was received.
439 * Uses the skb_get_hash() to select the upcall portid to send the
442 * Returns the portid of the target socket. Must be called with rcu_read_lock.
444 u32
ovs_vport_find_upcall_portid(const struct vport
*p
, struct sk_buff
*skb
)
446 struct vport_portids
*ids
;
449 ids
= rcu_dereference(p
->upcall_portids
);
451 if (ids
->n_ids
== 1 && ids
->ids
[0] == 0)
454 hash
= skb_get_hash(skb
);
455 return ids
->ids
[hash
- ids
->n_ids
* reciprocal_divide(hash
, ids
->rn_ids
)];
459 * ovs_vport_receive - pass up received packet to the datapath for processing
461 * @vport: vport that received the packet
462 * @skb: skb that was received
463 * @tun_info: tunnel (if any) that carried packet
465 * Must be called with rcu_read_lock. The packet cannot be shared and
466 * skb->data should point to the Ethernet header. The caller must have already
467 * called compute_ip_summed() to initialize the checksumming fields.
469 void ovs_vport_receive(struct vport
*vport
, struct sk_buff
*skb
,
470 struct ovs_tunnel_info
*tun_info
)
472 struct pcpu_sw_netstats
*stats
;
474 stats
= this_cpu_ptr(vport
->percpu_stats
);
475 u64_stats_update_begin(&stats
->syncp
);
477 stats
->rx_bytes
+= skb
->len
;
478 u64_stats_update_end(&stats
->syncp
);
480 OVS_CB(skb
)->tun_info
= tun_info
;
481 ovs_dp_process_received_packet(vport
, skb
);
485 * ovs_vport_send - send a packet on a device
487 * @vport: vport on which to send the packet
490 * Sends the given packet and returns the length of data sent. Either ovs
491 * lock or rcu_read_lock must be held.
493 int ovs_vport_send(struct vport
*vport
, struct sk_buff
*skb
)
495 int sent
= vport
->ops
->send(vport
, skb
);
497 if (likely(sent
> 0)) {
498 struct pcpu_sw_netstats
*stats
;
500 stats
= this_cpu_ptr(vport
->percpu_stats
);
502 u64_stats_update_begin(&stats
->syncp
);
504 stats
->tx_bytes
+= sent
;
505 u64_stats_update_end(&stats
->syncp
);
506 } else if (sent
< 0) {
507 ovs_vport_record_error(vport
, VPORT_E_TX_ERROR
);
510 ovs_vport_record_error(vport
, VPORT_E_TX_DROPPED
);
516 * ovs_vport_record_error - indicate device error to generic stats layer
518 * @vport: vport that encountered the error
519 * @err_type: one of enum vport_err_type types to indicate the error type
521 * If using the vport generic stats layer indicate that an error of the given
524 static void ovs_vport_record_error(struct vport
*vport
,
525 enum vport_err_type err_type
)
527 spin_lock(&vport
->stats_lock
);
530 case VPORT_E_RX_DROPPED
:
531 vport
->err_stats
.rx_dropped
++;
534 case VPORT_E_RX_ERROR
:
535 vport
->err_stats
.rx_errors
++;
538 case VPORT_E_TX_DROPPED
:
539 vport
->err_stats
.tx_dropped
++;
542 case VPORT_E_TX_ERROR
:
543 vport
->err_stats
.tx_errors
++;
547 spin_unlock(&vport
->stats_lock
);
550 static void free_vport_rcu(struct rcu_head
*rcu
)
552 struct vport
*vport
= container_of(rcu
, struct vport
, rcu
);
554 ovs_vport_free(vport
);
557 void ovs_vport_deferred_free(struct vport
*vport
)
562 call_rcu(&vport
->rcu
, free_vport_rcu
);