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/module.h>
27 #include <linux/mutex.h>
28 #include <linux/percpu.h>
29 #include <linux/rcupdate.h>
30 #include <linux/rtnetlink.h>
31 #include <linux/compat.h>
32 #include <linux/version.h>
33 #include <net/net_namespace.h>
38 #include "vport-internal_dev.h"
40 static void ovs_vport_record_error(struct vport
*,
41 enum vport_err_type err_type
);
43 static LIST_HEAD(vport_ops_list
);
45 /* Protected by RCU read lock for reading, ovs_mutex for writing. */
46 static struct hlist_head
*dev_table
;
47 #define VPORT_HASH_BUCKETS 1024
50 * ovs_vport_init - initialize vport subsystem
52 * Called at module load time to initialize the vport subsystem.
54 int ovs_vport_init(void)
56 dev_table
= kzalloc(VPORT_HASH_BUCKETS
* sizeof(struct hlist_head
),
65 * ovs_vport_exit - shutdown vport subsystem
67 * Called at module exit time to shutdown the vport subsystem.
69 void ovs_vport_exit(void)
74 static struct hlist_head
*hash_bucket(const struct net
*net
, const char *name
)
76 unsigned int hash
= jhash(name
, strlen(name
), (unsigned long) net
);
77 return &dev_table
[hash
& (VPORT_HASH_BUCKETS
- 1)];
80 int ovs_vport_ops_register(struct vport_ops
*ops
)
86 list_for_each_entry(o
, &vport_ops_list
, list
)
87 if (ops
->type
== o
->type
)
90 list_add_tail(&ops
->list
, &vport_ops_list
);
96 EXPORT_SYMBOL_GPL(ovs_vport_ops_register
);
98 void ovs_vport_ops_unregister(struct vport_ops
*ops
)
101 list_del(&ops
->list
);
104 EXPORT_SYMBOL_GPL(ovs_vport_ops_unregister
);
107 * ovs_vport_locate - find a port that has already been created
109 * @name: name of port to find
111 * Must be called with ovs or RCU read lock.
113 struct vport
*ovs_vport_locate(const struct net
*net
, const char *name
)
115 struct hlist_head
*bucket
= hash_bucket(net
, name
);
118 hlist_for_each_entry_rcu(vport
, bucket
, hash_node
)
119 if (!strcmp(name
, vport
->ops
->get_name(vport
)) &&
120 net_eq(ovs_dp_get_net(vport
->dp
), net
))
127 * ovs_vport_alloc - allocate and initialize new vport
129 * @priv_size: Size of private data area to allocate.
130 * @ops: vport device ops
132 * Allocate and initialize a new vport defined by @ops. The vport will contain
133 * a private data area of size @priv_size that can be accessed using
134 * vport_priv(). vports that are no longer needed should be released with
137 struct vport
*ovs_vport_alloc(int priv_size
, const struct vport_ops
*ops
,
138 const struct vport_parms
*parms
)
143 alloc_size
= sizeof(struct vport
);
145 alloc_size
= ALIGN(alloc_size
, VPORT_ALIGN
);
146 alloc_size
+= priv_size
;
149 vport
= kzalloc(alloc_size
, GFP_KERNEL
);
151 return ERR_PTR(-ENOMEM
);
153 vport
->dp
= parms
->dp
;
154 vport
->port_no
= parms
->port_no
;
156 INIT_HLIST_NODE(&vport
->dp_hash_node
);
158 if (ovs_vport_set_upcall_portids(vport
, parms
->upcall_portids
)) {
160 return ERR_PTR(-EINVAL
);
163 vport
->percpu_stats
= netdev_alloc_pcpu_stats(struct pcpu_sw_netstats
);
164 if (!vport
->percpu_stats
) {
166 return ERR_PTR(-ENOMEM
);
171 EXPORT_SYMBOL_GPL(ovs_vport_alloc
);
173 static struct vport_ops
*ovs_vport_lookup(const struct vport_parms
*parms
)
175 struct vport_ops
*ops
;
177 list_for_each_entry(ops
, &vport_ops_list
, list
)
178 if (ops
->type
== parms
->type
)
185 * ovs_vport_free - uninitialize and free vport
187 * @vport: vport to free
189 * Frees a vport allocated with ovs_vport_alloc() when it is no longer needed.
191 * The caller must ensure that an RCU grace period has passed since the last
192 * time @vport was in a datapath.
194 void ovs_vport_free(struct vport
*vport
)
196 kfree(rcu_dereference_raw(vport
->upcall_portids
));
197 free_percpu(vport
->percpu_stats
);
200 EXPORT_SYMBOL_GPL(ovs_vport_free
);
203 * ovs_vport_add - add vport device (for kernel callers)
205 * @parms: Information about new vport.
207 * Creates a new vport with the specified configuration (which is dependent on
208 * device type). ovs_mutex must be held.
210 struct vport
*ovs_vport_add(const struct vport_parms
*parms
)
212 struct vport_ops
*ops
;
215 ops
= ovs_vport_lookup(parms
);
217 struct hlist_head
*bucket
;
219 if (!try_module_get(ops
->owner
))
220 return ERR_PTR(-EAFNOSUPPORT
);
222 vport
= ops
->create(parms
);
224 module_put(ops
->owner
);
228 bucket
= hash_bucket(ovs_dp_get_net(vport
->dp
),
229 vport
->ops
->get_name(vport
));
230 hlist_add_head_rcu(&vport
->hash_node
, bucket
);
234 /* Unlock to attempt module load and return -EAGAIN if load
235 * was successful as we need to restart the port addition
239 request_module("vport-type-%d", parms
->type
);
242 if (!ovs_vport_lookup(parms
))
243 return ERR_PTR(-EAFNOSUPPORT
);
245 return ERR_PTR(-EAGAIN
);
249 * ovs_vport_set_options - modify existing vport device (for kernel callers)
251 * @vport: vport to modify.
252 * @options: New configuration.
254 * Modifies an existing device with the specified configuration (which is
255 * dependent on device type). ovs_mutex must be held.
257 int ovs_vport_set_options(struct vport
*vport
, struct nlattr
*options
)
259 if (!vport
->ops
->set_options
)
261 return vport
->ops
->set_options(vport
, options
);
265 * ovs_vport_del - delete existing vport device
267 * @vport: vport to delete.
269 * Detaches @vport from its datapath and destroys it. It is possible to fail
270 * for reasons such as lack of memory. ovs_mutex must be held.
272 void ovs_vport_del(struct vport
*vport
)
276 hlist_del_rcu(&vport
->hash_node
);
277 module_put(vport
->ops
->owner
);
278 vport
->ops
->destroy(vport
);
282 * ovs_vport_get_stats - retrieve device stats
284 * @vport: vport from which to retrieve the stats
285 * @stats: location to store stats
287 * Retrieves transmit, receive, and error stats for the given device.
289 * Must be called with ovs_mutex or rcu_read_lock.
291 void ovs_vport_get_stats(struct vport
*vport
, struct ovs_vport_stats
*stats
)
295 /* We potentially have two surces of stats that need to be
296 * combined: those we have collected (split into err_stats and
297 * percpu_stats), and device error stats from netdev->get_stats()
298 * (for errors that happen downstream and therefore aren't
299 * reported through our vport_record_error() function).
300 * Stats from first source are reported by ovs over
301 * OVS_VPORT_ATTR_STATS.
302 * netdev-stats can be directly read over netlink-ioctl.
305 stats
->rx_errors
= atomic_long_read(&vport
->err_stats
.rx_errors
);
306 stats
->tx_errors
= atomic_long_read(&vport
->err_stats
.tx_errors
);
307 stats
->tx_dropped
= atomic_long_read(&vport
->err_stats
.tx_dropped
);
308 stats
->rx_dropped
= atomic_long_read(&vport
->err_stats
.rx_dropped
);
311 stats
->rx_packets
= 0;
313 stats
->tx_packets
= 0;
315 for_each_possible_cpu(i
) {
316 const struct pcpu_sw_netstats
*percpu_stats
;
317 struct pcpu_sw_netstats local_stats
;
320 percpu_stats
= per_cpu_ptr(vport
->percpu_stats
, i
);
323 start
= u64_stats_fetch_begin_irq(&percpu_stats
->syncp
);
324 local_stats
= *percpu_stats
;
325 } while (u64_stats_fetch_retry_irq(&percpu_stats
->syncp
, start
));
327 stats
->rx_bytes
+= local_stats
.rx_bytes
;
328 stats
->rx_packets
+= local_stats
.rx_packets
;
329 stats
->tx_bytes
+= local_stats
.tx_bytes
;
330 stats
->tx_packets
+= local_stats
.tx_packets
;
335 * ovs_vport_get_options - retrieve device options
337 * @vport: vport from which to retrieve the options.
338 * @skb: sk_buff where options should be appended.
340 * Retrieves the configuration of the given device, appending an
341 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
342 * vport-specific attributes to @skb.
344 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
345 * negative error code if a real error occurred. If an error occurs, @skb is
348 * Must be called with ovs_mutex or rcu_read_lock.
350 int ovs_vport_get_options(const struct vport
*vport
, struct sk_buff
*skb
)
355 if (!vport
->ops
->get_options
)
358 nla
= nla_nest_start(skb
, OVS_VPORT_ATTR_OPTIONS
);
362 err
= vport
->ops
->get_options(vport
, skb
);
364 nla_nest_cancel(skb
, nla
);
368 nla_nest_end(skb
, nla
);
372 static void vport_portids_destroy_rcu_cb(struct rcu_head
*rcu
)
374 struct vport_portids
*ids
= container_of(rcu
, struct vport_portids
,
381 * ovs_vport_set_upcall_portids - set upcall portids of @vport.
383 * @vport: vport to modify.
384 * @ids: new configuration, an array of port ids.
386 * Sets the vport's upcall_portids to @ids.
388 * Returns 0 if successful, -EINVAL if @ids is zero length or cannot be parsed
389 * as an array of U32.
391 * Must be called with ovs_mutex.
393 int ovs_vport_set_upcall_portids(struct vport
*vport
, const struct nlattr
*ids
)
395 struct vport_portids
*old
, *vport_portids
;
397 if (!nla_len(ids
) || nla_len(ids
) % sizeof(u32
))
400 old
= ovsl_dereference(vport
->upcall_portids
);
402 vport_portids
= kmalloc(sizeof *vport_portids
+ nla_len(ids
),
407 vport_portids
->n_ids
= nla_len(ids
) / sizeof(u32
);
408 vport_portids
->rn_ids
= reciprocal_value(vport_portids
->n_ids
);
409 nla_memcpy(vport_portids
->ids
, ids
, nla_len(ids
));
411 rcu_assign_pointer(vport
->upcall_portids
, vport_portids
);
414 call_rcu(&old
->rcu
, vport_portids_destroy_rcu_cb
);
420 * ovs_vport_get_upcall_portids - get the upcall_portids of @vport.
422 * @vport: vport from which to retrieve the portids.
423 * @skb: sk_buff where portids should be appended.
425 * Retrieves the configuration of the given vport, appending the
426 * %OVS_VPORT_ATTR_UPCALL_PID attribute which is the array of upcall
429 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room.
430 * If an error occurs, @skb is left unmodified. Must be called with
431 * ovs_mutex or rcu_read_lock.
433 int ovs_vport_get_upcall_portids(const struct vport
*vport
,
436 struct vport_portids
*ids
;
438 ids
= rcu_dereference_ovsl(vport
->upcall_portids
);
440 if (vport
->dp
->user_features
& OVS_DP_F_VPORT_PIDS
)
441 return nla_put(skb
, OVS_VPORT_ATTR_UPCALL_PID
,
442 ids
->n_ids
* sizeof(u32
), (void *) ids
->ids
);
444 return nla_put_u32(skb
, OVS_VPORT_ATTR_UPCALL_PID
, ids
->ids
[0]);
448 * ovs_vport_find_upcall_portid - find the upcall portid to send upcall.
450 * @vport: vport from which the missed packet is received.
451 * @skb: skb that the missed packet was received.
453 * Uses the skb_get_hash() to select the upcall portid to send the
456 * Returns the portid of the target socket. Must be called with rcu_read_lock.
458 u32
ovs_vport_find_upcall_portid(const struct vport
*vport
, struct sk_buff
*skb
)
460 struct vport_portids
*ids
;
463 ids
= rcu_dereference(vport
->upcall_portids
);
465 if (ids
->n_ids
== 1 && ids
->ids
[0] == 0)
468 hash
= skb_get_hash(skb
);
469 return ids
->ids
[hash
- ids
->n_ids
* reciprocal_divide(hash
, ids
->rn_ids
)];
473 * ovs_vport_receive - pass up received packet to the datapath for processing
475 * @vport: vport that received the packet
476 * @skb: skb that was received
477 * @tun_info: tunnel (if any) that carried packet
479 * Must be called with rcu_read_lock. The packet cannot be shared and
480 * skb->data should point to the Ethernet header. The caller must have already
481 * called compute_ip_summed() to initialize the checksumming fields.
483 void ovs_vport_receive(struct vport
*vport
, struct sk_buff
*skb
,
484 const struct ovs_tunnel_info
*tun_info
)
486 struct pcpu_sw_netstats
*stats
;
487 struct sw_flow_key key
;
490 stats
= this_cpu_ptr(vport
->percpu_stats
);
491 u64_stats_update_begin(&stats
->syncp
);
493 stats
->rx_bytes
+= skb
->len
+ (skb_vlan_tag_present(skb
) ? VLAN_HLEN
: 0);
494 u64_stats_update_end(&stats
->syncp
);
496 ovs_skb_init_inner_protocol(skb
);
497 OVS_CB(skb
)->input_vport
= vport
;
498 OVS_CB(skb
)->egress_tun_info
= NULL
;
499 error
= ovs_flow_key_extract(tun_info
, skb
, &key
);
500 if (unlikely(error
)) {
505 ovs_dp_process_packet(skb
, &key
);
507 EXPORT_SYMBOL_GPL(ovs_vport_receive
);
510 * ovs_vport_send - send a packet on a device
512 * @vport: vport on which to send the packet
515 * Sends the given packet and returns the length of data sent. Either ovs
516 * lock or rcu_read_lock must be held.
518 int ovs_vport_send(struct vport
*vport
, struct sk_buff
*skb
)
520 int sent
= vport
->ops
->send(vport
, skb
);
522 if (likely(sent
> 0)) {
523 struct pcpu_sw_netstats
*stats
;
525 stats
= this_cpu_ptr(vport
->percpu_stats
);
527 u64_stats_update_begin(&stats
->syncp
);
529 stats
->tx_bytes
+= sent
;
530 u64_stats_update_end(&stats
->syncp
);
531 } else if (sent
< 0) {
532 ovs_vport_record_error(vport
, VPORT_E_TX_ERROR
);
534 ovs_vport_record_error(vport
, VPORT_E_TX_DROPPED
);
541 * ovs_vport_record_error - indicate device error to generic stats layer
543 * @vport: vport that encountered the error
544 * @err_type: one of enum vport_err_type types to indicate the error type
546 * If using the vport generic stats layer indicate that an error of the given
549 static void ovs_vport_record_error(struct vport
*vport
,
550 enum vport_err_type err_type
)
553 case VPORT_E_RX_DROPPED
:
554 atomic_long_inc(&vport
->err_stats
.rx_dropped
);
557 case VPORT_E_RX_ERROR
:
558 atomic_long_inc(&vport
->err_stats
.rx_errors
);
561 case VPORT_E_TX_DROPPED
:
562 atomic_long_inc(&vport
->err_stats
.tx_dropped
);
565 case VPORT_E_TX_ERROR
:
566 atomic_long_inc(&vport
->err_stats
.tx_errors
);
571 static void free_vport_rcu(struct rcu_head
*rcu
)
573 struct vport
*vport
= container_of(rcu
, struct vport
, rcu
);
575 ovs_vport_free(vport
);
578 void ovs_vport_deferred_free(struct vport
*vport
)
583 call_rcu(&vport
->rcu
, free_vport_rcu
);
585 EXPORT_SYMBOL_GPL(ovs_vport_deferred_free
);
587 int ovs_tunnel_get_egress_info(struct ovs_tunnel_info
*egress_tun_info
,
589 const struct ovs_tunnel_info
*tun_info
,
595 const struct ovs_key_ipv4_tunnel
*tun_key
;
599 if (unlikely(!tun_info
))
602 tun_key
= &tun_info
->tunnel
;
603 saddr
= tun_key
->ipv4_src
;
604 /* Route lookup to get srouce IP address: saddr.
605 * The process may need to be changed if the corresponding process
606 * in vports ops changed.
619 /* Generate egress_tun_info based on tun_info,
620 * saddr, tp_src and tp_dst
622 __ovs_flow_tun_info_init(egress_tun_info
,
623 saddr
, tun_key
->ipv4_dst
,
630 tun_info
->options_len
);
634 EXPORT_SYMBOL_GPL(ovs_tunnel_get_egress_info
);
636 int ovs_vport_get_egress_tun_info(struct vport
*vport
, struct sk_buff
*skb
,
637 struct ovs_tunnel_info
*info
)
639 /* get_egress_tun_info() is only implemented on tunnel ports. */
640 if (unlikely(!vport
->ops
->get_egress_tun_info
))
643 return vport
->ops
->get_egress_tun_info(vport
, skb
, info
);