2 * Copyright (c) 2007-2015 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/module.h>
31 #include <linux/if_link.h>
32 #include <net/net_namespace.h>
35 #include <net/geneve.h>
37 #include <net/vxlan.h>
42 #include "vport-internal_dev.h"
44 static LIST_HEAD(vport_ops_list
);
45 static bool compat_gre_loaded
= false;
46 static bool compat_ip6_tunnel_loaded
= false;
48 /* Protected by RCU read lock for reading, ovs_mutex for writing. */
49 static struct hlist_head
*dev_table
;
50 #define VPORT_HASH_BUCKETS 1024
53 * ovs_vport_init - initialize vport subsystem
55 * Called at module load time to initialize the vport subsystem.
57 int ovs_vport_init(void)
61 dev_table
= kcalloc(VPORT_HASH_BUCKETS
, sizeof(struct hlist_head
),
66 err
= lisp_init_module();
70 if (err
&& err
!= -EEXIST
) {
74 pr_warn("Cannot take GRE protocol rx entry"\
75 "- The GRE/ERSPAN rx feature not supported\n");
80 if (err
&& err
!= -EEXIST
)
82 compat_gre_loaded
= true;
85 if (err
&& err
!= -EEXIST
) {
89 pr_warn("IPv6 GRE/ERSPAN Rx mode is not supported\n");
90 goto skip_ip6_tunnel_init
;
94 err
= ip6_tunnel_init();
98 compat_ip6_tunnel_loaded
= true;
100 skip_ip6_tunnel_init
:
101 err
= geneve_init_module();
104 err
= vxlan_init_module();
107 err
= ovs_stt_init_module();
112 ovs_stt_cleanup_module();
114 vxlan_cleanup_module();
116 geneve_cleanup_module();
118 ip6_tunnel_cleanup();
126 lisp_cleanup_module();
133 * ovs_vport_exit - shutdown vport subsystem
135 * Called at module exit time to shutdown the vport subsystem.
137 void ovs_vport_exit(void)
139 if (compat_gre_loaded
) {
143 ovs_stt_cleanup_module();
144 vxlan_cleanup_module();
145 geneve_cleanup_module();
146 if (compat_ip6_tunnel_loaded
)
147 ip6_tunnel_cleanup();
149 lisp_cleanup_module();
153 static struct hlist_head
*hash_bucket(const struct net
*net
, const char *name
)
155 unsigned int hash
= jhash(name
, strlen(name
), (unsigned long) net
);
156 return &dev_table
[hash
& (VPORT_HASH_BUCKETS
- 1)];
159 int __ovs_vport_ops_register(struct vport_ops
*ops
)
165 list_for_each_entry(o
, &vport_ops_list
, list
)
166 if (ops
->type
== o
->type
)
169 list_add_tail(&ops
->list
, &vport_ops_list
);
175 EXPORT_SYMBOL_GPL(__ovs_vport_ops_register
);
177 void ovs_vport_ops_unregister(struct vport_ops
*ops
)
180 list_del(&ops
->list
);
183 EXPORT_SYMBOL_GPL(ovs_vport_ops_unregister
);
186 * ovs_vport_locate - find a port that has already been created
188 * @name: name of port to find
190 * Must be called with ovs or RCU read lock.
192 struct vport
*ovs_vport_locate(const struct net
*net
, const char *name
)
194 struct hlist_head
*bucket
= hash_bucket(net
, name
);
197 hlist_for_each_entry_rcu(vport
, bucket
, hash_node
)
198 if (!strcmp(name
, ovs_vport_name(vport
)) &&
199 net_eq(ovs_dp_get_net(vport
->dp
), net
))
206 * ovs_vport_alloc - allocate and initialize new vport
208 * @priv_size: Size of private data area to allocate.
209 * @ops: vport device ops
211 * Allocate and initialize a new vport defined by @ops. The vport will contain
212 * a private data area of size @priv_size that can be accessed using
213 * vport_priv(). vports that are no longer needed should be released with
216 struct vport
*ovs_vport_alloc(int priv_size
, const struct vport_ops
*ops
,
217 const struct vport_parms
*parms
)
222 alloc_size
= sizeof(struct vport
);
224 alloc_size
= ALIGN(alloc_size
, VPORT_ALIGN
);
225 alloc_size
+= priv_size
;
228 vport
= kzalloc(alloc_size
, GFP_KERNEL
);
230 return ERR_PTR(-ENOMEM
);
232 vport
->dp
= parms
->dp
;
233 vport
->port_no
= parms
->port_no
;
235 INIT_HLIST_NODE(&vport
->dp_hash_node
);
237 if (ovs_vport_set_upcall_portids(vport
, parms
->upcall_portids
)) {
239 return ERR_PTR(-EINVAL
);
244 EXPORT_SYMBOL_GPL(ovs_vport_alloc
);
247 * ovs_vport_free - uninitialize and free vport
249 * @vport: vport to free
251 * Frees a vport allocated with vport_alloc() when it is no longer needed.
253 * The caller must ensure that an RCU grace period has passed since the last
254 * time @vport was in a datapath.
256 void ovs_vport_free(struct vport
*vport
)
258 /* vport is freed from RCU callback or error path, Therefore
259 * it is safe to use raw dereference.
261 kfree(rcu_dereference_raw(vport
->upcall_portids
));
264 EXPORT_SYMBOL_GPL(ovs_vport_free
);
266 static struct vport_ops
*ovs_vport_lookup(const struct vport_parms
*parms
)
268 struct vport_ops
*ops
;
270 list_for_each_entry(ops
, &vport_ops_list
, list
)
271 if (ops
->type
== parms
->type
)
278 * ovs_vport_add - add vport device (for kernel callers)
280 * @parms: Information about new vport.
282 * Creates a new vport with the specified configuration (which is dependent on
283 * device type). ovs_mutex must be held.
285 struct vport
*ovs_vport_add(const struct vport_parms
*parms
)
287 struct vport_ops
*ops
;
290 ops
= ovs_vport_lookup(parms
);
292 struct hlist_head
*bucket
;
294 if (!try_module_get(ops
->owner
))
295 return ERR_PTR(-EAFNOSUPPORT
);
297 vport
= ops
->create(parms
);
299 module_put(ops
->owner
);
303 bucket
= hash_bucket(ovs_dp_get_net(vport
->dp
),
304 ovs_vport_name(vport
));
305 hlist_add_head_rcu(&vport
->hash_node
, bucket
);
309 if (parms
->type
== OVS_VPORT_TYPE_GRE
&& !compat_gre_loaded
) {
310 pr_warn("GRE protocol already loaded!\n");
311 return ERR_PTR(-EAFNOSUPPORT
);
313 /* Unlock to attempt module load and return -EAGAIN if load
314 * was successful as we need to restart the port addition
318 request_module("vport-type-%d", parms
->type
);
321 if (!ovs_vport_lookup(parms
))
322 return ERR_PTR(-EAFNOSUPPORT
);
324 return ERR_PTR(-EAGAIN
);
328 * ovs_vport_set_options - modify existing vport device (for kernel callers)
330 * @vport: vport to modify.
331 * @options: New configuration.
333 * Modifies an existing device with the specified configuration (which is
334 * dependent on device type). ovs_mutex must be held.
336 int ovs_vport_set_options(struct vport
*vport
, struct nlattr
*options
)
338 if (!vport
->ops
->set_options
)
340 return vport
->ops
->set_options(vport
, options
);
344 * ovs_vport_del - delete existing vport device
346 * @vport: vport to delete.
348 * Detaches @vport from its datapath and destroys it. ovs_mutex must be
351 void ovs_vport_del(struct vport
*vport
)
355 hlist_del_rcu(&vport
->hash_node
);
356 module_put(vport
->ops
->owner
);
357 vport
->ops
->destroy(vport
);
361 * ovs_vport_get_stats - retrieve device stats
363 * @vport: vport from which to retrieve the stats
364 * @stats: location to store stats
366 * Retrieves transmit, receive, and error stats for the given device.
368 * Must be called with ovs_mutex or rcu_read_lock.
370 void ovs_vport_get_stats(struct vport
*vport
, struct ovs_vport_stats
*stats
)
372 const struct rtnl_link_stats64
*dev_stats
;
373 struct rtnl_link_stats64 temp
;
375 dev_stats
= dev_get_stats(vport
->dev
, &temp
);
376 stats
->rx_errors
= dev_stats
->rx_errors
;
377 stats
->tx_errors
= dev_stats
->tx_errors
;
378 stats
->tx_dropped
= dev_stats
->tx_dropped
;
379 stats
->rx_dropped
= dev_stats
->rx_dropped
;
381 stats
->rx_bytes
= dev_stats
->rx_bytes
;
382 stats
->rx_packets
= dev_stats
->rx_packets
;
383 stats
->tx_bytes
= dev_stats
->tx_bytes
;
384 stats
->tx_packets
= dev_stats
->tx_packets
;
388 * ovs_vport_get_options - retrieve device options
390 * @vport: vport from which to retrieve the options.
391 * @skb: sk_buff where options should be appended.
393 * Retrieves the configuration of the given device, appending an
394 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
395 * vport-specific attributes to @skb.
397 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
398 * negative error code if a real error occurred. If an error occurs, @skb is
401 * Must be called with ovs_mutex or rcu_read_lock.
403 int ovs_vport_get_options(const struct vport
*vport
, struct sk_buff
*skb
)
408 if (!vport
->ops
->get_options
)
411 nla
= nla_nest_start_noflag(skb
, OVS_VPORT_ATTR_OPTIONS
);
415 err
= vport
->ops
->get_options(vport
, skb
);
417 nla_nest_cancel(skb
, nla
);
421 nla_nest_end(skb
, nla
);
426 * ovs_vport_set_upcall_portids - set upcall portids of @vport.
428 * @vport: vport to modify.
429 * @ids: new configuration, an array of port ids.
431 * Sets the vport's upcall_portids to @ids.
433 * Returns 0 if successful, -EINVAL if @ids is zero length or cannot be parsed
434 * as an array of U32.
436 * Must be called with ovs_mutex.
438 int ovs_vport_set_upcall_portids(struct vport
*vport
, const struct nlattr
*ids
)
440 struct vport_portids
*old
, *vport_portids
;
442 if (!nla_len(ids
) || nla_len(ids
) % sizeof(u32
))
445 old
= ovsl_dereference(vport
->upcall_portids
);
447 vport_portids
= kmalloc(sizeof(*vport_portids
) + nla_len(ids
),
452 vport_portids
->n_ids
= nla_len(ids
) / sizeof(u32
);
453 vport_portids
->rn_ids
= reciprocal_value(vport_portids
->n_ids
);
454 nla_memcpy(vport_portids
->ids
, ids
, nla_len(ids
));
456 rcu_assign_pointer(vport
->upcall_portids
, vport_portids
);
464 * ovs_vport_get_upcall_portids - get the upcall_portids of @vport.
466 * @vport: vport from which to retrieve the portids.
467 * @skb: sk_buff where portids should be appended.
469 * Retrieves the configuration of the given vport, appending the
470 * %OVS_VPORT_ATTR_UPCALL_PID attribute which is the array of upcall
473 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room.
474 * If an error occurs, @skb is left unmodified. Must be called with
475 * ovs_mutex or rcu_read_lock.
477 int ovs_vport_get_upcall_portids(const struct vport
*vport
,
480 struct vport_portids
*ids
;
482 ids
= rcu_dereference_ovsl(vport
->upcall_portids
);
484 if (vport
->dp
->user_features
& OVS_DP_F_VPORT_PIDS
)
485 return nla_put(skb
, OVS_VPORT_ATTR_UPCALL_PID
,
486 ids
->n_ids
* sizeof(u32
), (void *)ids
->ids
);
488 return nla_put_u32(skb
, OVS_VPORT_ATTR_UPCALL_PID
, ids
->ids
[0]);
492 * ovs_vport_find_upcall_portid - find the upcall portid to send upcall.
494 * @vport: vport from which the missed packet is received.
495 * @skb: skb that the missed packet was received.
497 * Uses the skb_get_hash() to select the upcall portid to send the
500 * Returns the portid of the target socket. Must be called with rcu_read_lock.
502 u32
ovs_vport_find_upcall_portid(const struct vport
*vport
, struct sk_buff
*skb
)
504 struct vport_portids
*ids
;
508 ids
= rcu_dereference(vport
->upcall_portids
);
510 if (ids
->n_ids
== 1 && ids
->ids
[0] == 0)
513 hash
= skb_get_hash(skb
);
514 ids_index
= hash
- ids
->n_ids
* reciprocal_divide(hash
, ids
->rn_ids
);
515 return ids
->ids
[ids_index
];
519 * ovs_vport_receive - pass up received packet to the datapath for processing
521 * @vport: vport that received the packet
522 * @skb: skb that was received
523 * @tun_key: tunnel (if any) that carried packet
525 * Must be called with rcu_read_lock. The packet cannot be shared and
526 * skb->data should point to the Ethernet header.
528 int ovs_vport_receive(struct vport
*vport
, struct sk_buff
*skb
,
529 const struct ip_tunnel_info
*tun_info
)
531 struct sw_flow_key key
;
534 OVS_CB(skb
)->input_vport
= vport
;
535 OVS_CB(skb
)->mru
= 0;
536 OVS_CB(skb
)->cutlen
= 0;
537 if (unlikely(dev_net(skb
->dev
) != ovs_dp_get_net(vport
->dp
))) {
541 skb_scrub_packet(skb
, true);
546 ovs_skb_init_inner_protocol(skb
);
547 skb_clear_ovs_gso_cb(skb
);
548 /* Extract flow from 'skb' into 'key'. */
549 error
= ovs_flow_key_extract(tun_info
, skb
, &key
);
550 if (unlikely(error
)) {
554 ovs_dp_process_packet(skb
, &key
);
558 static int packet_length(const struct sk_buff
*skb
,
559 struct net_device
*dev
)
561 int length
= skb
->len
- dev
->hard_header_len
;
563 if (!skb_vlan_tag_present(skb
) &&
564 eth_type_vlan(skb
->protocol
))
567 /* Don't subtract for multiple VLAN tags. Most (all?) drivers allow
568 * (ETH_LEN + VLAN_HLEN) in addition to the mtu value, but almost none
569 * account for 802.1ad. e.g. is_skb_forwardable().
572 return length
> 0 ? length
: 0;
575 void ovs_vport_send(struct vport
*vport
, struct sk_buff
*skb
, u8 mac_proto
)
577 int mtu
= vport
->dev
->mtu
;
579 switch (vport
->dev
->type
) {
581 if (mac_proto
== MAC_PROTO_ETHERNET
) {
582 skb_reset_network_header(skb
);
583 skb_reset_mac_len(skb
);
584 skb
->protocol
= htons(ETH_P_TEB
);
585 } else if (mac_proto
!= MAC_PROTO_NONE
) {
591 if (mac_proto
!= MAC_PROTO_ETHERNET
)
598 if (unlikely(packet_length(skb
, vport
->dev
) > mtu
&&
600 net_warn_ratelimited("%s: dropped over-mtu packet: %d > %d\n",
602 packet_length(skb
, vport
->dev
), mtu
);
603 vport
->dev
->stats
.tx_errors
++;
607 skb
->dev
= vport
->dev
;
608 vport
->ops
->send(skb
);