Linux Audio

Check our new training course

Loading...
v4.10.11
 
  1/*
  2 * Copyright (c) 2007-2014 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#include <linux/module.h>
 32
 33#include "datapath.h"
 34#include "vport.h"
 35#include "vport-internal_dev.h"
 36
 37static LIST_HEAD(vport_ops_list);
 38
 39/* Protected by RCU read lock for reading, ovs_mutex for writing. */
 40static struct hlist_head *dev_table;
 41#define VPORT_HASH_BUCKETS 1024
 42
 43/**
 44 *	ovs_vport_init - initialize vport subsystem
 45 *
 46 * Called at module load time to initialize the vport subsystem.
 47 */
 48int ovs_vport_init(void)
 49{
 50	dev_table = kzalloc(VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
 51			    GFP_KERNEL);
 52	if (!dev_table)
 53		return -ENOMEM;
 54
 55	return 0;
 56}
 57
 58/**
 59 *	ovs_vport_exit - shutdown vport subsystem
 60 *
 61 * Called at module exit time to shutdown the vport subsystem.
 62 */
 63void ovs_vport_exit(void)
 64{
 65	kfree(dev_table);
 66}
 67
 68static struct hlist_head *hash_bucket(const struct net *net, const char *name)
 69{
 70	unsigned int hash = jhash(name, strlen(name), (unsigned long) net);
 71	return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
 72}
 73
 74int __ovs_vport_ops_register(struct vport_ops *ops)
 75{
 76	int err = -EEXIST;
 77	struct vport_ops *o;
 78
 79	ovs_lock();
 80	list_for_each_entry(o, &vport_ops_list, list)
 81		if (ops->type == o->type)
 82			goto errout;
 83
 84	list_add_tail(&ops->list, &vport_ops_list);
 85	err = 0;
 86errout:
 87	ovs_unlock();
 88	return err;
 89}
 90EXPORT_SYMBOL_GPL(__ovs_vport_ops_register);
 91
 92void ovs_vport_ops_unregister(struct vport_ops *ops)
 93{
 94	ovs_lock();
 95	list_del(&ops->list);
 96	ovs_unlock();
 97}
 98EXPORT_SYMBOL_GPL(ovs_vport_ops_unregister);
 99
100/**
101 *	ovs_vport_locate - find a port that has already been created
102 *
 
103 * @name: name of port to find
104 *
105 * Must be called with ovs or RCU read lock.
106 */
107struct vport *ovs_vport_locate(const struct net *net, const char *name)
108{
109	struct hlist_head *bucket = hash_bucket(net, name);
110	struct vport *vport;
111
112	hlist_for_each_entry_rcu(vport, bucket, hash_node)
 
113		if (!strcmp(name, ovs_vport_name(vport)) &&
114		    net_eq(ovs_dp_get_net(vport->dp), net))
115			return vport;
116
117	return NULL;
118}
119
120/**
121 *	ovs_vport_alloc - allocate and initialize new vport
122 *
123 * @priv_size: Size of private data area to allocate.
124 * @ops: vport device ops
 
125 *
126 * Allocate and initialize a new vport defined by @ops.  The vport will contain
127 * a private data area of size @priv_size that can be accessed using
128 * vport_priv().  vports that are no longer needed should be released with
 
129 * vport_free().
130 */
131struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
132			  const struct vport_parms *parms)
133{
134	struct vport *vport;
135	size_t alloc_size;
136
137	alloc_size = sizeof(struct vport);
138	if (priv_size) {
139		alloc_size = ALIGN(alloc_size, VPORT_ALIGN);
140		alloc_size += priv_size;
141	}
142
143	vport = kzalloc(alloc_size, GFP_KERNEL);
144	if (!vport)
145		return ERR_PTR(-ENOMEM);
146
147	vport->dp = parms->dp;
148	vport->port_no = parms->port_no;
149	vport->ops = ops;
150	INIT_HLIST_NODE(&vport->dp_hash_node);
151
152	if (ovs_vport_set_upcall_portids(vport, parms->upcall_portids)) {
153		kfree(vport);
154		return ERR_PTR(-EINVAL);
155	}
156
157	return vport;
158}
159EXPORT_SYMBOL_GPL(ovs_vport_alloc);
160
161/**
162 *	ovs_vport_free - uninitialize and free vport
163 *
164 * @vport: vport to free
165 *
166 * Frees a vport allocated with vport_alloc() when it is no longer needed.
167 *
168 * The caller must ensure that an RCU grace period has passed since the last
169 * time @vport was in a datapath.
170 */
171void ovs_vport_free(struct vport *vport)
172{
173	/* vport is freed from RCU callback or error path, Therefore
174	 * it is safe to use raw dereference.
175	 */
176	kfree(rcu_dereference_raw(vport->upcall_portids));
177	kfree(vport);
178}
179EXPORT_SYMBOL_GPL(ovs_vport_free);
180
181static struct vport_ops *ovs_vport_lookup(const struct vport_parms *parms)
182{
183	struct vport_ops *ops;
184
185	list_for_each_entry(ops, &vport_ops_list, list)
186		if (ops->type == parms->type)
187			return ops;
188
189	return NULL;
190}
191
192/**
193 *	ovs_vport_add - add vport device (for kernel callers)
194 *
195 * @parms: Information about new vport.
196 *
197 * Creates a new vport with the specified configuration (which is dependent on
198 * device type).  ovs_mutex must be held.
199 */
200struct vport *ovs_vport_add(const struct vport_parms *parms)
201{
202	struct vport_ops *ops;
203	struct vport *vport;
204
205	ops = ovs_vport_lookup(parms);
206	if (ops) {
207		struct hlist_head *bucket;
208
209		if (!try_module_get(ops->owner))
210			return ERR_PTR(-EAFNOSUPPORT);
211
212		vport = ops->create(parms);
213		if (IS_ERR(vport)) {
214			module_put(ops->owner);
215			return vport;
216		}
217
218		bucket = hash_bucket(ovs_dp_get_net(vport->dp),
219				     ovs_vport_name(vport));
220		hlist_add_head_rcu(&vport->hash_node, bucket);
221		return vport;
222	}
223
224	/* Unlock to attempt module load and return -EAGAIN if load
225	 * was successful as we need to restart the port addition
226	 * workflow.
227	 */
228	ovs_unlock();
229	request_module("vport-type-%d", parms->type);
230	ovs_lock();
231
232	if (!ovs_vport_lookup(parms))
233		return ERR_PTR(-EAFNOSUPPORT);
234	else
235		return ERR_PTR(-EAGAIN);
236}
237
238/**
239 *	ovs_vport_set_options - modify existing vport device (for kernel callers)
240 *
241 * @vport: vport to modify.
242 * @options: New configuration.
243 *
244 * Modifies an existing device with the specified configuration (which is
245 * dependent on device type).  ovs_mutex must be held.
246 */
247int ovs_vport_set_options(struct vport *vport, struct nlattr *options)
248{
249	if (!vport->ops->set_options)
250		return -EOPNOTSUPP;
251	return vport->ops->set_options(vport, options);
252}
253
254/**
255 *	ovs_vport_del - delete existing vport device
256 *
257 * @vport: vport to delete.
258 *
259 * Detaches @vport from its datapath and destroys it.  ovs_mutex must
260 * be held.
261 */
262void ovs_vport_del(struct vport *vport)
263{
264	ASSERT_OVSL();
265
266	hlist_del_rcu(&vport->hash_node);
267	module_put(vport->ops->owner);
268	vport->ops->destroy(vport);
269}
270
271/**
272 *	ovs_vport_get_stats - retrieve device stats
273 *
274 * @vport: vport from which to retrieve the stats
275 * @stats: location to store stats
276 *
277 * Retrieves transmit, receive, and error stats for the given device.
278 *
279 * Must be called with ovs_mutex or rcu_read_lock.
280 */
281void ovs_vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
282{
283	const struct rtnl_link_stats64 *dev_stats;
284	struct rtnl_link_stats64 temp;
285
286	dev_stats = dev_get_stats(vport->dev, &temp);
287	stats->rx_errors  = dev_stats->rx_errors;
288	stats->tx_errors  = dev_stats->tx_errors;
289	stats->tx_dropped = dev_stats->tx_dropped;
290	stats->rx_dropped = dev_stats->rx_dropped;
291
292	stats->rx_bytes	  = dev_stats->rx_bytes;
293	stats->rx_packets = dev_stats->rx_packets;
294	stats->tx_bytes	  = dev_stats->tx_bytes;
295	stats->tx_packets = dev_stats->tx_packets;
296}
297
298/**
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
299 *	ovs_vport_get_options - retrieve device options
300 *
301 * @vport: vport from which to retrieve the options.
302 * @skb: sk_buff where options should be appended.
303 *
304 * Retrieves the configuration of the given device, appending an
305 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
306 * vport-specific attributes to @skb.
307 *
308 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
309 * negative error code if a real error occurred.  If an error occurs, @skb is
310 * left unmodified.
311 *
312 * Must be called with ovs_mutex or rcu_read_lock.
313 */
314int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
315{
316	struct nlattr *nla;
317	int err;
318
319	if (!vport->ops->get_options)
320		return 0;
321
322	nla = nla_nest_start(skb, OVS_VPORT_ATTR_OPTIONS);
323	if (!nla)
324		return -EMSGSIZE;
325
326	err = vport->ops->get_options(vport, skb);
327	if (err) {
328		nla_nest_cancel(skb, nla);
329		return err;
330	}
331
332	nla_nest_end(skb, nla);
333	return 0;
334}
335
336/**
337 *	ovs_vport_set_upcall_portids - set upcall portids of @vport.
338 *
339 * @vport: vport to modify.
340 * @ids: new configuration, an array of port ids.
341 *
342 * Sets the vport's upcall_portids to @ids.
343 *
344 * Returns 0 if successful, -EINVAL if @ids is zero length or cannot be parsed
345 * as an array of U32.
346 *
347 * Must be called with ovs_mutex.
348 */
349int ovs_vport_set_upcall_portids(struct vport *vport, const struct nlattr *ids)
350{
351	struct vport_portids *old, *vport_portids;
352
353	if (!nla_len(ids) || nla_len(ids) % sizeof(u32))
354		return -EINVAL;
355
356	old = ovsl_dereference(vport->upcall_portids);
357
358	vport_portids = kmalloc(sizeof(*vport_portids) + nla_len(ids),
359				GFP_KERNEL);
360	if (!vport_portids)
361		return -ENOMEM;
362
363	vport_portids->n_ids = nla_len(ids) / sizeof(u32);
364	vport_portids->rn_ids = reciprocal_value(vport_portids->n_ids);
365	nla_memcpy(vport_portids->ids, ids, nla_len(ids));
366
367	rcu_assign_pointer(vport->upcall_portids, vport_portids);
368
369	if (old)
370		kfree_rcu(old, rcu);
371	return 0;
372}
373
374/**
375 *	ovs_vport_get_upcall_portids - get the upcall_portids of @vport.
376 *
377 * @vport: vport from which to retrieve the portids.
378 * @skb: sk_buff where portids should be appended.
379 *
380 * Retrieves the configuration of the given vport, appending the
381 * %OVS_VPORT_ATTR_UPCALL_PID attribute which is the array of upcall
382 * portids to @skb.
383 *
384 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room.
385 * If an error occurs, @skb is left unmodified.  Must be called with
386 * ovs_mutex or rcu_read_lock.
387 */
388int ovs_vport_get_upcall_portids(const struct vport *vport,
389				 struct sk_buff *skb)
390{
391	struct vport_portids *ids;
392
393	ids = rcu_dereference_ovsl(vport->upcall_portids);
394
395	if (vport->dp->user_features & OVS_DP_F_VPORT_PIDS)
396		return nla_put(skb, OVS_VPORT_ATTR_UPCALL_PID,
397			       ids->n_ids * sizeof(u32), (void *)ids->ids);
398	else
399		return nla_put_u32(skb, OVS_VPORT_ATTR_UPCALL_PID, ids->ids[0]);
400}
401
402/**
403 *	ovs_vport_find_upcall_portid - find the upcall portid to send upcall.
404 *
405 * @vport: vport from which the missed packet is received.
406 * @skb: skb that the missed packet was received.
407 *
408 * Uses the skb_get_hash() to select the upcall portid to send the
409 * upcall.
410 *
411 * Returns the portid of the target socket.  Must be called with rcu_read_lock.
412 */
413u32 ovs_vport_find_upcall_portid(const struct vport *vport, struct sk_buff *skb)
 
414{
415	struct vport_portids *ids;
416	u32 ids_index;
417	u32 hash;
418
419	ids = rcu_dereference(vport->upcall_portids);
420
421	if (ids->n_ids == 1 && ids->ids[0] == 0)
422		return 0;
 
423
424	hash = skb_get_hash(skb);
425	ids_index = hash - ids->n_ids * reciprocal_divide(hash, ids->rn_ids);
426	return ids->ids[ids_index];
427}
428
429/**
430 *	ovs_vport_receive - pass up received packet to the datapath for processing
431 *
432 * @vport: vport that received the packet
433 * @skb: skb that was received
434 * @tun_key: tunnel (if any) that carried packet
435 *
436 * Must be called with rcu_read_lock.  The packet cannot be shared and
437 * skb->data should point to the Ethernet header.
438 */
439int ovs_vport_receive(struct vport *vport, struct sk_buff *skb,
440		      const struct ip_tunnel_info *tun_info)
441{
442	struct sw_flow_key key;
443	int error;
444
445	OVS_CB(skb)->input_vport = vport;
446	OVS_CB(skb)->mru = 0;
447	OVS_CB(skb)->cutlen = 0;
448	if (unlikely(dev_net(skb->dev) != ovs_dp_get_net(vport->dp))) {
449		u32 mark;
450
451		mark = skb->mark;
452		skb_scrub_packet(skb, true);
453		skb->mark = mark;
454		tun_info = NULL;
455	}
456
457	/* Extract flow from 'skb' into 'key'. */
458	error = ovs_flow_key_extract(tun_info, skb, &key);
459	if (unlikely(error)) {
460		kfree_skb(skb);
461		return error;
462	}
463	ovs_dp_process_packet(skb, &key);
464	return 0;
465}
466
467static unsigned int packet_length(const struct sk_buff *skb,
468				  struct net_device *dev)
469{
470	unsigned int length = skb->len - dev->hard_header_len;
471
472	if (!skb_vlan_tag_present(skb) &&
473	    eth_type_vlan(skb->protocol))
474		length -= VLAN_HLEN;
475
476	/* Don't subtract for multiple VLAN tags. Most (all?) drivers allow
477	 * (ETH_LEN + VLAN_HLEN) in addition to the mtu value, but almost none
478	 * account for 802.1ad. e.g. is_skb_forwardable().
479	 */
480
481	return length;
482}
483
484void ovs_vport_send(struct vport *vport, struct sk_buff *skb, u8 mac_proto)
485{
486	int mtu = vport->dev->mtu;
487
488	switch (vport->dev->type) {
489	case ARPHRD_NONE:
490		if (mac_proto == MAC_PROTO_ETHERNET) {
491			skb_reset_network_header(skb);
492			skb_reset_mac_len(skb);
493			skb->protocol = htons(ETH_P_TEB);
494		} else if (mac_proto != MAC_PROTO_NONE) {
495			WARN_ON_ONCE(1);
496			goto drop;
497		}
498		break;
499	case ARPHRD_ETHER:
500		if (mac_proto != MAC_PROTO_ETHERNET)
501			goto drop;
502		break;
503	default:
504		goto drop;
505	}
506
507	if (unlikely(packet_length(skb, vport->dev) > mtu &&
508		     !skb_is_gso(skb))) {
509		net_warn_ratelimited("%s: dropped over-mtu packet: %d > %d\n",
510				     vport->dev->name,
511				     packet_length(skb, vport->dev), mtu);
512		vport->dev->stats.tx_errors++;
 
 
 
 
 
513		goto drop;
514	}
515
516	skb->dev = vport->dev;
 
517	vport->ops->send(skb);
518	return;
519
520drop:
521	kfree_skb(skb);
522}
v6.2
  1// SPDX-License-Identifier: GPL-2.0-only
  2/*
  3 * Copyright (c) 2007-2014 Nicira, Inc.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
  4 */
  5
  6#include <linux/etherdevice.h>
  7#include <linux/if.h>
  8#include <linux/if_vlan.h>
  9#include <linux/jhash.h>
 10#include <linux/kernel.h>
 11#include <linux/list.h>
 12#include <linux/mutex.h>
 13#include <linux/percpu.h>
 14#include <linux/rcupdate.h>
 15#include <linux/rtnetlink.h>
 16#include <linux/compat.h>
 17#include <net/net_namespace.h>
 18#include <linux/module.h>
 19
 20#include "datapath.h"
 21#include "vport.h"
 22#include "vport-internal_dev.h"
 23
 24static LIST_HEAD(vport_ops_list);
 25
 26/* Protected by RCU read lock for reading, ovs_mutex for writing. */
 27static struct hlist_head *dev_table;
 28#define VPORT_HASH_BUCKETS 1024
 29
 30/**
 31 *	ovs_vport_init - initialize vport subsystem
 32 *
 33 * Called at module load time to initialize the vport subsystem.
 34 */
 35int ovs_vport_init(void)
 36{
 37	dev_table = kcalloc(VPORT_HASH_BUCKETS, sizeof(struct hlist_head),
 38			    GFP_KERNEL);
 39	if (!dev_table)
 40		return -ENOMEM;
 41
 42	return 0;
 43}
 44
 45/**
 46 *	ovs_vport_exit - shutdown vport subsystem
 47 *
 48 * Called at module exit time to shutdown the vport subsystem.
 49 */
 50void ovs_vport_exit(void)
 51{
 52	kfree(dev_table);
 53}
 54
 55static struct hlist_head *hash_bucket(const struct net *net, const char *name)
 56{
 57	unsigned int hash = jhash(name, strlen(name), (unsigned long) net);
 58	return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)];
 59}
 60
 61int __ovs_vport_ops_register(struct vport_ops *ops)
 62{
 63	int err = -EEXIST;
 64	struct vport_ops *o;
 65
 66	ovs_lock();
 67	list_for_each_entry(o, &vport_ops_list, list)
 68		if (ops->type == o->type)
 69			goto errout;
 70
 71	list_add_tail(&ops->list, &vport_ops_list);
 72	err = 0;
 73errout:
 74	ovs_unlock();
 75	return err;
 76}
 77EXPORT_SYMBOL_GPL(__ovs_vport_ops_register);
 78
 79void ovs_vport_ops_unregister(struct vport_ops *ops)
 80{
 81	ovs_lock();
 82	list_del(&ops->list);
 83	ovs_unlock();
 84}
 85EXPORT_SYMBOL_GPL(ovs_vport_ops_unregister);
 86
 87/**
 88 *	ovs_vport_locate - find a port that has already been created
 89 *
 90 * @net: network namespace
 91 * @name: name of port to find
 92 *
 93 * Must be called with ovs or RCU read lock.
 94 */
 95struct vport *ovs_vport_locate(const struct net *net, const char *name)
 96{
 97	struct hlist_head *bucket = hash_bucket(net, name);
 98	struct vport *vport;
 99
100	hlist_for_each_entry_rcu(vport, bucket, hash_node,
101				 lockdep_ovsl_is_held())
102		if (!strcmp(name, ovs_vport_name(vport)) &&
103		    net_eq(ovs_dp_get_net(vport->dp), net))
104			return vport;
105
106	return NULL;
107}
108
109/**
110 *	ovs_vport_alloc - allocate and initialize new vport
111 *
112 * @priv_size: Size of private data area to allocate.
113 * @ops: vport device ops
114 * @parms: information about new vport.
115 *
116 * Allocate and initialize a new vport defined by @ops.  The vport will contain
117 * a private data area of size @priv_size that can be accessed using
118 * vport_priv().  Some parameters of the vport will be initialized from @parms.
119 * @vports that are no longer needed should be released with
120 * vport_free().
121 */
122struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops,
123			      const struct vport_parms *parms)
124{
125	struct vport *vport;
126	size_t alloc_size;
127
128	alloc_size = sizeof(struct vport);
129	if (priv_size) {
130		alloc_size = ALIGN(alloc_size, VPORT_ALIGN);
131		alloc_size += priv_size;
132	}
133
134	vport = kzalloc(alloc_size, GFP_KERNEL);
135	if (!vport)
136		return ERR_PTR(-ENOMEM);
137
138	vport->dp = parms->dp;
139	vport->port_no = parms->port_no;
140	vport->ops = ops;
141	INIT_HLIST_NODE(&vport->dp_hash_node);
142
143	if (ovs_vport_set_upcall_portids(vport, parms->upcall_portids)) {
144		kfree(vport);
145		return ERR_PTR(-EINVAL);
146	}
147
148	return vport;
149}
150EXPORT_SYMBOL_GPL(ovs_vport_alloc);
151
152/**
153 *	ovs_vport_free - uninitialize and free vport
154 *
155 * @vport: vport to free
156 *
157 * Frees a vport allocated with vport_alloc() when it is no longer needed.
158 *
159 * The caller must ensure that an RCU grace period has passed since the last
160 * time @vport was in a datapath.
161 */
162void ovs_vport_free(struct vport *vport)
163{
164	/* vport is freed from RCU callback or error path, Therefore
165	 * it is safe to use raw dereference.
166	 */
167	kfree(rcu_dereference_raw(vport->upcall_portids));
168	kfree(vport);
169}
170EXPORT_SYMBOL_GPL(ovs_vport_free);
171
172static struct vport_ops *ovs_vport_lookup(const struct vport_parms *parms)
173{
174	struct vport_ops *ops;
175
176	list_for_each_entry(ops, &vport_ops_list, list)
177		if (ops->type == parms->type)
178			return ops;
179
180	return NULL;
181}
182
183/**
184 *	ovs_vport_add - add vport device (for kernel callers)
185 *
186 * @parms: Information about new vport.
187 *
188 * Creates a new vport with the specified configuration (which is dependent on
189 * device type).  ovs_mutex must be held.
190 */
191struct vport *ovs_vport_add(const struct vport_parms *parms)
192{
193	struct vport_ops *ops;
194	struct vport *vport;
195
196	ops = ovs_vport_lookup(parms);
197	if (ops) {
198		struct hlist_head *bucket;
199
200		if (!try_module_get(ops->owner))
201			return ERR_PTR(-EAFNOSUPPORT);
202
203		vport = ops->create(parms);
204		if (IS_ERR(vport)) {
205			module_put(ops->owner);
206			return vport;
207		}
208
209		bucket = hash_bucket(ovs_dp_get_net(vport->dp),
210				     ovs_vport_name(vport));
211		hlist_add_head_rcu(&vport->hash_node, bucket);
212		return vport;
213	}
214
215	/* Unlock to attempt module load and return -EAGAIN if load
216	 * was successful as we need to restart the port addition
217	 * workflow.
218	 */
219	ovs_unlock();
220	request_module("vport-type-%d", parms->type);
221	ovs_lock();
222
223	if (!ovs_vport_lookup(parms))
224		return ERR_PTR(-EAFNOSUPPORT);
225	else
226		return ERR_PTR(-EAGAIN);
227}
228
229/**
230 *	ovs_vport_set_options - modify existing vport device (for kernel callers)
231 *
232 * @vport: vport to modify.
233 * @options: New configuration.
234 *
235 * Modifies an existing device with the specified configuration (which is
236 * dependent on device type).  ovs_mutex must be held.
237 */
238int ovs_vport_set_options(struct vport *vport, struct nlattr *options)
239{
240	if (!vport->ops->set_options)
241		return -EOPNOTSUPP;
242	return vport->ops->set_options(vport, options);
243}
244
245/**
246 *	ovs_vport_del - delete existing vport device
247 *
248 * @vport: vport to delete.
249 *
250 * Detaches @vport from its datapath and destroys it.  ovs_mutex must
251 * be held.
252 */
253void ovs_vport_del(struct vport *vport)
254{
 
 
255	hlist_del_rcu(&vport->hash_node);
256	module_put(vport->ops->owner);
257	vport->ops->destroy(vport);
258}
259
260/**
261 *	ovs_vport_get_stats - retrieve device stats
262 *
263 * @vport: vport from which to retrieve the stats
264 * @stats: location to store stats
265 *
266 * Retrieves transmit, receive, and error stats for the given device.
267 *
268 * Must be called with ovs_mutex or rcu_read_lock.
269 */
270void ovs_vport_get_stats(struct vport *vport, struct ovs_vport_stats *stats)
271{
272	const struct rtnl_link_stats64 *dev_stats;
273	struct rtnl_link_stats64 temp;
274
275	dev_stats = dev_get_stats(vport->dev, &temp);
276	stats->rx_errors  = dev_stats->rx_errors;
277	stats->tx_errors  = dev_stats->tx_errors;
278	stats->tx_dropped = dev_stats->tx_dropped;
279	stats->rx_dropped = dev_stats->rx_dropped;
280
281	stats->rx_bytes	  = dev_stats->rx_bytes;
282	stats->rx_packets = dev_stats->rx_packets;
283	stats->tx_bytes	  = dev_stats->tx_bytes;
284	stats->tx_packets = dev_stats->tx_packets;
285}
286
287/**
288 *	ovs_vport_get_upcall_stats - retrieve upcall stats
289 *
290 * @vport: vport from which to retrieve the stats.
291 * @skb: sk_buff where upcall stats should be appended.
292 *
293 * Retrieves upcall stats for the given device.
294 *
295 * Must be called with ovs_mutex or rcu_read_lock.
296 */
297int ovs_vport_get_upcall_stats(struct vport *vport, struct sk_buff *skb)
298{
299	struct nlattr *nla;
300	int i;
301
302	__u64 tx_success = 0;
303	__u64 tx_fail = 0;
304
305	for_each_possible_cpu(i) {
306		const struct vport_upcall_stats_percpu *stats;
307		unsigned int start;
308
309		stats = per_cpu_ptr(vport->upcall_stats, i);
310		do {
311			start = u64_stats_fetch_begin(&stats->syncp);
312			tx_success += u64_stats_read(&stats->n_success);
313			tx_fail += u64_stats_read(&stats->n_fail);
314		} while (u64_stats_fetch_retry(&stats->syncp, start));
315	}
316
317	nla = nla_nest_start_noflag(skb, OVS_VPORT_ATTR_UPCALL_STATS);
318	if (!nla)
319		return -EMSGSIZE;
320
321	if (nla_put_u64_64bit(skb, OVS_VPORT_UPCALL_ATTR_SUCCESS, tx_success,
322			      OVS_VPORT_ATTR_PAD)) {
323		nla_nest_cancel(skb, nla);
324		return -EMSGSIZE;
325	}
326
327	if (nla_put_u64_64bit(skb, OVS_VPORT_UPCALL_ATTR_FAIL, tx_fail,
328			      OVS_VPORT_ATTR_PAD)) {
329		nla_nest_cancel(skb, nla);
330		return -EMSGSIZE;
331	}
332	nla_nest_end(skb, nla);
333
334	return 0;
335}
336
337/**
338 *	ovs_vport_get_options - retrieve device options
339 *
340 * @vport: vport from which to retrieve the options.
341 * @skb: sk_buff where options should be appended.
342 *
343 * Retrieves the configuration of the given device, appending an
344 * %OVS_VPORT_ATTR_OPTIONS attribute that in turn contains nested
345 * vport-specific attributes to @skb.
346 *
347 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room, or another
348 * negative error code if a real error occurred.  If an error occurs, @skb is
349 * left unmodified.
350 *
351 * Must be called with ovs_mutex or rcu_read_lock.
352 */
353int ovs_vport_get_options(const struct vport *vport, struct sk_buff *skb)
354{
355	struct nlattr *nla;
356	int err;
357
358	if (!vport->ops->get_options)
359		return 0;
360
361	nla = nla_nest_start_noflag(skb, OVS_VPORT_ATTR_OPTIONS);
362	if (!nla)
363		return -EMSGSIZE;
364
365	err = vport->ops->get_options(vport, skb);
366	if (err) {
367		nla_nest_cancel(skb, nla);
368		return err;
369	}
370
371	nla_nest_end(skb, nla);
372	return 0;
373}
374
375/**
376 *	ovs_vport_set_upcall_portids - set upcall portids of @vport.
377 *
378 * @vport: vport to modify.
379 * @ids: new configuration, an array of port ids.
380 *
381 * Sets the vport's upcall_portids to @ids.
382 *
383 * Returns 0 if successful, -EINVAL if @ids is zero length or cannot be parsed
384 * as an array of U32.
385 *
386 * Must be called with ovs_mutex.
387 */
388int ovs_vport_set_upcall_portids(struct vport *vport, const struct nlattr *ids)
389{
390	struct vport_portids *old, *vport_portids;
391
392	if (!nla_len(ids) || nla_len(ids) % sizeof(u32))
393		return -EINVAL;
394
395	old = ovsl_dereference(vport->upcall_portids);
396
397	vport_portids = kmalloc(sizeof(*vport_portids) + nla_len(ids),
398				GFP_KERNEL);
399	if (!vport_portids)
400		return -ENOMEM;
401
402	vport_portids->n_ids = nla_len(ids) / sizeof(u32);
403	vport_portids->rn_ids = reciprocal_value(vport_portids->n_ids);
404	nla_memcpy(vport_portids->ids, ids, nla_len(ids));
405
406	rcu_assign_pointer(vport->upcall_portids, vport_portids);
407
408	if (old)
409		kfree_rcu(old, rcu);
410	return 0;
411}
412
413/**
414 *	ovs_vport_get_upcall_portids - get the upcall_portids of @vport.
415 *
416 * @vport: vport from which to retrieve the portids.
417 * @skb: sk_buff where portids should be appended.
418 *
419 * Retrieves the configuration of the given vport, appending the
420 * %OVS_VPORT_ATTR_UPCALL_PID attribute which is the array of upcall
421 * portids to @skb.
422 *
423 * Returns 0 if successful, -EMSGSIZE if @skb has insufficient room.
424 * If an error occurs, @skb is left unmodified.  Must be called with
425 * ovs_mutex or rcu_read_lock.
426 */
427int ovs_vport_get_upcall_portids(const struct vport *vport,
428				 struct sk_buff *skb)
429{
430	struct vport_portids *ids;
431
432	ids = rcu_dereference_ovsl(vport->upcall_portids);
433
434	if (vport->dp->user_features & OVS_DP_F_VPORT_PIDS)
435		return nla_put(skb, OVS_VPORT_ATTR_UPCALL_PID,
436			       ids->n_ids * sizeof(u32), (void *)ids->ids);
437	else
438		return nla_put_u32(skb, OVS_VPORT_ATTR_UPCALL_PID, ids->ids[0]);
439}
440
441/**
442 *	ovs_vport_find_upcall_portid - find the upcall portid to send upcall.
443 *
444 * @vport: vport from which the missed packet is received.
445 * @skb: skb that the missed packet was received.
446 *
447 * Uses the skb_get_hash() to select the upcall portid to send the
448 * upcall.
449 *
450 * Returns the portid of the target socket.  Must be called with rcu_read_lock.
451 */
452u32 ovs_vport_find_upcall_portid(const struct vport *vport,
453				 struct sk_buff *skb)
454{
455	struct vport_portids *ids;
456	u32 ids_index;
457	u32 hash;
458
459	ids = rcu_dereference(vport->upcall_portids);
460
461	/* If there is only one portid, select it in the fast-path. */
462	if (ids->n_ids == 1)
463		return ids->ids[0];
464
465	hash = skb_get_hash(skb);
466	ids_index = hash - ids->n_ids * reciprocal_divide(hash, ids->rn_ids);
467	return ids->ids[ids_index];
468}
469
470/**
471 *	ovs_vport_receive - pass up received packet to the datapath for processing
472 *
473 * @vport: vport that received the packet
474 * @skb: skb that was received
475 * @tun_info: tunnel (if any) that carried packet
476 *
477 * Must be called with rcu_read_lock.  The packet cannot be shared and
478 * skb->data should point to the Ethernet header.
479 */
480int ovs_vport_receive(struct vport *vport, struct sk_buff *skb,
481		      const struct ip_tunnel_info *tun_info)
482{
483	struct sw_flow_key key;
484	int error;
485
486	OVS_CB(skb)->input_vport = vport;
487	OVS_CB(skb)->mru = 0;
488	OVS_CB(skb)->cutlen = 0;
489	if (unlikely(dev_net(skb->dev) != ovs_dp_get_net(vport->dp))) {
490		u32 mark;
491
492		mark = skb->mark;
493		skb_scrub_packet(skb, true);
494		skb->mark = mark;
495		tun_info = NULL;
496	}
497
498	/* Extract flow from 'skb' into 'key'. */
499	error = ovs_flow_key_extract(tun_info, skb, &key);
500	if (unlikely(error)) {
501		kfree_skb(skb);
502		return error;
503	}
504	ovs_dp_process_packet(skb, &key);
505	return 0;
506}
507
508static int packet_length(const struct sk_buff *skb,
509			 struct net_device *dev)
510{
511	int length = skb->len - dev->hard_header_len;
512
513	if (!skb_vlan_tag_present(skb) &&
514	    eth_type_vlan(skb->protocol))
515		length -= VLAN_HLEN;
516
517	/* Don't subtract for multiple VLAN tags. Most (all?) drivers allow
518	 * (ETH_LEN + VLAN_HLEN) in addition to the mtu value, but almost none
519	 * account for 802.1ad. e.g. is_skb_forwardable().
520	 */
521
522	return length > 0 ? length : 0;
523}
524
525void ovs_vport_send(struct vport *vport, struct sk_buff *skb, u8 mac_proto)
526{
527	int mtu = vport->dev->mtu;
528
529	switch (vport->dev->type) {
530	case ARPHRD_NONE:
531		if (mac_proto == MAC_PROTO_ETHERNET) {
532			skb_reset_network_header(skb);
533			skb_reset_mac_len(skb);
534			skb->protocol = htons(ETH_P_TEB);
535		} else if (mac_proto != MAC_PROTO_NONE) {
536			WARN_ON_ONCE(1);
537			goto drop;
538		}
539		break;
540	case ARPHRD_ETHER:
541		if (mac_proto != MAC_PROTO_ETHERNET)
542			goto drop;
543		break;
544	default:
545		goto drop;
546	}
547
548	if (unlikely(packet_length(skb, vport->dev) > mtu &&
549		     !skb_is_gso(skb))) {
 
 
 
550		vport->dev->stats.tx_errors++;
551		if (vport->dev->flags & IFF_UP)
552			net_warn_ratelimited("%s: dropped over-mtu packet: "
553					     "%d > %d\n", vport->dev->name,
554					     packet_length(skb, vport->dev),
555					     mtu);
556		goto drop;
557	}
558
559	skb->dev = vport->dev;
560	skb_clear_tstamp(skb);
561	vport->ops->send(skb);
562	return;
563
564drop:
565	kfree_skb(skb);
566}