2 * Copyright (c) 2007-2013 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
21 #include <linux/uaccess.h>
22 #include <linux/netdevice.h>
23 #include <linux/etherdevice.h>
24 #include <linux/if_ether.h>
25 #include <linux/if_vlan.h>
26 #include <net/llc_pdu.h>
27 #include <linux/kernel.h>
28 #include <linux/jhash.h>
29 #include <linux/jiffies.h>
30 #include <linux/llc.h>
31 #include <linux/module.h>
33 #include <linux/rcupdate.h>
34 #include <linux/if_arp.h>
36 #include <linux/ipv6.h>
37 #include <linux/sctp.h>
38 #include <linux/tcp.h>
39 #include <linux/udp.h>
40 #include <linux/icmp.h>
41 #include <linux/icmpv6.h>
42 #include <linux/rculist.h>
45 #include <net/ndisc.h>
50 #define TBL_MIN_BUCKETS 1024
51 #define REHASH_INTERVAL (10 * 60 * HZ)
53 static struct kmem_cache *flow_cache;
55 static u16 range_n_bytes(const struct sw_flow_key_range *range)
57 return range->end - range->start;
60 void ovs_flow_mask_key(struct sw_flow_key *dst, const struct sw_flow_key *src,
61 const struct sw_flow_mask *mask)
63 const long *m = (long *)((u8 *)&mask->key + mask->range.start);
64 const long *s = (long *)((u8 *)src + mask->range.start);
65 long *d = (long *)((u8 *)dst + mask->range.start);
68 /* The memory outside of the 'mask->range' are not set since
69 * further operations on 'dst' only uses contents within
72 for (i = 0; i < range_n_bytes(&mask->range); i += sizeof(long))
76 struct sw_flow *ovs_flow_alloc(void)
80 flow = kmem_cache_alloc(flow_cache, GFP_KERNEL);
82 return ERR_PTR(-ENOMEM);
84 spin_lock_init(&flow->lock);
91 int ovs_flow_tbl_count(struct flow_table *table)
96 static struct flex_array *alloc_buckets(unsigned int n_buckets)
98 struct flex_array *buckets;
101 buckets = flex_array_alloc(sizeof(struct hlist_head),
102 n_buckets, GFP_KERNEL);
106 err = flex_array_prealloc(buckets, 0, n_buckets, GFP_KERNEL);
108 flex_array_free(buckets);
112 for (i = 0; i < n_buckets; i++)
113 INIT_HLIST_HEAD((struct hlist_head *)
114 flex_array_get(buckets, i));
119 static void flow_free(struct sw_flow *flow)
121 kfree((struct sf_flow_acts __force *)flow->sf_acts);
122 kmem_cache_free(flow_cache, flow);
125 static void rcu_free_flow_callback(struct rcu_head *rcu)
127 struct sw_flow *flow = container_of(rcu, struct sw_flow, rcu);
132 void ovs_flow_free(struct sw_flow *flow, bool deferred)
137 ovs_sw_flow_mask_del_ref(flow->mask, deferred);
140 call_rcu(&flow->rcu, rcu_free_flow_callback);
145 static void free_buckets(struct flex_array *buckets)
147 flex_array_free(buckets);
150 static void __table_instance_destroy(struct table_instance *ti)
157 for (i = 0; i < ti->n_buckets; i++) {
158 struct sw_flow *flow;
159 struct hlist_head *head = flex_array_get(ti->buckets, i);
160 struct hlist_node *n;
161 int ver = ti->node_ver;
163 hlist_for_each_entry_safe(flow, n, head, hash_node[ver]) {
164 hlist_del(&flow->hash_node[ver]);
165 ovs_flow_free(flow, false);
170 free_buckets(ti->buckets);
174 static struct table_instance *table_instance_alloc(int new_size)
176 struct table_instance *ti = kmalloc(sizeof(*ti), GFP_KERNEL);
181 ti->buckets = alloc_buckets(new_size);
187 ti->n_buckets = new_size;
189 ti->keep_flows = false;
190 get_random_bytes(&ti->hash_seed, sizeof(u32));
195 int ovs_flow_tbl_init(struct flow_table *table)
197 struct table_instance *ti;
199 ti = table_instance_alloc(TBL_MIN_BUCKETS);
204 rcu_assign_pointer(table->ti, ti);
205 INIT_LIST_HEAD(&table->mask_list);
206 table->last_rehash = jiffies;
211 static void flow_tbl_destroy_rcu_cb(struct rcu_head *rcu)
213 struct table_instance *ti = container_of(rcu, struct table_instance, rcu);
215 __table_instance_destroy(ti);
218 static void table_instance_destroy(struct table_instance *ti, bool deferred)
224 call_rcu(&ti->rcu, flow_tbl_destroy_rcu_cb);
226 __table_instance_destroy(ti);
229 void ovs_flow_tbl_destroy(struct flow_table *table, bool deferred)
231 struct table_instance *ti = ovsl_dereference(table->ti);
233 table_instance_destroy(ti, deferred);
236 struct sw_flow *ovs_flow_tbl_dump_next(struct table_instance *ti,
237 u32 *bucket, u32 *last)
239 struct sw_flow *flow;
240 struct hlist_head *head;
245 while (*bucket < ti->n_buckets) {
247 head = flex_array_get(ti->buckets, *bucket);
248 hlist_for_each_entry_rcu(flow, head, hash_node[ver]) {
263 static struct hlist_head *find_bucket(struct table_instance *ti, u32 hash)
265 hash = jhash_1word(hash, ti->hash_seed);
266 return flex_array_get(ti->buckets,
267 (hash & (ti->n_buckets - 1)));
270 static void table_instance_insert(struct table_instance *ti, struct sw_flow *flow)
272 struct hlist_head *head;
274 head = find_bucket(ti, flow->hash);
275 hlist_add_head_rcu(&flow->hash_node[ti->node_ver], head);
278 static void flow_table_copy_flows(struct table_instance *old,
279 struct table_instance *new)
284 old_ver = old->node_ver;
285 new->node_ver = !old_ver;
287 /* Insert in new table. */
288 for (i = 0; i < old->n_buckets; i++) {
289 struct sw_flow *flow;
290 struct hlist_head *head;
292 head = flex_array_get(old->buckets, i);
294 hlist_for_each_entry(flow, head, hash_node[old_ver])
295 table_instance_insert(new, flow);
298 old->keep_flows = true;
301 static struct table_instance *table_instance_rehash(struct table_instance *ti,
304 struct table_instance *new_ti;
306 new_ti = table_instance_alloc(n_buckets);
308 return ERR_PTR(-ENOMEM);
310 flow_table_copy_flows(ti, new_ti);
315 int ovs_flow_tbl_flush(struct flow_table *flow_table)
317 struct table_instance *old_ti;
318 struct table_instance *new_ti;
320 old_ti = ovsl_dereference(flow_table->ti);
321 new_ti = table_instance_alloc(TBL_MIN_BUCKETS);
325 rcu_assign_pointer(flow_table->ti, new_ti);
326 flow_table->last_rehash = jiffies;
327 flow_table->count = 0;
329 table_instance_destroy(old_ti, true);
333 static u32 flow_hash(const struct sw_flow_key *key, int key_start,
336 u32 *hash_key = (u32 *)((u8 *)key + key_start);
337 int hash_u32s = (key_end - key_start) >> 2;
339 /* Make sure number of hash bytes are multiple of u32. */
340 BUILD_BUG_ON(sizeof(long) % sizeof(u32));
342 return jhash2(hash_key, hash_u32s, 0);
345 static int flow_key_start(const struct sw_flow_key *key)
347 if (key->tun_key.ipv4_dst)
350 return rounddown(offsetof(struct sw_flow_key, phy),
354 static bool cmp_key(const struct sw_flow_key *key1,
355 const struct sw_flow_key *key2,
356 int key_start, int key_end)
358 const long *cp1 = (long *)((u8 *)key1 + key_start);
359 const long *cp2 = (long *)((u8 *)key2 + key_start);
363 for (i = key_start; i < key_end; i += sizeof(long))
364 diffs |= *cp1++ ^ *cp2++;
369 static bool flow_cmp_masked_key(const struct sw_flow *flow,
370 const struct sw_flow_key *key,
371 int key_start, int key_end)
373 return cmp_key(&flow->key, key, key_start, key_end);
376 bool ovs_flow_cmp_unmasked_key(const struct sw_flow *flow,
377 struct sw_flow_match *match)
379 struct sw_flow_key *key = match->key;
380 int key_start = flow_key_start(key);
381 int key_end = match->range.end;
383 return cmp_key(&flow->unmasked_key, key, key_start, key_end);
386 static struct sw_flow *masked_flow_lookup(struct table_instance *ti,
387 const struct sw_flow_key *unmasked,
388 struct sw_flow_mask *mask)
390 struct sw_flow *flow;
391 struct hlist_head *head;
392 int key_start = mask->range.start;
393 int key_end = mask->range.end;
395 struct sw_flow_key masked_key;
397 ovs_flow_mask_key(&masked_key, unmasked, mask);
398 hash = flow_hash(&masked_key, key_start, key_end);
399 head = find_bucket(ti, hash);
400 hlist_for_each_entry_rcu(flow, head, hash_node[ti->node_ver]) {
401 if (flow->mask == mask &&
402 flow_cmp_masked_key(flow, &masked_key,
409 struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *tbl,
410 const struct sw_flow_key *key)
412 struct table_instance *ti = rcu_dereference(tbl->ti);
413 struct sw_flow_mask *mask;
414 struct sw_flow *flow;
416 list_for_each_entry_rcu(mask, &tbl->mask_list, list) {
417 flow = masked_flow_lookup(ti, key, mask);
418 if (flow) /* Found */
424 static struct table_instance *table_instance_expand(struct table_instance *ti)
426 return table_instance_rehash(ti, ti->n_buckets * 2);
429 void ovs_flow_tbl_insert(struct flow_table *table, struct sw_flow *flow)
431 struct table_instance *ti = NULL;
432 struct table_instance *new_ti = NULL;
434 ti = ovsl_dereference(table->ti);
436 /* Expand table, if necessary, to make room. */
437 if (table->count > ti->n_buckets)
438 new_ti = table_instance_expand(ti);
439 else if (time_after(jiffies, table->last_rehash + REHASH_INTERVAL))
440 new_ti = table_instance_rehash(ti, ti->n_buckets);
442 if (new_ti && !IS_ERR(new_ti)) {
443 rcu_assign_pointer(table->ti, new_ti);
444 ovs_flow_tbl_destroy(table, true);
445 ti = ovsl_dereference(table->ti);
446 table->last_rehash = jiffies;
449 flow->hash = flow_hash(&flow->key, flow->mask->range.start,
450 flow->mask->range.end);
451 table_instance_insert(ti, flow);
455 void ovs_flow_tbl_remove(struct flow_table *table, struct sw_flow *flow)
457 struct table_instance *ti = ovsl_dereference(table->ti);
459 BUG_ON(table->count == 0);
460 hlist_del_rcu(&flow->hash_node[ti->node_ver]);
464 struct sw_flow_mask *ovs_sw_flow_mask_alloc(void)
466 struct sw_flow_mask *mask;
468 mask = kmalloc(sizeof(*mask), GFP_KERNEL);
475 void ovs_sw_flow_mask_add_ref(struct sw_flow_mask *mask)
480 static void rcu_free_sw_flow_mask_cb(struct rcu_head *rcu)
482 struct sw_flow_mask *mask = container_of(rcu, struct sw_flow_mask, rcu);
487 void ovs_sw_flow_mask_del_ref(struct sw_flow_mask *mask, bool deferred)
492 BUG_ON(!mask->ref_count);
495 if (!mask->ref_count) {
496 list_del_rcu(&mask->list);
498 call_rcu(&mask->rcu, rcu_free_sw_flow_mask_cb);
504 static bool mask_equal(const struct sw_flow_mask *a,
505 const struct sw_flow_mask *b)
507 u8 *a_ = (u8 *)&a->key + a->range.start;
508 u8 *b_ = (u8 *)&b->key + b->range.start;
510 return (a->range.end == b->range.end)
511 && (a->range.start == b->range.start)
512 && (memcmp(a_, b_, range_n_bytes(&a->range)) == 0);
515 struct sw_flow_mask *ovs_sw_flow_mask_find(const struct flow_table *tbl,
516 const struct sw_flow_mask *mask)
518 struct list_head *ml;
520 list_for_each(ml, &tbl->mask_list) {
521 struct sw_flow_mask *m;
522 m = container_of(ml, struct sw_flow_mask, list);
523 if (mask_equal(mask, m))
531 * add a new mask into the mask list.
532 * The caller needs to make sure that 'mask' is not the same
533 * as any masks that are already on the list.
535 void ovs_sw_flow_mask_insert(struct flow_table *tbl, struct sw_flow_mask *mask)
537 list_add_rcu(&mask->list, &tbl->mask_list);
540 /* Initializes the flow module.
541 * Returns zero if successful or a negative error code. */
542 int ovs_flow_init(void)
544 BUILD_BUG_ON(__alignof__(struct sw_flow_key) % __alignof__(long));
545 BUILD_BUG_ON(sizeof(struct sw_flow_key) % sizeof(long));
547 flow_cache = kmem_cache_create("sw_flow", sizeof(struct sw_flow), 0,
549 if (flow_cache == NULL)
555 /* Uninitializes the flow module. */
556 void ovs_flow_exit(void)
558 kmem_cache_destroy(flow_cache);