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 static void rcu_free_sw_flow_mask_cb(struct rcu_head *rcu)
134 struct sw_flow_mask *mask = container_of(rcu, struct sw_flow_mask, rcu);
139 static void flow_mask_del_ref(struct sw_flow_mask *mask, bool deferred)
144 BUG_ON(!mask->ref_count);
147 if (!mask->ref_count) {
148 list_del_rcu(&mask->list);
150 call_rcu(&mask->rcu, rcu_free_sw_flow_mask_cb);
156 void ovs_flow_free(struct sw_flow *flow, bool deferred)
161 flow_mask_del_ref(flow->mask, deferred);
164 call_rcu(&flow->rcu, rcu_free_flow_callback);
169 static void free_buckets(struct flex_array *buckets)
171 flex_array_free(buckets);
174 static void __table_instance_destroy(struct table_instance *ti)
181 for (i = 0; i < ti->n_buckets; i++) {
182 struct sw_flow *flow;
183 struct hlist_head *head = flex_array_get(ti->buckets, i);
184 struct hlist_node *n;
185 int ver = ti->node_ver;
187 hlist_for_each_entry_safe(flow, n, head, hash_node[ver]) {
188 hlist_del(&flow->hash_node[ver]);
189 ovs_flow_free(flow, false);
194 free_buckets(ti->buckets);
198 static struct table_instance *table_instance_alloc(int new_size)
200 struct table_instance *ti = kmalloc(sizeof(*ti), GFP_KERNEL);
205 ti->buckets = alloc_buckets(new_size);
211 ti->n_buckets = new_size;
213 ti->keep_flows = false;
214 get_random_bytes(&ti->hash_seed, sizeof(u32));
219 int ovs_flow_tbl_init(struct flow_table *table)
221 struct table_instance *ti;
223 ti = table_instance_alloc(TBL_MIN_BUCKETS);
228 rcu_assign_pointer(table->ti, ti);
229 INIT_LIST_HEAD(&table->mask_list);
230 table->last_rehash = jiffies;
235 static void flow_tbl_destroy_rcu_cb(struct rcu_head *rcu)
237 struct table_instance *ti = container_of(rcu, struct table_instance, rcu);
239 __table_instance_destroy(ti);
242 static void table_instance_destroy(struct table_instance *ti, bool deferred)
248 call_rcu(&ti->rcu, flow_tbl_destroy_rcu_cb);
250 __table_instance_destroy(ti);
253 void ovs_flow_tbl_destroy(struct flow_table *table)
255 struct table_instance *ti = ovsl_dereference(table->ti);
257 table_instance_destroy(ti, false);
260 struct sw_flow *ovs_flow_tbl_dump_next(struct table_instance *ti,
261 u32 *bucket, u32 *last)
263 struct sw_flow *flow;
264 struct hlist_head *head;
269 while (*bucket < ti->n_buckets) {
271 head = flex_array_get(ti->buckets, *bucket);
272 hlist_for_each_entry_rcu(flow, head, hash_node[ver]) {
287 static struct hlist_head *find_bucket(struct table_instance *ti, u32 hash)
289 hash = jhash_1word(hash, ti->hash_seed);
290 return flex_array_get(ti->buckets,
291 (hash & (ti->n_buckets - 1)));
294 static void table_instance_insert(struct table_instance *ti, struct sw_flow *flow)
296 struct hlist_head *head;
298 head = find_bucket(ti, flow->hash);
299 hlist_add_head_rcu(&flow->hash_node[ti->node_ver], head);
302 static void flow_table_copy_flows(struct table_instance *old,
303 struct table_instance *new)
308 old_ver = old->node_ver;
309 new->node_ver = !old_ver;
311 /* Insert in new table. */
312 for (i = 0; i < old->n_buckets; i++) {
313 struct sw_flow *flow;
314 struct hlist_head *head;
316 head = flex_array_get(old->buckets, i);
318 hlist_for_each_entry(flow, head, hash_node[old_ver])
319 table_instance_insert(new, flow);
322 old->keep_flows = true;
325 static struct table_instance *table_instance_rehash(struct table_instance *ti,
328 struct table_instance *new_ti;
330 new_ti = table_instance_alloc(n_buckets);
334 flow_table_copy_flows(ti, new_ti);
339 int ovs_flow_tbl_flush(struct flow_table *flow_table)
341 struct table_instance *old_ti;
342 struct table_instance *new_ti;
344 old_ti = ovsl_dereference(flow_table->ti);
345 new_ti = table_instance_alloc(TBL_MIN_BUCKETS);
349 rcu_assign_pointer(flow_table->ti, new_ti);
350 flow_table->last_rehash = jiffies;
351 flow_table->count = 0;
353 table_instance_destroy(old_ti, true);
357 static u32 flow_hash(const struct sw_flow_key *key, int key_start,
360 u32 *hash_key = (u32 *)((u8 *)key + key_start);
361 int hash_u32s = (key_end - key_start) >> 2;
363 /* Make sure number of hash bytes are multiple of u32. */
364 BUILD_BUG_ON(sizeof(long) % sizeof(u32));
366 return jhash2(hash_key, hash_u32s, 0);
369 static int flow_key_start(const struct sw_flow_key *key)
371 if (key->tun_key.ipv4_dst)
374 return rounddown(offsetof(struct sw_flow_key, phy),
378 static bool cmp_key(const struct sw_flow_key *key1,
379 const struct sw_flow_key *key2,
380 int key_start, int key_end)
382 const long *cp1 = (long *)((u8 *)key1 + key_start);
383 const long *cp2 = (long *)((u8 *)key2 + key_start);
387 for (i = key_start; i < key_end; i += sizeof(long))
388 diffs |= *cp1++ ^ *cp2++;
393 static bool flow_cmp_masked_key(const struct sw_flow *flow,
394 const struct sw_flow_key *key,
395 int key_start, int key_end)
397 return cmp_key(&flow->key, key, key_start, key_end);
400 bool ovs_flow_cmp_unmasked_key(const struct sw_flow *flow,
401 struct sw_flow_match *match)
403 struct sw_flow_key *key = match->key;
404 int key_start = flow_key_start(key);
405 int key_end = match->range.end;
407 return cmp_key(&flow->unmasked_key, key, key_start, key_end);
410 static struct sw_flow *masked_flow_lookup(struct table_instance *ti,
411 const struct sw_flow_key *unmasked,
412 struct sw_flow_mask *mask)
414 struct sw_flow *flow;
415 struct hlist_head *head;
416 int key_start = mask->range.start;
417 int key_end = mask->range.end;
419 struct sw_flow_key masked_key;
421 ovs_flow_mask_key(&masked_key, unmasked, mask);
422 hash = flow_hash(&masked_key, key_start, key_end);
423 head = find_bucket(ti, hash);
424 hlist_for_each_entry_rcu(flow, head, hash_node[ti->node_ver]) {
425 if (flow->mask == mask &&
426 flow_cmp_masked_key(flow, &masked_key,
433 struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *tbl,
434 const struct sw_flow_key *key)
436 struct table_instance *ti = rcu_dereference(tbl->ti);
437 struct sw_flow_mask *mask;
438 struct sw_flow *flow;
440 list_for_each_entry_rcu(mask, &tbl->mask_list, list) {
441 flow = masked_flow_lookup(ti, key, mask);
442 if (flow) /* Found */
448 static struct table_instance *table_instance_expand(struct table_instance *ti)
450 return table_instance_rehash(ti, ti->n_buckets * 2);
453 void ovs_flow_tbl_remove(struct flow_table *table, struct sw_flow *flow)
455 struct table_instance *ti = ovsl_dereference(table->ti);
457 BUG_ON(table->count == 0);
458 hlist_del_rcu(&flow->hash_node[ti->node_ver]);
462 static struct sw_flow_mask *mask_alloc(void)
464 struct sw_flow_mask *mask;
466 mask = kmalloc(sizeof(*mask), GFP_KERNEL);
473 static void mask_add_ref(struct sw_flow_mask *mask)
478 static bool mask_equal(const struct sw_flow_mask *a,
479 const struct sw_flow_mask *b)
481 u8 *a_ = (u8 *)&a->key + a->range.start;
482 u8 *b_ = (u8 *)&b->key + b->range.start;
484 return (a->range.end == b->range.end)
485 && (a->range.start == b->range.start)
486 && (memcmp(a_, b_, range_n_bytes(&a->range)) == 0);
489 static struct sw_flow_mask *flow_mask_find(const struct flow_table *tbl,
490 const struct sw_flow_mask *mask)
492 struct list_head *ml;
494 list_for_each(ml, &tbl->mask_list) {
495 struct sw_flow_mask *m;
496 m = container_of(ml, struct sw_flow_mask, list);
497 if (mask_equal(mask, m))
505 * add a new mask into the mask list.
506 * The caller needs to make sure that 'mask' is not the same
507 * as any masks that are already on the list.
509 static int flow_mask_insert(struct flow_table *tbl, struct sw_flow *flow,
510 struct sw_flow_mask *new)
512 struct sw_flow_mask *mask;
513 mask = flow_mask_find(tbl, new);
515 /* Allocate a new mask if none exsits. */
519 mask->key = new->key;
520 mask->range = new->range;
521 list_add_rcu(&mask->list, &tbl->mask_list);
529 int ovs_flow_tbl_insert(struct flow_table *table, struct sw_flow *flow,
530 struct sw_flow_mask *mask)
532 struct table_instance *new_ti = NULL;
533 struct table_instance *ti;
536 err = flow_mask_insert(table, flow, mask);
540 flow->hash = flow_hash(&flow->key, flow->mask->range.start,
541 flow->mask->range.end);
542 ti = ovsl_dereference(table->ti);
543 table_instance_insert(ti, flow);
546 /* Expand table, if necessary, to make room. */
547 if (table->count > ti->n_buckets)
548 new_ti = table_instance_expand(ti);
549 else if (time_after(jiffies, table->last_rehash + REHASH_INTERVAL))
550 new_ti = table_instance_rehash(ti, ti->n_buckets);
553 rcu_assign_pointer(table->ti, new_ti);
554 table_instance_destroy(ti, true);
555 table->last_rehash = jiffies;
560 /* Initializes the flow module.
561 * Returns zero if successful or a negative error code. */
562 int ovs_flow_init(void)
564 BUILD_BUG_ON(__alignof__(struct sw_flow_key) % __alignof__(long));
565 BUILD_BUG_ON(sizeof(struct sw_flow_key) % sizeof(long));
567 flow_cache = kmem_cache_create("sw_flow", sizeof(struct sw_flow), 0,
569 if (flow_cache == NULL)
575 /* Uninitializes the flow module. */
576 void ovs_flow_exit(void)
578 kmem_cache_destroy(flow_cache);