X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;ds=sidebyside;f=datapath%2Fflow.c;h=823c6b5bc0f322c34e6421e56ea40c7d46cb6689;hb=816fd533f85923c03cf8d9d6450bd9a0845d5160;hp=78f737abda2d087a2cc505e1bd5b855b53a19900;hpb=850b6b3b9f8c38b42e315c2c07d232a33b82da3e;p=openvswitch diff --git a/datapath/flow.c b/datapath/flow.c index 78f737ab..823c6b5b 100644 --- a/datapath/flow.c +++ b/datapath/flow.c @@ -32,7 +32,6 @@ #include #include #include -#include #include #include #include @@ -47,7 +46,6 @@ #include "vlan.h" static struct kmem_cache *flow_cache; -static unsigned int hash_seed __read_mostly; static int check_header(struct sk_buff *skb, int len) { @@ -128,66 +126,6 @@ u64 ovs_flow_used_time(unsigned long flow_jiffies) (offsetof(struct sw_flow_key, field) + \ FIELD_SIZEOF(struct sw_flow_key, field)) -/** - * skip_exthdr - skip any IPv6 extension headers - * @skb: skbuff to parse - * @start: offset of first extension header - * @nexthdrp: Initially, points to the type of the extension header at @start. - * This function updates it to point to the extension header at the final - * offset. - * @frag: Points to the @frag member in a &struct sw_flow_key. This - * function sets an appropriate %OVS_FRAG_TYPE_* value. - * - * This is based on ipv6_skip_exthdr() but adds the updates to *@frag. - * - * When there is more than one fragment header, this version reports whether - * the final fragment header that it examines is a first fragment. - * - * Returns the final payload offset, or -1 on error. - */ -static int skip_exthdr(const struct sk_buff *skb, int start, u8 *nexthdrp, - u8 *frag) -{ - u8 nexthdr = *nexthdrp; - - while (ipv6_ext_hdr(nexthdr)) { - struct ipv6_opt_hdr _hdr, *hp; - int hdrlen; - - if (nexthdr == NEXTHDR_NONE) - return -1; - hp = skb_header_pointer(skb, start, sizeof(_hdr), &_hdr); - if (hp == NULL) - return -1; - if (nexthdr == NEXTHDR_FRAGMENT) { - __be16 _frag_off, *fp; - fp = skb_header_pointer(skb, - start+offsetof(struct frag_hdr, - frag_off), - sizeof(_frag_off), - &_frag_off); - if (fp == NULL) - return -1; - - if (ntohs(*fp) & ~0x7) { - *frag = OVS_FRAG_TYPE_LATER; - break; - } - *frag = OVS_FRAG_TYPE_FIRST; - hdrlen = 8; - } else if (nexthdr == NEXTHDR_AUTH) - hdrlen = (hp->hdrlen+2)<<2; - else - hdrlen = ipv6_optlen(hp); - - nexthdr = hp->nexthdr; - start += hdrlen; - } - - *nexthdrp = nexthdr; - return start; -} - static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key, int *key_lenp) { @@ -196,6 +134,7 @@ static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key, int payload_ofs; struct ipv6hdr *nh; uint8_t nexthdr; + __be16 frag_off; int err; *key_lenp = SW_FLOW_KEY_OFFSET(ipv6.label); @@ -212,13 +151,20 @@ static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key, key->ip.tos = ipv6_get_dsfield(nh); key->ip.ttl = nh->hop_limit; key->ipv6.label = *(__be32 *)nh & htonl(IPV6_FLOWINFO_FLOWLABEL); - ipv6_addr_copy(&key->ipv6.addr.src, &nh->saddr); - ipv6_addr_copy(&key->ipv6.addr.dst, &nh->daddr); + key->ipv6.addr.src = nh->saddr; + key->ipv6.addr.dst = nh->daddr; - payload_ofs = skip_exthdr(skb, payload_ofs, &nexthdr, &key->ip.frag); + payload_ofs = ipv6_skip_exthdr(skb, payload_ofs, &nexthdr, &frag_off); if (unlikely(payload_ofs < 0)) return -EINVAL; + if (frag_off) { + if (frag_off & htons(~0x7)) + key->ip.frag = OVS_FRAG_TYPE_LATER; + else + key->ip.frag = OVS_FRAG_TYPE_FIRST; + } + nh_len = payload_ofs - nh_ofs; skb_set_transport_header(skb, nh_ofs + nh_len); key->ip.proto = nexthdr; @@ -290,6 +236,7 @@ struct sw_flow *ovs_flow_alloc(void) static struct hlist_head *find_bucket(struct flow_table *table, u32 hash) { + hash = jhash_1word(hash, table->hash_seed); return flex_array_get(table->buckets, (hash & (table->n_buckets - 1))); } @@ -337,6 +284,9 @@ struct flow_table *ovs_flow_tbl_alloc(int new_size) } table->n_buckets = new_size; table->count = 0; + table->node_ver = 0; + table->keep_flows = false; + get_random_bytes(&table->hash_seed, sizeof(u32)); return table; } @@ -354,17 +304,22 @@ void ovs_flow_tbl_destroy(struct flow_table *table) if (!table) return; + if (table->keep_flows) + goto skip_flows; + for (i = 0; i < table->n_buckets; i++) { struct sw_flow *flow; struct hlist_head *head = flex_array_get(table->buckets, i); struct hlist_node *node, *n; + int ver = table->node_ver; - hlist_for_each_entry_safe(flow, node, n, head, hash_node) { - hlist_del_init_rcu(&flow->hash_node); + hlist_for_each_entry_safe(flow, node, n, head, hash_node[ver]) { + hlist_del_rcu(&flow->hash_node[ver]); flow_free(flow); } } +skip_flows: free_buckets(table->buckets); kfree(table); } @@ -389,12 +344,14 @@ struct sw_flow *ovs_flow_tbl_next(struct flow_table *table, u32 *bucket, u32 *la struct sw_flow *flow; struct hlist_head *head; struct hlist_node *n; + int ver; int i; + ver = table->node_ver; while (*bucket < table->n_buckets) { i = 0; head = flex_array_get(table->buckets, *bucket); - hlist_for_each_entry_rcu(flow, n, head, hash_node) { + hlist_for_each_entry_rcu(flow, n, head, hash_node[ver]) { if (i < *last) { i++; continue; @@ -409,32 +366,51 @@ struct sw_flow *ovs_flow_tbl_next(struct flow_table *table, u32 *bucket, u32 *la return NULL; } -struct flow_table *ovs_flow_tbl_expand(struct flow_table *table) +static void flow_table_copy_flows(struct flow_table *old, struct flow_table *new) { - struct flow_table *new_table; - int n_buckets = table->n_buckets * 2; + int old_ver; int i; - new_table = ovs_flow_tbl_alloc(n_buckets); - if (!new_table) - return ERR_PTR(-ENOMEM); + old_ver = old->node_ver; + new->node_ver = !old_ver; - for (i = 0; i < table->n_buckets; i++) { + /* Insert in new table. */ + for (i = 0; i < old->n_buckets; i++) { struct sw_flow *flow; struct hlist_head *head; - struct hlist_node *n, *pos; + struct hlist_node *n; - head = flex_array_get(table->buckets, i); + head = flex_array_get(old->buckets, i); - hlist_for_each_entry_safe(flow, n, pos, head, hash_node) { - hlist_del_init_rcu(&flow->hash_node); - ovs_flow_tbl_insert(new_table, flow); - } + hlist_for_each_entry(flow, n, head, hash_node[old_ver]) + ovs_flow_tbl_insert(new, flow); } + old->keep_flows = true; +} + +static struct flow_table *__flow_tbl_rehash(struct flow_table *table, int n_buckets) +{ + struct flow_table *new_table; + + new_table = ovs_flow_tbl_alloc(n_buckets); + if (!new_table) + return ERR_PTR(-ENOMEM); + + flow_table_copy_flows(table, new_table); return new_table; } +struct flow_table *ovs_flow_tbl_rehash(struct flow_table *table) +{ + return __flow_tbl_rehash(table, table->n_buckets); +} + +struct flow_table *ovs_flow_tbl_expand(struct flow_table *table) +{ + return __flow_tbl_rehash(table, table->n_buckets * 2); +} + /* RCU callback used by ovs_flow_deferred_free. */ static void rcu_free_flow_callback(struct rcu_head *rcu) { @@ -572,7 +548,7 @@ static int parse_icmpv6(struct sk_buff *skb, struct sw_flow_key *key, } nd = (struct nd_msg *)skb_transport_header(skb); - ipv6_addr_copy(&key->ipv6.nd.target, &nd->target); + key->ipv6.nd.target = nd->target; key_len = SW_FLOW_KEY_OFFSET(ipv6.nd); icmp_len -= sizeof(*nd); @@ -813,7 +789,7 @@ out: u32 ovs_flow_hash(const struct sw_flow_key *key, int key_len) { - return jhash2((u32 *)key, DIV_ROUND_UP(key_len, sizeof(u32)), hash_seed); + return jhash2((u32 *)key, DIV_ROUND_UP(key_len, sizeof(u32)), 0); } struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *table, @@ -827,7 +803,7 @@ struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *table, hash = ovs_flow_hash(key, key_len); head = find_bucket(table, hash); - hlist_for_each_entry_rcu(flow, n, head, hash_node) { + hlist_for_each_entry_rcu(flow, n, head, hash_node[table->node_ver]) { if (flow->hash == hash && !memcmp(&flow->key, key, key_len)) { @@ -842,17 +818,15 @@ void ovs_flow_tbl_insert(struct flow_table *table, struct sw_flow *flow) struct hlist_head *head; head = find_bucket(table, flow->hash); - hlist_add_head_rcu(&flow->hash_node, head); + hlist_add_head_rcu(&flow->hash_node[table->node_ver], head); table->count++; } void ovs_flow_tbl_remove(struct flow_table *table, struct sw_flow *flow) { - if (!hlist_unhashed(&flow->hash_node)) { - hlist_del_init_rcu(&flow->hash_node); - table->count--; - BUG_ON(table->count < 0); - } + hlist_del_rcu(&flow->hash_node[table->node_ver]); + table->count--; + BUG_ON(table->count < 0); } /* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */ @@ -1397,8 +1371,6 @@ int ovs_flow_init(void) if (flow_cache == NULL) return -ENOMEM; - get_random_bytes(&hash_seed, sizeof(hash_seed)); - return 0; }