X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=datapath%2Ftunnel.c;h=08906e4ca1ba4b02561311af5638febfd4cbf087;hb=c19e653509deb3f1002c4ed99714e16549ec74e9;hp=239e604c7dbf1e0d722f7cf292530d83b3ad05b7;hpb=ad9197112ae58f338491e6344a9f66073e4bce95;p=sliver-openvswitch.git diff --git a/datapath/tunnel.c b/datapath/tunnel.c index 239e604c7..08906e4ca 100644 --- a/datapath/tunnel.c +++ b/datapath/tunnel.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2010 Nicira Networks. + * Copyright (c) 2010, 2011 Nicira Networks. * Distributed under the terms of the GNU GPL version 2. * * Significant portions of this file may be copied from parts of the Linux @@ -99,6 +99,15 @@ static inline struct tnl_vport *tnl_vport_table_cast(const struct tbl_node *node return container_of(node, struct tnl_vport, tbl_node); } +/* This is analogous to rtnl_dereference for the tunnel cache. It checks that + * cache_lock is held, so it is only for update side code. + */ +static inline struct tnl_cache *cache_dereference(struct tnl_vport *tnl_vport) +{ + return rcu_dereference_protected(tnl_vport->cache, + lockdep_is_held(&tnl_vport->cache_lock)); +} + static inline void schedule_cache_cleaner(void) { schedule_delayed_work(&cache_cleaner_wq, CACHE_CLEANER_INTERVAL); @@ -142,7 +151,7 @@ static void assign_cache_rcu(struct vport *vport, struct tnl_cache *new_cache) struct tnl_vport *tnl_vport = tnl_vport_priv(vport); struct tnl_cache *old_cache; - old_cache = tnl_vport->cache; + old_cache = cache_dereference(tnl_vport); rcu_assign_pointer(tnl_vport->cache, new_cache); if (old_cache) @@ -151,13 +160,13 @@ static void assign_cache_rcu(struct vport *vport, struct tnl_cache *new_cache) static unsigned int *find_port_pool(const struct tnl_mutable_config *mutable) { - if (mutable->port_config.flags & TNL_F_IN_KEY_MATCH) { - if (mutable->port_config.saddr) + if (mutable->flags & TNL_F_IN_KEY_MATCH) { + if (mutable->saddr) return &local_remote_ports; else return &remote_ports; } else { - if (mutable->port_config.saddr) + if (mutable->saddr) return &key_local_remote_ports; else return &key_remote_ports; @@ -181,12 +190,12 @@ static int port_cmp(const struct tbl_node *node, void *target) const struct tnl_vport *tnl_vport = tnl_vport_table_cast(node); struct port_lookup_key *lookup = target; - lookup->mutable = rcu_dereference(tnl_vport->mutable); + lookup->mutable = rcu_dereference_rtnl(tnl_vport->mutable); return (lookup->mutable->tunnel_type == lookup->tunnel_type && - lookup->mutable->port_config.daddr == lookup->daddr && - lookup->mutable->port_config.in_key == lookup->key && - lookup->mutable->port_config.saddr == lookup->saddr); + lookup->mutable->daddr == lookup->daddr && + lookup->mutable->in_key == lookup->key && + lookup->mutable->saddr == lookup->saddr); } static u32 port_hash(struct port_lookup_key *k) @@ -200,9 +209,9 @@ static u32 mutable_hash(const struct tnl_mutable_config *mutable) { struct port_lookup_key lookup; - lookup.saddr = mutable->port_config.saddr; - lookup.daddr = mutable->port_config.daddr; - lookup.key = mutable->port_config.in_key; + lookup.saddr = mutable->saddr; + lookup.daddr = mutable->daddr; + lookup.key = mutable->in_key; lookup.tunnel_type = mutable->tunnel_type; return port_hash(&lookup); @@ -228,7 +237,7 @@ static int add_port(struct vport *vport) if (!port_table) { struct tbl *new_table; - new_table = tbl_create(0); + new_table = tbl_create(TBL_MIN_BUCKETS); if (!new_table) return -ENOMEM; @@ -313,7 +322,7 @@ struct vport *tnl_find_port(__be32 saddr, __be32 daddr, __be64 key, const struct tnl_mutable_config **mutable) { struct port_lookup_key lookup; - struct tbl *table = rcu_dereference(port_table); + struct tbl *table = rcu_dereference_rtnl(port_table); struct tbl_node *tbl_node; if (unlikely(!table)) @@ -692,7 +701,7 @@ bool tnl_frag_needed(struct vport *vport, const struct tnl_mutable_config *mutab * not symmetric then PMTUD needs to be disabled since we won't have * any way of synthesizing packets. */ - if ((mutable->port_config.flags & (TNL_F_IN_KEY_MATCH | TNL_F_OUT_KEY_ACTION)) == + if ((mutable->flags & (TNL_F_IN_KEY_MATCH | TNL_F_OUT_KEY_ACTION)) == (TNL_F_IN_KEY_MATCH | TNL_F_OUT_KEY_ACTION)) OVS_CB(nskb)->tun_id = flow_key; @@ -710,7 +719,7 @@ static bool check_mtu(struct sk_buff *skb, int mtu; __be16 frag_off; - frag_off = (mutable->port_config.flags & TNL_F_PMTUD) ? htons(IP_DF) : 0; + frag_off = (mutable->flags & TNL_F_PMTUD) ? htons(IP_DF) : 0; if (frag_off) mtu = dst_mtu(&rt_dst(rt)) - ETH_HLEN @@ -768,10 +777,10 @@ static void create_tunnel_header(const struct vport *vport, iph->ihl = sizeof(struct iphdr) >> 2; iph->frag_off = htons(IP_DF); iph->protocol = tnl_vport->tnl_ops->ipproto; - iph->tos = mutable->port_config.tos; + iph->tos = mutable->tos; iph->daddr = rt->rt_dst; iph->saddr = rt->rt_src; - iph->ttl = mutable->port_config.ttl; + iph->ttl = mutable->ttl; if (!iph->ttl) iph->ttl = dst_metric(&rt_dst(rt), RTAX_HOPLIMIT); @@ -857,7 +866,7 @@ static struct tnl_cache *build_cache(struct vport *vport, void *cache_data; int cache_len; - if (!(mutable->port_config.flags & TNL_F_HDR_CACHE)) + if (!(mutable->flags & TNL_F_HDR_CACHE)) return NULL; /* @@ -874,7 +883,7 @@ static struct tnl_cache *build_cache(struct vport *vport, if (!spin_trylock_bh(&tnl_vport->cache_lock)) return NULL; - cache = tnl_vport->cache; + cache = cache_dereference(tnl_vport); if (check_cache_valid(cache, mutable)) goto unlock; else @@ -901,15 +910,15 @@ static struct tnl_cache *build_cache(struct vport *vport, #endif if (is_internal_dev(rt_dst(rt).dev)) { - struct odp_flow_key flow_key; + struct sw_flow_key flow_key; struct tbl_node *flow_node; - struct vport *vport; + struct vport *dst_vport; struct sk_buff *skb; bool is_frag; int err; - vport = internal_dev_get_vport(rt_dst(rt).dev); - if (!vport) + dst_vport = internal_dev_get_vport(rt_dst(rt).dev); + if (!dst_vport) goto done; skb = alloc_skb(cache->len, GFP_ATOMIC); @@ -919,13 +928,13 @@ static struct tnl_cache *build_cache(struct vport *vport, __skb_put(skb, cache->len); memcpy(skb->data, get_cached_header(cache), cache->len); - err = flow_extract(skb, vport->port_no, &flow_key, &is_frag); + err = flow_extract(skb, dst_vport->port_no, &flow_key, &is_frag); kfree_skb(skb); if (err || is_frag) goto done; - flow_node = tbl_lookup(rcu_dereference(vport->dp->table), + flow_node = tbl_lookup(rcu_dereference(dst_vport->dp->table), &flow_key, flow_hash(&flow_key), flow_cmp); if (flow_node) { @@ -955,22 +964,21 @@ static struct rtable *find_route(struct vport *vport, *cache = NULL; tos = RT_TOS(tos); - if (likely(tos == mutable->port_config.tos && - check_cache_valid(cur_cache, mutable))) { + if (likely(tos == mutable->tos && check_cache_valid(cur_cache, mutable))) { *cache = cur_cache; return cur_cache->rt; } else { struct rtable *rt; struct flowi fl = { .nl_u = { .ip4_u = - { .daddr = mutable->port_config.daddr, - .saddr = mutable->port_config.saddr, + { .daddr = mutable->daddr, + .saddr = mutable->saddr, .tos = tos } }, .proto = tnl_vport->tnl_ops->ipproto }; if (unlikely(ip_route_output_key(&init_net, &rt, &fl))) return NULL; - if (likely(tos == mutable->port_config.tos)) + if (likely(tos == mutable->tos)) *cache = build_cache(vport, mutable, rt); return rt; @@ -1180,10 +1188,10 @@ int tnl_send(struct vport *vport, struct sk_buff *skb) else inner_tos = 0; - if (mutable->port_config.flags & TNL_F_TOS_INHERIT) + if (mutable->flags & TNL_F_TOS_INHERIT) tos = inner_tos; else - tos = mutable->port_config.tos; + tos = mutable->tos; tos = INET_ECN_encapsulate(tos, inner_tos); @@ -1222,11 +1230,11 @@ int tnl_send(struct vport *vport, struct sk_buff *skb) } /* TTL */ - ttl = mutable->port_config.ttl; + ttl = mutable->ttl; if (!ttl) ttl = dst_metric(&rt_dst(rt), RTAX_HOPLIMIT); - if (mutable->port_config.flags & TNL_F_TTL_INHERIT) { + if (mutable->flags & TNL_F_TTL_INHERIT) { if (skb->protocol == htons(ETH_P_IP)) ttl = ip_hdr(skb)->ttl; #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) @@ -1275,6 +1283,7 @@ int tnl_send(struct vport *vport, struct sk_buff *skb) struct vport *cache_vport = internal_dev_get_vport(rt_dst(rt).dev); skb->protocol = htons(ETH_P_IP); + iph = ip_hdr(skb); iph->tot_len = htons(skb->len - skb_network_offset(skb)); ip_send_check(iph); @@ -1284,12 +1293,12 @@ int tnl_send(struct vport *vport, struct sk_buff *skb) vport_receive(cache_vport, skb); sent_len += orig_len; } else { - int err; + int xmit_err; skb->dev = rt_dst(rt).dev; - err = dev_queue_xmit(skb); + xmit_err = dev_queue_xmit(skb); - if (likely(net_xmit_eval(err) == 0)) + if (likely(net_xmit_eval(xmit_err) == 0)) sent_len += orig_len; } } else @@ -1313,46 +1322,78 @@ out: return sent_len; } -static int set_config(const void *config, const struct tnl_ops *tnl_ops, - const struct vport *cur_vport, - struct tnl_mutable_config *mutable) +static const struct nla_policy tnl_policy[ODP_TUNNEL_ATTR_MAX + 1] = { + [ODP_TUNNEL_ATTR_FLAGS] = { .type = NLA_U32 }, + [ODP_TUNNEL_ATTR_DST_IPV4] = { .type = NLA_U32 }, + [ODP_TUNNEL_ATTR_SRC_IPV4] = { .type = NLA_U32 }, + [ODP_TUNNEL_ATTR_OUT_KEY] = { .type = NLA_U64 }, + [ODP_TUNNEL_ATTR_IN_KEY] = { .type = NLA_U64 }, + [ODP_TUNNEL_ATTR_TOS] = { .type = NLA_U8 }, + [ODP_TUNNEL_ATTR_TTL] = { .type = NLA_U8 }, +}; + +/* Sets ODP_TUNNEL_ATTR_* fields in 'mutable', which must initially be zeroed. */ +static int tnl_set_config(struct nlattr *options, const struct tnl_ops *tnl_ops, + const struct vport *cur_vport, + struct tnl_mutable_config *mutable) { const struct vport *old_vport; const struct tnl_mutable_config *old_mutable; + struct nlattr *a[ODP_TUNNEL_ATTR_MAX + 1]; + int err; - mutable->port_config = *(struct tnl_port_config *)config; - - if (mutable->port_config.daddr == 0) + if (!options) return -EINVAL; - if (mutable->port_config.tos != RT_TOS(mutable->port_config.tos)) + err = nla_parse_nested(a, ODP_TUNNEL_ATTR_MAX, options, tnl_policy); + if (err) + return err; + + if (!a[ODP_TUNNEL_ATTR_FLAGS] || !a[ODP_TUNNEL_ATTR_DST_IPV4]) return -EINVAL; - mutable->tunnel_hlen = tnl_ops->hdr_len(&mutable->port_config); + mutable->flags = nla_get_u32(a[ODP_TUNNEL_ATTR_FLAGS]) & TNL_F_PUBLIC; + + if (a[ODP_TUNNEL_ATTR_SRC_IPV4]) + mutable->saddr = nla_get_be32(a[ODP_TUNNEL_ATTR_SRC_IPV4]); + mutable->daddr = nla_get_be32(a[ODP_TUNNEL_ATTR_DST_IPV4]); + + if (a[ODP_TUNNEL_ATTR_TOS]) { + mutable->tos = nla_get_u8(a[ODP_TUNNEL_ATTR_TOS]); + if (mutable->tos != RT_TOS(mutable->tos)) + return -EINVAL; + } + + if (a[ODP_TUNNEL_ATTR_TTL]) + mutable->ttl = nla_get_u8(a[ODP_TUNNEL_ATTR_TTL]); + + mutable->tunnel_hlen = tnl_ops->hdr_len(mutable); if (mutable->tunnel_hlen < 0) return mutable->tunnel_hlen; mutable->tunnel_hlen += sizeof(struct iphdr); mutable->tunnel_type = tnl_ops->tunnel_type; - if (mutable->port_config.flags & TNL_F_IN_KEY_MATCH) { + if (!a[ODP_TUNNEL_ATTR_IN_KEY]) { mutable->tunnel_type |= TNL_T_KEY_MATCH; - mutable->port_config.in_key = 0; - } else + mutable->flags |= TNL_F_IN_KEY_MATCH; + } else { mutable->tunnel_type |= TNL_T_KEY_EXACT; + mutable->in_key = nla_get_be64(a[ODP_TUNNEL_ATTR_IN_KEY]); + } + + if (!a[ODP_TUNNEL_ATTR_OUT_KEY]) + mutable->flags |= TNL_F_OUT_KEY_ACTION; + else + mutable->out_key = nla_get_be64(a[ODP_TUNNEL_ATTR_OUT_KEY]); - old_vport = tnl_find_port(mutable->port_config.saddr, - mutable->port_config.daddr, - mutable->port_config.in_key, - mutable->tunnel_type, + old_vport = tnl_find_port(mutable->saddr, mutable->daddr, + mutable->in_key, mutable->tunnel_type, &old_mutable); if (old_vport && old_vport != cur_vport) return -EEXIST; - if (mutable->port_config.flags & TNL_F_OUT_KEY_ACTION) - mutable->port_config.out_key = 0; - return 0; } @@ -1362,6 +1403,7 @@ struct vport *tnl_create(const struct vport_parms *parms, { struct vport *vport; struct tnl_vport *tnl_vport; + struct tnl_mutable_config *mutable; int initial_frag_id; int err; @@ -1376,19 +1418,19 @@ struct vport *tnl_create(const struct vport_parms *parms, strcpy(tnl_vport->name, parms->name); tnl_vport->tnl_ops = tnl_ops; - tnl_vport->mutable = kzalloc(sizeof(struct tnl_mutable_config), GFP_KERNEL); - if (!tnl_vport->mutable) { + mutable = kzalloc(sizeof(struct tnl_mutable_config), GFP_KERNEL); + if (!mutable) { err = -ENOMEM; goto error_free_vport; } - vport_gen_rand_ether_addr(tnl_vport->mutable->eth_addr); - tnl_vport->mutable->mtu = ETH_DATA_LEN; + vport_gen_rand_ether_addr(mutable->eth_addr); + mutable->mtu = ETH_DATA_LEN; get_random_bytes(&initial_frag_id, sizeof(int)); atomic_set(&tnl_vport->frag_id, initial_frag_id); - err = set_config(parms->config, tnl_ops, NULL, tnl_vport->mutable); + err = tnl_set_config(parms->options, tnl_ops, NULL, mutable); if (err) goto error_free_mutable; @@ -1396,9 +1438,11 @@ struct vport *tnl_create(const struct vport_parms *parms, #ifdef NEED_CACHE_TIMEOUT tnl_vport->cache_exp_interval = MAX_CACHE_EXP - - (net_random() % (MAX_CACHE_EXP / 2)); + (net_random() % (MAX_CACHE_EXP / 2)); #endif + rcu_assign_pointer(tnl_vport->mutable, mutable); + err = add_port(vport); if (err) goto error_free_mutable; @@ -1406,31 +1450,37 @@ struct vport *tnl_create(const struct vport_parms *parms, return vport; error_free_mutable: - kfree(tnl_vport->mutable); + kfree(mutable); error_free_vport: vport_free(vport); error: return ERR_PTR(err); } -int tnl_modify(struct vport *vport, struct odp_port *port) +int tnl_set_options(struct vport *vport, struct nlattr *options) { struct tnl_vport *tnl_vport = tnl_vport_priv(vport); + const struct tnl_mutable_config *old_mutable; struct tnl_mutable_config *mutable; int err; - mutable = kmemdup(tnl_vport->mutable, sizeof(struct tnl_mutable_config), GFP_KERNEL); + mutable = kzalloc(sizeof(struct tnl_mutable_config), GFP_KERNEL); if (!mutable) { err = -ENOMEM; goto error; } - err = set_config(port->config, tnl_vport->tnl_ops, vport, mutable); + /* Copy fields whose values should be retained. */ + old_mutable = rtnl_dereference(tnl_vport->mutable); + mutable->seq = old_mutable->seq + 1; + memcpy(mutable->eth_addr, old_mutable->eth_addr, ETH_ALEN); + mutable->mtu = old_mutable->mtu; + + /* Parse the others configured by userspace. */ + err = tnl_set_config(options, tnl_vport->tnl_ops, vport, mutable); if (err) goto error_free; - mutable->seq++; - err = move_port(vport, mutable); if (err) goto error_free; @@ -1443,28 +1493,51 @@ error: return err; } -static void free_port_rcu(struct rcu_head *rcu) +int tnl_get_options(const struct vport *vport, struct sk_buff *skb) { - struct tnl_vport *tnl_vport = container_of(rcu, struct tnl_vport, rcu); + const struct tnl_vport *tnl_vport = tnl_vport_priv(vport); + const struct tnl_mutable_config *mutable = rcu_dereference_rtnl(tnl_vport->mutable); + + NLA_PUT_U32(skb, ODP_TUNNEL_ATTR_FLAGS, mutable->flags & TNL_F_PUBLIC); + NLA_PUT_BE32(skb, ODP_TUNNEL_ATTR_DST_IPV4, mutable->daddr); + + if (!(mutable->flags & TNL_F_IN_KEY_MATCH)) + NLA_PUT_BE64(skb, ODP_TUNNEL_ATTR_IN_KEY, mutable->in_key); + if (!(mutable->flags & TNL_F_OUT_KEY_ACTION)) + NLA_PUT_BE64(skb, ODP_TUNNEL_ATTR_OUT_KEY, mutable->out_key); + if (mutable->saddr) + NLA_PUT_BE32(skb, ODP_TUNNEL_ATTR_SRC_IPV4, mutable->saddr); + if (mutable->tos) + NLA_PUT_U8(skb, ODP_TUNNEL_ATTR_TOS, mutable->tos); + if (mutable->ttl) + NLA_PUT_U8(skb, ODP_TUNNEL_ATTR_TTL, mutable->ttl); - spin_lock_bh(&tnl_vport->cache_lock); - free_cache(tnl_vport->cache); - spin_unlock_bh(&tnl_vport->cache_lock); + return 0; - kfree(tnl_vport->mutable); +nla_put_failure: + return -EMSGSIZE; +} + +static void free_port_rcu(struct rcu_head *rcu) +{ + struct tnl_vport *tnl_vport = container_of(rcu, + struct tnl_vport, rcu); + + free_cache((struct tnl_cache __force *)tnl_vport->cache); + kfree((struct tnl_mutable __force *)tnl_vport->mutable); vport_free(tnl_vport_to_vport(tnl_vport)); } int tnl_destroy(struct vport *vport) { struct tnl_vport *tnl_vport = tnl_vport_priv(vport); - const struct tnl_mutable_config *old_mutable; + const struct tnl_mutable_config *mutable, *old_mutable; + + mutable = rtnl_dereference(tnl_vport->mutable); - if (vport == tnl_find_port(tnl_vport->mutable->port_config.saddr, - tnl_vport->mutable->port_config.daddr, - tnl_vport->mutable->port_config.in_key, - tnl_vport->mutable->tunnel_type, - &old_mutable)) + if (vport == tnl_find_port(mutable->saddr, mutable->daddr, + mutable->in_key, mutable->tunnel_type, + &old_mutable)) del_port(vport); call_rcu(&tnl_vport->rcu, free_port_rcu); @@ -1477,7 +1550,8 @@ int tnl_set_mtu(struct vport *vport, int mtu) struct tnl_vport *tnl_vport = tnl_vport_priv(vport); struct tnl_mutable_config *mutable; - mutable = kmemdup(tnl_vport->mutable, sizeof(struct tnl_mutable_config), GFP_KERNEL); + mutable = kmemdup(rtnl_dereference(tnl_vport->mutable), + sizeof(struct tnl_mutable_config), GFP_KERNEL); if (!mutable) return -ENOMEM; @@ -1492,7 +1566,8 @@ int tnl_set_addr(struct vport *vport, const unsigned char *addr) struct tnl_vport *tnl_vport = tnl_vport_priv(vport); struct tnl_mutable_config *mutable; - mutable = kmemdup(tnl_vport->mutable, sizeof(struct tnl_mutable_config), GFP_KERNEL); + mutable = kmemdup(rtnl_dereference(tnl_vport->mutable), + sizeof(struct tnl_mutable_config), GFP_KERNEL); if (!mutable) return -ENOMEM; @@ -1511,13 +1586,13 @@ const char *tnl_get_name(const struct vport *vport) const unsigned char *tnl_get_addr(const struct vport *vport) { const struct tnl_vport *tnl_vport = tnl_vport_priv(vport); - return rcu_dereference(tnl_vport->mutable)->eth_addr; + return rcu_dereference_rtnl(tnl_vport->mutable)->eth_addr; } int tnl_get_mtu(const struct vport *vport) { const struct tnl_vport *tnl_vport = tnl_vport_priv(vport); - return rcu_dereference(tnl_vport->mutable)->mtu; + return rcu_dereference_rtnl(tnl_vport->mutable)->mtu; } void tnl_free_linked_skbs(struct sk_buff *skb)