X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=datapath%2Ftunnel.c;h=ae99cc16bef66edc82eef6ab3350d33c16cc361c;hb=5ca1ba484bd9ade5116a49cf241cb98219d7d696;hp=ec3ad936f824fbaf6857a38ef9a79a1bb53b13df;hpb=dd8d6b8cd4d708553037e306d41d1c501d3cecad;p=sliver-openvswitch.git diff --git a/datapath/tunnel.c b/datapath/tunnel.c index ec3ad936f..ae99cc16b 100644 --- a/datapath/tunnel.c +++ b/datapath/tunnel.c @@ -1,21 +1,37 @@ /* - * Copyright (c) 2010 Nicira Networks. - * Distributed under the terms of the GNU GPL version 2. + * Copyright (c) 2007-2012 Nicira, Inc. * - * Significant portions of this file may be copied from parts of the Linux - * kernel, by Linus Torvalds and others. + * This program is free software; you can redistribute it and/or + * modify it under the terms of version 2 of the GNU General Public + * License as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA + * 02110-1301, USA */ +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + #include #include #include #include +#include #include #include +#include #include +#include #include #include #include +#include #include #include @@ -28,51 +44,16 @@ #include #include -#include "actions.h" #include "checksum.h" #include "datapath.h" -#include "table.h" #include "tunnel.h" +#include "vlan.h" #include "vport.h" -#include "vport-generic.h" #include "vport-internal_dev.h" -#ifdef NEED_CACHE_TIMEOUT -/* - * On kernels where we can't quickly detect changes in the rest of the system - * we use an expiration time to invalidate the cache. A shorter expiration - * reduces the length of time that we may potentially blackhole packets while - * a longer time increases performance by reducing the frequency that the - * cache needs to be rebuilt. A variety of factors may cause the cache to be - * invalidated before the expiration time but this is the maximum. The time - * is expressed in jiffies. - */ -#define MAX_CACHE_EXP HZ -#endif - -/* - * Interval to check for and remove caches that are no longer valid. Caches - * are checked for validity before they are used for packet encapsulation and - * old caches are removed at that time. However, if no packets are sent through - * the tunnel then the cache will never be destroyed. Since it holds - * references to a number of system objects, the cache will continue to use - * system resources by not allowing those objects to be destroyed. The cache - * cleaner is periodically run to free invalid caches. It does not - * significantly affect system performance. A lower interval will release - * resources faster but will itself consume resources by requiring more frequent - * checks. A longer interval may result in messages being printed to the kernel - * message buffer about unreleased resources. The interval is expressed in - * jiffies. - */ -#define CACHE_CLEANER_INTERVAL (5 * HZ) - -#define CACHE_DATA_ALIGN 16 - -/* Protected by RCU. */ -static struct tbl *port_table __read_mostly; +#define PORT_TABLE_SIZE 1024 -static void cache_cleaner(struct work_struct *work); -DECLARE_DELAYED_WORK(cache_cleaner_wq, cache_cleaner); +static struct hlist_head *port_table __read_mostly; /* * These are just used as an optimization: they don't require any kind of @@ -81,8 +62,11 @@ DECLARE_DELAYED_WORK(cache_cleaner_wq, cache_cleaner); */ static unsigned int key_local_remote_ports __read_mostly; static unsigned int key_remote_ports __read_mostly; +static unsigned int key_multicast_ports __read_mostly; static unsigned int local_remote_ports __read_mostly; static unsigned int remote_ports __read_mostly; +static unsigned int null_ports __read_mostly; +static unsigned int multicast_ports __read_mostly; #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,36) #define rt_dst(rt) (rt->dst) @@ -90,41 +74,30 @@ static unsigned int remote_ports __read_mostly; #define rt_dst(rt) (rt->u.dst) #endif -static inline struct vport *tnl_vport_to_vport(const struct tnl_vport *tnl_vport) +static struct vport *tnl_vport_to_vport(const struct tnl_vport *tnl_vport) { return vport_from_priv(tnl_vport); } -static inline struct tnl_vport *tnl_vport_table_cast(const struct tbl_node *node) -{ - return container_of(node, struct tnl_vport, tbl_node); -} - -static inline void schedule_cache_cleaner(void) -{ - schedule_delayed_work(&cache_cleaner_wq, CACHE_CLEANER_INTERVAL); -} - -static void free_cache(struct tnl_cache *cache) -{ - if (!cache) - return; - - flow_put(cache->flow); - ip_rt_put(cache->rt); - kfree(cache); -} - static void free_config_rcu(struct rcu_head *rcu) { struct tnl_mutable_config *c = container_of(rcu, struct tnl_mutable_config, rcu); kfree(c); } -static void free_cache_rcu(struct rcu_head *rcu) +/* Frees the portion of 'mutable' that requires RTNL and thus can't happen + * within an RCU callback. Fortunately this part doesn't require waiting for + * an RCU grace period. + */ +static void free_mutable_rtnl(struct tnl_mutable_config *mutable) { - struct tnl_cache *c = container_of(rcu, struct tnl_cache, rcu); - free_cache(c); + ASSERT_RTNL(); + if (ipv4_is_multicast(mutable->key.daddr) && mutable->mlink) { + struct in_device *in_dev; + in_dev = inetdev_by_index(port_key_get_net(&mutable->key), mutable->mlink); + if (in_dev) + ip_mc_dec_group(in_dev, mutable->key.daddr); + } } static void assign_config_rcu(struct vport *vport, @@ -133,288 +106,256 @@ static void assign_config_rcu(struct vport *vport, struct tnl_vport *tnl_vport = tnl_vport_priv(vport); struct tnl_mutable_config *old_config; - old_config = tnl_vport->mutable; + old_config = rtnl_dereference(tnl_vport->mutable); rcu_assign_pointer(tnl_vport->mutable, new_config); - call_rcu(&old_config->rcu, free_config_rcu); -} - -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; - rcu_assign_pointer(tnl_vport->cache, new_cache); - if (old_cache) - call_rcu(&old_cache->rcu, free_cache_rcu); + free_mutable_rtnl(old_config); + call_rcu(&old_config->rcu, free_config_rcu); } 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) + bool is_multicast = ipv4_is_multicast(mutable->key.daddr); + + if (mutable->flags & TNL_F_IN_KEY_MATCH) { + if (mutable->key.saddr) return &local_remote_ports; + else if (is_multicast) + return &multicast_ports; else return &remote_ports; } else { - if (mutable->port_config.saddr) + if (mutable->key.saddr) return &key_local_remote_ports; - else + else if (is_multicast) + return &key_multicast_ports; + else if (mutable->key.daddr) return &key_remote_ports; + else + return &null_ports; } } -struct port_lookup_key { - u32 tunnel_type; - __be32 saddr; - __be32 daddr; - __be32 key; - const struct tnl_mutable_config *mutable; -}; - -/* - * Modifies 'target' to store the rcu_dereferenced pointer that was used to do - * the comparision. - */ -static int port_cmp(const struct tbl_node *node, void *target) +static u32 port_hash(const struct port_lookup_key *key) { - const struct tnl_vport *tnl_vport = tnl_vport_table_cast(node); - struct port_lookup_key *lookup = target; - - lookup->mutable = rcu_dereference(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); + return jhash2((u32 *)key, (PORT_KEY_LEN / sizeof(u32)), 0); } -static u32 port_hash(struct port_lookup_key *k) +static struct hlist_head *find_bucket(u32 hash) { - return jhash_3words(k->key, k->saddr, k->daddr, k->tunnel_type); + return &port_table[(hash & (PORT_TABLE_SIZE - 1))]; } -static u32 mutable_hash(const struct tnl_mutable_config *mutable) +static void port_table_add_port(struct vport *vport) { - 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.tunnel_type = mutable->tunnel_type; - - return port_hash(&lookup); -} + struct tnl_vport *tnl_vport = tnl_vport_priv(vport); + const struct tnl_mutable_config *mutable; + u32 hash; -static void check_table_empty(void) -{ - if (tbl_count(port_table) == 0) { - struct tbl *old_table = port_table; + mutable = rtnl_dereference(tnl_vport->mutable); + hash = port_hash(&mutable->key); + hlist_add_head_rcu(&tnl_vport->hash_node, find_bucket(hash)); - cancel_delayed_work_sync(&cache_cleaner_wq); - rcu_assign_pointer(port_table, NULL); - tbl_deferred_destroy(old_table, NULL); - } + (*find_port_pool(rtnl_dereference(tnl_vport->mutable)))++; } -static int add_port(struct vport *vport) +static void port_table_move_port(struct vport *vport, + struct tnl_mutable_config *new_mutable) { struct tnl_vport *tnl_vport = tnl_vport_priv(vport); - int err; - - if (!port_table) { - struct tbl *new_table; - - new_table = tbl_create(0); - if (!new_table) - return -ENOMEM; - - rcu_assign_pointer(port_table, new_table); - schedule_cache_cleaner(); - - } else if (tbl_count(port_table) > tbl_n_buckets(port_table)) { - struct tbl *old_table = port_table; - struct tbl *new_table; - - new_table = tbl_expand(old_table); - if (IS_ERR(new_table)) - return PTR_ERR(new_table); - - rcu_assign_pointer(port_table, new_table); - tbl_deferred_destroy(old_table, NULL); - } - - err = tbl_insert(port_table, &tnl_vport->tbl_node, mutable_hash(tnl_vport->mutable)); - if (err) { - check_table_empty(); - return err; - } + u32 hash; - (*find_port_pool(tnl_vport->mutable))++; + hash = port_hash(&new_mutable->key); + hlist_del_init_rcu(&tnl_vport->hash_node); + hlist_add_head_rcu(&tnl_vport->hash_node, find_bucket(hash)); - return 0; + (*find_port_pool(rtnl_dereference(tnl_vport->mutable)))--; + assign_config_rcu(vport, new_mutable); + (*find_port_pool(rtnl_dereference(tnl_vport->mutable)))++; } -static int move_port(struct vport *vport, struct tnl_mutable_config *new_mutable) +static void port_table_remove_port(struct vport *vport) { - int err; struct tnl_vport *tnl_vport = tnl_vport_priv(vport); - u32 hash; - hash = mutable_hash(new_mutable); - if (hash == tnl_vport->tbl_node.hash) - goto table_updated; + hlist_del_init_rcu(&tnl_vport->hash_node); - /* - * Ideally we should make this move atomic to avoid having gaps in - * finding tunnels or the possibility of failure. However, if we do - * find a tunnel it will always be consistent. - */ - err = tbl_remove(port_table, &tnl_vport->tbl_node); - if (err) - return err; - - err = tbl_insert(port_table, &tnl_vport->tbl_node, hash); - if (err) { - check_table_empty(); - return err; - } - -table_updated: - assign_config_rcu(vport, new_mutable); - - return 0; + (*find_port_pool(rtnl_dereference(tnl_vport->mutable)))--; } -static int del_port(struct vport *vport) +static struct vport *port_table_lookup(struct port_lookup_key *key, + const struct tnl_mutable_config **pmutable) { - struct tnl_vport *tnl_vport = tnl_vport_priv(vport); - int err; + struct hlist_node *n; + struct hlist_head *bucket; + u32 hash = port_hash(key); + struct tnl_vport *tnl_vport; - err = tbl_remove(port_table, &tnl_vport->tbl_node); - if (err) - return err; + bucket = find_bucket(hash); - check_table_empty(); - (*find_port_pool(tnl_vport->mutable))--; + hlist_for_each_entry_rcu(tnl_vport, n, bucket, hash_node) { + struct tnl_mutable_config *mutable; - return 0; + mutable = rcu_dereference_rtnl(tnl_vport->mutable); + if (!memcmp(&mutable->key, key, PORT_KEY_LEN)) { + *pmutable = mutable; + return tnl_vport_to_vport(tnl_vport); + } + } + + return NULL; } -struct vport *tnl_find_port(__be32 saddr, __be32 daddr, __be32 key, - int tunnel_type, - const struct tnl_mutable_config **mutable) +struct vport *ovs_tnl_find_port(struct net *net, __be32 saddr, __be32 daddr, + __be64 key, int tunnel_type, + const struct tnl_mutable_config **mutable) { struct port_lookup_key lookup; - struct tbl *table = rcu_dereference(port_table); - struct tbl_node *tbl_node; - - if (unlikely(!table)) - return NULL; + struct vport *vport; + bool is_multicast = ipv4_is_multicast(saddr); + port_key_set_net(&lookup, net); lookup.saddr = saddr; lookup.daddr = daddr; - if (tunnel_type & TNL_T_KEY_EXACT) { - lookup.key = key; - lookup.tunnel_type = tunnel_type & ~TNL_T_KEY_MATCH; - - if (key_local_remote_ports) { - tbl_node = tbl_lookup(table, &lookup, port_hash(&lookup), port_cmp); - if (tbl_node) - goto found; - } - - if (key_remote_ports) { - lookup.saddr = 0; - - tbl_node = tbl_lookup(table, &lookup, port_hash(&lookup), port_cmp); - if (tbl_node) - goto found; + /* First try for exact match on in_key. */ + lookup.in_key = key; + lookup.tunnel_type = tunnel_type | TNL_T_KEY_EXACT; + if (!is_multicast && key_local_remote_ports) { + vport = port_table_lookup(&lookup, mutable); + if (vport) + return vport; + } + if (key_remote_ports) { + lookup.saddr = 0; + vport = port_table_lookup(&lookup, mutable); + if (vport) + return vport; - lookup.saddr = saddr; - } + lookup.saddr = saddr; } - if (tunnel_type & TNL_T_KEY_MATCH) { - lookup.key = 0; - lookup.tunnel_type = tunnel_type & ~TNL_T_KEY_EXACT; + /* Then try matches that wildcard in_key. */ + lookup.in_key = 0; + lookup.tunnel_type = tunnel_type | TNL_T_KEY_MATCH; + if (!is_multicast && local_remote_ports) { + vport = port_table_lookup(&lookup, mutable); + if (vport) + return vport; + } + if (remote_ports) { + lookup.saddr = 0; + vport = port_table_lookup(&lookup, mutable); + if (vport) + return vport; + } - if (local_remote_ports) { - tbl_node = tbl_lookup(table, &lookup, port_hash(&lookup), port_cmp); - if (tbl_node) - goto found; + if (is_multicast) { + lookup.saddr = 0; + lookup.daddr = saddr; + if (key_multicast_ports) { + lookup.tunnel_type = tunnel_type | TNL_T_KEY_EXACT; + lookup.in_key = key; + vport = port_table_lookup(&lookup, mutable); + if (vport) + return vport; } - - if (remote_ports) { - lookup.saddr = 0; - - tbl_node = tbl_lookup(table, &lookup, port_hash(&lookup), port_cmp); - if (tbl_node) - goto found; + if (multicast_ports) { + lookup.tunnel_type = tunnel_type | TNL_T_KEY_MATCH; + lookup.in_key = 0; + vport = port_table_lookup(&lookup, mutable); + if (vport) + return vport; } } + if (null_ports) { + lookup.daddr = 0; + lookup.saddr = 0; + lookup.in_key = 0; + lookup.tunnel_type = tunnel_type; + vport = port_table_lookup(&lookup, mutable); + if (vport) + return vport; + } return NULL; - -found: - *mutable = lookup.mutable; - return tnl_vport_to_vport(tnl_vport_table_cast(tbl_node)); } -static inline void ecn_decapsulate(struct sk_buff *skb) +static void ecn_decapsulate(struct sk_buff *skb) { - u8 tos = ip_hdr(skb)->tos; - - if (INET_ECN_is_ce(tos)) { + if (unlikely(INET_ECN_is_ce(OVS_CB(skb)->tun_key->ipv4_tos))) { __be16 protocol = skb->protocol; - unsigned int nw_header = skb_network_offset(skb); - if (skb->protocol == htons(ETH_P_8021Q)) { + skb_set_network_header(skb, ETH_HLEN); + + if (protocol == htons(ETH_P_8021Q)) { if (unlikely(!pskb_may_pull(skb, VLAN_ETH_HLEN))) return; protocol = vlan_eth_hdr(skb)->h_vlan_encapsulated_proto; - nw_header += VLAN_HLEN; + skb_set_network_header(skb, VLAN_ETH_HLEN); } if (protocol == htons(ETH_P_IP)) { - if (unlikely(!pskb_may_pull(skb, nw_header + if (unlikely(!pskb_may_pull(skb, skb_network_offset(skb) + sizeof(struct iphdr)))) return; - IP_ECN_set_ce((struct iphdr *)(skb->data + nw_header)); + IP_ECN_set_ce(ip_hdr(skb)); } #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) else if (protocol == htons(ETH_P_IPV6)) { - if (unlikely(!pskb_may_pull(skb, nw_header + if (unlikely(!pskb_may_pull(skb, skb_network_offset(skb) + sizeof(struct ipv6hdr)))) return; - IP6_ECN_set_ce((struct ipv6hdr *)(skb->data + nw_header)); + IP6_ECN_set_ce(ipv6_hdr(skb)); } #endif } } -/* Called with rcu_read_lock. */ -void tnl_rcv(struct vport *vport, struct sk_buff *skb) +/** + * ovs_tnl_rcv - ingress point for generic tunnel code + * + * @vport: port this packet was received on + * @skb: received packet + * @tos: ToS from encapsulating IP packet, used to copy ECN bits + * + * Must be called with rcu_read_lock. + * + * Packets received by this function are in the following state: + * - skb->data points to the inner Ethernet header. + * - The inner Ethernet header is in the linear data area. + * - skb->csum does not include the inner Ethernet header. + * - The layer pointers are undefined. + */ +void ovs_tnl_rcv(struct vport *vport, struct sk_buff *skb) { - skb->pkt_type = PACKET_HOST; - skb->protocol = eth_type_trans(skb, skb->dev); + struct ethhdr *eh; + + skb_reset_mac_header(skb); + eh = eth_hdr(skb); + + if (likely(ntohs(eh->h_proto) >= 1536)) + skb->protocol = eh->h_proto; + else + skb->protocol = htons(ETH_P_802_2); skb_dst_drop(skb); nf_reset(skb); + skb_clear_rxhash(skb); secpath_reset(skb); - skb_reset_network_header(skb); ecn_decapsulate(skb); + vlan_set_tci(skb, 0); - skb_push(skb, ETH_HLEN); - compute_ip_summed(skb, false); + if (unlikely(compute_ip_summed(skb, false))) { + kfree_skb(skb); + return; + } - vport_receive(vport, skb); + ovs_vport_receive(vport, skb); } static bool check_ipv4_address(__be32 addr) @@ -514,6 +455,7 @@ static bool ipv6_should_icmp(struct sk_buff *skb) int addr_type; int payload_off = (u8 *)(old_ipv6h + 1) - skb->data; u8 nexthdr = ipv6_hdr(skb)->nexthdr; + __be16 frag_off; /* Check source address is valid. */ addr_type = ipv6_addr_type(&old_ipv6h->saddr); @@ -525,7 +467,7 @@ static bool ipv6_should_icmp(struct sk_buff *skb) return false; /* Don't respond to ICMP error messages. */ - payload_off = ipv6_skip_exthdr(skb, payload_off, &nexthdr); + payload_off = ipv6_skip_exthdr(skb, payload_off, &nexthdr, &frag_off); if (payload_off < 0) return false; @@ -563,8 +505,8 @@ static void ipv6_build_icmp(struct sk_buff *skb, struct sk_buff *nskb, + payload_length); ipv6h->nexthdr = NEXTHDR_ICMP; ipv6h->hop_limit = IPV6_DEFAULT_HOPLIMIT; - ipv6_addr_copy(&ipv6h->daddr, &old_ipv6h->saddr); - ipv6_addr_copy(&ipv6h->saddr, &old_ipv6h->daddr); + ipv6h->daddr = old_ipv6h->saddr; + ipv6h->saddr = old_ipv6h->daddr; /* ICMPv6 */ icmp6h->icmp6_type = ICMPV6_PKT_TOOBIG; @@ -583,8 +525,9 @@ static void ipv6_build_icmp(struct sk_buff *skb, struct sk_buff *nskb, } #endif /* IPv6 */ -bool tnl_frag_needed(struct vport *vport, const struct tnl_mutable_config *mutable, - struct sk_buff *skb, unsigned int mtu, __be32 flow_key) +bool ovs_tnl_frag_needed(struct vport *vport, + const struct tnl_mutable_config *mutable, + struct sk_buff *skb, unsigned int mtu) { unsigned int eth_hdr_len = ETH_HLEN; unsigned int total_length = 0, header_length = 0, payload_length; @@ -637,7 +580,6 @@ bool tnl_frag_needed(struct vport *vport, const struct tnl_mutable_config *mutab } #endif - total_length = min(total_length, mutable->mtu); payload_length = total_length - header_length; nskb = dev_alloc_skb(NET_IP_ALIGN + eth_hdr_len + header_length + @@ -657,7 +599,8 @@ bool tnl_frag_needed(struct vport *vport, const struct tnl_mutable_config *mutab vh->h_vlan_TCI = vlan_eth_hdr(skb)->h_vlan_TCI; vh->h_vlan_encapsulated_proto = skb->protocol; - } + } else + vlan_set_tci(nskb, vlan_get_tci(skb)); skb_reset_mac_header(nskb); /* Protocol */ @@ -668,19 +611,12 @@ bool tnl_frag_needed(struct vport *vport, const struct tnl_mutable_config *mutab ipv6_build_icmp(skb, nskb, mtu, payload_length); #endif - /* - * Assume that flow based keys are symmetric with respect to input - * and output and use the key that we were going to put on the - * outgoing packet for the fake received packet. If the keys are - * 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)) == - (TNL_F_IN_KEY_MATCH | TNL_F_OUT_KEY_ACTION)) - OVS_CB(nskb)->tun_id = flow_key; + if (unlikely(compute_ip_summed(nskb, false))) { + kfree_skb(nskb); + return false; + } - compute_ip_summed(nskb, false); - vport_receive(vport, nskb); + ovs_vport_receive(vport, nskb); return true; } @@ -688,300 +624,115 @@ bool tnl_frag_needed(struct vport *vport, const struct tnl_mutable_config *mutab static bool check_mtu(struct sk_buff *skb, struct vport *vport, const struct tnl_mutable_config *mutable, - const struct rtable *rt, __be16 *frag_offp) + const struct rtable *rt, __be16 *frag_offp, + int tunnel_hlen) { - int mtu; + bool df_inherit; + bool pmtud; __be16 frag_off; + int mtu = 0; + unsigned int packet_length = skb->len - ETH_HLEN; + + if (OVS_CB(skb)->tun_key->ipv4_dst) { + df_inherit = false; + pmtud = false; + frag_off = OVS_CB(skb)->tun_key->tun_flags & OVS_TNL_F_DONT_FRAGMENT ? + htons(IP_DF) : 0; + } else { + df_inherit = mutable->flags & TNL_F_DF_INHERIT; + pmtud = mutable->flags & TNL_F_PMTUD; + frag_off = mutable->flags & TNL_F_DF_DEFAULT ? htons(IP_DF) : 0; + } + + /* Allow for one level of tagging in the packet length. */ + if (!vlan_tx_tag_present(skb) && + eth_hdr(skb)->h_proto == htons(ETH_P_8021Q)) + packet_length -= VLAN_HLEN; + + if (pmtud) { + int vlan_header = 0; + + /* The tag needs to go in packet regardless of where it + * currently is, so subtract it from the MTU. + */ + if (vlan_tx_tag_present(skb) || + eth_hdr(skb)->h_proto == htons(ETH_P_8021Q)) + vlan_header = VLAN_HLEN; - frag_off = (mutable->port_config.flags & TNL_F_PMTUD) ? htons(IP_DF) : 0; - if (frag_off) mtu = dst_mtu(&rt_dst(rt)) - ETH_HLEN - - mutable->tunnel_hlen - - (eth_hdr(skb)->h_proto == htons(ETH_P_8021Q) ? VLAN_HLEN : 0); - else - mtu = mutable->mtu; + - tunnel_hlen + - vlan_header; + } if (skb->protocol == htons(ETH_P_IP)) { - struct iphdr *old_iph = ip_hdr(skb); + struct iphdr *iph = ip_hdr(skb); + + if (df_inherit) + frag_off = iph->frag_off & htons(IP_DF); - frag_off |= old_iph->frag_off & htons(IP_DF); - mtu = max(mtu, IP_MIN_MTU); + if (pmtud && iph->frag_off & htons(IP_DF)) { + mtu = max(mtu, IP_MIN_MTU); - if ((old_iph->frag_off & htons(IP_DF)) && - mtu < ntohs(old_iph->tot_len)) { - if (tnl_frag_needed(vport, mutable, skb, mtu, OVS_CB(skb)->tun_id)) - goto drop; + if (packet_length > mtu && + ovs_tnl_frag_needed(vport, mutable, skb, mtu)) + return false; } } #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) else if (skb->protocol == htons(ETH_P_IPV6)) { - unsigned int packet_length = skb->len - ETH_HLEN - - (eth_hdr(skb)->h_proto == htons(ETH_P_8021Q) ? VLAN_HLEN : 0); - - mtu = max(mtu, IPV6_MIN_MTU); - - /* IPv6 requires PMTUD if the packet is above the minimum MTU. */ - if (packet_length > IPV6_MIN_MTU) + /* IPv6 requires end hosts to do fragmentation + * if the packet is above the minimum MTU. + */ + if (df_inherit && packet_length > IPV6_MIN_MTU) frag_off = htons(IP_DF); - if (mtu < packet_length) { - if (tnl_frag_needed(vport, mutable, skb, mtu, OVS_CB(skb)->tun_id)) - goto drop; + if (pmtud) { + mtu = max(mtu, IPV6_MIN_MTU); + + if (packet_length > mtu && + ovs_tnl_frag_needed(vport, mutable, skb, mtu)) + return false; } } #endif *frag_offp = frag_off; return true; - -drop: - *frag_offp = 0; - return false; -} - -static void create_tunnel_header(const struct vport *vport, - const struct tnl_mutable_config *mutable, - const struct rtable *rt, void *header) -{ - struct tnl_vport *tnl_vport = tnl_vport_priv(vport); - struct iphdr *iph = header; - - iph->version = 4; - 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->daddr = rt->rt_dst; - iph->saddr = rt->rt_src; - iph->ttl = mutable->port_config.ttl; - if (!iph->ttl) - iph->ttl = dst_metric(&rt_dst(rt), RTAX_HOPLIMIT); - - tnl_vport->tnl_ops->build_header(vport, mutable, iph + 1); -} - -static inline void *get_cached_header(const struct tnl_cache *cache) -{ - return (void *)cache + ALIGN(sizeof(struct tnl_cache), CACHE_DATA_ALIGN); -} - -static inline bool check_cache_valid(const struct tnl_cache *cache, - const struct tnl_mutable_config *mutable) -{ - return cache && -#ifdef NEED_CACHE_TIMEOUT - time_before(jiffies, cache->expiration) && -#endif -#ifdef HAVE_RT_GENID - atomic_read(&init_net.ipv4.rt_genid) == cache->rt->rt_genid && -#endif -#ifdef HAVE_HH_SEQ - rt_dst(cache->rt).hh->hh_lock.sequence == cache->hh_seq && -#endif - mutable->seq == cache->mutable_seq && - (!is_internal_dev(rt_dst(cache->rt).dev) || - (cache->flow && !cache->flow->dead)); -} - -static int cache_cleaner_cb(struct tbl_node *tbl_node, void *aux) -{ - struct tnl_vport *tnl_vport = tnl_vport_table_cast(tbl_node); - const struct tnl_mutable_config *mutable = rcu_dereference(tnl_vport->mutable); - const struct tnl_cache *cache = rcu_dereference(tnl_vport->cache); - - if (cache && !check_cache_valid(cache, mutable) && - spin_trylock_bh(&tnl_vport->cache_lock)) { - assign_cache_rcu(tnl_vport_to_vport(tnl_vport), NULL); - spin_unlock_bh(&tnl_vport->cache_lock); - } - - return 0; -} - -static void cache_cleaner(struct work_struct *work) -{ - schedule_cache_cleaner(); - - rcu_read_lock(); - tbl_foreach(port_table, cache_cleaner_cb, NULL); - rcu_read_unlock(); } -static inline void create_eth_hdr(struct tnl_cache *cache, - const struct rtable *rt) +static struct rtable *find_route(struct net *net, + __be32 *saddr, __be32 daddr, u8 ipproto, + u8 tos) { - void *cache_data = get_cached_header(cache); - int hh_len = rt_dst(rt).hh->hh_len; - int hh_off = HH_DATA_ALIGN(rt_dst(rt).hh->hh_len) - hh_len; - -#ifdef HAVE_HH_SEQ - unsigned hh_seq; - - do { - hh_seq = read_seqbegin(&rt_dst(rt).hh->hh_lock); - memcpy(cache_data, (void *)rt_dst(rt).hh->hh_data + hh_off, hh_len); - } while (read_seqretry(&rt_dst(rt).hh->hh_lock, hh_seq)); - - cache->hh_seq = hh_seq; + struct rtable *rt; + /* Tunnel configuration keeps DSCP part of TOS bits, But Linux + * router expect RT_TOS bits only. */ + +#if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,39) + struct flowi fl = { .nl_u = { .ip4_u = { + .daddr = daddr, + .saddr = *saddr, + .tos = RT_TOS(tos) } }, + .proto = ipproto }; + + if (unlikely(ip_route_output_key(net, &rt, &fl))) + return ERR_PTR(-EADDRNOTAVAIL); + *saddr = fl.nl_u.ip4_u.saddr; + return rt; #else - read_lock_bh(&rt_dst(rt).hh->hh_lock); - memcpy(cache_data, (void *)rt_dst(rt).hh->hh_data + hh_off, hh_len); - read_unlock_bh(&rt_dst(rt).hh->hh_lock); + struct flowi4 fl = { .daddr = daddr, + .saddr = *saddr, + .flowi4_tos = RT_TOS(tos), + .flowi4_proto = ipproto }; + + rt = ip_route_output_key(net, &fl); + *saddr = fl.saddr; + return rt; #endif } -static struct tnl_cache *build_cache(struct vport *vport, - const struct tnl_mutable_config *mutable, - struct rtable *rt) -{ - struct tnl_vport *tnl_vport = tnl_vport_priv(vport); - struct tnl_cache *cache; - void *cache_data; - int cache_len; - - if (!(mutable->port_config.flags & TNL_F_HDR_CACHE)) - return NULL; - - /* - * If there is no entry in the ARP cache or if this device does not - * support hard header caching just fall back to the IP stack. - */ - if (!rt_dst(rt).hh) - return NULL; - - /* - * If lock is contended fall back to directly building the header. - * We're not going to help performance by sitting here spinning. - */ - if (!spin_trylock_bh(&tnl_vport->cache_lock)) - return NULL; - - cache = tnl_vport->cache; - if (check_cache_valid(cache, mutable)) - goto unlock; - else - cache = NULL; - - cache_len = rt_dst(rt).hh->hh_len + mutable->tunnel_hlen; - - cache = kzalloc(ALIGN(sizeof(struct tnl_cache), CACHE_DATA_ALIGN) + - cache_len, GFP_ATOMIC); - if (!cache) - goto unlock; - - cache->len = cache_len; - - create_eth_hdr(cache, rt); - cache_data = get_cached_header(cache) + rt_dst(rt).hh->hh_len; - - create_tunnel_header(vport, mutable, rt, cache_data); - - cache->mutable_seq = mutable->seq; - cache->rt = rt; -#ifdef NEED_CACHE_TIMEOUT - cache->expiration = jiffies + tnl_vport->cache_exp_interval; -#endif - - if (is_internal_dev(rt_dst(rt).dev)) { - struct odp_flow_key flow_key; - struct tbl_node *flow_node; - struct vport *vport; - struct sk_buff *skb; - bool is_frag; - int err; - - vport = internal_dev_get_vport(rt_dst(rt).dev); - if (!vport) - goto done; - - skb = alloc_skb(cache->len, GFP_ATOMIC); - if (!skb) - goto done; - - __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); - - kfree_skb(skb); - if (err || is_frag) - goto done; - - flow_node = tbl_lookup(rcu_dereference(vport->dp->table), - &flow_key, flow_hash(&flow_key), - flow_cmp); - if (flow_node) { - struct sw_flow *flow = flow_cast(flow_node); - - cache->flow = flow; - flow_hold(flow); - } - } - -done: - assign_cache_rcu(vport, cache); - -unlock: - spin_unlock_bh(&tnl_vport->cache_lock); - - return cache; -} - -static struct rtable *find_route(struct vport *vport, - const struct tnl_mutable_config *mutable, - u8 tos, struct tnl_cache **cache) -{ - struct tnl_vport *tnl_vport = tnl_vport_priv(vport); - struct tnl_cache *cur_cache = rcu_dereference(tnl_vport->cache); - - *cache = NULL; - tos = RT_TOS(tos); - - if (likely(tos == mutable->port_config.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, - .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)) - *cache = build_cache(vport, mutable, rt); - - return rt; - } -} - -static struct sk_buff *check_headroom(struct sk_buff *skb, int headroom) -{ - if (skb_headroom(skb) < headroom || skb_header_cloned(skb)) { - struct sk_buff *nskb = skb_realloc_headroom(skb, headroom + 16); - if (unlikely(!nskb)) { - kfree_skb(skb); - return ERR_PTR(-ENOMEM); - } - - set_skb_csum_bits(skb, nskb); - - if (skb->sk) - skb_set_owner_w(nskb, skb->sk); - - kfree_skb(skb); - return nskb; - } - - return skb; -} - -static inline bool need_linearize(const struct sk_buff *skb) +static bool need_linearize(const struct sk_buff *skb) { int i; @@ -994,7 +745,7 @@ static inline bool need_linearize(const struct sk_buff *skb) * change them from underneath us and we can skip the linearization. */ for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) - if (unlikely(page_count(skb_shinfo(skb)->frags[0].page) > 1)) + if (unlikely(page_count(skb_frag_page(&skb_shinfo(skb)->frags[i])) > 1)) return true; return false; @@ -1002,75 +753,59 @@ static inline bool need_linearize(const struct sk_buff *skb) static struct sk_buff *handle_offloads(struct sk_buff *skb, const struct tnl_mutable_config *mutable, - const struct rtable *rt) + const struct rtable *rt, + int tunnel_hlen) { int min_headroom; int err; - forward_ip_summed(skb); - - err = vswitch_skb_checksum_setup(skb); - if (unlikely(err)) - goto error_free; - min_headroom = LL_RESERVED_SPACE(rt_dst(rt).dev) + rt_dst(rt).header_len - + mutable->tunnel_hlen; + + tunnel_hlen + + (vlan_tx_tag_present(skb) ? VLAN_HLEN : 0); + + if (skb_headroom(skb) < min_headroom || skb_header_cloned(skb)) { + int head_delta = SKB_DATA_ALIGN(min_headroom - + skb_headroom(skb) + + 16); + err = pskb_expand_head(skb, max_t(int, head_delta, 0), + 0, GFP_ATOMIC); + if (unlikely(err)) + goto error_free; + } + + forward_ip_summed(skb, true); if (skb_is_gso(skb)) { struct sk_buff *nskb; - /* - * If we are doing GSO on a pskb it is better to make sure that - * the headroom is correct now. We will only have to copy the - * portion in the linear data area and GSO will preserve - * headroom when it creates the segments. This is particularly - * beneficial on Xen where we get a lot of GSO pskbs. - * Conversely, we avoid copying if it is just to get our own - * writable clone because GSO will do the copy for us. - */ - if (skb_headroom(skb) < min_headroom) { - skb = check_headroom(skb, min_headroom); - if (unlikely(IS_ERR(skb))) { - err = PTR_ERR(skb); - goto error; - } - } - nskb = skb_gso_segment(skb, 0); - kfree_skb(skb); - if (unlikely(IS_ERR(nskb))) { + if (IS_ERR(nskb)) { + kfree_skb(skb); err = PTR_ERR(nskb); goto error; } + consume_skb(skb); skb = nskb; - } else { - skb = check_headroom(skb, min_headroom); - if (unlikely(IS_ERR(skb))) { - err = PTR_ERR(skb); - goto error; - } - - if (skb->ip_summed == CHECKSUM_PARTIAL) { - /* - * Pages aren't locked and could change at any time. - * If this happens after we compute the checksum, the - * checksum will be wrong. We linearize now to avoid - * this problem. - */ - if (unlikely(need_linearize(skb))) { - err = __skb_linearize(skb); - if (unlikely(err)) - goto error_free; - } - - err = skb_checksum_help(skb); + } else if (get_ip_summed(skb) == OVS_CSUM_PARTIAL) { + /* Pages aren't locked and could change at any time. + * If this happens after we compute the checksum, the + * checksum will be wrong. We linearize now to avoid + * this problem. + */ + if (unlikely(need_linearize(skb))) { + err = __skb_linearize(skb); if (unlikely(err)) goto error_free; - } else if (skb->ip_summed == CHECKSUM_COMPLETE) - skb->ip_summed = CHECKSUM_NONE; + } + + err = skb_checksum_help(skb); + if (unlikely(err)) + goto error_free; } + set_ip_summed(skb, OVS_CSUM_NONE); + return skb; error_free: @@ -1080,28 +815,24 @@ error: } static int send_frags(struct sk_buff *skb, - const struct tnl_mutable_config *mutable) + int tunnel_hlen) { int sent_len; - int err; sent_len = 0; while (skb) { struct sk_buff *next = skb->next; - int frag_len = skb->len - mutable->tunnel_hlen; + int frag_len = skb->len - tunnel_hlen; + int err; skb->next = NULL; memset(IPCB(skb), 0, sizeof(*IPCB(skb))); err = ip_local_out(skb); - if (likely(net_xmit_eval(err) == 0)) - sent_len += frag_len; - else { - skb = next; - goto free_frags; - } - skb = next; + if (unlikely(net_xmit_eval(err))) + goto free_frags; + sent_len += frag_len; } return sent_len; @@ -1112,27 +843,28 @@ free_frags: * dropped so just free the rest. This may help improve the congestion * that caused the first packet to be dropped. */ - tnl_free_linked_skbs(skb); + ovs_tnl_free_linked_skbs(skb); return sent_len; } -int tnl_send(struct vport *vport, struct sk_buff *skb) +int ovs_tnl_send(struct vport *vport, struct sk_buff *skb) { struct tnl_vport *tnl_vport = tnl_vport_priv(vport); const struct tnl_mutable_config *mutable = rcu_dereference(tnl_vport->mutable); - enum vport_err_type err = VPORT_E_TX_ERROR; struct rtable *rt; - struct dst_entry *unattached_dst = NULL; - struct tnl_cache *cache; + struct ovs_key_ipv4_tunnel tun_key; int sent_len = 0; - __be16 frag_off; + int tunnel_hlen; + __be16 frag_off = 0; + __be32 daddr; + __be32 saddr; u8 ttl; - u8 inner_tos; u8 tos; /* Validate the protocol headers before we try to use them. */ - if (skb->protocol == htons(ETH_P_8021Q)) { + if (skb->protocol == htons(ETH_P_8021Q) && + !vlan_tx_tag_present(skb)) { if (unlikely(!pskb_may_pull(skb, VLAN_ETH_HLEN))) goto error_free; @@ -1153,69 +885,99 @@ int tnl_send(struct vport *vport, struct sk_buff *skb) } #endif - /* ToS */ - if (skb->protocol == htons(ETH_P_IP)) - inner_tos = ip_hdr(skb)->tos; + /* If OVS_CB(skb)->tun_key is NULL, point it at the local tun_key here, + * and zero it out. + */ + if (!OVS_CB(skb)->tun_key) { + memset(&tun_key, 0, sizeof(tun_key)); + OVS_CB(skb)->tun_key = &tun_key; + } + + tunnel_hlen = tnl_vport->tnl_ops->hdr_len(mutable, OVS_CB(skb)->tun_key); + if (unlikely(tunnel_hlen < 0)) { + err = VPORT_E_TX_DROPPED; + goto error_free; + } + tunnel_hlen += sizeof(struct iphdr); + + if (OVS_CB(skb)->tun_key->ipv4_dst) { + daddr = OVS_CB(skb)->tun_key->ipv4_dst; + saddr = OVS_CB(skb)->tun_key->ipv4_src; + tos = OVS_CB(skb)->tun_key->ipv4_tos; + ttl = OVS_CB(skb)->tun_key->ipv4_ttl; + } else { + u8 inner_tos; + daddr = mutable->key.daddr; + saddr = mutable->key.saddr; + + if (unlikely(!daddr)) { + /* Trying to sent packet from Null-port without + * tunnel info? Drop this packet. */ + err = VPORT_E_TX_DROPPED; + goto error_free; + } + + /* ToS */ + if (skb->protocol == htons(ETH_P_IP)) + inner_tos = ip_hdr(skb)->tos; #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE) - else if (skb->protocol == htons(ETH_P_IPV6)) - inner_tos = ipv6_get_dsfield(ipv6_hdr(skb)); + else if (skb->protocol == htons(ETH_P_IPV6)) + inner_tos = ipv6_get_dsfield(ipv6_hdr(skb)); #endif - else - inner_tos = 0; + else + inner_tos = 0; - if (mutable->port_config.flags & TNL_F_TOS_INHERIT) - tos = inner_tos; - else - tos = mutable->port_config.tos; + if (mutable->flags & TNL_F_TOS_INHERIT) + tos = inner_tos; + else + tos = mutable->tos; + + tos = INET_ECN_encapsulate(tos, inner_tos); - tos = INET_ECN_encapsulate(tos, inner_tos); + /* TTL */ + ttl = mutable->ttl; + 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) + else if (skb->protocol == htons(ETH_P_IPV6)) + ttl = ipv6_hdr(skb)->hop_limit; +#endif + } + + } /* Route lookup */ - rt = find_route(vport, mutable, tos, &cache); - if (unlikely(!rt)) + rt = find_route(port_key_get_net(&mutable->key), &saddr, daddr, + tnl_vport->tnl_ops->ipproto, tos); + if (IS_ERR(rt)) goto error_free; - if (unlikely(!cache)) - unattached_dst = &rt_dst(rt); /* Reset SKB */ nf_reset(skb); secpath_reset(skb); skb_dst_drop(skb); + skb_clear_rxhash(skb); /* Offloading */ - skb = handle_offloads(skb, mutable, rt); - if (unlikely(IS_ERR(skb))) - goto error; + skb = handle_offloads(skb, mutable, rt, tunnel_hlen); + if (IS_ERR(skb)) { + skb = NULL; + goto err_free_rt; + } /* MTU */ - if (unlikely(!check_mtu(skb, vport, mutable, rt, &frag_off))) { + if (unlikely(!check_mtu(skb, vport, mutable, rt, &frag_off, tunnel_hlen))) { err = VPORT_E_TX_DROPPED; - goto error_free; + goto err_free_rt; } - /* - * If we are over the MTU, allow the IP stack to handle fragmentation. - * Fragmentation is a slow path anyways. - */ - if (unlikely(skb->len + mutable->tunnel_hlen > dst_mtu(&rt_dst(rt)) && - cache)) { - unattached_dst = &rt_dst(rt); - dst_hold(unattached_dst); - cache = NULL; - } - - /* TTL */ - ttl = mutable->port_config.ttl; - if (!ttl) - ttl = dst_metric(&rt_dst(rt), RTAX_HOPLIMIT); - - if (mutable->port_config.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) - else if (skb->protocol == htons(ETH_P_IPV6)) - ttl = ipv6_hdr(skb)->hop_limit; -#endif + /* TTL Fixup. */ + if (!OVS_CB(skb)->tun_key->ipv4_dst) { + if (!(mutable->flags & TNL_F_TTL_INHERIT)) { + if (!ttl) + ttl = ip4_dst_hoplimit(&rt_dst(rt)); + } } while (skb) { @@ -1223,132 +985,164 @@ int tnl_send(struct vport *vport, struct sk_buff *skb) struct sk_buff *next_skb = skb->next; skb->next = NULL; - if (likely(cache)) { - skb_push(skb, cache->len); - memcpy(skb->data, get_cached_header(cache), cache->len); - skb_reset_mac_header(skb); - skb_set_network_header(skb, rt_dst(rt).hh->hh_len); - - } else { - skb_push(skb, mutable->tunnel_hlen); - create_tunnel_header(vport, mutable, rt, skb->data); - skb_reset_network_header(skb); - - if (next_skb) - skb_dst_set(skb, dst_clone(unattached_dst)); - else { - skb_dst_set(skb, unattached_dst); - unattached_dst = NULL; - } - } - skb_set_transport_header(skb, skb_network_offset(skb) + sizeof(struct iphdr)); + if (unlikely(vlan_deaccel_tag(skb))) + goto next; + + skb_push(skb, tunnel_hlen); + skb_reset_network_header(skb); + skb_set_transport_header(skb, sizeof(struct iphdr)); + if (next_skb) + skb_dst_set(skb, dst_clone(&rt_dst(rt))); + else + skb_dst_set(skb, &rt_dst(rt)); + + /* Push IP header. */ iph = ip_hdr(skb); - iph->tos = tos; - iph->ttl = ttl; - iph->frag_off = frag_off; + iph->version = 4; + iph->ihl = sizeof(struct iphdr) >> 2; + iph->protocol = tnl_vport->tnl_ops->ipproto; + iph->daddr = daddr; + iph->saddr = saddr; + iph->tos = tos; + iph->ttl = ttl; + iph->frag_off = frag_off; ip_select_ident(iph, &rt_dst(rt), NULL); - skb = tnl_vport->tnl_ops->update_header(vport, mutable, &rt_dst(rt), skb); + /* Push Tunnel header. */ + skb = tnl_vport->tnl_ops->build_header(vport, mutable, + &rt_dst(rt), skb, tunnel_hlen); if (unlikely(!skb)) goto next; - if (likely(cache)) { - int orig_len = skb->len - cache->len; - struct vport *cache_vport = internal_dev_get_vport(rt_dst(rt).dev); - - skb->protocol = htons(ETH_P_IP); - iph->tot_len = htons(skb->len - skb_network_offset(skb)); - ip_send_check(iph); - - if (cache_vport) { - OVS_CB(skb)->flow = cache->flow; - compute_ip_summed(skb, true); - vport_receive(cache_vport, skb); - sent_len += orig_len; - } else { - int err; - - skb->dev = rt_dst(rt).dev; - err = dev_queue_xmit(skb); - - if (likely(net_xmit_eval(err) == 0)) - sent_len += orig_len; - } - } else - sent_len += send_frags(skb, mutable); + sent_len += send_frags(skb, tunnel_hlen); next: skb = next_skb; } if (unlikely(sent_len == 0)) - vport_record_error(vport, VPORT_E_TX_DROPPED); + ovs_vport_record_error(vport, VPORT_E_TX_DROPPED); - goto out; + return sent_len; +err_free_rt: + ip_rt_put(rt); error_free: - tnl_free_linked_skbs(skb); -error: - dst_release(unattached_dst); - vport_record_error(vport, err); -out: + ovs_tnl_free_linked_skbs(skb); + ovs_vport_record_error(vport, err); 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[OVS_TUNNEL_ATTR_MAX + 1] = { + [OVS_TUNNEL_ATTR_FLAGS] = { .type = NLA_U32 }, + [OVS_TUNNEL_ATTR_DST_IPV4] = { .type = NLA_U32 }, + [OVS_TUNNEL_ATTR_SRC_IPV4] = { .type = NLA_U32 }, + [OVS_TUNNEL_ATTR_OUT_KEY] = { .type = NLA_U64 }, + [OVS_TUNNEL_ATTR_IN_KEY] = { .type = NLA_U64 }, + [OVS_TUNNEL_ATTR_TOS] = { .type = NLA_U8 }, + [OVS_TUNNEL_ATTR_TTL] = { .type = NLA_U8 }, + [OVS_TUNNEL_ATTR_DST_PORT] = { .type = NLA_U16 }, +}; + +/* Sets OVS_TUNNEL_ATTR_* fields in 'mutable', which must initially be + * zeroed. */ +static int tnl_set_config(struct net *net, 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[OVS_TUNNEL_ATTR_MAX + 1]; + int err; - mutable->port_config = *(struct tnl_port_config *)config; + port_key_set_net(&mutable->key, net); + mutable->key.tunnel_type = tnl_ops->tunnel_type; + if (!options) + goto out; - if (mutable->port_config.daddr == 0) - return -EINVAL; + err = nla_parse_nested(a, OVS_TUNNEL_ATTR_MAX, options, tnl_policy); + if (err) + return err; - if (mutable->port_config.tos != RT_TOS(mutable->port_config.tos)) + if (!a[OVS_TUNNEL_ATTR_FLAGS] || !a[OVS_TUNNEL_ATTR_DST_IPV4]) return -EINVAL; - mutable->tunnel_hlen = tnl_ops->hdr_len(&mutable->port_config); - if (mutable->tunnel_hlen < 0) - return mutable->tunnel_hlen; + mutable->flags = nla_get_u32(a[OVS_TUNNEL_ATTR_FLAGS]) & TNL_F_PUBLIC; + mutable->key.daddr = nla_get_be32(a[OVS_TUNNEL_ATTR_DST_IPV4]); - mutable->tunnel_hlen += sizeof(struct iphdr); + if (a[OVS_TUNNEL_ATTR_SRC_IPV4]) { + if (ipv4_is_multicast(mutable->key.daddr)) + return -EINVAL; + mutable->key.saddr = nla_get_be32(a[OVS_TUNNEL_ATTR_SRC_IPV4]); + } - mutable->tunnel_type = tnl_ops->tunnel_type; - if (mutable->port_config.flags & TNL_F_IN_KEY_MATCH) { - mutable->tunnel_type |= TNL_T_KEY_MATCH; - mutable->port_config.in_key = 0; - } else - mutable->tunnel_type |= TNL_T_KEY_EXACT; + if (a[OVS_TUNNEL_ATTR_TOS]) { + mutable->tos = nla_get_u8(a[OVS_TUNNEL_ATTR_TOS]); + /* Reject ToS config with ECN bits set. */ + if (mutable->tos & INET_ECN_MASK) + return -EINVAL; + } - old_vport = tnl_find_port(mutable->port_config.saddr, - mutable->port_config.daddr, - mutable->port_config.in_key, - mutable->tunnel_type, - &old_mutable); + if (a[OVS_TUNNEL_ATTR_TTL]) + mutable->ttl = nla_get_u8(a[OVS_TUNNEL_ATTR_TTL]); + + if (a[OVS_TUNNEL_ATTR_DST_PORT]) + mutable->dst_port = + htons(nla_get_u16(a[OVS_TUNNEL_ATTR_DST_PORT])); + + if (!a[OVS_TUNNEL_ATTR_IN_KEY]) { + mutable->key.tunnel_type |= TNL_T_KEY_MATCH; + mutable->flags |= TNL_F_IN_KEY_MATCH; + } else { + mutable->key.tunnel_type |= TNL_T_KEY_EXACT; + mutable->key.in_key = nla_get_be64(a[OVS_TUNNEL_ATTR_IN_KEY]); + } + + if (!a[OVS_TUNNEL_ATTR_OUT_KEY]) + mutable->flags |= TNL_F_OUT_KEY_ACTION; + else + mutable->out_key = nla_get_be64(a[OVS_TUNNEL_ATTR_OUT_KEY]); + mutable->mlink = 0; + if (ipv4_is_multicast(mutable->key.daddr)) { + struct net_device *dev; + struct rtable *rt; + __be32 saddr = mutable->key.saddr; + + rt = find_route(port_key_get_net(&mutable->key), + &saddr, mutable->key.daddr, + tnl_ops->ipproto, mutable->tos); + if (IS_ERR(rt)) + return -EADDRNOTAVAIL; + dev = rt_dst(rt).dev; + ip_rt_put(rt); + if (__in_dev_get_rtnl(dev) == NULL) + return -EADDRNOTAVAIL; + mutable->mlink = dev->ifindex; + ip_mc_inc_group(__in_dev_get_rtnl(dev), mutable->key.daddr); + } + +out: + old_vport = port_table_lookup(&mutable->key, &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; } -struct vport *tnl_create(const struct vport_parms *parms, - const struct vport_ops *vport_ops, - const struct tnl_ops *tnl_ops) +struct vport *ovs_tnl_create(const struct vport_parms *parms, + const struct vport_ops *vport_ops, + const struct tnl_ops *tnl_ops) { struct vport *vport; struct tnl_vport *tnl_vport; + struct tnl_mutable_config *mutable; int initial_frag_id; int err; - vport = vport_alloc(sizeof(struct tnl_vport), vport_ops, parms); + vport = ovs_vport_alloc(sizeof(struct tnl_vport), vport_ops, parms); if (IS_ERR(vport)) { err = PTR_ERR(vport); goto error; @@ -1359,158 +1153,185 @@ 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; + random_ether_addr(mutable->eth_addr); 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(ovs_dp_get_net(parms->dp), parms->options, tnl_ops, + NULL, mutable); if (err) goto error_free_mutable; - spin_lock_init(&tnl_vport->cache_lock); - -#ifdef NEED_CACHE_TIMEOUT - tnl_vport->cache_exp_interval = MAX_CACHE_EXP - - (net_random() % (MAX_CACHE_EXP / 2)); -#endif - - err = add_port(vport); - if (err) - goto error_free_mutable; + rcu_assign_pointer(tnl_vport->mutable, mutable); + port_table_add_port(vport); return vport; error_free_mutable: - kfree(tnl_vport->mutable); + free_mutable_rtnl(mutable); + kfree(mutable); error_free_vport: - vport_free(vport); + ovs_vport_free(vport); error: return ERR_PTR(err); } -int tnl_modify(struct vport *vport, struct odp_port *port) +int ovs_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); + old_mutable = rtnl_dereference(tnl_vport->mutable); + if (!old_mutable->key.daddr) + return -EINVAL; + + 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); - if (err) - goto error_free; - - mutable->seq++; + /* Copy fields whose values should be retained. */ + mutable->seq = old_mutable->seq + 1; + memcpy(mutable->eth_addr, old_mutable->eth_addr, ETH_ALEN); - err = move_port(vport, mutable); + /* Parse the others configured by userspace. */ + err = tnl_set_config(ovs_dp_get_net(vport->dp), options, tnl_vport->tnl_ops, + vport, mutable); if (err) goto error_free; + if (port_hash(&mutable->key) != port_hash(&old_mutable->key)) + port_table_move_port(vport, mutable); + else + assign_config_rcu(vport, mutable); + return 0; error_free: + free_mutable_rtnl(mutable); kfree(mutable); error: return err; } -static void free_port_rcu(struct rcu_head *rcu) +int ovs_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); + + if (nla_put_u32(skb, OVS_TUNNEL_ATTR_FLAGS, + mutable->flags & TNL_F_PUBLIC) || + nla_put_be32(skb, OVS_TUNNEL_ATTR_DST_IPV4, mutable->key.daddr)) + goto nla_put_failure; + + if (!(mutable->flags & TNL_F_IN_KEY_MATCH) && + nla_put_be64(skb, OVS_TUNNEL_ATTR_IN_KEY, mutable->key.in_key)) + goto nla_put_failure; + if (!(mutable->flags & TNL_F_OUT_KEY_ACTION) && + nla_put_be64(skb, OVS_TUNNEL_ATTR_OUT_KEY, mutable->out_key)) + goto nla_put_failure; + if (mutable->key.saddr && + nla_put_be32(skb, OVS_TUNNEL_ATTR_SRC_IPV4, mutable->key.saddr)) + goto nla_put_failure; + if (mutable->tos && nla_put_u8(skb, OVS_TUNNEL_ATTR_TOS, mutable->tos)) + goto nla_put_failure; + if (mutable->ttl && nla_put_u8(skb, OVS_TUNNEL_ATTR_TTL, mutable->ttl)) + goto nla_put_failure; + if (mutable->dst_port && nla_put_u16(skb, OVS_TUNNEL_ATTR_DST_PORT, + ntohs(mutable->dst_port))) + goto nla_put_failure; - 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); - vport_free(tnl_vport_to_vport(tnl_vport)); +nla_put_failure: + return -EMSGSIZE; } -int tnl_destroy(struct vport *vport) +static void free_port_rcu(struct rcu_head *rcu) { - struct tnl_vport *tnl_vport = tnl_vport_priv(vport); - const struct tnl_mutable_config *old_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)) - del_port(vport); - - call_rcu(&tnl_vport->rcu, free_port_rcu); + struct tnl_vport *tnl_vport = container_of(rcu, + struct tnl_vport, rcu); - return 0; + kfree((struct tnl_mutable __force *)tnl_vport->mutable); + ovs_vport_free(tnl_vport_to_vport(tnl_vport)); } -int tnl_set_mtu(struct vport *vport, int mtu) +void ovs_tnl_destroy(struct vport *vport) { 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); - if (!mutable) - return -ENOMEM; - - mutable->mtu = mtu; - assign_config_rcu(vport, mutable); - - return 0; + mutable = rtnl_dereference(tnl_vport->mutable); + port_table_remove_port(vport); + free_mutable_rtnl(mutable); + call_rcu(&tnl_vport->rcu, free_port_rcu); } -int tnl_set_addr(struct vport *vport, const unsigned char *addr) +int ovs_tnl_set_addr(struct vport *vport, const unsigned char *addr) { struct tnl_vport *tnl_vport = tnl_vport_priv(vport); - struct tnl_mutable_config *mutable; + struct tnl_mutable_config *old_mutable, *mutable; - mutable = kmemdup(tnl_vport->mutable, sizeof(struct tnl_mutable_config), GFP_KERNEL); + old_mutable = rtnl_dereference(tnl_vport->mutable); + mutable = kmemdup(old_mutable, sizeof(struct tnl_mutable_config), GFP_KERNEL); if (!mutable) return -ENOMEM; + old_mutable->mlink = 0; + memcpy(mutable->eth_addr, addr, ETH_ALEN); assign_config_rcu(vport, mutable); return 0; } -const char *tnl_get_name(const struct vport *vport) +const char *ovs_tnl_get_name(const struct vport *vport) { const struct tnl_vport *tnl_vport = tnl_vport_priv(vport); return tnl_vport->name; } -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; -} - -int tnl_get_mtu(const struct vport *vport) +const unsigned char *ovs_tnl_get_addr(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)->eth_addr; } -void tnl_free_linked_skbs(struct sk_buff *skb) +void ovs_tnl_free_linked_skbs(struct sk_buff *skb) { - if (unlikely(!skb)) - return; - while (skb) { struct sk_buff *next = skb->next; kfree_skb(skb); skb = next; } } + +int ovs_tnl_init(void) +{ + int i; + + port_table = kmalloc(PORT_TABLE_SIZE * sizeof(struct hlist_head *), + GFP_KERNEL); + if (!port_table) + return -ENOMEM; + + for (i = 0; i < PORT_TABLE_SIZE; i++) + INIT_HLIST_HEAD(&port_table[i]); + + return 0; +} + +void ovs_tnl_exit(void) +{ + kfree(port_table); +}