/*
- * 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 <linux/if_arp.h>
#include <linux/if_ether.h>
#include <linux/ip.h>
#include <linux/if_vlan.h>
+#include <linux/igmp.h>
#include <linux/in.h>
#include <linux/in_route.h>
+#include <linux/inetdevice.h>
#include <linux/jhash.h>
+#include <linux/list.h>
#include <linux/kernel.h>
#include <linux/version.h>
+#include <linux/workqueue.h>
+#include <linux/rculist.h>
#include <net/dsfield.h>
#include <net/dst.h>
#include <net/route.h>
#include <net/xfrm.h>
-#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"
-/* Protected by RCU. */
-static struct tbl *port_table;
+#define PORT_TABLE_SIZE 1024
+
+static struct hlist_head *port_table __read_mostly;
/*
* These are just used as an optimization: they don't require any kind of
* synchronization because we could have just as easily read the value before
* the port change happened.
*/
-static unsigned int key_local_remote_ports;
-static unsigned int key_remote_ports;
-static unsigned int local_remote_ports;
-static unsigned int remote_ports;
+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)
#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)
+static void free_config_rcu(struct rcu_head *rcu)
{
- return container_of(node, struct tnl_vport, tbl_node);
+ struct tnl_mutable_config *c = container_of(rcu, struct tnl_mutable_config, rcu);
+ kfree(c);
}
-/* RCU callback. */
-static void free_config(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_mutable_config *c = container_of(rcu, struct tnl_mutable_config, rcu);
- kfree(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,
struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
struct tnl_mutable_config *old_config;
- old_config = rcu_dereference(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);
+
+ 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;
}
}
-enum lookup_key {
- LOOKUP_TUNNEL_TYPE = 0,
- LOOKUP_SADDR = 1,
- LOOKUP_DADDR = 2,
- LOOKUP_KEY = 3,
-};
+static u32 port_hash(const struct port_lookup_key *key)
+{
+ return jhash2((u32 *)key, (PORT_KEY_LEN / sizeof(u32)), 0);
+}
-struct port_lookup_key {
- u32 vals[4]; /* Contains enum lookup_key keys. */
- const struct tnl_mutable_config *mutable;
-};
+static struct hlist_head *find_bucket(u32 hash)
+{
+ return &port_table[(hash & (PORT_TABLE_SIZE - 1))];
+}
-/*
- * 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 void port_table_add_port(struct vport *vport)
{
- const struct tnl_vport *tnl_vport = tnl_vport_table_cast(node);
- struct port_lookup_key *lookup = target;
+ struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
+ const struct tnl_mutable_config *mutable;
+ u32 hash;
- lookup->mutable = rcu_dereference(tnl_vport->mutable);
+ mutable = rtnl_dereference(tnl_vport->mutable);
+ hash = port_hash(&mutable->key);
+ hlist_add_head_rcu(&tnl_vport->hash_node, find_bucket(hash));
- return (lookup->mutable->tunnel_type == lookup->vals[LOOKUP_TUNNEL_TYPE]) &&
- lookup->mutable->port_config.daddr == lookup->vals[LOOKUP_DADDR] &&
- lookup->mutable->port_config.in_key == lookup->vals[LOOKUP_KEY] &&
- lookup->mutable->port_config.saddr == lookup->vals[LOOKUP_SADDR];
+ (*find_port_pool(rtnl_dereference(tnl_vport->mutable)))++;
}
-static u32 port_hash(struct port_lookup_key *lookup)
+static void port_table_move_port(struct vport *vport,
+ struct tnl_mutable_config *new_mutable)
{
- return jhash2(lookup->vals, ARRAY_SIZE(lookup->vals), 0);
+ struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
+ u32 hash;
+
+ 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));
+
+ (*find_port_pool(rtnl_dereference(tnl_vport->mutable)))--;
+ assign_config_rcu(vport, new_mutable);
+ (*find_port_pool(rtnl_dereference(tnl_vport->mutable)))++;
}
-static int add_port(struct vport *vport)
+static void port_table_remove_port(struct vport *vport)
{
struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
- struct port_lookup_key lookup;
- int err;
- if (!port_table) {
- struct tbl *new_table;
+ hlist_del_init_rcu(&tnl_vport->hash_node);
- new_table = tbl_create(0);
- if (!new_table)
- return -ENOMEM;
+ (*find_port_pool(rtnl_dereference(tnl_vport->mutable)))--;
+}
- rcu_assign_pointer(port_table, new_table);
+static struct vport *port_table_lookup(struct port_lookup_key *key,
+ const struct tnl_mutable_config **pmutable)
+{
+ struct hlist_node *n;
+ struct hlist_head *bucket;
+ u32 hash = port_hash(key);
+ struct tnl_vport *tnl_vport;
- } else if (tbl_count(port_table) > tbl_n_buckets(port_table)) {
- struct tbl *old_table = port_table;
- struct tbl *new_table;
+ bucket = find_bucket(hash);
- new_table = tbl_expand(old_table);
- if (IS_ERR(new_table))
- return PTR_ERR(new_table);
+ hlist_for_each_entry_rcu(tnl_vport, n, bucket, hash_node) {
+ struct tnl_mutable_config *mutable;
- rcu_assign_pointer(port_table, new_table);
- tbl_deferred_destroy(old_table, NULL);
+ mutable = rcu_dereference_rtnl(tnl_vport->mutable);
+ if (!memcmp(&mutable->key, key, PORT_KEY_LEN)) {
+ *pmutable = mutable;
+ return tnl_vport_to_vport(tnl_vport);
+ }
}
- lookup.vals[LOOKUP_SADDR] = tnl_vport->mutable->port_config.saddr;
- lookup.vals[LOOKUP_DADDR] = tnl_vport->mutable->port_config.daddr;
- lookup.vals[LOOKUP_KEY] = tnl_vport->mutable->port_config.in_key;
- lookup.vals[LOOKUP_TUNNEL_TYPE] = tnl_vport->mutable->tunnel_type;
-
- err = tbl_insert(port_table, &tnl_vport->tbl_node, port_hash(&lookup));
- if (err)
- return err;
-
- (*find_port_pool(tnl_vport->mutable))++;
-
- return 0;
+ return NULL;
}
-static int del_port(struct vport *vport)
+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 tnl_vport *tnl_vport = tnl_vport_priv(vport);
- int err;
+ struct port_lookup_key lookup;
+ struct vport *vport;
+ bool is_multicast = ipv4_is_multicast(saddr);
+
+ port_key_set_net(&lookup, net);
+ lookup.saddr = saddr;
+ lookup.daddr = daddr;
+
+ /* 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;
- err = tbl_remove(port_table, &tnl_vport->tbl_node);
- if (err)
- return err;
+ lookup.saddr = saddr;
+ }
+
+ /* 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;
+ }
- (*find_port_pool(tnl_vport->mutable))--;
+ 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 (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;
+ }
+ }
- return 0;
+ 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;
}
-struct vport *tnl_find_port(__be32 saddr, __be32 daddr, __be32 key,
- int tunnel_type,
- const struct tnl_mutable_config **mutable)
+static void ecn_decapsulate(struct sk_buff *skb)
{
- struct port_lookup_key lookup;
- struct tbl *table = rcu_dereference(port_table);
- struct tbl_node *tbl_node;
-
- if (!table)
- return NULL;
+ if (unlikely(INET_ECN_is_ce(OVS_CB(skb)->tun_key->ipv4_tos))) {
+ __be16 protocol = skb->protocol;
- lookup.vals[LOOKUP_SADDR] = saddr;
- lookup.vals[LOOKUP_DADDR] = daddr;
+ skb_set_network_header(skb, ETH_HLEN);
- if (tunnel_type & TNL_T_KEY_EXACT) {
- lookup.vals[LOOKUP_KEY] = key;
- lookup.vals[LOOKUP_TUNNEL_TYPE] = tunnel_type & ~TNL_T_KEY_MATCH;
+ if (protocol == htons(ETH_P_8021Q)) {
+ if (unlikely(!pskb_may_pull(skb, VLAN_ETH_HLEN)))
+ return;
- if (key_local_remote_ports) {
- tbl_node = tbl_lookup(table, &lookup, port_hash(&lookup), port_cmp);
- if (tbl_node)
- goto found;
+ protocol = vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
+ skb_set_network_header(skb, VLAN_ETH_HLEN);
}
- if (key_remote_ports) {
- lookup.vals[LOOKUP_SADDR] = 0;
+ if (protocol == htons(ETH_P_IP)) {
+ if (unlikely(!pskb_may_pull(skb, skb_network_offset(skb)
+ + sizeof(struct iphdr))))
+ return;
- tbl_node = tbl_lookup(table, &lookup, port_hash(&lookup), port_cmp);
- if (tbl_node)
- goto found;
+ 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, skb_network_offset(skb)
+ + sizeof(struct ipv6hdr))))
+ return;
- lookup.vals[LOOKUP_SADDR] = saddr;
+ IP6_ECN_set_ce(ipv6_hdr(skb));
}
+#endif
}
+}
- if (tunnel_type & TNL_T_KEY_MATCH) {
- lookup.vals[LOOKUP_KEY] = 0;
- lookup.vals[LOOKUP_TUNNEL_TYPE] = tunnel_type & ~TNL_T_KEY_EXACT;
+/**
+ * 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)
+{
+ struct ethhdr *eh;
- if (local_remote_ports) {
- tbl_node = tbl_lookup(table, &lookup, port_hash(&lookup), port_cmp);
- if (tbl_node)
- goto found;
- }
+ skb_reset_mac_header(skb);
+ eh = eth_hdr(skb);
- if (remote_ports) {
- lookup.vals[LOOKUP_SADDR] = 0;
+ if (likely(ntohs(eh->h_proto) >= 1536))
+ skb->protocol = eh->h_proto;
+ else
+ skb->protocol = htons(ETH_P_802_2);
- tbl_node = tbl_lookup(table, &lookup, port_hash(&lookup), port_cmp);
- if (tbl_node)
- goto found;
- }
- }
+ skb_dst_drop(skb);
+ nf_reset(skb);
+ skb_clear_rxhash(skb);
+ secpath_reset(skb);
- return NULL;
+ ecn_decapsulate(skb);
+ vlan_set_tci(skb, 0);
+
+ if (unlikely(compute_ip_summed(skb, false))) {
+ kfree_skb(skb);
+ return;
+ }
-found:
- *mutable = lookup.mutable;
- return tnl_vport_to_vport(tnl_vport_table_cast(tbl_node));
+ ovs_vport_receive(vport, skb);
}
static bool check_ipv4_address(__be32 addr)
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);
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;
+ 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;
}
#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;
}
#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 +
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 */
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;
}
-static struct sk_buff *check_headroom(struct sk_buff *skb, int headroom)
+static bool check_mtu(struct sk_buff *skb,
+ struct vport *vport,
+ const struct tnl_mutable_config *mutable,
+ const struct rtable *rt, __be16 *frag_offp,
+ int tunnel_hlen)
{
- 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);
-
- dev_kfree_skb(skb);
- return nskb;
+ 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;
}
- return skb;
-}
+ /* 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;
-static inline u8 ecn_encapsulate(u8 tos, struct sk_buff *skb)
-{
- u8 inner;
+ if (pmtud) {
+ int vlan_header = 0;
- if (skb->protocol == htons(ETH_P_IP))
- inner = ((struct iphdr *)skb_network_header(skb))->tos;
-#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
- else if (skb->protocol == htons(ETH_P_IPV6))
- inner = ipv6_get_dsfield((struct ipv6hdr *)skb_network_header(skb));
-#endif
- else
- inner = 0;
-
- return INET_ECN_encapsulate(tos, inner);
-}
-
-static inline void ecn_decapsulate(struct sk_buff *skb)
-{
- u8 tos = ip_hdr(skb)->tos;
-
- if (INET_ECN_is_ce(tos)) {
- __be16 protocol = skb->protocol;
- unsigned int nw_header = skb_network_header(skb) - skb->data;
+ /* 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;
- if (skb->protocol == htons(ETH_P_8021Q)) {
- if (unlikely(!pskb_may_pull(skb, VLAN_ETH_HLEN)))
- return;
+ mtu = dst_mtu(&rt_dst(rt))
+ - ETH_HLEN
+ - tunnel_hlen
+ - vlan_header;
+ }
- protocol = vlan_eth_hdr(skb)->h_vlan_encapsulated_proto;
- nw_header += VLAN_HLEN;
- }
+ if (skb->protocol == htons(ETH_P_IP)) {
+ struct iphdr *iph = ip_hdr(skb);
- if (protocol == htons(ETH_P_IP)) {
- if (unlikely(!pskb_may_pull(skb, nw_header
- + sizeof(struct iphdr))))
- return;
+ if (df_inherit)
+ frag_off = iph->frag_off & htons(IP_DF);
- IP_ECN_set_ce((struct iphdr *)(nw_header + skb->data));
- }
-#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
- else if (protocol == htons(ETH_P_IPV6)) {
- if (unlikely(!pskb_may_pull(skb, nw_header
- + sizeof(struct ipv6hdr))))
- return;
+ if (pmtud && iph->frag_off & htons(IP_DF)) {
+ mtu = max(mtu, IP_MIN_MTU);
- IP6_ECN_set_ce((struct ipv6hdr *)(nw_header
- + skb->data));
+ if (packet_length > mtu &&
+ ovs_tnl_frag_needed(vport, mutable, skb, mtu))
+ return false;
}
-#endif
}
-}
+#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
+ else if (skb->protocol == htons(ETH_P_IPV6)) {
+ /* 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);
-static struct sk_buff *handle_gso(struct sk_buff *skb)
-{
- if (skb_is_gso(skb)) {
- struct sk_buff *nskb = skb_gso_segment(skb, 0);
+ if (pmtud) {
+ mtu = max(mtu, IPV6_MIN_MTU);
- dev_kfree_skb(skb);
- return nskb;
+ if (packet_length > mtu &&
+ ovs_tnl_frag_needed(vport, mutable, skb, mtu))
+ return false;
+ }
}
+#endif
- return skb;
+ *frag_offp = frag_off;
+ return true;
}
-static int handle_csum_offload(struct sk_buff *skb)
+static struct rtable *find_route(struct net *net,
+ __be32 *saddr, __be32 daddr, u8 ipproto,
+ u8 tos)
{
- if (skb->ip_summed == CHECKSUM_PARTIAL)
- return skb_checksum_help(skb);
- else {
- skb->ip_summed = CHECKSUM_NONE;
- return 0;
- }
+ 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
+ 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
}
-/* Called with rcu_read_lock. */
-void tnl_rcv(struct vport *vport, struct sk_buff *skb)
+static bool need_linearize(const struct sk_buff *skb)
{
- skb->pkt_type = PACKET_HOST;
- skb->protocol = eth_type_trans(skb, skb->dev);
+ int i;
- skb_dst_drop(skb);
- nf_reset(skb);
- secpath_reset(skb);
- skb_reset_network_header(skb);
+ if (unlikely(skb_shinfo(skb)->frag_list))
+ return true;
- ecn_decapsulate(skb);
-
- skb_push(skb, ETH_HLEN);
- compute_ip_summed(skb, false);
+ /*
+ * Generally speaking we should linearize if there are paged frags.
+ * However, if all of the refcounts are 1 we know nobody else can
+ * 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_frag_page(&skb_shinfo(skb)->frags[i])) > 1))
+ return true;
- vport_receive(vport, skb);
+ return false;
}
-static int build_packet(struct vport *vport, const struct tnl_mutable_config *mutable,
- struct iphdr *iph, struct rtable *rt, int max_headroom,
- int mtu, struct sk_buff *skb)
+static struct sk_buff *handle_offloads(struct sk_buff *skb,
+ const struct tnl_mutable_config *mutable,
+ const struct rtable *rt,
+ int tunnel_hlen)
{
- struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
+ int min_headroom;
int err;
- struct iphdr *new_iph;
- int orig_len = skb->len;
- __be16 frag_off = iph->frag_off;
- skb = check_headroom(skb, max_headroom);
- if (unlikely(IS_ERR(skb)))
- goto error;
+ min_headroom = LL_RESERVED_SPACE(rt_dst(rt).dev) + rt_dst(rt).header_len
+ + 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;
+ }
- err = handle_csum_offload(skb);
- if (unlikely(err))
- goto error_free;
+ forward_ip_summed(skb, true);
- if (skb->protocol == htons(ETH_P_IP)) {
- struct iphdr *old_iph = ip_hdr(skb);
+ if (skb_is_gso(skb)) {
+ struct sk_buff *nskb;
- 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 error_free;
+ nskb = skb_gso_segment(skb, 0);
+ if (IS_ERR(nskb)) {
+ kfree_skb(skb);
+ err = PTR_ERR(nskb);
+ goto error;
}
- }
-#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);
-
- /* IPv6 requires PMTUD if the packet is above the minimum MTU. */
- if (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))
+ consume_skb(skb);
+ skb = nskb;
+ } 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;
}
+
+ err = skb_checksum_help(skb);
+ if (unlikely(err))
+ goto error_free;
}
-#endif
- new_iph = (struct iphdr *)skb_push(skb, mutable->tunnel_hlen);
- skb_reset_network_header(skb);
- skb_set_transport_header(skb, sizeof(struct iphdr));
+ set_ip_summed(skb, OVS_CSUM_NONE);
- memcpy(new_iph, iph, sizeof(struct iphdr));
- new_iph->frag_off = frag_off;
- ip_select_ident(new_iph, &rt_dst(rt), NULL);
+ return skb;
+
+error_free:
+ kfree_skb(skb);
+error:
+ return ERR_PTR(err);
+}
- tnl_vport->tnl_ops->build_header(skb, vport, mutable);
+static int send_frags(struct sk_buff *skb,
+ int tunnel_hlen)
+{
+ int sent_len;
- /* Allow our local IP stack to fragment the outer packet even if the
- * DF bit is set as a last resort. */
- skb->local_df = 1;
+ sent_len = 0;
+ while (skb) {
+ struct sk_buff *next = skb->next;
+ int frag_len = skb->len - tunnel_hlen;
+ int err;
- memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
- IPCB(skb)->flags = 0;
+ skb->next = NULL;
+ memset(IPCB(skb), 0, sizeof(*IPCB(skb)));
- err = ip_local_out(skb);
- if (likely(net_xmit_eval(err) == 0))
- return orig_len;
- else {
- vport_record_error(vport, VPORT_E_TX_ERROR);
- return 0;
+ err = ip_local_out(skb);
+ skb = next;
+ if (unlikely(net_xmit_eval(err)))
+ goto free_frags;
+ sent_len += frag_len;
}
-error_free:
- kfree_skb(skb);
-error:
- vport_record_error(vport, VPORT_E_TX_DROPPED);
+ return sent_len;
- return 0;
+free_frags:
+ /*
+ * There's no point in continuing to send fragments once one has been
+ * dropped so just free the rest. This may help improve the congestion
+ * that caused the first packet to be dropped.
+ */
+ 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);
-
- struct iphdr *old_iph;
- int orig_len;
- struct iphdr iph;
+ enum vport_err_type err = VPORT_E_TX_ERROR;
struct rtable *rt;
- int max_headroom;
- int mtu;
+ struct ovs_key_ipv4_tunnel tun_key;
+ int sent_len = 0;
+ int tunnel_hlen;
+ __be16 frag_off = 0;
+ __be32 daddr;
+ __be32 saddr;
+ u8 ttl;
+ 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;
}
if (skb->protocol == htons(ETH_P_IP)) {
- if (unlikely(!pskb_may_pull(skb, skb_network_header(skb)
- + sizeof(struct iphdr) - skb->data)))
+ if (unlikely(!pskb_may_pull(skb, skb_network_offset(skb)
+ + sizeof(struct iphdr))))
skb->protocol = 0;
}
#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
else if (skb->protocol == htons(ETH_P_IPV6)) {
- if (unlikely(!pskb_may_pull(skb, skb_network_header(skb)
- + sizeof(struct ipv6hdr) - skb->data)))
+ if (unlikely(!pskb_may_pull(skb, skb_network_offset(skb)
+ + sizeof(struct ipv6hdr))))
skb->protocol = 0;
}
#endif
- old_iph = ip_hdr(skb);
- iph.tos = mutable->port_config.tos;
- if (mutable->port_config.flags & TNL_F_TOS_INHERIT) {
- if (skb->protocol == htons(ETH_P_IP))
- iph.tos = old_iph->tos;
-#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
- else if (skb->protocol == htons(ETH_P_IPV6))
- iph.tos = ipv6_get_dsfield(ipv6_hdr(skb));
-#endif
+ /* 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;
}
- iph.tos = ecn_encapsulate(iph.tos, skb);
- {
- struct flowi fl = { .nl_u = { .ip4_u =
- { .daddr = mutable->port_config.daddr,
- .saddr = mutable->port_config.saddr,
- .tos = RT_TOS(iph.tos) } },
- .proto = tnl_vport->tnl_ops->ipproto };
+ 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 (unlikely(ip_route_output_key(&init_net, &rt, &fl)))
+ 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;
- }
+ }
- iph.ttl = mutable->port_config.ttl;
- if (mutable->port_config.flags & TNL_F_TTL_INHERIT) {
+ /* ToS */
if (skb->protocol == htons(ETH_P_IP))
- iph.ttl = old_iph->ttl;
+ inner_tos = ip_hdr(skb)->tos;
#if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
else if (skb->protocol == htons(ETH_P_IPV6))
- iph.ttl = ipv6_hdr(skb)->hop_limit;
+ inner_tos = ipv6_get_dsfield(ipv6_hdr(skb));
#endif
- }
- if (!iph.ttl)
- iph.ttl = dst_metric(&rt_dst(rt), RTAX_HOPLIMIT);
+ else
+ inner_tos = 0;
- iph.frag_off = (mutable->port_config.flags & TNL_F_PMTUD) ? htons(IP_DF) : 0;
- if (iph.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;
+ if (mutable->flags & TNL_F_TOS_INHERIT)
+ tos = inner_tos;
+ else
+ tos = mutable->tos;
- if (skb->protocol == htons(ETH_P_IP)) {
- iph.frag_off |= old_iph->frag_off & htons(IP_DF);
- mtu = max(mtu, IP_MIN_MTU);
- }
+ 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))
- mtu = max(mtu, IPV6_MIN_MTU);
+ else if (skb->protocol == htons(ETH_P_IPV6))
+ ttl = ipv6_hdr(skb)->hop_limit;
#endif
+ }
- iph.version = 4;
- iph.ihl = sizeof(struct iphdr) >> 2;
- iph.protocol = tnl_vport->tnl_ops->ipproto;
- iph.daddr = rt->rt_dst;
- iph.saddr = rt->rt_src;
+ }
+
+ /* Route lookup */
+ rt = find_route(port_key_get_net(&mutable->key), &saddr, daddr,
+ tnl_vport->tnl_ops->ipproto, tos);
+ if (IS_ERR(rt))
+ goto error_free;
+ /* Reset SKB */
nf_reset(skb);
secpath_reset(skb);
skb_dst_drop(skb);
- skb_dst_set(skb, &rt_dst(rt));
+ skb_clear_rxhash(skb);
- /*
- * 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
- * lots of GSO pskbs. Conversely, we delay copying if it is just to
- * get our own writable clone because GSO may do the copy for us.
- */
- max_headroom = LL_RESERVED_SPACE(rt_dst(rt).dev) + rt_dst(rt).header_len
- + mutable->tunnel_hlen;
-
- if (skb_headroom(skb) < max_headroom) {
- skb = check_headroom(skb, max_headroom);
- if (unlikely(IS_ERR(skb))) {
- vport_record_error(vport, VPORT_E_TX_DROPPED);
- goto error;
- }
+ /* Offloading */
+ skb = handle_offloads(skb, mutable, rt, tunnel_hlen);
+ if (IS_ERR(skb)) {
+ skb = NULL;
+ goto err_free_rt;
}
- forward_ip_summed(skb);
-
- if (unlikely(vswitch_skb_checksum_setup(skb)))
- goto error_free;
+ /* MTU */
+ if (unlikely(!check_mtu(skb, vport, mutable, rt, &frag_off, tunnel_hlen))) {
+ err = VPORT_E_TX_DROPPED;
+ goto err_free_rt;
+ }
- skb = handle_gso(skb);
- if (unlikely(IS_ERR(skb))) {
- vport_record_error(vport, VPORT_E_TX_DROPPED);
- goto error;
+ /* 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));
+ }
}
- /*
- * Process GSO segments. Try to do any work for the entire packet that
- * doesn't involve actually writing to it before this point.
- */
- orig_len = 0;
- do {
+ while (skb) {
+ struct iphdr *iph;
struct sk_buff *next_skb = skb->next;
skb->next = NULL;
- orig_len += build_packet(vport, mutable, &iph, rt, max_headroom, mtu, skb);
+ 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->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);
+
+ /* Push Tunnel header. */
+ skb = tnl_vport->tnl_ops->build_header(vport, mutable,
+ &rt_dst(rt), skb, tunnel_hlen);
+ if (unlikely(!skb))
+ goto next;
+
+ sent_len += send_frags(skb, tunnel_hlen);
+
+next:
skb = next_skb;
- } while (skb);
+ }
- return orig_len;
+ if (unlikely(sent_len == 0))
+ ovs_vport_record_error(vport, VPORT_E_TX_DROPPED);
-error_free:
- kfree_skb(skb);
- vport_record_error(vport, VPORT_E_TX_ERROR);
-error:
- return 0;
-}
+ return sent_len;
-int tnl_init(void)
-{
- return 0;
+err_free_rt:
+ ip_rt_put(rt);
+error_free:
+ ovs_tnl_free_linked_skbs(skb);
+ ovs_vport_record_error(vport, err);
+ return sent_len;
}
-void tnl_exit(void)
-{
- tbl_destroy(port_table, NULL);
- port_table = NULL;
-}
+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 },
+};
-static int set_config(const void __user *uconfig, const struct tnl_ops *tnl_ops,
- const struct vport *cur_vport,
- struct tnl_mutable_config *mutable)
+/* 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;
- if (copy_from_user(&mutable->port_config, uconfig, sizeof(struct tnl_port_config)))
- return -EFAULT;
-
- mutable->tunnel_hlen = tnl_ops->hdr_len(&mutable->port_config);
- if (mutable->tunnel_hlen < 0)
- return mutable->tunnel_hlen;
+ port_key_set_net(&mutable->key, net);
+ mutable->key.tunnel_type = tnl_ops->tunnel_type;
+ if (!options)
+ goto out;
- mutable->tunnel_hlen += sizeof(struct iphdr);
+ err = nla_parse_nested(a, OVS_TUNNEL_ATTR_MAX, options, tnl_policy);
+ if (err)
+ return err;
- if (mutable->port_config.daddr == 0)
+ if (!a[OVS_TUNNEL_ATTR_FLAGS] || !a[OVS_TUNNEL_ATTR_DST_IPV4])
return -EINVAL;
- 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;
+ 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]);
+
+ 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]);
+ }
+
+ 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;
+ }
+
+ 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]));
- 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_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 char *name, const void __user *config,
- 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);
+ vport = ovs_vport_alloc(sizeof(struct tnl_vport), vport_ops, parms);
if (IS_ERR(vport)) {
err = PTR_ERR(vport);
goto error;
tnl_vport = tnl_vport_priv(vport);
- strcpy(tnl_vport->name, name);
+ strcpy(tnl_vport->name, parms->name);
tnl_vport->tnl_ops = tnl_ops;
- tnl_vport->mutable = kmalloc(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);
- err = set_config(config, tnl_ops, NULL, tnl_vport->mutable);
- if (err)
- goto error_free_mutable;
+ get_random_bytes(&initial_frag_id, sizeof(int));
+ atomic_set(&tnl_vport->frag_id, initial_frag_id);
- err = add_port(vport);
+ err = tnl_set_config(ovs_dp_get_net(parms->dp), parms->options, tnl_ops,
+ NULL, mutable);
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, const void __user *config)
+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;
- bool update_hash = false;
- 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(config, tnl_vport->tnl_ops, vport, mutable);
+ /* Copy fields whose values should be retained. */
+ mutable->seq = old_mutable->seq + 1;
+ memcpy(mutable->eth_addr, old_mutable->eth_addr, ETH_ALEN);
+
+ /* 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;
- /*
- * Only remove the port from the hash table if something that would
- * affect the lookup has changed.
- */
- if (tnl_vport->mutable->port_config.saddr != mutable->port_config.saddr ||
- tnl_vport->mutable->port_config.daddr != mutable->port_config.daddr ||
- tnl_vport->mutable->port_config.in_key != mutable->port_config.in_key ||
- (tnl_vport->mutable->port_config.flags & TNL_F_IN_KEY_MATCH) !=
- (mutable->port_config.flags & TNL_F_IN_KEY_MATCH))
- update_hash = true;
-
-
- /*
- * This update is not atomic but the lookup uses the config, which
- * serves as an inherent double check.
- */
- if (update_hash) {
- err = del_port(vport);
- if (err)
- goto error_free;
- }
-
- assign_config_rcu(vport, mutable);
-
- if (update_hash) {
- err = add_port(vport);
- 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(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;
- kfree(tnl_vport->mutable);
- vport_free(tnl_vport_to_vport(tnl_vport));
+ return 0;
+
+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);
+ struct tnl_vport *tnl_vport = container_of(rcu,
+ struct tnl_vport, rcu);
- call_rcu(&tnl_vport->rcu, free_port);
-
- 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 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)->eth_addr;
+ return rcu_dereference_rtnl(tnl_vport->mutable)->eth_addr;
}
-int tnl_get_mtu(const struct vport *vport)
+void ovs_tnl_free_linked_skbs(struct sk_buff *skb)
{
- const struct tnl_vport *tnl_vport = tnl_vport_priv(vport);
- return rcu_dereference(tnl_vport->mutable)->mtu;
+ 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);
}
git://git.onelab.eu / sliver-openvswitch.git / blobdiff