X-Git-Url: https://pintos-os.org/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=datapath%2Fdatapath.c;h=cd29482c508eb50c042f1484d33651ef2ce55fa4;hb=b1184a78452e7160db81e379783cd18937866382;hp=157e106944e1050bb7d38d85766a2cbe5d01dd27;hpb=df2c07f4338faac04f4969f243fe4e8083b309ac;p=openvswitch diff --git a/datapath/datapath.c b/datapath/datapath.c index 157e1069..cd29482c 100644 --- a/datapath/datapath.c +++ b/datapath/datapath.c @@ -39,18 +39,18 @@ #include #include #include +#include #include #include #include #include -#include "openvswitch/datapath-protocol.h" #include "checksum.h" #include "datapath.h" #include "actions.h" #include "flow.h" -#include "table.h" #include "vlan.h" +#include "tunnel.h" #include "vport-internal_dev.h" #if LINUX_VERSION_CODE < KERNEL_VERSION(2,6,18) || \ @@ -107,7 +107,7 @@ struct datapath *get_dp(int dp_ifindex) EXPORT_SYMBOL_GPL(get_dp); /* Must be called with genl_mutex. */ -static struct tbl *get_table_protected(struct datapath *dp) +static struct flow_table *get_table_protected(struct datapath *dp) { return rcu_dereference_protected(dp->table, lockdep_genl_is_held()); } @@ -124,6 +124,24 @@ const char *dp_name(const struct datapath *dp) return vport_get_name(rcu_dereference_rtnl(dp->ports[OVSP_LOCAL])); } +static int get_dpifindex(struct datapath *dp) +{ + struct vport *local; + int ifindex; + + rcu_read_lock(); + + local = get_vport_protected(dp, OVSP_LOCAL); + if (local) + ifindex = vport_get_ifindex(local); + else + ifindex = 0; + + rcu_read_unlock(); + + return ifindex; +} + static inline size_t br_nlmsg_size(void) { return NLMSG_ALIGN(sizeof(struct ifinfomsg)) @@ -131,7 +149,6 @@ static inline size_t br_nlmsg_size(void) + nla_total_size(MAX_ADDR_LEN) /* IFLA_ADDRESS */ + nla_total_size(4) /* IFLA_MASTER */ + nla_total_size(4) /* IFLA_MTU */ - + nla_total_size(4) /* IFLA_LINK */ + nla_total_size(1); /* IFLA_OPERSTATE */ } @@ -142,16 +159,12 @@ static int dp_fill_ifinfo(struct sk_buff *skb, { struct datapath *dp = port->dp; int ifindex = vport_get_ifindex(port); - int iflink = vport_get_iflink(port); struct ifinfomsg *hdr; struct nlmsghdr *nlh; if (ifindex < 0) return ifindex; - if (iflink < 0) - return iflink; - nlh = nlmsg_put(skb, 0, 0, event, sizeof(*hdr), flags); if (nlh == NULL) return -EMSGSIZE; @@ -165,8 +178,7 @@ static int dp_fill_ifinfo(struct sk_buff *skb, hdr->ifi_change = 0; NLA_PUT_STRING(skb, IFLA_IFNAME, vport_get_name(port)); - NLA_PUT_U32(skb, IFLA_MASTER, - vport_get_ifindex(get_vport_protected(dp, OVSP_LOCAL))); + NLA_PUT_U32(skb, IFLA_MASTER, get_dpifindex(dp)); NLA_PUT_U32(skb, IFLA_MTU, vport_get_mtu(port)); #ifdef IFLA_OPERSTATE NLA_PUT_U8(skb, IFLA_OPERSTATE, @@ -177,9 +189,6 @@ static int dp_fill_ifinfo(struct sk_buff *skb, NLA_PUT(skb, IFLA_ADDRESS, ETH_ALEN, vport_get_addr(port)); - if (ifindex != iflink) - NLA_PUT_U32(skb, IFLA_LINK,iflink); - return nlmsg_end(skb, nlh); nla_put_failure: @@ -225,7 +234,7 @@ static void destroy_dp_rcu(struct rcu_head *rcu) { struct datapath *dp = container_of(rcu, struct datapath, rcu); - tbl_destroy((struct tbl __force *)dp->table, flow_free_tbl); + flow_tbl_destroy(dp->table); free_percpu(dp->stats_percpu); kobject_put(&dp->ifobj); } @@ -249,7 +258,7 @@ static struct vport *new_vport(const struct vport_parms *parms) } /* Called with RTNL lock. */ -int dp_detach_port(struct vport *p) +void dp_detach_port(struct vport *p) { ASSERT_RTNL(); @@ -262,22 +271,23 @@ int dp_detach_port(struct vport *p) rcu_assign_pointer(p->dp->ports[p->port_no], NULL); /* Then destroy it. */ - return vport_del(p); + vport_del(p); } /* Must be called with rcu_read_lock. */ void dp_process_received_packet(struct vport *p, struct sk_buff *skb) { struct datapath *dp = p->dp; + struct sw_flow *flow; struct dp_stats_percpu *stats; - int stats_counter_off; + u64 *stats_counter; int error; + stats = per_cpu_ptr(dp->stats_percpu, smp_processor_id()); OVS_CB(skb)->vport = p; if (!OVS_CB(skb)->flow) { struct sw_flow_key key; - struct tbl_node *flow_node; int key_len; bool is_frag; @@ -290,44 +300,38 @@ void dp_process_received_packet(struct vport *p, struct sk_buff *skb) if (is_frag && dp->drop_frags) { consume_skb(skb); - stats_counter_off = offsetof(struct dp_stats_percpu, n_frags); + stats_counter = &stats->n_frags; goto out; } /* Look up flow. */ - flow_node = tbl_lookup(rcu_dereference(dp->table), &key, key_len, - flow_hash(&key, key_len), flow_cmp); - if (unlikely(!flow_node)) { + flow = flow_tbl_lookup(rcu_dereference(dp->table), &key, key_len); + if (unlikely(!flow)) { struct dp_upcall_info upcall; upcall.cmd = OVS_PACKET_CMD_MISS; upcall.key = &key; - upcall.userdata = 0; - upcall.sample_pool = 0; - upcall.actions = NULL; - upcall.actions_len = 0; + upcall.userdata = NULL; + upcall.pid = p->upcall_pid; dp_upcall(dp, skb, &upcall); - stats_counter_off = offsetof(struct dp_stats_percpu, n_missed); + kfree_skb(skb); + stats_counter = &stats->n_missed; goto out; } - OVS_CB(skb)->flow = flow_cast(flow_node); + OVS_CB(skb)->flow = flow; } - stats_counter_off = offsetof(struct dp_stats_percpu, n_hit); + stats_counter = &stats->n_hit; flow_used(OVS_CB(skb)->flow, skb); execute_actions(dp, skb); out: /* Update datapath statistics. */ - local_bh_disable(); - stats = per_cpu_ptr(dp->stats_percpu, smp_processor_id()); write_seqcount_begin(&stats->seqlock); - (*(u64 *)((u8 *)stats + stats_counter_off))++; + (*stats_counter)++; write_seqcount_end(&stats->seqlock); - - local_bh_enable(); } static void copy_and_csum_skb(struct sk_buff *skb, void *to) @@ -353,84 +357,54 @@ static struct genl_family dp_packet_genl_family = { .maxattr = OVS_PACKET_ATTR_MAX }; -/* Generic Netlink multicast groups for upcalls. - * - * We really want three unique multicast groups per datapath, but we can't even - * get one, because genl_register_mc_group() takes genl_lock, which is also - * held during Generic Netlink message processing, so trying to acquire - * multicast groups during OVS_DP_NEW processing deadlocks. Instead, we - * preallocate a few groups and use them round-robin for datapaths. Collision - * isn't fatal--multicast listeners should check that the family is the one - * that they want and discard others--but it wastes time and memory to receive - * unwanted messages. - */ -#define PACKET_N_MC_GROUPS 16 -static struct genl_multicast_group packet_mc_groups[PACKET_N_MC_GROUPS]; - -static u32 packet_mc_group(struct datapath *dp, u8 cmd) -{ - u32 idx; - BUILD_BUG_ON_NOT_POWER_OF_2(PACKET_N_MC_GROUPS); - - idx = jhash_2words(dp->dp_ifindex, cmd, 0) & (PACKET_N_MC_GROUPS - 1); - return packet_mc_groups[idx].id; -} - -static int packet_register_mc_groups(void) -{ - int i; - - for (i = 0; i < PACKET_N_MC_GROUPS; i++) { - struct genl_multicast_group *group = &packet_mc_groups[i]; - int error; - - sprintf(group->name, "packet%d", i); - error = genl_register_mc_group(&dp_packet_genl_family, group); - if (error) - return error; - } - return 0; -} - -int dp_upcall(struct datapath *dp, struct sk_buff *skb, const struct dp_upcall_info *upcall_info) +int dp_upcall(struct datapath *dp, struct sk_buff *skb, + const struct dp_upcall_info *upcall_info) { + struct sk_buff *segs = NULL; struct dp_stats_percpu *stats; int err; - WARN_ON_ONCE(skb_shared(skb)); + if (upcall_info->pid == 0) { + err = -ENOTCONN; + goto err; + } forward_ip_summed(skb, true); /* Break apart GSO packets into their component pieces. Otherwise * userspace may try to stuff a 64kB packet into a 1500-byte MTU. */ if (skb_is_gso(skb)) { - struct sk_buff *nskb = skb_gso_segment(skb, NETIF_F_SG | NETIF_F_HW_CSUM); + segs = skb_gso_segment(skb, NETIF_F_SG | NETIF_F_HW_CSUM); - if (IS_ERR(nskb)) { - kfree_skb(skb); - err = PTR_ERR(nskb); + if (IS_ERR(segs)) { + err = PTR_ERR(segs); goto err; } - consume_skb(skb); - skb = nskb; + skb = segs; } err = queue_userspace_packets(dp, skb, upcall_info); + if (segs) { + struct sk_buff *next; + /* Free GSO-segments */ + do { + next = segs->next; + kfree_skb(segs); + } while ((segs = next) != NULL); + } + if (err) goto err; return 0; err: - local_bh_disable(); stats = per_cpu_ptr(dp->stats_percpu, smp_processor_id()); write_seqcount_begin(&stats->seqlock); stats->n_lost++; write_seqcount_end(&stats->seqlock); - local_bh_enable(); - return err; } @@ -439,63 +413,49 @@ err: * packet. */ static int queue_userspace_packets(struct datapath *dp, struct sk_buff *skb, - const struct dp_upcall_info *upcall_info) + const struct dp_upcall_info *upcall_info) { - u32 group = packet_mc_group(dp, upcall_info->cmd); - struct sk_buff *nskb; - int err; + int dp_ifindex; + + dp_ifindex = get_dpifindex(dp); + if (!dp_ifindex) + return -ENODEV; do { struct ovs_header *upcall; struct sk_buff *user_skb; /* to be queued to userspace */ struct nlattr *nla; unsigned int len; - - nskb = skb->next; - skb->next = NULL; + int err; err = vlan_deaccel_tag(skb); if (unlikely(err)) - goto err_kfree_skbs; + return err; if (nla_attr_size(skb->len) > USHRT_MAX) - goto err_kfree_skbs; + return -EFBIG; len = sizeof(struct ovs_header); len += nla_total_size(skb->len); len += nla_total_size(FLOW_BUFSIZE); - if (upcall_info->userdata) + if (upcall_info->cmd == OVS_PACKET_CMD_ACTION) len += nla_total_size(8); - if (upcall_info->sample_pool) - len += nla_total_size(4); - if (upcall_info->actions_len) - len += nla_total_size(upcall_info->actions_len); user_skb = genlmsg_new(len, GFP_ATOMIC); - if (!user_skb) { - netlink_set_err(INIT_NET_GENL_SOCK, 0, group, -ENOBUFS); - goto err_kfree_skbs; - } + if (!user_skb) + return -ENOMEM; - upcall = genlmsg_put(user_skb, 0, 0, &dp_packet_genl_family, 0, upcall_info->cmd); - upcall->dp_ifindex = dp->dp_ifindex; + upcall = genlmsg_put(user_skb, 0, 0, &dp_packet_genl_family, + 0, upcall_info->cmd); + upcall->dp_ifindex = dp_ifindex; nla = nla_nest_start(user_skb, OVS_PACKET_ATTR_KEY); flow_to_nlattrs(upcall_info->key, user_skb); nla_nest_end(user_skb, nla); if (upcall_info->userdata) - nla_put_u64(user_skb, OVS_PACKET_ATTR_USERDATA, upcall_info->userdata); - if (upcall_info->sample_pool) - nla_put_u32(user_skb, OVS_PACKET_ATTR_SAMPLE_POOL, upcall_info->sample_pool); - if (upcall_info->actions_len) { - const struct nlattr *actions = upcall_info->actions; - u32 actions_len = upcall_info->actions_len; - - nla = nla_nest_start(user_skb, OVS_PACKET_ATTR_ACTIONS); - memcpy(__skb_put(user_skb, actions_len), actions, actions_len); - nla_nest_end(user_skb, nla); - } + nla_put_u64(user_skb, OVS_PACKET_ATTR_USERDATA, + nla_get_u64(upcall_info->userdata)); nla = __nla_reserve(user_skb, OVS_PACKET_ATTR_PACKET, skb->len); if (skb->ip_summed == CHECKSUM_PARTIAL) @@ -503,29 +463,20 @@ static int queue_userspace_packets(struct datapath *dp, struct sk_buff *skb, else skb_copy_bits(skb, 0, nla_data(nla), skb->len); - err = genlmsg_multicast(user_skb, 0, group, GFP_ATOMIC); + err = genlmsg_unicast(&init_net, user_skb, upcall_info->pid); if (err) - goto err_kfree_skbs; + return err; - consume_skb(skb); - skb = nskb; - } while (skb); - return 0; + } while ((skb = skb->next)); -err_kfree_skbs: - kfree_skb(skb); - while ((skb = nskb) != NULL) { - nskb = skb->next; - kfree_skb(skb); - } - return err; + return 0; } /* Called with genl_mutex. */ static int flush_flows(int dp_ifindex) { - struct tbl *old_table; - struct tbl *new_table; + struct flow_table *old_table; + struct flow_table *new_table; struct datapath *dp; dp = get_dp(dp_ifindex); @@ -533,28 +484,75 @@ static int flush_flows(int dp_ifindex) return -ENODEV; old_table = get_table_protected(dp); - new_table = tbl_create(TBL_MIN_BUCKETS); + new_table = flow_tbl_alloc(TBL_MIN_BUCKETS); if (!new_table) return -ENOMEM; rcu_assign_pointer(dp->table, new_table); - tbl_deferred_destroy(old_table, flow_free_tbl); + flow_tbl_deferred_destroy(old_table); + return 0; +} + +static int validate_actions(const struct nlattr *attr, int depth); + +static int validate_sample(const struct nlattr *attr, int depth) +{ + static const struct nla_policy sample_policy[OVS_SAMPLE_ATTR_MAX + 1] = + { + [OVS_SAMPLE_ATTR_PROBABILITY] = {.type = NLA_U32 }, + [OVS_SAMPLE_ATTR_ACTIONS] = {.type = NLA_UNSPEC }, + }; + struct nlattr *a[OVS_SAMPLE_ATTR_MAX + 1]; + int error; + + error = nla_parse_nested(a, OVS_SAMPLE_ATTR_MAX, attr, sample_policy); + if (error) + return error; + + if (!a[OVS_SAMPLE_ATTR_PROBABILITY]) + return -EINVAL; + if (!a[OVS_SAMPLE_ATTR_ACTIONS]) + return -EINVAL; + + return validate_actions(a[OVS_SAMPLE_ATTR_ACTIONS], (depth + 1)); +} + +static int validate_userspace(const struct nlattr *attr) +{ + static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = + { + [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 }, + [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_U64 }, + }; + struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1]; + int error; + + error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX, attr, userspace_policy); + if (error) + return error; + + if (!a[OVS_USERSPACE_ATTR_PID] || !nla_get_u32(a[OVS_USERSPACE_ATTR_PID])) + return -EINVAL; return 0; } -static int validate_actions(const struct nlattr *attr) +static int validate_actions(const struct nlattr *attr, int depth) { const struct nlattr *a; - int rem; + int rem, err; + + if (depth >= SAMPLE_ACTION_DEPTH) + return -EOVERFLOW; nla_for_each_nested(a, attr, rem) { + /* Expected argument lengths, (u32)-1 for variable length. */ static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = { [OVS_ACTION_ATTR_OUTPUT] = 4, - [OVS_ACTION_ATTR_USERSPACE] = 8, - [OVS_ACTION_ATTR_SET_DL_TCI] = 2, - [OVS_ACTION_ATTR_STRIP_VLAN] = 0, + [OVS_ACTION_ATTR_USERSPACE] = (u32)-1, + [OVS_ACTION_ATTR_PUSH_VLAN] = 2, + [OVS_ACTION_ATTR_POP_VLAN] = 0, [OVS_ACTION_ATTR_SET_DL_SRC] = ETH_ALEN, [OVS_ACTION_ATTR_SET_DL_DST] = ETH_ALEN, [OVS_ACTION_ATTR_SET_NW_SRC] = 4, @@ -565,18 +563,20 @@ static int validate_actions(const struct nlattr *attr) [OVS_ACTION_ATTR_SET_TUNNEL] = 8, [OVS_ACTION_ATTR_SET_PRIORITY] = 4, [OVS_ACTION_ATTR_POP_PRIORITY] = 0, + [OVS_ACTION_ATTR_SAMPLE] = (u32)-1 }; int type = nla_type(a); - if (type > OVS_ACTION_ATTR_MAX || nla_len(a) != action_lens[type]) + if (type > OVS_ACTION_ATTR_MAX || + (action_lens[type] != nla_len(a) && + action_lens[type] != (u32)-1)) return -EINVAL; switch (type) { case OVS_ACTION_ATTR_UNSPEC: return -EINVAL; - case OVS_ACTION_ATTR_USERSPACE: - case OVS_ACTION_ATTR_STRIP_VLAN: + case OVS_ACTION_ATTR_POP_VLAN: case OVS_ACTION_ATTR_SET_DL_SRC: case OVS_ACTION_ATTR_SET_DL_DST: case OVS_ACTION_ATTR_SET_NW_SRC: @@ -589,12 +589,18 @@ static int validate_actions(const struct nlattr *attr) /* No validation needed. */ break; + case OVS_ACTION_ATTR_USERSPACE: + err = validate_userspace(a); + if (err) + return err; + break; + case OVS_ACTION_ATTR_OUTPUT: if (nla_get_u32(a) >= DP_MAX_PORTS) return -EINVAL; break; - case OVS_ACTION_ATTR_SET_DL_TCI: + case OVS_ACTION_ATTR_PUSH_VLAN: if (nla_get_be16(a) & htons(VLAN_CFI_MASK)) return -EINVAL; break; @@ -604,6 +610,12 @@ static int validate_actions(const struct nlattr *attr) return -EINVAL; break; + case OVS_ACTION_ATTR_SAMPLE: + err = validate_sample(a, depth); + if (err) + return err; + break; + default: return -EOPNOTSUPP; } @@ -622,24 +634,6 @@ static void clear_stats(struct sw_flow *flow) flow->byte_count = 0; } -/* Called with genl_mutex. */ -static int expand_table(struct datapath *dp) -{ - struct tbl *old_table = get_table_protected(dp); - struct tbl *new_table; - - new_table = tbl_expand(old_table); - if (IS_ERR(new_table)) { - if (PTR_ERR(new_table) != -ENOSPC) - return PTR_ERR(new_table); - } else { - rcu_assign_pointer(dp->table, new_table); - tbl_deferred_destroy(old_table, NULL); - } - - return 0; -} - static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info) { struct ovs_header *ovs_header = info->userhdr; @@ -660,7 +654,7 @@ static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info) nla_len(a[OVS_PACKET_ATTR_PACKET]) < ETH_HLEN) goto err; - err = validate_actions(a[OVS_PACKET_ATTR_ACTIONS]); + err = validate_actions(a[OVS_PACKET_ATTR_ACTIONS], 0); if (err) goto err; @@ -693,7 +687,6 @@ static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info) err = flow_extract(packet, -1, &flow->key, &key_len, &is_frag); if (err) goto err_flow_put; - flow->tbl_node.hash = flow_hash(&flow->key, key_len); err = flow_metadata_from_nlattrs(&flow->key.eth.in_port, &flow->key.eth.tun_id, @@ -701,6 +694,8 @@ static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info) if (err) goto err_flow_put; + flow->hash = flow_hash(&flow->key, key_len); + acts = flow_actions_alloc(a[OVS_PACKET_ATTR_ACTIONS]); err = PTR_ERR(acts); if (IS_ERR(acts)) @@ -714,7 +709,14 @@ static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info) err = -ENODEV; if (!dp) goto err_unlock; + + if (flow->key.eth.in_port < DP_MAX_PORTS) + OVS_CB(packet)->vport = get_vport_protected(dp, + flow->key.eth.in_port); + + local_bh_disable(); err = execute_actions(dp, packet); + local_bh_enable(); rcu_read_unlock(); flow_put(flow); @@ -747,9 +749,9 @@ static struct genl_ops dp_packet_genl_ops[] = { static void get_dp_stats(struct datapath *dp, struct ovs_dp_stats *stats) { int i; - struct tbl *table = get_table_protected(dp); + struct flow_table *table = get_table_protected(dp); - stats->n_flows = tbl_count(table); + stats->n_flows = flow_tbl_count(table); stats->n_frags = stats->n_hit = stats->n_missed = stats->n_lost = 0; for_each_possible_cpu(i) { @@ -771,52 +773,6 @@ static void get_dp_stats(struct datapath *dp, struct ovs_dp_stats *stats) } } -/* MTU of the dp pseudo-device: ETH_DATA_LEN or the minimum of the ports. - * Called with RTNL lock. - */ -int dp_min_mtu(const struct datapath *dp) -{ - struct vport *p; - int mtu = 0; - - ASSERT_RTNL(); - - list_for_each_entry (p, &dp->port_list, node) { - int dev_mtu; - - /* Skip any internal ports, since that's what we're trying to - * set. */ - if (is_internal_vport(p)) - continue; - - dev_mtu = vport_get_mtu(p); - if (!dev_mtu) - continue; - if (!mtu || dev_mtu < mtu) - mtu = dev_mtu; - } - - return mtu ? mtu : ETH_DATA_LEN; -} - -/* Sets the MTU of all datapath devices to the minimum of the ports - * Called with RTNL lock. - */ -void set_internal_devs_mtu(const struct datapath *dp) -{ - struct vport *p; - int mtu; - - ASSERT_RTNL(); - - mtu = dp_min_mtu(dp); - - list_for_each_entry (p, &dp->port_list, node) { - if (is_internal_vport(p)) - vport_set_mtu(p, mtu); - } -} - static const struct nla_policy flow_policy[OVS_FLOW_ATTR_MAX + 1] = { [OVS_FLOW_ATTR_KEY] = { .type = NLA_NESTED }, [OVS_FLOW_ATTR_ACTIONS] = { .type = NLA_NESTED }, @@ -855,7 +811,7 @@ static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp, if (!ovs_header) return -EMSGSIZE; - ovs_header->dp_ifindex = dp->dp_ifindex; + ovs_header->dp_ifindex = get_dpifindex(dp); nla = nla_nest_start(skb, OVS_FLOW_ATTR_KEY); if (!nla) @@ -940,13 +896,11 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info) { struct nlattr **a = info->attrs; struct ovs_header *ovs_header = info->userhdr; - struct tbl_node *flow_node; struct sw_flow_key key; struct sw_flow *flow; struct sk_buff *reply; struct datapath *dp; - struct tbl *table; - u32 hash; + struct flow_table *table; int error; int key_len; @@ -960,7 +914,7 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info) /* Validate actions. */ if (a[OVS_FLOW_ATTR_ACTIONS]) { - error = validate_actions(a[OVS_FLOW_ATTR_ACTIONS]); + error = validate_actions(a[OVS_FLOW_ATTR_ACTIONS], 0); if (error) goto error; } else if (info->genlhdr->cmd == OVS_FLOW_CMD_NEW) { @@ -973,10 +927,9 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info) if (!dp) goto error; - hash = flow_hash(&key, key_len); table = get_table_protected(dp); - flow_node = tbl_lookup(table, &key, key_len, hash, flow_cmp); - if (!flow_node) { + flow = flow_tbl_lookup(table, &key, key_len); + if (!flow) { struct sw_flow_actions *acts; /* Bail out if we're not allowed to create a new flow. */ @@ -985,11 +938,15 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info) goto error; /* Expand table, if necessary, to make room. */ - if (tbl_count(table) >= tbl_n_buckets(table)) { - error = expand_table(dp); - if (error) - goto error; - table = get_table_protected(dp); + if (flow_tbl_need_to_expand(table)) { + struct flow_table *new_table; + + new_table = flow_tbl_expand(table); + if (!IS_ERR(new_table)) { + rcu_assign_pointer(dp->table, new_table); + flow_tbl_deferred_destroy(table); + table = get_table_protected(dp); + } } /* Allocate flow. */ @@ -1009,9 +966,8 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info) rcu_assign_pointer(flow->sf_acts, acts); /* Put flow in bucket. */ - error = tbl_insert(table, &flow->tbl_node, hash); - if (error) - goto error_free_flow; + flow->hash = flow_hash(&key, key_len); + flow_tbl_insert(table, flow); reply = ovs_flow_cmd_build_info(flow, dp, info->snd_pid, info->snd_seq, OVS_FLOW_CMD_NEW); @@ -1031,7 +987,6 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info) goto error; /* Update actions. */ - flow = flow_cast(flow_node); old_acts = rcu_dereference_protected(flow->sf_acts, lockdep_genl_is_held()); if (a[OVS_FLOW_ATTR_ACTIONS] && @@ -1079,11 +1034,10 @@ static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info) struct nlattr **a = info->attrs; struct ovs_header *ovs_header = info->userhdr; struct sw_flow_key key; - struct tbl_node *flow_node; struct sk_buff *reply; struct sw_flow *flow; struct datapath *dp; - struct tbl *table; + struct flow_table *table; int err; int key_len; @@ -1098,12 +1052,10 @@ static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info) return -ENODEV; table = get_table_protected(dp); - flow_node = tbl_lookup(table, &key, key_len, flow_hash(&key, key_len), - flow_cmp); - if (!flow_node) + flow = flow_tbl_lookup(table, &key, key_len); + if (!flow) return -ENOENT; - flow = flow_cast(flow_node); reply = ovs_flow_cmd_build_info(flow, dp, info->snd_pid, info->snd_seq, OVS_FLOW_CMD_NEW); if (IS_ERR(reply)) return PTR_ERR(reply); @@ -1116,11 +1068,10 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info) struct nlattr **a = info->attrs; struct ovs_header *ovs_header = info->userhdr; struct sw_flow_key key; - struct tbl_node *flow_node; struct sk_buff *reply; struct sw_flow *flow; struct datapath *dp; - struct tbl *table; + struct flow_table *table; int err; int key_len; @@ -1135,21 +1086,15 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info) return -ENODEV; table = get_table_protected(dp); - flow_node = tbl_lookup(table, &key, key_len, flow_hash(&key, key_len), - flow_cmp); - if (!flow_node) + flow = flow_tbl_lookup(table, &key, key_len); + if (!flow) return -ENOENT; - flow = flow_cast(flow_node); reply = ovs_flow_cmd_alloc_info(flow); if (!reply) return -ENOMEM; - err = tbl_remove(table, flow_node); - if (err) { - kfree_skb(reply); - return err; - } + flow_tbl_remove(table, flow); err = ovs_flow_cmd_fill_info(flow, dp, reply, info->snd_pid, info->snd_seq, 0, OVS_FLOW_CMD_DEL); @@ -1172,17 +1117,15 @@ static int ovs_flow_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb) return -ENODEV; for (;;) { - struct tbl_node *flow_node; struct sw_flow *flow; u32 bucket, obj; bucket = cb->args[0]; obj = cb->args[1]; - flow_node = tbl_next(get_table_protected(dp), &bucket, &obj); - if (!flow_node) + flow = flow_tbl_next(get_table_protected(dp), &bucket, &obj); + if (!flow) break; - flow = flow_cast(flow_node); if (ovs_flow_cmd_fill_info(flow, dp, skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq, NLM_F_MULTI, OVS_FLOW_CMD_NEW) < 0) @@ -1222,8 +1165,8 @@ static const struct nla_policy datapath_policy[OVS_DP_ATTR_MAX + 1] = { #ifdef HAVE_NLA_NUL_STRING [OVS_DP_ATTR_NAME] = { .type = NLA_NUL_STRING, .len = IFNAMSIZ - 1 }, #endif + [OVS_DP_ATTR_UPCALL_PID] = { .type = NLA_U32 }, [OVS_DP_ATTR_IPV4_FRAGS] = { .type = NLA_U32 }, - [OVS_DP_ATTR_SAMPLING] = { .type = NLA_U32 }, }; static struct genl_family dp_datapath_genl_family = { @@ -1250,7 +1193,7 @@ static int ovs_dp_cmd_fill_info(struct datapath *dp, struct sk_buff *skb, if (!ovs_header) goto error; - ovs_header->dp_ifindex = dp->dp_ifindex; + ovs_header->dp_ifindex = get_dpifindex(dp); rcu_read_lock(); err = nla_put_string(skb, OVS_DP_ATTR_NAME, dp_name(dp)); @@ -1266,17 +1209,6 @@ static int ovs_dp_cmd_fill_info(struct datapath *dp, struct sk_buff *skb, NLA_PUT_U32(skb, OVS_DP_ATTR_IPV4_FRAGS, dp->drop_frags ? OVS_DP_FRAG_DROP : OVS_DP_FRAG_ZERO); - if (dp->sflow_probability) - NLA_PUT_U32(skb, OVS_DP_ATTR_SAMPLING, dp->sflow_probability); - - nla = nla_nest_start(skb, OVS_DP_ATTR_MCGROUPS); - if (!nla) - goto nla_put_failure; - NLA_PUT_U32(skb, OVS_PACKET_CMD_MISS, packet_mc_group(dp, OVS_PACKET_CMD_MISS)); - NLA_PUT_U32(skb, OVS_PACKET_CMD_ACTION, packet_mc_group(dp, OVS_PACKET_CMD_ACTION)); - NLA_PUT_U32(skb, OVS_PACKET_CMD_SAMPLE, packet_mc_group(dp, OVS_PACKET_CMD_SAMPLE)); - nla_nest_end(skb, nla); - return genlmsg_end(skb, ovs_header); nla_put_failure: @@ -1338,8 +1270,6 @@ static void change_datapath(struct datapath *dp, struct nlattr *a[OVS_DP_ATTR_MA { if (a[OVS_DP_ATTR_IPV4_FRAGS]) dp->drop_frags = nla_get_u32(a[OVS_DP_ATTR_IPV4_FRAGS]) == OVS_DP_FRAG_DROP; - if (a[OVS_DP_ATTR_SAMPLING]) - dp->sflow_probability = nla_get_u32(a[OVS_DP_ATTR_SAMPLING]); } static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info) @@ -1352,7 +1282,7 @@ static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info) int err; err = -EINVAL; - if (!a[OVS_DP_ATTR_NAME]) + if (!a[OVS_DP_ATTR_NAME] || !a[OVS_DP_ATTR_UPCALL_PID]) goto err; err = ovs_dp_cmd_validate(a); @@ -1377,34 +1307,35 @@ static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info) /* Allocate table. */ err = -ENOMEM; - rcu_assign_pointer(dp->table, tbl_create(TBL_MIN_BUCKETS)); + rcu_assign_pointer(dp->table, flow_tbl_alloc(TBL_MIN_BUCKETS)); if (!dp->table) goto err_free_dp; + dp->drop_frags = 0; + dp->stats_percpu = alloc_percpu(struct dp_stats_percpu); + if (!dp->stats_percpu) { + err = -ENOMEM; + goto err_destroy_table; + } + + change_datapath(dp, a); + /* Set up our datapath device. */ parms.name = nla_data(a[OVS_DP_ATTR_NAME]); parms.type = OVS_VPORT_TYPE_INTERNAL; parms.options = NULL; parms.dp = dp; parms.port_no = OVSP_LOCAL; + parms.upcall_pid = nla_get_u32(a[OVS_DP_ATTR_UPCALL_PID]); + vport = new_vport(&parms); if (IS_ERR(vport)) { err = PTR_ERR(vport); if (err == -EBUSY) err = -EEXIST; - goto err_destroy_table; + goto err_destroy_percpu; } - dp->dp_ifindex = vport_get_ifindex(vport); - - dp->drop_frags = 0; - dp->stats_percpu = alloc_percpu(struct dp_stats_percpu); - if (!dp->stats_percpu) { - err = -ENOMEM; - goto err_destroy_local_port; - } - - change_datapath(dp, a); reply = ovs_dp_cmd_build_info(dp, info->snd_pid, info->snd_seq, OVS_DP_CMD_NEW); err = PTR_ERR(reply); @@ -1422,8 +1353,10 @@ static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info) err_destroy_local_port: dp_detach_port(get_vport_protected(dp, OVSP_LOCAL)); +err_destroy_percpu: + free_percpu(dp->stats_percpu); err_destroy_table: - tbl_destroy(get_table_protected(dp), NULL); + flow_tbl_destroy(get_table_protected(dp)); err_free_dp: kfree(dp); err_put_module: @@ -1583,15 +1516,15 @@ static struct genl_ops dp_datapath_genl_ops[] = { static const struct nla_policy vport_policy[OVS_VPORT_ATTR_MAX + 1] = { #ifdef HAVE_NLA_NUL_STRING [OVS_VPORT_ATTR_NAME] = { .type = NLA_NUL_STRING, .len = IFNAMSIZ - 1 }, - [OVS_VPORT_ATTR_PORT_NO] = { .type = NLA_U32 }, - [OVS_VPORT_ATTR_TYPE] = { .type = NLA_U32 }, - [OVS_VPORT_ATTR_STATS] = { .len = sizeof(struct rtnl_link_stats64) }, + [OVS_VPORT_ATTR_STATS] = { .len = sizeof(struct ovs_vport_stats) }, [OVS_VPORT_ATTR_ADDRESS] = { .len = ETH_ALEN }, #else - [OVS_VPORT_ATTR_STATS] = { .minlen = sizeof(struct rtnl_link_stats64) }, + [OVS_VPORT_ATTR_STATS] = { .minlen = sizeof(struct ovs_vport_stats) }, [OVS_VPORT_ATTR_ADDRESS] = { .minlen = ETH_ALEN }, #endif - [OVS_VPORT_ATTR_MTU] = { .type = NLA_U32 }, + [OVS_VPORT_ATTR_PORT_NO] = { .type = NLA_U32 }, + [OVS_VPORT_ATTR_TYPE] = { .type = NLA_U32 }, + [OVS_VPORT_ATTR_UPCALL_PID] = { .type = NLA_U32 }, [OVS_VPORT_ATTR_OPTIONS] = { .type = NLA_NESTED }, }; @@ -1603,7 +1536,7 @@ static struct genl_family dp_vport_genl_family = { .maxattr = OVS_VPORT_ATTR_MAX }; -static struct genl_multicast_group dp_vport_multicast_group = { +struct genl_multicast_group dp_vport_multicast_group = { .name = OVS_VPORT_MCGROUP }; @@ -1613,8 +1546,6 @@ static int ovs_vport_cmd_fill_info(struct vport *vport, struct sk_buff *skb, { struct ovs_header *ovs_header; struct nlattr *nla; - int ifindex, iflink; - int mtu; int err; ovs_header = genlmsg_put(skb, pid, seq, &dp_vport_genl_family, @@ -1622,36 +1553,25 @@ static int ovs_vport_cmd_fill_info(struct vport *vport, struct sk_buff *skb, if (!ovs_header) return -EMSGSIZE; - ovs_header->dp_ifindex = vport->dp->dp_ifindex; + ovs_header->dp_ifindex = get_dpifindex(vport->dp); NLA_PUT_U32(skb, OVS_VPORT_ATTR_PORT_NO, vport->port_no); NLA_PUT_U32(skb, OVS_VPORT_ATTR_TYPE, vport_get_type(vport)); NLA_PUT_STRING(skb, OVS_VPORT_ATTR_NAME, vport_get_name(vport)); + NLA_PUT_U32(skb, OVS_VPORT_ATTR_UPCALL_PID, vport->upcall_pid); - nla = nla_reserve(skb, OVS_VPORT_ATTR_STATS, sizeof(struct rtnl_link_stats64)); + nla = nla_reserve(skb, OVS_VPORT_ATTR_STATS, sizeof(struct ovs_vport_stats)); if (!nla) goto nla_put_failure; - if (vport_get_stats(vport, nla_data(nla))) - __skb_trim(skb, skb->len - nla->nla_len); - NLA_PUT(skb, OVS_VPORT_ATTR_ADDRESS, ETH_ALEN, vport_get_addr(vport)); + vport_get_stats(vport, nla_data(nla)); - mtu = vport_get_mtu(vport); - if (mtu) - NLA_PUT_U32(skb, OVS_VPORT_ATTR_MTU, mtu); + NLA_PUT(skb, OVS_VPORT_ATTR_ADDRESS, ETH_ALEN, vport_get_addr(vport)); err = vport_get_options(vport, skb); if (err == -EMSGSIZE) goto error; - ifindex = vport_get_ifindex(vport); - if (ifindex > 0) - NLA_PUT_U32(skb, OVS_VPORT_ATTR_IFINDEX, ifindex); - - iflink = vport_get_iflink(vport); - if (iflink > 0) - NLA_PUT_U32(skb, OVS_VPORT_ATTR_IFLINK, iflink); - return genlmsg_end(skb, ovs_header); nla_put_failure: @@ -1662,8 +1582,8 @@ error: } /* Called with RTNL lock or RCU read lock. */ -static struct sk_buff *ovs_vport_cmd_build_info(struct vport *vport, u32 pid, - u32 seq, u8 cmd) +struct sk_buff *ovs_vport_cmd_build_info(struct vport *vport, u32 pid, + u32 seq, u8 cmd) { struct sk_buff *skb; int retval; @@ -1719,12 +1639,13 @@ static struct vport *lookup_vport(struct ovs_header *ovs_header, static int change_vport(struct vport *vport, struct nlattr *a[OVS_VPORT_ATTR_MAX + 1]) { int err = 0; + if (a[OVS_VPORT_ATTR_STATS]) - err = vport_set_stats(vport, nla_data(a[OVS_VPORT_ATTR_STATS])); - if (!err && a[OVS_VPORT_ATTR_ADDRESS]) + vport_set_stats(vport, nla_data(a[OVS_VPORT_ATTR_STATS])); + + if (a[OVS_VPORT_ATTR_ADDRESS]) err = vport_set_addr(vport, nla_data(a[OVS_VPORT_ATTR_ADDRESS])); - if (!err && a[OVS_VPORT_ATTR_MTU]) - err = vport_set_mtu(vport, nla_get_u32(a[OVS_VPORT_ATTR_MTU])); + return err; } @@ -1740,7 +1661,8 @@ static int ovs_vport_cmd_new(struct sk_buff *skb, struct genl_info *info) int err; err = -EINVAL; - if (!a[OVS_VPORT_ATTR_NAME] || !a[OVS_VPORT_ATTR_TYPE]) + if (!a[OVS_VPORT_ATTR_NAME] || !a[OVS_VPORT_ATTR_TYPE] || + !a[OVS_VPORT_ATTR_UPCALL_PID]) goto exit; err = ovs_vport_cmd_validate(a); @@ -1781,13 +1703,13 @@ static int ovs_vport_cmd_new(struct sk_buff *skb, struct genl_info *info) parms.options = a[OVS_VPORT_ATTR_OPTIONS]; parms.dp = dp; parms.port_no = port_no; + parms.upcall_pid = nla_get_u32(a[OVS_VPORT_ATTR_UPCALL_PID]); vport = new_vport(&parms); err = PTR_ERR(vport); if (IS_ERR(vport)) goto exit_unlock; - set_internal_devs_mtu(dp); dp_sysfs_add_if(vport); err = change_vport(vport, a); @@ -1829,10 +1751,14 @@ static int ovs_vport_cmd_set(struct sk_buff *skb, struct genl_info *info) goto exit_unlock; err = 0; - if (a[OVS_VPORT_ATTR_OPTIONS]) + if (a[OVS_VPORT_ATTR_TYPE] && nla_get_u32(a[OVS_VPORT_ATTR_TYPE]) != vport_get_type(vport)) + err = -EINVAL; + if (!err && a[OVS_VPORT_ATTR_OPTIONS]) err = vport_set_options(vport, a[OVS_VPORT_ATTR_OPTIONS]); if (!err) err = change_vport(vport, a); + if (!err && a[OVS_VPORT_ATTR_UPCALL_PID]) + vport->upcall_pid = nla_get_u32(a[OVS_VPORT_ATTR_UPCALL_PID]); reply = ovs_vport_cmd_build_info(vport, info->snd_pid, info->snd_seq, OVS_VPORT_CMD_NEW); @@ -1880,7 +1806,7 @@ static int ovs_vport_cmd_del(struct sk_buff *skb, struct genl_info *info) if (IS_ERR(reply)) goto exit_unlock; - err = dp_detach_port(vport); + dp_detach_port(vport); genl_notify(reply, genl_info_net(info), info->snd_pid, dp_vport_multicast_group.id, info->nlhdr, GFP_KERNEL); @@ -2007,9 +1933,8 @@ static void dp_unregister_genl(int n_families) { int i; - for (i = 0; i < n_families; i++) { + for (i = 0; i < n_families; i++) genl_unregister_family(dp_genl_families[i].family); - } } static int dp_register_genl(void) @@ -2035,9 +1960,6 @@ static int dp_register_genl(void) } } - err = packet_register_mc_groups(); - if (err) - goto error; return 0; error: @@ -2054,10 +1976,14 @@ static int __init dp_init(void) printk("Open vSwitch %s, built "__DATE__" "__TIME__"\n", VERSION BUILDNR); - err = flow_init(); + err = tnl_init(); if (err) goto error; + err = flow_init(); + if (err) + goto error_tnl_exit; + err = vport_init(); if (err) goto error_flow_exit; @@ -2078,6 +2004,8 @@ error_vport_exit: vport_exit(); error_flow_exit: flow_exit(); +error_tnl_exit: + tnl_exit(); error: return err; } @@ -2089,6 +2017,7 @@ static void dp_cleanup(void) unregister_netdevice_notifier(&dp_device_notifier); vport_exit(); flow_exit(); + tnl_exit(); } module_init(dp_init);