X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=datapath%2Fdatapath.c;h=49d93aac5c2ff15e6e5c57b36a0e840f1c3f4f00;hb=eba9291a25a9d38f3382510de52e619a65b2f711;hp=7c9ec3b28e101225630eb0206b567559171add2c;hpb=f14d80834e73e5e91892e4763e01ccfd3a2d3be2;p=sliver-openvswitch.git diff --git a/datapath/datapath.c b/datapath/datapath.c index 7c9ec3b28..49d93aac5 100644 --- a/datapath/datapath.c +++ b/datapath/datapath.c @@ -33,29 +33,26 @@ #include #include #include -#include #include #include #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) || \ - LINUX_VERSION_CODE >= KERNEL_VERSION(3,1,0) -#error Kernels before 2.6.18 or after 3.0 are not supported by this version of Open vSwitch. + LINUX_VERSION_CODE > KERNEL_VERSION(3,2,0) +#error Kernels before 2.6.18 or after 3.2 are not supported by this version of Open vSwitch. #endif int (*dp_ioctl_hook)(struct net_device *dev, struct ifreq *rq, int cmd); @@ -68,8 +65,8 @@ EXPORT_SYMBOL(dp_ioctl_hook); * etc.) are protected by RTNL. * * Writes to other state (flow table modifications, set miscellaneous datapath - * parameters such as drop frags, etc.) are protected by genl_mutex. The RTNL - * lock nests inside genl_mutex. + * parameters, etc.) are protected by genl_mutex. The RTNL lock nests inside + * genl_mutex. * * Reads are protected by RCU. * @@ -84,8 +81,10 @@ EXPORT_SYMBOL(dp_ioctl_hook); static LIST_HEAD(dps); static struct vport *new_vport(const struct vport_parms *); -static int queue_userspace_packets(struct datapath *, struct sk_buff *, - const struct dp_upcall_info *); +static int queue_gso_packets(int dp_ifindex, struct sk_buff *, + const struct dp_upcall_info *); +static int queue_userspace_packet(int dp_ifindex, struct sk_buff *, + const struct dp_upcall_info *); /* Must be called with rcu_read_lock, genl_mutex, or RTNL lock. */ struct datapath *get_dp(int dp_ifindex) @@ -107,7 +106,7 @@ 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()); } @@ -121,17 +120,35 @@ static struct vport *get_vport_protected(struct datapath *dp, u16 port_no) /* Must be called with rcu_read_lock or RTNL lock. */ const char *dp_name(const struct datapath *dp) { - return vport_get_name(rcu_dereference_rtnl(dp->ports[OVSP_LOCAL])); + struct vport *vport = rcu_dereference_rtnl(dp->ports[OVSP_LOCAL]); + return vport->ops->get_name(vport); } -static inline size_t br_nlmsg_size(void) +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 = local->ops->get_ifindex(local); + else + ifindex = 0; + + rcu_read_unlock(); + + return ifindex; +} + +static size_t br_nlmsg_size(void) { return NLMSG_ALIGN(sizeof(struct ifinfomsg)) + nla_total_size(IFNAMSIZ) /* IFLA_IFNAME */ + 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 */ } @@ -141,16 +158,11 @@ static int dp_fill_ifinfo(struct sk_buff *skb, int event, unsigned int flags) { 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; + if (!port->ops->get_ifindex) + return -ENODEV; nlh = nlmsg_put(skb, 0, 0, event, sizeof(*hdr), flags); if (nlh == NULL) @@ -160,25 +172,21 @@ static int dp_fill_ifinfo(struct sk_buff *skb, hdr->ifi_family = AF_BRIDGE; hdr->__ifi_pad = 0; hdr->ifi_type = ARPHRD_ETHER; - hdr->ifi_index = ifindex; - hdr->ifi_flags = vport_get_flags(port); + hdr->ifi_index = port->ops->get_ifindex(port); + hdr->ifi_flags = port->ops->get_dev_flags(port); 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_MTU, vport_get_mtu(port)); + NLA_PUT_STRING(skb, IFLA_IFNAME, port->ops->get_name(port)); + NLA_PUT_U32(skb, IFLA_MASTER, get_dpifindex(dp)); + NLA_PUT_U32(skb, IFLA_MTU, port->ops->get_mtu(port)); #ifdef IFLA_OPERSTATE NLA_PUT_U8(skb, IFLA_OPERSTATE, - vport_is_running(port) - ? vport_get_operstate(port) + port->ops->is_running(port) + ? port->ops->get_operstate(port) : IF_OPER_DOWN); #endif - NLA_PUT(skb, IFLA_ADDRESS, ETH_ALEN, vport_get_addr(port)); - - if (ifindex != iflink) - NLA_PUT_U32(skb, IFLA_LINK,iflink); + NLA_PUT(skb, IFLA_ADDRESS, ETH_ALEN, port->ops->get_addr(port)); return nlmsg_end(skb, nlh); @@ -191,24 +199,32 @@ nla_put_failure: static void dp_ifinfo_notify(int event, struct vport *port) { struct sk_buff *skb; - int err = -ENOBUFS; + int err; skb = nlmsg_new(br_nlmsg_size(), GFP_KERNEL); - if (skb == NULL) - goto errout; + if (!skb) { + err = -ENOBUFS; + goto err; + } err = dp_fill_ifinfo(skb, port, event, 0); if (err < 0) { - /* -EMSGSIZE implies BUG in br_nlmsg_size() */ - WARN_ON(err == -EMSGSIZE); - kfree_skb(skb); - goto errout; + if (err == -ENODEV) { + goto out; + } else { + /* -EMSGSIZE implies BUG in br_nlmsg_size() */ + WARN_ON(err == -EMSGSIZE); + goto err; + } } + rtnl_notify(skb, &init_net, 0, RTNLGRP_LINK, NULL, GFP_KERNEL); + return; -errout: - if (err < 0) - rtnl_set_sk_err(&init_net, RTNLGRP_LINK, err); +err: + rtnl_set_sk_err(&init_net, RTNLGRP_LINK, err); +out: + kfree_skb(skb); } static void release_dp(struct kobject *kobj) @@ -225,7 +241,7 @@ 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 +265,7 @@ 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(); @@ -262,270 +278,204 @@ 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; - OVS_CB(skb)->vport = p; + stats = per_cpu_ptr(dp->stats_percpu, smp_processor_id()); if (!OVS_CB(skb)->flow) { struct sw_flow_key key; - struct tbl_node *flow_node; int key_len; - bool is_frag; /* Extract flow from 'skb' into 'key'. */ - error = flow_extract(skb, p->port_no, &key, &key_len, &is_frag); + error = flow_extract(skb, p->port_no, &key, &key_len); if (unlikely(error)) { kfree_skb(skb); return; } - if (is_frag && dp->drop_frags) { - consume_skb(skb); - stats_counter_off = offsetof(struct dp_stats_percpu, 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); + consume_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) -{ - u16 csum_start, csum_offset; - __wsum csum; - - get_skb_csum_pointers(skb, &csum_start, &csum_offset); - csum_start -= skb_headroom(skb); - - skb_copy_bits(skb, 0, to, csum_start); - - csum = skb_copy_and_csum_bits(skb, csum_start, to + csum_start, - skb->len - csum_start, 0); - *(__sum16 *)(to + csum_start + csum_offset) = csum_fold(csum); } static struct genl_family dp_packet_genl_family = { .id = GENL_ID_GENERATE, .hdrsize = sizeof(struct ovs_header), .name = OVS_PACKET_FAMILY, - .version = 1, + .version = OVS_PACKET_VERSION, .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 dp_stats_percpu *stats; + int dp_ifindex; int err; - WARN_ON_ONCE(skb_shared(skb)); - - forward_ip_summed(skb, true); + if (upcall_info->pid == 0) { + err = -ENOTCONN; + goto err; + } - /* 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); - - if (IS_ERR(nskb)) { - kfree_skb(skb); - err = PTR_ERR(nskb); - goto err; - } - consume_skb(skb); - skb = nskb; + dp_ifindex = get_dpifindex(dp); + if (!dp_ifindex) { + err = -ENODEV; + goto err; } - err = queue_userspace_packets(dp, skb, upcall_info); + forward_ip_summed(skb, true); + + if (!skb_is_gso(skb)) + err = queue_userspace_packet(dp_ifindex, skb, upcall_info); + else + err = queue_gso_packets(dp_ifindex, skb, upcall_info); 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; } -/* Send each packet in the 'skb' list to userspace for 'dp' as directed by - * 'upcall_info'. There will be only one packet unless we broke up a GSO - * packet. - */ -static int queue_userspace_packets(struct datapath *dp, struct sk_buff *skb, - const struct dp_upcall_info *upcall_info) +static int queue_gso_packets(int dp_ifindex, struct sk_buff *skb, + const struct dp_upcall_info *upcall_info) { - u32 group = packet_mc_group(dp, upcall_info->cmd); - struct sk_buff *nskb; + struct dp_upcall_info later_info; + struct sw_flow_key later_key; + struct sk_buff *segs, *nskb; int err; + segs = skb_gso_segment(skb, NETIF_F_SG | NETIF_F_HW_CSUM); + if (IS_ERR(skb)) + return PTR_ERR(skb); + + /* Queue all of the segments. */ + skb = segs; do { - struct ovs_header *upcall; - struct sk_buff *user_skb; /* to be queued to userspace */ - struct nlattr *nla; - unsigned int len; + err = queue_userspace_packet(dp_ifindex, skb, upcall_info); + if (err) + break; - nskb = skb->next; - skb->next = NULL; - - err = vlan_deaccel_tag(skb); - if (unlikely(err)) - goto err_kfree_skbs; - - if (nla_attr_size(skb->len) > USHRT_MAX) - goto err_kfree_skbs; - - len = sizeof(struct ovs_header); - len += nla_total_size(skb->len); - len += nla_total_size(FLOW_BUFSIZE); - if (upcall_info->userdata) - 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 (skb == segs && skb_shinfo(skb)->gso_type & SKB_GSO_UDP) { + /* The initial flow key extracted by flow_extract() in + * this case is for a first fragment, so we need to + * properly mark later fragments. + */ + later_key = *upcall_info->key; + later_key.ip.frag = OVS_FRAG_TYPE_LATER; + + later_info = *upcall_info; + later_info.key = &later_key; + upcall_info = &later_info; } + } while ((skb = skb->next)); - upcall = genlmsg_put(user_skb, 0, 0, &dp_packet_genl_family, 0, upcall_info->cmd); - upcall->dp_ifindex = dp->dp_ifindex; + /* Free all of the segments. */ + skb = segs; + do { + nskb = skb->next; + if (err) + kfree_skb(skb); + else + consume_skb(skb); + } while ((skb = nskb)); + return err; +} - nla = nla_nest_start(user_skb, OVS_PACKET_ATTR_KEY); - flow_to_nlattrs(upcall_info->key, user_skb); - nla_nest_end(user_skb, nla); +static int queue_userspace_packet(int dp_ifindex, struct sk_buff *skb, + const struct dp_upcall_info *upcall_info) +{ + struct ovs_header *upcall; + struct sk_buff *user_skb; /* to be queued to userspace */ + struct nlattr *nla; + unsigned int len; + int err; - 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; + err = vlan_deaccel_tag(skb); + if (unlikely(err)) + return err; - 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); - } + if (nla_attr_size(skb->len) > USHRT_MAX) + return -EFBIG; - nla = __nla_reserve(user_skb, OVS_PACKET_ATTR_PACKET, skb->len); - if (skb->ip_summed == CHECKSUM_PARTIAL) - copy_and_csum_skb(skb, nla_data(nla)); - else - skb_copy_bits(skb, 0, nla_data(nla), skb->len); + len = sizeof(struct ovs_header); + len += nla_total_size(skb->len); + len += nla_total_size(FLOW_BUFSIZE); + if (upcall_info->cmd == OVS_PACKET_CMD_ACTION) + len += nla_total_size(8); - err = genlmsg_multicast(user_skb, 0, group, GFP_ATOMIC); - if (err) - goto err_kfree_skbs; + user_skb = genlmsg_new(len, GFP_ATOMIC); + if (!user_skb) + return -ENOMEM; - consume_skb(skb); - skb = nskb; - } while (skb); - return 0; + upcall = genlmsg_put(user_skb, 0, 0, &dp_packet_genl_family, + 0, upcall_info->cmd); + upcall->dp_ifindex = dp_ifindex; -err_kfree_skbs: - kfree_skb(skb); - while ((skb = nskb) != NULL) { - nskb = skb->next; - kfree_skb(skb); - } - return err; + 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, + nla_get_u64(upcall_info->userdata)); + + nla = __nla_reserve(user_skb, OVS_PACKET_ATTR_PACKET, skb->len); + + skb_copy_and_csum_dev(skb, nla_data(nla)); + + return genlmsg_unicast(&init_net, user_skb, upcall_info->pid); } /* 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); @@ -533,42 +483,167 @@ 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) +static int validate_actions(const struct nlattr *attr, + const struct sw_flow_key *key, int depth); + +static int validate_sample(const struct nlattr *attr, + const struct sw_flow_key *key, int depth) { + const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1]; + const struct nlattr *probability, *actions; const struct nlattr *a; int rem; + memset(attrs, 0, sizeof(attrs)); nla_for_each_nested(a, attr, rem) { + int type = nla_type(a); + if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type]) + return -EINVAL; + attrs[type] = a; + } + if (rem) + return -EINVAL; + + probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY]; + if (!probability || nla_len(probability) != sizeof(u32)) + return -EINVAL; + + actions = attrs[OVS_SAMPLE_ATTR_ACTIONS]; + if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN)) + return -EINVAL; + return validate_actions(actions, key, depth + 1); +} + +static int validate_action_key(const struct nlattr *a, + const struct sw_flow_key *flow_key) +{ + int act_type = nla_type(a); + const struct nlattr *ovs_key = nla_data(a); + int key_type = nla_type(ovs_key); + + /* There can be only one key in a action */ + if (nla_total_size(nla_len(ovs_key)) != nla_len(a)) + return -EINVAL; + + if (key_type > OVS_KEY_ATTR_MAX || + nla_len(ovs_key) != ovs_key_lens[key_type]) + return -EINVAL; + +#define ACTION(act, key) (((act) << 8) | (key)) + + switch (ACTION(act_type, key_type)) { + const struct ovs_key_ipv4 *ipv4_key; + const struct ovs_key_8021q *q_key; + + case ACTION(OVS_ACTION_ATTR_SET, OVS_KEY_ATTR_PRIORITY): + case ACTION(OVS_ACTION_ATTR_SET, OVS_KEY_ATTR_TUN_ID): + case ACTION(OVS_ACTION_ATTR_SET, OVS_KEY_ATTR_ETHERNET): + break; + + case ACTION(OVS_ACTION_ATTR_PUSH, OVS_KEY_ATTR_8021Q): + q_key = nla_data(ovs_key); + if (q_key->q_tpid != htons(ETH_P_8021Q)) + return -EINVAL; + + if (q_key->q_tci & htons(VLAN_TAG_PRESENT)) + return -EINVAL; + break; + + case ACTION(OVS_ACTION_ATTR_SET, OVS_KEY_ATTR_IPV4): + if (flow_key->eth.type != htons(ETH_P_IP)) + return -EINVAL; + + if (!flow_key->ipv4.addr.src || !flow_key->ipv4.addr.dst) + return -EINVAL; + + ipv4_key = nla_data(ovs_key); + if (ipv4_key->ipv4_proto != flow_key->ip.proto) + return -EINVAL; + + if (ipv4_key->ipv4_frag != flow_key->ip.frag) + return -EINVAL; + + break; + + case ACTION(OVS_ACTION_ATTR_SET, OVS_KEY_ATTR_TCP): + if (flow_key->ip.proto != IPPROTO_TCP) + return -EINVAL; + + if (!flow_key->ipv4.tp.src || !flow_key->ipv4.tp.dst) + return -EINVAL; + + break; + + case ACTION(OVS_ACTION_ATTR_SET, OVS_KEY_ATTR_UDP): + if (flow_key->ip.proto != IPPROTO_UDP) + return -EINVAL; + + if (!flow_key->ipv4.tp.src || !flow_key->ipv4.tp.dst) + return -EINVAL; + break; + + default: + return -EINVAL; + } +#undef ACTION + return 0; +} + +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, + const struct sw_flow_key *key, int depth) +{ + const struct nlattr *a; + 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_SET_DL_SRC] = ETH_ALEN, - [OVS_ACTION_ATTR_SET_DL_DST] = ETH_ALEN, - [OVS_ACTION_ATTR_SET_NW_SRC] = 4, - [OVS_ACTION_ATTR_SET_NW_DST] = 4, - [OVS_ACTION_ATTR_SET_NW_TOS] = 1, - [OVS_ACTION_ATTR_SET_TP_SRC] = 2, - [OVS_ACTION_ATTR_SET_TP_DST] = 2, - [OVS_ACTION_ATTR_SET_TUNNEL] = 8, - [OVS_ACTION_ATTR_SET_PRIORITY] = 4, - [OVS_ACTION_ATTR_POP_PRIORITY] = 0, + [OVS_ACTION_ATTR_USERSPACE] = (u32)-1, + [OVS_ACTION_ATTR_PUSH] = (u32)-1, + [OVS_ACTION_ATTR_POP] = 2, + [OVS_ACTION_ATTR_SET] = (u32)-1, + [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) { @@ -576,17 +651,9 @@ static int validate_actions(const struct nlattr *attr) return -EINVAL; case OVS_ACTION_ATTR_USERSPACE: - case OVS_ACTION_ATTR_STRIP_VLAN: - case OVS_ACTION_ATTR_SET_DL_SRC: - case OVS_ACTION_ATTR_SET_DL_DST: - case OVS_ACTION_ATTR_SET_NW_SRC: - case OVS_ACTION_ATTR_SET_NW_DST: - case OVS_ACTION_ATTR_SET_TP_SRC: - case OVS_ACTION_ATTR_SET_TP_DST: - case OVS_ACTION_ATTR_SET_TUNNEL: - case OVS_ACTION_ATTR_SET_PRIORITY: - case OVS_ACTION_ATTR_POP_PRIORITY: - /* No validation needed. */ + err = validate_userspace(a); + if (err) + return err; break; case OVS_ACTION_ATTR_OUTPUT: @@ -594,18 +661,27 @@ static int validate_actions(const struct nlattr *attr) return -EINVAL; break; - case OVS_ACTION_ATTR_SET_DL_TCI: - if (nla_get_be16(a) & htons(VLAN_CFI_MASK)) + + case OVS_ACTION_ATTR_POP: + if (nla_get_u16(a) != OVS_KEY_ATTR_8021Q) return -EINVAL; break; - case OVS_ACTION_ATTR_SET_NW_TOS: - if (nla_get_u8(a) & INET_ECN_MASK) - return -EINVAL; + case OVS_ACTION_ATTR_SET: + case OVS_ACTION_ATTR_PUSH: + err = validate_action_key(a, key); + if (err) + return err; + break; + + case OVS_ACTION_ATTR_SAMPLE: + err = validate_sample(a, key, depth); + if (err) + return err; break; default: - return -EOPNOTSUPP; + return -EINVAL; } } @@ -614,6 +690,7 @@ static int validate_actions(const struct nlattr *attr) return 0; } + static void clear_stats(struct sw_flow *flow) { flow->used = 0; @@ -622,24 +699,6 @@ 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; @@ -649,7 +708,6 @@ static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info) struct sw_flow *flow; struct datapath *dp; struct ethhdr *eth; - bool is_frag; int len; int err; int key_len; @@ -660,10 +718,6 @@ 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]); - if (err) - goto err; - len = nla_len(a[OVS_PACKET_ATTR_PACKET]); packet = __dev_alloc_skb(NET_IP_ALIGN + len, GFP_KERNEL); err = -ENOMEM; @@ -690,17 +744,23 @@ static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info) if (IS_ERR(flow)) goto err_kfree_skb; - err = flow_extract(packet, -1, &flow->key, &key_len, &is_frag); + err = flow_extract(packet, -1, &flow->key, &key_len); 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, + err = flow_metadata_from_nlattrs(&flow->key.phy.priority, + &flow->key.phy.in_port, + &flow->key.phy.tun_id, a[OVS_PACKET_ATTR_KEY]); if (err) goto err_flow_put; + err = validate_actions(a[OVS_PACKET_ATTR_ACTIONS], &flow->key, 0); + 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)) @@ -708,13 +768,17 @@ static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info) rcu_assign_pointer(flow->sf_acts, acts); OVS_CB(packet)->flow = flow; + packet->priority = flow->key.phy.priority; rcu_read_lock(); dp = get_dp(ovs_header->dp_ifindex); err = -ENODEV; if (!dp) goto err_unlock; + + local_bh_disable(); err = execute_actions(dp, packet); + local_bh_enable(); rcu_read_unlock(); flow_put(flow); @@ -747,11 +811,11 @@ 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; + stats->n_hit = stats->n_missed = stats->n_lost = 0; for_each_possible_cpu(i) { const struct dp_stats_percpu *percpu_stats; struct dp_stats_percpu local_stats; @@ -764,59 +828,12 @@ static void get_dp_stats(struct datapath *dp, struct ovs_dp_stats *stats) local_stats = *percpu_stats; } while (read_seqcount_retry(&percpu_stats->seqlock, seqcount)); - stats->n_frags += local_stats.n_frags; stats->n_hit += local_stats.n_hit; stats->n_missed += local_stats.n_missed; stats->n_lost += local_stats.n_lost; } } -/* 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 }, @@ -827,7 +844,7 @@ static struct genl_family dp_flow_genl_family = { .id = GENL_ID_GENERATE, .hdrsize = sizeof(struct ovs_header), .name = OVS_FLOW_FAMILY, - .version = 1, + .version = OVS_FLOW_VERSION, .maxattr = OVS_FLOW_ATTR_MAX }; @@ -837,7 +854,8 @@ static struct genl_multicast_group dp_flow_multicast_group = { /* Called with genl_lock. */ static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp, - struct sk_buff *skb, u32 pid, u32 seq, u32 flags, u8 cmd) + struct sk_buff *skb, u32 pid, + u32 seq, u32 flags, u8 cmd) { const int skb_orig_len = skb->len; const struct sw_flow_actions *sf_acts; @@ -855,7 +873,7 @@ 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) @@ -876,7 +894,8 @@ static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp, NLA_PUT_U64(skb, OVS_FLOW_ATTR_USED, flow_used_time(used)); if (stats.n_packets) - NLA_PUT(skb, OVS_FLOW_ATTR_STATS, sizeof(struct ovs_flow_stats), &stats); + NLA_PUT(skb, OVS_FLOW_ATTR_STATS, + sizeof(struct ovs_flow_stats), &stats); if (tcp_flags) NLA_PUT_U8(skb, OVS_FLOW_ATTR_TCP_FLAGS, tcp_flags); @@ -913,15 +932,24 @@ static struct sk_buff *ovs_flow_cmd_alloc_info(struct sw_flow *flow) sf_acts = rcu_dereference_protected(flow->sf_acts, lockdep_genl_is_held()); - len = nla_total_size(FLOW_BUFSIZE); /* OVS_FLOW_ATTR_KEY */ - len += nla_total_size(sf_acts->actions_len); /* OVS_FLOW_ATTR_ACTIONS */ - len += nla_total_size(sizeof(struct ovs_flow_stats)); /* OVS_FLOW_ATTR_STATS */ - len += nla_total_size(1); /* OVS_FLOW_ATTR_TCP_FLAGS */ - len += nla_total_size(8); /* OVS_FLOW_ATTR_USED */ - return genlmsg_new(NLMSG_ALIGN(sizeof(struct ovs_header)) + len, GFP_KERNEL); + /* OVS_FLOW_ATTR_KEY */ + len = nla_total_size(FLOW_BUFSIZE); + /* OVS_FLOW_ATTR_ACTIONS */ + len += nla_total_size(sf_acts->actions_len); + /* OVS_FLOW_ATTR_STATS */ + len += nla_total_size(sizeof(struct ovs_flow_stats)); + /* OVS_FLOW_ATTR_TCP_FLAGS */ + len += nla_total_size(1); + /* OVS_FLOW_ATTR_USED */ + len += nla_total_size(8); + + len += NLMSG_ALIGN(sizeof(struct ovs_header)); + + return genlmsg_new(len, GFP_KERNEL); } -static struct sk_buff *ovs_flow_cmd_build_info(struct sw_flow *flow, struct datapath *dp, +static struct sk_buff *ovs_flow_cmd_build_info(struct sw_flow *flow, + struct datapath *dp, u32 pid, u32 seq, u8 cmd) { struct sk_buff *skb; @@ -940,13 +968,11 @@ 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 +986,7 @@ 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], &key, 0); if (error) goto error; } else if (info->genlhdr->cmd == OVS_FLOW_CMD_NEW) { @@ -973,10 +999,9 @@ 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 +1010,15 @@ 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. */ @@ -1009,15 +1038,16 @@ 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); + info->snd_seq, + OVS_FLOW_CMD_NEW); } else { /* We found a matching flow. */ struct sw_flow_actions *old_acts; + struct nlattr *acts_attrs; /* Bail out if we're not allowed to modify an existing flow. * We accept NLM_F_CREATE in place of the intended NLM_F_EXCL @@ -1031,16 +1061,16 @@ 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] && - (old_acts->actions_len != nla_len(a[OVS_FLOW_ATTR_ACTIONS]) || - memcmp(old_acts->actions, nla_data(a[OVS_FLOW_ATTR_ACTIONS]), - old_acts->actions_len))) { + acts_attrs = a[OVS_FLOW_ATTR_ACTIONS]; + if (acts_attrs && + (old_acts->actions_len != nla_len(acts_attrs) || + memcmp(old_acts->actions, nla_data(acts_attrs), + old_acts->actions_len))) { struct sw_flow_actions *new_acts; - new_acts = flow_actions_alloc(a[OVS_FLOW_ATTR_ACTIONS]); + new_acts = flow_actions_alloc(acts_attrs); error = PTR_ERR(new_acts); if (IS_ERR(new_acts)) goto error; @@ -1050,7 +1080,7 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info) } reply = ovs_flow_cmd_build_info(flow, dp, info->snd_pid, - info->snd_seq, OVS_FLOW_CMD_NEW); + info->snd_seq, OVS_FLOW_CMD_NEW); /* Clear stats. */ if (a[OVS_FLOW_ATTR_CLEAR]) { @@ -1062,7 +1092,7 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info) if (!IS_ERR(reply)) genl_notify(reply, genl_info_net(info), info->snd_pid, - dp_flow_multicast_group.id, info->nlhdr, GFP_KERNEL); + dp_flow_multicast_group.id, info->nlhdr, GFP_KERNEL); else netlink_set_err(INIT_NET_GENL_SOCK, 0, dp_flow_multicast_group.id, PTR_ERR(reply)); @@ -1079,11 +1109,10 @@ 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; @@ -1098,13 +1127,12 @@ 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); + 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 +1144,10 @@ 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; @@ -1132,24 +1159,18 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info) dp = get_dp(ovs_header->dp_ifindex); if (!dp) - return -ENODEV; + 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); @@ -1172,18 +1193,17 @@ 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, + 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) break; @@ -1222,15 +1242,14 @@ 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_IPV4_FRAGS] = { .type = NLA_U32 }, - [OVS_DP_ATTR_SAMPLING] = { .type = NLA_U32 }, + [OVS_DP_ATTR_UPCALL_PID] = { .type = NLA_U32 }, }; static struct genl_family dp_datapath_genl_family = { .id = GENL_ID_GENERATE, .hdrsize = sizeof(struct ovs_header), .name = OVS_DATAPATH_FAMILY, - .version = 1, + .version = OVS_DATAPATH_VERSION, .maxattr = OVS_DP_ATTR_MAX }; @@ -1250,7 +1269,7 @@ 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)); @@ -1263,20 +1282,6 @@ static int ovs_dp_cmd_fill_info(struct datapath *dp, struct sk_buff *skb, goto nla_put_failure; get_dp_stats(dp, nla_data(nla)); - 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: @@ -1305,18 +1310,12 @@ static struct sk_buff *ovs_dp_cmd_build_info(struct datapath *dp, u32 pid, static int ovs_dp_cmd_validate(struct nlattr *a[OVS_DP_ATTR_MAX + 1]) { - if (a[OVS_DP_ATTR_IPV4_FRAGS]) { - u32 frags = nla_get_u32(a[OVS_DP_ATTR_IPV4_FRAGS]); - - if (frags != OVS_DP_FRAG_ZERO && frags != OVS_DP_FRAG_DROP) - return -EINVAL; - } - return CHECK_NUL_STRING(a[OVS_DP_ATTR_NAME], IFNAMSIZ - 1); } /* Called with genl_mutex and optionally with RTNL lock also. */ -static struct datapath *lookup_datapath(struct ovs_header *ovs_header, struct nlattr *a[OVS_DP_ATTR_MAX + 1]) +static struct datapath *lookup_datapath(struct ovs_header *ovs_header, + struct nlattr *a[OVS_DP_ATTR_MAX + 1]) { struct datapath *dp; @@ -1333,15 +1332,6 @@ static struct datapath *lookup_datapath(struct ovs_header *ovs_header, struct nl return dp ? dp : ERR_PTR(-ENODEV); } -/* Called with genl_mutex. */ -static void change_datapath(struct datapath *dp, struct nlattr *a[OVS_DP_ATTR_MAX + 1]) -{ - 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) { struct nlattr **a = info->attrs; @@ -1352,7 +1342,7 @@ 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); @@ -1377,36 +1367,35 @@ 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->stats_percpu = alloc_percpu(struct dp_stats_percpu); + if (!dp->stats_percpu) { + err = -ENOMEM; + goto err_destroy_table; + } + /* 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); + reply = ovs_dp_cmd_build_info(dp, info->snd_pid, + info->snd_seq, OVS_DP_CMD_NEW); err = PTR_ERR(reply); if (IS_ERR(reply)) goto err_destroy_local_port; @@ -1422,8 +1411,10 @@ 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: @@ -1451,12 +1442,13 @@ static int ovs_dp_cmd_del(struct sk_buff *skb, struct genl_info *info) if (IS_ERR(dp)) goto exit_unlock; - reply = ovs_dp_cmd_build_info(dp, info->snd_pid, info->snd_seq, OVS_DP_CMD_DEL); + reply = ovs_dp_cmd_build_info(dp, info->snd_pid, + info->snd_seq, OVS_DP_CMD_DEL); err = PTR_ERR(reply); if (IS_ERR(reply)) goto exit_unlock; - list_for_each_entry_safe (vport, next_vport, &dp->port_list, node) + list_for_each_entry_safe(vport, next_vport, &dp->port_list, node) if (vport->port_no != OVSP_LOCAL) dp_detach_port(vport); @@ -1499,9 +1491,8 @@ static int ovs_dp_cmd_set(struct sk_buff *skb, struct genl_info *info) if (IS_ERR(dp)) return PTR_ERR(dp); - change_datapath(dp, info->attrs); - - reply = ovs_dp_cmd_build_info(dp, info->snd_pid, info->snd_seq, OVS_DP_CMD_NEW); + reply = ovs_dp_cmd_build_info(dp, info->snd_pid, + info->snd_seq, OVS_DP_CMD_NEW); if (IS_ERR(reply)) { err = PTR_ERR(reply); netlink_set_err(INIT_NET_GENL_SOCK, 0, @@ -1528,7 +1519,8 @@ static int ovs_dp_cmd_get(struct sk_buff *skb, struct genl_info *info) if (IS_ERR(dp)) return PTR_ERR(dp); - reply = ovs_dp_cmd_build_info(dp, info->snd_pid, info->snd_seq, OVS_DP_CMD_NEW); + reply = ovs_dp_cmd_build_info(dp, info->snd_pid, + info->snd_seq, OVS_DP_CMD_NEW); if (IS_ERR(reply)) return PTR_ERR(reply); @@ -1541,7 +1533,7 @@ static int ovs_dp_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb) int skip = cb->args[0]; int i = 0; - list_for_each_entry (dp, &dps, list_node) { + list_for_each_entry(dp, &dps, list_node) { if (i < skip) continue; if (ovs_dp_cmd_fill_info(dp, skb, NETLINK_CB(cb->skb).pid, @@ -1583,15 +1575,15 @@ 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 }, }; @@ -1599,7 +1591,7 @@ static struct genl_family dp_vport_genl_family = { .id = GENL_ID_GENERATE, .hdrsize = sizeof(struct ovs_header), .name = OVS_VPORT_FAMILY, - .version = 1, + .version = OVS_VPORT_VERSION, .maxattr = OVS_VPORT_ATTR_MAX }; @@ -1613,8 +1605,6 @@ 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 +1612,27 @@ 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_TYPE, vport->ops->type); + NLA_PUT_STRING(skb, OVS_VPORT_ATTR_NAME, vport->ops->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->ops->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: @@ -1716,15 +1697,17 @@ static struct vport *lookup_vport(struct ovs_header *ovs_header, } /* Called with RTNL lock. */ -static int change_vport(struct vport *vport, struct nlattr *a[OVS_VPORT_ATTR_MAX + 1]) +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 +1723,8 @@ 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); @@ -1781,19 +1765,20 @@ 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); + dp_sysfs_add_if(vport); err = change_vport(vport, a); if (!err) { reply = ovs_vport_cmd_build_info(vport, info->snd_pid, - info->snd_seq, OVS_VPORT_CMD_NEW); + info->snd_seq, + OVS_VPORT_CMD_NEW); if (IS_ERR(reply)) err = PTR_ERR(reply); } @@ -1829,10 +1814,16 @@ 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->ops->type) + 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 +1871,7 @@ 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); @@ -2034,9 +2025,6 @@ static int dp_register_genl(void) } } - err = packet_register_mc_groups(); - if (err) - goto error; return 0; error: @@ -2051,12 +2039,17 @@ static int __init dp_init(void) BUILD_BUG_ON(sizeof(struct ovs_skb_cb) > sizeof(dummy_skb->cb)); - printk("Open vSwitch %s, built "__DATE__" "__TIME__"\n", VERSION BUILDNR); + pr_info("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; @@ -2077,6 +2070,8 @@ error_vport_exit: vport_exit(); error_flow_exit: flow_exit(); +error_tnl_exit: + tnl_exit(); error: return err; } @@ -2088,6 +2083,7 @@ static void dp_cleanup(void) unregister_netdevice_notifier(&dp_device_notifier); vport_exit(); flow_exit(); + tnl_exit(); } module_init(dp_init);