X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=datapath%2Fdatapath.c;h=98c2bafd306634e6eb2d7e34163f2afb0d31f5fa;hb=f613a0d72c521ca3a4eeb2c29ac523f6fdf72667;hp=8931456afd4d99a23a49b8217ab2ca7d64596ba1;hpb=37a1300c3ca05a97160d44dc34298c447b50ed9d;p=sliver-openvswitch.git diff --git a/datapath/datapath.c b/datapath/datapath.c index 8931456af..98c2bafd3 100644 --- a/datapath/datapath.c +++ b/datapath/datapath.c @@ -49,10 +49,15 @@ #include "datapath.h" #include "actions.h" #include "flow.h" -#include "loop_counter.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. +#endif + int (*dp_ioctl_hook)(struct net_device *dev, struct ifreq *rq, int cmd); EXPORT_SYMBOL(dp_ioctl_hook); @@ -73,25 +78,36 @@ EXPORT_SYMBOL(dp_ioctl_hook); * each other. */ -/* Protected by genl_mutex. */ -static struct datapath __rcu *dps[256]; +/* Global list of datapaths to enable dumping them all out. + * Protected by genl_mutex. + */ +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 *); /* Must be called with rcu_read_lock, genl_mutex, or RTNL lock. */ -struct datapath *get_dp(int dp_idx) +struct datapath *get_dp(int dp_ifindex) { - if (dp_idx < 0 || dp_idx >= ARRAY_SIZE(dps)) - return NULL; + struct datapath *dp = NULL; + struct net_device *dev; - return rcu_dereference_check(dps[dp_idx], rcu_read_lock_held() || - lockdep_rtnl_is_held() || - lockdep_genl_is_held()); + rcu_read_lock(); + dev = dev_get_by_index_rcu(&init_net, dp_ifindex); + if (dev) { + struct vport *vport = internal_dev_get_vport(dev); + if (vport) + dp = vport->dp; + } + rcu_read_unlock(); + + return dp; } 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()); } @@ -105,7 +121,7 @@ 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[ODPP_LOCAL])); + return vport_get_name(rcu_dereference_rtnl(dp->ports[OVSP_LOCAL])); } static inline size_t br_nlmsg_size(void) @@ -115,7 +131,6 @@ 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 */ } @@ -126,16 +141,12 @@ 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; @@ -150,7 +161,7 @@ static int dp_fill_ifinfo(struct sk_buff *skb, NLA_PUT_STRING(skb, IFLA_IFNAME, vport_get_name(port)); NLA_PUT_U32(skb, IFLA_MASTER, - vport_get_ifindex(get_vport_protected(dp, ODPP_LOCAL))); + vport_get_ifindex(get_vport_protected(dp, OVSP_LOCAL))); NLA_PUT_U32(skb, IFLA_MTU, vport_get_mtu(port)); #ifdef IFLA_OPERSTATE NLA_PUT_U8(skb, IFLA_OPERSTATE, @@ -161,9 +172,6 @@ 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: @@ -209,7 +217,7 @@ 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); } @@ -233,11 +241,11 @@ 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(); - if (p->port_no != ODPP_LOCAL) + if (p->port_no != OVSP_LOCAL) dp_sysfs_del_if(p); dp_ifinfo_notify(RTM_DELLINK, p); @@ -246,46 +254,44 @@ 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; - struct sw_flow_actions *acts; - struct loop_counter *loop; int error; 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; /* Extract flow from 'skb' into 'key'. */ - error = flow_extract(skb, p->port_no, &key, &is_frag); + error = flow_extract(skb, p->port_no, &key, &key_len, &is_frag); if (unlikely(error)) { kfree_skb(skb); return; } if (is_frag && dp->drop_frags) { - kfree_skb(skb); + 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, - flow_hash(&key), 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 = ODP_PACKET_CMD_MISS; + upcall.cmd = OVS_PACKET_CMD_MISS; upcall.key = &key; upcall.userdata = 0; upcall.sample_pool = 0; @@ -296,37 +302,12 @@ void dp_process_received_packet(struct vport *p, struct sk_buff *skb) 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); flow_used(OVS_CB(skb)->flow, skb); - - acts = rcu_dereference(OVS_CB(skb)->flow->sf_acts); - - /* Check whether we've looped too much. */ - loop = loop_get_counter(); - if (unlikely(++loop->count > MAX_LOOPS)) - loop->looping = true; - if (unlikely(loop->looping)) { - loop_suppress(dp, acts); - kfree_skb(skb); - goto out_loop; - } - - /* Execute actions. */ - execute_actions(dp, skb, &OVS_CB(skb)->flow->key, acts->actions, - acts->actions_len); - - /* Check whether sub-actions looped too much. */ - if (unlikely(loop->looping)) - loop_suppress(dp, acts); - -out_loop: - /* Decrement loop counter. */ - if (!--loop->count) - loop->looping = false; - loop_put_counter(); + execute_actions(dp, skb); out: /* Update datapath statistics. */ @@ -347,7 +328,6 @@ static void copy_and_csum_skb(struct sk_buff *skb, void *to) get_skb_csum_pointers(skb, &csum_start, &csum_offset); csum_start -= skb_headroom(skb); - BUG_ON(csum_start >= skb_headlen(skb)); skb_copy_bits(skb, 0, to, csum_start); @@ -356,34 +336,108 @@ static void copy_and_csum_skb(struct sk_buff *skb, void *to) *(__sum16 *)(to + csum_start + csum_offset) = csum_fold(csum); } -static struct genl_family dp_packet_genl_family; +static struct genl_family dp_packet_genl_family = { + .id = GENL_ID_GENERATE, + .hdrsize = sizeof(struct ovs_header), + .name = OVS_PACKET_FAMILY, + .version = 1, + .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 int packet_mc_group(struct datapath *dp, u8 cmd) +static u32 packet_mc_group(struct datapath *dp, u8 cmd) { + u32 idx; BUILD_BUG_ON_NOT_POWER_OF_2(PACKET_N_MC_GROUPS); - return jhash_2words(dp->dp_idx, cmd, 0) & (PACKET_N_MC_GROUPS - 1); + + 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) +{ + struct dp_stats_percpu *stats; + int err; + + WARN_ON_ONCE(skb_shared(skb)); + + 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); + + if (IS_ERR(nskb)) { + kfree_skb(skb); + err = PTR_ERR(nskb); + goto err; + } + consume_skb(skb); + skb = nskb; + } + + err = queue_userspace_packets(dp, 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_control_packets(struct datapath *dp, struct sk_buff *skb, +static int queue_userspace_packets(struct datapath *dp, struct sk_buff *skb, const struct dp_upcall_info *upcall_info) { u32 group = packet_mc_group(dp, upcall_info->cmd); struct sk_buff *nskb; - int port_no; int err; - if (OVS_CB(skb)->vport) - port_no = OVS_CB(skb)->vport->port_no; - else - port_no = ODPP_LOCAL; - do { - struct odp_header *upcall; + struct ovs_header *upcall; struct sk_buff *user_skb; /* to be queued to userspace */ struct nlattr *nla; unsigned int len; @@ -391,8 +445,14 @@ static int queue_control_packets(struct datapath *dp, struct sk_buff *skb, nskb = skb->next; skb->next = NULL; - len = sizeof(struct odp_header); - len += nla_total_size(4); /* ODP_PACKET_ATTR_TYPE. */ + 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) @@ -409,26 +469,26 @@ static int queue_control_packets(struct datapath *dp, struct sk_buff *skb, } upcall = genlmsg_put(user_skb, 0, 0, &dp_packet_genl_family, 0, upcall_info->cmd); - upcall->dp_idx = dp->dp_idx; + upcall->dp_ifindex = dp->dp_ifindex; - nla = nla_nest_start(user_skb, ODP_PACKET_ATTR_KEY); + 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, ODP_PACKET_ATTR_USERDATA, 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, ODP_PACKET_ATTR_SAMPLE_POOL, 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, ODP_PACKET_ATTR_ACTIONS); + 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 = __nla_reserve(user_skb, ODP_PACKET_ATTR_PACKET, skb->len); + 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 @@ -438,7 +498,7 @@ static int queue_control_packets(struct datapath *dp, struct sk_buff *skb, if (err) goto err_kfree_skbs; - kfree_skb(skb); + consume_skb(skb); skb = nskb; } while (skb); return 0; @@ -452,106 +512,25 @@ err_kfree_skbs: return err; } -/* 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 ODP_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. - */ -static struct genl_multicast_group packet_mc_groups[PACKET_N_MC_GROUPS]; - -static struct genl_family dp_packet_genl_family = { - .id = GENL_ID_GENERATE, - .hdrsize = sizeof(struct odp_header), - .name = ODP_PACKET_FAMILY, - .version = 1, - .maxattr = ODP_PACKET_ATTR_MAX -}; - -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) -{ - struct dp_stats_percpu *stats; - int err; - - WARN_ON_ONCE(skb_shared(skb)); - - forward_ip_summed(skb); - - err = vswitch_skb_checksum_setup(skb); - if (err) - goto err_kfree_skb; - - /* 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); - - kfree_skb(skb); - skb = nskb; - if (IS_ERR(skb)) { - err = PTR_ERR(skb); - goto err; - } - } - - return queue_control_packets(dp, skb, upcall_info); - -err_kfree_skb: - kfree_skb(skb); -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; -} - /* Called with genl_mutex. */ -static int flush_flows(int dp_idx) +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_idx); + dp = get_dp(dp_ifindex); if (!dp) 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; } @@ -561,58 +540,56 @@ static int validate_actions(const struct nlattr *attr) int rem; nla_for_each_nested(a, attr, rem) { - static const u32 action_lens[ODPAT_MAX + 1] = { - [ODPAT_OUTPUT] = 4, - [ODPAT_CONTROLLER] = 8, - [ODPAT_SET_DL_TCI] = 2, - [ODPAT_STRIP_VLAN] = 0, - [ODPAT_SET_DL_SRC] = ETH_ALEN, - [ODPAT_SET_DL_DST] = ETH_ALEN, - [ODPAT_SET_NW_SRC] = 4, - [ODPAT_SET_NW_DST] = 4, - [ODPAT_SET_NW_TOS] = 1, - [ODPAT_SET_TP_SRC] = 2, - [ODPAT_SET_TP_DST] = 2, - [ODPAT_SET_TUNNEL] = 8, - [ODPAT_SET_PRIORITY] = 4, - [ODPAT_POP_PRIORITY] = 0, - [ODPAT_DROP_SPOOFED_ARP] = 0, + static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = { + [OVS_ACTION_ATTR_OUTPUT] = 4, + [OVS_ACTION_ATTR_USERSPACE] = 8, + [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, + [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, }; int type = nla_type(a); - if (type > ODPAT_MAX || nla_len(a) != action_lens[type]) + if (type > OVS_ACTION_ATTR_MAX || nla_len(a) != action_lens[type]) return -EINVAL; switch (type) { - case ODPAT_UNSPEC: + case OVS_ACTION_ATTR_UNSPEC: return -EINVAL; - case ODPAT_CONTROLLER: - case ODPAT_STRIP_VLAN: - case ODPAT_SET_DL_SRC: - case ODPAT_SET_DL_DST: - case ODPAT_SET_NW_SRC: - case ODPAT_SET_NW_DST: - case ODPAT_SET_TP_SRC: - case ODPAT_SET_TP_DST: - case ODPAT_SET_TUNNEL: - case ODPAT_SET_PRIORITY: - case ODPAT_POP_PRIORITY: - case ODPAT_DROP_SPOOFED_ARP: + case OVS_ACTION_ATTR_USERSPACE: + 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: + 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. */ break; - case ODPAT_OUTPUT: + case OVS_ACTION_ATTR_OUTPUT: if (nla_get_u32(a) >= DP_MAX_PORTS) return -EINVAL; break; - case ODPAT_SET_DL_TCI: + case OVS_ACTION_ATTR_PUSH_VLAN: if (nla_get_be16(a) & htons(VLAN_CFI_MASK)) return -EINVAL; break; - case ODPAT_SET_NW_TOS: + case OVS_ACTION_ATTR_SET_NW_TOS: if (nla_get_u8(a) & INET_ECN_MASK) return -EINVAL; break; @@ -635,48 +612,38 @@ 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)) - return PTR_ERR(new_table); - - rcu_assign_pointer(dp->table, new_table); - tbl_deferred_destroy(old_table, NULL); - - return 0; -} - -static int odp_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info) +static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info) { - struct odp_header *odp_header = info->userhdr; + struct ovs_header *ovs_header = info->userhdr; struct nlattr **a = info->attrs; + struct sw_flow_actions *acts; struct sk_buff *packet; - struct sw_flow_key key; + struct sw_flow *flow; struct datapath *dp; struct ethhdr *eth; bool is_frag; + int len; int err; + int key_len; err = -EINVAL; - if (!a[ODP_PACKET_ATTR_PACKET] || !a[ODP_PACKET_ATTR_ACTIONS] || - nla_len(a[ODP_PACKET_ATTR_PACKET]) < ETH_HLEN) - goto exit; + if (!a[OVS_PACKET_ATTR_PACKET] || !a[OVS_PACKET_ATTR_KEY] || + !a[OVS_PACKET_ATTR_ACTIONS] || + nla_len(a[OVS_PACKET_ATTR_PACKET]) < ETH_HLEN) + goto err; - err = validate_actions(a[ODP_PACKET_ATTR_ACTIONS]); + err = validate_actions(a[OVS_PACKET_ATTR_ACTIONS]); if (err) - goto exit; + goto err; - packet = skb_clone(skb, GFP_KERNEL); + len = nla_len(a[OVS_PACKET_ATTR_PACKET]); + packet = __dev_alloc_skb(NET_IP_ALIGN + len, GFP_KERNEL); err = -ENOMEM; if (!packet) - goto exit; - packet->data = nla_data(a[ODP_PACKET_ATTR_PACKET]); - packet->len = nla_len(a[ODP_PACKET_ATTR_PACKET]); + goto err; + skb_reserve(packet, NET_IP_ALIGN); + + memcpy(__skb_put(packet, len), nla_data(a[OVS_PACKET_ATTR_PACKET]), len); skb_reset_mac_header(packet); eth = eth_hdr(packet); @@ -689,39 +656,78 @@ static int odp_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info) else packet->protocol = htons(ETH_P_802_2); - err = flow_extract(packet, -1, &key, &is_frag); + /* Build an sw_flow for sending this packet. */ + flow = flow_alloc(); + err = PTR_ERR(flow); + if (IS_ERR(flow)) + goto err_kfree_skb; + + err = flow_extract(packet, -1, &flow->key, &key_len, &is_frag); if (err) - goto exit; + goto err_flow_put; + + err = flow_metadata_from_nlattrs(&flow->key.eth.in_port, + &flow->key.eth.tun_id, + a[OVS_PACKET_ATTR_KEY]); + 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)) + goto err_flow_put; + rcu_assign_pointer(flow->sf_acts, acts); + + OVS_CB(packet)->flow = flow; rcu_read_lock(); - dp = get_dp(odp_header->dp_idx); + dp = get_dp(ovs_header->dp_ifindex); err = -ENODEV; - if (dp) - err = execute_actions(dp, packet, &key, - nla_data(a[ODP_PACKET_ATTR_ACTIONS]), - nla_len(a[ODP_PACKET_ATTR_ACTIONS])); + 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); + + err = execute_actions(dp, packet); rcu_read_unlock(); -exit: + flow_put(flow); + return err; + +err_unlock: + rcu_read_unlock(); +err_flow_put: + flow_put(flow); +err_kfree_skb: + kfree_skb(packet); +err: return err; } -static const struct nla_policy packet_policy[ODP_PACKET_ATTR_MAX + 1] = { - [ODP_PACKET_ATTR_PACKET] = { .type = NLA_UNSPEC }, - [ODP_PACKET_ATTR_ACTIONS] = { .type = NLA_NESTED }, +static const struct nla_policy packet_policy[OVS_PACKET_ATTR_MAX + 1] = { + [OVS_PACKET_ATTR_PACKET] = { .type = NLA_UNSPEC }, + [OVS_PACKET_ATTR_KEY] = { .type = NLA_NESTED }, + [OVS_PACKET_ATTR_ACTIONS] = { .type = NLA_NESTED }, }; static struct genl_ops dp_packet_genl_ops[] = { - { .cmd = ODP_PACKET_CMD_EXECUTE, + { .cmd = OVS_PACKET_CMD_EXECUTE, .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */ .policy = packet_policy, - .doit = odp_packet_cmd_execute + .doit = ovs_packet_cmd_execute } }; -static void get_dp_stats(struct datapath *dp, struct odp_stats *stats) +static void get_dp_stats(struct datapath *dp, struct ovs_dp_stats *stats) { int i; + struct flow_table *table = get_table_protected(dp); + + 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) { @@ -743,92 +749,47 @@ static void get_dp_stats(struct datapath *dp, struct odp_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 (!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[ODP_FLOW_ATTR_MAX + 1] = { - [ODP_FLOW_ATTR_KEY] = { .type = NLA_NESTED }, - [ODP_FLOW_ATTR_ACTIONS] = { .type = NLA_NESTED }, - [ODP_FLOW_ATTR_CLEAR] = { .type = NLA_FLAG }, +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 }, + [OVS_FLOW_ATTR_CLEAR] = { .type = NLA_FLAG }, }; static struct genl_family dp_flow_genl_family = { .id = GENL_ID_GENERATE, - .hdrsize = sizeof(struct odp_header), - .name = ODP_FLOW_FAMILY, + .hdrsize = sizeof(struct ovs_header), + .name = OVS_FLOW_FAMILY, .version = 1, - .maxattr = ODP_FLOW_ATTR_MAX + .maxattr = OVS_FLOW_ATTR_MAX }; static struct genl_multicast_group dp_flow_multicast_group = { - .name = ODP_FLOW_MCGROUP + .name = OVS_FLOW_MCGROUP }; /* Called with genl_lock. */ -static int odp_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp, +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) { const int skb_orig_len = skb->len; const struct sw_flow_actions *sf_acts; - struct odp_flow_stats stats; - struct odp_header *odp_header; + struct ovs_flow_stats stats; + struct ovs_header *ovs_header; struct nlattr *nla; unsigned long used; u8 tcp_flags; - int nla_len; int err; sf_acts = rcu_dereference_protected(flow->sf_acts, lockdep_genl_is_held()); - odp_header = genlmsg_put(skb, pid, seq, &dp_flow_genl_family, flags, cmd); - if (!odp_header) + ovs_header = genlmsg_put(skb, pid, seq, &dp_flow_genl_family, flags, cmd); + if (!ovs_header) return -EMSGSIZE; - odp_header->dp_idx = dp->dp_idx; + ovs_header->dp_ifindex = dp->dp_ifindex; - nla = nla_nest_start(skb, ODP_FLOW_ATTR_KEY); + nla = nla_nest_start(skb, OVS_FLOW_ATTR_KEY); if (!nla) goto nla_put_failure; err = flow_to_nlattrs(&flow->key, skb); @@ -844,42 +805,39 @@ static int odp_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp, spin_unlock_bh(&flow->lock); if (used) - NLA_PUT_MSECS(skb, ODP_FLOW_ATTR_USED, used); + NLA_PUT_U64(skb, OVS_FLOW_ATTR_USED, flow_used_time(used)); if (stats.n_packets) - NLA_PUT(skb, ODP_FLOW_ATTR_STATS, sizeof(struct odp_flow_stats), &stats); + NLA_PUT(skb, OVS_FLOW_ATTR_STATS, sizeof(struct ovs_flow_stats), &stats); if (tcp_flags) - NLA_PUT_U8(skb, ODP_FLOW_ATTR_TCP_FLAGS, tcp_flags); - - /* If ODP_FLOW_ATTR_ACTIONS doesn't fit, and this is the first flow to - * be dumped into 'skb', then expand the skb. This is unusual for - * Netlink but individual action lists can be longer than a page and - * thus entirely undumpable if we didn't do this. */ - nla_len = nla_total_size(sf_acts->actions_len); - if (nla_len > skb_tailroom(skb) && !skb_orig_len) { - int hdr_off = (unsigned char *)odp_header - skb->data; - - err = pskb_expand_head(skb, 0, nla_len - skb_tailroom(skb), GFP_KERNEL); - if (err) - goto error; - - odp_header = (struct odp_header *)(skb->data + hdr_off); - } - nla = nla_nest_start(skb, ODP_FLOW_ATTR_ACTIONS); - memcpy(__skb_put(skb, sf_acts->actions_len), sf_acts->actions, sf_acts->actions_len); - nla_nest_end(skb, nla); + NLA_PUT_U8(skb, OVS_FLOW_ATTR_TCP_FLAGS, tcp_flags); + + /* If OVS_FLOW_ATTR_ACTIONS doesn't fit, skip dumping the actions if + * this is the first flow to be dumped into 'skb'. This is unusual for + * Netlink but individual action lists can be longer than + * NLMSG_GOODSIZE and thus entirely undumpable if we didn't do this. + * The userspace caller can always fetch the actions separately if it + * really wants them. (Most userspace callers in fact don't care.) + * + * This can only fail for dump operations because the skb is always + * properly sized for single flows. + */ + err = nla_put(skb, OVS_FLOW_ATTR_ACTIONS, sf_acts->actions_len, + sf_acts->actions); + if (err < 0 && skb_orig_len) + goto error; - return genlmsg_end(skb, odp_header); + return genlmsg_end(skb, ovs_header); nla_put_failure: err = -EMSGSIZE; error: - genlmsg_cancel(skb, odp_header); + genlmsg_cancel(skb, ovs_header); return err; } -static struct sk_buff *odp_flow_cmd_alloc_info(struct sw_flow *flow) +static struct sk_buff *ovs_flow_cmd_alloc_info(struct sw_flow *flow) { const struct sw_flow_actions *sf_acts; int len; @@ -887,82 +845,84 @@ static struct sk_buff *odp_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); /* ODP_FLOW_ATTR_KEY */ - len += nla_total_size(sf_acts->actions_len); /* ODP_FLOW_ATTR_ACTIONS */ - len += nla_total_size(sizeof(struct odp_flow_stats)); /* ODP_FLOW_ATTR_STATS */ - len += nla_total_size(1); /* ODP_FLOW_ATTR_TCP_FLAGS */ - len += nla_total_size(8); /* ODP_FLOW_ATTR_USED */ - return genlmsg_new(NLMSG_ALIGN(sizeof(struct odp_header)) + len, GFP_KERNEL); + 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); } -static struct sk_buff *odp_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; int retval; - skb = odp_flow_cmd_alloc_info(flow); + skb = ovs_flow_cmd_alloc_info(flow); if (!skb) return ERR_PTR(-ENOMEM); - retval = odp_flow_cmd_fill_info(flow, dp, skb, pid, seq, 0, cmd); + retval = ovs_flow_cmd_fill_info(flow, dp, skb, pid, seq, 0, cmd); BUG_ON(retval < 0); return skb; } -static int odp_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info) +static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info) { struct nlattr **a = info->attrs; - struct odp_header *odp_header = info->userhdr; - struct tbl_node *flow_node; + struct ovs_header *ovs_header = info->userhdr; 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; /* Extract key. */ error = -EINVAL; - if (!a[ODP_FLOW_ATTR_KEY]) + if (!a[OVS_FLOW_ATTR_KEY]) goto error; - error = flow_from_nlattrs(&key, a[ODP_FLOW_ATTR_KEY]); + error = flow_from_nlattrs(&key, &key_len, a[OVS_FLOW_ATTR_KEY]); if (error) goto error; /* Validate actions. */ - if (a[ODP_FLOW_ATTR_ACTIONS]) { - error = validate_actions(a[ODP_FLOW_ATTR_ACTIONS]); + if (a[OVS_FLOW_ATTR_ACTIONS]) { + error = validate_actions(a[OVS_FLOW_ATTR_ACTIONS]); if (error) goto error; - } else if (info->genlhdr->cmd == ODP_FLOW_CMD_NEW) { + } else if (info->genlhdr->cmd == OVS_FLOW_CMD_NEW) { error = -EINVAL; goto error; } - dp = get_dp(odp_header->dp_idx); + dp = get_dp(ovs_header->dp_ifindex); error = -ENODEV; if (!dp) goto error; - hash = flow_hash(&key); table = get_table_protected(dp); - flow_node = tbl_lookup(table, &key, 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. */ error = -ENOENT; - if (info->genlhdr->cmd == ODP_FLOW_CMD_SET) + if (info->genlhdr->cmd == OVS_FLOW_CMD_SET) 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. */ @@ -975,19 +935,18 @@ static int odp_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info) clear_stats(flow); /* Obtain actions. */ - acts = flow_actions_alloc(a[ODP_FLOW_ATTR_ACTIONS]); + acts = flow_actions_alloc(a[OVS_FLOW_ATTR_ACTIONS]); error = PTR_ERR(acts); if (IS_ERR(acts)) goto error_free_flow; 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 = odp_flow_cmd_build_info(flow, dp, info->snd_pid, - info->snd_seq, ODP_FLOW_CMD_NEW); + reply = ovs_flow_cmd_build_info(flow, dp, info->snd_pid, + info->snd_seq, OVS_FLOW_CMD_NEW); } else { /* We found a matching flow. */ struct sw_flow_actions *old_acts; @@ -999,21 +958,20 @@ static int odp_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info) * gets fixed. */ error = -EEXIST; - if (info->genlhdr->cmd == ODP_FLOW_CMD_NEW && + if (info->genlhdr->cmd == OVS_FLOW_CMD_NEW && info->nlhdr->nlmsg_flags & (NLM_F_CREATE | NLM_F_EXCL)) goto error; /* Update actions. */ - flow = flow_cast(flow_node); old_acts = rcu_dereference_protected(flow->sf_acts, lockdep_genl_is_held()); - if (a[ODP_FLOW_ATTR_ACTIONS] && - (old_acts->actions_len != nla_len(a[ODP_FLOW_ATTR_ACTIONS]) || - memcmp(old_acts->actions, nla_data(a[ODP_FLOW_ATTR_ACTIONS]), + 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))) { struct sw_flow_actions *new_acts; - new_acts = flow_actions_alloc(a[ODP_FLOW_ATTR_ACTIONS]); + new_acts = flow_actions_alloc(a[OVS_FLOW_ATTR_ACTIONS]); error = PTR_ERR(new_acts); if (IS_ERR(new_acts)) goto error; @@ -1022,11 +980,11 @@ static int odp_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info) flow_deferred_free_acts(old_acts); } - reply = odp_flow_cmd_build_info(flow, dp, info->snd_pid, - info->snd_seq, ODP_FLOW_CMD_NEW); + reply = ovs_flow_cmd_build_info(flow, dp, info->snd_pid, + info->snd_seq, OVS_FLOW_CMD_NEW); /* Clear stats. */ - if (a[ODP_FLOW_ATTR_CLEAR]) { + if (a[OVS_FLOW_ATTR_CLEAR]) { spin_lock_bh(&flow->lock); clear_stats(flow); spin_unlock_bh(&flow->lock); @@ -1047,81 +1005,75 @@ error: return error; } -static int odp_flow_cmd_get(struct sk_buff *skb, struct genl_info *info) +static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info) { struct nlattr **a = info->attrs; - struct odp_header *odp_header = info->userhdr; + 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; - if (!a[ODP_FLOW_ATTR_KEY]) + if (!a[OVS_FLOW_ATTR_KEY]) return -EINVAL; - err = flow_from_nlattrs(&key, a[ODP_FLOW_ATTR_KEY]); + err = flow_from_nlattrs(&key, &key_len, a[OVS_FLOW_ATTR_KEY]); if (err) return err; - dp = get_dp(odp_header->dp_idx); + dp = get_dp(ovs_header->dp_ifindex); if (!dp) return -ENODEV; table = get_table_protected(dp); - flow_node = tbl_lookup(table, &key, flow_hash(&key), flow_cmp); - if (!flow_node) + flow = flow_tbl_lookup(table, &key, key_len); + if (!flow) return -ENOENT; - flow = flow_cast(flow_node); - reply = odp_flow_cmd_build_info(flow, dp, info->snd_pid, info->snd_seq, ODP_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); return genlmsg_reply(reply, info); } -static int odp_flow_cmd_del(struct sk_buff *skb, struct genl_info *info) +static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info) { struct nlattr **a = info->attrs; - struct odp_header *odp_header = info->userhdr; + 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; - if (!a[ODP_FLOW_ATTR_KEY]) - return flush_flows(odp_header->dp_idx); - err = flow_from_nlattrs(&key, a[ODP_FLOW_ATTR_KEY]); + if (!a[OVS_FLOW_ATTR_KEY]) + return flush_flows(ovs_header->dp_ifindex); + err = flow_from_nlattrs(&key, &key_len, a[OVS_FLOW_ATTR_KEY]); if (err) return err; - dp = get_dp(odp_header->dp_idx); + dp = get_dp(ovs_header->dp_ifindex); if (!dp) return -ENODEV; table = get_table_protected(dp); - flow_node = tbl_lookup(table, &key, flow_hash(&key), flow_cmp); - if (!flow_node) + flow = flow_tbl_lookup(table, &key, key_len); + if (!flow) return -ENOENT; - flow = flow_cast(flow_node); - reply = odp_flow_cmd_alloc_info(flow); + 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 = odp_flow_cmd_fill_info(flow, dp, reply, info->snd_pid, - info->snd_seq, 0, ODP_FLOW_CMD_DEL); + err = ovs_flow_cmd_fill_info(flow, dp, reply, info->snd_pid, + info->snd_seq, 0, OVS_FLOW_CMD_DEL); BUG_ON(err < 0); flow_deferred_free(flow); @@ -1131,30 +1083,28 @@ static int odp_flow_cmd_del(struct sk_buff *skb, struct genl_info *info) return 0; } -static int odp_flow_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb) +static int ovs_flow_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb) { - struct odp_header *odp_header = genlmsg_data(nlmsg_data(cb->nlh)); + struct ovs_header *ovs_header = genlmsg_data(nlmsg_data(cb->nlh)); struct datapath *dp; - dp = get_dp(odp_header->dp_idx); + dp = get_dp(ovs_header->dp_ifindex); if (!dp) 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 (odp_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, - ODP_FLOW_CMD_NEW) < 0) + OVS_FLOW_CMD_NEW) < 0) break; cb->args[0] = bucket; @@ -1164,97 +1114,97 @@ static int odp_flow_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb) } static struct genl_ops dp_flow_genl_ops[] = { - { .cmd = ODP_FLOW_CMD_NEW, + { .cmd = OVS_FLOW_CMD_NEW, .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */ .policy = flow_policy, - .doit = odp_flow_cmd_new_or_set + .doit = ovs_flow_cmd_new_or_set }, - { .cmd = ODP_FLOW_CMD_DEL, + { .cmd = OVS_FLOW_CMD_DEL, .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */ .policy = flow_policy, - .doit = odp_flow_cmd_del + .doit = ovs_flow_cmd_del }, - { .cmd = ODP_FLOW_CMD_GET, + { .cmd = OVS_FLOW_CMD_GET, .flags = 0, /* OK for unprivileged users. */ .policy = flow_policy, - .doit = odp_flow_cmd_get, - .dumpit = odp_flow_cmd_dump + .doit = ovs_flow_cmd_get, + .dumpit = ovs_flow_cmd_dump }, - { .cmd = ODP_FLOW_CMD_SET, + { .cmd = OVS_FLOW_CMD_SET, .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */ .policy = flow_policy, - .doit = odp_flow_cmd_new_or_set, + .doit = ovs_flow_cmd_new_or_set, }, }; -static const struct nla_policy datapath_policy[ODP_DP_ATTR_MAX + 1] = { +static const struct nla_policy datapath_policy[OVS_DP_ATTR_MAX + 1] = { #ifdef HAVE_NLA_NUL_STRING - [ODP_DP_ATTR_NAME] = { .type = NLA_NUL_STRING, .len = IFNAMSIZ - 1 }, + [OVS_DP_ATTR_NAME] = { .type = NLA_NUL_STRING, .len = IFNAMSIZ - 1 }, #endif - [ODP_DP_ATTR_IPV4_FRAGS] = { .type = NLA_U32 }, - [ODP_DP_ATTR_SAMPLING] = { .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 = { .id = GENL_ID_GENERATE, - .hdrsize = sizeof(struct odp_header), - .name = ODP_DATAPATH_FAMILY, + .hdrsize = sizeof(struct ovs_header), + .name = OVS_DATAPATH_FAMILY, .version = 1, - .maxattr = ODP_DP_ATTR_MAX + .maxattr = OVS_DP_ATTR_MAX }; static struct genl_multicast_group dp_datapath_multicast_group = { - .name = ODP_DATAPATH_MCGROUP + .name = OVS_DATAPATH_MCGROUP }; -static int odp_dp_cmd_fill_info(struct datapath *dp, struct sk_buff *skb, +static int ovs_dp_cmd_fill_info(struct datapath *dp, struct sk_buff *skb, u32 pid, u32 seq, u32 flags, u8 cmd) { - struct odp_header *odp_header; + struct ovs_header *ovs_header; struct nlattr *nla; int err; - odp_header = genlmsg_put(skb, pid, seq, &dp_datapath_genl_family, + ovs_header = genlmsg_put(skb, pid, seq, &dp_datapath_genl_family, flags, cmd); - if (!odp_header) + if (!ovs_header) goto error; - odp_header->dp_idx = dp->dp_idx; + ovs_header->dp_ifindex = dp->dp_ifindex; rcu_read_lock(); - err = nla_put_string(skb, ODP_DP_ATTR_NAME, dp_name(dp)); + err = nla_put_string(skb, OVS_DP_ATTR_NAME, dp_name(dp)); rcu_read_unlock(); if (err) goto nla_put_failure; - nla = nla_reserve(skb, ODP_DP_ATTR_STATS, sizeof(struct odp_stats)); + nla = nla_reserve(skb, OVS_DP_ATTR_STATS, sizeof(struct ovs_dp_stats)); if (!nla) goto nla_put_failure; get_dp_stats(dp, nla_data(nla)); - NLA_PUT_U32(skb, ODP_DP_ATTR_IPV4_FRAGS, - dp->drop_frags ? ODP_DP_FRAG_DROP : ODP_DP_FRAG_ZERO); + 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, ODP_DP_ATTR_SAMPLING, dp->sflow_probability); + NLA_PUT_U32(skb, OVS_DP_ATTR_SAMPLING, dp->sflow_probability); - nla = nla_nest_start(skb, ODP_DP_ATTR_MCGROUPS); + nla = nla_nest_start(skb, OVS_DP_ATTR_MCGROUPS); if (!nla) goto nla_put_failure; - NLA_PUT_U32(skb, ODP_PACKET_CMD_MISS, packet_mc_group(dp, ODP_PACKET_CMD_MISS)); - NLA_PUT_U32(skb, ODP_PACKET_CMD_ACTION, packet_mc_group(dp, ODP_PACKET_CMD_ACTION)); - NLA_PUT_U32(skb, ODP_PACKET_CMD_SAMPLE, packet_mc_group(dp, ODP_PACKET_CMD_SAMPLE)); + 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, odp_header); + return genlmsg_end(skb, ovs_header); nla_put_failure: - genlmsg_cancel(skb, odp_header); + genlmsg_cancel(skb, ovs_header); error: return -EMSGSIZE; } -static struct sk_buff *odp_dp_cmd_build_info(struct datapath *dp, u32 pid, +static struct sk_buff *ovs_dp_cmd_build_info(struct datapath *dp, u32 pid, u32 seq, u8 cmd) { struct sk_buff *skb; @@ -1264,7 +1214,7 @@ static struct sk_buff *odp_dp_cmd_build_info(struct datapath *dp, u32 pid, if (!skb) return ERR_PTR(-ENOMEM); - retval = odp_dp_cmd_fill_info(dp, skb, pid, seq, 0, cmd); + retval = ovs_dp_cmd_fill_info(dp, skb, pid, seq, 0, cmd); if (retval < 0) { kfree_skb(skb); return ERR_PTR(retval); @@ -1272,66 +1222,59 @@ static struct sk_buff *odp_dp_cmd_build_info(struct datapath *dp, u32 pid, return skb; } -static int odp_dp_cmd_validate(struct nlattr *a[ODP_DP_ATTR_MAX + 1]) +static int ovs_dp_cmd_validate(struct nlattr *a[OVS_DP_ATTR_MAX + 1]) { - if (a[ODP_DP_ATTR_IPV4_FRAGS]) { - u32 frags = nla_get_u32(a[ODP_DP_ATTR_IPV4_FRAGS]); + if (a[OVS_DP_ATTR_IPV4_FRAGS]) { + u32 frags = nla_get_u32(a[OVS_DP_ATTR_IPV4_FRAGS]); - if (frags != ODP_DP_FRAG_ZERO && frags != ODP_DP_FRAG_DROP) + if (frags != OVS_DP_FRAG_ZERO && frags != OVS_DP_FRAG_DROP) return -EINVAL; } - return VERIFY_NUL_STRING(a[ODP_DP_ATTR_NAME], IFNAMSIZ - 1); + 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 odp_header *odp_header, struct nlattr *a[ODP_DP_ATTR_MAX + 1]) +static struct datapath *lookup_datapath(struct ovs_header *ovs_header, struct nlattr *a[OVS_DP_ATTR_MAX + 1]) { - if (!a[ODP_DP_ATTR_NAME]) { - struct datapath *dp = get_dp(odp_header->dp_idx); - if (!dp) - return ERR_PTR(-ENODEV); - return dp; - } else { + struct datapath *dp; + + if (!a[OVS_DP_ATTR_NAME]) + dp = get_dp(ovs_header->dp_ifindex); + else { struct vport *vport; - int dp_idx; rcu_read_lock(); - vport = vport_locate(nla_data(a[ODP_DP_ATTR_NAME])); - dp_idx = vport && vport->port_no == ODPP_LOCAL ? vport->dp->dp_idx : -1; + vport = vport_locate(nla_data(a[OVS_DP_ATTR_NAME])); + dp = vport && vport->port_no == OVSP_LOCAL ? vport->dp : NULL; rcu_read_unlock(); - - if (dp_idx < 0) - return ERR_PTR(-ENODEV); - return vport->dp; } + return dp ? dp : ERR_PTR(-ENODEV); } /* Called with genl_mutex. */ -static void change_datapath(struct datapath *dp, struct nlattr *a[ODP_DP_ATTR_MAX + 1]) +static void change_datapath(struct datapath *dp, struct nlattr *a[OVS_DP_ATTR_MAX + 1]) { - if (a[ODP_DP_ATTR_IPV4_FRAGS]) - dp->drop_frags = nla_get_u32(a[ODP_DP_ATTR_IPV4_FRAGS]) == ODP_DP_FRAG_DROP; - if (a[ODP_DP_ATTR_SAMPLING]) - dp->sflow_probability = nla_get_u32(a[ODP_DP_ATTR_SAMPLING]); + 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 odp_dp_cmd_new(struct sk_buff *skb, struct genl_info *info) +static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info) { struct nlattr **a = info->attrs; - struct odp_header *odp_header = info->userhdr; struct vport_parms parms; struct sk_buff *reply; struct datapath *dp; struct vport *vport; - int dp_idx; int err; err = -EINVAL; - if (!a[ODP_DP_ATTR_NAME]) + if (!a[OVS_DP_ATTR_NAME]) goto err; - err = odp_dp_cmd_validate(a); + err = ovs_dp_cmd_validate(a); if (err) goto err; @@ -1340,28 +1283,11 @@ static int odp_dp_cmd_new(struct sk_buff *skb, struct genl_info *info) if (!try_module_get(THIS_MODULE)) goto err_unlock_rtnl; - dp_idx = odp_header->dp_idx; - if (dp_idx < 0) { - err = -EFBIG; - for (dp_idx = 0; dp_idx < ARRAY_SIZE(dps); dp_idx++) { - if (get_dp(dp_idx)) - continue; - err = 0; - break; - } - } else if (dp_idx < ARRAY_SIZE(dps)) - err = get_dp(dp_idx) ? -EBUSY : 0; - else - err = -EINVAL; - if (err) - goto err_put_module; - err = -ENOMEM; dp = kzalloc(sizeof(*dp), GFP_KERNEL); if (dp == NULL) goto err_put_module; INIT_LIST_HEAD(&dp->port_list); - dp->dp_idx = dp_idx; /* Initialize kobject for bridge. This will be added as * /sys/class/net//brif later, if sysfs is enabled. */ @@ -1370,16 +1296,16 @@ static int odp_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; /* Set up our datapath device. */ - parms.name = nla_data(a[ODP_DP_ATTR_NAME]); - parms.type = ODP_VPORT_TYPE_INTERNAL; + parms.name = nla_data(a[OVS_DP_ATTR_NAME]); + parms.type = OVS_VPORT_TYPE_INTERNAL; parms.options = NULL; parms.dp = dp; - parms.port_no = ODPP_LOCAL; + parms.port_no = OVSP_LOCAL; vport = new_vport(&parms); if (IS_ERR(vport)) { err = PTR_ERR(vport); @@ -1388,6 +1314,7 @@ static int odp_dp_cmd_new(struct sk_buff *skb, struct genl_info *info) goto err_destroy_table; } + dp->dp_ifindex = vport_get_ifindex(vport); dp->drop_frags = 0; dp->stats_percpu = alloc_percpu(struct dp_stats_percpu); @@ -1398,12 +1325,12 @@ static int odp_dp_cmd_new(struct sk_buff *skb, struct genl_info *info) change_datapath(dp, a); - reply = odp_dp_cmd_build_info(dp, info->snd_pid, info->snd_seq, ODP_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; - rcu_assign_pointer(dps[dp_idx], dp); + list_add_tail(&dp->list_node, &dps); dp_sysfs_add_dp(dp); rtnl_unlock(); @@ -1413,9 +1340,9 @@ static int odp_dp_cmd_new(struct sk_buff *skb, struct genl_info *info) return 0; err_destroy_local_port: - dp_detach_port(get_vport_protected(dp, ODPP_LOCAL)); + dp_detach_port(get_vport_protected(dp, OVSP_LOCAL)); 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: @@ -1426,14 +1353,14 @@ err: return err; } -static int odp_dp_cmd_del(struct sk_buff *skb, struct genl_info *info) +static int ovs_dp_cmd_del(struct sk_buff *skb, struct genl_info *info) { struct vport *vport, *next_vport; struct sk_buff *reply; struct datapath *dp; int err; - err = odp_dp_cmd_validate(info->attrs); + err = ovs_dp_cmd_validate(info->attrs); if (err) goto exit; @@ -1443,25 +1370,33 @@ static int odp_dp_cmd_del(struct sk_buff *skb, struct genl_info *info) if (IS_ERR(dp)) goto exit_unlock; - reply = odp_dp_cmd_build_info(dp, info->snd_pid, info->snd_seq, ODP_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) - if (vport->port_no != ODPP_LOCAL) + if (vport->port_no != OVSP_LOCAL) dp_detach_port(vport); dp_sysfs_del_dp(dp); - rcu_assign_pointer(dps[dp->dp_idx], NULL); - dp_detach_port(get_vport_protected(dp, ODPP_LOCAL)); + list_del(&dp->list_node); + dp_detach_port(get_vport_protected(dp, OVSP_LOCAL)); + + /* rtnl_unlock() will wait until all the references to devices that + * are pending unregistration have been dropped. We do it here to + * ensure that any internal devices (which contain DP pointers) are + * fully destroyed before freeing the datapath. + */ + rtnl_unlock(); call_rcu(&dp->rcu, destroy_dp_rcu); module_put(THIS_MODULE); genl_notify(reply, genl_info_net(info), info->snd_pid, dp_datapath_multicast_group.id, info->nlhdr, GFP_KERNEL); - err = 0; + + return 0; exit_unlock: rtnl_unlock(); @@ -1469,13 +1404,13 @@ exit: return err; } -static int odp_dp_cmd_set(struct sk_buff *skb, struct genl_info *info) +static int ovs_dp_cmd_set(struct sk_buff *skb, struct genl_info *info) { struct sk_buff *reply; struct datapath *dp; int err; - err = odp_dp_cmd_validate(info->attrs); + err = ovs_dp_cmd_validate(info->attrs); if (err) return err; @@ -1485,7 +1420,7 @@ static int odp_dp_cmd_set(struct sk_buff *skb, struct genl_info *info) change_datapath(dp, info->attrs); - reply = odp_dp_cmd_build_info(dp, info->snd_pid, info->snd_seq, ODP_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, @@ -1498,13 +1433,13 @@ static int odp_dp_cmd_set(struct sk_buff *skb, struct genl_info *info) return 0; } -static int odp_dp_cmd_get(struct sk_buff *skb, struct genl_info *info) +static int ovs_dp_cmd_get(struct sk_buff *skb, struct genl_info *info) { struct sk_buff *reply; struct datapath *dp; int err; - err = odp_dp_cmd_validate(info->attrs); + err = ovs_dp_cmd_validate(info->attrs); if (err) return err; @@ -1512,111 +1447,111 @@ static int odp_dp_cmd_get(struct sk_buff *skb, struct genl_info *info) if (IS_ERR(dp)) return PTR_ERR(dp); - reply = odp_dp_cmd_build_info(dp, info->snd_pid, info->snd_seq, ODP_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); return genlmsg_reply(reply, info); } -static int odp_dp_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb) +static int ovs_dp_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb) { - u32 dp_idx; + struct datapath *dp; + int skip = cb->args[0]; + int i = 0; - for (dp_idx = cb->args[0]; dp_idx < ARRAY_SIZE(dps); dp_idx++) { - struct datapath *dp = get_dp(dp_idx); - if (!dp) + list_for_each_entry (dp, &dps, list_node) { + if (i < skip) continue; - if (odp_dp_cmd_fill_info(dp, skb, NETLINK_CB(cb->skb).pid, + if (ovs_dp_cmd_fill_info(dp, skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq, NLM_F_MULTI, - ODP_DP_CMD_NEW) < 0) + OVS_DP_CMD_NEW) < 0) break; + i++; } - cb->args[0] = dp_idx; + cb->args[0] = i; + return skb->len; } static struct genl_ops dp_datapath_genl_ops[] = { - { .cmd = ODP_DP_CMD_NEW, + { .cmd = OVS_DP_CMD_NEW, .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */ .policy = datapath_policy, - .doit = odp_dp_cmd_new + .doit = ovs_dp_cmd_new }, - { .cmd = ODP_DP_CMD_DEL, + { .cmd = OVS_DP_CMD_DEL, .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */ .policy = datapath_policy, - .doit = odp_dp_cmd_del + .doit = ovs_dp_cmd_del }, - { .cmd = ODP_DP_CMD_GET, + { .cmd = OVS_DP_CMD_GET, .flags = 0, /* OK for unprivileged users. */ .policy = datapath_policy, - .doit = odp_dp_cmd_get, - .dumpit = odp_dp_cmd_dump + .doit = ovs_dp_cmd_get, + .dumpit = ovs_dp_cmd_dump }, - { .cmd = ODP_DP_CMD_SET, + { .cmd = OVS_DP_CMD_SET, .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */ .policy = datapath_policy, - .doit = odp_dp_cmd_set, + .doit = ovs_dp_cmd_set, }, }; -static const struct nla_policy vport_policy[ODP_VPORT_ATTR_MAX + 1] = { +static const struct nla_policy vport_policy[OVS_VPORT_ATTR_MAX + 1] = { #ifdef HAVE_NLA_NUL_STRING - [ODP_VPORT_ATTR_NAME] = { .type = NLA_NUL_STRING, .len = IFNAMSIZ - 1 }, - [ODP_VPORT_ATTR_PORT_NO] = { .type = NLA_U32 }, - [ODP_VPORT_ATTR_TYPE] = { .type = NLA_U32 }, - [ODP_VPORT_ATTR_STATS] = { .len = sizeof(struct rtnl_link_stats64) }, - [ODP_VPORT_ATTR_ADDRESS] = { .len = ETH_ALEN }, + [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 ovs_vport_stats) }, + [OVS_VPORT_ATTR_ADDRESS] = { .len = ETH_ALEN }, #else - [ODP_VPORT_ATTR_STATS] = { .minlen = sizeof(struct rtnl_link_stats64) }, - [ODP_VPORT_ATTR_ADDRESS] = { .minlen = ETH_ALEN }, + [OVS_VPORT_ATTR_STATS] = { .minlen = sizeof(struct ovs_vport_stats) }, + [OVS_VPORT_ATTR_ADDRESS] = { .minlen = ETH_ALEN }, #endif - [ODP_VPORT_ATTR_MTU] = { .type = NLA_U32 }, - [ODP_VPORT_ATTR_OPTIONS] = { .type = NLA_NESTED }, + [OVS_VPORT_ATTR_OPTIONS] = { .type = NLA_NESTED }, }; static struct genl_family dp_vport_genl_family = { .id = GENL_ID_GENERATE, - .hdrsize = sizeof(struct odp_header), - .name = ODP_VPORT_FAMILY, + .hdrsize = sizeof(struct ovs_header), + .name = OVS_VPORT_FAMILY, .version = 1, - .maxattr = ODP_VPORT_ATTR_MAX + .maxattr = OVS_VPORT_ATTR_MAX }; -static struct genl_multicast_group dp_vport_multicast_group = { - .name = ODP_VPORT_MCGROUP +struct genl_multicast_group dp_vport_multicast_group = { + .name = OVS_VPORT_MCGROUP }; /* Called with RTNL lock or RCU read lock. */ -static int odp_vport_cmd_fill_info(struct vport *vport, struct sk_buff *skb, +static int ovs_vport_cmd_fill_info(struct vport *vport, struct sk_buff *skb, u32 pid, u32 seq, u32 flags, u8 cmd) { - struct odp_header *odp_header; + struct ovs_header *ovs_header; struct nlattr *nla; - int ifindex, iflink; + int ifindex; int err; - odp_header = genlmsg_put(skb, pid, seq, &dp_vport_genl_family, + ovs_header = genlmsg_put(skb, pid, seq, &dp_vport_genl_family, flags, cmd); - if (!odp_header) + if (!ovs_header) return -EMSGSIZE; - odp_header->dp_idx = vport->dp->dp_idx; + ovs_header->dp_ifindex = vport->dp->dp_ifindex; - NLA_PUT_U32(skb, ODP_VPORT_ATTR_PORT_NO, vport->port_no); - NLA_PUT_U32(skb, ODP_VPORT_ATTR_TYPE, vport_get_type(vport)); - NLA_PUT_STRING(skb, ODP_VPORT_ATTR_NAME, vport_get_name(vport)); + 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 = nla_reserve(skb, ODP_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, ODP_VPORT_ATTR_ADDRESS, ETH_ALEN, vport_get_addr(vport)); + vport_get_stats(vport, nla_data(nla)); - NLA_PUT_U32(skb, ODP_VPORT_ATTR_MTU, vport_get_mtu(vport)); + NLA_PUT(skb, OVS_VPORT_ATTR_ADDRESS, ETH_ALEN, vport_get_addr(vport)); err = vport_get_options(vport, skb); if (err == -EMSGSIZE) @@ -1624,24 +1559,20 @@ static int odp_vport_cmd_fill_info(struct vport *vport, struct sk_buff *skb, ifindex = vport_get_ifindex(vport); if (ifindex > 0) - NLA_PUT_U32(skb, ODP_VPORT_ATTR_IFINDEX, ifindex); + NLA_PUT_U32(skb, OVS_VPORT_ATTR_IFINDEX, ifindex); - iflink = vport_get_iflink(vport); - if (iflink > 0) - NLA_PUT_U32(skb, ODP_VPORT_ATTR_IFLINK, iflink); - - return genlmsg_end(skb, odp_header); + return genlmsg_end(skb, ovs_header); nla_put_failure: err = -EMSGSIZE; error: - genlmsg_cancel(skb, odp_header); + genlmsg_cancel(skb, ovs_header); return err; } /* Called with RTNL lock or RCU read lock. */ -static struct sk_buff *odp_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; @@ -1650,7 +1581,7 @@ static struct sk_buff *odp_vport_cmd_build_info(struct vport *vport, u32 pid, if (!skb) return ERR_PTR(-ENOMEM); - retval = odp_vport_cmd_fill_info(vport, skb, pid, seq, 0, cmd); + retval = ovs_vport_cmd_fill_info(vport, skb, pid, seq, 0, cmd); if (retval < 0) { kfree_skb(skb); return ERR_PTR(retval); @@ -1658,30 +1589,30 @@ static struct sk_buff *odp_vport_cmd_build_info(struct vport *vport, u32 pid, return skb; } -static int odp_vport_cmd_validate(struct nlattr *a[ODP_VPORT_ATTR_MAX + 1]) +static int ovs_vport_cmd_validate(struct nlattr *a[OVS_VPORT_ATTR_MAX + 1]) { - return VERIFY_NUL_STRING(a[ODP_VPORT_ATTR_NAME], IFNAMSIZ - 1); + return CHECK_NUL_STRING(a[OVS_VPORT_ATTR_NAME], IFNAMSIZ - 1); } /* Called with RTNL lock or RCU read lock. */ -static struct vport *lookup_vport(struct odp_header *odp_header, - struct nlattr *a[ODP_VPORT_ATTR_MAX + 1]) +static struct vport *lookup_vport(struct ovs_header *ovs_header, + struct nlattr *a[OVS_VPORT_ATTR_MAX + 1]) { struct datapath *dp; struct vport *vport; - if (a[ODP_VPORT_ATTR_NAME]) { - vport = vport_locate(nla_data(a[ODP_VPORT_ATTR_NAME])); + if (a[OVS_VPORT_ATTR_NAME]) { + vport = vport_locate(nla_data(a[OVS_VPORT_ATTR_NAME])); if (!vport) return ERR_PTR(-ENODEV); return vport; - } else if (a[ODP_VPORT_ATTR_PORT_NO]) { - u32 port_no = nla_get_u32(a[ODP_VPORT_ATTR_PORT_NO]); + } else if (a[OVS_VPORT_ATTR_PORT_NO]) { + u32 port_no = nla_get_u32(a[OVS_VPORT_ATTR_PORT_NO]); if (port_no >= DP_MAX_PORTS) return ERR_PTR(-EFBIG); - dp = get_dp(odp_header->dp_idx); + dp = get_dp(ovs_header->dp_ifindex); if (!dp) return ERR_PTR(-ENODEV); @@ -1694,22 +1625,23 @@ static struct vport *lookup_vport(struct odp_header *odp_header, } /* Called with RTNL lock. */ -static int change_vport(struct vport *vport, struct nlattr *a[ODP_VPORT_ATTR_MAX + 1]) +static int change_vport(struct vport *vport, struct nlattr *a[OVS_VPORT_ATTR_MAX + 1]) { int err = 0; - if (a[ODP_VPORT_ATTR_STATS]) - err = vport_set_stats(vport, nla_data(a[ODP_VPORT_ATTR_STATS])); - if (!err && a[ODP_VPORT_ATTR_ADDRESS]) - err = vport_set_addr(vport, nla_data(a[ODP_VPORT_ATTR_ADDRESS])); - if (!err && a[ODP_VPORT_ATTR_MTU]) - err = vport_set_mtu(vport, nla_get_u32(a[ODP_VPORT_ATTR_MTU])); + + if (a[OVS_VPORT_ATTR_STATS]) + 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])); + return err; } -static int odp_vport_cmd_new(struct sk_buff *skb, struct genl_info *info) +static int ovs_vport_cmd_new(struct sk_buff *skb, struct genl_info *info) { struct nlattr **a = info->attrs; - struct odp_header *odp_header = info->userhdr; + struct ovs_header *ovs_header = info->userhdr; struct vport_parms parms; struct sk_buff *reply; struct vport *vport; @@ -1718,21 +1650,21 @@ static int odp_vport_cmd_new(struct sk_buff *skb, struct genl_info *info) int err; err = -EINVAL; - if (!a[ODP_VPORT_ATTR_NAME] || !a[ODP_VPORT_ATTR_TYPE]) + if (!a[OVS_VPORT_ATTR_NAME] || !a[OVS_VPORT_ATTR_TYPE]) goto exit; - err = odp_vport_cmd_validate(a); + err = ovs_vport_cmd_validate(a); if (err) goto exit; rtnl_lock(); - dp = get_dp(odp_header->dp_idx); + dp = get_dp(ovs_header->dp_ifindex); err = -ENODEV; if (!dp) goto exit_unlock; - if (a[ODP_VPORT_ATTR_PORT_NO]) { - port_no = nla_get_u32(a[ODP_VPORT_ATTR_PORT_NO]); + if (a[OVS_VPORT_ATTR_PORT_NO]) { + port_no = nla_get_u32(a[OVS_VPORT_ATTR_PORT_NO]); err = -EFBIG; if (port_no >= DP_MAX_PORTS) @@ -1754,9 +1686,9 @@ static int odp_vport_cmd_new(struct sk_buff *skb, struct genl_info *info) } } - parms.name = nla_data(a[ODP_VPORT_ATTR_NAME]); - parms.type = nla_get_u32(a[ODP_VPORT_ATTR_TYPE]); - parms.options = a[ODP_VPORT_ATTR_OPTIONS]; + parms.name = nla_data(a[OVS_VPORT_ATTR_NAME]); + parms.type = nla_get_u32(a[OVS_VPORT_ATTR_TYPE]); + parms.options = a[OVS_VPORT_ATTR_OPTIONS]; parms.dp = dp; parms.port_no = port_no; @@ -1765,13 +1697,12 @@ static int odp_vport_cmd_new(struct sk_buff *skb, struct genl_info *info) if (IS_ERR(vport)) goto exit_unlock; - set_internal_devs_mtu(dp); dp_sysfs_add_if(vport); err = change_vport(vport, a); if (!err) { - reply = odp_vport_cmd_build_info(vport, info->snd_pid, - info->snd_seq, ODP_VPORT_CMD_NEW); + reply = ovs_vport_cmd_build_info(vport, info->snd_pid, + info->snd_seq, OVS_VPORT_CMD_NEW); if (IS_ERR(reply)) err = PTR_ERR(reply); } @@ -1789,14 +1720,14 @@ exit: return err; } -static int odp_vport_cmd_set(struct sk_buff *skb, struct genl_info *info) +static int ovs_vport_cmd_set(struct sk_buff *skb, struct genl_info *info) { struct nlattr **a = info->attrs; struct sk_buff *reply; struct vport *vport; int err; - err = odp_vport_cmd_validate(a); + err = ovs_vport_cmd_validate(a); if (err) goto exit; @@ -1807,13 +1738,13 @@ static int odp_vport_cmd_set(struct sk_buff *skb, struct genl_info *info) goto exit_unlock; err = 0; - if (a[ODP_VPORT_ATTR_OPTIONS]) - err = vport_set_options(vport, a[ODP_VPORT_ATTR_OPTIONS]); + if (a[OVS_VPORT_ATTR_OPTIONS]) + err = vport_set_options(vport, a[OVS_VPORT_ATTR_OPTIONS]); if (!err) err = change_vport(vport, a); - reply = odp_vport_cmd_build_info(vport, info->snd_pid, info->snd_seq, - ODP_VPORT_CMD_NEW); + reply = ovs_vport_cmd_build_info(vport, info->snd_pid, info->snd_seq, + OVS_VPORT_CMD_NEW); if (IS_ERR(reply)) { err = PTR_ERR(reply); netlink_set_err(INIT_NET_GENL_SOCK, 0, @@ -1830,14 +1761,14 @@ exit: return err; } -static int odp_vport_cmd_del(struct sk_buff *skb, struct genl_info *info) +static int ovs_vport_cmd_del(struct sk_buff *skb, struct genl_info *info) { struct nlattr **a = info->attrs; struct sk_buff *reply; struct vport *vport; int err; - err = odp_vport_cmd_validate(a); + err = ovs_vport_cmd_validate(a); if (err) goto exit; @@ -1847,18 +1778,18 @@ static int odp_vport_cmd_del(struct sk_buff *skb, struct genl_info *info) if (IS_ERR(vport)) goto exit_unlock; - if (vport->port_no == ODPP_LOCAL) { + if (vport->port_no == OVSP_LOCAL) { err = -EINVAL; goto exit_unlock; } - reply = odp_vport_cmd_build_info(vport, info->snd_pid, info->snd_seq, - ODP_VPORT_CMD_DEL); + reply = ovs_vport_cmd_build_info(vport, info->snd_pid, info->snd_seq, + OVS_VPORT_CMD_DEL); err = PTR_ERR(reply); 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); @@ -1869,31 +1800,33 @@ exit: return err; } -static int odp_vport_cmd_get(struct sk_buff *skb, struct genl_info *info) +static int ovs_vport_cmd_get(struct sk_buff *skb, struct genl_info *info) { struct nlattr **a = info->attrs; - struct odp_header *odp_header = info->userhdr; + struct ovs_header *ovs_header = info->userhdr; struct sk_buff *reply; struct vport *vport; int err; - err = odp_vport_cmd_validate(a); + err = ovs_vport_cmd_validate(a); if (err) goto exit; rcu_read_lock(); - vport = lookup_vport(odp_header, a); + vport = lookup_vport(ovs_header, a); err = PTR_ERR(vport); if (IS_ERR(vport)) goto exit_unlock; - reply = odp_vport_cmd_build_info(vport, info->snd_pid, info->snd_seq, - ODP_VPORT_CMD_NEW); + reply = ovs_vport_cmd_build_info(vport, info->snd_pid, info->snd_seq, + OVS_VPORT_CMD_NEW); err = PTR_ERR(reply); if (IS_ERR(reply)) goto exit_unlock; - err = genlmsg_reply(reply, info); + rcu_read_unlock(); + + return genlmsg_reply(reply, info); exit_unlock: rcu_read_unlock(); @@ -1901,14 +1834,14 @@ exit: return err; } -static int odp_vport_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb) +static int ovs_vport_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb) { - struct odp_header *odp_header = genlmsg_data(nlmsg_data(cb->nlh)); + struct ovs_header *ovs_header = genlmsg_data(nlmsg_data(cb->nlh)); struct datapath *dp; u32 port_no; int retval; - dp = get_dp(odp_header->dp_idx); + dp = get_dp(ovs_header->dp_ifindex); if (!dp) return -ENODEV; @@ -1920,9 +1853,9 @@ static int odp_vport_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb) if (!vport) continue; - if (odp_vport_cmd_fill_info(vport, skb, NETLINK_CB(cb->skb).pid, + if (ovs_vport_cmd_fill_info(vport, skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq, NLM_F_MULTI, - ODP_VPORT_CMD_NEW) < 0) + OVS_VPORT_CMD_NEW) < 0) break; } rcu_read_unlock(); @@ -1934,26 +1867,26 @@ static int odp_vport_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb) } static struct genl_ops dp_vport_genl_ops[] = { - { .cmd = ODP_VPORT_CMD_NEW, + { .cmd = OVS_VPORT_CMD_NEW, .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */ .policy = vport_policy, - .doit = odp_vport_cmd_new + .doit = ovs_vport_cmd_new }, - { .cmd = ODP_VPORT_CMD_DEL, + { .cmd = OVS_VPORT_CMD_DEL, .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */ .policy = vport_policy, - .doit = odp_vport_cmd_del + .doit = ovs_vport_cmd_del }, - { .cmd = ODP_VPORT_CMD_GET, + { .cmd = OVS_VPORT_CMD_GET, .flags = 0, /* OK for unprivileged users. */ .policy = vport_policy, - .doit = odp_vport_cmd_get, - .dumpit = odp_vport_cmd_dump + .doit = ovs_vport_cmd_get, + .dumpit = ovs_vport_cmd_dump }, - { .cmd = ODP_VPORT_CMD_SET, + { .cmd = OVS_VPORT_CMD_SET, .flags = GENL_ADMIN_PERM, /* Requires CAP_NET_ADMIN privilege. */ .policy = vport_policy, - .doit = odp_vport_cmd_set, + .doit = ovs_vport_cmd_set, }, }; @@ -1983,9 +1916,8 @@ 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) @@ -2030,10 +1962,14 @@ 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; @@ -2054,6 +1990,8 @@ error_vport_exit: vport_exit(); error_flow_exit: flow_exit(); +error_tnl_exit: + tnl_exit(); error: return err; } @@ -2065,6 +2003,7 @@ static void dp_cleanup(void) unregister_netdevice_notifier(&dp_device_notifier); vport_exit(); flow_exit(); + tnl_exit(); } module_init(dp_init);