diff --git a/net/openvswitch/actions.c b/net/openvswitch/actions.c index f3f96badf5a..0da687769f5 100644 --- a/net/openvswitch/actions.c +++ b/net/openvswitch/actions.c @@ -266,7 +266,7 @@ static int do_output(struct datapath *dp, struct sk_buff *skb, int out_port) if (unlikely(!skb)) return -ENOMEM; - vport = rcu_dereference(dp->ports[out_port]); + vport = ovs_vport_rcu(dp, out_port); if (unlikely(!vport)) { kfree_skb(skb); return -ENODEV; diff --git a/net/openvswitch/datapath.c b/net/openvswitch/datapath.c index d8277d29e71..105a0b5adc5 100644 --- a/net/openvswitch/datapath.c +++ b/net/openvswitch/datapath.c @@ -49,11 +49,28 @@ #include #include #include +#include +#include #include "datapath.h" #include "flow.h" #include "vport-internal_dev.h" +/** + * struct ovs_net - Per net-namespace data for ovs. + * @dps: List of datapaths to enable dumping them all out. + * Protected by genl_mutex. + */ +struct ovs_net { + struct list_head dps; +}; + +static int ovs_net_id __read_mostly; + +#define REHASH_FLOW_INTERVAL (10 * 60 * HZ) +static void rehash_flow_table(struct work_struct *work); +static DECLARE_DELAYED_WORK(rehash_flow_wq, rehash_flow_table); + /** * DOC: Locking: * @@ -71,29 +88,21 @@ * each other. */ -/* Global list of datapaths to enable dumping them all out. - * Protected by genl_mutex. - */ -static LIST_HEAD(dps); - -#define REHASH_FLOW_INTERVAL (10 * 60 * HZ) -static void rehash_flow_table(struct work_struct *work); -static DECLARE_DELAYED_WORK(rehash_flow_wq, rehash_flow_table); - static struct vport *new_vport(const struct vport_parms *); -static int queue_gso_packets(int dp_ifindex, struct sk_buff *, +static int queue_gso_packets(struct net *, int dp_ifindex, struct sk_buff *, const struct dp_upcall_info *); -static int queue_userspace_packet(int dp_ifindex, struct sk_buff *, +static int queue_userspace_packet(struct net *, int dp_ifindex, + struct sk_buff *, const struct dp_upcall_info *); /* Must be called with rcu_read_lock, genl_mutex, or RTNL lock. */ -static struct datapath *get_dp(int dp_ifindex) +static struct datapath *get_dp(struct net *net, int dp_ifindex) { struct datapath *dp = NULL; struct net_device *dev; rcu_read_lock(); - dev = dev_get_by_index_rcu(&init_net, dp_ifindex); + dev = dev_get_by_index_rcu(net, dp_ifindex); if (dev) { struct vport *vport = ovs_internal_dev_get_vport(dev); if (vport) @@ -107,7 +116,7 @@ static struct datapath *get_dp(int dp_ifindex) /* Must be called with rcu_read_lock or RTNL lock. */ const char *ovs_dp_name(const struct datapath *dp) { - struct vport *vport = rcu_dereference_rtnl(dp->ports[OVSP_LOCAL]); + struct vport *vport = ovs_vport_rtnl_rcu(dp, OVSP_LOCAL); return vport->ops->get_name(vport); } @@ -118,7 +127,7 @@ static int get_dpifindex(struct datapath *dp) rcu_read_lock(); - local = rcu_dereference(dp->ports[OVSP_LOCAL]); + local = ovs_vport_rcu(dp, OVSP_LOCAL); if (local) ifindex = local->ops->get_ifindex(local); else @@ -135,9 +144,31 @@ static void destroy_dp_rcu(struct rcu_head *rcu) ovs_flow_tbl_destroy((__force struct flow_table *)dp->table); free_percpu(dp->stats_percpu); + release_net(ovs_dp_get_net(dp)); + kfree(dp->ports); kfree(dp); } +static struct hlist_head *vport_hash_bucket(const struct datapath *dp, + u16 port_no) +{ + return &dp->ports[port_no & (DP_VPORT_HASH_BUCKETS - 1)]; +} + +struct vport *ovs_lookup_vport(const struct datapath *dp, u16 port_no) +{ + struct vport *vport; + struct hlist_node *n; + struct hlist_head *head; + + head = vport_hash_bucket(dp, port_no); + hlist_for_each_entry_rcu(vport, n, head, dp_hash_node) { + if (vport->port_no == port_no) + return vport; + } + return NULL; +} + /* Called with RTNL lock and genl_lock. */ static struct vport *new_vport(const struct vport_parms *parms) { @@ -146,9 +177,9 @@ static struct vport *new_vport(const struct vport_parms *parms) vport = ovs_vport_add(parms); if (!IS_ERR(vport)) { struct datapath *dp = parms->dp; + struct hlist_head *head = vport_hash_bucket(dp, vport->port_no); - rcu_assign_pointer(dp->ports[parms->port_no], vport); - list_add(&vport->node, &dp->port_list); + hlist_add_head_rcu(&vport->dp_hash_node, head); } return vport; @@ -160,8 +191,7 @@ void ovs_dp_detach_port(struct vport *p) ASSERT_RTNL(); /* First drop references to device. */ - list_del(&p->node); - rcu_assign_pointer(p->dp->ports[p->port_no], NULL); + hlist_del_rcu(&p->dp_hash_node); /* Then destroy it. */ ovs_vport_del(p); @@ -220,11 +250,12 @@ static struct genl_family dp_packet_genl_family = { .hdrsize = sizeof(struct ovs_header), .name = OVS_PACKET_FAMILY, .version = OVS_PACKET_VERSION, - .maxattr = OVS_PACKET_ATTR_MAX + .maxattr = OVS_PACKET_ATTR_MAX, + .netnsok = true }; int ovs_dp_upcall(struct datapath *dp, struct sk_buff *skb, - const struct dp_upcall_info *upcall_info) + const struct dp_upcall_info *upcall_info) { struct dp_stats_percpu *stats; int dp_ifindex; @@ -242,9 +273,9 @@ int ovs_dp_upcall(struct datapath *dp, struct sk_buff *skb, } if (!skb_is_gso(skb)) - err = queue_userspace_packet(dp_ifindex, skb, upcall_info); + err = queue_userspace_packet(ovs_dp_get_net(dp), dp_ifindex, skb, upcall_info); else - err = queue_gso_packets(dp_ifindex, skb, upcall_info); + err = queue_gso_packets(ovs_dp_get_net(dp), dp_ifindex, skb, upcall_info); if (err) goto err; @@ -260,7 +291,8 @@ err: return err; } -static int queue_gso_packets(int dp_ifindex, struct sk_buff *skb, +static int queue_gso_packets(struct net *net, int dp_ifindex, + struct sk_buff *skb, const struct dp_upcall_info *upcall_info) { unsigned short gso_type = skb_shinfo(skb)->gso_type; @@ -276,7 +308,7 @@ static int queue_gso_packets(int dp_ifindex, struct sk_buff *skb, /* Queue all of the segments. */ skb = segs; do { - err = queue_userspace_packet(dp_ifindex, skb, upcall_info); + err = queue_userspace_packet(net, dp_ifindex, skb, upcall_info); if (err) break; @@ -306,7 +338,8 @@ static int queue_gso_packets(int dp_ifindex, struct sk_buff *skb, return err; } -static int queue_userspace_packet(int dp_ifindex, struct sk_buff *skb, +static int queue_userspace_packet(struct net *net, int dp_ifindex, + struct sk_buff *skb, const struct dp_upcall_info *upcall_info) { struct ovs_header *upcall; @@ -362,7 +395,7 @@ static int queue_userspace_packet(int dp_ifindex, struct sk_buff *skb, skb_copy_and_csum_dev(skb, nla_data(nla)); - err = genlmsg_unicast(&init_net, user_skb, upcall_info->pid); + err = genlmsg_unicast(net, user_skb, upcall_info->pid); out: kfree_skb(nskb); @@ -370,15 +403,10 @@ out: } /* Called with genl_mutex. */ -static int flush_flows(int dp_ifindex) +static int flush_flows(struct datapath *dp) { struct flow_table *old_table; struct flow_table *new_table; - struct datapath *dp; - - dp = get_dp(dp_ifindex); - if (!dp) - return -ENODEV; old_table = genl_dereference(dp->table); new_table = ovs_flow_tbl_alloc(TBL_MIN_BUCKETS); @@ -668,7 +696,7 @@ static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info) packet->priority = flow->key.phy.priority; rcu_read_lock(); - dp = get_dp(ovs_header->dp_ifindex); + dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex); err = -ENODEV; if (!dp) goto err_unlock; @@ -742,7 +770,8 @@ static struct genl_family dp_flow_genl_family = { .hdrsize = sizeof(struct ovs_header), .name = OVS_FLOW_FAMILY, .version = OVS_FLOW_VERSION, - .maxattr = OVS_FLOW_ATTR_MAX + .maxattr = OVS_FLOW_ATTR_MAX, + .netnsok = true }; static struct genl_multicast_group ovs_dp_flow_multicast_group = { @@ -894,7 +923,7 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info) goto error; } - dp = get_dp(ovs_header->dp_ifindex); + dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex); error = -ENODEV; if (!dp) goto error; @@ -995,7 +1024,7 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info) ovs_dp_flow_multicast_group.id, info->nlhdr, GFP_KERNEL); else - netlink_set_err(init_net.genl_sock, 0, + netlink_set_err(sock_net(skb->sk)->genl_sock, 0, ovs_dp_flow_multicast_group.id, PTR_ERR(reply)); return 0; @@ -1023,7 +1052,7 @@ static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info) if (err) return err; - dp = get_dp(ovs_header->dp_ifindex); + dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex); if (!dp) return -ENODEV; @@ -1052,16 +1081,17 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info) int err; int key_len; + dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex); + if (!dp) + return -ENODEV; + if (!a[OVS_FLOW_ATTR_KEY]) - return flush_flows(ovs_header->dp_ifindex); + return flush_flows(dp); + err = ovs_flow_from_nlattrs(&key, &key_len, a[OVS_FLOW_ATTR_KEY]); if (err) return err; - dp = get_dp(ovs_header->dp_ifindex); - if (!dp) - return -ENODEV; - table = genl_dereference(dp->table); flow = ovs_flow_tbl_lookup(table, &key, key_len); if (!flow) @@ -1090,7 +1120,7 @@ static int ovs_flow_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb) struct datapath *dp; struct flow_table *table; - dp = get_dp(ovs_header->dp_ifindex); + dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex); if (!dp) return -ENODEV; @@ -1152,7 +1182,8 @@ static struct genl_family dp_datapath_genl_family = { .hdrsize = sizeof(struct ovs_header), .name = OVS_DATAPATH_FAMILY, .version = OVS_DATAPATH_VERSION, - .maxattr = OVS_DP_ATTR_MAX + .maxattr = OVS_DP_ATTR_MAX, + .netnsok = true }; static struct genl_multicast_group ovs_dp_datapath_multicast_group = { @@ -1210,18 +1241,19 @@ static struct sk_buff *ovs_dp_cmd_build_info(struct datapath *dp, u32 pid, } /* Called with genl_mutex and optionally with RTNL lock also. */ -static struct datapath *lookup_datapath(struct ovs_header *ovs_header, +static struct datapath *lookup_datapath(struct net *net, + struct ovs_header *ovs_header, struct nlattr *a[OVS_DP_ATTR_MAX + 1]) { struct datapath *dp; if (!a[OVS_DP_ATTR_NAME]) - dp = get_dp(ovs_header->dp_ifindex); + dp = get_dp(net, ovs_header->dp_ifindex); else { struct vport *vport; rcu_read_lock(); - vport = ovs_vport_locate(nla_data(a[OVS_DP_ATTR_NAME])); + vport = ovs_vport_locate(net, nla_data(a[OVS_DP_ATTR_NAME])); dp = vport && vport->port_no == OVSP_LOCAL ? vport->dp : NULL; rcu_read_unlock(); } @@ -1235,22 +1267,21 @@ static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info) struct sk_buff *reply; struct datapath *dp; struct vport *vport; - int err; + struct ovs_net *ovs_net; + int err, i; err = -EINVAL; if (!a[OVS_DP_ATTR_NAME] || !a[OVS_DP_ATTR_UPCALL_PID]) goto err; rtnl_lock(); - err = -ENODEV; - if (!try_module_get(THIS_MODULE)) - goto err_unlock_rtnl; err = -ENOMEM; dp = kzalloc(sizeof(*dp), GFP_KERNEL); if (dp == NULL) - goto err_put_module; - INIT_LIST_HEAD(&dp->port_list); + goto err_unlock_rtnl; + + ovs_dp_set_net(dp, hold_net(sock_net(skb->sk))); /* Allocate table. */ err = -ENOMEM; @@ -1264,6 +1295,16 @@ static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info) goto err_destroy_table; } + dp->ports = kmalloc(DP_VPORT_HASH_BUCKETS * sizeof(struct hlist_head), + GFP_KERNEL); + if (!dp->ports) { + err = -ENOMEM; + goto err_destroy_percpu; + } + + for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++) + INIT_HLIST_HEAD(&dp->ports[i]); + /* Set up our datapath device. */ parms.name = nla_data(a[OVS_DP_ATTR_NAME]); parms.type = OVS_VPORT_TYPE_INTERNAL; @@ -1278,7 +1319,7 @@ static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info) if (err == -EBUSY) err = -EEXIST; - goto err_destroy_percpu; + goto err_destroy_ports_array; } reply = ovs_dp_cmd_build_info(dp, info->snd_pid, @@ -1287,7 +1328,8 @@ static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info) if (IS_ERR(reply)) goto err_destroy_local_port; - list_add_tail(&dp->list_node, &dps); + ovs_net = net_generic(ovs_dp_get_net(dp), ovs_net_id); + list_add_tail(&dp->list_node, &ovs_net->dps); rtnl_unlock(); genl_notify(reply, genl_info_net(info), info->snd_pid, @@ -1296,46 +1338,40 @@ static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info) return 0; err_destroy_local_port: - ovs_dp_detach_port(rtnl_dereference(dp->ports[OVSP_LOCAL])); + ovs_dp_detach_port(ovs_vport_rtnl(dp, OVSP_LOCAL)); +err_destroy_ports_array: + kfree(dp->ports); err_destroy_percpu: free_percpu(dp->stats_percpu); err_destroy_table: ovs_flow_tbl_destroy(genl_dereference(dp->table)); err_free_dp: + release_net(ovs_dp_get_net(dp)); kfree(dp); -err_put_module: - module_put(THIS_MODULE); err_unlock_rtnl: rtnl_unlock(); err: return err; } -static int ovs_dp_cmd_del(struct sk_buff *skb, struct genl_info *info) +/* Called with genl_mutex. */ +static void __dp_destroy(struct datapath *dp) { - struct vport *vport, *next_vport; - struct sk_buff *reply; - struct datapath *dp; - int err; + int i; rtnl_lock(); - dp = lookup_datapath(info->userhdr, info->attrs); - err = PTR_ERR(dp); - if (IS_ERR(dp)) - goto exit_unlock; - 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; + for (i = 0; i < DP_VPORT_HASH_BUCKETS; i++) { + struct vport *vport; + struct hlist_node *node, *n; - list_for_each_entry_safe(vport, next_vport, &dp->port_list, node) - if (vport->port_no != OVSP_LOCAL) - ovs_dp_detach_port(vport); + hlist_for_each_entry_safe(vport, node, n, &dp->ports[i], dp_hash_node) + if (vport->port_no != OVSP_LOCAL) + ovs_dp_detach_port(vport); + } list_del(&dp->list_node); - ovs_dp_detach_port(rtnl_dereference(dp->ports[OVSP_LOCAL])); + ovs_dp_detach_port(ovs_vport_rtnl(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 @@ -1345,17 +1381,32 @@ static int ovs_dp_cmd_del(struct sk_buff *skb, struct genl_info *info) rtnl_unlock(); call_rcu(&dp->rcu, destroy_dp_rcu); - module_put(THIS_MODULE); +} + +static int ovs_dp_cmd_del(struct sk_buff *skb, struct genl_info *info) +{ + struct sk_buff *reply; + struct datapath *dp; + int err; + + dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs); + err = PTR_ERR(dp); + if (IS_ERR(dp)) + return err; + + 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)) + return err; + + __dp_destroy(dp); genl_notify(reply, genl_info_net(info), info->snd_pid, ovs_dp_datapath_multicast_group.id, info->nlhdr, GFP_KERNEL); return 0; - -exit_unlock: - rtnl_unlock(); - return err; } static int ovs_dp_cmd_set(struct sk_buff *skb, struct genl_info *info) @@ -1364,7 +1415,7 @@ static int ovs_dp_cmd_set(struct sk_buff *skb, struct genl_info *info) struct datapath *dp; int err; - dp = lookup_datapath(info->userhdr, info->attrs); + dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs); if (IS_ERR(dp)) return PTR_ERR(dp); @@ -1372,7 +1423,7 @@ static int ovs_dp_cmd_set(struct sk_buff *skb, struct genl_info *info) info->snd_seq, OVS_DP_CMD_NEW); if (IS_ERR(reply)) { err = PTR_ERR(reply); - netlink_set_err(init_net.genl_sock, 0, + netlink_set_err(sock_net(skb->sk)->genl_sock, 0, ovs_dp_datapath_multicast_group.id, err); return 0; } @@ -1389,7 +1440,7 @@ static int ovs_dp_cmd_get(struct sk_buff *skb, struct genl_info *info) struct sk_buff *reply; struct datapath *dp; - dp = lookup_datapath(info->userhdr, info->attrs); + dp = lookup_datapath(sock_net(skb->sk), info->userhdr, info->attrs); if (IS_ERR(dp)) return PTR_ERR(dp); @@ -1403,11 +1454,12 @@ static int ovs_dp_cmd_get(struct sk_buff *skb, struct genl_info *info) static int ovs_dp_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb) { + struct ovs_net *ovs_net = net_generic(sock_net(skb->sk), ovs_net_id); struct datapath *dp; int skip = cb->args[0]; int i = 0; - list_for_each_entry(dp, &dps, list_node) { + list_for_each_entry(dp, &ovs_net->dps, list_node) { if (i >= skip && ovs_dp_cmd_fill_info(dp, skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq, NLM_F_MULTI, @@ -1459,7 +1511,8 @@ static struct genl_family dp_vport_genl_family = { .hdrsize = sizeof(struct ovs_header), .name = OVS_VPORT_FAMILY, .version = OVS_VPORT_VERSION, - .maxattr = OVS_VPORT_ATTR_MAX + .maxattr = OVS_VPORT_ATTR_MAX, + .netnsok = true }; struct genl_multicast_group ovs_dp_vport_multicast_group = { @@ -1525,14 +1578,15 @@ struct sk_buff *ovs_vport_cmd_build_info(struct vport *vport, u32 pid, } /* Called with RTNL lock or RCU read lock. */ -static struct vport *lookup_vport(struct ovs_header *ovs_header, +static struct vport *lookup_vport(struct net *net, + struct ovs_header *ovs_header, struct nlattr *a[OVS_VPORT_ATTR_MAX + 1]) { struct datapath *dp; struct vport *vport; if (a[OVS_VPORT_ATTR_NAME]) { - vport = ovs_vport_locate(nla_data(a[OVS_VPORT_ATTR_NAME])); + vport = ovs_vport_locate(net, nla_data(a[OVS_VPORT_ATTR_NAME])); if (!vport) return ERR_PTR(-ENODEV); if (ovs_header->dp_ifindex && @@ -1545,11 +1599,11 @@ static struct vport *lookup_vport(struct ovs_header *ovs_header, if (port_no >= DP_MAX_PORTS) return ERR_PTR(-EFBIG); - dp = get_dp(ovs_header->dp_ifindex); + dp = get_dp(net, ovs_header->dp_ifindex); if (!dp) return ERR_PTR(-ENODEV); - vport = rcu_dereference_rtnl(dp->ports[port_no]); + vport = ovs_vport_rtnl_rcu(dp, port_no); if (!vport) return ERR_PTR(-ENOENT); return vport; @@ -1574,7 +1628,7 @@ static int ovs_vport_cmd_new(struct sk_buff *skb, struct genl_info *info) goto exit; rtnl_lock(); - dp = get_dp(ovs_header->dp_ifindex); + dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex); err = -ENODEV; if (!dp) goto exit_unlock; @@ -1586,7 +1640,7 @@ static int ovs_vport_cmd_new(struct sk_buff *skb, struct genl_info *info) if (port_no >= DP_MAX_PORTS) goto exit_unlock; - vport = rtnl_dereference(dp->ports[port_no]); + vport = ovs_vport_rtnl_rcu(dp, port_no); err = -EBUSY; if (vport) goto exit_unlock; @@ -1596,7 +1650,7 @@ static int ovs_vport_cmd_new(struct sk_buff *skb, struct genl_info *info) err = -EFBIG; goto exit_unlock; } - vport = rtnl_dereference(dp->ports[port_no]); + vport = ovs_vport_rtnl(dp, port_no); if (!vport) break; } @@ -1638,7 +1692,7 @@ static int ovs_vport_cmd_set(struct sk_buff *skb, struct genl_info *info) int err; rtnl_lock(); - vport = lookup_vport(info->userhdr, a); + vport = lookup_vport(sock_net(skb->sk), info->userhdr, a); err = PTR_ERR(vport); if (IS_ERR(vport)) goto exit_unlock; @@ -1658,7 +1712,7 @@ static int ovs_vport_cmd_set(struct sk_buff *skb, struct genl_info *info) reply = ovs_vport_cmd_build_info(vport, info->snd_pid, info->snd_seq, OVS_VPORT_CMD_NEW); if (IS_ERR(reply)) { - netlink_set_err(init_net.genl_sock, 0, + netlink_set_err(sock_net(skb->sk)->genl_sock, 0, ovs_dp_vport_multicast_group.id, PTR_ERR(reply)); goto exit_unlock; } @@ -1679,7 +1733,7 @@ static int ovs_vport_cmd_del(struct sk_buff *skb, struct genl_info *info) int err; rtnl_lock(); - vport = lookup_vport(info->userhdr, a); + vport = lookup_vport(sock_net(skb->sk), info->userhdr, a); err = PTR_ERR(vport); if (IS_ERR(vport)) goto exit_unlock; @@ -1714,7 +1768,7 @@ static int ovs_vport_cmd_get(struct sk_buff *skb, struct genl_info *info) int err; rcu_read_lock(); - vport = lookup_vport(ovs_header, a); + vport = lookup_vport(sock_net(skb->sk), ovs_header, a); err = PTR_ERR(vport); if (IS_ERR(vport)) goto exit_unlock; @@ -1738,54 +1792,39 @@ static int ovs_vport_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb) { struct ovs_header *ovs_header = genlmsg_data(nlmsg_data(cb->nlh)); struct datapath *dp; - u32 port_no; - int retval; + int bucket = cb->args[0], skip = cb->args[1]; + int i, j = 0; - dp = get_dp(ovs_header->dp_ifindex); + dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex); if (!dp) return -ENODEV; rcu_read_lock(); - for (port_no = cb->args[0]; port_no < DP_MAX_PORTS; port_no++) { + for (i = bucket; i < DP_VPORT_HASH_BUCKETS; i++) { struct vport *vport; + struct hlist_node *n; - vport = rcu_dereference(dp->ports[port_no]); - if (!vport) - continue; + j = 0; + hlist_for_each_entry_rcu(vport, n, &dp->ports[i], dp_hash_node) { + if (j >= skip && + ovs_vport_cmd_fill_info(vport, skb, + NETLINK_CB(cb->skb).pid, + cb->nlh->nlmsg_seq, + NLM_F_MULTI, + OVS_VPORT_CMD_NEW) < 0) + goto out; - if (ovs_vport_cmd_fill_info(vport, skb, NETLINK_CB(cb->skb).pid, - cb->nlh->nlmsg_seq, NLM_F_MULTI, - OVS_VPORT_CMD_NEW) < 0) - break; + j++; + } + skip = 0; } +out: rcu_read_unlock(); - cb->args[0] = port_no; - retval = skb->len; + cb->args[0] = i; + cb->args[1] = j; - return retval; -} - -static void rehash_flow_table(struct work_struct *work) -{ - struct datapath *dp; - - genl_lock(); - - list_for_each_entry(dp, &dps, list_node) { - struct flow_table *old_table = genl_dereference(dp->table); - struct flow_table *new_table; - - new_table = ovs_flow_tbl_rehash(old_table); - if (!IS_ERR(new_table)) { - rcu_assign_pointer(dp->table, new_table); - ovs_flow_tbl_deferred_destroy(old_table); - } - } - - genl_unlock(); - - schedule_delayed_work(&rehash_flow_wq, REHASH_FLOW_INTERVAL); + return skb->len; } static struct genl_ops dp_vport_genl_ops[] = { @@ -1872,6 +1911,59 @@ error: return err; } +static void rehash_flow_table(struct work_struct *work) +{ + struct datapath *dp; + struct net *net; + + genl_lock(); + rtnl_lock(); + for_each_net(net) { + struct ovs_net *ovs_net = net_generic(net, ovs_net_id); + + list_for_each_entry(dp, &ovs_net->dps, list_node) { + struct flow_table *old_table = genl_dereference(dp->table); + struct flow_table *new_table; + + new_table = ovs_flow_tbl_rehash(old_table); + if (!IS_ERR(new_table)) { + rcu_assign_pointer(dp->table, new_table); + ovs_flow_tbl_deferred_destroy(old_table); + } + } + } + rtnl_unlock(); + genl_unlock(); + + schedule_delayed_work(&rehash_flow_wq, REHASH_FLOW_INTERVAL); +} + +static int __net_init ovs_init_net(struct net *net) +{ + struct ovs_net *ovs_net = net_generic(net, ovs_net_id); + + INIT_LIST_HEAD(&ovs_net->dps); + return 0; +} + +static void __net_exit ovs_exit_net(struct net *net) +{ + struct ovs_net *ovs_net = net_generic(net, ovs_net_id); + struct datapath *dp, *dp_next; + + genl_lock(); + list_for_each_entry_safe(dp, dp_next, &ovs_net->dps, list_node) + __dp_destroy(dp); + genl_unlock(); +} + +static struct pernet_operations ovs_net_ops = { + .init = ovs_init_net, + .exit = ovs_exit_net, + .id = &ovs_net_id, + .size = sizeof(struct ovs_net), +}; + static int __init dp_init(void) { struct sk_buff *dummy_skb; @@ -1889,10 +1981,14 @@ static int __init dp_init(void) if (err) goto error_flow_exit; - err = register_netdevice_notifier(&ovs_dp_device_notifier); + err = register_pernet_device(&ovs_net_ops); if (err) goto error_vport_exit; + err = register_netdevice_notifier(&ovs_dp_device_notifier); + if (err) + goto error_netns_exit; + err = dp_register_genl(); if (err < 0) goto error_unreg_notifier; @@ -1903,6 +1999,8 @@ static int __init dp_init(void) error_unreg_notifier: unregister_netdevice_notifier(&ovs_dp_device_notifier); +error_netns_exit: + unregister_pernet_device(&ovs_net_ops); error_vport_exit: ovs_vport_exit(); error_flow_exit: @@ -1914,9 +2012,10 @@ error: static void dp_cleanup(void) { cancel_delayed_work_sync(&rehash_flow_wq); - rcu_barrier(); dp_unregister_genl(ARRAY_SIZE(dp_genl_families)); unregister_netdevice_notifier(&ovs_dp_device_notifier); + unregister_pernet_device(&ovs_net_ops); + rcu_barrier(); ovs_vport_exit(); ovs_flow_exit(); } diff --git a/net/openvswitch/datapath.h b/net/openvswitch/datapath.h index c1105c14753..129ec548075 100644 --- a/net/openvswitch/datapath.h +++ b/net/openvswitch/datapath.h @@ -27,10 +27,11 @@ #include #include "flow.h" +#include "vport.h" -struct vport; +#define DP_MAX_PORTS USHRT_MAX +#define DP_VPORT_HASH_BUCKETS 1024 -#define DP_MAX_PORTS 1024 #define SAMPLE_ACTION_DEPTH 3 /** @@ -58,11 +59,10 @@ struct dp_stats_percpu { * @list_node: Element in global 'dps' list. * @n_flows: Number of flows currently in flow table. * @table: Current flow table. Protected by genl_lock and RCU. - * @ports: Map from port number to &struct vport. %OVSP_LOCAL port - * always exists, other ports may be %NULL. Protected by RTNL and RCU. - * @port_list: List of all ports in @ports in arbitrary order. RTNL required - * to iterate or modify. + * @ports: Hash table for ports. %OVSP_LOCAL port always exists. Protected by + * RTNL and RCU. * @stats_percpu: Per-CPU datapath statistics. + * @net: Reference to net namespace. * * Context: See the comment on locking at the top of datapath.c for additional * locking information. @@ -75,13 +75,37 @@ struct datapath { struct flow_table __rcu *table; /* Switch ports. */ - struct vport __rcu *ports[DP_MAX_PORTS]; - struct list_head port_list; + struct hlist_head *ports; /* Stats. */ struct dp_stats_percpu __percpu *stats_percpu; + +#ifdef CONFIG_NET_NS + /* Network namespace ref. */ + struct net *net; +#endif }; +struct vport *ovs_lookup_vport(const struct datapath *dp, u16 port_no); + +static inline struct vport *ovs_vport_rcu(const struct datapath *dp, int port_no) +{ + WARN_ON_ONCE(!rcu_read_lock_held()); + return ovs_lookup_vport(dp, port_no); +} + +static inline struct vport *ovs_vport_rtnl_rcu(const struct datapath *dp, int port_no) +{ + WARN_ON_ONCE(!rcu_read_lock_held() && !rtnl_is_locked()); + return ovs_lookup_vport(dp, port_no); +} + +static inline struct vport *ovs_vport_rtnl(const struct datapath *dp, int port_no) +{ + ASSERT_RTNL(); + return ovs_lookup_vport(dp, port_no); +} + /** * struct ovs_skb_cb - OVS data in skb CB * @flow: The flow associated with this packet. May be %NULL if no flow. @@ -108,6 +132,16 @@ struct dp_upcall_info { u32 pid; }; +static inline struct net *ovs_dp_get_net(struct datapath *dp) +{ + return read_pnet(&dp->net); +} + +static inline void ovs_dp_set_net(struct datapath *dp, struct net *net) +{ + write_pnet(&dp->net, net); +} + extern struct notifier_block ovs_dp_device_notifier; extern struct genl_multicast_group ovs_dp_vport_multicast_group; diff --git a/net/openvswitch/dp_notify.c b/net/openvswitch/dp_notify.c index 36dcee8fc84..5558350e0d3 100644 --- a/net/openvswitch/dp_notify.c +++ b/net/openvswitch/dp_notify.c @@ -41,19 +41,21 @@ static int dp_device_event(struct notifier_block *unused, unsigned long event, case NETDEV_UNREGISTER: if (!ovs_is_internal_dev(dev)) { struct sk_buff *notify; + struct datapath *dp = vport->dp; notify = ovs_vport_cmd_build_info(vport, 0, 0, OVS_VPORT_CMD_DEL); ovs_dp_detach_port(vport); if (IS_ERR(notify)) { - netlink_set_err(init_net.genl_sock, 0, + netlink_set_err(ovs_dp_get_net(dp)->genl_sock, 0, ovs_dp_vport_multicast_group.id, PTR_ERR(notify)); break; } - genlmsg_multicast(notify, 0, ovs_dp_vport_multicast_group.id, - GFP_KERNEL); + genlmsg_multicast_netns(ovs_dp_get_net(dp), notify, 0, + ovs_dp_vport_multicast_group.id, + GFP_KERNEL); } break; } diff --git a/net/openvswitch/flow.c b/net/openvswitch/flow.c index c7bf2f26525..98c70630ad0 100644 --- a/net/openvswitch/flow.c +++ b/net/openvswitch/flow.c @@ -203,10 +203,7 @@ struct sw_flow_actions *ovs_flow_actions_alloc(const struct nlattr *actions) int actions_len = nla_len(actions); struct sw_flow_actions *sfa; - /* At least DP_MAX_PORTS actions are required to be able to flood a - * packet to every port. Factor of 2 allows for setting VLAN tags, - * etc. */ - if (actions_len > 2 * DP_MAX_PORTS * nla_total_size(4)) + if (actions_len > MAX_ACTIONS_BUFSIZE) return ERR_PTR(-EINVAL); sfa = kmalloc(sizeof(*sfa) + actions_len, GFP_KERNEL); @@ -992,7 +989,7 @@ int ovs_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_lenp, swkey->phy.in_port = in_port; attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT); } else { - swkey->phy.in_port = USHRT_MAX; + swkey->phy.in_port = DP_MAX_PORTS; } /* Data attributes. */ @@ -1135,7 +1132,7 @@ int ovs_flow_metadata_from_nlattrs(u32 *priority, u16 *in_port, const struct nlattr *nla; int rem; - *in_port = USHRT_MAX; + *in_port = DP_MAX_PORTS; *priority = 0; nla_for_each_nested(nla, attr, rem) { @@ -1172,7 +1169,7 @@ int ovs_flow_to_nlattrs(const struct sw_flow_key *swkey, struct sk_buff *skb) nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, swkey->phy.priority)) goto nla_put_failure; - if (swkey->phy.in_port != USHRT_MAX && + if (swkey->phy.in_port != DP_MAX_PORTS && nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, swkey->phy.in_port)) goto nla_put_failure; diff --git a/net/openvswitch/flow.h b/net/openvswitch/flow.h index 9b75617ca4e..d92e22a638c 100644 --- a/net/openvswitch/flow.h +++ b/net/openvswitch/flow.h @@ -43,7 +43,7 @@ struct sw_flow_actions { struct sw_flow_key { struct { u32 priority; /* Packet QoS priority. */ - u16 in_port; /* Input switch port (or USHRT_MAX). */ + u16 in_port; /* Input switch port (or DP_MAX_PORTS). */ } phy; struct { u8 src[ETH_ALEN]; /* Ethernet source address. */ @@ -161,6 +161,7 @@ int ovs_flow_from_nlattrs(struct sw_flow_key *swkey, int *key_lenp, int ovs_flow_metadata_from_nlattrs(u32 *priority, u16 *in_port, const struct nlattr *); +#define MAX_ACTIONS_BUFSIZE (16 * 1024) #define TBL_MIN_BUCKETS 1024 struct flow_table { diff --git a/net/openvswitch/vport-internal_dev.c b/net/openvswitch/vport-internal_dev.c index 4061b9ee07f..5d460c37df0 100644 --- a/net/openvswitch/vport-internal_dev.c +++ b/net/openvswitch/vport-internal_dev.c @@ -144,7 +144,7 @@ static void do_setup(struct net_device *netdev) netdev->tx_queue_len = 0; netdev->features = NETIF_F_LLTX | NETIF_F_SG | NETIF_F_FRAGLIST | - NETIF_F_HIGHDMA | NETIF_F_HW_CSUM | NETIF_F_TSO; + NETIF_F_HIGHDMA | NETIF_F_HW_CSUM | NETIF_F_TSO; netdev->vlan_features = netdev->features; netdev->features |= NETIF_F_HW_VLAN_TX; @@ -175,9 +175,14 @@ static struct vport *internal_dev_create(const struct vport_parms *parms) goto error_free_vport; } + dev_net_set(netdev_vport->dev, ovs_dp_get_net(vport->dp)); internal_dev = internal_dev_priv(netdev_vport->dev); internal_dev->vport = vport; + /* Restrict bridge port to current netns. */ + if (vport->port_no == OVSP_LOCAL) + netdev_vport->dev->features |= NETIF_F_NETNS_LOCAL; + err = register_netdevice(netdev_vport->dev); if (err) goto error_free_netdev; diff --git a/net/openvswitch/vport-netdev.c b/net/openvswitch/vport-netdev.c index 6ea3551cc78..3c1e58ba714 100644 --- a/net/openvswitch/vport-netdev.c +++ b/net/openvswitch/vport-netdev.c @@ -83,7 +83,7 @@ static struct vport *netdev_create(const struct vport_parms *parms) netdev_vport = netdev_vport_priv(vport); - netdev_vport->dev = dev_get_by_name(&init_net, parms->name); + netdev_vport->dev = dev_get_by_name(ovs_dp_get_net(vport->dp), parms->name); if (!netdev_vport->dev) { err = -ENODEV; goto error_free_vport; diff --git a/net/openvswitch/vport.c b/net/openvswitch/vport.c index 6140336e79d..1abd9609ba7 100644 --- a/net/openvswitch/vport.c +++ b/net/openvswitch/vport.c @@ -16,10 +16,10 @@ * 02110-1301, USA */ -#include #include #include #include +#include #include #include #include @@ -27,7 +27,9 @@ #include #include #include +#include +#include "datapath.h" #include "vport.h" #include "vport-internal_dev.h" @@ -67,9 +69,9 @@ void ovs_vport_exit(void) kfree(dev_table); } -static struct hlist_head *hash_bucket(const char *name) +static struct hlist_head *hash_bucket(struct net *net, const char *name) { - unsigned int hash = full_name_hash(name, strlen(name)); + unsigned int hash = jhash(name, strlen(name), (unsigned long) net); return &dev_table[hash & (VPORT_HASH_BUCKETS - 1)]; } @@ -80,14 +82,15 @@ static struct hlist_head *hash_bucket(const char *name) * * Must be called with RTNL or RCU read lock. */ -struct vport *ovs_vport_locate(const char *name) +struct vport *ovs_vport_locate(struct net *net, const char *name) { - struct hlist_head *bucket = hash_bucket(name); + struct hlist_head *bucket = hash_bucket(net, name); struct vport *vport; struct hlist_node *node; hlist_for_each_entry_rcu(vport, node, bucket, hash_node) - if (!strcmp(name, vport->ops->get_name(vport))) + if (!strcmp(name, vport->ops->get_name(vport)) && + net_eq(ovs_dp_get_net(vport->dp), net)) return vport; return NULL; @@ -124,6 +127,7 @@ struct vport *ovs_vport_alloc(int priv_size, const struct vport_ops *ops, vport->port_no = parms->port_no; vport->upcall_pid = parms->upcall_pid; vport->ops = ops; + INIT_HLIST_NODE(&vport->dp_hash_node); vport->percpu_stats = alloc_percpu(struct vport_percpu_stats); if (!vport->percpu_stats) { @@ -170,14 +174,17 @@ struct vport *ovs_vport_add(const struct vport_parms *parms) for (i = 0; i < ARRAY_SIZE(vport_ops_list); i++) { if (vport_ops_list[i]->type == parms->type) { + struct hlist_head *bucket; + vport = vport_ops_list[i]->create(parms); if (IS_ERR(vport)) { err = PTR_ERR(vport); goto out; } - hlist_add_head_rcu(&vport->hash_node, - hash_bucket(vport->ops->get_name(vport))); + bucket = hash_bucket(ovs_dp_get_net(vport->dp), + vport->ops->get_name(vport)); + hlist_add_head_rcu(&vport->hash_node, bucket); return vport; } } diff --git a/net/openvswitch/vport.h b/net/openvswitch/vport.h index aac680ca2b0..c56e4836e93 100644 --- a/net/openvswitch/vport.h +++ b/net/openvswitch/vport.h @@ -20,6 +20,7 @@ #define VPORT_H 1 #include +#include #include #include #include @@ -38,7 +39,7 @@ void ovs_vport_exit(void); struct vport *ovs_vport_add(const struct vport_parms *); void ovs_vport_del(struct vport *); -struct vport *ovs_vport_locate(const char *name); +struct vport *ovs_vport_locate(struct net *net, const char *name); void ovs_vport_get_stats(struct vport *, struct ovs_vport_stats *); @@ -69,10 +70,10 @@ struct vport_err_stats { * @rcu: RCU callback head for deferred destruction. * @port_no: Index into @dp's @ports array. * @dp: Datapath to which this port belongs. - * @node: Element in @dp's @port_list. * @upcall_pid: The Netlink port to use for packets received on this port that * miss the flow table. * @hash_node: Element in @dev_table hash table in vport.c. + * @dp_hash_node: Element in @datapath->ports hash table in datapath.c. * @ops: Class structure. * @percpu_stats: Points to per-CPU statistics used and maintained by vport * @stats_lock: Protects @err_stats; @@ -82,10 +83,10 @@ struct vport { struct rcu_head rcu; u16 port_no; struct datapath *dp; - struct list_head node; u32 upcall_pid; struct hlist_node hash_node; + struct hlist_node dp_hash_node; const struct vport_ops *ops; struct vport_percpu_stats __percpu *percpu_stats;