aboutsummaryrefslogtreecommitdiffstats
path: root/net/openvswitch
diff options
context:
space:
mode:
authorPravin B Shelar <pshelar@nicira.com>2013-10-03 21:16:47 -0400
committerJesse Gross <jesse@nicira.com>2013-10-03 21:16:47 -0400
commite64457191a259537bbbfaebeba9a8043786af96f (patch)
tree579f47d3124b69a94b07878d65d27f71c4f7c9c5 /net/openvswitch
parentf0627cfa24389cab25c67bb7ca902912216a8a2d (diff)
openvswitch: Restructure datapath.c and flow.c
Over the time datapath.c and flow.c has became pretty large files. Following patch restructures functionality of component into three different components: flow.c: contains flow extract. flow_netlink.c: netlink flow api. flow_table.c: flow table api. This patch restructures code without changing logic. Signed-off-by: Pravin B Shelar <pshelar@nicira.com> Signed-off-by: Jesse Gross <jesse@nicira.com>
Diffstat (limited to 'net/openvswitch')
-rw-r--r--net/openvswitch/Makefile2
-rw-r--r--net/openvswitch/datapath.c528
-rw-r--r--net/openvswitch/datapath.h1
-rw-r--r--net/openvswitch/flow.c1605
-rw-r--r--net/openvswitch/flow.h128
-rw-r--r--net/openvswitch/flow_netlink.c1603
-rw-r--r--net/openvswitch/flow_netlink.h60
-rw-r--r--net/openvswitch/flow_table.c517
-rw-r--r--net/openvswitch/flow_table.h91
9 files changed, 2354 insertions, 2181 deletions
diff --git a/net/openvswitch/Makefile b/net/openvswitch/Makefile
index ea36e99089af..3591cb5dae91 100644
--- a/net/openvswitch/Makefile
+++ b/net/openvswitch/Makefile
@@ -9,6 +9,8 @@ openvswitch-y := \
9 datapath.o \ 9 datapath.o \
10 dp_notify.o \ 10 dp_notify.o \
11 flow.o \ 11 flow.o \
12 flow_netlink.o \
13 flow_table.o \
12 vport.o \ 14 vport.o \
13 vport-internal_dev.o \ 15 vport-internal_dev.o \
14 vport-netdev.o 16 vport-netdev.o
diff --git a/net/openvswitch/datapath.c b/net/openvswitch/datapath.c
index 2e1a9c24e380..72e68743c643 100644
--- a/net/openvswitch/datapath.c
+++ b/net/openvswitch/datapath.c
@@ -55,10 +55,10 @@
55 55
56#include "datapath.h" 56#include "datapath.h"
57#include "flow.h" 57#include "flow.h"
58#include "flow_netlink.h"
58#include "vport-internal_dev.h" 59#include "vport-internal_dev.h"
59#include "vport-netdev.h" 60#include "vport-netdev.h"
60 61
61
62#define REHASH_FLOW_INTERVAL (10 * 60 * HZ) 62#define REHASH_FLOW_INTERVAL (10 * 60 * HZ)
63 63
64int ovs_net_id __read_mostly; 64int ovs_net_id __read_mostly;
@@ -235,7 +235,7 @@ void ovs_dp_process_received_packet(struct vport *p, struct sk_buff *skb)
235 } 235 }
236 236
237 /* Look up flow. */ 237 /* Look up flow. */
238 flow = ovs_flow_lookup(rcu_dereference(dp->table), &key); 238 flow = ovs_flow_tbl_lookup(rcu_dereference(dp->table), &key);
239 if (unlikely(!flow)) { 239 if (unlikely(!flow)) {
240 struct dp_upcall_info upcall; 240 struct dp_upcall_info upcall;
241 241
@@ -433,7 +433,7 @@ static int queue_userspace_packet(struct net *net, int dp_ifindex,
433 upcall->dp_ifindex = dp_ifindex; 433 upcall->dp_ifindex = dp_ifindex;
434 434
435 nla = nla_nest_start(user_skb, OVS_PACKET_ATTR_KEY); 435 nla = nla_nest_start(user_skb, OVS_PACKET_ATTR_KEY);
436 ovs_flow_to_nlattrs(upcall_info->key, upcall_info->key, user_skb); 436 ovs_nla_put_flow(upcall_info->key, upcall_info->key, user_skb);
437 nla_nest_end(user_skb, nla); 437 nla_nest_end(user_skb, nla);
438 438
439 if (upcall_info->userdata) 439 if (upcall_info->userdata)
@@ -470,381 +470,6 @@ static int flush_flows(struct datapath *dp)
470 return 0; 470 return 0;
471} 471}
472 472
473static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa, int attr_len)
474{
475
476 struct sw_flow_actions *acts;
477 int new_acts_size;
478 int req_size = NLA_ALIGN(attr_len);
479 int next_offset = offsetof(struct sw_flow_actions, actions) +
480 (*sfa)->actions_len;
481
482 if (req_size <= (ksize(*sfa) - next_offset))
483 goto out;
484
485 new_acts_size = ksize(*sfa) * 2;
486
487 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
488 if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size)
489 return ERR_PTR(-EMSGSIZE);
490 new_acts_size = MAX_ACTIONS_BUFSIZE;
491 }
492
493 acts = ovs_flow_actions_alloc(new_acts_size);
494 if (IS_ERR(acts))
495 return (void *)acts;
496
497 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
498 acts->actions_len = (*sfa)->actions_len;
499 kfree(*sfa);
500 *sfa = acts;
501
502out:
503 (*sfa)->actions_len += req_size;
504 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
505}
506
507static int add_action(struct sw_flow_actions **sfa, int attrtype, void *data, int len)
508{
509 struct nlattr *a;
510
511 a = reserve_sfa_size(sfa, nla_attr_size(len));
512 if (IS_ERR(a))
513 return PTR_ERR(a);
514
515 a->nla_type = attrtype;
516 a->nla_len = nla_attr_size(len);
517
518 if (data)
519 memcpy(nla_data(a), data, len);
520 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
521
522 return 0;
523}
524
525static inline int add_nested_action_start(struct sw_flow_actions **sfa, int attrtype)
526{
527 int used = (*sfa)->actions_len;
528 int err;
529
530 err = add_action(sfa, attrtype, NULL, 0);
531 if (err)
532 return err;
533
534 return used;
535}
536
537static inline void add_nested_action_end(struct sw_flow_actions *sfa, int st_offset)
538{
539 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions + st_offset);
540
541 a->nla_len = sfa->actions_len - st_offset;
542}
543
544static int validate_and_copy_actions(const struct nlattr *attr,
545 const struct sw_flow_key *key, int depth,
546 struct sw_flow_actions **sfa);
547
548static int validate_and_copy_sample(const struct nlattr *attr,
549 const struct sw_flow_key *key, int depth,
550 struct sw_flow_actions **sfa)
551{
552 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
553 const struct nlattr *probability, *actions;
554 const struct nlattr *a;
555 int rem, start, err, st_acts;
556
557 memset(attrs, 0, sizeof(attrs));
558 nla_for_each_nested(a, attr, rem) {
559 int type = nla_type(a);
560 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
561 return -EINVAL;
562 attrs[type] = a;
563 }
564 if (rem)
565 return -EINVAL;
566
567 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
568 if (!probability || nla_len(probability) != sizeof(u32))
569 return -EINVAL;
570
571 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
572 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
573 return -EINVAL;
574
575 /* validation done, copy sample action. */
576 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE);
577 if (start < 0)
578 return start;
579 err = add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY, nla_data(probability), sizeof(u32));
580 if (err)
581 return err;
582 st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS);
583 if (st_acts < 0)
584 return st_acts;
585
586 err = validate_and_copy_actions(actions, key, depth + 1, sfa);
587 if (err)
588 return err;
589
590 add_nested_action_end(*sfa, st_acts);
591 add_nested_action_end(*sfa, start);
592
593 return 0;
594}
595
596static int validate_tp_port(const struct sw_flow_key *flow_key)
597{
598 if (flow_key->eth.type == htons(ETH_P_IP)) {
599 if (flow_key->ipv4.tp.src || flow_key->ipv4.tp.dst)
600 return 0;
601 } else if (flow_key->eth.type == htons(ETH_P_IPV6)) {
602 if (flow_key->ipv6.tp.src || flow_key->ipv6.tp.dst)
603 return 0;
604 }
605
606 return -EINVAL;
607}
608
609static int validate_and_copy_set_tun(const struct nlattr *attr,
610 struct sw_flow_actions **sfa)
611{
612 struct sw_flow_match match;
613 struct sw_flow_key key;
614 int err, start;
615
616 ovs_match_init(&match, &key, NULL);
617 err = ovs_ipv4_tun_from_nlattr(nla_data(attr), &match, false);
618 if (err)
619 return err;
620
621 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET);
622 if (start < 0)
623 return start;
624
625 err = add_action(sfa, OVS_KEY_ATTR_IPV4_TUNNEL, &match.key->tun_key,
626 sizeof(match.key->tun_key));
627 add_nested_action_end(*sfa, start);
628
629 return err;
630}
631
632static int validate_set(const struct nlattr *a,
633 const struct sw_flow_key *flow_key,
634 struct sw_flow_actions **sfa,
635 bool *set_tun)
636{
637 const struct nlattr *ovs_key = nla_data(a);
638 int key_type = nla_type(ovs_key);
639
640 /* There can be only one key in a action */
641 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
642 return -EINVAL;
643
644 if (key_type > OVS_KEY_ATTR_MAX ||
645 (ovs_key_lens[key_type] != nla_len(ovs_key) &&
646 ovs_key_lens[key_type] != -1))
647 return -EINVAL;
648
649 switch (key_type) {
650 const struct ovs_key_ipv4 *ipv4_key;
651 const struct ovs_key_ipv6 *ipv6_key;
652 int err;
653
654 case OVS_KEY_ATTR_PRIORITY:
655 case OVS_KEY_ATTR_SKB_MARK:
656 case OVS_KEY_ATTR_ETHERNET:
657 break;
658
659 case OVS_KEY_ATTR_TUNNEL:
660 *set_tun = true;
661 err = validate_and_copy_set_tun(a, sfa);
662 if (err)
663 return err;
664 break;
665
666 case OVS_KEY_ATTR_IPV4:
667 if (flow_key->eth.type != htons(ETH_P_IP))
668 return -EINVAL;
669
670 if (!flow_key->ip.proto)
671 return -EINVAL;
672
673 ipv4_key = nla_data(ovs_key);
674 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
675 return -EINVAL;
676
677 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
678 return -EINVAL;
679
680 break;
681
682 case OVS_KEY_ATTR_IPV6:
683 if (flow_key->eth.type != htons(ETH_P_IPV6))
684 return -EINVAL;
685
686 if (!flow_key->ip.proto)
687 return -EINVAL;
688
689 ipv6_key = nla_data(ovs_key);
690 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
691 return -EINVAL;
692
693 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
694 return -EINVAL;
695
696 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
697 return -EINVAL;
698
699 break;
700
701 case OVS_KEY_ATTR_TCP:
702 if (flow_key->ip.proto != IPPROTO_TCP)
703 return -EINVAL;
704
705 return validate_tp_port(flow_key);
706
707 case OVS_KEY_ATTR_UDP:
708 if (flow_key->ip.proto != IPPROTO_UDP)
709 return -EINVAL;
710
711 return validate_tp_port(flow_key);
712
713 case OVS_KEY_ATTR_SCTP:
714 if (flow_key->ip.proto != IPPROTO_SCTP)
715 return -EINVAL;
716
717 return validate_tp_port(flow_key);
718
719 default:
720 return -EINVAL;
721 }
722
723 return 0;
724}
725
726static int validate_userspace(const struct nlattr *attr)
727{
728 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
729 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
730 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
731 };
732 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
733 int error;
734
735 error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
736 attr, userspace_policy);
737 if (error)
738 return error;
739
740 if (!a[OVS_USERSPACE_ATTR_PID] ||
741 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
742 return -EINVAL;
743
744 return 0;
745}
746
747static int copy_action(const struct nlattr *from,
748 struct sw_flow_actions **sfa)
749{
750 int totlen = NLA_ALIGN(from->nla_len);
751 struct nlattr *to;
752
753 to = reserve_sfa_size(sfa, from->nla_len);
754 if (IS_ERR(to))
755 return PTR_ERR(to);
756
757 memcpy(to, from, totlen);
758 return 0;
759}
760
761static int validate_and_copy_actions(const struct nlattr *attr,
762 const struct sw_flow_key *key,
763 int depth,
764 struct sw_flow_actions **sfa)
765{
766 const struct nlattr *a;
767 int rem, err;
768
769 if (depth >= SAMPLE_ACTION_DEPTH)
770 return -EOVERFLOW;
771
772 nla_for_each_nested(a, attr, rem) {
773 /* Expected argument lengths, (u32)-1 for variable length. */
774 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
775 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
776 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
777 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
778 [OVS_ACTION_ATTR_POP_VLAN] = 0,
779 [OVS_ACTION_ATTR_SET] = (u32)-1,
780 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1
781 };
782 const struct ovs_action_push_vlan *vlan;
783 int type = nla_type(a);
784 bool skip_copy;
785
786 if (type > OVS_ACTION_ATTR_MAX ||
787 (action_lens[type] != nla_len(a) &&
788 action_lens[type] != (u32)-1))
789 return -EINVAL;
790
791 skip_copy = false;
792 switch (type) {
793 case OVS_ACTION_ATTR_UNSPEC:
794 return -EINVAL;
795
796 case OVS_ACTION_ATTR_USERSPACE:
797 err = validate_userspace(a);
798 if (err)
799 return err;
800 break;
801
802 case OVS_ACTION_ATTR_OUTPUT:
803 if (nla_get_u32(a) >= DP_MAX_PORTS)
804 return -EINVAL;
805 break;
806
807
808 case OVS_ACTION_ATTR_POP_VLAN:
809 break;
810
811 case OVS_ACTION_ATTR_PUSH_VLAN:
812 vlan = nla_data(a);
813 if (vlan->vlan_tpid != htons(ETH_P_8021Q))
814 return -EINVAL;
815 if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
816 return -EINVAL;
817 break;
818
819 case OVS_ACTION_ATTR_SET:
820 err = validate_set(a, key, sfa, &skip_copy);
821 if (err)
822 return err;
823 break;
824
825 case OVS_ACTION_ATTR_SAMPLE:
826 err = validate_and_copy_sample(a, key, depth, sfa);
827 if (err)
828 return err;
829 skip_copy = true;
830 break;
831
832 default:
833 return -EINVAL;
834 }
835 if (!skip_copy) {
836 err = copy_action(a, sfa);
837 if (err)
838 return err;
839 }
840 }
841
842 if (rem > 0)
843 return -EINVAL;
844
845 return 0;
846}
847
848static void clear_stats(struct sw_flow *flow) 473static void clear_stats(struct sw_flow *flow)
849{ 474{
850 flow->used = 0; 475 flow->used = 0;
@@ -900,15 +525,16 @@ static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info)
900 if (err) 525 if (err)
901 goto err_flow_free; 526 goto err_flow_free;
902 527
903 err = ovs_flow_metadata_from_nlattrs(flow, a[OVS_PACKET_ATTR_KEY]); 528 err = ovs_nla_get_flow_metadata(flow, a[OVS_PACKET_ATTR_KEY]);
904 if (err) 529 if (err)
905 goto err_flow_free; 530 goto err_flow_free;
906 acts = ovs_flow_actions_alloc(nla_len(a[OVS_PACKET_ATTR_ACTIONS])); 531 acts = ovs_nla_alloc_flow_actions(nla_len(a[OVS_PACKET_ATTR_ACTIONS]));
907 err = PTR_ERR(acts); 532 err = PTR_ERR(acts);
908 if (IS_ERR(acts)) 533 if (IS_ERR(acts))
909 goto err_flow_free; 534 goto err_flow_free;
910 535
911 err = validate_and_copy_actions(a[OVS_PACKET_ATTR_ACTIONS], &flow->key, 0, &acts); 536 err = ovs_nla_copy_actions(a[OVS_PACKET_ATTR_ACTIONS],
537 &flow->key, 0, &acts);
912 rcu_assign_pointer(flow->sf_acts, acts); 538 rcu_assign_pointer(flow->sf_acts, acts);
913 if (err) 539 if (err)
914 goto err_flow_free; 540 goto err_flow_free;
@@ -1003,100 +629,6 @@ static struct genl_multicast_group ovs_dp_flow_multicast_group = {
1003 .name = OVS_FLOW_MCGROUP 629 .name = OVS_FLOW_MCGROUP
1004}; 630};
1005 631
1006static int actions_to_attr(const struct nlattr *attr, int len, struct sk_buff *skb);
1007static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
1008{
1009 const struct nlattr *a;
1010 struct nlattr *start;
1011 int err = 0, rem;
1012
1013 start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
1014 if (!start)
1015 return -EMSGSIZE;
1016
1017 nla_for_each_nested(a, attr, rem) {
1018 int type = nla_type(a);
1019 struct nlattr *st_sample;
1020
1021 switch (type) {
1022 case OVS_SAMPLE_ATTR_PROBABILITY:
1023 if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY, sizeof(u32), nla_data(a)))
1024 return -EMSGSIZE;
1025 break;
1026 case OVS_SAMPLE_ATTR_ACTIONS:
1027 st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
1028 if (!st_sample)
1029 return -EMSGSIZE;
1030 err = actions_to_attr(nla_data(a), nla_len(a), skb);
1031 if (err)
1032 return err;
1033 nla_nest_end(skb, st_sample);
1034 break;
1035 }
1036 }
1037
1038 nla_nest_end(skb, start);
1039 return err;
1040}
1041
1042static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
1043{
1044 const struct nlattr *ovs_key = nla_data(a);
1045 int key_type = nla_type(ovs_key);
1046 struct nlattr *start;
1047 int err;
1048
1049 switch (key_type) {
1050 case OVS_KEY_ATTR_IPV4_TUNNEL:
1051 start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
1052 if (!start)
1053 return -EMSGSIZE;
1054
1055 err = ovs_ipv4_tun_to_nlattr(skb, nla_data(ovs_key),
1056 nla_data(ovs_key));
1057 if (err)
1058 return err;
1059 nla_nest_end(skb, start);
1060 break;
1061 default:
1062 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
1063 return -EMSGSIZE;
1064 break;
1065 }
1066
1067 return 0;
1068}
1069
1070static int actions_to_attr(const struct nlattr *attr, int len, struct sk_buff *skb)
1071{
1072 const struct nlattr *a;
1073 int rem, err;
1074
1075 nla_for_each_attr(a, attr, len, rem) {
1076 int type = nla_type(a);
1077
1078 switch (type) {
1079 case OVS_ACTION_ATTR_SET:
1080 err = set_action_to_attr(a, skb);
1081 if (err)
1082 return err;
1083 break;
1084
1085 case OVS_ACTION_ATTR_SAMPLE:
1086 err = sample_action_to_attr(a, skb);
1087 if (err)
1088 return err;
1089 break;
1090 default:
1091 if (nla_put(skb, type, nla_len(a), nla_data(a)))
1092 return -EMSGSIZE;
1093 break;
1094 }
1095 }
1096
1097 return 0;
1098}
1099
1100static size_t ovs_flow_cmd_msg_size(const struct sw_flow_actions *acts) 632static size_t ovs_flow_cmd_msg_size(const struct sw_flow_actions *acts)
1101{ 633{
1102 return NLMSG_ALIGN(sizeof(struct ovs_header)) 634 return NLMSG_ALIGN(sizeof(struct ovs_header))
@@ -1133,8 +665,7 @@ static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp,
1133 if (!nla) 665 if (!nla)
1134 goto nla_put_failure; 666 goto nla_put_failure;
1135 667
1136 err = ovs_flow_to_nlattrs(&flow->unmasked_key, 668 err = ovs_nla_put_flow(&flow->unmasked_key, &flow->unmasked_key, skb);
1137 &flow->unmasked_key, skb);
1138 if (err) 669 if (err)
1139 goto error; 670 goto error;
1140 nla_nest_end(skb, nla); 671 nla_nest_end(skb, nla);
@@ -1143,7 +674,7 @@ static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp,
1143 if (!nla) 674 if (!nla)
1144 goto nla_put_failure; 675 goto nla_put_failure;
1145 676
1146 err = ovs_flow_to_nlattrs(&flow->key, &flow->mask->key, skb); 677 err = ovs_nla_put_flow(&flow->key, &flow->mask->key, skb);
1147 if (err) 678 if (err)
1148 goto error; 679 goto error;
1149 680
@@ -1186,7 +717,8 @@ static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp,
1186 sf_acts = rcu_dereference_check(flow->sf_acts, 717 sf_acts = rcu_dereference_check(flow->sf_acts,
1187 lockdep_ovsl_is_held()); 718 lockdep_ovsl_is_held());
1188 719
1189 err = actions_to_attr(sf_acts->actions, sf_acts->actions_len, skb); 720 err = ovs_nla_put_actions(sf_acts->actions,
721 sf_acts->actions_len, skb);
1190 if (!err) 722 if (!err)
1191 nla_nest_end(skb, start); 723 nla_nest_end(skb, start);
1192 else { 724 else {
@@ -1252,21 +784,21 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
1252 goto error; 784 goto error;
1253 785
1254 ovs_match_init(&match, &key, &mask); 786 ovs_match_init(&match, &key, &mask);
1255 error = ovs_match_from_nlattrs(&match, 787 error = ovs_nla_get_match(&match,
1256 a[OVS_FLOW_ATTR_KEY], a[OVS_FLOW_ATTR_MASK]); 788 a[OVS_FLOW_ATTR_KEY], a[OVS_FLOW_ATTR_MASK]);
1257 if (error) 789 if (error)
1258 goto error; 790 goto error;
1259 791
1260 /* Validate actions. */ 792 /* Validate actions. */
1261 if (a[OVS_FLOW_ATTR_ACTIONS]) { 793 if (a[OVS_FLOW_ATTR_ACTIONS]) {
1262 acts = ovs_flow_actions_alloc(nla_len(a[OVS_FLOW_ATTR_ACTIONS])); 794 acts = ovs_nla_alloc_flow_actions(nla_len(a[OVS_FLOW_ATTR_ACTIONS]));
1263 error = PTR_ERR(acts); 795 error = PTR_ERR(acts);
1264 if (IS_ERR(acts)) 796 if (IS_ERR(acts))
1265 goto error; 797 goto error;
1266 798
1267 ovs_flow_key_mask(&masked_key, &key, &mask); 799 ovs_flow_mask_key(&masked_key, &key, &mask);
1268 error = validate_and_copy_actions(a[OVS_FLOW_ATTR_ACTIONS], 800 error = ovs_nla_copy_actions(a[OVS_FLOW_ATTR_ACTIONS],
1269 &masked_key, 0, &acts); 801 &masked_key, 0, &acts);
1270 if (error) { 802 if (error) {
1271 OVS_NLERR("Flow actions may not be safe on all matching packets.\n"); 803 OVS_NLERR("Flow actions may not be safe on all matching packets.\n");
1272 goto err_kfree; 804 goto err_kfree;
@@ -1285,7 +817,7 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
1285 table = ovsl_dereference(dp->table); 817 table = ovsl_dereference(dp->table);
1286 818
1287 /* Check if this is a duplicate flow */ 819 /* Check if this is a duplicate flow */
1288 flow = ovs_flow_lookup(table, &key); 820 flow = ovs_flow_tbl_lookup(table, &key);
1289 if (!flow) { 821 if (!flow) {
1290 struct flow_table *new_table = NULL; 822 struct flow_table *new_table = NULL;
1291 struct sw_flow_mask *mask_p; 823 struct sw_flow_mask *mask_p;
@@ -1336,7 +868,7 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
1336 rcu_assign_pointer(flow->sf_acts, acts); 868 rcu_assign_pointer(flow->sf_acts, acts);
1337 869
1338 /* Put flow in bucket. */ 870 /* Put flow in bucket. */
1339 ovs_flow_insert(table, flow); 871 ovs_flow_tbl_insert(table, flow);
1340 872
1341 reply = ovs_flow_cmd_build_info(flow, dp, info->snd_portid, 873 reply = ovs_flow_cmd_build_info(flow, dp, info->snd_portid,
1342 info->snd_seq, OVS_FLOW_CMD_NEW); 874 info->snd_seq, OVS_FLOW_CMD_NEW);
@@ -1357,7 +889,7 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
1357 889
1358 /* The unmasked key has to be the same for flow updates. */ 890 /* The unmasked key has to be the same for flow updates. */
1359 error = -EINVAL; 891 error = -EINVAL;
1360 if (!ovs_flow_cmp_unmasked_key(flow, &key, match.range.end)) { 892 if (!ovs_flow_cmp_unmasked_key(flow, &match)) {
1361 OVS_NLERR("Flow modification message rejected, unmasked key does not match.\n"); 893 OVS_NLERR("Flow modification message rejected, unmasked key does not match.\n");
1362 goto err_unlock_ovs; 894 goto err_unlock_ovs;
1363 } 895 }
@@ -1365,7 +897,7 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
1365 /* Update actions. */ 897 /* Update actions. */
1366 old_acts = ovsl_dereference(flow->sf_acts); 898 old_acts = ovsl_dereference(flow->sf_acts);
1367 rcu_assign_pointer(flow->sf_acts, acts); 899 rcu_assign_pointer(flow->sf_acts, acts);
1368 ovs_flow_deferred_free_acts(old_acts); 900 ovs_nla_free_flow_actions(old_acts);
1369 901
1370 reply = ovs_flow_cmd_build_info(flow, dp, info->snd_portid, 902 reply = ovs_flow_cmd_build_info(flow, dp, info->snd_portid,
1371 info->snd_seq, OVS_FLOW_CMD_NEW); 903 info->snd_seq, OVS_FLOW_CMD_NEW);
@@ -1414,7 +946,7 @@ static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info)
1414 } 946 }
1415 947
1416 ovs_match_init(&match, &key, NULL); 948 ovs_match_init(&match, &key, NULL);
1417 err = ovs_match_from_nlattrs(&match, a[OVS_FLOW_ATTR_KEY], NULL); 949 err = ovs_nla_get_match(&match, a[OVS_FLOW_ATTR_KEY], NULL);
1418 if (err) 950 if (err)
1419 return err; 951 return err;
1420 952
@@ -1426,8 +958,8 @@ static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info)
1426 } 958 }
1427 959
1428 table = ovsl_dereference(dp->table); 960 table = ovsl_dereference(dp->table);
1429 flow = ovs_flow_lookup_unmasked_key(table, &match); 961 flow = ovs_flow_tbl_lookup(table, &key);
1430 if (!flow) { 962 if (!flow || !ovs_flow_cmp_unmasked_key(flow, &match)) {
1431 err = -ENOENT; 963 err = -ENOENT;
1432 goto unlock; 964 goto unlock;
1433 } 965 }
@@ -1471,13 +1003,13 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
1471 } 1003 }
1472 1004
1473 ovs_match_init(&match, &key, NULL); 1005 ovs_match_init(&match, &key, NULL);
1474 err = ovs_match_from_nlattrs(&match, a[OVS_FLOW_ATTR_KEY], NULL); 1006 err = ovs_nla_get_match(&match, a[OVS_FLOW_ATTR_KEY], NULL);
1475 if (err) 1007 if (err)
1476 goto unlock; 1008 goto unlock;
1477 1009
1478 table = ovsl_dereference(dp->table); 1010 table = ovsl_dereference(dp->table);
1479 flow = ovs_flow_lookup_unmasked_key(table, &match); 1011 flow = ovs_flow_tbl_lookup(table, &key);
1480 if (!flow) { 1012 if (!flow || !ovs_flow_cmp_unmasked_key(flow, &match)) {
1481 err = -ENOENT; 1013 err = -ENOENT;
1482 goto unlock; 1014 goto unlock;
1483 } 1015 }
@@ -1488,7 +1020,7 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
1488 goto unlock; 1020 goto unlock;
1489 } 1021 }
1490 1022
1491 ovs_flow_remove(table, flow); 1023 ovs_flow_tbl_remove(table, flow);
1492 1024
1493 err = ovs_flow_cmd_fill_info(flow, dp, reply, info->snd_portid, 1025 err = ovs_flow_cmd_fill_info(flow, dp, reply, info->snd_portid,
1494 info->snd_seq, 0, OVS_FLOW_CMD_DEL); 1026 info->snd_seq, 0, OVS_FLOW_CMD_DEL);
@@ -1524,7 +1056,7 @@ static int ovs_flow_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
1524 1056
1525 bucket = cb->args[0]; 1057 bucket = cb->args[0];
1526 obj = cb->args[1]; 1058 obj = cb->args[1];
1527 flow = ovs_flow_dump_next(table, &bucket, &obj); 1059 flow = ovs_flow_tbl_dump_next(table, &bucket, &obj);
1528 if (!flow) 1060 if (!flow)
1529 break; 1061 break;
1530 1062
@@ -1700,7 +1232,7 @@ static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info)
1700 } 1232 }
1701 1233
1702 dp->ports = kmalloc(DP_VPORT_HASH_BUCKETS * sizeof(struct hlist_head), 1234 dp->ports = kmalloc(DP_VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
1703 GFP_KERNEL); 1235 GFP_KERNEL);
1704 if (!dp->ports) { 1236 if (!dp->ports) {
1705 err = -ENOMEM; 1237 err = -ENOMEM;
1706 goto err_destroy_percpu; 1238 goto err_destroy_percpu;
diff --git a/net/openvswitch/datapath.h b/net/openvswitch/datapath.h
index 2c15541f3b46..a6982ef84f20 100644
--- a/net/openvswitch/datapath.h
+++ b/net/openvswitch/datapath.h
@@ -27,6 +27,7 @@
27#include <linux/u64_stats_sync.h> 27#include <linux/u64_stats_sync.h>
28 28
29#include "flow.h" 29#include "flow.h"
30#include "flow_table.h"
30#include "vport.h" 31#include "vport.h"
31 32
32#define DP_MAX_PORTS USHRT_MAX 33#define DP_MAX_PORTS USHRT_MAX
diff --git a/net/openvswitch/flow.c b/net/openvswitch/flow.c
index 410db90db73d..617810f1a21e 100644
--- a/net/openvswitch/flow.c
+++ b/net/openvswitch/flow.c
@@ -45,202 +45,40 @@
45#include <net/ipv6.h> 45#include <net/ipv6.h>
46#include <net/ndisc.h> 46#include <net/ndisc.h>
47 47
48static struct kmem_cache *flow_cache; 48u64 ovs_flow_used_time(unsigned long flow_jiffies)
49
50static void ovs_sw_flow_mask_set(struct sw_flow_mask *mask,
51 struct sw_flow_key_range *range, u8 val);
52
53static void update_range__(struct sw_flow_match *match,
54 size_t offset, size_t size, bool is_mask)
55{ 49{
56 struct sw_flow_key_range *range = NULL; 50 struct timespec cur_ts;
57 size_t start = rounddown(offset, sizeof(long)); 51 u64 cur_ms, idle_ms;
58 size_t end = roundup(offset + size, sizeof(long));
59
60 if (!is_mask)
61 range = &match->range;
62 else if (match->mask)
63 range = &match->mask->range;
64
65 if (!range)
66 return;
67
68 if (range->start == range->end) {
69 range->start = start;
70 range->end = end;
71 return;
72 }
73
74 if (range->start > start)
75 range->start = start;
76 52
77 if (range->end < end) 53 ktime_get_ts(&cur_ts);
78 range->end = end; 54 idle_ms = jiffies_to_msecs(jiffies - flow_jiffies);
79} 55 cur_ms = (u64)cur_ts.tv_sec * MSEC_PER_SEC +
56 cur_ts.tv_nsec / NSEC_PER_MSEC;
80 57
81#define SW_FLOW_KEY_PUT(match, field, value, is_mask) \ 58 return cur_ms - idle_ms;
82 do { \
83 update_range__(match, offsetof(struct sw_flow_key, field), \
84 sizeof((match)->key->field), is_mask); \
85 if (is_mask) { \
86 if ((match)->mask) \
87 (match)->mask->key.field = value; \
88 } else { \
89 (match)->key->field = value; \
90 } \
91 } while (0)
92
93#define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask) \
94 do { \
95 update_range__(match, offsetof(struct sw_flow_key, field), \
96 len, is_mask); \
97 if (is_mask) { \
98 if ((match)->mask) \
99 memcpy(&(match)->mask->key.field, value_p, len);\
100 } else { \
101 memcpy(&(match)->key->field, value_p, len); \
102 } \
103 } while (0)
104
105static u16 range_n_bytes(const struct sw_flow_key_range *range)
106{
107 return range->end - range->start;
108} 59}
109 60
110void ovs_match_init(struct sw_flow_match *match, 61#define TCP_FLAGS_OFFSET 13
111 struct sw_flow_key *key, 62#define TCP_FLAG_MASK 0x3f
112 struct sw_flow_mask *mask)
113{
114 memset(match, 0, sizeof(*match));
115 match->key = key;
116 match->mask = mask;
117
118 memset(key, 0, sizeof(*key));
119
120 if (mask) {
121 memset(&mask->key, 0, sizeof(mask->key));
122 mask->range.start = mask->range.end = 0;
123 }
124}
125 63
126static bool ovs_match_validate(const struct sw_flow_match *match, 64void ovs_flow_used(struct sw_flow *flow, struct sk_buff *skb)
127 u64 key_attrs, u64 mask_attrs)
128{ 65{
129 u64 key_expected = 1 << OVS_KEY_ATTR_ETHERNET; 66 u8 tcp_flags = 0;
130 u64 mask_allowed = key_attrs; /* At most allow all key attributes */
131
132 /* The following mask attributes allowed only if they
133 * pass the validation tests. */
134 mask_allowed &= ~((1 << OVS_KEY_ATTR_IPV4)
135 | (1 << OVS_KEY_ATTR_IPV6)
136 | (1 << OVS_KEY_ATTR_TCP)
137 | (1 << OVS_KEY_ATTR_UDP)
138 | (1 << OVS_KEY_ATTR_SCTP)
139 | (1 << OVS_KEY_ATTR_ICMP)
140 | (1 << OVS_KEY_ATTR_ICMPV6)
141 | (1 << OVS_KEY_ATTR_ARP)
142 | (1 << OVS_KEY_ATTR_ND));
143
144 /* Always allowed mask fields. */
145 mask_allowed |= ((1 << OVS_KEY_ATTR_TUNNEL)
146 | (1 << OVS_KEY_ATTR_IN_PORT)
147 | (1 << OVS_KEY_ATTR_ETHERTYPE));
148
149 /* Check key attributes. */
150 if (match->key->eth.type == htons(ETH_P_ARP)
151 || match->key->eth.type == htons(ETH_P_RARP)) {
152 key_expected |= 1 << OVS_KEY_ATTR_ARP;
153 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
154 mask_allowed |= 1 << OVS_KEY_ATTR_ARP;
155 }
156
157 if (match->key->eth.type == htons(ETH_P_IP)) {
158 key_expected |= 1 << OVS_KEY_ATTR_IPV4;
159 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
160 mask_allowed |= 1 << OVS_KEY_ATTR_IPV4;
161
162 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
163 if (match->key->ip.proto == IPPROTO_UDP) {
164 key_expected |= 1 << OVS_KEY_ATTR_UDP;
165 if (match->mask && (match->mask->key.ip.proto == 0xff))
166 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
167 }
168
169 if (match->key->ip.proto == IPPROTO_SCTP) {
170 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
171 if (match->mask && (match->mask->key.ip.proto == 0xff))
172 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
173 }
174
175 if (match->key->ip.proto == IPPROTO_TCP) {
176 key_expected |= 1 << OVS_KEY_ATTR_TCP;
177 if (match->mask && (match->mask->key.ip.proto == 0xff))
178 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
179 }
180
181 if (match->key->ip.proto == IPPROTO_ICMP) {
182 key_expected |= 1 << OVS_KEY_ATTR_ICMP;
183 if (match->mask && (match->mask->key.ip.proto == 0xff))
184 mask_allowed |= 1 << OVS_KEY_ATTR_ICMP;
185 }
186 }
187 }
188
189 if (match->key->eth.type == htons(ETH_P_IPV6)) {
190 key_expected |= 1 << OVS_KEY_ATTR_IPV6;
191 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
192 mask_allowed |= 1 << OVS_KEY_ATTR_IPV6;
193
194 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
195 if (match->key->ip.proto == IPPROTO_UDP) {
196 key_expected |= 1 << OVS_KEY_ATTR_UDP;
197 if (match->mask && (match->mask->key.ip.proto == 0xff))
198 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
199 }
200
201 if (match->key->ip.proto == IPPROTO_SCTP) {
202 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
203 if (match->mask && (match->mask->key.ip.proto == 0xff))
204 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
205 }
206
207 if (match->key->ip.proto == IPPROTO_TCP) {
208 key_expected |= 1 << OVS_KEY_ATTR_TCP;
209 if (match->mask && (match->mask->key.ip.proto == 0xff))
210 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
211 }
212
213 if (match->key->ip.proto == IPPROTO_ICMPV6) {
214 key_expected |= 1 << OVS_KEY_ATTR_ICMPV6;
215 if (match->mask && (match->mask->key.ip.proto == 0xff))
216 mask_allowed |= 1 << OVS_KEY_ATTR_ICMPV6;
217
218 if (match->key->ipv6.tp.src ==
219 htons(NDISC_NEIGHBOUR_SOLICITATION) ||
220 match->key->ipv6.tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
221 key_expected |= 1 << OVS_KEY_ATTR_ND;
222 if (match->mask && (match->mask->key.ipv6.tp.src == htons(0xffff)))
223 mask_allowed |= 1 << OVS_KEY_ATTR_ND;
224 }
225 }
226 }
227 }
228
229 if ((key_attrs & key_expected) != key_expected) {
230 /* Key attributes check failed. */
231 OVS_NLERR("Missing expected key attributes (key_attrs=%llx, expected=%llx).\n",
232 key_attrs, key_expected);
233 return false;
234 }
235 67
236 if ((mask_attrs & mask_allowed) != mask_attrs) { 68 if ((flow->key.eth.type == htons(ETH_P_IP) ||
237 /* Mask attributes check failed. */ 69 flow->key.eth.type == htons(ETH_P_IPV6)) &&
238 OVS_NLERR("Contain more than allowed mask fields (mask_attrs=%llx, mask_allowed=%llx).\n", 70 flow->key.ip.proto == IPPROTO_TCP &&
239 mask_attrs, mask_allowed); 71 likely(skb->len >= skb_transport_offset(skb) + sizeof(struct tcphdr))) {
240 return false; 72 u8 *tcp = (u8 *)tcp_hdr(skb);
73 tcp_flags = *(tcp + TCP_FLAGS_OFFSET) & TCP_FLAG_MASK;
241 } 74 }
242 75
243 return true; 76 spin_lock(&flow->lock);
77 flow->used = jiffies;
78 flow->packet_count++;
79 flow->byte_count += skb->len;
80 flow->tcp_flags |= tcp_flags;
81 spin_unlock(&flow->lock);
244} 82}
245 83
246static int check_header(struct sk_buff *skb, int len) 84static int check_header(struct sk_buff *skb, int len)
@@ -311,19 +149,6 @@ static bool icmphdr_ok(struct sk_buff *skb)
311 sizeof(struct icmphdr)); 149 sizeof(struct icmphdr));
312} 150}
313 151
314u64 ovs_flow_used_time(unsigned long flow_jiffies)
315{
316 struct timespec cur_ts;
317 u64 cur_ms, idle_ms;
318
319 ktime_get_ts(&cur_ts);
320 idle_ms = jiffies_to_msecs(jiffies - flow_jiffies);
321 cur_ms = (u64)cur_ts.tv_sec * MSEC_PER_SEC +
322 cur_ts.tv_nsec / NSEC_PER_MSEC;
323
324 return cur_ms - idle_ms;
325}
326
327static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key) 152static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key)
328{ 153{
329 unsigned int nh_ofs = skb_network_offset(skb); 154 unsigned int nh_ofs = skb_network_offset(skb);
@@ -372,311 +197,6 @@ static bool icmp6hdr_ok(struct sk_buff *skb)
372 sizeof(struct icmp6hdr)); 197 sizeof(struct icmp6hdr));
373} 198}
374 199
375void ovs_flow_key_mask(struct sw_flow_key *dst, const struct sw_flow_key *src,
376 const struct sw_flow_mask *mask)
377{
378 const long *m = (long *)((u8 *)&mask->key + mask->range.start);
379 const long *s = (long *)((u8 *)src + mask->range.start);
380 long *d = (long *)((u8 *)dst + mask->range.start);
381 int i;
382
383 /* The memory outside of the 'mask->range' are not set since
384 * further operations on 'dst' only uses contents within
385 * 'mask->range'.
386 */
387 for (i = 0; i < range_n_bytes(&mask->range); i += sizeof(long))
388 *d++ = *s++ & *m++;
389}
390
391#define TCP_FLAGS_OFFSET 13
392#define TCP_FLAG_MASK 0x3f
393
394void ovs_flow_used(struct sw_flow *flow, struct sk_buff *skb)
395{
396 u8 tcp_flags = 0;
397
398 if ((flow->key.eth.type == htons(ETH_P_IP) ||
399 flow->key.eth.type == htons(ETH_P_IPV6)) &&
400 flow->key.ip.proto == IPPROTO_TCP &&
401 likely(skb->len >= skb_transport_offset(skb) + sizeof(struct tcphdr))) {
402 u8 *tcp = (u8 *)tcp_hdr(skb);
403 tcp_flags = *(tcp + TCP_FLAGS_OFFSET) & TCP_FLAG_MASK;
404 }
405
406 spin_lock(&flow->lock);
407 flow->used = jiffies;
408 flow->packet_count++;
409 flow->byte_count += skb->len;
410 flow->tcp_flags |= tcp_flags;
411 spin_unlock(&flow->lock);
412}
413
414struct sw_flow_actions *ovs_flow_actions_alloc(int size)
415{
416 struct sw_flow_actions *sfa;
417
418 if (size > MAX_ACTIONS_BUFSIZE)
419 return ERR_PTR(-EINVAL);
420
421 sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
422 if (!sfa)
423 return ERR_PTR(-ENOMEM);
424
425 sfa->actions_len = 0;
426 return sfa;
427}
428
429struct sw_flow *ovs_flow_alloc(void)
430{
431 struct sw_flow *flow;
432
433 flow = kmem_cache_alloc(flow_cache, GFP_KERNEL);
434 if (!flow)
435 return ERR_PTR(-ENOMEM);
436
437 spin_lock_init(&flow->lock);
438 flow->sf_acts = NULL;
439 flow->mask = NULL;
440
441 return flow;
442}
443
444static struct hlist_head *find_bucket(struct flow_table *table, u32 hash)
445{
446 hash = jhash_1word(hash, table->hash_seed);
447 return flex_array_get(table->buckets,
448 (hash & (table->n_buckets - 1)));
449}
450
451static struct flex_array *alloc_buckets(unsigned int n_buckets)
452{
453 struct flex_array *buckets;
454 int i, err;
455
456 buckets = flex_array_alloc(sizeof(struct hlist_head),
457 n_buckets, GFP_KERNEL);
458 if (!buckets)
459 return NULL;
460
461 err = flex_array_prealloc(buckets, 0, n_buckets, GFP_KERNEL);
462 if (err) {
463 flex_array_free(buckets);
464 return NULL;
465 }
466
467 for (i = 0; i < n_buckets; i++)
468 INIT_HLIST_HEAD((struct hlist_head *)
469 flex_array_get(buckets, i));
470
471 return buckets;
472}
473
474static void free_buckets(struct flex_array *buckets)
475{
476 flex_array_free(buckets);
477}
478
479static struct flow_table *__flow_tbl_alloc(int new_size)
480{
481 struct flow_table *table = kmalloc(sizeof(*table), GFP_KERNEL);
482
483 if (!table)
484 return NULL;
485
486 table->buckets = alloc_buckets(new_size);
487
488 if (!table->buckets) {
489 kfree(table);
490 return NULL;
491 }
492 table->n_buckets = new_size;
493 table->count = 0;
494 table->node_ver = 0;
495 table->keep_flows = false;
496 get_random_bytes(&table->hash_seed, sizeof(u32));
497 table->mask_list = NULL;
498
499 return table;
500}
501
502static void __flow_tbl_destroy(struct flow_table *table)
503{
504 int i;
505
506 if (table->keep_flows)
507 goto skip_flows;
508
509 for (i = 0; i < table->n_buckets; i++) {
510 struct sw_flow *flow;
511 struct hlist_head *head = flex_array_get(table->buckets, i);
512 struct hlist_node *n;
513 int ver = table->node_ver;
514
515 hlist_for_each_entry_safe(flow, n, head, hash_node[ver]) {
516 hlist_del(&flow->hash_node[ver]);
517 ovs_flow_free(flow, false);
518 }
519 }
520
521 BUG_ON(!list_empty(table->mask_list));
522 kfree(table->mask_list);
523
524skip_flows:
525 free_buckets(table->buckets);
526 kfree(table);
527}
528
529struct flow_table *ovs_flow_tbl_alloc(int new_size)
530{
531 struct flow_table *table = __flow_tbl_alloc(new_size);
532
533 if (!table)
534 return NULL;
535
536 table->mask_list = kmalloc(sizeof(struct list_head), GFP_KERNEL);
537 if (!table->mask_list) {
538 table->keep_flows = true;
539 __flow_tbl_destroy(table);
540 return NULL;
541 }
542 INIT_LIST_HEAD(table->mask_list);
543
544 return table;
545}
546
547static void flow_tbl_destroy_rcu_cb(struct rcu_head *rcu)
548{
549 struct flow_table *table = container_of(rcu, struct flow_table, rcu);
550
551 __flow_tbl_destroy(table);
552}
553
554void ovs_flow_tbl_destroy(struct flow_table *table, bool deferred)
555{
556 if (!table)
557 return;
558
559 if (deferred)
560 call_rcu(&table->rcu, flow_tbl_destroy_rcu_cb);
561 else
562 __flow_tbl_destroy(table);
563}
564
565struct sw_flow *ovs_flow_dump_next(struct flow_table *table, u32 *bucket, u32 *last)
566{
567 struct sw_flow *flow;
568 struct hlist_head *head;
569 int ver;
570 int i;
571
572 ver = table->node_ver;
573 while (*bucket < table->n_buckets) {
574 i = 0;
575 head = flex_array_get(table->buckets, *bucket);
576 hlist_for_each_entry_rcu(flow, head, hash_node[ver]) {
577 if (i < *last) {
578 i++;
579 continue;
580 }
581 *last = i + 1;
582 return flow;
583 }
584 (*bucket)++;
585 *last = 0;
586 }
587
588 return NULL;
589}
590
591static void __tbl_insert(struct flow_table *table, struct sw_flow *flow)
592{
593 struct hlist_head *head;
594
595 head = find_bucket(table, flow->hash);
596 hlist_add_head_rcu(&flow->hash_node[table->node_ver], head);
597
598 table->count++;
599}
600
601static void flow_table_copy_flows(struct flow_table *old, struct flow_table *new)
602{
603 int old_ver;
604 int i;
605
606 old_ver = old->node_ver;
607 new->node_ver = !old_ver;
608
609 /* Insert in new table. */
610 for (i = 0; i < old->n_buckets; i++) {
611 struct sw_flow *flow;
612 struct hlist_head *head;
613
614 head = flex_array_get(old->buckets, i);
615
616 hlist_for_each_entry(flow, head, hash_node[old_ver])
617 __tbl_insert(new, flow);
618 }
619
620 new->mask_list = old->mask_list;
621 old->keep_flows = true;
622}
623
624static struct flow_table *__flow_tbl_rehash(struct flow_table *table, int n_buckets)
625{
626 struct flow_table *new_table;
627
628 new_table = __flow_tbl_alloc(n_buckets);
629 if (!new_table)
630 return ERR_PTR(-ENOMEM);
631
632 flow_table_copy_flows(table, new_table);
633
634 return new_table;
635}
636
637struct flow_table *ovs_flow_tbl_rehash(struct flow_table *table)
638{
639 return __flow_tbl_rehash(table, table->n_buckets);
640}
641
642struct flow_table *ovs_flow_tbl_expand(struct flow_table *table)
643{
644 return __flow_tbl_rehash(table, table->n_buckets * 2);
645}
646
647static void __flow_free(struct sw_flow *flow)
648{
649 kfree((struct sf_flow_acts __force *)flow->sf_acts);
650 kmem_cache_free(flow_cache, flow);
651}
652
653static void rcu_free_flow_callback(struct rcu_head *rcu)
654{
655 struct sw_flow *flow = container_of(rcu, struct sw_flow, rcu);
656
657 __flow_free(flow);
658}
659
660void ovs_flow_free(struct sw_flow *flow, bool deferred)
661{
662 if (!flow)
663 return;
664
665 ovs_sw_flow_mask_del_ref(flow->mask, deferred);
666
667 if (deferred)
668 call_rcu(&flow->rcu, rcu_free_flow_callback);
669 else
670 __flow_free(flow);
671}
672
673/* Schedules 'sf_acts' to be freed after the next RCU grace period.
674 * The caller must hold rcu_read_lock for this to be sensible. */
675void ovs_flow_deferred_free_acts(struct sw_flow_actions *sf_acts)
676{
677 kfree_rcu(sf_acts, rcu);
678}
679
680static int parse_vlan(struct sk_buff *skb, struct sw_flow_key *key) 200static int parse_vlan(struct sk_buff *skb, struct sw_flow_key *key)
681{ 201{
682 struct qtag_prefix { 202 struct qtag_prefix {
@@ -1002,1080 +522,3 @@ int ovs_flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key)
1002 522
1003 return 0; 523 return 0;
1004} 524}
1005
1006static u32 ovs_flow_hash(const struct sw_flow_key *key, int key_start,
1007 int key_end)
1008{
1009 u32 *hash_key = (u32 *)((u8 *)key + key_start);
1010 int hash_u32s = (key_end - key_start) >> 2;
1011
1012 /* Make sure number of hash bytes are multiple of u32. */
1013 BUILD_BUG_ON(sizeof(long) % sizeof(u32));
1014
1015 return jhash2(hash_key, hash_u32s, 0);
1016}
1017
1018static int flow_key_start(const struct sw_flow_key *key)
1019{
1020 if (key->tun_key.ipv4_dst)
1021 return 0;
1022 else
1023 return rounddown(offsetof(struct sw_flow_key, phy),
1024 sizeof(long));
1025}
1026
1027static bool __cmp_key(const struct sw_flow_key *key1,
1028 const struct sw_flow_key *key2, int key_start, int key_end)
1029{
1030 const long *cp1 = (long *)((u8 *)key1 + key_start);
1031 const long *cp2 = (long *)((u8 *)key2 + key_start);
1032 long diffs = 0;
1033 int i;
1034
1035 for (i = key_start; i < key_end; i += sizeof(long))
1036 diffs |= *cp1++ ^ *cp2++;
1037
1038 return diffs == 0;
1039}
1040
1041static bool __flow_cmp_masked_key(const struct sw_flow *flow,
1042 const struct sw_flow_key *key, int key_start, int key_end)
1043{
1044 return __cmp_key(&flow->key, key, key_start, key_end);
1045}
1046
1047static bool __flow_cmp_unmasked_key(const struct sw_flow *flow,
1048 const struct sw_flow_key *key, int key_start, int key_end)
1049{
1050 return __cmp_key(&flow->unmasked_key, key, key_start, key_end);
1051}
1052
1053bool ovs_flow_cmp_unmasked_key(const struct sw_flow *flow,
1054 const struct sw_flow_key *key, int key_end)
1055{
1056 int key_start;
1057 key_start = flow_key_start(key);
1058
1059 return __flow_cmp_unmasked_key(flow, key, key_start, key_end);
1060
1061}
1062
1063struct sw_flow *ovs_flow_lookup_unmasked_key(struct flow_table *table,
1064 struct sw_flow_match *match)
1065{
1066 struct sw_flow_key *unmasked = match->key;
1067 int key_end = match->range.end;
1068 struct sw_flow *flow;
1069
1070 flow = ovs_flow_lookup(table, unmasked);
1071 if (flow && (!ovs_flow_cmp_unmasked_key(flow, unmasked, key_end)))
1072 flow = NULL;
1073
1074 return flow;
1075}
1076
1077static struct sw_flow *ovs_masked_flow_lookup(struct flow_table *table,
1078 const struct sw_flow_key *unmasked,
1079 struct sw_flow_mask *mask)
1080{
1081 struct sw_flow *flow;
1082 struct hlist_head *head;
1083 int key_start = mask->range.start;
1084 int key_end = mask->range.end;
1085 u32 hash;
1086 struct sw_flow_key masked_key;
1087
1088 ovs_flow_key_mask(&masked_key, unmasked, mask);
1089 hash = ovs_flow_hash(&masked_key, key_start, key_end);
1090 head = find_bucket(table, hash);
1091 hlist_for_each_entry_rcu(flow, head, hash_node[table->node_ver]) {
1092 if (flow->mask == mask &&
1093 __flow_cmp_masked_key(flow, &masked_key,
1094 key_start, key_end))
1095 return flow;
1096 }
1097 return NULL;
1098}
1099
1100struct sw_flow *ovs_flow_lookup(struct flow_table *tbl,
1101 const struct sw_flow_key *key)
1102{
1103 struct sw_flow *flow = NULL;
1104 struct sw_flow_mask *mask;
1105
1106 list_for_each_entry_rcu(mask, tbl->mask_list, list) {
1107 flow = ovs_masked_flow_lookup(tbl, key, mask);
1108 if (flow) /* Found */
1109 break;
1110 }
1111
1112 return flow;
1113}
1114
1115
1116void ovs_flow_insert(struct flow_table *table, struct sw_flow *flow)
1117{
1118 flow->hash = ovs_flow_hash(&flow->key, flow->mask->range.start,
1119 flow->mask->range.end);
1120 __tbl_insert(table, flow);
1121}
1122
1123void ovs_flow_remove(struct flow_table *table, struct sw_flow *flow)
1124{
1125 BUG_ON(table->count == 0);
1126 hlist_del_rcu(&flow->hash_node[table->node_ver]);
1127 table->count--;
1128}
1129
1130/* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
1131const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
1132 [OVS_KEY_ATTR_ENCAP] = -1,
1133 [OVS_KEY_ATTR_PRIORITY] = sizeof(u32),
1134 [OVS_KEY_ATTR_IN_PORT] = sizeof(u32),
1135 [OVS_KEY_ATTR_SKB_MARK] = sizeof(u32),
1136 [OVS_KEY_ATTR_ETHERNET] = sizeof(struct ovs_key_ethernet),
1137 [OVS_KEY_ATTR_VLAN] = sizeof(__be16),
1138 [OVS_KEY_ATTR_ETHERTYPE] = sizeof(__be16),
1139 [OVS_KEY_ATTR_IPV4] = sizeof(struct ovs_key_ipv4),
1140 [OVS_KEY_ATTR_IPV6] = sizeof(struct ovs_key_ipv6),
1141 [OVS_KEY_ATTR_TCP] = sizeof(struct ovs_key_tcp),
1142 [OVS_KEY_ATTR_UDP] = sizeof(struct ovs_key_udp),
1143 [OVS_KEY_ATTR_SCTP] = sizeof(struct ovs_key_sctp),
1144 [OVS_KEY_ATTR_ICMP] = sizeof(struct ovs_key_icmp),
1145 [OVS_KEY_ATTR_ICMPV6] = sizeof(struct ovs_key_icmpv6),
1146 [OVS_KEY_ATTR_ARP] = sizeof(struct ovs_key_arp),
1147 [OVS_KEY_ATTR_ND] = sizeof(struct ovs_key_nd),
1148 [OVS_KEY_ATTR_TUNNEL] = -1,
1149};
1150
1151static bool is_all_zero(const u8 *fp, size_t size)
1152{
1153 int i;
1154
1155 if (!fp)
1156 return false;
1157
1158 for (i = 0; i < size; i++)
1159 if (fp[i])
1160 return false;
1161
1162 return true;
1163}
1164
1165static int __parse_flow_nlattrs(const struct nlattr *attr,
1166 const struct nlattr *a[],
1167 u64 *attrsp, bool nz)
1168{
1169 const struct nlattr *nla;
1170 u32 attrs;
1171 int rem;
1172
1173 attrs = *attrsp;
1174 nla_for_each_nested(nla, attr, rem) {
1175 u16 type = nla_type(nla);
1176 int expected_len;
1177
1178 if (type > OVS_KEY_ATTR_MAX) {
1179 OVS_NLERR("Unknown key attribute (type=%d, max=%d).\n",
1180 type, OVS_KEY_ATTR_MAX);
1181 return -EINVAL;
1182 }
1183
1184 if (attrs & (1 << type)) {
1185 OVS_NLERR("Duplicate key attribute (type %d).\n", type);
1186 return -EINVAL;
1187 }
1188
1189 expected_len = ovs_key_lens[type];
1190 if (nla_len(nla) != expected_len && expected_len != -1) {
1191 OVS_NLERR("Key attribute has unexpected length (type=%d"
1192 ", length=%d, expected=%d).\n", type,
1193 nla_len(nla), expected_len);
1194 return -EINVAL;
1195 }
1196
1197 if (!nz || !is_all_zero(nla_data(nla), expected_len)) {
1198 attrs |= 1 << type;
1199 a[type] = nla;
1200 }
1201 }
1202 if (rem) {
1203 OVS_NLERR("Message has %d unknown bytes.\n", rem);
1204 return -EINVAL;
1205 }
1206
1207 *attrsp = attrs;
1208 return 0;
1209}
1210
1211static int parse_flow_mask_nlattrs(const struct nlattr *attr,
1212 const struct nlattr *a[], u64 *attrsp)
1213{
1214 return __parse_flow_nlattrs(attr, a, attrsp, true);
1215}
1216
1217static int parse_flow_nlattrs(const struct nlattr *attr,
1218 const struct nlattr *a[], u64 *attrsp)
1219{
1220 return __parse_flow_nlattrs(attr, a, attrsp, false);
1221}
1222
1223int ovs_ipv4_tun_from_nlattr(const struct nlattr *attr,
1224 struct sw_flow_match *match, bool is_mask)
1225{
1226 struct nlattr *a;
1227 int rem;
1228 bool ttl = false;
1229 __be16 tun_flags = 0;
1230
1231 nla_for_each_nested(a, attr, rem) {
1232 int type = nla_type(a);
1233 static const u32 ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
1234 [OVS_TUNNEL_KEY_ATTR_ID] = sizeof(u64),
1235 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = sizeof(u32),
1236 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = sizeof(u32),
1237 [OVS_TUNNEL_KEY_ATTR_TOS] = 1,
1238 [OVS_TUNNEL_KEY_ATTR_TTL] = 1,
1239 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = 0,
1240 [OVS_TUNNEL_KEY_ATTR_CSUM] = 0,
1241 };
1242
1243 if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
1244 OVS_NLERR("Unknown IPv4 tunnel attribute (type=%d, max=%d).\n",
1245 type, OVS_TUNNEL_KEY_ATTR_MAX);
1246 return -EINVAL;
1247 }
1248
1249 if (ovs_tunnel_key_lens[type] != nla_len(a)) {
1250 OVS_NLERR("IPv4 tunnel attribute type has unexpected "
1251 " length (type=%d, length=%d, expected=%d).\n",
1252 type, nla_len(a), ovs_tunnel_key_lens[type]);
1253 return -EINVAL;
1254 }
1255
1256 switch (type) {
1257 case OVS_TUNNEL_KEY_ATTR_ID:
1258 SW_FLOW_KEY_PUT(match, tun_key.tun_id,
1259 nla_get_be64(a), is_mask);
1260 tun_flags |= TUNNEL_KEY;
1261 break;
1262 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
1263 SW_FLOW_KEY_PUT(match, tun_key.ipv4_src,
1264 nla_get_be32(a), is_mask);
1265 break;
1266 case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
1267 SW_FLOW_KEY_PUT(match, tun_key.ipv4_dst,
1268 nla_get_be32(a), is_mask);
1269 break;
1270 case OVS_TUNNEL_KEY_ATTR_TOS:
1271 SW_FLOW_KEY_PUT(match, tun_key.ipv4_tos,
1272 nla_get_u8(a), is_mask);
1273 break;
1274 case OVS_TUNNEL_KEY_ATTR_TTL:
1275 SW_FLOW_KEY_PUT(match, tun_key.ipv4_ttl,
1276 nla_get_u8(a), is_mask);
1277 ttl = true;
1278 break;
1279 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
1280 tun_flags |= TUNNEL_DONT_FRAGMENT;
1281 break;
1282 case OVS_TUNNEL_KEY_ATTR_CSUM:
1283 tun_flags |= TUNNEL_CSUM;
1284 break;
1285 default:
1286 return -EINVAL;
1287 }
1288 }
1289
1290 SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
1291
1292 if (rem > 0) {
1293 OVS_NLERR("IPv4 tunnel attribute has %d unknown bytes.\n", rem);
1294 return -EINVAL;
1295 }
1296
1297 if (!is_mask) {
1298 if (!match->key->tun_key.ipv4_dst) {
1299 OVS_NLERR("IPv4 tunnel destination address is zero.\n");
1300 return -EINVAL;
1301 }
1302
1303 if (!ttl) {
1304 OVS_NLERR("IPv4 tunnel TTL not specified.\n");
1305 return -EINVAL;
1306 }
1307 }
1308
1309 return 0;
1310}
1311
1312int ovs_ipv4_tun_to_nlattr(struct sk_buff *skb,
1313 const struct ovs_key_ipv4_tunnel *tun_key,
1314 const struct ovs_key_ipv4_tunnel *output)
1315{
1316 struct nlattr *nla;
1317
1318 nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
1319 if (!nla)
1320 return -EMSGSIZE;
1321
1322 if (output->tun_flags & TUNNEL_KEY &&
1323 nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id))
1324 return -EMSGSIZE;
1325 if (output->ipv4_src &&
1326 nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC, output->ipv4_src))
1327 return -EMSGSIZE;
1328 if (output->ipv4_dst &&
1329 nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST, output->ipv4_dst))
1330 return -EMSGSIZE;
1331 if (output->ipv4_tos &&
1332 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->ipv4_tos))
1333 return -EMSGSIZE;
1334 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ipv4_ttl))
1335 return -EMSGSIZE;
1336 if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
1337 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
1338 return -EMSGSIZE;
1339 if ((output->tun_flags & TUNNEL_CSUM) &&
1340 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
1341 return -EMSGSIZE;
1342
1343 nla_nest_end(skb, nla);
1344 return 0;
1345}
1346
1347static int metadata_from_nlattrs(struct sw_flow_match *match, u64 *attrs,
1348 const struct nlattr **a, bool is_mask)
1349{
1350 if (*attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
1351 SW_FLOW_KEY_PUT(match, phy.priority,
1352 nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
1353 *attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
1354 }
1355
1356 if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
1357 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
1358
1359 if (is_mask)
1360 in_port = 0xffffffff; /* Always exact match in_port. */
1361 else if (in_port >= DP_MAX_PORTS)
1362 return -EINVAL;
1363
1364 SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
1365 *attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
1366 } else if (!is_mask) {
1367 SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
1368 }
1369
1370 if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
1371 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
1372
1373 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
1374 *attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
1375 }
1376 if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
1377 if (ovs_ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
1378 is_mask))
1379 return -EINVAL;
1380 *attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
1381 }
1382 return 0;
1383}
1384
1385static int ovs_key_from_nlattrs(struct sw_flow_match *match, u64 attrs,
1386 const struct nlattr **a, bool is_mask)
1387{
1388 int err;
1389 u64 orig_attrs = attrs;
1390
1391 err = metadata_from_nlattrs(match, &attrs, a, is_mask);
1392 if (err)
1393 return err;
1394
1395 if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
1396 const struct ovs_key_ethernet *eth_key;
1397
1398 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
1399 SW_FLOW_KEY_MEMCPY(match, eth.src,
1400 eth_key->eth_src, ETH_ALEN, is_mask);
1401 SW_FLOW_KEY_MEMCPY(match, eth.dst,
1402 eth_key->eth_dst, ETH_ALEN, is_mask);
1403 attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
1404 }
1405
1406 if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
1407 __be16 tci;
1408
1409 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1410 if (!(tci & htons(VLAN_TAG_PRESENT))) {
1411 if (is_mask)
1412 OVS_NLERR("VLAN TCI mask does not have exact match for VLAN_TAG_PRESENT bit.\n");
1413 else
1414 OVS_NLERR("VLAN TCI does not have VLAN_TAG_PRESENT bit set.\n");
1415
1416 return -EINVAL;
1417 }
1418
1419 SW_FLOW_KEY_PUT(match, eth.tci, tci, is_mask);
1420 attrs &= ~(1 << OVS_KEY_ATTR_VLAN);
1421 } else if (!is_mask)
1422 SW_FLOW_KEY_PUT(match, eth.tci, htons(0xffff), true);
1423
1424 if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
1425 __be16 eth_type;
1426
1427 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1428 if (is_mask) {
1429 /* Always exact match EtherType. */
1430 eth_type = htons(0xffff);
1431 } else if (ntohs(eth_type) < ETH_P_802_3_MIN) {
1432 OVS_NLERR("EtherType is less than minimum (type=%x, min=%x).\n",
1433 ntohs(eth_type), ETH_P_802_3_MIN);
1434 return -EINVAL;
1435 }
1436
1437 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
1438 attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1439 } else if (!is_mask) {
1440 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
1441 }
1442
1443 if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
1444 const struct ovs_key_ipv4 *ipv4_key;
1445
1446 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
1447 if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
1448 OVS_NLERR("Unknown IPv4 fragment type (value=%d, max=%d).\n",
1449 ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
1450 return -EINVAL;
1451 }
1452 SW_FLOW_KEY_PUT(match, ip.proto,
1453 ipv4_key->ipv4_proto, is_mask);
1454 SW_FLOW_KEY_PUT(match, ip.tos,
1455 ipv4_key->ipv4_tos, is_mask);
1456 SW_FLOW_KEY_PUT(match, ip.ttl,
1457 ipv4_key->ipv4_ttl, is_mask);
1458 SW_FLOW_KEY_PUT(match, ip.frag,
1459 ipv4_key->ipv4_frag, is_mask);
1460 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1461 ipv4_key->ipv4_src, is_mask);
1462 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1463 ipv4_key->ipv4_dst, is_mask);
1464 attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
1465 }
1466
1467 if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
1468 const struct ovs_key_ipv6 *ipv6_key;
1469
1470 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
1471 if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
1472 OVS_NLERR("Unknown IPv6 fragment type (value=%d, max=%d).\n",
1473 ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
1474 return -EINVAL;
1475 }
1476 SW_FLOW_KEY_PUT(match, ipv6.label,
1477 ipv6_key->ipv6_label, is_mask);
1478 SW_FLOW_KEY_PUT(match, ip.proto,
1479 ipv6_key->ipv6_proto, is_mask);
1480 SW_FLOW_KEY_PUT(match, ip.tos,
1481 ipv6_key->ipv6_tclass, is_mask);
1482 SW_FLOW_KEY_PUT(match, ip.ttl,
1483 ipv6_key->ipv6_hlimit, is_mask);
1484 SW_FLOW_KEY_PUT(match, ip.frag,
1485 ipv6_key->ipv6_frag, is_mask);
1486 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
1487 ipv6_key->ipv6_src,
1488 sizeof(match->key->ipv6.addr.src),
1489 is_mask);
1490 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
1491 ipv6_key->ipv6_dst,
1492 sizeof(match->key->ipv6.addr.dst),
1493 is_mask);
1494
1495 attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
1496 }
1497
1498 if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
1499 const struct ovs_key_arp *arp_key;
1500
1501 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
1502 if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
1503 OVS_NLERR("Unknown ARP opcode (opcode=%d).\n",
1504 arp_key->arp_op);
1505 return -EINVAL;
1506 }
1507
1508 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
1509 arp_key->arp_sip, is_mask);
1510 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
1511 arp_key->arp_tip, is_mask);
1512 SW_FLOW_KEY_PUT(match, ip.proto,
1513 ntohs(arp_key->arp_op), is_mask);
1514 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
1515 arp_key->arp_sha, ETH_ALEN, is_mask);
1516 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
1517 arp_key->arp_tha, ETH_ALEN, is_mask);
1518
1519 attrs &= ~(1 << OVS_KEY_ATTR_ARP);
1520 }
1521
1522 if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
1523 const struct ovs_key_tcp *tcp_key;
1524
1525 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
1526 if (orig_attrs & (1 << OVS_KEY_ATTR_IPV4)) {
1527 SW_FLOW_KEY_PUT(match, ipv4.tp.src,
1528 tcp_key->tcp_src, is_mask);
1529 SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
1530 tcp_key->tcp_dst, is_mask);
1531 } else {
1532 SW_FLOW_KEY_PUT(match, ipv6.tp.src,
1533 tcp_key->tcp_src, is_mask);
1534 SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
1535 tcp_key->tcp_dst, is_mask);
1536 }
1537 attrs &= ~(1 << OVS_KEY_ATTR_TCP);
1538 }
1539
1540 if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
1541 const struct ovs_key_udp *udp_key;
1542
1543 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
1544 if (orig_attrs & (1 << OVS_KEY_ATTR_IPV4)) {
1545 SW_FLOW_KEY_PUT(match, ipv4.tp.src,
1546 udp_key->udp_src, is_mask);
1547 SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
1548 udp_key->udp_dst, is_mask);
1549 } else {
1550 SW_FLOW_KEY_PUT(match, ipv6.tp.src,
1551 udp_key->udp_src, is_mask);
1552 SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
1553 udp_key->udp_dst, is_mask);
1554 }
1555 attrs &= ~(1 << OVS_KEY_ATTR_UDP);
1556 }
1557
1558 if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
1559 const struct ovs_key_sctp *sctp_key;
1560
1561 sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
1562 if (orig_attrs & (1 << OVS_KEY_ATTR_IPV4)) {
1563 SW_FLOW_KEY_PUT(match, ipv4.tp.src,
1564 sctp_key->sctp_src, is_mask);
1565 SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
1566 sctp_key->sctp_dst, is_mask);
1567 } else {
1568 SW_FLOW_KEY_PUT(match, ipv6.tp.src,
1569 sctp_key->sctp_src, is_mask);
1570 SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
1571 sctp_key->sctp_dst, is_mask);
1572 }
1573 attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
1574 }
1575
1576 if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
1577 const struct ovs_key_icmp *icmp_key;
1578
1579 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
1580 SW_FLOW_KEY_PUT(match, ipv4.tp.src,
1581 htons(icmp_key->icmp_type), is_mask);
1582 SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
1583 htons(icmp_key->icmp_code), is_mask);
1584 attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
1585 }
1586
1587 if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
1588 const struct ovs_key_icmpv6 *icmpv6_key;
1589
1590 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
1591 SW_FLOW_KEY_PUT(match, ipv6.tp.src,
1592 htons(icmpv6_key->icmpv6_type), is_mask);
1593 SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
1594 htons(icmpv6_key->icmpv6_code), is_mask);
1595 attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
1596 }
1597
1598 if (attrs & (1 << OVS_KEY_ATTR_ND)) {
1599 const struct ovs_key_nd *nd_key;
1600
1601 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
1602 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
1603 nd_key->nd_target,
1604 sizeof(match->key->ipv6.nd.target),
1605 is_mask);
1606 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
1607 nd_key->nd_sll, ETH_ALEN, is_mask);
1608 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
1609 nd_key->nd_tll, ETH_ALEN, is_mask);
1610 attrs &= ~(1 << OVS_KEY_ATTR_ND);
1611 }
1612
1613 if (attrs != 0)
1614 return -EINVAL;
1615
1616 return 0;
1617}
1618
1619/**
1620 * ovs_match_from_nlattrs - parses Netlink attributes into a flow key and
1621 * mask. In case the 'mask' is NULL, the flow is treated as exact match
1622 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
1623 * does not include any don't care bit.
1624 * @match: receives the extracted flow match information.
1625 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1626 * sequence. The fields should of the packet that triggered the creation
1627 * of this flow.
1628 * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
1629 * attribute specifies the mask field of the wildcarded flow.
1630 */
1631int ovs_match_from_nlattrs(struct sw_flow_match *match,
1632 const struct nlattr *key,
1633 const struct nlattr *mask)
1634{
1635 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1636 const struct nlattr *encap;
1637 u64 key_attrs = 0;
1638 u64 mask_attrs = 0;
1639 bool encap_valid = false;
1640 int err;
1641
1642 err = parse_flow_nlattrs(key, a, &key_attrs);
1643 if (err)
1644 return err;
1645
1646 if ((key_attrs & (1 << OVS_KEY_ATTR_ETHERNET)) &&
1647 (key_attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) &&
1648 (nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]) == htons(ETH_P_8021Q))) {
1649 __be16 tci;
1650
1651 if (!((key_attrs & (1 << OVS_KEY_ATTR_VLAN)) &&
1652 (key_attrs & (1 << OVS_KEY_ATTR_ENCAP)))) {
1653 OVS_NLERR("Invalid Vlan frame.\n");
1654 return -EINVAL;
1655 }
1656
1657 key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1658 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1659 encap = a[OVS_KEY_ATTR_ENCAP];
1660 key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
1661 encap_valid = true;
1662
1663 if (tci & htons(VLAN_TAG_PRESENT)) {
1664 err = parse_flow_nlattrs(encap, a, &key_attrs);
1665 if (err)
1666 return err;
1667 } else if (!tci) {
1668 /* Corner case for truncated 802.1Q header. */
1669 if (nla_len(encap)) {
1670 OVS_NLERR("Truncated 802.1Q header has non-zero encap attribute.\n");
1671 return -EINVAL;
1672 }
1673 } else {
1674 OVS_NLERR("Encap attribute is set for a non-VLAN frame.\n");
1675 return -EINVAL;
1676 }
1677 }
1678
1679 err = ovs_key_from_nlattrs(match, key_attrs, a, false);
1680 if (err)
1681 return err;
1682
1683 if (mask) {
1684 err = parse_flow_mask_nlattrs(mask, a, &mask_attrs);
1685 if (err)
1686 return err;
1687
1688 if (mask_attrs & 1ULL << OVS_KEY_ATTR_ENCAP) {
1689 __be16 eth_type = 0;
1690 __be16 tci = 0;
1691
1692 if (!encap_valid) {
1693 OVS_NLERR("Encap mask attribute is set for non-VLAN frame.\n");
1694 return -EINVAL;
1695 }
1696
1697 mask_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
1698 if (a[OVS_KEY_ATTR_ETHERTYPE])
1699 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
1700
1701 if (eth_type == htons(0xffff)) {
1702 mask_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
1703 encap = a[OVS_KEY_ATTR_ENCAP];
1704 err = parse_flow_mask_nlattrs(encap, a, &mask_attrs);
1705 } else {
1706 OVS_NLERR("VLAN frames must have an exact match on the TPID (mask=%x).\n",
1707 ntohs(eth_type));
1708 return -EINVAL;
1709 }
1710
1711 if (a[OVS_KEY_ATTR_VLAN])
1712 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
1713
1714 if (!(tci & htons(VLAN_TAG_PRESENT))) {
1715 OVS_NLERR("VLAN tag present bit must have an exact match (tci_mask=%x).\n", ntohs(tci));
1716 return -EINVAL;
1717 }
1718 }
1719
1720 err = ovs_key_from_nlattrs(match, mask_attrs, a, true);
1721 if (err)
1722 return err;
1723 } else {
1724 /* Populate exact match flow's key mask. */
1725 if (match->mask)
1726 ovs_sw_flow_mask_set(match->mask, &match->range, 0xff);
1727 }
1728
1729 if (!ovs_match_validate(match, key_attrs, mask_attrs))
1730 return -EINVAL;
1731
1732 return 0;
1733}
1734
1735/**
1736 * ovs_flow_metadata_from_nlattrs - parses Netlink attributes into a flow key.
1737 * @flow: Receives extracted in_port, priority, tun_key and skb_mark.
1738 * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
1739 * sequence.
1740 *
1741 * This parses a series of Netlink attributes that form a flow key, which must
1742 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
1743 * get the metadata, that is, the parts of the flow key that cannot be
1744 * extracted from the packet itself.
1745 */
1746
1747int ovs_flow_metadata_from_nlattrs(struct sw_flow *flow,
1748 const struct nlattr *attr)
1749{
1750 struct ovs_key_ipv4_tunnel *tun_key = &flow->key.tun_key;
1751 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
1752 u64 attrs = 0;
1753 int err;
1754 struct sw_flow_match match;
1755
1756 flow->key.phy.in_port = DP_MAX_PORTS;
1757 flow->key.phy.priority = 0;
1758 flow->key.phy.skb_mark = 0;
1759 memset(tun_key, 0, sizeof(flow->key.tun_key));
1760
1761 err = parse_flow_nlattrs(attr, a, &attrs);
1762 if (err)
1763 return -EINVAL;
1764
1765 memset(&match, 0, sizeof(match));
1766 match.key = &flow->key;
1767
1768 err = metadata_from_nlattrs(&match, &attrs, a, false);
1769 if (err)
1770 return err;
1771
1772 return 0;
1773}
1774
1775int ovs_flow_to_nlattrs(const struct sw_flow_key *swkey,
1776 const struct sw_flow_key *output, struct sk_buff *skb)
1777{
1778 struct ovs_key_ethernet *eth_key;
1779 struct nlattr *nla, *encap;
1780 bool is_mask = (swkey != output);
1781
1782 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
1783 goto nla_put_failure;
1784
1785 if ((swkey->tun_key.ipv4_dst || is_mask) &&
1786 ovs_ipv4_tun_to_nlattr(skb, &swkey->tun_key, &output->tun_key))
1787 goto nla_put_failure;
1788
1789 if (swkey->phy.in_port == DP_MAX_PORTS) {
1790 if (is_mask && (output->phy.in_port == 0xffff))
1791 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
1792 goto nla_put_failure;
1793 } else {
1794 u16 upper_u16;
1795 upper_u16 = !is_mask ? 0 : 0xffff;
1796
1797 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
1798 (upper_u16 << 16) | output->phy.in_port))
1799 goto nla_put_failure;
1800 }
1801
1802 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
1803 goto nla_put_failure;
1804
1805 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
1806 if (!nla)
1807 goto nla_put_failure;
1808
1809 eth_key = nla_data(nla);
1810 memcpy(eth_key->eth_src, output->eth.src, ETH_ALEN);
1811 memcpy(eth_key->eth_dst, output->eth.dst, ETH_ALEN);
1812
1813 if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) {
1814 __be16 eth_type;
1815 eth_type = !is_mask ? htons(ETH_P_8021Q) : htons(0xffff);
1816 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
1817 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, output->eth.tci))
1818 goto nla_put_failure;
1819 encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
1820 if (!swkey->eth.tci)
1821 goto unencap;
1822 } else
1823 encap = NULL;
1824
1825 if (swkey->eth.type == htons(ETH_P_802_2)) {
1826 /*
1827 * Ethertype 802.2 is represented in the netlink with omitted
1828 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
1829 * 0xffff in the mask attribute. Ethertype can also
1830 * be wildcarded.
1831 */
1832 if (is_mask && output->eth.type)
1833 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
1834 output->eth.type))
1835 goto nla_put_failure;
1836 goto unencap;
1837 }
1838
1839 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
1840 goto nla_put_failure;
1841
1842 if (swkey->eth.type == htons(ETH_P_IP)) {
1843 struct ovs_key_ipv4 *ipv4_key;
1844
1845 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
1846 if (!nla)
1847 goto nla_put_failure;
1848 ipv4_key = nla_data(nla);
1849 ipv4_key->ipv4_src = output->ipv4.addr.src;
1850 ipv4_key->ipv4_dst = output->ipv4.addr.dst;
1851 ipv4_key->ipv4_proto = output->ip.proto;
1852 ipv4_key->ipv4_tos = output->ip.tos;
1853 ipv4_key->ipv4_ttl = output->ip.ttl;
1854 ipv4_key->ipv4_frag = output->ip.frag;
1855 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1856 struct ovs_key_ipv6 *ipv6_key;
1857
1858 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
1859 if (!nla)
1860 goto nla_put_failure;
1861 ipv6_key = nla_data(nla);
1862 memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
1863 sizeof(ipv6_key->ipv6_src));
1864 memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
1865 sizeof(ipv6_key->ipv6_dst));
1866 ipv6_key->ipv6_label = output->ipv6.label;
1867 ipv6_key->ipv6_proto = output->ip.proto;
1868 ipv6_key->ipv6_tclass = output->ip.tos;
1869 ipv6_key->ipv6_hlimit = output->ip.ttl;
1870 ipv6_key->ipv6_frag = output->ip.frag;
1871 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
1872 swkey->eth.type == htons(ETH_P_RARP)) {
1873 struct ovs_key_arp *arp_key;
1874
1875 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
1876 if (!nla)
1877 goto nla_put_failure;
1878 arp_key = nla_data(nla);
1879 memset(arp_key, 0, sizeof(struct ovs_key_arp));
1880 arp_key->arp_sip = output->ipv4.addr.src;
1881 arp_key->arp_tip = output->ipv4.addr.dst;
1882 arp_key->arp_op = htons(output->ip.proto);
1883 memcpy(arp_key->arp_sha, output->ipv4.arp.sha, ETH_ALEN);
1884 memcpy(arp_key->arp_tha, output->ipv4.arp.tha, ETH_ALEN);
1885 }
1886
1887 if ((swkey->eth.type == htons(ETH_P_IP) ||
1888 swkey->eth.type == htons(ETH_P_IPV6)) &&
1889 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1890
1891 if (swkey->ip.proto == IPPROTO_TCP) {
1892 struct ovs_key_tcp *tcp_key;
1893
1894 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1895 if (!nla)
1896 goto nla_put_failure;
1897 tcp_key = nla_data(nla);
1898 if (swkey->eth.type == htons(ETH_P_IP)) {
1899 tcp_key->tcp_src = output->ipv4.tp.src;
1900 tcp_key->tcp_dst = output->ipv4.tp.dst;
1901 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1902 tcp_key->tcp_src = output->ipv6.tp.src;
1903 tcp_key->tcp_dst = output->ipv6.tp.dst;
1904 }
1905 } else if (swkey->ip.proto == IPPROTO_UDP) {
1906 struct ovs_key_udp *udp_key;
1907
1908 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1909 if (!nla)
1910 goto nla_put_failure;
1911 udp_key = nla_data(nla);
1912 if (swkey->eth.type == htons(ETH_P_IP)) {
1913 udp_key->udp_src = output->ipv4.tp.src;
1914 udp_key->udp_dst = output->ipv4.tp.dst;
1915 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1916 udp_key->udp_src = output->ipv6.tp.src;
1917 udp_key->udp_dst = output->ipv6.tp.dst;
1918 }
1919 } else if (swkey->ip.proto == IPPROTO_SCTP) {
1920 struct ovs_key_sctp *sctp_key;
1921
1922 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
1923 if (!nla)
1924 goto nla_put_failure;
1925 sctp_key = nla_data(nla);
1926 if (swkey->eth.type == htons(ETH_P_IP)) {
1927 sctp_key->sctp_src = swkey->ipv4.tp.src;
1928 sctp_key->sctp_dst = swkey->ipv4.tp.dst;
1929 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1930 sctp_key->sctp_src = swkey->ipv6.tp.src;
1931 sctp_key->sctp_dst = swkey->ipv6.tp.dst;
1932 }
1933 } else if (swkey->eth.type == htons(ETH_P_IP) &&
1934 swkey->ip.proto == IPPROTO_ICMP) {
1935 struct ovs_key_icmp *icmp_key;
1936
1937 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1938 if (!nla)
1939 goto nla_put_failure;
1940 icmp_key = nla_data(nla);
1941 icmp_key->icmp_type = ntohs(output->ipv4.tp.src);
1942 icmp_key->icmp_code = ntohs(output->ipv4.tp.dst);
1943 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1944 swkey->ip.proto == IPPROTO_ICMPV6) {
1945 struct ovs_key_icmpv6 *icmpv6_key;
1946
1947 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1948 sizeof(*icmpv6_key));
1949 if (!nla)
1950 goto nla_put_failure;
1951 icmpv6_key = nla_data(nla);
1952 icmpv6_key->icmpv6_type = ntohs(output->ipv6.tp.src);
1953 icmpv6_key->icmpv6_code = ntohs(output->ipv6.tp.dst);
1954
1955 if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1956 icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1957 struct ovs_key_nd *nd_key;
1958
1959 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1960 if (!nla)
1961 goto nla_put_failure;
1962 nd_key = nla_data(nla);
1963 memcpy(nd_key->nd_target, &output->ipv6.nd.target,
1964 sizeof(nd_key->nd_target));
1965 memcpy(nd_key->nd_sll, output->ipv6.nd.sll, ETH_ALEN);
1966 memcpy(nd_key->nd_tll, output->ipv6.nd.tll, ETH_ALEN);
1967 }
1968 }
1969 }
1970
1971unencap:
1972 if (encap)
1973 nla_nest_end(skb, encap);
1974
1975 return 0;
1976
1977nla_put_failure:
1978 return -EMSGSIZE;
1979}
1980
1981/* Initializes the flow module.
1982 * Returns zero if successful or a negative error code. */
1983int ovs_flow_init(void)
1984{
1985 BUILD_BUG_ON(__alignof__(struct sw_flow_key) % __alignof__(long));
1986 BUILD_BUG_ON(sizeof(struct sw_flow_key) % sizeof(long));
1987
1988 flow_cache = kmem_cache_create("sw_flow", sizeof(struct sw_flow), 0,
1989 0, NULL);
1990 if (flow_cache == NULL)
1991 return -ENOMEM;
1992
1993 return 0;
1994}
1995
1996/* Uninitializes the flow module. */
1997void ovs_flow_exit(void)
1998{
1999 kmem_cache_destroy(flow_cache);
2000}
2001
2002struct sw_flow_mask *ovs_sw_flow_mask_alloc(void)
2003{
2004 struct sw_flow_mask *mask;
2005
2006 mask = kmalloc(sizeof(*mask), GFP_KERNEL);
2007 if (mask)
2008 mask->ref_count = 0;
2009
2010 return mask;
2011}
2012
2013void ovs_sw_flow_mask_add_ref(struct sw_flow_mask *mask)
2014{
2015 mask->ref_count++;
2016}
2017
2018void ovs_sw_flow_mask_del_ref(struct sw_flow_mask *mask, bool deferred)
2019{
2020 if (!mask)
2021 return;
2022
2023 BUG_ON(!mask->ref_count);
2024 mask->ref_count--;
2025
2026 if (!mask->ref_count) {
2027 list_del_rcu(&mask->list);
2028 if (deferred)
2029 kfree_rcu(mask, rcu);
2030 else
2031 kfree(mask);
2032 }
2033}
2034
2035static bool ovs_sw_flow_mask_equal(const struct sw_flow_mask *a,
2036 const struct sw_flow_mask *b)
2037{
2038 u8 *a_ = (u8 *)&a->key + a->range.start;
2039 u8 *b_ = (u8 *)&b->key + b->range.start;
2040
2041 return (a->range.end == b->range.end)
2042 && (a->range.start == b->range.start)
2043 && (memcmp(a_, b_, range_n_bytes(&a->range)) == 0);
2044}
2045
2046struct sw_flow_mask *ovs_sw_flow_mask_find(const struct flow_table *tbl,
2047 const struct sw_flow_mask *mask)
2048{
2049 struct list_head *ml;
2050
2051 list_for_each(ml, tbl->mask_list) {
2052 struct sw_flow_mask *m;
2053 m = container_of(ml, struct sw_flow_mask, list);
2054 if (ovs_sw_flow_mask_equal(mask, m))
2055 return m;
2056 }
2057
2058 return NULL;
2059}
2060
2061/**
2062 * add a new mask into the mask list.
2063 * The caller needs to make sure that 'mask' is not the same
2064 * as any masks that are already on the list.
2065 */
2066void ovs_sw_flow_mask_insert(struct flow_table *tbl, struct sw_flow_mask *mask)
2067{
2068 list_add_rcu(&mask->list, tbl->mask_list);
2069}
2070
2071/**
2072 * Set 'range' fields in the mask to the value of 'val'.
2073 */
2074static void ovs_sw_flow_mask_set(struct sw_flow_mask *mask,
2075 struct sw_flow_key_range *range, u8 val)
2076{
2077 u8 *m = (u8 *)&mask->key + range->start;
2078
2079 mask->range = *range;
2080 memset(m, val, range_n_bytes(range));
2081}
diff --git a/net/openvswitch/flow.h b/net/openvswitch/flow.h
index 212fbf7510c4..098fd1db6a23 100644
--- a/net/openvswitch/flow.h
+++ b/net/openvswitch/flow.h
@@ -33,14 +33,6 @@
33#include <net/inet_ecn.h> 33#include <net/inet_ecn.h>
34 34
35struct sk_buff; 35struct sk_buff;
36struct sw_flow_mask;
37struct flow_table;
38
39struct sw_flow_actions {
40 struct rcu_head rcu;
41 u32 actions_len;
42 struct nlattr actions[];
43};
44 36
45/* Used to memset ovs_key_ipv4_tunnel padding. */ 37/* Used to memset ovs_key_ipv4_tunnel padding. */
46#define OVS_TUNNEL_KEY_SIZE \ 38#define OVS_TUNNEL_KEY_SIZE \
@@ -127,6 +119,31 @@ struct sw_flow_key {
127 }; 119 };
128} __aligned(BITS_PER_LONG/8); /* Ensure that we can do comparisons as longs. */ 120} __aligned(BITS_PER_LONG/8); /* Ensure that we can do comparisons as longs. */
129 121
122struct sw_flow_key_range {
123 size_t start;
124 size_t end;
125};
126
127struct sw_flow_mask {
128 int ref_count;
129 struct rcu_head rcu;
130 struct list_head list;
131 struct sw_flow_key_range range;
132 struct sw_flow_key key;
133};
134
135struct sw_flow_match {
136 struct sw_flow_key *key;
137 struct sw_flow_key_range range;
138 struct sw_flow_mask *mask;
139};
140
141struct sw_flow_actions {
142 struct rcu_head rcu;
143 u32 actions_len;
144 struct nlattr actions[];
145};
146
130struct sw_flow { 147struct sw_flow {
131 struct rcu_head rcu; 148 struct rcu_head rcu;
132 struct hlist_node hash_node[2]; 149 struct hlist_node hash_node[2];
@@ -144,20 +161,6 @@ struct sw_flow {
144 u8 tcp_flags; /* Union of seen TCP flags. */ 161 u8 tcp_flags; /* Union of seen TCP flags. */
145}; 162};
146 163
147struct sw_flow_key_range {
148 size_t start;
149 size_t end;
150};
151
152struct sw_flow_match {
153 struct sw_flow_key *key;
154 struct sw_flow_key_range range;
155 struct sw_flow_mask *mask;
156};
157
158void ovs_match_init(struct sw_flow_match *match,
159 struct sw_flow_key *key, struct sw_flow_mask *mask);
160
161struct arp_eth_header { 164struct arp_eth_header {
162 __be16 ar_hrd; /* format of hardware address */ 165 __be16 ar_hrd; /* format of hardware address */
163 __be16 ar_pro; /* format of protocol address */ 166 __be16 ar_pro; /* format of protocol address */
@@ -172,88 +175,9 @@ struct arp_eth_header {
172 unsigned char ar_tip[4]; /* target IP address */ 175 unsigned char ar_tip[4]; /* target IP address */
173} __packed; 176} __packed;
174 177
175int ovs_flow_init(void);
176void ovs_flow_exit(void);
177
178struct sw_flow *ovs_flow_alloc(void);
179void ovs_flow_deferred_free(struct sw_flow *);
180void ovs_flow_free(struct sw_flow *, bool deferred);
181
182struct sw_flow_actions *ovs_flow_actions_alloc(int actions_len);
183void ovs_flow_deferred_free_acts(struct sw_flow_actions *);
184
185int ovs_flow_extract(struct sk_buff *, u16 in_port, struct sw_flow_key *);
186void ovs_flow_used(struct sw_flow *, struct sk_buff *); 178void ovs_flow_used(struct sw_flow *, struct sk_buff *);
187u64 ovs_flow_used_time(unsigned long flow_jiffies); 179u64 ovs_flow_used_time(unsigned long flow_jiffies);
188int ovs_flow_to_nlattrs(const struct sw_flow_key *,
189 const struct sw_flow_key *, struct sk_buff *);
190int ovs_match_from_nlattrs(struct sw_flow_match *match,
191 const struct nlattr *,
192 const struct nlattr *);
193int ovs_flow_metadata_from_nlattrs(struct sw_flow *flow,
194 const struct nlattr *attr);
195 180
196#define MAX_ACTIONS_BUFSIZE (32 * 1024) 181int ovs_flow_extract(struct sk_buff *, u16 in_port, struct sw_flow_key *);
197#define TBL_MIN_BUCKETS 1024
198
199struct flow_table {
200 struct flex_array *buckets;
201 unsigned int count, n_buckets;
202 struct rcu_head rcu;
203 struct list_head *mask_list;
204 int node_ver;
205 u32 hash_seed;
206 bool keep_flows;
207};
208
209static inline int ovs_flow_tbl_count(struct flow_table *table)
210{
211 return table->count;
212}
213
214static inline int ovs_flow_tbl_need_to_expand(struct flow_table *table)
215{
216 return (table->count > table->n_buckets);
217}
218
219struct sw_flow *ovs_flow_lookup(struct flow_table *,
220 const struct sw_flow_key *);
221struct sw_flow *ovs_flow_lookup_unmasked_key(struct flow_table *table,
222 struct sw_flow_match *match);
223
224void ovs_flow_tbl_destroy(struct flow_table *table, bool deferred);
225struct flow_table *ovs_flow_tbl_alloc(int new_size);
226struct flow_table *ovs_flow_tbl_expand(struct flow_table *table);
227struct flow_table *ovs_flow_tbl_rehash(struct flow_table *table);
228
229void ovs_flow_insert(struct flow_table *table, struct sw_flow *flow);
230void ovs_flow_remove(struct flow_table *table, struct sw_flow *flow);
231
232struct sw_flow *ovs_flow_dump_next(struct flow_table *table, u32 *bucket, u32 *idx);
233extern const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1];
234int ovs_ipv4_tun_from_nlattr(const struct nlattr *attr,
235 struct sw_flow_match *match, bool is_mask);
236int ovs_ipv4_tun_to_nlattr(struct sk_buff *skb,
237 const struct ovs_key_ipv4_tunnel *tun_key,
238 const struct ovs_key_ipv4_tunnel *output);
239
240bool ovs_flow_cmp_unmasked_key(const struct sw_flow *flow,
241 const struct sw_flow_key *key, int key_end);
242
243struct sw_flow_mask {
244 int ref_count;
245 struct rcu_head rcu;
246 struct list_head list;
247 struct sw_flow_key_range range;
248 struct sw_flow_key key;
249};
250 182
251struct sw_flow_mask *ovs_sw_flow_mask_alloc(void);
252void ovs_sw_flow_mask_add_ref(struct sw_flow_mask *);
253void ovs_sw_flow_mask_del_ref(struct sw_flow_mask *, bool deferred);
254void ovs_sw_flow_mask_insert(struct flow_table *, struct sw_flow_mask *);
255struct sw_flow_mask *ovs_sw_flow_mask_find(const struct flow_table *,
256 const struct sw_flow_mask *);
257void ovs_flow_key_mask(struct sw_flow_key *dst, const struct sw_flow_key *src,
258 const struct sw_flow_mask *mask);
259#endif /* flow.h */ 183#endif /* flow.h */
diff --git a/net/openvswitch/flow_netlink.c b/net/openvswitch/flow_netlink.c
new file mode 100644
index 000000000000..e04649c56a96
--- /dev/null
+++ b/net/openvswitch/flow_netlink.c
@@ -0,0 +1,1603 @@
1/*
2 * Copyright (c) 2007-2013 Nicira, Inc.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public License
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
16 * 02110-1301, USA
17 */
18
19#include "flow.h"
20#include "datapath.h"
21#include <linux/uaccess.h>
22#include <linux/netdevice.h>
23#include <linux/etherdevice.h>
24#include <linux/if_ether.h>
25#include <linux/if_vlan.h>
26#include <net/llc_pdu.h>
27#include <linux/kernel.h>
28#include <linux/jhash.h>
29#include <linux/jiffies.h>
30#include <linux/llc.h>
31#include <linux/module.h>
32#include <linux/in.h>
33#include <linux/rcupdate.h>
34#include <linux/if_arp.h>
35#include <linux/ip.h>
36#include <linux/ipv6.h>
37#include <linux/sctp.h>
38#include <linux/tcp.h>
39#include <linux/udp.h>
40#include <linux/icmp.h>
41#include <linux/icmpv6.h>
42#include <linux/rculist.h>
43#include <net/ip.h>
44#include <net/ipv6.h>
45#include <net/ndisc.h>
46
47#include "flow_netlink.h"
48
49static void update_range__(struct sw_flow_match *match,
50 size_t offset, size_t size, bool is_mask)
51{
52 struct sw_flow_key_range *range = NULL;
53 size_t start = rounddown(offset, sizeof(long));
54 size_t end = roundup(offset + size, sizeof(long));
55
56 if (!is_mask)
57 range = &match->range;
58 else if (match->mask)
59 range = &match->mask->range;
60
61 if (!range)
62 return;
63
64 if (range->start == range->end) {
65 range->start = start;
66 range->end = end;
67 return;
68 }
69
70 if (range->start > start)
71 range->start = start;
72
73 if (range->end < end)
74 range->end = end;
75}
76
77#define SW_FLOW_KEY_PUT(match, field, value, is_mask) \
78 do { \
79 update_range__(match, offsetof(struct sw_flow_key, field), \
80 sizeof((match)->key->field), is_mask); \
81 if (is_mask) { \
82 if ((match)->mask) \
83 (match)->mask->key.field = value; \
84 } else { \
85 (match)->key->field = value; \
86 } \
87 } while (0)
88
89#define SW_FLOW_KEY_MEMCPY(match, field, value_p, len, is_mask) \
90 do { \
91 update_range__(match, offsetof(struct sw_flow_key, field), \
92 len, is_mask); \
93 if (is_mask) { \
94 if ((match)->mask) \
95 memcpy(&(match)->mask->key.field, value_p, len);\
96 } else { \
97 memcpy(&(match)->key->field, value_p, len); \
98 } \
99 } while (0)
100
101static u16 range_n_bytes(const struct sw_flow_key_range *range)
102{
103 return range->end - range->start;
104}
105
106static bool match_validate(const struct sw_flow_match *match,
107 u64 key_attrs, u64 mask_attrs)
108{
109 u64 key_expected = 1 << OVS_KEY_ATTR_ETHERNET;
110 u64 mask_allowed = key_attrs; /* At most allow all key attributes */
111
112 /* The following mask attributes allowed only if they
113 * pass the validation tests. */
114 mask_allowed &= ~((1 << OVS_KEY_ATTR_IPV4)
115 | (1 << OVS_KEY_ATTR_IPV6)
116 | (1 << OVS_KEY_ATTR_TCP)
117 | (1 << OVS_KEY_ATTR_UDP)
118 | (1 << OVS_KEY_ATTR_SCTP)
119 | (1 << OVS_KEY_ATTR_ICMP)
120 | (1 << OVS_KEY_ATTR_ICMPV6)
121 | (1 << OVS_KEY_ATTR_ARP)
122 | (1 << OVS_KEY_ATTR_ND));
123
124 /* Always allowed mask fields. */
125 mask_allowed |= ((1 << OVS_KEY_ATTR_TUNNEL)
126 | (1 << OVS_KEY_ATTR_IN_PORT)
127 | (1 << OVS_KEY_ATTR_ETHERTYPE));
128
129 /* Check key attributes. */
130 if (match->key->eth.type == htons(ETH_P_ARP)
131 || match->key->eth.type == htons(ETH_P_RARP)) {
132 key_expected |= 1 << OVS_KEY_ATTR_ARP;
133 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
134 mask_allowed |= 1 << OVS_KEY_ATTR_ARP;
135 }
136
137 if (match->key->eth.type == htons(ETH_P_IP)) {
138 key_expected |= 1 << OVS_KEY_ATTR_IPV4;
139 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
140 mask_allowed |= 1 << OVS_KEY_ATTR_IPV4;
141
142 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
143 if (match->key->ip.proto == IPPROTO_UDP) {
144 key_expected |= 1 << OVS_KEY_ATTR_UDP;
145 if (match->mask && (match->mask->key.ip.proto == 0xff))
146 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
147 }
148
149 if (match->key->ip.proto == IPPROTO_SCTP) {
150 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
151 if (match->mask && (match->mask->key.ip.proto == 0xff))
152 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
153 }
154
155 if (match->key->ip.proto == IPPROTO_TCP) {
156 key_expected |= 1 << OVS_KEY_ATTR_TCP;
157 if (match->mask && (match->mask->key.ip.proto == 0xff))
158 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
159 }
160
161 if (match->key->ip.proto == IPPROTO_ICMP) {
162 key_expected |= 1 << OVS_KEY_ATTR_ICMP;
163 if (match->mask && (match->mask->key.ip.proto == 0xff))
164 mask_allowed |= 1 << OVS_KEY_ATTR_ICMP;
165 }
166 }
167 }
168
169 if (match->key->eth.type == htons(ETH_P_IPV6)) {
170 key_expected |= 1 << OVS_KEY_ATTR_IPV6;
171 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
172 mask_allowed |= 1 << OVS_KEY_ATTR_IPV6;
173
174 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
175 if (match->key->ip.proto == IPPROTO_UDP) {
176 key_expected |= 1 << OVS_KEY_ATTR_UDP;
177 if (match->mask && (match->mask->key.ip.proto == 0xff))
178 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
179 }
180
181 if (match->key->ip.proto == IPPROTO_SCTP) {
182 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
183 if (match->mask && (match->mask->key.ip.proto == 0xff))
184 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
185 }
186
187 if (match->key->ip.proto == IPPROTO_TCP) {
188 key_expected |= 1 << OVS_KEY_ATTR_TCP;
189 if (match->mask && (match->mask->key.ip.proto == 0xff))
190 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
191 }
192
193 if (match->key->ip.proto == IPPROTO_ICMPV6) {
194 key_expected |= 1 << OVS_KEY_ATTR_ICMPV6;
195 if (match->mask && (match->mask->key.ip.proto == 0xff))
196 mask_allowed |= 1 << OVS_KEY_ATTR_ICMPV6;
197
198 if (match->key->ipv6.tp.src ==
199 htons(NDISC_NEIGHBOUR_SOLICITATION) ||
200 match->key->ipv6.tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
201 key_expected |= 1 << OVS_KEY_ATTR_ND;
202 if (match->mask && (match->mask->key.ipv6.tp.src == htons(0xffff)))
203 mask_allowed |= 1 << OVS_KEY_ATTR_ND;
204 }
205 }
206 }
207 }
208
209 if ((key_attrs & key_expected) != key_expected) {
210 /* Key attributes check failed. */
211 OVS_NLERR("Missing expected key attributes (key_attrs=%llx, expected=%llx).\n",
212 key_attrs, key_expected);
213 return false;
214 }
215
216 if ((mask_attrs & mask_allowed) != mask_attrs) {
217 /* Mask attributes check failed. */
218 OVS_NLERR("Contain more than allowed mask fields (mask_attrs=%llx, mask_allowed=%llx).\n",
219 mask_attrs, mask_allowed);
220 return false;
221 }
222
223 return true;
224}
225
226/* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
227static const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
228 [OVS_KEY_ATTR_ENCAP] = -1,
229 [OVS_KEY_ATTR_PRIORITY] = sizeof(u32),
230 [OVS_KEY_ATTR_IN_PORT] = sizeof(u32),
231 [OVS_KEY_ATTR_SKB_MARK] = sizeof(u32),
232 [OVS_KEY_ATTR_ETHERNET] = sizeof(struct ovs_key_ethernet),
233 [OVS_KEY_ATTR_VLAN] = sizeof(__be16),
234 [OVS_KEY_ATTR_ETHERTYPE] = sizeof(__be16),
235 [OVS_KEY_ATTR_IPV4] = sizeof(struct ovs_key_ipv4),
236 [OVS_KEY_ATTR_IPV6] = sizeof(struct ovs_key_ipv6),
237 [OVS_KEY_ATTR_TCP] = sizeof(struct ovs_key_tcp),
238 [OVS_KEY_ATTR_UDP] = sizeof(struct ovs_key_udp),
239 [OVS_KEY_ATTR_SCTP] = sizeof(struct ovs_key_sctp),
240 [OVS_KEY_ATTR_ICMP] = sizeof(struct ovs_key_icmp),
241 [OVS_KEY_ATTR_ICMPV6] = sizeof(struct ovs_key_icmpv6),
242 [OVS_KEY_ATTR_ARP] = sizeof(struct ovs_key_arp),
243 [OVS_KEY_ATTR_ND] = sizeof(struct ovs_key_nd),
244 [OVS_KEY_ATTR_TUNNEL] = -1,
245};
246
247static bool is_all_zero(const u8 *fp, size_t size)
248{
249 int i;
250
251 if (!fp)
252 return false;
253
254 for (i = 0; i < size; i++)
255 if (fp[i])
256 return false;
257
258 return true;
259}
260
261static int __parse_flow_nlattrs(const struct nlattr *attr,
262 const struct nlattr *a[],
263 u64 *attrsp, bool nz)
264{
265 const struct nlattr *nla;
266 u64 attrs;
267 int rem;
268
269 attrs = *attrsp;
270 nla_for_each_nested(nla, attr, rem) {
271 u16 type = nla_type(nla);
272 int expected_len;
273
274 if (type > OVS_KEY_ATTR_MAX) {
275 OVS_NLERR("Unknown key attribute (type=%d, max=%d).\n",
276 type, OVS_KEY_ATTR_MAX);
277 return -EINVAL;
278 }
279
280 if (attrs & (1 << type)) {
281 OVS_NLERR("Duplicate key attribute (type %d).\n", type);
282 return -EINVAL;
283 }
284
285 expected_len = ovs_key_lens[type];
286 if (nla_len(nla) != expected_len && expected_len != -1) {
287 OVS_NLERR("Key attribute has unexpected length (type=%d"
288 ", length=%d, expected=%d).\n", type,
289 nla_len(nla), expected_len);
290 return -EINVAL;
291 }
292
293 if (!nz || !is_all_zero(nla_data(nla), expected_len)) {
294 attrs |= 1 << type;
295 a[type] = nla;
296 }
297 }
298 if (rem) {
299 OVS_NLERR("Message has %d unknown bytes.\n", rem);
300 return -EINVAL;
301 }
302
303 *attrsp = attrs;
304 return 0;
305}
306
307static int parse_flow_mask_nlattrs(const struct nlattr *attr,
308 const struct nlattr *a[], u64 *attrsp)
309{
310 return __parse_flow_nlattrs(attr, a, attrsp, true);
311}
312
313static int parse_flow_nlattrs(const struct nlattr *attr,
314 const struct nlattr *a[], u64 *attrsp)
315{
316 return __parse_flow_nlattrs(attr, a, attrsp, false);
317}
318
319static int ipv4_tun_from_nlattr(const struct nlattr *attr,
320 struct sw_flow_match *match, bool is_mask)
321{
322 struct nlattr *a;
323 int rem;
324 bool ttl = false;
325 __be16 tun_flags = 0;
326
327 nla_for_each_nested(a, attr, rem) {
328 int type = nla_type(a);
329 static const u32 ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
330 [OVS_TUNNEL_KEY_ATTR_ID] = sizeof(u64),
331 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = sizeof(u32),
332 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = sizeof(u32),
333 [OVS_TUNNEL_KEY_ATTR_TOS] = 1,
334 [OVS_TUNNEL_KEY_ATTR_TTL] = 1,
335 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = 0,
336 [OVS_TUNNEL_KEY_ATTR_CSUM] = 0,
337 };
338
339 if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
340 OVS_NLERR("Unknown IPv4 tunnel attribute (type=%d, max=%d).\n",
341 type, OVS_TUNNEL_KEY_ATTR_MAX);
342 return -EINVAL;
343 }
344
345 if (ovs_tunnel_key_lens[type] != nla_len(a)) {
346 OVS_NLERR("IPv4 tunnel attribute type has unexpected "
347 " length (type=%d, length=%d, expected=%d).\n",
348 type, nla_len(a), ovs_tunnel_key_lens[type]);
349 return -EINVAL;
350 }
351
352 switch (type) {
353 case OVS_TUNNEL_KEY_ATTR_ID:
354 SW_FLOW_KEY_PUT(match, tun_key.tun_id,
355 nla_get_be64(a), is_mask);
356 tun_flags |= TUNNEL_KEY;
357 break;
358 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
359 SW_FLOW_KEY_PUT(match, tun_key.ipv4_src,
360 nla_get_be32(a), is_mask);
361 break;
362 case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
363 SW_FLOW_KEY_PUT(match, tun_key.ipv4_dst,
364 nla_get_be32(a), is_mask);
365 break;
366 case OVS_TUNNEL_KEY_ATTR_TOS:
367 SW_FLOW_KEY_PUT(match, tun_key.ipv4_tos,
368 nla_get_u8(a), is_mask);
369 break;
370 case OVS_TUNNEL_KEY_ATTR_TTL:
371 SW_FLOW_KEY_PUT(match, tun_key.ipv4_ttl,
372 nla_get_u8(a), is_mask);
373 ttl = true;
374 break;
375 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
376 tun_flags |= TUNNEL_DONT_FRAGMENT;
377 break;
378 case OVS_TUNNEL_KEY_ATTR_CSUM:
379 tun_flags |= TUNNEL_CSUM;
380 break;
381 default:
382 return -EINVAL;
383 }
384 }
385
386 SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
387
388 if (rem > 0) {
389 OVS_NLERR("IPv4 tunnel attribute has %d unknown bytes.\n", rem);
390 return -EINVAL;
391 }
392
393 if (!is_mask) {
394 if (!match->key->tun_key.ipv4_dst) {
395 OVS_NLERR("IPv4 tunnel destination address is zero.\n");
396 return -EINVAL;
397 }
398
399 if (!ttl) {
400 OVS_NLERR("IPv4 tunnel TTL not specified.\n");
401 return -EINVAL;
402 }
403 }
404
405 return 0;
406}
407
408static int ipv4_tun_to_nlattr(struct sk_buff *skb,
409 const struct ovs_key_ipv4_tunnel *tun_key,
410 const struct ovs_key_ipv4_tunnel *output)
411{
412 struct nlattr *nla;
413
414 nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
415 if (!nla)
416 return -EMSGSIZE;
417
418 if (output->tun_flags & TUNNEL_KEY &&
419 nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id))
420 return -EMSGSIZE;
421 if (output->ipv4_src &&
422 nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC, output->ipv4_src))
423 return -EMSGSIZE;
424 if (output->ipv4_dst &&
425 nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST, output->ipv4_dst))
426 return -EMSGSIZE;
427 if (output->ipv4_tos &&
428 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->ipv4_tos))
429 return -EMSGSIZE;
430 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ipv4_ttl))
431 return -EMSGSIZE;
432 if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
433 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
434 return -EMSGSIZE;
435 if ((output->tun_flags & TUNNEL_CSUM) &&
436 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
437 return -EMSGSIZE;
438
439 nla_nest_end(skb, nla);
440 return 0;
441}
442
443
444static int metadata_from_nlattrs(struct sw_flow_match *match, u64 *attrs,
445 const struct nlattr **a, bool is_mask)
446{
447 if (*attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
448 SW_FLOW_KEY_PUT(match, phy.priority,
449 nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
450 *attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
451 }
452
453 if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
454 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
455
456 if (is_mask)
457 in_port = 0xffffffff; /* Always exact match in_port. */
458 else if (in_port >= DP_MAX_PORTS)
459 return -EINVAL;
460
461 SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
462 *attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
463 } else if (!is_mask) {
464 SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
465 }
466
467 if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
468 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
469
470 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
471 *attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
472 }
473 if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
474 if (ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
475 is_mask))
476 return -EINVAL;
477 *attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
478 }
479 return 0;
480}
481
482static int ovs_key_from_nlattrs(struct sw_flow_match *match, u64 attrs,
483 const struct nlattr **a, bool is_mask)
484{
485 int err;
486 u64 orig_attrs = attrs;
487
488 err = metadata_from_nlattrs(match, &attrs, a, is_mask);
489 if (err)
490 return err;
491
492 if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
493 const struct ovs_key_ethernet *eth_key;
494
495 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
496 SW_FLOW_KEY_MEMCPY(match, eth.src,
497 eth_key->eth_src, ETH_ALEN, is_mask);
498 SW_FLOW_KEY_MEMCPY(match, eth.dst,
499 eth_key->eth_dst, ETH_ALEN, is_mask);
500 attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
501 }
502
503 if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
504 __be16 tci;
505
506 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
507 if (!(tci & htons(VLAN_TAG_PRESENT))) {
508 if (is_mask)
509 OVS_NLERR("VLAN TCI mask does not have exact match for VLAN_TAG_PRESENT bit.\n");
510 else
511 OVS_NLERR("VLAN TCI does not have VLAN_TAG_PRESENT bit set.\n");
512
513 return -EINVAL;
514 }
515
516 SW_FLOW_KEY_PUT(match, eth.tci, tci, is_mask);
517 attrs &= ~(1 << OVS_KEY_ATTR_VLAN);
518 } else if (!is_mask)
519 SW_FLOW_KEY_PUT(match, eth.tci, htons(0xffff), true);
520
521 if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
522 __be16 eth_type;
523
524 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
525 if (is_mask) {
526 /* Always exact match EtherType. */
527 eth_type = htons(0xffff);
528 } else if (ntohs(eth_type) < ETH_P_802_3_MIN) {
529 OVS_NLERR("EtherType is less than minimum (type=%x, min=%x).\n",
530 ntohs(eth_type), ETH_P_802_3_MIN);
531 return -EINVAL;
532 }
533
534 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
535 attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
536 } else if (!is_mask) {
537 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
538 }
539
540 if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
541 const struct ovs_key_ipv4 *ipv4_key;
542
543 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
544 if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
545 OVS_NLERR("Unknown IPv4 fragment type (value=%d, max=%d).\n",
546 ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
547 return -EINVAL;
548 }
549 SW_FLOW_KEY_PUT(match, ip.proto,
550 ipv4_key->ipv4_proto, is_mask);
551 SW_FLOW_KEY_PUT(match, ip.tos,
552 ipv4_key->ipv4_tos, is_mask);
553 SW_FLOW_KEY_PUT(match, ip.ttl,
554 ipv4_key->ipv4_ttl, is_mask);
555 SW_FLOW_KEY_PUT(match, ip.frag,
556 ipv4_key->ipv4_frag, is_mask);
557 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
558 ipv4_key->ipv4_src, is_mask);
559 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
560 ipv4_key->ipv4_dst, is_mask);
561 attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
562 }
563
564 if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
565 const struct ovs_key_ipv6 *ipv6_key;
566
567 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
568 if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
569 OVS_NLERR("Unknown IPv6 fragment type (value=%d, max=%d).\n",
570 ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
571 return -EINVAL;
572 }
573 SW_FLOW_KEY_PUT(match, ipv6.label,
574 ipv6_key->ipv6_label, is_mask);
575 SW_FLOW_KEY_PUT(match, ip.proto,
576 ipv6_key->ipv6_proto, is_mask);
577 SW_FLOW_KEY_PUT(match, ip.tos,
578 ipv6_key->ipv6_tclass, is_mask);
579 SW_FLOW_KEY_PUT(match, ip.ttl,
580 ipv6_key->ipv6_hlimit, is_mask);
581 SW_FLOW_KEY_PUT(match, ip.frag,
582 ipv6_key->ipv6_frag, is_mask);
583 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
584 ipv6_key->ipv6_src,
585 sizeof(match->key->ipv6.addr.src),
586 is_mask);
587 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
588 ipv6_key->ipv6_dst,
589 sizeof(match->key->ipv6.addr.dst),
590 is_mask);
591
592 attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
593 }
594
595 if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
596 const struct ovs_key_arp *arp_key;
597
598 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
599 if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
600 OVS_NLERR("Unknown ARP opcode (opcode=%d).\n",
601 arp_key->arp_op);
602 return -EINVAL;
603 }
604
605 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
606 arp_key->arp_sip, is_mask);
607 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
608 arp_key->arp_tip, is_mask);
609 SW_FLOW_KEY_PUT(match, ip.proto,
610 ntohs(arp_key->arp_op), is_mask);
611 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
612 arp_key->arp_sha, ETH_ALEN, is_mask);
613 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
614 arp_key->arp_tha, ETH_ALEN, is_mask);
615
616 attrs &= ~(1 << OVS_KEY_ATTR_ARP);
617 }
618
619 if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
620 const struct ovs_key_tcp *tcp_key;
621
622 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
623 if (orig_attrs & (1 << OVS_KEY_ATTR_IPV4)) {
624 SW_FLOW_KEY_PUT(match, ipv4.tp.src,
625 tcp_key->tcp_src, is_mask);
626 SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
627 tcp_key->tcp_dst, is_mask);
628 } else {
629 SW_FLOW_KEY_PUT(match, ipv6.tp.src,
630 tcp_key->tcp_src, is_mask);
631 SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
632 tcp_key->tcp_dst, is_mask);
633 }
634 attrs &= ~(1 << OVS_KEY_ATTR_TCP);
635 }
636
637 if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
638 const struct ovs_key_udp *udp_key;
639
640 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
641 if (orig_attrs & (1 << OVS_KEY_ATTR_IPV4)) {
642 SW_FLOW_KEY_PUT(match, ipv4.tp.src,
643 udp_key->udp_src, is_mask);
644 SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
645 udp_key->udp_dst, is_mask);
646 } else {
647 SW_FLOW_KEY_PUT(match, ipv6.tp.src,
648 udp_key->udp_src, is_mask);
649 SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
650 udp_key->udp_dst, is_mask);
651 }
652 attrs &= ~(1 << OVS_KEY_ATTR_UDP);
653 }
654
655 if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
656 const struct ovs_key_sctp *sctp_key;
657
658 sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
659 if (orig_attrs & (1 << OVS_KEY_ATTR_IPV4)) {
660 SW_FLOW_KEY_PUT(match, ipv4.tp.src,
661 sctp_key->sctp_src, is_mask);
662 SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
663 sctp_key->sctp_dst, is_mask);
664 } else {
665 SW_FLOW_KEY_PUT(match, ipv6.tp.src,
666 sctp_key->sctp_src, is_mask);
667 SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
668 sctp_key->sctp_dst, is_mask);
669 }
670 attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
671 }
672
673 if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
674 const struct ovs_key_icmp *icmp_key;
675
676 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
677 SW_FLOW_KEY_PUT(match, ipv4.tp.src,
678 htons(icmp_key->icmp_type), is_mask);
679 SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
680 htons(icmp_key->icmp_code), is_mask);
681 attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
682 }
683
684 if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
685 const struct ovs_key_icmpv6 *icmpv6_key;
686
687 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
688 SW_FLOW_KEY_PUT(match, ipv6.tp.src,
689 htons(icmpv6_key->icmpv6_type), is_mask);
690 SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
691 htons(icmpv6_key->icmpv6_code), is_mask);
692 attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
693 }
694
695 if (attrs & (1 << OVS_KEY_ATTR_ND)) {
696 const struct ovs_key_nd *nd_key;
697
698 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
699 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
700 nd_key->nd_target,
701 sizeof(match->key->ipv6.nd.target),
702 is_mask);
703 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
704 nd_key->nd_sll, ETH_ALEN, is_mask);
705 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
706 nd_key->nd_tll, ETH_ALEN, is_mask);
707 attrs &= ~(1 << OVS_KEY_ATTR_ND);
708 }
709
710 if (attrs != 0)
711 return -EINVAL;
712
713 return 0;
714}
715
716static void sw_flow_mask_set(struct sw_flow_mask *mask,
717 struct sw_flow_key_range *range, u8 val)
718{
719 u8 *m = (u8 *)&mask->key + range->start;
720
721 mask->range = *range;
722 memset(m, val, range_n_bytes(range));
723}
724
725/**
726 * ovs_nla_get_match - parses Netlink attributes into a flow key and
727 * mask. In case the 'mask' is NULL, the flow is treated as exact match
728 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
729 * does not include any don't care bit.
730 * @match: receives the extracted flow match information.
731 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
732 * sequence. The fields should of the packet that triggered the creation
733 * of this flow.
734 * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
735 * attribute specifies the mask field of the wildcarded flow.
736 */
737int ovs_nla_get_match(struct sw_flow_match *match,
738 const struct nlattr *key,
739 const struct nlattr *mask)
740{
741 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
742 const struct nlattr *encap;
743 u64 key_attrs = 0;
744 u64 mask_attrs = 0;
745 bool encap_valid = false;
746 int err;
747
748 err = parse_flow_nlattrs(key, a, &key_attrs);
749 if (err)
750 return err;
751
752 if ((key_attrs & (1 << OVS_KEY_ATTR_ETHERNET)) &&
753 (key_attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) &&
754 (nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]) == htons(ETH_P_8021Q))) {
755 __be16 tci;
756
757 if (!((key_attrs & (1 << OVS_KEY_ATTR_VLAN)) &&
758 (key_attrs & (1 << OVS_KEY_ATTR_ENCAP)))) {
759 OVS_NLERR("Invalid Vlan frame.\n");
760 return -EINVAL;
761 }
762
763 key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
764 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
765 encap = a[OVS_KEY_ATTR_ENCAP];
766 key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
767 encap_valid = true;
768
769 if (tci & htons(VLAN_TAG_PRESENT)) {
770 err = parse_flow_nlattrs(encap, a, &key_attrs);
771 if (err)
772 return err;
773 } else if (!tci) {
774 /* Corner case for truncated 802.1Q header. */
775 if (nla_len(encap)) {
776 OVS_NLERR("Truncated 802.1Q header has non-zero encap attribute.\n");
777 return -EINVAL;
778 }
779 } else {
780 OVS_NLERR("Encap attribute is set for a non-VLAN frame.\n");
781 return -EINVAL;
782 }
783 }
784
785 err = ovs_key_from_nlattrs(match, key_attrs, a, false);
786 if (err)
787 return err;
788
789 if (mask) {
790 err = parse_flow_mask_nlattrs(mask, a, &mask_attrs);
791 if (err)
792 return err;
793
794 if (mask_attrs & 1 << OVS_KEY_ATTR_ENCAP) {
795 __be16 eth_type = 0;
796 __be16 tci = 0;
797
798 if (!encap_valid) {
799 OVS_NLERR("Encap mask attribute is set for non-VLAN frame.\n");
800 return -EINVAL;
801 }
802
803 mask_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
804 if (a[OVS_KEY_ATTR_ETHERTYPE])
805 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
806
807 if (eth_type == htons(0xffff)) {
808 mask_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
809 encap = a[OVS_KEY_ATTR_ENCAP];
810 err = parse_flow_mask_nlattrs(encap, a, &mask_attrs);
811 } else {
812 OVS_NLERR("VLAN frames must have an exact match on the TPID (mask=%x).\n",
813 ntohs(eth_type));
814 return -EINVAL;
815 }
816
817 if (a[OVS_KEY_ATTR_VLAN])
818 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
819
820 if (!(tci & htons(VLAN_TAG_PRESENT))) {
821 OVS_NLERR("VLAN tag present bit must have an exact match (tci_mask=%x).\n", ntohs(tci));
822 return -EINVAL;
823 }
824 }
825
826 err = ovs_key_from_nlattrs(match, mask_attrs, a, true);
827 if (err)
828 return err;
829 } else {
830 /* Populate exact match flow's key mask. */
831 if (match->mask)
832 sw_flow_mask_set(match->mask, &match->range, 0xff);
833 }
834
835 if (!match_validate(match, key_attrs, mask_attrs))
836 return -EINVAL;
837
838 return 0;
839}
840
841/**
842 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
843 * @flow: Receives extracted in_port, priority, tun_key and skb_mark.
844 * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
845 * sequence.
846 *
847 * This parses a series of Netlink attributes that form a flow key, which must
848 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
849 * get the metadata, that is, the parts of the flow key that cannot be
850 * extracted from the packet itself.
851 */
852
853int ovs_nla_get_flow_metadata(struct sw_flow *flow,
854 const struct nlattr *attr)
855{
856 struct ovs_key_ipv4_tunnel *tun_key = &flow->key.tun_key;
857 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
858 u64 attrs = 0;
859 int err;
860 struct sw_flow_match match;
861
862 flow->key.phy.in_port = DP_MAX_PORTS;
863 flow->key.phy.priority = 0;
864 flow->key.phy.skb_mark = 0;
865 memset(tun_key, 0, sizeof(flow->key.tun_key));
866
867 err = parse_flow_nlattrs(attr, a, &attrs);
868 if (err)
869 return -EINVAL;
870
871 memset(&match, 0, sizeof(match));
872 match.key = &flow->key;
873
874 err = metadata_from_nlattrs(&match, &attrs, a, false);
875 if (err)
876 return err;
877
878 return 0;
879}
880
881int ovs_nla_put_flow(const struct sw_flow_key *swkey,
882 const struct sw_flow_key *output, struct sk_buff *skb)
883{
884 struct ovs_key_ethernet *eth_key;
885 struct nlattr *nla, *encap;
886 bool is_mask = (swkey != output);
887
888 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
889 goto nla_put_failure;
890
891 if ((swkey->tun_key.ipv4_dst || is_mask) &&
892 ipv4_tun_to_nlattr(skb, &swkey->tun_key, &output->tun_key))
893 goto nla_put_failure;
894
895 if (swkey->phy.in_port == DP_MAX_PORTS) {
896 if (is_mask && (output->phy.in_port == 0xffff))
897 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
898 goto nla_put_failure;
899 } else {
900 u16 upper_u16;
901 upper_u16 = !is_mask ? 0 : 0xffff;
902
903 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
904 (upper_u16 << 16) | output->phy.in_port))
905 goto nla_put_failure;
906 }
907
908 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
909 goto nla_put_failure;
910
911 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
912 if (!nla)
913 goto nla_put_failure;
914
915 eth_key = nla_data(nla);
916 memcpy(eth_key->eth_src, output->eth.src, ETH_ALEN);
917 memcpy(eth_key->eth_dst, output->eth.dst, ETH_ALEN);
918
919 if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) {
920 __be16 eth_type;
921 eth_type = !is_mask ? htons(ETH_P_8021Q) : htons(0xffff);
922 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
923 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, output->eth.tci))
924 goto nla_put_failure;
925 encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
926 if (!swkey->eth.tci)
927 goto unencap;
928 } else
929 encap = NULL;
930
931 if (swkey->eth.type == htons(ETH_P_802_2)) {
932 /*
933 * Ethertype 802.2 is represented in the netlink with omitted
934 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
935 * 0xffff in the mask attribute. Ethertype can also
936 * be wildcarded.
937 */
938 if (is_mask && output->eth.type)
939 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
940 output->eth.type))
941 goto nla_put_failure;
942 goto unencap;
943 }
944
945 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
946 goto nla_put_failure;
947
948 if (swkey->eth.type == htons(ETH_P_IP)) {
949 struct ovs_key_ipv4 *ipv4_key;
950
951 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
952 if (!nla)
953 goto nla_put_failure;
954 ipv4_key = nla_data(nla);
955 ipv4_key->ipv4_src = output->ipv4.addr.src;
956 ipv4_key->ipv4_dst = output->ipv4.addr.dst;
957 ipv4_key->ipv4_proto = output->ip.proto;
958 ipv4_key->ipv4_tos = output->ip.tos;
959 ipv4_key->ipv4_ttl = output->ip.ttl;
960 ipv4_key->ipv4_frag = output->ip.frag;
961 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
962 struct ovs_key_ipv6 *ipv6_key;
963
964 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
965 if (!nla)
966 goto nla_put_failure;
967 ipv6_key = nla_data(nla);
968 memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
969 sizeof(ipv6_key->ipv6_src));
970 memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
971 sizeof(ipv6_key->ipv6_dst));
972 ipv6_key->ipv6_label = output->ipv6.label;
973 ipv6_key->ipv6_proto = output->ip.proto;
974 ipv6_key->ipv6_tclass = output->ip.tos;
975 ipv6_key->ipv6_hlimit = output->ip.ttl;
976 ipv6_key->ipv6_frag = output->ip.frag;
977 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
978 swkey->eth.type == htons(ETH_P_RARP)) {
979 struct ovs_key_arp *arp_key;
980
981 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
982 if (!nla)
983 goto nla_put_failure;
984 arp_key = nla_data(nla);
985 memset(arp_key, 0, sizeof(struct ovs_key_arp));
986 arp_key->arp_sip = output->ipv4.addr.src;
987 arp_key->arp_tip = output->ipv4.addr.dst;
988 arp_key->arp_op = htons(output->ip.proto);
989 memcpy(arp_key->arp_sha, output->ipv4.arp.sha, ETH_ALEN);
990 memcpy(arp_key->arp_tha, output->ipv4.arp.tha, ETH_ALEN);
991 }
992
993 if ((swkey->eth.type == htons(ETH_P_IP) ||
994 swkey->eth.type == htons(ETH_P_IPV6)) &&
995 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
996
997 if (swkey->ip.proto == IPPROTO_TCP) {
998 struct ovs_key_tcp *tcp_key;
999
1000 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1001 if (!nla)
1002 goto nla_put_failure;
1003 tcp_key = nla_data(nla);
1004 if (swkey->eth.type == htons(ETH_P_IP)) {
1005 tcp_key->tcp_src = output->ipv4.tp.src;
1006 tcp_key->tcp_dst = output->ipv4.tp.dst;
1007 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1008 tcp_key->tcp_src = output->ipv6.tp.src;
1009 tcp_key->tcp_dst = output->ipv6.tp.dst;
1010 }
1011 } else if (swkey->ip.proto == IPPROTO_UDP) {
1012 struct ovs_key_udp *udp_key;
1013
1014 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1015 if (!nla)
1016 goto nla_put_failure;
1017 udp_key = nla_data(nla);
1018 if (swkey->eth.type == htons(ETH_P_IP)) {
1019 udp_key->udp_src = output->ipv4.tp.src;
1020 udp_key->udp_dst = output->ipv4.tp.dst;
1021 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1022 udp_key->udp_src = output->ipv6.tp.src;
1023 udp_key->udp_dst = output->ipv6.tp.dst;
1024 }
1025 } else if (swkey->ip.proto == IPPROTO_SCTP) {
1026 struct ovs_key_sctp *sctp_key;
1027
1028 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
1029 if (!nla)
1030 goto nla_put_failure;
1031 sctp_key = nla_data(nla);
1032 if (swkey->eth.type == htons(ETH_P_IP)) {
1033 sctp_key->sctp_src = swkey->ipv4.tp.src;
1034 sctp_key->sctp_dst = swkey->ipv4.tp.dst;
1035 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1036 sctp_key->sctp_src = swkey->ipv6.tp.src;
1037 sctp_key->sctp_dst = swkey->ipv6.tp.dst;
1038 }
1039 } else if (swkey->eth.type == htons(ETH_P_IP) &&
1040 swkey->ip.proto == IPPROTO_ICMP) {
1041 struct ovs_key_icmp *icmp_key;
1042
1043 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1044 if (!nla)
1045 goto nla_put_failure;
1046 icmp_key = nla_data(nla);
1047 icmp_key->icmp_type = ntohs(output->ipv4.tp.src);
1048 icmp_key->icmp_code = ntohs(output->ipv4.tp.dst);
1049 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1050 swkey->ip.proto == IPPROTO_ICMPV6) {
1051 struct ovs_key_icmpv6 *icmpv6_key;
1052
1053 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1054 sizeof(*icmpv6_key));
1055 if (!nla)
1056 goto nla_put_failure;
1057 icmpv6_key = nla_data(nla);
1058 icmpv6_key->icmpv6_type = ntohs(output->ipv6.tp.src);
1059 icmpv6_key->icmpv6_code = ntohs(output->ipv6.tp.dst);
1060
1061 if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1062 icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1063 struct ovs_key_nd *nd_key;
1064
1065 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1066 if (!nla)
1067 goto nla_put_failure;
1068 nd_key = nla_data(nla);
1069 memcpy(nd_key->nd_target, &output->ipv6.nd.target,
1070 sizeof(nd_key->nd_target));
1071 memcpy(nd_key->nd_sll, output->ipv6.nd.sll, ETH_ALEN);
1072 memcpy(nd_key->nd_tll, output->ipv6.nd.tll, ETH_ALEN);
1073 }
1074 }
1075 }
1076
1077unencap:
1078 if (encap)
1079 nla_nest_end(skb, encap);
1080
1081 return 0;
1082
1083nla_put_failure:
1084 return -EMSGSIZE;
1085}
1086
1087#define MAX_ACTIONS_BUFSIZE (32 * 1024)
1088
1089struct sw_flow_actions *ovs_nla_alloc_flow_actions(int size)
1090{
1091 struct sw_flow_actions *sfa;
1092
1093 if (size > MAX_ACTIONS_BUFSIZE)
1094 return ERR_PTR(-EINVAL);
1095
1096 sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
1097 if (!sfa)
1098 return ERR_PTR(-ENOMEM);
1099
1100 sfa->actions_len = 0;
1101 return sfa;
1102}
1103
1104/* RCU callback used by ovs_nla_free_flow_actions. */
1105static void rcu_free_acts_callback(struct rcu_head *rcu)
1106{
1107 struct sw_flow_actions *sf_acts = container_of(rcu,
1108 struct sw_flow_actions, rcu);
1109 kfree(sf_acts);
1110}
1111
1112/* Schedules 'sf_acts' to be freed after the next RCU grace period.
1113 * The caller must hold rcu_read_lock for this to be sensible. */
1114void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
1115{
1116 call_rcu(&sf_acts->rcu, rcu_free_acts_callback);
1117}
1118
1119static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
1120 int attr_len)
1121{
1122
1123 struct sw_flow_actions *acts;
1124 int new_acts_size;
1125 int req_size = NLA_ALIGN(attr_len);
1126 int next_offset = offsetof(struct sw_flow_actions, actions) +
1127 (*sfa)->actions_len;
1128
1129 if (req_size <= (ksize(*sfa) - next_offset))
1130 goto out;
1131
1132 new_acts_size = ksize(*sfa) * 2;
1133
1134 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
1135 if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size)
1136 return ERR_PTR(-EMSGSIZE);
1137 new_acts_size = MAX_ACTIONS_BUFSIZE;
1138 }
1139
1140 acts = ovs_nla_alloc_flow_actions(new_acts_size);
1141 if (IS_ERR(acts))
1142 return (void *)acts;
1143
1144 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
1145 acts->actions_len = (*sfa)->actions_len;
1146 kfree(*sfa);
1147 *sfa = acts;
1148
1149out:
1150 (*sfa)->actions_len += req_size;
1151 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
1152}
1153
1154static int add_action(struct sw_flow_actions **sfa, int attrtype, void *data, int len)
1155{
1156 struct nlattr *a;
1157
1158 a = reserve_sfa_size(sfa, nla_attr_size(len));
1159 if (IS_ERR(a))
1160 return PTR_ERR(a);
1161
1162 a->nla_type = attrtype;
1163 a->nla_len = nla_attr_size(len);
1164
1165 if (data)
1166 memcpy(nla_data(a), data, len);
1167 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
1168
1169 return 0;
1170}
1171
1172static inline int add_nested_action_start(struct sw_flow_actions **sfa,
1173 int attrtype)
1174{
1175 int used = (*sfa)->actions_len;
1176 int err;
1177
1178 err = add_action(sfa, attrtype, NULL, 0);
1179 if (err)
1180 return err;
1181
1182 return used;
1183}
1184
1185static inline void add_nested_action_end(struct sw_flow_actions *sfa,
1186 int st_offset)
1187{
1188 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
1189 st_offset);
1190
1191 a->nla_len = sfa->actions_len - st_offset;
1192}
1193
1194static int validate_and_copy_sample(const struct nlattr *attr,
1195 const struct sw_flow_key *key, int depth,
1196 struct sw_flow_actions **sfa)
1197{
1198 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
1199 const struct nlattr *probability, *actions;
1200 const struct nlattr *a;
1201 int rem, start, err, st_acts;
1202
1203 memset(attrs, 0, sizeof(attrs));
1204 nla_for_each_nested(a, attr, rem) {
1205 int type = nla_type(a);
1206 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
1207 return -EINVAL;
1208 attrs[type] = a;
1209 }
1210 if (rem)
1211 return -EINVAL;
1212
1213 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
1214 if (!probability || nla_len(probability) != sizeof(u32))
1215 return -EINVAL;
1216
1217 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
1218 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
1219 return -EINVAL;
1220
1221 /* validation done, copy sample action. */
1222 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE);
1223 if (start < 0)
1224 return start;
1225 err = add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY,
1226 nla_data(probability), sizeof(u32));
1227 if (err)
1228 return err;
1229 st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS);
1230 if (st_acts < 0)
1231 return st_acts;
1232
1233 err = ovs_nla_copy_actions(actions, key, depth + 1, sfa);
1234 if (err)
1235 return err;
1236
1237 add_nested_action_end(*sfa, st_acts);
1238 add_nested_action_end(*sfa, start);
1239
1240 return 0;
1241}
1242
1243static int validate_tp_port(const struct sw_flow_key *flow_key)
1244{
1245 if (flow_key->eth.type == htons(ETH_P_IP)) {
1246 if (flow_key->ipv4.tp.src || flow_key->ipv4.tp.dst)
1247 return 0;
1248 } else if (flow_key->eth.type == htons(ETH_P_IPV6)) {
1249 if (flow_key->ipv6.tp.src || flow_key->ipv6.tp.dst)
1250 return 0;
1251 }
1252
1253 return -EINVAL;
1254}
1255
1256void ovs_match_init(struct sw_flow_match *match,
1257 struct sw_flow_key *key,
1258 struct sw_flow_mask *mask)
1259{
1260 memset(match, 0, sizeof(*match));
1261 match->key = key;
1262 match->mask = mask;
1263
1264 memset(key, 0, sizeof(*key));
1265
1266 if (mask) {
1267 memset(&mask->key, 0, sizeof(mask->key));
1268 mask->range.start = mask->range.end = 0;
1269 }
1270}
1271
1272static int validate_and_copy_set_tun(const struct nlattr *attr,
1273 struct sw_flow_actions **sfa)
1274{
1275 struct sw_flow_match match;
1276 struct sw_flow_key key;
1277 int err, start;
1278
1279 ovs_match_init(&match, &key, NULL);
1280 err = ipv4_tun_from_nlattr(nla_data(attr), &match, false);
1281 if (err)
1282 return err;
1283
1284 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET);
1285 if (start < 0)
1286 return start;
1287
1288 err = add_action(sfa, OVS_KEY_ATTR_IPV4_TUNNEL, &match.key->tun_key,
1289 sizeof(match.key->tun_key));
1290 add_nested_action_end(*sfa, start);
1291
1292 return err;
1293}
1294
1295static int validate_set(const struct nlattr *a,
1296 const struct sw_flow_key *flow_key,
1297 struct sw_flow_actions **sfa,
1298 bool *set_tun)
1299{
1300 const struct nlattr *ovs_key = nla_data(a);
1301 int key_type = nla_type(ovs_key);
1302
1303 /* There can be only one key in a action */
1304 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
1305 return -EINVAL;
1306
1307 if (key_type > OVS_KEY_ATTR_MAX ||
1308 (ovs_key_lens[key_type] != nla_len(ovs_key) &&
1309 ovs_key_lens[key_type] != -1))
1310 return -EINVAL;
1311
1312 switch (key_type) {
1313 const struct ovs_key_ipv4 *ipv4_key;
1314 const struct ovs_key_ipv6 *ipv6_key;
1315 int err;
1316
1317 case OVS_KEY_ATTR_PRIORITY:
1318 case OVS_KEY_ATTR_SKB_MARK:
1319 case OVS_KEY_ATTR_ETHERNET:
1320 break;
1321
1322 case OVS_KEY_ATTR_TUNNEL:
1323 *set_tun = true;
1324 err = validate_and_copy_set_tun(a, sfa);
1325 if (err)
1326 return err;
1327 break;
1328
1329 case OVS_KEY_ATTR_IPV4:
1330 if (flow_key->eth.type != htons(ETH_P_IP))
1331 return -EINVAL;
1332
1333 if (!flow_key->ip.proto)
1334 return -EINVAL;
1335
1336 ipv4_key = nla_data(ovs_key);
1337 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
1338 return -EINVAL;
1339
1340 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
1341 return -EINVAL;
1342
1343 break;
1344
1345 case OVS_KEY_ATTR_IPV6:
1346 if (flow_key->eth.type != htons(ETH_P_IPV6))
1347 return -EINVAL;
1348
1349 if (!flow_key->ip.proto)
1350 return -EINVAL;
1351
1352 ipv6_key = nla_data(ovs_key);
1353 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
1354 return -EINVAL;
1355
1356 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
1357 return -EINVAL;
1358
1359 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
1360 return -EINVAL;
1361
1362 break;
1363
1364 case OVS_KEY_ATTR_TCP:
1365 if (flow_key->ip.proto != IPPROTO_TCP)
1366 return -EINVAL;
1367
1368 return validate_tp_port(flow_key);
1369
1370 case OVS_KEY_ATTR_UDP:
1371 if (flow_key->ip.proto != IPPROTO_UDP)
1372 return -EINVAL;
1373
1374 return validate_tp_port(flow_key);
1375
1376 case OVS_KEY_ATTR_SCTP:
1377 if (flow_key->ip.proto != IPPROTO_SCTP)
1378 return -EINVAL;
1379
1380 return validate_tp_port(flow_key);
1381
1382 default:
1383 return -EINVAL;
1384 }
1385
1386 return 0;
1387}
1388
1389static int validate_userspace(const struct nlattr *attr)
1390{
1391 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
1392 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
1393 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
1394 };
1395 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
1396 int error;
1397
1398 error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
1399 attr, userspace_policy);
1400 if (error)
1401 return error;
1402
1403 if (!a[OVS_USERSPACE_ATTR_PID] ||
1404 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
1405 return -EINVAL;
1406
1407 return 0;
1408}
1409
1410static int copy_action(const struct nlattr *from,
1411 struct sw_flow_actions **sfa)
1412{
1413 int totlen = NLA_ALIGN(from->nla_len);
1414 struct nlattr *to;
1415
1416 to = reserve_sfa_size(sfa, from->nla_len);
1417 if (IS_ERR(to))
1418 return PTR_ERR(to);
1419
1420 memcpy(to, from, totlen);
1421 return 0;
1422}
1423
1424int ovs_nla_copy_actions(const struct nlattr *attr,
1425 const struct sw_flow_key *key,
1426 int depth,
1427 struct sw_flow_actions **sfa)
1428{
1429 const struct nlattr *a;
1430 int rem, err;
1431
1432 if (depth >= SAMPLE_ACTION_DEPTH)
1433 return -EOVERFLOW;
1434
1435 nla_for_each_nested(a, attr, rem) {
1436 /* Expected argument lengths, (u32)-1 for variable length. */
1437 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
1438 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
1439 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
1440 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
1441 [OVS_ACTION_ATTR_POP_VLAN] = 0,
1442 [OVS_ACTION_ATTR_SET] = (u32)-1,
1443 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1
1444 };
1445 const struct ovs_action_push_vlan *vlan;
1446 int type = nla_type(a);
1447 bool skip_copy;
1448
1449 if (type > OVS_ACTION_ATTR_MAX ||
1450 (action_lens[type] != nla_len(a) &&
1451 action_lens[type] != (u32)-1))
1452 return -EINVAL;
1453
1454 skip_copy = false;
1455 switch (type) {
1456 case OVS_ACTION_ATTR_UNSPEC:
1457 return -EINVAL;
1458
1459 case OVS_ACTION_ATTR_USERSPACE:
1460 err = validate_userspace(a);
1461 if (err)
1462 return err;
1463 break;
1464
1465 case OVS_ACTION_ATTR_OUTPUT:
1466 if (nla_get_u32(a) >= DP_MAX_PORTS)
1467 return -EINVAL;
1468 break;
1469
1470
1471 case OVS_ACTION_ATTR_POP_VLAN:
1472 break;
1473
1474 case OVS_ACTION_ATTR_PUSH_VLAN:
1475 vlan = nla_data(a);
1476 if (vlan->vlan_tpid != htons(ETH_P_8021Q))
1477 return -EINVAL;
1478 if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
1479 return -EINVAL;
1480 break;
1481
1482 case OVS_ACTION_ATTR_SET:
1483 err = validate_set(a, key, sfa, &skip_copy);
1484 if (err)
1485 return err;
1486 break;
1487
1488 case OVS_ACTION_ATTR_SAMPLE:
1489 err = validate_and_copy_sample(a, key, depth, sfa);
1490 if (err)
1491 return err;
1492 skip_copy = true;
1493 break;
1494
1495 default:
1496 return -EINVAL;
1497 }
1498 if (!skip_copy) {
1499 err = copy_action(a, sfa);
1500 if (err)
1501 return err;
1502 }
1503 }
1504
1505 if (rem > 0)
1506 return -EINVAL;
1507
1508 return 0;
1509}
1510
1511static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
1512{
1513 const struct nlattr *a;
1514 struct nlattr *start;
1515 int err = 0, rem;
1516
1517 start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
1518 if (!start)
1519 return -EMSGSIZE;
1520
1521 nla_for_each_nested(a, attr, rem) {
1522 int type = nla_type(a);
1523 struct nlattr *st_sample;
1524
1525 switch (type) {
1526 case OVS_SAMPLE_ATTR_PROBABILITY:
1527 if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY,
1528 sizeof(u32), nla_data(a)))
1529 return -EMSGSIZE;
1530 break;
1531 case OVS_SAMPLE_ATTR_ACTIONS:
1532 st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
1533 if (!st_sample)
1534 return -EMSGSIZE;
1535 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
1536 if (err)
1537 return err;
1538 nla_nest_end(skb, st_sample);
1539 break;
1540 }
1541 }
1542
1543 nla_nest_end(skb, start);
1544 return err;
1545}
1546
1547static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
1548{
1549 const struct nlattr *ovs_key = nla_data(a);
1550 int key_type = nla_type(ovs_key);
1551 struct nlattr *start;
1552 int err;
1553
1554 switch (key_type) {
1555 case OVS_KEY_ATTR_IPV4_TUNNEL:
1556 start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
1557 if (!start)
1558 return -EMSGSIZE;
1559
1560 err = ipv4_tun_to_nlattr(skb, nla_data(ovs_key),
1561 nla_data(ovs_key));
1562 if (err)
1563 return err;
1564 nla_nest_end(skb, start);
1565 break;
1566 default:
1567 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
1568 return -EMSGSIZE;
1569 break;
1570 }
1571
1572 return 0;
1573}
1574
1575int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
1576{
1577 const struct nlattr *a;
1578 int rem, err;
1579
1580 nla_for_each_attr(a, attr, len, rem) {
1581 int type = nla_type(a);
1582
1583 switch (type) {
1584 case OVS_ACTION_ATTR_SET:
1585 err = set_action_to_attr(a, skb);
1586 if (err)
1587 return err;
1588 break;
1589
1590 case OVS_ACTION_ATTR_SAMPLE:
1591 err = sample_action_to_attr(a, skb);
1592 if (err)
1593 return err;
1594 break;
1595 default:
1596 if (nla_put(skb, type, nla_len(a), nla_data(a)))
1597 return -EMSGSIZE;
1598 break;
1599 }
1600 }
1601
1602 return 0;
1603}
diff --git a/net/openvswitch/flow_netlink.h b/net/openvswitch/flow_netlink.h
new file mode 100644
index 000000000000..440151045d39
--- /dev/null
+++ b/net/openvswitch/flow_netlink.h
@@ -0,0 +1,60 @@
1/*
2 * Copyright (c) 2007-2013 Nicira, Inc.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public License
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
16 * 02110-1301, USA
17 */
18
19
20#ifndef FLOW_NETLINK_H
21#define FLOW_NETLINK_H 1
22
23#include <linux/kernel.h>
24#include <linux/netlink.h>
25#include <linux/openvswitch.h>
26#include <linux/spinlock.h>
27#include <linux/types.h>
28#include <linux/rcupdate.h>
29#include <linux/if_ether.h>
30#include <linux/in6.h>
31#include <linux/jiffies.h>
32#include <linux/time.h>
33#include <linux/flex_array.h>
34
35#include <net/inet_ecn.h>
36#include <net/ip_tunnels.h>
37
38#include "flow.h"
39
40void ovs_match_init(struct sw_flow_match *match,
41 struct sw_flow_key *key, struct sw_flow_mask *mask);
42
43int ovs_nla_put_flow(const struct sw_flow_key *,
44 const struct sw_flow_key *, struct sk_buff *);
45int ovs_nla_get_flow_metadata(struct sw_flow *flow,
46 const struct nlattr *attr);
47int ovs_nla_get_match(struct sw_flow_match *match,
48 const struct nlattr *,
49 const struct nlattr *);
50
51int ovs_nla_copy_actions(const struct nlattr *attr,
52 const struct sw_flow_key *key, int depth,
53 struct sw_flow_actions **sfa);
54int ovs_nla_put_actions(const struct nlattr *attr,
55 int len, struct sk_buff *skb);
56
57struct sw_flow_actions *ovs_nla_alloc_flow_actions(int actions_len);
58void ovs_nla_free_flow_actions(struct sw_flow_actions *);
59
60#endif /* flow_netlink.h */
diff --git a/net/openvswitch/flow_table.c b/net/openvswitch/flow_table.c
new file mode 100644
index 000000000000..dcadb75bb173
--- /dev/null
+++ b/net/openvswitch/flow_table.c
@@ -0,0 +1,517 @@
1/*
2 * Copyright (c) 2007-2013 Nicira, Inc.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public License
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
16 * 02110-1301, USA
17 */
18
19#include "flow.h"
20#include "datapath.h"
21#include <linux/uaccess.h>
22#include <linux/netdevice.h>
23#include <linux/etherdevice.h>
24#include <linux/if_ether.h>
25#include <linux/if_vlan.h>
26#include <net/llc_pdu.h>
27#include <linux/kernel.h>
28#include <linux/jhash.h>
29#include <linux/jiffies.h>
30#include <linux/llc.h>
31#include <linux/module.h>
32#include <linux/in.h>
33#include <linux/rcupdate.h>
34#include <linux/if_arp.h>
35#include <linux/ip.h>
36#include <linux/ipv6.h>
37#include <linux/sctp.h>
38#include <linux/tcp.h>
39#include <linux/udp.h>
40#include <linux/icmp.h>
41#include <linux/icmpv6.h>
42#include <linux/rculist.h>
43#include <net/ip.h>
44#include <net/ipv6.h>
45#include <net/ndisc.h>
46
47static struct kmem_cache *flow_cache;
48
49static u16 range_n_bytes(const struct sw_flow_key_range *range)
50{
51 return range->end - range->start;
52}
53
54void ovs_flow_mask_key(struct sw_flow_key *dst, const struct sw_flow_key *src,
55 const struct sw_flow_mask *mask)
56{
57 const long *m = (long *)((u8 *)&mask->key + mask->range.start);
58 const long *s = (long *)((u8 *)src + mask->range.start);
59 long *d = (long *)((u8 *)dst + mask->range.start);
60 int i;
61
62 /* The memory outside of the 'mask->range' are not set since
63 * further operations on 'dst' only uses contents within
64 * 'mask->range'.
65 */
66 for (i = 0; i < range_n_bytes(&mask->range); i += sizeof(long))
67 *d++ = *s++ & *m++;
68}
69
70struct sw_flow *ovs_flow_alloc(void)
71{
72 struct sw_flow *flow;
73
74 flow = kmem_cache_alloc(flow_cache, GFP_KERNEL);
75 if (!flow)
76 return ERR_PTR(-ENOMEM);
77
78 spin_lock_init(&flow->lock);
79 flow->sf_acts = NULL;
80 flow->mask = NULL;
81
82 return flow;
83}
84
85static struct flex_array *alloc_buckets(unsigned int n_buckets)
86{
87 struct flex_array *buckets;
88 int i, err;
89
90 buckets = flex_array_alloc(sizeof(struct hlist_head),
91 n_buckets, GFP_KERNEL);
92 if (!buckets)
93 return NULL;
94
95 err = flex_array_prealloc(buckets, 0, n_buckets, GFP_KERNEL);
96 if (err) {
97 flex_array_free(buckets);
98 return NULL;
99 }
100
101 for (i = 0; i < n_buckets; i++)
102 INIT_HLIST_HEAD((struct hlist_head *)
103 flex_array_get(buckets, i));
104
105 return buckets;
106}
107
108static void flow_free(struct sw_flow *flow)
109{
110 kfree((struct sf_flow_acts __force *)flow->sf_acts);
111 kmem_cache_free(flow_cache, flow);
112}
113
114static void rcu_free_flow_callback(struct rcu_head *rcu)
115{
116 struct sw_flow *flow = container_of(rcu, struct sw_flow, rcu);
117
118 flow_free(flow);
119}
120
121void ovs_flow_free(struct sw_flow *flow, bool deferred)
122{
123 if (!flow)
124 return;
125
126 ovs_sw_flow_mask_del_ref(flow->mask, deferred);
127
128 if (deferred)
129 call_rcu(&flow->rcu, rcu_free_flow_callback);
130 else
131 flow_free(flow);
132}
133
134static void free_buckets(struct flex_array *buckets)
135{
136 flex_array_free(buckets);
137}
138
139static void __flow_tbl_destroy(struct flow_table *table)
140{
141 int i;
142
143 if (table->keep_flows)
144 goto skip_flows;
145
146 for (i = 0; i < table->n_buckets; i++) {
147 struct sw_flow *flow;
148 struct hlist_head *head = flex_array_get(table->buckets, i);
149 struct hlist_node *n;
150 int ver = table->node_ver;
151
152 hlist_for_each_entry_safe(flow, n, head, hash_node[ver]) {
153 hlist_del(&flow->hash_node[ver]);
154 ovs_flow_free(flow, false);
155 }
156 }
157
158 BUG_ON(!list_empty(table->mask_list));
159 kfree(table->mask_list);
160
161skip_flows:
162 free_buckets(table->buckets);
163 kfree(table);
164}
165
166static struct flow_table *__flow_tbl_alloc(int new_size)
167{
168 struct flow_table *table = kmalloc(sizeof(*table), GFP_KERNEL);
169
170 if (!table)
171 return NULL;
172
173 table->buckets = alloc_buckets(new_size);
174
175 if (!table->buckets) {
176 kfree(table);
177 return NULL;
178 }
179 table->n_buckets = new_size;
180 table->count = 0;
181 table->node_ver = 0;
182 table->keep_flows = false;
183 get_random_bytes(&table->hash_seed, sizeof(u32));
184 table->mask_list = NULL;
185
186 return table;
187}
188
189struct flow_table *ovs_flow_tbl_alloc(int new_size)
190{
191 struct flow_table *table = __flow_tbl_alloc(new_size);
192
193 if (!table)
194 return NULL;
195
196 table->mask_list = kmalloc(sizeof(struct list_head), GFP_KERNEL);
197 if (!table->mask_list) {
198 table->keep_flows = true;
199 __flow_tbl_destroy(table);
200 return NULL;
201 }
202 INIT_LIST_HEAD(table->mask_list);
203
204 return table;
205}
206
207static void flow_tbl_destroy_rcu_cb(struct rcu_head *rcu)
208{
209 struct flow_table *table = container_of(rcu, struct flow_table, rcu);
210
211 __flow_tbl_destroy(table);
212}
213
214void ovs_flow_tbl_destroy(struct flow_table *table, bool deferred)
215{
216 if (!table)
217 return;
218
219 if (deferred)
220 call_rcu(&table->rcu, flow_tbl_destroy_rcu_cb);
221 else
222 __flow_tbl_destroy(table);
223}
224
225struct sw_flow *ovs_flow_tbl_dump_next(struct flow_table *table,
226 u32 *bucket, u32 *last)
227{
228 struct sw_flow *flow;
229 struct hlist_head *head;
230 int ver;
231 int i;
232
233 ver = table->node_ver;
234 while (*bucket < table->n_buckets) {
235 i = 0;
236 head = flex_array_get(table->buckets, *bucket);
237 hlist_for_each_entry_rcu(flow, head, hash_node[ver]) {
238 if (i < *last) {
239 i++;
240 continue;
241 }
242 *last = i + 1;
243 return flow;
244 }
245 (*bucket)++;
246 *last = 0;
247 }
248
249 return NULL;
250}
251
252static struct hlist_head *find_bucket(struct flow_table *table, u32 hash)
253{
254 hash = jhash_1word(hash, table->hash_seed);
255 return flex_array_get(table->buckets,
256 (hash & (table->n_buckets - 1)));
257}
258
259static void __tbl_insert(struct flow_table *table, struct sw_flow *flow)
260{
261 struct hlist_head *head;
262
263 head = find_bucket(table, flow->hash);
264 hlist_add_head_rcu(&flow->hash_node[table->node_ver], head);
265
266 table->count++;
267}
268
269static void flow_table_copy_flows(struct flow_table *old,
270 struct flow_table *new)
271{
272 int old_ver;
273 int i;
274
275 old_ver = old->node_ver;
276 new->node_ver = !old_ver;
277
278 /* Insert in new table. */
279 for (i = 0; i < old->n_buckets; i++) {
280 struct sw_flow *flow;
281 struct hlist_head *head;
282
283 head = flex_array_get(old->buckets, i);
284
285 hlist_for_each_entry(flow, head, hash_node[old_ver])
286 __tbl_insert(new, flow);
287 }
288
289 new->mask_list = old->mask_list;
290 old->keep_flows = true;
291}
292
293static struct flow_table *__flow_tbl_rehash(struct flow_table *table,
294 int n_buckets)
295{
296 struct flow_table *new_table;
297
298 new_table = __flow_tbl_alloc(n_buckets);
299 if (!new_table)
300 return ERR_PTR(-ENOMEM);
301
302 flow_table_copy_flows(table, new_table);
303
304 return new_table;
305}
306
307struct flow_table *ovs_flow_tbl_rehash(struct flow_table *table)
308{
309 return __flow_tbl_rehash(table, table->n_buckets);
310}
311
312struct flow_table *ovs_flow_tbl_expand(struct flow_table *table)
313{
314 return __flow_tbl_rehash(table, table->n_buckets * 2);
315}
316
317static u32 flow_hash(const struct sw_flow_key *key, int key_start,
318 int key_end)
319{
320 u32 *hash_key = (u32 *)((u8 *)key + key_start);
321 int hash_u32s = (key_end - key_start) >> 2;
322
323 /* Make sure number of hash bytes are multiple of u32. */
324 BUILD_BUG_ON(sizeof(long) % sizeof(u32));
325
326 return jhash2(hash_key, hash_u32s, 0);
327}
328
329static int flow_key_start(const struct sw_flow_key *key)
330{
331 if (key->tun_key.ipv4_dst)
332 return 0;
333 else
334 return rounddown(offsetof(struct sw_flow_key, phy),
335 sizeof(long));
336}
337
338static bool cmp_key(const struct sw_flow_key *key1,
339 const struct sw_flow_key *key2,
340 int key_start, int key_end)
341{
342 const long *cp1 = (long *)((u8 *)key1 + key_start);
343 const long *cp2 = (long *)((u8 *)key2 + key_start);
344 long diffs = 0;
345 int i;
346
347 for (i = key_start; i < key_end; i += sizeof(long))
348 diffs |= *cp1++ ^ *cp2++;
349
350 return diffs == 0;
351}
352
353static bool flow_cmp_masked_key(const struct sw_flow *flow,
354 const struct sw_flow_key *key,
355 int key_start, int key_end)
356{
357 return cmp_key(&flow->key, key, key_start, key_end);
358}
359
360bool ovs_flow_cmp_unmasked_key(const struct sw_flow *flow,
361 struct sw_flow_match *match)
362{
363 struct sw_flow_key *key = match->key;
364 int key_start = flow_key_start(key);
365 int key_end = match->range.end;
366
367 return cmp_key(&flow->unmasked_key, key, key_start, key_end);
368}
369
370static struct sw_flow *masked_flow_lookup(struct flow_table *table,
371 const struct sw_flow_key *unmasked,
372 struct sw_flow_mask *mask)
373{
374 struct sw_flow *flow;
375 struct hlist_head *head;
376 int key_start = mask->range.start;
377 int key_end = mask->range.end;
378 u32 hash;
379 struct sw_flow_key masked_key;
380
381 ovs_flow_mask_key(&masked_key, unmasked, mask);
382 hash = flow_hash(&masked_key, key_start, key_end);
383 head = find_bucket(table, hash);
384 hlist_for_each_entry_rcu(flow, head, hash_node[table->node_ver]) {
385 if (flow->mask == mask &&
386 flow_cmp_masked_key(flow, &masked_key,
387 key_start, key_end))
388 return flow;
389 }
390 return NULL;
391}
392
393struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *tbl,
394 const struct sw_flow_key *key)
395{
396 struct sw_flow *flow = NULL;
397 struct sw_flow_mask *mask;
398
399 list_for_each_entry_rcu(mask, tbl->mask_list, list) {
400 flow = masked_flow_lookup(tbl, key, mask);
401 if (flow) /* Found */
402 break;
403 }
404
405 return flow;
406}
407
408void ovs_flow_tbl_insert(struct flow_table *table, struct sw_flow *flow)
409{
410 flow->hash = flow_hash(&flow->key, flow->mask->range.start,
411 flow->mask->range.end);
412 __tbl_insert(table, flow);
413}
414
415void ovs_flow_tbl_remove(struct flow_table *table, struct sw_flow *flow)
416{
417 BUG_ON(table->count == 0);
418 hlist_del_rcu(&flow->hash_node[table->node_ver]);
419 table->count--;
420}
421
422struct sw_flow_mask *ovs_sw_flow_mask_alloc(void)
423{
424 struct sw_flow_mask *mask;
425
426 mask = kmalloc(sizeof(*mask), GFP_KERNEL);
427 if (mask)
428 mask->ref_count = 0;
429
430 return mask;
431}
432
433void ovs_sw_flow_mask_add_ref(struct sw_flow_mask *mask)
434{
435 mask->ref_count++;
436}
437
438static void rcu_free_sw_flow_mask_cb(struct rcu_head *rcu)
439{
440 struct sw_flow_mask *mask = container_of(rcu, struct sw_flow_mask, rcu);
441
442 kfree(mask);
443}
444
445void ovs_sw_flow_mask_del_ref(struct sw_flow_mask *mask, bool deferred)
446{
447 if (!mask)
448 return;
449
450 BUG_ON(!mask->ref_count);
451 mask->ref_count--;
452
453 if (!mask->ref_count) {
454 list_del_rcu(&mask->list);
455 if (deferred)
456 call_rcu(&mask->rcu, rcu_free_sw_flow_mask_cb);
457 else
458 kfree(mask);
459 }
460}
461
462static bool mask_equal(const struct sw_flow_mask *a,
463 const struct sw_flow_mask *b)
464{
465 u8 *a_ = (u8 *)&a->key + a->range.start;
466 u8 *b_ = (u8 *)&b->key + b->range.start;
467
468 return (a->range.end == b->range.end)
469 && (a->range.start == b->range.start)
470 && (memcmp(a_, b_, range_n_bytes(&a->range)) == 0);
471}
472
473struct sw_flow_mask *ovs_sw_flow_mask_find(const struct flow_table *tbl,
474 const struct sw_flow_mask *mask)
475{
476 struct list_head *ml;
477
478 list_for_each(ml, tbl->mask_list) {
479 struct sw_flow_mask *m;
480 m = container_of(ml, struct sw_flow_mask, list);
481 if (mask_equal(mask, m))
482 return m;
483 }
484
485 return NULL;
486}
487
488/**
489 * add a new mask into the mask list.
490 * The caller needs to make sure that 'mask' is not the same
491 * as any masks that are already on the list.
492 */
493void ovs_sw_flow_mask_insert(struct flow_table *tbl, struct sw_flow_mask *mask)
494{
495 list_add_rcu(&mask->list, tbl->mask_list);
496}
497
498/* Initializes the flow module.
499 * Returns zero if successful or a negative error code. */
500int ovs_flow_init(void)
501{
502 BUILD_BUG_ON(__alignof__(struct sw_flow_key) % __alignof__(long));
503 BUILD_BUG_ON(sizeof(struct sw_flow_key) % sizeof(long));
504
505 flow_cache = kmem_cache_create("sw_flow", sizeof(struct sw_flow), 0,
506 0, NULL);
507 if (flow_cache == NULL)
508 return -ENOMEM;
509
510 return 0;
511}
512
513/* Uninitializes the flow module. */
514void ovs_flow_exit(void)
515{
516 kmem_cache_destroy(flow_cache);
517}
diff --git a/net/openvswitch/flow_table.h b/net/openvswitch/flow_table.h
new file mode 100644
index 000000000000..d7a114457cde
--- /dev/null
+++ b/net/openvswitch/flow_table.h
@@ -0,0 +1,91 @@
1/*
2 * Copyright (c) 2007-2013 Nicira, Inc.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 *
13 * You should have received a copy of the GNU General Public License
14 * along with this program; if not, write to the Free Software
15 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
16 * 02110-1301, USA
17 */
18
19#ifndef FLOW_TABLE_H
20#define FLOW_TABLE_H 1
21
22#include <linux/kernel.h>
23#include <linux/netlink.h>
24#include <linux/openvswitch.h>
25#include <linux/spinlock.h>
26#include <linux/types.h>
27#include <linux/rcupdate.h>
28#include <linux/if_ether.h>
29#include <linux/in6.h>
30#include <linux/jiffies.h>
31#include <linux/time.h>
32#include <linux/flex_array.h>
33
34#include <net/inet_ecn.h>
35#include <net/ip_tunnels.h>
36
37#include "flow.h"
38
39#define TBL_MIN_BUCKETS 1024
40
41struct flow_table {
42 struct flex_array *buckets;
43 unsigned int count, n_buckets;
44 struct rcu_head rcu;
45 struct list_head *mask_list;
46 int node_ver;
47 u32 hash_seed;
48 bool keep_flows;
49};
50
51int ovs_flow_init(void);
52void ovs_flow_exit(void);
53
54struct sw_flow *ovs_flow_alloc(void);
55void ovs_flow_free(struct sw_flow *, bool deferred);
56
57static inline int ovs_flow_tbl_count(struct flow_table *table)
58{
59 return table->count;
60}
61
62static inline int ovs_flow_tbl_need_to_expand(struct flow_table *table)
63{
64 return (table->count > table->n_buckets);
65}
66
67struct flow_table *ovs_flow_tbl_alloc(int new_size);
68struct flow_table *ovs_flow_tbl_expand(struct flow_table *table);
69struct flow_table *ovs_flow_tbl_rehash(struct flow_table *table);
70void ovs_flow_tbl_destroy(struct flow_table *table, bool deferred);
71
72void ovs_flow_tbl_insert(struct flow_table *table, struct sw_flow *flow);
73void ovs_flow_tbl_remove(struct flow_table *table, struct sw_flow *flow);
74struct sw_flow *ovs_flow_tbl_dump_next(struct flow_table *table,
75 u32 *bucket, u32 *idx);
76struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *,
77 const struct sw_flow_key *);
78
79bool ovs_flow_cmp_unmasked_key(const struct sw_flow *flow,
80 struct sw_flow_match *match);
81
82struct sw_flow_mask *ovs_sw_flow_mask_alloc(void);
83void ovs_sw_flow_mask_add_ref(struct sw_flow_mask *);
84void ovs_sw_flow_mask_del_ref(struct sw_flow_mask *, bool deferred);
85void ovs_sw_flow_mask_insert(struct flow_table *, struct sw_flow_mask *);
86struct sw_flow_mask *ovs_sw_flow_mask_find(const struct flow_table *,
87 const struct sw_flow_mask *);
88void ovs_flow_mask_key(struct sw_flow_key *dst, const struct sw_flow_key *src,
89 const struct sw_flow_mask *mask);
90
91#endif /* flow_table.h */
generated by cgit v1.2.2 (git 2.25.0) at 2026-05-29 14:45:10 -0400