aboutsummaryrefslogtreecommitdiffstats
path: root/net/openvswitch
diff options
context:
space:
mode:
authorDavid S. Miller <davem@davemloft.net>2013-11-04 16:25:04 -0500
committerDavid S. Miller <davem@davemloft.net>2013-11-04 16:25:04 -0500
commit6fcf018ae4491dc11b080892fa9f3dbd928fdbb9 (patch)
treef28592676e295183581d59149fd1eac913b229a2 /net/openvswitch
parent5a6e55c461db3364aa5be919101db972bc859133 (diff)
parent8ddd094675cfd453fc9838caa46ea108a4107183 (diff)
Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/jesse/openvswitch
Jesse Gross says: ==================== Open vSwitch A set of updates for net-next/3.13. Major changes are: * Restructure flow handling code to be more logically organized and easier to read. * Rehashing of the flow table is moved from a workqueue to flow installation time. Before, heavy load could block the workqueue for excessive periods of time. * Additional debugging information is provided to help diagnose megaflows. * It's now possible to match on TCP flags. ==================== Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'net/openvswitch')
-rw-r--r--net/openvswitch/Makefile2
-rw-r--r--net/openvswitch/datapath.c668
-rw-r--r--net/openvswitch/datapath.h9
-rw-r--r--net/openvswitch/flow.c1605
-rw-r--r--net/openvswitch/flow.h132
-rw-r--r--net/openvswitch/flow_netlink.c1630
-rw-r--r--net/openvswitch/flow_netlink.h60
-rw-r--r--net/openvswitch/flow_table.c592
-rw-r--r--net/openvswitch/flow_table.h81
-rw-r--r--net/openvswitch/vport-gre.c2
-rw-r--r--net/openvswitch/vport-internal_dev.c2
-rw-r--r--net/openvswitch/vport-vxlan.c1
12 files changed, 2496 insertions, 2288 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 2aa13bd7f2b2..1408adc2a2a7 100644
--- a/net/openvswitch/datapath.c
+++ b/net/openvswitch/datapath.c
@@ -55,14 +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)
63static void rehash_flow_table(struct work_struct *work);
64static DECLARE_DELAYED_WORK(rehash_flow_wq, rehash_flow_table);
65
66int ovs_net_id __read_mostly; 62int ovs_net_id __read_mostly;
67 63
68static void ovs_notify(struct sk_buff *skb, struct genl_info *info, 64static void ovs_notify(struct sk_buff *skb, struct genl_info *info,
@@ -165,7 +161,7 @@ static void destroy_dp_rcu(struct rcu_head *rcu)
165{ 161{
166 struct datapath *dp = container_of(rcu, struct datapath, rcu); 162 struct datapath *dp = container_of(rcu, struct datapath, rcu);
167 163
168 ovs_flow_tbl_destroy((__force struct flow_table *)dp->table, false); 164 ovs_flow_tbl_destroy(&dp->table);
169 free_percpu(dp->stats_percpu); 165 free_percpu(dp->stats_percpu);
170 release_net(ovs_dp_get_net(dp)); 166 release_net(ovs_dp_get_net(dp));
171 kfree(dp->ports); 167 kfree(dp->ports);
@@ -225,6 +221,7 @@ void ovs_dp_process_received_packet(struct vport *p, struct sk_buff *skb)
225 struct dp_stats_percpu *stats; 221 struct dp_stats_percpu *stats;
226 struct sw_flow_key key; 222 struct sw_flow_key key;
227 u64 *stats_counter; 223 u64 *stats_counter;
224 u32 n_mask_hit;
228 int error; 225 int error;
229 226
230 stats = this_cpu_ptr(dp->stats_percpu); 227 stats = this_cpu_ptr(dp->stats_percpu);
@@ -237,7 +234,7 @@ void ovs_dp_process_received_packet(struct vport *p, struct sk_buff *skb)
237 } 234 }
238 235
239 /* Look up flow. */ 236 /* Look up flow. */
240 flow = ovs_flow_lookup(rcu_dereference(dp->table), &key); 237 flow = ovs_flow_tbl_lookup(&dp->table, &key, &n_mask_hit);
241 if (unlikely(!flow)) { 238 if (unlikely(!flow)) {
242 struct dp_upcall_info upcall; 239 struct dp_upcall_info upcall;
243 240
@@ -262,6 +259,7 @@ out:
262 /* Update datapath statistics. */ 259 /* Update datapath statistics. */
263 u64_stats_update_begin(&stats->sync); 260 u64_stats_update_begin(&stats->sync);
264 (*stats_counter)++; 261 (*stats_counter)++;
262 stats->n_mask_hit += n_mask_hit;
265 u64_stats_update_end(&stats->sync); 263 u64_stats_update_end(&stats->sync);
266} 264}
267 265
@@ -435,7 +433,7 @@ static int queue_userspace_packet(struct net *net, int dp_ifindex,
435 upcall->dp_ifindex = dp_ifindex; 433 upcall->dp_ifindex = dp_ifindex;
436 434
437 nla = nla_nest_start(user_skb, OVS_PACKET_ATTR_KEY); 435 nla = nla_nest_start(user_skb, OVS_PACKET_ATTR_KEY);
438 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);
439 nla_nest_end(user_skb, nla); 437 nla_nest_end(user_skb, nla);
440 438
441 if (upcall_info->userdata) 439 if (upcall_info->userdata)
@@ -455,398 +453,6 @@ out:
455 return err; 453 return err;
456} 454}
457 455
458/* Called with ovs_mutex. */
459static int flush_flows(struct datapath *dp)
460{
461 struct flow_table *old_table;
462 struct flow_table *new_table;
463
464 old_table = ovsl_dereference(dp->table);
465 new_table = ovs_flow_tbl_alloc(TBL_MIN_BUCKETS);
466 if (!new_table)
467 return -ENOMEM;
468
469 rcu_assign_pointer(dp->table, new_table);
470
471 ovs_flow_tbl_destroy(old_table, true);
472 return 0;
473}
474
475static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa, int attr_len)
476{
477
478 struct sw_flow_actions *acts;
479 int new_acts_size;
480 int req_size = NLA_ALIGN(attr_len);
481 int next_offset = offsetof(struct sw_flow_actions, actions) +
482 (*sfa)->actions_len;
483
484 if (req_size <= (ksize(*sfa) - next_offset))
485 goto out;
486
487 new_acts_size = ksize(*sfa) * 2;
488
489 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
490 if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size)
491 return ERR_PTR(-EMSGSIZE);
492 new_acts_size = MAX_ACTIONS_BUFSIZE;
493 }
494
495 acts = ovs_flow_actions_alloc(new_acts_size);
496 if (IS_ERR(acts))
497 return (void *)acts;
498
499 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
500 acts->actions_len = (*sfa)->actions_len;
501 kfree(*sfa);
502 *sfa = acts;
503
504out:
505 (*sfa)->actions_len += req_size;
506 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
507}
508
509static int add_action(struct sw_flow_actions **sfa, int attrtype, void *data, int len)
510{
511 struct nlattr *a;
512
513 a = reserve_sfa_size(sfa, nla_attr_size(len));
514 if (IS_ERR(a))
515 return PTR_ERR(a);
516
517 a->nla_type = attrtype;
518 a->nla_len = nla_attr_size(len);
519
520 if (data)
521 memcpy(nla_data(a), data, len);
522 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
523
524 return 0;
525}
526
527static inline int add_nested_action_start(struct sw_flow_actions **sfa, int attrtype)
528{
529 int used = (*sfa)->actions_len;
530 int err;
531
532 err = add_action(sfa, attrtype, NULL, 0);
533 if (err)
534 return err;
535
536 return used;
537}
538
539static inline void add_nested_action_end(struct sw_flow_actions *sfa, int st_offset)
540{
541 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions + st_offset);
542
543 a->nla_len = sfa->actions_len - st_offset;
544}
545
546static int validate_and_copy_actions(const struct nlattr *attr,
547 const struct sw_flow_key *key, int depth,
548 struct sw_flow_actions **sfa);
549
550static int validate_and_copy_sample(const struct nlattr *attr,
551 const struct sw_flow_key *key, int depth,
552 struct sw_flow_actions **sfa)
553{
554 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
555 const struct nlattr *probability, *actions;
556 const struct nlattr *a;
557 int rem, start, err, st_acts;
558
559 memset(attrs, 0, sizeof(attrs));
560 nla_for_each_nested(a, attr, rem) {
561 int type = nla_type(a);
562 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
563 return -EINVAL;
564 attrs[type] = a;
565 }
566 if (rem)
567 return -EINVAL;
568
569 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
570 if (!probability || nla_len(probability) != sizeof(u32))
571 return -EINVAL;
572
573 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
574 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
575 return -EINVAL;
576
577 /* validation done, copy sample action. */
578 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE);
579 if (start < 0)
580 return start;
581 err = add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY, nla_data(probability), sizeof(u32));
582 if (err)
583 return err;
584 st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS);
585 if (st_acts < 0)
586 return st_acts;
587
588 err = validate_and_copy_actions(actions, key, depth + 1, sfa);
589 if (err)
590 return err;
591
592 add_nested_action_end(*sfa, st_acts);
593 add_nested_action_end(*sfa, start);
594
595 return 0;
596}
597
598static int validate_tp_port(const struct sw_flow_key *flow_key)
599{
600 if (flow_key->eth.type == htons(ETH_P_IP)) {
601 if (flow_key->ipv4.tp.src || flow_key->ipv4.tp.dst)
602 return 0;
603 } else if (flow_key->eth.type == htons(ETH_P_IPV6)) {
604 if (flow_key->ipv6.tp.src || flow_key->ipv6.tp.dst)
605 return 0;
606 }
607
608 return -EINVAL;
609}
610
611static int validate_and_copy_set_tun(const struct nlattr *attr,
612 struct sw_flow_actions **sfa)
613{
614 struct sw_flow_match match;
615 struct sw_flow_key key;
616 int err, start;
617
618 ovs_match_init(&match, &key, NULL);
619 err = ovs_ipv4_tun_from_nlattr(nla_data(attr), &match, false);
620 if (err)
621 return err;
622
623 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET);
624 if (start < 0)
625 return start;
626
627 err = add_action(sfa, OVS_KEY_ATTR_IPV4_TUNNEL, &match.key->tun_key,
628 sizeof(match.key->tun_key));
629 add_nested_action_end(*sfa, start);
630
631 return err;
632}
633
634static int validate_set(const struct nlattr *a,
635 const struct sw_flow_key *flow_key,
636 struct sw_flow_actions **sfa,
637 bool *set_tun)
638{
639 const struct nlattr *ovs_key = nla_data(a);
640 int key_type = nla_type(ovs_key);
641
642 /* There can be only one key in a action */
643 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
644 return -EINVAL;
645
646 if (key_type > OVS_KEY_ATTR_MAX ||
647 (ovs_key_lens[key_type] != nla_len(ovs_key) &&
648 ovs_key_lens[key_type] != -1))
649 return -EINVAL;
650
651 switch (key_type) {
652 const struct ovs_key_ipv4 *ipv4_key;
653 const struct ovs_key_ipv6 *ipv6_key;
654 int err;
655
656 case OVS_KEY_ATTR_PRIORITY:
657 case OVS_KEY_ATTR_SKB_MARK:
658 case OVS_KEY_ATTR_ETHERNET:
659 break;
660
661 case OVS_KEY_ATTR_TUNNEL:
662 *set_tun = true;
663 err = validate_and_copy_set_tun(a, sfa);
664 if (err)
665 return err;
666 break;
667
668 case OVS_KEY_ATTR_IPV4:
669 if (flow_key->eth.type != htons(ETH_P_IP))
670 return -EINVAL;
671
672 if (!flow_key->ip.proto)
673 return -EINVAL;
674
675 ipv4_key = nla_data(ovs_key);
676 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
677 return -EINVAL;
678
679 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
680 return -EINVAL;
681
682 break;
683
684 case OVS_KEY_ATTR_IPV6:
685 if (flow_key->eth.type != htons(ETH_P_IPV6))
686 return -EINVAL;
687
688 if (!flow_key->ip.proto)
689 return -EINVAL;
690
691 ipv6_key = nla_data(ovs_key);
692 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
693 return -EINVAL;
694
695 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
696 return -EINVAL;
697
698 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
699 return -EINVAL;
700
701 break;
702
703 case OVS_KEY_ATTR_TCP:
704 if (flow_key->ip.proto != IPPROTO_TCP)
705 return -EINVAL;
706
707 return validate_tp_port(flow_key);
708
709 case OVS_KEY_ATTR_UDP:
710 if (flow_key->ip.proto != IPPROTO_UDP)
711 return -EINVAL;
712
713 return validate_tp_port(flow_key);
714
715 case OVS_KEY_ATTR_SCTP:
716 if (flow_key->ip.proto != IPPROTO_SCTP)
717 return -EINVAL;
718
719 return validate_tp_port(flow_key);
720
721 default:
722 return -EINVAL;
723 }
724
725 return 0;
726}
727
728static int validate_userspace(const struct nlattr *attr)
729{
730 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
731 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
732 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
733 };
734 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
735 int error;
736
737 error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
738 attr, userspace_policy);
739 if (error)
740 return error;
741
742 if (!a[OVS_USERSPACE_ATTR_PID] ||
743 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
744 return -EINVAL;
745
746 return 0;
747}
748
749static int copy_action(const struct nlattr *from,
750 struct sw_flow_actions **sfa)
751{
752 int totlen = NLA_ALIGN(from->nla_len);
753 struct nlattr *to;
754
755 to = reserve_sfa_size(sfa, from->nla_len);
756 if (IS_ERR(to))
757 return PTR_ERR(to);
758
759 memcpy(to, from, totlen);
760 return 0;
761}
762
763static int validate_and_copy_actions(const struct nlattr *attr,
764 const struct sw_flow_key *key,
765 int depth,
766 struct sw_flow_actions **sfa)
767{
768 const struct nlattr *a;
769 int rem, err;
770
771 if (depth >= SAMPLE_ACTION_DEPTH)
772 return -EOVERFLOW;
773
774 nla_for_each_nested(a, attr, rem) {
775 /* Expected argument lengths, (u32)-1 for variable length. */
776 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
777 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
778 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
779 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
780 [OVS_ACTION_ATTR_POP_VLAN] = 0,
781 [OVS_ACTION_ATTR_SET] = (u32)-1,
782 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1
783 };
784 const struct ovs_action_push_vlan *vlan;
785 int type = nla_type(a);
786 bool skip_copy;
787
788 if (type > OVS_ACTION_ATTR_MAX ||
789 (action_lens[type] != nla_len(a) &&
790 action_lens[type] != (u32)-1))
791 return -EINVAL;
792
793 skip_copy = false;
794 switch (type) {
795 case OVS_ACTION_ATTR_UNSPEC:
796 return -EINVAL;
797
798 case OVS_ACTION_ATTR_USERSPACE:
799 err = validate_userspace(a);
800 if (err)
801 return err;
802 break;
803
804 case OVS_ACTION_ATTR_OUTPUT:
805 if (nla_get_u32(a) >= DP_MAX_PORTS)
806 return -EINVAL;
807 break;
808
809
810 case OVS_ACTION_ATTR_POP_VLAN:
811 break;
812
813 case OVS_ACTION_ATTR_PUSH_VLAN:
814 vlan = nla_data(a);
815 if (vlan->vlan_tpid != htons(ETH_P_8021Q))
816 return -EINVAL;
817 if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
818 return -EINVAL;
819 break;
820
821 case OVS_ACTION_ATTR_SET:
822 err = validate_set(a, key, sfa, &skip_copy);
823 if (err)
824 return err;
825 break;
826
827 case OVS_ACTION_ATTR_SAMPLE:
828 err = validate_and_copy_sample(a, key, depth, sfa);
829 if (err)
830 return err;
831 skip_copy = true;
832 break;
833
834 default:
835 return -EINVAL;
836 }
837 if (!skip_copy) {
838 err = copy_action(a, sfa);
839 if (err)
840 return err;
841 }
842 }
843
844 if (rem > 0)
845 return -EINVAL;
846
847 return 0;
848}
849
850static void clear_stats(struct sw_flow *flow) 456static void clear_stats(struct sw_flow *flow)
851{ 457{
852 flow->used = 0; 458 flow->used = 0;
@@ -902,15 +508,16 @@ static int ovs_packet_cmd_execute(struct sk_buff *skb, struct genl_info *info)
902 if (err) 508 if (err)
903 goto err_flow_free; 509 goto err_flow_free;
904 510
905 err = ovs_flow_metadata_from_nlattrs(flow, a[OVS_PACKET_ATTR_KEY]); 511 err = ovs_nla_get_flow_metadata(flow, a[OVS_PACKET_ATTR_KEY]);
906 if (err) 512 if (err)
907 goto err_flow_free; 513 goto err_flow_free;
908 acts = ovs_flow_actions_alloc(nla_len(a[OVS_PACKET_ATTR_ACTIONS])); 514 acts = ovs_nla_alloc_flow_actions(nla_len(a[OVS_PACKET_ATTR_ACTIONS]));
909 err = PTR_ERR(acts); 515 err = PTR_ERR(acts);
910 if (IS_ERR(acts)) 516 if (IS_ERR(acts))
911 goto err_flow_free; 517 goto err_flow_free;
912 518
913 err = validate_and_copy_actions(a[OVS_PACKET_ATTR_ACTIONS], &flow->key, 0, &acts); 519 err = ovs_nla_copy_actions(a[OVS_PACKET_ATTR_ACTIONS],
520 &flow->key, 0, &acts);
914 rcu_assign_pointer(flow->sf_acts, acts); 521 rcu_assign_pointer(flow->sf_acts, acts);
915 if (err) 522 if (err)
916 goto err_flow_free; 523 goto err_flow_free;
@@ -958,15 +565,18 @@ static struct genl_ops dp_packet_genl_ops[] = {
958 } 565 }
959}; 566};
960 567
961static void get_dp_stats(struct datapath *dp, struct ovs_dp_stats *stats) 568static void get_dp_stats(struct datapath *dp, struct ovs_dp_stats *stats,
569 struct ovs_dp_megaflow_stats *mega_stats)
962{ 570{
963 struct flow_table *table;
964 int i; 571 int i;
965 572
966 table = rcu_dereference_check(dp->table, lockdep_ovsl_is_held()); 573 memset(mega_stats, 0, sizeof(*mega_stats));
967 stats->n_flows = ovs_flow_tbl_count(table); 574
575 stats->n_flows = ovs_flow_tbl_count(&dp->table);
576 mega_stats->n_masks = ovs_flow_tbl_num_masks(&dp->table);
968 577
969 stats->n_hit = stats->n_missed = stats->n_lost = 0; 578 stats->n_hit = stats->n_missed = stats->n_lost = 0;
579
970 for_each_possible_cpu(i) { 580 for_each_possible_cpu(i) {
971 const struct dp_stats_percpu *percpu_stats; 581 const struct dp_stats_percpu *percpu_stats;
972 struct dp_stats_percpu local_stats; 582 struct dp_stats_percpu local_stats;
@@ -982,6 +592,7 @@ static void get_dp_stats(struct datapath *dp, struct ovs_dp_stats *stats)
982 stats->n_hit += local_stats.n_hit; 592 stats->n_hit += local_stats.n_hit;
983 stats->n_missed += local_stats.n_missed; 593 stats->n_missed += local_stats.n_missed;
984 stats->n_lost += local_stats.n_lost; 594 stats->n_lost += local_stats.n_lost;
595 mega_stats->n_mask_hit += local_stats.n_mask_hit;
985 } 596 }
986} 597}
987 598
@@ -1005,100 +616,6 @@ static struct genl_multicast_group ovs_dp_flow_multicast_group = {
1005 .name = OVS_FLOW_MCGROUP 616 .name = OVS_FLOW_MCGROUP
1006}; 617};
1007 618
1008static int actions_to_attr(const struct nlattr *attr, int len, struct sk_buff *skb);
1009static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
1010{
1011 const struct nlattr *a;
1012 struct nlattr *start;
1013 int err = 0, rem;
1014
1015 start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
1016 if (!start)
1017 return -EMSGSIZE;
1018
1019 nla_for_each_nested(a, attr, rem) {
1020 int type = nla_type(a);
1021 struct nlattr *st_sample;
1022
1023 switch (type) {
1024 case OVS_SAMPLE_ATTR_PROBABILITY:
1025 if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY, sizeof(u32), nla_data(a)))
1026 return -EMSGSIZE;
1027 break;
1028 case OVS_SAMPLE_ATTR_ACTIONS:
1029 st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
1030 if (!st_sample)
1031 return -EMSGSIZE;
1032 err = actions_to_attr(nla_data(a), nla_len(a), skb);
1033 if (err)
1034 return err;
1035 nla_nest_end(skb, st_sample);
1036 break;
1037 }
1038 }
1039
1040 nla_nest_end(skb, start);
1041 return err;
1042}
1043
1044static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
1045{
1046 const struct nlattr *ovs_key = nla_data(a);
1047 int key_type = nla_type(ovs_key);
1048 struct nlattr *start;
1049 int err;
1050
1051 switch (key_type) {
1052 case OVS_KEY_ATTR_IPV4_TUNNEL:
1053 start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
1054 if (!start)
1055 return -EMSGSIZE;
1056
1057 err = ovs_ipv4_tun_to_nlattr(skb, nla_data(ovs_key),
1058 nla_data(ovs_key));
1059 if (err)
1060 return err;
1061 nla_nest_end(skb, start);
1062 break;
1063 default:
1064 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
1065 return -EMSGSIZE;
1066 break;
1067 }
1068
1069 return 0;
1070}
1071
1072static int actions_to_attr(const struct nlattr *attr, int len, struct sk_buff *skb)
1073{
1074 const struct nlattr *a;
1075 int rem, err;
1076
1077 nla_for_each_attr(a, attr, len, rem) {
1078 int type = nla_type(a);
1079
1080 switch (type) {
1081 case OVS_ACTION_ATTR_SET:
1082 err = set_action_to_attr(a, skb);
1083 if (err)
1084 return err;
1085 break;
1086
1087 case OVS_ACTION_ATTR_SAMPLE:
1088 err = sample_action_to_attr(a, skb);
1089 if (err)
1090 return err;
1091 break;
1092 default:
1093 if (nla_put(skb, type, nla_len(a), nla_data(a)))
1094 return -EMSGSIZE;
1095 break;
1096 }
1097 }
1098
1099 return 0;
1100}
1101
1102static size_t ovs_flow_cmd_msg_size(const struct sw_flow_actions *acts) 619static size_t ovs_flow_cmd_msg_size(const struct sw_flow_actions *acts)
1103{ 620{
1104 return NLMSG_ALIGN(sizeof(struct ovs_header)) 621 return NLMSG_ALIGN(sizeof(struct ovs_header))
@@ -1135,8 +652,7 @@ static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp,
1135 if (!nla) 652 if (!nla)
1136 goto nla_put_failure; 653 goto nla_put_failure;
1137 654
1138 err = ovs_flow_to_nlattrs(&flow->unmasked_key, 655 err = ovs_nla_put_flow(&flow->unmasked_key, &flow->unmasked_key, skb);
1139 &flow->unmasked_key, skb);
1140 if (err) 656 if (err)
1141 goto error; 657 goto error;
1142 nla_nest_end(skb, nla); 658 nla_nest_end(skb, nla);
@@ -1145,7 +661,7 @@ static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp,
1145 if (!nla) 661 if (!nla)
1146 goto nla_put_failure; 662 goto nla_put_failure;
1147 663
1148 err = ovs_flow_to_nlattrs(&flow->key, &flow->mask->key, skb); 664 err = ovs_nla_put_flow(&flow->key, &flow->mask->key, skb);
1149 if (err) 665 if (err)
1150 goto error; 666 goto error;
1151 667
@@ -1155,7 +671,7 @@ static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp,
1155 used = flow->used; 671 used = flow->used;
1156 stats.n_packets = flow->packet_count; 672 stats.n_packets = flow->packet_count;
1157 stats.n_bytes = flow->byte_count; 673 stats.n_bytes = flow->byte_count;
1158 tcp_flags = flow->tcp_flags; 674 tcp_flags = (u8)ntohs(flow->tcp_flags);
1159 spin_unlock_bh(&flow->lock); 675 spin_unlock_bh(&flow->lock);
1160 676
1161 if (used && 677 if (used &&
@@ -1188,7 +704,8 @@ static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp,
1188 sf_acts = rcu_dereference_check(flow->sf_acts, 704 sf_acts = rcu_dereference_check(flow->sf_acts,
1189 lockdep_ovsl_is_held()); 705 lockdep_ovsl_is_held());
1190 706
1191 err = actions_to_attr(sf_acts->actions, sf_acts->actions_len, skb); 707 err = ovs_nla_put_actions(sf_acts->actions,
708 sf_acts->actions_len, skb);
1192 if (!err) 709 if (!err)
1193 nla_nest_end(skb, start); 710 nla_nest_end(skb, start);
1194 else { 711 else {
@@ -1234,6 +751,14 @@ static struct sk_buff *ovs_flow_cmd_build_info(struct sw_flow *flow,
1234 return skb; 751 return skb;
1235} 752}
1236 753
754static struct sw_flow *__ovs_flow_tbl_lookup(struct flow_table *tbl,
755 const struct sw_flow_key *key)
756{
757 u32 __always_unused n_mask_hit;
758
759 return ovs_flow_tbl_lookup(tbl, key, &n_mask_hit);
760}
761
1237static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info) 762static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
1238{ 763{
1239 struct nlattr **a = info->attrs; 764 struct nlattr **a = info->attrs;
@@ -1243,7 +768,6 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
1243 struct sw_flow_mask mask; 768 struct sw_flow_mask mask;
1244 struct sk_buff *reply; 769 struct sk_buff *reply;
1245 struct datapath *dp; 770 struct datapath *dp;
1246 struct flow_table *table;
1247 struct sw_flow_actions *acts = NULL; 771 struct sw_flow_actions *acts = NULL;
1248 struct sw_flow_match match; 772 struct sw_flow_match match;
1249 int error; 773 int error;
@@ -1254,21 +778,21 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
1254 goto error; 778 goto error;
1255 779
1256 ovs_match_init(&match, &key, &mask); 780 ovs_match_init(&match, &key, &mask);
1257 error = ovs_match_from_nlattrs(&match, 781 error = ovs_nla_get_match(&match,
1258 a[OVS_FLOW_ATTR_KEY], a[OVS_FLOW_ATTR_MASK]); 782 a[OVS_FLOW_ATTR_KEY], a[OVS_FLOW_ATTR_MASK]);
1259 if (error) 783 if (error)
1260 goto error; 784 goto error;
1261 785
1262 /* Validate actions. */ 786 /* Validate actions. */
1263 if (a[OVS_FLOW_ATTR_ACTIONS]) { 787 if (a[OVS_FLOW_ATTR_ACTIONS]) {
1264 acts = ovs_flow_actions_alloc(nla_len(a[OVS_FLOW_ATTR_ACTIONS])); 788 acts = ovs_nla_alloc_flow_actions(nla_len(a[OVS_FLOW_ATTR_ACTIONS]));
1265 error = PTR_ERR(acts); 789 error = PTR_ERR(acts);
1266 if (IS_ERR(acts)) 790 if (IS_ERR(acts))
1267 goto error; 791 goto error;
1268 792
1269 ovs_flow_key_mask(&masked_key, &key, &mask); 793 ovs_flow_mask_key(&masked_key, &key, &mask);
1270 error = validate_and_copy_actions(a[OVS_FLOW_ATTR_ACTIONS], 794 error = ovs_nla_copy_actions(a[OVS_FLOW_ATTR_ACTIONS],
1271 &masked_key, 0, &acts); 795 &masked_key, 0, &acts);
1272 if (error) { 796 if (error) {
1273 OVS_NLERR("Flow actions may not be safe on all matching packets.\n"); 797 OVS_NLERR("Flow actions may not be safe on all matching packets.\n");
1274 goto err_kfree; 798 goto err_kfree;
@@ -1284,29 +808,14 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
1284 if (!dp) 808 if (!dp)
1285 goto err_unlock_ovs; 809 goto err_unlock_ovs;
1286 810
1287 table = ovsl_dereference(dp->table);
1288
1289 /* Check if this is a duplicate flow */ 811 /* Check if this is a duplicate flow */
1290 flow = ovs_flow_lookup(table, &key); 812 flow = __ovs_flow_tbl_lookup(&dp->table, &key);
1291 if (!flow) { 813 if (!flow) {
1292 struct sw_flow_mask *mask_p;
1293 /* Bail out if we're not allowed to create a new flow. */ 814 /* Bail out if we're not allowed to create a new flow. */
1294 error = -ENOENT; 815 error = -ENOENT;
1295 if (info->genlhdr->cmd == OVS_FLOW_CMD_SET) 816 if (info->genlhdr->cmd == OVS_FLOW_CMD_SET)
1296 goto err_unlock_ovs; 817 goto err_unlock_ovs;
1297 818
1298 /* Expand table, if necessary, to make room. */
1299 if (ovs_flow_tbl_need_to_expand(table)) {
1300 struct flow_table *new_table;
1301
1302 new_table = ovs_flow_tbl_expand(table);
1303 if (!IS_ERR(new_table)) {
1304 rcu_assign_pointer(dp->table, new_table);
1305 ovs_flow_tbl_destroy(table, true);
1306 table = ovsl_dereference(dp->table);
1307 }
1308 }
1309
1310 /* Allocate flow. */ 819 /* Allocate flow. */
1311 flow = ovs_flow_alloc(); 820 flow = ovs_flow_alloc();
1312 if (IS_ERR(flow)) { 821 if (IS_ERR(flow)) {
@@ -1317,25 +826,14 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
1317 826
1318 flow->key = masked_key; 827 flow->key = masked_key;
1319 flow->unmasked_key = key; 828 flow->unmasked_key = key;
1320
1321 /* Make sure mask is unique in the system */
1322 mask_p = ovs_sw_flow_mask_find(table, &mask);
1323 if (!mask_p) {
1324 /* Allocate a new mask if none exsits. */
1325 mask_p = ovs_sw_flow_mask_alloc();
1326 if (!mask_p)
1327 goto err_flow_free;
1328 mask_p->key = mask.key;
1329 mask_p->range = mask.range;
1330 ovs_sw_flow_mask_insert(table, mask_p);
1331 }
1332
1333 ovs_sw_flow_mask_add_ref(mask_p);
1334 flow->mask = mask_p;
1335 rcu_assign_pointer(flow->sf_acts, acts); 829 rcu_assign_pointer(flow->sf_acts, acts);
1336 830
1337 /* Put flow in bucket. */ 831 /* Put flow in bucket. */
1338 ovs_flow_insert(table, flow); 832 error = ovs_flow_tbl_insert(&dp->table, flow, &mask);
833 if (error) {
834 acts = NULL;
835 goto err_flow_free;
836 }
1339 837
1340 reply = ovs_flow_cmd_build_info(flow, dp, info->snd_portid, 838 reply = ovs_flow_cmd_build_info(flow, dp, info->snd_portid,
1341 info->snd_seq, OVS_FLOW_CMD_NEW); 839 info->snd_seq, OVS_FLOW_CMD_NEW);
@@ -1356,7 +854,7 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
1356 854
1357 /* The unmasked key has to be the same for flow updates. */ 855 /* The unmasked key has to be the same for flow updates. */
1358 error = -EINVAL; 856 error = -EINVAL;
1359 if (!ovs_flow_cmp_unmasked_key(flow, &key, match.range.end)) { 857 if (!ovs_flow_cmp_unmasked_key(flow, &match)) {
1360 OVS_NLERR("Flow modification message rejected, unmasked key does not match.\n"); 858 OVS_NLERR("Flow modification message rejected, unmasked key does not match.\n");
1361 goto err_unlock_ovs; 859 goto err_unlock_ovs;
1362 } 860 }
@@ -1364,7 +862,7 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
1364 /* Update actions. */ 862 /* Update actions. */
1365 old_acts = ovsl_dereference(flow->sf_acts); 863 old_acts = ovsl_dereference(flow->sf_acts);
1366 rcu_assign_pointer(flow->sf_acts, acts); 864 rcu_assign_pointer(flow->sf_acts, acts);
1367 ovs_flow_deferred_free_acts(old_acts); 865 ovs_nla_free_flow_actions(old_acts);
1368 866
1369 reply = ovs_flow_cmd_build_info(flow, dp, info->snd_portid, 867 reply = ovs_flow_cmd_build_info(flow, dp, info->snd_portid,
1370 info->snd_seq, OVS_FLOW_CMD_NEW); 868 info->snd_seq, OVS_FLOW_CMD_NEW);
@@ -1403,7 +901,6 @@ static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info)
1403 struct sk_buff *reply; 901 struct sk_buff *reply;
1404 struct sw_flow *flow; 902 struct sw_flow *flow;
1405 struct datapath *dp; 903 struct datapath *dp;
1406 struct flow_table *table;
1407 struct sw_flow_match match; 904 struct sw_flow_match match;
1408 int err; 905 int err;
1409 906
@@ -1413,7 +910,7 @@ static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info)
1413 } 910 }
1414 911
1415 ovs_match_init(&match, &key, NULL); 912 ovs_match_init(&match, &key, NULL);
1416 err = ovs_match_from_nlattrs(&match, a[OVS_FLOW_ATTR_KEY], NULL); 913 err = ovs_nla_get_match(&match, a[OVS_FLOW_ATTR_KEY], NULL);
1417 if (err) 914 if (err)
1418 return err; 915 return err;
1419 916
@@ -1424,9 +921,8 @@ static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info)
1424 goto unlock; 921 goto unlock;
1425 } 922 }
1426 923
1427 table = ovsl_dereference(dp->table); 924 flow = __ovs_flow_tbl_lookup(&dp->table, &key);
1428 flow = ovs_flow_lookup_unmasked_key(table, &match); 925 if (!flow || !ovs_flow_cmp_unmasked_key(flow, &match)) {
1429 if (!flow) {
1430 err = -ENOENT; 926 err = -ENOENT;
1431 goto unlock; 927 goto unlock;
1432 } 928 }
@@ -1453,7 +949,6 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
1453 struct sk_buff *reply; 949 struct sk_buff *reply;
1454 struct sw_flow *flow; 950 struct sw_flow *flow;
1455 struct datapath *dp; 951 struct datapath *dp;
1456 struct flow_table *table;
1457 struct sw_flow_match match; 952 struct sw_flow_match match;
1458 int err; 953 int err;
1459 954
@@ -1465,18 +960,17 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
1465 } 960 }
1466 961
1467 if (!a[OVS_FLOW_ATTR_KEY]) { 962 if (!a[OVS_FLOW_ATTR_KEY]) {
1468 err = flush_flows(dp); 963 err = ovs_flow_tbl_flush(&dp->table);
1469 goto unlock; 964 goto unlock;
1470 } 965 }
1471 966
1472 ovs_match_init(&match, &key, NULL); 967 ovs_match_init(&match, &key, NULL);
1473 err = ovs_match_from_nlattrs(&match, a[OVS_FLOW_ATTR_KEY], NULL); 968 err = ovs_nla_get_match(&match, a[OVS_FLOW_ATTR_KEY], NULL);
1474 if (err) 969 if (err)
1475 goto unlock; 970 goto unlock;
1476 971
1477 table = ovsl_dereference(dp->table); 972 flow = __ovs_flow_tbl_lookup(&dp->table, &key);
1478 flow = ovs_flow_lookup_unmasked_key(table, &match); 973 if (!flow || !ovs_flow_cmp_unmasked_key(flow, &match)) {
1479 if (!flow) {
1480 err = -ENOENT; 974 err = -ENOENT;
1481 goto unlock; 975 goto unlock;
1482 } 976 }
@@ -1487,7 +981,7 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
1487 goto unlock; 981 goto unlock;
1488 } 982 }
1489 983
1490 ovs_flow_remove(table, flow); 984 ovs_flow_tbl_remove(&dp->table, flow);
1491 985
1492 err = ovs_flow_cmd_fill_info(flow, dp, reply, info->snd_portid, 986 err = ovs_flow_cmd_fill_info(flow, dp, reply, info->snd_portid,
1493 info->snd_seq, 0, OVS_FLOW_CMD_DEL); 987 info->snd_seq, 0, OVS_FLOW_CMD_DEL);
@@ -1506,8 +1000,8 @@ unlock:
1506static int ovs_flow_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb) 1000static int ovs_flow_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
1507{ 1001{
1508 struct ovs_header *ovs_header = genlmsg_data(nlmsg_data(cb->nlh)); 1002 struct ovs_header *ovs_header = genlmsg_data(nlmsg_data(cb->nlh));
1003 struct table_instance *ti;
1509 struct datapath *dp; 1004 struct datapath *dp;
1510 struct flow_table *table;
1511 1005
1512 rcu_read_lock(); 1006 rcu_read_lock();
1513 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex); 1007 dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
@@ -1516,14 +1010,14 @@ static int ovs_flow_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
1516 return -ENODEV; 1010 return -ENODEV;
1517 } 1011 }
1518 1012
1519 table = rcu_dereference(dp->table); 1013 ti = rcu_dereference(dp->table.ti);
1520 for (;;) { 1014 for (;;) {
1521 struct sw_flow *flow; 1015 struct sw_flow *flow;
1522 u32 bucket, obj; 1016 u32 bucket, obj;
1523 1017
1524 bucket = cb->args[0]; 1018 bucket = cb->args[0];
1525 obj = cb->args[1]; 1019 obj = cb->args[1];
1526 flow = ovs_flow_dump_next(table, &bucket, &obj); 1020 flow = ovs_flow_tbl_dump_next(ti, &bucket, &obj);
1527 if (!flow) 1021 if (!flow)
1528 break; 1022 break;
1529 1023
@@ -1589,6 +1083,7 @@ static size_t ovs_dp_cmd_msg_size(void)
1589 1083
1590 msgsize += nla_total_size(IFNAMSIZ); 1084 msgsize += nla_total_size(IFNAMSIZ);
1591 msgsize += nla_total_size(sizeof(struct ovs_dp_stats)); 1085 msgsize += nla_total_size(sizeof(struct ovs_dp_stats));
1086 msgsize += nla_total_size(sizeof(struct ovs_dp_megaflow_stats));
1592 1087
1593 return msgsize; 1088 return msgsize;
1594} 1089}
@@ -1598,6 +1093,7 @@ static int ovs_dp_cmd_fill_info(struct datapath *dp, struct sk_buff *skb,
1598{ 1093{
1599 struct ovs_header *ovs_header; 1094 struct ovs_header *ovs_header;
1600 struct ovs_dp_stats dp_stats; 1095 struct ovs_dp_stats dp_stats;
1096 struct ovs_dp_megaflow_stats dp_megaflow_stats;
1601 int err; 1097 int err;
1602 1098
1603 ovs_header = genlmsg_put(skb, portid, seq, &dp_datapath_genl_family, 1099 ovs_header = genlmsg_put(skb, portid, seq, &dp_datapath_genl_family,
@@ -1613,8 +1109,14 @@ static int ovs_dp_cmd_fill_info(struct datapath *dp, struct sk_buff *skb,
1613 if (err) 1109 if (err)
1614 goto nla_put_failure; 1110 goto nla_put_failure;
1615 1111
1616 get_dp_stats(dp, &dp_stats); 1112 get_dp_stats(dp, &dp_stats, &dp_megaflow_stats);
1617 if (nla_put(skb, OVS_DP_ATTR_STATS, sizeof(struct ovs_dp_stats), &dp_stats)) 1113 if (nla_put(skb, OVS_DP_ATTR_STATS, sizeof(struct ovs_dp_stats),
1114 &dp_stats))
1115 goto nla_put_failure;
1116
1117 if (nla_put(skb, OVS_DP_ATTR_MEGAFLOW_STATS,
1118 sizeof(struct ovs_dp_megaflow_stats),
1119 &dp_megaflow_stats))
1618 goto nla_put_failure; 1120 goto nla_put_failure;
1619 1121
1620 return genlmsg_end(skb, ovs_header); 1122 return genlmsg_end(skb, ovs_header);
@@ -1687,9 +1189,8 @@ static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info)
1687 ovs_dp_set_net(dp, hold_net(sock_net(skb->sk))); 1189 ovs_dp_set_net(dp, hold_net(sock_net(skb->sk)));
1688 1190
1689 /* Allocate table. */ 1191 /* Allocate table. */
1690 err = -ENOMEM; 1192 err = ovs_flow_tbl_init(&dp->table);
1691 rcu_assign_pointer(dp->table, ovs_flow_tbl_alloc(TBL_MIN_BUCKETS)); 1193 if (err)
1692 if (!dp->table)
1693 goto err_free_dp; 1194 goto err_free_dp;
1694 1195
1695 dp->stats_percpu = alloc_percpu(struct dp_stats_percpu); 1196 dp->stats_percpu = alloc_percpu(struct dp_stats_percpu);
@@ -1699,7 +1200,7 @@ static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info)
1699 } 1200 }
1700 1201
1701 dp->ports = kmalloc(DP_VPORT_HASH_BUCKETS * sizeof(struct hlist_head), 1202 dp->ports = kmalloc(DP_VPORT_HASH_BUCKETS * sizeof(struct hlist_head),
1702 GFP_KERNEL); 1203 GFP_KERNEL);
1703 if (!dp->ports) { 1204 if (!dp->ports) {
1704 err = -ENOMEM; 1205 err = -ENOMEM;
1705 goto err_destroy_percpu; 1206 goto err_destroy_percpu;
@@ -1746,7 +1247,7 @@ err_destroy_ports_array:
1746err_destroy_percpu: 1247err_destroy_percpu:
1747 free_percpu(dp->stats_percpu); 1248 free_percpu(dp->stats_percpu);
1748err_destroy_table: 1249err_destroy_table:
1749 ovs_flow_tbl_destroy(ovsl_dereference(dp->table), false); 1250 ovs_flow_tbl_destroy(&dp->table);
1750err_free_dp: 1251err_free_dp:
1751 release_net(ovs_dp_get_net(dp)); 1252 release_net(ovs_dp_get_net(dp));
1752 kfree(dp); 1253 kfree(dp);
@@ -2336,32 +1837,6 @@ error:
2336 return err; 1837 return err;
2337} 1838}
2338 1839
2339static void rehash_flow_table(struct work_struct *work)
2340{
2341 struct datapath *dp;
2342 struct net *net;
2343
2344 ovs_lock();
2345 rtnl_lock();
2346 for_each_net(net) {
2347 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
2348
2349 list_for_each_entry(dp, &ovs_net->dps, list_node) {
2350 struct flow_table *old_table = ovsl_dereference(dp->table);
2351 struct flow_table *new_table;
2352
2353 new_table = ovs_flow_tbl_rehash(old_table);
2354 if (!IS_ERR(new_table)) {
2355 rcu_assign_pointer(dp->table, new_table);
2356 ovs_flow_tbl_destroy(old_table, true);
2357 }
2358 }
2359 }
2360 rtnl_unlock();
2361 ovs_unlock();
2362 schedule_delayed_work(&rehash_flow_wq, REHASH_FLOW_INTERVAL);
2363}
2364
2365static int __net_init ovs_init_net(struct net *net) 1840static int __net_init ovs_init_net(struct net *net)
2366{ 1841{
2367 struct ovs_net *ovs_net = net_generic(net, ovs_net_id); 1842 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
@@ -2419,8 +1894,6 @@ static int __init dp_init(void)
2419 if (err < 0) 1894 if (err < 0)
2420 goto error_unreg_notifier; 1895 goto error_unreg_notifier;
2421 1896
2422 schedule_delayed_work(&rehash_flow_wq, REHASH_FLOW_INTERVAL);
2423
2424 return 0; 1897 return 0;
2425 1898
2426error_unreg_notifier: 1899error_unreg_notifier:
@@ -2437,7 +1910,6 @@ error:
2437 1910
2438static void dp_cleanup(void) 1911static void dp_cleanup(void)
2439{ 1912{
2440 cancel_delayed_work_sync(&rehash_flow_wq);
2441 dp_unregister_genl(ARRAY_SIZE(dp_genl_families)); 1913 dp_unregister_genl(ARRAY_SIZE(dp_genl_families));
2442 unregister_netdevice_notifier(&ovs_dp_device_notifier); 1914 unregister_netdevice_notifier(&ovs_dp_device_notifier);
2443 unregister_pernet_device(&ovs_net_ops); 1915 unregister_pernet_device(&ovs_net_ops);
diff --git a/net/openvswitch/datapath.h b/net/openvswitch/datapath.h
index 4d109c176ef3..d3d14a58aa91 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
@@ -45,11 +46,15 @@
45 * @n_lost: Number of received packets that had no matching flow in the flow 46 * @n_lost: Number of received packets that had no matching flow in the flow
46 * table that could not be sent to userspace (normally due to an overflow in 47 * table that could not be sent to userspace (normally due to an overflow in
47 * one of the datapath's queues). 48 * one of the datapath's queues).
49 * @n_mask_hit: Number of masks looked up for flow match.
50 * @n_mask_hit / (@n_hit + @n_missed) will be the average masks looked
51 * up per packet.
48 */ 52 */
49struct dp_stats_percpu { 53struct dp_stats_percpu {
50 u64 n_hit; 54 u64 n_hit;
51 u64 n_missed; 55 u64 n_missed;
52 u64 n_lost; 56 u64 n_lost;
57 u64 n_mask_hit;
53 struct u64_stats_sync sync; 58 struct u64_stats_sync sync;
54}; 59};
55 60
@@ -57,7 +62,7 @@ struct dp_stats_percpu {
57 * struct datapath - datapath for flow-based packet switching 62 * struct datapath - datapath for flow-based packet switching
58 * @rcu: RCU callback head for deferred destruction. 63 * @rcu: RCU callback head for deferred destruction.
59 * @list_node: Element in global 'dps' list. 64 * @list_node: Element in global 'dps' list.
60 * @table: Current flow table. Protected by ovs_mutex and RCU. 65 * @table: flow table.
61 * @ports: Hash table for ports. %OVSP_LOCAL port always exists. Protected by 66 * @ports: Hash table for ports. %OVSP_LOCAL port always exists. Protected by
62 * ovs_mutex and RCU. 67 * ovs_mutex and RCU.
63 * @stats_percpu: Per-CPU datapath statistics. 68 * @stats_percpu: Per-CPU datapath statistics.
@@ -71,7 +76,7 @@ struct datapath {
71 struct list_head list_node; 76 struct list_head list_node;
72 77
73 /* Flow table. */ 78 /* Flow table. */
74 struct flow_table __rcu *table; 79 struct flow_table table;
75 80
76 /* Switch ports. */ 81 /* Switch ports. */
77 struct hlist_head *ports; 82 struct hlist_head *ports;
diff --git a/net/openvswitch/flow.c b/net/openvswitch/flow.c
index 410db90db73d..b409f5279601 100644
--- a/net/openvswitch/flow.c
+++ b/net/openvswitch/flow.c
@@ -45,202 +45,38 @@
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_BE16(tp) (*(__be16 *)&tcp_flag_word(tp) & htons(0x0FFF))
111 struct sw_flow_key *key,
112 struct sw_flow_mask *mask)
113{
114 memset(match, 0, sizeof(*match));
115 match->key = key;
116 match->mask = mask;
117 62
118 memset(key, 0, sizeof(*key)); 63void ovs_flow_used(struct sw_flow *flow, struct sk_buff *skb)
119
120 if (mask) {
121 memset(&mask->key, 0, sizeof(mask->key));
122 mask->range.start = mask->range.end = 0;
123 }
124}
125
126static bool ovs_match_validate(const struct sw_flow_match *match,
127 u64 key_attrs, u64 mask_attrs)
128{ 64{
129 u64 key_expected = 1 << OVS_KEY_ATTR_ETHERNET; 65 __be16 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 66
157 if (match->key->eth.type == htons(ETH_P_IP)) { 67 if ((flow->key.eth.type == htons(ETH_P_IP) ||
158 key_expected |= 1 << OVS_KEY_ATTR_IPV4; 68 flow->key.eth.type == htons(ETH_P_IPV6)) &&
159 if (match->mask && (match->mask->key.eth.type == htons(0xffff))) 69 flow->key.ip.proto == IPPROTO_TCP &&
160 mask_allowed |= 1 << OVS_KEY_ATTR_IPV4; 70 likely(skb->len >= skb_transport_offset(skb) + sizeof(struct tcphdr))) {
161 71 tcp_flags = TCP_FLAGS_BE16(tcp_hdr(skb));
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
236 if ((mask_attrs & mask_allowed) != mask_attrs) {
237 /* Mask attributes check failed. */
238 OVS_NLERR("Contain more than allowed mask fields (mask_attrs=%llx, mask_allowed=%llx).\n",
239 mask_attrs, mask_allowed);
240 return false;
241 } 72 }
242 73
243 return true; 74 spin_lock(&flow->lock);
75 flow->used = jiffies;
76 flow->packet_count++;
77 flow->byte_count += skb->len;
78 flow->tcp_flags |= tcp_flags;
79 spin_unlock(&flow->lock);
244} 80}
245 81
246static int check_header(struct sk_buff *skb, int len) 82static int check_header(struct sk_buff *skb, int len)
@@ -311,19 +147,6 @@ static bool icmphdr_ok(struct sk_buff *skb)
311 sizeof(struct icmphdr)); 147 sizeof(struct icmphdr));
312} 148}
313 149
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) 150static int parse_ipv6hdr(struct sk_buff *skb, struct sw_flow_key *key)
328{ 151{
329 unsigned int nh_ofs = skb_network_offset(skb); 152 unsigned int nh_ofs = skb_network_offset(skb);
@@ -372,311 +195,6 @@ static bool icmp6hdr_ok(struct sk_buff *skb)
372 sizeof(struct icmp6hdr)); 195 sizeof(struct icmp6hdr));
373} 196}
374 197
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) 198static int parse_vlan(struct sk_buff *skb, struct sw_flow_key *key)
681{ 199{
682 struct qtag_prefix { 200 struct qtag_prefix {
@@ -910,6 +428,7 @@ int ovs_flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key)
910 struct tcphdr *tcp = tcp_hdr(skb); 428 struct tcphdr *tcp = tcp_hdr(skb);
911 key->ipv4.tp.src = tcp->source; 429 key->ipv4.tp.src = tcp->source;
912 key->ipv4.tp.dst = tcp->dest; 430 key->ipv4.tp.dst = tcp->dest;
431 key->ipv4.tp.flags = TCP_FLAGS_BE16(tcp);
913 } 432 }
914 } else if (key->ip.proto == IPPROTO_UDP) { 433 } else if (key->ip.proto == IPPROTO_UDP) {
915 if (udphdr_ok(skb)) { 434 if (udphdr_ok(skb)) {
@@ -978,6 +497,7 @@ int ovs_flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key)
978 struct tcphdr *tcp = tcp_hdr(skb); 497 struct tcphdr *tcp = tcp_hdr(skb);
979 key->ipv6.tp.src = tcp->source; 498 key->ipv6.tp.src = tcp->source;
980 key->ipv6.tp.dst = tcp->dest; 499 key->ipv6.tp.dst = tcp->dest;
500 key->ipv6.tp.flags = TCP_FLAGS_BE16(tcp);
981 } 501 }
982 } else if (key->ip.proto == NEXTHDR_UDP) { 502 } else if (key->ip.proto == NEXTHDR_UDP) {
983 if (udphdr_ok(skb)) { 503 if (udphdr_ok(skb)) {
@@ -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..1510f51dbf74 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 \
@@ -101,6 +93,7 @@ struct sw_flow_key {
101 struct { 93 struct {
102 __be16 src; /* TCP/UDP/SCTP source port. */ 94 __be16 src; /* TCP/UDP/SCTP source port. */
103 __be16 dst; /* TCP/UDP/SCTP destination port. */ 95 __be16 dst; /* TCP/UDP/SCTP destination port. */
96 __be16 flags; /* TCP flags. */
104 } tp; 97 } tp;
105 struct { 98 struct {
106 u8 sha[ETH_ALEN]; /* ARP source hardware address. */ 99 u8 sha[ETH_ALEN]; /* ARP source hardware address. */
@@ -117,6 +110,7 @@ struct sw_flow_key {
117 struct { 110 struct {
118 __be16 src; /* TCP/UDP/SCTP source port. */ 111 __be16 src; /* TCP/UDP/SCTP source port. */
119 __be16 dst; /* TCP/UDP/SCTP destination port. */ 112 __be16 dst; /* TCP/UDP/SCTP destination port. */
113 __be16 flags; /* TCP flags. */
120 } tp; 114 } tp;
121 struct { 115 struct {
122 struct in6_addr target; /* ND target address. */ 116 struct in6_addr target; /* ND target address. */
@@ -127,6 +121,31 @@ struct sw_flow_key {
127 }; 121 };
128} __aligned(BITS_PER_LONG/8); /* Ensure that we can do comparisons as longs. */ 122} __aligned(BITS_PER_LONG/8); /* Ensure that we can do comparisons as longs. */
129 123
124struct sw_flow_key_range {
125 size_t start;
126 size_t end;
127};
128
129struct sw_flow_mask {
130 int ref_count;
131 struct rcu_head rcu;
132 struct list_head list;
133 struct sw_flow_key_range range;
134 struct sw_flow_key key;
135};
136
137struct sw_flow_match {
138 struct sw_flow_key *key;
139 struct sw_flow_key_range range;
140 struct sw_flow_mask *mask;
141};
142
143struct sw_flow_actions {
144 struct rcu_head rcu;
145 u32 actions_len;
146 struct nlattr actions[];
147};
148
130struct sw_flow { 149struct sw_flow {
131 struct rcu_head rcu; 150 struct rcu_head rcu;
132 struct hlist_node hash_node[2]; 151 struct hlist_node hash_node[2];
@@ -141,23 +160,9 @@ struct sw_flow {
141 unsigned long used; /* Last used time (in jiffies). */ 160 unsigned long used; /* Last used time (in jiffies). */
142 u64 packet_count; /* Number of packets matched. */ 161 u64 packet_count; /* Number of packets matched. */
143 u64 byte_count; /* Number of bytes matched. */ 162 u64 byte_count; /* Number of bytes matched. */
144 u8 tcp_flags; /* Union of seen TCP flags. */ 163 __be16 tcp_flags; /* Union of seen TCP flags. */
145};
146
147struct sw_flow_key_range {
148 size_t start;
149 size_t end;
150}; 164};
151 165
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 { 166struct arp_eth_header {
162 __be16 ar_hrd; /* format of hardware address */ 167 __be16 ar_hrd; /* format of hardware address */
163 __be16 ar_pro; /* format of protocol address */ 168 __be16 ar_pro; /* format of protocol address */
@@ -172,88 +177,9 @@ struct arp_eth_header {
172 unsigned char ar_tip[4]; /* target IP address */ 177 unsigned char ar_tip[4]; /* target IP address */
173} __packed; 178} __packed;
174 179
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 *); 180void ovs_flow_used(struct sw_flow *, struct sk_buff *);
187u64 ovs_flow_used_time(unsigned long flow_jiffies); 181u64 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);
195 182
196#define MAX_ACTIONS_BUFSIZE (32 * 1024) 183int 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};
250 184
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 */ 185#endif /* flow.h */
diff --git a/net/openvswitch/flow_netlink.c b/net/openvswitch/flow_netlink.c
new file mode 100644
index 000000000000..2bc1bc1aca3b
--- /dev/null
+++ b/net/openvswitch/flow_netlink.c
@@ -0,0 +1,1630 @@
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_TCP_FLAGS)
118 | (1 << OVS_KEY_ATTR_UDP)
119 | (1 << OVS_KEY_ATTR_SCTP)
120 | (1 << OVS_KEY_ATTR_ICMP)
121 | (1 << OVS_KEY_ATTR_ICMPV6)
122 | (1 << OVS_KEY_ATTR_ARP)
123 | (1 << OVS_KEY_ATTR_ND));
124
125 /* Always allowed mask fields. */
126 mask_allowed |= ((1 << OVS_KEY_ATTR_TUNNEL)
127 | (1 << OVS_KEY_ATTR_IN_PORT)
128 | (1 << OVS_KEY_ATTR_ETHERTYPE));
129
130 /* Check key attributes. */
131 if (match->key->eth.type == htons(ETH_P_ARP)
132 || match->key->eth.type == htons(ETH_P_RARP)) {
133 key_expected |= 1 << OVS_KEY_ATTR_ARP;
134 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
135 mask_allowed |= 1 << OVS_KEY_ATTR_ARP;
136 }
137
138 if (match->key->eth.type == htons(ETH_P_IP)) {
139 key_expected |= 1 << OVS_KEY_ATTR_IPV4;
140 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
141 mask_allowed |= 1 << OVS_KEY_ATTR_IPV4;
142
143 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
144 if (match->key->ip.proto == IPPROTO_UDP) {
145 key_expected |= 1 << OVS_KEY_ATTR_UDP;
146 if (match->mask && (match->mask->key.ip.proto == 0xff))
147 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
148 }
149
150 if (match->key->ip.proto == IPPROTO_SCTP) {
151 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
152 if (match->mask && (match->mask->key.ip.proto == 0xff))
153 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
154 }
155
156 if (match->key->ip.proto == IPPROTO_TCP) {
157 key_expected |= 1 << OVS_KEY_ATTR_TCP;
158 key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
159 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
160 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
161 mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
162 }
163 }
164
165 if (match->key->ip.proto == IPPROTO_ICMP) {
166 key_expected |= 1 << OVS_KEY_ATTR_ICMP;
167 if (match->mask && (match->mask->key.ip.proto == 0xff))
168 mask_allowed |= 1 << OVS_KEY_ATTR_ICMP;
169 }
170 }
171 }
172
173 if (match->key->eth.type == htons(ETH_P_IPV6)) {
174 key_expected |= 1 << OVS_KEY_ATTR_IPV6;
175 if (match->mask && (match->mask->key.eth.type == htons(0xffff)))
176 mask_allowed |= 1 << OVS_KEY_ATTR_IPV6;
177
178 if (match->key->ip.frag != OVS_FRAG_TYPE_LATER) {
179 if (match->key->ip.proto == IPPROTO_UDP) {
180 key_expected |= 1 << OVS_KEY_ATTR_UDP;
181 if (match->mask && (match->mask->key.ip.proto == 0xff))
182 mask_allowed |= 1 << OVS_KEY_ATTR_UDP;
183 }
184
185 if (match->key->ip.proto == IPPROTO_SCTP) {
186 key_expected |= 1 << OVS_KEY_ATTR_SCTP;
187 if (match->mask && (match->mask->key.ip.proto == 0xff))
188 mask_allowed |= 1 << OVS_KEY_ATTR_SCTP;
189 }
190
191 if (match->key->ip.proto == IPPROTO_TCP) {
192 key_expected |= 1 << OVS_KEY_ATTR_TCP;
193 key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
194 if (match->mask && (match->mask->key.ip.proto == 0xff)) {
195 mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
196 mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
197 }
198 }
199
200 if (match->key->ip.proto == IPPROTO_ICMPV6) {
201 key_expected |= 1 << OVS_KEY_ATTR_ICMPV6;
202 if (match->mask && (match->mask->key.ip.proto == 0xff))
203 mask_allowed |= 1 << OVS_KEY_ATTR_ICMPV6;
204
205 if (match->key->ipv6.tp.src ==
206 htons(NDISC_NEIGHBOUR_SOLICITATION) ||
207 match->key->ipv6.tp.src == htons(NDISC_NEIGHBOUR_ADVERTISEMENT)) {
208 key_expected |= 1 << OVS_KEY_ATTR_ND;
209 if (match->mask && (match->mask->key.ipv6.tp.src == htons(0xffff)))
210 mask_allowed |= 1 << OVS_KEY_ATTR_ND;
211 }
212 }
213 }
214 }
215
216 if ((key_attrs & key_expected) != key_expected) {
217 /* Key attributes check failed. */
218 OVS_NLERR("Missing expected key attributes (key_attrs=%llx, expected=%llx).\n",
219 key_attrs, key_expected);
220 return false;
221 }
222
223 if ((mask_attrs & mask_allowed) != mask_attrs) {
224 /* Mask attributes check failed. */
225 OVS_NLERR("Contain more than allowed mask fields (mask_attrs=%llx, mask_allowed=%llx).\n",
226 mask_attrs, mask_allowed);
227 return false;
228 }
229
230 return true;
231}
232
233/* The size of the argument for each %OVS_KEY_ATTR_* Netlink attribute. */
234static const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
235 [OVS_KEY_ATTR_ENCAP] = -1,
236 [OVS_KEY_ATTR_PRIORITY] = sizeof(u32),
237 [OVS_KEY_ATTR_IN_PORT] = sizeof(u32),
238 [OVS_KEY_ATTR_SKB_MARK] = sizeof(u32),
239 [OVS_KEY_ATTR_ETHERNET] = sizeof(struct ovs_key_ethernet),
240 [OVS_KEY_ATTR_VLAN] = sizeof(__be16),
241 [OVS_KEY_ATTR_ETHERTYPE] = sizeof(__be16),
242 [OVS_KEY_ATTR_IPV4] = sizeof(struct ovs_key_ipv4),
243 [OVS_KEY_ATTR_IPV6] = sizeof(struct ovs_key_ipv6),
244 [OVS_KEY_ATTR_TCP] = sizeof(struct ovs_key_tcp),
245 [OVS_KEY_ATTR_TCP_FLAGS] = sizeof(__be16),
246 [OVS_KEY_ATTR_UDP] = sizeof(struct ovs_key_udp),
247 [OVS_KEY_ATTR_SCTP] = sizeof(struct ovs_key_sctp),
248 [OVS_KEY_ATTR_ICMP] = sizeof(struct ovs_key_icmp),
249 [OVS_KEY_ATTR_ICMPV6] = sizeof(struct ovs_key_icmpv6),
250 [OVS_KEY_ATTR_ARP] = sizeof(struct ovs_key_arp),
251 [OVS_KEY_ATTR_ND] = sizeof(struct ovs_key_nd),
252 [OVS_KEY_ATTR_TUNNEL] = -1,
253};
254
255static bool is_all_zero(const u8 *fp, size_t size)
256{
257 int i;
258
259 if (!fp)
260 return false;
261
262 for (i = 0; i < size; i++)
263 if (fp[i])
264 return false;
265
266 return true;
267}
268
269static int __parse_flow_nlattrs(const struct nlattr *attr,
270 const struct nlattr *a[],
271 u64 *attrsp, bool nz)
272{
273 const struct nlattr *nla;
274 u64 attrs;
275 int rem;
276
277 attrs = *attrsp;
278 nla_for_each_nested(nla, attr, rem) {
279 u16 type = nla_type(nla);
280 int expected_len;
281
282 if (type > OVS_KEY_ATTR_MAX) {
283 OVS_NLERR("Unknown key attribute (type=%d, max=%d).\n",
284 type, OVS_KEY_ATTR_MAX);
285 return -EINVAL;
286 }
287
288 if (attrs & (1 << type)) {
289 OVS_NLERR("Duplicate key attribute (type %d).\n", type);
290 return -EINVAL;
291 }
292
293 expected_len = ovs_key_lens[type];
294 if (nla_len(nla) != expected_len && expected_len != -1) {
295 OVS_NLERR("Key attribute has unexpected length (type=%d"
296 ", length=%d, expected=%d).\n", type,
297 nla_len(nla), expected_len);
298 return -EINVAL;
299 }
300
301 if (!nz || !is_all_zero(nla_data(nla), expected_len)) {
302 attrs |= 1 << type;
303 a[type] = nla;
304 }
305 }
306 if (rem) {
307 OVS_NLERR("Message has %d unknown bytes.\n", rem);
308 return -EINVAL;
309 }
310
311 *attrsp = attrs;
312 return 0;
313}
314
315static int parse_flow_mask_nlattrs(const struct nlattr *attr,
316 const struct nlattr *a[], u64 *attrsp)
317{
318 return __parse_flow_nlattrs(attr, a, attrsp, true);
319}
320
321static int parse_flow_nlattrs(const struct nlattr *attr,
322 const struct nlattr *a[], u64 *attrsp)
323{
324 return __parse_flow_nlattrs(attr, a, attrsp, false);
325}
326
327static int ipv4_tun_from_nlattr(const struct nlattr *attr,
328 struct sw_flow_match *match, bool is_mask)
329{
330 struct nlattr *a;
331 int rem;
332 bool ttl = false;
333 __be16 tun_flags = 0;
334
335 nla_for_each_nested(a, attr, rem) {
336 int type = nla_type(a);
337 static const u32 ovs_tunnel_key_lens[OVS_TUNNEL_KEY_ATTR_MAX + 1] = {
338 [OVS_TUNNEL_KEY_ATTR_ID] = sizeof(u64),
339 [OVS_TUNNEL_KEY_ATTR_IPV4_SRC] = sizeof(u32),
340 [OVS_TUNNEL_KEY_ATTR_IPV4_DST] = sizeof(u32),
341 [OVS_TUNNEL_KEY_ATTR_TOS] = 1,
342 [OVS_TUNNEL_KEY_ATTR_TTL] = 1,
343 [OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT] = 0,
344 [OVS_TUNNEL_KEY_ATTR_CSUM] = 0,
345 };
346
347 if (type > OVS_TUNNEL_KEY_ATTR_MAX) {
348 OVS_NLERR("Unknown IPv4 tunnel attribute (type=%d, max=%d).\n",
349 type, OVS_TUNNEL_KEY_ATTR_MAX);
350 return -EINVAL;
351 }
352
353 if (ovs_tunnel_key_lens[type] != nla_len(a)) {
354 OVS_NLERR("IPv4 tunnel attribute type has unexpected "
355 " length (type=%d, length=%d, expected=%d).\n",
356 type, nla_len(a), ovs_tunnel_key_lens[type]);
357 return -EINVAL;
358 }
359
360 switch (type) {
361 case OVS_TUNNEL_KEY_ATTR_ID:
362 SW_FLOW_KEY_PUT(match, tun_key.tun_id,
363 nla_get_be64(a), is_mask);
364 tun_flags |= TUNNEL_KEY;
365 break;
366 case OVS_TUNNEL_KEY_ATTR_IPV4_SRC:
367 SW_FLOW_KEY_PUT(match, tun_key.ipv4_src,
368 nla_get_be32(a), is_mask);
369 break;
370 case OVS_TUNNEL_KEY_ATTR_IPV4_DST:
371 SW_FLOW_KEY_PUT(match, tun_key.ipv4_dst,
372 nla_get_be32(a), is_mask);
373 break;
374 case OVS_TUNNEL_KEY_ATTR_TOS:
375 SW_FLOW_KEY_PUT(match, tun_key.ipv4_tos,
376 nla_get_u8(a), is_mask);
377 break;
378 case OVS_TUNNEL_KEY_ATTR_TTL:
379 SW_FLOW_KEY_PUT(match, tun_key.ipv4_ttl,
380 nla_get_u8(a), is_mask);
381 ttl = true;
382 break;
383 case OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT:
384 tun_flags |= TUNNEL_DONT_FRAGMENT;
385 break;
386 case OVS_TUNNEL_KEY_ATTR_CSUM:
387 tun_flags |= TUNNEL_CSUM;
388 break;
389 default:
390 return -EINVAL;
391 }
392 }
393
394 SW_FLOW_KEY_PUT(match, tun_key.tun_flags, tun_flags, is_mask);
395
396 if (rem > 0) {
397 OVS_NLERR("IPv4 tunnel attribute has %d unknown bytes.\n", rem);
398 return -EINVAL;
399 }
400
401 if (!is_mask) {
402 if (!match->key->tun_key.ipv4_dst) {
403 OVS_NLERR("IPv4 tunnel destination address is zero.\n");
404 return -EINVAL;
405 }
406
407 if (!ttl) {
408 OVS_NLERR("IPv4 tunnel TTL not specified.\n");
409 return -EINVAL;
410 }
411 }
412
413 return 0;
414}
415
416static int ipv4_tun_to_nlattr(struct sk_buff *skb,
417 const struct ovs_key_ipv4_tunnel *tun_key,
418 const struct ovs_key_ipv4_tunnel *output)
419{
420 struct nlattr *nla;
421
422 nla = nla_nest_start(skb, OVS_KEY_ATTR_TUNNEL);
423 if (!nla)
424 return -EMSGSIZE;
425
426 if (output->tun_flags & TUNNEL_KEY &&
427 nla_put_be64(skb, OVS_TUNNEL_KEY_ATTR_ID, output->tun_id))
428 return -EMSGSIZE;
429 if (output->ipv4_src &&
430 nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_SRC, output->ipv4_src))
431 return -EMSGSIZE;
432 if (output->ipv4_dst &&
433 nla_put_be32(skb, OVS_TUNNEL_KEY_ATTR_IPV4_DST, output->ipv4_dst))
434 return -EMSGSIZE;
435 if (output->ipv4_tos &&
436 nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TOS, output->ipv4_tos))
437 return -EMSGSIZE;
438 if (nla_put_u8(skb, OVS_TUNNEL_KEY_ATTR_TTL, output->ipv4_ttl))
439 return -EMSGSIZE;
440 if ((output->tun_flags & TUNNEL_DONT_FRAGMENT) &&
441 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_DONT_FRAGMENT))
442 return -EMSGSIZE;
443 if ((output->tun_flags & TUNNEL_CSUM) &&
444 nla_put_flag(skb, OVS_TUNNEL_KEY_ATTR_CSUM))
445 return -EMSGSIZE;
446
447 nla_nest_end(skb, nla);
448 return 0;
449}
450
451
452static int metadata_from_nlattrs(struct sw_flow_match *match, u64 *attrs,
453 const struct nlattr **a, bool is_mask)
454{
455 if (*attrs & (1 << OVS_KEY_ATTR_PRIORITY)) {
456 SW_FLOW_KEY_PUT(match, phy.priority,
457 nla_get_u32(a[OVS_KEY_ATTR_PRIORITY]), is_mask);
458 *attrs &= ~(1 << OVS_KEY_ATTR_PRIORITY);
459 }
460
461 if (*attrs & (1 << OVS_KEY_ATTR_IN_PORT)) {
462 u32 in_port = nla_get_u32(a[OVS_KEY_ATTR_IN_PORT]);
463
464 if (is_mask)
465 in_port = 0xffffffff; /* Always exact match in_port. */
466 else if (in_port >= DP_MAX_PORTS)
467 return -EINVAL;
468
469 SW_FLOW_KEY_PUT(match, phy.in_port, in_port, is_mask);
470 *attrs &= ~(1 << OVS_KEY_ATTR_IN_PORT);
471 } else if (!is_mask) {
472 SW_FLOW_KEY_PUT(match, phy.in_port, DP_MAX_PORTS, is_mask);
473 }
474
475 if (*attrs & (1 << OVS_KEY_ATTR_SKB_MARK)) {
476 uint32_t mark = nla_get_u32(a[OVS_KEY_ATTR_SKB_MARK]);
477
478 SW_FLOW_KEY_PUT(match, phy.skb_mark, mark, is_mask);
479 *attrs &= ~(1 << OVS_KEY_ATTR_SKB_MARK);
480 }
481 if (*attrs & (1 << OVS_KEY_ATTR_TUNNEL)) {
482 if (ipv4_tun_from_nlattr(a[OVS_KEY_ATTR_TUNNEL], match,
483 is_mask))
484 return -EINVAL;
485 *attrs &= ~(1 << OVS_KEY_ATTR_TUNNEL);
486 }
487 return 0;
488}
489
490static int ovs_key_from_nlattrs(struct sw_flow_match *match, u64 attrs,
491 const struct nlattr **a, bool is_mask)
492{
493 int err;
494 u64 orig_attrs = attrs;
495
496 err = metadata_from_nlattrs(match, &attrs, a, is_mask);
497 if (err)
498 return err;
499
500 if (attrs & (1 << OVS_KEY_ATTR_ETHERNET)) {
501 const struct ovs_key_ethernet *eth_key;
502
503 eth_key = nla_data(a[OVS_KEY_ATTR_ETHERNET]);
504 SW_FLOW_KEY_MEMCPY(match, eth.src,
505 eth_key->eth_src, ETH_ALEN, is_mask);
506 SW_FLOW_KEY_MEMCPY(match, eth.dst,
507 eth_key->eth_dst, ETH_ALEN, is_mask);
508 attrs &= ~(1 << OVS_KEY_ATTR_ETHERNET);
509 }
510
511 if (attrs & (1 << OVS_KEY_ATTR_VLAN)) {
512 __be16 tci;
513
514 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
515 if (!(tci & htons(VLAN_TAG_PRESENT))) {
516 if (is_mask)
517 OVS_NLERR("VLAN TCI mask does not have exact match for VLAN_TAG_PRESENT bit.\n");
518 else
519 OVS_NLERR("VLAN TCI does not have VLAN_TAG_PRESENT bit set.\n");
520
521 return -EINVAL;
522 }
523
524 SW_FLOW_KEY_PUT(match, eth.tci, tci, is_mask);
525 attrs &= ~(1 << OVS_KEY_ATTR_VLAN);
526 } else if (!is_mask)
527 SW_FLOW_KEY_PUT(match, eth.tci, htons(0xffff), true);
528
529 if (attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) {
530 __be16 eth_type;
531
532 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
533 if (is_mask) {
534 /* Always exact match EtherType. */
535 eth_type = htons(0xffff);
536 } else if (ntohs(eth_type) < ETH_P_802_3_MIN) {
537 OVS_NLERR("EtherType is less than minimum (type=%x, min=%x).\n",
538 ntohs(eth_type), ETH_P_802_3_MIN);
539 return -EINVAL;
540 }
541
542 SW_FLOW_KEY_PUT(match, eth.type, eth_type, is_mask);
543 attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
544 } else if (!is_mask) {
545 SW_FLOW_KEY_PUT(match, eth.type, htons(ETH_P_802_2), is_mask);
546 }
547
548 if (attrs & (1 << OVS_KEY_ATTR_IPV4)) {
549 const struct ovs_key_ipv4 *ipv4_key;
550
551 ipv4_key = nla_data(a[OVS_KEY_ATTR_IPV4]);
552 if (!is_mask && ipv4_key->ipv4_frag > OVS_FRAG_TYPE_MAX) {
553 OVS_NLERR("Unknown IPv4 fragment type (value=%d, max=%d).\n",
554 ipv4_key->ipv4_frag, OVS_FRAG_TYPE_MAX);
555 return -EINVAL;
556 }
557 SW_FLOW_KEY_PUT(match, ip.proto,
558 ipv4_key->ipv4_proto, is_mask);
559 SW_FLOW_KEY_PUT(match, ip.tos,
560 ipv4_key->ipv4_tos, is_mask);
561 SW_FLOW_KEY_PUT(match, ip.ttl,
562 ipv4_key->ipv4_ttl, is_mask);
563 SW_FLOW_KEY_PUT(match, ip.frag,
564 ipv4_key->ipv4_frag, is_mask);
565 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
566 ipv4_key->ipv4_src, is_mask);
567 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
568 ipv4_key->ipv4_dst, is_mask);
569 attrs &= ~(1 << OVS_KEY_ATTR_IPV4);
570 }
571
572 if (attrs & (1 << OVS_KEY_ATTR_IPV6)) {
573 const struct ovs_key_ipv6 *ipv6_key;
574
575 ipv6_key = nla_data(a[OVS_KEY_ATTR_IPV6]);
576 if (!is_mask && ipv6_key->ipv6_frag > OVS_FRAG_TYPE_MAX) {
577 OVS_NLERR("Unknown IPv6 fragment type (value=%d, max=%d).\n",
578 ipv6_key->ipv6_frag, OVS_FRAG_TYPE_MAX);
579 return -EINVAL;
580 }
581 SW_FLOW_KEY_PUT(match, ipv6.label,
582 ipv6_key->ipv6_label, is_mask);
583 SW_FLOW_KEY_PUT(match, ip.proto,
584 ipv6_key->ipv6_proto, is_mask);
585 SW_FLOW_KEY_PUT(match, ip.tos,
586 ipv6_key->ipv6_tclass, is_mask);
587 SW_FLOW_KEY_PUT(match, ip.ttl,
588 ipv6_key->ipv6_hlimit, is_mask);
589 SW_FLOW_KEY_PUT(match, ip.frag,
590 ipv6_key->ipv6_frag, is_mask);
591 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.src,
592 ipv6_key->ipv6_src,
593 sizeof(match->key->ipv6.addr.src),
594 is_mask);
595 SW_FLOW_KEY_MEMCPY(match, ipv6.addr.dst,
596 ipv6_key->ipv6_dst,
597 sizeof(match->key->ipv6.addr.dst),
598 is_mask);
599
600 attrs &= ~(1 << OVS_KEY_ATTR_IPV6);
601 }
602
603 if (attrs & (1 << OVS_KEY_ATTR_ARP)) {
604 const struct ovs_key_arp *arp_key;
605
606 arp_key = nla_data(a[OVS_KEY_ATTR_ARP]);
607 if (!is_mask && (arp_key->arp_op & htons(0xff00))) {
608 OVS_NLERR("Unknown ARP opcode (opcode=%d).\n",
609 arp_key->arp_op);
610 return -EINVAL;
611 }
612
613 SW_FLOW_KEY_PUT(match, ipv4.addr.src,
614 arp_key->arp_sip, is_mask);
615 SW_FLOW_KEY_PUT(match, ipv4.addr.dst,
616 arp_key->arp_tip, is_mask);
617 SW_FLOW_KEY_PUT(match, ip.proto,
618 ntohs(arp_key->arp_op), is_mask);
619 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.sha,
620 arp_key->arp_sha, ETH_ALEN, is_mask);
621 SW_FLOW_KEY_MEMCPY(match, ipv4.arp.tha,
622 arp_key->arp_tha, ETH_ALEN, is_mask);
623
624 attrs &= ~(1 << OVS_KEY_ATTR_ARP);
625 }
626
627 if (attrs & (1 << OVS_KEY_ATTR_TCP)) {
628 const struct ovs_key_tcp *tcp_key;
629
630 tcp_key = nla_data(a[OVS_KEY_ATTR_TCP]);
631 if (orig_attrs & (1 << OVS_KEY_ATTR_IPV4)) {
632 SW_FLOW_KEY_PUT(match, ipv4.tp.src,
633 tcp_key->tcp_src, is_mask);
634 SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
635 tcp_key->tcp_dst, is_mask);
636 } else {
637 SW_FLOW_KEY_PUT(match, ipv6.tp.src,
638 tcp_key->tcp_src, is_mask);
639 SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
640 tcp_key->tcp_dst, is_mask);
641 }
642 attrs &= ~(1 << OVS_KEY_ATTR_TCP);
643 }
644
645 if (attrs & (1 << OVS_KEY_ATTR_TCP_FLAGS)) {
646 if (orig_attrs & (1 << OVS_KEY_ATTR_IPV4)) {
647 SW_FLOW_KEY_PUT(match, ipv4.tp.flags,
648 nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
649 is_mask);
650 } else {
651 SW_FLOW_KEY_PUT(match, ipv6.tp.flags,
652 nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
653 is_mask);
654 }
655 attrs &= ~(1 << OVS_KEY_ATTR_TCP_FLAGS);
656 }
657
658 if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
659 const struct ovs_key_udp *udp_key;
660
661 udp_key = nla_data(a[OVS_KEY_ATTR_UDP]);
662 if (orig_attrs & (1 << OVS_KEY_ATTR_IPV4)) {
663 SW_FLOW_KEY_PUT(match, ipv4.tp.src,
664 udp_key->udp_src, is_mask);
665 SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
666 udp_key->udp_dst, is_mask);
667 } else {
668 SW_FLOW_KEY_PUT(match, ipv6.tp.src,
669 udp_key->udp_src, is_mask);
670 SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
671 udp_key->udp_dst, is_mask);
672 }
673 attrs &= ~(1 << OVS_KEY_ATTR_UDP);
674 }
675
676 if (attrs & (1 << OVS_KEY_ATTR_SCTP)) {
677 const struct ovs_key_sctp *sctp_key;
678
679 sctp_key = nla_data(a[OVS_KEY_ATTR_SCTP]);
680 if (orig_attrs & (1 << OVS_KEY_ATTR_IPV4)) {
681 SW_FLOW_KEY_PUT(match, ipv4.tp.src,
682 sctp_key->sctp_src, is_mask);
683 SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
684 sctp_key->sctp_dst, is_mask);
685 } else {
686 SW_FLOW_KEY_PUT(match, ipv6.tp.src,
687 sctp_key->sctp_src, is_mask);
688 SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
689 sctp_key->sctp_dst, is_mask);
690 }
691 attrs &= ~(1 << OVS_KEY_ATTR_SCTP);
692 }
693
694 if (attrs & (1 << OVS_KEY_ATTR_ICMP)) {
695 const struct ovs_key_icmp *icmp_key;
696
697 icmp_key = nla_data(a[OVS_KEY_ATTR_ICMP]);
698 SW_FLOW_KEY_PUT(match, ipv4.tp.src,
699 htons(icmp_key->icmp_type), is_mask);
700 SW_FLOW_KEY_PUT(match, ipv4.tp.dst,
701 htons(icmp_key->icmp_code), is_mask);
702 attrs &= ~(1 << OVS_KEY_ATTR_ICMP);
703 }
704
705 if (attrs & (1 << OVS_KEY_ATTR_ICMPV6)) {
706 const struct ovs_key_icmpv6 *icmpv6_key;
707
708 icmpv6_key = nla_data(a[OVS_KEY_ATTR_ICMPV6]);
709 SW_FLOW_KEY_PUT(match, ipv6.tp.src,
710 htons(icmpv6_key->icmpv6_type), is_mask);
711 SW_FLOW_KEY_PUT(match, ipv6.tp.dst,
712 htons(icmpv6_key->icmpv6_code), is_mask);
713 attrs &= ~(1 << OVS_KEY_ATTR_ICMPV6);
714 }
715
716 if (attrs & (1 << OVS_KEY_ATTR_ND)) {
717 const struct ovs_key_nd *nd_key;
718
719 nd_key = nla_data(a[OVS_KEY_ATTR_ND]);
720 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.target,
721 nd_key->nd_target,
722 sizeof(match->key->ipv6.nd.target),
723 is_mask);
724 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.sll,
725 nd_key->nd_sll, ETH_ALEN, is_mask);
726 SW_FLOW_KEY_MEMCPY(match, ipv6.nd.tll,
727 nd_key->nd_tll, ETH_ALEN, is_mask);
728 attrs &= ~(1 << OVS_KEY_ATTR_ND);
729 }
730
731 if (attrs != 0)
732 return -EINVAL;
733
734 return 0;
735}
736
737static void sw_flow_mask_set(struct sw_flow_mask *mask,
738 struct sw_flow_key_range *range, u8 val)
739{
740 u8 *m = (u8 *)&mask->key + range->start;
741
742 mask->range = *range;
743 memset(m, val, range_n_bytes(range));
744}
745
746/**
747 * ovs_nla_get_match - parses Netlink attributes into a flow key and
748 * mask. In case the 'mask' is NULL, the flow is treated as exact match
749 * flow. Otherwise, it is treated as a wildcarded flow, except the mask
750 * does not include any don't care bit.
751 * @match: receives the extracted flow match information.
752 * @key: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
753 * sequence. The fields should of the packet that triggered the creation
754 * of this flow.
755 * @mask: Optional. Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink
756 * attribute specifies the mask field of the wildcarded flow.
757 */
758int ovs_nla_get_match(struct sw_flow_match *match,
759 const struct nlattr *key,
760 const struct nlattr *mask)
761{
762 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
763 const struct nlattr *encap;
764 u64 key_attrs = 0;
765 u64 mask_attrs = 0;
766 bool encap_valid = false;
767 int err;
768
769 err = parse_flow_nlattrs(key, a, &key_attrs);
770 if (err)
771 return err;
772
773 if ((key_attrs & (1 << OVS_KEY_ATTR_ETHERNET)) &&
774 (key_attrs & (1 << OVS_KEY_ATTR_ETHERTYPE)) &&
775 (nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]) == htons(ETH_P_8021Q))) {
776 __be16 tci;
777
778 if (!((key_attrs & (1 << OVS_KEY_ATTR_VLAN)) &&
779 (key_attrs & (1 << OVS_KEY_ATTR_ENCAP)))) {
780 OVS_NLERR("Invalid Vlan frame.\n");
781 return -EINVAL;
782 }
783
784 key_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
785 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
786 encap = a[OVS_KEY_ATTR_ENCAP];
787 key_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
788 encap_valid = true;
789
790 if (tci & htons(VLAN_TAG_PRESENT)) {
791 err = parse_flow_nlattrs(encap, a, &key_attrs);
792 if (err)
793 return err;
794 } else if (!tci) {
795 /* Corner case for truncated 802.1Q header. */
796 if (nla_len(encap)) {
797 OVS_NLERR("Truncated 802.1Q header has non-zero encap attribute.\n");
798 return -EINVAL;
799 }
800 } else {
801 OVS_NLERR("Encap attribute is set for a non-VLAN frame.\n");
802 return -EINVAL;
803 }
804 }
805
806 err = ovs_key_from_nlattrs(match, key_attrs, a, false);
807 if (err)
808 return err;
809
810 if (mask) {
811 err = parse_flow_mask_nlattrs(mask, a, &mask_attrs);
812 if (err)
813 return err;
814
815 if (mask_attrs & 1 << OVS_KEY_ATTR_ENCAP) {
816 __be16 eth_type = 0;
817 __be16 tci = 0;
818
819 if (!encap_valid) {
820 OVS_NLERR("Encap mask attribute is set for non-VLAN frame.\n");
821 return -EINVAL;
822 }
823
824 mask_attrs &= ~(1 << OVS_KEY_ATTR_ENCAP);
825 if (a[OVS_KEY_ATTR_ETHERTYPE])
826 eth_type = nla_get_be16(a[OVS_KEY_ATTR_ETHERTYPE]);
827
828 if (eth_type == htons(0xffff)) {
829 mask_attrs &= ~(1 << OVS_KEY_ATTR_ETHERTYPE);
830 encap = a[OVS_KEY_ATTR_ENCAP];
831 err = parse_flow_mask_nlattrs(encap, a, &mask_attrs);
832 } else {
833 OVS_NLERR("VLAN frames must have an exact match on the TPID (mask=%x).\n",
834 ntohs(eth_type));
835 return -EINVAL;
836 }
837
838 if (a[OVS_KEY_ATTR_VLAN])
839 tci = nla_get_be16(a[OVS_KEY_ATTR_VLAN]);
840
841 if (!(tci & htons(VLAN_TAG_PRESENT))) {
842 OVS_NLERR("VLAN tag present bit must have an exact match (tci_mask=%x).\n", ntohs(tci));
843 return -EINVAL;
844 }
845 }
846
847 err = ovs_key_from_nlattrs(match, mask_attrs, a, true);
848 if (err)
849 return err;
850 } else {
851 /* Populate exact match flow's key mask. */
852 if (match->mask)
853 sw_flow_mask_set(match->mask, &match->range, 0xff);
854 }
855
856 if (!match_validate(match, key_attrs, mask_attrs))
857 return -EINVAL;
858
859 return 0;
860}
861
862/**
863 * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
864 * @flow: Receives extracted in_port, priority, tun_key and skb_mark.
865 * @attr: Netlink attribute holding nested %OVS_KEY_ATTR_* Netlink attribute
866 * sequence.
867 *
868 * This parses a series of Netlink attributes that form a flow key, which must
869 * take the same form accepted by flow_from_nlattrs(), but only enough of it to
870 * get the metadata, that is, the parts of the flow key that cannot be
871 * extracted from the packet itself.
872 */
873
874int ovs_nla_get_flow_metadata(struct sw_flow *flow,
875 const struct nlattr *attr)
876{
877 struct ovs_key_ipv4_tunnel *tun_key = &flow->key.tun_key;
878 const struct nlattr *a[OVS_KEY_ATTR_MAX + 1];
879 u64 attrs = 0;
880 int err;
881 struct sw_flow_match match;
882
883 flow->key.phy.in_port = DP_MAX_PORTS;
884 flow->key.phy.priority = 0;
885 flow->key.phy.skb_mark = 0;
886 memset(tun_key, 0, sizeof(flow->key.tun_key));
887
888 err = parse_flow_nlattrs(attr, a, &attrs);
889 if (err)
890 return -EINVAL;
891
892 memset(&match, 0, sizeof(match));
893 match.key = &flow->key;
894
895 err = metadata_from_nlattrs(&match, &attrs, a, false);
896 if (err)
897 return err;
898
899 return 0;
900}
901
902int ovs_nla_put_flow(const struct sw_flow_key *swkey,
903 const struct sw_flow_key *output, struct sk_buff *skb)
904{
905 struct ovs_key_ethernet *eth_key;
906 struct nlattr *nla, *encap;
907 bool is_mask = (swkey != output);
908
909 if (nla_put_u32(skb, OVS_KEY_ATTR_PRIORITY, output->phy.priority))
910 goto nla_put_failure;
911
912 if ((swkey->tun_key.ipv4_dst || is_mask) &&
913 ipv4_tun_to_nlattr(skb, &swkey->tun_key, &output->tun_key))
914 goto nla_put_failure;
915
916 if (swkey->phy.in_port == DP_MAX_PORTS) {
917 if (is_mask && (output->phy.in_port == 0xffff))
918 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT, 0xffffffff))
919 goto nla_put_failure;
920 } else {
921 u16 upper_u16;
922 upper_u16 = !is_mask ? 0 : 0xffff;
923
924 if (nla_put_u32(skb, OVS_KEY_ATTR_IN_PORT,
925 (upper_u16 << 16) | output->phy.in_port))
926 goto nla_put_failure;
927 }
928
929 if (nla_put_u32(skb, OVS_KEY_ATTR_SKB_MARK, output->phy.skb_mark))
930 goto nla_put_failure;
931
932 nla = nla_reserve(skb, OVS_KEY_ATTR_ETHERNET, sizeof(*eth_key));
933 if (!nla)
934 goto nla_put_failure;
935
936 eth_key = nla_data(nla);
937 memcpy(eth_key->eth_src, output->eth.src, ETH_ALEN);
938 memcpy(eth_key->eth_dst, output->eth.dst, ETH_ALEN);
939
940 if (swkey->eth.tci || swkey->eth.type == htons(ETH_P_8021Q)) {
941 __be16 eth_type;
942 eth_type = !is_mask ? htons(ETH_P_8021Q) : htons(0xffff);
943 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, eth_type) ||
944 nla_put_be16(skb, OVS_KEY_ATTR_VLAN, output->eth.tci))
945 goto nla_put_failure;
946 encap = nla_nest_start(skb, OVS_KEY_ATTR_ENCAP);
947 if (!swkey->eth.tci)
948 goto unencap;
949 } else
950 encap = NULL;
951
952 if (swkey->eth.type == htons(ETH_P_802_2)) {
953 /*
954 * Ethertype 802.2 is represented in the netlink with omitted
955 * OVS_KEY_ATTR_ETHERTYPE in the flow key attribute, and
956 * 0xffff in the mask attribute. Ethertype can also
957 * be wildcarded.
958 */
959 if (is_mask && output->eth.type)
960 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE,
961 output->eth.type))
962 goto nla_put_failure;
963 goto unencap;
964 }
965
966 if (nla_put_be16(skb, OVS_KEY_ATTR_ETHERTYPE, output->eth.type))
967 goto nla_put_failure;
968
969 if (swkey->eth.type == htons(ETH_P_IP)) {
970 struct ovs_key_ipv4 *ipv4_key;
971
972 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV4, sizeof(*ipv4_key));
973 if (!nla)
974 goto nla_put_failure;
975 ipv4_key = nla_data(nla);
976 ipv4_key->ipv4_src = output->ipv4.addr.src;
977 ipv4_key->ipv4_dst = output->ipv4.addr.dst;
978 ipv4_key->ipv4_proto = output->ip.proto;
979 ipv4_key->ipv4_tos = output->ip.tos;
980 ipv4_key->ipv4_ttl = output->ip.ttl;
981 ipv4_key->ipv4_frag = output->ip.frag;
982 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
983 struct ovs_key_ipv6 *ipv6_key;
984
985 nla = nla_reserve(skb, OVS_KEY_ATTR_IPV6, sizeof(*ipv6_key));
986 if (!nla)
987 goto nla_put_failure;
988 ipv6_key = nla_data(nla);
989 memcpy(ipv6_key->ipv6_src, &output->ipv6.addr.src,
990 sizeof(ipv6_key->ipv6_src));
991 memcpy(ipv6_key->ipv6_dst, &output->ipv6.addr.dst,
992 sizeof(ipv6_key->ipv6_dst));
993 ipv6_key->ipv6_label = output->ipv6.label;
994 ipv6_key->ipv6_proto = output->ip.proto;
995 ipv6_key->ipv6_tclass = output->ip.tos;
996 ipv6_key->ipv6_hlimit = output->ip.ttl;
997 ipv6_key->ipv6_frag = output->ip.frag;
998 } else if (swkey->eth.type == htons(ETH_P_ARP) ||
999 swkey->eth.type == htons(ETH_P_RARP)) {
1000 struct ovs_key_arp *arp_key;
1001
1002 nla = nla_reserve(skb, OVS_KEY_ATTR_ARP, sizeof(*arp_key));
1003 if (!nla)
1004 goto nla_put_failure;
1005 arp_key = nla_data(nla);
1006 memset(arp_key, 0, sizeof(struct ovs_key_arp));
1007 arp_key->arp_sip = output->ipv4.addr.src;
1008 arp_key->arp_tip = output->ipv4.addr.dst;
1009 arp_key->arp_op = htons(output->ip.proto);
1010 memcpy(arp_key->arp_sha, output->ipv4.arp.sha, ETH_ALEN);
1011 memcpy(arp_key->arp_tha, output->ipv4.arp.tha, ETH_ALEN);
1012 }
1013
1014 if ((swkey->eth.type == htons(ETH_P_IP) ||
1015 swkey->eth.type == htons(ETH_P_IPV6)) &&
1016 swkey->ip.frag != OVS_FRAG_TYPE_LATER) {
1017
1018 if (swkey->ip.proto == IPPROTO_TCP) {
1019 struct ovs_key_tcp *tcp_key;
1020
1021 nla = nla_reserve(skb, OVS_KEY_ATTR_TCP, sizeof(*tcp_key));
1022 if (!nla)
1023 goto nla_put_failure;
1024 tcp_key = nla_data(nla);
1025 if (swkey->eth.type == htons(ETH_P_IP)) {
1026 tcp_key->tcp_src = output->ipv4.tp.src;
1027 tcp_key->tcp_dst = output->ipv4.tp.dst;
1028 if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
1029 output->ipv4.tp.flags))
1030 goto nla_put_failure;
1031 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1032 tcp_key->tcp_src = output->ipv6.tp.src;
1033 tcp_key->tcp_dst = output->ipv6.tp.dst;
1034 if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
1035 output->ipv6.tp.flags))
1036 goto nla_put_failure;
1037 }
1038 } else if (swkey->ip.proto == IPPROTO_UDP) {
1039 struct ovs_key_udp *udp_key;
1040
1041 nla = nla_reserve(skb, OVS_KEY_ATTR_UDP, sizeof(*udp_key));
1042 if (!nla)
1043 goto nla_put_failure;
1044 udp_key = nla_data(nla);
1045 if (swkey->eth.type == htons(ETH_P_IP)) {
1046 udp_key->udp_src = output->ipv4.tp.src;
1047 udp_key->udp_dst = output->ipv4.tp.dst;
1048 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1049 udp_key->udp_src = output->ipv6.tp.src;
1050 udp_key->udp_dst = output->ipv6.tp.dst;
1051 }
1052 } else if (swkey->ip.proto == IPPROTO_SCTP) {
1053 struct ovs_key_sctp *sctp_key;
1054
1055 nla = nla_reserve(skb, OVS_KEY_ATTR_SCTP, sizeof(*sctp_key));
1056 if (!nla)
1057 goto nla_put_failure;
1058 sctp_key = nla_data(nla);
1059 if (swkey->eth.type == htons(ETH_P_IP)) {
1060 sctp_key->sctp_src = swkey->ipv4.tp.src;
1061 sctp_key->sctp_dst = swkey->ipv4.tp.dst;
1062 } else if (swkey->eth.type == htons(ETH_P_IPV6)) {
1063 sctp_key->sctp_src = swkey->ipv6.tp.src;
1064 sctp_key->sctp_dst = swkey->ipv6.tp.dst;
1065 }
1066 } else if (swkey->eth.type == htons(ETH_P_IP) &&
1067 swkey->ip.proto == IPPROTO_ICMP) {
1068 struct ovs_key_icmp *icmp_key;
1069
1070 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMP, sizeof(*icmp_key));
1071 if (!nla)
1072 goto nla_put_failure;
1073 icmp_key = nla_data(nla);
1074 icmp_key->icmp_type = ntohs(output->ipv4.tp.src);
1075 icmp_key->icmp_code = ntohs(output->ipv4.tp.dst);
1076 } else if (swkey->eth.type == htons(ETH_P_IPV6) &&
1077 swkey->ip.proto == IPPROTO_ICMPV6) {
1078 struct ovs_key_icmpv6 *icmpv6_key;
1079
1080 nla = nla_reserve(skb, OVS_KEY_ATTR_ICMPV6,
1081 sizeof(*icmpv6_key));
1082 if (!nla)
1083 goto nla_put_failure;
1084 icmpv6_key = nla_data(nla);
1085 icmpv6_key->icmpv6_type = ntohs(output->ipv6.tp.src);
1086 icmpv6_key->icmpv6_code = ntohs(output->ipv6.tp.dst);
1087
1088 if (icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_SOLICITATION ||
1089 icmpv6_key->icmpv6_type == NDISC_NEIGHBOUR_ADVERTISEMENT) {
1090 struct ovs_key_nd *nd_key;
1091
1092 nla = nla_reserve(skb, OVS_KEY_ATTR_ND, sizeof(*nd_key));
1093 if (!nla)
1094 goto nla_put_failure;
1095 nd_key = nla_data(nla);
1096 memcpy(nd_key->nd_target, &output->ipv6.nd.target,
1097 sizeof(nd_key->nd_target));
1098 memcpy(nd_key->nd_sll, output->ipv6.nd.sll, ETH_ALEN);
1099 memcpy(nd_key->nd_tll, output->ipv6.nd.tll, ETH_ALEN);
1100 }
1101 }
1102 }
1103
1104unencap:
1105 if (encap)
1106 nla_nest_end(skb, encap);
1107
1108 return 0;
1109
1110nla_put_failure:
1111 return -EMSGSIZE;
1112}
1113
1114#define MAX_ACTIONS_BUFSIZE (32 * 1024)
1115
1116struct sw_flow_actions *ovs_nla_alloc_flow_actions(int size)
1117{
1118 struct sw_flow_actions *sfa;
1119
1120 if (size > MAX_ACTIONS_BUFSIZE)
1121 return ERR_PTR(-EINVAL);
1122
1123 sfa = kmalloc(sizeof(*sfa) + size, GFP_KERNEL);
1124 if (!sfa)
1125 return ERR_PTR(-ENOMEM);
1126
1127 sfa->actions_len = 0;
1128 return sfa;
1129}
1130
1131/* RCU callback used by ovs_nla_free_flow_actions. */
1132static void rcu_free_acts_callback(struct rcu_head *rcu)
1133{
1134 struct sw_flow_actions *sf_acts = container_of(rcu,
1135 struct sw_flow_actions, rcu);
1136 kfree(sf_acts);
1137}
1138
1139/* Schedules 'sf_acts' to be freed after the next RCU grace period.
1140 * The caller must hold rcu_read_lock for this to be sensible. */
1141void ovs_nla_free_flow_actions(struct sw_flow_actions *sf_acts)
1142{
1143 call_rcu(&sf_acts->rcu, rcu_free_acts_callback);
1144}
1145
1146static struct nlattr *reserve_sfa_size(struct sw_flow_actions **sfa,
1147 int attr_len)
1148{
1149
1150 struct sw_flow_actions *acts;
1151 int new_acts_size;
1152 int req_size = NLA_ALIGN(attr_len);
1153 int next_offset = offsetof(struct sw_flow_actions, actions) +
1154 (*sfa)->actions_len;
1155
1156 if (req_size <= (ksize(*sfa) - next_offset))
1157 goto out;
1158
1159 new_acts_size = ksize(*sfa) * 2;
1160
1161 if (new_acts_size > MAX_ACTIONS_BUFSIZE) {
1162 if ((MAX_ACTIONS_BUFSIZE - next_offset) < req_size)
1163 return ERR_PTR(-EMSGSIZE);
1164 new_acts_size = MAX_ACTIONS_BUFSIZE;
1165 }
1166
1167 acts = ovs_nla_alloc_flow_actions(new_acts_size);
1168 if (IS_ERR(acts))
1169 return (void *)acts;
1170
1171 memcpy(acts->actions, (*sfa)->actions, (*sfa)->actions_len);
1172 acts->actions_len = (*sfa)->actions_len;
1173 kfree(*sfa);
1174 *sfa = acts;
1175
1176out:
1177 (*sfa)->actions_len += req_size;
1178 return (struct nlattr *) ((unsigned char *)(*sfa) + next_offset);
1179}
1180
1181static int add_action(struct sw_flow_actions **sfa, int attrtype, void *data, int len)
1182{
1183 struct nlattr *a;
1184
1185 a = reserve_sfa_size(sfa, nla_attr_size(len));
1186 if (IS_ERR(a))
1187 return PTR_ERR(a);
1188
1189 a->nla_type = attrtype;
1190 a->nla_len = nla_attr_size(len);
1191
1192 if (data)
1193 memcpy(nla_data(a), data, len);
1194 memset((unsigned char *) a + a->nla_len, 0, nla_padlen(len));
1195
1196 return 0;
1197}
1198
1199static inline int add_nested_action_start(struct sw_flow_actions **sfa,
1200 int attrtype)
1201{
1202 int used = (*sfa)->actions_len;
1203 int err;
1204
1205 err = add_action(sfa, attrtype, NULL, 0);
1206 if (err)
1207 return err;
1208
1209 return used;
1210}
1211
1212static inline void add_nested_action_end(struct sw_flow_actions *sfa,
1213 int st_offset)
1214{
1215 struct nlattr *a = (struct nlattr *) ((unsigned char *)sfa->actions +
1216 st_offset);
1217
1218 a->nla_len = sfa->actions_len - st_offset;
1219}
1220
1221static int validate_and_copy_sample(const struct nlattr *attr,
1222 const struct sw_flow_key *key, int depth,
1223 struct sw_flow_actions **sfa)
1224{
1225 const struct nlattr *attrs[OVS_SAMPLE_ATTR_MAX + 1];
1226 const struct nlattr *probability, *actions;
1227 const struct nlattr *a;
1228 int rem, start, err, st_acts;
1229
1230 memset(attrs, 0, sizeof(attrs));
1231 nla_for_each_nested(a, attr, rem) {
1232 int type = nla_type(a);
1233 if (!type || type > OVS_SAMPLE_ATTR_MAX || attrs[type])
1234 return -EINVAL;
1235 attrs[type] = a;
1236 }
1237 if (rem)
1238 return -EINVAL;
1239
1240 probability = attrs[OVS_SAMPLE_ATTR_PROBABILITY];
1241 if (!probability || nla_len(probability) != sizeof(u32))
1242 return -EINVAL;
1243
1244 actions = attrs[OVS_SAMPLE_ATTR_ACTIONS];
1245 if (!actions || (nla_len(actions) && nla_len(actions) < NLA_HDRLEN))
1246 return -EINVAL;
1247
1248 /* validation done, copy sample action. */
1249 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SAMPLE);
1250 if (start < 0)
1251 return start;
1252 err = add_action(sfa, OVS_SAMPLE_ATTR_PROBABILITY,
1253 nla_data(probability), sizeof(u32));
1254 if (err)
1255 return err;
1256 st_acts = add_nested_action_start(sfa, OVS_SAMPLE_ATTR_ACTIONS);
1257 if (st_acts < 0)
1258 return st_acts;
1259
1260 err = ovs_nla_copy_actions(actions, key, depth + 1, sfa);
1261 if (err)
1262 return err;
1263
1264 add_nested_action_end(*sfa, st_acts);
1265 add_nested_action_end(*sfa, start);
1266
1267 return 0;
1268}
1269
1270static int validate_tp_port(const struct sw_flow_key *flow_key)
1271{
1272 if (flow_key->eth.type == htons(ETH_P_IP)) {
1273 if (flow_key->ipv4.tp.src || flow_key->ipv4.tp.dst)
1274 return 0;
1275 } else if (flow_key->eth.type == htons(ETH_P_IPV6)) {
1276 if (flow_key->ipv6.tp.src || flow_key->ipv6.tp.dst)
1277 return 0;
1278 }
1279
1280 return -EINVAL;
1281}
1282
1283void ovs_match_init(struct sw_flow_match *match,
1284 struct sw_flow_key *key,
1285 struct sw_flow_mask *mask)
1286{
1287 memset(match, 0, sizeof(*match));
1288 match->key = key;
1289 match->mask = mask;
1290
1291 memset(key, 0, sizeof(*key));
1292
1293 if (mask) {
1294 memset(&mask->key, 0, sizeof(mask->key));
1295 mask->range.start = mask->range.end = 0;
1296 }
1297}
1298
1299static int validate_and_copy_set_tun(const struct nlattr *attr,
1300 struct sw_flow_actions **sfa)
1301{
1302 struct sw_flow_match match;
1303 struct sw_flow_key key;
1304 int err, start;
1305
1306 ovs_match_init(&match, &key, NULL);
1307 err = ipv4_tun_from_nlattr(nla_data(attr), &match, false);
1308 if (err)
1309 return err;
1310
1311 start = add_nested_action_start(sfa, OVS_ACTION_ATTR_SET);
1312 if (start < 0)
1313 return start;
1314
1315 err = add_action(sfa, OVS_KEY_ATTR_IPV4_TUNNEL, &match.key->tun_key,
1316 sizeof(match.key->tun_key));
1317 add_nested_action_end(*sfa, start);
1318
1319 return err;
1320}
1321
1322static int validate_set(const struct nlattr *a,
1323 const struct sw_flow_key *flow_key,
1324 struct sw_flow_actions **sfa,
1325 bool *set_tun)
1326{
1327 const struct nlattr *ovs_key = nla_data(a);
1328 int key_type = nla_type(ovs_key);
1329
1330 /* There can be only one key in a action */
1331 if (nla_total_size(nla_len(ovs_key)) != nla_len(a))
1332 return -EINVAL;
1333
1334 if (key_type > OVS_KEY_ATTR_MAX ||
1335 (ovs_key_lens[key_type] != nla_len(ovs_key) &&
1336 ovs_key_lens[key_type] != -1))
1337 return -EINVAL;
1338
1339 switch (key_type) {
1340 const struct ovs_key_ipv4 *ipv4_key;
1341 const struct ovs_key_ipv6 *ipv6_key;
1342 int err;
1343
1344 case OVS_KEY_ATTR_PRIORITY:
1345 case OVS_KEY_ATTR_SKB_MARK:
1346 case OVS_KEY_ATTR_ETHERNET:
1347 break;
1348
1349 case OVS_KEY_ATTR_TUNNEL:
1350 *set_tun = true;
1351 err = validate_and_copy_set_tun(a, sfa);
1352 if (err)
1353 return err;
1354 break;
1355
1356 case OVS_KEY_ATTR_IPV4:
1357 if (flow_key->eth.type != htons(ETH_P_IP))
1358 return -EINVAL;
1359
1360 if (!flow_key->ip.proto)
1361 return -EINVAL;
1362
1363 ipv4_key = nla_data(ovs_key);
1364 if (ipv4_key->ipv4_proto != flow_key->ip.proto)
1365 return -EINVAL;
1366
1367 if (ipv4_key->ipv4_frag != flow_key->ip.frag)
1368 return -EINVAL;
1369
1370 break;
1371
1372 case OVS_KEY_ATTR_IPV6:
1373 if (flow_key->eth.type != htons(ETH_P_IPV6))
1374 return -EINVAL;
1375
1376 if (!flow_key->ip.proto)
1377 return -EINVAL;
1378
1379 ipv6_key = nla_data(ovs_key);
1380 if (ipv6_key->ipv6_proto != flow_key->ip.proto)
1381 return -EINVAL;
1382
1383 if (ipv6_key->ipv6_frag != flow_key->ip.frag)
1384 return -EINVAL;
1385
1386 if (ntohl(ipv6_key->ipv6_label) & 0xFFF00000)
1387 return -EINVAL;
1388
1389 break;
1390
1391 case OVS_KEY_ATTR_TCP:
1392 if (flow_key->ip.proto != IPPROTO_TCP)
1393 return -EINVAL;
1394
1395 return validate_tp_port(flow_key);
1396
1397 case OVS_KEY_ATTR_UDP:
1398 if (flow_key->ip.proto != IPPROTO_UDP)
1399 return -EINVAL;
1400
1401 return validate_tp_port(flow_key);
1402
1403 case OVS_KEY_ATTR_SCTP:
1404 if (flow_key->ip.proto != IPPROTO_SCTP)
1405 return -EINVAL;
1406
1407 return validate_tp_port(flow_key);
1408
1409 default:
1410 return -EINVAL;
1411 }
1412
1413 return 0;
1414}
1415
1416static int validate_userspace(const struct nlattr *attr)
1417{
1418 static const struct nla_policy userspace_policy[OVS_USERSPACE_ATTR_MAX + 1] = {
1419 [OVS_USERSPACE_ATTR_PID] = {.type = NLA_U32 },
1420 [OVS_USERSPACE_ATTR_USERDATA] = {.type = NLA_UNSPEC },
1421 };
1422 struct nlattr *a[OVS_USERSPACE_ATTR_MAX + 1];
1423 int error;
1424
1425 error = nla_parse_nested(a, OVS_USERSPACE_ATTR_MAX,
1426 attr, userspace_policy);
1427 if (error)
1428 return error;
1429
1430 if (!a[OVS_USERSPACE_ATTR_PID] ||
1431 !nla_get_u32(a[OVS_USERSPACE_ATTR_PID]))
1432 return -EINVAL;
1433
1434 return 0;
1435}
1436
1437static int copy_action(const struct nlattr *from,
1438 struct sw_flow_actions **sfa)
1439{
1440 int totlen = NLA_ALIGN(from->nla_len);
1441 struct nlattr *to;
1442
1443 to = reserve_sfa_size(sfa, from->nla_len);
1444 if (IS_ERR(to))
1445 return PTR_ERR(to);
1446
1447 memcpy(to, from, totlen);
1448 return 0;
1449}
1450
1451int ovs_nla_copy_actions(const struct nlattr *attr,
1452 const struct sw_flow_key *key,
1453 int depth,
1454 struct sw_flow_actions **sfa)
1455{
1456 const struct nlattr *a;
1457 int rem, err;
1458
1459 if (depth >= SAMPLE_ACTION_DEPTH)
1460 return -EOVERFLOW;
1461
1462 nla_for_each_nested(a, attr, rem) {
1463 /* Expected argument lengths, (u32)-1 for variable length. */
1464 static const u32 action_lens[OVS_ACTION_ATTR_MAX + 1] = {
1465 [OVS_ACTION_ATTR_OUTPUT] = sizeof(u32),
1466 [OVS_ACTION_ATTR_USERSPACE] = (u32)-1,
1467 [OVS_ACTION_ATTR_PUSH_VLAN] = sizeof(struct ovs_action_push_vlan),
1468 [OVS_ACTION_ATTR_POP_VLAN] = 0,
1469 [OVS_ACTION_ATTR_SET] = (u32)-1,
1470 [OVS_ACTION_ATTR_SAMPLE] = (u32)-1
1471 };
1472 const struct ovs_action_push_vlan *vlan;
1473 int type = nla_type(a);
1474 bool skip_copy;
1475
1476 if (type > OVS_ACTION_ATTR_MAX ||
1477 (action_lens[type] != nla_len(a) &&
1478 action_lens[type] != (u32)-1))
1479 return -EINVAL;
1480
1481 skip_copy = false;
1482 switch (type) {
1483 case OVS_ACTION_ATTR_UNSPEC:
1484 return -EINVAL;
1485
1486 case OVS_ACTION_ATTR_USERSPACE:
1487 err = validate_userspace(a);
1488 if (err)
1489 return err;
1490 break;
1491
1492 case OVS_ACTION_ATTR_OUTPUT:
1493 if (nla_get_u32(a) >= DP_MAX_PORTS)
1494 return -EINVAL;
1495 break;
1496
1497
1498 case OVS_ACTION_ATTR_POP_VLAN:
1499 break;
1500
1501 case OVS_ACTION_ATTR_PUSH_VLAN:
1502 vlan = nla_data(a);
1503 if (vlan->vlan_tpid != htons(ETH_P_8021Q))
1504 return -EINVAL;
1505 if (!(vlan->vlan_tci & htons(VLAN_TAG_PRESENT)))
1506 return -EINVAL;
1507 break;
1508
1509 case OVS_ACTION_ATTR_SET:
1510 err = validate_set(a, key, sfa, &skip_copy);
1511 if (err)
1512 return err;
1513 break;
1514
1515 case OVS_ACTION_ATTR_SAMPLE:
1516 err = validate_and_copy_sample(a, key, depth, sfa);
1517 if (err)
1518 return err;
1519 skip_copy = true;
1520 break;
1521
1522 default:
1523 return -EINVAL;
1524 }
1525 if (!skip_copy) {
1526 err = copy_action(a, sfa);
1527 if (err)
1528 return err;
1529 }
1530 }
1531
1532 if (rem > 0)
1533 return -EINVAL;
1534
1535 return 0;
1536}
1537
1538static int sample_action_to_attr(const struct nlattr *attr, struct sk_buff *skb)
1539{
1540 const struct nlattr *a;
1541 struct nlattr *start;
1542 int err = 0, rem;
1543
1544 start = nla_nest_start(skb, OVS_ACTION_ATTR_SAMPLE);
1545 if (!start)
1546 return -EMSGSIZE;
1547
1548 nla_for_each_nested(a, attr, rem) {
1549 int type = nla_type(a);
1550 struct nlattr *st_sample;
1551
1552 switch (type) {
1553 case OVS_SAMPLE_ATTR_PROBABILITY:
1554 if (nla_put(skb, OVS_SAMPLE_ATTR_PROBABILITY,
1555 sizeof(u32), nla_data(a)))
1556 return -EMSGSIZE;
1557 break;
1558 case OVS_SAMPLE_ATTR_ACTIONS:
1559 st_sample = nla_nest_start(skb, OVS_SAMPLE_ATTR_ACTIONS);
1560 if (!st_sample)
1561 return -EMSGSIZE;
1562 err = ovs_nla_put_actions(nla_data(a), nla_len(a), skb);
1563 if (err)
1564 return err;
1565 nla_nest_end(skb, st_sample);
1566 break;
1567 }
1568 }
1569
1570 nla_nest_end(skb, start);
1571 return err;
1572}
1573
1574static int set_action_to_attr(const struct nlattr *a, struct sk_buff *skb)
1575{
1576 const struct nlattr *ovs_key = nla_data(a);
1577 int key_type = nla_type(ovs_key);
1578 struct nlattr *start;
1579 int err;
1580
1581 switch (key_type) {
1582 case OVS_KEY_ATTR_IPV4_TUNNEL:
1583 start = nla_nest_start(skb, OVS_ACTION_ATTR_SET);
1584 if (!start)
1585 return -EMSGSIZE;
1586
1587 err = ipv4_tun_to_nlattr(skb, nla_data(ovs_key),
1588 nla_data(ovs_key));
1589 if (err)
1590 return err;
1591 nla_nest_end(skb, start);
1592 break;
1593 default:
1594 if (nla_put(skb, OVS_ACTION_ATTR_SET, nla_len(a), ovs_key))
1595 return -EMSGSIZE;
1596 break;
1597 }
1598
1599 return 0;
1600}
1601
1602int ovs_nla_put_actions(const struct nlattr *attr, int len, struct sk_buff *skb)
1603{
1604 const struct nlattr *a;
1605 int rem, err;
1606
1607 nla_for_each_attr(a, attr, len, rem) {
1608 int type = nla_type(a);
1609
1610 switch (type) {
1611 case OVS_ACTION_ATTR_SET:
1612 err = set_action_to_attr(a, skb);
1613 if (err)
1614 return err;
1615 break;
1616
1617 case OVS_ACTION_ATTR_SAMPLE:
1618 err = sample_action_to_attr(a, skb);
1619 if (err)
1620 return err;
1621 break;
1622 default:
1623 if (nla_put(skb, type, nla_len(a), nla_data(a)))
1624 return -EMSGSIZE;
1625 break;
1626 }
1627 }
1628
1629 return 0;
1630}
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..e42542706087
--- /dev/null
+++ b/net/openvswitch/flow_table.c
@@ -0,0 +1,592 @@
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 "datapath.h"
48
49#define TBL_MIN_BUCKETS 1024
50#define REHASH_INTERVAL (10 * 60 * HZ)
51
52static struct kmem_cache *flow_cache;
53
54static u16 range_n_bytes(const struct sw_flow_key_range *range)
55{
56 return range->end - range->start;
57}
58
59void ovs_flow_mask_key(struct sw_flow_key *dst, const struct sw_flow_key *src,
60 const struct sw_flow_mask *mask)
61{
62 const long *m = (long *)((u8 *)&mask->key + mask->range.start);
63 const long *s = (long *)((u8 *)src + mask->range.start);
64 long *d = (long *)((u8 *)dst + mask->range.start);
65 int i;
66
67 /* The memory outside of the 'mask->range' are not set since
68 * further operations on 'dst' only uses contents within
69 * 'mask->range'.
70 */
71 for (i = 0; i < range_n_bytes(&mask->range); i += sizeof(long))
72 *d++ = *s++ & *m++;
73}
74
75struct sw_flow *ovs_flow_alloc(void)
76{
77 struct sw_flow *flow;
78
79 flow = kmem_cache_alloc(flow_cache, GFP_KERNEL);
80 if (!flow)
81 return ERR_PTR(-ENOMEM);
82
83 spin_lock_init(&flow->lock);
84 flow->sf_acts = NULL;
85 flow->mask = NULL;
86
87 return flow;
88}
89
90int ovs_flow_tbl_count(struct flow_table *table)
91{
92 return table->count;
93}
94
95static struct flex_array *alloc_buckets(unsigned int n_buckets)
96{
97 struct flex_array *buckets;
98 int i, err;
99
100 buckets = flex_array_alloc(sizeof(struct hlist_head),
101 n_buckets, GFP_KERNEL);
102 if (!buckets)
103 return NULL;
104
105 err = flex_array_prealloc(buckets, 0, n_buckets, GFP_KERNEL);
106 if (err) {
107 flex_array_free(buckets);
108 return NULL;
109 }
110
111 for (i = 0; i < n_buckets; i++)
112 INIT_HLIST_HEAD((struct hlist_head *)
113 flex_array_get(buckets, i));
114
115 return buckets;
116}
117
118static void flow_free(struct sw_flow *flow)
119{
120 kfree((struct sf_flow_acts __force *)flow->sf_acts);
121 kmem_cache_free(flow_cache, flow);
122}
123
124static void rcu_free_flow_callback(struct rcu_head *rcu)
125{
126 struct sw_flow *flow = container_of(rcu, struct sw_flow, rcu);
127
128 flow_free(flow);
129}
130
131static void rcu_free_sw_flow_mask_cb(struct rcu_head *rcu)
132{
133 struct sw_flow_mask *mask = container_of(rcu, struct sw_flow_mask, rcu);
134
135 kfree(mask);
136}
137
138static void flow_mask_del_ref(struct sw_flow_mask *mask, bool deferred)
139{
140 if (!mask)
141 return;
142
143 BUG_ON(!mask->ref_count);
144 mask->ref_count--;
145
146 if (!mask->ref_count) {
147 list_del_rcu(&mask->list);
148 if (deferred)
149 call_rcu(&mask->rcu, rcu_free_sw_flow_mask_cb);
150 else
151 kfree(mask);
152 }
153}
154
155void ovs_flow_free(struct sw_flow *flow, bool deferred)
156{
157 if (!flow)
158 return;
159
160 flow_mask_del_ref(flow->mask, deferred);
161
162 if (deferred)
163 call_rcu(&flow->rcu, rcu_free_flow_callback);
164 else
165 flow_free(flow);
166}
167
168static void free_buckets(struct flex_array *buckets)
169{
170 flex_array_free(buckets);
171}
172
173static void __table_instance_destroy(struct table_instance *ti)
174{
175 int i;
176
177 if (ti->keep_flows)
178 goto skip_flows;
179
180 for (i = 0; i < ti->n_buckets; i++) {
181 struct sw_flow *flow;
182 struct hlist_head *head = flex_array_get(ti->buckets, i);
183 struct hlist_node *n;
184 int ver = ti->node_ver;
185
186 hlist_for_each_entry_safe(flow, n, head, hash_node[ver]) {
187 hlist_del(&flow->hash_node[ver]);
188 ovs_flow_free(flow, false);
189 }
190 }
191
192skip_flows:
193 free_buckets(ti->buckets);
194 kfree(ti);
195}
196
197static struct table_instance *table_instance_alloc(int new_size)
198{
199 struct table_instance *ti = kmalloc(sizeof(*ti), GFP_KERNEL);
200
201 if (!ti)
202 return NULL;
203
204 ti->buckets = alloc_buckets(new_size);
205
206 if (!ti->buckets) {
207 kfree(ti);
208 return NULL;
209 }
210 ti->n_buckets = new_size;
211 ti->node_ver = 0;
212 ti->keep_flows = false;
213 get_random_bytes(&ti->hash_seed, sizeof(u32));
214
215 return ti;
216}
217
218int ovs_flow_tbl_init(struct flow_table *table)
219{
220 struct table_instance *ti;
221
222 ti = table_instance_alloc(TBL_MIN_BUCKETS);
223
224 if (!ti)
225 return -ENOMEM;
226
227 rcu_assign_pointer(table->ti, ti);
228 INIT_LIST_HEAD(&table->mask_list);
229 table->last_rehash = jiffies;
230 table->count = 0;
231 return 0;
232}
233
234static void flow_tbl_destroy_rcu_cb(struct rcu_head *rcu)
235{
236 struct table_instance *ti = container_of(rcu, struct table_instance, rcu);
237
238 __table_instance_destroy(ti);
239}
240
241static void table_instance_destroy(struct table_instance *ti, bool deferred)
242{
243 if (!ti)
244 return;
245
246 if (deferred)
247 call_rcu(&ti->rcu, flow_tbl_destroy_rcu_cb);
248 else
249 __table_instance_destroy(ti);
250}
251
252void ovs_flow_tbl_destroy(struct flow_table *table)
253{
254 struct table_instance *ti = ovsl_dereference(table->ti);
255
256 table_instance_destroy(ti, false);
257}
258
259struct sw_flow *ovs_flow_tbl_dump_next(struct table_instance *ti,
260 u32 *bucket, u32 *last)
261{
262 struct sw_flow *flow;
263 struct hlist_head *head;
264 int ver;
265 int i;
266
267 ver = ti->node_ver;
268 while (*bucket < ti->n_buckets) {
269 i = 0;
270 head = flex_array_get(ti->buckets, *bucket);
271 hlist_for_each_entry_rcu(flow, head, hash_node[ver]) {
272 if (i < *last) {
273 i++;
274 continue;
275 }
276 *last = i + 1;
277 return flow;
278 }
279 (*bucket)++;
280 *last = 0;
281 }
282
283 return NULL;
284}
285
286static struct hlist_head *find_bucket(struct table_instance *ti, u32 hash)
287{
288 hash = jhash_1word(hash, ti->hash_seed);
289 return flex_array_get(ti->buckets,
290 (hash & (ti->n_buckets - 1)));
291}
292
293static void table_instance_insert(struct table_instance *ti, struct sw_flow *flow)
294{
295 struct hlist_head *head;
296
297 head = find_bucket(ti, flow->hash);
298 hlist_add_head_rcu(&flow->hash_node[ti->node_ver], head);
299}
300
301static void flow_table_copy_flows(struct table_instance *old,
302 struct table_instance *new)
303{
304 int old_ver;
305 int i;
306
307 old_ver = old->node_ver;
308 new->node_ver = !old_ver;
309
310 /* Insert in new table. */
311 for (i = 0; i < old->n_buckets; i++) {
312 struct sw_flow *flow;
313 struct hlist_head *head;
314
315 head = flex_array_get(old->buckets, i);
316
317 hlist_for_each_entry(flow, head, hash_node[old_ver])
318 table_instance_insert(new, flow);
319 }
320
321 old->keep_flows = true;
322}
323
324static struct table_instance *table_instance_rehash(struct table_instance *ti,
325 int n_buckets)
326{
327 struct table_instance *new_ti;
328
329 new_ti = table_instance_alloc(n_buckets);
330 if (!new_ti)
331 return NULL;
332
333 flow_table_copy_flows(ti, new_ti);
334
335 return new_ti;
336}
337
338int ovs_flow_tbl_flush(struct flow_table *flow_table)
339{
340 struct table_instance *old_ti;
341 struct table_instance *new_ti;
342
343 old_ti = ovsl_dereference(flow_table->ti);
344 new_ti = table_instance_alloc(TBL_MIN_BUCKETS);
345 if (!new_ti)
346 return -ENOMEM;
347
348 rcu_assign_pointer(flow_table->ti, new_ti);
349 flow_table->last_rehash = jiffies;
350 flow_table->count = 0;
351
352 table_instance_destroy(old_ti, true);
353 return 0;
354}
355
356static u32 flow_hash(const struct sw_flow_key *key, int key_start,
357 int key_end)
358{
359 u32 *hash_key = (u32 *)((u8 *)key + key_start);
360 int hash_u32s = (key_end - key_start) >> 2;
361
362 /* Make sure number of hash bytes are multiple of u32. */
363 BUILD_BUG_ON(sizeof(long) % sizeof(u32));
364
365 return jhash2(hash_key, hash_u32s, 0);
366}
367
368static int flow_key_start(const struct sw_flow_key *key)
369{
370 if (key->tun_key.ipv4_dst)
371 return 0;
372 else
373 return rounddown(offsetof(struct sw_flow_key, phy),
374 sizeof(long));
375}
376
377static bool cmp_key(const struct sw_flow_key *key1,
378 const struct sw_flow_key *key2,
379 int key_start, int key_end)
380{
381 const long *cp1 = (long *)((u8 *)key1 + key_start);
382 const long *cp2 = (long *)((u8 *)key2 + key_start);
383 long diffs = 0;
384 int i;
385
386 for (i = key_start; i < key_end; i += sizeof(long))
387 diffs |= *cp1++ ^ *cp2++;
388
389 return diffs == 0;
390}
391
392static bool flow_cmp_masked_key(const struct sw_flow *flow,
393 const struct sw_flow_key *key,
394 int key_start, int key_end)
395{
396 return cmp_key(&flow->key, key, key_start, key_end);
397}
398
399bool ovs_flow_cmp_unmasked_key(const struct sw_flow *flow,
400 struct sw_flow_match *match)
401{
402 struct sw_flow_key *key = match->key;
403 int key_start = flow_key_start(key);
404 int key_end = match->range.end;
405
406 return cmp_key(&flow->unmasked_key, key, key_start, key_end);
407}
408
409static struct sw_flow *masked_flow_lookup(struct table_instance *ti,
410 const struct sw_flow_key *unmasked,
411 struct sw_flow_mask *mask)
412{
413 struct sw_flow *flow;
414 struct hlist_head *head;
415 int key_start = mask->range.start;
416 int key_end = mask->range.end;
417 u32 hash;
418 struct sw_flow_key masked_key;
419
420 ovs_flow_mask_key(&masked_key, unmasked, mask);
421 hash = flow_hash(&masked_key, key_start, key_end);
422 head = find_bucket(ti, hash);
423 hlist_for_each_entry_rcu(flow, head, hash_node[ti->node_ver]) {
424 if (flow->mask == mask && flow->hash == hash &&
425 flow_cmp_masked_key(flow, &masked_key,
426 key_start, key_end))
427 return flow;
428 }
429 return NULL;
430}
431
432struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *tbl,
433 const struct sw_flow_key *key,
434 u32 *n_mask_hit)
435{
436 struct table_instance *ti = rcu_dereference(tbl->ti);
437 struct sw_flow_mask *mask;
438 struct sw_flow *flow;
439
440 *n_mask_hit = 0;
441 list_for_each_entry_rcu(mask, &tbl->mask_list, list) {
442 (*n_mask_hit)++;
443 flow = masked_flow_lookup(ti, key, mask);
444 if (flow) /* Found */
445 return flow;
446 }
447 return NULL;
448}
449
450int ovs_flow_tbl_num_masks(const struct flow_table *table)
451{
452 struct sw_flow_mask *mask;
453 int num = 0;
454
455 list_for_each_entry(mask, &table->mask_list, list)
456 num++;
457
458 return num;
459}
460
461static struct table_instance *table_instance_expand(struct table_instance *ti)
462{
463 return table_instance_rehash(ti, ti->n_buckets * 2);
464}
465
466void ovs_flow_tbl_remove(struct flow_table *table, struct sw_flow *flow)
467{
468 struct table_instance *ti = ovsl_dereference(table->ti);
469
470 BUG_ON(table->count == 0);
471 hlist_del_rcu(&flow->hash_node[ti->node_ver]);
472 table->count--;
473}
474
475static struct sw_flow_mask *mask_alloc(void)
476{
477 struct sw_flow_mask *mask;
478
479 mask = kmalloc(sizeof(*mask), GFP_KERNEL);
480 if (mask)
481 mask->ref_count = 0;
482
483 return mask;
484}
485
486static void mask_add_ref(struct sw_flow_mask *mask)
487{
488 mask->ref_count++;
489}
490
491static bool mask_equal(const struct sw_flow_mask *a,
492 const struct sw_flow_mask *b)
493{
494 u8 *a_ = (u8 *)&a->key + a->range.start;
495 u8 *b_ = (u8 *)&b->key + b->range.start;
496
497 return (a->range.end == b->range.end)
498 && (a->range.start == b->range.start)
499 && (memcmp(a_, b_, range_n_bytes(&a->range)) == 0);
500}
501
502static struct sw_flow_mask *flow_mask_find(const struct flow_table *tbl,
503 const struct sw_flow_mask *mask)
504{
505 struct list_head *ml;
506
507 list_for_each(ml, &tbl->mask_list) {
508 struct sw_flow_mask *m;
509 m = container_of(ml, struct sw_flow_mask, list);
510 if (mask_equal(mask, m))
511 return m;
512 }
513
514 return NULL;
515}
516
517/**
518 * add a new mask into the mask list.
519 * The caller needs to make sure that 'mask' is not the same
520 * as any masks that are already on the list.
521 */
522static int flow_mask_insert(struct flow_table *tbl, struct sw_flow *flow,
523 struct sw_flow_mask *new)
524{
525 struct sw_flow_mask *mask;
526 mask = flow_mask_find(tbl, new);
527 if (!mask) {
528 /* Allocate a new mask if none exsits. */
529 mask = mask_alloc();
530 if (!mask)
531 return -ENOMEM;
532 mask->key = new->key;
533 mask->range = new->range;
534 list_add_rcu(&mask->list, &tbl->mask_list);
535 }
536
537 mask_add_ref(mask);
538 flow->mask = mask;
539 return 0;
540}
541
542int ovs_flow_tbl_insert(struct flow_table *table, struct sw_flow *flow,
543 struct sw_flow_mask *mask)
544{
545 struct table_instance *new_ti = NULL;
546 struct table_instance *ti;
547 int err;
548
549 err = flow_mask_insert(table, flow, mask);
550 if (err)
551 return err;
552
553 flow->hash = flow_hash(&flow->key, flow->mask->range.start,
554 flow->mask->range.end);
555 ti = ovsl_dereference(table->ti);
556 table_instance_insert(ti, flow);
557 table->count++;
558
559 /* Expand table, if necessary, to make room. */
560 if (table->count > ti->n_buckets)
561 new_ti = table_instance_expand(ti);
562 else if (time_after(jiffies, table->last_rehash + REHASH_INTERVAL))
563 new_ti = table_instance_rehash(ti, ti->n_buckets);
564
565 if (new_ti) {
566 rcu_assign_pointer(table->ti, new_ti);
567 table_instance_destroy(ti, true);
568 table->last_rehash = jiffies;
569 }
570 return 0;
571}
572
573/* Initializes the flow module.
574 * Returns zero if successful or a negative error code. */
575int ovs_flow_init(void)
576{
577 BUILD_BUG_ON(__alignof__(struct sw_flow_key) % __alignof__(long));
578 BUILD_BUG_ON(sizeof(struct sw_flow_key) % sizeof(long));
579
580 flow_cache = kmem_cache_create("sw_flow", sizeof(struct sw_flow), 0,
581 0, NULL);
582 if (flow_cache == NULL)
583 return -ENOMEM;
584
585 return 0;
586}
587
588/* Uninitializes the flow module. */
589void ovs_flow_exit(void)
590{
591 kmem_cache_destroy(flow_cache);
592}
diff --git a/net/openvswitch/flow_table.h b/net/openvswitch/flow_table.h
new file mode 100644
index 000000000000..fbe45d5ad07d
--- /dev/null
+++ b/net/openvswitch/flow_table.h
@@ -0,0 +1,81 @@
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
39struct table_instance {
40 struct flex_array *buckets;
41 unsigned int n_buckets;
42 struct rcu_head rcu;
43 int node_ver;
44 u32 hash_seed;
45 bool keep_flows;
46};
47
48struct flow_table {
49 struct table_instance __rcu *ti;
50 struct list_head mask_list;
51 unsigned long last_rehash;
52 unsigned int count;
53};
54
55int ovs_flow_init(void);
56void ovs_flow_exit(void);
57
58struct sw_flow *ovs_flow_alloc(void);
59void ovs_flow_free(struct sw_flow *, bool deferred);
60
61int ovs_flow_tbl_init(struct flow_table *);
62int ovs_flow_tbl_count(struct flow_table *table);
63void ovs_flow_tbl_destroy(struct flow_table *table);
64int ovs_flow_tbl_flush(struct flow_table *flow_table);
65
66int ovs_flow_tbl_insert(struct flow_table *table, struct sw_flow *flow,
67 struct sw_flow_mask *mask);
68void ovs_flow_tbl_remove(struct flow_table *table, struct sw_flow *flow);
69int ovs_flow_tbl_num_masks(const struct flow_table *table);
70struct sw_flow *ovs_flow_tbl_dump_next(struct table_instance *table,
71 u32 *bucket, u32 *idx);
72struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *,
73 const struct sw_flow_key *,
74 u32 *n_mask_hit);
75
76bool ovs_flow_cmp_unmasked_key(const struct sw_flow *flow,
77 struct sw_flow_match *match);
78
79void ovs_flow_mask_key(struct sw_flow_key *dst, const struct sw_flow_key *src,
80 const struct sw_flow_mask *mask);
81#endif /* flow_table.h */
diff --git a/net/openvswitch/vport-gre.c b/net/openvswitch/vport-gre.c
index c99dea543d64..a3d6951602db 100644
--- a/net/openvswitch/vport-gre.c
+++ b/net/openvswitch/vport-gre.c
@@ -24,8 +24,6 @@
24#include <linux/if_tunnel.h> 24#include <linux/if_tunnel.h>
25#include <linux/if_vlan.h> 25#include <linux/if_vlan.h>
26#include <linux/in.h> 26#include <linux/in.h>
27#include <linux/if_vlan.h>
28#include <linux/in.h>
29#include <linux/in_route.h> 27#include <linux/in_route.h>
30#include <linux/inetdevice.h> 28#include <linux/inetdevice.h>
31#include <linux/jhash.h> 29#include <linux/jhash.h>
diff --git a/net/openvswitch/vport-internal_dev.c b/net/openvswitch/vport-internal_dev.c
index 98d3edbbc235..729c68763fe7 100644
--- a/net/openvswitch/vport-internal_dev.c
+++ b/net/openvswitch/vport-internal_dev.c
@@ -134,7 +134,7 @@ static void do_setup(struct net_device *netdev)
134 netdev->tx_queue_len = 0; 134 netdev->tx_queue_len = 0;
135 135
136 netdev->features = NETIF_F_LLTX | NETIF_F_SG | NETIF_F_FRAGLIST | 136 netdev->features = NETIF_F_LLTX | NETIF_F_SG | NETIF_F_FRAGLIST |
137 NETIF_F_HIGHDMA | NETIF_F_HW_CSUM | NETIF_F_TSO; 137 NETIF_F_HIGHDMA | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE;
138 138
139 netdev->vlan_features = netdev->features; 139 netdev->vlan_features = netdev->features;
140 netdev->features |= NETIF_F_HW_VLAN_CTAG_TX; 140 netdev->features |= NETIF_F_HW_VLAN_CTAG_TX;
diff --git a/net/openvswitch/vport-vxlan.c b/net/openvswitch/vport-vxlan.c
index 56e22b74cf96..e797a50ac2be 100644
--- a/net/openvswitch/vport-vxlan.c
+++ b/net/openvswitch/vport-vxlan.c
@@ -29,7 +29,6 @@
29#include <net/ip.h> 29#include <net/ip.h>
30#include <net/udp.h> 30#include <net/udp.h>
31#include <net/ip_tunnels.h> 31#include <net/ip_tunnels.h>
32#include <net/udp.h>
33#include <net/rtnetlink.h> 32#include <net/rtnetlink.h>
34#include <net/route.h> 33#include <net/route.h>
35#include <net/dsfield.h> 34#include <net/dsfield.h>
generated by cgit v1.2.2 (git 2.25.0) at 2025-02-26 09:46:15 -0500