[mirror_ubuntu-jammy-kernel.git] / net / netfilter / nf_flow_table_core.c

// SPDX-License-Identifier: GPL-2.0-only
#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/module.h>
#include <linux/netfilter.h>
#include <linux/rhashtable.h>
#include <linux/netdevice.h>
#include <net/ip.h>
#include <net/ip6_route.h>
#include <net/netfilter/nf_tables.h>
#include <net/netfilter/nf_flow_table.h>
#include <net/netfilter/nf_conntrack.h>
#include <net/netfilter/nf_conntrack_core.h>
#include <net/netfilter/nf_conntrack_l4proto.h>
#include <net/netfilter/nf_conntrack_tuple.h>

static DEFINE_MUTEX(flowtable_lock);
static LIST_HEAD(flowtables);

static void
flow_offload_fill_dir(struct flow_offload *flow,
		      enum flow_offload_tuple_dir dir)
{
	struct flow_offload_tuple *ft = &flow->tuplehash[dir].tuple;
	struct nf_conntrack_tuple *ctt = &flow->ct->tuplehash[dir].tuple;

	ft->dir = dir;

	switch (ctt->src.l3num) {
	case NFPROTO_IPV4:
		ft->src_v4 = ctt->src.u3.in;
		ft->dst_v4 = ctt->dst.u3.in;
		break;
	case NFPROTO_IPV6:
		ft->src_v6 = ctt->src.u3.in6;
		ft->dst_v6 = ctt->dst.u3.in6;
		break;
	}

	ft->l3proto = ctt->src.l3num;
	ft->l4proto = ctt->dst.protonum;
	ft->src_port = ctt->src.u.tcp.port;
	ft->dst_port = ctt->dst.u.tcp.port;
}

struct flow_offload *flow_offload_alloc(struct nf_conn *ct)
{
	struct flow_offload *flow;

	if (unlikely(nf_ct_is_dying(ct) ||
	    !atomic_inc_not_zero(&ct->ct_general.use)))
		return NULL;

	flow = kzalloc(sizeof(*flow), GFP_ATOMIC);
	if (!flow)
		goto err_ct_refcnt;

	flow->ct = ct;

	flow_offload_fill_dir(flow, FLOW_OFFLOAD_DIR_ORIGINAL);
	flow_offload_fill_dir(flow, FLOW_OFFLOAD_DIR_REPLY);

	if (ct->status & IPS_SRC_NAT)
		__set_bit(NF_FLOW_SNAT, &flow->flags);
	if (ct->status & IPS_DST_NAT)
		__set_bit(NF_FLOW_DNAT, &flow->flags);

	return flow;

err_ct_refcnt:
	nf_ct_put(ct);

	return NULL;
}
EXPORT_SYMBOL_GPL(flow_offload_alloc);

static int flow_offload_fill_route(struct flow_offload *flow,
				   const struct nf_flow_route *route,
				   enum flow_offload_tuple_dir dir)
{
	struct flow_offload_tuple *flow_tuple = &flow->tuplehash[dir].tuple;
	struct dst_entry *other_dst = route->tuple[!dir].dst;
	struct dst_entry *dst = route->tuple[dir].dst;

	if (!dst_hold_safe(route->tuple[dir].dst))
		return -1;

	switch (flow_tuple->l3proto) {
	case NFPROTO_IPV4:
		flow_tuple->mtu = ip_dst_mtu_maybe_forward(dst, true);
		break;
	case NFPROTO_IPV6:
		flow_tuple->mtu = ip6_dst_mtu_forward(dst);
		break;
	}

	flow_tuple->iifidx = other_dst->dev->ifindex;
	flow_tuple->dst_cache = dst;

	return 0;
}

int flow_offload_route_init(struct flow_offload *flow,
			    const struct nf_flow_route *route)
{
	int err;

	err = flow_offload_fill_route(flow, route, FLOW_OFFLOAD_DIR_ORIGINAL);
	if (err < 0)
		return err;

	err = flow_offload_fill_route(flow, route, FLOW_OFFLOAD_DIR_REPLY);
	if (err < 0)
		goto err_route_reply;

	flow->type = NF_FLOW_OFFLOAD_ROUTE;

	return 0;

err_route_reply:
	dst_release(route->tuple[FLOW_OFFLOAD_DIR_ORIGINAL].dst);

	return err;
}
EXPORT_SYMBOL_GPL(flow_offload_route_init);

static void flow_offload_fixup_tcp(struct ip_ct_tcp *tcp)
{
	tcp->state = TCP_CONNTRACK_ESTABLISHED;
	tcp->seen[0].td_maxwin = 0;
	tcp->seen[1].td_maxwin = 0;
}

#define NF_FLOWTABLE_TCP_PICKUP_TIMEOUT	(120 * HZ)
#define NF_FLOWTABLE_UDP_PICKUP_TIMEOUT	(30 * HZ)

static void flow_offload_fixup_ct_timeout(struct nf_conn *ct)
{
	const struct nf_conntrack_l4proto *l4proto;
	int l4num = nf_ct_protonum(ct);
	unsigned int timeout;

	l4proto = nf_ct_l4proto_find(l4num);
	if (!l4proto)
		return;

	if (l4num == IPPROTO_TCP)
		timeout = NF_FLOWTABLE_TCP_PICKUP_TIMEOUT;
	else if (l4num == IPPROTO_UDP)
		timeout = NF_FLOWTABLE_UDP_PICKUP_TIMEOUT;
	else
		return;

	if (nf_flow_timeout_delta(ct->timeout) > (__s32)timeout)
		ct->timeout = nfct_time_stamp + timeout;
}

static void flow_offload_fixup_ct_state(struct nf_conn *ct)
{
	if (nf_ct_protonum(ct) == IPPROTO_TCP)
		flow_offload_fixup_tcp(&ct->proto.tcp);
}

static void flow_offload_fixup_ct(struct nf_conn *ct)
{
	flow_offload_fixup_ct_state(ct);
	flow_offload_fixup_ct_timeout(ct);
}

static void flow_offload_route_release(struct flow_offload *flow)
{
	dst_release(flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.dst_cache);
	dst_release(flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple.dst_cache);
}

void flow_offload_free(struct flow_offload *flow)
{
	switch (flow->type) {
	case NF_FLOW_OFFLOAD_ROUTE:
		flow_offload_route_release(flow);
		break;
	default:
		break;
	}
	nf_ct_put(flow->ct);
	kfree_rcu(flow, rcu_head);
}
EXPORT_SYMBOL_GPL(flow_offload_free);

static u32 flow_offload_hash(const void *data, u32 len, u32 seed)
{
	const struct flow_offload_tuple *tuple = data;

	return jhash(tuple, offsetof(struct flow_offload_tuple, dir), seed);
}

static u32 flow_offload_hash_obj(const void *data, u32 len, u32 seed)
{
	const struct flow_offload_tuple_rhash *tuplehash = data;

	return jhash(&tuplehash->tuple, offsetof(struct flow_offload_tuple, dir), seed);
}

static int flow_offload_hash_cmp(struct rhashtable_compare_arg *arg,
					const void *ptr)
{
	const struct flow_offload_tuple *tuple = arg->key;
	const struct flow_offload_tuple_rhash *x = ptr;

	if (memcmp(&x->tuple, tuple, offsetof(struct flow_offload_tuple, dir)))
		return 1;

	return 0;
}

static const struct rhashtable_params nf_flow_offload_rhash_params = {
	.head_offset		= offsetof(struct flow_offload_tuple_rhash, node),
	.hashfn			= flow_offload_hash,
	.obj_hashfn		= flow_offload_hash_obj,
	.obj_cmpfn		= flow_offload_hash_cmp,
	.automatic_shrinking	= true,
};

int flow_offload_add(struct nf_flowtable *flow_table, struct flow_offload *flow)
{
	int err;

	flow->timeout = nf_flowtable_time_stamp + NF_FLOW_TIMEOUT;

	err = rhashtable_insert_fast(&flow_table->rhashtable,
				     &flow->tuplehash[0].node,
				     nf_flow_offload_rhash_params);
	if (err < 0)
		return err;

	err = rhashtable_insert_fast(&flow_table->rhashtable,
				     &flow->tuplehash[1].node,
				     nf_flow_offload_rhash_params);
	if (err < 0) {
		rhashtable_remove_fast(&flow_table->rhashtable,
				       &flow->tuplehash[0].node,
				       nf_flow_offload_rhash_params);
		return err;
	}

	if (nf_flowtable_hw_offload(flow_table)) {
		__set_bit(NF_FLOW_HW, &flow->flags);
		nf_flow_offload_add(flow_table, flow);
	}

	return 0;
}
EXPORT_SYMBOL_GPL(flow_offload_add);

void flow_offload_refresh(struct nf_flowtable *flow_table,
			  struct flow_offload *flow)
{
	flow->timeout = nf_flowtable_time_stamp + NF_FLOW_TIMEOUT;

	if (likely(!nf_flowtable_hw_offload(flow_table) ||
		   !test_and_clear_bit(NF_FLOW_HW_REFRESH, &flow->flags)))
		return;

	nf_flow_offload_add(flow_table, flow);
}
EXPORT_SYMBOL_GPL(flow_offload_refresh);

static inline bool nf_flow_has_expired(const struct flow_offload *flow)
{
	return nf_flow_timeout_delta(flow->timeout) <= 0;
}

static void flow_offload_del(struct nf_flowtable *flow_table,
			     struct flow_offload *flow)
{
	rhashtable_remove_fast(&flow_table->rhashtable,
			       &flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].node,
			       nf_flow_offload_rhash_params);
	rhashtable_remove_fast(&flow_table->rhashtable,
			       &flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].node,
			       nf_flow_offload_rhash_params);

	clear_bit(IPS_OFFLOAD_BIT, &flow->ct->status);

	if (nf_flow_has_expired(flow))
		flow_offload_fixup_ct(flow->ct);
	else
		flow_offload_fixup_ct_timeout(flow->ct);

	flow_offload_free(flow);
}

void flow_offload_teardown(struct flow_offload *flow)
{
	set_bit(NF_FLOW_TEARDOWN, &flow->flags);

	flow_offload_fixup_ct_state(flow->ct);
}
EXPORT_SYMBOL_GPL(flow_offload_teardown);

struct flow_offload_tuple_rhash *
flow_offload_lookup(struct nf_flowtable *flow_table,
		    struct flow_offload_tuple *tuple)
{
	struct flow_offload_tuple_rhash *tuplehash;
	struct flow_offload *flow;
	int dir;

	tuplehash = rhashtable_lookup(&flow_table->rhashtable, tuple,
				      nf_flow_offload_rhash_params);
	if (!tuplehash)
		return NULL;

	dir = tuplehash->tuple.dir;
	flow = container_of(tuplehash, struct flow_offload, tuplehash[dir]);
	if (test_bit(NF_FLOW_TEARDOWN, &flow->flags))
		return NULL;

	if (unlikely(nf_ct_is_dying(flow->ct)))
		return NULL;

	return tuplehash;
}
EXPORT_SYMBOL_GPL(flow_offload_lookup);

static int
nf_flow_table_iterate(struct nf_flowtable *flow_table,
		      void (*iter)(struct flow_offload *flow, void *data),
		      void *data)
{
	struct flow_offload_tuple_rhash *tuplehash;
	struct rhashtable_iter hti;
	struct flow_offload *flow;
	int err = 0;

	rhashtable_walk_enter(&flow_table->rhashtable, &hti);
	rhashtable_walk_start(&hti);

	while ((tuplehash = rhashtable_walk_next(&hti))) {
		if (IS_ERR(tuplehash)) {
			if (PTR_ERR(tuplehash) != -EAGAIN) {
				err = PTR_ERR(tuplehash);
				break;
			}
			continue;
		}
		if (tuplehash->tuple.dir)
			continue;

		flow = container_of(tuplehash, struct flow_offload, tuplehash[0]);

		iter(flow, data);
	}
	rhashtable_walk_stop(&hti);
	rhashtable_walk_exit(&hti);

	return err;
}

static void nf_flow_offload_gc_step(struct flow_offload *flow, void *data)
{
	struct nf_flowtable *flow_table = data;

	if (nf_flow_has_expired(flow) || nf_ct_is_dying(flow->ct))
		set_bit(NF_FLOW_TEARDOWN, &flow->flags);

	if (test_bit(NF_FLOW_TEARDOWN, &flow->flags)) {
		if (test_bit(NF_FLOW_HW, &flow->flags)) {
			if (!test_bit(NF_FLOW_HW_DYING, &flow->flags))
				nf_flow_offload_del(flow_table, flow);
			else if (test_bit(NF_FLOW_HW_DEAD, &flow->flags))
				flow_offload_del(flow_table, flow);
		} else {
			flow_offload_del(flow_table, flow);
		}
	} else if (test_bit(NF_FLOW_HW, &flow->flags)) {
		nf_flow_offload_stats(flow_table, flow);
	}
}

static void nf_flow_offload_work_gc(struct work_struct *work)
{
	struct nf_flowtable *flow_table;

	flow_table = container_of(work, struct nf_flowtable, gc_work.work);
	nf_flow_table_iterate(flow_table, nf_flow_offload_gc_step, flow_table);
	queue_delayed_work(system_power_efficient_wq, &flow_table->gc_work, HZ);
}


static int nf_flow_nat_port_tcp(struct sk_buff *skb, unsigned int thoff,
				__be16 port, __be16 new_port)
{
	struct tcphdr *tcph;

	if (!pskb_may_pull(skb, thoff + sizeof(*tcph)) ||
	    skb_try_make_writable(skb, thoff + sizeof(*tcph)))
		return -1;

	tcph = (void *)(skb_network_header(skb) + thoff);
	inet_proto_csum_replace2(&tcph->check, skb, port, new_port, true);

	return 0;
}

static int nf_flow_nat_port_udp(struct sk_buff *skb, unsigned int thoff,
				__be16 port, __be16 new_port)
{
	struct udphdr *udph;

	if (!pskb_may_pull(skb, thoff + sizeof(*udph)) ||
	    skb_try_make_writable(skb, thoff + sizeof(*udph)))
		return -1;

	udph = (void *)(skb_network_header(skb) + thoff);
	if (udph->check || skb->ip_summed == CHECKSUM_PARTIAL) {
		inet_proto_csum_replace2(&udph->check, skb, port,
					 new_port, true);
		if (!udph->check)
			udph->check = CSUM_MANGLED_0;
	}

	return 0;
}

static int nf_flow_nat_port(struct sk_buff *skb, unsigned int thoff,
			    u8 protocol, __be16 port, __be16 new_port)
{
	switch (protocol) {
	case IPPROTO_TCP:
		if (nf_flow_nat_port_tcp(skb, thoff, port, new_port) < 0)
			return NF_DROP;
		break;
	case IPPROTO_UDP:
		if (nf_flow_nat_port_udp(skb, thoff, port, new_port) < 0)
			return NF_DROP;
		break;
	}

	return 0;
}

int nf_flow_snat_port(const struct flow_offload *flow,
		      struct sk_buff *skb, unsigned int thoff,
		      u8 protocol, enum flow_offload_tuple_dir dir)
{
	struct flow_ports *hdr;
	__be16 port, new_port;

	if (!pskb_may_pull(skb, thoff + sizeof(*hdr)) ||
	    skb_try_make_writable(skb, thoff + sizeof(*hdr)))
		return -1;

	hdr = (void *)(skb_network_header(skb) + thoff);

	switch (dir) {
	case FLOW_OFFLOAD_DIR_ORIGINAL:
		port = hdr->source;
		new_port = flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple.dst_port;
		hdr->source = new_port;
		break;
	case FLOW_OFFLOAD_DIR_REPLY:
		port = hdr->dest;
		new_port = flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.src_port;
		hdr->dest = new_port;
		break;
	default:
		return -1;
	}

	return nf_flow_nat_port(skb, thoff, protocol, port, new_port);
}
EXPORT_SYMBOL_GPL(nf_flow_snat_port);

int nf_flow_dnat_port(const struct flow_offload *flow,
		      struct sk_buff *skb, unsigned int thoff,
		      u8 protocol, enum flow_offload_tuple_dir dir)
{
	struct flow_ports *hdr;
	__be16 port, new_port;

	if (!pskb_may_pull(skb, thoff + sizeof(*hdr)) ||
	    skb_try_make_writable(skb, thoff + sizeof(*hdr)))
		return -1;

	hdr = (void *)(skb_network_header(skb) + thoff);

	switch (dir) {
	case FLOW_OFFLOAD_DIR_ORIGINAL:
		port = hdr->dest;
		new_port = flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple.src_port;
		hdr->dest = new_port;
		break;
	case FLOW_OFFLOAD_DIR_REPLY:
		port = hdr->source;
		new_port = flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.dst_port;
		hdr->source = new_port;
		break;
	default:
		return -1;
	}

	return nf_flow_nat_port(skb, thoff, protocol, port, new_port);
}
EXPORT_SYMBOL_GPL(nf_flow_dnat_port);

int nf_flow_table_init(struct nf_flowtable *flowtable)
{
	int err;

	INIT_DEFERRABLE_WORK(&flowtable->gc_work, nf_flow_offload_work_gc);
	flow_block_init(&flowtable->flow_block);
	init_rwsem(&flowtable->flow_block_lock);

	err = rhashtable_init(&flowtable->rhashtable,
			      &nf_flow_offload_rhash_params);
	if (err < 0)
		return err;

	queue_delayed_work(system_power_efficient_wq,
			   &flowtable->gc_work, HZ);

	mutex_lock(&flowtable_lock);
	list_add(&flowtable->list, &flowtables);
	mutex_unlock(&flowtable_lock);

	return 0;
}
EXPORT_SYMBOL_GPL(nf_flow_table_init);

static void nf_flow_table_do_cleanup(struct flow_offload *flow, void *data)
{
	struct net_device *dev = data;

	if (!dev) {
		flow_offload_teardown(flow);
		return;
	}

	if (net_eq(nf_ct_net(flow->ct), dev_net(dev)) &&
	    (flow->tuplehash[0].tuple.iifidx == dev->ifindex ||
	     flow->tuplehash[1].tuple.iifidx == dev->ifindex))
		flow_offload_teardown(flow);
}

void nf_flow_table_gc_cleanup(struct nf_flowtable *flowtable,
			      struct net_device *dev)
{
	nf_flow_table_iterate(flowtable, nf_flow_table_do_cleanup, dev);
	flush_delayed_work(&flowtable->gc_work);
	nf_flow_table_offload_flush(flowtable);
}

void nf_flow_table_cleanup(struct net_device *dev)
{
	struct nf_flowtable *flowtable;

	mutex_lock(&flowtable_lock);
	list_for_each_entry(flowtable, &flowtables, list)
		nf_flow_table_gc_cleanup(flowtable, dev);
	mutex_unlock(&flowtable_lock);
}
EXPORT_SYMBOL_GPL(nf_flow_table_cleanup);

void nf_flow_table_free(struct nf_flowtable *flow_table)
{
	mutex_lock(&flowtable_lock);
	list_del(&flow_table->list);
	mutex_unlock(&flowtable_lock);

	cancel_delayed_work_sync(&flow_table->gc_work);
	nf_flow_table_iterate(flow_table, nf_flow_table_do_cleanup, NULL);
	nf_flow_table_iterate(flow_table, nf_flow_offload_gc_step, flow_table);
	nf_flow_table_offload_flush(flow_table);
	if (nf_flowtable_hw_offload(flow_table))
		nf_flow_table_iterate(flow_table, nf_flow_offload_gc_step,
				      flow_table);
	rhashtable_destroy(&flow_table->rhashtable);
}
EXPORT_SYMBOL_GPL(nf_flow_table_free);

static int __init nf_flow_table_module_init(void)
{
	return nf_flow_table_offload_init();
}

static void __exit nf_flow_table_module_exit(void)
{
	nf_flow_table_offload_exit();
}

module_init(nf_flow_table_module_init);
module_exit(nf_flow_table_module_exit);

MODULE_LICENSE("GPL");
MODULE_AUTHOR("Pablo Neira Ayuso <pablo@netfilter.org>");
Commit	Line	Data
09c434b8	1	// SPDX-License-Identifier: GPL-2.0-only
ac2a6666 PNA	2	#include <linux/kernel.h>
	3	#include <linux/init.h>
	4	#include <linux/module.h>
	5	#include <linux/netfilter.h>
	6	#include <linux/rhashtable.h>
	7	#include <linux/netdevice.h>
4f3780c0 FF	8	#include <net/ip.h>
4f3780c0 FF	9	#include <net/ip6_route.h>
c0ea1bcb	10	#include <net/netfilter/nf_tables.h>
ac2a6666 PNA	11	#include <net/netfilter/nf_flow_table.h>
	12	#include <net/netfilter/nf_conntrack.h>
	13	#include <net/netfilter/nf_conntrack_core.h>
40d102cd	14	#include <net/netfilter/nf_conntrack_l4proto.h>
ac2a6666 PNA	15	#include <net/netfilter/nf_conntrack_tuple.h>
ac2a6666 PNA	16
84453a90 FF	17	static DEFINE_MUTEX(flowtable_lock);
	18	static LIST_HEAD(flowtables);
	19
047b300e	20	static void
458a1828	21	flow_offload_fill_dir(struct flow_offload *flow,
047b300e FF	22	enum flow_offload_tuple_dir dir)
	23	{
	24	struct flow_offload_tuple *ft = &flow->tuplehash[dir].tuple;
458a1828	25	struct nf_conntrack_tuple *ctt = &flow->ct->tuplehash[dir].tuple;
047b300e FF	26
	27	ft->dir = dir;
	28
	29	switch (ctt->src.l3num) {
	30	case NFPROTO_IPV4:
	31	ft->src_v4 = ctt->src.u3.in;
	32	ft->dst_v4 = ctt->dst.u3.in;
	33	break;
	34	case NFPROTO_IPV6:
	35	ft->src_v6 = ctt->src.u3.in6;
	36	ft->dst_v6 = ctt->dst.u3.in6;
	37	break;
	38	}
	39
	40	ft->l3proto = ctt->src.l3num;
	41	ft->l4proto = ctt->dst.protonum;
	42	ft->src_port = ctt->src.u.tcp.port;
	43	ft->dst_port = ctt->dst.u.tcp.port;
047b300e FF	44	}
047b300e FF	45
f1363e05	46	struct flow_offload flow_offload_alloc(struct nf_conn ct)
ac2a6666	47	{
ac2a6666 PNA	48	struct flow_offload *flow;
	49
	50	if (unlikely(nf_ct_is_dying(ct) \|\|
	51	!atomic_inc_not_zero(&ct->ct_general.use)))
	52	return NULL;
	53
62248df8 PNA	54	flow = kzalloc(sizeof(*flow), GFP_ATOMIC);
62248df8 PNA	55	if (!flow)
ac2a6666 PNA	56	goto err_ct_refcnt;
ac2a6666 PNA	57
b32d2f34	58	flow->ct = ct;
ac2a6666	59
458a1828	60	flow_offload_fill_dir(flow, FLOW_OFFLOAD_DIR_ORIGINAL);
458a1828	61	flow_offload_fill_dir(flow, FLOW_OFFLOAD_DIR_REPLY);
ac2a6666 PNA	62
ac2a6666 PNA	63	if (ct->status & IPS_SRC_NAT)
355a8b13	64	__set_bit(NF_FLOW_SNAT, &flow->flags);
df1e2025	65	if (ct->status & IPS_DST_NAT)
355a8b13	66	__set_bit(NF_FLOW_DNAT, &flow->flags);
ac2a6666 PNA	67
	68	return flow;
	69
ac2a6666 PNA	70	err_ct_refcnt:
	71	nf_ct_put(ct);
	72
	73	return NULL;
	74	}
	75	EXPORT_SYMBOL_GPL(flow_offload_alloc);
	76
f1363e05 PNA	77	static int flow_offload_fill_route(struct flow_offload *flow,
	78	const struct nf_flow_route *route,
	79	enum flow_offload_tuple_dir dir)
	80	{
	81	struct flow_offload_tuple *flow_tuple = &flow->tuplehash[dir].tuple;
	82	struct dst_entry *other_dst = route->tuple[!dir].dst;
	83	struct dst_entry *dst = route->tuple[dir].dst;
	84
	85	if (!dst_hold_safe(route->tuple[dir].dst))
	86	return -1;
	87
	88	switch (flow_tuple->l3proto) {
	89	case NFPROTO_IPV4:
	90	flow_tuple->mtu = ip_dst_mtu_maybe_forward(dst, true);
	91	break;
	92	case NFPROTO_IPV6:
	93	flow_tuple->mtu = ip6_dst_mtu_forward(dst);
	94	break;
	95	}
	96
	97	flow_tuple->iifidx = other_dst->dev->ifindex;
	98	flow_tuple->dst_cache = dst;
	99
	100	return 0;
	101	}
	102
	103	int flow_offload_route_init(struct flow_offload *flow,
	104	const struct nf_flow_route *route)
	105	{
	106	int err;
	107
	108	err = flow_offload_fill_route(flow, route, FLOW_OFFLOAD_DIR_ORIGINAL);
	109	if (err < 0)
	110	return err;
	111
	112	err = flow_offload_fill_route(flow, route, FLOW_OFFLOAD_DIR_REPLY);
	113	if (err < 0)
	114	goto err_route_reply;
	115
	116	flow->type = NF_FLOW_OFFLOAD_ROUTE;
	117
	118	return 0;
	119
	120	err_route_reply:
	121	dst_release(route->tuple[FLOW_OFFLOAD_DIR_ORIGINAL].dst);
	122
	123	return err;
	124	}
	125	EXPORT_SYMBOL_GPL(flow_offload_route_init);
	126
da5984e5 FF	127	static void flow_offload_fixup_tcp(struct ip_ct_tcp *tcp)
	128	{
	129	tcp->state = TCP_CONNTRACK_ESTABLISHED;
	130	tcp->seen[0].td_maxwin = 0;
	131	tcp->seen[1].td_maxwin = 0;
	132	}
	133
e97d9404 FW	134	#define NF_FLOWTABLE_TCP_PICKUP_TIMEOUT (120 * HZ)
	135	#define NF_FLOWTABLE_UDP_PICKUP_TIMEOUT (30 * HZ)
	136
1e5b2471	137	static void flow_offload_fixup_ct_timeout(struct nf_conn *ct)
da5984e5 FF	138	{
da5984e5 FF	139	const struct nf_conntrack_l4proto *l4proto;
1e5b2471	140	int l4num = nf_ct_protonum(ct);
da5984e5	141	unsigned int timeout;
da5984e5	142
4a60dc74	143	l4proto = nf_ct_l4proto_find(l4num);
da5984e5 FF	144	if (!l4proto)
	145	return;
	146
da5984e5	147	if (l4num == IPPROTO_TCP)
e97d9404	148	timeout = NF_FLOWTABLE_TCP_PICKUP_TIMEOUT;
da5984e5	149	else if (l4num == IPPROTO_UDP)
e97d9404	150	timeout = NF_FLOWTABLE_UDP_PICKUP_TIMEOUT;
da5984e5 FF	151	else
	152	return;
	153
1e5b2471 PNA	154	if (nf_flow_timeout_delta(ct->timeout) > (__s32)timeout)
	155	ct->timeout = nfct_time_stamp + timeout;
	156	}
	157
	158	static void flow_offload_fixup_ct_state(struct nf_conn *ct)
	159	{
	160	if (nf_ct_protonum(ct) == IPPROTO_TCP)
	161	flow_offload_fixup_tcp(&ct->proto.tcp);
	162	}
	163
	164	static void flow_offload_fixup_ct(struct nf_conn *ct)
	165	{
	166	flow_offload_fixup_ct_state(ct);
	167	flow_offload_fixup_ct_timeout(ct);
da5984e5 FF	168	}
da5984e5 FF	169
f1363e05	170	static void flow_offload_route_release(struct flow_offload *flow)
ac2a6666	171	{
ac2a6666 PNA	172	dst_release(flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.dst_cache);
ac2a6666 PNA	173	dst_release(flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple.dst_cache);
f1363e05 PNA	174	}
	175
	176	void flow_offload_free(struct flow_offload *flow)
	177	{
	178	switch (flow->type) {
	179	case NF_FLOW_OFFLOAD_ROUTE:
	180	flow_offload_route_release(flow);
	181	break;
	182	default:
	183	break;
	184	}
b32d2f34	185	nf_ct_put(flow->ct);
62248df8	186	kfree_rcu(flow, rcu_head);
ac2a6666 PNA	187	}
	188	EXPORT_SYMBOL_GPL(flow_offload_free);
	189
a268de77 FF	190	static u32 flow_offload_hash(const void *data, u32 len, u32 seed)
	191	{
	192	const struct flow_offload_tuple *tuple = data;
	193
	194	return jhash(tuple, offsetof(struct flow_offload_tuple, dir), seed);
	195	}
	196
	197	static u32 flow_offload_hash_obj(const void *data, u32 len, u32 seed)
	198	{
	199	const struct flow_offload_tuple_rhash *tuplehash = data;
	200
	201	return jhash(&tuplehash->tuple, offsetof(struct flow_offload_tuple, dir), seed);
	202	}
	203
	204	static int flow_offload_hash_cmp(struct rhashtable_compare_arg *arg,
	205	const void *ptr)
	206	{
	207	const struct flow_offload_tuple *tuple = arg->key;
	208	const struct flow_offload_tuple_rhash *x = ptr;
	209
	210	if (memcmp(&x->tuple, tuple, offsetof(struct flow_offload_tuple, dir)))
	211	return 1;
	212
	213	return 0;
	214	}
	215
	216	static const struct rhashtable_params nf_flow_offload_rhash_params = {
	217	.head_offset = offsetof(struct flow_offload_tuple_rhash, node),
	218	.hashfn = flow_offload_hash,
	219	.obj_hashfn = flow_offload_hash_obj,
	220	.obj_cmpfn = flow_offload_hash_cmp,
	221	.automatic_shrinking = true,
	222	};
	223
ac2a6666 PNA	224	int flow_offload_add(struct nf_flowtable flow_table, struct flow_offload flow)
ac2a6666 PNA	225	{
43c8f131 TY	226	int err;
43c8f131 TY	227
fb46f1b7	228	flow->timeout = nf_flowtable_time_stamp + NF_FLOW_TIMEOUT;
daf61b02	229
43c8f131 TY	230	err = rhashtable_insert_fast(&flow_table->rhashtable,
	231	&flow->tuplehash[0].node,
	232	nf_flow_offload_rhash_params);
	233	if (err < 0)
	234	return err;
	235
	236	err = rhashtable_insert_fast(&flow_table->rhashtable,
	237	&flow->tuplehash[1].node,
	238	nf_flow_offload_rhash_params);
	239	if (err < 0) {
	240	rhashtable_remove_fast(&flow_table->rhashtable,
	241	&flow->tuplehash[0].node,
	242	nf_flow_offload_rhash_params);
	243	return err;
	244	}
ac2a6666	245
f698fe40 PNA	246	if (nf_flowtable_hw_offload(flow_table)) {
f698fe40 PNA	247	__set_bit(NF_FLOW_HW, &flow->flags);
c29f74e0	248	nf_flow_offload_add(flow_table, flow);
f698fe40	249	}
c29f74e0	250
ac2a6666 PNA	251	return 0;
	252	}
	253	EXPORT_SYMBOL_GPL(flow_offload_add);
	254
8b3646d6 PB	255	void flow_offload_refresh(struct nf_flowtable *flow_table,
	256	struct flow_offload *flow)
	257	{
	258	flow->timeout = nf_flowtable_time_stamp + NF_FLOW_TIMEOUT;
	259
	260	if (likely(!nf_flowtable_hw_offload(flow_table) \|\|
	261	!test_and_clear_bit(NF_FLOW_HW_REFRESH, &flow->flags)))
	262	return;
	263
	264	nf_flow_offload_add(flow_table, flow);
	265	}
	266	EXPORT_SYMBOL_GPL(flow_offload_refresh);
	267
3e68db2f PNA	268	static inline bool nf_flow_has_expired(const struct flow_offload *flow)
3e68db2f PNA	269	{
1e5b2471	270	return nf_flow_timeout_delta(flow->timeout) <= 0;
3e68db2f PNA	271	}
3e68db2f PNA	272
0ff90b6c FF	273	static void flow_offload_del(struct nf_flowtable *flow_table,
0ff90b6c FF	274	struct flow_offload *flow)
ac2a6666	275	{
ac2a6666 PNA	276	rhashtable_remove_fast(&flow_table->rhashtable,
ac2a6666 PNA	277	&flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].node,
a268de77	278	nf_flow_offload_rhash_params);
ac2a6666 PNA	279	rhashtable_remove_fast(&flow_table->rhashtable,
ac2a6666 PNA	280	&flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].node,
a268de77	281	nf_flow_offload_rhash_params);
ac2a6666	282
b32d2f34	283	clear_bit(IPS_OFFLOAD_BIT, &flow->ct->status);
da5984e5	284
3e68db2f	285	if (nf_flow_has_expired(flow))
b32d2f34	286	flow_offload_fixup_ct(flow->ct);
9ed81c8e	287	else
b32d2f34	288	flow_offload_fixup_ct_timeout(flow->ct);
3e68db2f	289
0ff90b6c	290	flow_offload_free(flow);
ac2a6666	291	}
ac2a6666	292
59c466dd FF	293	void flow_offload_teardown(struct flow_offload *flow)
59c466dd FF	294	{
355a8b13	295	set_bit(NF_FLOW_TEARDOWN, &flow->flags);
da5984e5	296
b32d2f34	297	flow_offload_fixup_ct_state(flow->ct);
59c466dd FF	298	}
	299	EXPORT_SYMBOL_GPL(flow_offload_teardown);
	300
ac2a6666 PNA	301	struct flow_offload_tuple_rhash *
	302	flow_offload_lookup(struct nf_flowtable *flow_table,
	303	struct flow_offload_tuple *tuple)
	304	{
ba03137f FF	305	struct flow_offload_tuple_rhash *tuplehash;
	306	struct flow_offload *flow;
	307	int dir;
	308
a2d88182 TY	309	tuplehash = rhashtable_lookup(&flow_table->rhashtable, tuple,
a2d88182 TY	310	nf_flow_offload_rhash_params);
ba03137f FF	311	if (!tuplehash)
	312	return NULL;
	313
	314	dir = tuplehash->tuple.dir;
	315	flow = container_of(tuplehash, struct flow_offload, tuplehash[dir]);
355a8b13	316	if (test_bit(NF_FLOW_TEARDOWN, &flow->flags))
ba03137f FF	317	return NULL;
ba03137f FF	318
b32d2f34	319	if (unlikely(nf_ct_is_dying(flow->ct)))
8cd2bc98 TY	320	return NULL;
8cd2bc98 TY	321
ba03137f	322	return tuplehash;
ac2a6666 PNA	323	}
	324	EXPORT_SYMBOL_GPL(flow_offload_lookup);
	325
49de9c09 TY	326	static int
	327	nf_flow_table_iterate(struct nf_flowtable *flow_table,
	328	void (iter)(struct flow_offload flow, void *data),
	329	void *data)
ac2a6666 PNA	330	{
	331	struct flow_offload_tuple_rhash *tuplehash;
	332	struct rhashtable_iter hti;
	333	struct flow_offload *flow;
0de22baa	334	int err = 0;
ac2a6666	335
0de22baa	336	rhashtable_walk_enter(&flow_table->rhashtable, &hti);
ac2a6666 PNA	337	rhashtable_walk_start(&hti);
	338
	339	while ((tuplehash = rhashtable_walk_next(&hti))) {
	340	if (IS_ERR(tuplehash)) {
0de22baa TY	341	if (PTR_ERR(tuplehash) != -EAGAIN) {
	342	err = PTR_ERR(tuplehash);
	343	break;
	344	}
ac2a6666 PNA	345	continue;
	346	}
	347	if (tuplehash->tuple.dir)
	348	continue;
	349
	350	flow = container_of(tuplehash, struct flow_offload, tuplehash[0]);
	351
	352	iter(flow, data);
	353	}
ac2a6666 PNA	354	rhashtable_walk_stop(&hti);
	355	rhashtable_walk_exit(&hti);
	356
	357	return err;
	358	}
ac2a6666	359
b9660987	360	static void nf_flow_offload_gc_step(struct flow_offload flow, void data)
ac2a6666	361	{
b9660987	362	struct nf_flowtable *flow_table = data;
ac2a6666	363
9ed81c8e PNA	364	if (nf_flow_has_expired(flow) \|\| nf_ct_is_dying(flow->ct))
	365	set_bit(NF_FLOW_TEARDOWN, &flow->flags);
	366
	367	if (test_bit(NF_FLOW_TEARDOWN, &flow->flags)) {
355a8b13 PNA	368	if (test_bit(NF_FLOW_HW, &flow->flags)) {
355a8b13 PNA	369	if (!test_bit(NF_FLOW_HW_DYING, &flow->flags))
c29f74e0	370	nf_flow_offload_del(flow_table, flow);
355a8b13	371	else if (test_bit(NF_FLOW_HW_DEAD, &flow->flags))
c29f74e0 PNA	372	flow_offload_del(flow_table, flow);
	373	} else {
	374	flow_offload_del(flow_table, flow);
	375	}
355a8b13	376	} else if (test_bit(NF_FLOW_HW, &flow->flags)) {
79b9b685	377	nf_flow_offload_stats(flow_table, flow);
c29f74e0	378	}
b408c5b0 PNA	379	}
b408c5b0 PNA	380
a268de77	381	static void nf_flow_offload_work_gc(struct work_struct *work)
b408c5b0 PNA	382	{
	383	struct nf_flowtable *flow_table;
	384
	385	flow_table = container_of(work, struct nf_flowtable, gc_work.work);
b9660987	386	nf_flow_table_iterate(flow_table, nf_flow_offload_gc_step, flow_table);
ac2a6666 PNA	387	queue_delayed_work(system_power_efficient_wq, &flow_table->gc_work, HZ);
ac2a6666 PNA	388	}
ac2a6666	389
978703f4	390
ac2a6666 PNA	391	static int nf_flow_nat_port_tcp(struct sk_buff *skb, unsigned int thoff,
	392	__be16 port, __be16 new_port)
	393	{
	394	struct tcphdr *tcph;
	395
	396	if (!pskb_may_pull(skb, thoff + sizeof(*tcph)) \|\|
	397	skb_try_make_writable(skb, thoff + sizeof(*tcph)))
	398	return -1;
	399
	400	tcph = (void *)(skb_network_header(skb) + thoff);
	401	inet_proto_csum_replace2(&tcph->check, skb, port, new_port, true);
	402
	403	return 0;
	404	}
	405
	406	static int nf_flow_nat_port_udp(struct sk_buff *skb, unsigned int thoff,
	407	__be16 port, __be16 new_port)
	408	{
	409	struct udphdr *udph;
	410
	411	if (!pskb_may_pull(skb, thoff + sizeof(*udph)) \|\|
	412	skb_try_make_writable(skb, thoff + sizeof(*udph)))
	413	return -1;
	414
	415	udph = (void *)(skb_network_header(skb) + thoff);
	416	if (udph->check \|\| skb->ip_summed == CHECKSUM_PARTIAL) {
	417	inet_proto_csum_replace2(&udph->check, skb, port,
	418	new_port, true);
	419	if (!udph->check)
	420	udph->check = CSUM_MANGLED_0;
	421	}
	422
	423	return 0;
	424	}
	425
	426	static int nf_flow_nat_port(struct sk_buff *skb, unsigned int thoff,
	427	u8 protocol, __be16 port, __be16 new_port)
	428	{
	429	switch (protocol) {
	430	case IPPROTO_TCP:
	431	if (nf_flow_nat_port_tcp(skb, thoff, port, new_port) < 0)
	432	return NF_DROP;
	433	break;
	434	case IPPROTO_UDP:
	435	if (nf_flow_nat_port_udp(skb, thoff, port, new_port) < 0)
	436	return NF_DROP;
	437	break;
	438	}
	439
	440	return 0;
	441	}
	442
	443	int nf_flow_snat_port(const struct flow_offload *flow,
	444	struct sk_buff *skb, unsigned int thoff,
	445	u8 protocol, enum flow_offload_tuple_dir dir)
	446	{
	447	struct flow_ports *hdr;
	448	__be16 port, new_port;
	449
	450	if (!pskb_may_pull(skb, thoff + sizeof(*hdr)) \|\|
	451	skb_try_make_writable(skb, thoff + sizeof(*hdr)))
	452	return -1;
	453
	454	hdr = (void *)(skb_network_header(skb) + thoff);
455
456	switch (dir) {
457	case FLOW_OFFLOAD_DIR_ORIGINAL:
458	port = hdr->source;
459	new_port = flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple.dst_port;
460	hdr->source = new_port;
461	break;
462	case FLOW_OFFLOAD_DIR_REPLY:
463	port = hdr->dest;
464	new_port = flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.src_port;
465	hdr->dest = new_port;
466	break;
467	default:
468	return -1;
469	}
470
471	return nf_flow_nat_port(skb, thoff, protocol, port, new_port);
472	}
473	EXPORT_SYMBOL_GPL(nf_flow_snat_port);
474
475	int nf_flow_dnat_port(const struct flow_offload *flow,
476	struct sk_buff *skb, unsigned int thoff,
477	u8 protocol, enum flow_offload_tuple_dir dir)
478	{
479	struct flow_ports *hdr;
480	__be16 port, new_port;
481
482	if (!pskb_may_pull(skb, thoff + sizeof(*hdr)) \|\|
483	skb_try_make_writable(skb, thoff + sizeof(*hdr)))
484	return -1;
485
486	hdr = (void *)(skb_network_header(skb) + thoff);
487
488	switch (dir) {
489	case FLOW_OFFLOAD_DIR_ORIGINAL:
490	port = hdr->dest;
491	new_port = flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple.src_port;
492	hdr->dest = new_port;
493	break;
494	case FLOW_OFFLOAD_DIR_REPLY:
495	port = hdr->source;
496	new_port = flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple.dst_port;
497	hdr->source = new_port;
498	break;
499	default:
500	return -1;
501	}
502
503	return nf_flow_nat_port(skb, thoff, protocol, port, new_port);
504	}
505	EXPORT_SYMBOL_GPL(nf_flow_dnat_port);
506
a268de77 FF	507	int nf_flow_table_init(struct nf_flowtable *flowtable)
	508	{
	509	int err;
	510
	511	INIT_DEFERRABLE_WORK(&flowtable->gc_work, nf_flow_offload_work_gc);
c29f74e0	512	flow_block_init(&flowtable->flow_block);
422c032a	513	init_rwsem(&flowtable->flow_block_lock);
a268de77 FF	514
	515	err = rhashtable_init(&flowtable->rhashtable,
	516	&nf_flow_offload_rhash_params);
	517	if (err < 0)
	518	return err;
	519
	520	queue_delayed_work(system_power_efficient_wq,
	521	&flowtable->gc_work, HZ);
	522
84453a90 FF	523	mutex_lock(&flowtable_lock);
	524	list_add(&flowtable->list, &flowtables);
	525	mutex_unlock(&flowtable_lock);
	526
a268de77 FF	527	return 0;
	528	}
	529	EXPORT_SYMBOL_GPL(nf_flow_table_init);
	530
c0ea1bcb PNA	531	static void nf_flow_table_do_cleanup(struct flow_offload flow, void data)
	532	{
	533	struct net_device *dev = data;
	534
59c466dd FF	535	if (!dev) {
59c466dd FF	536	flow_offload_teardown(flow);
c0ea1bcb	537	return;
59c466dd	538	}
b32d2f34 PNA	539
b32d2f34 PNA	540	if (net_eq(nf_ct_net(flow->ct), dev_net(dev)) &&
a3fb3698 TY	541	(flow->tuplehash[0].tuple.iifidx == dev->ifindex \|\|
a3fb3698 TY	542	flow->tuplehash[1].tuple.iifidx == dev->ifindex))
445db8d0	543	flow_offload_teardown(flow);
c0ea1bcb PNA	544	}
c0ea1bcb PNA	545
a8284c68 PNA	546	void nf_flow_table_gc_cleanup(struct nf_flowtable *flowtable,
a8284c68 PNA	547	struct net_device *dev)
c0ea1bcb	548	{
84453a90	549	nf_flow_table_iterate(flowtable, nf_flow_table_do_cleanup, dev);
c0ea1bcb	550	flush_delayed_work(&flowtable->gc_work);
91bfaa15	551	nf_flow_table_offload_flush(flowtable);
c0ea1bcb PNA	552	}
c0ea1bcb PNA	553
5f1be84a	554	void nf_flow_table_cleanup(struct net_device *dev)
c0ea1bcb	555	{
84453a90 FF	556	struct nf_flowtable *flowtable;
	557
	558	mutex_lock(&flowtable_lock);
	559	list_for_each_entry(flowtable, &flowtables, list)
a8284c68	560	nf_flow_table_gc_cleanup(flowtable, dev);
84453a90	561	mutex_unlock(&flowtable_lock);
c0ea1bcb PNA	562	}
	563	EXPORT_SYMBOL_GPL(nf_flow_table_cleanup);
	564
b408c5b0 PNA	565	void nf_flow_table_free(struct nf_flowtable *flow_table)
b408c5b0 PNA	566	{
84453a90 FF	567	mutex_lock(&flowtable_lock);
	568	list_del(&flow_table->list);
	569	mutex_unlock(&flowtable_lock);
978703f4	570
a268de77	571	cancel_delayed_work_sync(&flow_table->gc_work);
b408c5b0	572	nf_flow_table_iterate(flow_table, nf_flow_table_do_cleanup, NULL);
b9660987	573	nf_flow_table_iterate(flow_table, nf_flow_offload_gc_step, flow_table);
0f34f30a	574	nf_flow_table_offload_flush(flow_table);
c921ffe8 PB	575	if (nf_flowtable_hw_offload(flow_table))
	576	nf_flow_table_iterate(flow_table, nf_flow_offload_gc_step,
	577	flow_table);
a268de77	578	rhashtable_destroy(&flow_table->rhashtable);
b408c5b0 PNA	579	}
	580	EXPORT_SYMBOL_GPL(nf_flow_table_free);
	581
c29f74e0 PNA	582	static int __init nf_flow_table_module_init(void)
	583	{
	584	return nf_flow_table_offload_init();
	585	}
	586
	587	static void __exit nf_flow_table_module_exit(void)
	588	{
	589	nf_flow_table_offload_exit();
	590	}
	591
	592	module_init(nf_flow_table_module_init);
	593	module_exit(nf_flow_table_module_exit);
	594
ac2a6666 PNA	595	MODULE_LICENSE("GPL");
ac2a6666 PNA	596	MODULE_AUTHOR("Pablo Neira Ayuso <pablo@netfilter.org>");