1 files changed, 737 insertions, 164 deletions

diff --git a/net/xfrm/xfrm_policy.c b/net/xfrm/xfrm_policy.c
index 3da67ca2c3c..b6e2e79d726 100644
--- a/net/xfrm/xfrm_policy.c
+++ b/net/xfrm/xfrm_policy.c
@@ -22,16 +22,19 @@
 #include <linux/netdevice.h>
 #include <linux/netfilter.h>
 #include <linux/module.h>
+#include <linux/cache.h>
 #include <net/xfrm.h>
 #include <net/ip.h>
 
+#include "xfrm_hash.h"
+
 DEFINE_MUTEX(xfrm_cfg_mutex);
 EXPORT_SYMBOL(xfrm_cfg_mutex);
 
 static DEFINE_RWLOCK(xfrm_policy_lock);
 
-struct xfrm_policy *xfrm_policy_list[XFRM_POLICY_MAX*2];
-EXPORT_SYMBOL(xfrm_policy_list);
+unsigned int xfrm_policy_count[XFRM_POLICY_MAX*2];
+EXPORT_SYMBOL(xfrm_policy_count);
 
 static DEFINE_RWLOCK(xfrm_policy_afinfo_lock);
 static struct xfrm_policy_afinfo *xfrm_policy_afinfo[NPROTO];
@@ -39,8 +42,7 @@ static struct xfrm_policy_afinfo *xfrm_policy_afinfo[NPROTO];
 static kmem_cache_t *xfrm_dst_cache __read_mostly;
 
 static struct work_struct xfrm_policy_gc_work;
-static struct list_head xfrm_policy_gc_list =
-	LIST_HEAD_INIT(xfrm_policy_gc_list);
+static HLIST_HEAD(xfrm_policy_gc_list);
 static DEFINE_SPINLOCK(xfrm_policy_gc_lock);
 
 static struct xfrm_policy_afinfo *xfrm_policy_get_afinfo(unsigned short family);
@@ -310,8 +312,10 @@ struct xfrm_policy *xfrm_policy_alloc(gfp_t gfp)
 	policy = kzalloc(sizeof(struct xfrm_policy), gfp);
 
 	if (policy) {
-		atomic_set(&policy->refcnt, 1);
+		INIT_HLIST_NODE(&policy->bydst);
+		INIT_HLIST_NODE(&policy->byidx);
 		rwlock_init(&policy->lock);
+		atomic_set(&policy->refcnt, 1);
 		init_timer(&policy->timer);
 		policy->timer.data = (unsigned long)policy;
 		policy->timer.function = xfrm_policy_timer;
@@ -357,17 +361,16 @@ static void xfrm_policy_gc_kill(struct xfrm_policy *policy)
 static void xfrm_policy_gc_task(void *data)
 {
 	struct xfrm_policy *policy;
-	struct list_head *entry, *tmp;
-	struct list_head gc_list = LIST_HEAD_INIT(gc_list);
+	struct hlist_node *entry, *tmp;
+	struct hlist_head gc_list;
 
 	spin_lock_bh(&xfrm_policy_gc_lock);
-	list_splice_init(&xfrm_policy_gc_list, &gc_list);
+	gc_list.first = xfrm_policy_gc_list.first;
+	INIT_HLIST_HEAD(&xfrm_policy_gc_list);
 	spin_unlock_bh(&xfrm_policy_gc_lock);
 
-	list_for_each_safe(entry, tmp, &gc_list) {
-		policy = list_entry(entry, struct xfrm_policy, list);
+	hlist_for_each_entry_safe(policy, entry, tmp, &gc_list, bydst)
 		xfrm_policy_gc_kill(policy);
-	}
 }
 
 /* Rule must be locked. Release descentant resources, announce
@@ -389,70 +392,275 @@ static void xfrm_policy_kill(struct xfrm_policy *policy)
 	}
 
 	spin_lock(&xfrm_policy_gc_lock);
-	list_add(&policy->list, &xfrm_policy_gc_list);
+	hlist_add_head(&policy->bydst, &xfrm_policy_gc_list);
 	spin_unlock(&xfrm_policy_gc_lock);
 
 	schedule_work(&xfrm_policy_gc_work);
 }
 
+struct xfrm_policy_hash {
+	struct hlist_head	*table;
+	unsigned int		hmask;
+};
+
+static struct hlist_head xfrm_policy_inexact[XFRM_POLICY_MAX*2];
+static struct xfrm_policy_hash xfrm_policy_bydst[XFRM_POLICY_MAX*2] __read_mostly;
+static struct hlist_head *xfrm_policy_byidx __read_mostly;
+static unsigned int xfrm_idx_hmask __read_mostly;
+static unsigned int xfrm_policy_hashmax __read_mostly = 1 * 1024 * 1024;
+
+static inline unsigned int idx_hash(u32 index)
+{
+	return __idx_hash(index, xfrm_idx_hmask);
+}
+
+static struct hlist_head *policy_hash_bysel(struct xfrm_selector *sel, unsigned short family, int dir)
+{
+	unsigned int hmask = xfrm_policy_bydst[dir].hmask;
+	unsigned int hash = __sel_hash(sel, family, hmask);
+
+	return (hash == hmask + 1 ?
+		&xfrm_policy_inexact[dir] :
+		xfrm_policy_bydst[dir].table + hash);
+}
+
+static struct hlist_head *policy_hash_direct(xfrm_address_t *daddr, xfrm_address_t *saddr, unsigned short family, int dir)
+{
+	unsigned int hmask = xfrm_policy_bydst[dir].hmask;
+	unsigned int hash = __addr_hash(daddr, saddr, family, hmask);
+
+	return xfrm_policy_bydst[dir].table + hash;
+}
+
+static void xfrm_dst_hash_transfer(struct hlist_head *list,
+				   struct hlist_head *ndsttable,
+				   unsigned int nhashmask)
+{
+	struct hlist_node *entry, *tmp;
+	struct xfrm_policy *pol;
+
+	hlist_for_each_entry_safe(pol, entry, tmp, list, bydst) {
+		unsigned int h;
+
+		h = __addr_hash(&pol->selector.daddr, &pol->selector.saddr,
+				pol->family, nhashmask);
+		hlist_add_head(&pol->bydst, ndsttable+h);
+	}
+}
+
+static void xfrm_idx_hash_transfer(struct hlist_head *list,
+				   struct hlist_head *nidxtable,
+				   unsigned int nhashmask)
+{
+	struct hlist_node *entry, *tmp;
+	struct xfrm_policy *pol;
+
+	hlist_for_each_entry_safe(pol, entry, tmp, list, byidx) {
+		unsigned int h;
+
+		h = __idx_hash(pol->index, nhashmask);
+		hlist_add_head(&pol->byidx, nidxtable+h);
+	}
+}
+
+static unsigned long xfrm_new_hash_mask(unsigned int old_hmask)
+{
+	return ((old_hmask + 1) << 1) - 1;
+}
+
+static void xfrm_bydst_resize(int dir)
+{
+	unsigned int hmask = xfrm_policy_bydst[dir].hmask;
+	unsigned int nhashmask = xfrm_new_hash_mask(hmask);
+	unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
+	struct hlist_head *odst = xfrm_policy_bydst[dir].table;
+	struct hlist_head *ndst = xfrm_hash_alloc(nsize);
+	int i;
+
+	if (!ndst)
+		return;
+
+	write_lock_bh(&xfrm_policy_lock);
+
+	for (i = hmask; i >= 0; i--)
+		xfrm_dst_hash_transfer(odst + i, ndst, nhashmask);
+
+	xfrm_policy_bydst[dir].table = ndst;
+	xfrm_policy_bydst[dir].hmask = nhashmask;
+
+	write_unlock_bh(&xfrm_policy_lock);
+
+	xfrm_hash_free(odst, (hmask + 1) * sizeof(struct hlist_head));
+}
+
+static void xfrm_byidx_resize(int total)
+{
+	unsigned int hmask = xfrm_idx_hmask;
+	unsigned int nhashmask = xfrm_new_hash_mask(hmask);
+	unsigned int nsize = (nhashmask + 1) * sizeof(struct hlist_head);
+	struct hlist_head *oidx = xfrm_policy_byidx;
+	struct hlist_head *nidx = xfrm_hash_alloc(nsize);
+	int i;
+
+	if (!nidx)
+		return;
+
+	write_lock_bh(&xfrm_policy_lock);
+
+	for (i = hmask; i >= 0; i--)
+		xfrm_idx_hash_transfer(oidx + i, nidx, nhashmask);
+
+	xfrm_policy_byidx = nidx;
+	xfrm_idx_hmask = nhashmask;
+
+	write_unlock_bh(&xfrm_policy_lock);
+
+	xfrm_hash_free(oidx, (hmask + 1) * sizeof(struct hlist_head));
+}
+
+static inline int xfrm_bydst_should_resize(int dir, int *total)
+{
+	unsigned int cnt = xfrm_policy_count[dir];
+	unsigned int hmask = xfrm_policy_bydst[dir].hmask;
+
+	if (total)
+		*total += cnt;
+
+	if ((hmask + 1) < xfrm_policy_hashmax &&
+	    cnt > hmask)
+		return 1;
+
+	return 0;
+}
+
+static inline int xfrm_byidx_should_resize(int total)
+{
+	unsigned int hmask = xfrm_idx_hmask;
+
+	if ((hmask + 1) < xfrm_policy_hashmax &&
+	    total > hmask)
+		return 1;
+
+	return 0;
+}
+
+static DEFINE_MUTEX(hash_resize_mutex);
+
+static void xfrm_hash_resize(void *__unused)
+{
+	int dir, total;
+
+	mutex_lock(&hash_resize_mutex);
+
+	total = 0;
+	for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
+		if (xfrm_bydst_should_resize(dir, &total))
+			xfrm_bydst_resize(dir);
+	}
+	if (xfrm_byidx_should_resize(total))
+		xfrm_byidx_resize(total);
+
+	mutex_unlock(&hash_resize_mutex);
+}
+
+static DECLARE_WORK(xfrm_hash_work, xfrm_hash_resize, NULL);
+
 /* Generate new index... KAME seems to generate them ordered by cost
  * of an absolute inpredictability of ordering of rules. This will not pass. */
-static u32 xfrm_gen_index(int dir)
+static u32 xfrm_gen_index(u8 type, int dir)
 {
-	u32 idx;
-	struct xfrm_policy *p;
 	static u32 idx_generator;
 
 	for (;;) {
+		struct hlist_node *entry;
+		struct hlist_head *list;
+		struct xfrm_policy *p;
+		u32 idx;
+		int found;
+
 		idx = (idx_generator | dir);
 		idx_generator += 8;
 		if (idx == 0)
 			idx = 8;
-		for (p = xfrm_policy_list[dir]; p; p = p->next) {
-			if (p->index == idx)
+		list = xfrm_policy_byidx + idx_hash(idx);
+		found = 0;
+		hlist_for_each_entry(p, entry, list, byidx) {
+			if (p->index == idx) {
+				found = 1;
 				break;
+			}
 		}
-		if (!p)
+		if (!found)
 			return idx;
 	}
 }
 
+static inline int selector_cmp(struct xfrm_selector *s1, struct xfrm_selector *s2)
+{
+	u32 *p1 = (u32 *) s1;
+	u32 *p2 = (u32 *) s2;
+	int len = sizeof(struct xfrm_selector) / sizeof(u32);
+	int i;
+
+	for (i = 0; i < len; i++) {
+		if (p1[i] != p2[i])
+			return 1;
+	}
+
+	return 0;
+}
+
 int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
 {
-	struct xfrm_policy *pol, **p;
-	struct xfrm_policy *delpol = NULL;
-	struct xfrm_policy **newpos = NULL;
+	struct xfrm_policy *pol;
+	struct xfrm_policy *delpol;
+	struct hlist_head *chain;
+	struct hlist_node *entry, *newpos, *last;
 	struct dst_entry *gc_list;
 
 	write_lock_bh(&xfrm_policy_lock);
-	for (p = &xfrm_policy_list[dir]; (pol=*p)!=NULL;) {
-		if (!delpol && memcmp(&policy->selector, &pol->selector, sizeof(pol->selector)) == 0 &&
+	chain = policy_hash_bysel(&policy->selector, policy->family, dir);
+	delpol = NULL;
+	newpos = NULL;
+	last = NULL;
+	hlist_for_each_entry(pol, entry, chain, bydst) {
+		if (!delpol &&
+		    pol->type == policy->type &&
+		    !selector_cmp(&pol->selector, &policy->selector) &&
 		    xfrm_sec_ctx_match(pol->security, policy->security)) {
 			if (excl) {
 				write_unlock_bh(&xfrm_policy_lock);
 				return -EEXIST;
 			}
-			*p = pol->next;
 			delpol = pol;
 			if (policy->priority > pol->priority)
 				continue;
 		} else if (policy->priority >= pol->priority) {
-			p = &pol->next;
+			last = &pol->bydst;
 			continue;
 		}
 		if (!newpos)
-			newpos = p;
+			newpos = &pol->bydst;
 		if (delpol)
 			break;
-		p = &pol->next;
+		last = &pol->bydst;
 	}
+	if (!newpos)
+		newpos = last;
 	if (newpos)
-		p = newpos;
+		hlist_add_after(newpos, &policy->bydst);
+	else
+		hlist_add_head(&policy->bydst, chain);
 	xfrm_pol_hold(policy);
-	policy->next = *p;
-	*p = policy;
+	xfrm_policy_count[dir]++;
 	atomic_inc(&flow_cache_genid);
-	policy->index = delpol ? delpol->index : xfrm_gen_index(dir);
+	if (delpol) {
+		hlist_del(&delpol->bydst);
+		hlist_del(&delpol->byidx);
+		xfrm_policy_count[dir]--;
+	}
+	policy->index = delpol ? delpol->index : xfrm_gen_index(policy->type, dir);
+	hlist_add_head(&policy->byidx, xfrm_policy_byidx+idx_hash(policy->index));
 	policy->curlft.add_time = (unsigned long)xtime.tv_sec;
 	policy->curlft.use_time = 0;
 	if (!mod_timer(&policy->timer, jiffies + HZ))
@@ -461,10 +669,13 @@ int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
 
 	if (delpol)
 		xfrm_policy_kill(delpol);
+	else if (xfrm_bydst_should_resize(dir, NULL))
+		schedule_work(&xfrm_hash_work);
 
 	read_lock_bh(&xfrm_policy_lock);
 	gc_list = NULL;
-	for (policy = policy->next; policy; policy = policy->next) {
+	entry = &policy->bydst;
+	hlist_for_each_entry_continue(policy, entry, bydst) {
 		struct dst_entry *dst;
 
 		write_lock(&policy->lock);
@@ -493,87 +704,146 @@ int xfrm_policy_insert(int dir, struct xfrm_policy *policy, int excl)
 }
 EXPORT_SYMBOL(xfrm_policy_insert);
 
-struct xfrm_policy *xfrm_policy_bysel_ctx(int dir, struct xfrm_selector *sel,
+struct xfrm_policy *xfrm_policy_bysel_ctx(u8 type, int dir,
+					  struct xfrm_selector *sel,
 					  struct xfrm_sec_ctx *ctx, int delete)
 {
-	struct xfrm_policy *pol, **p;
+	struct xfrm_policy *pol, *ret;
+	struct hlist_head *chain;
+	struct hlist_node *entry;
 
 	write_lock_bh(&xfrm_policy_lock);
-	for (p = &xfrm_policy_list[dir]; (pol=*p)!=NULL; p = &pol->next) {
-		if ((memcmp(sel, &pol->selector, sizeof(*sel)) == 0) &&
-		    (xfrm_sec_ctx_match(ctx, pol->security))) {
+	chain = policy_hash_bysel(sel, sel->family, dir);
+	ret = NULL;
+	hlist_for_each_entry(pol, entry, chain, bydst) {
+		if (pol->type == type &&
+		    !selector_cmp(sel, &pol->selector) &&
+		    xfrm_sec_ctx_match(ctx, pol->security)) {
 			xfrm_pol_hold(pol);
-			if (delete)
-				*p = pol->next;
+			if (delete) {
+				hlist_del(&pol->bydst);
+				hlist_del(&pol->byidx);
+				xfrm_policy_count[dir]--;
+			}
+			ret = pol;
 			break;
 		}
 	}
 	write_unlock_bh(&xfrm_policy_lock);
 
-	if (pol && delete) {
+	if (ret && delete) {
 		atomic_inc(&flow_cache_genid);
-		xfrm_policy_kill(pol);
+		xfrm_policy_kill(ret);
 	}
-	return pol;
+	return ret;
 }
 EXPORT_SYMBOL(xfrm_policy_bysel_ctx);
 
-struct xfrm_policy *xfrm_policy_byid(int dir, u32 id, int delete)
+struct xfrm_policy *xfrm_policy_byid(u8 type, int dir, u32 id, int delete)
 {
-	struct xfrm_policy *pol, **p;
+	struct xfrm_policy *pol, *ret;
+	struct hlist_head *chain;
+	struct hlist_node *entry;
 
 	write_lock_bh(&xfrm_policy_lock);
-	for (p = &xfrm_policy_list[dir]; (pol=*p)!=NULL; p = &pol->next) {
-		if (pol->index == id) {
+	chain = xfrm_policy_byidx + idx_hash(id);
+	ret = NULL;
+	hlist_for_each_entry(pol, entry, chain, byidx) {
+		if (pol->type == type && pol->index == id) {
 			xfrm_pol_hold(pol);
-			if (delete)
-				*p = pol->next;
+			if (delete) {
+				hlist_del(&pol->bydst);
+				hlist_del(&pol->byidx);
+				xfrm_policy_count[dir]--;
+			}
+			ret = pol;
 			break;
 		}
 	}
 	write_unlock_bh(&xfrm_policy_lock);
 
-	if (pol && delete) {
+	if (ret && delete) {
 		atomic_inc(&flow_cache_genid);
-		xfrm_policy_kill(pol);
+		xfrm_policy_kill(ret);
 	}
-	return pol;
+	return ret;
 }
 EXPORT_SYMBOL(xfrm_policy_byid);
 
-void xfrm_policy_flush(void)
+void xfrm_policy_flush(u8 type)
 {
-	struct xfrm_policy *xp;
 	int dir;
 
 	write_lock_bh(&xfrm_policy_lock);
 	for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
-		while ((xp = xfrm_policy_list[dir]) != NULL) {
-			xfrm_policy_list[dir] = xp->next;
+		struct xfrm_policy *pol;
+		struct hlist_node *entry;
+		int i;
+
+	again1:
+		hlist_for_each_entry(pol, entry,
+				     &xfrm_policy_inexact[dir], bydst) {
+			if (pol->type != type)
+				continue;
+			hlist_del(&pol->bydst);
+			hlist_del(&pol->byidx);
 			write_unlock_bh(&xfrm_policy_lock);
 
-			xfrm_policy_kill(xp);
+			xfrm_policy_kill(pol);
 
 			write_lock_bh(&xfrm_policy_lock);
+			goto again1;
 		}
+
+		for (i = xfrm_policy_bydst[dir].hmask; i >= 0; i--) {
+	again2:
+			hlist_for_each_entry(pol, entry,
+					     xfrm_policy_bydst[dir].table + i,
+					     bydst) {
+				if (pol->type != type)
+					continue;
+				hlist_del(&pol->bydst);
+				hlist_del(&pol->byidx);
+				write_unlock_bh(&xfrm_policy_lock);
+
+				xfrm_policy_kill(pol);
+
+				write_lock_bh(&xfrm_policy_lock);
+				goto again2;
+			}
+		}
+
+		xfrm_policy_count[dir] = 0;
 	}
 	atomic_inc(&flow_cache_genid);
 	write_unlock_bh(&xfrm_policy_lock);
 }
 EXPORT_SYMBOL(xfrm_policy_flush);
 
-int xfrm_policy_walk(int (*func)(struct xfrm_policy *, int, int, void*),
+int xfrm_policy_walk(u8 type, int (*func)(struct xfrm_policy *, int, int, void*),
 		     void *data)
 {
-	struct xfrm_policy *xp;
-	int dir;
-	int count = 0;
-	int error = 0;
+	struct xfrm_policy *pol;
+	struct hlist_node *entry;
+	int dir, count, error;
 
 	read_lock_bh(&xfrm_policy_lock);
+	count = 0;
 	for (dir = 0; dir < 2*XFRM_POLICY_MAX; dir++) {
-		for (xp = xfrm_policy_list[dir]; xp; xp = xp->next)
-			count++;
+		struct hlist_head *table = xfrm_policy_bydst[dir].table;
+		int i;
+
+		hlist_for_each_entry(pol, entry,
+				     &xfrm_policy_inexact[dir], bydst) {
+			if (pol->type == type)
+				count++;
+		}
+		for (i = xfrm_policy_bydst[dir].hmask; i >= 0; i--) {
+			hlist_for_each_entry(pol, entry, table + i, bydst) {
+				if (pol->type == type)
+					count++;
+			}
+		}
 	}
 
 	if (count == 0) {
@@ -582,13 +852,28 @@ int xfrm_policy_walk(int (*func)(struct xfrm_policy *, int, int, void*),
 	}
 
 	for (dir = 0; dir < 2*XFRM_POLICY_MAX; dir++) {
-		for (xp = xfrm_policy_list[dir]; xp; xp = xp->next) {
-			error = func(xp, dir%XFRM_POLICY_MAX, --count, data);
+		struct hlist_head *table = xfrm_policy_bydst[dir].table;
+		int i;
+
+		hlist_for_each_entry(pol, entry,
+				     &xfrm_policy_inexact[dir], bydst) {
+			if (pol->type != type)
+				continue;
+			error = func(pol, dir % XFRM_POLICY_MAX, --count, data);
 			if (error)
 				goto out;
 		}
+		for (i = xfrm_policy_bydst[dir].hmask; i >= 0; i--) {
+			hlist_for_each_entry(pol, entry, table + i, bydst) {
+				if (pol->type != type)
+					continue;
+				error = func(pol, dir % XFRM_POLICY_MAX, --count, data);
+				if (error)
+					goto out;
+			}
+		}
 	}
-
+	error = 0;
 out:
 	read_unlock_bh(&xfrm_policy_lock);
 	return error;
@@ -597,29 +882,79 @@ EXPORT_SYMBOL(xfrm_policy_walk);
 
 /* Find policy to apply to this flow. */
 
-static void xfrm_policy_lookup(struct flowi *fl, u32 sk_sid, u16 family, u8 dir,
-			       void **objp, atomic_t **obj_refp)
+static int xfrm_policy_match(struct xfrm_policy *pol, struct flowi *fl,
+			     u8 type, u16 family, int dir)
 {
-	struct xfrm_policy *pol;
+	struct xfrm_selector *sel = &pol->selector;
+	int match;
 
-	read_lock_bh(&xfrm_policy_lock);
-	for (pol = xfrm_policy_list[dir]; pol; pol = pol->next) {
-		struct xfrm_selector *sel = &pol->selector;
-		int match;
+	if (pol->family != family ||
+	    pol->type != type)
+		return 0;
 
-		if (pol->family != family)
-			continue;
+	match = xfrm_selector_match(sel, fl, family);
+	if (match) {
+		if (!security_xfrm_policy_lookup(pol, fl->secid, dir))
+			return 1;
+	}
 
-		match = xfrm_selector_match(sel, fl, family);
+	return 0;
+}
 
-		if (match) {
- 			if (!security_xfrm_policy_lookup(pol, sk_sid, dir)) {
-				xfrm_pol_hold(pol);
-				break;
-			}
+static struct xfrm_policy *xfrm_policy_lookup_bytype(u8 type, struct flowi *fl,
+						     u16 family, u8 dir)
+{
+	struct xfrm_policy *pol, *ret;
+	xfrm_address_t *daddr, *saddr;
+	struct hlist_node *entry;
+	struct hlist_head *chain;
+	u32 priority = ~0U;
+
+	daddr = xfrm_flowi_daddr(fl, family);
+	saddr = xfrm_flowi_saddr(fl, family);
+	if (unlikely(!daddr || !saddr))
+		return NULL;
+
+	read_lock_bh(&xfrm_policy_lock);
+	chain = policy_hash_direct(daddr, saddr, family, dir);
+	ret = NULL;
+	hlist_for_each_entry(pol, entry, chain, bydst) {
+		if (xfrm_policy_match(pol, fl, type, family, dir)) {
+			ret = pol;
+			priority = ret->priority;
+			break;
+		}
+	}
+	chain = &xfrm_policy_inexact[dir];
+	hlist_for_each_entry(pol, entry, chain, bydst) {
+		if (xfrm_policy_match(pol, fl, type, family, dir) &&
+		    pol->priority < priority) {
+			ret = pol;
+			break;
 		}
 	}
+	if (ret)
+		xfrm_pol_hold(ret);
 	read_unlock_bh(&xfrm_policy_lock);
+
+	return ret;
+}
+
+static void xfrm_policy_lookup(struct flowi *fl, u16 family, u8 dir,
+			       void **objp, atomic_t **obj_refp)
+{
+	struct xfrm_policy *pol;
+
+#ifdef CONFIG_XFRM_SUB_POLICY
+	pol = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_SUB, fl, family, dir);
+	if (pol)
+		goto end;
+#endif
+	pol = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_MAIN, fl, family, dir);
+
+#ifdef CONFIG_XFRM_SUB_POLICY
+end:
+#endif
 	if ((*objp = (void *) pol) != NULL)
 		*obj_refp = &pol->refcnt;
 }
@@ -641,7 +976,7 @@ static inline int policy_to_flow_dir(int dir)
 	};
 }
 
-static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir, struct flowi *fl, u32 sk_sid)
+static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir, struct flowi *fl)
 {
 	struct xfrm_policy *pol;
 
@@ -652,7 +987,7 @@ static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir, struc
  		int err = 0;
 
 		if (match)
-		  err = security_xfrm_policy_lookup(pol, sk_sid, policy_to_flow_dir(dir));
+		  err = security_xfrm_policy_lookup(pol, fl->secid, policy_to_flow_dir(dir));
 
  		if (match && !err)
 			xfrm_pol_hold(pol);
@@ -665,24 +1000,29 @@ static struct xfrm_policy *xfrm_sk_policy_lookup(struct sock *sk, int dir, struc
 
 static void __xfrm_policy_link(struct xfrm_policy *pol, int dir)
 {
-	pol->next = xfrm_policy_list[dir];
-	xfrm_policy_list[dir] = pol;
+	struct hlist_head *chain = policy_hash_bysel(&pol->selector,
+						     pol->family, dir);
+
+	hlist_add_head(&pol->bydst, chain);
+	hlist_add_head(&pol->byidx, xfrm_policy_byidx+idx_hash(pol->index));
+	xfrm_policy_count[dir]++;
 	xfrm_pol_hold(pol);
+
+	if (xfrm_bydst_should_resize(dir, NULL))
+		schedule_work(&xfrm_hash_work);
 }
 
 static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
 						int dir)
 {
-	struct xfrm_policy **polp;
+	if (hlist_unhashed(&pol->bydst))
+		return NULL;
 
-	for (polp = &xfrm_policy_list[dir];
-	     *polp != NULL; polp = &(*polp)->next) {
-		if (*polp == pol) {
-			*polp = pol->next;
-			return pol;
-		}
-	}
-	return NULL;
+	hlist_del(&pol->bydst);
+	hlist_del(&pol->byidx);
+	xfrm_policy_count[dir]--;
+
+	return pol;
 }
 
 int xfrm_policy_delete(struct xfrm_policy *pol, int dir)
@@ -704,12 +1044,17 @@ int xfrm_sk_policy_insert(struct sock *sk, int dir, struct xfrm_policy *pol)
 {
 	struct xfrm_policy *old_pol;
 
+#ifdef CONFIG_XFRM_SUB_POLICY
+	if (pol && pol->type != XFRM_POLICY_TYPE_MAIN)
+		return -EINVAL;
+#endif
+
 	write_lock_bh(&xfrm_policy_lock);
 	old_pol = sk->sk_policy[dir];
 	sk->sk_policy[dir] = pol;
 	if (pol) {
 		pol->curlft.add_time = (unsigned long)xtime.tv_sec;
-		pol->index = xfrm_gen_index(XFRM_POLICY_MAX+dir);
+		pol->index = xfrm_gen_index(pol->type, XFRM_POLICY_MAX+dir);
 		__xfrm_policy_link(pol, XFRM_POLICY_MAX+dir);
 	}
 	if (old_pol)
@@ -738,6 +1083,7 @@ static struct xfrm_policy *clone_policy(struct xfrm_policy *old, int dir)
 		newp->flags = old->flags;
 		newp->xfrm_nr = old->xfrm_nr;
 		newp->index = old->index;
+		newp->type = old->type;
 		memcpy(newp->xfrm_vec, old->xfrm_vec,
 		       newp->xfrm_nr*sizeof(struct xfrm_tmpl));
 		write_lock_bh(&xfrm_policy_lock);
@@ -761,17 +1107,32 @@ int __xfrm_sk_clone_policy(struct sock *sk)
 	return 0;
 }
 
+static int
+xfrm_get_saddr(xfrm_address_t *local, xfrm_address_t *remote,
+	       unsigned short family)
+{
+	int err;
+	struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
+
+	if (unlikely(afinfo == NULL))
+		return -EINVAL;
+	err = afinfo->get_saddr(local, remote);
+	xfrm_policy_put_afinfo(afinfo);
+	return err;
+}
+
 /* Resolve list of templates for the flow, given policy. */
 
 static int
-xfrm_tmpl_resolve(struct xfrm_policy *policy, struct flowi *fl,
-		  struct xfrm_state **xfrm,
-		  unsigned short family)
+xfrm_tmpl_resolve_one(struct xfrm_policy *policy, struct flowi *fl,
+		      struct xfrm_state **xfrm,
+		      unsigned short family)
 {
 	int nx;
 	int i, error;
 	xfrm_address_t *daddr = xfrm_flowi_daddr(fl, family);
 	xfrm_address_t *saddr = xfrm_flowi_saddr(fl, family);
+	xfrm_address_t tmp;
 
 	for (nx=0, i = 0; i < policy->xfrm_nr; i++) {
 		struct xfrm_state *x;
@@ -779,9 +1140,15 @@ xfrm_tmpl_resolve(struct xfrm_policy *policy, struct flowi *fl,
 		xfrm_address_t *local  = saddr;
 		struct xfrm_tmpl *tmpl = &policy->xfrm_vec[i];
 
-		if (tmpl->mode) {
+		if (tmpl->mode == XFRM_MODE_TUNNEL) {
 			remote = &tmpl->id.daddr;
 			local = &tmpl->saddr;
+			if (xfrm_addr_any(local, family)) {
+				error = xfrm_get_saddr(&tmp, remote, family);
+				if (error)
+					goto fail;
+				local = &tmp;
+			}
 		}
 
 		x = xfrm_state_find(remote, local, fl, tmpl, policy, &error, family);
@@ -809,6 +1176,45 @@ fail:
 	return error;
 }
 
+static int
+xfrm_tmpl_resolve(struct xfrm_policy **pols, int npols, struct flowi *fl,
+		  struct xfrm_state **xfrm,
+		  unsigned short family)
+{
+	struct xfrm_state *tp[XFRM_MAX_DEPTH];
+	struct xfrm_state **tpp = (npols > 1) ? tp : xfrm;
+	int cnx = 0;
+	int error;
+	int ret;
+	int i;
+
+	for (i = 0; i < npols; i++) {
+		if (cnx + pols[i]->xfrm_nr >= XFRM_MAX_DEPTH) {
+			error = -ENOBUFS;
+			goto fail;
+		}
+
+		ret = xfrm_tmpl_resolve_one(pols[i], fl, &tpp[cnx], family);
+		if (ret < 0) {
+			error = ret;
+			goto fail;
+		} else
+			cnx += ret;
+	}
+
+	/* found states are sorted for outbound processing */
+	if (npols > 1)
+		xfrm_state_sort(xfrm, tpp, cnx, family);
+
+	return cnx;
+
+ fail:
+	for (cnx--; cnx>=0; cnx--)
+		xfrm_state_put(tpp[cnx]);
+	return error;
+
+}
+
 /* Check that the bundle accepts the flow and its components are
  * still valid.
  */
@@ -855,6 +1261,11 @@ int xfrm_lookup(struct dst_entry **dst_p, struct flowi *fl,
 		struct sock *sk, int flags)
 {
 	struct xfrm_policy *policy;
+	struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
+	int npols;
+	int pol_dead;
+	int xfrm_nr;
+	int pi;
 	struct xfrm_state *xfrm[XFRM_MAX_DEPTH];
 	struct dst_entry *dst, *dst_orig = *dst_p;
 	int nx = 0;
@@ -862,19 +1273,26 @@ int xfrm_lookup(struct dst_entry **dst_p, struct flowi *fl,
 	u32 genid;
 	u16 family;
 	u8 dir = policy_to_flow_dir(XFRM_POLICY_OUT);
-	u32 sk_sid = security_sk_sid(sk, fl, dir);
+
 restart:
 	genid = atomic_read(&flow_cache_genid);
 	policy = NULL;
+	for (pi = 0; pi < ARRAY_SIZE(pols); pi++)
+		pols[pi] = NULL;
+	npols = 0;
+	pol_dead = 0;
+	xfrm_nr = 0;
+
 	if (sk && sk->sk_policy[1])
-		policy = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl, sk_sid);
+		policy = xfrm_sk_policy_lookup(sk, XFRM_POLICY_OUT, fl);
 
 	if (!policy) {
 		/* To accelerate a bit...  */
-		if ((dst_orig->flags & DST_NOXFRM) || !xfrm_policy_list[XFRM_POLICY_OUT])
+		if ((dst_orig->flags & DST_NOXFRM) ||
+		    !xfrm_policy_count[XFRM_POLICY_OUT])
 			return 0;
 
-		policy = flow_cache_lookup(fl, sk_sid, dst_orig->ops->family,
+		policy = flow_cache_lookup(fl, dst_orig->ops->family,
 					   dir, xfrm_policy_lookup);
 	}
 
@@ -883,6 +1301,9 @@ restart:
 
 	family = dst_orig->ops->family;
 	policy->curlft.use_time = (unsigned long)xtime.tv_sec;
+	pols[0] = policy;
+	npols ++;
+	xfrm_nr += pols[0]->xfrm_nr;
 
 	switch (policy->action) {
 	case XFRM_POLICY_BLOCK:
@@ -891,11 +1312,13 @@ restart:
 		goto error;
 
 	case XFRM_POLICY_ALLOW:
+#ifndef CONFIG_XFRM_SUB_POLICY
 		if (policy->xfrm_nr == 0) {
 			/* Flow passes not transformed. */
 			xfrm_pol_put(policy);
 			return 0;
 		}
+#endif
 
 		/* Try to find matching bundle.
 		 *
@@ -911,7 +1334,36 @@ restart:
 		if (dst)
 			break;
 
-		nx = xfrm_tmpl_resolve(policy, fl, xfrm, family);
+#ifdef CONFIG_XFRM_SUB_POLICY
+		if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
+			pols[1] = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_MAIN,
+							    fl, family,
+							    XFRM_POLICY_OUT);
+			if (pols[1]) {
+				if (pols[1]->action == XFRM_POLICY_BLOCK) {
+					err = -EPERM;
+					goto error;
+				}
+				npols ++;
+				xfrm_nr += pols[1]->xfrm_nr;
+			}
+		}
+
+		/*
+		 * Because neither flowi nor bundle information knows about
+		 * transformation template size. On more than one policy usage
+		 * we can realize whether all of them is bypass or not after
+		 * they are searched. See above not-transformed bypass
+		 * is surrounded by non-sub policy configuration, too.
+		 */
+		if (xfrm_nr == 0) {
+			/* Flow passes not transformed. */
+			xfrm_pols_put(pols, npols);
+			return 0;
+		}
+
+#endif
+		nx = xfrm_tmpl_resolve(pols, npols, fl, xfrm, family);
 
 		if (unlikely(nx<0)) {
 			err = nx;
@@ -924,7 +1376,7 @@ restart:
 				set_current_state(TASK_RUNNING);
 				remove_wait_queue(&km_waitq, &wait);
 
-				nx = xfrm_tmpl_resolve(policy, fl, xfrm, family);
+				nx = xfrm_tmpl_resolve(pols, npols, fl, xfrm, family);
 
 				if (nx == -EAGAIN && signal_pending(current)) {
 					err = -ERESTART;
@@ -932,7 +1384,7 @@ restart:
 				}
 				if (nx == -EAGAIN ||
 				    genid != atomic_read(&flow_cache_genid)) {
-					xfrm_pol_put(policy);
+					xfrm_pols_put(pols, npols);
 					goto restart;
 				}
 				err = nx;
@@ -942,7 +1394,7 @@ restart:
 		}
 		if (nx == 0) {
 			/* Flow passes not transformed. */
-			xfrm_pol_put(policy);
+			xfrm_pols_put(pols, npols);
 			return 0;
 		}
 
@@ -956,8 +1408,14 @@ restart:
 			goto error;
 		}
 
+		for (pi = 0; pi < npols; pi++) {
+			read_lock_bh(&pols[pi]->lock);
+			pol_dead |= pols[pi]->dead;
+			read_unlock_bh(&pols[pi]->lock);
+		}
+
 		write_lock_bh(&policy->lock);
-		if (unlikely(policy->dead || stale_bundle(dst))) {
+		if (unlikely(pol_dead || stale_bundle(dst))) {
 			/* Wow! While we worked on resolving, this
 			 * policy has gone. Retry. It is not paranoia,
 			 * we just cannot enlist new bundle to dead object.
@@ -977,17 +1435,34 @@ restart:
 	}
 	*dst_p = dst;
 	dst_release(dst_orig);
-	xfrm_pol_put(policy);
+ 	xfrm_pols_put(pols, npols);
 	return 0;
 
 error:
 	dst_release(dst_orig);
-	xfrm_pol_put(policy);
+	xfrm_pols_put(pols, npols);
 	*dst_p = NULL;
 	return err;
 }
 EXPORT_SYMBOL(xfrm_lookup);
 
+static inline int
+xfrm_secpath_reject(int idx, struct sk_buff *skb, struct flowi *fl)
+{
+	struct xfrm_state *x;
+	int err;
+
+	if (!skb->sp || idx < 0 || idx >= skb->sp->len)
+		return 0;
+	x = skb->sp->xvec[idx];
+	if (!x->type->reject)
+		return 0;
+	xfrm_state_hold(x);
+	err = x->type->reject(x, skb, fl);
+	xfrm_state_put(x);
+	return err;
+}
+
 /* When skb is transformed back to its "native" form, we have to
  * check policy restrictions. At the moment we make this in maximally
  * stupid way. Shame on me. :-) Of course, connected sockets must
@@ -1004,10 +1479,19 @@ xfrm_state_ok(struct xfrm_tmpl *tmpl, struct xfrm_state *x,
 		(x->id.spi == tmpl->id.spi || !tmpl->id.spi) &&
 		(x->props.reqid == tmpl->reqid || !tmpl->reqid) &&
 		x->props.mode == tmpl->mode &&
-		(tmpl->aalgos & (1<<x->props.aalgo)) &&
-		!(x->props.mode && xfrm_state_addr_cmp(tmpl, x, family));
+		((tmpl->aalgos & (1<<x->props.aalgo)) ||
+		 !(xfrm_id_proto_match(tmpl->id.proto, IPSEC_PROTO_ANY))) &&
+		!(x->props.mode != XFRM_MODE_TRANSPORT &&
+		  xfrm_state_addr_cmp(tmpl, x, family));
 }
 
+/*
+ * 0 or more than 0 is returned when validation is succeeded (either bypass
+ * because of optional transport mode, or next index of the mathced secpath
+ * state with the template.
+ * -1 is returned when no matching template is found.
+ * Otherwise "-2 - errored_index" is returned.
+ */
 static inline int
 xfrm_policy_ok(struct xfrm_tmpl *tmpl, struct sec_path *sp, int start,
 	       unsigned short family)
@@ -1015,15 +1499,18 @@ xfrm_policy_ok(struct xfrm_tmpl *tmpl, struct sec_path *sp, int start,
 	int idx = start;
 
 	if (tmpl->optional) {
-		if (!tmpl->mode)
+		if (tmpl->mode == XFRM_MODE_TRANSPORT)
 			return start;
 	} else
 		start = -1;
 	for (; idx < sp->len; idx++) {
 		if (xfrm_state_ok(tmpl, sp->xvec[idx], family))
 			return ++idx;
-		if (sp->xvec[idx]->props.mode)
+		if (sp->xvec[idx]->props.mode != XFRM_MODE_TRANSPORT) {
+			if (start == -1)
+				start = -2-idx;
 			break;
+		}
 	}
 	return start;
 }
@@ -1032,21 +1519,25 @@ int
 xfrm_decode_session(struct sk_buff *skb, struct flowi *fl, unsigned short family)
 {
 	struct xfrm_policy_afinfo *afinfo = xfrm_policy_get_afinfo(family);
+	int err;
 
 	if (unlikely(afinfo == NULL))
 		return -EAFNOSUPPORT;
 
 	afinfo->decode_session(skb, fl);
+	err = security_xfrm_decode_session(skb, &fl->secid);
 	xfrm_policy_put_afinfo(afinfo);
-	return 0;
+	return err;
 }
 EXPORT_SYMBOL(xfrm_decode_session);
 
-static inline int secpath_has_tunnel(struct sec_path *sp, int k)
+static inline int secpath_has_nontransport(struct sec_path *sp, int k, int *idxp)
 {
 	for (; k < sp->len; k++) {
-		if (sp->xvec[k]->props.mode)
+		if (sp->xvec[k]->props.mode != XFRM_MODE_TRANSPORT) {
+			*idxp = k;
 			return 1;
+		}
 	}
 
 	return 0;
@@ -1056,16 +1547,18 @@ int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
 			unsigned short family)
 {
 	struct xfrm_policy *pol;
+	struct xfrm_policy *pols[XFRM_POLICY_TYPE_MAX];
+	int npols = 0;
+	int xfrm_nr;
+	int pi;
 	struct flowi fl;
 	u8 fl_dir = policy_to_flow_dir(dir);
-	u32 sk_sid;
+	int xerr_idx = -1;
 
 	if (xfrm_decode_session(skb, &fl, family) < 0)
 		return 0;
 	nf_nat_decode_session(skb, &fl, family);
 
-	sk_sid = security_sk_sid(sk, &fl, fl_dir);
-
 	/* First, check used SA against their selectors. */
 	if (skb->sp) {
 		int i;
@@ -1079,46 +1572,90 @@ int __xfrm_policy_check(struct sock *sk, int dir, struct sk_buff *skb,
 
 	pol = NULL;
 	if (sk && sk->sk_policy[dir])
-		pol = xfrm_sk_policy_lookup(sk, dir, &fl, sk_sid);
+		pol = xfrm_sk_policy_lookup(sk, dir, &fl);
 
 	if (!pol)
-		pol = flow_cache_lookup(&fl, sk_sid, family, fl_dir,
+		pol = flow_cache_lookup(&fl, family, fl_dir,
 					xfrm_policy_lookup);
 
-	if (!pol)
-		return !skb->sp || !secpath_has_tunnel(skb->sp, 0);
+	if (!pol) {
+		if (skb->sp && secpath_has_nontransport(skb->sp, 0, &xerr_idx)) {
+			xfrm_secpath_reject(xerr_idx, skb, &fl);
+			return 0;
+		}
+		return 1;
+	}
 
 	pol->curlft.use_time = (unsigned long)xtime.tv_sec;
 
+	pols[0] = pol;
+	npols ++;
+#ifdef CONFIG_XFRM_SUB_POLICY
+	if (pols[0]->type != XFRM_POLICY_TYPE_MAIN) {
+		pols[1] = xfrm_policy_lookup_bytype(XFRM_POLICY_TYPE_MAIN,
+						    &fl, family,
+						    XFRM_POLICY_IN);
+		if (pols[1]) {
+			pols[1]->curlft.use_time = (unsigned long)xtime.tv_sec;
+			npols ++;
+		}
+	}
+#endif
+
 	if (pol->action == XFRM_POLICY_ALLOW) {
 		struct sec_path *sp;
 		static struct sec_path dummy;
+		struct xfrm_tmpl *tp[XFRM_MAX_DEPTH];
+		struct xfrm_tmpl *stp[XFRM_MAX_DEPTH];
+		struct xfrm_tmpl **tpp = tp;
+		int ti = 0;
 		int i, k;
 
 		if ((sp = skb->sp) == NULL)
 			sp = &dummy;
 
+		for (pi = 0; pi < npols; pi++) {
+			if (pols[pi] != pol &&
+			    pols[pi]->action != XFRM_POLICY_ALLOW)
+				goto reject;
+			if (ti + pols[pi]->xfrm_nr >= XFRM_MAX_DEPTH)
+				goto reject_error;
+			for (i = 0; i < pols[pi]->xfrm_nr; i++)
+				tpp[ti++] = &pols[pi]->xfrm_vec[i];
+		}
+		xfrm_nr = ti;
+		if (npols > 1) {
+			xfrm_tmpl_sort(stp, tpp, xfrm_nr, family);
+			tpp = stp;
+		}
+
 		/* For each tunnel xfrm, find the first matching tmpl.
 		 * For each tmpl before that, find corresponding xfrm.
 		 * Order is _important_. Later we will implement
 		 * some barriers, but at the moment barriers
 		 * are implied between each two transformations.
 		 */
-		for (i = pol->xfrm_nr-1, k = 0; i >= 0; i--) {
-			k = xfrm_policy_ok(pol->xfrm_vec+i, sp, k, family);
-			if (k < 0)
+		for (i = xfrm_nr-1, k = 0; i >= 0; i--) {
+			k = xfrm_policy_ok(tpp[i], sp, k, family);
+			if (k < 0) {
+				if (k < -1)
+					/* "-2 - errored_index" returned */
+					xerr_idx = -(2+k);
 				goto reject;
+			}
 		}
 
-		if (secpath_has_tunnel(sp, k))
+		if (secpath_has_nontransport(sp, k, &xerr_idx))
 			goto reject;
 
-		xfrm_pol_put(pol);
+		xfrm_pols_put(pols, npols);
 		return 1;
 	}
 
 reject:
-	xfrm_pol_put(pol);
+	xfrm_secpath_reject(xerr_idx, skb, &fl);
+reject_error:
+	xfrm_pols_put(pols, npols);
 	return 0;
 }
 EXPORT_SYMBOL(__xfrm_policy_check);
@@ -1166,7 +1703,7 @@ static struct dst_entry *xfrm_dst_check(struct dst_entry *dst, u32 cookie)
 
 static int stale_bundle(struct dst_entry *dst)
 {
-	return !xfrm_bundle_ok((struct xfrm_dst *)dst, NULL, AF_UNSPEC);
+	return !xfrm_bundle_ok((struct xfrm_dst *)dst, NULL, AF_UNSPEC, 0);
 }
 
 void xfrm_dst_ifdown(struct dst_entry *dst, struct net_device *dev)
@@ -1196,33 +1733,50 @@ static struct dst_entry *xfrm_negative_advice(struct dst_entry *dst)
 	return dst;
 }
 
+static void prune_one_bundle(struct xfrm_policy *pol, int (*func)(struct dst_entry *), struct dst_entry **gc_list_p)
+{
+	struct dst_entry *dst, **dstp;
+
+	write_lock(&pol->lock);
+	dstp = &pol->bundles;
+	while ((dst=*dstp) != NULL) {
+		if (func(dst)) {
+			*dstp = dst->next;
+			dst->next = *gc_list_p;
+			*gc_list_p = dst;
+		} else {
+			dstp = &dst->next;
+		}
+	}
+	write_unlock(&pol->lock);
+}
+
 static void xfrm_prune_bundles(int (*func)(struct dst_entry *))
 {
-	int i;
-	struct xfrm_policy *pol;
-	struct dst_entry *dst, **dstp, *gc_list = NULL;
+	struct dst_entry *gc_list = NULL;
+	int dir;
 
 	read_lock_bh(&xfrm_policy_lock);
-	for (i=0; i<2*XFRM_POLICY_MAX; i++) {
-		for (pol = xfrm_policy_list[i]; pol; pol = pol->next) {
-			write_lock(&pol->lock);
-			dstp = &pol->bundles;
-			while ((dst=*dstp) != NULL) {
-				if (func(dst)) {
-					*dstp = dst->next;
-					dst->next = gc_list;
-					gc_list = dst;
-				} else {
-					dstp = &dst->next;
-				}
-			}
-			write_unlock(&pol->lock);
+	for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
+		struct xfrm_policy *pol;
+		struct hlist_node *entry;
+		struct hlist_head *table;
+		int i;
+
+		hlist_for_each_entry(pol, entry,
+				     &xfrm_policy_inexact[dir], bydst)
+			prune_one_bundle(pol, func, &gc_list);
+
+		table = xfrm_policy_bydst[dir].table;
+		for (i = xfrm_policy_bydst[dir].hmask; i >= 0; i--) {
+			hlist_for_each_entry(pol, entry, table + i, bydst)
+				prune_one_bundle(pol, func, &gc_list);
 		}
 	}
 	read_unlock_bh(&xfrm_policy_lock);
 
 	while (gc_list) {
-		dst = gc_list;
+		struct dst_entry *dst = gc_list;
 		gc_list = dst->next;
 		dst_free(dst);
 	}
@@ -1238,22 +1792,12 @@ static void __xfrm_garbage_collect(void)
 	xfrm_prune_bundles(unused_bundle);
 }
 
-int xfrm_flush_bundles(void)
+static int xfrm_flush_bundles(void)
 {
 	xfrm_prune_bundles(stale_bundle);
 	return 0;
 }
 
-static int always_true(struct dst_entry *dst)
-{
-	return 1;
-}
-
-void xfrm_flush_all_bundles(void)
-{
-	xfrm_prune_bundles(always_true);
-}
-
 void xfrm_init_pmtu(struct dst_entry *dst)
 {
 	do {
@@ -1281,7 +1825,7 @@ EXPORT_SYMBOL(xfrm_init_pmtu);
  * still valid.
  */
 
-int xfrm_bundle_ok(struct xfrm_dst *first, struct flowi *fl, int family)
+int xfrm_bundle_ok(struct xfrm_dst *first, struct flowi *fl, int family, int strict)
 {
 	struct dst_entry *dst = &first->u.dst;
 	struct xfrm_dst *last;
@@ -1298,8 +1842,16 @@ int xfrm_bundle_ok(struct xfrm_dst *first, struct flowi *fl, int family)
 
 		if (fl && !xfrm_selector_match(&dst->xfrm->sel, fl, family))
 			return 0;
+		if (fl && !security_xfrm_flow_state_match(fl, dst->xfrm))
+			return 0;
 		if (dst->xfrm->km.state != XFRM_STATE_VALID)
 			return 0;
+		if (xdst->genid != dst->xfrm->genid)
+			return 0;
+
+		if (strict && fl && dst->xfrm->props.mode != XFRM_MODE_TUNNEL &&
+		    !xfrm_state_addr_flow_check(dst->xfrm, fl, family))
+			return 0;
 
 		mtu = dst_mtu(dst->child);
 		if (xdst->child_mtu_cached != mtu) {
@@ -1448,12 +2000,33 @@ static struct notifier_block xfrm_dev_notifier = {
 
 static void __init xfrm_policy_init(void)
 {
+	unsigned int hmask, sz;
+	int dir;
+
 	xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
 					   sizeof(struct xfrm_dst),
-					   0, SLAB_HWCACHE_ALIGN,
+					   0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
 					   NULL, NULL);
-	if (!xfrm_dst_cache)
-		panic("XFRM: failed to allocate xfrm_dst_cache\n");
+
+	hmask = 8 - 1;
+	sz = (hmask+1) * sizeof(struct hlist_head);
+
+	xfrm_policy_byidx = xfrm_hash_alloc(sz);
+	xfrm_idx_hmask = hmask;
+	if (!xfrm_policy_byidx)
+		panic("XFRM: failed to allocate byidx hash\n");
+
+	for (dir = 0; dir < XFRM_POLICY_MAX * 2; dir++) {
+		struct xfrm_policy_hash *htab;
+
+		INIT_HLIST_HEAD(&xfrm_policy_inexact[dir]);
+
+		htab = &xfrm_policy_bydst[dir];
+		htab->table = xfrm_hash_alloc(sz);
+		htab->hmask = hmask;
+		if (!htab->table)
+			panic("XFRM: failed to allocate bydst hash\n");
+	}
 
 	INIT_WORK(&xfrm_policy_gc_work, xfrm_policy_gc_task, NULL);
 	register_netdevice_notifier(&xfrm_dev_notifier);