xfrm: policy: store inexact policies in an rhashtable

Switch packet-path lookups for inexact policies to rhashtable.

In this initial version, we now no longer need to search policies with
non-matching address family and type.

Next patch will add the if_id as well so lookups from the xfrm interface
driver only need to search inexact policies for that device.

Future patches will augment the hlist in each rhash bucket with a tree
and pre-sort policies according to daddr/prefix.

A single rhashtable is used. In order to avoid a full rhashtable walk on
netns exit, the bins get placed on a pernet list, i.e. we add almost no
cost for network namespaces that had no xfrm policies.

The inexact lists are kept in place, and policies are added to both the
per-rhash-inexact list and a pernet one.

The latter is needed for the control plane to handle migrate -- these
requests do not consider the if_id, so if we'd remove the inexact_list
now we would have to search all hash buckets and then figure
out which matching policy candidate is the most recent one -- this appears
a bit harder than just keeping the 'old' inexact list for this purpose.

Signed-off-by: Florian Westphal <[e-mail]>
Acked-by: David S. Miller <[e-mail]>
Signed-off-by: Steffen Klassert <[e-mail]>

include/net/netns/xfrm.h

···
 #include <linux/list.h>
 #include <linux/wait.h>
 #include <linux/workqueue.h>
+#include <linux/rhashtable-types.h>
 #include <linux/xfrm.h>
 #include <net/dst_ops.h>
 
···
 	unsigned int		policy_count[XFRM_POLICY_MAX * 2];
 	struct work_struct	policy_hash_work;
 	struct xfrm_policy_hthresh policy_hthresh;
+	struct list_head	inexact_bins;
 
 
 	struct sock		*nlsk;

include/net/xfrm.h

···
 	u16			family;
 	struct xfrm_sec_ctx	*security;
 	struct xfrm_tmpl       	xfrm_vec[XFRM_MAX_DEPTH];
+	struct hlist_node	bydst_inexact_list;
 	struct rcu_head		rcu;
 };
 

net/xfrm/xfrm_policy.c

···
 #include <linux/cache.h>
 #include <linux/cpu.h>
 #include <linux/audit.h>
+#include <linux/rhashtable.h>
 #include <net/dst.h>
 #include <net/flow.h>
 #include <net/xfrm.h>
···
 	u8 flags;
 };
 
+struct xfrm_pol_inexact_key {
+	possible_net_t net;
+	u16 family;
+	u8 dir, type;
+};
+
+struct xfrm_pol_inexact_bin {
+	struct xfrm_pol_inexact_key k;
+	struct rhash_head head;
+	struct hlist_head hhead;
+
+	/* slow path below */
+	struct list_head inexact_bins;
+	struct rcu_head rcu;
+};
+
 static DEFINE_SPINLOCK(xfrm_if_cb_lock);
 static struct xfrm_if_cb const __rcu *xfrm_if_cb __read_mostly;
 
···
 static struct kmem_cache *xfrm_dst_cache __ro_after_init;
 static __read_mostly seqcount_t xfrm_policy_hash_generation;
 
+static struct rhashtable xfrm_policy_inexact_table;
+static const struct rhashtable_params xfrm_pol_inexact_params;
+
 static void xfrm_init_pmtu(struct xfrm_dst **bundle, int nr);
 static int stale_bundle(struct dst_entry *dst);
 static int xfrm_bundle_ok(struct xfrm_dst *xdst);
···
 static void __xfrm_policy_link(struct xfrm_policy *pol, int dir);
 static struct xfrm_policy *__xfrm_policy_unlink(struct xfrm_policy *pol,
 						int dir);
+
+static struct xfrm_pol_inexact_bin *
+xfrm_policy_inexact_lookup(struct net *net, u8 type, u16 family, u8 dir);
+
+static struct xfrm_pol_inexact_bin *
+xfrm_policy_inexact_lookup_rcu(struct net *net,
+			       u8 type, u16 family, u8 dir);
+static struct xfrm_policy *
+xfrm_policy_insert_list(struct hlist_head *chain, struct xfrm_policy *policy,
+			bool excl);
+static void xfrm_policy_insert_inexact_list(struct hlist_head *chain,
+					    struct xfrm_policy *policy);
 
 static inline bool xfrm_pol_hold_rcu(struct xfrm_policy *policy)
 {
···
 	if (policy) {
 		write_pnet(&policy->xp_net, net);
 		INIT_LIST_HEAD(&policy->walk.all);
+		INIT_HLIST_NODE(&policy->bydst_inexact_list);
 		INIT_HLIST_NODE(&policy->bydst);
 		INIT_HLIST_NODE(&policy->byidx);
 		rwlock_init(&policy->lock);
···
 	mutex_unlock(&hash_resize_mutex);
 }
 
+static void xfrm_hash_reset_inexact_table(struct net *net)
+{
+	struct xfrm_pol_inexact_bin *b;
+
+	lockdep_assert_held(&net->xfrm.xfrm_policy_lock);
+
+	list_for_each_entry(b, &net->xfrm.inexact_bins, inexact_bins)
+		INIT_HLIST_HEAD(&b->hhead);
+}
+
+/* Make sure *pol can be inserted into fastbin.
+ * Useful to check that later insert requests will be sucessful
+ * (provided xfrm_policy_lock is held throughout).
+ */
+static struct xfrm_pol_inexact_bin *
+xfrm_policy_inexact_alloc_bin(const struct xfrm_policy *pol, u8 dir)
+{
+	struct xfrm_pol_inexact_bin *bin, *prev;
+	struct xfrm_pol_inexact_key k = {
+		.family = pol->family,
+		.type = pol->type,
+		.dir = dir,
+	};
+	struct net *net = xp_net(pol);
+
+	lockdep_assert_held(&net->xfrm.xfrm_policy_lock);
+
+	write_pnet(&k.net, net);
+	bin = rhashtable_lookup_fast(&xfrm_policy_inexact_table, &k,
+				     xfrm_pol_inexact_params);
+	if (bin)
+		return bin;
+
+	bin = kzalloc(sizeof(*bin), GFP_ATOMIC);
+	if (!bin)
+		return NULL;
+
+	bin->k = k;
+	INIT_HLIST_HEAD(&bin->hhead);
+
+	prev = rhashtable_lookup_get_insert_key(&xfrm_policy_inexact_table,
+						&bin->k, &bin->head,
+						xfrm_pol_inexact_params);
+	if (!prev) {
+		list_add(&bin->inexact_bins, &net->xfrm.inexact_bins);
+		return bin;
+	}
+
+	kfree(bin);
+
+	return IS_ERR(prev) ? NULL : prev;
+}
+
+static void xfrm_policy_inexact_delete_bin(struct net *net,
+					   struct xfrm_pol_inexact_bin *b)
+{
+	lockdep_assert_held(&net->xfrm.xfrm_policy_lock);
+
+	if (!hlist_empty(&b->hhead))
+		return;
+
+	if (rhashtable_remove_fast(&xfrm_policy_inexact_table, &b->head,
+				   xfrm_pol_inexact_params) == 0) {
+		list_del(&b->inexact_bins);
+		kfree_rcu(b, rcu);
+	}
+}
+
+static void __xfrm_policy_inexact_flush(struct net *net)
+{
+	struct xfrm_pol_inexact_bin *bin;
+
+	lockdep_assert_held(&net->xfrm.xfrm_policy_lock);
+
+	list_for_each_entry(bin, &net->xfrm.inexact_bins, inexact_bins)
+		xfrm_policy_inexact_delete_bin(net, bin);
+}
+
+static struct xfrm_policy *
+xfrm_policy_inexact_insert(struct xfrm_policy *policy, u8 dir, int excl)
+{
+	struct xfrm_pol_inexact_bin *bin;
+	struct xfrm_policy *delpol;
+	struct hlist_head *chain;
+	struct net *net;
+
+	bin = xfrm_policy_inexact_alloc_bin(policy, dir);
+	if (!bin)
+		return ERR_PTR(-ENOMEM);
+
+	delpol = xfrm_policy_insert_list(&bin->hhead, policy, excl);
+	if (delpol && excl)
+		return ERR_PTR(-EEXIST);
+
+	net = xp_net(policy);
+	chain = &net->xfrm.policy_inexact[dir];
+	xfrm_policy_insert_inexact_list(chain, policy);
+
+	return delpol;
+}
+
 static void xfrm_hash_rebuild(struct work_struct *work)
 {
 	struct net *net = container_of(work, struct net,
···
 
 	spin_lock_bh(&net->xfrm.xfrm_policy_lock);
 
+	/* make sure that we can insert the indirect policies again before
+	 * we start with destructive action.
+	 */
+	list_for_each_entry(policy, &net->xfrm.policy_all, walk.all) {
+		u8 dbits, sbits;
+
+		dir = xfrm_policy_id2dir(policy->index);
+		if (policy->walk.dead || dir >= XFRM_POLICY_MAX)
+			continue;
+
+		if ((dir & XFRM_POLICY_MASK) == XFRM_POLICY_OUT) {
+			if (policy->family == AF_INET) {
+				dbits = rbits4;
+				sbits = lbits4;
+			} else {
+				dbits = rbits6;
+				sbits = lbits6;
+			}
+		} else {
+			if (policy->family == AF_INET) {
+				dbits = lbits4;
+				sbits = rbits4;
+			} else {
+				dbits = lbits6;
+				sbits = rbits6;
+			}
+		}
+
+		if (policy->selector.prefixlen_d < dbits ||
+		    policy->selector.prefixlen_s < sbits)
+			continue;
+
+		if (!xfrm_policy_inexact_alloc_bin(policy, dir))
+			goto out_unlock;
+	}
+
 	/* reset the bydst and inexact table in all directions */
+	xfrm_hash_reset_inexact_table(net);
+
 	for (dir = 0; dir < XFRM_POLICY_MAX; dir++) {
 		INIT_HLIST_HEAD(&net->xfrm.policy_inexact[dir]);
 		hmask = net->xfrm.policy_bydst[dir].hmask;
···
 		chain = policy_hash_bysel(net, &policy->selector,
 					  policy->family,
 					  xfrm_policy_id2dir(policy->index));
-		if (!chain)
-			chain = &net->xfrm.policy_inexact[dir];
+		if (!chain) {
+			void *p = xfrm_policy_inexact_insert(policy, dir, 0);
+
+			WARN_ONCE(IS_ERR(p), "reinsert: %ld\n", PTR_ERR(p));
+			continue;
+		}
+
 		hlist_for_each_entry(pol, chain, bydst) {
 			if (policy->priority >= pol->priority)
 				newpos = &pol->bydst;
···
 			hlist_add_head_rcu(&policy->bydst, chain);
 	}
 
+out_unlock:
 	spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
 
 	mutex_unlock(&hash_resize_mutex);
···
 	return false;
 }
 
+static u32 xfrm_pol_bin_key(const void *data, u32 len, u32 seed)
+{
+	const struct xfrm_pol_inexact_key *k = data;
+	u32 a = k->type << 24 | k->dir << 16 | k->family;
+
+	return jhash_2words(a, net_hash_mix(read_pnet(&k->net)), seed);
+}
+
+static u32 xfrm_pol_bin_obj(const void *data, u32 len, u32 seed)
+{
+	const struct xfrm_pol_inexact_bin *b = data;
+
+	return xfrm_pol_bin_key(&b->k, 0, seed);
+}
+
+static int xfrm_pol_bin_cmp(struct rhashtable_compare_arg *arg,
+			    const void *ptr)
+{
+	const struct xfrm_pol_inexact_key *key = arg->key;
+	const struct xfrm_pol_inexact_bin *b = ptr;
+	int ret;
+
+	if (!net_eq(read_pnet(&b->k.net), read_pnet(&key->net)))
+		return -1;
+
+	ret = b->k.dir ^ key->dir;
+	if (ret)
+		return ret;
+
+	ret = b->k.type ^ key->type;
+	if (ret)
+		return ret;
+
+	ret = b->k.family ^ key->family;
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static const struct rhashtable_params xfrm_pol_inexact_params = {
+	.head_offset		= offsetof(struct xfrm_pol_inexact_bin, head),
+	.hashfn			= xfrm_pol_bin_key,
+	.obj_hashfn		= xfrm_pol_bin_obj,
+	.obj_cmpfn		= xfrm_pol_bin_cmp,
+	.automatic_shrinking	= true,
+};
+
+static void xfrm_policy_insert_inexact_list(struct hlist_head *chain,
+					    struct xfrm_policy *policy)
+{
+	struct xfrm_policy *pol, *delpol = NULL;
+	struct hlist_node *newpos = NULL;
+
+	hlist_for_each_entry(pol, chain, bydst_inexact_list) {
+		if (pol->type == policy->type &&
+		    pol->if_id == policy->if_id &&
+		    !selector_cmp(&pol->selector, &policy->selector) &&
+		    xfrm_policy_mark_match(policy, pol) &&
+		    xfrm_sec_ctx_match(pol->security, policy->security) &&
+		    !WARN_ON(delpol)) {
+			delpol = pol;
+			if (policy->priority > pol->priority)
+				continue;
+		} else if (policy->priority >= pol->priority) {
+			newpos = &pol->bydst_inexact_list;
+			continue;
+		}
+		if (delpol)
+			break;
+	}
+
+	if (newpos)
+		hlist_add_behind_rcu(&policy->bydst_inexact_list, newpos);
+	else
+		hlist_add_head_rcu(&policy->bydst_inexact_list, chain);
+}
+
 static struct xfrm_policy *xfrm_policy_insert_list(struct hlist_head *chain,
 						   struct xfrm_policy *policy,
 						   bool excl)
···
 		if (delpol)
 			break;
 	}
+
 	if (newpos)
 		hlist_add_behind_rcu(&policy->bydst, &newpos->bydst);
 	else
···
 
 	spin_lock_bh(&net->xfrm.xfrm_policy_lock);
 	chain = policy_hash_bysel(net, &policy->selector, policy->family, dir);
-	if (chain) {
+	if (chain)
 		delpol = xfrm_policy_insert_list(chain, policy, excl);
-	} else {
-		chain = &net->xfrm.policy_inexact[dir];
-		delpol = xfrm_policy_insert_list(chain, policy, excl);
-	}
+	else
+		delpol = xfrm_policy_inexact_insert(policy, dir, excl);
 
 	if (IS_ERR(delpol)) {
 		spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
···
 					  struct xfrm_sec_ctx *ctx, int delete,
 					  int *err)
 {
-	struct xfrm_policy *pol, *ret;
+	struct xfrm_pol_inexact_bin *bin = NULL;
+	struct xfrm_policy *pol, *ret = NULL;
 	struct hlist_head *chain;
 
 	*err = 0;
 	spin_lock_bh(&net->xfrm.xfrm_policy_lock);
 	chain = policy_hash_bysel(net, sel, sel->family, dir);
-	if (!chain)
-		chain = &net->xfrm.policy_inexact[dir];
+	if (!chain) {
+		bin = xfrm_policy_inexact_lookup(net, type,
+						 sel->family, dir);
+		if (!bin) {
+			spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
+			return NULL;
+		}
+
+		chain = &bin->hhead;
+	}
+
 	ret = NULL;
 	hlist_for_each_entry(pol, chain, bydst) {
 		if (pol->type == type &&
···
 					return pol;
 				}
 				__xfrm_policy_unlink(pol, dir);
+				xfrm_policy_inexact_delete_bin(net, bin);
 			}
 			ret = pol;
 			break;
···
 		spin_lock_bh(&net->xfrm.xfrm_policy_lock);
 		goto again;
 	}
-	if (!cnt)
+	if (cnt)
+		__xfrm_policy_inexact_flush(net);
+	else
 		err = -ESRCH;
 out:
 	spin_unlock_bh(&net->xfrm.xfrm_policy_lock);
···
 	if (match)
 		ret = security_xfrm_policy_lookup(pol->security, fl->flowi_secid,
 						  dir);
-
 	return ret;
+}
+
+static struct xfrm_pol_inexact_bin *
+xfrm_policy_inexact_lookup_rcu(struct net *net, u8 type, u16 family, u8 dir)
+{
+	struct xfrm_pol_inexact_key k = {
+		.family = family,
+		.type = type,
+		.dir = dir,
+	};
+
+	write_pnet(&k.net, net);
+
+	return rhashtable_lookup(&xfrm_policy_inexact_table, &k,
+				 xfrm_pol_inexact_params);
+}
+
+static struct xfrm_pol_inexact_bin *
+xfrm_policy_inexact_lookup(struct net *net, u8 type, u16 family, u8 dir)
+{
+	struct xfrm_pol_inexact_bin *bin;
+
+	lockdep_assert_held(&net->xfrm.xfrm_policy_lock);
+
+	rcu_read_lock();
+	bin = xfrm_policy_inexact_lookup_rcu(net, type, family, dir);
+	rcu_read_unlock();
+
+	return bin;
 }
 
 static struct xfrm_policy *xfrm_policy_lookup_bytype(struct net *net, u8 type,
···
 						     u16 family, u8 dir,
 						     u32 if_id)
 {
-	int err;
-	struct xfrm_policy *pol, *ret;
 	const xfrm_address_t *daddr, *saddr;
+	struct xfrm_pol_inexact_bin *bin;
+	struct xfrm_policy *pol, *ret;
 	struct hlist_head *chain;
 	unsigned int sequence;
 	u32 priority;
+	int err;
 
 	daddr = xfrm_flowi_daddr(fl, family);
 	saddr = xfrm_flowi_saddr(fl, family);
···
 			break;
 		}
 	}
-	chain = &net->xfrm.policy_inexact[dir];
+	bin = xfrm_policy_inexact_lookup_rcu(net, type, family, dir);
+	if (!bin)
+		goto skip_inexact;
+	chain = &bin->hhead;
 	hlist_for_each_entry_rcu(pol, chain, bydst) {
 		if ((pol->priority >= priority) && ret)
 			break;
···
 		}
 	}
 
+skip_inexact:
 	if (read_seqcount_retry(&xfrm_policy_hash_generation, sequence))
 		goto retry;
 
···
 	/* Socket policies are not hashed. */
 	if (!hlist_unhashed(&pol->bydst)) {
 		hlist_del_rcu(&pol->bydst);
+		hlist_del_init(&pol->bydst_inexact_list);
 		hlist_del(&pol->byidx);
 	}
 
···
 static int __net_init xfrm_policy_init(struct net *net)
 {
 	unsigned int hmask, sz;
-	int dir;
+	int dir, err;
 
-	if (net_eq(net, &init_net))
+	if (net_eq(net, &init_net)) {
 		xfrm_dst_cache = kmem_cache_create("xfrm_dst_cache",
 					   sizeof(struct xfrm_dst),
 					   0, SLAB_HWCACHE_ALIGN|SLAB_PANIC,
 					   NULL);
+		err = rhashtable_init(&xfrm_policy_inexact_table,
+				      &xfrm_pol_inexact_params);
+		BUG_ON(err);
+	}
 
 	hmask = 8 - 1;
 	sz = (hmask+1) * sizeof(struct hlist_head);
···
 	seqlock_init(&net->xfrm.policy_hthresh.lock);
 
 	INIT_LIST_HEAD(&net->xfrm.policy_all);
+	INIT_LIST_HEAD(&net->xfrm.inexact_bins);
 	INIT_WORK(&net->xfrm.policy_hash_work, xfrm_hash_resize);
 	INIT_WORK(&net->xfrm.policy_hthresh.work, xfrm_hash_rebuild);
 	return 0;
···
 
 static void xfrm_policy_fini(struct net *net)
 {
+	struct xfrm_pol_inexact_bin *bin, *tmp;
 	unsigned int sz;
 	int dir;
 
···
 	sz = (net->xfrm.policy_idx_hmask + 1) * sizeof(struct hlist_head);
 	WARN_ON(!hlist_empty(net->xfrm.policy_byidx));
 	xfrm_hash_free(net->xfrm.policy_byidx, sz);
+
+	list_for_each_entry_safe(bin, tmp, &net->xfrm.inexact_bins,
+				 inexact_bins) {
+		WARN_ON(!hlist_empty(&bin->hhead));
+		xfrm_policy_inexact_delete_bin(net, bin);
+	}
 }
 
 static int __net_init xfrm_net_init(struct net *net)
···
 		}
 	}
 	chain = &net->xfrm.policy_inexact[dir];
-	hlist_for_each_entry(pol, chain, bydst) {
+	hlist_for_each_entry(pol, chain, bydst_inexact_list) {
 		if ((pol->priority >= priority) && ret)
 			break;