Merge branch 'rhashtable-dups'

Herbert Xu says:

====================
rhashtable: rhashtable with duplicate objects

v3 fixes a bug in the remove path that causes the element count
to decrease when it shouldn't, leading to a gigantic hash table
when it underflows.

v2 contains a reworked insertion slowpath to ensure that the
spinlock for the table we're inserting into is taken.

This series contains two patches.  The first adds the rhlist
interface and the second converts mac80211 to use it.  If this
works out I'll then proceed to convert the other insecure_elasticity
users over to this.

I've tested the rhlist code with test_rhashtable but I haven't
tested the mac80211 conversion.  So please give it a go and see
if it still works.
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

lib/rhashtable.c

@@ -378,21 +378,7 @@ static void rht_deferred_worker(struct work_struct *work)
 		schedule_work(&ht->run_work);
 }
 
-static bool rhashtable_check_elasticity(struct rhashtable *ht,
-					struct bucket_table *tbl,
-					unsigned int hash)
-{
-	unsigned int elasticity = ht->elasticity;
-	struct rhash_head *head;
-
-	rht_for_each(head, tbl, hash)
-		if (!--elasticity)
-			return true;
-
-	return false;
-}
-
-int rhashtable_insert_rehash(struct rhashtable *ht,
+static int rhashtable_insert_rehash(struct rhashtable *ht,
 				    struct bucket_table *tbl)
 {
 	struct bucket_table *old_tbl;
@@ -439,57 +425,165 @@ int rhashtable_insert_rehash(struct rhashtable *ht,
 
 	return err;
 }
-EXPORT_SYMBOL_GPL(rhashtable_insert_rehash);
 
-struct bucket_table *rhashtable_insert_slow(struct rhashtable *ht,
-					    const void *key,
-					    struct rhash_head *obj,
-					    struct bucket_table *tbl,
-					    void **data)
+static void *rhashtable_lookup_one(struct rhashtable *ht,
+				   struct bucket_table *tbl, unsigned int hash,
+				   const void *key, struct rhash_head *obj)
 {
+	struct rhashtable_compare_arg arg = {
+		.ht = ht,
+		.key = key,
+	};
+	struct rhash_head __rcu **pprev;
 	struct rhash_head *head;
-	unsigned int hash;
-	int err;
+	int elasticity;
 
-	tbl = rhashtable_last_table(ht, tbl);
-	hash = head_hashfn(ht, tbl, obj);
-	spin_lock_nested(rht_bucket_lock(tbl, hash), SINGLE_DEPTH_NESTING);
-
-	err = -EEXIST;
-	if (key) {
-		*data = rhashtable_lookup_fast(ht, key, ht->p);
-		if (*data)
-			goto exit;
+	elasticity = ht->elasticity;
+	pprev = &tbl->buckets[hash];
+	rht_for_each(head, tbl, hash) {
+		struct rhlist_head *list;
+		struct rhlist_head *plist;
+
+		elasticity--;
+		if (!key ||
+		    (ht->p.obj_cmpfn ?
+		     ht->p.obj_cmpfn(&arg, rht_obj(ht, head)) :
+		     rhashtable_compare(&arg, rht_obj(ht, head))))
+			continue;
+
+		if (!ht->rhlist)
+			return rht_obj(ht, head);
+
+		list = container_of(obj, struct rhlist_head, rhead);
+		plist = container_of(head, struct rhlist_head, rhead);
+
+		RCU_INIT_POINTER(list->next, plist);
+		head = rht_dereference_bucket(head->next, tbl, hash);
+		RCU_INIT_POINTER(list->rhead.next, head);
+		rcu_assign_pointer(*pprev, obj);
+
+		return NULL;
 	}
 
-	err = -E2BIG;
-	if (unlikely(rht_grow_above_max(ht, tbl)))
-		goto exit;
+	if (elasticity <= 0)
+		return ERR_PTR(-EAGAIN);
 
-	err = -EAGAIN;
-	if (rhashtable_check_elasticity(ht, tbl, hash) ||
-	    rht_grow_above_100(ht, tbl))
-		goto exit;
+	return ERR_PTR(-ENOENT);
+}
 
-	err = 0;
+static struct bucket_table *rhashtable_insert_one(struct rhashtable *ht,
+						  struct bucket_table *tbl,
+						  unsigned int hash,
+						  struct rhash_head *obj,
+						  void *data)
+{
+	struct bucket_table *new_tbl;
+	struct rhash_head *head;
+
+	if (!IS_ERR_OR_NULL(data))
+		return ERR_PTR(-EEXIST);
+
+	if (PTR_ERR(data) != -EAGAIN && PTR_ERR(data) != -ENOENT)
+		return ERR_CAST(data);
+
+	new_tbl = rcu_dereference(tbl->future_tbl);
+	if (new_tbl)
+		return new_tbl;
+
+	if (PTR_ERR(data) != -ENOENT)
+		return ERR_CAST(data);
+
+	if (unlikely(rht_grow_above_max(ht, tbl)))
+		return ERR_PTR(-E2BIG);
+
+	if (unlikely(rht_grow_above_100(ht, tbl)))
+		return ERR_PTR(-EAGAIN);
 
 	head = rht_dereference_bucket(tbl->buckets[hash], tbl, hash);
 
 	RCU_INIT_POINTER(obj->next, head);
+	if (ht->rhlist) {
+		struct rhlist_head *list;
+
+		list = container_of(obj, struct rhlist_head, rhead);
+		RCU_INIT_POINTER(list->next, NULL);
+	}
 
 	rcu_assign_pointer(tbl->buckets[hash], obj);
 
 	atomic_inc(&ht->nelems);
+	if (rht_grow_above_75(ht, tbl))
+		schedule_work(&ht->run_work);
+
+	return NULL;
+}
+
+static void *rhashtable_try_insert(struct rhashtable *ht, const void *key,
+				   struct rhash_head *obj)
+{
+	struct bucket_table *new_tbl;
+	struct bucket_table *tbl;
+	unsigned int hash;
+	spinlock_t *lock;
+	void *data;
+
+	tbl = rcu_dereference(ht->tbl);
+
+	/* All insertions must grab the oldest table containing
+	 * the hashed bucket that is yet to be rehashed.
+	 */
+	for (;;) {
+		hash = rht_head_hashfn(ht, tbl, obj, ht->p);
+		lock = rht_bucket_lock(tbl, hash);
+		spin_lock_bh(lock);
+
+		if (tbl->rehash <= hash)
+			break;
+
+		spin_unlock_bh(lock);
+		tbl = rcu_dereference(tbl->future_tbl);
+	}
+
+	data = rhashtable_lookup_one(ht, tbl, hash, key, obj);
+	new_tbl = rhashtable_insert_one(ht, tbl, hash, obj, data);
+	if (PTR_ERR(new_tbl) != -EEXIST)
+		data = ERR_CAST(new_tbl);
+
+	while (!IS_ERR_OR_NULL(new_tbl)) {
+		tbl = new_tbl;
+		hash = rht_head_hashfn(ht, tbl, obj, ht->p);
+		spin_lock_nested(rht_bucket_lock(tbl, hash),
+				 SINGLE_DEPTH_NESTING);
+
+		data = rhashtable_lookup_one(ht, tbl, hash, key, obj);
+		new_tbl = rhashtable_insert_one(ht, tbl, hash, obj, data);
+		if (PTR_ERR(new_tbl) != -EEXIST)
+			data = ERR_CAST(new_tbl);
 
-exit:
 		spin_unlock(rht_bucket_lock(tbl, hash));
+	}
 
-	if (err == 0)
-		return NULL;
-	else if (err == -EAGAIN)
-		return tbl;
-	else
-		return ERR_PTR(err);
+	spin_unlock_bh(lock);
+
+	if (PTR_ERR(data) == -EAGAIN)
+		data = ERR_PTR(rhashtable_insert_rehash(ht, tbl) ?:
+			       -EAGAIN);
+
+	return data;
+}
+
+void *rhashtable_insert_slow(struct rhashtable *ht, const void *key,
+			     struct rhash_head *obj)
+{
+	void *data;
+
+	do {
+		rcu_read_lock();
+		data = rhashtable_try_insert(ht, key, obj);
+		rcu_read_unlock();
+	} while (PTR_ERR(data) == -EAGAIN);
+
+	return data;
 }
 EXPORT_SYMBOL_GPL(rhashtable_insert_slow);
 
@@ -593,11 +687,16 @@ EXPORT_SYMBOL_GPL(rhashtable_walk_start);
 void *rhashtable_walk_next(struct rhashtable_iter *iter)
 {
 	struct bucket_table *tbl = iter->walker.tbl;
+	struct rhlist_head *list = iter->list;
 	struct rhashtable *ht = iter->ht;
 	struct rhash_head *p = iter->p;
+	bool rhlist = ht->rhlist;
 
 	if (p) {
-		p = rht_dereference_bucket_rcu(p->next, tbl, iter->slot);
+		if (!rhlist || !(list = rcu_dereference(list->next))) {
+			p = rcu_dereference(p->next);
+			list = container_of(p, struct rhlist_head, rhead);
+		}
 		goto next;
 	}
 
@@ -605,6 +704,18 @@ void *rhashtable_walk_next(struct rhashtable_iter *iter)
 		int skip = iter->skip;
 
 		rht_for_each_rcu(p, tbl, iter->slot) {
+			if (rhlist) {
+				list = container_of(p, struct rhlist_head,
+						    rhead);
+				do {
+					if (!skip)
+						goto next;
+					skip--;
+					list = rcu_dereference(list->next);
+				} while (list);
+
+				continue;
+			}
 			if (!skip)
 				break;
 			skip--;
@@ -614,7 +725,8 @@ void *rhashtable_walk_next(struct rhashtable_iter *iter)
 		if (!rht_is_a_nulls(p)) {
 			iter->skip++;
 			iter->p = p;
-			return rht_obj(ht, p);
+			iter->list = list;
+			return rht_obj(ht, rhlist ? &list->rhead : p);
 		}
 
 		iter->skip = 0;
@@ -802,6 +914,48 @@ int rhashtable_init(struct rhashtable *ht,
 }
 EXPORT_SYMBOL_GPL(rhashtable_init);
 
+/**
+ * rhltable_init - initialize a new hash list table
+ * @hlt:	hash list table to be initialized
+ * @params:	configuration parameters
+ *
+ * Initializes a new hash list table.
+ *
+ * See documentation for rhashtable_init.
+ */
+int rhltable_init(struct rhltable *hlt, const struct rhashtable_params *params)
+{
+	int err;
+
+	/* No rhlist NULLs marking for now. */
+	if (params->nulls_base)
+		return -EINVAL;
+
+	err = rhashtable_init(&hlt->ht, params);
+	hlt->ht.rhlist = true;
+	return err;
+}
+EXPORT_SYMBOL_GPL(rhltable_init);
+
+static void rhashtable_free_one(struct rhashtable *ht, struct rhash_head *obj,
+				void (*free_fn)(void *ptr, void *arg),
+				void *arg)
+{
+	struct rhlist_head *list;
+
+	if (!ht->rhlist) {
+		free_fn(rht_obj(ht, obj), arg);
+		return;
+	}
+
+	list = container_of(obj, struct rhlist_head, rhead);
+	do {
+		obj = &list->rhead;
+		list = rht_dereference(list->next, ht);
+		free_fn(rht_obj(ht, obj), arg);
+	} while (list);
+}
+
 /**
  * rhashtable_free_and_destroy - free elements and destroy hash table
  * @ht:		the hash table to destroy
@@ -839,7 +993,7 @@ void rhashtable_free_and_destroy(struct rhashtable *ht,
 			     pos = next,
 			     next = !rht_is_a_nulls(pos) ?
 					rht_dereference(pos->next, ht) : NULL)
-				free_fn(rht_obj(ht, pos), arg);
+				rhashtable_free_one(ht, pos, free_fn, arg);
 		}
 	}
 

net/mac80211/ieee80211_i.h

@@ -1213,7 +1213,7 @@ struct ieee80211_local {
 	spinlock_t tim_lock;
 	unsigned long num_sta;
 	struct list_head sta_list;
-	struct rhashtable sta_hash;
+	struct rhltable sta_hash;
 	struct timer_list sta_cleanup;
 	int sta_generation;
 

net/mac80211/rx.c

@@ -4003,7 +4003,7 @@ static void __ieee80211_rx_handle_packet(struct ieee80211_hw *hw,
 	__le16 fc;
 	struct ieee80211_rx_data rx;
 	struct ieee80211_sub_if_data *prev;
-	struct rhash_head *tmp;
+	struct rhlist_head *tmp;
 	int err = 0;
 
 	fc = ((struct ieee80211_hdr *)skb->data)->frame_control;
@@ -4046,13 +4046,10 @@ static void __ieee80211_rx_handle_packet(struct ieee80211_hw *hw,
 		goto out;
 	} else if (ieee80211_is_data(fc)) {
 		struct sta_info *sta, *prev_sta;
-		const struct bucket_table *tbl;
 
 		prev_sta = NULL;
 
-		tbl = rht_dereference_rcu(local->sta_hash.tbl, &local->sta_hash);
-
-		for_each_sta_info(local, tbl, hdr->addr2, sta, tmp) {
+		for_each_sta_info(local, hdr->addr2, sta, tmp) {
 			if (!prev_sta) {
 				prev_sta = sta;
 				continue;

net/mac80211/sta_info.c

@@ -67,12 +67,10 @@
 
 static const struct rhashtable_params sta_rht_params = {
 	.nelem_hint = 3, /* start small */
-	.insecure_elasticity = true, /* Disable chain-length checks. */
 	.automatic_shrinking = true,
 	.head_offset = offsetof(struct sta_info, hash_node),
 	.key_offset = offsetof(struct sta_info, addr),
 	.key_len = ETH_ALEN,
-	.hashfn = sta_addr_hash,
 	.max_size = CONFIG_MAC80211_STA_HASH_MAX_SIZE,
 };
 
@@ -80,7 +78,7 @@ static const struct rhashtable_params sta_rht_params = {
 static int sta_info_hash_del(struct ieee80211_local *local,
 			     struct sta_info *sta)
 {
-	return rhashtable_remove_fast(&local->sta_hash, &sta->hash_node,
+	return rhltable_remove(&local->sta_hash, &sta->hash_node,
 			       sta_rht_params);
 }
 
@@ -157,19 +155,22 @@ static void cleanup_single_sta(struct sta_info *sta)
 	sta_info_free(local, sta);
 }
 
+struct rhlist_head *sta_info_hash_lookup(struct ieee80211_local *local,
+					 const u8 *addr)
+{
+	return rhltable_lookup(&local->sta_hash, addr, sta_rht_params);
+}
+
 /* protected by RCU */
 struct sta_info *sta_info_get(struct ieee80211_sub_if_data *sdata,
 			      const u8 *addr)
 {
 	struct ieee80211_local *local = sdata->local;
+	struct rhlist_head *tmp;
 	struct sta_info *sta;
-	struct rhash_head *tmp;
-	const struct bucket_table *tbl;
 
 	rcu_read_lock();
-	tbl = rht_dereference_rcu(local->sta_hash.tbl, &local->sta_hash);
-
-	for_each_sta_info(local, tbl, addr, sta, tmp) {
+	for_each_sta_info(local, addr, sta, tmp) {
 		if (sta->sdata == sdata) {
 			rcu_read_unlock();
 			/* this is safe as the caller must already hold
@@ -190,14 +191,11 @@ struct sta_info *sta_info_get_bss(struct ieee80211_sub_if_data *sdata,
 				  const u8 *addr)
 {
 	struct ieee80211_local *local = sdata->local;
+	struct rhlist_head *tmp;
 	struct sta_info *sta;
-	struct rhash_head *tmp;
-	const struct bucket_table *tbl;
 
 	rcu_read_lock();
-	tbl = rht_dereference_rcu(local->sta_hash.tbl, &local->sta_hash);
-
-	for_each_sta_info(local, tbl, addr, sta, tmp) {
+	for_each_sta_info(local, addr, sta, tmp) {
 		if (sta->sdata == sdata ||
 		    (sta->sdata->bss && sta->sdata->bss == sdata->bss)) {
 			rcu_read_unlock();
@@ -263,7 +261,7 @@ void sta_info_free(struct ieee80211_local *local, struct sta_info *sta)
 static int sta_info_hash_add(struct ieee80211_local *local,
 			     struct sta_info *sta)
 {
-	return rhashtable_insert_fast(&local->sta_hash, &sta->hash_node,
+	return rhltable_insert(&local->sta_hash, &sta->hash_node,
 			       sta_rht_params);
 }
 
@@ -453,9 +451,9 @@ static int sta_info_insert_check(struct sta_info *sta)
 		    is_multicast_ether_addr(sta->sta.addr)))
 		return -EINVAL;
 
-	/* Strictly speaking this isn't necessary as we hold the mutex, but
-	 * the rhashtable code can't really deal with that distinction. We
-	 * do require the mutex for correctness though.
+	/* The RCU read lock is required by rhashtable due to
+	 * asynchronous resize/rehash.  We also require the mutex
+	 * for correctness.
 	 */
 	rcu_read_lock();
 	lockdep_assert_held(&sdata->local->sta_mtx);
@@ -1043,16 +1041,11 @@ static void sta_info_cleanup(unsigned long data)
 		  round_jiffies(jiffies + STA_INFO_CLEANUP_INTERVAL));
 }
 
-u32 sta_addr_hash(const void *key, u32 length, u32 seed)
-{
-	return jhash(key, ETH_ALEN, seed);
-}
-
 int sta_info_init(struct ieee80211_local *local)
 {
 	int err;
 
-	err = rhashtable_init(&local->sta_hash, &sta_rht_params);
+	err = rhltable_init(&local->sta_hash, &sta_rht_params);
 	if (err)
 		return err;
 
@@ -1068,7 +1061,7 @@ int sta_info_init(struct ieee80211_local *local)
 void sta_info_stop(struct ieee80211_local *local)
 {
 	del_timer_sync(&local->sta_cleanup);
-	rhashtable_destroy(&local->sta_hash);
+	rhltable_destroy(&local->sta_hash);
 }
 
 
@@ -1138,17 +1131,14 @@ struct ieee80211_sta *ieee80211_find_sta_by_ifaddr(struct ieee80211_hw *hw,
 						   const u8 *localaddr)
 {
 	struct ieee80211_local *local = hw_to_local(hw);
+	struct rhlist_head *tmp;
 	struct sta_info *sta;
-	struct rhash_head *tmp;
-	const struct bucket_table *tbl;
-
-	tbl = rht_dereference_rcu(local->sta_hash.tbl, &local->sta_hash);
 
 	/*
 	 * Just return a random station if localaddr is NULL
 	 * ... first in list.
 	 */
-	for_each_sta_info(local, tbl, addr, sta, tmp) {
+	for_each_sta_info(local, addr, sta, tmp) {
 		if (localaddr &&
 		    !ether_addr_equal(sta->sdata->vif.addr, localaddr))
 			continue;

net/mac80211/sta_info.h

@@ -455,7 +455,7 @@ struct sta_info {
 	/* General information, mostly static */
 	struct list_head list, free_list;
 	struct rcu_head rcu_head;
-	struct rhash_head hash_node;
+	struct rhlist_head hash_node;
 	u8 addr[ETH_ALEN];
 	struct ieee80211_local *local;
 	struct ieee80211_sub_if_data *sdata;
@@ -638,6 +638,9 @@ rcu_dereference_protected_tid_tx(struct sta_info *sta, int tid)
  */
 #define STA_INFO_CLEANUP_INTERVAL (10 * HZ)
 
+struct rhlist_head *sta_info_hash_lookup(struct ieee80211_local *local,
+					 const u8 *addr);
+
 /*
  * Get a STA info, must be under RCU read lock.
  */
@@ -647,17 +650,9 @@ struct sta_info *sta_info_get(struct ieee80211_sub_if_data *sdata,
 struct sta_info *sta_info_get_bss(struct ieee80211_sub_if_data *sdata,
 				  const u8 *addr);
 
-u32 sta_addr_hash(const void *key, u32 length, u32 seed);
-
-#define _sta_bucket_idx(_tbl, _a)					\
-	rht_bucket_index(_tbl, sta_addr_hash(_a, ETH_ALEN, (_tbl)->hash_rnd))
-
-#define for_each_sta_info(local, tbl, _addr, _sta, _tmp)		\
-	rht_for_each_entry_rcu(_sta, _tmp, tbl, 			\
-			       _sta_bucket_idx(tbl, _addr),		\
-			       hash_node)				\
-	/* compare address and run code only if it matches */		\
-	if (ether_addr_equal(_sta->addr, (_addr)))
+#define for_each_sta_info(local, _addr, _sta, _tmp)			\
+	rhl_for_each_entry_rcu(_sta, _tmp,				\
+			       sta_info_hash_lookup(local, _addr), hash_node)
 
 /*
  * Get STA info by index, BROKEN!

net/mac80211/status.c

@@ -746,8 +746,8 @@ void ieee80211_tx_status(struct ieee80211_hw *hw, struct sk_buff *skb)
 	struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb);
 	__le16 fc;
 	struct ieee80211_supported_band *sband;
+	struct rhlist_head *tmp;
 	struct sta_info *sta;
-	struct rhash_head *tmp;
 	int retry_count;
 	int rates_idx;
 	bool send_to_cooked;
@@ -755,7 +755,6 @@ void ieee80211_tx_status(struct ieee80211_hw *hw, struct sk_buff *skb)
 	struct ieee80211_bar *bar;
 	int shift = 0;
 	int tid = IEEE80211_NUM_TIDS;
-	const struct bucket_table *tbl;
 
 	rates_idx = ieee80211_tx_get_rates(hw, info, &retry_count);
 
@@ -764,9 +763,7 @@ void ieee80211_tx_status(struct ieee80211_hw *hw, struct sk_buff *skb)
 	sband = local->hw.wiphy->bands[info->band];
 	fc = hdr->frame_control;
 
-	tbl = rht_dereference_rcu(local->sta_hash.tbl, &local->sta_hash);
-
-	for_each_sta_info(local, tbl, hdr->addr1, sta, tmp) {
+	for_each_sta_info(local, hdr->addr1, sta, tmp) {
 		/* skip wrong virtual interface */
 		if (!ether_addr_equal(hdr->addr2, sta->sdata->vif.addr))
 			continue;