]> git.openfabrics.org - compat-rdma/compat.git/commitdiff
Added functionality to support SLES12 SP3
authorVladimir Sokolovsky <vlad@mellanox.com>
Wed, 8 Aug 2018 20:09:40 +0000 (15:09 -0500)
committerVladimir Sokolovsky <vlad@mellanox.com>
Wed, 8 Aug 2018 20:09:40 +0000 (15:09 -0500)
Signed-off-by: Vladimir Sokolovsky <vlad@mellanox.com>
compat/rhashtable.c [new file with mode: 0644]
config/rdma.m4
include/linux/ethtool.h [new file with mode: 0644]
include/linux/kernel.h [new file with mode: 0644]
include/linux/rhashtable.h [new file with mode: 0644]

diff --git a/compat/rhashtable.c b/compat/rhashtable.c
new file mode 100644 (file)
index 0000000..55f2fbb
--- /dev/null
@@ -0,0 +1,1015 @@
+/*
+ * Resizable, Scalable, Concurrent Hash Table
+ *
+ * Copyright (c) 2015 Herbert Xu <herbert@gondor.apana.org.au>
+ * Copyright (c) 2014-2015 Thomas Graf <tgraf@suug.ch>
+ * Copyright (c) 2008-2014 Patrick McHardy <kaber@trash.net>
+ *
+ * Code partially derived from nft_hash
+ * Rewritten with rehash code from br_multicast plus single list
+ * pointer as suggested by Josh Triplett
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/rhashtable.h>
+
+#ifndef HAVE_RHLTABLE
+
+#include <linux/atomic.h>
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/log2.h>
+#include <linux/sched.h>
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
+#include <linux/mm.h>
+#include <linux/jhash.h>
+#include <linux/random.h>
+#include <linux/err.h>
+#include <linux/export.h>
+
+#define HASH_DEFAULT_SIZE      64UL
+#define HASH_MIN_SIZE          4U
+#define BUCKET_LOCKS_PER_CPU   32UL
+
+static u32 head_hashfn(struct rhashtable *ht,
+                      const struct bucket_table *tbl,
+                      const struct rhash_head *he)
+{
+       return rht_head_hashfn(ht, tbl, he, ht->p);
+}
+
+#ifdef CONFIG_PROVE_LOCKING
+#define ASSERT_RHT_MUTEX(HT) BUG_ON(!lockdep_rht_mutex_is_held(HT))
+
+int lockdep_rht_mutex_is_held(struct rhashtable *ht)
+{
+       return (debug_locks) ? lockdep_is_held(&ht->mutex) : 1;
+}
+EXPORT_SYMBOL_GPL(lockdep_rht_mutex_is_held);
+
+int lockdep_rht_bucket_is_held(const struct bucket_table *tbl, u32 hash)
+{
+       spinlock_t *lock = rht_bucket_lock(tbl, hash);
+
+       return (debug_locks) ? lockdep_is_held(lock) : 1;
+}
+EXPORT_SYMBOL_GPL(lockdep_rht_bucket_is_held);
+#else
+#define ASSERT_RHT_MUTEX(HT)
+#endif
+
+
+static int alloc_bucket_locks(struct rhashtable *ht, struct bucket_table *tbl,
+                             gfp_t gfp)
+{
+       unsigned int i, size;
+#if defined(CONFIG_PROVE_LOCKING)
+       unsigned int nr_pcpus = 2;
+#else
+       unsigned int nr_pcpus = num_possible_cpus();
+#endif
+
+       nr_pcpus = min_t(unsigned int, nr_pcpus, 64UL);
+       size = roundup_pow_of_two(nr_pcpus * ht->p.locks_mul);
+
+       /* Never allocate more than 0.5 locks per bucket */
+       size = min_t(unsigned int, size, tbl->size >> 1);
+
+       if (sizeof(spinlock_t) != 0) {
+               tbl->locks = NULL;
+#ifdef CONFIG_NUMA
+               if (size * sizeof(spinlock_t) > PAGE_SIZE &&
+                   gfp == GFP_KERNEL)
+                       tbl->locks = vmalloc(size * sizeof(spinlock_t));
+#endif
+               if (gfp != GFP_KERNEL)
+                       gfp |= __GFP_NOWARN | __GFP_NORETRY;
+
+               if (!tbl->locks)
+                       tbl->locks = kmalloc_array(size, sizeof(spinlock_t),
+                                                  gfp);
+               if (!tbl->locks)
+                       return -ENOMEM;
+               for (i = 0; i < size; i++)
+                       spin_lock_init(&tbl->locks[i]);
+       }
+       tbl->locks_mask = size - 1;
+
+       return 0;
+}
+
+static void bucket_table_free(const struct bucket_table *tbl)
+{
+       if (tbl)
+               kvfree(tbl->locks);
+
+       kvfree(tbl);
+}
+
+static void bucket_table_free_rcu(struct rcu_head *head)
+{
+       bucket_table_free(container_of(head, struct bucket_table, rcu));
+}
+
+static struct bucket_table *bucket_table_alloc(struct rhashtable *ht,
+                                              size_t nbuckets,
+                                              gfp_t gfp)
+{
+       struct bucket_table *tbl = NULL;
+       size_t size;
+       int i;
+
+       size = sizeof(*tbl) + nbuckets * sizeof(tbl->buckets[0]);
+       if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER) ||
+           gfp != GFP_KERNEL)
+               tbl = kzalloc(size, gfp | __GFP_NOWARN | __GFP_NORETRY);
+       if (tbl == NULL && gfp == GFP_KERNEL)
+               tbl = vzalloc(size);
+       if (tbl == NULL)
+               return NULL;
+
+       tbl->size = nbuckets;
+
+       if (alloc_bucket_locks(ht, tbl, gfp) < 0) {
+               bucket_table_free(tbl);
+               return NULL;
+       }
+
+       INIT_LIST_HEAD(&tbl->walkers);
+
+       get_random_bytes(&tbl->hash_rnd, sizeof(tbl->hash_rnd));
+
+       for (i = 0; i < nbuckets; i++)
+               INIT_RHT_NULLS_HEAD(tbl->buckets[i], ht, i);
+
+       return tbl;
+}
+
+static struct bucket_table *rhashtable_last_table(struct rhashtable *ht,
+                                                 struct bucket_table *tbl)
+{
+       struct bucket_table *new_tbl;
+
+       do {
+               new_tbl = tbl;
+               tbl = rht_dereference_rcu(tbl->future_tbl, ht);
+       } while (tbl);
+
+       return new_tbl;
+}
+
+static int rhashtable_rehash_one(struct rhashtable *ht, unsigned int old_hash)
+{
+       struct bucket_table *old_tbl = rht_dereference(ht->tbl, ht);
+       struct bucket_table *new_tbl = rhashtable_last_table(ht,
+               rht_dereference_rcu(old_tbl->future_tbl, ht));
+       struct rhash_head __rcu **pprev = &old_tbl->buckets[old_hash];
+       int err = -ENOENT;
+       struct rhash_head *head, *next, *entry;
+       spinlock_t *new_bucket_lock;
+       unsigned int new_hash;
+
+       rht_for_each(entry, old_tbl, old_hash) {
+               err = 0;
+               next = rht_dereference_bucket(entry->next, old_tbl, old_hash);
+
+               if (rht_is_a_nulls(next))
+                       break;
+
+               pprev = &entry->next;
+       }
+
+       if (err)
+               goto out;
+
+       new_hash = head_hashfn(ht, new_tbl, entry);
+
+       new_bucket_lock = rht_bucket_lock(new_tbl, new_hash);
+
+       spin_lock_nested(new_bucket_lock, SINGLE_DEPTH_NESTING);
+       head = rht_dereference_bucket(new_tbl->buckets[new_hash],
+                                     new_tbl, new_hash);
+
+       RCU_INIT_POINTER(entry->next, head);
+
+       rcu_assign_pointer(new_tbl->buckets[new_hash], entry);
+       spin_unlock(new_bucket_lock);
+
+       rcu_assign_pointer(*pprev, next);
+
+out:
+       return err;
+}
+
+static void rhashtable_rehash_chain(struct rhashtable *ht,
+                                   unsigned int old_hash)
+{
+       struct bucket_table *old_tbl = rht_dereference(ht->tbl, ht);
+       spinlock_t *old_bucket_lock;
+
+       old_bucket_lock = rht_bucket_lock(old_tbl, old_hash);
+
+       spin_lock_bh(old_bucket_lock);
+       while (!rhashtable_rehash_one(ht, old_hash))
+               ;
+       old_tbl->rehash++;
+       spin_unlock_bh(old_bucket_lock);
+}
+
+static int rhashtable_rehash_attach(struct rhashtable *ht,
+                                   struct bucket_table *old_tbl,
+                                   struct bucket_table *new_tbl)
+{
+       /* Protect future_tbl using the first bucket lock. */
+       spin_lock_bh(old_tbl->locks);
+
+       /* Did somebody beat us to it? */
+       if (rcu_access_pointer(old_tbl->future_tbl)) {
+               spin_unlock_bh(old_tbl->locks);
+               return -EEXIST;
+       }
+
+       /* Make insertions go into the new, empty table right away. Deletions
+        * and lookups will be attempted in both tables until we synchronize.
+        */
+       rcu_assign_pointer(old_tbl->future_tbl, new_tbl);
+
+       spin_unlock_bh(old_tbl->locks);
+
+       return 0;
+}
+
+static int rhashtable_rehash_table(struct rhashtable *ht)
+{
+       struct bucket_table *old_tbl = rht_dereference(ht->tbl, ht);
+       struct bucket_table *new_tbl;
+       struct rhashtable_walker *walker;
+       unsigned int old_hash;
+
+       new_tbl = rht_dereference(old_tbl->future_tbl, ht);
+       if (!new_tbl)
+               return 0;
+
+       for (old_hash = 0; old_hash < old_tbl->size; old_hash++)
+               rhashtable_rehash_chain(ht, old_hash);
+
+       /* Publish the new table pointer. */
+       rcu_assign_pointer(ht->tbl, new_tbl);
+
+       spin_lock(&ht->lock);
+       list_for_each_entry(walker, &old_tbl->walkers, list)
+               walker->tbl = NULL;
+       spin_unlock(&ht->lock);
+
+       /* Wait for readers. All new readers will see the new
+        * table, and thus no references to the old table will
+        * remain.
+        */
+       call_rcu(&old_tbl->rcu, bucket_table_free_rcu);
+
+       return rht_dereference(new_tbl->future_tbl, ht) ? -EAGAIN : 0;
+}
+
+/**
+ * rhashtable_expand - Expand hash table while allowing concurrent lookups
+ * @ht:                the hash table to expand
+ *
+ * A secondary bucket array is allocated and the hash entries are migrated.
+ *
+ * This function may only be called in a context where it is safe to call
+ * synchronize_rcu(), e.g. not within a rcu_read_lock() section.
+ *
+ * The caller must ensure that no concurrent resizing occurs by holding
+ * ht->mutex.
+ *
+ * It is valid to have concurrent insertions and deletions protected by per
+ * bucket locks or concurrent RCU protected lookups and traversals.
+ */
+static int rhashtable_expand(struct rhashtable *ht)
+{
+       struct bucket_table *new_tbl, *old_tbl = rht_dereference(ht->tbl, ht);
+       int err;
+
+       ASSERT_RHT_MUTEX(ht);
+
+       old_tbl = rhashtable_last_table(ht, old_tbl);
+
+       new_tbl = bucket_table_alloc(ht, old_tbl->size * 2, GFP_KERNEL);
+       if (new_tbl == NULL)
+               return -ENOMEM;
+
+       err = rhashtable_rehash_attach(ht, old_tbl, new_tbl);
+       if (err)
+               bucket_table_free(new_tbl);
+
+       return err;
+}
+
+/**
+ * rhashtable_shrink - Shrink hash table while allowing concurrent lookups
+ * @ht:                the hash table to shrink
+ *
+ * This function shrinks the hash table to fit, i.e., the smallest
+ * size would not cause it to expand right away automatically.
+ *
+ * The caller must ensure that no concurrent resizing occurs by holding
+ * ht->mutex.
+ *
+ * The caller must ensure that no concurrent table mutations take place.
+ * It is however valid to have concurrent lookups if they are RCU protected.
+ *
+ * It is valid to have concurrent insertions and deletions protected by per
+ * bucket locks or concurrent RCU protected lookups and traversals.
+ */
+static int rhashtable_shrink(struct rhashtable *ht)
+{
+       struct bucket_table *new_tbl, *old_tbl = rht_dereference(ht->tbl, ht);
+       unsigned int nelems = atomic_read(&ht->nelems);
+       unsigned int size = 0;
+       int err;
+
+       ASSERT_RHT_MUTEX(ht);
+
+       if (nelems)
+               size = roundup_pow_of_two(nelems * 3 / 2);
+       if (size < ht->p.min_size)
+               size = ht->p.min_size;
+
+       if (old_tbl->size <= size)
+               return 0;
+
+       if (rht_dereference(old_tbl->future_tbl, ht))
+               return -EEXIST;
+
+       new_tbl = bucket_table_alloc(ht, size, GFP_KERNEL);
+       if (new_tbl == NULL)
+               return -ENOMEM;
+
+       err = rhashtable_rehash_attach(ht, old_tbl, new_tbl);
+       if (err)
+               bucket_table_free(new_tbl);
+
+       return err;
+}
+
+static void rht_deferred_worker(struct work_struct *work)
+{
+       struct rhashtable *ht;
+       struct bucket_table *tbl;
+       int err = 0;
+
+       ht = container_of(work, struct rhashtable, run_work);
+       mutex_lock(&ht->mutex);
+
+       tbl = rht_dereference(ht->tbl, ht);
+       tbl = rhashtable_last_table(ht, tbl);
+
+       if (rht_grow_above_75(ht, tbl))
+               rhashtable_expand(ht);
+       else if (ht->p.automatic_shrinking && rht_shrink_below_30(ht, tbl))
+               rhashtable_shrink(ht);
+
+       err = rhashtable_rehash_table(ht);
+
+       mutex_unlock(&ht->mutex);
+
+       if (err)
+               schedule_work(&ht->run_work);
+}
+
+static int rhashtable_insert_rehash(struct rhashtable *ht,
+                                   struct bucket_table *tbl)
+{
+       struct bucket_table *old_tbl;
+       struct bucket_table *new_tbl;
+       unsigned int size;
+       int err;
+
+       old_tbl = rht_dereference_rcu(ht->tbl, ht);
+
+       size = tbl->size;
+
+       err = -EBUSY;
+
+       if (rht_grow_above_75(ht, tbl))
+               size *= 2;
+       /* Do not schedule more than one rehash */
+       else if (old_tbl != tbl)
+               goto fail;
+
+       err = -ENOMEM;
+
+       new_tbl = bucket_table_alloc(ht, size, GFP_ATOMIC);
+       if (new_tbl == NULL)
+               goto fail;
+
+       err = rhashtable_rehash_attach(ht, tbl, new_tbl);
+       if (err) {
+               bucket_table_free(new_tbl);
+               if (err == -EEXIST)
+                       err = 0;
+       } else
+               schedule_work(&ht->run_work);
+
+       return err;
+
+fail:
+       /* Do not fail the insert if someone else did a rehash. */
+       if (likely(rcu_dereference_raw(tbl->future_tbl)))
+               return 0;
+
+       /* Schedule async rehash to retry allocation in process context. */
+       if (err == -ENOMEM)
+               schedule_work(&ht->run_work);
+
+       return err;
+}
+
+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;
+       int elasticity;
+
+       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)))) {
+                       pprev = &head->next;
+                       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;
+       }
+
+       if (elasticity <= 0)
+               return ERR_PTR(-EAGAIN);
+
+       return ERR_PTR(-ENOENT);
+}
+
+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);
+
+               spin_unlock(rht_bucket_lock(tbl, hash));
+       }
+
+       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);
+
+/**
+ * rhashtable_walk_enter - Initialise an iterator
+ * @ht:                Table to walk over
+ * @iter:      Hash table Iterator
+ *
+ * This function prepares a hash table walk.
+ *
+ * Note that if you restart a walk after rhashtable_walk_stop you
+ * may see the same object twice.  Also, you may miss objects if
+ * there are removals in between rhashtable_walk_stop and the next
+ * call to rhashtable_walk_start.
+ *
+ * For a completely stable walk you should construct your own data
+ * structure outside the hash table.
+ *
+ * This function may sleep so you must not call it from interrupt
+ * context or with spin locks held.
+ *
+ * You must call rhashtable_walk_exit after this function returns.
+ */
+void rhashtable_walk_enter(struct rhashtable *ht, struct rhashtable_iter *iter)
+{
+       iter->ht = ht;
+       iter->p = NULL;
+       iter->slot = 0;
+       iter->skip = 0;
+
+       spin_lock(&ht->lock);
+       iter->walker.tbl =
+               rcu_dereference_protected(ht->tbl, lockdep_is_held(&ht->lock));
+       list_add(&iter->walker.list, &iter->walker.tbl->walkers);
+       spin_unlock(&ht->lock);
+}
+EXPORT_SYMBOL_GPL(rhashtable_walk_enter);
+
+/**
+ * rhashtable_walk_exit - Free an iterator
+ * @iter:      Hash table Iterator
+ *
+ * This function frees resources allocated by rhashtable_walk_init.
+ */
+void rhashtable_walk_exit(struct rhashtable_iter *iter)
+{
+       spin_lock(&iter->ht->lock);
+       if (iter->walker.tbl)
+               list_del(&iter->walker.list);
+       spin_unlock(&iter->ht->lock);
+}
+EXPORT_SYMBOL_GPL(rhashtable_walk_exit);
+
+/**
+ * rhashtable_walk_start - Start a hash table walk
+ * @iter:      Hash table iterator
+ *
+ * Start a hash table walk.  Note that we take the RCU lock in all
+ * cases including when we return an error.  So you must always call
+ * rhashtable_walk_stop to clean up.
+ *
+ * Returns zero if successful.
+ *
+ * Returns -EAGAIN if resize event occured.  Note that the iterator
+ * will rewind back to the beginning and you may use it immediately
+ * by calling rhashtable_walk_next.
+ */
+int rhashtable_walk_start(struct rhashtable_iter *iter)
+       __acquires(RCU)
+{
+       struct rhashtable *ht = iter->ht;
+
+       rcu_read_lock();
+
+       spin_lock(&ht->lock);
+       if (iter->walker.tbl)
+               list_del(&iter->walker.list);
+       spin_unlock(&ht->lock);
+
+       if (!iter->walker.tbl) {
+               iter->walker.tbl = rht_dereference_rcu(ht->tbl, ht);
+               return -EAGAIN;
+       }
+
+       return 0;
+}
+EXPORT_SYMBOL_GPL(rhashtable_walk_start);
+
+/**
+ * rhashtable_walk_next - Return the next object and advance the iterator
+ * @iter:      Hash table iterator
+ *
+ * Note that you must call rhashtable_walk_stop when you are finished
+ * with the walk.
+ *
+ * Returns the next object or NULL when the end of the table is reached.
+ *
+ * Returns -EAGAIN if resize event occured.  Note that the iterator
+ * will rewind back to the beginning and you may continue to use it.
+ */
+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) {
+               if (!rhlist || !(list = rcu_dereference(list->next))) {
+                       p = rcu_dereference(p->next);
+                       list = container_of(p, struct rhlist_head, rhead);
+               }
+               goto next;
+       }
+
+       for (; iter->slot < tbl->size; iter->slot++) {
+               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--;
+               }
+
+next:
+               if (!rht_is_a_nulls(p)) {
+                       iter->skip++;
+                       iter->p = p;
+                       iter->list = list;
+                       return rht_obj(ht, rhlist ? &list->rhead : p);
+               }
+
+               iter->skip = 0;
+       }
+
+       iter->p = NULL;
+
+       /* Ensure we see any new tables. */
+       smp_rmb();
+
+       iter->walker.tbl = rht_dereference_rcu(tbl->future_tbl, ht);
+       if (iter->walker.tbl) {
+               iter->slot = 0;
+               iter->skip = 0;
+               return ERR_PTR(-EAGAIN);
+       }
+
+       return NULL;
+}
+EXPORT_SYMBOL_GPL(rhashtable_walk_next);
+
+/**
+ * rhashtable_walk_stop - Finish a hash table walk
+ * @iter:      Hash table iterator
+ *
+ * Finish a hash table walk.
+ */
+void rhashtable_walk_stop(struct rhashtable_iter *iter)
+       __releases(RCU)
+{
+       struct rhashtable *ht;
+       struct bucket_table *tbl = iter->walker.tbl;
+
+       if (!tbl)
+               goto out;
+
+       ht = iter->ht;
+
+       spin_lock(&ht->lock);
+       if (tbl->rehash < tbl->size)
+               list_add(&iter->walker.list, &tbl->walkers);
+       else
+               iter->walker.tbl = NULL;
+       spin_unlock(&ht->lock);
+
+       iter->p = NULL;
+
+out:
+       rcu_read_unlock();
+}
+EXPORT_SYMBOL_GPL(rhashtable_walk_stop);
+
+static size_t rounded_hashtable_size(const struct rhashtable_params *params)
+{
+       return max(roundup_pow_of_two(params->nelem_hint * 4 / 3),
+                  (unsigned long)params->min_size);
+}
+
+static u32 rhashtable_jhash2(const void *key, u32 length, u32 seed)
+{
+       return jhash2(key, length, seed);
+}
+
+/**
+ * rhashtable_init - initialize a new hash table
+ * @ht:                hash table to be initialized
+ * @params:    configuration parameters
+ *
+ * Initializes a new hash table based on the provided configuration
+ * parameters. A table can be configured either with a variable or
+ * fixed length key:
+ *
+ * Configuration Example 1: Fixed length keys
+ * struct test_obj {
+ *     int                     key;
+ *     void *                  my_member;
+ *     struct rhash_head       node;
+ * };
+ *
+ * struct rhashtable_params params = {
+ *     .head_offset = offsetof(struct test_obj, node),
+ *     .key_offset = offsetof(struct test_obj, key),
+ *     .key_len = sizeof(int),
+ *     .hashfn = jhash,
+ *     .nulls_base = (1U << RHT_BASE_SHIFT),
+ * };
+ *
+ * Configuration Example 2: Variable length keys
+ * struct test_obj {
+ *     [...]
+ *     struct rhash_head       node;
+ * };
+ *
+ * u32 my_hash_fn(const void *data, u32 len, u32 seed)
+ * {
+ *     struct test_obj *obj = data;
+ *
+ *     return [... hash ...];
+ * }
+ *
+ * struct rhashtable_params params = {
+ *     .head_offset = offsetof(struct test_obj, node),
+ *     .hashfn = jhash,
+ *     .obj_hashfn = my_hash_fn,
+ * };
+ */
+int rhashtable_init(struct rhashtable *ht,
+                   const struct rhashtable_params *params)
+{
+       struct bucket_table *tbl;
+       size_t size;
+
+       size = HASH_DEFAULT_SIZE;
+
+       if ((!params->key_len && !params->obj_hashfn) ||
+           (params->obj_hashfn && !params->obj_cmpfn))
+               return -EINVAL;
+
+       if (params->nulls_base && params->nulls_base < (1U << RHT_BASE_SHIFT))
+               return -EINVAL;
+
+       memset(ht, 0, sizeof(*ht));
+       mutex_init(&ht->mutex);
+       spin_lock_init(&ht->lock);
+       memcpy(&ht->p, params, sizeof(*params));
+
+       if (params->min_size)
+               ht->p.min_size = roundup_pow_of_two(params->min_size);
+
+       if (params->max_size)
+               ht->p.max_size = rounddown_pow_of_two(params->max_size);
+
+       if (params->insecure_max_entries)
+               ht->p.insecure_max_entries =
+                       rounddown_pow_of_two(params->insecure_max_entries);
+       else
+               ht->p.insecure_max_entries = ht->p.max_size * 2;
+
+       ht->p.min_size = max(ht->p.min_size, HASH_MIN_SIZE);
+
+       if (params->nelem_hint)
+               size = rounded_hashtable_size(&ht->p);
+
+       /* The maximum (not average) chain length grows with the
+        * size of the hash table, at a rate of (log N)/(log log N).
+        * The value of 16 is selected so that even if the hash
+        * table grew to 2^32 you would not expect the maximum
+        * chain length to exceed it unless we are under attack
+        * (or extremely unlucky).
+        *
+        * As this limit is only to detect attacks, we don't need
+        * to set it to a lower value as you'd need the chain
+        * length to vastly exceed 16 to have any real effect
+        * on the system.
+        */
+       if (!params->insecure_elasticity)
+               ht->elasticity = 16;
+
+       if (params->locks_mul)
+               ht->p.locks_mul = roundup_pow_of_two(params->locks_mul);
+       else
+               ht->p.locks_mul = BUCKET_LOCKS_PER_CPU;
+
+       ht->key_len = ht->p.key_len;
+       if (!params->hashfn) {
+               ht->p.hashfn = jhash;
+
+               if (!(ht->key_len & (sizeof(u32) - 1))) {
+                       ht->key_len /= sizeof(u32);
+                       ht->p.hashfn = rhashtable_jhash2;
+               }
+       }
+
+       tbl = bucket_table_alloc(ht, size, GFP_KERNEL);
+       if (tbl == NULL)
+               return -ENOMEM;
+
+       atomic_set(&ht->nelems, 0);
+
+       RCU_INIT_POINTER(ht->tbl, tbl);
+
+       INIT_WORK(&ht->run_work, rht_deferred_worker);
+
+       return 0;
+}
+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
+ * @free_fn:   callback to release resources of element
+ * @arg:       pointer passed to free_fn
+ *
+ * Stops an eventual async resize. If defined, invokes free_fn for each
+ * element to releasal resources. Please note that RCU protected
+ * readers may still be accessing the elements. Releasing of resources
+ * must occur in a compatible manner. Then frees the bucket array.
+ *
+ * This function will eventually sleep to wait for an async resize
+ * to complete. The caller is responsible that no further write operations
+ * occurs in parallel.
+ */
+void rhashtable_free_and_destroy(struct rhashtable *ht,
+                                void (*free_fn)(void *ptr, void *arg),
+                                void *arg)
+{
+       const struct bucket_table *tbl;
+       unsigned int i;
+
+       cancel_work_sync(&ht->run_work);
+
+       mutex_lock(&ht->mutex);
+       tbl = rht_dereference(ht->tbl, ht);
+       if (free_fn) {
+               for (i = 0; i < tbl->size; i++) {
+                       struct rhash_head *pos, *next;
+
+                       for (pos = rht_dereference(tbl->buckets[i], ht),
+                            next = !rht_is_a_nulls(pos) ?
+                                       rht_dereference(pos->next, ht) : NULL;
+                            !rht_is_a_nulls(pos);
+                            pos = next,
+                            next = !rht_is_a_nulls(pos) ?
+                                       rht_dereference(pos->next, ht) : NULL)
+                               rhashtable_free_one(ht, pos, free_fn, arg);
+               }
+       }
+
+       bucket_table_free(tbl);
+       mutex_unlock(&ht->mutex);
+}
+EXPORT_SYMBOL_GPL(rhashtable_free_and_destroy);
+
+void rhashtable_destroy(struct rhashtable *ht)
+{
+       return rhashtable_free_and_destroy(ht, NULL, NULL);
+}
+EXPORT_SYMBOL_GPL(rhashtable_destroy);
+#endif
index aab7dbe74d97c5862c42d208e2905663985914d4..929256bb2c8df4333fa2b01a9f5605c3691c7ff4 100644 (file)
@@ -6871,6 +6871,98 @@ AC_DEFUN([LINUX_CONFIG_COMPAT],
        ],[
                AC_MSG_RESULT(no)
        ])
+
+       AC_MSG_CHECKING([if netdevice.h has netdev_walk_all_upper_dev_rcu])
+       LB_LINUX_TRY_COMPILE([
+               #include <linux/netdevice.h>
+
+               struct upper_list {
+                       struct list_head list;
+                       struct net_device *upper;
+               };
+
+               static int netdev_upper_walk(struct net_device *upper, void *data) {
+                       return 0;
+               }
+       ],[
+               struct net_device *ndev;
+               struct list_head upper_list;
+
+               netdev_walk_all_upper_dev_rcu(ndev, netdev_upper_walk, &upper_list);
+               return 0;
+       ],[
+               AC_MSG_RESULT(yes)
+               AC_DEFINE(HAVE_NETDEV_WALK_ALL_UPPER_DEV_RCU, 1,
+                         [netdev_walk_all_upper_dev_rcu is defined])
+       ],[
+               AC_MSG_RESULT(no)
+       ])
+
+       AC_MSG_CHECKING([if netdevice.h has netdev_has_upper_dev_all_rcu])
+       LB_LINUX_TRY_COMPILE([
+               #include <linux/netdevice.h>
+       ],[
+               struct net_device *dev;
+               struct net_device *upper;
+
+               netdev_has_upper_dev_all_rcu(dev, upper);
+               return 0;
+       ],[
+               AC_MSG_RESULT(yes)
+               AC_DEFINE(HAVE_NETDEV_HAS_UPPER_DEV_ALL_RCU, 1,
+                         [netdev_has_upper_dev_all_rcu is defined])
+       ],[
+               AC_MSG_RESULT(no)
+       ])
+
+       AC_MSG_CHECKING([if page_ref.h has page_ref_count/add/sub/inc])
+       LB_LINUX_TRY_COMPILE([
+               #include <linux/page_ref.h>
+       ],[
+               page_ref_count(NULL);
+               page_ref_add(NULL, 0);
+               page_ref_sub(NULL, 0);
+               page_ref_inc(NULL);
+
+               return 0;
+       ],[
+               AC_MSG_RESULT(yes)
+               AC_DEFINE(HAVE_PAGE_REF_COUNT_ADD_SUB_INC, 1,
+                         [page_ref_count/add/sub/inc defined])
+       ],[
+               AC_MSG_RESULT(no)
+       ])
+
+       AC_MSG_CHECKING([if struct netdev_features.h has NETIF_F_GSO_PARTIAL])
+       LB_LINUX_TRY_COMPILE([
+               #include <linux/netdev_features.h>
+       ],[
+               int x = NETIF_F_GSO_PARTIAL;
+
+               return 0;
+       ],[
+               AC_MSG_RESULT(yes)
+               AC_DEFINE(HAVE_NETIF_F_GSO_PARTIAL, 1,
+                         [NETIF_F_GSO_PARTIAL is defined in netdev_features.h])
+       ],[
+               AC_MSG_RESULT(no)
+       ])
+
+       AC_MSG_CHECKING([if  struct rhltable exists])
+       LB_LINUX_TRY_COMPILE([
+               #include <linux/rhashtable.h>
+       ],[
+               struct rhltable x;
+               x = x;
+
+               return 0;
+       ],[
+               AC_MSG_RESULT(yes)
+               AC_DEFINE(HAVE_RHLTABLE, 1,
+                         [struct rhltable is defined])
+       ],[
+               AC_MSG_RESULT(no)
+       ])
 ])
 #
 # COMPAT_CONFIG_HEADERS
diff --git a/include/linux/ethtool.h b/include/linux/ethtool.h
new file mode 100644 (file)
index 0000000..a8ba978
--- /dev/null
@@ -0,0 +1,11 @@
+#ifndef __COMPAT_ETHTOOL_H
+#define __COMPAT_ETHTOOL_H
+
+#include_next <linux/ethtool.h>
+
+
+#ifndef SPEED_14000
+#define SPEED_14000    14000
+#endif
+
+#endif
diff --git a/include/linux/kernel.h b/include/linux/kernel.h
new file mode 100644 (file)
index 0000000..52cca3a
--- /dev/null
@@ -0,0 +1,18 @@
+#ifndef COMPAT_KERNEL_H
+#define COMPAT_KERNEL_H
+
+#include "../../compat/config.h"
+
+#include_next <linux/kernel.h>
+
+
+#ifndef u64_to_user_ptr
+#define u64_to_user_ptr(x) (           \
+{                                      \
+       typecheck(u64, x);              \
+       (void __user *)(uintptr_t)x;    \
+}                                      \
+)
+#endif
+
+#endif /* COMPAT_KERNEL_H */
diff --git a/include/linux/rhashtable.h b/include/linux/rhashtable.h
new file mode 100644 (file)
index 0000000..8475712
--- /dev/null
@@ -0,0 +1,1256 @@
+/*
+ * Resizable, Scalable, Concurrent Hash Table
+ *
+ * Copyright (c) 2015-2016 Herbert Xu <herbert@gondor.apana.org.au>
+ * Copyright (c) 2014-2015 Thomas Graf <tgraf@suug.ch>
+ * Copyright (c) 2008-2014 Patrick McHardy <kaber@trash.net>
+ *
+ * Code partially derived from nft_hash
+ * Rewritten with rehash code from br_multicast plus single list
+ * pointer as suggested by Josh Triplett
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#ifndef __COMPAT_LINUX_RHASHTABLE_H
+#define __COMPAT_LINUX_RHASHTABLE_H
+
+#include "../../compat/config.h"
+
+#ifdef CONFIG__COMPAT_RHLTABLE
+#undef HAVE_RHLTABLE
+#endif
+
+#ifdef HAVE_RHLTABLE
+#include_next <linux/rhashtable.h>
+#else /* HAVE_RHLTABLE */
+/* use our own implementaiton when the rhltable struct is not avaialbe in the
+ * kerenl
+*/
+
+#include <linux/atomic.h>
+#include <linux/compiler.h>
+#include <linux/err.h>
+#include <linux/errno.h>
+#include <linux/jhash.h>
+#include <linux/list_nulls.h>
+#include <linux/workqueue.h>
+#include <linux/mutex.h>
+#include <linux/rcupdate.h>
+
+#ifndef NULLS_MARKER
+#define NULLS_MARKER(value) (1UL | (((long)value) << 1))
+#endif
+
+/*
+ * The end of the chain is marked with a special nulls marks which has
+ * the following format:
+ *
+ * +-------+-----------------------------------------------------+-+
+ * | Base  |                      Hash                           |1|
+ * +-------+-----------------------------------------------------+-+
+ *
+ * Base (4 bits) : Reserved to distinguish between multiple tables.
+ *                 Specified via &struct rhashtable_params.nulls_base.
+ * Hash (27 bits): Full hash (unmasked) of first element added to bucket
+ * 1 (1 bit)     : Nulls marker (always set)
+ *
+ * The remaining bits of the next pointer remain unused for now.
+ */
+#define RHT_BASE_BITS          4
+#define RHT_HASH_BITS          27
+#define RHT_BASE_SHIFT         RHT_HASH_BITS
+
+/* Base bits plus 1 bit for nulls marker */
+#define RHT_HASH_RESERVED_SPACE        (RHT_BASE_BITS + 1)
+
+struct rhash_head {
+       struct rhash_head __rcu         *next;
+};
+
+struct rhlist_head {
+       struct rhash_head               rhead;
+       struct rhlist_head __rcu        *next;
+};
+
+/**
+ * struct bucket_table - Table of hash buckets
+ * @size: Number of hash buckets
+ * @rehash: Current bucket being rehashed
+ * @hash_rnd: Random seed to fold into hash
+ * @locks_mask: Mask to apply before accessing locks[]
+ * @locks: Array of spinlocks protecting individual buckets
+ * @walkers: List of active walkers
+ * @rcu: RCU structure for freeing the table
+ * @future_tbl: Table under construction during rehashing
+ * @buckets: size * hash buckets
+ */
+struct bucket_table {
+       unsigned int            size;
+       unsigned int            rehash;
+       u32                     hash_rnd;
+       unsigned int            locks_mask;
+       spinlock_t              *locks;
+       struct list_head        walkers;
+       struct rcu_head         rcu;
+
+       struct bucket_table __rcu *future_tbl;
+
+       struct rhash_head __rcu *buckets[] ____cacheline_aligned_in_smp;
+};
+
+/**
+ * struct rhashtable_compare_arg - Key for the function rhashtable_compare
+ * @ht: Hash table
+ * @key: Key to compare against
+ */
+struct rhashtable_compare_arg {
+       struct rhashtable *ht;
+       const void *key;
+};
+
+typedef u32 (*rht_hashfn_t)(const void *data, u32 len, u32 seed);
+typedef u32 (*rht_obj_hashfn_t)(const void *data, u32 len, u32 seed);
+typedef int (*rht_obj_cmpfn_t)(struct rhashtable_compare_arg *arg,
+                              const void *obj);
+
+struct rhashtable;
+
+/**
+ * struct rhashtable_params - Hash table construction parameters
+ * @nelem_hint: Hint on number of elements, should be 75% of desired size
+ * @key_len: Length of key
+ * @key_offset: Offset of key in struct to be hashed
+ * @head_offset: Offset of rhash_head in struct to be hashed
+ * @insecure_max_entries: Maximum number of entries (may be exceeded)
+ * @max_size: Maximum size while expanding
+ * @min_size: Minimum size while shrinking
+ * @nulls_base: Base value to generate nulls marker
+ * @insecure_elasticity: Set to true to disable chain length checks
+ * @automatic_shrinking: Enable automatic shrinking of tables
+ * @locks_mul: Number of bucket locks to allocate per cpu (default: 128)
+ * @hashfn: Hash function (default: jhash2 if !(key_len % 4), or jhash)
+ * @obj_hashfn: Function to hash object
+ * @obj_cmpfn: Function to compare key with object
+ */
+struct rhashtable_params {
+       size_t                  nelem_hint;
+       size_t                  key_len;
+       size_t                  key_offset;
+       size_t                  head_offset;
+       unsigned int            insecure_max_entries;
+       unsigned int            max_size;
+       unsigned int            min_size;
+       u32                     nulls_base;
+       bool                    insecure_elasticity;
+       bool                    automatic_shrinking;
+       size_t                  locks_mul;
+       rht_hashfn_t            hashfn;
+       rht_obj_hashfn_t        obj_hashfn;
+       rht_obj_cmpfn_t         obj_cmpfn;
+};
+
+/**
+ * struct rhashtable - Hash table handle
+ * @tbl: Bucket table
+ * @nelems: Number of elements in table
+ * @key_len: Key length for hashfn
+ * @elasticity: Maximum chain length before rehash
+ * @p: Configuration parameters
+ * @rhlist: True if this is an rhltable
+ * @run_work: Deferred worker to expand/shrink asynchronously
+ * @mutex: Mutex to protect current/future table swapping
+ * @lock: Spin lock to protect walker list
+ */
+struct rhashtable {
+       struct bucket_table __rcu       *tbl;
+       atomic_t                        nelems;
+       unsigned int                    key_len;
+       unsigned int                    elasticity;
+       struct rhashtable_params        p;
+       bool                            rhlist;
+       struct work_struct              run_work;
+       struct mutex                    mutex;
+       spinlock_t                      lock;
+};
+
+/**
+ * struct rhltable - Hash table with duplicate objects in a list
+ * @ht: Underlying rhtable
+ */
+struct rhltable {
+       struct rhashtable ht;
+};
+
+/**
+ * struct rhashtable_walker - Hash table walker
+ * @list: List entry on list of walkers
+ * @tbl: The table that we were walking over
+ */
+struct rhashtable_walker {
+       struct list_head list;
+       struct bucket_table *tbl;
+};
+
+/**
+ * struct rhashtable_iter - Hash table iterator
+ * @ht: Table to iterate through
+ * @p: Current pointer
+ * @list: Current hash list pointer
+ * @walker: Associated rhashtable walker
+ * @slot: Current slot
+ * @skip: Number of entries to skip in slot
+ */
+struct rhashtable_iter {
+       struct rhashtable *ht;
+       struct rhash_head *p;
+       struct rhlist_head *list;
+       struct rhashtable_walker walker;
+       unsigned int slot;
+       unsigned int skip;
+};
+
+static inline unsigned long rht_marker(const struct rhashtable *ht, u32 hash)
+{
+       return NULLS_MARKER(ht->p.nulls_base + hash);
+}
+
+#define INIT_RHT_NULLS_HEAD(ptr, ht, hash) \
+       ((ptr) = (typeof(ptr)) rht_marker(ht, hash))
+
+static inline bool rht_is_a_nulls(const struct rhash_head *ptr)
+{
+       return ((unsigned long) ptr & 1);
+}
+
+static inline unsigned long rht_get_nulls_value(const struct rhash_head *ptr)
+{
+       return ((unsigned long) ptr) >> 1;
+}
+
+static inline void *rht_obj(const struct rhashtable *ht,
+                           const struct rhash_head *he)
+{
+       return (char *)he - ht->p.head_offset;
+}
+
+static inline unsigned int rht_bucket_index(const struct bucket_table *tbl,
+                                           unsigned int hash)
+{
+       return (hash >> RHT_HASH_RESERVED_SPACE) & (tbl->size - 1);
+}
+
+static inline unsigned int rht_key_hashfn(
+       struct rhashtable *ht, const struct bucket_table *tbl,
+       const void *key, const struct rhashtable_params params)
+{
+       unsigned int hash;
+
+       /* params must be equal to ht->p if it isn't constant. */
+       if (!__builtin_constant_p(params.key_len))
+               hash = ht->p.hashfn(key, ht->key_len, tbl->hash_rnd);
+       else if (params.key_len) {
+               unsigned int key_len = params.key_len;
+
+               if (params.hashfn)
+                       hash = params.hashfn(key, key_len, tbl->hash_rnd);
+               else if (key_len & (sizeof(u32) - 1))
+                       hash = jhash(key, key_len, tbl->hash_rnd);
+               else
+                       hash = jhash2(key, key_len / sizeof(u32),
+                                     tbl->hash_rnd);
+       } else {
+               unsigned int key_len = ht->p.key_len;
+
+               if (params.hashfn)
+                       hash = params.hashfn(key, key_len, tbl->hash_rnd);
+               else
+                       hash = jhash(key, key_len, tbl->hash_rnd);
+       }
+
+       return rht_bucket_index(tbl, hash);
+}
+
+static inline unsigned int rht_head_hashfn(
+       struct rhashtable *ht, const struct bucket_table *tbl,
+       const struct rhash_head *he, const struct rhashtable_params params)
+{
+       const char *ptr = rht_obj(ht, he);
+
+       return likely(params.obj_hashfn) ?
+              rht_bucket_index(tbl, params.obj_hashfn(ptr, params.key_len ?:
+                                                           ht->p.key_len,
+                                                      tbl->hash_rnd)) :
+              rht_key_hashfn(ht, tbl, ptr + params.key_offset, params);
+}
+
+/**
+ * rht_grow_above_75 - returns true if nelems > 0.75 * table-size
+ * @ht:                hash table
+ * @tbl:       current table
+ */
+static inline bool rht_grow_above_75(const struct rhashtable *ht,
+                                    const struct bucket_table *tbl)
+{
+       /* Expand table when exceeding 75% load */
+       return atomic_read(&ht->nelems) > (tbl->size / 4 * 3) &&
+              (!ht->p.max_size || tbl->size < ht->p.max_size);
+}
+
+/**
+ * rht_shrink_below_30 - returns true if nelems < 0.3 * table-size
+ * @ht:                hash table
+ * @tbl:       current table
+ */
+static inline bool rht_shrink_below_30(const struct rhashtable *ht,
+                                      const struct bucket_table *tbl)
+{
+       /* Shrink table beneath 30% load */
+       return atomic_read(&ht->nelems) < (tbl->size * 3 / 10) &&
+              tbl->size > ht->p.min_size;
+}
+
+/**
+ * rht_grow_above_100 - returns true if nelems > table-size
+ * @ht:                hash table
+ * @tbl:       current table
+ */
+static inline bool rht_grow_above_100(const struct rhashtable *ht,
+                                     const struct bucket_table *tbl)
+{
+       return atomic_read(&ht->nelems) > tbl->size &&
+               (!ht->p.max_size || tbl->size < ht->p.max_size);
+}
+
+/**
+ * rht_grow_above_max - returns true if table is above maximum
+ * @ht:                hash table
+ * @tbl:       current table
+ */
+static inline bool rht_grow_above_max(const struct rhashtable *ht,
+                                     const struct bucket_table *tbl)
+{
+       return ht->p.insecure_max_entries &&
+              atomic_read(&ht->nelems) >= ht->p.insecure_max_entries;
+}
+
+/* The bucket lock is selected based on the hash and protects mutations
+ * on a group of hash buckets.
+ *
+ * A maximum of tbl->size/2 bucket locks is allocated. This ensures that
+ * a single lock always covers both buckets which may both contains
+ * entries which link to the same bucket of the old table during resizing.
+ * This allows to simplify the locking as locking the bucket in both
+ * tables during resize always guarantee protection.
+ *
+ * IMPORTANT: When holding the bucket lock of both the old and new table
+ * during expansions and shrinking, the old bucket lock must always be
+ * acquired first.
+ */
+static inline spinlock_t *rht_bucket_lock(const struct bucket_table *tbl,
+                                         unsigned int hash)
+{
+       return &tbl->locks[hash & tbl->locks_mask];
+}
+
+#ifdef CONFIG_PROVE_LOCKING
+#define lockdep_rht_mutex_is_held LINUX_BACKPORT(lockdep_rht_mutex_is_held)
+int lockdep_rht_mutex_is_held(struct rhashtable *ht);
+#define lockdep_rht_bucket_is_held LINUX_BACKPORT(lockdep_rht_bucket_is_held)
+int lockdep_rht_bucket_is_held(const struct bucket_table *tbl, u32 hash);
+#else
+static inline int lockdep_rht_mutex_is_held(struct rhashtable *ht)
+{
+       return 1;
+}
+
+static inline int lockdep_rht_bucket_is_held(const struct bucket_table *tbl,
+                                            u32 hash)
+{
+       return 1;
+}
+#endif /* CONFIG_PROVE_LOCKING */
+
+#define rhashtable_init LINUX_BACKPORT(rhashtable_init)
+int rhashtable_init(struct rhashtable *ht,
+                   const struct rhashtable_params *params);
+#define rhltable_init LINUX_BACKPORT(rhltable_init)
+int rhltable_init(struct rhltable *hlt,
+                 const struct rhashtable_params *params);
+
+#define rhashtable_insert_slow LINUX_BACKPORT(rhashtable_insert_slow)
+void *rhashtable_insert_slow(struct rhashtable *ht, const void *key,
+                            struct rhash_head *obj);
+
+#define rhashtable_walk_enter LINUX_BACKPORT(rhashtable_walk_enter)
+void rhashtable_walk_enter(struct rhashtable *ht,
+                          struct rhashtable_iter *iter);
+#define rhashtable_walk_exit LINUX_BACKPORT(rhashtable_walk_exit)
+void rhashtable_walk_exit(struct rhashtable_iter *iter);
+#define rhashtable_walk_start LINUX_BACKPORT(rhashtable_walk_start)
+int rhashtable_walk_start(struct rhashtable_iter *iter) __acquires(RCU);
+#define rhashtable_walk_next LINUX_BACKPORT(rhashtable_walk_next)
+void *rhashtable_walk_next(struct rhashtable_iter *iter);
+#define rhashtable_walk_stop LINUX_BACKPORT(rhashtable_walk_stop)
+void rhashtable_walk_stop(struct rhashtable_iter *iter) __releases(RCU);
+
+#define rhashtable_free_and_destroy LINUX_BACKPORT(rhashtable_free_and_destroy)
+void rhashtable_free_and_destroy(struct rhashtable *ht,
+                                void (*free_fn)(void *ptr, void *arg),
+                                void *arg);
+#define rhashtable_destroy LINUX_BACKPORT(rhashtable_destroy)
+void rhashtable_destroy(struct rhashtable *ht);
+
+#define rht_dereference(p, ht) \
+       rcu_dereference_protected(p, lockdep_rht_mutex_is_held(ht))
+
+#define rht_dereference_rcu(p, ht) \
+       rcu_dereference_check(p, lockdep_rht_mutex_is_held(ht))
+
+#define rht_dereference_bucket(p, tbl, hash) \
+       rcu_dereference_protected(p, lockdep_rht_bucket_is_held(tbl, hash))
+
+#define rht_dereference_bucket_rcu(p, tbl, hash) \
+       rcu_dereference_check(p, lockdep_rht_bucket_is_held(tbl, hash))
+
+#define rht_entry(tpos, pos, member) \
+       ({ tpos = container_of(pos, typeof(*tpos), member); 1; })
+
+/**
+ * rht_for_each_continue - continue iterating over hash chain
+ * @pos:       the &struct rhash_head to use as a loop cursor.
+ * @head:      the previous &struct rhash_head to continue from
+ * @tbl:       the &struct bucket_table
+ * @hash:      the hash value / bucket index
+ */
+#define rht_for_each_continue(pos, head, tbl, hash) \
+       for (pos = rht_dereference_bucket(head, tbl, hash); \
+            !rht_is_a_nulls(pos); \
+            pos = rht_dereference_bucket((pos)->next, tbl, hash))
+
+/**
+ * rht_for_each - iterate over hash chain
+ * @pos:       the &struct rhash_head to use as a loop cursor.
+ * @tbl:       the &struct bucket_table
+ * @hash:      the hash value / bucket index
+ */
+#define rht_for_each(pos, tbl, hash) \
+       rht_for_each_continue(pos, (tbl)->buckets[hash], tbl, hash)
+
+/**
+ * rht_for_each_entry_continue - continue iterating over hash chain
+ * @tpos:      the type * to use as a loop cursor.
+ * @pos:       the &struct rhash_head to use as a loop cursor.
+ * @head:      the previous &struct rhash_head to continue from
+ * @tbl:       the &struct bucket_table
+ * @hash:      the hash value / bucket index
+ * @member:    name of the &struct rhash_head within the hashable struct.
+ */
+#define rht_for_each_entry_continue(tpos, pos, head, tbl, hash, member)        \
+       for (pos = rht_dereference_bucket(head, tbl, hash);             \
+            (!rht_is_a_nulls(pos)) && rht_entry(tpos, pos, member);    \
+            pos = rht_dereference_bucket((pos)->next, tbl, hash))
+
+/**
+ * rht_for_each_entry - iterate over hash chain of given type
+ * @tpos:      the type * to use as a loop cursor.
+ * @pos:       the &struct rhash_head to use as a loop cursor.
+ * @tbl:       the &struct bucket_table
+ * @hash:      the hash value / bucket index
+ * @member:    name of the &struct rhash_head within the hashable struct.
+ */
+#define rht_for_each_entry(tpos, pos, tbl, hash, member)               \
+       rht_for_each_entry_continue(tpos, pos, (tbl)->buckets[hash],    \
+                                   tbl, hash, member)
+
+/**
+ * rht_for_each_entry_safe - safely iterate over hash chain of given type
+ * @tpos:      the type * to use as a loop cursor.
+ * @pos:       the &struct rhash_head to use as a loop cursor.
+ * @next:      the &struct rhash_head to use as next in loop cursor.
+ * @tbl:       the &struct bucket_table
+ * @hash:      the hash value / bucket index
+ * @member:    name of the &struct rhash_head within the hashable struct.
+ *
+ * This hash chain list-traversal primitive allows for the looped code to
+ * remove the loop cursor from the list.
+ */
+#define rht_for_each_entry_safe(tpos, pos, next, tbl, hash, member)        \
+       for (pos = rht_dereference_bucket((tbl)->buckets[hash], tbl, hash), \
+            next = !rht_is_a_nulls(pos) ?                                  \
+                      rht_dereference_bucket(pos->next, tbl, hash) : NULL; \
+            (!rht_is_a_nulls(pos)) && rht_entry(tpos, pos, member);        \
+            pos = next,                                                    \
+            next = !rht_is_a_nulls(pos) ?                                  \
+                      rht_dereference_bucket(pos->next, tbl, hash) : NULL)
+
+/**
+ * rht_for_each_rcu_continue - continue iterating over rcu hash chain
+ * @pos:       the &struct rhash_head to use as a loop cursor.
+ * @head:      the previous &struct rhash_head to continue from
+ * @tbl:       the &struct bucket_table
+ * @hash:      the hash value / bucket index
+ *
+ * This hash chain list-traversal primitive may safely run concurrently with
+ * the _rcu mutation primitives such as rhashtable_insert() as long as the
+ * traversal is guarded by rcu_read_lock().
+ */
+#define rht_for_each_rcu_continue(pos, head, tbl, hash)                        \
+       for (({barrier(); }),                                           \
+            pos = rht_dereference_bucket_rcu(head, tbl, hash);         \
+            !rht_is_a_nulls(pos);                                      \
+            pos = rcu_dereference_raw(pos->next))
+
+/**
+ * rht_for_each_rcu - iterate over rcu hash chain
+ * @pos:       the &struct rhash_head to use as a loop cursor.
+ * @tbl:       the &struct bucket_table
+ * @hash:      the hash value / bucket index
+ *
+ * This hash chain list-traversal primitive may safely run concurrently with
+ * the _rcu mutation primitives such as rhashtable_insert() as long as the
+ * traversal is guarded by rcu_read_lock().
+ */
+#define rht_for_each_rcu(pos, tbl, hash)                               \
+       rht_for_each_rcu_continue(pos, (tbl)->buckets[hash], tbl, hash)
+
+/**
+ * rht_for_each_entry_rcu_continue - continue iterating over rcu hash chain
+ * @tpos:      the type * to use as a loop cursor.
+ * @pos:       the &struct rhash_head to use as a loop cursor.
+ * @head:      the previous &struct rhash_head to continue from
+ * @tbl:       the &struct bucket_table
+ * @hash:      the hash value / bucket index
+ * @member:    name of the &struct rhash_head within the hashable struct.
+ *
+ * This hash chain list-traversal primitive may safely run concurrently with
+ * the _rcu mutation primitives such as rhashtable_insert() as long as the
+ * traversal is guarded by rcu_read_lock().
+ */
+#define rht_for_each_entry_rcu_continue(tpos, pos, head, tbl, hash, member) \
+       for (({barrier(); }),                                               \
+            pos = rht_dereference_bucket_rcu(head, tbl, hash);             \
+            (!rht_is_a_nulls(pos)) && rht_entry(tpos, pos, member);        \
+            pos = rht_dereference_bucket_rcu(pos->next, tbl, hash))
+
+/**
+ * rht_for_each_entry_rcu - iterate over rcu hash chain of given type
+ * @tpos:      the type * to use as a loop cursor.
+ * @pos:       the &struct rhash_head to use as a loop cursor.
+ * @tbl:       the &struct bucket_table
+ * @hash:      the hash value / bucket index
+ * @member:    name of the &struct rhash_head within the hashable struct.
+ *
+ * This hash chain list-traversal primitive may safely run concurrently with
+ * the _rcu mutation primitives such as rhashtable_insert() as long as the
+ * traversal is guarded by rcu_read_lock().
+ */
+#define rht_for_each_entry_rcu(tpos, pos, tbl, hash, member)           \
+       rht_for_each_entry_rcu_continue(tpos, pos, (tbl)->buckets[hash],\
+                                       tbl, hash, member)
+
+/**
+ * rhl_for_each_rcu - iterate over rcu hash table list
+ * @pos:       the &struct rlist_head to use as a loop cursor.
+ * @list:      the head of the list
+ *
+ * This hash chain list-traversal primitive should be used on the
+ * list returned by rhltable_lookup.
+ */
+#define rhl_for_each_rcu(pos, list)                                    \
+       for (pos = list; pos; pos = rcu_dereference_raw(pos->next))
+
+/**
+ * rhl_for_each_entry_rcu - iterate over rcu hash table list of given type
+ * @tpos:      the type * to use as a loop cursor.
+ * @pos:       the &struct rlist_head to use as a loop cursor.
+ * @list:      the head of the list
+ * @member:    name of the &struct rlist_head within the hashable struct.
+ *
+ * This hash chain list-traversal primitive should be used on the
+ * list returned by rhltable_lookup.
+ */
+#define rhl_for_each_entry_rcu(tpos, pos, list, member)                        \
+       for (pos = list; pos && rht_entry(tpos, pos, member);           \
+            pos = rcu_dereference_raw(pos->next))
+
+static inline int rhashtable_compare(struct rhashtable_compare_arg *arg,
+                                    const void *obj)
+{
+       struct rhashtable *ht = arg->ht;
+       const char *ptr = obj;
+
+       return memcmp(ptr + ht->p.key_offset, arg->key, ht->p.key_len);
+}
+
+/* Internal function, do not use. */
+static inline struct rhash_head *__rhashtable_lookup(
+       struct rhashtable *ht, const void *key,
+       const struct rhashtable_params params)
+{
+       struct rhashtable_compare_arg arg = {
+               .ht = ht,
+               .key = key,
+       };
+       const struct bucket_table *tbl;
+       struct rhash_head *he;
+       unsigned int hash;
+
+       tbl = rht_dereference_rcu(ht->tbl, ht);
+restart:
+       hash = rht_key_hashfn(ht, tbl, key, params);
+       rht_for_each_rcu(he, tbl, hash) {
+               if (params.obj_cmpfn ?
+                   params.obj_cmpfn(&arg, rht_obj(ht, he)) :
+                   rhashtable_compare(&arg, rht_obj(ht, he)))
+                       continue;
+               return he;
+       }
+
+       /* Ensure we see any new tables. */
+       smp_rmb();
+
+       tbl = rht_dereference_rcu(tbl->future_tbl, ht);
+       if (unlikely(tbl))
+               goto restart;
+
+       return NULL;
+}
+
+/**
+ * rhashtable_lookup - search hash table
+ * @ht:                hash table
+ * @key:       the pointer to the key
+ * @params:    hash table parameters
+ *
+ * Computes the hash value for the key and traverses the bucket chain looking
+ * for a entry with an identical key. The first matching entry is returned.
+ *
+ * This must only be called under the RCU read lock.
+ *
+ * Returns the first entry on which the compare function returned true.
+ */
+static inline void *rhashtable_lookup(
+       struct rhashtable *ht, const void *key,
+       const struct rhashtable_params params)
+{
+       struct rhash_head *he = __rhashtable_lookup(ht, key, params);
+
+       return he ? rht_obj(ht, he) : NULL;
+}
+
+/**
+ * rhashtable_lookup_fast - search hash table, without RCU read lock
+ * @ht:                hash table
+ * @key:       the pointer to the key
+ * @params:    hash table parameters
+ *
+ * Computes the hash value for the key and traverses the bucket chain looking
+ * for a entry with an identical key. The first matching entry is returned.
+ *
+ * Only use this function when you have other mechanisms guaranteeing
+ * that the object won't go away after the RCU read lock is released.
+ *
+ * Returns the first entry on which the compare function returned true.
+ */
+static inline void *rhashtable_lookup_fast(
+       struct rhashtable *ht, const void *key,
+       const struct rhashtable_params params)
+{
+       void *obj;
+
+       rcu_read_lock();
+       obj = rhashtable_lookup(ht, key, params);
+       rcu_read_unlock();
+
+       return obj;
+}
+
+/**
+ * rhltable_lookup - search hash list table
+ * @hlt:       hash table
+ * @key:       the pointer to the key
+ * @params:    hash table parameters
+ *
+ * Computes the hash value for the key and traverses the bucket chain looking
+ * for a entry with an identical key.  All matching entries are returned
+ * in a list.
+ *
+ * This must only be called under the RCU read lock.
+ *
+ * Returns the list of entries that match the given key.
+ */
+static inline struct rhlist_head *rhltable_lookup(
+       struct rhltable *hlt, const void *key,
+       const struct rhashtable_params params)
+{
+       struct rhash_head *he = __rhashtable_lookup(&hlt->ht, key, params);
+
+       return he ? container_of(he, struct rhlist_head, rhead) : NULL;
+}
+
+/* Internal function, please use rhashtable_insert_fast() instead. This
+ * function returns the existing element already in hashes in there is a clash,
+ * otherwise it returns an error via ERR_PTR().
+ */
+static inline void *__rhashtable_insert_fast(
+       struct rhashtable *ht, const void *key, struct rhash_head *obj,
+       const struct rhashtable_params params, bool rhlist)
+{
+       struct rhashtable_compare_arg arg = {
+               .ht = ht,
+               .key = key,
+       };
+       struct rhash_head __rcu **pprev;
+       struct bucket_table *tbl;
+       struct rhash_head *head;
+       spinlock_t *lock;
+       unsigned int hash;
+       int elasticity;
+       void *data;
+
+       rcu_read_lock();
+
+       tbl = rht_dereference_rcu(ht->tbl, ht);
+       hash = rht_head_hashfn(ht, tbl, obj, params);
+       lock = rht_bucket_lock(tbl, hash);
+       spin_lock_bh(lock);
+
+       if (unlikely(rht_dereference_bucket(tbl->future_tbl, tbl, hash))) {
+slow_path:
+               spin_unlock_bh(lock);
+               rcu_read_unlock();
+               return rhashtable_insert_slow(ht, key, obj);
+       }
+
+       elasticity = ht->elasticity;
+       pprev = &tbl->buckets[hash];
+       rht_for_each(head, tbl, hash) {
+               struct rhlist_head *plist;
+               struct rhlist_head *list;
+
+               elasticity--;
+               if (!key ||
+                   (params.obj_cmpfn ?
+                    params.obj_cmpfn(&arg, rht_obj(ht, head)) :
+                    rhashtable_compare(&arg, rht_obj(ht, head)))) {
+                       pprev = &head->next;
+                       continue;
+               }
+
+               data = rht_obj(ht, head);
+
+               if (!rhlist)
+                       goto out;
+
+
+               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);
+
+               goto good;
+       }
+
+       if (elasticity <= 0)
+               goto slow_path;
+
+       data = ERR_PTR(-E2BIG);
+       if (unlikely(rht_grow_above_max(ht, tbl)))
+               goto out;
+
+       if (unlikely(rht_grow_above_100(ht, tbl)))
+               goto slow_path;
+
+       head = rht_dereference_bucket(tbl->buckets[hash], tbl, hash);
+
+       RCU_INIT_POINTER(obj->next, head);
+       if (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);
+
+good:
+       data = NULL;
+
+out:
+       spin_unlock_bh(lock);
+       rcu_read_unlock();
+
+       return data;
+}
+
+/**
+ * rhashtable_insert_fast - insert object into hash table
+ * @ht:                hash table
+ * @obj:       pointer to hash head inside object
+ * @params:    hash table parameters
+ *
+ * Will take a per bucket spinlock to protect against mutual mutations
+ * on the same bucket. Multiple insertions may occur in parallel unless
+ * they map to the same bucket lock.
+ *
+ * It is safe to call this function from atomic context.
+ *
+ * Will trigger an automatic deferred table resizing if the size grows
+ * beyond the watermark indicated by grow_decision() which can be passed
+ * to rhashtable_init().
+ */
+static inline int rhashtable_insert_fast(
+       struct rhashtable *ht, struct rhash_head *obj,
+       const struct rhashtable_params params)
+{
+       void *ret;
+
+       ret = __rhashtable_insert_fast(ht, NULL, obj, params, false);
+       if (IS_ERR(ret))
+               return PTR_ERR(ret);
+
+       return ret == NULL ? 0 : -EEXIST;
+}
+
+/**
+ * rhltable_insert_key - insert object into hash list table
+ * @hlt:       hash list table
+ * @key:       the pointer to the key
+ * @list:      pointer to hash list head inside object
+ * @params:    hash table parameters
+ *
+ * Will take a per bucket spinlock to protect against mutual mutations
+ * on the same bucket. Multiple insertions may occur in parallel unless
+ * they map to the same bucket lock.
+ *
+ * It is safe to call this function from atomic context.
+ *
+ * Will trigger an automatic deferred table resizing if the size grows
+ * beyond the watermark indicated by grow_decision() which can be passed
+ * to rhashtable_init().
+ */
+static inline int rhltable_insert_key(
+       struct rhltable *hlt, const void *key, struct rhlist_head *list,
+       const struct rhashtable_params params)
+{
+       return PTR_ERR(__rhashtable_insert_fast(&hlt->ht, key, &list->rhead,
+                                               params, true));
+}
+
+/**
+ * rhltable_insert - insert object into hash list table
+ * @hlt:       hash list table
+ * @list:      pointer to hash list head inside object
+ * @params:    hash table parameters
+ *
+ * Will take a per bucket spinlock to protect against mutual mutations
+ * on the same bucket. Multiple insertions may occur in parallel unless
+ * they map to the same bucket lock.
+ *
+ * It is safe to call this function from atomic context.
+ *
+ * Will trigger an automatic deferred table resizing if the size grows
+ * beyond the watermark indicated by grow_decision() which can be passed
+ * to rhashtable_init().
+ */
+static inline int rhltable_insert(
+       struct rhltable *hlt, struct rhlist_head *list,
+       const struct rhashtable_params params)
+{
+       const char *key = rht_obj(&hlt->ht, &list->rhead);
+
+       key += params.key_offset;
+
+       return rhltable_insert_key(hlt, key, list, params);
+}
+
+/**
+ * rhashtable_lookup_insert_fast - lookup and insert object into hash table
+ * @ht:                hash table
+ * @obj:       pointer to hash head inside object
+ * @params:    hash table parameters
+ *
+ * Locks down the bucket chain in both the old and new table if a resize
+ * is in progress to ensure that writers can't remove from the old table
+ * and can't insert to the new table during the atomic operation of search
+ * and insertion. Searches for duplicates in both the old and new table if
+ * a resize is in progress.
+ *
+ * This lookup function may only be used for fixed key hash table (key_len
+ * parameter set). It will BUG() if used inappropriately.
+ *
+ * It is safe to call this function from atomic context.
+ *
+ * Will trigger an automatic deferred table resizing if the size grows
+ * beyond the watermark indicated by grow_decision() which can be passed
+ * to rhashtable_init().
+ */
+static inline int rhashtable_lookup_insert_fast(
+       struct rhashtable *ht, struct rhash_head *obj,
+       const struct rhashtable_params params)
+{
+       const char *key = rht_obj(ht, obj);
+       void *ret;
+
+       BUG_ON(ht->p.obj_hashfn);
+
+       ret = __rhashtable_insert_fast(ht, key + ht->p.key_offset, obj, params,
+                                      false);
+       if (IS_ERR(ret))
+               return PTR_ERR(ret);
+
+       return ret == NULL ? 0 : -EEXIST;
+}
+
+/**
+ * rhashtable_lookup_insert_key - search and insert object to hash table
+ *                               with explicit key
+ * @ht:                hash table
+ * @key:       key
+ * @obj:       pointer to hash head inside object
+ * @params:    hash table parameters
+ *
+ * Locks down the bucket chain in both the old and new table if a resize
+ * is in progress to ensure that writers can't remove from the old table
+ * and can't insert to the new table during the atomic operation of search
+ * and insertion. Searches for duplicates in both the old and new table if
+ * a resize is in progress.
+ *
+ * Lookups may occur in parallel with hashtable mutations and resizing.
+ *
+ * Will trigger an automatic deferred table resizing if the size grows
+ * beyond the watermark indicated by grow_decision() which can be passed
+ * to rhashtable_init().
+ *
+ * Returns zero on success.
+ */
+static inline int rhashtable_lookup_insert_key(
+       struct rhashtable *ht, const void *key, struct rhash_head *obj,
+       const struct rhashtable_params params)
+{
+       void *ret;
+
+       BUG_ON(!ht->p.obj_hashfn || !key);
+
+       ret = __rhashtable_insert_fast(ht, key, obj, params, false);
+       if (IS_ERR(ret))
+               return PTR_ERR(ret);
+
+       return ret == NULL ? 0 : -EEXIST;
+}
+
+/**
+ * rhashtable_lookup_get_insert_key - lookup and insert object into hash table
+ * @ht:                hash table
+ * @obj:       pointer to hash head inside object
+ * @params:    hash table parameters
+ * @data:      pointer to element data already in hashes
+ *
+ * Just like rhashtable_lookup_insert_key(), but this function returns the
+ * object if it exists, NULL if it does not and the insertion was successful,
+ * and an ERR_PTR otherwise.
+ */
+static inline void *rhashtable_lookup_get_insert_key(
+       struct rhashtable *ht, const void *key, struct rhash_head *obj,
+       const struct rhashtable_params params)
+{
+       BUG_ON(!ht->p.obj_hashfn || !key);
+
+       return __rhashtable_insert_fast(ht, key, obj, params, false);
+}
+
+/* Internal function, please use rhashtable_remove_fast() instead */
+static inline int __rhashtable_remove_fast_one(
+       struct rhashtable *ht, struct bucket_table *tbl,
+       struct rhash_head *obj, const struct rhashtable_params params,
+       bool rhlist)
+{
+       struct rhash_head __rcu **pprev;
+       struct rhash_head *he;
+       spinlock_t * lock;
+       unsigned int hash;
+       int err = -ENOENT;
+
+       hash = rht_head_hashfn(ht, tbl, obj, params);
+       lock = rht_bucket_lock(tbl, hash);
+
+       spin_lock_bh(lock);
+
+       pprev = &tbl->buckets[hash];
+       rht_for_each(he, tbl, hash) {
+               struct rhlist_head *list;
+
+               list = container_of(he, struct rhlist_head, rhead);
+
+               if (he != obj) {
+                       struct rhlist_head __rcu **lpprev;
+
+                       pprev = &he->next;
+
+                       if (!rhlist)
+                               continue;
+
+                       do {
+                               lpprev = &list->next;
+                               list = rht_dereference_bucket(list->next,
+                                                             tbl, hash);
+                       } while (list && obj != &list->rhead);
+
+                       if (!list)
+                               continue;
+
+                       list = rht_dereference_bucket(list->next, tbl, hash);
+                       RCU_INIT_POINTER(*lpprev, list);
+                       err = 0;
+                       break;
+               }
+
+               obj = rht_dereference_bucket(obj->next, tbl, hash);
+               err = 1;
+
+               if (rhlist) {
+                       list = rht_dereference_bucket(list->next, tbl, hash);
+                       if (list) {
+                               RCU_INIT_POINTER(list->rhead.next, obj);
+                               obj = &list->rhead;
+                               err = 0;
+                       }
+               }
+
+               rcu_assign_pointer(*pprev, obj);
+               break;
+       }
+
+       spin_unlock_bh(lock);
+
+       if (err > 0) {
+               atomic_dec(&ht->nelems);
+               if (unlikely(ht->p.automatic_shrinking &&
+                            rht_shrink_below_30(ht, tbl)))
+                       schedule_work(&ht->run_work);
+               err = 0;
+       }
+
+       return err;
+}
+
+/* Internal function, please use rhashtable_remove_fast() instead */
+static inline int __rhashtable_remove_fast(
+       struct rhashtable *ht, struct rhash_head *obj,
+       const struct rhashtable_params params, bool rhlist)
+{
+       struct bucket_table *tbl;
+       int err;
+
+       rcu_read_lock();
+
+       tbl = rht_dereference_rcu(ht->tbl, ht);
+
+       /* Because we have already taken (and released) the bucket
+        * lock in old_tbl, if we find that future_tbl is not yet
+        * visible then that guarantees the entry to still be in
+        * the old tbl if it exists.
+        */
+       while ((err = __rhashtable_remove_fast_one(ht, tbl, obj, params,
+                                                  rhlist)) &&
+              (tbl = rht_dereference_rcu(tbl->future_tbl, ht)))
+               ;
+
+       rcu_read_unlock();
+
+       return err;
+}
+
+/**
+ * rhashtable_remove_fast - remove object from hash table
+ * @ht:                hash table
+ * @obj:       pointer to hash head inside object
+ * @params:    hash table parameters
+ *
+ * Since the hash chain is single linked, the removal operation needs to
+ * walk the bucket chain upon removal. The removal operation is thus
+ * considerable slow if the hash table is not correctly sized.
+ *
+ * Will automatically shrink the table via rhashtable_expand() if the
+ * shrink_decision function specified at rhashtable_init() returns true.
+ *
+ * Returns zero on success, -ENOENT if the entry could not be found.
+ */
+static inline int rhashtable_remove_fast(
+       struct rhashtable *ht, struct rhash_head *obj,
+       const struct rhashtable_params params)
+{
+       return __rhashtable_remove_fast(ht, obj, params, false);
+}
+
+/**
+ * rhltable_remove - remove object from hash list table
+ * @hlt:       hash list table
+ * @list:      pointer to hash list head inside object
+ * @params:    hash table parameters
+ *
+ * Since the hash chain is single linked, the removal operation needs to
+ * walk the bucket chain upon removal. The removal operation is thus
+ * considerable slow if the hash table is not correctly sized.
+ *
+ * Will automatically shrink the table via rhashtable_expand() if the
+ * shrink_decision function specified at rhashtable_init() returns true.
+ *
+ * Returns zero on success, -ENOENT if the entry could not be found.
+ */
+static inline int rhltable_remove(
+       struct rhltable *hlt, struct rhlist_head *list,
+       const struct rhashtable_params params)
+{
+       return __rhashtable_remove_fast(&hlt->ht, &list->rhead, params, true);
+}
+
+/* Internal function, please use rhashtable_replace_fast() instead */
+static inline int __rhashtable_replace_fast(
+       struct rhashtable *ht, struct bucket_table *tbl,
+       struct rhash_head *obj_old, struct rhash_head *obj_new,
+       const struct rhashtable_params params)
+{
+       struct rhash_head __rcu **pprev;
+       struct rhash_head *he;
+       spinlock_t *lock;
+       unsigned int hash;
+       int err = -ENOENT;
+
+       /* Minimally, the old and new objects must have same hash
+        * (which should mean identifiers are the same).
+        */
+       hash = rht_head_hashfn(ht, tbl, obj_old, params);
+       if (hash != rht_head_hashfn(ht, tbl, obj_new, params))
+               return -EINVAL;
+
+       lock = rht_bucket_lock(tbl, hash);
+
+       spin_lock_bh(lock);
+
+       pprev = &tbl->buckets[hash];
+       rht_for_each(he, tbl, hash) {
+               if (he != obj_old) {
+                       pprev = &he->next;
+                       continue;
+               }
+
+               rcu_assign_pointer(obj_new->next, obj_old->next);
+               rcu_assign_pointer(*pprev, obj_new);
+               err = 0;
+               break;
+       }
+
+       spin_unlock_bh(lock);
+
+       return err;
+}
+
+/**
+ * rhashtable_replace_fast - replace an object in hash table
+ * @ht:                hash table
+ * @obj_old:   pointer to hash head inside object being replaced
+ * @obj_new:   pointer to hash head inside object which is new
+ * @params:    hash table parameters
+ *
+ * Replacing an object doesn't affect the number of elements in the hash table
+ * or bucket, so we don't need to worry about shrinking or expanding the
+ * table here.
+ *
+ * Returns zero on success, -ENOENT if the entry could not be found,
+ * -EINVAL if hash is not the same for the old and new objects.
+ */
+static inline int rhashtable_replace_fast(
+       struct rhashtable *ht, struct rhash_head *obj_old,
+       struct rhash_head *obj_new,
+       const struct rhashtable_params params)
+{
+       struct bucket_table *tbl;
+       int err;
+
+       rcu_read_lock();
+
+       tbl = rht_dereference_rcu(ht->tbl, ht);
+
+       /* Because we have already taken (and released) the bucket
+        * lock in old_tbl, if we find that future_tbl is not yet
+        * visible then that guarantees the entry to still be in
+        * the old tbl if it exists.
+        */
+       while ((err = __rhashtable_replace_fast(ht, tbl, obj_old,
+                                               obj_new, params)) &&
+              (tbl = rht_dereference_rcu(tbl->future_tbl, ht)))
+               ;
+
+       rcu_read_unlock();
+
+       return err;
+}
+
+/* Obsolete function, do not use in new code. */
+static inline int rhashtable_walk_init(struct rhashtable *ht,
+                                      struct rhashtable_iter *iter, gfp_t gfp)
+{
+       rhashtable_walk_enter(ht, iter);
+       return 0;
+}
+
+/**
+ * rhltable_walk_enter - Initialise an iterator
+ * @hlt:       Table to walk over
+ * @iter:      Hash table Iterator
+ *
+ * This function prepares a hash table walk.
+ *
+ * Note that if you restart a walk after rhashtable_walk_stop you
+ * may see the same object twice.  Also, you may miss objects if
+ * there are removals in between rhashtable_walk_stop and the next
+ * call to rhashtable_walk_start.
+ *
+ * For a completely stable walk you should construct your own data
+ * structure outside the hash table.
+ *
+ * This function may sleep so you must not call it from interrupt
+ * context or with spin locks held.
+ *
+ * You must call rhashtable_walk_exit after this function returns.
+ */
+static inline void rhltable_walk_enter(struct rhltable *hlt,
+                                      struct rhashtable_iter *iter)
+{
+       return rhashtable_walk_enter(&hlt->ht, iter);
+}
+
+/**
+ * rhltable_free_and_destroy - free elements and destroy hash list table
+ * @hlt:       the hash list table to destroy
+ * @free_fn:   callback to release resources of element
+ * @arg:       pointer passed to free_fn
+ *
+ * See documentation for rhashtable_free_and_destroy.
+ */
+static inline void rhltable_free_and_destroy(struct rhltable *hlt,
+                                            void (*free_fn)(void *ptr,
+                                                            void *arg),
+                                            void *arg)
+{
+       return rhashtable_free_and_destroy(&hlt->ht, free_fn, arg);
+}
+
+static inline void rhltable_destroy(struct rhltable *hlt)
+{
+       return rhltable_free_and_destroy(hlt, NULL, NULL);
+}
+#endif /* HAVE_RHLTABLE */
+
+#endif /* __COMPAT_LINUX_RHASHTABLE_H */