Fix handling of systems without sysconf nr possible cpu support
[urcu.git] / rculfhash.c
index 6ac53ea5bae5b1f3fea8ec1ecf98f5939110f25e..638432a90b3121ef3c4b1a9d870680e1f6ba98f9 100644 (file)
 #include <urcu-call-rcu.h>
 #include <urcu/arch.h>
 #include <urcu/uatomic.h>
-#include <urcu/jhash.h>
 #include <urcu/compiler.h>
 #include <urcu/rculfhash.h>
 #include <stdio.h>
 #define MAX_TABLE_ORDER                        64
 #endif
 
+/*
+ * Minimum number of dummy nodes to touch per thread to parallelize grow/shrink.
+ */
+#define MIN_PARTITION_PER_THREAD_ORDER 12
+#define MIN_PARTITION_PER_THREAD       (1UL << MIN_PARTITION_PER_THREAD_ORDER)
+
 #ifndef min
 #define min(a, b)      ((a) < (b) ? (a) : (b))
 #endif
 
 /*
  * The removed flag needs to be updated atomically with the pointer.
+ * It indicates that no node must attach to the node scheduled for
+ * removal, and that node garbage collection must be performed.
  * The dummy flag does not require to be updated atomically with the
  * pointer, but it is added as a pointer low bit flag to save space.
  */
 #define END_VALUE              NULL
 
 struct ht_items_count {
-       unsigned long add, remove;
+       unsigned long add, del;
 } __attribute__((aligned(CAA_CACHE_LINE_SIZE)));
 
 struct rcu_level {
@@ -235,7 +242,10 @@ struct cds_lfht {
        void (*cds_lfht_rcu_read_unlock)(void);
        void (*cds_lfht_rcu_thread_offline)(void);
        void (*cds_lfht_rcu_thread_online)(void);
-       unsigned long count;            /* global approximate item count */
+       void (*cds_lfht_rcu_register_thread)(void);
+       void (*cds_lfht_rcu_unregister_thread)(void);
+       pthread_attr_t *resize_attr;    /* Resize threads attributes */
+       long count;                     /* global approximate item count */
        struct ht_items_count *percpu_count;    /* per-cpu item count */
 };
 
@@ -244,11 +254,25 @@ struct rcu_resize_work {
        struct cds_lfht *ht;
 };
 
+struct partition_resize_work {
+       struct rcu_head head;
+       struct cds_lfht *ht;
+       unsigned long i, start, len;
+       void (*fct)(struct cds_lfht *ht, unsigned long i,
+                   unsigned long start, unsigned long len);
+};
+
+enum add_mode {
+       ADD_DEFAULT = 0,
+       ADD_UNIQUE = 1,
+       ADD_REPLACE = 2,
+};
+
 static
 struct cds_lfht_node *_cds_lfht_add(struct cds_lfht *ht,
                                unsigned long size,
                                struct cds_lfht_node *node,
-                               int unique, int dummy);
+                               enum add_mode mode, int dummy);
 
 /*
  * Algorithm to reverse bits in a word by lookup table, extended to
@@ -531,7 +555,7 @@ void ht_count_add(struct cds_lfht *ht, unsigned long size)
                return;
        percpu_count = uatomic_add_return(&ht->percpu_count[cpu].add, 1);
        if (unlikely(!(percpu_count & ((1UL << COUNT_COMMIT_ORDER) - 1)))) {
-               unsigned long count;
+               long count;
 
                dbg_printf("add percpu %lu\n", percpu_count);
                count = uatomic_add_return(&ht->count,
@@ -540,7 +564,7 @@ void ht_count_add(struct cds_lfht *ht, unsigned long size)
                if (!(count & (count - 1))) {
                        if ((count >> CHAIN_LEN_RESIZE_THRESHOLD) < size)
                                return;
-                       dbg_printf("add set global %lu\n", count);
+                       dbg_printf("add set global %ld\n", count);
                        cds_lfht_resize_lazy_count(ht, size,
                                count >> (CHAIN_LEN_TARGET - 1));
                }
@@ -548,7 +572,7 @@ void ht_count_add(struct cds_lfht *ht, unsigned long size)
 }
 
 static
-void ht_count_remove(struct cds_lfht *ht, unsigned long size)
+void ht_count_del(struct cds_lfht *ht, unsigned long size)
 {
        unsigned long percpu_count;
        int cpu;
@@ -558,18 +582,24 @@ void ht_count_remove(struct cds_lfht *ht, unsigned long size)
        cpu = ht_get_cpu();
        if (unlikely(cpu < 0))
                return;
-       percpu_count = uatomic_add_return(&ht->percpu_count[cpu].remove, -1);
+       percpu_count = uatomic_add_return(&ht->percpu_count[cpu].del, 1);
        if (unlikely(!(percpu_count & ((1UL << COUNT_COMMIT_ORDER) - 1)))) {
-               unsigned long count;
+               long count;
 
-               dbg_printf("remove percpu %lu\n", percpu_count);
+               dbg_printf("del percpu %lu\n", percpu_count);
                count = uatomic_add_return(&ht->count,
                                           -(1UL << COUNT_COMMIT_ORDER));
                /* If power of 2 */
                if (!(count & (count - 1))) {
                        if ((count >> CHAIN_LEN_RESIZE_THRESHOLD) >= size)
                                return;
-                       dbg_printf("remove set global %lu\n", count);
+                       dbg_printf("del set global %ld\n", count);
+                       /*
+                        * Don't shrink table if the number of nodes is below a
+                        * certain threshold.
+                        */
+                       if (count < (1UL << COUNT_COMMIT_ORDER) * (nr_cpus_mask + 1))
+                               return;
                        cds_lfht_resize_lazy_count(ht, size,
                                count >> (CHAIN_LEN_TARGET - 1));
                }
@@ -578,7 +608,7 @@ void ht_count_remove(struct cds_lfht *ht, unsigned long size)
 
 #else /* #if defined(HAVE_SCHED_GETCPU) && defined(HAVE_SYSCONF) */
 
-static const long nr_cpus_mask = -1;
+static const long nr_cpus_mask = -2;
 
 static
 struct ht_items_count *alloc_per_cpu_items_count(void)
@@ -597,7 +627,7 @@ void ht_count_add(struct cds_lfht *ht, unsigned long size)
 }
 
 static
-void ht_count_remove(struct cds_lfht *ht, unsigned long size)
+void ht_count_del(struct cds_lfht *ht, unsigned long size)
 {
 }
 
@@ -730,19 +760,95 @@ void _cds_lfht_gc_bucket(struct cds_lfht_node *dummy, struct cds_lfht_node *node
                        new_next = flag_dummy(clear_flag(next));
                else
                        new_next = clear_flag(next);
+               if (is_removed(iter))
+                       new_next = flag_removed(new_next);
                (void) uatomic_cmpxchg(&iter_prev->p.next, iter, new_next);
        }
        return;
 }
 
+static
+int _cds_lfht_replace(struct cds_lfht *ht, unsigned long size,
+               struct cds_lfht_node *old_node,
+               struct cds_lfht_node *ret_next,
+               struct cds_lfht_node *new_node)
+{
+       struct cds_lfht_node *dummy, *old_next;
+       struct _cds_lfht_node *lookup;
+       int flagged = 0;
+       unsigned long hash, index, order;
+
+       if (!old_node)  /* Return -ENOENT if asked to replace NULL node */
+               goto end;
+
+       assert(!is_removed(old_node));
+       assert(!is_dummy(old_node));
+       assert(!is_removed(new_node));
+       assert(!is_dummy(new_node));
+       assert(new_node != old_node);
+       do {
+               /* Insert after node to be replaced */
+               old_next = ret_next;
+               if (is_removed(old_next)) {
+                       /*
+                        * Too late, the old node has been removed under us
+                        * between lookup and replace. Fail.
+                        */
+                       goto end;
+               }
+               assert(!is_dummy(old_next));
+               assert(new_node != clear_flag(old_next));
+               new_node->p.next = clear_flag(old_next);
+               /*
+                * Here is the whole trick for lock-free replace: we add
+                * the replacement node _after_ the node we want to
+                * replace by atomically setting its next pointer at the
+                * same time we set its removal flag. Given that
+                * the lookups/get next use an iterator aware of the
+                * next pointer, they will either skip the old node due
+                * to the removal flag and see the new node, or use
+                * the old node, but will not see the new one.
+                */
+               ret_next = uatomic_cmpxchg(&old_node->p.next,
+                             old_next, flag_removed(new_node));
+       } while (ret_next != old_next);
+
+       /* We performed the replacement. */
+       flagged = 1;
+
+       /*
+        * Ensure that the old node is not visible to readers anymore:
+        * lookup for the node, and remove it (along with any other
+        * logically removed node) if found.
+        */
+       hash = bit_reverse_ulong(old_node->p.reverse_hash);
+       assert(size > 0);
+       index = hash & (size - 1);
+       order = get_count_order_ulong(index + 1);
+       lookup = &ht->t.tbl[order]->nodes[index & (!order ? 0 : ((1UL << (order - 1)) - 1))];
+       dummy = (struct cds_lfht_node *) lookup;
+       _cds_lfht_gc_bucket(dummy, new_node);
+end:
+       /*
+        * Only the flagging action indicated that we (and no other)
+        * replaced the node from the hash table.
+        */
+       if (flagged) {
+               assert(is_removed(rcu_dereference(old_node->p.next)));
+               return 0;
+       } else {
+               return -ENOENT;
+       }
+}
+
 static
 struct cds_lfht_node *_cds_lfht_add(struct cds_lfht *ht,
                                unsigned long size,
                                struct cds_lfht_node *node,
-                               int unique, int dummy)
+                               enum add_mode mode, int dummy)
 {
        struct cds_lfht_node *iter_prev, *iter, *next, *new_node, *new_next,
-                       *dummy_node;
+                       *dummy_node, *return_node;
        struct _cds_lfht_node *lookup;
        unsigned long hash, index, order;
 
@@ -776,12 +882,16 @@ struct cds_lfht_node *_cds_lfht_add(struct cds_lfht *ht,
                        next = rcu_dereference(clear_flag(iter)->p.next);
                        if (unlikely(is_removed(next)))
                                goto gc_node;
-                       if (unique
+                       if ((mode == ADD_UNIQUE || mode == ADD_REPLACE)
                            && !is_dummy(next)
                            && !ht->compare_fct(node->key, node->key_len,
                                                clear_flag(iter)->key,
-                                               clear_flag(iter)->key_len))
-                               return clear_flag(iter);
+                                               clear_flag(iter)->key_len)) {
+                               if (mode == ADD_UNIQUE)
+                                       return clear_flag(iter);
+                               else /* mode == ADD_REPLACE */
+                                       goto replace;
+                       }
                        /* Only account for identical reverse hash once */
                        if (iter_prev->p.reverse_hash != clear_flag(iter)->p.reverse_hash
                            && !is_dummy(next))
@@ -789,6 +899,7 @@ struct cds_lfht_node *_cds_lfht_add(struct cds_lfht *ht,
                        iter_prev = clear_flag(iter);
                        iter = next;
                }
+
        insert:
                assert(node != clear_flag(iter));
                assert(!is_removed(iter_prev));
@@ -803,10 +914,26 @@ struct cds_lfht_node *_cds_lfht_add(struct cds_lfht *ht,
                else
                        new_node = node;
                if (uatomic_cmpxchg(&iter_prev->p.next, iter,
-                                   new_node) != iter)
+                                   new_node) != iter) {
                        continue;       /* retry */
-               else
+               } else {
+                       if (mode == ADD_REPLACE)
+                               return_node = NULL;
+                       else    /* ADD_DEFAULT and ADD_UNIQUE */
+                               return_node = node;
                        goto gc_end;
+               }
+
+       replace:
+
+               if (!_cds_lfht_replace(ht, size, clear_flag(iter), next,
+                                   node)) {
+                       return_node = clear_flag(iter);
+                       goto end;       /* gc already done */
+               } else {
+                       continue;       /* retry */
+               }
+
        gc_node:
                assert(!is_removed(iter));
                if (is_dummy(iter))
@@ -823,11 +950,12 @@ gc_end:
        lookup = &ht->t.tbl[order]->nodes[index & (!order ? 0 : ((1UL << (order - 1)) - 1))];
        dummy_node = (struct cds_lfht_node *) lookup;
        _cds_lfht_gc_bucket(dummy_node, node);
-       return node;
+end:
+       return return_node;
 }
 
 static
-int _cds_lfht_remove(struct cds_lfht *ht, unsigned long size,
+int _cds_lfht_del(struct cds_lfht *ht, unsigned long size,
                struct cds_lfht_node *node,
                int dummy_removal)
 {
@@ -836,11 +964,16 @@ int _cds_lfht_remove(struct cds_lfht *ht, unsigned long size,
        int flagged = 0;
        unsigned long hash, index, order;
 
+       if (!node)      /* Return -ENOENT if asked to delete NULL node */
+               goto end;
+
        /* logically delete the node */
        assert(!is_dummy(node));
        assert(!is_removed(node));
        old = rcu_dereference(node->p.next);
        do {
+               struct cds_lfht_node *new_next;
+
                next = old;
                if (unlikely(is_removed(next)))
                        goto end;
@@ -848,8 +981,8 @@ int _cds_lfht_remove(struct cds_lfht *ht, unsigned long size,
                        assert(is_dummy(next));
                else
                        assert(!is_dummy(next));
-               old = uatomic_cmpxchg(&node->p.next, next,
-                                     flag_removed(next));
+               new_next = flag_removed(next);
+               old = uatomic_cmpxchg(&node->p.next, next, new_next);
        } while (old != next);
 
        /* We performed the (logical) deletion. */
@@ -875,8 +1008,65 @@ end:
        if (flagged) {
                assert(is_removed(rcu_dereference(node->p.next)));
                return 0;
-       } else
+       } else {
                return -ENOENT;
+       }
+}
+
+static
+void *partition_resize_thread(void *arg)
+{
+       struct partition_resize_work *work = arg;
+
+       work->ht->cds_lfht_rcu_register_thread();
+       work->fct(work->ht, work->i, work->start, work->len);
+       work->ht->cds_lfht_rcu_unregister_thread();
+       return NULL;
+}
+
+static
+void partition_resize_helper(struct cds_lfht *ht, unsigned long i,
+               unsigned long len,
+               void (*fct)(struct cds_lfht *ht, unsigned long i,
+                       unsigned long start, unsigned long len))
+{
+       unsigned long partition_len;
+       struct partition_resize_work *work;
+       int thread, ret;
+       unsigned long nr_threads;
+       pthread_t *thread_id;
+
+       /*
+        * Note: nr_cpus_mask + 1 is always power of 2.
+        * We spawn just the number of threads we need to satisfy the minimum
+        * partition size, up to the number of CPUs in the system.
+        */
+       if (nr_cpus_mask > 0) {
+               nr_threads = min(nr_cpus_mask + 1,
+                                len >> MIN_PARTITION_PER_THREAD_ORDER);
+       } else {
+               nr_threads = 1;
+       }
+       partition_len = len >> get_count_order_ulong(nr_threads);
+       work = calloc(nr_threads, sizeof(*work));
+       thread_id = calloc(nr_threads, sizeof(*thread_id));
+       assert(work);
+       for (thread = 0; thread < nr_threads; thread++) {
+               work[thread].ht = ht;
+               work[thread].i = i;
+               work[thread].len = partition_len;
+               work[thread].start = thread * partition_len;
+               work[thread].fct = fct;
+               ret = pthread_create(&thread_id[thread], ht->resize_attr,
+                       partition_resize_thread, &work[thread]);
+               assert(!ret);
+       }
+       for (thread = 0; thread < nr_threads; thread++) {
+               ret = pthread_join(thread_id[thread], NULL);
+               assert(!ret);
+       }
+       free(work);
+       free(thread_id);
 }
 
 /*
@@ -884,20 +1074,20 @@ end:
  * reclaim that could be performed by other call_rcu worker threads (ABA
  * problem).
  *
- * TODO: when we reach a certain length, we can split this population phase over
+ * When we reach a certain length, we can split this population phase over
  * many worker threads, based on the number of CPUs available in the system.
  * This should therefore take care of not having the expand lagging behind too
  * many concurrent insertion threads by using the scheduler's ability to
  * schedule dummy node population fairly with insertions.
  */
 static
-void init_table_populate(struct cds_lfht *ht, unsigned long i, unsigned long len)
+void init_table_populate_partition(struct cds_lfht *ht, unsigned long i,
+                                  unsigned long start, unsigned long len)
 {
        unsigned long j;
 
-       ht->cds_lfht_rcu_thread_online();
        ht->cds_lfht_rcu_read_lock();
-       for (j = 0; j < len; j++) {
+       for (j = start; j < start + len; j++) {
                struct cds_lfht_node *new_node =
                        (struct cds_lfht_node *) &ht->t.tbl[i]->nodes[j];
 
@@ -906,12 +1096,25 @@ void init_table_populate(struct cds_lfht *ht, unsigned long i, unsigned long len
                new_node->p.reverse_hash =
                        bit_reverse_ulong(!i ? 0 : (1UL << (i - 1)) + j);
                (void) _cds_lfht_add(ht, !i ? 0 : (1UL << (i - 1)),
-                               new_node, 0, 1);
+                               new_node, ADD_DEFAULT, 1);
                if (CMM_LOAD_SHARED(ht->in_progress_destroy))
                        break;
        }
        ht->cds_lfht_rcu_read_unlock();
-       ht->cds_lfht_rcu_thread_offline();
+}
+
+static
+void init_table_populate(struct cds_lfht *ht, unsigned long i,
+                        unsigned long len)
+{
+       assert(nr_cpus_mask != -1);
+       if (nr_cpus_mask < 0 || len < 2 * MIN_PARTITION_PER_THREAD) {
+               ht->cds_lfht_rcu_thread_online();
+               init_table_populate_partition(ht, i, 0, len);
+               ht->cds_lfht_rcu_thread_offline();
+               return;
+       }
+       partition_resize_helper(ht, i, len, init_table_populate_partition);
 }
 
 static
@@ -935,6 +1138,7 @@ void init_table(struct cds_lfht *ht,
 
                ht->t.tbl[i] = calloc(1, sizeof(struct rcu_level)
                                + (len * sizeof(struct _cds_lfht_node)));
+               assert(ht->t.tbl[i]);
 
                /*
                 * Set all dummy nodes reverse hash values for a level and
@@ -974,19 +1178,19 @@ void init_table(struct cds_lfht *ht,
  * Logical removal and garbage collection can therefore be done in batch or on a
  * node-per-node basis, as long as the guarantee above holds.
  *
- * TODO: when we reach a certain length, we can split this removal over many
- * worker threads, based on the number of CPUs available in the system. This
- * should take care of not letting resize process lag behind too many concurrent
+ * When we reach a certain length, we can split this removal over many worker
+ * threads, based on the number of CPUs available in the system. This should
+ * take care of not letting resize process lag behind too many concurrent
  * updater threads actively inserting into the hash table.
  */
 static
-void remove_table(struct cds_lfht *ht, unsigned long i, unsigned long len)
+void remove_table_partition(struct cds_lfht *ht, unsigned long i,
+                           unsigned long start, unsigned long len)
 {
        unsigned long j;
 
-       ht->cds_lfht_rcu_thread_online();
        ht->cds_lfht_rcu_read_lock();
-       for (j = 0; j < len; j++) {
+       for (j = start; j < start + len; j++) {
                struct cds_lfht_node *fini_node =
                        (struct cds_lfht_node *) &ht->t.tbl[i]->nodes[j];
 
@@ -994,13 +1198,26 @@ void remove_table(struct cds_lfht *ht, unsigned long i, unsigned long len)
                           i, j, !i ? 0 : (1UL << (i - 1)) + j);
                fini_node->p.reverse_hash =
                        bit_reverse_ulong(!i ? 0 : (1UL << (i - 1)) + j);
-               (void) _cds_lfht_remove(ht, !i ? 0 : (1UL << (i - 1)),
+               (void) _cds_lfht_del(ht, !i ? 0 : (1UL << (i - 1)),
                                fini_node, 1);
                if (CMM_LOAD_SHARED(ht->in_progress_destroy))
                        break;
        }
        ht->cds_lfht_rcu_read_unlock();
-       ht->cds_lfht_rcu_thread_offline();
+}
+
+static
+void remove_table(struct cds_lfht *ht, unsigned long i, unsigned long len)
+{
+
+       assert(nr_cpus_mask != -1);
+       if (nr_cpus_mask < 0 || len < 2 * MIN_PARTITION_PER_THREAD) {
+               ht->cds_lfht_rcu_thread_online();
+               remove_table_partition(ht, i, 0, len);
+               ht->cds_lfht_rcu_thread_offline();
+               return;
+       }
+       partition_resize_helper(ht, i, len, remove_table_partition);
 }
 
 static
@@ -1050,7 +1267,7 @@ void fini_table(struct cds_lfht *ht,
        }
 }
 
-struct cds_lfht *cds_lfht_new(cds_lfht_hash_fct hash_fct,
+struct cds_lfht *_cds_lfht_new(cds_lfht_hash_fct hash_fct,
                        cds_lfht_compare_fct compare_fct,
                        unsigned long hash_seed,
                        unsigned long init_size,
@@ -1061,7 +1278,10 @@ struct cds_lfht *cds_lfht_new(cds_lfht_hash_fct hash_fct,
                        void (*cds_lfht_rcu_read_lock)(void),
                        void (*cds_lfht_rcu_read_unlock)(void),
                        void (*cds_lfht_rcu_thread_offline)(void),
-                       void (*cds_lfht_rcu_thread_online)(void))
+                       void (*cds_lfht_rcu_thread_online)(void),
+                       void (*cds_lfht_rcu_register_thread)(void),
+                       void (*cds_lfht_rcu_unregister_thread)(void),
+                       pthread_attr_t *attr)
 {
        struct cds_lfht *ht;
        unsigned long order;
@@ -1070,6 +1290,7 @@ struct cds_lfht *cds_lfht_new(cds_lfht_hash_fct hash_fct,
        if (init_size && (init_size & (init_size - 1)))
                return NULL;
        ht = calloc(1, sizeof(struct cds_lfht));
+       assert(ht);
        ht->hash_fct = hash_fct;
        ht->compare_fct = compare_fct;
        ht->hash_seed = hash_seed;
@@ -1079,6 +1300,9 @@ struct cds_lfht *cds_lfht_new(cds_lfht_hash_fct hash_fct,
        ht->cds_lfht_rcu_read_unlock = cds_lfht_rcu_read_unlock;
        ht->cds_lfht_rcu_thread_offline = cds_lfht_rcu_thread_offline;
        ht->cds_lfht_rcu_thread_online = cds_lfht_rcu_thread_online;
+       ht->cds_lfht_rcu_register_thread = cds_lfht_rcu_register_thread;
+       ht->cds_lfht_rcu_unregister_thread = cds_lfht_rcu_unregister_thread;
+       ht->resize_attr = attr;
        ht->percpu_count = alloc_per_cpu_items_count();
        /* this mutex should not nest in read-side C.S. */
        pthread_mutex_init(&ht->resize_mutex, NULL);
@@ -1093,7 +1317,8 @@ struct cds_lfht *cds_lfht_new(cds_lfht_hash_fct hash_fct,
        return ht;
 }
 
-struct cds_lfht_node *cds_lfht_lookup(struct cds_lfht *ht, void *key, size_t key_len)
+void cds_lfht_lookup(struct cds_lfht *ht, void *key, size_t key_len,
+               struct cds_lfht_iter *iter)
 {
        struct cds_lfht_node *node, *next, *dummy_node;
        struct _cds_lfht_node *lookup;
@@ -1114,11 +1339,11 @@ struct cds_lfht_node *cds_lfht_lookup(struct cds_lfht *ht, void *key, size_t key
        node = clear_flag(node);
        for (;;) {
                if (unlikely(is_end(node))) {
-                       node = NULL;
+                       node = next = NULL;
                        break;
                }
                if (unlikely(node->p.reverse_hash > reverse_hash)) {
-                       node = NULL;
+                       node = next = NULL;
                        break;
                }
                next = rcu_dereference(node->p.next);
@@ -1130,30 +1355,31 @@ struct cds_lfht_node *cds_lfht_lookup(struct cds_lfht *ht, void *key, size_t key
                node = clear_flag(next);
        }
        assert(!node || !is_dummy(rcu_dereference(node->p.next)));
-       return node;
+       iter->node = node;
+       iter->next = next;
 }
 
-struct cds_lfht_node *cds_lfht_next(struct cds_lfht *ht,
-                               struct cds_lfht_node *node)
+void cds_lfht_next_duplicate(struct cds_lfht *ht, struct cds_lfht_iter *iter)
 {
-       struct cds_lfht_node *next;
+       struct cds_lfht_node *node, *next;
        unsigned long reverse_hash;
        void *key;
        size_t key_len;
 
+       node = iter->node;
        reverse_hash = node->p.reverse_hash;
        key = node->key;
        key_len = node->key_len;
-       next = rcu_dereference(node->p.next);
+       next = iter->next;
        node = clear_flag(next);
 
        for (;;) {
                if (unlikely(is_end(node))) {
-                       node = NULL;
+                       node = next = NULL;
                        break;
                }
                if (unlikely(node->p.reverse_hash > reverse_hash)) {
-                       node = NULL;
+                       node = next = NULL;
                        break;
                }
                next = rcu_dereference(node->p.next);
@@ -1165,7 +1391,46 @@ struct cds_lfht_node *cds_lfht_next(struct cds_lfht *ht,
                node = clear_flag(next);
        }
        assert(!node || !is_dummy(rcu_dereference(node->p.next)));
-       return node;
+       iter->node = node;
+       iter->next = next;
+}
+
+void cds_lfht_next(struct cds_lfht *ht, struct cds_lfht_iter *iter)
+{
+       struct cds_lfht_node *node, *next;
+
+       node = iter->node;
+       next = iter->next;
+       node = clear_flag(next);
+
+       for (;;) {
+               if (unlikely(is_end(node))) {
+                       node = next = NULL;
+                       break;
+               }
+               next = rcu_dereference(node->p.next);
+               if (likely(!is_removed(next))
+                   && !is_dummy(next)) {
+                               break;
+               }
+               node = clear_flag(next);
+       }
+       assert(!node || !is_dummy(rcu_dereference(node->p.next)));
+       iter->node = node;
+       iter->next = next;
+}
+
+void cds_lfht_first(struct cds_lfht *ht, struct cds_lfht_iter *iter)
+{
+       struct _cds_lfht_node *lookup;
+
+       /*
+        * Get next after first dummy node. The first dummy node is the
+        * first node of the linked list.
+        */
+       lookup = &ht->t.tbl[0]->nodes[0];
+       iter->node = (struct cds_lfht_node *) lookup;
+       cds_lfht_next(ht, iter);
 }
 
 void cds_lfht_add(struct cds_lfht *ht, struct cds_lfht_node *node)
@@ -1176,12 +1441,12 @@ void cds_lfht_add(struct cds_lfht *ht, struct cds_lfht_node *node)
        node->p.reverse_hash = bit_reverse_ulong((unsigned long) hash);
 
        size = rcu_dereference(ht->t.size);
-       (void) _cds_lfht_add(ht, size, node, 0, 0);
+       (void) _cds_lfht_add(ht, size, node, ADD_DEFAULT, 0);
        ht_count_add(ht, size);
 }
 
 struct cds_lfht_node *cds_lfht_add_unique(struct cds_lfht *ht,
-                                       struct cds_lfht_node *node)
+                               struct cds_lfht_node *node)
 {
        unsigned long hash, size;
        struct cds_lfht_node *ret;
@@ -1190,21 +1455,47 @@ struct cds_lfht_node *cds_lfht_add_unique(struct cds_lfht *ht,
        node->p.reverse_hash = bit_reverse_ulong((unsigned long) hash);
 
        size = rcu_dereference(ht->t.size);
-       ret = _cds_lfht_add(ht, size, node, 1, 0);
+       ret = _cds_lfht_add(ht, size, node, ADD_UNIQUE, 0);
        if (ret == node)
                ht_count_add(ht, size);
        return ret;
 }
 
-int cds_lfht_remove(struct cds_lfht *ht, struct cds_lfht_node *node)
+struct cds_lfht_node *cds_lfht_add_replace(struct cds_lfht *ht,
+                               struct cds_lfht_node *node)
+{
+       unsigned long hash, size;
+       struct cds_lfht_node *ret;
+
+       hash = ht->hash_fct(node->key, node->key_len, ht->hash_seed);
+       node->p.reverse_hash = bit_reverse_ulong((unsigned long) hash);
+
+       size = rcu_dereference(ht->t.size);
+       ret = _cds_lfht_add(ht, size, node, ADD_REPLACE, 0);
+       if (ret == NULL)
+               ht_count_add(ht, size);
+       return ret;
+}
+
+int cds_lfht_replace(struct cds_lfht *ht, struct cds_lfht_iter *old_iter,
+               struct cds_lfht_node *new_node)
+{
+       unsigned long size;
+
+       size = rcu_dereference(ht->t.size);
+       return _cds_lfht_replace(ht, size, old_iter->node, old_iter->next,
+                       new_node);
+}
+
+int cds_lfht_del(struct cds_lfht *ht, struct cds_lfht_iter *iter)
 {
        unsigned long size;
        int ret;
 
        size = rcu_dereference(ht->t.size);
-       ret = _cds_lfht_remove(ht, size, node, 0);
+       ret = _cds_lfht_del(ht, size, iter->node, 0);
        if (!ret)
-               ht_count_remove(ht, size);
+               ht_count_del(ht, size);
        return ret;
 }
 
@@ -1249,7 +1540,7 @@ int cds_lfht_delete_dummy(struct cds_lfht *ht)
  * Should only be called when no more concurrent readers nor writers can
  * possibly access the table.
  */
-int cds_lfht_destroy(struct cds_lfht *ht)
+int cds_lfht_destroy(struct cds_lfht *ht, pthread_attr_t **attr)
 {
        int ret;
 
@@ -1261,18 +1552,32 @@ int cds_lfht_destroy(struct cds_lfht *ht)
        if (ret)
                return ret;
        free_per_cpu_items_count(ht->percpu_count);
+       if (attr)
+               *attr = ht->resize_attr;
        poison_free(ht);
        return ret;
 }
 
 void cds_lfht_count_nodes(struct cds_lfht *ht,
+               long *approx_before,
                unsigned long *count,
-               unsigned long *removed)
+               unsigned long *removed,
+               long *approx_after)
 {
        struct cds_lfht_node *node, *next;
        struct _cds_lfht_node *lookup;
        unsigned long nr_dummy = 0;
 
+       *approx_before = 0;
+       if (nr_cpus_mask >= 0) {
+               int i;
+
+               for (i = 0; i < nr_cpus_mask + 1; i++) {
+                       *approx_before += uatomic_read(&ht->percpu_count[i].add);
+                       *approx_before -= uatomic_read(&ht->percpu_count[i].del);
+               }
+       }
+
        *count = 0;
        *removed = 0;
 
@@ -1282,8 +1587,10 @@ void cds_lfht_count_nodes(struct cds_lfht *ht,
        do {
                next = rcu_dereference(node->p.next);
                if (is_removed(next)) {
-                       assert(!is_dummy(next));
-                       (*removed)++;
+                       if (!is_dummy(next))
+                               (*removed)++;
+                       else
+                               (nr_dummy)++;
                } else if (!is_dummy(next))
                        (*count)++;
                else
@@ -1291,6 +1598,15 @@ void cds_lfht_count_nodes(struct cds_lfht *ht,
                node = clear_flag(next);
        } while (!is_end(node));
        dbg_printf("number of dummy nodes: %lu\n", nr_dummy);
+       *approx_after = 0;
+       if (nr_cpus_mask >= 0) {
+               int i;
+
+               for (i = 0; i < nr_cpus_mask + 1; i++) {
+                       *approx_after += uatomic_read(&ht->percpu_count[i].add);
+                       *approx_after -= uatomic_read(&ht->percpu_count[i].del);
+               }
+       }
 }
 
 /* called with resize mutex held */
This page took 0.032634 seconds and 4 git commands to generate.