| /* |
| * Copyright (c) 2015 Cisco and/or its affiliates. |
| * Licensed under the Apache License, Version 2.0 (the "License"); |
| * you may not use this file except in compliance with the License. |
| * You may obtain a copy of the License at: |
| * |
| * http://www.apache.org/licenses/LICENSE-2.0 |
| * |
| * Unless required by applicable law or agreed to in writing, software |
| * distributed under the License is distributed on an "AS IS" BASIS, |
| * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| * See the License for the specific language governing permissions and |
| * limitations under the License. |
| */ |
| |
| /** @cond DOCUMENTATION_IS_IN_BIHASH_DOC_H */ |
| |
| static inline void *BV (alloc_aligned) (BVT (clib_bihash) * h, uword nbytes) |
| { |
| uword rv; |
| |
| /* Round to an even number of cache lines */ |
| nbytes += CLIB_CACHE_LINE_BYTES - 1; |
| nbytes &= ~(CLIB_CACHE_LINE_BYTES - 1); |
| |
| rv = h->alloc_arena_next; |
| h->alloc_arena_next += nbytes; |
| |
| if (rv >= (h->alloc_arena + h->alloc_arena_size)) |
| os_out_of_memory (); |
| |
| return (void *) rv; |
| } |
| |
| |
| void BV (clib_bihash_init) |
| (BVT (clib_bihash) * h, char *name, u32 nbuckets, uword memory_size) |
| { |
| uword bucket_size; |
| |
| nbuckets = 1 << (max_log2 (nbuckets)); |
| |
| h->name = (u8 *) name; |
| h->nbuckets = nbuckets; |
| h->log2_nbuckets = max_log2 (nbuckets); |
| h->cache_hits = 0; |
| h->cache_misses = 0; |
| |
| /* |
| * Make sure the requested size is rational. The max table |
| * size without playing the alignment card is 64 Gbytes. |
| * If someone starts complaining that's not enough, we can shift |
| * the offset by CLIB_LOG2_CACHE_LINE_BYTES... |
| */ |
| ASSERT (memory_size < (1ULL << BIHASH_BUCKET_OFFSET_BITS)); |
| |
| h->alloc_arena = (uword) clib_mem_vm_alloc (memory_size); |
| h->alloc_arena_next = h->alloc_arena; |
| h->alloc_arena_size = memory_size; |
| |
| bucket_size = nbuckets * sizeof (h->buckets[0]); |
| h->buckets = BV (alloc_aligned) (h, bucket_size); |
| |
| h->alloc_lock = BV (alloc_aligned) (h, CLIB_CACHE_LINE_BYTES); |
| h->alloc_lock[0] = 0; |
| |
| h->fmt_fn = NULL; |
| } |
| |
| void BV (clib_bihash_set_kvp_format_fn) (BVT (clib_bihash) * h, |
| format_function_t * fmt_fn) |
| { |
| h->fmt_fn = fmt_fn; |
| } |
| |
| void BV (clib_bihash_free) (BVT (clib_bihash) * h) |
| { |
| vec_free (h->working_copies); |
| vec_free (h->freelists); |
| clib_mem_vm_free ((void *) (h->alloc_arena), h->alloc_arena_size); |
| memset (h, 0, sizeof (*h)); |
| } |
| |
| static |
| BVT (clib_bihash_value) * |
| BV (value_alloc) (BVT (clib_bihash) * h, u32 log2_pages) |
| { |
| BVT (clib_bihash_value) * rv = 0; |
| |
| ASSERT (h->alloc_lock[0]); |
| if (log2_pages >= vec_len (h->freelists) || h->freelists[log2_pages] == 0) |
| { |
| vec_validate_init_empty (h->freelists, log2_pages, 0); |
| rv = BV (alloc_aligned) (h, (sizeof (*rv) * (1 << log2_pages))); |
| goto initialize; |
| } |
| rv = h->freelists[log2_pages]; |
| h->freelists[log2_pages] = rv->next_free; |
| |
| initialize: |
| ASSERT (rv); |
| /* |
| * Latest gcc complains that the length arg is zero |
| * if we replace (1<<log2_pages) with vec_len(rv). |
| * No clue. |
| */ |
| memset (rv, 0xff, sizeof (*rv) * (1 << log2_pages)); |
| return rv; |
| } |
| |
| static void |
| BV (value_free) (BVT (clib_bihash) * h, BVT (clib_bihash_value) * v, |
| u32 log2_pages) |
| { |
| ASSERT (h->alloc_lock[0]); |
| |
| ASSERT (vec_len (h->freelists) > log2_pages); |
| |
| if (CLIB_DEBUG > 0) |
| memset (v, 0xFE, sizeof (*v) * (1 << log2_pages)); |
| |
| v->next_free = h->freelists[log2_pages]; |
| h->freelists[log2_pages] = v; |
| } |
| |
| static inline void |
| BV (make_working_copy) (BVT (clib_bihash) * h, BVT (clib_bihash_bucket) * b) |
| { |
| BVT (clib_bihash_value) * v; |
| BVT (clib_bihash_bucket) working_bucket __attribute__ ((aligned (8))); |
| BVT (clib_bihash_value) * working_copy; |
| u32 thread_index = os_get_thread_index (); |
| int log2_working_copy_length; |
| |
| ASSERT (h->alloc_lock[0]); |
| |
| if (thread_index >= vec_len (h->working_copies)) |
| { |
| vec_validate (h->working_copies, thread_index); |
| vec_validate_init_empty (h->working_copy_lengths, thread_index, ~0); |
| } |
| |
| /* |
| * working_copies are per-cpu so that near-simultaneous |
| * updates from multiple threads will not result in sporadic, spurious |
| * lookup failures. |
| */ |
| working_copy = h->working_copies[thread_index]; |
| log2_working_copy_length = h->working_copy_lengths[thread_index]; |
| |
| h->saved_bucket.as_u64 = b->as_u64; |
| |
| if (b->log2_pages > log2_working_copy_length) |
| { |
| /* |
| * It's not worth the bookkeeping to free working copies |
| * if (working_copy) |
| * clib_mem_free (working_copy); |
| */ |
| working_copy = BV (alloc_aligned) |
| (h, sizeof (working_copy[0]) * (1 << b->log2_pages)); |
| h->working_copy_lengths[thread_index] = b->log2_pages; |
| h->working_copies[thread_index] = working_copy; |
| } |
| |
| v = BV (clib_bihash_get_value) (h, b->offset); |
| |
| clib_memcpy (working_copy, v, sizeof (*v) * (1 << b->log2_pages)); |
| working_bucket.as_u64 = b->as_u64; |
| working_bucket.offset = BV (clib_bihash_get_offset) (h, working_copy); |
| CLIB_MEMORY_BARRIER (); |
| b->as_u64 = working_bucket.as_u64; |
| h->working_copies[thread_index] = working_copy; |
| } |
| |
| static |
| BVT (clib_bihash_value) * |
| BV (split_and_rehash) |
| (BVT (clib_bihash) * h, |
| BVT (clib_bihash_value) * old_values, u32 old_log2_pages, |
| u32 new_log2_pages) |
| { |
| BVT (clib_bihash_value) * new_values, *new_v; |
| int i, j, length_in_kvs; |
| |
| ASSERT (h->alloc_lock[0]); |
| |
| new_values = BV (value_alloc) (h, new_log2_pages); |
| length_in_kvs = (1 << old_log2_pages) * BIHASH_KVP_PER_PAGE; |
| |
| for (i = 0; i < length_in_kvs; i++) |
| { |
| u64 new_hash; |
| |
| /* Entry not in use? Forget it */ |
| if (BV (clib_bihash_is_free) (&(old_values->kvp[i]))) |
| continue; |
| |
| /* rehash the item onto its new home-page */ |
| new_hash = BV (clib_bihash_hash) (&(old_values->kvp[i])); |
| new_hash >>= h->log2_nbuckets; |
| new_hash &= (1 << new_log2_pages) - 1; |
| new_v = &new_values[new_hash]; |
| |
| /* Across the new home-page */ |
| for (j = 0; j < BIHASH_KVP_PER_PAGE; j++) |
| { |
| /* Empty slot */ |
| if (BV (clib_bihash_is_free) (&(new_v->kvp[j]))) |
| { |
| clib_memcpy (&(new_v->kvp[j]), &(old_values->kvp[i]), |
| sizeof (new_v->kvp[j])); |
| goto doublebreak; |
| } |
| } |
| /* Crap. Tell caller to try again */ |
| BV (value_free) (h, new_values, new_log2_pages); |
| return 0; |
| doublebreak:; |
| } |
| |
| return new_values; |
| } |
| |
| static |
| BVT (clib_bihash_value) * |
| BV (split_and_rehash_linear) |
| (BVT (clib_bihash) * h, |
| BVT (clib_bihash_value) * old_values, u32 old_log2_pages, |
| u32 new_log2_pages) |
| { |
| BVT (clib_bihash_value) * new_values; |
| int i, j, new_length, old_length; |
| |
| ASSERT (h->alloc_lock[0]); |
| |
| new_values = BV (value_alloc) (h, new_log2_pages); |
| new_length = (1 << new_log2_pages) * BIHASH_KVP_PER_PAGE; |
| old_length = (1 << old_log2_pages) * BIHASH_KVP_PER_PAGE; |
| |
| j = 0; |
| /* Across the old value array */ |
| for (i = 0; i < old_length; i++) |
| { |
| /* Find a free slot in the new linear scan bucket */ |
| for (; j < new_length; j++) |
| { |
| /* Old value not in use? Forget it. */ |
| if (BV (clib_bihash_is_free) (&(old_values->kvp[i]))) |
| goto doublebreak; |
| |
| /* New value should never be in use */ |
| if (BV (clib_bihash_is_free) (&(new_values->kvp[j]))) |
| { |
| /* Copy the old value and move along */ |
| clib_memcpy (&(new_values->kvp[j]), &(old_values->kvp[i]), |
| sizeof (new_values->kvp[j])); |
| j++; |
| goto doublebreak; |
| } |
| } |
| /* This should never happen... */ |
| clib_warning ("BUG: linear rehash failed!"); |
| BV (value_free) (h, new_values, new_log2_pages); |
| return 0; |
| |
| doublebreak:; |
| } |
| return new_values; |
| } |
| |
| int BV (clib_bihash_add_del) |
| (BVT (clib_bihash) * h, BVT (clib_bihash_kv) * add_v, int is_add) |
| { |
| u32 bucket_index; |
| BVT (clib_bihash_bucket) * b, tmp_b; |
| BVT (clib_bihash_value) * v, *new_v, *save_new_v, *working_copy; |
| int i, limit; |
| u64 hash, new_hash; |
| u32 new_log2_pages, old_log2_pages; |
| u32 thread_index = os_get_thread_index (); |
| int mark_bucket_linear; |
| int resplit_once; |
| |
| hash = BV (clib_bihash_hash) (add_v); |
| |
| bucket_index = hash & (h->nbuckets - 1); |
| b = &h->buckets[bucket_index]; |
| |
| hash >>= h->log2_nbuckets; |
| |
| BV (clib_bihash_lock_bucket) (b); |
| |
| /* First elt in the bucket? */ |
| if (BV (clib_bihash_bucket_is_empty) (b)) |
| { |
| if (is_add == 0) |
| { |
| BV (clib_bihash_unlock_bucket) (b); |
| return (-1); |
| } |
| |
| BV (clib_bihash_alloc_lock) (h); |
| v = BV (value_alloc) (h, 0); |
| BV (clib_bihash_alloc_unlock) (h); |
| |
| *v->kvp = *add_v; |
| tmp_b.as_u64 = 0; /* clears bucket lock */ |
| tmp_b.offset = BV (clib_bihash_get_offset) (h, v); |
| tmp_b.refcnt = 1; |
| CLIB_MEMORY_BARRIER (); |
| |
| b->as_u64 = tmp_b.as_u64; |
| BV (clib_bihash_unlock_bucket) (b); |
| return (0); |
| } |
| |
| /* WARNING: we're still looking at the live copy... */ |
| limit = BIHASH_KVP_PER_PAGE; |
| v = BV (clib_bihash_get_value) (h, b->offset); |
| |
| v += (b->linear_search == 0) ? hash & ((1 << b->log2_pages) - 1) : 0; |
| if (b->linear_search) |
| limit <<= b->log2_pages; |
| |
| if (is_add) |
| { |
| /* |
| * Because reader threads are looking at live data, |
| * we have to be extra careful. Readers do NOT hold the |
| * bucket lock. We need to be SLOWER than a search, past the |
| * point where readers CHECK the bucket lock. |
| */ |
| |
| /* |
| * For obvious (in hindsight) reasons, see if we're supposed to |
| * replace an existing key, then look for an empty slot. |
| */ |
| for (i = 0; i < limit; i++) |
| { |
| if (!memcmp (&(v->kvp[i]), &add_v->key, sizeof (add_v->key))) |
| { |
| CLIB_MEMORY_BARRIER (); /* Add a delay */ |
| clib_memcpy (&(v->kvp[i]), add_v, sizeof (*add_v)); |
| BV (clib_bihash_unlock_bucket) (b); |
| return (0); |
| } |
| } |
| /* |
| * Look for an empty slot. If found, use it |
| */ |
| for (i = 0; i < limit; i++) |
| { |
| if (BV (clib_bihash_is_free) (&(v->kvp[i]))) |
| { |
| /* |
| * Copy the value first, so that if a reader manages |
| * to match the new key, the value will be right... |
| */ |
| clib_memcpy (&(v->kvp[i].value), |
| &add_v->value, sizeof (add_v->value)); |
| CLIB_MEMORY_BARRIER (); /* Make sure the value has settled */ |
| clib_memcpy (&(v->kvp[i]), &add_v->key, sizeof (add_v->key)); |
| b->refcnt++; |
| BV (clib_bihash_unlock_bucket) (b); |
| return (0); |
| } |
| } |
| /* Out of space in this bucket, split the bucket... */ |
| } |
| else /* delete case */ |
| { |
| for (i = 0; i < limit; i++) |
| { |
| /* Found the key? Kill it... */ |
| if (!memcmp (&(v->kvp[i]), &add_v->key, sizeof (add_v->key))) |
| { |
| memset (&(v->kvp[i]), 0xff, sizeof (*(add_v))); |
| /* Is the bucket empty? */ |
| if (PREDICT_TRUE (b->refcnt > 1)) |
| { |
| b->refcnt--; |
| BV (clib_bihash_unlock_bucket) (b); |
| return (0); |
| } |
| else /* yes, free it */ |
| { |
| /* Save old bucket value, need log2_pages to free it */ |
| tmp_b.as_u64 = b->as_u64; |
| CLIB_MEMORY_BARRIER (); |
| |
| /* Kill and unlock the bucket */ |
| b->as_u64 = 0; |
| |
| /* And free the backing storage */ |
| BV (clib_bihash_alloc_lock) (h); |
| /* Note: v currently points into the middle of the bucket */ |
| v = BV (clib_bihash_get_value) (h, tmp_b.offset); |
| BV (value_free) (h, v, tmp_b.log2_pages); |
| BV (clib_bihash_alloc_unlock) (h); |
| return (0); |
| } |
| } |
| } |
| /* Not found... */ |
| BV (clib_bihash_unlock_bucket) (b); |
| return (-3); |
| } |
| |
| /* Move readers to a (locked) temp copy of the bucket */ |
| BV (clib_bihash_alloc_lock) (h); |
| BV (make_working_copy) (h, b); |
| |
| v = BV (clib_bihash_get_value) (h, h->saved_bucket.offset); |
| |
| old_log2_pages = h->saved_bucket.log2_pages; |
| new_log2_pages = old_log2_pages + 1; |
| mark_bucket_linear = 0; |
| |
| working_copy = h->working_copies[thread_index]; |
| resplit_once = 0; |
| |
| new_v = BV (split_and_rehash) (h, working_copy, old_log2_pages, |
| new_log2_pages); |
| if (new_v == 0) |
| { |
| try_resplit: |
| resplit_once = 1; |
| new_log2_pages++; |
| /* Try re-splitting. If that fails, fall back to linear search */ |
| new_v = BV (split_and_rehash) (h, working_copy, old_log2_pages, |
| new_log2_pages); |
| if (new_v == 0) |
| { |
| mark_linear: |
| new_log2_pages--; |
| /* pinned collisions, use linear search */ |
| new_v = |
| BV (split_and_rehash_linear) (h, working_copy, old_log2_pages, |
| new_log2_pages); |
| mark_bucket_linear = 1; |
| } |
| } |
| |
| /* Try to add the new entry */ |
| save_new_v = new_v; |
| new_hash = BV (clib_bihash_hash) (add_v); |
| limit = BIHASH_KVP_PER_PAGE; |
| if (mark_bucket_linear) |
| limit <<= new_log2_pages; |
| new_hash >>= h->log2_nbuckets; |
| new_hash &= (1 << new_log2_pages) - 1; |
| new_v += mark_bucket_linear ? 0 : new_hash; |
| |
| for (i = 0; i < limit; i++) |
| { |
| if (BV (clib_bihash_is_free) (&(new_v->kvp[i]))) |
| { |
| clib_memcpy (&(new_v->kvp[i]), add_v, sizeof (*add_v)); |
| goto expand_ok; |
| } |
| } |
| |
| /* Crap. Try again */ |
| BV (value_free) (h, save_new_v, new_log2_pages); |
| /* |
| * If we've already doubled the size of the bucket once, |
| * fall back to linear search now. |
| */ |
| if (resplit_once) |
| goto mark_linear; |
| else |
| goto try_resplit; |
| |
| expand_ok: |
| tmp_b.log2_pages = new_log2_pages; |
| tmp_b.offset = BV (clib_bihash_get_offset) (h, save_new_v); |
| tmp_b.linear_search = mark_bucket_linear; |
| tmp_b.refcnt = h->saved_bucket.refcnt + 1; |
| tmp_b.lock = 0; |
| CLIB_MEMORY_BARRIER (); |
| b->as_u64 = tmp_b.as_u64; |
| BV (clib_bihash_alloc_unlock) (h); |
| return (0); |
| } |
| |
| int BV (clib_bihash_search) |
| (BVT (clib_bihash) * h, |
| BVT (clib_bihash_kv) * search_key, BVT (clib_bihash_kv) * valuep) |
| { |
| u64 hash; |
| u32 bucket_index; |
| BVT (clib_bihash_value) * v; |
| BVT (clib_bihash_bucket) * b; |
| int i, limit; |
| |
| ASSERT (valuep); |
| |
| hash = BV (clib_bihash_hash) (search_key); |
| |
| bucket_index = hash & (h->nbuckets - 1); |
| b = &h->buckets[bucket_index]; |
| |
| if (BV (clib_bihash_bucket_is_empty) (b)) |
| return -1; |
| |
| if (PREDICT_FALSE (b->lock)) |
| { |
| volatile BVT (clib_bihash_bucket) * bv = b; |
| while (bv->lock) |
| CLIB_PAUSE (); |
| } |
| |
| hash >>= h->log2_nbuckets; |
| |
| v = BV (clib_bihash_get_value) (h, b->offset); |
| limit = BIHASH_KVP_PER_PAGE; |
| v += (b->linear_search == 0) ? hash & ((1 << b->log2_pages) - 1) : 0; |
| if (PREDICT_FALSE (b->linear_search)) |
| limit <<= b->log2_pages; |
| |
| for (i = 0; i < limit; i++) |
| { |
| if (BV (clib_bihash_key_compare) (v->kvp[i].key, search_key->key)) |
| { |
| *valuep = v->kvp[i]; |
| return 0; |
| } |
| } |
| return -1; |
| } |
| |
| u8 *BV (format_bihash) (u8 * s, va_list * args) |
| { |
| BVT (clib_bihash) * h = va_arg (*args, BVT (clib_bihash) *); |
| int verbose = va_arg (*args, int); |
| BVT (clib_bihash_bucket) * b; |
| BVT (clib_bihash_value) * v; |
| int i, j, k; |
| u64 active_elements = 0; |
| u64 active_buckets = 0; |
| u64 linear_buckets = 0; |
| u64 used_bytes; |
| |
| s = format (s, "Hash table %s\n", h->name ? h->name : (u8 *) "(unnamed)"); |
| |
| for (i = 0; i < h->nbuckets; i++) |
| { |
| b = &h->buckets[i]; |
| if (BV (clib_bihash_bucket_is_empty) (b)) |
| { |
| if (verbose > 1) |
| s = format (s, "[%d]: empty\n", i); |
| continue; |
| } |
| |
| active_buckets++; |
| |
| if (b->linear_search) |
| linear_buckets++; |
| |
| if (verbose) |
| { |
| s = format (s, "[%d]: heap offset %d, len %d, linear %d\n", i, |
| b->offset, (1 << b->log2_pages), b->linear_search); |
| } |
| |
| v = BV (clib_bihash_get_value) (h, b->offset); |
| for (j = 0; j < (1 << b->log2_pages); j++) |
| { |
| for (k = 0; k < BIHASH_KVP_PER_PAGE; k++) |
| { |
| if (BV (clib_bihash_is_free) (&v->kvp[k])) |
| { |
| if (verbose > 1) |
| s = format (s, " %d: empty\n", |
| j * BIHASH_KVP_PER_PAGE + k); |
| continue; |
| } |
| if (verbose) |
| { |
| if (h->fmt_fn) |
| { |
| s = format (s, " %d: %U\n", |
| j * BIHASH_KVP_PER_PAGE + k, |
| h->fmt_fn, &(v->kvp[k])); |
| } |
| else |
| { |
| s = format (s, " %d: %U\n", |
| j * BIHASH_KVP_PER_PAGE + k, |
| BV (format_bihash_kvp), &(v->kvp[k])); |
| } |
| } |
| active_elements++; |
| } |
| v++; |
| } |
| } |
| |
| s = format (s, " %lld active elements %lld active buckets\n", |
| active_elements, active_buckets); |
| s = format (s, " %d free lists\n", vec_len (h->freelists)); |
| |
| for (i = 0; i < vec_len (h->freelists); i++) |
| { |
| u32 nfree = 0; |
| BVT (clib_bihash_value) * free_elt; |
| |
| free_elt = h->freelists[i]; |
| while (free_elt) |
| { |
| nfree++; |
| free_elt = free_elt->next_free; |
| } |
| |
| s = format (s, " [len %d] %u free elts\n", 1 << i, nfree); |
| } |
| |
| s = format (s, " %lld linear search buckets\n", linear_buckets); |
| s = format (s, " %lld cache hits, %lld cache misses\n", |
| h->cache_hits, h->cache_misses); |
| used_bytes = h->alloc_arena_next - h->alloc_arena; |
| s = format (s, |
| " arena: base %llx, next %llx\n" |
| " used %lld b (%lld Mbytes) of %lld b (%lld Mbytes)\n", |
| h->alloc_arena, h->alloc_arena_next, |
| used_bytes, used_bytes >> 20, |
| h->alloc_arena_size, h->alloc_arena_size >> 20); |
| return s; |
| } |
| |
| void BV (clib_bihash_foreach_key_value_pair) |
| (BVT (clib_bihash) * h, void *callback, void *arg) |
| { |
| int i, j, k; |
| BVT (clib_bihash_bucket) * b; |
| BVT (clib_bihash_value) * v; |
| void (*fp) (BVT (clib_bihash_kv) *, void *) = callback; |
| |
| for (i = 0; i < h->nbuckets; i++) |
| { |
| b = &h->buckets[i]; |
| if (BV (clib_bihash_bucket_is_empty) (b)) |
| continue; |
| |
| v = BV (clib_bihash_get_value) (h, b->offset); |
| for (j = 0; j < (1 << b->log2_pages); j++) |
| { |
| for (k = 0; k < BIHASH_KVP_PER_PAGE; k++) |
| { |
| if (BV (clib_bihash_is_free) (&v->kvp[k])) |
| continue; |
| |
| (*fp) (&v->kvp[k], arg); |
| /* |
| * In case the callback deletes the last entry in the bucket... |
| */ |
| if (BV (clib_bihash_bucket_is_empty) (b)) |
| goto doublebreak; |
| } |
| v++; |
| } |
| doublebreak: |
| ; |
| } |
| } |
| |
| /** @endcond */ |
| |
| /* |
| * fd.io coding-style-patch-verification: ON |
| * |
| * Local Variables: |
| * eval: (c-set-style "gnu") |
| * End: |
| */ |