add leveldb

lib/leveldb/util/arena.cc

@@ -0,0 +1,66 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "util/arena.h"
+
+namespace leveldb {
+
+static const int kBlockSize = 4096;
+
+Arena::Arena()
+    : alloc_ptr_(nullptr), alloc_bytes_remaining_(0), memory_usage_(0) {}
+
+Arena::~Arena() {
+  for (size_t i = 0; i < blocks_.size(); i++) {
+    delete[] blocks_[i];
+  }
+}
+
+char* Arena::AllocateFallback(size_t bytes) {
+  if (bytes > kBlockSize / 4) {
+    // Object is more than a quarter of our block size.  Allocate it separately
+    // to avoid wasting too much space in leftover bytes.
+    char* result = AllocateNewBlock(bytes);
+    return result;
+  }
+
+  // We waste the remaining space in the current block.
+  alloc_ptr_ = AllocateNewBlock(kBlockSize);
+  alloc_bytes_remaining_ = kBlockSize;
+
+  char* result = alloc_ptr_;
+  alloc_ptr_ += bytes;
+  alloc_bytes_remaining_ -= bytes;
+  return result;
+}
+
+char* Arena::AllocateAligned(size_t bytes) {
+  const int align = (sizeof(void*) > 8) ? sizeof(void*) : 8;
+  static_assert((align & (align - 1)) == 0,
+                "Pointer size should be a power of 2");
+  size_t current_mod = reinterpret_cast<uintptr_t>(alloc_ptr_) & (align - 1);
+  size_t slop = (current_mod == 0 ? 0 : align - current_mod);
+  size_t needed = bytes + slop;
+  char* result;
+  if (needed <= alloc_bytes_remaining_) {
+    result = alloc_ptr_ + slop;
+    alloc_ptr_ += needed;
+    alloc_bytes_remaining_ -= needed;
+  } else {
+    // AllocateFallback always returned aligned memory
+    result = AllocateFallback(bytes);
+  }
+  assert((reinterpret_cast<uintptr_t>(result) & (align - 1)) == 0);
+  return result;
+}
+
+char* Arena::AllocateNewBlock(size_t block_bytes) {
+  char* result = new char[block_bytes];
+  blocks_.push_back(result);
+  memory_usage_.fetch_add(block_bytes + sizeof(char*),
+                          std::memory_order_relaxed);
+  return result;
+}
+
+}  // namespace leveldb

lib/leveldb/util/arena.h

@@ -0,0 +1,71 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#ifndef STORAGE_LEVELDB_UTIL_ARENA_H_
+#define STORAGE_LEVELDB_UTIL_ARENA_H_
+
+#include <atomic>
+#include <cassert>
+#include <cstddef>
+#include <cstdint>
+#include <vector>
+
+namespace leveldb {
+
+class Arena {
+ public:
+  Arena();
+
+  Arena(const Arena&) = delete;
+  Arena& operator=(const Arena&) = delete;
+
+  ~Arena();
+
+  // Return a pointer to a newly allocated memory block of "bytes" bytes.
+  char* Allocate(size_t bytes);
+
+  // Allocate memory with the normal alignment guarantees provided by malloc.
+  char* AllocateAligned(size_t bytes);
+
+  // Returns an estimate of the total memory usage of data allocated
+  // by the arena.
+  size_t MemoryUsage() const {
+    return memory_usage_.load(std::memory_order_relaxed);
+  }
+
+ private:
+  char* AllocateFallback(size_t bytes);
+  char* AllocateNewBlock(size_t block_bytes);
+
+  // Allocation state
+  char* alloc_ptr_;
+  size_t alloc_bytes_remaining_;
+
+  // Array of new[] allocated memory blocks
+  std::vector<char*> blocks_;
+
+  // Total memory usage of the arena.
+  //
+  // TODO(costan): This member is accessed via atomics, but the others are
+  //               accessed without any locking. Is this OK?
+  std::atomic<size_t> memory_usage_;
+};
+
+inline char* Arena::Allocate(size_t bytes) {
+  // The semantics of what to return are a bit messy if we allow
+  // 0-byte allocations, so we disallow them here (we don't need
+  // them for our internal use).
+  assert(bytes > 0);
+  if (bytes <= alloc_bytes_remaining_) {
+    char* result = alloc_ptr_;
+    alloc_ptr_ += bytes;
+    alloc_bytes_remaining_ -= bytes;
+    return result;
+  }
+  return AllocateFallback(bytes);
+}
+
+}  // namespace leveldb
+
+#endif  // STORAGE_LEVELDB_UTIL_ARENA_H_

lib/leveldb/util/arena_test.cc

@@ -0,0 +1,66 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "util/arena.h"
+
+#include "gtest/gtest.h"
+#include "util/random.h"
+
+namespace leveldb {
+
+TEST(ArenaTest, Empty) { Arena arena; }
+
+TEST(ArenaTest, Simple) {
+  std::vector<std::pair<size_t, char*>> allocated;
+  Arena arena;
+  const int N = 100000;
+  size_t bytes = 0;
+  Random rnd(301);
+  for (int i = 0; i < N; i++) {
+    size_t s;
+    if (i % (N / 10) == 0) {
+      s = i;
+    } else {
+      s = rnd.OneIn(4000)
+              ? rnd.Uniform(6000)
+              : (rnd.OneIn(10) ? rnd.Uniform(100) : rnd.Uniform(20));
+    }
+    if (s == 0) {
+      // Our arena disallows size 0 allocations.
+      s = 1;
+    }
+    char* r;
+    if (rnd.OneIn(10)) {
+      r = arena.AllocateAligned(s);
+    } else {
+      r = arena.Allocate(s);
+    }
+
+    for (size_t b = 0; b < s; b++) {
+      // Fill the "i"th allocation with a known bit pattern
+      r[b] = i % 256;
+    }
+    bytes += s;
+    allocated.push_back(std::make_pair(s, r));
+    ASSERT_GE(arena.MemoryUsage(), bytes);
+    if (i > N / 10) {
+      ASSERT_LE(arena.MemoryUsage(), bytes * 1.10);
+    }
+  }
+  for (size_t i = 0; i < allocated.size(); i++) {
+    size_t num_bytes = allocated[i].first;
+    const char* p = allocated[i].second;
+    for (size_t b = 0; b < num_bytes; b++) {
+      // Check the "i"th allocation for the known bit pattern
+      ASSERT_EQ(int(p[b]) & 0xff, i % 256);
+    }
+  }
+}
+
+}  // namespace leveldb
+
+int main(int argc, char** argv) {
+  testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}

lib/leveldb/util/bloom.cc

@@ -0,0 +1,92 @@
+// Copyright (c) 2012 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "leveldb/filter_policy.h"
+
+#include "leveldb/slice.h"
+#include "util/hash.h"
+
+namespace leveldb {
+
+namespace {
+static uint32_t BloomHash(const Slice& key) {
+  return Hash(key.data(), key.size(), 0xbc9f1d34);
+}
+
+class BloomFilterPolicy : public FilterPolicy {
+ public:
+  explicit BloomFilterPolicy(int bits_per_key) : bits_per_key_(bits_per_key) {
+    // We intentionally round down to reduce probing cost a little bit
+    k_ = static_cast<size_t>(bits_per_key * 0.69);  // 0.69 =~ ln(2)
+    if (k_ < 1) k_ = 1;
+    if (k_ > 30) k_ = 30;
+  }
+
+  const char* Name() const override { return "leveldb.BuiltinBloomFilter2"; }
+
+  void CreateFilter(const Slice* keys, int n, std::string* dst) const override {
+    // Compute bloom filter size (in both bits and bytes)
+    size_t bits = n * bits_per_key_;
+
+    // For small n, we can see a very high false positive rate.  Fix it
+    // by enforcing a minimum bloom filter length.
+    if (bits < 64) bits = 64;
+
+    size_t bytes = (bits + 7) / 8;
+    bits = bytes * 8;
+
+    const size_t init_size = dst->size();
+    dst->resize(init_size + bytes, 0);
+    dst->push_back(static_cast<char>(k_));  // Remember # of probes in filter
+    char* array = &(*dst)[init_size];
+    for (int i = 0; i < n; i++) {
+      // Use double-hashing to generate a sequence of hash values.
+      // See analysis in [Kirsch,Mitzenmacher 2006].
+      uint32_t h = BloomHash(keys[i]);
+      const uint32_t delta = (h >> 17) | (h << 15);  // Rotate right 17 bits
+      for (size_t j = 0; j < k_; j++) {
+        const uint32_t bitpos = h % bits;
+        array[bitpos / 8] |= (1 << (bitpos % 8));
+        h += delta;
+      }
+    }
+  }
+
+  bool KeyMayMatch(const Slice& key, const Slice& bloom_filter) const override {
+    const size_t len = bloom_filter.size();
+    if (len < 2) return false;
+
+    const char* array = bloom_filter.data();
+    const size_t bits = (len - 1) * 8;
+
+    // Use the encoded k so that we can read filters generated by
+    // bloom filters created using different parameters.
+    const size_t k = array[len - 1];
+    if (k > 30) {
+      // Reserved for potentially new encodings for short bloom filters.
+      // Consider it a match.
+      return true;
+    }
+
+    uint32_t h = BloomHash(key);
+    const uint32_t delta = (h >> 17) | (h << 15);  // Rotate right 17 bits
+    for (size_t j = 0; j < k; j++) {
+      const uint32_t bitpos = h % bits;
+      if ((array[bitpos / 8] & (1 << (bitpos % 8))) == 0) return false;
+      h += delta;
+    }
+    return true;
+  }
+
+ private:
+  size_t bits_per_key_;
+  size_t k_;
+};
+}  // namespace
+
+const FilterPolicy* NewBloomFilterPolicy(int bits_per_key) {
+  return new BloomFilterPolicy(bits_per_key);
+}
+
+}  // namespace leveldb

lib/leveldb/util/bloom_test.cc

@@ -0,0 +1,159 @@
+// Copyright (c) 2012 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "gtest/gtest.h"
+#include "leveldb/filter_policy.h"
+#include "util/coding.h"
+#include "util/logging.h"
+#include "util/testutil.h"
+
+namespace leveldb {
+
+static const int kVerbose = 1;
+
+static Slice Key(int i, char* buffer) {
+  EncodeFixed32(buffer, i);
+  return Slice(buffer, sizeof(uint32_t));
+}
+
+class BloomTest : public testing::Test {
+ public:
+  BloomTest() : policy_(NewBloomFilterPolicy(10)) {}
+
+  ~BloomTest() { delete policy_; }
+
+  void Reset() {
+    keys_.clear();
+    filter_.clear();
+  }
+
+  void Add(const Slice& s) { keys_.push_back(s.ToString()); }
+
+  void Build() {
+    std::vector<Slice> key_slices;
+    for (size_t i = 0; i < keys_.size(); i++) {
+      key_slices.push_back(Slice(keys_[i]));
+    }
+    filter_.clear();
+    policy_->CreateFilter(&key_slices[0], static_cast<int>(key_slices.size()),
+                          &filter_);
+    keys_.clear();
+    if (kVerbose >= 2) DumpFilter();
+  }
+
+  size_t FilterSize() const { return filter_.size(); }
+
+  void DumpFilter() {
+    std::fprintf(stderr, "F(");
+    for (size_t i = 0; i + 1 < filter_.size(); i++) {
+      const unsigned int c = static_cast<unsigned int>(filter_[i]);
+      for (int j = 0; j < 8; j++) {
+        std::fprintf(stderr, "%c", (c & (1 << j)) ? '1' : '.');
+      }
+    }
+    std::fprintf(stderr, ")\n");
+  }
+
+  bool Matches(const Slice& s) {
+    if (!keys_.empty()) {
+      Build();
+    }
+    return policy_->KeyMayMatch(s, filter_);
+  }
+
+  double FalsePositiveRate() {
+    char buffer[sizeof(int)];
+    int result = 0;
+    for (int i = 0; i < 10000; i++) {
+      if (Matches(Key(i + 1000000000, buffer))) {
+        result++;
+      }
+    }
+    return result / 10000.0;
+  }
+
+ private:
+  const FilterPolicy* policy_;
+  std::string filter_;
+  std::vector<std::string> keys_;
+};
+
+TEST_F(BloomTest, EmptyFilter) {
+  ASSERT_TRUE(!Matches("hello"));
+  ASSERT_TRUE(!Matches("world"));
+}
+
+TEST_F(BloomTest, Small) {
+  Add("hello");
+  Add("world");
+  ASSERT_TRUE(Matches("hello"));
+  ASSERT_TRUE(Matches("world"));
+  ASSERT_TRUE(!Matches("x"));
+  ASSERT_TRUE(!Matches("foo"));
+}
+
+static int NextLength(int length) {
+  if (length < 10) {
+    length += 1;
+  } else if (length < 100) {
+    length += 10;
+  } else if (length < 1000) {
+    length += 100;
+  } else {
+    length += 1000;
+  }
+  return length;
+}
+
+TEST_F(BloomTest, VaryingLengths) {
+  char buffer[sizeof(int)];
+
+  // Count number of filters that significantly exceed the false positive rate
+  int mediocre_filters = 0;
+  int good_filters = 0;
+
+  for (int length = 1; length <= 10000; length = NextLength(length)) {
+    Reset();
+    for (int i = 0; i < length; i++) {
+      Add(Key(i, buffer));
+    }
+    Build();
+
+    ASSERT_LE(FilterSize(), static_cast<size_t>((length * 10 / 8) + 40))
+        << length;
+
+    // All added keys must match
+    for (int i = 0; i < length; i++) {
+      ASSERT_TRUE(Matches(Key(i, buffer)))
+          << "Length " << length << "; key " << i;
+    }
+
+    // Check false positive rate
+    double rate = FalsePositiveRate();
+    if (kVerbose >= 1) {
+      std::fprintf(stderr,
+                   "False positives: %5.2f%% @ length = %6d ; bytes = %6d\n",
+                   rate * 100.0, length, static_cast<int>(FilterSize()));
+    }
+    ASSERT_LE(rate, 0.02);  // Must not be over 2%
+    if (rate > 0.0125)
+      mediocre_filters++;  // Allowed, but not too often
+    else
+      good_filters++;
+  }
+  if (kVerbose >= 1) {
+    std::fprintf(stderr, "Filters: %d good, %d mediocre\n", good_filters,
+                 mediocre_filters);
+  }
+  ASSERT_LE(mediocre_filters, good_filters / 5);
+}
+
+// Different bits-per-byte
+
+}  // namespace leveldb
+
+int main(int argc, char** argv) {
+  testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}

lib/leveldb/util/cache.cc

@@ -0,0 +1,401 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "leveldb/cache.h"
+
+#include <cassert>
+#include <cstdio>
+#include <cstdlib>
+
+#include "port/port.h"
+#include "port/thread_annotations.h"
+#include "util/hash.h"
+#include "util/mutexlock.h"
+
+namespace leveldb {
+
+Cache::~Cache() {}
+
+namespace {
+
+// LRU cache implementation
+//
+// Cache entries have an "in_cache" boolean indicating whether the cache has a
+// reference on the entry.  The only ways that this can become false without the
+// entry being passed to its "deleter" are via Erase(), via Insert() when
+// an element with a duplicate key is inserted, or on destruction of the cache.
+//
+// The cache keeps two linked lists of items in the cache.  All items in the
+// cache are in one list or the other, and never both.  Items still referenced
+// by clients but erased from the cache are in neither list.  The lists are:
+// - in-use:  contains the items currently referenced by clients, in no
+//   particular order.  (This list is used for invariant checking.  If we
+//   removed the check, elements that would otherwise be on this list could be
+//   left as disconnected singleton lists.)
+// - LRU:  contains the items not currently referenced by clients, in LRU order
+// Elements are moved between these lists by the Ref() and Unref() methods,
+// when they detect an element in the cache acquiring or losing its only
+// external reference.
+
+// An entry is a variable length heap-allocated structure.  Entries
+// are kept in a circular doubly linked list ordered by access time.
+struct LRUHandle {
+  void* value;
+  void (*deleter)(const Slice&, void* value);
+  LRUHandle* next_hash;
+  LRUHandle* next;
+  LRUHandle* prev;
+  size_t charge;  // TODO(opt): Only allow uint32_t?
+  size_t key_length;
+  bool in_cache;     // Whether entry is in the cache.
+  uint32_t refs;     // References, including cache reference, if present.
+  uint32_t hash;     // Hash of key(); used for fast sharding and comparisons
+  char key_data[1];  // Beginning of key
+
+  Slice key() const {
+    // next_ is only equal to this if the LRU handle is the list head of an
+    // empty list. List heads never have meaningful keys.
+    assert(next != this);
+
+    return Slice(key_data, key_length);
+  }
+};
+
+// We provide our own simple hash table since it removes a whole bunch
+// of porting hacks and is also faster than some of the built-in hash
+// table implementations in some of the compiler/runtime combinations
+// we have tested.  E.g., readrandom speeds up by ~5% over the g++
+// 4.4.3's builtin hashtable.
+class HandleTable {
+ public:
+  HandleTable() : length_(0), elems_(0), list_(nullptr) { Resize(); }
+  ~HandleTable() { delete[] list_; }
+
+  LRUHandle* Lookup(const Slice& key, uint32_t hash) {
+    return *FindPointer(key, hash);
+  }
+
+  LRUHandle* Insert(LRUHandle* h) {
+    LRUHandle** ptr = FindPointer(h->key(), h->hash);
+    LRUHandle* old = *ptr;
+    h->next_hash = (old == nullptr ? nullptr : old->next_hash);
+    *ptr = h;
+    if (old == nullptr) {
+      ++elems_;
+      if (elems_ > length_) {
+        // Since each cache entry is fairly large, we aim for a small
+        // average linked list length (<= 1).
+        Resize();
+      }
+    }
+    return old;
+  }
+
+  LRUHandle* Remove(const Slice& key, uint32_t hash) {
+    LRUHandle** ptr = FindPointer(key, hash);
+    LRUHandle* result = *ptr;
+    if (result != nullptr) {
+      *ptr = result->next_hash;
+      --elems_;
+    }
+    return result;
+  }
+
+ private:
+  // The table consists of an array of buckets where each bucket is
+  // a linked list of cache entries that hash into the bucket.
+  uint32_t length_;
+  uint32_t elems_;
+  LRUHandle** list_;
+
+  // Return a pointer to slot that points to a cache entry that
+  // matches key/hash.  If there is no such cache entry, return a
+  // pointer to the trailing slot in the corresponding linked list.
+  LRUHandle** FindPointer(const Slice& key, uint32_t hash) {
+    LRUHandle** ptr = &list_[hash & (length_ - 1)];
+    while (*ptr != nullptr && ((*ptr)->hash != hash || key != (*ptr)->key())) {
+      ptr = &(*ptr)->next_hash;
+    }
+    return ptr;
+  }
+
+  void Resize() {
+    uint32_t new_length = 4;
+    while (new_length < elems_) {
+      new_length *= 2;
+    }
+    LRUHandle** new_list = new LRUHandle*[new_length];
+    memset(new_list, 0, sizeof(new_list[0]) * new_length);
+    uint32_t count = 0;
+    for (uint32_t i = 0; i < length_; i++) {
+      LRUHandle* h = list_[i];
+      while (h != nullptr) {
+        LRUHandle* next = h->next_hash;
+        uint32_t hash = h->hash;
+        LRUHandle** ptr = &new_list[hash & (new_length - 1)];
+        h->next_hash = *ptr;
+        *ptr = h;
+        h = next;
+        count++;
+      }
+    }
+    assert(elems_ == count);
+    delete[] list_;
+    list_ = new_list;
+    length_ = new_length;
+  }
+};
+
+// A single shard of sharded cache.
+class LRUCache {
+ public:
+  LRUCache();
+  ~LRUCache();
+
+  // Separate from constructor so caller can easily make an array of LRUCache
+  void SetCapacity(size_t capacity) { capacity_ = capacity; }
+
+  // Like Cache methods, but with an extra "hash" parameter.
+  Cache::Handle* Insert(const Slice& key, uint32_t hash, void* value,
+                        size_t charge,
+                        void (*deleter)(const Slice& key, void* value));
+  Cache::Handle* Lookup(const Slice& key, uint32_t hash);
+  void Release(Cache::Handle* handle);
+  void Erase(const Slice& key, uint32_t hash);
+  void Prune();
+  size_t TotalCharge() const {
+    MutexLock l(&mutex_);
+    return usage_;
+  }
+
+ private:
+  void LRU_Remove(LRUHandle* e);
+  void LRU_Append(LRUHandle* list, LRUHandle* e);
+  void Ref(LRUHandle* e);
+  void Unref(LRUHandle* e);
+  bool FinishErase(LRUHandle* e) EXCLUSIVE_LOCKS_REQUIRED(mutex_);
+
+  // Initialized before use.
+  size_t capacity_;
+
+  // mutex_ protects the following state.
+  mutable port::Mutex mutex_;
+  size_t usage_ GUARDED_BY(mutex_);
+
+  // Dummy head of LRU list.
+  // lru.prev is newest entry, lru.next is oldest entry.
+  // Entries have refs==1 and in_cache==true.
+  LRUHandle lru_ GUARDED_BY(mutex_);
+
+  // Dummy head of in-use list.
+  // Entries are in use by clients, and have refs >= 2 and in_cache==true.
+  LRUHandle in_use_ GUARDED_BY(mutex_);
+
+  HandleTable table_ GUARDED_BY(mutex_);
+};
+
+LRUCache::LRUCache() : capacity_(0), usage_(0) {
+  // Make empty circular linked lists.
+  lru_.next = &lru_;
+  lru_.prev = &lru_;
+  in_use_.next = &in_use_;
+  in_use_.prev = &in_use_;
+}
+
+LRUCache::~LRUCache() {
+  assert(in_use_.next == &in_use_);  // Error if caller has an unreleased handle
+  for (LRUHandle* e = lru_.next; e != &lru_;) {
+    LRUHandle* next = e->next;
+    assert(e->in_cache);
+    e->in_cache = false;
+    assert(e->refs == 1);  // Invariant of lru_ list.
+    Unref(e);
+    e = next;
+  }
+}
+
+void LRUCache::Ref(LRUHandle* e) {
+  if (e->refs == 1 && e->in_cache) {  // If on lru_ list, move to in_use_ list.
+    LRU_Remove(e);
+    LRU_Append(&in_use_, e);
+  }
+  e->refs++;
+}
+
+void LRUCache::Unref(LRUHandle* e) {
+  assert(e->refs > 0);
+  e->refs--;
+  if (e->refs == 0) {  // Deallocate.
+    assert(!e->in_cache);
+    (*e->deleter)(e->key(), e->value);
+    free(e);
+  } else if (e->in_cache && e->refs == 1) {
+    // No longer in use; move to lru_ list.
+    LRU_Remove(e);
+    LRU_Append(&lru_, e);
+  }
+}
+
+void LRUCache::LRU_Remove(LRUHandle* e) {
+  e->next->prev = e->prev;
+  e->prev->next = e->next;
+}
+
+void LRUCache::LRU_Append(LRUHandle* list, LRUHandle* e) {
+  // Make "e" newest entry by inserting just before *list
+  e->next = list;
+  e->prev = list->prev;
+  e->prev->next = e;
+  e->next->prev = e;
+}
+
+Cache::Handle* LRUCache::Lookup(const Slice& key, uint32_t hash) {
+  MutexLock l(&mutex_);
+  LRUHandle* e = table_.Lookup(key, hash);
+  if (e != nullptr) {
+    Ref(e);
+  }
+  return reinterpret_cast<Cache::Handle*>(e);
+}
+
+void LRUCache::Release(Cache::Handle* handle) {
+  MutexLock l(&mutex_);
+  Unref(reinterpret_cast<LRUHandle*>(handle));
+}
+
+Cache::Handle* LRUCache::Insert(const Slice& key, uint32_t hash, void* value,
+                                size_t charge,
+                                void (*deleter)(const Slice& key,
+                                                void* value)) {
+  MutexLock l(&mutex_);
+
+  LRUHandle* e =
+      reinterpret_cast<LRUHandle*>(malloc(sizeof(LRUHandle) - 1 + key.size()));
+  e->value = value;
+  e->deleter = deleter;
+  e->charge = charge;
+  e->key_length = key.size();
+  e->hash = hash;
+  e->in_cache = false;
+  e->refs = 1;  // for the returned handle.
+  std::memcpy(e->key_data, key.data(), key.size());
+
+  if (capacity_ > 0) {
+    e->refs++;  // for the cache's reference.
+    e->in_cache = true;
+    LRU_Append(&in_use_, e);
+    usage_ += charge;
+    FinishErase(table_.Insert(e));
+  } else {  // don't cache. (capacity_==0 is supported and turns off caching.)
+    // next is read by key() in an assert, so it must be initialized
+    e->next = nullptr;
+  }
+  while (usage_ > capacity_ && lru_.next != &lru_) {
+    LRUHandle* old = lru_.next;
+    assert(old->refs == 1);
+    bool erased = FinishErase(table_.Remove(old->key(), old->hash));
+    if (!erased) {  // to avoid unused variable when compiled NDEBUG
+      assert(erased);
+    }
+  }
+
+  return reinterpret_cast<Cache::Handle*>(e);
+}
+
+// If e != nullptr, finish removing *e from the cache; it has already been
+// removed from the hash table.  Return whether e != nullptr.
+bool LRUCache::FinishErase(LRUHandle* e) {
+  if (e != nullptr) {
+    assert(e->in_cache);
+    LRU_Remove(e);
+    e->in_cache = false;
+    usage_ -= e->charge;
+    Unref(e);
+  }
+  return e != nullptr;
+}
+
+void LRUCache::Erase(const Slice& key, uint32_t hash) {
+  MutexLock l(&mutex_);
+  FinishErase(table_.Remove(key, hash));
+}
+
+void LRUCache::Prune() {
+  MutexLock l(&mutex_);
+  while (lru_.next != &lru_) {
+    LRUHandle* e = lru_.next;
+    assert(e->refs == 1);
+    bool erased = FinishErase(table_.Remove(e->key(), e->hash));
+    if (!erased) {  // to avoid unused variable when compiled NDEBUG
+      assert(erased);
+    }
+  }
+}
+
+static const int kNumShardBits = 4;
+static const int kNumShards = 1 << kNumShardBits;
+
+class ShardedLRUCache : public Cache {
+ private:
+  LRUCache shard_[kNumShards];
+  port::Mutex id_mutex_;
+  uint64_t last_id_;
+
+  static inline uint32_t HashSlice(const Slice& s) {
+    return Hash(s.data(), s.size(), 0);
+  }
+
+  static uint32_t Shard(uint32_t hash) { return hash >> (32 - kNumShardBits); }
+
+ public:
+  explicit ShardedLRUCache(size_t capacity) : last_id_(0) {
+    const size_t per_shard = (capacity + (kNumShards - 1)) / kNumShards;
+    for (int s = 0; s < kNumShards; s++) {
+      shard_[s].SetCapacity(per_shard);
+    }
+  }
+  ~ShardedLRUCache() override {}
+  Handle* Insert(const Slice& key, void* value, size_t charge,
+                 void (*deleter)(const Slice& key, void* value)) override {
+    const uint32_t hash = HashSlice(key);
+    return shard_[Shard(hash)].Insert(key, hash, value, charge, deleter);
+  }
+  Handle* Lookup(const Slice& key) override {
+    const uint32_t hash = HashSlice(key);
+    return shard_[Shard(hash)].Lookup(key, hash);
+  }
+  void Release(Handle* handle) override {
+    LRUHandle* h = reinterpret_cast<LRUHandle*>(handle);
+    shard_[Shard(h->hash)].Release(handle);
+  }
+  void Erase(const Slice& key) override {
+    const uint32_t hash = HashSlice(key);
+    shard_[Shard(hash)].Erase(key, hash);
+  }
+  void* Value(Handle* handle) override {
+    return reinterpret_cast<LRUHandle*>(handle)->value;
+  }
+  uint64_t NewId() override {
+    MutexLock l(&id_mutex_);
+    return ++(last_id_);
+  }
+  void Prune() override {
+    for (int s = 0; s < kNumShards; s++) {
+      shard_[s].Prune();
+    }
+  }
+  size_t TotalCharge() const override {
+    size_t total = 0;
+    for (int s = 0; s < kNumShards; s++) {
+      total += shard_[s].TotalCharge();
+    }
+    return total;
+  }
+};
+
+}  // end anonymous namespace
+
+Cache* NewLRUCache(size_t capacity) { return new ShardedLRUCache(capacity); }
+
+}  // namespace leveldb

lib/leveldb/util/cache_test.cc

@@ -0,0 +1,229 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "leveldb/cache.h"
+
+#include <vector>
+
+#include "gtest/gtest.h"
+#include "util/coding.h"
+
+namespace leveldb {
+
+// Conversions between numeric keys/values and the types expected by Cache.
+static std::string EncodeKey(int k) {
+  std::string result;
+  PutFixed32(&result, k);
+  return result;
+}
+static int DecodeKey(const Slice& k) {
+  assert(k.size() == 4);
+  return DecodeFixed32(k.data());
+}
+static void* EncodeValue(uintptr_t v) { return reinterpret_cast<void*>(v); }
+static int DecodeValue(void* v) { return reinterpret_cast<uintptr_t>(v); }
+
+class CacheTest : public testing::Test {
+ public:
+  static void Deleter(const Slice& key, void* v) {
+    current_->deleted_keys_.push_back(DecodeKey(key));
+    current_->deleted_values_.push_back(DecodeValue(v));
+  }
+
+  static constexpr int kCacheSize = 1000;
+  std::vector<int> deleted_keys_;
+  std::vector<int> deleted_values_;
+  Cache* cache_;
+
+  CacheTest() : cache_(NewLRUCache(kCacheSize)) { current_ = this; }
+
+  ~CacheTest() { delete cache_; }
+
+  int Lookup(int key) {
+    Cache::Handle* handle = cache_->Lookup(EncodeKey(key));
+    const int r = (handle == nullptr) ? -1 : DecodeValue(cache_->Value(handle));
+    if (handle != nullptr) {
+      cache_->Release(handle);
+    }
+    return r;
+  }
+
+  void Insert(int key, int value, int charge = 1) {
+    cache_->Release(cache_->Insert(EncodeKey(key), EncodeValue(value), charge,
+                                   &CacheTest::Deleter));
+  }
+
+  Cache::Handle* InsertAndReturnHandle(int key, int value, int charge = 1) {
+    return cache_->Insert(EncodeKey(key), EncodeValue(value), charge,
+                          &CacheTest::Deleter);
+  }
+
+  void Erase(int key) { cache_->Erase(EncodeKey(key)); }
+  static CacheTest* current_;
+};
+CacheTest* CacheTest::current_;
+
+TEST_F(CacheTest, HitAndMiss) {
+  ASSERT_EQ(-1, Lookup(100));
+
+  Insert(100, 101);
+  ASSERT_EQ(101, Lookup(100));
+  ASSERT_EQ(-1, Lookup(200));
+  ASSERT_EQ(-1, Lookup(300));
+
+  Insert(200, 201);
+  ASSERT_EQ(101, Lookup(100));
+  ASSERT_EQ(201, Lookup(200));
+  ASSERT_EQ(-1, Lookup(300));
+
+  Insert(100, 102);
+  ASSERT_EQ(102, Lookup(100));
+  ASSERT_EQ(201, Lookup(200));
+  ASSERT_EQ(-1, Lookup(300));
+
+  ASSERT_EQ(1, deleted_keys_.size());
+  ASSERT_EQ(100, deleted_keys_[0]);
+  ASSERT_EQ(101, deleted_values_[0]);
+}
+
+TEST_F(CacheTest, Erase) {
+  Erase(200);
+  ASSERT_EQ(0, deleted_keys_.size());
+
+  Insert(100, 101);
+  Insert(200, 201);
+  Erase(100);
+  ASSERT_EQ(-1, Lookup(100));
+  ASSERT_EQ(201, Lookup(200));
+  ASSERT_EQ(1, deleted_keys_.size());
+  ASSERT_EQ(100, deleted_keys_[0]);
+  ASSERT_EQ(101, deleted_values_[0]);
+
+  Erase(100);
+  ASSERT_EQ(-1, Lookup(100));
+  ASSERT_EQ(201, Lookup(200));
+  ASSERT_EQ(1, deleted_keys_.size());
+}
+
+TEST_F(CacheTest, EntriesArePinned) {
+  Insert(100, 101);
+  Cache::Handle* h1 = cache_->Lookup(EncodeKey(100));
+  ASSERT_EQ(101, DecodeValue(cache_->Value(h1)));
+
+  Insert(100, 102);
+  Cache::Handle* h2 = cache_->Lookup(EncodeKey(100));
+  ASSERT_EQ(102, DecodeValue(cache_->Value(h2)));
+  ASSERT_EQ(0, deleted_keys_.size());
+
+  cache_->Release(h1);
+  ASSERT_EQ(1, deleted_keys_.size());
+  ASSERT_EQ(100, deleted_keys_[0]);
+  ASSERT_EQ(101, deleted_values_[0]);
+
+  Erase(100);
+  ASSERT_EQ(-1, Lookup(100));
+  ASSERT_EQ(1, deleted_keys_.size());
+
+  cache_->Release(h2);
+  ASSERT_EQ(2, deleted_keys_.size());
+  ASSERT_EQ(100, deleted_keys_[1]);
+  ASSERT_EQ(102, deleted_values_[1]);
+}
+
+TEST_F(CacheTest, EvictionPolicy) {
+  Insert(100, 101);
+  Insert(200, 201);
+  Insert(300, 301);
+  Cache::Handle* h = cache_->Lookup(EncodeKey(300));
+
+  // Frequently used entry must be kept around,
+  // as must things that are still in use.
+  for (int i = 0; i < kCacheSize + 100; i++) {
+    Insert(1000 + i, 2000 + i);
+    ASSERT_EQ(2000 + i, Lookup(1000 + i));
+    ASSERT_EQ(101, Lookup(100));
+  }
+  ASSERT_EQ(101, Lookup(100));
+  ASSERT_EQ(-1, Lookup(200));
+  ASSERT_EQ(301, Lookup(300));
+  cache_->Release(h);
+}
+
+TEST_F(CacheTest, UseExceedsCacheSize) {
+  // Overfill the cache, keeping handles on all inserted entries.
+  std::vector<Cache::Handle*> h;
+  for (int i = 0; i < kCacheSize + 100; i++) {
+    h.push_back(InsertAndReturnHandle(1000 + i, 2000 + i));
+  }
+
+  // Check that all the entries can be found in the cache.
+  for (int i = 0; i < h.size(); i++) {
+    ASSERT_EQ(2000 + i, Lookup(1000 + i));
+  }
+
+  for (int i = 0; i < h.size(); i++) {
+    cache_->Release(h[i]);
+  }
+}
+
+TEST_F(CacheTest, HeavyEntries) {
+  // Add a bunch of light and heavy entries and then count the combined
+  // size of items still in the cache, which must be approximately the
+  // same as the total capacity.
+  const int kLight = 1;
+  const int kHeavy = 10;
+  int added = 0;
+  int index = 0;
+  while (added < 2 * kCacheSize) {
+    const int weight = (index & 1) ? kLight : kHeavy;
+    Insert(index, 1000 + index, weight);
+    added += weight;
+    index++;
+  }
+
+  int cached_weight = 0;
+  for (int i = 0; i < index; i++) {
+    const int weight = (i & 1 ? kLight : kHeavy);
+    int r = Lookup(i);
+    if (r >= 0) {
+      cached_weight += weight;
+      ASSERT_EQ(1000 + i, r);
+    }
+  }
+  ASSERT_LE(cached_weight, kCacheSize + kCacheSize / 10);
+}
+
+TEST_F(CacheTest, NewId) {
+  uint64_t a = cache_->NewId();
+  uint64_t b = cache_->NewId();
+  ASSERT_NE(a, b);
+}
+
+TEST_F(CacheTest, Prune) {
+  Insert(1, 100);
+  Insert(2, 200);
+
+  Cache::Handle* handle = cache_->Lookup(EncodeKey(1));
+  ASSERT_TRUE(handle);
+  cache_->Prune();
+  cache_->Release(handle);
+
+  ASSERT_EQ(100, Lookup(1));
+  ASSERT_EQ(-1, Lookup(2));
+}
+
+TEST_F(CacheTest, ZeroSizeCache) {
+  delete cache_;
+  cache_ = NewLRUCache(0);
+
+  Insert(1, 100);
+  ASSERT_EQ(-1, Lookup(1));
+}
+
+}  // namespace leveldb
+
+int main(int argc, char** argv) {
+  testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}

lib/leveldb/util/coding.cc

@@ -0,0 +1,166 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "util/coding.h"
+
+namespace leveldb {
+
+void PutFixed32(std::string* dst, uint32_t value) {
+  char buf[sizeof(value)];
+  EncodeFixed32(buf, value);
+  dst->append(buf, sizeof(buf));
+}
+
+void PutFixed64(std::string* dst, uint64_t value) {
+  char buf[sizeof(value)];
+  EncodeFixed64(buf, value);
+  dst->append(buf, sizeof(buf));
+}
+
+char* EncodeVarint32(char* dst, uint32_t v) {
+  // Operate on characters as unsigneds
+  uint8_t* ptr = reinterpret_cast<uint8_t*>(dst);
+  static const int B = 128;
+  if (v < (1 << 7)) {
+    *(ptr++) = v;
+  } else if (v < (1 << 14)) {
+    *(ptr++) = v | B;
+    *(ptr++) = v >> 7;
+  } else if (v < (1 << 21)) {
+    *(ptr++) = v | B;
+    *(ptr++) = (v >> 7) | B;
+    *(ptr++) = v >> 14;
+  } else if (v < (1 << 28)) {
+    *(ptr++) = v | B;
+    *(ptr++) = (v >> 7) | B;
+    *(ptr++) = (v >> 14) | B;
+    *(ptr++) = v >> 21;
+  } else {
+    *(ptr++) = v | B;
+    *(ptr++) = (v >> 7) | B;
+    *(ptr++) = (v >> 14) | B;
+    *(ptr++) = (v >> 21) | B;
+    *(ptr++) = v >> 28;
+  }
+  return reinterpret_cast<char*>(ptr);
+}
+
+void PutVarint32(std::string* dst, uint32_t v) {
+  char buf[5];
+  char* ptr = EncodeVarint32(buf, v);
+  dst->append(buf, ptr - buf);
+}
+
+char* EncodeVarint64(char* dst, uint64_t v) {
+  static const int B = 128;
+  uint8_t* ptr = reinterpret_cast<uint8_t*>(dst);
+  while (v >= B) {
+    *(ptr++) = v | B;
+    v >>= 7;
+  }
+  *(ptr++) = static_cast<uint8_t>(v);
+  return reinterpret_cast<char*>(ptr);
+}
+
+void PutVarint64(std::string* dst, uint64_t v) {
+  char buf[10];
+  char* ptr = EncodeVarint64(buf, v);
+  dst->append(buf, ptr - buf);
+}
+
+void PutLengthPrefixedSlice(std::string* dst, const Slice& value) {
+  PutVarint32(dst, value.size());
+  dst->append(value.data(), value.size());
+}
+
+int VarintLength(uint64_t v) {
+  int len = 1;
+  while (v >= 128) {
+    v >>= 7;
+    len++;
+  }
+  return len;
+}
+
+const char* GetVarint32PtrFallback(const char* p, const char* limit,
+                                   uint32_t* value) {
+  uint32_t result = 0;
+  for (uint32_t shift = 0; shift <= 28 && p < limit; shift += 7) {
+    uint32_t byte = *(reinterpret_cast<const uint8_t*>(p));
+    p++;
+    if (byte & 128) {
+      // More bytes are present
+      result |= ((byte & 127) << shift);
+    } else {
+      result |= (byte << shift);
+      *value = result;
+      return reinterpret_cast<const char*>(p);
+    }
+  }
+  return nullptr;
+}
+
+bool GetVarint32(Slice* input, uint32_t* value) {
+  const char* p = input->data();
+  const char* limit = p + input->size();
+  const char* q = GetVarint32Ptr(p, limit, value);
+  if (q == nullptr) {
+    return false;
+  } else {
+    *input = Slice(q, limit - q);
+    return true;
+  }
+}
+
+const char* GetVarint64Ptr(const char* p, const char* limit, uint64_t* value) {
+  uint64_t result = 0;
+  for (uint32_t shift = 0; shift <= 63 && p < limit; shift += 7) {
+    uint64_t byte = *(reinterpret_cast<const uint8_t*>(p));
+    p++;
+    if (byte & 128) {
+      // More bytes are present
+      result |= ((byte & 127) << shift);
+    } else {
+      result |= (byte << shift);
+      *value = result;
+      return reinterpret_cast<const char*>(p);
+    }
+  }
+  return nullptr;
+}
+
+bool GetVarint64(Slice* input, uint64_t* value) {
+  const char* p = input->data();
+  const char* limit = p + input->size();
+  const char* q = GetVarint64Ptr(p, limit, value);
+  if (q == nullptr) {
+    return false;
+  } else {
+    *input = Slice(q, limit - q);
+    return true;
+  }
+}
+
+const char* GetLengthPrefixedSlice(const char* p, const char* limit,
+                                   Slice* result) {
+  uint32_t len;
+  p = GetVarint32Ptr(p, limit, &len);
+  if (p == nullptr) return nullptr;
+  if (p + len > limit) return nullptr;
+  *result = Slice(p, len);
+  return p + len;
+}
+
+bool GetLengthPrefixedSlice(Slice* input, Slice* result) {
+  uint32_t len;
+  if (GetVarint32(input, &len) && input->size() >= len) {
+    *result = Slice(input->data(), len);
+    input->remove_prefix(len);
+    return true;
+  } else {
+    return false;
+  }
+}
+
+}  // namespace leveldb

lib/leveldb/util/coding.h

@@ -0,0 +1,122 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+//
+// Endian-neutral encoding:
+// * Fixed-length numbers are encoded with least-significant byte first
+// * In addition we support variable length "varint" encoding
+// * Strings are encoded prefixed by their length in varint format
+
+#ifndef STORAGE_LEVELDB_UTIL_CODING_H_
+#define STORAGE_LEVELDB_UTIL_CODING_H_
+
+#include <cstdint>
+#include <cstring>
+#include <string>
+
+#include "leveldb/slice.h"
+#include "port/port.h"
+
+namespace leveldb {
+
+// Standard Put... routines append to a string
+void PutFixed32(std::string* dst, uint32_t value);
+void PutFixed64(std::string* dst, uint64_t value);
+void PutVarint32(std::string* dst, uint32_t value);
+void PutVarint64(std::string* dst, uint64_t value);
+void PutLengthPrefixedSlice(std::string* dst, const Slice& value);
+
+// Standard Get... routines parse a value from the beginning of a Slice
+// and advance the slice past the parsed value.
+bool GetVarint32(Slice* input, uint32_t* value);
+bool GetVarint64(Slice* input, uint64_t* value);
+bool GetLengthPrefixedSlice(Slice* input, Slice* result);
+
+// Pointer-based variants of GetVarint...  These either store a value
+// in *v and return a pointer just past the parsed value, or return
+// nullptr on error.  These routines only look at bytes in the range
+// [p..limit-1]
+const char* GetVarint32Ptr(const char* p, const char* limit, uint32_t* v);
+const char* GetVarint64Ptr(const char* p, const char* limit, uint64_t* v);
+
+// Returns the length of the varint32 or varint64 encoding of "v"
+int VarintLength(uint64_t v);
+
+// Lower-level versions of Put... that write directly into a character buffer
+// and return a pointer just past the last byte written.
+// REQUIRES: dst has enough space for the value being written
+char* EncodeVarint32(char* dst, uint32_t value);
+char* EncodeVarint64(char* dst, uint64_t value);
+
+// Lower-level versions of Put... that write directly into a character buffer
+// REQUIRES: dst has enough space for the value being written
+
+inline void EncodeFixed32(char* dst, uint32_t value) {
+  uint8_t* const buffer = reinterpret_cast<uint8_t*>(dst);
+
+  // Recent clang and gcc optimize this to a single mov / str instruction.
+  buffer[0] = static_cast<uint8_t>(value);
+  buffer[1] = static_cast<uint8_t>(value >> 8);
+  buffer[2] = static_cast<uint8_t>(value >> 16);
+  buffer[3] = static_cast<uint8_t>(value >> 24);
+}
+
+inline void EncodeFixed64(char* dst, uint64_t value) {
+  uint8_t* const buffer = reinterpret_cast<uint8_t*>(dst);
+
+  // Recent clang and gcc optimize this to a single mov / str instruction.
+  buffer[0] = static_cast<uint8_t>(value);
+  buffer[1] = static_cast<uint8_t>(value >> 8);
+  buffer[2] = static_cast<uint8_t>(value >> 16);
+  buffer[3] = static_cast<uint8_t>(value >> 24);
+  buffer[4] = static_cast<uint8_t>(value >> 32);
+  buffer[5] = static_cast<uint8_t>(value >> 40);
+  buffer[6] = static_cast<uint8_t>(value >> 48);
+  buffer[7] = static_cast<uint8_t>(value >> 56);
+}
+
+// Lower-level versions of Get... that read directly from a character buffer
+// without any bounds checking.
+
+inline uint32_t DecodeFixed32(const char* ptr) {
+  const uint8_t* const buffer = reinterpret_cast<const uint8_t*>(ptr);
+
+  // Recent clang and gcc optimize this to a single mov / ldr instruction.
+  return (static_cast<uint32_t>(buffer[0])) |
+         (static_cast<uint32_t>(buffer[1]) << 8) |
+         (static_cast<uint32_t>(buffer[2]) << 16) |
+         (static_cast<uint32_t>(buffer[3]) << 24);
+}
+
+inline uint64_t DecodeFixed64(const char* ptr) {
+  const uint8_t* const buffer = reinterpret_cast<const uint8_t*>(ptr);
+
+  // Recent clang and gcc optimize this to a single mov / ldr instruction.
+  return (static_cast<uint64_t>(buffer[0])) |
+         (static_cast<uint64_t>(buffer[1]) << 8) |
+         (static_cast<uint64_t>(buffer[2]) << 16) |
+         (static_cast<uint64_t>(buffer[3]) << 24) |
+         (static_cast<uint64_t>(buffer[4]) << 32) |
+         (static_cast<uint64_t>(buffer[5]) << 40) |
+         (static_cast<uint64_t>(buffer[6]) << 48) |
+         (static_cast<uint64_t>(buffer[7]) << 56);
+}
+
+// Internal routine for use by fallback path of GetVarint32Ptr
+const char* GetVarint32PtrFallback(const char* p, const char* limit,
+                                   uint32_t* value);
+inline const char* GetVarint32Ptr(const char* p, const char* limit,
+                                  uint32_t* value) {
+  if (p < limit) {
+    uint32_t result = *(reinterpret_cast<const uint8_t*>(p));
+    if ((result & 128) == 0) {
+      *value = result;
+      return p + 1;
+    }
+  }
+  return GetVarint32PtrFallback(p, limit, value);
+}
+
+}  // namespace leveldb
+
+#endif  // STORAGE_LEVELDB_UTIL_CODING_H_

lib/leveldb/util/coding_test.cc

@@ -0,0 +1,198 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "util/coding.h"
+
+#include <vector>
+
+#include "gtest/gtest.h"
+
+namespace leveldb {
+
+TEST(Coding, Fixed32) {
+  std::string s;
+  for (uint32_t v = 0; v < 100000; v++) {
+    PutFixed32(&s, v);
+  }
+
+  const char* p = s.data();
+  for (uint32_t v = 0; v < 100000; v++) {
+    uint32_t actual = DecodeFixed32(p);
+    ASSERT_EQ(v, actual);
+    p += sizeof(uint32_t);
+  }
+}
+
+TEST(Coding, Fixed64) {
+  std::string s;
+  for (int power = 0; power <= 63; power++) {
+    uint64_t v = static_cast<uint64_t>(1) << power;
+    PutFixed64(&s, v - 1);
+    PutFixed64(&s, v + 0);
+    PutFixed64(&s, v + 1);
+  }
+
+  const char* p = s.data();
+  for (int power = 0; power <= 63; power++) {
+    uint64_t v = static_cast<uint64_t>(1) << power;
+    uint64_t actual;
+    actual = DecodeFixed64(p);
+    ASSERT_EQ(v - 1, actual);
+    p += sizeof(uint64_t);
+
+    actual = DecodeFixed64(p);
+    ASSERT_EQ(v + 0, actual);
+    p += sizeof(uint64_t);
+
+    actual = DecodeFixed64(p);
+    ASSERT_EQ(v + 1, actual);
+    p += sizeof(uint64_t);
+  }
+}
+
+// Test that encoding routines generate little-endian encodings
+TEST(Coding, EncodingOutput) {
+  std::string dst;
+  PutFixed32(&dst, 0x04030201);
+  ASSERT_EQ(4, dst.size());
+  ASSERT_EQ(0x01, static_cast<int>(dst[0]));
+  ASSERT_EQ(0x02, static_cast<int>(dst[1]));
+  ASSERT_EQ(0x03, static_cast<int>(dst[2]));
+  ASSERT_EQ(0x04, static_cast<int>(dst[3]));
+
+  dst.clear();
+  PutFixed64(&dst, 0x0807060504030201ull);
+  ASSERT_EQ(8, dst.size());
+  ASSERT_EQ(0x01, static_cast<int>(dst[0]));
+  ASSERT_EQ(0x02, static_cast<int>(dst[1]));
+  ASSERT_EQ(0x03, static_cast<int>(dst[2]));
+  ASSERT_EQ(0x04, static_cast<int>(dst[3]));
+  ASSERT_EQ(0x05, static_cast<int>(dst[4]));
+  ASSERT_EQ(0x06, static_cast<int>(dst[5]));
+  ASSERT_EQ(0x07, static_cast<int>(dst[6]));
+  ASSERT_EQ(0x08, static_cast<int>(dst[7]));
+}
+
+TEST(Coding, Varint32) {
+  std::string s;
+  for (uint32_t i = 0; i < (32 * 32); i++) {
+    uint32_t v = (i / 32) << (i % 32);
+    PutVarint32(&s, v);
+  }
+
+  const char* p = s.data();
+  const char* limit = p + s.size();
+  for (uint32_t i = 0; i < (32 * 32); i++) {
+    uint32_t expected = (i / 32) << (i % 32);
+    uint32_t actual;
+    const char* start = p;
+    p = GetVarint32Ptr(p, limit, &actual);
+    ASSERT_TRUE(p != nullptr);
+    ASSERT_EQ(expected, actual);
+    ASSERT_EQ(VarintLength(actual), p - start);
+  }
+  ASSERT_EQ(p, s.data() + s.size());
+}
+
+TEST(Coding, Varint64) {
+  // Construct the list of values to check
+  std::vector<uint64_t> values;
+  // Some special values
+  values.push_back(0);
+  values.push_back(100);
+  values.push_back(~static_cast<uint64_t>(0));
+  values.push_back(~static_cast<uint64_t>(0) - 1);
+  for (uint32_t k = 0; k < 64; k++) {
+    // Test values near powers of two
+    const uint64_t power = 1ull << k;
+    values.push_back(power);
+    values.push_back(power - 1);
+    values.push_back(power + 1);
+  }
+
+  std::string s;
+  for (size_t i = 0; i < values.size(); i++) {
+    PutVarint64(&s, values[i]);
+  }
+
+  const char* p = s.data();
+  const char* limit = p + s.size();
+  for (size_t i = 0; i < values.size(); i++) {
+    ASSERT_TRUE(p < limit);
+    uint64_t actual;
+    const char* start = p;
+    p = GetVarint64Ptr(p, limit, &actual);
+    ASSERT_TRUE(p != nullptr);
+    ASSERT_EQ(values[i], actual);
+    ASSERT_EQ(VarintLength(actual), p - start);
+  }
+  ASSERT_EQ(p, limit);
+}
+
+TEST(Coding, Varint32Overflow) {
+  uint32_t result;
+  std::string input("\x81\x82\x83\x84\x85\x11");
+  ASSERT_TRUE(GetVarint32Ptr(input.data(), input.data() + input.size(),
+                             &result) == nullptr);
+}
+
+TEST(Coding, Varint32Truncation) {
+  uint32_t large_value = (1u << 31) + 100;
+  std::string s;
+  PutVarint32(&s, large_value);
+  uint32_t result;
+  for (size_t len = 0; len < s.size() - 1; len++) {
+    ASSERT_TRUE(GetVarint32Ptr(s.data(), s.data() + len, &result) == nullptr);
+  }
+  ASSERT_TRUE(GetVarint32Ptr(s.data(), s.data() + s.size(), &result) !=
+              nullptr);
+  ASSERT_EQ(large_value, result);
+}
+
+TEST(Coding, Varint64Overflow) {
+  uint64_t result;
+  std::string input("\x81\x82\x83\x84\x85\x81\x82\x83\x84\x85\x11");
+  ASSERT_TRUE(GetVarint64Ptr(input.data(), input.data() + input.size(),
+                             &result) == nullptr);
+}
+
+TEST(Coding, Varint64Truncation) {
+  uint64_t large_value = (1ull << 63) + 100ull;
+  std::string s;
+  PutVarint64(&s, large_value);
+  uint64_t result;
+  for (size_t len = 0; len < s.size() - 1; len++) {
+    ASSERT_TRUE(GetVarint64Ptr(s.data(), s.data() + len, &result) == nullptr);
+  }
+  ASSERT_TRUE(GetVarint64Ptr(s.data(), s.data() + s.size(), &result) !=
+              nullptr);
+  ASSERT_EQ(large_value, result);
+}
+
+TEST(Coding, Strings) {
+  std::string s;
+  PutLengthPrefixedSlice(&s, Slice(""));
+  PutLengthPrefixedSlice(&s, Slice("foo"));
+  PutLengthPrefixedSlice(&s, Slice("bar"));
+  PutLengthPrefixedSlice(&s, Slice(std::string(200, 'x')));
+
+  Slice input(s);
+  Slice v;
+  ASSERT_TRUE(GetLengthPrefixedSlice(&input, &v));
+  ASSERT_EQ("", v.ToString());
+  ASSERT_TRUE(GetLengthPrefixedSlice(&input, &v));
+  ASSERT_EQ("foo", v.ToString());
+  ASSERT_TRUE(GetLengthPrefixedSlice(&input, &v));
+  ASSERT_EQ("bar", v.ToString());
+  ASSERT_TRUE(GetLengthPrefixedSlice(&input, &v));
+  ASSERT_EQ(std::string(200, 'x'), v.ToString());
+  ASSERT_EQ("", input.ToString());
+}
+
+}  // namespace leveldb
+
+int main(int argc, char** argv) {
+  testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}

lib/leveldb/util/comparator.cc

@@ -0,0 +1,75 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "leveldb/comparator.h"
+
+#include <algorithm>
+#include <cstdint>
+#include <string>
+#include <type_traits>
+
+#include "leveldb/slice.h"
+#include "util/logging.h"
+#include "util/no_destructor.h"
+
+namespace leveldb {
+
+Comparator::~Comparator() = default;
+
+namespace {
+class BytewiseComparatorImpl : public Comparator {
+ public:
+  BytewiseComparatorImpl() = default;
+
+  const char* Name() const override { return "leveldb.BytewiseComparator"; }
+
+  int Compare(const Slice& a, const Slice& b) const override {
+    return a.compare(b);
+  }
+
+  void FindShortestSeparator(std::string* start,
+                             const Slice& limit) const override {
+    // Find length of common prefix
+    size_t min_length = std::min(start->size(), limit.size());
+    size_t diff_index = 0;
+    while ((diff_index < min_length) &&
+           ((*start)[diff_index] == limit[diff_index])) {
+      diff_index++;
+    }
+
+    if (diff_index >= min_length) {
+      // Do not shorten if one string is a prefix of the other
+    } else {
+      uint8_t diff_byte = static_cast<uint8_t>((*start)[diff_index]);
+      if (diff_byte < static_cast<uint8_t>(0xff) &&
+          diff_byte + 1 < static_cast<uint8_t>(limit[diff_index])) {
+        (*start)[diff_index]++;
+        start->resize(diff_index + 1);
+        assert(Compare(*start, limit) < 0);
+      }
+    }
+  }
+
+  void FindShortSuccessor(std::string* key) const override {
+    // Find first character that can be incremented
+    size_t n = key->size();
+    for (size_t i = 0; i < n; i++) {
+      const uint8_t byte = (*key)[i];
+      if (byte != static_cast<uint8_t>(0xff)) {
+        (*key)[i] = byte + 1;
+        key->resize(i + 1);
+        return;
+      }
+    }
+    // *key is a run of 0xffs.  Leave it alone.
+  }
+};
+}  // namespace
+
+const Comparator* BytewiseComparator() {
+  static NoDestructor<BytewiseComparatorImpl> singleton;
+  return singleton.get();
+}
+
+}  // namespace leveldb

lib/leveldb/util/crc32c.cc

@@ -0,0 +1,380 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+//
+// A portable implementation of crc32c.
+
+#include "util/crc32c.h"
+
+#include <cstddef>
+#include <cstdint>
+
+#include "port/port.h"
+#include "util/coding.h"
+
+namespace leveldb {
+namespace crc32c {
+
+namespace {
+
+const uint32_t kByteExtensionTable[256] = {
+    0x00000000, 0xf26b8303, 0xe13b70f7, 0x1350f3f4, 0xc79a971f, 0x35f1141c,
+    0x26a1e7e8, 0xd4ca64eb, 0x8ad958cf, 0x78b2dbcc, 0x6be22838, 0x9989ab3b,
+    0x4d43cfd0, 0xbf284cd3, 0xac78bf27, 0x5e133c24, 0x105ec76f, 0xe235446c,
+    0xf165b798, 0x030e349b, 0xd7c45070, 0x25afd373, 0x36ff2087, 0xc494a384,
+    0x9a879fa0, 0x68ec1ca3, 0x7bbcef57, 0x89d76c54, 0x5d1d08bf, 0xaf768bbc,
+    0xbc267848, 0x4e4dfb4b, 0x20bd8ede, 0xd2d60ddd, 0xc186fe29, 0x33ed7d2a,
+    0xe72719c1, 0x154c9ac2, 0x061c6936, 0xf477ea35, 0xaa64d611, 0x580f5512,
+    0x4b5fa6e6, 0xb93425e5, 0x6dfe410e, 0x9f95c20d, 0x8cc531f9, 0x7eaeb2fa,
+    0x30e349b1, 0xc288cab2, 0xd1d83946, 0x23b3ba45, 0xf779deae, 0x05125dad,
+    0x1642ae59, 0xe4292d5a, 0xba3a117e, 0x4851927d, 0x5b016189, 0xa96ae28a,
+    0x7da08661, 0x8fcb0562, 0x9c9bf696, 0x6ef07595, 0x417b1dbc, 0xb3109ebf,
+    0xa0406d4b, 0x522bee48, 0x86e18aa3, 0x748a09a0, 0x67dafa54, 0x95b17957,
+    0xcba24573, 0x39c9c670, 0x2a993584, 0xd8f2b687, 0x0c38d26c, 0xfe53516f,
+    0xed03a29b, 0x1f682198, 0x5125dad3, 0xa34e59d0, 0xb01eaa24, 0x42752927,
+    0x96bf4dcc, 0x64d4cecf, 0x77843d3b, 0x85efbe38, 0xdbfc821c, 0x2997011f,
+    0x3ac7f2eb, 0xc8ac71e8, 0x1c661503, 0xee0d9600, 0xfd5d65f4, 0x0f36e6f7,
+    0x61c69362, 0x93ad1061, 0x80fde395, 0x72966096, 0xa65c047d, 0x5437877e,
+    0x4767748a, 0xb50cf789, 0xeb1fcbad, 0x197448ae, 0x0a24bb5a, 0xf84f3859,
+    0x2c855cb2, 0xdeeedfb1, 0xcdbe2c45, 0x3fd5af46, 0x7198540d, 0x83f3d70e,
+    0x90a324fa, 0x62c8a7f9, 0xb602c312, 0x44694011, 0x5739b3e5, 0xa55230e6,
+    0xfb410cc2, 0x092a8fc1, 0x1a7a7c35, 0xe811ff36, 0x3cdb9bdd, 0xceb018de,
+    0xdde0eb2a, 0x2f8b6829, 0x82f63b78, 0x709db87b, 0x63cd4b8f, 0x91a6c88c,
+    0x456cac67, 0xb7072f64, 0xa457dc90, 0x563c5f93, 0x082f63b7, 0xfa44e0b4,
+    0xe9141340, 0x1b7f9043, 0xcfb5f4a8, 0x3dde77ab, 0x2e8e845f, 0xdce5075c,
+    0x92a8fc17, 0x60c37f14, 0x73938ce0, 0x81f80fe3, 0x55326b08, 0xa759e80b,
+    0xb4091bff, 0x466298fc, 0x1871a4d8, 0xea1a27db, 0xf94ad42f, 0x0b21572c,
+    0xdfeb33c7, 0x2d80b0c4, 0x3ed04330, 0xccbbc033, 0xa24bb5a6, 0x502036a5,
+    0x4370c551, 0xb11b4652, 0x65d122b9, 0x97baa1ba, 0x84ea524e, 0x7681d14d,
+    0x2892ed69, 0xdaf96e6a, 0xc9a99d9e, 0x3bc21e9d, 0xef087a76, 0x1d63f975,
+    0x0e330a81, 0xfc588982, 0xb21572c9, 0x407ef1ca, 0x532e023e, 0xa145813d,
+    0x758fe5d6, 0x87e466d5, 0x94b49521, 0x66df1622, 0x38cc2a06, 0xcaa7a905,
+    0xd9f75af1, 0x2b9cd9f2, 0xff56bd19, 0x0d3d3e1a, 0x1e6dcdee, 0xec064eed,
+    0xc38d26c4, 0x31e6a5c7, 0x22b65633, 0xd0ddd530, 0x0417b1db, 0xf67c32d8,
+    0xe52cc12c, 0x1747422f, 0x49547e0b, 0xbb3ffd08, 0xa86f0efc, 0x5a048dff,
+    0x8ecee914, 0x7ca56a17, 0x6ff599e3, 0x9d9e1ae0, 0xd3d3e1ab, 0x21b862a8,
+    0x32e8915c, 0xc083125f, 0x144976b4, 0xe622f5b7, 0xf5720643, 0x07198540,
+    0x590ab964, 0xab613a67, 0xb831c993, 0x4a5a4a90, 0x9e902e7b, 0x6cfbad78,
+    0x7fab5e8c, 0x8dc0dd8f, 0xe330a81a, 0x115b2b19, 0x020bd8ed, 0xf0605bee,
+    0x24aa3f05, 0xd6c1bc06, 0xc5914ff2, 0x37faccf1, 0x69e9f0d5, 0x9b8273d6,
+    0x88d28022, 0x7ab90321, 0xae7367ca, 0x5c18e4c9, 0x4f48173d, 0xbd23943e,
+    0xf36e6f75, 0x0105ec76, 0x12551f82, 0xe03e9c81, 0x34f4f86a, 0xc69f7b69,
+    0xd5cf889d, 0x27a40b9e, 0x79b737ba, 0x8bdcb4b9, 0x988c474d, 0x6ae7c44e,
+    0xbe2da0a5, 0x4c4623a6, 0x5f16d052, 0xad7d5351};
+
+const uint32_t kStrideExtensionTable0[256] = {
+    0x00000000, 0x30d23865, 0x61a470ca, 0x517648af, 0xc348e194, 0xf39ad9f1,
+    0xa2ec915e, 0x923ea93b, 0x837db5d9, 0xb3af8dbc, 0xe2d9c513, 0xd20bfd76,
+    0x4035544d, 0x70e76c28, 0x21912487, 0x11431ce2, 0x03171d43, 0x33c52526,
+    0x62b36d89, 0x526155ec, 0xc05ffcd7, 0xf08dc4b2, 0xa1fb8c1d, 0x9129b478,
+    0x806aa89a, 0xb0b890ff, 0xe1ced850, 0xd11ce035, 0x4322490e, 0x73f0716b,
+    0x228639c4, 0x125401a1, 0x062e3a86, 0x36fc02e3, 0x678a4a4c, 0x57587229,
+    0xc566db12, 0xf5b4e377, 0xa4c2abd8, 0x941093bd, 0x85538f5f, 0xb581b73a,
+    0xe4f7ff95, 0xd425c7f0, 0x461b6ecb, 0x76c956ae, 0x27bf1e01, 0x176d2664,
+    0x053927c5, 0x35eb1fa0, 0x649d570f, 0x544f6f6a, 0xc671c651, 0xf6a3fe34,
+    0xa7d5b69b, 0x97078efe, 0x8644921c, 0xb696aa79, 0xe7e0e2d6, 0xd732dab3,
+    0x450c7388, 0x75de4bed, 0x24a80342, 0x147a3b27, 0x0c5c750c, 0x3c8e4d69,
+    0x6df805c6, 0x5d2a3da3, 0xcf149498, 0xffc6acfd, 0xaeb0e452, 0x9e62dc37,
+    0x8f21c0d5, 0xbff3f8b0, 0xee85b01f, 0xde57887a, 0x4c692141, 0x7cbb1924,
+    0x2dcd518b, 0x1d1f69ee, 0x0f4b684f, 0x3f99502a, 0x6eef1885, 0x5e3d20e0,
+    0xcc0389db, 0xfcd1b1be, 0xada7f911, 0x9d75c174, 0x8c36dd96, 0xbce4e5f3,
+    0xed92ad5c, 0xdd409539, 0x4f7e3c02, 0x7fac0467, 0x2eda4cc8, 0x1e0874ad,
+    0x0a724f8a, 0x3aa077ef, 0x6bd63f40, 0x5b040725, 0xc93aae1e, 0xf9e8967b,
+    0xa89eded4, 0x984ce6b1, 0x890ffa53, 0xb9ddc236, 0xe8ab8a99, 0xd879b2fc,
+    0x4a471bc7, 0x7a9523a2, 0x2be36b0d, 0x1b315368, 0x096552c9, 0x39b76aac,
+    0x68c12203, 0x58131a66, 0xca2db35d, 0xfaff8b38, 0xab89c397, 0x9b5bfbf2,
+    0x8a18e710, 0xbacadf75, 0xebbc97da, 0xdb6eafbf, 0x49500684, 0x79823ee1,
+    0x28f4764e, 0x18264e2b, 0x18b8ea18, 0x286ad27d, 0x791c9ad2, 0x49cea2b7,
+    0xdbf00b8c, 0xeb2233e9, 0xba547b46, 0x8a864323, 0x9bc55fc1, 0xab1767a4,
+    0xfa612f0b, 0xcab3176e, 0x588dbe55, 0x685f8630, 0x3929ce9f, 0x09fbf6fa,
+    0x1baff75b, 0x2b7dcf3e, 0x7a0b8791, 0x4ad9bff4, 0xd8e716cf, 0xe8352eaa,
+    0xb9436605, 0x89915e60, 0x98d24282, 0xa8007ae7, 0xf9763248, 0xc9a40a2d,
+    0x5b9aa316, 0x6b489b73, 0x3a3ed3dc, 0x0aecebb9, 0x1e96d09e, 0x2e44e8fb,
+    0x7f32a054, 0x4fe09831, 0xddde310a, 0xed0c096f, 0xbc7a41c0, 0x8ca879a5,
+    0x9deb6547, 0xad395d22, 0xfc4f158d, 0xcc9d2de8, 0x5ea384d3, 0x6e71bcb6,
+    0x3f07f419, 0x0fd5cc7c, 0x1d81cddd, 0x2d53f5b8, 0x7c25bd17, 0x4cf78572,
+    0xdec92c49, 0xee1b142c, 0xbf6d5c83, 0x8fbf64e6, 0x9efc7804, 0xae2e4061,
+    0xff5808ce, 0xcf8a30ab, 0x5db49990, 0x6d66a1f5, 0x3c10e95a, 0x0cc2d13f,
+    0x14e49f14, 0x2436a771, 0x7540efde, 0x4592d7bb, 0xd7ac7e80, 0xe77e46e5,
+    0xb6080e4a, 0x86da362f, 0x97992acd, 0xa74b12a8, 0xf63d5a07, 0xc6ef6262,
+    0x54d1cb59, 0x6403f33c, 0x3575bb93, 0x05a783f6, 0x17f38257, 0x2721ba32,
+    0x7657f29d, 0x4685caf8, 0xd4bb63c3, 0xe4695ba6, 0xb51f1309, 0x85cd2b6c,
+    0x948e378e, 0xa45c0feb, 0xf52a4744, 0xc5f87f21, 0x57c6d61a, 0x6714ee7f,
+    0x3662a6d0, 0x06b09eb5, 0x12caa592, 0x22189df7, 0x736ed558, 0x43bced3d,
+    0xd1824406, 0xe1507c63, 0xb02634cc, 0x80f40ca9, 0x91b7104b, 0xa165282e,
+    0xf0136081, 0xc0c158e4, 0x52fff1df, 0x622dc9ba, 0x335b8115, 0x0389b970,
+    0x11ddb8d1, 0x210f80b4, 0x7079c81b, 0x40abf07e, 0xd2955945, 0xe2476120,
+    0xb331298f, 0x83e311ea, 0x92a00d08, 0xa272356d, 0xf3047dc2, 0xc3d645a7,
+    0x51e8ec9c, 0x613ad4f9, 0x304c9c56, 0x009ea433};
+
+const uint32_t kStrideExtensionTable1[256] = {
+    0x00000000, 0x54075546, 0xa80eaa8c, 0xfc09ffca, 0x55f123e9, 0x01f676af,
+    0xfdff8965, 0xa9f8dc23, 0xabe247d2, 0xffe51294, 0x03eced5e, 0x57ebb818,
+    0xfe13643b, 0xaa14317d, 0x561dceb7, 0x021a9bf1, 0x5228f955, 0x062fac13,
+    0xfa2653d9, 0xae21069f, 0x07d9dabc, 0x53de8ffa, 0xafd77030, 0xfbd02576,
+    0xf9cabe87, 0xadcdebc1, 0x51c4140b, 0x05c3414d, 0xac3b9d6e, 0xf83cc828,
+    0x043537e2, 0x503262a4, 0xa451f2aa, 0xf056a7ec, 0x0c5f5826, 0x58580d60,
+    0xf1a0d143, 0xa5a78405, 0x59ae7bcf, 0x0da92e89, 0x0fb3b578, 0x5bb4e03e,
+    0xa7bd1ff4, 0xf3ba4ab2, 0x5a429691, 0x0e45c3d7, 0xf24c3c1d, 0xa64b695b,
+    0xf6790bff, 0xa27e5eb9, 0x5e77a173, 0x0a70f435, 0xa3882816, 0xf78f7d50,
+    0x0b86829a, 0x5f81d7dc, 0x5d9b4c2d, 0x099c196b, 0xf595e6a1, 0xa192b3e7,
+    0x086a6fc4, 0x5c6d3a82, 0xa064c548, 0xf463900e, 0x4d4f93a5, 0x1948c6e3,
+    0xe5413929, 0xb1466c6f, 0x18beb04c, 0x4cb9e50a, 0xb0b01ac0, 0xe4b74f86,
+    0xe6add477, 0xb2aa8131, 0x4ea37efb, 0x1aa42bbd, 0xb35cf79e, 0xe75ba2d8,
+    0x1b525d12, 0x4f550854, 0x1f676af0, 0x4b603fb6, 0xb769c07c, 0xe36e953a,
+    0x4a964919, 0x1e911c5f, 0xe298e395, 0xb69fb6d3, 0xb4852d22, 0xe0827864,
+    0x1c8b87ae, 0x488cd2e8, 0xe1740ecb, 0xb5735b8d, 0x497aa447, 0x1d7df101,
+    0xe91e610f, 0xbd193449, 0x4110cb83, 0x15179ec5, 0xbcef42e6, 0xe8e817a0,
+    0x14e1e86a, 0x40e6bd2c, 0x42fc26dd, 0x16fb739b, 0xeaf28c51, 0xbef5d917,
+    0x170d0534, 0x430a5072, 0xbf03afb8, 0xeb04fafe, 0xbb36985a, 0xef31cd1c,
+    0x133832d6, 0x473f6790, 0xeec7bbb3, 0xbac0eef5, 0x46c9113f, 0x12ce4479,
+    0x10d4df88, 0x44d38ace, 0xb8da7504, 0xecdd2042, 0x4525fc61, 0x1122a927,
+    0xed2b56ed, 0xb92c03ab, 0x9a9f274a, 0xce98720c, 0x32918dc6, 0x6696d880,
+    0xcf6e04a3, 0x9b6951e5, 0x6760ae2f, 0x3367fb69, 0x317d6098, 0x657a35de,
+    0x9973ca14, 0xcd749f52, 0x648c4371, 0x308b1637, 0xcc82e9fd, 0x9885bcbb,
+    0xc8b7de1f, 0x9cb08b59, 0x60b97493, 0x34be21d5, 0x9d46fdf6, 0xc941a8b0,
+    0x3548577a, 0x614f023c, 0x635599cd, 0x3752cc8b, 0xcb5b3341, 0x9f5c6607,
+    0x36a4ba24, 0x62a3ef62, 0x9eaa10a8, 0xcaad45ee, 0x3eced5e0, 0x6ac980a6,
+    0x96c07f6c, 0xc2c72a2a, 0x6b3ff609, 0x3f38a34f, 0xc3315c85, 0x973609c3,
+    0x952c9232, 0xc12bc774, 0x3d2238be, 0x69256df8, 0xc0ddb1db, 0x94dae49d,
+    0x68d31b57, 0x3cd44e11, 0x6ce62cb5, 0x38e179f3, 0xc4e88639, 0x90efd37f,
+    0x39170f5c, 0x6d105a1a, 0x9119a5d0, 0xc51ef096, 0xc7046b67, 0x93033e21,
+    0x6f0ac1eb, 0x3b0d94ad, 0x92f5488e, 0xc6f21dc8, 0x3afbe202, 0x6efcb744,
+    0xd7d0b4ef, 0x83d7e1a9, 0x7fde1e63, 0x2bd94b25, 0x82219706, 0xd626c240,
+    0x2a2f3d8a, 0x7e2868cc, 0x7c32f33d, 0x2835a67b, 0xd43c59b1, 0x803b0cf7,
+    0x29c3d0d4, 0x7dc48592, 0x81cd7a58, 0xd5ca2f1e, 0x85f84dba, 0xd1ff18fc,
+    0x2df6e736, 0x79f1b270, 0xd0096e53, 0x840e3b15, 0x7807c4df, 0x2c009199,
+    0x2e1a0a68, 0x7a1d5f2e, 0x8614a0e4, 0xd213f5a2, 0x7beb2981, 0x2fec7cc7,
+    0xd3e5830d, 0x87e2d64b, 0x73814645, 0x27861303, 0xdb8fecc9, 0x8f88b98f,
+    0x267065ac, 0x727730ea, 0x8e7ecf20, 0xda799a66, 0xd8630197, 0x8c6454d1,
+    0x706dab1b, 0x246afe5d, 0x8d92227e, 0xd9957738, 0x259c88f2, 0x719bddb4,
+    0x21a9bf10, 0x75aeea56, 0x89a7159c, 0xdda040da, 0x74589cf9, 0x205fc9bf,
+    0xdc563675, 0x88516333, 0x8a4bf8c2, 0xde4cad84, 0x2245524e, 0x76420708,
+    0xdfbadb2b, 0x8bbd8e6d, 0x77b471a7, 0x23b324e1};
+
+const uint32_t kStrideExtensionTable2[256] = {
+    0x00000000, 0x678efd01, 0xcf1dfa02, 0xa8930703, 0x9bd782f5, 0xfc597ff4,
+    0x54ca78f7, 0x334485f6, 0x3243731b, 0x55cd8e1a, 0xfd5e8919, 0x9ad07418,
+    0xa994f1ee, 0xce1a0cef, 0x66890bec, 0x0107f6ed, 0x6486e636, 0x03081b37,
+    0xab9b1c34, 0xcc15e135, 0xff5164c3, 0x98df99c2, 0x304c9ec1, 0x57c263c0,
+    0x56c5952d, 0x314b682c, 0x99d86f2f, 0xfe56922e, 0xcd1217d8, 0xaa9cead9,
+    0x020fedda, 0x658110db, 0xc90dcc6c, 0xae83316d, 0x0610366e, 0x619ecb6f,
+    0x52da4e99, 0x3554b398, 0x9dc7b49b, 0xfa49499a, 0xfb4ebf77, 0x9cc04276,
+    0x34534575, 0x53ddb874, 0x60993d82, 0x0717c083, 0xaf84c780, 0xc80a3a81,
+    0xad8b2a5a, 0xca05d75b, 0x6296d058, 0x05182d59, 0x365ca8af, 0x51d255ae,
+    0xf94152ad, 0x9ecfafac, 0x9fc85941, 0xf846a440, 0x50d5a343, 0x375b5e42,
+    0x041fdbb4, 0x639126b5, 0xcb0221b6, 0xac8cdcb7, 0x97f7ee29, 0xf0791328,
+    0x58ea142b, 0x3f64e92a, 0x0c206cdc, 0x6bae91dd, 0xc33d96de, 0xa4b36bdf,
+    0xa5b49d32, 0xc23a6033, 0x6aa96730, 0x0d279a31, 0x3e631fc7, 0x59ede2c6,
+    0xf17ee5c5, 0x96f018c4, 0xf371081f, 0x94fff51e, 0x3c6cf21d, 0x5be20f1c,
+    0x68a68aea, 0x0f2877eb, 0xa7bb70e8, 0xc0358de9, 0xc1327b04, 0xa6bc8605,
+    0x0e2f8106, 0x69a17c07, 0x5ae5f9f1, 0x3d6b04f0, 0x95f803f3, 0xf276fef2,
+    0x5efa2245, 0x3974df44, 0x91e7d847, 0xf6692546, 0xc52da0b0, 0xa2a35db1,
+    0x0a305ab2, 0x6dbea7b3, 0x6cb9515e, 0x0b37ac5f, 0xa3a4ab5c, 0xc42a565d,
+    0xf76ed3ab, 0x90e02eaa, 0x387329a9, 0x5ffdd4a8, 0x3a7cc473, 0x5df23972,
+    0xf5613e71, 0x92efc370, 0xa1ab4686, 0xc625bb87, 0x6eb6bc84, 0x09384185,
+    0x083fb768, 0x6fb14a69, 0xc7224d6a, 0xa0acb06b, 0x93e8359d, 0xf466c89c,
+    0x5cf5cf9f, 0x3b7b329e, 0x2a03aaa3, 0x4d8d57a2, 0xe51e50a1, 0x8290ada0,
+    0xb1d42856, 0xd65ad557, 0x7ec9d254, 0x19472f55, 0x1840d9b8, 0x7fce24b9,
+    0xd75d23ba, 0xb0d3debb, 0x83975b4d, 0xe419a64c, 0x4c8aa14f, 0x2b045c4e,
+    0x4e854c95, 0x290bb194, 0x8198b697, 0xe6164b96, 0xd552ce60, 0xb2dc3361,
+    0x1a4f3462, 0x7dc1c963, 0x7cc63f8e, 0x1b48c28f, 0xb3dbc58c, 0xd455388d,
+    0xe711bd7b, 0x809f407a, 0x280c4779, 0x4f82ba78, 0xe30e66cf, 0x84809bce,
+    0x2c139ccd, 0x4b9d61cc, 0x78d9e43a, 0x1f57193b, 0xb7c41e38, 0xd04ae339,
+    0xd14d15d4, 0xb6c3e8d5, 0x1e50efd6, 0x79de12d7, 0x4a9a9721, 0x2d146a20,
+    0x85876d23, 0xe2099022, 0x878880f9, 0xe0067df8, 0x48957afb, 0x2f1b87fa,
+    0x1c5f020c, 0x7bd1ff0d, 0xd342f80e, 0xb4cc050f, 0xb5cbf3e2, 0xd2450ee3,
+    0x7ad609e0, 0x1d58f4e1, 0x2e1c7117, 0x49928c16, 0xe1018b15, 0x868f7614,
+    0xbdf4448a, 0xda7ab98b, 0x72e9be88, 0x15674389, 0x2623c67f, 0x41ad3b7e,
+    0xe93e3c7d, 0x8eb0c17c, 0x8fb73791, 0xe839ca90, 0x40aacd93, 0x27243092,
+    0x1460b564, 0x73ee4865, 0xdb7d4f66, 0xbcf3b267, 0xd972a2bc, 0xbefc5fbd,
+    0x166f58be, 0x71e1a5bf, 0x42a52049, 0x252bdd48, 0x8db8da4b, 0xea36274a,
+    0xeb31d1a7, 0x8cbf2ca6, 0x242c2ba5, 0x43a2d6a4, 0x70e65352, 0x1768ae53,
+    0xbffba950, 0xd8755451, 0x74f988e6, 0x137775e7, 0xbbe472e4, 0xdc6a8fe5,
+    0xef2e0a13, 0x88a0f712, 0x2033f011, 0x47bd0d10, 0x46bafbfd, 0x213406fc,
+    0x89a701ff, 0xee29fcfe, 0xdd6d7908, 0xbae38409, 0x1270830a, 0x75fe7e0b,
+    0x107f6ed0, 0x77f193d1, 0xdf6294d2, 0xb8ec69d3, 0x8ba8ec25, 0xec261124,
+    0x44b51627, 0x233beb26, 0x223c1dcb, 0x45b2e0ca, 0xed21e7c9, 0x8aaf1ac8,
+    0xb9eb9f3e, 0xde65623f, 0x76f6653c, 0x1178983d};
+
+const uint32_t kStrideExtensionTable3[256] = {
+    0x00000000, 0xf20c0dfe, 0xe1f46d0d, 0x13f860f3, 0xc604aceb, 0x3408a115,
+    0x27f0c1e6, 0xd5fccc18, 0x89e52f27, 0x7be922d9, 0x6811422a, 0x9a1d4fd4,
+    0x4fe183cc, 0xbded8e32, 0xae15eec1, 0x5c19e33f, 0x162628bf, 0xe42a2541,
+    0xf7d245b2, 0x05de484c, 0xd0228454, 0x222e89aa, 0x31d6e959, 0xc3dae4a7,
+    0x9fc30798, 0x6dcf0a66, 0x7e376a95, 0x8c3b676b, 0x59c7ab73, 0xabcba68d,
+    0xb833c67e, 0x4a3fcb80, 0x2c4c517e, 0xde405c80, 0xcdb83c73, 0x3fb4318d,
+    0xea48fd95, 0x1844f06b, 0x0bbc9098, 0xf9b09d66, 0xa5a97e59, 0x57a573a7,
+    0x445d1354, 0xb6511eaa, 0x63add2b2, 0x91a1df4c, 0x8259bfbf, 0x7055b241,
+    0x3a6a79c1, 0xc866743f, 0xdb9e14cc, 0x29921932, 0xfc6ed52a, 0x0e62d8d4,
+    0x1d9ab827, 0xef96b5d9, 0xb38f56e6, 0x41835b18, 0x527b3beb, 0xa0773615,
+    0x758bfa0d, 0x8787f7f3, 0x947f9700, 0x66739afe, 0x5898a2fc, 0xaa94af02,
+    0xb96ccff1, 0x4b60c20f, 0x9e9c0e17, 0x6c9003e9, 0x7f68631a, 0x8d646ee4,
+    0xd17d8ddb, 0x23718025, 0x3089e0d6, 0xc285ed28, 0x17792130, 0xe5752cce,
+    0xf68d4c3d, 0x048141c3, 0x4ebe8a43, 0xbcb287bd, 0xaf4ae74e, 0x5d46eab0,
+    0x88ba26a8, 0x7ab62b56, 0x694e4ba5, 0x9b42465b, 0xc75ba564, 0x3557a89a,
+    0x26afc869, 0xd4a3c597, 0x015f098f, 0xf3530471, 0xe0ab6482, 0x12a7697c,
+    0x74d4f382, 0x86d8fe7c, 0x95209e8f, 0x672c9371, 0xb2d05f69, 0x40dc5297,
+    0x53243264, 0xa1283f9a, 0xfd31dca5, 0x0f3dd15b, 0x1cc5b1a8, 0xeec9bc56,
+    0x3b35704e, 0xc9397db0, 0xdac11d43, 0x28cd10bd, 0x62f2db3d, 0x90fed6c3,
+    0x8306b630, 0x710abbce, 0xa4f677d6, 0x56fa7a28, 0x45021adb, 0xb70e1725,
+    0xeb17f41a, 0x191bf9e4, 0x0ae39917, 0xf8ef94e9, 0x2d1358f1, 0xdf1f550f,
+    0xcce735fc, 0x3eeb3802, 0xb13145f8, 0x433d4806, 0x50c528f5, 0xa2c9250b,
+    0x7735e913, 0x8539e4ed, 0x96c1841e, 0x64cd89e0, 0x38d46adf, 0xcad86721,
+    0xd92007d2, 0x2b2c0a2c, 0xfed0c634, 0x0cdccbca, 0x1f24ab39, 0xed28a6c7,
+    0xa7176d47, 0x551b60b9, 0x46e3004a, 0xb4ef0db4, 0x6113c1ac, 0x931fcc52,
+    0x80e7aca1, 0x72eba15f, 0x2ef24260, 0xdcfe4f9e, 0xcf062f6d, 0x3d0a2293,
+    0xe8f6ee8b, 0x1afae375, 0x09028386, 0xfb0e8e78, 0x9d7d1486, 0x6f711978,
+    0x7c89798b, 0x8e857475, 0x5b79b86d, 0xa975b593, 0xba8dd560, 0x4881d89e,
+    0x14983ba1, 0xe694365f, 0xf56c56ac, 0x07605b52, 0xd29c974a, 0x20909ab4,
+    0x3368fa47, 0xc164f7b9, 0x8b5b3c39, 0x795731c7, 0x6aaf5134, 0x98a35cca,
+    0x4d5f90d2, 0xbf539d2c, 0xacabfddf, 0x5ea7f021, 0x02be131e, 0xf0b21ee0,
+    0xe34a7e13, 0x114673ed, 0xc4babff5, 0x36b6b20b, 0x254ed2f8, 0xd742df06,
+    0xe9a9e704, 0x1ba5eafa, 0x085d8a09, 0xfa5187f7, 0x2fad4bef, 0xdda14611,
+    0xce5926e2, 0x3c552b1c, 0x604cc823, 0x9240c5dd, 0x81b8a52e, 0x73b4a8d0,
+    0xa64864c8, 0x54446936, 0x47bc09c5, 0xb5b0043b, 0xff8fcfbb, 0x0d83c245,
+    0x1e7ba2b6, 0xec77af48, 0x398b6350, 0xcb876eae, 0xd87f0e5d, 0x2a7303a3,
+    0x766ae09c, 0x8466ed62, 0x979e8d91, 0x6592806f, 0xb06e4c77, 0x42624189,
+    0x519a217a, 0xa3962c84, 0xc5e5b67a, 0x37e9bb84, 0x2411db77, 0xd61dd689,
+    0x03e11a91, 0xf1ed176f, 0xe215779c, 0x10197a62, 0x4c00995d, 0xbe0c94a3,
+    0xadf4f450, 0x5ff8f9ae, 0x8a0435b6, 0x78083848, 0x6bf058bb, 0x99fc5545,
+    0xd3c39ec5, 0x21cf933b, 0x3237f3c8, 0xc03bfe36, 0x15c7322e, 0xe7cb3fd0,
+    0xf4335f23, 0x063f52dd, 0x5a26b1e2, 0xa82abc1c, 0xbbd2dcef, 0x49ded111,
+    0x9c221d09, 0x6e2e10f7, 0x7dd67004, 0x8fda7dfa};
+
+// CRCs are pre- and post- conditioned by xoring with all ones.
+static constexpr const uint32_t kCRC32Xor = static_cast<uint32_t>(0xffffffffU);
+
+// Reads a little-endian 32-bit integer from a 32-bit-aligned buffer.
+inline uint32_t ReadUint32LE(const uint8_t* buffer) {
+  return DecodeFixed32(reinterpret_cast<const char*>(buffer));
+}
+
+// Returns the smallest address >= the given address that is aligned to N bytes.
+//
+// N must be a power of two.
+template <int N>
+constexpr inline const uint8_t* RoundUp(const uint8_t* pointer) {
+  return reinterpret_cast<uint8_t*>(
+      (reinterpret_cast<uintptr_t>(pointer) + (N - 1)) &
+      ~static_cast<uintptr_t>(N - 1));
+}
+
+}  // namespace
+
+// Determine if the CPU running this program can accelerate the CRC32C
+// calculation.
+static bool CanAccelerateCRC32C() {
+  // port::AcceleretedCRC32C returns zero when unable to accelerate.
+  static const char kTestCRCBuffer[] = "TestCRCBuffer";
+  static const char kBufSize = sizeof(kTestCRCBuffer) - 1;
+  static const uint32_t kTestCRCValue = 0xdcbc59fa;
+
+  return port::AcceleratedCRC32C(0, kTestCRCBuffer, kBufSize) == kTestCRCValue;
+}
+
+uint32_t Extend(uint32_t crc, const char* data, size_t n) {
+  static bool accelerate = CanAccelerateCRC32C();
+  if (accelerate) {
+    return port::AcceleratedCRC32C(crc, data, n);
+  }
+
+  const uint8_t* p = reinterpret_cast<const uint8_t*>(data);
+  const uint8_t* e = p + n;
+  uint32_t l = crc ^ kCRC32Xor;
+
+// Process one byte at a time.
+#define STEP1                              \
+  do {                                     \
+    int c = (l & 0xff) ^ *p++;             \
+    l = kByteExtensionTable[c] ^ (l >> 8); \
+  } while (0)
+
+// Process one of the 4 strides of 4-byte data.
+#define STEP4(s)                                                               \
+  do {                                                                         \
+    crc##s = ReadUint32LE(p + s * 4) ^ kStrideExtensionTable3[crc##s & 0xff] ^ \
+             kStrideExtensionTable2[(crc##s >> 8) & 0xff] ^                    \
+             kStrideExtensionTable1[(crc##s >> 16) & 0xff] ^                   \
+             kStrideExtensionTable0[crc##s >> 24];                             \
+  } while (0)
+
+// Process a 16-byte swath of 4 strides, each of which has 4 bytes of data.
+#define STEP16 \
+  do {         \
+    STEP4(0);  \
+    STEP4(1);  \
+    STEP4(2);  \
+    STEP4(3);  \
+    p += 16;   \
+  } while (0)
+
+// Process 4 bytes that were already loaded into a word.
+#define STEP4W(w)                                   \
+  do {                                              \
+    w ^= l;                                         \
+    for (size_t i = 0; i < 4; ++i) {                \
+      w = (w >> 8) ^ kByteExtensionTable[w & 0xff]; \
+    }                                               \
+    l = w;                                          \
+  } while (0)
+
+  // Point x at first 4-byte aligned byte in the buffer. This might be past the
+  // end of the buffer.
+  const uint8_t* x = RoundUp<4>(p);
+  if (x <= e) {
+    // Process bytes p is 4-byte aligned.
+    while (p != x) {
+      STEP1;
+    }
+  }
+
+  if ((e - p) >= 16) {
+    // Load a 16-byte swath into the stride partial results.
+    uint32_t crc0 = ReadUint32LE(p + 0 * 4) ^ l;
+    uint32_t crc1 = ReadUint32LE(p + 1 * 4);
+    uint32_t crc2 = ReadUint32LE(p + 2 * 4);
+    uint32_t crc3 = ReadUint32LE(p + 3 * 4);
+    p += 16;
+
+    // It is possible to get better speeds (at least on x86) by interleaving
+    // prefetching 256 bytes ahead with processing 64 bytes at a time. See the
+    // portable implementation in https://github.com/google/crc32c/.
+
+    // Process one 16-byte swath at a time.
+    while ((e - p) >= 16) {
+      STEP16;
+    }
+
+    // Advance one word at a time as far as possible.
+    while ((e - p) >= 4) {
+      STEP4(0);
+      uint32_t tmp = crc0;
+      crc0 = crc1;
+      crc1 = crc2;
+      crc2 = crc3;
+      crc3 = tmp;
+      p += 4;
+    }
+
+    // Combine the 4 partial stride results.
+    l = 0;
+    STEP4W(crc0);
+    STEP4W(crc1);
+    STEP4W(crc2);
+    STEP4W(crc3);
+  }
+
+  // Process the last few bytes.
+  while (p != e) {
+    STEP1;
+  }
+#undef STEP4W
+#undef STEP16
+#undef STEP4
+#undef STEP1
+  return l ^ kCRC32Xor;
+}
+
+}  // namespace crc32c
+}  // namespace leveldb

lib/leveldb/util/crc32c.h

@@ -0,0 +1,43 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#ifndef STORAGE_LEVELDB_UTIL_CRC32C_H_
+#define STORAGE_LEVELDB_UTIL_CRC32C_H_
+
+#include <cstddef>
+#include <cstdint>
+
+namespace leveldb {
+namespace crc32c {
+
+// Return the crc32c of concat(A, data[0,n-1]) where init_crc is the
+// crc32c of some string A.  Extend() is often used to maintain the
+// crc32c of a stream of data.
+uint32_t Extend(uint32_t init_crc, const char* data, size_t n);
+
+// Return the crc32c of data[0,n-1]
+inline uint32_t Value(const char* data, size_t n) { return Extend(0, data, n); }
+
+static const uint32_t kMaskDelta = 0xa282ead8ul;
+
+// Return a masked representation of crc.
+//
+// Motivation: it is problematic to compute the CRC of a string that
+// contains embedded CRCs.  Therefore we recommend that CRCs stored
+// somewhere (e.g., in files) should be masked before being stored.
+inline uint32_t Mask(uint32_t crc) {
+  // Rotate right by 15 bits and add a constant.
+  return ((crc >> 15) | (crc << 17)) + kMaskDelta;
+}
+
+// Return the crc whose masked representation is masked_crc.
+inline uint32_t Unmask(uint32_t masked_crc) {
+  uint32_t rot = masked_crc - kMaskDelta;
+  return ((rot >> 17) | (rot << 15));
+}
+
+}  // namespace crc32c
+}  // namespace leveldb
+
+#endif  // STORAGE_LEVELDB_UTIL_CRC32C_H_

lib/leveldb/util/crc32c_test.cc

@@ -0,0 +1,61 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "util/crc32c.h"
+
+#include "gtest/gtest.h"
+
+namespace leveldb {
+namespace crc32c {
+
+TEST(CRC, StandardResults) {
+  // From rfc3720 section B.4.
+  char buf[32];
+
+  memset(buf, 0, sizeof(buf));
+  ASSERT_EQ(0x8a9136aa, Value(buf, sizeof(buf)));
+
+  memset(buf, 0xff, sizeof(buf));
+  ASSERT_EQ(0x62a8ab43, Value(buf, sizeof(buf)));
+
+  for (int i = 0; i < 32; i++) {
+    buf[i] = i;
+  }
+  ASSERT_EQ(0x46dd794e, Value(buf, sizeof(buf)));
+
+  for (int i = 0; i < 32; i++) {
+    buf[i] = 31 - i;
+  }
+  ASSERT_EQ(0x113fdb5c, Value(buf, sizeof(buf)));
+
+  uint8_t data[48] = {
+      0x01, 0xc0, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+      0x00, 0x00, 0x00, 0x00, 0x14, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00,
+      0x00, 0x00, 0x00, 0x14, 0x00, 0x00, 0x00, 0x18, 0x28, 0x00, 0x00, 0x00,
+      0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+  };
+  ASSERT_EQ(0xd9963a56, Value(reinterpret_cast<char*>(data), sizeof(data)));
+}
+
+TEST(CRC, Values) { ASSERT_NE(Value("a", 1), Value("foo", 3)); }
+
+TEST(CRC, Extend) {
+  ASSERT_EQ(Value("hello world", 11), Extend(Value("hello ", 6), "world", 5));
+}
+
+TEST(CRC, Mask) {
+  uint32_t crc = Value("foo", 3);
+  ASSERT_NE(crc, Mask(crc));
+  ASSERT_NE(crc, Mask(Mask(crc)));
+  ASSERT_EQ(crc, Unmask(Mask(crc)));
+  ASSERT_EQ(crc, Unmask(Unmask(Mask(Mask(crc)))));
+}
+
+}  // namespace crc32c
+}  // namespace leveldb
+
+int main(int argc, char** argv) {
+  testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}

lib/leveldb/util/env.cc

@@ -0,0 +1,108 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "leveldb/env.h"
+
+#include <cstdarg>
+
+// This workaround can be removed when leveldb::Env::DeleteFile is removed.
+// See env.h for justification.
+#if defined(_WIN32) && defined(LEVELDB_DELETEFILE_UNDEFINED)
+#undef DeleteFile
+#endif
+
+namespace leveldb {
+
+Env::Env() = default;
+
+Env::~Env() = default;
+
+Status Env::NewAppendableFile(const std::string& fname, WritableFile** result) {
+  return Status::NotSupported("NewAppendableFile", fname);
+}
+
+Status Env::RemoveDir(const std::string& dirname) { return DeleteDir(dirname); }
+Status Env::DeleteDir(const std::string& dirname) { return RemoveDir(dirname); }
+
+Status Env::RemoveFile(const std::string& fname) { return DeleteFile(fname); }
+Status Env::DeleteFile(const std::string& fname) { return RemoveFile(fname); }
+
+SequentialFile::~SequentialFile() = default;
+
+RandomAccessFile::~RandomAccessFile() = default;
+
+WritableFile::~WritableFile() = default;
+
+Logger::~Logger() = default;
+
+FileLock::~FileLock() = default;
+
+void Log(Logger* info_log, const char* format, ...) {
+  if (info_log != nullptr) {
+    std::va_list ap;
+    va_start(ap, format);
+    info_log->Logv(format, ap);
+    va_end(ap);
+  }
+}
+
+static Status DoWriteStringToFile(Env* env, const Slice& data,
+                                  const std::string& fname, bool should_sync) {
+  WritableFile* file;
+  Status s = env->NewWritableFile(fname, &file);
+  if (!s.ok()) {
+    return s;
+  }
+  s = file->Append(data);
+  if (s.ok() && should_sync) {
+    s = file->Sync();
+  }
+  if (s.ok()) {
+    s = file->Close();
+  }
+  delete file;  // Will auto-close if we did not close above
+  if (!s.ok()) {
+    env->RemoveFile(fname);
+  }
+  return s;
+}
+
+Status WriteStringToFile(Env* env, const Slice& data,
+                         const std::string& fname) {
+  return DoWriteStringToFile(env, data, fname, false);
+}
+
+Status WriteStringToFileSync(Env* env, const Slice& data,
+                             const std::string& fname) {
+  return DoWriteStringToFile(env, data, fname, true);
+}
+
+Status ReadFileToString(Env* env, const std::string& fname, std::string* data) {
+  data->clear();
+  SequentialFile* file;
+  Status s = env->NewSequentialFile(fname, &file);
+  if (!s.ok()) {
+    return s;
+  }
+  static const int kBufferSize = 8192;
+  char* space = new char[kBufferSize];
+  while (true) {
+    Slice fragment;
+    s = file->Read(kBufferSize, &fragment, space);
+    if (!s.ok()) {
+      break;
+    }
+    data->append(fragment.data(), fragment.size());
+    if (fragment.empty()) {
+      break;
+    }
+  }
+  delete[] space;
+  delete file;
+  return s;
+}
+
+EnvWrapper::~EnvWrapper() {}
+
+}  // namespace leveldb

lib/leveldb/util/env_posix.cc

@@ -0,0 +1,893 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include <dirent.h>
+#include <fcntl.h>
+#include <pthread.h>
+#include <sys/mman.h>
+#include <sys/resource.h>
+#include <sys/stat.h>
+#include <sys/time.h>
+#include <sys/types.h>
+#include <unistd.h>
+
+#include <atomic>
+#include <cerrno>
+#include <cstddef>
+#include <cstdint>
+#include <cstdio>
+#include <cstdlib>
+#include <cstring>
+#include <limits>
+#include <queue>
+#include <set>
+#include <string>
+#include <thread>
+#include <type_traits>
+#include <utility>
+
+#include "leveldb/env.h"
+#include "leveldb/slice.h"
+#include "leveldb/status.h"
+#include "port/port.h"
+#include "port/thread_annotations.h"
+#include "util/env_posix_test_helper.h"
+#include "util/posix_logger.h"
+
+namespace leveldb {
+
+namespace {
+
+// Set by EnvPosixTestHelper::SetReadOnlyMMapLimit() and MaxOpenFiles().
+int g_open_read_only_file_limit = -1;
+
+// Up to 1000 mmap regions for 64-bit binaries; none for 32-bit.
+constexpr const int kDefaultMmapLimit = (sizeof(void*) >= 8) ? 1000 : 0;
+
+// Can be set using EnvPosixTestHelper::SetReadOnlyMMapLimit().
+int g_mmap_limit = kDefaultMmapLimit;
+
+// Common flags defined for all posix open operations
+#if defined(HAVE_O_CLOEXEC)
+constexpr const int kOpenBaseFlags = O_CLOEXEC;
+#else
+constexpr const int kOpenBaseFlags = 0;
+#endif  // defined(HAVE_O_CLOEXEC)
+
+constexpr const size_t kWritableFileBufferSize = 65536;
+
+Status PosixError(const std::string& context, int error_number) {
+  if (error_number == ENOENT) {
+    return Status::NotFound(context, std::strerror(error_number));
+  } else {
+    return Status::IOError(context, std::strerror(error_number));
+  }
+}
+
+// Helper class to limit resource usage to avoid exhaustion.
+// Currently used to limit read-only file descriptors and mmap file usage
+// so that we do not run out of file descriptors or virtual memory, or run into
+// kernel performance problems for very large databases.
+class Limiter {
+ public:
+  // Limit maximum number of resources to |max_acquires|.
+  Limiter(int max_acquires) : acquires_allowed_(max_acquires) {}
+
+  Limiter(const Limiter&) = delete;
+  Limiter operator=(const Limiter&) = delete;
+
+  // If another resource is available, acquire it and return true.
+  // Else return false.
+  bool Acquire() {
+    int old_acquires_allowed =
+        acquires_allowed_.fetch_sub(1, std::memory_order_relaxed);
+
+    if (old_acquires_allowed > 0) return true;
+
+    acquires_allowed_.fetch_add(1, std::memory_order_relaxed);
+    return false;
+  }
+
+  // Release a resource acquired by a previous call to Acquire() that returned
+  // true.
+  void Release() { acquires_allowed_.fetch_add(1, std::memory_order_relaxed); }
+
+ private:
+  // The number of available resources.
+  //
+  // This is a counter and is not tied to the invariants of any other class, so
+  // it can be operated on safely using std::memory_order_relaxed.
+  std::atomic<int> acquires_allowed_;
+};
+
+// Implements sequential read access in a file using read().
+//
+// Instances of this class are thread-friendly but not thread-safe, as required
+// by the SequentialFile API.
+class PosixSequentialFile final : public SequentialFile {
+ public:
+  PosixSequentialFile(std::string filename, int fd)
+      : fd_(fd), filename_(filename) {}
+  ~PosixSequentialFile() override { close(fd_); }
+
+  Status Read(size_t n, Slice* result, char* scratch) override {
+    Status status;
+    while (true) {
+      ::ssize_t read_size = ::read(fd_, scratch, n);
+      if (read_size < 0) {  // Read error.
+        if (errno == EINTR) {
+          continue;  // Retry
+        }
+        status = PosixError(filename_, errno);
+        break;
+      }
+      *result = Slice(scratch, read_size);
+      break;
+    }
+    return status;
+  }
+
+  Status Skip(uint64_t n) override {
+    if (::lseek(fd_, n, SEEK_CUR) == static_cast<off_t>(-1)) {
+      return PosixError(filename_, errno);
+    }
+    return Status::OK();
+  }
+
+ private:
+  const int fd_;
+  const std::string filename_;
+};
+
+// Implements random read access in a file using pread().
+//
+// Instances of this class are thread-safe, as required by the RandomAccessFile
+// API. Instances are immutable and Read() only calls thread-safe library
+// functions.
+class PosixRandomAccessFile final : public RandomAccessFile {
+ public:
+  // The new instance takes ownership of |fd|. |fd_limiter| must outlive this
+  // instance, and will be used to determine if .
+  PosixRandomAccessFile(std::string filename, int fd, Limiter* fd_limiter)
+      : has_permanent_fd_(fd_limiter->Acquire()),
+        fd_(has_permanent_fd_ ? fd : -1),
+        fd_limiter_(fd_limiter),
+        filename_(std::move(filename)) {
+    if (!has_permanent_fd_) {
+      assert(fd_ == -1);
+      ::close(fd);  // The file will be opened on every read.
+    }
+  }
+
+  ~PosixRandomAccessFile() override {
+    if (has_permanent_fd_) {
+      assert(fd_ != -1);
+      ::close(fd_);
+      fd_limiter_->Release();
+    }
+  }
+
+  Status Read(uint64_t offset, size_t n, Slice* result,
+              char* scratch) const override {
+    int fd = fd_;
+    if (!has_permanent_fd_) {
+      fd = ::open(filename_.c_str(), O_RDONLY | kOpenBaseFlags);
+      if (fd < 0) {
+        return PosixError(filename_, errno);
+      }
+    }
+
+    assert(fd != -1);
+
+    Status status;
+    ssize_t read_size = ::pread(fd, scratch, n, static_cast<off_t>(offset));
+    *result = Slice(scratch, (read_size < 0) ? 0 : read_size);
+    if (read_size < 0) {
+      // An error: return a non-ok status.
+      status = PosixError(filename_, errno);
+    }
+    if (!has_permanent_fd_) {
+      // Close the temporary file descriptor opened earlier.
+      assert(fd != fd_);
+      ::close(fd);
+    }
+    return status;
+  }
+
+ private:
+  const bool has_permanent_fd_;  // If false, the file is opened on every read.
+  const int fd_;                 // -1 if has_permanent_fd_ is false.
+  Limiter* const fd_limiter_;
+  const std::string filename_;
+};
+
+// Implements random read access in a file using mmap().
+//
+// Instances of this class are thread-safe, as required by the RandomAccessFile
+// API. Instances are immutable and Read() only calls thread-safe library
+// functions.
+class PosixMmapReadableFile final : public RandomAccessFile {
+ public:
+  // mmap_base[0, length-1] points to the memory-mapped contents of the file. It
+  // must be the result of a successful call to mmap(). This instances takes
+  // over the ownership of the region.
+  //
+  // |mmap_limiter| must outlive this instance. The caller must have already
+  // aquired the right to use one mmap region, which will be released when this
+  // instance is destroyed.
+  PosixMmapReadableFile(std::string filename, char* mmap_base, size_t length,
+                        Limiter* mmap_limiter)
+      : mmap_base_(mmap_base),
+        length_(length),
+        mmap_limiter_(mmap_limiter),
+        filename_(std::move(filename)) {}
+
+  ~PosixMmapReadableFile() override {
+    ::munmap(static_cast<void*>(mmap_base_), length_);
+    mmap_limiter_->Release();
+  }
+
+  Status Read(uint64_t offset, size_t n, Slice* result,
+              char* scratch) const override {
+    if (offset + n > length_) {
+      *result = Slice();
+      return PosixError(filename_, EINVAL);
+    }
+
+    *result = Slice(mmap_base_ + offset, n);
+    return Status::OK();
+  }
+
+ private:
+  char* const mmap_base_;
+  const size_t length_;
+  Limiter* const mmap_limiter_;
+  const std::string filename_;
+};
+
+class PosixWritableFile final : public WritableFile {
+ public:
+  PosixWritableFile(std::string filename, int fd)
+      : pos_(0),
+        fd_(fd),
+        is_manifest_(IsManifest(filename)),
+        filename_(std::move(filename)),
+        dirname_(Dirname(filename_)) {}
+
+  ~PosixWritableFile() override {
+    if (fd_ >= 0) {
+      // Ignoring any potential errors
+      Close();
+    }
+  }
+
+  Status Append(const Slice& data) override {
+    size_t write_size = data.size();
+    const char* write_data = data.data();
+
+    // Fit as much as possible into buffer.
+    size_t copy_size = std::min(write_size, kWritableFileBufferSize - pos_);
+    std::memcpy(buf_ + pos_, write_data, copy_size);
+    write_data += copy_size;
+    write_size -= copy_size;
+    pos_ += copy_size;
+    if (write_size == 0) {
+      return Status::OK();
+    }
+
+    // Can't fit in buffer, so need to do at least one write.
+    Status status = FlushBuffer();
+    if (!status.ok()) {
+      return status;
+    }
+
+    // Small writes go to buffer, large writes are written directly.
+    if (write_size < kWritableFileBufferSize) {
+      std::memcpy(buf_, write_data, write_size);
+      pos_ = write_size;
+      return Status::OK();
+    }
+    return WriteUnbuffered(write_data, write_size);
+  }
+
+  Status Close() override {
+    Status status = FlushBuffer();
+    const int close_result = ::close(fd_);
+    if (close_result < 0 && status.ok()) {
+      status = PosixError(filename_, errno);
+    }
+    fd_ = -1;
+    return status;
+  }
+
+  Status Flush() override { return FlushBuffer(); }
+
+  Status Sync() override {
+    // Ensure new files referred to by the manifest are in the filesystem.
+    //
+    // This needs to happen before the manifest file is flushed to disk, to
+    // avoid crashing in a state where the manifest refers to files that are not
+    // yet on disk.
+    Status status = SyncDirIfManifest();
+    if (!status.ok()) {
+      return status;
+    }
+
+    status = FlushBuffer();
+    if (!status.ok()) {
+      return status;
+    }
+
+    return SyncFd(fd_, filename_);
+  }
+
+ private:
+  Status FlushBuffer() {
+    Status status = WriteUnbuffered(buf_, pos_);
+    pos_ = 0;
+    return status;
+  }
+
+  Status WriteUnbuffered(const char* data, size_t size) {
+    while (size > 0) {
+      ssize_t write_result = ::write(fd_, data, size);
+      if (write_result < 0) {
+        if (errno == EINTR) {
+          continue;  // Retry
+        }
+        return PosixError(filename_, errno);
+      }
+      data += write_result;
+      size -= write_result;
+    }
+    return Status::OK();
+  }
+
+  Status SyncDirIfManifest() {
+    Status status;
+    if (!is_manifest_) {
+      return status;
+    }
+
+    int fd = ::open(dirname_.c_str(), O_RDONLY | kOpenBaseFlags);
+    if (fd < 0) {
+      status = PosixError(dirname_, errno);
+    } else {
+      status = SyncFd(fd, dirname_);
+      ::close(fd);
+    }
+    return status;
+  }
+
+  // Ensures that all the caches associated with the given file descriptor's
+  // data are flushed all the way to durable media, and can withstand power
+  // failures.
+  //
+  // The path argument is only used to populate the description string in the
+  // returned Status if an error occurs.
+  static Status SyncFd(int fd, const std::string& fd_path) {
+#if HAVE_FULLFSYNC
+    // On macOS and iOS, fsync() doesn't guarantee durability past power
+    // failures. fcntl(F_FULLFSYNC) is required for that purpose. Some
+    // filesystems don't support fcntl(F_FULLFSYNC), and require a fallback to
+    // fsync().
+    if (::fcntl(fd, F_FULLFSYNC) == 0) {
+      return Status::OK();
+    }
+#endif  // HAVE_FULLFSYNC
+
+#if HAVE_FDATASYNC
+    bool sync_success = ::fdatasync(fd) == 0;
+#else
+    bool sync_success = ::fsync(fd) == 0;
+#endif  // HAVE_FDATASYNC
+
+    if (sync_success) {
+      return Status::OK();
+    }
+    return PosixError(fd_path, errno);
+  }
+
+  // Returns the directory name in a path pointing to a file.
+  //
+  // Returns "." if the path does not contain any directory separator.
+  static std::string Dirname(const std::string& filename) {
+    std::string::size_type separator_pos = filename.rfind('/');
+    if (separator_pos == std::string::npos) {
+      return std::string(".");
+    }
+    // The filename component should not contain a path separator. If it does,
+    // the splitting was done incorrectly.
+    assert(filename.find('/', separator_pos + 1) == std::string::npos);
+
+    return filename.substr(0, separator_pos);
+  }
+
+  // Extracts the file name from a path pointing to a file.
+  //
+  // The returned Slice points to |filename|'s data buffer, so it is only valid
+  // while |filename| is alive and unchanged.
+  static Slice Basename(const std::string& filename) {
+    std::string::size_type separator_pos = filename.rfind('/');
+    if (separator_pos == std::string::npos) {
+      return Slice(filename);
+    }
+    // The filename component should not contain a path separator. If it does,
+    // the splitting was done incorrectly.
+    assert(filename.find('/', separator_pos + 1) == std::string::npos);
+
+    return Slice(filename.data() + separator_pos + 1,
+                 filename.length() - separator_pos - 1);
+  }
+
+  // True if the given file is a manifest file.
+  static bool IsManifest(const std::string& filename) {
+    return Basename(filename).starts_with("MANIFEST");
+  }
+
+  // buf_[0, pos_ - 1] contains data to be written to fd_.
+  char buf_[kWritableFileBufferSize];
+  size_t pos_;
+  int fd_;
+
+  const bool is_manifest_;  // True if the file's name starts with MANIFEST.
+  const std::string filename_;
+  const std::string dirname_;  // The directory of filename_.
+};
+
+int LockOrUnlock(int fd, bool lock) {
+  errno = 0;
+  struct ::flock file_lock_info;
+  std::memset(&file_lock_info, 0, sizeof(file_lock_info));
+  file_lock_info.l_type = (lock ? F_WRLCK : F_UNLCK);
+  file_lock_info.l_whence = SEEK_SET;
+  file_lock_info.l_start = 0;
+  file_lock_info.l_len = 0;  // Lock/unlock entire file.
+  return ::fcntl(fd, F_SETLK, &file_lock_info);
+}
+
+// Instances are thread-safe because they are immutable.
+class PosixFileLock : public FileLock {
+ public:
+  PosixFileLock(int fd, std::string filename)
+      : fd_(fd), filename_(std::move(filename)) {}
+
+  int fd() const { return fd_; }
+  const std::string& filename() const { return filename_; }
+
+ private:
+  const int fd_;
+  const std::string filename_;
+};
+
+// Tracks the files locked by PosixEnv::LockFile().
+//
+// We maintain a separate set instead of relying on fcntl(F_SETLK) because
+// fcntl(F_SETLK) does not provide any protection against multiple uses from the
+// same process.
+//
+// Instances are thread-safe because all member data is guarded by a mutex.
+class PosixLockTable {
+ public:
+  bool Insert(const std::string& fname) LOCKS_EXCLUDED(mu_) {
+    mu_.Lock();
+    bool succeeded = locked_files_.insert(fname).second;
+    mu_.Unlock();
+    return succeeded;
+  }
+  void Remove(const std::string& fname) LOCKS_EXCLUDED(mu_) {
+    mu_.Lock();
+    locked_files_.erase(fname);
+    mu_.Unlock();
+  }
+
+ private:
+  port::Mutex mu_;
+  std::set<std::string> locked_files_ GUARDED_BY(mu_);
+};
+
+class PosixEnv : public Env {
+ public:
+  PosixEnv();
+  ~PosixEnv() override {
+    static const char msg[] =
+        "PosixEnv singleton destroyed. Unsupported behavior!\n";
+    std::fwrite(msg, 1, sizeof(msg), stderr);
+    std::abort();
+  }
+
+  Status NewSequentialFile(const std::string& filename,
+                           SequentialFile** result) override {
+    int fd = ::open(filename.c_str(), O_RDONLY | kOpenBaseFlags);
+    if (fd < 0) {
+      *result = nullptr;
+      return PosixError(filename, errno);
+    }
+
+    *result = new PosixSequentialFile(filename, fd);
+    return Status::OK();
+  }
+
+  Status NewRandomAccessFile(const std::string& filename,
+                             RandomAccessFile** result) override {
+    *result = nullptr;
+    int fd = ::open(filename.c_str(), O_RDONLY | kOpenBaseFlags);
+    if (fd < 0) {
+      return PosixError(filename, errno);
+    }
+
+    if (!mmap_limiter_.Acquire()) {
+      *result = new PosixRandomAccessFile(filename, fd, &fd_limiter_);
+      return Status::OK();
+    }
+
+    uint64_t file_size;
+    Status status = GetFileSize(filename, &file_size);
+    if (status.ok()) {
+      void* mmap_base =
+          ::mmap(/*addr=*/nullptr, file_size, PROT_READ, MAP_SHARED, fd, 0);
+      if (mmap_base != MAP_FAILED) {
+        *result = new PosixMmapReadableFile(filename,
+                                            reinterpret_cast<char*>(mmap_base),
+                                            file_size, &mmap_limiter_);
+      } else {
+        status = PosixError(filename, errno);
+      }
+    }
+    ::close(fd);
+    if (!status.ok()) {
+      mmap_limiter_.Release();
+    }
+    return status;
+  }
+
+  Status NewWritableFile(const std::string& filename,
+                         WritableFile** result) override {
+    int fd = ::open(filename.c_str(),
+                    O_TRUNC | O_WRONLY | O_CREAT | kOpenBaseFlags, 0644);
+    if (fd < 0) {
+      *result = nullptr;
+      return PosixError(filename, errno);
+    }
+
+    *result = new PosixWritableFile(filename, fd);
+    return Status::OK();
+  }
+
+  Status NewAppendableFile(const std::string& filename,
+                           WritableFile** result) override {
+    int fd = ::open(filename.c_str(),
+                    O_APPEND | O_WRONLY | O_CREAT | kOpenBaseFlags, 0644);
+    if (fd < 0) {
+      *result = nullptr;
+      return PosixError(filename, errno);
+    }
+
+    *result = new PosixWritableFile(filename, fd);
+    return Status::OK();
+  }
+
+  bool FileExists(const std::string& filename) override {
+    return ::access(filename.c_str(), F_OK) == 0;
+  }
+
+  Status GetChildren(const std::string& directory_path,
+                     std::vector<std::string>* result) override {
+    result->clear();
+    ::DIR* dir = ::opendir(directory_path.c_str());
+    if (dir == nullptr) {
+      return PosixError(directory_path, errno);
+    }
+    struct ::dirent* entry;
+    while ((entry = ::readdir(dir)) != nullptr) {
+      result->emplace_back(entry->d_name);
+    }
+    ::closedir(dir);
+    return Status::OK();
+  }
+
+  Status RemoveFile(const std::string& filename) override {
+    if (::unlink(filename.c_str()) != 0) {
+      return PosixError(filename, errno);
+    }
+    return Status::OK();
+  }
+
+  Status CreateDir(const std::string& dirname) override {
+    if (::mkdir(dirname.c_str(), 0755) != 0) {
+      return PosixError(dirname, errno);
+    }
+    return Status::OK();
+  }
+
+  Status RemoveDir(const std::string& dirname) override {
+    if (::rmdir(dirname.c_str()) != 0) {
+      return PosixError(dirname, errno);
+    }
+    return Status::OK();
+  }
+
+  Status GetFileSize(const std::string& filename, uint64_t* size) override {
+    struct ::stat file_stat;
+    if (::stat(filename.c_str(), &file_stat) != 0) {
+      *size = 0;
+      return PosixError(filename, errno);
+    }
+    *size = file_stat.st_size;
+    return Status::OK();
+  }
+
+  Status RenameFile(const std::string& from, const std::string& to) override {
+    if (std::rename(from.c_str(), to.c_str()) != 0) {
+      return PosixError(from, errno);
+    }
+    return Status::OK();
+  }
+
+  Status LockFile(const std::string& filename, FileLock** lock) override {
+    *lock = nullptr;
+
+    int fd = ::open(filename.c_str(), O_RDWR | O_CREAT | kOpenBaseFlags, 0644);
+    if (fd < 0) {
+      return PosixError(filename, errno);
+    }
+
+    if (!locks_.Insert(filename)) {
+      ::close(fd);
+      return Status::IOError("lock " + filename, "already held by process");
+    }
+
+    if (LockOrUnlock(fd, true) == -1) {
+      int lock_errno = errno;
+      ::close(fd);
+      locks_.Remove(filename);
+      return PosixError("lock " + filename, lock_errno);
+    }
+
+    *lock = new PosixFileLock(fd, filename);
+    return Status::OK();
+  }
+
+  Status UnlockFile(FileLock* lock) override {
+    PosixFileLock* posix_file_lock = static_cast<PosixFileLock*>(lock);
+    if (LockOrUnlock(posix_file_lock->fd(), false) == -1) {
+      return PosixError("unlock " + posix_file_lock->filename(), errno);
+    }
+    locks_.Remove(posix_file_lock->filename());
+    ::close(posix_file_lock->fd());
+    delete posix_file_lock;
+    return Status::OK();
+  }
+
+  void Schedule(void (*background_work_function)(void* background_work_arg),
+                void* background_work_arg) override;
+
+  void StartThread(void (*thread_main)(void* thread_main_arg),
+                   void* thread_main_arg) override {
+    std::thread new_thread(thread_main, thread_main_arg);
+    new_thread.detach();
+  }
+
+  Status GetTestDirectory(std::string* result) override {
+    const char* env = std::getenv("TEST_TMPDIR");
+    if (env && env[0] != '\0') {
+      *result = env;
+    } else {
+      char buf[100];
+      std::snprintf(buf, sizeof(buf), "/tmp/leveldbtest-%d",
+                    static_cast<int>(::geteuid()));
+      *result = buf;
+    }
+
+    // The CreateDir status is ignored because the directory may already exist.
+    CreateDir(*result);
+
+    return Status::OK();
+  }
+
+  Status NewLogger(const std::string& filename, Logger** result) override {
+    int fd = ::open(filename.c_str(),
+                    O_APPEND | O_WRONLY | O_CREAT | kOpenBaseFlags, 0644);
+    if (fd < 0) {
+      *result = nullptr;
+      return PosixError(filename, errno);
+    }
+
+    std::FILE* fp = ::fdopen(fd, "w");
+    if (fp == nullptr) {
+      ::close(fd);
+      *result = nullptr;
+      return PosixError(filename, errno);
+    } else {
+      *result = new PosixLogger(fp);
+      return Status::OK();
+    }
+  }
+
+  uint64_t NowMicros() override {
+    static constexpr uint64_t kUsecondsPerSecond = 1000000;
+    struct ::timeval tv;
+    ::gettimeofday(&tv, nullptr);
+    return static_cast<uint64_t>(tv.tv_sec) * kUsecondsPerSecond + tv.tv_usec;
+  }
+
+  void SleepForMicroseconds(int micros) override {
+    std::this_thread::sleep_for(std::chrono::microseconds(micros));
+  }
+
+ private:
+  void BackgroundThreadMain();
+
+  static void BackgroundThreadEntryPoint(PosixEnv* env) {
+    env->BackgroundThreadMain();
+  }
+
+  // Stores the work item data in a Schedule() call.
+  //
+  // Instances are constructed on the thread calling Schedule() and used on the
+  // background thread.
+  //
+  // This structure is thread-safe beacuse it is immutable.
+  struct BackgroundWorkItem {
+    explicit BackgroundWorkItem(void (*function)(void* arg), void* arg)
+        : function(function), arg(arg) {}
+
+    void (*const function)(void*);
+    void* const arg;
+  };
+
+  port::Mutex background_work_mutex_;
+  port::CondVar background_work_cv_ GUARDED_BY(background_work_mutex_);
+  bool started_background_thread_ GUARDED_BY(background_work_mutex_);
+
+  std::queue<BackgroundWorkItem> background_work_queue_
+      GUARDED_BY(background_work_mutex_);
+
+  PosixLockTable locks_;  // Thread-safe.
+  Limiter mmap_limiter_;  // Thread-safe.
+  Limiter fd_limiter_;    // Thread-safe.
+};
+
+// Return the maximum number of concurrent mmaps.
+int MaxMmaps() { return g_mmap_limit; }
+
+// Return the maximum number of read-only files to keep open.
+int MaxOpenFiles() {
+  if (g_open_read_only_file_limit >= 0) {
+    return g_open_read_only_file_limit;
+  }
+  struct ::rlimit rlim;
+  if (::getrlimit(RLIMIT_NOFILE, &rlim)) {
+    // getrlimit failed, fallback to hard-coded default.
+    g_open_read_only_file_limit = 50;
+  } else if (rlim.rlim_cur == RLIM_INFINITY) {
+    g_open_read_only_file_limit = std::numeric_limits<int>::max();
+  } else {
+    // Allow use of 20% of available file descriptors for read-only files.
+    g_open_read_only_file_limit = rlim.rlim_cur / 5;
+  }
+  return g_open_read_only_file_limit;
+}
+
+}  // namespace
+
+PosixEnv::PosixEnv()
+    : background_work_cv_(&background_work_mutex_),
+      started_background_thread_(false),
+      mmap_limiter_(MaxMmaps()),
+      fd_limiter_(MaxOpenFiles()) {}
+
+void PosixEnv::Schedule(
+    void (*background_work_function)(void* background_work_arg),
+    void* background_work_arg) {
+  background_work_mutex_.Lock();
+
+  // Start the background thread, if we haven't done so already.
+  if (!started_background_thread_) {
+    started_background_thread_ = true;
+    std::thread background_thread(PosixEnv::BackgroundThreadEntryPoint, this);
+    background_thread.detach();
+  }
+
+  // If the queue is empty, the background thread may be waiting for work.
+  if (background_work_queue_.empty()) {
+    background_work_cv_.Signal();
+  }
+
+  background_work_queue_.emplace(background_work_function, background_work_arg);
+  background_work_mutex_.Unlock();
+}
+
+void PosixEnv::BackgroundThreadMain() {
+  while (true) {
+    background_work_mutex_.Lock();
+
+    // Wait until there is work to be done.
+    while (background_work_queue_.empty()) {
+      background_work_cv_.Wait();
+    }
+
+    assert(!background_work_queue_.empty());
+    auto background_work_function = background_work_queue_.front().function;
+    void* background_work_arg = background_work_queue_.front().arg;
+    background_work_queue_.pop();
+
+    background_work_mutex_.Unlock();
+    background_work_function(background_work_arg);
+  }
+}
+
+namespace {
+
+// Wraps an Env instance whose destructor is never created.
+//
+// Intended usage:
+//   using PlatformSingletonEnv = SingletonEnv<PlatformEnv>;
+//   void ConfigurePosixEnv(int param) {
+//     PlatformSingletonEnv::AssertEnvNotInitialized();
+//     // set global configuration flags.
+//   }
+//   Env* Env::Default() {
+//     static PlatformSingletonEnv default_env;
+//     return default_env.env();
+//   }
+template <typename EnvType>
+class SingletonEnv {
+ public:
+  SingletonEnv() {
+#if !defined(NDEBUG)
+    env_initialized_.store(true, std::memory_order::memory_order_relaxed);
+#endif  // !defined(NDEBUG)
+    static_assert(sizeof(env_storage_) >= sizeof(EnvType),
+                  "env_storage_ will not fit the Env");
+    static_assert(alignof(decltype(env_storage_)) >= alignof(EnvType),
+                  "env_storage_ does not meet the Env's alignment needs");
+    new (&env_storage_) EnvType();
+  }
+  ~SingletonEnv() = default;
+
+  SingletonEnv(const SingletonEnv&) = delete;
+  SingletonEnv& operator=(const SingletonEnv&) = delete;
+
+  Env* env() { return reinterpret_cast<Env*>(&env_storage_); }
+
+  static void AssertEnvNotInitialized() {
+#if !defined(NDEBUG)
+    assert(!env_initialized_.load(std::memory_order::memory_order_relaxed));
+#endif  // !defined(NDEBUG)
+  }
+
+ private:
+  typename std::aligned_storage<sizeof(EnvType), alignof(EnvType)>::type
+      env_storage_;
+#if !defined(NDEBUG)
+  static std::atomic<bool> env_initialized_;
+#endif  // !defined(NDEBUG)
+};
+
+#if !defined(NDEBUG)
+template <typename EnvType>
+std::atomic<bool> SingletonEnv<EnvType>::env_initialized_;
+#endif  // !defined(NDEBUG)
+
+using PosixDefaultEnv = SingletonEnv<PosixEnv>;
+
+}  // namespace
+
+void EnvPosixTestHelper::SetReadOnlyFDLimit(int limit) {
+  PosixDefaultEnv::AssertEnvNotInitialized();
+  g_open_read_only_file_limit = limit;
+}
+
+void EnvPosixTestHelper::SetReadOnlyMMapLimit(int limit) {
+  PosixDefaultEnv::AssertEnvNotInitialized();
+  g_mmap_limit = limit;
+}
+
+Env* Env::Default() {
+  static PosixDefaultEnv env_container;
+  return env_container.env();
+}
+
+}  // namespace leveldb

lib/leveldb/util/env_posix_test.cc

@@ -0,0 +1,353 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include <sys/resource.h>
+#include <sys/wait.h>
+#include <unistd.h>
+
+#include <cstdio>
+#include <cstdlib>
+#include <cstring>
+#include <string>
+#include <unordered_set>
+#include <vector>
+
+#include "gtest/gtest.h"
+#include "leveldb/env.h"
+#include "port/port.h"
+#include "util/env_posix_test_helper.h"
+#include "util/testutil.h"
+
+#if HAVE_O_CLOEXEC
+
+namespace {
+
+// Exit codes for the helper process spawned by TestCloseOnExec* tests.
+// Useful for debugging test failures.
+constexpr int kTextCloseOnExecHelperExecFailedCode = 61;
+constexpr int kTextCloseOnExecHelperDup2FailedCode = 62;
+constexpr int kTextCloseOnExecHelperFoundOpenFdCode = 63;
+
+// Global set by main() and read in TestCloseOnExec.
+//
+// The argv[0] value is stored in a std::vector instead of a std::string because
+// std::string does not return a mutable pointer to its buffer until C++17.
+//
+// The vector stores the string pointed to by argv[0], plus the trailing null.
+std::vector<char>* GetArgvZero() {
+  static std::vector<char> program_name;
+  return &program_name;
+}
+
+// Command-line switch used to run this test as the CloseOnExecSwitch helper.
+static const char kTestCloseOnExecSwitch[] = "--test-close-on-exec-helper";
+
+// Executed in a separate process by TestCloseOnExec* tests.
+//
+// main() delegates to this function when the test executable is launched with
+// a special command-line switch. TestCloseOnExec* tests fork()+exec() the test
+// executable and pass the special command-line switch.
+//
+
+// main() delegates to this function when the test executable is launched with
+// a special command-line switch. TestCloseOnExec* tests fork()+exec() the test
+// executable and pass the special command-line switch.
+//
+// When main() delegates to this function, the process probes whether a given
+// file descriptor is open, and communicates the result via its exit code.
+int TestCloseOnExecHelperMain(char* pid_arg) {
+  int fd = std::atoi(pid_arg);
+  // When given the same file descriptor twice, dup2() returns -1 if the
+  // file descriptor is closed, or the given file descriptor if it is open.
+  if (::dup2(fd, fd) == fd) {
+    std::fprintf(stderr, "Unexpected open fd %d\n", fd);
+    return kTextCloseOnExecHelperFoundOpenFdCode;
+  }
+  // Double-check that dup2() is saying the file descriptor is closed.
+  if (errno != EBADF) {
+    std::fprintf(stderr, "Unexpected errno after calling dup2 on fd %d: %s\n",
+                 fd, std::strerror(errno));
+    return kTextCloseOnExecHelperDup2FailedCode;
+  }
+  return 0;
+}
+
+// File descriptors are small non-negative integers.
+//
+// Returns void so the implementation can use ASSERT_EQ.
+void GetMaxFileDescriptor(int* result_fd) {
+  // Get the maximum file descriptor number.
+  ::rlimit fd_rlimit;
+  ASSERT_EQ(0, ::getrlimit(RLIMIT_NOFILE, &fd_rlimit));
+  *result_fd = fd_rlimit.rlim_cur;
+}
+
+// Iterates through all possible FDs and returns the currently open ones.
+//
+// Returns void so the implementation can use ASSERT_EQ.
+void GetOpenFileDescriptors(std::unordered_set<int>* open_fds) {
+  int max_fd = 0;
+  GetMaxFileDescriptor(&max_fd);
+
+  for (int fd = 0; fd < max_fd; ++fd) {
+    if (::dup2(fd, fd) != fd) {
+      // When given the same file descriptor twice, dup2() returns -1 if the
+      // file descriptor is closed, or the given file descriptor if it is open.
+      //
+      // Double-check that dup2() is saying the fd is closed.
+      ASSERT_EQ(EBADF, errno)
+          << "dup2() should set errno to EBADF on closed file descriptors";
+      continue;
+    }
+    open_fds->insert(fd);
+  }
+}
+
+// Finds an FD open since a previous call to GetOpenFileDescriptors().
+//
+// |baseline_open_fds| is the result of a previous GetOpenFileDescriptors()
+// call. Assumes that exactly one FD was opened since that call.
+//
+// Returns void so the implementation can use ASSERT_EQ.
+void GetNewlyOpenedFileDescriptor(
+    const std::unordered_set<int>& baseline_open_fds, int* result_fd) {
+  std::unordered_set<int> open_fds;
+  GetOpenFileDescriptors(&open_fds);
+  for (int fd : baseline_open_fds) {
+    ASSERT_EQ(1, open_fds.count(fd))
+        << "Previously opened file descriptor was closed during test setup";
+    open_fds.erase(fd);
+  }
+  ASSERT_EQ(1, open_fds.size())
+      << "Expected exactly one newly opened file descriptor during test setup";
+  *result_fd = *open_fds.begin();
+}
+
+// Check that a fork()+exec()-ed child process does not have an extra open FD.
+void CheckCloseOnExecDoesNotLeakFDs(
+    const std::unordered_set<int>& baseline_open_fds) {
+  // Prepare the argument list for the child process.
+  // execv() wants mutable buffers.
+  char switch_buffer[sizeof(kTestCloseOnExecSwitch)];
+  std::memcpy(switch_buffer, kTestCloseOnExecSwitch,
+              sizeof(kTestCloseOnExecSwitch));
+
+  int probed_fd;
+  GetNewlyOpenedFileDescriptor(baseline_open_fds, &probed_fd);
+  std::string fd_string = std::to_string(probed_fd);
+  std::vector<char> fd_buffer(fd_string.begin(), fd_string.end());
+  fd_buffer.emplace_back('\0');
+
+  // The helper process is launched with the command below.
+  //      env_posix_tests --test-close-on-exec-helper 3
+  char* child_argv[] = {GetArgvZero()->data(), switch_buffer, fd_buffer.data(),
+                        nullptr};
+
+  constexpr int kForkInChildProcessReturnValue = 0;
+  int child_pid = fork();
+  if (child_pid == kForkInChildProcessReturnValue) {
+    ::execv(child_argv[0], child_argv);
+    std::fprintf(stderr, "Error spawning child process: %s\n", strerror(errno));
+    std::exit(kTextCloseOnExecHelperExecFailedCode);
+  }
+
+  int child_status = 0;
+  ASSERT_EQ(child_pid, ::waitpid(child_pid, &child_status, 0));
+  ASSERT_TRUE(WIFEXITED(child_status))
+      << "The helper process did not exit with an exit code";
+  ASSERT_EQ(0, WEXITSTATUS(child_status))
+      << "The helper process encountered an error";
+}
+
+}  // namespace
+
+#endif  // HAVE_O_CLOEXEC
+
+namespace leveldb {
+
+static const int kReadOnlyFileLimit = 4;
+static const int kMMapLimit = 4;
+
+class EnvPosixTest : public testing::Test {
+ public:
+  static void SetFileLimits(int read_only_file_limit, int mmap_limit) {
+    EnvPosixTestHelper::SetReadOnlyFDLimit(read_only_file_limit);
+    EnvPosixTestHelper::SetReadOnlyMMapLimit(mmap_limit);
+  }
+
+  EnvPosixTest() : env_(Env::Default()) {}
+
+  Env* env_;
+};
+
+TEST_F(EnvPosixTest, TestOpenOnRead) {
+  // Write some test data to a single file that will be opened |n| times.
+  std::string test_dir;
+  ASSERT_LEVELDB_OK(env_->GetTestDirectory(&test_dir));
+  std::string test_file = test_dir + "/open_on_read.txt";
+
+  FILE* f = std::fopen(test_file.c_str(), "we");
+  ASSERT_TRUE(f != nullptr);
+  const char kFileData[] = "abcdefghijklmnopqrstuvwxyz";
+  fputs(kFileData, f);
+  std::fclose(f);
+
+  // Open test file some number above the sum of the two limits to force
+  // open-on-read behavior of POSIX Env leveldb::RandomAccessFile.
+  const int kNumFiles = kReadOnlyFileLimit + kMMapLimit + 5;
+  leveldb::RandomAccessFile* files[kNumFiles] = {0};
+  for (int i = 0; i < kNumFiles; i++) {
+    ASSERT_LEVELDB_OK(env_->NewRandomAccessFile(test_file, &files[i]));
+  }
+  char scratch;
+  Slice read_result;
+  for (int i = 0; i < kNumFiles; i++) {
+    ASSERT_LEVELDB_OK(files[i]->Read(i, 1, &read_result, &scratch));
+    ASSERT_EQ(kFileData[i], read_result[0]);
+  }
+  for (int i = 0; i < kNumFiles; i++) {
+    delete files[i];
+  }
+  ASSERT_LEVELDB_OK(env_->RemoveFile(test_file));
+}
+
+#if HAVE_O_CLOEXEC
+
+TEST_F(EnvPosixTest, TestCloseOnExecSequentialFile) {
+  std::unordered_set<int> open_fds;
+  GetOpenFileDescriptors(&open_fds);
+
+  std::string test_dir;
+  ASSERT_LEVELDB_OK(env_->GetTestDirectory(&test_dir));
+  std::string file_path = test_dir + "/close_on_exec_sequential.txt";
+  ASSERT_LEVELDB_OK(WriteStringToFile(env_, "0123456789", file_path));
+
+  leveldb::SequentialFile* file = nullptr;
+  ASSERT_LEVELDB_OK(env_->NewSequentialFile(file_path, &file));
+  CheckCloseOnExecDoesNotLeakFDs(open_fds);
+  delete file;
+
+  ASSERT_LEVELDB_OK(env_->RemoveFile(file_path));
+}
+
+TEST_F(EnvPosixTest, TestCloseOnExecRandomAccessFile) {
+  std::unordered_set<int> open_fds;
+  GetOpenFileDescriptors(&open_fds);
+
+  std::string test_dir;
+  ASSERT_LEVELDB_OK(env_->GetTestDirectory(&test_dir));
+  std::string file_path = test_dir + "/close_on_exec_random_access.txt";
+  ASSERT_LEVELDB_OK(WriteStringToFile(env_, "0123456789", file_path));
+
+  // Exhaust the RandomAccessFile mmap limit. This way, the test
+  // RandomAccessFile instance below is backed by a file descriptor, not by an
+  // mmap region.
+  leveldb::RandomAccessFile* mmapped_files[kReadOnlyFileLimit] = {nullptr};
+  for (int i = 0; i < kReadOnlyFileLimit; i++) {
+    ASSERT_LEVELDB_OK(env_->NewRandomAccessFile(file_path, &mmapped_files[i]));
+  }
+
+  leveldb::RandomAccessFile* file = nullptr;
+  ASSERT_LEVELDB_OK(env_->NewRandomAccessFile(file_path, &file));
+  CheckCloseOnExecDoesNotLeakFDs(open_fds);
+  delete file;
+
+  for (int i = 0; i < kReadOnlyFileLimit; i++) {
+    delete mmapped_files[i];
+  }
+  ASSERT_LEVELDB_OK(env_->RemoveFile(file_path));
+}
+
+TEST_F(EnvPosixTest, TestCloseOnExecWritableFile) {
+  std::unordered_set<int> open_fds;
+  GetOpenFileDescriptors(&open_fds);
+
+  std::string test_dir;
+  ASSERT_LEVELDB_OK(env_->GetTestDirectory(&test_dir));
+  std::string file_path = test_dir + "/close_on_exec_writable.txt";
+  ASSERT_LEVELDB_OK(WriteStringToFile(env_, "0123456789", file_path));
+
+  leveldb::WritableFile* file = nullptr;
+  ASSERT_LEVELDB_OK(env_->NewWritableFile(file_path, &file));
+  CheckCloseOnExecDoesNotLeakFDs(open_fds);
+  delete file;
+
+  ASSERT_LEVELDB_OK(env_->RemoveFile(file_path));
+}
+
+TEST_F(EnvPosixTest, TestCloseOnExecAppendableFile) {
+  std::unordered_set<int> open_fds;
+  GetOpenFileDescriptors(&open_fds);
+
+  std::string test_dir;
+  ASSERT_LEVELDB_OK(env_->GetTestDirectory(&test_dir));
+  std::string file_path = test_dir + "/close_on_exec_appendable.txt";
+  ASSERT_LEVELDB_OK(WriteStringToFile(env_, "0123456789", file_path));
+
+  leveldb::WritableFile* file = nullptr;
+  ASSERT_LEVELDB_OK(env_->NewAppendableFile(file_path, &file));
+  CheckCloseOnExecDoesNotLeakFDs(open_fds);
+  delete file;
+
+  ASSERT_LEVELDB_OK(env_->RemoveFile(file_path));
+}
+
+TEST_F(EnvPosixTest, TestCloseOnExecLockFile) {
+  std::unordered_set<int> open_fds;
+  GetOpenFileDescriptors(&open_fds);
+
+  std::string test_dir;
+  ASSERT_LEVELDB_OK(env_->GetTestDirectory(&test_dir));
+  std::string file_path = test_dir + "/close_on_exec_lock.txt";
+  ASSERT_LEVELDB_OK(WriteStringToFile(env_, "0123456789", file_path));
+
+  leveldb::FileLock* lock = nullptr;
+  ASSERT_LEVELDB_OK(env_->LockFile(file_path, &lock));
+  CheckCloseOnExecDoesNotLeakFDs(open_fds);
+  ASSERT_LEVELDB_OK(env_->UnlockFile(lock));
+
+  ASSERT_LEVELDB_OK(env_->RemoveFile(file_path));
+}
+
+TEST_F(EnvPosixTest, TestCloseOnExecLogger) {
+  std::unordered_set<int> open_fds;
+  GetOpenFileDescriptors(&open_fds);
+
+  std::string test_dir;
+  ASSERT_LEVELDB_OK(env_->GetTestDirectory(&test_dir));
+  std::string file_path = test_dir + "/close_on_exec_logger.txt";
+  ASSERT_LEVELDB_OK(WriteStringToFile(env_, "0123456789", file_path));
+
+  leveldb::Logger* file = nullptr;
+  ASSERT_LEVELDB_OK(env_->NewLogger(file_path, &file));
+  CheckCloseOnExecDoesNotLeakFDs(open_fds);
+  delete file;
+
+  ASSERT_LEVELDB_OK(env_->RemoveFile(file_path));
+}
+
+#endif  // HAVE_O_CLOEXEC
+
+}  // namespace leveldb
+
+int main(int argc, char** argv) {
+#if HAVE_O_CLOEXEC
+  // Check if we're invoked as a helper program, or as the test suite.
+  for (int i = 1; i < argc; ++i) {
+    if (!std::strcmp(argv[i], kTestCloseOnExecSwitch)) {
+      return TestCloseOnExecHelperMain(argv[i + 1]);
+    }
+  }
+
+  // Save argv[0] early, because googletest may modify argv.
+  GetArgvZero()->assign(argv[0], argv[0] + std::strlen(argv[0]) + 1);
+#endif  // HAVE_O_CLOEXEC
+
+  // All tests currently run with the same read-only file limits.
+  leveldb::EnvPosixTest::SetFileLimits(leveldb::kReadOnlyFileLimit,
+                                       leveldb::kMMapLimit);
+
+  testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}

lib/leveldb/util/env_posix_test_helper.h

@@ -0,0 +1,28 @@
+// Copyright 2017 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#ifndef STORAGE_LEVELDB_UTIL_ENV_POSIX_TEST_HELPER_H_
+#define STORAGE_LEVELDB_UTIL_ENV_POSIX_TEST_HELPER_H_
+
+namespace leveldb {
+
+class EnvPosixTest;
+
+// A helper for the POSIX Env to facilitate testing.
+class EnvPosixTestHelper {
+ private:
+  friend class EnvPosixTest;
+
+  // Set the maximum number of read-only files that will be opened.
+  // Must be called before creating an Env.
+  static void SetReadOnlyFDLimit(int limit);
+
+  // Set the maximum number of read-only files that will be mapped via mmap.
+  // Must be called before creating an Env.
+  static void SetReadOnlyMMapLimit(int limit);
+};
+
+}  // namespace leveldb
+
+#endif  // STORAGE_LEVELDB_UTIL_ENV_POSIX_TEST_HELPER_H_

lib/leveldb/util/env_test.cc

@@ -0,0 +1,240 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "leveldb/env.h"
+
+#include <algorithm>
+
+#include "gtest/gtest.h"
+#include "port/port.h"
+#include "port/thread_annotations.h"
+#include "util/mutexlock.h"
+#include "util/testutil.h"
+
+namespace leveldb {
+
+static const int kDelayMicros = 100000;
+
+class EnvTest : public testing::Test {
+ public:
+  EnvTest() : env_(Env::Default()) {}
+
+  Env* env_;
+};
+
+TEST_F(EnvTest, ReadWrite) {
+  Random rnd(test::RandomSeed());
+
+  // Get file to use for testing.
+  std::string test_dir;
+  ASSERT_LEVELDB_OK(env_->GetTestDirectory(&test_dir));
+  std::string test_file_name = test_dir + "/open_on_read.txt";
+  WritableFile* writable_file;
+  ASSERT_LEVELDB_OK(env_->NewWritableFile(test_file_name, &writable_file));
+
+  // Fill a file with data generated via a sequence of randomly sized writes.
+  static const size_t kDataSize = 10 * 1048576;
+  std::string data;
+  while (data.size() < kDataSize) {
+    int len = rnd.Skewed(18);  // Up to 2^18 - 1, but typically much smaller
+    std::string r;
+    test::RandomString(&rnd, len, &r);
+    ASSERT_LEVELDB_OK(writable_file->Append(r));
+    data += r;
+    if (rnd.OneIn(10)) {
+      ASSERT_LEVELDB_OK(writable_file->Flush());
+    }
+  }
+  ASSERT_LEVELDB_OK(writable_file->Sync());
+  ASSERT_LEVELDB_OK(writable_file->Close());
+  delete writable_file;
+
+  // Read all data using a sequence of randomly sized reads.
+  SequentialFile* sequential_file;
+  ASSERT_LEVELDB_OK(env_->NewSequentialFile(test_file_name, &sequential_file));
+  std::string read_result;
+  std::string scratch;
+  while (read_result.size() < data.size()) {
+    int len = std::min<int>(rnd.Skewed(18), data.size() - read_result.size());
+    scratch.resize(std::max(len, 1));  // at least 1 so &scratch[0] is legal
+    Slice read;
+    ASSERT_LEVELDB_OK(sequential_file->Read(len, &read, &scratch[0]));
+    if (len > 0) {
+      ASSERT_GT(read.size(), 0);
+    }
+    ASSERT_LE(read.size(), len);
+    read_result.append(read.data(), read.size());
+  }
+  ASSERT_EQ(read_result, data);
+  delete sequential_file;
+}
+
+TEST_F(EnvTest, RunImmediately) {
+  struct RunState {
+    port::Mutex mu;
+    port::CondVar cvar{&mu};
+    bool called = false;
+
+    static void Run(void* arg) {
+      RunState* state = reinterpret_cast<RunState*>(arg);
+      MutexLock l(&state->mu);
+      ASSERT_EQ(state->called, false);
+      state->called = true;
+      state->cvar.Signal();
+    }
+  };
+
+  RunState state;
+  env_->Schedule(&RunState::Run, &state);
+
+  MutexLock l(&state.mu);
+  while (!state.called) {
+    state.cvar.Wait();
+  }
+}
+
+TEST_F(EnvTest, RunMany) {
+  struct RunState {
+    port::Mutex mu;
+    port::CondVar cvar{&mu};
+    int last_id = 0;
+  };
+
+  struct Callback {
+    RunState* state_;  // Pointer to shared state.
+    const int id_;  // Order# for the execution of this callback.
+
+    Callback(RunState* s, int id) : state_(s), id_(id) {}
+
+    static void Run(void* arg) {
+      Callback* callback = reinterpret_cast<Callback*>(arg);
+      RunState* state = callback->state_;
+
+      MutexLock l(&state->mu);
+      ASSERT_EQ(state->last_id, callback->id_ - 1);
+      state->last_id = callback->id_;
+      state->cvar.Signal();
+    }
+  };
+
+  RunState state;
+  Callback callback1(&state, 1);
+  Callback callback2(&state, 2);
+  Callback callback3(&state, 3);
+  Callback callback4(&state, 4);
+  env_->Schedule(&Callback::Run, &callback1);
+  env_->Schedule(&Callback::Run, &callback2);
+  env_->Schedule(&Callback::Run, &callback3);
+  env_->Schedule(&Callback::Run, &callback4);
+
+  MutexLock l(&state.mu);
+  while (state.last_id != 4) {
+    state.cvar.Wait();
+  }
+}
+
+struct State {
+  port::Mutex mu;
+  port::CondVar cvar{&mu};
+
+  int val GUARDED_BY(mu);
+  int num_running GUARDED_BY(mu);
+
+  State(int val, int num_running) : val(val), num_running(num_running) {}
+};
+
+static void ThreadBody(void* arg) {
+  State* s = reinterpret_cast<State*>(arg);
+  s->mu.Lock();
+  s->val += 1;
+  s->num_running -= 1;
+  s->cvar.Signal();
+  s->mu.Unlock();
+}
+
+TEST_F(EnvTest, StartThread) {
+  State state(0, 3);
+  for (int i = 0; i < 3; i++) {
+    env_->StartThread(&ThreadBody, &state);
+  }
+
+  MutexLock l(&state.mu);
+  while (state.num_running != 0) {
+    state.cvar.Wait();
+  }
+  ASSERT_EQ(state.val, 3);
+}
+
+TEST_F(EnvTest, TestOpenNonExistentFile) {
+  // Write some test data to a single file that will be opened |n| times.
+  std::string test_dir;
+  ASSERT_LEVELDB_OK(env_->GetTestDirectory(&test_dir));
+
+  std::string non_existent_file = test_dir + "/non_existent_file";
+  ASSERT_TRUE(!env_->FileExists(non_existent_file));
+
+  RandomAccessFile* random_access_file;
+  Status status =
+      env_->NewRandomAccessFile(non_existent_file, &random_access_file);
+  ASSERT_TRUE(status.IsNotFound());
+
+  SequentialFile* sequential_file;
+  status = env_->NewSequentialFile(non_existent_file, &sequential_file);
+  ASSERT_TRUE(status.IsNotFound());
+}
+
+TEST_F(EnvTest, ReopenWritableFile) {
+  std::string test_dir;
+  ASSERT_LEVELDB_OK(env_->GetTestDirectory(&test_dir));
+  std::string test_file_name = test_dir + "/reopen_writable_file.txt";
+  env_->RemoveFile(test_file_name);
+
+  WritableFile* writable_file;
+  ASSERT_LEVELDB_OK(env_->NewWritableFile(test_file_name, &writable_file));
+  std::string data("hello world!");
+  ASSERT_LEVELDB_OK(writable_file->Append(data));
+  ASSERT_LEVELDB_OK(writable_file->Close());
+  delete writable_file;
+
+  ASSERT_LEVELDB_OK(env_->NewWritableFile(test_file_name, &writable_file));
+  data = "42";
+  ASSERT_LEVELDB_OK(writable_file->Append(data));
+  ASSERT_LEVELDB_OK(writable_file->Close());
+  delete writable_file;
+
+  ASSERT_LEVELDB_OK(ReadFileToString(env_, test_file_name, &data));
+  ASSERT_EQ(std::string("42"), data);
+  env_->RemoveFile(test_file_name);
+}
+
+TEST_F(EnvTest, ReopenAppendableFile) {
+  std::string test_dir;
+  ASSERT_LEVELDB_OK(env_->GetTestDirectory(&test_dir));
+  std::string test_file_name = test_dir + "/reopen_appendable_file.txt";
+  env_->RemoveFile(test_file_name);
+
+  WritableFile* appendable_file;
+  ASSERT_LEVELDB_OK(env_->NewAppendableFile(test_file_name, &appendable_file));
+  std::string data("hello world!");
+  ASSERT_LEVELDB_OK(appendable_file->Append(data));
+  ASSERT_LEVELDB_OK(appendable_file->Close());
+  delete appendable_file;
+
+  ASSERT_LEVELDB_OK(env_->NewAppendableFile(test_file_name, &appendable_file));
+  data = "42";
+  ASSERT_LEVELDB_OK(appendable_file->Append(data));
+  ASSERT_LEVELDB_OK(appendable_file->Close());
+  delete appendable_file;
+
+  ASSERT_LEVELDB_OK(ReadFileToString(env_, test_file_name, &data));
+  ASSERT_EQ(std::string("hello world!42"), data);
+  env_->RemoveFile(test_file_name);
+}
+
+}  // namespace leveldb
+
+int main(int argc, char** argv) {
+  testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}

lib/leveldb/util/env_windows.cc

@@ -0,0 +1,796 @@
+// Copyright (c) 2018 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+// Prevent Windows headers from defining min/max macros and instead
+// use STL.
+#ifndef NOMINMAX
+#define NOMINMAX
+#endif  // ifndef NOMINMAX
+#include <windows.h>
+
+#include <algorithm>
+#include <atomic>
+#include <chrono>
+#include <condition_variable>
+#include <cstddef>
+#include <cstdint>
+#include <cstdlib>
+#include <cstring>
+#include <memory>
+#include <mutex>
+#include <queue>
+#include <sstream>
+#include <string>
+#include <vector>
+
+#include "leveldb/env.h"
+#include "leveldb/slice.h"
+#include "port/port.h"
+#include "port/thread_annotations.h"
+#include "util/env_windows_test_helper.h"
+#include "util/logging.h"
+#include "util/mutexlock.h"
+#include "util/windows_logger.h"
+
+namespace leveldb {
+
+namespace {
+
+constexpr const size_t kWritableFileBufferSize = 65536;
+
+// Up to 1000 mmaps for 64-bit binaries; none for 32-bit.
+constexpr int kDefaultMmapLimit = (sizeof(void*) >= 8) ? 1000 : 0;
+
+// Can be set by by EnvWindowsTestHelper::SetReadOnlyMMapLimit().
+int g_mmap_limit = kDefaultMmapLimit;
+
+std::string GetWindowsErrorMessage(DWORD error_code) {
+  std::string message;
+  char* error_text = nullptr;
+  // Use MBCS version of FormatMessage to match return value.
+  size_t error_text_size = ::FormatMessageA(
+      FORMAT_MESSAGE_FROM_SYSTEM | FORMAT_MESSAGE_ALLOCATE_BUFFER |
+          FORMAT_MESSAGE_IGNORE_INSERTS,
+      nullptr, error_code, MAKELANGID(LANG_NEUTRAL, SUBLANG_DEFAULT),
+      reinterpret_cast<char*>(&error_text), 0, nullptr);
+  if (!error_text) {
+    return message;
+  }
+  message.assign(error_text, error_text_size);
+  ::LocalFree(error_text);
+  return message;
+}
+
+Status WindowsError(const std::string& context, DWORD error_code) {
+  if (error_code == ERROR_FILE_NOT_FOUND || error_code == ERROR_PATH_NOT_FOUND)
+    return Status::NotFound(context, GetWindowsErrorMessage(error_code));
+  return Status::IOError(context, GetWindowsErrorMessage(error_code));
+}
+
+class ScopedHandle {
+ public:
+  ScopedHandle(HANDLE handle) : handle_(handle) {}
+  ScopedHandle(const ScopedHandle&) = delete;
+  ScopedHandle(ScopedHandle&& other) noexcept : handle_(other.Release()) {}
+  ~ScopedHandle() { Close(); }
+
+  ScopedHandle& operator=(const ScopedHandle&) = delete;
+
+  ScopedHandle& operator=(ScopedHandle&& rhs) noexcept {
+    if (this != &rhs) handle_ = rhs.Release();
+    return *this;
+  }
+
+  bool Close() {
+    if (!is_valid()) {
+      return true;
+    }
+    HANDLE h = handle_;
+    handle_ = INVALID_HANDLE_VALUE;
+    return ::CloseHandle(h);
+  }
+
+  bool is_valid() const {
+    return handle_ != INVALID_HANDLE_VALUE && handle_ != nullptr;
+  }
+
+  HANDLE get() const { return handle_; }
+
+  HANDLE Release() {
+    HANDLE h = handle_;
+    handle_ = INVALID_HANDLE_VALUE;
+    return h;
+  }
+
+ private:
+  HANDLE handle_;
+};
+
+// Helper class to limit resource usage to avoid exhaustion.
+// Currently used to limit read-only file descriptors and mmap file usage
+// so that we do not run out of file descriptors or virtual memory, or run into
+// kernel performance problems for very large databases.
+class Limiter {
+ public:
+  // Limit maximum number of resources to |max_acquires|.
+  Limiter(int max_acquires) : acquires_allowed_(max_acquires) {}
+
+  Limiter(const Limiter&) = delete;
+  Limiter operator=(const Limiter&) = delete;
+
+  // If another resource is available, acquire it and return true.
+  // Else return false.
+  bool Acquire() {
+    int old_acquires_allowed =
+        acquires_allowed_.fetch_sub(1, std::memory_order_relaxed);
+
+    if (old_acquires_allowed > 0) return true;
+
+    acquires_allowed_.fetch_add(1, std::memory_order_relaxed);
+    return false;
+  }
+
+  // Release a resource acquired by a previous call to Acquire() that returned
+  // true.
+  void Release() { acquires_allowed_.fetch_add(1, std::memory_order_relaxed); }
+
+ private:
+  // The number of available resources.
+  //
+  // This is a counter and is not tied to the invariants of any other class, so
+  // it can be operated on safely using std::memory_order_relaxed.
+  std::atomic<int> acquires_allowed_;
+};
+
+class WindowsSequentialFile : public SequentialFile {
+ public:
+  WindowsSequentialFile(std::string filename, ScopedHandle handle)
+      : handle_(std::move(handle)), filename_(std::move(filename)) {}
+  ~WindowsSequentialFile() override {}
+
+  Status Read(size_t n, Slice* result, char* scratch) override {
+    DWORD bytes_read;
+    // DWORD is 32-bit, but size_t could technically be larger. However leveldb
+    // files are limited to leveldb::Options::max_file_size which is clamped to
+    // 1<<30 or 1 GiB.
+    assert(n <= std::numeric_limits<DWORD>::max());
+    if (!::ReadFile(handle_.get(), scratch, static_cast<DWORD>(n), &bytes_read,
+                    nullptr)) {
+      return WindowsError(filename_, ::GetLastError());
+    }
+
+    *result = Slice(scratch, bytes_read);
+    return Status::OK();
+  }
+
+  Status Skip(uint64_t n) override {
+    LARGE_INTEGER distance;
+    distance.QuadPart = n;
+    if (!::SetFilePointerEx(handle_.get(), distance, nullptr, FILE_CURRENT)) {
+      return WindowsError(filename_, ::GetLastError());
+    }
+    return Status::OK();
+  }
+
+ private:
+  const ScopedHandle handle_;
+  const std::string filename_;
+};
+
+class WindowsRandomAccessFile : public RandomAccessFile {
+ public:
+  WindowsRandomAccessFile(std::string filename, ScopedHandle handle)
+      : handle_(std::move(handle)), filename_(std::move(filename)) {}
+
+  ~WindowsRandomAccessFile() override = default;
+
+  Status Read(uint64_t offset, size_t n, Slice* result,
+              char* scratch) const override {
+    DWORD bytes_read = 0;
+    OVERLAPPED overlapped = {0};
+
+    overlapped.OffsetHigh = static_cast<DWORD>(offset >> 32);
+    overlapped.Offset = static_cast<DWORD>(offset);
+    if (!::ReadFile(handle_.get(), scratch, static_cast<DWORD>(n), &bytes_read,
+                    &overlapped)) {
+      DWORD error_code = ::GetLastError();
+      if (error_code != ERROR_HANDLE_EOF) {
+        *result = Slice(scratch, 0);
+        return Status::IOError(filename_, GetWindowsErrorMessage(error_code));
+      }
+    }
+
+    *result = Slice(scratch, bytes_read);
+    return Status::OK();
+  }
+
+ private:
+  const ScopedHandle handle_;
+  const std::string filename_;
+};
+
+class WindowsMmapReadableFile : public RandomAccessFile {
+ public:
+  // base[0,length-1] contains the mmapped contents of the file.
+  WindowsMmapReadableFile(std::string filename, char* mmap_base, size_t length,
+                          Limiter* mmap_limiter)
+      : mmap_base_(mmap_base),
+        length_(length),
+        mmap_limiter_(mmap_limiter),
+        filename_(std::move(filename)) {}
+
+  ~WindowsMmapReadableFile() override {
+    ::UnmapViewOfFile(mmap_base_);
+    mmap_limiter_->Release();
+  }
+
+  Status Read(uint64_t offset, size_t n, Slice* result,
+              char* scratch) const override {
+    if (offset + n > length_) {
+      *result = Slice();
+      return WindowsError(filename_, ERROR_INVALID_PARAMETER);
+    }
+
+    *result = Slice(mmap_base_ + offset, n);
+    return Status::OK();
+  }
+
+ private:
+  char* const mmap_base_;
+  const size_t length_;
+  Limiter* const mmap_limiter_;
+  const std::string filename_;
+};
+
+class WindowsWritableFile : public WritableFile {
+ public:
+  WindowsWritableFile(std::string filename, ScopedHandle handle)
+      : pos_(0), handle_(std::move(handle)), filename_(std::move(filename)) {}
+
+  ~WindowsWritableFile() override = default;
+
+  Status Append(const Slice& data) override {
+    size_t write_size = data.size();
+    const char* write_data = data.data();
+
+    // Fit as much as possible into buffer.
+    size_t copy_size = std::min(write_size, kWritableFileBufferSize - pos_);
+    std::memcpy(buf_ + pos_, write_data, copy_size);
+    write_data += copy_size;
+    write_size -= copy_size;
+    pos_ += copy_size;
+    if (write_size == 0) {
+      return Status::OK();
+    }
+
+    // Can't fit in buffer, so need to do at least one write.
+    Status status = FlushBuffer();
+    if (!status.ok()) {
+      return status;
+    }
+
+    // Small writes go to buffer, large writes are written directly.
+    if (write_size < kWritableFileBufferSize) {
+      std::memcpy(buf_, write_data, write_size);
+      pos_ = write_size;
+      return Status::OK();
+    }
+    return WriteUnbuffered(write_data, write_size);
+  }
+
+  Status Close() override {
+    Status status = FlushBuffer();
+    if (!handle_.Close() && status.ok()) {
+      status = WindowsError(filename_, ::GetLastError());
+    }
+    return status;
+  }
+
+  Status Flush() override { return FlushBuffer(); }
+
+  Status Sync() override {
+    // On Windows no need to sync parent directory. Its metadata will be updated
+    // via the creation of the new file, without an explicit sync.
+
+    Status status = FlushBuffer();
+    if (!status.ok()) {
+      return status;
+    }
+
+    if (!::FlushFileBuffers(handle_.get())) {
+      return Status::IOError(filename_,
+                             GetWindowsErrorMessage(::GetLastError()));
+    }
+    return Status::OK();
+  }
+
+ private:
+  Status FlushBuffer() {
+    Status status = WriteUnbuffered(buf_, pos_);
+    pos_ = 0;
+    return status;
+  }
+
+  Status WriteUnbuffered(const char* data, size_t size) {
+    DWORD bytes_written;
+    if (!::WriteFile(handle_.get(), data, static_cast<DWORD>(size),
+                     &bytes_written, nullptr)) {
+      return Status::IOError(filename_,
+                             GetWindowsErrorMessage(::GetLastError()));
+    }
+    return Status::OK();
+  }
+
+  // buf_[0, pos_-1] contains data to be written to handle_.
+  char buf_[kWritableFileBufferSize];
+  size_t pos_;
+
+  ScopedHandle handle_;
+  const std::string filename_;
+};
+
+// Lock or unlock the entire file as specified by |lock|. Returns true
+// when successful, false upon failure. Caller should call ::GetLastError()
+// to determine cause of failure
+bool LockOrUnlock(HANDLE handle, bool lock) {
+  if (lock) {
+    return ::LockFile(handle,
+                      /*dwFileOffsetLow=*/0, /*dwFileOffsetHigh=*/0,
+                      /*nNumberOfBytesToLockLow=*/MAXDWORD,
+                      /*nNumberOfBytesToLockHigh=*/MAXDWORD);
+  } else {
+    return ::UnlockFile(handle,
+                        /*dwFileOffsetLow=*/0, /*dwFileOffsetHigh=*/0,
+                        /*nNumberOfBytesToLockLow=*/MAXDWORD,
+                        /*nNumberOfBytesToLockHigh=*/MAXDWORD);
+  }
+}
+
+class WindowsFileLock : public FileLock {
+ public:
+  WindowsFileLock(ScopedHandle handle, std::string filename)
+      : handle_(std::move(handle)), filename_(std::move(filename)) {}
+
+  const ScopedHandle& handle() const { return handle_; }
+  const std::string& filename() const { return filename_; }
+
+ private:
+  const ScopedHandle handle_;
+  const std::string filename_;
+};
+
+class WindowsEnv : public Env {
+ public:
+  WindowsEnv();
+  ~WindowsEnv() override {
+    static const char msg[] =
+        "WindowsEnv singleton destroyed. Unsupported behavior!\n";
+    std::fwrite(msg, 1, sizeof(msg), stderr);
+    std::abort();
+  }
+
+  Status NewSequentialFile(const std::string& filename,
+                           SequentialFile** result) override {
+    *result = nullptr;
+    DWORD desired_access = GENERIC_READ;
+    DWORD share_mode = FILE_SHARE_READ;
+    ScopedHandle handle = ::CreateFileA(
+        filename.c_str(), desired_access, share_mode,
+        /*lpSecurityAttributes=*/nullptr, OPEN_EXISTING, FILE_ATTRIBUTE_NORMAL,
+        /*hTemplateFile=*/nullptr);
+    if (!handle.is_valid()) {
+      return WindowsError(filename, ::GetLastError());
+    }
+
+    *result = new WindowsSequentialFile(filename, std::move(handle));
+    return Status::OK();
+  }
+
+  Status NewRandomAccessFile(const std::string& filename,
+                             RandomAccessFile** result) override {
+    *result = nullptr;
+    DWORD desired_access = GENERIC_READ;
+    DWORD share_mode = FILE_SHARE_READ;
+    ScopedHandle handle =
+        ::CreateFileA(filename.c_str(), desired_access, share_mode,
+                      /*lpSecurityAttributes=*/nullptr, OPEN_EXISTING,
+                      FILE_ATTRIBUTE_READONLY,
+                      /*hTemplateFile=*/nullptr);
+    if (!handle.is_valid()) {
+      return WindowsError(filename, ::GetLastError());
+    }
+    if (!mmap_limiter_.Acquire()) {
+      *result = new WindowsRandomAccessFile(filename, std::move(handle));
+      return Status::OK();
+    }
+
+    LARGE_INTEGER file_size;
+    Status status;
+    if (!::GetFileSizeEx(handle.get(), &file_size)) {
+      mmap_limiter_.Release();
+      return WindowsError(filename, ::GetLastError());
+    }
+
+    ScopedHandle mapping =
+        ::CreateFileMappingA(handle.get(),
+                             /*security attributes=*/nullptr, PAGE_READONLY,
+                             /*dwMaximumSizeHigh=*/0,
+                             /*dwMaximumSizeLow=*/0,
+                             /*lpName=*/nullptr);
+    if (mapping.is_valid()) {
+      void* mmap_base = ::MapViewOfFile(mapping.get(), FILE_MAP_READ,
+                                        /*dwFileOffsetHigh=*/0,
+                                        /*dwFileOffsetLow=*/0,
+                                        /*dwNumberOfBytesToMap=*/0);
+      if (mmap_base) {
+        *result = new WindowsMmapReadableFile(
+            filename, reinterpret_cast<char*>(mmap_base),
+            static_cast<size_t>(file_size.QuadPart), &mmap_limiter_);
+        return Status::OK();
+      }
+    }
+    mmap_limiter_.Release();
+    return WindowsError(filename, ::GetLastError());
+  }
+
+  Status NewWritableFile(const std::string& filename,
+                         WritableFile** result) override {
+    DWORD desired_access = GENERIC_WRITE;
+    DWORD share_mode = 0;  // Exclusive access.
+    ScopedHandle handle = ::CreateFileA(
+        filename.c_str(), desired_access, share_mode,
+        /*lpSecurityAttributes=*/nullptr, CREATE_ALWAYS, FILE_ATTRIBUTE_NORMAL,
+        /*hTemplateFile=*/nullptr);
+    if (!handle.is_valid()) {
+      *result = nullptr;
+      return WindowsError(filename, ::GetLastError());
+    }
+
+    *result = new WindowsWritableFile(filename, std::move(handle));
+    return Status::OK();
+  }
+
+  Status NewAppendableFile(const std::string& filename,
+                           WritableFile** result) override {
+    DWORD desired_access = FILE_APPEND_DATA;
+    DWORD share_mode = 0;  // Exclusive access.
+    ScopedHandle handle = ::CreateFileA(
+        filename.c_str(), desired_access, share_mode,
+        /*lpSecurityAttributes=*/nullptr, OPEN_ALWAYS, FILE_ATTRIBUTE_NORMAL,
+        /*hTemplateFile=*/nullptr);
+    if (!handle.is_valid()) {
+      *result = nullptr;
+      return WindowsError(filename, ::GetLastError());
+    }
+
+    *result = new WindowsWritableFile(filename, std::move(handle));
+    return Status::OK();
+  }
+
+  bool FileExists(const std::string& filename) override {
+    return GetFileAttributesA(filename.c_str()) != INVALID_FILE_ATTRIBUTES;
+  }
+
+  Status GetChildren(const std::string& directory_path,
+                     std::vector<std::string>* result) override {
+    const std::string find_pattern = directory_path + "\\*";
+    WIN32_FIND_DATAA find_data;
+    HANDLE dir_handle = ::FindFirstFileA(find_pattern.c_str(), &find_data);
+    if (dir_handle == INVALID_HANDLE_VALUE) {
+      DWORD last_error = ::GetLastError();
+      if (last_error == ERROR_FILE_NOT_FOUND) {
+        return Status::OK();
+      }
+      return WindowsError(directory_path, last_error);
+    }
+    do {
+      char base_name[_MAX_FNAME];
+      char ext[_MAX_EXT];
+
+      if (!_splitpath_s(find_data.cFileName, nullptr, 0, nullptr, 0, base_name,
+                        ARRAYSIZE(base_name), ext, ARRAYSIZE(ext))) {
+        result->emplace_back(std::string(base_name) + ext);
+      }
+    } while (::FindNextFileA(dir_handle, &find_data));
+    DWORD last_error = ::GetLastError();
+    ::FindClose(dir_handle);
+    if (last_error != ERROR_NO_MORE_FILES) {
+      return WindowsError(directory_path, last_error);
+    }
+    return Status::OK();
+  }
+
+  Status RemoveFile(const std::string& filename) override {
+    if (!::DeleteFileA(filename.c_str())) {
+      return WindowsError(filename, ::GetLastError());
+    }
+    return Status::OK();
+  }
+
+  Status CreateDir(const std::string& dirname) override {
+    if (!::CreateDirectoryA(dirname.c_str(), nullptr)) {
+      return WindowsError(dirname, ::GetLastError());
+    }
+    return Status::OK();
+  }
+
+  Status RemoveDir(const std::string& dirname) override {
+    if (!::RemoveDirectoryA(dirname.c_str())) {
+      return WindowsError(dirname, ::GetLastError());
+    }
+    return Status::OK();
+  }
+
+  Status GetFileSize(const std::string& filename, uint64_t* size) override {
+    WIN32_FILE_ATTRIBUTE_DATA file_attributes;
+    if (!::GetFileAttributesExA(filename.c_str(), GetFileExInfoStandard,
+                                &file_attributes)) {
+      return WindowsError(filename, ::GetLastError());
+    }
+    ULARGE_INTEGER file_size;
+    file_size.HighPart = file_attributes.nFileSizeHigh;
+    file_size.LowPart = file_attributes.nFileSizeLow;
+    *size = file_size.QuadPart;
+    return Status::OK();
+  }
+
+  Status RenameFile(const std::string& from, const std::string& to) override {
+    // Try a simple move first. It will only succeed when |to| doesn't already
+    // exist.
+    if (::MoveFileA(from.c_str(), to.c_str())) {
+      return Status::OK();
+    }
+    DWORD move_error = ::GetLastError();
+
+    // Try the full-blown replace if the move fails, as ReplaceFile will only
+    // succeed when |to| does exist. When writing to a network share, we may not
+    // be able to change the ACLs. Ignore ACL errors then
+    // (REPLACEFILE_IGNORE_MERGE_ERRORS).
+    if (::ReplaceFileA(to.c_str(), from.c_str(), /*lpBackupFileName=*/nullptr,
+                       REPLACEFILE_IGNORE_MERGE_ERRORS,
+                       /*lpExclude=*/nullptr, /*lpReserved=*/nullptr)) {
+      return Status::OK();
+    }
+    DWORD replace_error = ::GetLastError();
+    // In the case of FILE_ERROR_NOT_FOUND from ReplaceFile, it is likely that
+    // |to| does not exist. In this case, the more relevant error comes from the
+    // call to MoveFile.
+    if (replace_error == ERROR_FILE_NOT_FOUND ||
+        replace_error == ERROR_PATH_NOT_FOUND) {
+      return WindowsError(from, move_error);
+    } else {
+      return WindowsError(from, replace_error);
+    }
+  }
+
+  Status LockFile(const std::string& filename, FileLock** lock) override {
+    *lock = nullptr;
+    Status result;
+    ScopedHandle handle = ::CreateFileA(
+        filename.c_str(), GENERIC_READ | GENERIC_WRITE, FILE_SHARE_READ,
+        /*lpSecurityAttributes=*/nullptr, OPEN_ALWAYS, FILE_ATTRIBUTE_NORMAL,
+        nullptr);
+    if (!handle.is_valid()) {
+      result = WindowsError(filename, ::GetLastError());
+    } else if (!LockOrUnlock(handle.get(), true)) {
+      result = WindowsError("lock " + filename, ::GetLastError());
+    } else {
+      *lock = new WindowsFileLock(std::move(handle), filename);
+    }
+    return result;
+  }
+
+  Status UnlockFile(FileLock* lock) override {
+    WindowsFileLock* windows_file_lock =
+        reinterpret_cast<WindowsFileLock*>(lock);
+    if (!LockOrUnlock(windows_file_lock->handle().get(), false)) {
+      return WindowsError("unlock " + windows_file_lock->filename(),
+                          ::GetLastError());
+    }
+    delete windows_file_lock;
+    return Status::OK();
+  }
+
+  void Schedule(void (*background_work_function)(void* background_work_arg),
+                void* background_work_arg) override;
+
+  void StartThread(void (*thread_main)(void* thread_main_arg),
+                   void* thread_main_arg) override {
+    std::thread new_thread(thread_main, thread_main_arg);
+    new_thread.detach();
+  }
+
+  Status GetTestDirectory(std::string* result) override {
+    const char* env = getenv("TEST_TMPDIR");
+    if (env && env[0] != '\0') {
+      *result = env;
+      return Status::OK();
+    }
+
+    char tmp_path[MAX_PATH];
+    if (!GetTempPathA(ARRAYSIZE(tmp_path), tmp_path)) {
+      return WindowsError("GetTempPath", ::GetLastError());
+    }
+    std::stringstream ss;
+    ss << tmp_path << "leveldbtest-" << std::this_thread::get_id();
+    *result = ss.str();
+
+    // Directory may already exist
+    CreateDir(*result);
+    return Status::OK();
+  }
+
+  Status NewLogger(const std::string& filename, Logger** result) override {
+    std::FILE* fp = std::fopen(filename.c_str(), "w");
+    if (fp == nullptr) {
+      *result = nullptr;
+      return WindowsError(filename, ::GetLastError());
+    } else {
+      *result = new WindowsLogger(fp);
+      return Status::OK();
+    }
+  }
+
+  uint64_t NowMicros() override {
+    // GetSystemTimeAsFileTime typically has a resolution of 10-20 msec.
+    // TODO(cmumford): Switch to GetSystemTimePreciseAsFileTime which is
+    // available in Windows 8 and later.
+    FILETIME ft;
+    ::GetSystemTimeAsFileTime(&ft);
+    // Each tick represents a 100-nanosecond intervals since January 1, 1601
+    // (UTC).
+    uint64_t num_ticks =
+        (static_cast<uint64_t>(ft.dwHighDateTime) << 32) + ft.dwLowDateTime;
+    return num_ticks / 10;
+  }
+
+  void SleepForMicroseconds(int micros) override {
+    std::this_thread::sleep_for(std::chrono::microseconds(micros));
+  }
+
+ private:
+  void BackgroundThreadMain();
+
+  static void BackgroundThreadEntryPoint(WindowsEnv* env) {
+    env->BackgroundThreadMain();
+  }
+
+  // Stores the work item data in a Schedule() call.
+  //
+  // Instances are constructed on the thread calling Schedule() and used on the
+  // background thread.
+  //
+  // This structure is thread-safe beacuse it is immutable.
+  struct BackgroundWorkItem {
+    explicit BackgroundWorkItem(void (*function)(void* arg), void* arg)
+        : function(function), arg(arg) {}
+
+    void (*const function)(void*);
+    void* const arg;
+  };
+
+  port::Mutex background_work_mutex_;
+  port::CondVar background_work_cv_ GUARDED_BY(background_work_mutex_);
+  bool started_background_thread_ GUARDED_BY(background_work_mutex_);
+
+  std::queue<BackgroundWorkItem> background_work_queue_
+      GUARDED_BY(background_work_mutex_);
+
+  Limiter mmap_limiter_;  // Thread-safe.
+};
+
+// Return the maximum number of concurrent mmaps.
+int MaxMmaps() { return g_mmap_limit; }
+
+WindowsEnv::WindowsEnv()
+    : background_work_cv_(&background_work_mutex_),
+      started_background_thread_(false),
+      mmap_limiter_(MaxMmaps()) {}
+
+void WindowsEnv::Schedule(
+    void (*background_work_function)(void* background_work_arg),
+    void* background_work_arg) {
+  background_work_mutex_.Lock();
+
+  // Start the background thread, if we haven't done so already.
+  if (!started_background_thread_) {
+    started_background_thread_ = true;
+    std::thread background_thread(WindowsEnv::BackgroundThreadEntryPoint, this);
+    background_thread.detach();
+  }
+
+  // If the queue is empty, the background thread may be waiting for work.
+  if (background_work_queue_.empty()) {
+    background_work_cv_.Signal();
+  }
+
+  background_work_queue_.emplace(background_work_function, background_work_arg);
+  background_work_mutex_.Unlock();
+}
+
+void WindowsEnv::BackgroundThreadMain() {
+  while (true) {
+    background_work_mutex_.Lock();
+
+    // Wait until there is work to be done.
+    while (background_work_queue_.empty()) {
+      background_work_cv_.Wait();
+    }
+
+    assert(!background_work_queue_.empty());
+    auto background_work_function = background_work_queue_.front().function;
+    void* background_work_arg = background_work_queue_.front().arg;
+    background_work_queue_.pop();
+
+    background_work_mutex_.Unlock();
+    background_work_function(background_work_arg);
+  }
+}
+
+// Wraps an Env instance whose destructor is never created.
+//
+// Intended usage:
+//   using PlatformSingletonEnv = SingletonEnv<PlatformEnv>;
+//   void ConfigurePosixEnv(int param) {
+//     PlatformSingletonEnv::AssertEnvNotInitialized();
+//     // set global configuration flags.
+//   }
+//   Env* Env::Default() {
+//     static PlatformSingletonEnv default_env;
+//     return default_env.env();
+//   }
+template <typename EnvType>
+class SingletonEnv {
+ public:
+  SingletonEnv() {
+#if !defined(NDEBUG)
+    env_initialized_.store(true, std::memory_order::memory_order_relaxed);
+#endif  // !defined(NDEBUG)
+    static_assert(sizeof(env_storage_) >= sizeof(EnvType),
+                  "env_storage_ will not fit the Env");
+    static_assert(alignof(decltype(env_storage_)) >= alignof(EnvType),
+                  "env_storage_ does not meet the Env's alignment needs");
+    new (&env_storage_) EnvType();
+  }
+  ~SingletonEnv() = default;
+
+  SingletonEnv(const SingletonEnv&) = delete;
+  SingletonEnv& operator=(const SingletonEnv&) = delete;
+
+  Env* env() { return reinterpret_cast<Env*>(&env_storage_); }
+
+  static void AssertEnvNotInitialized() {
+#if !defined(NDEBUG)
+    assert(!env_initialized_.load(std::memory_order::memory_order_relaxed));
+#endif  // !defined(NDEBUG)
+  }
+
+ private:
+  typename std::aligned_storage<sizeof(EnvType), alignof(EnvType)>::type
+      env_storage_;
+#if !defined(NDEBUG)
+  static std::atomic<bool> env_initialized_;
+#endif  // !defined(NDEBUG)
+};
+
+#if !defined(NDEBUG)
+template <typename EnvType>
+std::atomic<bool> SingletonEnv<EnvType>::env_initialized_;
+#endif  // !defined(NDEBUG)
+
+using WindowsDefaultEnv = SingletonEnv<WindowsEnv>;
+
+}  // namespace
+
+void EnvWindowsTestHelper::SetReadOnlyMMapLimit(int limit) {
+  WindowsDefaultEnv::AssertEnvNotInitialized();
+  g_mmap_limit = limit;
+}
+
+Env* Env::Default() {
+  static WindowsDefaultEnv env_container;
+  return env_container.env();
+}
+
+}  // namespace leveldb

lib/leveldb/util/env_windows_test.cc

@@ -0,0 +1,65 @@
+// Copyright (c) 2018 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "gtest/gtest.h"
+#include "leveldb/env.h"
+#include "port/port.h"
+#include "util/env_windows_test_helper.h"
+#include "util/testutil.h"
+
+namespace leveldb {
+
+static const int kMMapLimit = 4;
+
+class EnvWindowsTest : public testing::Test {
+ public:
+  static void SetFileLimits(int mmap_limit) {
+    EnvWindowsTestHelper::SetReadOnlyMMapLimit(mmap_limit);
+  }
+
+  EnvWindowsTest() : env_(Env::Default()) {}
+
+  Env* env_;
+};
+
+TEST_F(EnvWindowsTest, TestOpenOnRead) {
+  // Write some test data to a single file that will be opened |n| times.
+  std::string test_dir;
+  ASSERT_LEVELDB_OK(env_->GetTestDirectory(&test_dir));
+  std::string test_file = test_dir + "/open_on_read.txt";
+
+  FILE* f = std::fopen(test_file.c_str(), "w");
+  ASSERT_TRUE(f != nullptr);
+  const char kFileData[] = "abcdefghijklmnopqrstuvwxyz";
+  fputs(kFileData, f);
+  std::fclose(f);
+
+  // Open test file some number above the sum of the two limits to force
+  // leveldb::WindowsEnv to switch from mapping the file into memory
+  // to basic file reading.
+  const int kNumFiles = kMMapLimit + 5;
+  leveldb::RandomAccessFile* files[kNumFiles] = {0};
+  for (int i = 0; i < kNumFiles; i++) {
+    ASSERT_LEVELDB_OK(env_->NewRandomAccessFile(test_file, &files[i]));
+  }
+  char scratch;
+  Slice read_result;
+  for (int i = 0; i < kNumFiles; i++) {
+    ASSERT_LEVELDB_OK(files[i]->Read(i, 1, &read_result, &scratch));
+    ASSERT_EQ(kFileData[i], read_result[0]);
+  }
+  for (int i = 0; i < kNumFiles; i++) {
+    delete files[i];
+  }
+  ASSERT_LEVELDB_OK(env_->RemoveFile(test_file));
+}
+
+}  // namespace leveldb
+
+int main(int argc, char** argv) {
+  // All tests currently run with the same read-only file limits.
+  leveldb::EnvWindowsTest::SetFileLimits(leveldb::kMMapLimit);
+  testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}

lib/leveldb/util/env_windows_test_helper.h

@@ -0,0 +1,25 @@
+// Copyright 2018 (c) The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#ifndef STORAGE_LEVELDB_UTIL_ENV_WINDOWS_TEST_HELPER_H_
+#define STORAGE_LEVELDB_UTIL_ENV_WINDOWS_TEST_HELPER_H_
+
+namespace leveldb {
+
+class EnvWindowsTest;
+
+// A helper for the Windows Env to facilitate testing.
+class EnvWindowsTestHelper {
+ private:
+  friend class CorruptionTest;
+  friend class EnvWindowsTest;
+
+  // Set the maximum number of read-only files that will be mapped via mmap.
+  // Must be called before creating an Env.
+  static void SetReadOnlyMMapLimit(int limit);
+};
+
+}  // namespace leveldb
+
+#endif  // STORAGE_LEVELDB_UTIL_ENV_WINDOWS_TEST_HELPER_H_

lib/leveldb/util/filter_policy.cc

@@ -0,0 +1,11 @@
+// Copyright (c) 2012 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "leveldb/filter_policy.h"
+
+namespace leveldb {
+
+FilterPolicy::~FilterPolicy() {}
+
+}  // namespace leveldb

lib/leveldb/util/hash.cc

@@ -0,0 +1,55 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "util/hash.h"
+
+#include <cstring>
+
+#include "util/coding.h"
+
+// The FALLTHROUGH_INTENDED macro can be used to annotate implicit fall-through
+// between switch labels. The real definition should be provided externally.
+// This one is a fallback version for unsupported compilers.
+#ifndef FALLTHROUGH_INTENDED
+#define FALLTHROUGH_INTENDED \
+  do {                       \
+  } while (0)
+#endif
+
+namespace leveldb {
+
+uint32_t Hash(const char* data, size_t n, uint32_t seed) {
+  // Similar to murmur hash
+  const uint32_t m = 0xc6a4a793;
+  const uint32_t r = 24;
+  const char* limit = data + n;
+  uint32_t h = seed ^ (n * m);
+
+  // Pick up four bytes at a time
+  while (data + 4 <= limit) {
+    uint32_t w = DecodeFixed32(data);
+    data += 4;
+    h += w;
+    h *= m;
+    h ^= (h >> 16);
+  }
+
+  // Pick up remaining bytes
+  switch (limit - data) {
+    case 3:
+      h += static_cast<uint8_t>(data[2]) << 16;
+      FALLTHROUGH_INTENDED;
+    case 2:
+      h += static_cast<uint8_t>(data[1]) << 8;
+      FALLTHROUGH_INTENDED;
+    case 1:
+      h += static_cast<uint8_t>(data[0]);
+      h *= m;
+      h ^= (h >> r);
+      break;
+  }
+  return h;
+}
+
+}  // namespace leveldb

lib/leveldb/util/hash.h

@@ -0,0 +1,19 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+//
+// Simple hash function used for internal data structures
+
+#ifndef STORAGE_LEVELDB_UTIL_HASH_H_
+#define STORAGE_LEVELDB_UTIL_HASH_H_
+
+#include <cstddef>
+#include <cstdint>
+
+namespace leveldb {
+
+uint32_t Hash(const char* data, size_t n, uint32_t seed);
+
+}  // namespace leveldb
+
+#endif  // STORAGE_LEVELDB_UTIL_HASH_H_

lib/leveldb/util/hash_test.cc

@@ -0,0 +1,46 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "util/hash.h"
+
+#include "gtest/gtest.h"
+
+namespace leveldb {
+
+TEST(HASH, SignedUnsignedIssue) {
+  const uint8_t data1[1] = {0x62};
+  const uint8_t data2[2] = {0xc3, 0x97};
+  const uint8_t data3[3] = {0xe2, 0x99, 0xa5};
+  const uint8_t data4[4] = {0xe1, 0x80, 0xb9, 0x32};
+  const uint8_t data5[48] = {
+      0x01, 0xc0, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+      0x00, 0x00, 0x00, 0x00, 0x14, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00,
+      0x00, 0x00, 0x00, 0x14, 0x00, 0x00, 0x00, 0x18, 0x28, 0x00, 0x00, 0x00,
+      0x00, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+  };
+
+  ASSERT_EQ(Hash(0, 0, 0xbc9f1d34), 0xbc9f1d34);
+  ASSERT_EQ(
+      Hash(reinterpret_cast<const char*>(data1), sizeof(data1), 0xbc9f1d34),
+      0xef1345c4);
+  ASSERT_EQ(
+      Hash(reinterpret_cast<const char*>(data2), sizeof(data2), 0xbc9f1d34),
+      0x5b663814);
+  ASSERT_EQ(
+      Hash(reinterpret_cast<const char*>(data3), sizeof(data3), 0xbc9f1d34),
+      0x323c078f);
+  ASSERT_EQ(
+      Hash(reinterpret_cast<const char*>(data4), sizeof(data4), 0xbc9f1d34),
+      0xed21633a);
+  ASSERT_EQ(
+      Hash(reinterpret_cast<const char*>(data5), sizeof(data5), 0x12345678),
+      0xf333dabb);
+}
+
+}  // namespace leveldb
+
+int main(int argc, char** argv) {
+  testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}

lib/leveldb/util/histogram.cc

@@ -0,0 +1,272 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "util/histogram.h"
+
+#include <cmath>
+#include <cstdio>
+
+#include "port/port.h"
+
+namespace leveldb {
+
+const double Histogram::kBucketLimit[kNumBuckets] = {
+    1,
+    2,
+    3,
+    4,
+    5,
+    6,
+    7,
+    8,
+    9,
+    10,
+    12,
+    14,
+    16,
+    18,
+    20,
+    25,
+    30,
+    35,
+    40,
+    45,
+    50,
+    60,
+    70,
+    80,
+    90,
+    100,
+    120,
+    140,
+    160,
+    180,
+    200,
+    250,
+    300,
+    350,
+    400,
+    450,
+    500,
+    600,
+    700,
+    800,
+    900,
+    1000,
+    1200,
+    1400,
+    1600,
+    1800,
+    2000,
+    2500,
+    3000,
+    3500,
+    4000,
+    4500,
+    5000,
+    6000,
+    7000,
+    8000,
+    9000,
+    10000,
+    12000,
+    14000,
+    16000,
+    18000,
+    20000,
+    25000,
+    30000,
+    35000,
+    40000,
+    45000,
+    50000,
+    60000,
+    70000,
+    80000,
+    90000,
+    100000,
+    120000,
+    140000,
+    160000,
+    180000,
+    200000,
+    250000,
+    300000,
+    350000,
+    400000,
+    450000,
+    500000,
+    600000,
+    700000,
+    800000,
+    900000,
+    1000000,
+    1200000,
+    1400000,
+    1600000,
+    1800000,
+    2000000,
+    2500000,
+    3000000,
+    3500000,
+    4000000,
+    4500000,
+    5000000,
+    6000000,
+    7000000,
+    8000000,
+    9000000,
+    10000000,
+    12000000,
+    14000000,
+    16000000,
+    18000000,
+    20000000,
+    25000000,
+    30000000,
+    35000000,
+    40000000,
+    45000000,
+    50000000,
+    60000000,
+    70000000,
+    80000000,
+    90000000,
+    100000000,
+    120000000,
+    140000000,
+    160000000,
+    180000000,
+    200000000,
+    250000000,
+    300000000,
+    350000000,
+    400000000,
+    450000000,
+    500000000,
+    600000000,
+    700000000,
+    800000000,
+    900000000,
+    1000000000,
+    1200000000,
+    1400000000,
+    1600000000,
+    1800000000,
+    2000000000,
+    2500000000.0,
+    3000000000.0,
+    3500000000.0,
+    4000000000.0,
+    4500000000.0,
+    5000000000.0,
+    6000000000.0,
+    7000000000.0,
+    8000000000.0,
+    9000000000.0,
+    1e200,
+};
+
+void Histogram::Clear() {
+  min_ = kBucketLimit[kNumBuckets - 1];
+  max_ = 0;
+  num_ = 0;
+  sum_ = 0;
+  sum_squares_ = 0;
+  for (int i = 0; i < kNumBuckets; i++) {
+    buckets_[i] = 0;
+  }
+}
+
+void Histogram::Add(double value) {
+  // Linear search is fast enough for our usage in db_bench
+  int b = 0;
+  while (b < kNumBuckets - 1 && kBucketLimit[b] <= value) {
+    b++;
+  }
+  buckets_[b] += 1.0;
+  if (min_ > value) min_ = value;
+  if (max_ < value) max_ = value;
+  num_++;
+  sum_ += value;
+  sum_squares_ += (value * value);
+}
+
+void Histogram::Merge(const Histogram& other) {
+  if (other.min_ < min_) min_ = other.min_;
+  if (other.max_ > max_) max_ = other.max_;
+  num_ += other.num_;
+  sum_ += other.sum_;
+  sum_squares_ += other.sum_squares_;
+  for (int b = 0; b < kNumBuckets; b++) {
+    buckets_[b] += other.buckets_[b];
+  }
+}
+
+double Histogram::Median() const { return Percentile(50.0); }
+
+double Histogram::Percentile(double p) const {
+  double threshold = num_ * (p / 100.0);
+  double sum = 0;
+  for (int b = 0; b < kNumBuckets; b++) {
+    sum += buckets_[b];
+    if (sum >= threshold) {
+      // Scale linearly within this bucket
+      double left_point = (b == 0) ? 0 : kBucketLimit[b - 1];
+      double right_point = kBucketLimit[b];
+      double left_sum = sum - buckets_[b];
+      double right_sum = sum;
+      double pos = (threshold - left_sum) / (right_sum - left_sum);
+      double r = left_point + (right_point - left_point) * pos;
+      if (r < min_) r = min_;
+      if (r > max_) r = max_;
+      return r;
+    }
+  }
+  return max_;
+}
+
+double Histogram::Average() const {
+  if (num_ == 0.0) return 0;
+  return sum_ / num_;
+}
+
+double Histogram::StandardDeviation() const {
+  if (num_ == 0.0) return 0;
+  double variance = (sum_squares_ * num_ - sum_ * sum_) / (num_ * num_);
+  return sqrt(variance);
+}
+
+std::string Histogram::ToString() const {
+  std::string r;
+  char buf[200];
+  std::snprintf(buf, sizeof(buf), "Count: %.0f  Average: %.4f  StdDev: %.2f\n",
+                num_, Average(), StandardDeviation());
+  r.append(buf);
+  std::snprintf(buf, sizeof(buf), "Min: %.4f  Median: %.4f  Max: %.4f\n",
+                (num_ == 0.0 ? 0.0 : min_), Median(), max_);
+  r.append(buf);
+  r.append("------------------------------------------------------\n");
+  const double mult = 100.0 / num_;
+  double sum = 0;
+  for (int b = 0; b < kNumBuckets; b++) {
+    if (buckets_[b] <= 0.0) continue;
+    sum += buckets_[b];
+    std::snprintf(buf, sizeof(buf), "[ %7.0f, %7.0f ) %7.0f %7.3f%% %7.3f%% ",
+                  ((b == 0) ? 0.0 : kBucketLimit[b - 1]),  // left
+                  kBucketLimit[b],                         // right
+                  buckets_[b],                             // count
+                  mult * buckets_[b],                      // percentage
+                  mult * sum);  // cumulative percentage
+    r.append(buf);
+
+    // Add hash marks based on percentage; 20 marks for 100%.
+    int marks = static_cast<int>(20 * (buckets_[b] / num_) + 0.5);
+    r.append(marks, '#');
+    r.push_back('\n');
+  }
+  return r;
+}
+
+}  // namespace leveldb

lib/leveldb/util/histogram.h

@@ -0,0 +1,44 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#ifndef STORAGE_LEVELDB_UTIL_HISTOGRAM_H_
+#define STORAGE_LEVELDB_UTIL_HISTOGRAM_H_
+
+#include <string>
+
+namespace leveldb {
+
+class Histogram {
+ public:
+  Histogram() {}
+  ~Histogram() {}
+
+  void Clear();
+  void Add(double value);
+  void Merge(const Histogram& other);
+
+  std::string ToString() const;
+
+ private:
+  enum { kNumBuckets = 154 };
+
+  double Median() const;
+  double Percentile(double p) const;
+  double Average() const;
+  double StandardDeviation() const;
+
+  static const double kBucketLimit[kNumBuckets];
+
+  double min_;
+  double max_;
+  double num_;
+  double sum_;
+  double sum_squares_;
+
+  double buckets_[kNumBuckets];
+};
+
+}  // namespace leveldb
+
+#endif  // STORAGE_LEVELDB_UTIL_HISTOGRAM_H_

lib/leveldb/util/logging.cc

@@ -0,0 +1,82 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "util/logging.h"
+
+#include <cstdarg>
+#include <cstdio>
+#include <cstdlib>
+#include <limits>
+
+#include "leveldb/env.h"
+#include "leveldb/slice.h"
+
+namespace leveldb {
+
+void AppendNumberTo(std::string* str, uint64_t num) {
+  char buf[30];
+  std::snprintf(buf, sizeof(buf), "%llu", static_cast<unsigned long long>(num));
+  str->append(buf);
+}
+
+void AppendEscapedStringTo(std::string* str, const Slice& value) {
+  for (size_t i = 0; i < value.size(); i++) {
+    char c = value[i];
+    if (c >= ' ' && c <= '~') {
+      str->push_back(c);
+    } else {
+      char buf[10];
+      std::snprintf(buf, sizeof(buf), "\\x%02x",
+                    static_cast<unsigned int>(c) & 0xff);
+      str->append(buf);
+    }
+  }
+}
+
+std::string NumberToString(uint64_t num) {
+  std::string r;
+  AppendNumberTo(&r, num);
+  return r;
+}
+
+std::string EscapeString(const Slice& value) {
+  std::string r;
+  AppendEscapedStringTo(&r, value);
+  return r;
+}
+
+bool ConsumeDecimalNumber(Slice* in, uint64_t* val) {
+  // Constants that will be optimized away.
+  constexpr const uint64_t kMaxUint64 = std::numeric_limits<uint64_t>::max();
+  constexpr const char kLastDigitOfMaxUint64 =
+      '0' + static_cast<char>(kMaxUint64 % 10);
+
+  uint64_t value = 0;
+
+  // reinterpret_cast-ing from char* to uint8_t* to avoid signedness.
+  const uint8_t* start = reinterpret_cast<const uint8_t*>(in->data());
+
+  const uint8_t* end = start + in->size();
+  const uint8_t* current = start;
+  for (; current != end; ++current) {
+    const uint8_t ch = *current;
+    if (ch < '0' || ch > '9') break;
+
+    // Overflow check.
+    // kMaxUint64 / 10 is also constant and will be optimized away.
+    if (value > kMaxUint64 / 10 ||
+        (value == kMaxUint64 / 10 && ch > kLastDigitOfMaxUint64)) {
+      return false;
+    }
+
+    value = (value * 10) + (ch - '0');
+  }
+
+  *val = value;
+  const size_t digits_consumed = current - start;
+  in->remove_prefix(digits_consumed);
+  return digits_consumed != 0;
+}
+
+}  // namespace leveldb

lib/leveldb/util/logging.h

@@ -0,0 +1,44 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+//
+// Must not be included from any .h files to avoid polluting the namespace
+// with macros.
+
+#ifndef STORAGE_LEVELDB_UTIL_LOGGING_H_
+#define STORAGE_LEVELDB_UTIL_LOGGING_H_
+
+#include <cstdint>
+#include <cstdio>
+#include <string>
+
+#include "port/port.h"
+
+namespace leveldb {
+
+class Slice;
+class WritableFile;
+
+// Append a human-readable printout of "num" to *str
+void AppendNumberTo(std::string* str, uint64_t num);
+
+// Append a human-readable printout of "value" to *str.
+// Escapes any non-printable characters found in "value".
+void AppendEscapedStringTo(std::string* str, const Slice& value);
+
+// Return a human-readable printout of "num"
+std::string NumberToString(uint64_t num);
+
+// Return a human-readable version of "value".
+// Escapes any non-printable characters found in "value".
+std::string EscapeString(const Slice& value);
+
+// Parse a human-readable number from "*in" into *value.  On success,
+// advances "*in" past the consumed number and sets "*val" to the
+// numeric value.  Otherwise, returns false and leaves *in in an
+// unspecified state.
+bool ConsumeDecimalNumber(Slice* in, uint64_t* val);
+
+}  // namespace leveldb
+
+#endif  // STORAGE_LEVELDB_UTIL_LOGGING_H_

lib/leveldb/util/logging_test.cc

@@ -0,0 +1,145 @@
+// Copyright (c) 2018 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "util/logging.h"
+
+#include <limits>
+#include <string>
+
+#include "gtest/gtest.h"
+#include "leveldb/slice.h"
+
+namespace leveldb {
+
+TEST(Logging, NumberToString) {
+  ASSERT_EQ("0", NumberToString(0));
+  ASSERT_EQ("1", NumberToString(1));
+  ASSERT_EQ("9", NumberToString(9));
+
+  ASSERT_EQ("10", NumberToString(10));
+  ASSERT_EQ("11", NumberToString(11));
+  ASSERT_EQ("19", NumberToString(19));
+  ASSERT_EQ("99", NumberToString(99));
+
+  ASSERT_EQ("100", NumberToString(100));
+  ASSERT_EQ("109", NumberToString(109));
+  ASSERT_EQ("190", NumberToString(190));
+  ASSERT_EQ("123", NumberToString(123));
+  ASSERT_EQ("12345678", NumberToString(12345678));
+
+  static_assert(std::numeric_limits<uint64_t>::max() == 18446744073709551615U,
+                "Test consistency check");
+  ASSERT_EQ("18446744073709551000", NumberToString(18446744073709551000U));
+  ASSERT_EQ("18446744073709551600", NumberToString(18446744073709551600U));
+  ASSERT_EQ("18446744073709551610", NumberToString(18446744073709551610U));
+  ASSERT_EQ("18446744073709551614", NumberToString(18446744073709551614U));
+  ASSERT_EQ("18446744073709551615", NumberToString(18446744073709551615U));
+}
+
+void ConsumeDecimalNumberRoundtripTest(uint64_t number,
+                                       const std::string& padding = "") {
+  std::string decimal_number = NumberToString(number);
+  std::string input_string = decimal_number + padding;
+  Slice input(input_string);
+  Slice output = input;
+  uint64_t result;
+  ASSERT_TRUE(ConsumeDecimalNumber(&output, &result));
+  ASSERT_EQ(number, result);
+  ASSERT_EQ(decimal_number.size(), output.data() - input.data());
+  ASSERT_EQ(padding.size(), output.size());
+}
+
+TEST(Logging, ConsumeDecimalNumberRoundtrip) {
+  ConsumeDecimalNumberRoundtripTest(0);
+  ConsumeDecimalNumberRoundtripTest(1);
+  ConsumeDecimalNumberRoundtripTest(9);
+
+  ConsumeDecimalNumberRoundtripTest(10);
+  ConsumeDecimalNumberRoundtripTest(11);
+  ConsumeDecimalNumberRoundtripTest(19);
+  ConsumeDecimalNumberRoundtripTest(99);
+
+  ConsumeDecimalNumberRoundtripTest(100);
+  ConsumeDecimalNumberRoundtripTest(109);
+  ConsumeDecimalNumberRoundtripTest(190);
+  ConsumeDecimalNumberRoundtripTest(123);
+  ASSERT_EQ("12345678", NumberToString(12345678));
+
+  for (uint64_t i = 0; i < 100; ++i) {
+    uint64_t large_number = std::numeric_limits<uint64_t>::max() - i;
+    ConsumeDecimalNumberRoundtripTest(large_number);
+  }
+}
+
+TEST(Logging, ConsumeDecimalNumberRoundtripWithPadding) {
+  ConsumeDecimalNumberRoundtripTest(0, " ");
+  ConsumeDecimalNumberRoundtripTest(1, "abc");
+  ConsumeDecimalNumberRoundtripTest(9, "x");
+
+  ConsumeDecimalNumberRoundtripTest(10, "_");
+  ConsumeDecimalNumberRoundtripTest(11, std::string("\0\0\0", 3));
+  ConsumeDecimalNumberRoundtripTest(19, "abc");
+  ConsumeDecimalNumberRoundtripTest(99, "padding");
+
+  ConsumeDecimalNumberRoundtripTest(100, " ");
+
+  for (uint64_t i = 0; i < 100; ++i) {
+    uint64_t large_number = std::numeric_limits<uint64_t>::max() - i;
+    ConsumeDecimalNumberRoundtripTest(large_number, "pad");
+  }
+}
+
+void ConsumeDecimalNumberOverflowTest(const std::string& input_string) {
+  Slice input(input_string);
+  Slice output = input;
+  uint64_t result;
+  ASSERT_EQ(false, ConsumeDecimalNumber(&output, &result));
+}
+
+TEST(Logging, ConsumeDecimalNumberOverflow) {
+  static_assert(std::numeric_limits<uint64_t>::max() == 18446744073709551615U,
+                "Test consistency check");
+  ConsumeDecimalNumberOverflowTest("18446744073709551616");
+  ConsumeDecimalNumberOverflowTest("18446744073709551617");
+  ConsumeDecimalNumberOverflowTest("18446744073709551618");
+  ConsumeDecimalNumberOverflowTest("18446744073709551619");
+  ConsumeDecimalNumberOverflowTest("18446744073709551620");
+  ConsumeDecimalNumberOverflowTest("18446744073709551621");
+  ConsumeDecimalNumberOverflowTest("18446744073709551622");
+  ConsumeDecimalNumberOverflowTest("18446744073709551623");
+  ConsumeDecimalNumberOverflowTest("18446744073709551624");
+  ConsumeDecimalNumberOverflowTest("18446744073709551625");
+  ConsumeDecimalNumberOverflowTest("18446744073709551626");
+
+  ConsumeDecimalNumberOverflowTest("18446744073709551700");
+
+  ConsumeDecimalNumberOverflowTest("99999999999999999999");
+}
+
+void ConsumeDecimalNumberNoDigitsTest(const std::string& input_string) {
+  Slice input(input_string);
+  Slice output = input;
+  uint64_t result;
+  ASSERT_EQ(false, ConsumeDecimalNumber(&output, &result));
+  ASSERT_EQ(input.data(), output.data());
+  ASSERT_EQ(input.size(), output.size());
+}
+
+TEST(Logging, ConsumeDecimalNumberNoDigits) {
+  ConsumeDecimalNumberNoDigitsTest("");
+  ConsumeDecimalNumberNoDigitsTest(" ");
+  ConsumeDecimalNumberNoDigitsTest("a");
+  ConsumeDecimalNumberNoDigitsTest(" 123");
+  ConsumeDecimalNumberNoDigitsTest("a123");
+  ConsumeDecimalNumberNoDigitsTest(std::string("\000123", 4));
+  ConsumeDecimalNumberNoDigitsTest(std::string("\177123", 4));
+  ConsumeDecimalNumberNoDigitsTest(std::string("\377123", 4));
+}
+
+}  // namespace leveldb
+
+int main(int argc, char** argv) {
+  testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}

lib/leveldb/util/mutexlock.h

@@ -0,0 +1,39 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#ifndef STORAGE_LEVELDB_UTIL_MUTEXLOCK_H_
+#define STORAGE_LEVELDB_UTIL_MUTEXLOCK_H_
+
+#include "port/port.h"
+#include "port/thread_annotations.h"
+
+namespace leveldb {
+
+// Helper class that locks a mutex on construction and unlocks the mutex when
+// the destructor of the MutexLock object is invoked.
+//
+// Typical usage:
+//
+//   void MyClass::MyMethod() {
+//     MutexLock l(&mu_);       // mu_ is an instance variable
+//     ... some complex code, possibly with multiple return paths ...
+//   }
+
+class SCOPED_LOCKABLE MutexLock {
+ public:
+  explicit MutexLock(port::Mutex* mu) EXCLUSIVE_LOCK_FUNCTION(mu) : mu_(mu) {
+    this->mu_->Lock();
+  }
+  ~MutexLock() UNLOCK_FUNCTION() { this->mu_->Unlock(); }
+
+  MutexLock(const MutexLock&) = delete;
+  MutexLock& operator=(const MutexLock&) = delete;
+
+ private:
+  port::Mutex* const mu_;
+};
+
+}  // namespace leveldb
+
+#endif  // STORAGE_LEVELDB_UTIL_MUTEXLOCK_H_

lib/leveldb/util/no_destructor.h

@@ -0,0 +1,46 @@
+// Copyright (c) 2018 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#ifndef STORAGE_LEVELDB_UTIL_NO_DESTRUCTOR_H_
+#define STORAGE_LEVELDB_UTIL_NO_DESTRUCTOR_H_
+
+#include <type_traits>
+#include <utility>
+
+namespace leveldb {
+
+// Wraps an instance whose destructor is never called.
+//
+// This is intended for use with function-level static variables.
+template <typename InstanceType>
+class NoDestructor {
+ public:
+  template <typename... ConstructorArgTypes>
+  explicit NoDestructor(ConstructorArgTypes&&... constructor_args) {
+    static_assert(sizeof(instance_storage_) >= sizeof(InstanceType),
+                  "instance_storage_ is not large enough to hold the instance");
+    static_assert(
+        alignof(decltype(instance_storage_)) >= alignof(InstanceType),
+        "instance_storage_ does not meet the instance's alignment requirement");
+    new (&instance_storage_)
+        InstanceType(std::forward<ConstructorArgTypes>(constructor_args)...);
+  }
+
+  ~NoDestructor() = default;
+
+  NoDestructor(const NoDestructor&) = delete;
+  NoDestructor& operator=(const NoDestructor&) = delete;
+
+  InstanceType* get() {
+    return reinterpret_cast<InstanceType*>(&instance_storage_);
+  }
+
+ private:
+  typename std::aligned_storage<sizeof(InstanceType),
+                                alignof(InstanceType)>::type instance_storage_;
+};
+
+}  // namespace leveldb
+
+#endif  // STORAGE_LEVELDB_UTIL_NO_DESTRUCTOR_H_

lib/leveldb/util/no_destructor_test.cc

@@ -0,0 +1,49 @@
+// Copyright (c) 2018 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "util/no_destructor.h"
+
+#include <cstdint>
+#include <cstdlib>
+#include <utility>
+
+#include "gtest/gtest.h"
+
+namespace leveldb {
+
+namespace {
+
+struct DoNotDestruct {
+ public:
+  DoNotDestruct(uint32_t a, uint64_t b) : a(a), b(b) {}
+  ~DoNotDestruct() { std::abort(); }
+
+  // Used to check constructor argument forwarding.
+  uint32_t a;
+  uint64_t b;
+};
+
+constexpr const uint32_t kGoldenA = 0xdeadbeef;
+constexpr const uint64_t kGoldenB = 0xaabbccddeeffaabb;
+
+}  // namespace
+
+TEST(NoDestructorTest, StackInstance) {
+  NoDestructor<DoNotDestruct> instance(kGoldenA, kGoldenB);
+  ASSERT_EQ(kGoldenA, instance.get()->a);
+  ASSERT_EQ(kGoldenB, instance.get()->b);
+}
+
+TEST(NoDestructorTest, StaticInstance) {
+  static NoDestructor<DoNotDestruct> instance(kGoldenA, kGoldenB);
+  ASSERT_EQ(kGoldenA, instance.get()->a);
+  ASSERT_EQ(kGoldenB, instance.get()->b);
+}
+
+}  // namespace leveldb
+
+int main(int argc, char** argv) {
+  testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}

lib/leveldb/util/options.cc

@@ -0,0 +1,14 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "leveldb/options.h"
+
+#include "leveldb/comparator.h"
+#include "leveldb/env.h"
+
+namespace leveldb {
+
+Options::Options() : comparator(BytewiseComparator()), env(nullptr) {}
+
+}  // namespace leveldb

lib/leveldb/util/posix_logger.h

@@ -0,0 +1,130 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+//
+// Logger implementation that can be shared by all environments
+// where enough posix functionality is available.
+
+#ifndef STORAGE_LEVELDB_UTIL_POSIX_LOGGER_H_
+#define STORAGE_LEVELDB_UTIL_POSIX_LOGGER_H_
+
+#include <sys/time.h>
+
+#include <cassert>
+#include <cstdarg>
+#include <cstdio>
+#include <ctime>
+#include <sstream>
+#include <thread>
+
+#include "leveldb/env.h"
+
+namespace leveldb {
+
+class PosixLogger final : public Logger {
+ public:
+  // Creates a logger that writes to the given file.
+  //
+  // The PosixLogger instance takes ownership of the file handle.
+  explicit PosixLogger(std::FILE* fp) : fp_(fp) { assert(fp != nullptr); }
+
+  ~PosixLogger() override { std::fclose(fp_); }
+
+  void Logv(const char* format, std::va_list arguments) override {
+    // Record the time as close to the Logv() call as possible.
+    struct ::timeval now_timeval;
+    ::gettimeofday(&now_timeval, nullptr);
+    const std::time_t now_seconds = now_timeval.tv_sec;
+    struct std::tm now_components;
+    ::localtime_r(&now_seconds, &now_components);
+
+    // Record the thread ID.
+    constexpr const int kMaxThreadIdSize = 32;
+    std::ostringstream thread_stream;
+    thread_stream << std::this_thread::get_id();
+    std::string thread_id = thread_stream.str();
+    if (thread_id.size() > kMaxThreadIdSize) {
+      thread_id.resize(kMaxThreadIdSize);
+    }
+
+    // We first attempt to print into a stack-allocated buffer. If this attempt
+    // fails, we make a second attempt with a dynamically allocated buffer.
+    constexpr const int kStackBufferSize = 512;
+    char stack_buffer[kStackBufferSize];
+    static_assert(sizeof(stack_buffer) == static_cast<size_t>(kStackBufferSize),
+                  "sizeof(char) is expected to be 1 in C++");
+
+    int dynamic_buffer_size = 0;  // Computed in the first iteration.
+    for (int iteration = 0; iteration < 2; ++iteration) {
+      const int buffer_size =
+          (iteration == 0) ? kStackBufferSize : dynamic_buffer_size;
+      char* const buffer =
+          (iteration == 0) ? stack_buffer : new char[dynamic_buffer_size];
+
+      // Print the header into the buffer.
+      int buffer_offset = std::snprintf(
+          buffer, buffer_size, "%04d/%02d/%02d-%02d:%02d:%02d.%06d %s ",
+          now_components.tm_year + 1900, now_components.tm_mon + 1,
+          now_components.tm_mday, now_components.tm_hour, now_components.tm_min,
+          now_components.tm_sec, static_cast<int>(now_timeval.tv_usec),
+          thread_id.c_str());
+
+      // The header can be at most 28 characters (10 date + 15 time +
+      // 3 delimiters) plus the thread ID, which should fit comfortably into the
+      // static buffer.
+      assert(buffer_offset <= 28 + kMaxThreadIdSize);
+      static_assert(28 + kMaxThreadIdSize < kStackBufferSize,
+                    "stack-allocated buffer may not fit the message header");
+      assert(buffer_offset < buffer_size);
+
+      // Print the message into the buffer.
+      std::va_list arguments_copy;
+      va_copy(arguments_copy, arguments);
+      buffer_offset +=
+          std::vsnprintf(buffer + buffer_offset, buffer_size - buffer_offset,
+                         format, arguments_copy);
+      va_end(arguments_copy);
+
+      // The code below may append a newline at the end of the buffer, which
+      // requires an extra character.
+      if (buffer_offset >= buffer_size - 1) {
+        // The message did not fit into the buffer.
+        if (iteration == 0) {
+          // Re-run the loop and use a dynamically-allocated buffer. The buffer
+          // will be large enough for the log message, an extra newline and a
+          // null terminator.
+          dynamic_buffer_size = buffer_offset + 2;
+          continue;
+        }
+
+        // The dynamically-allocated buffer was incorrectly sized. This should
+        // not happen, assuming a correct implementation of std::(v)snprintf.
+        // Fail in tests, recover by truncating the log message in production.
+        assert(false);
+        buffer_offset = buffer_size - 1;
+      }
+
+      // Add a newline if necessary.
+      if (buffer[buffer_offset - 1] != '\n') {
+        buffer[buffer_offset] = '\n';
+        ++buffer_offset;
+      }
+
+      assert(buffer_offset <= buffer_size);
+      std::fwrite(buffer, 1, buffer_offset, fp_);
+      std::fflush(fp_);
+
+      if (iteration != 0) {
+        delete[] buffer;
+      }
+      break;
+    }
+  }
+
+ private:
+  std::FILE* const fp_;
+};
+
+}  // namespace leveldb
+
+#endif  // STORAGE_LEVELDB_UTIL_POSIX_LOGGER_H_

lib/leveldb/util/random.h

@@ -0,0 +1,63 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#ifndef STORAGE_LEVELDB_UTIL_RANDOM_H_
+#define STORAGE_LEVELDB_UTIL_RANDOM_H_
+
+#include <cstdint>
+
+namespace leveldb {
+
+// A very simple random number generator.  Not especially good at
+// generating truly random bits, but good enough for our needs in this
+// package.
+class Random {
+ private:
+  uint32_t seed_;
+
+ public:
+  explicit Random(uint32_t s) : seed_(s & 0x7fffffffu) {
+    // Avoid bad seeds.
+    if (seed_ == 0 || seed_ == 2147483647L) {
+      seed_ = 1;
+    }
+  }
+  uint32_t Next() {
+    static const uint32_t M = 2147483647L;  // 2^31-1
+    static const uint64_t A = 16807;        // bits 14, 8, 7, 5, 2, 1, 0
+    // We are computing
+    //       seed_ = (seed_ * A) % M,    where M = 2^31-1
+    //
+    // seed_ must not be zero or M, or else all subsequent computed values
+    // will be zero or M respectively.  For all other values, seed_ will end
+    // up cycling through every number in [1,M-1]
+    uint64_t product = seed_ * A;
+
+    // Compute (product % M) using the fact that ((x << 31) % M) == x.
+    seed_ = static_cast<uint32_t>((product >> 31) + (product & M));
+    // The first reduction may overflow by 1 bit, so we may need to
+    // repeat.  mod == M is not possible; using > allows the faster
+    // sign-bit-based test.
+    if (seed_ > M) {
+      seed_ -= M;
+    }
+    return seed_;
+  }
+  // Returns a uniformly distributed value in the range [0..n-1]
+  // REQUIRES: n > 0
+  uint32_t Uniform(int n) { return Next() % n; }
+
+  // Randomly returns true ~"1/n" of the time, and false otherwise.
+  // REQUIRES: n > 0
+  bool OneIn(int n) { return (Next() % n) == 0; }
+
+  // Skewed: pick "base" uniformly from range [0,max_log] and then
+  // return "base" random bits.  The effect is to pick a number in the
+  // range [0,2^max_log-1] with exponential bias towards smaller numbers.
+  uint32_t Skewed(int max_log) { return Uniform(1 << Uniform(max_log + 1)); }
+};
+
+}  // namespace leveldb
+
+#endif  // STORAGE_LEVELDB_UTIL_RANDOM_H_

lib/leveldb/util/status.cc

@@ -0,0 +1,77 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "leveldb/status.h"
+
+#include <cstdio>
+
+#include "port/port.h"
+
+namespace leveldb {
+
+const char* Status::CopyState(const char* state) {
+  uint32_t size;
+  std::memcpy(&size, state, sizeof(size));
+  char* result = new char[size + 5];
+  std::memcpy(result, state, size + 5);
+  return result;
+}
+
+Status::Status(Code code, const Slice& msg, const Slice& msg2) {
+  assert(code != kOk);
+  const uint32_t len1 = static_cast<uint32_t>(msg.size());
+  const uint32_t len2 = static_cast<uint32_t>(msg2.size());
+  const uint32_t size = len1 + (len2 ? (2 + len2) : 0);
+  char* result = new char[size + 5];
+  std::memcpy(result, &size, sizeof(size));
+  result[4] = static_cast<char>(code);
+  std::memcpy(result + 5, msg.data(), len1);
+  if (len2) {
+    result[5 + len1] = ':';
+    result[6 + len1] = ' ';
+    std::memcpy(result + 7 + len1, msg2.data(), len2);
+  }
+  state_ = result;
+}
+
+std::string Status::ToString() const {
+  if (state_ == nullptr) {
+    return "OK";
+  } else {
+    char tmp[30];
+    const char* type;
+    switch (code()) {
+      case kOk:
+        type = "OK";
+        break;
+      case kNotFound:
+        type = "NotFound: ";
+        break;
+      case kCorruption:
+        type = "Corruption: ";
+        break;
+      case kNotSupported:
+        type = "Not implemented: ";
+        break;
+      case kInvalidArgument:
+        type = "Invalid argument: ";
+        break;
+      case kIOError:
+        type = "IO error: ";
+        break;
+      default:
+        std::snprintf(tmp, sizeof(tmp),
+                      "Unknown code(%d): ", static_cast<int>(code()));
+        type = tmp;
+        break;
+    }
+    std::string result(type);
+    uint32_t length;
+    std::memcpy(&length, state_, sizeof(length));
+    result.append(state_ + 5, length);
+    return result;
+  }
+}
+
+}  // namespace leveldb

lib/leveldb/util/status_test.cc

@@ -0,0 +1,44 @@
+// Copyright (c) 2018 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "leveldb/status.h"
+
+#include <utility>
+
+#include "gtest/gtest.h"
+#include "leveldb/slice.h"
+
+namespace leveldb {
+
+TEST(Status, MoveConstructor) {
+  {
+    Status ok = Status::OK();
+    Status ok2 = std::move(ok);
+
+    ASSERT_TRUE(ok2.ok());
+  }
+
+  {
+    Status status = Status::NotFound("custom NotFound status message");
+    Status status2 = std::move(status);
+
+    ASSERT_TRUE(status2.IsNotFound());
+    ASSERT_EQ("NotFound: custom NotFound status message", status2.ToString());
+  }
+
+  {
+    Status self_moved = Status::IOError("custom IOError status message");
+
+    // Needed to bypass compiler warning about explicit move-assignment.
+    Status& self_moved_reference = self_moved;
+    self_moved_reference = std::move(self_moved);
+  }
+}
+
+}  // namespace leveldb
+
+int main(int argc, char** argv) {
+  testing::InitGoogleTest(&argc, argv);
+  return RUN_ALL_TESTS();
+}

lib/leveldb/util/testutil.cc

@@ -0,0 +1,51 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#include "util/testutil.h"
+
+#include <string>
+
+#include "util/random.h"
+
+namespace leveldb {
+namespace test {
+
+Slice RandomString(Random* rnd, int len, std::string* dst) {
+  dst->resize(len);
+  for (int i = 0; i < len; i++) {
+    (*dst)[i] = static_cast<char>(' ' + rnd->Uniform(95));  // ' ' .. '~'
+  }
+  return Slice(*dst);
+}
+
+std::string RandomKey(Random* rnd, int len) {
+  // Make sure to generate a wide variety of characters so we
+  // test the boundary conditions for short-key optimizations.
+  static const char kTestChars[] = {'\0', '\1', 'a',    'b',    'c',
+                                    'd',  'e',  '\xfd', '\xfe', '\xff'};
+  std::string result;
+  for (int i = 0; i < len; i++) {
+    result += kTestChars[rnd->Uniform(sizeof(kTestChars))];
+  }
+  return result;
+}
+
+Slice CompressibleString(Random* rnd, double compressed_fraction, size_t len,
+                         std::string* dst) {
+  int raw = static_cast<int>(len * compressed_fraction);
+  if (raw < 1) raw = 1;
+  std::string raw_data;
+  RandomString(rnd, raw, &raw_data);
+
+  // Duplicate the random data until we have filled "len" bytes
+  dst->clear();
+  while (dst->size() < len) {
+    dst->append(raw_data);
+  }
+  dst->resize(len);
+  return Slice(*dst);
+}
+
+}  // namespace test
+}  // namespace leveldb

lib/leveldb/util/testutil.h

@@ -0,0 +1,82 @@
+// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+
+#ifndef STORAGE_LEVELDB_UTIL_TESTUTIL_H_
+#define STORAGE_LEVELDB_UTIL_TESTUTIL_H_
+
+#include "gmock/gmock.h"
+#include "gtest/gtest.h"
+#include "helpers/memenv/memenv.h"
+#include "leveldb/env.h"
+#include "leveldb/slice.h"
+#include "util/random.h"
+
+namespace leveldb {
+namespace test {
+
+MATCHER(IsOK, "") { return arg.ok(); }
+
+// Macros for testing the results of functions that return leveldb::Status or
+// absl::StatusOr<T> (for any type T).
+#define EXPECT_LEVELDB_OK(expression) \
+  EXPECT_THAT(expression, leveldb::test::IsOK())
+#define ASSERT_LEVELDB_OK(expression) \
+  ASSERT_THAT(expression, leveldb::test::IsOK())
+
+// Returns the random seed used at the start of the current test run.
+inline int RandomSeed() {
+  return testing::UnitTest::GetInstance()->random_seed();
+}
+
+// Store in *dst a random string of length "len" and return a Slice that
+// references the generated data.
+Slice RandomString(Random* rnd, int len, std::string* dst);
+
+// Return a random key with the specified length that may contain interesting
+// characters (e.g. \x00, \xff, etc.).
+std::string RandomKey(Random* rnd, int len);
+
+// Store in *dst a string of length "len" that will compress to
+// "N*compressed_fraction" bytes and return a Slice that references
+// the generated data.
+Slice CompressibleString(Random* rnd, double compressed_fraction, size_t len,
+                         std::string* dst);
+
+// A wrapper that allows injection of errors.
+class ErrorEnv : public EnvWrapper {
+ public:
+  bool writable_file_error_;
+  int num_writable_file_errors_;
+
+  ErrorEnv()
+      : EnvWrapper(NewMemEnv(Env::Default())),
+        writable_file_error_(false),
+        num_writable_file_errors_(0) {}
+  ~ErrorEnv() override { delete target(); }
+
+  Status NewWritableFile(const std::string& fname,
+                         WritableFile** result) override {
+    if (writable_file_error_) {
+      ++num_writable_file_errors_;
+      *result = nullptr;
+      return Status::IOError(fname, "fake error");
+    }
+    return target()->NewWritableFile(fname, result);
+  }
+
+  Status NewAppendableFile(const std::string& fname,
+                           WritableFile** result) override {
+    if (writable_file_error_) {
+      ++num_writable_file_errors_;
+      *result = nullptr;
+      return Status::IOError(fname, "fake error");
+    }
+    return target()->NewAppendableFile(fname, result);
+  }
+};
+
+}  // namespace test
+}  // namespace leveldb
+
+#endif  // STORAGE_LEVELDB_UTIL_TESTUTIL_H_

lib/leveldb/util/windows_logger.h

@@ -0,0 +1,124 @@
+// Copyright (c) 2018 The LevelDB Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file. See the AUTHORS file for names of contributors.
+//
+// Logger implementation for the Windows platform.
+
+#ifndef STORAGE_LEVELDB_UTIL_WINDOWS_LOGGER_H_
+#define STORAGE_LEVELDB_UTIL_WINDOWS_LOGGER_H_
+
+#include <cassert>
+#include <cstdarg>
+#include <cstdio>
+#include <ctime>
+#include <sstream>
+#include <thread>
+
+#include "leveldb/env.h"
+
+namespace leveldb {
+
+class WindowsLogger final : public Logger {
+ public:
+  // Creates a logger that writes to the given file.
+  //
+  // The PosixLogger instance takes ownership of the file handle.
+  explicit WindowsLogger(std::FILE* fp) : fp_(fp) { assert(fp != nullptr); }
+
+  ~WindowsLogger() override { std::fclose(fp_); }
+
+  void Logv(const char* format, std::va_list arguments) override {
+    // Record the time as close to the Logv() call as possible.
+    SYSTEMTIME now_components;
+    ::GetLocalTime(&now_components);
+
+    // Record the thread ID.
+    constexpr const int kMaxThreadIdSize = 32;
+    std::ostringstream thread_stream;
+    thread_stream << std::this_thread::get_id();
+    std::string thread_id = thread_stream.str();
+    if (thread_id.size() > kMaxThreadIdSize) {
+      thread_id.resize(kMaxThreadIdSize);
+    }
+
+    // We first attempt to print into a stack-allocated buffer. If this attempt
+    // fails, we make a second attempt with a dynamically allocated buffer.
+    constexpr const int kStackBufferSize = 512;
+    char stack_buffer[kStackBufferSize];
+    static_assert(sizeof(stack_buffer) == static_cast<size_t>(kStackBufferSize),
+                  "sizeof(char) is expected to be 1 in C++");
+
+    int dynamic_buffer_size = 0;  // Computed in the first iteration.
+    for (int iteration = 0; iteration < 2; ++iteration) {
+      const int buffer_size =
+          (iteration == 0) ? kStackBufferSize : dynamic_buffer_size;
+      char* const buffer =
+          (iteration == 0) ? stack_buffer : new char[dynamic_buffer_size];
+
+      // Print the header into the buffer.
+      int buffer_offset = std::snprintf(
+          buffer, buffer_size, "%04d/%02d/%02d-%02d:%02d:%02d.%06d %s ",
+          now_components.wYear, now_components.wMonth, now_components.wDay,
+          now_components.wHour, now_components.wMinute, now_components.wSecond,
+          static_cast<int>(now_components.wMilliseconds * 1000),
+          thread_id.c_str());
+
+      // The header can be at most 28 characters (10 date + 15 time +
+      // 3 delimiters) plus the thread ID, which should fit comfortably into the
+      // static buffer.
+      assert(buffer_offset <= 28 + kMaxThreadIdSize);
+      static_assert(28 + kMaxThreadIdSize < kStackBufferSize,
+                    "stack-allocated buffer may not fit the message header");
+      assert(buffer_offset < buffer_size);
+
+      // Print the message into the buffer.
+      std::va_list arguments_copy;
+      va_copy(arguments_copy, arguments);
+      buffer_offset +=
+          std::vsnprintf(buffer + buffer_offset, buffer_size - buffer_offset,
+                         format, arguments_copy);
+      va_end(arguments_copy);
+
+      // The code below may append a newline at the end of the buffer, which
+      // requires an extra character.
+      if (buffer_offset >= buffer_size - 1) {
+        // The message did not fit into the buffer.
+        if (iteration == 0) {
+          // Re-run the loop and use a dynamically-allocated buffer. The buffer
+          // will be large enough for the log message, an extra newline and a
+          // null terminator.
+          dynamic_buffer_size = buffer_offset + 2;
+          continue;
+        }
+
+        // The dynamically-allocated buffer was incorrectly sized. This should
+        // not happen, assuming a correct implementation of std::(v)snprintf.
+        // Fail in tests, recover by truncating the log message in production.
+        assert(false);
+        buffer_offset = buffer_size - 1;
+      }
+
+      // Add a newline if necessary.
+      if (buffer[buffer_offset - 1] != '\n') {
+        buffer[buffer_offset] = '\n';
+        ++buffer_offset;
+      }
+
+      assert(buffer_offset <= buffer_size);
+      std::fwrite(buffer, 1, buffer_offset, fp_);
+      std::fflush(fp_);
+
+      if (iteration != 0) {
+        delete[] buffer;
+      }
+      break;
+    }
+  }
+
+ private:
+  std::FILE* const fp_;
+};
+
+}  // namespace leveldb
+
+#endif  // STORAGE_LEVELDB_UTIL_WINDOWS_LOGGER_H_