diff options
Diffstat (limited to 'util')
43 files changed, 3078 insertions, 1419 deletions
diff --git a/util/arena.cc b/util/arena.cc index 7407821..46e3b2e 100644 --- a/util/arena.cc +++ b/util/arena.cc @@ -3,16 +3,13 @@ // found in the LICENSE file. See the AUTHORS file for names of contributors. #include "util/arena.h" -#include <assert.h> namespace leveldb { static const int kBlockSize = 4096; -Arena::Arena() : memory_usage_(0) { - alloc_ptr_ = NULL; // First allocation will allocate a block - alloc_bytes_remaining_ = 0; -} +Arena::Arena() + : alloc_ptr_(nullptr), alloc_bytes_remaining_(0), memory_usage_(0) {} Arena::~Arena() { for (size_t i = 0; i < blocks_.size(); i++) { @@ -40,8 +37,9 @@ char* Arena::AllocateFallback(size_t bytes) { char* Arena::AllocateAligned(size_t bytes) { const int align = (sizeof(void*) > 8) ? sizeof(void*) : 8; - 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); + 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; @@ -53,15 +51,15 @@ char* Arena::AllocateAligned(size_t bytes) { // AllocateFallback always returned aligned memory result = AllocateFallback(bytes); } - assert((reinterpret_cast<uintptr_t>(result) & (align-1)) == 0); + 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_.NoBarrier_Store( - reinterpret_cast<void*>(MemoryUsage() + block_bytes + sizeof(char*))); + memory_usage_.fetch_add(block_bytes + sizeof(char*), + std::memory_order_relaxed); return result; } diff --git a/util/arena.h b/util/arena.h index 48bab33..68fc55d 100644 --- a/util/arena.h +++ b/util/arena.h @@ -5,29 +5,33 @@ #ifndef STORAGE_LEVELDB_UTIL_ARENA_H_ #define STORAGE_LEVELDB_UTIL_ARENA_H_ +#include <atomic> +#include <cassert> +#include <cstddef> +#include <cstdint> #include <vector> -#include <assert.h> -#include <stddef.h> -#include <stdint.h> -#include "port/port.h" 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 + // 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 reinterpret_cast<uintptr_t>(memory_usage_.NoBarrier_Load()); + return memory_usage_.load(std::memory_order_relaxed); } private: @@ -42,11 +46,10 @@ class Arena { std::vector<char*> blocks_; // Total memory usage of the arena. - port::AtomicPointer memory_usage_; - - // No copying allowed - Arena(const Arena&); - void operator=(const 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) { diff --git a/util/arena_test.cc b/util/arena_test.cc index 58e870e..3e2011e 100644 --- a/util/arena_test.cc +++ b/util/arena_test.cc @@ -4,19 +4,15 @@ #include "util/arena.h" +#include "gtest/gtest.h" #include "util/random.h" -#include "util/testharness.h" namespace leveldb { -class ArenaTest { }; - -TEST(ArenaTest, Empty) { - Arena arena; -} +TEST(ArenaTest, Empty) { Arena arena; } TEST(ArenaTest, Simple) { - std::vector<std::pair<size_t, char*> > allocated; + std::vector<std::pair<size_t, char*>> allocated; Arena arena; const int N = 100000; size_t bytes = 0; @@ -26,8 +22,9 @@ TEST(ArenaTest, Simple) { if (i % (N / 10) == 0) { s = i; } else { - s = rnd.OneIn(4000) ? rnd.Uniform(6000) : - (rnd.OneIn(10) ? rnd.Uniform(100) : rnd.Uniform(20)); + 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. @@ -47,7 +44,7 @@ TEST(ArenaTest, Simple) { bytes += s; allocated.push_back(std::make_pair(s, r)); ASSERT_GE(arena.MemoryUsage(), bytes); - if (i > N/10) { + if (i > N / 10) { ASSERT_LE(arena.MemoryUsage(), bytes * 1.10); } } @@ -62,7 +59,3 @@ TEST(ArenaTest, Simple) { } } // namespace leveldb - -int main(int argc, char** argv) { - return leveldb::test::RunAllTests(); -} diff --git a/util/bloom.cc b/util/bloom.cc index bf3e4ca..87547a7 100644 --- a/util/bloom.cc +++ b/util/bloom.cc @@ -15,24 +15,17 @@ static uint32_t BloomHash(const Slice& key) { } class BloomFilterPolicy : public FilterPolicy { - private: - size_t bits_per_key_; - size_t k_; - public: - explicit BloomFilterPolicy(int bits_per_key) - : bits_per_key_(bits_per_key) { + 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; } - virtual const char* Name() const { - return "leveldb.BuiltinBloomFilter2"; - } + const char* Name() const override { return "leveldb.BuiltinBloomFilter2"; } - virtual void CreateFilter(const Slice* keys, int n, std::string* dst) const { + 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_; @@ -54,13 +47,13 @@ class BloomFilterPolicy : public FilterPolicy { 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)); + array[bitpos / 8] |= (1 << (bitpos % 8)); h += delta; } } } - virtual bool KeyMayMatch(const Slice& key, const Slice& bloom_filter) const { + bool KeyMayMatch(const Slice& key, const Slice& bloom_filter) const override { const size_t len = bloom_filter.size(); if (len < 2) return false; @@ -69,7 +62,7 @@ class BloomFilterPolicy : public FilterPolicy { // 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]; + const size_t k = array[len - 1]; if (k > 30) { // Reserved for potentially new encodings for short bloom filters. // Consider it a match. @@ -80,13 +73,17 @@ class BloomFilterPolicy : public FilterPolicy { 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; + 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); diff --git a/util/bloom_test.cc b/util/bloom_test.cc index 1b87a2b..9f11108 100644 --- a/util/bloom_test.cc +++ b/util/bloom_test.cc @@ -2,11 +2,10 @@ // 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/testharness.h" #include "util/testutil.h" namespace leveldb { @@ -18,27 +17,18 @@ static Slice Key(int i, char* buffer) { return Slice(buffer, sizeof(uint32_t)); } -class BloomTest { - private: - const FilterPolicy* policy_; - std::string filter_; - std::vector<std::string> keys_; - +class BloomTest : public testing::Test { public: - BloomTest() : policy_(NewBloomFilterPolicy(10)) { } + BloomTest() : policy_(NewBloomFilterPolicy(10)) {} - ~BloomTest() { - delete policy_; - } + ~BloomTest() { delete policy_; } void Reset() { keys_.clear(); filter_.clear(); } - void Add(const Slice& s) { - keys_.push_back(s.ToString()); - } + void Add(const Slice& s) { keys_.push_back(s.ToString()); } void Build() { std::vector<Slice> key_slices; @@ -52,19 +42,17 @@ class BloomTest { if (kVerbose >= 2) DumpFilter(); } - size_t FilterSize() const { - return filter_.size(); - } + size_t FilterSize() const { return filter_.size(); } void DumpFilter() { - fprintf(stderr, "F("); - for (size_t i = 0; i+1 < filter_.size(); i++) { + 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++) { - fprintf(stderr, "%c", (c & (1 <<j)) ? '1' : '.'); + std::fprintf(stderr, "%c", (c & (1 << j)) ? '1' : '.'); } } - fprintf(stderr, ")\n"); + std::fprintf(stderr, ")\n"); } bool Matches(const Slice& s) { @@ -84,20 +72,25 @@ class BloomTest { } return result / 10000.0; } + + private: + const FilterPolicy* policy_; + std::string filter_; + std::vector<std::string> keys_; }; -TEST(BloomTest, EmptyFilter) { - ASSERT_TRUE(! Matches("hello")); - ASSERT_TRUE(! Matches("world")); +TEST_F(BloomTest, EmptyFilter) { + ASSERT_TRUE(!Matches("hello")); + ASSERT_TRUE(!Matches("world")); } -TEST(BloomTest, Small) { +TEST_F(BloomTest, Small) { Add("hello"); Add("world"); ASSERT_TRUE(Matches("hello")); ASSERT_TRUE(Matches("world")); - ASSERT_TRUE(! Matches("x")); - ASSERT_TRUE(! Matches("foo")); + ASSERT_TRUE(!Matches("x")); + ASSERT_TRUE(!Matches("foo")); } static int NextLength(int length) { @@ -113,7 +106,7 @@ static int NextLength(int length) { return length; } -TEST(BloomTest, VaryingLengths) { +TEST_F(BloomTest, VaryingLengths) { char buffer[sizeof(int)]; // Count number of filters that significantly exceed the false positive rate @@ -139,24 +132,23 @@ TEST(BloomTest, VaryingLengths) { // Check false positive rate double rate = FalsePositiveRate(); if (kVerbose >= 1) { - fprintf(stderr, "False positives: %5.2f%% @ length = %6d ; bytes = %6d\n", - rate*100.0, length, static_cast<int>(FilterSize())); + 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++; + 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) { - fprintf(stderr, "Filters: %d good, %d mediocre\n", - good_filters, mediocre_filters); + std::fprintf(stderr, "Filters: %d good, %d mediocre\n", good_filters, + mediocre_filters); } - ASSERT_LE(mediocre_filters, good_filters/5); + ASSERT_LE(mediocre_filters, good_filters / 5); } // Different bits-per-byte } // namespace leveldb - -int main(int argc, char** argv) { - return leveldb::test::RunAllTests(); -} diff --git a/util/cache.cc b/util/cache.cc index 10b7103..fc3d154 100644 --- a/util/cache.cc +++ b/util/cache.cc @@ -2,11 +2,12 @@ // 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 <assert.h> -#include <stdio.h> -#include <stdlib.h> - #include "leveldb/cache.h" + +#include <cassert> +#include <cstdio> +#include <cstdlib> + #include "port/port.h" #include "port/thread_annotations.h" #include "util/hash.h" @@ -14,8 +15,7 @@ namespace leveldb { -Cache::~Cache() { -} +Cache::~Cache() {} namespace { @@ -46,15 +46,15 @@ struct LRUHandle { LRUHandle* next_hash; LRUHandle* next; LRUHandle* prev; - size_t charge; // TODO(opt): Only allow uint32_t? + 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 + 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 + // 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); @@ -69,7 +69,7 @@ struct LRUHandle { // 4.4.3's builtin hashtable. class HandleTable { public: - HandleTable() : length_(0), elems_(0), list_(NULL) { Resize(); } + HandleTable() : length_(0), elems_(0), list_(nullptr) { Resize(); } ~HandleTable() { delete[] list_; } LRUHandle* Lookup(const Slice& key, uint32_t hash) { @@ -79,9 +79,9 @@ class HandleTable { LRUHandle* Insert(LRUHandle* h) { LRUHandle** ptr = FindPointer(h->key(), h->hash); LRUHandle* old = *ptr; - h->next_hash = (old == NULL ? NULL : old->next_hash); + h->next_hash = (old == nullptr ? nullptr : old->next_hash); *ptr = h; - if (old == NULL) { + if (old == nullptr) { ++elems_; if (elems_ > length_) { // Since each cache entry is fairly large, we aim for a small @@ -95,7 +95,7 @@ class HandleTable { LRUHandle* Remove(const Slice& key, uint32_t hash) { LRUHandle** ptr = FindPointer(key, hash); LRUHandle* result = *ptr; - if (result != NULL) { + if (result != nullptr) { *ptr = result->next_hash; --elems_; } @@ -114,8 +114,7 @@ class HandleTable { // 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 != NULL && - ((*ptr)->hash != hash || key != (*ptr)->key())) { + while (*ptr != nullptr && ((*ptr)->hash != hash || key != (*ptr)->key())) { ptr = &(*ptr)->next_hash; } return ptr; @@ -131,7 +130,7 @@ class HandleTable { uint32_t count = 0; for (uint32_t i = 0; i < length_; i++) { LRUHandle* h = list_[i]; - while (h != NULL) { + while (h != nullptr) { LRUHandle* next = h->next_hash; uint32_t hash = h->hash; LRUHandle** ptr = &new_list[hash & (new_length - 1)]; @@ -158,8 +157,8 @@ class 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, + 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); @@ -172,7 +171,7 @@ class LRUCache { private: void LRU_Remove(LRUHandle* e); - void LRU_Append(LRUHandle*list, 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_); @@ -196,8 +195,7 @@ class LRUCache { HandleTable table_ GUARDED_BY(mutex_); }; -LRUCache::LRUCache() - : usage_(0) { +LRUCache::LRUCache() : capacity_(0), usage_(0) { // Make empty circular linked lists. lru_.next = &lru_; lru_.prev = &lru_; @@ -207,7 +205,7 @@ LRUCache::LRUCache() LRUCache::~LRUCache() { assert(in_use_.next == &in_use_); // Error if caller has an unreleased handle - for (LRUHandle* e = lru_.next; e != &lru_; ) { + for (LRUHandle* e = lru_.next; e != &lru_;) { LRUHandle* next = e->next; assert(e->in_cache); e->in_cache = false; @@ -255,7 +253,7 @@ void LRUCache::LRU_Append(LRUHandle* list, LRUHandle* e) { Cache::Handle* LRUCache::Lookup(const Slice& key, uint32_t hash) { MutexLock l(&mutex_); LRUHandle* e = table_.Lookup(key, hash); - if (e != NULL) { + if (e != nullptr) { Ref(e); } return reinterpret_cast<Cache::Handle*>(e); @@ -266,13 +264,14 @@ void LRUCache::Release(Cache::Handle* handle) { 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)) { +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())); + LRUHandle* e = + reinterpret_cast<LRUHandle*>(malloc(sizeof(LRUHandle) - 1 + key.size())); e->value = value; e->deleter = deleter; e->charge = charge; @@ -280,7 +279,7 @@ Cache::Handle* LRUCache::Insert( e->hash = hash; e->in_cache = false; e->refs = 1; // for the returned handle. - memcpy(e->key_data, key.data(), key.size()); + std::memcpy(e->key_data, key.data(), key.size()); if (capacity_ > 0) { e->refs++; // for the cache's reference. @@ -290,7 +289,7 @@ Cache::Handle* LRUCache::Insert( 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 = NULL; + e->next = nullptr; } while (usage_ > capacity_ && lru_.next != &lru_) { LRUHandle* old = lru_.next; @@ -304,17 +303,17 @@ Cache::Handle* LRUCache::Insert( return reinterpret_cast<Cache::Handle*>(e); } -// If e != NULL, finish removing *e from the cache; it has already been removed -// from the hash table. Return whether e != NULL. Requires mutex_ held. +// 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 != NULL) { + if (e != nullptr) { assert(e->in_cache); LRU_Remove(e); e->in_cache = false; usage_ -= e->charge; Unref(e); } - return e != NULL; + return e != nullptr; } void LRUCache::Erase(const Slice& key, uint32_t hash) { @@ -347,49 +346,46 @@ class ShardedLRUCache : public Cache { return Hash(s.data(), s.size(), 0); } - static uint32_t Shard(uint32_t hash) { - return hash >> (32 - kNumShardBits); - } + static uint32_t Shard(uint32_t hash) { return hash >> (32 - kNumShardBits); } public: - explicit ShardedLRUCache(size_t capacity) - : last_id_(0) { + 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); } } - virtual ~ShardedLRUCache() { } - virtual Handle* Insert(const Slice& key, void* value, size_t charge, - void (*deleter)(const Slice& key, void* value)) { + ~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); } - virtual Handle* Lookup(const Slice& key) { + Handle* Lookup(const Slice& key) override { const uint32_t hash = HashSlice(key); return shard_[Shard(hash)].Lookup(key, hash); } - virtual void Release(Handle* handle) { + void Release(Handle* handle) override { LRUHandle* h = reinterpret_cast<LRUHandle*>(handle); shard_[Shard(h->hash)].Release(handle); } - virtual void Erase(const Slice& key) { + void Erase(const Slice& key) override { const uint32_t hash = HashSlice(key); shard_[Shard(hash)].Erase(key, hash); } - virtual void* Value(Handle* handle) { + void* Value(Handle* handle) override { return reinterpret_cast<LRUHandle*>(handle)->value; } - virtual uint64_t NewId() { + uint64_t NewId() override { MutexLock l(&id_mutex_); return ++(last_id_); } - virtual void Prune() { + void Prune() override { for (int s = 0; s < kNumShards; s++) { shard_[s].Prune(); } } - virtual size_t TotalCharge() const { + size_t TotalCharge() const override { size_t total = 0; for (int s = 0; s < kNumShards; s++) { total += shard_[s].TotalCharge(); @@ -400,8 +396,6 @@ class ShardedLRUCache : public Cache { } // end anonymous namespace -Cache* NewLRUCache(size_t capacity) { - return new ShardedLRUCache(capacity); -} +Cache* NewLRUCache(size_t capacity) { return new ShardedLRUCache(capacity); } } // namespace leveldb diff --git a/util/cache_test.cc b/util/cache_test.cc index 246ab8e..e68da34 100644 --- a/util/cache_test.cc +++ b/util/cache_test.cc @@ -5,8 +5,9 @@ #include "leveldb/cache.h" #include <vector> + +#include "gtest/gtest.h" #include "util/coding.h" -#include "util/testharness.h" namespace leveldb { @@ -23,32 +24,26 @@ static int DecodeKey(const Slice& k) { static void* EncodeValue(uintptr_t v) { return reinterpret_cast<void*>(v); } static int DecodeValue(void* v) { return reinterpret_cast<uintptr_t>(v); } -class CacheTest { +class CacheTest : public testing::Test { public: - static CacheTest* current_; - static void Deleter(const Slice& key, void* v) { current_->deleted_keys_.push_back(DecodeKey(key)); current_->deleted_values_.push_back(DecodeValue(v)); } - static const int kCacheSize = 1000; + static constexpr int kCacheSize = 1000; std::vector<int> deleted_keys_; std::vector<int> deleted_values_; Cache* cache_; - CacheTest() : cache_(NewLRUCache(kCacheSize)) { - current_ = this; - } + CacheTest() : cache_(NewLRUCache(kCacheSize)) { current_ = this; } - ~CacheTest() { - delete cache_; - } + ~CacheTest() { delete cache_; } int Lookup(int key) { Cache::Handle* handle = cache_->Lookup(EncodeKey(key)); - const int r = (handle == NULL) ? -1 : DecodeValue(cache_->Value(handle)); - if (handle != NULL) { + const int r = (handle == nullptr) ? -1 : DecodeValue(cache_->Value(handle)); + if (handle != nullptr) { cache_->Release(handle); } return r; @@ -64,55 +59,54 @@ class CacheTest { &CacheTest::Deleter); } - void Erase(int key) { - cache_->Erase(EncodeKey(key)); - } + void Erase(int key) { cache_->Erase(EncodeKey(key)); } + static CacheTest* current_; }; CacheTest* CacheTest::current_; -TEST(CacheTest, HitAndMiss) { +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)); + 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)); + 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, Lookup(300)); ASSERT_EQ(1, deleted_keys_.size()); ASSERT_EQ(100, deleted_keys_[0]); ASSERT_EQ(101, deleted_values_[0]); } -TEST(CacheTest, Erase) { +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(-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(-1, Lookup(100)); ASSERT_EQ(201, Lookup(200)); ASSERT_EQ(1, deleted_keys_.size()); } -TEST(CacheTest, EntriesArePinned) { +TEST_F(CacheTest, EntriesArePinned) { Insert(100, 101); Cache::Handle* h1 = cache_->Lookup(EncodeKey(100)); ASSERT_EQ(101, DecodeValue(cache_->Value(h1))); @@ -137,7 +131,7 @@ TEST(CacheTest, EntriesArePinned) { ASSERT_EQ(102, deleted_values_[1]); } -TEST(CacheTest, EvictionPolicy) { +TEST_F(CacheTest, EvictionPolicy) { Insert(100, 101); Insert(200, 201); Insert(300, 301); @@ -146,8 +140,8 @@ TEST(CacheTest, EvictionPolicy) { // 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)); + Insert(1000 + i, 2000 + i); + ASSERT_EQ(2000 + i, Lookup(1000 + i)); ASSERT_EQ(101, Lookup(100)); } ASSERT_EQ(101, Lookup(100)); @@ -156,16 +150,16 @@ TEST(CacheTest, EvictionPolicy) { cache_->Release(h); } -TEST(CacheTest, UseExceedsCacheSize) { +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)); + 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)); + ASSERT_EQ(2000 + i, Lookup(1000 + i)); } for (int i = 0; i < h.size(); i++) { @@ -173,7 +167,7 @@ TEST(CacheTest, UseExceedsCacheSize) { } } -TEST(CacheTest, HeavyEntries) { +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. @@ -181,9 +175,9 @@ TEST(CacheTest, HeavyEntries) { const int kHeavy = 10; int added = 0; int index = 0; - while (added < 2*kCacheSize) { + while (added < 2 * kCacheSize) { const int weight = (index & 1) ? kLight : kHeavy; - Insert(index, 1000+index, weight); + Insert(index, 1000 + index, weight); added += weight; index++; } @@ -194,19 +188,19 @@ TEST(CacheTest, HeavyEntries) { int r = Lookup(i); if (r >= 0) { cached_weight += weight; - ASSERT_EQ(1000+i, r); + ASSERT_EQ(1000 + i, r); } } - ASSERT_LE(cached_weight, kCacheSize + kCacheSize/10); + ASSERT_LE(cached_weight, kCacheSize + kCacheSize / 10); } -TEST(CacheTest, NewId) { +TEST_F(CacheTest, NewId) { uint64_t a = cache_->NewId(); uint64_t b = cache_->NewId(); ASSERT_NE(a, b); } -TEST(CacheTest, Prune) { +TEST_F(CacheTest, Prune) { Insert(1, 100); Insert(2, 200); @@ -219,7 +213,7 @@ TEST(CacheTest, Prune) { ASSERT_EQ(-1, Lookup(2)); } -TEST(CacheTest, ZeroSizeCache) { +TEST_F(CacheTest, ZeroSizeCache) { delete cache_; cache_ = NewLRUCache(0); @@ -228,7 +222,3 @@ TEST(CacheTest, ZeroSizeCache) { } } // namespace leveldb - -int main(int argc, char** argv) { - return leveldb::test::RunAllTests(); -} diff --git a/util/coding.cc b/util/coding.cc index 21e3186..df3fa10 100644 --- a/util/coding.cc +++ b/util/coding.cc @@ -6,32 +6,6 @@ namespace leveldb { -void EncodeFixed32(char* buf, uint32_t value) { - if (port::kLittleEndian) { - memcpy(buf, &value, sizeof(value)); - } else { - buf[0] = value & 0xff; - buf[1] = (value >> 8) & 0xff; - buf[2] = (value >> 16) & 0xff; - buf[3] = (value >> 24) & 0xff; - } -} - -void EncodeFixed64(char* buf, uint64_t value) { - if (port::kLittleEndian) { - memcpy(buf, &value, sizeof(value)); - } else { - buf[0] = value & 0xff; - buf[1] = (value >> 8) & 0xff; - buf[2] = (value >> 16) & 0xff; - buf[3] = (value >> 24) & 0xff; - buf[4] = (value >> 32) & 0xff; - buf[5] = (value >> 40) & 0xff; - buf[6] = (value >> 48) & 0xff; - buf[7] = (value >> 56) & 0xff; - } -} - void PutFixed32(std::string* dst, uint32_t value) { char buf[sizeof(value)]; EncodeFixed32(buf, value); @@ -46,28 +20,28 @@ void PutFixed64(std::string* dst, uint64_t value) { char* EncodeVarint32(char* dst, uint32_t v) { // Operate on characters as unsigneds - unsigned char* ptr = reinterpret_cast<unsigned char*>(dst); + uint8_t* ptr = reinterpret_cast<uint8_t*>(dst); static const int B = 128; - if (v < (1<<7)) { + if (v < (1 << 7)) { *(ptr++) = v; - } else if (v < (1<<14)) { + } else if (v < (1 << 14)) { *(ptr++) = v | B; - *(ptr++) = v>>7; - } else if (v < (1<<21)) { + *(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 >> 7) | B; + *(ptr++) = v >> 14; + } else if (v < (1 << 28)) { *(ptr++) = v | B; - *(ptr++) = (v>>7) | B; - *(ptr++) = (v>>14) | B; - *(ptr++) = v>>21; + *(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; + *(ptr++) = (v >> 7) | B; + *(ptr++) = (v >> 14) | B; + *(ptr++) = (v >> 21) | B; + *(ptr++) = v >> 28; } return reinterpret_cast<char*>(ptr); } @@ -80,12 +54,12 @@ void PutVarint32(std::string* dst, uint32_t v) { char* EncodeVarint64(char* dst, uint64_t v) { static const int B = 128; - unsigned char* ptr = reinterpret_cast<unsigned char*>(dst); + uint8_t* ptr = reinterpret_cast<uint8_t*>(dst); while (v >= B) { - *(ptr++) = (v & (B-1)) | B; + *(ptr++) = v | B; v >>= 7; } - *(ptr++) = static_cast<unsigned char>(v); + *(ptr++) = static_cast<uint8_t>(v); return reinterpret_cast<char*>(ptr); } @@ -109,12 +83,11 @@ int VarintLength(uint64_t v) { return len; } -const char* GetVarint32PtrFallback(const char* p, - const char* limit, +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 unsigned char*>(p)); + uint32_t byte = *(reinterpret_cast<const uint8_t*>(p)); p++; if (byte & 128) { // More bytes are present @@ -125,14 +98,14 @@ const char* GetVarint32PtrFallback(const char* p, return reinterpret_cast<const char*>(p); } } - return NULL; + 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 == NULL) { + if (q == nullptr) { return false; } else { *input = Slice(q, limit - q); @@ -143,7 +116,7 @@ bool GetVarint32(Slice* input, uint32_t* value) { 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 unsigned char*>(p)); + uint64_t byte = *(reinterpret_cast<const uint8_t*>(p)); p++; if (byte & 128) { // More bytes are present @@ -154,14 +127,14 @@ const char* GetVarint64Ptr(const char* p, const char* limit, uint64_t* value) { return reinterpret_cast<const char*>(p); } } - return NULL; + 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 == NULL) { + if (q == nullptr) { return false; } else { *input = Slice(q, limit - q); @@ -173,16 +146,15 @@ const char* GetLengthPrefixedSlice(const char* p, const char* limit, Slice* result) { uint32_t len; p = GetVarint32Ptr(p, limit, &len); - if (p == NULL) return NULL; - if (p + len > limit) return NULL; + 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) { + if (GetVarint32(input, &len) && input->size() >= len) { *result = Slice(input->data(), len); input->remove_prefix(len); return true; diff --git a/util/coding.h b/util/coding.h index 1fb3d66..f0bb57b 100644 --- a/util/coding.h +++ b/util/coding.h @@ -10,9 +10,8 @@ #ifndef STORAGE_LEVELDB_UTIL_CODING_H_ #define STORAGE_LEVELDB_UTIL_CODING_H_ -#include <stdint.h> -#include <string.h> - +#include <cstdint> +#include <cstring> #include <string> #include "leveldb/slice.h" @@ -35,7 +34,7 @@ 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 -// NULL on error. These routines only look at bytes in the range +// 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); @@ -44,55 +43,72 @@ const char* GetVarint64Ptr(const char* p, const char* limit, uint64_t* v); int VarintLength(uint64_t v); // Lower-level versions of Put... that write directly into a character buffer -// REQUIRES: dst has enough space for the value being written -void EncodeFixed32(char* dst, uint32_t value); -void EncodeFixed64(char* dst, uint64_t value); - -// 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) { - if (port::kLittleEndian) { - // Load the raw bytes - uint32_t result; - memcpy(&result, ptr, sizeof(result)); // gcc optimizes this to a plain load - return result; - } else { - return ((static_cast<uint32_t>(static_cast<unsigned char>(ptr[0]))) - | (static_cast<uint32_t>(static_cast<unsigned char>(ptr[1])) << 8) - | (static_cast<uint32_t>(static_cast<unsigned char>(ptr[2])) << 16) - | (static_cast<uint32_t>(static_cast<unsigned char>(ptr[3])) << 24)); - } + 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) { - if (port::kLittleEndian) { - // Load the raw bytes - uint64_t result; - memcpy(&result, ptr, sizeof(result)); // gcc optimizes this to a plain load - return result; - } else { - uint64_t lo = DecodeFixed32(ptr); - uint64_t hi = DecodeFixed32(ptr + 4); - return (hi << 32) | lo; - } + 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, +const char* GetVarint32PtrFallback(const char* p, const char* limit, uint32_t* value); -inline const char* GetVarint32Ptr(const char* p, - const char* limit, +inline const char* GetVarint32Ptr(const char* p, const char* limit, uint32_t* value) { if (p < limit) { - uint32_t result = *(reinterpret_cast<const unsigned char*>(p)); + uint32_t result = *(reinterpret_cast<const uint8_t*>(p)); if ((result & 128) == 0) { *value = result; return p + 1; diff --git a/util/coding_test.cc b/util/coding_test.cc index 22f6adc..cceda14 100644 --- a/util/coding_test.cc +++ b/util/coding_test.cc @@ -2,15 +2,14 @@ // 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 "util/coding.h" -#include "util/testharness.h" +#include "gtest/gtest.h" namespace leveldb { -class Coding { }; - TEST(Coding, Fixed32) { std::string s; for (uint32_t v = 0; v < 100000; v++) { @@ -39,15 +38,15 @@ TEST(Coding, Fixed64) { uint64_t v = static_cast<uint64_t>(1) << power; uint64_t actual; actual = DecodeFixed64(p); - ASSERT_EQ(v-1, actual); + ASSERT_EQ(v - 1, actual); p += sizeof(uint64_t); actual = DecodeFixed64(p); - ASSERT_EQ(v+0, actual); + ASSERT_EQ(v + 0, actual); p += sizeof(uint64_t); actual = DecodeFixed64(p); - ASSERT_EQ(v+1, actual); + ASSERT_EQ(v + 1, actual); p += sizeof(uint64_t); } } @@ -89,7 +88,7 @@ TEST(Coding, Varint32) { uint32_t actual; const char* start = p; p = GetVarint32Ptr(p, limit, &actual); - ASSERT_TRUE(p != NULL); + ASSERT_TRUE(p != nullptr); ASSERT_EQ(expected, actual); ASSERT_EQ(VarintLength(actual), p - start); } @@ -108,8 +107,8 @@ TEST(Coding, Varint64) { // 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); + values.push_back(power - 1); + values.push_back(power + 1); } std::string s; @@ -124,19 +123,18 @@ TEST(Coding, Varint64) { uint64_t actual; const char* start = p; p = GetVarint64Ptr(p, limit, &actual); - ASSERT_TRUE(p != NULL); + 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) - == NULL); + ASSERT_TRUE(GetVarint32Ptr(input.data(), input.data() + input.size(), + &result) == nullptr); } TEST(Coding, Varint32Truncation) { @@ -145,17 +143,18 @@ TEST(Coding, Varint32Truncation) { 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) == NULL); + ASSERT_TRUE(GetVarint32Ptr(s.data(), s.data() + len, &result) == nullptr); } - ASSERT_TRUE(GetVarint32Ptr(s.data(), s.data() + s.size(), &result) != NULL); + 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) - == NULL); + ASSERT_TRUE(GetVarint64Ptr(input.data(), input.data() + input.size(), + &result) == nullptr); } TEST(Coding, Varint64Truncation) { @@ -164,9 +163,10 @@ TEST(Coding, Varint64Truncation) { 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) == NULL); + ASSERT_TRUE(GetVarint64Ptr(s.data(), s.data() + len, &result) == nullptr); } - ASSERT_TRUE(GetVarint64Ptr(s.data(), s.data() + s.size(), &result) != NULL); + ASSERT_TRUE(GetVarint64Ptr(s.data(), s.data() + s.size(), &result) != + nullptr); ASSERT_EQ(large_value, result); } @@ -191,7 +191,3 @@ TEST(Coding, Strings) { } } // namespace leveldb - -int main(int argc, char** argv) { - return leveldb::test::RunAllTests(); -} diff --git a/util/comparator.cc b/util/comparator.cc index 4b7b572..c5766e9 100644 --- a/util/comparator.cc +++ b/util/comparator.cc @@ -2,33 +2,34 @@ // 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 <algorithm> -#include <stdint.h> #include "leveldb/comparator.h" + +#include <algorithm> +#include <cstdint> +#include <string> +#include <type_traits> + #include "leveldb/slice.h" -#include "port/port.h" #include "util/logging.h" +#include "util/no_destructor.h" namespace leveldb { -Comparator::~Comparator() { } +Comparator::~Comparator() = default; namespace { class BytewiseComparatorImpl : public Comparator { public: - BytewiseComparatorImpl() { } + BytewiseComparatorImpl() = default; - virtual const char* Name() const { - return "leveldb.BytewiseComparator"; - } + const char* Name() const override { return "leveldb.BytewiseComparator"; } - virtual int Compare(const Slice& a, const Slice& b) const { + int Compare(const Slice& a, const Slice& b) const override { return a.compare(b); } - virtual void FindShortestSeparator( - std::string* start, - const Slice& limit) const { + 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; @@ -50,14 +51,14 @@ class BytewiseComparatorImpl : public Comparator { } } - virtual void FindShortSuccessor(std::string* key) const { + 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); + key->resize(i + 1); return; } } @@ -66,16 +67,9 @@ class BytewiseComparatorImpl : public Comparator { }; } // namespace -static port::OnceType once = LEVELDB_ONCE_INIT; -static const Comparator* bytewise; - -static void InitModule() { - bytewise = new BytewiseComparatorImpl; -} - const Comparator* BytewiseComparator() { - port::InitOnce(&once, InitModule); - return bytewise; + static NoDestructor<BytewiseComparatorImpl> singleton; + return singleton.get(); } } // namespace leveldb diff --git a/util/crc32c.cc b/util/crc32c.cc index 4f1d80f..3f18908 100644 --- a/util/crc32c.cc +++ b/util/crc32c.cc @@ -6,8 +6,8 @@ #include "util/crc32c.h" -#include <stddef.h> -#include <stdint.h> +#include <cstddef> +#include <cstdint> #include "port/port.h" #include "util/coding.h" @@ -256,8 +256,8 @@ inline uint32_t ReadUint32LE(const uint8_t* buffer) { 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)); + (reinterpret_cast<uintptr_t>(pointer) + (N - 1)) & + ~static_cast<uintptr_t>(N - 1)); } } // namespace @@ -273,14 +273,14 @@ static bool CanAccelerateCRC32C() { return port::AcceleratedCRC32C(0, kTestCRCBuffer, kBufSize) == kTestCRCValue; } -uint32_t Extend(uint32_t crc, const char* buf, size_t size) { +uint32_t Extend(uint32_t crc, const char* data, size_t n) { static bool accelerate = CanAccelerateCRC32C(); if (accelerate) { - return port::AcceleratedCRC32C(crc, buf, size); + return port::AcceleratedCRC32C(crc, data, n); } - const uint8_t* p = reinterpret_cast<const uint8_t*>(buf); - const uint8_t* e = p + size; + 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. diff --git a/util/crc32c.h b/util/crc32c.h index 7864688..b420b5f 100644 --- a/util/crc32c.h +++ b/util/crc32c.h @@ -5,8 +5,8 @@ #ifndef STORAGE_LEVELDB_UTIL_CRC32C_H_ #define STORAGE_LEVELDB_UTIL_CRC32C_H_ -#include <stddef.h> -#include <stdint.h> +#include <cstddef> +#include <cstdint> namespace leveldb { namespace crc32c { @@ -17,9 +17,7 @@ namespace crc32c { 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); -} +inline uint32_t Value(const char* data, size_t n) { return Extend(0, data, n); } static const uint32_t kMaskDelta = 0xa282ead8ul; diff --git a/util/crc32c_test.cc b/util/crc32c_test.cc index 4b957ee..2fe1c41 100644 --- a/util/crc32c_test.cc +++ b/util/crc32c_test.cc @@ -3,13 +3,12 @@ // found in the LICENSE file. See the AUTHORS file for names of contributors. #include "util/crc32c.h" -#include "util/testharness.h" + +#include "gtest/gtest.h" namespace leveldb { namespace crc32c { -class CRC { }; - TEST(CRC, StandardResults) { // From rfc3720 section B.4. char buf[32]; @@ -30,30 +29,19 @@ TEST(CRC, StandardResults) { } ASSERT_EQ(0x113fdb5c, Value(buf, sizeof(buf))); - unsigned char 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, + 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, Values) { ASSERT_NE(Value("a", 1), Value("foo", 3)); } TEST(CRC, Extend) { - ASSERT_EQ(Value("hello world", 11), - Extend(Value("hello ", 6), "world", 5)); + ASSERT_EQ(Value("hello world", 11), Extend(Value("hello ", 6), "world", 5)); } TEST(CRC, Mask) { @@ -66,7 +54,3 @@ TEST(CRC, Mask) { } // namespace crc32c } // namespace leveldb - -int main(int argc, char** argv) { - return leveldb::test::RunAllTests(); -} diff --git a/util/env.cc b/util/env.cc index c58a082..a53b230 100644 --- a/util/env.cc +++ b/util/env.cc @@ -4,33 +4,43 @@ #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() { -} +Env::Env() = default; + +Env::~Env() = default; Status Env::NewAppendableFile(const std::string& fname, WritableFile** result) { return Status::NotSupported("NewAppendableFile", fname); } -SequentialFile::~SequentialFile() { -} +Status Env::RemoveDir(const std::string& dirname) { return DeleteDir(dirname); } +Status Env::DeleteDir(const std::string& dirname) { return RemoveDir(dirname); } -RandomAccessFile::~RandomAccessFile() { -} +Status Env::RemoveFile(const std::string& fname) { return DeleteFile(fname); } +Status Env::DeleteFile(const std::string& fname) { return RemoveFile(fname); } -WritableFile::~WritableFile() { -} +SequentialFile::~SequentialFile() = default; -Logger::~Logger() { -} +RandomAccessFile::~RandomAccessFile() = default; -FileLock::~FileLock() { -} +WritableFile::~WritableFile() = default; + +Logger::~Logger() = default; + +FileLock::~FileLock() = default; void Log(Logger* info_log, const char* format, ...) { - if (info_log != NULL) { - va_list ap; + if (info_log != nullptr) { + std::va_list ap; va_start(ap, format); info_log->Logv(format, ap); va_end(ap); @@ -38,8 +48,7 @@ void Log(Logger* info_log, const char* format, ...) { } static Status DoWriteStringToFile(Env* env, const Slice& data, - const std::string& fname, - bool should_sync) { + const std::string& fname, bool should_sync) { WritableFile* file; Status s = env->NewWritableFile(fname, &file); if (!s.ok()) { @@ -54,7 +63,7 @@ static Status DoWriteStringToFile(Env* env, const Slice& data, } delete file; // Will auto-close if we did not close above if (!s.ok()) { - env->DeleteFile(fname); + env->RemoveFile(fname); } return s; } @@ -94,7 +103,6 @@ Status ReadFileToString(Env* env, const std::string& fname, std::string* data) { return s; } -EnvWrapper::~EnvWrapper() { -} +EnvWrapper::~EnvWrapper() {} } // namespace leveldb diff --git a/util/env_posix.cc b/util/env_posix.cc index 4bfaf6c..ffd06c4 100644 --- a/util/env_posix.cc +++ b/util/env_posix.cc @@ -3,747 +3,924 @@ // found in the LICENSE file. See the AUTHORS file for names of contributors. #include <dirent.h> -#include <errno.h> #include <fcntl.h> -#include <pthread.h> -#include <stdio.h> -#include <stdlib.h> -#include <string.h> #include <sys/mman.h> +#ifndef __Fuchsia__ #include <sys/resource.h> +#endif #include <sys/stat.h> #include <sys/time.h> #include <sys/types.h> -#include <time.h> #include <unistd.h> -#include <deque> + +#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/logging.h" -#include "util/mutexlock.h" -#include "util/posix_logger.h" #include "util/env_posix_test_helper.h" +#include "util/posix_logger.h" namespace leveldb { namespace { -static int open_read_only_file_limit = -1; -static int mmap_limit = -1; +// 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; -static const size_t kBufSize = 65536; +// Can be set using EnvPosixTestHelper::SetReadOnlyMMapLimit(). +int g_mmap_limit = kDefaultMmapLimit; -static Status PosixError(const std::string& context, int err_number) { - if (err_number == ENOENT) { - return Status::NotFound(context, strerror(err_number)); +// 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, strerror(err_number)); + 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 end up running out of file descriptors, virtual memory, -// or running into kernel performance problems for very large databases. +// 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 |n|. - Limiter(intptr_t n) { - SetAllowed(n); + // Limit maximum number of resources to |max_acquires|. + Limiter(int max_acquires) + : +#if !defined(NDEBUG) + max_acquires_(max_acquires), +#endif // !defined(NDEBUG) + acquires_allowed_(max_acquires) { + assert(max_acquires >= 0); } + Limiter(const Limiter&) = delete; + Limiter operator=(const Limiter&) = delete; + // If another resource is available, acquire it and return true. // Else return false. - bool Acquire() LOCKS_EXCLUDED(mu_) { - if (GetAllowed() <= 0) { - return false; - } - MutexLock l(&mu_); - intptr_t x = GetAllowed(); - if (x <= 0) { - return false; - } else { - SetAllowed(x - 1); - return true; - } - } + bool Acquire() { + int old_acquires_allowed = + acquires_allowed_.fetch_sub(1, std::memory_order_relaxed); - // Release a resource acquired by a previous call to Acquire() that returned - // true. - void Release() LOCKS_EXCLUDED(mu_) { - MutexLock l(&mu_); - SetAllowed(GetAllowed() + 1); - } + if (old_acquires_allowed > 0) return true; - private: - port::Mutex mu_; - port::AtomicPointer allowed_; + int pre_increment_acquires_allowed = + acquires_allowed_.fetch_add(1, std::memory_order_relaxed); - intptr_t GetAllowed() const { - return reinterpret_cast<intptr_t>(allowed_.Acquire_Load()); - } + // Silence compiler warnings about unused arguments when NDEBUG is defined. + (void)pre_increment_acquires_allowed; + // If the check below fails, Release() was called more times than acquire. + assert(pre_increment_acquires_allowed < max_acquires_); - void SetAllowed(intptr_t v) EXCLUSIVE_LOCKS_REQUIRED(mu_) { - allowed_.Release_Store(reinterpret_cast<void*>(v)); + return false; } - Limiter(const Limiter&); - void operator=(const Limiter&); -}; + // Release a resource acquired by a previous call to Acquire() that returned + // true. + void Release() { + int old_acquires_allowed = + acquires_allowed_.fetch_add(1, std::memory_order_relaxed); + + // Silence compiler warnings about unused arguments when NDEBUG is defined. + (void)old_acquires_allowed; + // If the check below fails, Release() was called more times than acquire. + assert(old_acquires_allowed < max_acquires_); + } -class PosixSequentialFile: public SequentialFile { private: - std::string filename_; - int fd_; +#if !defined(NDEBUG) + // Catches an excessive number of Release() calls. + const int max_acquires_; +#endif // !defined(NDEBUG) + + // 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(const std::string& fname, int fd) - : filename_(fname), fd_(fd) {} - virtual ~PosixSequentialFile() { close(fd_); } + PosixSequentialFile(std::string filename, int fd) + : fd_(fd), filename_(std::move(filename)) {} + ~PosixSequentialFile() override { close(fd_); } - virtual Status Read(size_t n, Slice* result, char* scratch) { - Status s; + Status Read(size_t n, Slice* result, char* scratch) override { + Status status; while (true) { - ssize_t r = read(fd_, scratch, n); - if (r < 0) { + ::ssize_t read_size = ::read(fd_, scratch, n); + if (read_size < 0) { // Read error. if (errno == EINTR) { continue; // Retry } - s = PosixError(filename_, errno); + status = PosixError(filename_, errno); break; } - *result = Slice(scratch, r); + *result = Slice(scratch, read_size); break; } - return s; + return status; } - virtual Status Skip(uint64_t n) { - if (lseek(fd_, n, SEEK_CUR) == static_cast<off_t>(-1)) { + Status Skip(uint64_t n) override { + if (::lseek(fd_, n, SEEK_CUR) == static_cast<off_t>(-1)) { return PosixError(filename_, errno); } return Status::OK(); } -}; -// pread() based random-access -class PosixRandomAccessFile: public RandomAccessFile { private: - std::string filename_; - bool temporary_fd_; // If true, fd_ is -1 and we open on every read. - int fd_; - Limiter* limiter_; + 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: - PosixRandomAccessFile(const std::string& fname, int fd, Limiter* limiter) - : filename_(fname), fd_(fd), limiter_(limiter) { - temporary_fd_ = !limiter->Acquire(); - if (temporary_fd_) { - // Open file on every access. - close(fd_); - fd_ = -1; + // 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. } } - virtual ~PosixRandomAccessFile() { - if (!temporary_fd_) { - close(fd_); - limiter_->Release(); + ~PosixRandomAccessFile() override { + if (has_permanent_fd_) { + assert(fd_ != -1); + ::close(fd_); + fd_limiter_->Release(); } } - virtual Status Read(uint64_t offset, size_t n, Slice* result, - char* scratch) const { + Status Read(uint64_t offset, size_t n, Slice* result, + char* scratch) const override { int fd = fd_; - if (temporary_fd_) { - fd = open(filename_.c_str(), O_RDONLY); + if (!has_permanent_fd_) { + fd = ::open(filename_.c_str(), O_RDONLY | kOpenBaseFlags); if (fd < 0) { return PosixError(filename_, errno); } } - Status s; - ssize_t r = pread(fd, scratch, n, static_cast<off_t>(offset)); - *result = Slice(scratch, (r < 0) ? 0 : r); - if (r < 0) { - // An error: return a non-ok status - s = 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 (temporary_fd_) { + if (!has_permanent_fd_) { // Close the temporary file descriptor opened earlier. - close(fd); + assert(fd != fd_); + ::close(fd); } - return s; + return status; } -}; -// mmap() based random-access -class PosixMmapReadableFile: public RandomAccessFile { private: - std::string filename_; - void* mmapped_region_; - size_t length_; - Limiter* limiter_; + 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: - // base[0,length-1] contains the mmapped contents of the file. - PosixMmapReadableFile(const std::string& fname, void* base, size_t length, - Limiter* limiter) - : filename_(fname), mmapped_region_(base), length_(length), - limiter_(limiter) { - } - - virtual ~PosixMmapReadableFile() { - munmap(mmapped_region_, length_); - limiter_->Release(); - } - - virtual Status Read(uint64_t offset, size_t n, Slice* result, - char* scratch) const { - Status s; + // 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 + // acquired 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(); - s = PosixError(filename_, EINVAL); - } else { - *result = Slice(reinterpret_cast<char*>(mmapped_region_) + offset, n); + return PosixError(filename_, EINVAL); } - return s; + + *result = Slice(mmap_base_ + offset, n); + return Status::OK(); } -}; -class PosixWritableFile : public WritableFile { private: - // buf_[0, pos_-1] contains data to be written to fd_. - std::string filename_; - int fd_; - char buf_[kBufSize]; - size_t pos_; + char* const mmap_base_; + const size_t length_; + Limiter* const mmap_limiter_; + const std::string filename_; +}; +class PosixWritableFile final : public WritableFile { public: - PosixWritableFile(const std::string& fname, int fd) - : filename_(fname), fd_(fd), pos_(0) { } - - ~PosixWritableFile() { + 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(); } } - virtual Status Append(const Slice& data) { - size_t n = data.size(); - const char* p = data.data(); + 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 = std::min(n, kBufSize - pos_); - memcpy(buf_ + pos_, p, copy); - p += copy; - n -= copy; - pos_ += copy; - if (n == 0) { + 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 s = FlushBuffered(); - if (!s.ok()) { - return s; + Status status = FlushBuffer(); + if (!status.ok()) { + return status; } // Small writes go to buffer, large writes are written directly. - if (n < kBufSize) { - memcpy(buf_, p, n); - pos_ = n; + if (write_size < kWritableFileBufferSize) { + std::memcpy(buf_, write_data, write_size); + pos_ = write_size; return Status::OK(); } - return WriteRaw(p, n); + return WriteUnbuffered(write_data, write_size); } - virtual Status Close() { - Status result = FlushBuffered(); - const int r = close(fd_); - if (r < 0 && result.ok()) { - result = PosixError(filename_, errno); + 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 result; + return status; } - virtual Status Flush() { - return FlushBuffered(); - } + Status Flush() override { return FlushBuffer(); } - Status SyncDirIfManifest() { - const char* f = filename_.c_str(); - const char* sep = strrchr(f, '/'); - Slice basename; - std::string dir; - if (sep == NULL) { - dir = "."; - basename = f; - } else { - dir = std::string(f, sep - f); - basename = sep + 1; - } - Status s; - if (basename.starts_with("MANIFEST")) { - int fd = open(dir.c_str(), O_RDONLY); - if (fd < 0) { - s = PosixError(dir, errno); - } else { - if (fsync(fd) < 0) { - s = PosixError(dir, errno); - } - close(fd); - } + 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; } - return s; - } - virtual Status Sync() { - // Ensure new files referred to by the manifest are in the filesystem. - Status s = SyncDirIfManifest(); - if (!s.ok()) { - return s; - } - s = FlushBuffered(); - if (s.ok()) { - if (fdatasync(fd_) != 0) { - s = PosixError(filename_, errno); - } + status = FlushBuffer(); + if (!status.ok()) { + return status; } - return s; + + return SyncFd(fd_, filename_); } private: - Status FlushBuffered() { - Status s = WriteRaw(buf_, pos_); + Status FlushBuffer() { + Status status = WriteUnbuffered(buf_, pos_); pos_ = 0; - return s; + return status; } - Status WriteRaw(const char* p, size_t n) { - while (n > 0) { - ssize_t r = write(fd_, p, n); - if (r < 0) { + 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); } - p += r; - n -= r; + 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_. }; -static int LockOrUnlock(int fd, bool lock) { +int LockOrUnlock(int fd, bool lock) { errno = 0; - struct flock f; - memset(&f, 0, sizeof(f)); - f.l_type = (lock ? F_WRLCK : F_UNLCK); - f.l_whence = SEEK_SET; - f.l_start = 0; - f.l_len = 0; // Lock/unlock entire file - return fcntl(fd, F_SETLK, &f); + 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: - int fd_; - std::string name_; + 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_; }; -// Set of locked files. We keep a separate set instead of just -// relying on fcntrl(F_SETLK) since fcntl(F_SETLK) does not provide -// any protection against multiple uses from the same process. +// 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 { - private: - port::Mutex mu_; - std::set<std::string> locked_files_ GUARDED_BY(mu_); public: bool Insert(const std::string& fname) LOCKS_EXCLUDED(mu_) { - MutexLock l(&mu_); - return locked_files_.insert(fname).second; + mu_.Lock(); + bool succeeded = locked_files_.insert(fname).second; + mu_.Unlock(); + return succeeded; } void Remove(const std::string& fname) LOCKS_EXCLUDED(mu_) { - MutexLock l(&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(); - virtual ~PosixEnv() { - char msg[] = "Destroying Env::Default()\n"; - fwrite(msg, 1, sizeof(msg), stderr); - abort(); + ~PosixEnv() override { + static const char msg[] = + "PosixEnv singleton destroyed. Unsupported behavior!\n"; + std::fwrite(msg, 1, sizeof(msg), stderr); + std::abort(); } - virtual Status NewSequentialFile(const std::string& fname, - SequentialFile** result) { - int fd = open(fname.c_str(), O_RDONLY); + Status NewSequentialFile(const std::string& filename, + SequentialFile** result) override { + int fd = ::open(filename.c_str(), O_RDONLY | kOpenBaseFlags); if (fd < 0) { - *result = NULL; - return PosixError(fname, errno); - } else { - *result = new PosixSequentialFile(fname, fd); - return Status::OK(); + *result = nullptr; + return PosixError(filename, errno); } + + *result = new PosixSequentialFile(filename, fd); + return Status::OK(); } - virtual Status NewRandomAccessFile(const std::string& fname, - RandomAccessFile** result) { - *result = NULL; - Status s; - int fd = open(fname.c_str(), O_RDONLY); + Status NewRandomAccessFile(const std::string& filename, + RandomAccessFile** result) override { + *result = nullptr; + int fd = ::open(filename.c_str(), O_RDONLY | kOpenBaseFlags); if (fd < 0) { - s = PosixError(fname, errno); - } else if (mmap_limit_.Acquire()) { - uint64_t size; - s = GetFileSize(fname, &size); - if (s.ok()) { - void* base = mmap(NULL, size, PROT_READ, MAP_SHARED, fd, 0); - if (base != MAP_FAILED) { - *result = new PosixMmapReadableFile(fname, base, size, &mmap_limit_); - } else { - s = PosixError(fname, errno); - } - } - close(fd); - if (!s.ok()) { - mmap_limit_.Release(); + 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); } - } else { - *result = new PosixRandomAccessFile(fname, fd, &fd_limit_); } - return s; + ::close(fd); + if (!status.ok()) { + mmap_limiter_.Release(); + } + return status; } - virtual Status NewWritableFile(const std::string& fname, - WritableFile** result) { - Status s; - int fd = open(fname.c_str(), O_TRUNC | O_WRONLY | O_CREAT, 0644); + 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 = NULL; - s = PosixError(fname, errno); - } else { - *result = new PosixWritableFile(fname, fd); + *result = nullptr; + return PosixError(filename, errno); } - return s; + + *result = new PosixWritableFile(filename, fd); + return Status::OK(); } - virtual Status NewAppendableFile(const std::string& fname, - WritableFile** result) { - Status s; - int fd = open(fname.c_str(), O_APPEND | O_WRONLY | O_CREAT, 0644); + 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 = NULL; - s = PosixError(fname, errno); - } else { - *result = new PosixWritableFile(fname, fd); + *result = nullptr; + return PosixError(filename, errno); } - return s; + + *result = new PosixWritableFile(filename, fd); + return Status::OK(); } - virtual bool FileExists(const std::string& fname) { - return access(fname.c_str(), F_OK) == 0; + bool FileExists(const std::string& filename) override { + return ::access(filename.c_str(), F_OK) == 0; } - virtual Status GetChildren(const std::string& dir, - std::vector<std::string>* result) { + Status GetChildren(const std::string& directory_path, + std::vector<std::string>* result) override { result->clear(); - DIR* d = opendir(dir.c_str()); - if (d == NULL) { - return PosixError(dir, errno); + ::DIR* dir = ::opendir(directory_path.c_str()); + if (dir == nullptr) { + return PosixError(directory_path, errno); } - struct dirent* entry; - while ((entry = readdir(d)) != NULL) { - result->push_back(entry->d_name); + struct ::dirent* entry; + while ((entry = ::readdir(dir)) != nullptr) { + result->emplace_back(entry->d_name); } - closedir(d); + ::closedir(dir); return Status::OK(); } - virtual Status DeleteFile(const std::string& fname) { - Status result; - if (unlink(fname.c_str()) != 0) { - result = PosixError(fname, errno); + Status RemoveFile(const std::string& filename) override { + if (::unlink(filename.c_str()) != 0) { + return PosixError(filename, errno); } - return result; + return Status::OK(); } - virtual Status CreateDir(const std::string& name) { - Status result; - if (mkdir(name.c_str(), 0755) != 0) { - result = PosixError(name, errno); + Status CreateDir(const std::string& dirname) override { + if (::mkdir(dirname.c_str(), 0755) != 0) { + return PosixError(dirname, errno); } - return result; + return Status::OK(); } - virtual Status DeleteDir(const std::string& name) { - Status result; - if (rmdir(name.c_str()) != 0) { - result = PosixError(name, errno); + Status RemoveDir(const std::string& dirname) override { + if (::rmdir(dirname.c_str()) != 0) { + return PosixError(dirname, errno); } - return result; + return Status::OK(); } - virtual Status GetFileSize(const std::string& fname, uint64_t* size) { - Status s; - struct stat sbuf; - if (stat(fname.c_str(), &sbuf) != 0) { + Status GetFileSize(const std::string& filename, uint64_t* size) override { + struct ::stat file_stat; + if (::stat(filename.c_str(), &file_stat) != 0) { *size = 0; - s = PosixError(fname, errno); - } else { - *size = sbuf.st_size; + return PosixError(filename, errno); } - return s; + *size = file_stat.st_size; + return Status::OK(); } - virtual Status RenameFile(const std::string& src, const std::string& target) { - Status result; - if (rename(src.c_str(), target.c_str()) != 0) { - result = PosixError(src, errno); + 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 result; + return Status::OK(); } - virtual Status LockFile(const std::string& fname, FileLock** lock) { - *lock = NULL; - Status result; - int fd = open(fname.c_str(), O_RDWR | O_CREAT, 0644); + 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) { - result = PosixError(fname, errno); - } else if (!locks_.Insert(fname)) { - close(fd); - result = Status::IOError("lock " + fname, "already held by process"); - } else if (LockOrUnlock(fd, true) == -1) { - result = PosixError("lock " + fname, errno); - close(fd); - locks_.Remove(fname); - } else { - PosixFileLock* my_lock = new PosixFileLock; - my_lock->fd_ = fd; - my_lock->name_ = fname; - *lock = my_lock; + 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); } - return result; + + *lock = new PosixFileLock(fd, filename); + return Status::OK(); } - virtual Status UnlockFile(FileLock* lock) { - PosixFileLock* my_lock = reinterpret_cast<PosixFileLock*>(lock); - Status result; - if (LockOrUnlock(my_lock->fd_, false) == -1) { - result = PosixError("unlock", errno); + 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(my_lock->name_); - close(my_lock->fd_); - delete my_lock; - return result; + locks_.Remove(posix_file_lock->filename()); + ::close(posix_file_lock->fd()); + delete posix_file_lock; + return Status::OK(); } - virtual void Schedule(void (*function)(void*), void* arg); + void Schedule(void (*background_work_function)(void* background_work_arg), + void* background_work_arg) override; - virtual void StartThread(void (*function)(void* arg), void* arg); + 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(); + } - virtual Status GetTestDirectory(std::string* result) { - const char* env = getenv("TEST_TMPDIR"); + Status GetTestDirectory(std::string* result) override { + const char* env = std::getenv("TEST_TMPDIR"); if (env && env[0] != '\0') { *result = env; } else { char buf[100]; - snprintf(buf, sizeof(buf), "/tmp/leveldbtest-%d", int(geteuid())); + std::snprintf(buf, sizeof(buf), "/tmp/leveldbtest-%d", + static_cast<int>(::geteuid())); *result = buf; } - // Directory may already exist + + // The CreateDir status is ignored because the directory may already exist. CreateDir(*result); + return Status::OK(); } - static uint64_t gettid() { - pthread_t tid = pthread_self(); - uint64_t thread_id = 0; - memcpy(&thread_id, &tid, std::min(sizeof(thread_id), sizeof(tid))); - return thread_id; - } + 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); + } - virtual Status NewLogger(const std::string& fname, Logger** result) { - FILE* f = fopen(fname.c_str(), "w"); - if (f == NULL) { - *result = NULL; - return PosixError(fname, errno); + std::FILE* fp = ::fdopen(fd, "w"); + if (fp == nullptr) { + ::close(fd); + *result = nullptr; + return PosixError(filename, errno); } else { - *result = new PosixLogger(f, &PosixEnv::gettid); + *result = new PosixLogger(fp); return Status::OK(); } } - virtual uint64_t NowMicros() { - struct timeval tv; - gettimeofday(&tv, NULL); - return static_cast<uint64_t>(tv.tv_sec) * 1000000 + tv.tv_usec; + 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; } - virtual void SleepForMicroseconds(int micros) { - usleep(micros); + void SleepForMicroseconds(int micros) override { + std::this_thread::sleep_for(std::chrono::microseconds(micros)); } private: - void PthreadCall(const char* label, int result) { - if (result != 0) { - fprintf(stderr, "pthread %s: %s\n", label, strerror(result)); - abort(); - } - } + void BackgroundThreadMain(); - // BGThread() is the body of the background thread - void BGThread(); - static void* BGThreadWrapper(void* arg) { - reinterpret_cast<PosixEnv*>(arg)->BGThread(); - return NULL; + static void BackgroundThreadEntryPoint(PosixEnv* env) { + env->BackgroundThreadMain(); } - pthread_mutex_t mu_; - pthread_cond_t bgsignal_; - pthread_t bgthread_; - bool started_bgthread_; + // 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 because 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_); - // Entry per Schedule() call - struct BGItem { void* arg; void (*function)(void*); }; - typedef std::deque<BGItem> BGQueue; - BGQueue queue_; + std::queue<BackgroundWorkItem> background_work_queue_ + GUARDED_BY(background_work_mutex_); - PosixLockTable locks_; - Limiter mmap_limit_; - Limiter fd_limit_; + PosixLockTable locks_; // Thread-safe. + Limiter mmap_limiter_; // Thread-safe. + Limiter fd_limiter_; // Thread-safe. }; // Return the maximum number of concurrent mmaps. -static int MaxMmaps() { - if (mmap_limit >= 0) { - return mmap_limit; - } - // Up to 1000 mmaps for 64-bit binaries; none for smaller pointer sizes. - mmap_limit = sizeof(void*) >= 8 ? 1000 : 0; - return mmap_limit; -} +int MaxMmaps() { return g_mmap_limit; } // Return the maximum number of read-only files to keep open. -static intptr_t MaxOpenFiles() { - if (open_read_only_file_limit >= 0) { - return open_read_only_file_limit; - } - struct rlimit rlim; - if (getrlimit(RLIMIT_NOFILE, &rlim)) { +int MaxOpenFiles() { + if (g_open_read_only_file_limit >= 0) { + return g_open_read_only_file_limit; + } +#ifdef __Fuchsia__ + // Fuchsia doesn't implement getrlimit. + g_open_read_only_file_limit = 50; +#else + struct ::rlimit rlim; + if (::getrlimit(RLIMIT_NOFILE, &rlim)) { // getrlimit failed, fallback to hard-coded default. - open_read_only_file_limit = 50; + g_open_read_only_file_limit = 50; } else if (rlim.rlim_cur == RLIM_INFINITY) { - open_read_only_file_limit = std::numeric_limits<int>::max(); + g_open_read_only_file_limit = std::numeric_limits<int>::max(); } else { // Allow use of 20% of available file descriptors for read-only files. - open_read_only_file_limit = rlim.rlim_cur / 5; + g_open_read_only_file_limit = rlim.rlim_cur / 5; } - return open_read_only_file_limit; +#endif + return g_open_read_only_file_limit; } +} // namespace + PosixEnv::PosixEnv() - : started_bgthread_(false), - mmap_limit_(MaxMmaps()), - fd_limit_(MaxOpenFiles()) { - PthreadCall("mutex_init", pthread_mutex_init(&mu_, NULL)); - PthreadCall("cvar_init", pthread_cond_init(&bgsignal_, NULL)); -} + : background_work_cv_(&background_work_mutex_), + started_background_thread_(false), + mmap_limiter_(MaxMmaps()), + fd_limiter_(MaxOpenFiles()) {} -void PosixEnv::Schedule(void (*function)(void*), void* arg) { - PthreadCall("lock", pthread_mutex_lock(&mu_)); +void PosixEnv::Schedule( + void (*background_work_function)(void* background_work_arg), + void* background_work_arg) { + background_work_mutex_.Lock(); - // Start background thread if necessary - if (!started_bgthread_) { - started_bgthread_ = true; - PthreadCall( - "create thread", - pthread_create(&bgthread_, NULL, &PosixEnv::BGThreadWrapper, this)); + // 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 currently empty, the background thread may currently be - // waiting. - if (queue_.empty()) { - PthreadCall("signal", pthread_cond_signal(&bgsignal_)); + // If the queue is empty, the background thread may be waiting for work. + if (background_work_queue_.empty()) { + background_work_cv_.Signal(); } - // Add to priority queue - queue_.push_back(BGItem()); - queue_.back().function = function; - queue_.back().arg = arg; - - PthreadCall("unlock", pthread_mutex_unlock(&mu_)); + background_work_queue_.emplace(background_work_function, background_work_arg); + background_work_mutex_.Unlock(); } -void PosixEnv::BGThread() { +void PosixEnv::BackgroundThreadMain() { while (true) { - // Wait until there is an item that is ready to run - PthreadCall("lock", pthread_mutex_lock(&mu_)); - while (queue_.empty()) { - PthreadCall("wait", pthread_cond_wait(&bgsignal_, &mu_)); + background_work_mutex_.Lock(); + + // Wait until there is work to be done. + while (background_work_queue_.empty()) { + background_work_cv_.Wait(); } - void (*function)(void*) = queue_.front().function; - void* arg = queue_.front().arg; - queue_.pop_front(); + 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(); - PthreadCall("unlock", pthread_mutex_unlock(&mu_)); - (*function)(arg); + background_work_mutex_.Unlock(); + background_work_function(background_work_arg); } } namespace { -struct StartThreadState { - void (*user_function)(void*); - void* 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_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_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) }; -} -static void* StartThreadWrapper(void* arg) { - StartThreadState* state = reinterpret_cast<StartThreadState*>(arg); - state->user_function(state->arg); - delete state; - return NULL; -} -void PosixEnv::StartThread(void (*function)(void* arg), void* arg) { - pthread_t t; - StartThreadState* state = new StartThreadState; - state->user_function = function; - state->arg = arg; - PthreadCall("start thread", - pthread_create(&t, NULL, &StartThreadWrapper, state)); -} +#if !defined(NDEBUG) +template <typename EnvType> +std::atomic<bool> SingletonEnv<EnvType>::env_initialized_; +#endif // !defined(NDEBUG) -} // namespace +using PosixDefaultEnv = SingletonEnv<PosixEnv>; -static pthread_once_t once = PTHREAD_ONCE_INIT; -static Env* default_env; -static void InitDefaultEnv() { default_env = new PosixEnv; } +} // namespace void EnvPosixTestHelper::SetReadOnlyFDLimit(int limit) { - assert(default_env == NULL); - open_read_only_file_limit = limit; + PosixDefaultEnv::AssertEnvNotInitialized(); + g_open_read_only_file_limit = limit; } void EnvPosixTestHelper::SetReadOnlyMMapLimit(int limit) { - assert(default_env == NULL); - mmap_limit = limit; + PosixDefaultEnv::AssertEnvNotInitialized(); + g_mmap_limit = limit; } Env* Env::Default() { - pthread_once(&once, InitDefaultEnv); - return default_env; + static PosixDefaultEnv env_container; + return env_container.env(); } } // namespace leveldb diff --git a/util/env_posix_test.cc b/util/env_posix_test.cc index 295f8ae..34bda62 100644 --- a/util/env_posix_test.cc +++ b/util/env_posix_test.cc @@ -2,65 +2,352 @@ // 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 <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/testharness.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 kDelayMicros = 100000; static const int kReadOnlyFileLimit = 4; static const int kMMapLimit = 4; -class EnvPosixTest { +class EnvPosixTest : public testing::Test { public: - Env* env_; - EnvPosixTest() : env_(Env::Default()) { } - 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(EnvPosixTest, TestOpenOnRead) { +TEST_F(EnvPosixTest, TestOpenOnRead) { // Write some test data to a single file that will be opened |n| times. std::string test_dir; - ASSERT_OK(env_->GetTestDirectory(&test_dir)); + ASSERT_LEVELDB_OK(env_->GetTestDirectory(&test_dir)); std::string test_file = test_dir + "/open_on_read.txt"; - FILE* f = fopen(test_file.c_str(), "w"); - ASSERT_TRUE(f != NULL); + FILE* f = std::fopen(test_file.c_str(), "we"); + ASSERT_TRUE(f != nullptr); const char kFileData[] = "abcdefghijklmnopqrstuvwxyz"; fputs(kFileData, f); - fclose(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_OK(env_->NewRandomAccessFile(test_file, &files[i])); + ASSERT_LEVELDB_OK(env_->NewRandomAccessFile(test_file, &files[i])); } char scratch; Slice read_result; for (int i = 0; i < kNumFiles; i++) { - ASSERT_OK(files[i]->Read(i, 1, &read_result, &scratch)); + 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_OK(env_->DeleteFile(test_file)); + 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[kMMapLimit]; + for (int i = 0; i < kMMapLimit; 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 < kMMapLimit; 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); - return leveldb::test::RunAllTests(); + + testing::InitGoogleTest(&argc, argv); + return RUN_ALL_TESTS(); } diff --git a/util/env_test.cc b/util/env_test.cc index fd89b4c..47174f5 100644 --- a/util/env_test.cc +++ b/util/env_test.cc @@ -6,38 +6,30 @@ #include <algorithm> +#include "gtest/gtest.h" #include "port/port.h" #include "port/thread_annotations.h" #include "util/mutexlock.h" -#include "util/testharness.h" #include "util/testutil.h" namespace leveldb { -static const int kDelayMicros = 100000; -static const int kReadOnlyFileLimit = 4; -static const int kMMapLimit = 4; - -class EnvTest { +class EnvTest : public testing::Test { public: + EnvTest() : env_(Env::Default()) {} + Env* env_; - EnvTest() : env_(Env::Default()) { } }; -static void SetBool(void* ptr) { - reinterpret_cast<port::AtomicPointer*>(ptr)->NoBarrier_Store(ptr); -} - - -TEST(EnvTest, ReadWrite) { +TEST_F(EnvTest, ReadWrite) { Random rnd(test::RandomSeed()); // Get file to use for testing. std::string test_dir; - ASSERT_OK(env_->GetTestDirectory(&test_dir)); + ASSERT_LEVELDB_OK(env_->GetTestDirectory(&test_dir)); std::string test_file_name = test_dir + "/open_on_read.txt"; WritableFile* writable_file; - ASSERT_OK(env_->NewWritableFile(test_file_name, &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; @@ -46,26 +38,26 @@ TEST(EnvTest, ReadWrite) { int len = rnd.Skewed(18); // Up to 2^18 - 1, but typically much smaller std::string r; test::RandomString(&rnd, len, &r); - ASSERT_OK(writable_file->Append(r)); + ASSERT_LEVELDB_OK(writable_file->Append(r)); data += r; if (rnd.OneIn(10)) { - ASSERT_OK(writable_file->Flush()); + ASSERT_LEVELDB_OK(writable_file->Flush()); } } - ASSERT_OK(writable_file->Sync()); - ASSERT_OK(writable_file->Close()); + 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_OK(env_->NewSequentialFile(test_file_name, &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_OK(sequential_file->Read(len, &read, &scratch[0])); + ASSERT_LEVELDB_OK(sequential_file->Read(len, &read, &scratch[0])); if (len > 0) { ASSERT_GT(read.size(), 0); } @@ -76,51 +68,78 @@ TEST(EnvTest, ReadWrite) { delete sequential_file; } -TEST(EnvTest, RunImmediately) { - port::AtomicPointer called (NULL); - env_->Schedule(&SetBool, &called); - env_->SleepForMicroseconds(kDelayMicros); - ASSERT_TRUE(called.NoBarrier_Load() != NULL); +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(EnvTest, RunMany) { - port::AtomicPointer last_id (NULL); +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. - struct CB { - port::AtomicPointer* last_id_ptr; // Pointer to shared slot - uintptr_t id; // Order# for the execution of this callback + Callback(RunState* s, int id) : state_(s), id_(id) {} - CB(port::AtomicPointer* p, int i) : last_id_ptr(p), id(i) { } + static void Run(void* arg) { + Callback* callback = reinterpret_cast<Callback*>(arg); + RunState* state = callback->state_; - static void Run(void* v) { - CB* cb = reinterpret_cast<CB*>(v); - void* cur = cb->last_id_ptr->NoBarrier_Load(); - ASSERT_EQ(cb->id-1, reinterpret_cast<uintptr_t>(cur)); - cb->last_id_ptr->Release_Store(reinterpret_cast<void*>(cb->id)); + MutexLock l(&state->mu); + ASSERT_EQ(state->last_id, callback->id_ - 1); + state->last_id = callback->id_; + state->cvar.Signal(); } }; - // Schedule in different order than start time - CB cb1(&last_id, 1); - CB cb2(&last_id, 2); - CB cb3(&last_id, 3); - CB cb4(&last_id, 4); - env_->Schedule(&CB::Run, &cb1); - env_->Schedule(&CB::Run, &cb2); - env_->Schedule(&CB::Run, &cb3); - env_->Schedule(&CB::Run, &cb4); - - env_->SleepForMicroseconds(kDelayMicros); - void* cur = last_id.Acquire_Load(); - ASSERT_EQ(4, reinterpret_cast<uintptr_t>(cur)); + 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) { } + State(int val, int num_running) : val(val), num_running(num_running) {} }; static void ThreadBody(void* arg) { @@ -128,39 +147,34 @@ static void ThreadBody(void* arg) { s->mu.Lock(); s->val += 1; s->num_running -= 1; + s->cvar.Signal(); s->mu.Unlock(); } -TEST(EnvTest, StartThread) { +TEST_F(EnvTest, StartThread) { State state(0, 3); for (int i = 0; i < 3; i++) { env_->StartThread(&ThreadBody, &state); } - while (true) { - state.mu.Lock(); - int num = state.num_running; - state.mu.Unlock(); - if (num == 0) { - break; - } - env_->SleepForMicroseconds(kDelayMicros); - } MutexLock l(&state.mu); + while (state.num_running != 0) { + state.cvar.Wait(); + } ASSERT_EQ(state.val, 3); } -TEST(EnvTest, TestOpenNonExistentFile) { +TEST_F(EnvTest, TestOpenNonExistentFile) { // Write some test data to a single file that will be opened |n| times. std::string test_dir; - ASSERT_OK(env_->GetTestDirectory(&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); + Status status = + env_->NewRandomAccessFile(non_existent_file, &random_access_file); ASSERT_TRUE(status.IsNotFound()); SequentialFile* sequential_file; @@ -168,56 +182,52 @@ TEST(EnvTest, TestOpenNonExistentFile) { ASSERT_TRUE(status.IsNotFound()); } -TEST(EnvTest, ReopenWritableFile) { +TEST_F(EnvTest, ReopenWritableFile) { std::string test_dir; - ASSERT_OK(env_->GetTestDirectory(&test_dir)); + ASSERT_LEVELDB_OK(env_->GetTestDirectory(&test_dir)); std::string test_file_name = test_dir + "/reopen_writable_file.txt"; - env_->DeleteFile(test_file_name); + env_->RemoveFile(test_file_name); WritableFile* writable_file; - ASSERT_OK(env_->NewWritableFile(test_file_name, &writable_file)); + ASSERT_LEVELDB_OK(env_->NewWritableFile(test_file_name, &writable_file)); std::string data("hello world!"); - ASSERT_OK(writable_file->Append(data)); - ASSERT_OK(writable_file->Close()); + ASSERT_LEVELDB_OK(writable_file->Append(data)); + ASSERT_LEVELDB_OK(writable_file->Close()); delete writable_file; - ASSERT_OK(env_->NewWritableFile(test_file_name, &writable_file)); + ASSERT_LEVELDB_OK(env_->NewWritableFile(test_file_name, &writable_file)); data = "42"; - ASSERT_OK(writable_file->Append(data)); - ASSERT_OK(writable_file->Close()); + ASSERT_LEVELDB_OK(writable_file->Append(data)); + ASSERT_LEVELDB_OK(writable_file->Close()); delete writable_file; - ASSERT_OK(ReadFileToString(env_, test_file_name, &data)); + ASSERT_LEVELDB_OK(ReadFileToString(env_, test_file_name, &data)); ASSERT_EQ(std::string("42"), data); - env_->DeleteFile(test_file_name); + env_->RemoveFile(test_file_name); } -TEST(EnvTest, ReopenAppendableFile) { +TEST_F(EnvTest, ReopenAppendableFile) { std::string test_dir; - ASSERT_OK(env_->GetTestDirectory(&test_dir)); + ASSERT_LEVELDB_OK(env_->GetTestDirectory(&test_dir)); std::string test_file_name = test_dir + "/reopen_appendable_file.txt"; - env_->DeleteFile(test_file_name); + env_->RemoveFile(test_file_name); WritableFile* appendable_file; - ASSERT_OK(env_->NewAppendableFile(test_file_name, &appendable_file)); + ASSERT_LEVELDB_OK(env_->NewAppendableFile(test_file_name, &appendable_file)); std::string data("hello world!"); - ASSERT_OK(appendable_file->Append(data)); - ASSERT_OK(appendable_file->Close()); + ASSERT_LEVELDB_OK(appendable_file->Append(data)); + ASSERT_LEVELDB_OK(appendable_file->Close()); delete appendable_file; - ASSERT_OK(env_->NewAppendableFile(test_file_name, &appendable_file)); + ASSERT_LEVELDB_OK(env_->NewAppendableFile(test_file_name, &appendable_file)); data = "42"; - ASSERT_OK(appendable_file->Append(data)); - ASSERT_OK(appendable_file->Close()); + ASSERT_LEVELDB_OK(appendable_file->Append(data)); + ASSERT_LEVELDB_OK(appendable_file->Close()); delete appendable_file; - ASSERT_OK(ReadFileToString(env_, test_file_name, &data)); + ASSERT_LEVELDB_OK(ReadFileToString(env_, test_file_name, &data)); ASSERT_EQ(std::string("hello world!42"), data); - env_->DeleteFile(test_file_name); + env_->RemoveFile(test_file_name); } } // namespace leveldb - -int main(int argc, char** argv) { - return leveldb::test::RunAllTests(); -} diff --git a/util/env_windows.cc b/util/env_windows.cc new file mode 100644 index 0000000..1c74b02 --- /dev/null +++ b/util/env_windows.cc @@ -0,0 +1,816 @@ +// 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) + : +#if !defined(NDEBUG) + max_acquires_(max_acquires), +#endif // !defined(NDEBUG) + acquires_allowed_(max_acquires) { + assert(max_acquires >= 0); + } + + 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() { + int old_acquires_allowed = + acquires_allowed_.fetch_add(1, std::memory_order_relaxed); + + // Silence compiler warnings about unused arguments when NDEBUG is defined. + (void)old_acquires_allowed; + // If the check below fails, Release() was called more times than acquire. + assert(old_acquires_allowed < max_acquires_); + } + + private: +#if !defined(NDEBUG) + // Catches an excessive number of Release() calls. + const int max_acquires_; +#endif // !defined(NDEBUG) + + // 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(), "wN"); + 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 because 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_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_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 diff --git a/util/env_windows_test.cc b/util/env_windows_test.cc new file mode 100644 index 0000000..d6822d2 --- /dev/null +++ b/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(); +} diff --git a/util/env_windows_test_helper.h b/util/env_windows_test_helper.h new file mode 100644 index 0000000..e6f6020 --- /dev/null +++ b/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_ diff --git a/util/filter_policy.cc b/util/filter_policy.cc index 7b045c8..90fd754 100644 --- a/util/filter_policy.cc +++ b/util/filter_policy.cc @@ -6,6 +6,6 @@ namespace leveldb { -FilterPolicy::~FilterPolicy() { } +FilterPolicy::~FilterPolicy() {} } // namespace leveldb diff --git a/util/hash.cc b/util/hash.cc index ed439ce..8122fa8 100644 --- a/util/hash.cc +++ b/util/hash.cc @@ -2,15 +2,19 @@ // 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 <string.h> -#include "util/coding.h" #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) +#define FALLTHROUGH_INTENDED \ + do { \ + } while (0) #endif namespace leveldb { @@ -34,13 +38,13 @@ uint32_t Hash(const char* data, size_t n, uint32_t seed) { // Pick up remaining bytes switch (limit - data) { case 3: - h += static_cast<unsigned char>(data[2]) << 16; + h += static_cast<uint8_t>(data[2]) << 16; FALLTHROUGH_INTENDED; case 2: - h += static_cast<unsigned char>(data[1]) << 8; + h += static_cast<uint8_t>(data[1]) << 8; FALLTHROUGH_INTENDED; case 1: - h += static_cast<unsigned char>(data[0]); + h += static_cast<uint8_t>(data[0]); h *= m; h ^= (h >> r); break; @@ -48,5 +52,4 @@ uint32_t Hash(const char* data, size_t n, uint32_t seed) { return h; } - } // namespace leveldb diff --git a/util/hash.h b/util/hash.h index 74bdb6e..87ab279 100644 --- a/util/hash.h +++ b/util/hash.h @@ -7,8 +7,8 @@ #ifndef STORAGE_LEVELDB_UTIL_HASH_H_ #define STORAGE_LEVELDB_UTIL_HASH_H_ -#include <stddef.h> -#include <stdint.h> +#include <cstddef> +#include <cstdint> namespace leveldb { diff --git a/util/hash_test.cc b/util/hash_test.cc index eaa1c92..0ea5977 100644 --- a/util/hash_test.cc +++ b/util/hash_test.cc @@ -3,30 +3,21 @@ // found in the LICENSE file. See the AUTHORS file for names of contributors. #include "util/hash.h" -#include "util/testharness.h" -namespace leveldb { +#include "gtest/gtest.h" -class HASH { }; +namespace leveldb { TEST(HASH, SignedUnsignedIssue) { - const unsigned char data1[1] = {0x62}; - const unsigned char data2[2] = {0xc3, 0x97}; - const unsigned char data3[3] = {0xe2, 0x99, 0xa5}; - const unsigned char data4[4] = {0xe1, 0x80, 0xb9, 0x32}; - const unsigned char 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, + 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); @@ -48,7 +39,3 @@ TEST(HASH, SignedUnsignedIssue) { } } // namespace leveldb - -int main(int argc, char** argv) { - return leveldb::test::RunAllTests(); -} diff --git a/util/histogram.cc b/util/histogram.cc index bb95f58..7af4030 100644 --- a/util/histogram.cc +++ b/util/histogram.cc @@ -2,36 +2,174 @@ // 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 <math.h> -#include <stdio.h> -#include "port/port.h" #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, + 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]; + min_ = kBucketLimit[kNumBuckets - 1]; max_ = 0; num_ = 0; sum_ = 0; @@ -66,9 +204,7 @@ void Histogram::Merge(const Histogram& other) { } } -double Histogram::Median() const { - return Percentile(50.0); -} +double Histogram::Median() const { return Percentile(50.0); } double Histogram::Percentile(double p) const { double threshold = num_ * (p / 100.0); @@ -77,7 +213,7 @@ double Histogram::Percentile(double p) const { sum += buckets_[b]; if (sum >= threshold) { // Scale linearly within this bucket - double left_point = (b == 0) ? 0 : kBucketLimit[b-1]; + double left_point = (b == 0) ? 0 : kBucketLimit[b - 1]; double right_point = kBucketLimit[b]; double left_sum = sum - buckets_[b]; double right_sum = sum; @@ -105,13 +241,11 @@ double Histogram::StandardDeviation() const { std::string Histogram::ToString() const { std::string r; char buf[200]; - snprintf(buf, sizeof(buf), - "Count: %.0f Average: %.4f StdDev: %.2f\n", - num_, Average(), StandardDeviation()); + std::snprintf(buf, sizeof(buf), "Count: %.0f Average: %.4f StdDev: %.2f\n", + num_, Average(), StandardDeviation()); r.append(buf); - snprintf(buf, sizeof(buf), - "Min: %.4f Median: %.4f Max: %.4f\n", - (num_ == 0.0 ? 0.0 : min_), Median(), max_); + 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_; @@ -119,17 +253,16 @@ std::string Histogram::ToString() const { for (int b = 0; b < kNumBuckets; b++) { if (buckets_[b] <= 0.0) continue; sum += buckets_[b]; - 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 + 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); + int marks = static_cast<int>(20 * (buckets_[b] / num_) + 0.5); r.append(marks, '#'); r.push_back('\n'); } diff --git a/util/histogram.h b/util/histogram.h index 1ef9f3c..4da60fb 100644 --- a/util/histogram.h +++ b/util/histogram.h @@ -11,8 +11,8 @@ namespace leveldb { class Histogram { public: - Histogram() { } - ~Histogram() { } + Histogram() {} + ~Histogram() {} void Clear(); void Add(double value); @@ -21,20 +21,22 @@ class Histogram { 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_; - enum { kNumBuckets = 154 }; - static const double kBucketLimit[kNumBuckets]; double buckets_[kNumBuckets]; - - double Median() const; - double Percentile(double p) const; - double Average() const; - double StandardDeviation() const; }; } // namespace leveldb diff --git a/util/logging.cc b/util/logging.cc index d48b6dd..8d6fb5b 100644 --- a/util/logging.cc +++ b/util/logging.cc @@ -4,10 +4,11 @@ #include "util/logging.h" -#include <errno.h> -#include <stdarg.h> -#include <stdio.h> -#include <stdlib.h> +#include <cstdarg> +#include <cstdio> +#include <cstdlib> +#include <limits> + #include "leveldb/env.h" #include "leveldb/slice.h" @@ -15,7 +16,7 @@ namespace leveldb { void AppendNumberTo(std::string* str, uint64_t num) { char buf[30]; - snprintf(buf, sizeof(buf), "%llu", (unsigned long long) num); + std::snprintf(buf, sizeof(buf), "%llu", static_cast<unsigned long long>(num)); str->append(buf); } @@ -26,8 +27,8 @@ void AppendEscapedStringTo(std::string* str, const Slice& value) { str->push_back(c); } else { char buf[10]; - snprintf(buf, sizeof(buf), "\\x%02x", - static_cast<unsigned int>(c) & 0xff); + std::snprintf(buf, sizeof(buf), "\\x%02x", + static_cast<unsigned int>(c) & 0xff); str->append(buf); } } @@ -46,28 +47,36 @@ std::string EscapeString(const Slice& value) { } bool ConsumeDecimalNumber(Slice* in, uint64_t* val) { - uint64_t v = 0; - int digits = 0; - while (!in->empty()) { - char c = (*in)[0]; - if (c >= '0' && c <= '9') { - ++digits; - // |delta| intentionally unit64_t to avoid Android crash (see log). - const uint64_t delta = (c - '0'); - static const uint64_t kMaxUint64 = ~static_cast<uint64_t>(0); - if (v > kMaxUint64/10 || - (v == kMaxUint64/10 && delta > kMaxUint64%10)) { - // Overflow - return false; - } - v = (v * 10) + delta; - in->remove_prefix(1); - } else { - break; + // 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 = v; - return (digits > 0); + + *val = value; + const size_t digits_consumed = current - start; + in->remove_prefix(digits_consumed); + return digits_consumed != 0; } } // namespace leveldb diff --git a/util/logging.h b/util/logging.h index 13351a2..a0394b2 100644 --- a/util/logging.h +++ b/util/logging.h @@ -8,9 +8,10 @@ #ifndef STORAGE_LEVELDB_UTIL_LOGGING_H_ #define STORAGE_LEVELDB_UTIL_LOGGING_H_ -#include <stdio.h> -#include <stdint.h> +#include <cstdint> +#include <cstdio> #include <string> + #include "port/port.h" namespace leveldb { diff --git a/util/logging_test.cc b/util/logging_test.cc new file mode 100644 index 0000000..1746c57 --- /dev/null +++ b/util/logging_test.cc @@ -0,0 +1,140 @@ +// 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 diff --git a/util/mutexlock.h b/util/mutexlock.h index 1ff5a9e..0cb2e25 100644 --- a/util/mutexlock.h +++ b/util/mutexlock.h @@ -22,20 +22,18 @@ namespace leveldb { class SCOPED_LOCKABLE MutexLock { public: - explicit MutexLock(port::Mutex *mu) EXCLUSIVE_LOCK_FUNCTION(mu) - : mu_(mu) { + 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_; - // No copying allowed - MutexLock(const MutexLock&); - void operator=(const MutexLock&); + port::Mutex* const mu_; }; } // namespace leveldb - #endif // STORAGE_LEVELDB_UTIL_MUTEXLOCK_H_ diff --git a/util/no_destructor.h b/util/no_destructor.h new file mode 100644 index 0000000..a0d3b87 --- /dev/null +++ b/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_ diff --git a/util/no_destructor_test.cc b/util/no_destructor_test.cc new file mode 100644 index 0000000..e3602cc --- /dev/null +++ b/util/no_destructor_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 "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 diff --git a/util/options.cc b/util/options.cc index b5e6227..62de5bf 100644 --- a/util/options.cc +++ b/util/options.cc @@ -9,22 +9,6 @@ namespace leveldb { -Options::Options() - : comparator(BytewiseComparator()), - create_if_missing(false), - error_if_exists(false), - paranoid_checks(false), - env(Env::Default()), - info_log(NULL), - write_buffer_size(4<<20), - max_open_files(1000), - block_cache(NULL), - block_size(4096), - block_restart_interval(16), - max_file_size(2<<20), - compression(kSnappyCompression), - reuse_logs(false), - filter_policy(NULL) { -} +Options::Options() : comparator(BytewiseComparator()), env(Env::Default()) {} } // namespace leveldb diff --git a/util/posix_logger.h b/util/posix_logger.h index 9741b1a..6bbc1a0 100644 --- a/util/posix_logger.h +++ b/util/posix_logger.h @@ -8,89 +8,121 @@ #ifndef STORAGE_LEVELDB_UTIL_POSIX_LOGGER_H_ #define STORAGE_LEVELDB_UTIL_POSIX_LOGGER_H_ -#include <algorithm> -#include <stdio.h> #include <sys/time.h> -#include <time.h> + +#include <cassert> +#include <cstdarg> +#include <cstdio> +#include <ctime> +#include <sstream> +#include <thread> + #include "leveldb/env.h" namespace leveldb { -class PosixLogger : public Logger { - private: - FILE* file_; - uint64_t (*gettid_)(); // Return the thread id for the current thread +class PosixLogger final : public Logger { public: - PosixLogger(FILE* f, uint64_t (*gettid)()) : file_(f), gettid_(gettid) { } - virtual ~PosixLogger() { - fclose(file_); - } - virtual void Logv(const char* format, va_list ap) { - const uint64_t thread_id = (*gettid_)(); - - // We try twice: the first time with a fixed-size stack allocated buffer, - // and the second time with a much larger dynamically allocated buffer. - char buffer[500]; - for (int iter = 0; iter < 2; iter++) { - char* base; - int bufsize; - if (iter == 0) { - bufsize = sizeof(buffer); - base = buffer; - } else { - bufsize = 30000; - base = new char[bufsize]; - } - char* p = base; - char* limit = base + bufsize; - - struct timeval now_tv; - gettimeofday(&now_tv, NULL); - const time_t seconds = now_tv.tv_sec; - struct tm t; - localtime_r(&seconds, &t); - p += snprintf(p, limit - p, - "%04d/%02d/%02d-%02d:%02d:%02d.%06d %llx ", - t.tm_year + 1900, - t.tm_mon + 1, - t.tm_mday, - t.tm_hour, - t.tm_min, - t.tm_sec, - static_cast<int>(now_tv.tv_usec), - static_cast<long long unsigned int>(thread_id)); - - // Print the message - if (p < limit) { - va_list backup_ap; - va_copy(backup_ap, ap); - p += vsnprintf(p, limit - p, format, backup_ap); - va_end(backup_ap); - } + // 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); + } - // Truncate to available space if necessary - if (p >= limit) { - if (iter == 0) { - continue; // Try again with larger buffer - } else { - p = limit - 1; + // 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 newline if necessary - if (p == base || p[-1] != '\n') { - *p++ = '\n'; + // Add a newline if necessary. + if (buffer[buffer_offset - 1] != '\n') { + buffer[buffer_offset] = '\n'; + ++buffer_offset; } - assert(p <= limit); - fwrite(base, 1, p - base, file_); - fflush(file_); - if (base != buffer) { - delete[] base; + 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 diff --git a/util/random.h b/util/random.h index ddd51b1..fe76ab4 100644 --- a/util/random.h +++ b/util/random.h @@ -5,7 +5,7 @@ #ifndef STORAGE_LEVELDB_UTIL_RANDOM_H_ #define STORAGE_LEVELDB_UTIL_RANDOM_H_ -#include <stdint.h> +#include <cstdint> namespace leveldb { @@ -15,6 +15,7 @@ namespace leveldb { class Random { private: uint32_t seed_; + public: explicit Random(uint32_t s) : seed_(s & 0x7fffffffu) { // Avoid bad seeds. @@ -23,8 +24,8 @@ class Random { } } 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 + 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 // @@ -54,9 +55,7 @@ class Random { // 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)); - } + uint32_t Skewed(int max_log) { return Uniform(1 << Uniform(max_log + 1)); } }; } // namespace leveldb diff --git a/util/status.cc b/util/status.cc index a44f35b..0559f5b 100644 --- a/util/status.cc +++ b/util/status.cc @@ -2,39 +2,41 @@ // 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 <stdio.h> -#include "port/port.h" #include "leveldb/status.h" +#include <cstdio> + +#include "port/port.h" + namespace leveldb { const char* Status::CopyState(const char* state) { uint32_t size; - memcpy(&size, state, sizeof(size)); + std::memcpy(&size, state, sizeof(size)); char* result = new char[size + 5]; - memcpy(result, state, 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 = msg.size(); - const uint32_t len2 = msg2.size(); + 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]; - memcpy(result, &size, sizeof(size)); + std::memcpy(result, &size, sizeof(size)); result[4] = static_cast<char>(code); - memcpy(result + 5, msg.data(), len1); + std::memcpy(result + 5, msg.data(), len1); if (len2) { result[5 + len1] = ':'; result[6 + len1] = ' '; - memcpy(result + 7 + len1, msg2.data(), len2); + std::memcpy(result + 7 + len1, msg2.data(), len2); } state_ = result; } std::string Status::ToString() const { - if (state_ == NULL) { + if (state_ == nullptr) { return "OK"; } else { char tmp[30]; @@ -59,14 +61,14 @@ std::string Status::ToString() const { type = "IO error: "; break; default: - snprintf(tmp, sizeof(tmp), "Unknown code(%d): ", - static_cast<int>(code())); + std::snprintf(tmp, sizeof(tmp), + "Unknown code(%d): ", static_cast<int>(code())); type = tmp; break; } std::string result(type); uint32_t length; - memcpy(&length, state_, sizeof(length)); + std::memcpy(&length, state_, sizeof(length)); result.append(state_ + 5, length); return result; } diff --git a/util/status_test.cc b/util/status_test.cc new file mode 100644 index 0000000..dbf5faa --- /dev/null +++ b/util/status_test.cc @@ -0,0 +1,39 @@ +// 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 diff --git a/util/testharness.cc b/util/testharness.cc deleted file mode 100644 index 95f025f..0000000 --- a/util/testharness.cc +++ /dev/null @@ -1,81 +0,0 @@ -// 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/testharness.h" - -#include <stdlib.h> -#include <sys/stat.h> -#include <sys/types.h> - -#include <string> -#include <vector> - -#include "leveldb/env.h" - -namespace leveldb { -namespace test { - -namespace { -struct Test { - const char* base; - const char* name; - void (*func)(); -}; -std::vector<Test>* tests; -} - -bool RegisterTest(const char* base, const char* name, void (*func)()) { - if (tests == NULL) { - tests = new std::vector<Test>; - } - Test t; - t.base = base; - t.name = name; - t.func = func; - tests->push_back(t); - return true; -} - -int RunAllTests() { - const char* matcher = getenv("LEVELDB_TESTS"); - - int num = 0; - if (tests != NULL) { - for (size_t i = 0; i < tests->size(); i++) { - const Test& t = (*tests)[i]; - if (matcher != NULL) { - std::string name = t.base; - name.push_back('.'); - name.append(t.name); - if (strstr(name.c_str(), matcher) == NULL) { - continue; - } - } - fprintf(stderr, "==== Test %s.%s\n", t.base, t.name); - (*t.func)(); - ++num; - } - } - fprintf(stderr, "==== PASSED %d tests\n", num); - return 0; -} - -std::string TmpDir() { - std::string dir; - Status s = Env::Default()->GetTestDirectory(&dir); - ASSERT_TRUE(s.ok()) << s.ToString(); - return dir; -} - -int RandomSeed() { - const char* env = getenv("TEST_RANDOM_SEED"); - int result = (env != NULL ? atoi(env) : 301); - if (result <= 0) { - result = 301; - } - return result; -} - -} // namespace test -} // namespace leveldb diff --git a/util/testharness.h b/util/testharness.h deleted file mode 100644 index 8ee7972..0000000 --- a/util/testharness.h +++ /dev/null @@ -1,137 +0,0 @@ -// 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_TESTHARNESS_H_ -#define STORAGE_LEVELDB_UTIL_TESTHARNESS_H_ - -#include <stdio.h> -#include <stdlib.h> - -#include <sstream> - -#include "leveldb/status.h" - -namespace leveldb { -namespace test { - -// Run some of the tests registered by the TEST() macro. If the -// environment variable "LEVELDB_TESTS" is not set, runs all tests. -// Otherwise, runs only the tests whose name contains the value of -// "LEVELDB_TESTS" as a substring. E.g., suppose the tests are: -// TEST(Foo, Hello) { ... } -// TEST(Foo, World) { ... } -// LEVELDB_TESTS=Hello will run the first test -// LEVELDB_TESTS=o will run both tests -// LEVELDB_TESTS=Junk will run no tests -// -// Returns 0 if all tests pass. -// Dies or returns a non-zero value if some test fails. -int RunAllTests(); - -// Return the directory to use for temporary storage. -std::string TmpDir(); - -// Return a randomization seed for this run. Typically returns the -// same number on repeated invocations of this binary, but automated -// runs may be able to vary the seed. -int RandomSeed(); - -// An instance of Tester is allocated to hold temporary state during -// the execution of an assertion. -class Tester { - private: - bool ok_; - const char* fname_; - int line_; - std::stringstream ss_; - - public: - Tester(const char* f, int l) - : ok_(true), fname_(f), line_(l) { - } - - ~Tester() { - if (!ok_) { - fprintf(stderr, "%s:%d:%s\n", fname_, line_, ss_.str().c_str()); - exit(1); - } - } - - Tester& Is(bool b, const char* msg) { - if (!b) { - ss_ << " Assertion failure " << msg; - ok_ = false; - } - return *this; - } - - Tester& IsOk(const Status& s) { - if (!s.ok()) { - ss_ << " " << s.ToString(); - ok_ = false; - } - return *this; - } - -#define BINARY_OP(name, op) \ - template <class X, class Y> \ - Tester& name(const X& x, const Y& y) { \ - if (!(x op y)) { \ - ss_ << " failed: " << x << (" " #op " ") << y; \ - ok_ = false; \ - } \ - return *this; \ - } - - BINARY_OP(IsEq, ==) - BINARY_OP(IsNe, !=) - BINARY_OP(IsGe, >=) - BINARY_OP(IsGt, >) - BINARY_OP(IsLe, <=) - BINARY_OP(IsLt, <) -#undef BINARY_OP - - // Attach the specified value to the error message if an error has occurred - template <class V> - Tester& operator<<(const V& value) { - if (!ok_) { - ss_ << " " << value; - } - return *this; - } -}; - -#define ASSERT_TRUE(c) ::leveldb::test::Tester(__FILE__, __LINE__).Is((c), #c) -#define ASSERT_OK(s) ::leveldb::test::Tester(__FILE__, __LINE__).IsOk((s)) -#define ASSERT_EQ(a,b) ::leveldb::test::Tester(__FILE__, __LINE__).IsEq((a),(b)) -#define ASSERT_NE(a,b) ::leveldb::test::Tester(__FILE__, __LINE__).IsNe((a),(b)) -#define ASSERT_GE(a,b) ::leveldb::test::Tester(__FILE__, __LINE__).IsGe((a),(b)) -#define ASSERT_GT(a,b) ::leveldb::test::Tester(__FILE__, __LINE__).IsGt((a),(b)) -#define ASSERT_LE(a,b) ::leveldb::test::Tester(__FILE__, __LINE__).IsLe((a),(b)) -#define ASSERT_LT(a,b) ::leveldb::test::Tester(__FILE__, __LINE__).IsLt((a),(b)) - -#define TCONCAT(a, b) TCONCAT1(a, b) -#define TCONCAT1(a, b) a##b - -#define TEST(base, name) \ -class TCONCAT(_Test_, name) : public base { \ - public: \ - void _Run(); \ - static void _RunIt() { \ - TCONCAT(_Test_, name) t; \ - t._Run(); \ - } \ -}; \ -bool TCONCAT(_Test_ignored_, name) = \ - ::leveldb::test::RegisterTest(#base, #name, &TCONCAT(_Test_, name)::_RunIt); \ -void TCONCAT(_Test_, name)::_Run() - -// Register the specified test. Typically not used directly, but -// invoked via the macro expansion of TEST. -bool RegisterTest(const char* base, const char* name, void (*func)()); - -} // namespace test -} // namespace leveldb - -#endif // STORAGE_LEVELDB_UTIL_TESTHARNESS_H_ diff --git a/util/testutil.cc b/util/testutil.cc index 9d8079c..5f77b08 100644 --- a/util/testutil.cc +++ b/util/testutil.cc @@ -4,6 +4,8 @@ #include "util/testutil.h" +#include <string> + #include "util/random.h" namespace leveldb { @@ -12,7 +14,7 @@ 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)); // ' ' .. '~' + (*dst)[i] = static_cast<char>(' ' + rnd->Uniform(95)); // ' ' .. '~' } return Slice(*dst); } @@ -20,9 +22,8 @@ Slice RandomString(Random* rnd, int len, std::string* 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' - }; + 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))]; @@ -30,9 +31,8 @@ std::string RandomKey(Random* rnd, int len) { return result; } - -Slice CompressibleString(Random* rnd, double compressed_fraction, - size_t len, std::string* dst) { +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; diff --git a/util/testutil.h b/util/testutil.h index 8726bf7..e0e2d64 100644 --- a/util/testutil.h +++ b/util/testutil.h @@ -5,6 +5,9 @@ #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" @@ -12,6 +15,20 @@ 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); @@ -23,8 +40,8 @@ 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); +Slice CompressibleString(Random* rnd, double compressed_fraction, size_t len, + std::string* dst); // A wrapper that allows injection of errors. class ErrorEnv : public EnvWrapper { @@ -32,25 +49,27 @@ class ErrorEnv : public EnvWrapper { bool writable_file_error_; int num_writable_file_errors_; - ErrorEnv() : EnvWrapper(Env::Default()), - writable_file_error_(false), - num_writable_file_errors_(0) { } + ErrorEnv() + : EnvWrapper(NewMemEnv(Env::Default())), + writable_file_error_(false), + num_writable_file_errors_(0) {} + ~ErrorEnv() override { delete target(); } - virtual Status NewWritableFile(const std::string& fname, - WritableFile** result) { + Status NewWritableFile(const std::string& fname, + WritableFile** result) override { if (writable_file_error_) { ++num_writable_file_errors_; - *result = NULL; + *result = nullptr; return Status::IOError(fname, "fake error"); } return target()->NewWritableFile(fname, result); } - virtual Status NewAppendableFile(const std::string& fname, - WritableFile** result) { + Status NewAppendableFile(const std::string& fname, + WritableFile** result) override { if (writable_file_error_) { ++num_writable_file_errors_; - *result = NULL; + *result = nullptr; return Status::IOError(fname, "fake error"); } return target()->NewAppendableFile(fname, result); diff --git a/util/windows_logger.h b/util/windows_logger.h new file mode 100644 index 0000000..26e6c7b --- /dev/null +++ b/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_ |