/tmp/bitcoin/src/coins.h

Source
// Copyright (c) 2009-2010 Satoshi Nakamoto
// Copyright (c) 2009-present The Bitcoin Core developers
// Distributed under the MIT software license, see the accompanying
// file COPYING or http://www.opensource.org/licenses/mit-license.php.

#ifndef BITCOIN_COINS_H
#define BITCOIN_COINS_H

#include <attributes.h>
#include <compressor.h>
#include <core_memusage.h>
#include <crypto/siphash.h>
#include <memusage.h>
#include <primitives/transaction.h>
#include <primitives/transaction_identifier.h>
#include <serialize.h>
#include <support/allocators/pool.h>
#include <uint256.h>
#include <util/check.h>
#include <util/log.h>
#include <util/overflow.h>

#include <cassert>
#include <cstdint>

#include <atomic>
#include <functional>
#include <future>
#include <memory>
#include <optional>
#include <unordered_map>
#include <utility>
#include <vector>

class CBlock;
class ThreadPool;

/**
 * A UTXO entry.
 *
 * Serialized format:
 * - VARINT((height << 1) | (coinbase ? 1 : 0))
 * - the non-spent CTxOut (via TxOutCompression)
 */
class Coin
{
public:
    //! unspent transaction output
    CTxOut out;

    //! whether containing transaction was a coinbase
    bool fCoinBase : 1;

    //! at which height this containing transaction was included in the active block chain
    uint32_t nHeight : 31;

    //! construct a Coin from a CTxOut and height/coinbase information.
    Coin(CTxOut&& outIn, int nHeightIn, bool fCoinBaseIn) : out(std::move(outIn)), fCoinBase(fCoinBaseIn), nHeight(nHeightIn) {}
    Coin(const CTxOut& outIn, int nHeightIn, bool fCoinBaseIn) : out(outIn), fCoinBase(fCoinBaseIn),nHeight(nHeightIn) {}

    void Clear() {
        out.SetNull();
        fCoinBase = false;
        nHeight = 0;
    }

    //! empty constructor
    Coin() : fCoinBase(false), nHeight(0) { }

    bool IsCoinBase() const {
        return fCoinBase;
    }

    template<typename Stream>
    void Serialize(Stream &s) const {
        assert(!IsSpent());
        uint32_t code{(uint32_t{nHeight} << 1) | uint32_t{fCoinBase}};
        ::Serialize(s, VARINT(code));
        ::Serialize(s, Using<TxOutCompression>(out));
    }

    template<typename Stream>
    void Unserialize(Stream &s) {
        uint32_t code = 0;
        ::Unserialize(s, VARINT(code));
        nHeight = code >> 1;
        fCoinBase = code & 1;
        ::Unserialize(s, Using<TxOutCompression>(out));
    }

    /** Either this coin never existed (see e.g. coinEmpty in coins.cpp), or it
      * did exist and has been spent.
      */
    bool IsSpent() const {
        return out.IsNull();
    }

    size_t DynamicMemoryUsage() const {
        return memusage::DynamicUsage(out.scriptPubKey);
    }
};

struct CCoinsCacheEntry;
using CoinsCachePair = std::pair<const COutPoint, CCoinsCacheEntry>;

/**
 * A Coin in one level of the coins database caching hierarchy.
 *
 * A coin can either be:
 * - unspent or spent (in which case the Coin object will be nulled out - see Coin.Clear())
 * - DIRTY or not DIRTY
 * - FRESH or not FRESH
 *
 * Out of these 2^3 = 8 states, only some combinations are valid:
 * - unspent, FRESH, DIRTY (e.g. a new coin created in the cache)
 * - unspent, not FRESH, DIRTY (e.g. a coin changed in the cache during a reorg)
 * - unspent, not FRESH, not DIRTY (e.g. an unspent coin fetched from the parent cache)
 * - spent, not FRESH, DIRTY (e.g. a coin is spent and spentness needs to be flushed to the parent)
 */
struct CCoinsCacheEntry
{
private:
    /**
     * These are used to create a doubly linked list of flagged entries.
     * They are set in SetDirty, SetFresh, and unset in SetClean.
     * A flagged entry is any entry that is either DIRTY, FRESH, or both.
     *
     * DIRTY entries are tracked so that only modified entries can be passed to
     * the parent cache for batch writing. This is a performance optimization
     * compared to giving all entries in the cache to the parent and having the
     * parent scan for only modified entries.
     */
    CoinsCachePair* m_prev{nullptr};
    CoinsCachePair* m_next{nullptr};
    uint8_t m_flags{0};

    //! Adding a flag requires a reference to the sentinel of the flagged pair linked list.
    static void AddFlags(uint8_t flags, CoinsCachePair& pair, CoinsCachePair& sentinel) noexcept
    {
        Assume(flags & (DIRTY | FRESH));
        if (!pair.second.m_flags) {
            Assume(!pair.second.m_prev && !pair.second.m_next);
            pair.second.m_prev = sentinel.second.m_prev;
            pair.second.m_next = &sentinel;
            sentinel.second.m_prev = &pair;
            pair.second.m_prev->second.m_next = &pair;
        }
        Assume(pair.second.m_prev && pair.second.m_next);
        pair.second.m_flags |= flags;
    }

public:
    Coin coin; // The actual cached data.

    enum Flags {
        /**
         * DIRTY means the CCoinsCacheEntry is potentially different from the
         * version in the parent cache. Failure to mark a coin as DIRTY when
         * it is potentially different from the parent cache will cause a
         * consensus failure, since the coin's state won't get written to the
         * parent when the cache is flushed.
         */
        DIRTY = (1 << 0),
        /**
         * FRESH means the parent cache does not have this coin or that it is a
         * spent coin in the parent cache. If a FRESH coin in the cache is
         * later spent, it can be deleted entirely and doesn't ever need to be
         * flushed to the parent. This is a performance optimization. Marking a
         * coin as FRESH when it exists unspent in the parent cache will cause a
         * consensus failure, since it might not be deleted from the parent
         * when this cache is flushed.
         */
        FRESH = (1 << 1),
    };

    CCoinsCacheEntry() noexcept = default;
    explicit CCoinsCacheEntry(Coin&& coin_) noexcept : coin(std::move(coin_)) {}
    ~CCoinsCacheEntry()
    {
        SetClean();
    }

    static void SetDirty(CoinsCachePair& pair, CoinsCachePair& sentinel) noexcept { AddFlags(DIRTY, pair, sentinel); }
    static void SetFresh(CoinsCachePair& pair, CoinsCachePair& sentinel) noexcept { AddFlags(FRESH, pair, sentinel); }

    void SetClean() noexcept
    {
        if (!m_flags) return;
        m_next->second.m_prev = m_prev;
        m_prev->second.m_next = m_next;
        m_flags = 0;
        m_prev = m_next = nullptr;
    }
    bool IsDirty() const noexcept { return m_flags & DIRTY; }
    bool IsFresh() const noexcept { return m_flags & FRESH; }

    //! Only call Next when this entry is DIRTY, FRESH, or both
    CoinsCachePair* Next() const noexcept
    {
        Assume(m_flags);
        return m_next;
    }

    //! Only call Prev when this entry is DIRTY, FRESH, or both
    CoinsCachePair* Prev() const noexcept
    {
        Assume(m_flags);
        return m_prev;
    }

    //! Only use this for initializing the linked list sentinel
    void SelfRef(CoinsCachePair& pair) noexcept
    {
        Assume(&pair.second == this);
        m_prev = &pair;
        m_next = &pair;
        // Set sentinel to DIRTY so we can call Next on it
        m_flags = DIRTY;
    }
};

/**
 * SipHash-1-3-UJ based hasher for the coins cache and related coins containers.
 *
 * Retained entries identify real transaction outputs, so their keys contain computed txids.
 * Missing-input lookups may contain arbitrary claimed prevouts, but FetchCoin() immediately
 * erases their temporary entries when the backend lookup fails, so non-hash keys cannot
 * accumulate.
 *
 * The assumeutxo loader assumes snapshot txids are valid while loading and verifies the
 * complete snapshot's content hash before activation.
 *
 * Hash values are process-local and must not be persisted, serialized, or compared across
 * processes.
 *
 * Having the hash noexcept lets libstdc++ recalculate it during rehash instead of storing it in
 * each node.
 */
class SaltedCoinsCacheHasher
{
    const SipHasher13UJ m_hasher;

public:
    SaltedCoinsCacheHasher(bool deterministic = false);

    /** Hash a transaction ID, itself a cryptographic hash, as one jumbo block. */
    size_t operator()(const Txid& id) const noexcept
    {
        return m_hasher.Hash(id.ToUint256());
    }

    /** Hash an outpoint as its txid jumbo block followed by the zero-extended index as one normal block. */
    size_t operator()(const COutPoint& id) const noexcept
    {
        return m_hasher.Hash(id.hash.ToUint256(), uint64_t{id.n});
    }
};

/**
 * PoolAllocator's MAX_BLOCK_SIZE_BYTES parameter here uses sizeof the data, and adds the size
 * of 4 pointers. We do not know the exact node size used in the std::unordered_node implementation
 * because it is implementation defined. Most implementations have an overhead of 1 or 2 pointers,
 * so nodes can be connected in a linked list, and in some cases the hash value is stored as well.
 * Using an additional sizeof(void*)*4 for MAX_BLOCK_SIZE_BYTES should thus be sufficient so that
 * all implementations can allocate the nodes from the PoolAllocator.
 */
using CCoinsMap = std::unordered_map<COutPoint,
                                     CCoinsCacheEntry,
                                     SaltedCoinsCacheHasher,
                                     std::equal_to<COutPoint>,
                                     PoolAllocator<CoinsCachePair,
                                                   sizeof(CoinsCachePair) + sizeof(void*) * 4>>;

using CCoinsMapMemoryResource = CCoinsMap::allocator_type::ResourceType;

/** Cursor for iterating over CoinsView state */
class CCoinsViewCursor
{
public:
    CCoinsViewCursor(const uint256& in_block_hash) : block_hash(in_block_hash) {}
    virtual ~CCoinsViewCursor() = default;

    virtual bool GetKey(COutPoint &key) const = 0;
    virtual bool GetValue(Coin &coin) const = 0;

    virtual bool Valid() const = 0;
    virtual void Next() = 0;

    //! Get best block at the time this cursor was created
    const uint256& GetBestBlock() const { return block_hash; }
private:
    uint256 block_hash;
};

/**
 * Cursor for iterating over the linked list of flagged entries in CCoinsViewCache.
 *
 * This is a helper struct to encapsulate the diverging logic between a non-erasing
 * CCoinsViewCache::Sync and an erasing CCoinsViewCache::Flush. This allows the receiver
 * of CCoinsView::BatchWrite to iterate through the flagged entries without knowing
 * the caller's intent.
 *
 * However, the receiver can still call CoinsViewCacheCursor::WillErase to see if the
 * caller will erase the entry after BatchWrite returns. If so, the receiver can
 * perform optimizations such as moving the coin out of the CCoinsCachEntry instead
 * of copying it.
 */
struct CoinsViewCacheCursor
{
    //! If will_erase is not set, iterating through the cursor will erase spent coins from the map,
    //! and other coins will be unflagged (removing them from the linked list).
    //! If will_erase is set, the underlying map and linked list will not be modified,
    //! as the caller is expected to wipe the entire map anyway.
    //! This is an optimization compared to erasing all entries as the cursor iterates them when will_erase is set.
    //! Calling CCoinsMap::clear() afterwards is faster because a CoinsCachePair cannot be coerced back into a
    //! CCoinsMap::iterator to be erased, and must therefore be looked up again by key in the CCoinsMap before being erased.
    CoinsViewCacheCursor(size_t& dirty_count LIFETIMEBOUND,
                         CoinsCachePair& sentinel LIFETIMEBOUND,
                         CCoinsMap& map LIFETIMEBOUND,
                         bool will_erase) noexcept
        : m_dirty_count(dirty_count), m_sentinel(sentinel), m_map(map), m_will_erase(will_erase) {}

    inline CoinsCachePair* Begin() const noexcept { return m_sentinel.second.Next(); }
    inline CoinsCachePair* End() const noexcept { return &m_sentinel; }

    //! Return the next entry after current, possibly erasing current
    inline CoinsCachePair* NextAndMaybeErase(CoinsCachePair& current) noexcept
    {
        const auto next_entry{current.second.Next()};
        Assume(TrySub(m_dirty_count, current.second.IsDirty()));
        // If we are not going to erase the cache, we must still erase spent entries.
        // Otherwise, clear the state of the entry.
        if (!m_will_erase) {
            if (current.second.coin.IsSpent()) {
                assert(current.second.coin.DynamicMemoryUsage() == 0); // scriptPubKey was already cleared in SpendCoin
                m_map.erase(current.first);
            } else {
                current.second.SetClean();
            }
        }
        return next_entry;
    }

    inline bool WillErase(CoinsCachePair& current) const noexcept { return m_will_erase || current.second.coin.IsSpent(); }
    size_t GetDirtyCount() const noexcept { return m_dirty_count; }
    size_t GetTotalCount() const noexcept { return m_map.size(); }
private:
    size_t& m_dirty_count;
    CoinsCachePair& m_sentinel;
    CCoinsMap& m_map;
    bool m_will_erase;
};

/** Pure abstract view on the open txout dataset. */
class CCoinsView
{
public:
    //! As we use CCoinsViews polymorphically, have a virtual destructor
    virtual ~CCoinsView() = default;

    //! Retrieve the Coin (unspent transaction output) for a given outpoint.
    //! May populate the cache. Use PeekCoin() to perform a non-caching lookup.
    virtual std::optional<Coin> GetCoin(const COutPoint& outpoint) const = 0;

    //! Retrieve the Coin (unspent transaction output) for a given outpoint, without caching results.
    //! Does not populate the cache. Use GetCoin() to cache the result.
    virtual std::optional<Coin> PeekCoin(const COutPoint& outpoint) const = 0;

    //! Just check whether a given outpoint is unspent.
    //! May populate the cache. Use PeekCoin() to perform a non-caching lookup.
    virtual bool HaveCoin(const COutPoint& outpoint) const = 0;

    //! Retrieve the block hash whose state this CCoinsView currently represents
    virtual uint256 GetBestBlock() const = 0;

    //! Retrieve the range of blocks that may have been only partially written.
    //! If the database is in a consistent state, the result is the empty vector.
    //! Otherwise, a two-element vector is returned consisting of the new and
    //! the old block hash, in that order.
    virtual std::vector<uint256> GetHeadBlocks() const = 0;

    //! Do a bulk modification (multiple Coin changes + BestBlock change).
    //! The passed cursor is used to iterate through the coins.
    virtual void BatchWrite(CoinsViewCacheCursor& cursor, const uint256& block_hash) = 0;

    //! Estimate database size
    virtual size_t EstimateSize() const = 0;
};

/** Noop coins view. */
class CoinsViewEmpty : public CCoinsView
{
protected:
    CoinsViewEmpty() = default;

public:
    static CoinsViewEmpty& Get();

    CoinsViewEmpty(const CoinsViewEmpty&) = delete;
    CoinsViewEmpty& operator=(const CoinsViewEmpty&) = delete;

    std::optional<Coin> GetCoin(const COutPoint&) const override { return {}; }
    std::optional<Coin> PeekCoin(const COutPoint& outpoint) const override { return GetCoin(outpoint); }
    bool HaveCoin(const COutPoint& outpoint) const override { return !!GetCoin(outpoint); }
    uint256 GetBestBlock() const override { return {}; }
    std::vector<uint256> GetHeadBlocks() const override { return {}; }
    void BatchWrite(CoinsViewCacheCursor& cursor, const uint256&) override
    {
        for (auto it{cursor.Begin()}; it != cursor.End(); it = cursor.NextAndMaybeErase(*it)) { }
    }
    size_t EstimateSize() const override { return 0; }
};

/** CCoinsView backed by another CCoinsView */
class CCoinsViewBacked : public CCoinsView
{
protected:
    CCoinsView* base;

public:
    explicit CCoinsViewBacked(CCoinsView* in_view) : base{Assert(in_view)} {}

    void SetBackend(CCoinsView& in_view) { base = &in_view; }

    std::optional<Coin> GetCoin(const COutPoint& outpoint) const override { return base->GetCoin(outpoint); }
    std::optional<Coin> PeekCoin(const COutPoint& outpoint) const override { return base->PeekCoin(outpoint); }
    bool HaveCoin(const COutPoint& outpoint) const override { return base->HaveCoin(outpoint); }
    uint256 GetBestBlock() const override { return base->GetBestBlock(); }
    std::vector<uint256> GetHeadBlocks() const override { return base->GetHeadBlocks(); }
    void BatchWrite(CoinsViewCacheCursor& cursor, const uint256& block_hash) override { base->BatchWrite(cursor, block_hash); }
    size_t EstimateSize() const override { return base->EstimateSize(); }
};


/** CCoinsView that adds a memory cache for transactions to another CCoinsView */
class CCoinsViewCache : public CCoinsViewBacked
{
private:
    const bool m_deterministic;

protected:
    /**
     * Make mutable so that we can "fill the cache" even from Get-methods
     * declared as "const".
     */
    mutable uint256 m_block_hash;
    mutable CCoinsMapMemoryResource m_cache_coins_memory_resource{};
    /* The starting sentinel of the flagged entry circular doubly linked list. */
    mutable CoinsCachePair m_sentinel;
    mutable CCoinsMap cacheCoins;

    /* Cached dynamic memory usage for the inner Coin objects. */
    mutable size_t cachedCoinsUsage{0};
    /* Running count of dirty Coin cache entries. */
    mutable size_t m_dirty_count{0};

    /**
     * Discard all modifications made to this cache without flushing to the base view.
     * This can be used to efficiently reuse a cache instance across multiple operations.
     */
    virtual void Reset() noexcept;

    /* Fetch the coin from base. Used for cache misses in FetchCoin. */
    virtual std::optional<Coin> FetchCoinFromBase(const COutPoint& outpoint) const;

public:
    CCoinsViewCache(CCoinsView* in_base, bool deterministic = false);

    /**
     * By deleting the copy constructor, we prevent accidentally using it when one intends to create a cache on top of a base cache.
     */
    CCoinsViewCache(const CCoinsViewCache &) = delete;

    // Standard CCoinsView methods
    std::optional<Coin> GetCoin(const COutPoint& outpoint) const override;
    std::optional<Coin> PeekCoin(const COutPoint& outpoint) const override;
    bool HaveCoin(const COutPoint& outpoint) const override;
    uint256 GetBestBlock() const override;
    void SetBestBlock(const uint256& block_hash);
    void BatchWrite(CoinsViewCacheCursor& cursor, const uint256& block_hash) override;

    /**
     * Check if we have the given utxo already loaded in this cache.
     * The semantics are the same as HaveCoin(), but no calls to
     * the backing CCoinsView are made.
     */
    bool HaveCoinInCache(const COutPoint &outpoint) const;

    /**
     * Return a reference to Coin in the cache, or coinEmpty if not found. This is
     * more efficient than GetCoin.
     *
     * Generally, do not hold the reference returned for more than a short scope.
     * While the current implementation allows for modifications to the contents
     * of the cache while holding the reference, this behavior should not be relied
     * on! To be safe, best to not hold the returned reference through any other
     * calls to this cache.
     */
    const Coin& AccessCoin(const COutPoint &output) const;

    /**
     * Add a coin. Set possible_overwrite to true if an unspent version may
     * already exist in the cache.
     */
    void AddCoin(const COutPoint& outpoint, Coin&& coin, bool possible_overwrite);

    /**
     * Emplace a coin into cacheCoins without performing any checks, marking
     * the emplaced coin as dirty.
     *
     * NOT FOR GENERAL USE. Used only when loading coins from a UTXO snapshot.
     * @sa ChainstateManager::PopulateAndValidateSnapshot()
     */
    void EmplaceCoinInternalDANGER(const COutPoint& outpoint, Coin&& coin);

    /**
     * Spend a coin. Pass moveto in order to get the deleted data.
     * If no unspent output exists for the passed outpoint, this call
     * has no effect.
     */
    bool SpendCoin(const COutPoint &outpoint, Coin* moveto = nullptr);

    /**
     * Push the modifications applied to this cache to its base and wipe local state.
     * Failure to call this method or Sync() before destruction will cause the changes
     * to be forgotten.
     * If reallocate_cache is false, the cache will retain the same memory footprint
     * after flushing and should be destroyed to deallocate.
     */
    virtual void Flush(bool reallocate_cache = true);

    /**
     * Push the modifications applied to this cache to its base while retaining
     * the contents of this cache (except for spent coins, which we erase).
     * Failure to call this method or Flush() before destruction will cause the changes
     * to be forgotten.
     */
    void Sync();

    /**
     * Removes the UTXO with the given outpoint from the cache, if it is
     * not modified.
     */
    void Uncache(const COutPoint &outpoint);

    //! Size of the cache (in number of transaction outputs)
    unsigned int GetCacheSize() const;

    //! Number of dirty cache entries (transaction outputs)
    size_t GetDirtyCount() const noexcept { return m_dirty_count; }

    //! Calculate the size of the cache (in bytes)
    size_t DynamicMemoryUsage() const;

    //! Check whether all prevouts of the transaction are present in the UTXO set represented by this view
    bool HaveInputs(const CTransaction& tx) const;

    //! Force a reallocation of the cache map. This is required when downsizing
    //! the cache because the map's allocator may be hanging onto a lot of
    //! memory despite having called .clear().
    //!
    //! See: https://stackoverflow.com/questions/42114044/how-to-release-unordered-map-memory
    void ReallocateCache();

    //! Run an internal sanity check on the cache data structure. */
    void SanityCheck() const;

    class ResetGuard
    {
    private:
        friend CCoinsViewCache;
        CCoinsViewCache& m_cache;
        explicit ResetGuard(CCoinsViewCache& cache LIFETIMEBOUND) noexcept : m_cache{cache} {}

    public:
        ResetGuard(const ResetGuard&) = delete;
        ResetGuard& operator=(const ResetGuard&) = delete;
        ResetGuard(ResetGuard&&) = delete;
        ResetGuard& operator=(ResetGuard&&) = delete;

        ~ResetGuard() { m_cache.Reset(); }
    };

    //! Create a scoped guard that will call `Reset()` on this cache when it goes out of scope.
    [[nodiscard]] ResetGuard CreateResetGuard() noexcept { return ResetGuard{*this}; }

private:
    /**
     * @note this is marked const, but may actually append to `cacheCoins`, increasing
     * memory usage.
     */
    CCoinsMap::iterator FetchCoin(const COutPoint &outpoint) const;
};

/**
 * CCoinsViewCache subclass that asynchronously fetches most block input prevouts in parallel during ConnectBlock without
 * mutating the base cache.
 *
 * Only used in ConnectBlock to pass as an ephemeral view that can be reset if the block is invalid.
 * It provides the same interface as CCoinsViewCache.
 * It adds an additional StartFetching method to provide the block.
 *
 * When a block is passed to StartFetching, the inputs of the block are flattened into a vector of InputToFetch
 * objects. StartFetching then submits worker tasks to a ThreadPool and keeps the returned futures alive until fetching
 * is stopped.
 *
 * ProcessInput() atomically fetches and increments m_input_head, so each thread can only access a single element of the
 * m_inputs vector at a time. Workers race to claim inputs, so they may fetch elements in any order. If the fetched
 * index is greater than or equal to the size of m_inputs, no more inputs can be fetched and false is returned.
 *
 * The worker claims the InputToFetch at this index, fetches the coin from the base cache and moves it into the
 * InputToFetch object. The ready flag is then set with a release memory order. This allows the ready flag to be
 * used as a memory fence, guaranteeing the coin being written to the object will have happened before another
 * thread tests the flag with an acquire memory order.
 * This assumes all base->PeekCoin() paths are safe for concurrent readers and do not mutate lower cache layers.
 *
 * When a coin is requested from the cache on the main thread and is not already in cacheCoins map, FetchCoinFromBase
 * checks whether the next unconsumed entry in m_inputs has the requested outpoint. On a match, m_input_tail is advanced
 * and the entry's ready flag is waited on with an acquire memory order until a worker has finished fetching it. The
 * coin is then moved out and returned. Since the main thread is the only consumer of validation results, it blocks
 * on the specific input it needs rather than racing workers for other inputs.
 *
 * StopFetching() is called in Flush() and in Reset() (the per-block teardown) so workers stop before the block they
 * reference goes away. It stops fetching by moving m_input_head to the end of m_inputs (so workers quickly exit),
 * then waits for all futures to complete and clears the per-block state (m_inputs and the head/tail counters).
 *
 *       Workers advance m_input_head to fetch inputs. Main thread advances m_input_tail to consume.
 *
 *       Before workers start:
 *
 *                 m_input_head
 *                 m_input_tail
 *                      │
 *                      ▼
 *                 ┌─────────┬─────────┬─────────┬─────────┬─────────┬─────────┬─────────┬─────────┬─────────┐
 *       m_inputs: │ waiting │ waiting │ waiting │ waiting │ waiting │ waiting │ waiting │ waiting │ waiting │
 *                 │         │         │         │         │         │         │         │         │         │
 *                 └─────────┴─────────┴─────────┴─────────┴─────────┴─────────┴─────────┴─────────┴─────────┘
 *
 *       After workers start:
 *
 *                                       Worker 2            Worker 0  Worker 3  Worker 1  m_input_head
 *                                          │                   │         │         │         │
 *                                          ▼                   ▼         ▼         ▼         ▼
 *                 ┌─────────┬─────────┬─────────┬─────────┬─────────┬─────────┬─────────┬─────────┬─────────┐
 *       m_inputs: │  ready  │  ready  │fetching │  ready  │fetching │fetching │fetching │ waiting │ waiting │
 *                 │consumed │    ✓    │    ●    │    ✓    │    ●    │    ●    │    ●    │         │         │
 *                 └─────────┴─────────┴─────────┴─────────┴─────────┴─────────┴─────────┴─────────┴─────────┘
 *                                ▲
 *                                │
 *                           m_input_tail
 */
class CoinsViewOverlay : public CCoinsViewCache
{
private:
    //! The latest input not yet being fetched. Workers atomically increment this when fetching.
    std::atomic_uint32_t m_input_head{0};
    //! The latest input not yet accessed by a consumer. Only the main thread increments this.
    mutable uint32_t m_input_tail{0};

    //! The inputs of the block which is being fetched.
    struct InputToFetch {
        //! Workers set this after setting the coin. The main thread tests this before reading the coin.
        std::atomic_flag ready{};
        //! The outpoint of the input to fetch.
        const COutPoint& outpoint;
        //! The coin that workers will fetch and main thread will insert into cache.
        //! Mutable so it can be moved in FetchCoinFromBase.
        mutable std::optional<Coin> coin{std::nullopt};

        explicit InputToFetch(const COutPoint& o LIFETIMEBOUND) noexcept : outpoint{o} {}

        //! Move ctor is required for resizing m_inputs in StartFetching. Elements will never move once parallel tasks
        //! are started, so we can assert that coin is nullopt and ready is false.
        InputToFetch(InputToFetch&& other) noexcept : outpoint{other.outpoint}
        {
            Assert(!other.coin);
            Assert(!other.ready.test(std::memory_order_relaxed));
        }
    };
    //! Must only be mutated when m_futures is empty. Elements may be mutated when m_futures is not empty.
    std::vector<InputToFetch> m_inputs{};

    /**
     * Claim and fetch the next input in the queue.
     *
     * @return true if an input prevout was fetched
     * @return false if there are no more input prevouts in the queue to fetch
     */
    bool ProcessInput() noexcept
    {
        const auto i{m_input_head.fetch_add(1, std::memory_order_relaxed)};
        if (i >= m_inputs.size()) return false;

        auto& input{m_inputs[i]};
        input.coin = base->PeekCoin(input.outpoint);
        // Use release so writing coin above happens before the main thread acquires.
        Assert(!input.ready.test_and_set(std::memory_order_release));
        input.ready.notify_one();
        return true;
    }

    //! Stop all worker threads and clear fetching data.
    //! Calling this is idempotent, and may safely be called if not fetching.
    void StopFetching() noexcept
    {
        if (m_futures.empty()) {
            Assert(m_inputs.empty());
            Assert(m_input_head.load(std::memory_order_relaxed) == 0);
            Assert(m_input_tail == 0);
            return;
        }
        // Skip fetching the rest of the inputs by moving the head to the end.
        m_input_head.store(m_inputs.size(), std::memory_order_relaxed);
        // Wait for all threads to stop.
        for (auto& future : m_futures) future.wait();
        m_futures.clear();
        m_inputs.clear();
        m_input_head.store(0, std::memory_order_relaxed);
        m_input_tail = 0;
    }

    std::optional<Coin> FetchCoinFromBase(const COutPoint& outpoint) const override
    {
        // This assumes ConnectBlock accesses all inputs in the same order as
        // they are added to m_inputs in StartFetching.
        if (m_input_tail < m_inputs.size() && m_inputs[m_input_tail].outpoint == outpoint) {
            // We advance the tail since the input is cached and not accessed through this method again.
            auto& input{m_inputs[m_input_tail++]};
            // Wait until the coin is ready to be read. We need acquire so we match the worker thread's release.
            input.ready.wait(/*old=*/false, std::memory_order_acquire);
            // We can move the coin since we won't access this input again.
            return std::move(input.coin);
        }

        // We will only get here for BIP30 checks, an invalid block, or if the threadpool has not been started.
        return base->PeekCoin(outpoint);
    }

    //! Non-null. May have zero workers when input fetching is disabled.
    std::shared_ptr<ThreadPool> m_thread_pool;
    std::vector<std::future<void>> m_futures{};

protected:
    void Reset() noexcept override
    {
        StopFetching();
        CCoinsViewCache::Reset();
    }

public:
    explicit CoinsViewOverlay(CCoinsView* in_base, std::shared_ptr<ThreadPool> thread_pool,
                              bool deterministic = false) noexcept
        : CCoinsViewCache{in_base, deterministic}, m_thread_pool{std::move(thread_pool)}
    {
        Assert(m_thread_pool);
    }

    ~CoinsViewOverlay() noexcept override { StopFetching(); }

    //! Start fetching inputs from block.
    [[nodiscard]] ResetGuard StartFetching(const CBlock& block LIFETIMEBOUND) noexcept;

    void Flush(bool reallocate_cache = true) override
    {
        if (!Assume(AllInputsConsumed())) {
            LogWarning("Block %s input prevout prefetch queue was not fully consumed; inputs were accessed out of order, so prefetching degraded to serial lookups for this block.", GetBestBlock().ToString());
        }
        StopFetching();
        CCoinsViewCache::Flush(reallocate_cache);
    }

    //! Verify that all parallel fetched input prevouts have been consumed.
    bool AllInputsConsumed() const noexcept { return m_input_tail == m_inputs.size(); }
};

//! Utility function to add all of a transaction's outputs to a cache.
//! When check is false, this assumes that overwrites are only possible for coinbase transactions.
//! When check is true, the underlying view may be queried to determine whether an addition is
//! an overwrite.
// TODO: pass in a boolean to limit these possible overwrites to known
// (pre-BIP34) cases.
void AddCoins(CCoinsViewCache& cache, const CTransaction& tx, int nHeight, bool check = false);

//! Utility function to find any unspent output with a given txid.
//! This function can be quite expensive because in the event of a transaction
//! which is not found in the cache, it can cause up to MAX_OUTPUTS_PER_BLOCK
//! lookups to database, so it should be used with care.
const Coin& AccessByTxid(const CCoinsViewCache& cache, const Txid& txid);

/**
 * This is a minimally invasive approach to shutdown on LevelDB read errors from the
 * chainstate, while keeping user interface out of the common library, which is shared
 * between bitcoind, and bitcoin-qt and non-server tools.
 *
 * Writes do not need similar protection, as failure to write is handled by the caller.
*/
class CCoinsViewErrorCatcher final : public CCoinsViewBacked
{
public:
    explicit CCoinsViewErrorCatcher(CCoinsView* view) : CCoinsViewBacked(view) {}

    void AddReadErrCallback(std::function<void()> f) {
        m_err_callbacks.emplace_back(std::move(f));
    }

    std::optional<Coin> GetCoin(const COutPoint& outpoint) const override;
    bool HaveCoin(const COutPoint& outpoint) const override;
    std::optional<Coin> PeekCoin(const COutPoint& outpoint) const override;

private:
    /** A list of callbacks to execute upon leveldb read error. */
    std::vector<std::function<void()>> m_err_callbacks;

};

#endif // BITCOIN_COINS_H

Coverage Report

Created: 2026-07-23 20:35