CoopAllTheThings/common/include/coop/log_ring.hpp

// Shared-memory log channel: the injected hook (coop_hook.dll) streams its log
// lines to the host (coop_host.exe), which shows them in a Log window. Separate
// mapping from the input/status SharedBlock, named coop_log_<pid>.
//
// Lossy multi-producer / single-consumer ring: the hook logs from several threads
// (worker, audio render thread, window thread), so producers claim a slot with an
// atomic fetch_add and publish each record with a release store of its sequence;
// the host consumer reads in order and tolerates losing the oldest lines if it
// ever falls a whole ring behind (fine for diagnostics). POD + version-locked.
#pragma once

#include <atomic>
#include <cstdint>
#include <cstring>
#include <string>

namespace coop
{

// 'CLOG' little-endian.
inline constexpr std::uint32_t kLogRingMagic = 0x474F4C43u;
inline constexpr std::uint32_t kLogRingVersion = 1;

// Per-pid mapping name, mirroring the other channels: coop_log_<pid>.
inline constexpr wchar_t kLogRingPrefix[] = L"Local\\coop_log_";

inline constexpr std::uint32_t kLogMsgLen = 192;  // chars per line (incl. NUL)
inline constexpr std::uint32_t kLogCapacity = 1024; // ring records

// Severity of a log line; drives the host Log window's colour. Stored in
// LogRecord::level. Info is 0 so existing/zero-filled records read as Info.
enum LogLevel : std::uint32_t
{
	LogLevel_Info = 0,
	LogLevel_Warn = 1,
	LogLevel_Error = 2,
};

struct LogRecord
{
	std::atomic<std::uint64_t> seq; // 0 = empty; else (global index + 1) once written
	std::uint32_t pid;
	std::uint32_t level; // LogLevel
	std::uint64_t millis; // producer timestamp (GetTickCount64)
	char text[kLogMsgLen];
};

struct LogRing
{
	std::uint32_t magic;
	std::uint32_t version;
	std::uint32_t capacity; // number of records
	std::uint32_t msg_len;  // kLogMsgLen (sanity)
	std::atomic<std::uint64_t> write_index; // total records ever claimed (free-running)
	std::uint8_t reserved[32];
	// LogRecord records[capacity] follows immediately.
};

static_assert(std::atomic<std::uint64_t>::is_always_lock_free,
              "log ring needs a lock-free 64-bit atomic for cross-process use");

inline constexpr std::size_t log_ring_total_size(std::uint32_t capacity)
{
	return sizeof(LogRing) + static_cast<std::size_t>(capacity) * sizeof(LogRecord);
}

inline LogRecord* log_ring_records(LogRing* r)
{
	return reinterpret_cast<LogRecord*>(reinterpret_cast<std::uint8_t*>(r) + sizeof(LogRing));
}

// Host: stamp a freshly created (zero-filled) mapping. Records start empty (seq 0).
inline void log_ring_init(LogRing& r, std::uint32_t capacity)
{
	r.capacity = capacity;
	r.msg_len = kLogMsgLen;
	r.write_index.store(0, std::memory_order_relaxed);
	std::memset(r.reserved, 0, sizeof(r.reserved));
	r.version = kLogRingVersion;
	r.magic = kLogRingMagic; // last
}

inline bool log_ring_valid(const LogRing& r)
{
	return r.magic == kLogRingMagic && r.version == kLogRingVersion && r.capacity != 0 &&
	       r.msg_len == kLogMsgLen;
}

// Producer (hook): append a line at severity `level` (LogLevel). Multi-producer safe.
inline void log_ring_push(LogRing& r, std::uint32_t pid, std::uint32_t level, std::uint64_t millis,
                          const char* text)
{
	const std::uint64_t idx = r.write_index.fetch_add(1, std::memory_order_acq_rel);
	LogRecord& rec = log_ring_records(&r)[idx % r.capacity];
	// Seqlock write. Mark the slot in-progress (seq 0) and fence BEFORE touching the record, so a
	// consumer still reading the slot's previous occupant sees seq change and bails instead of reading
	// half-overwritten text; publish the new generation only after the text is fully written. Without
	// this the consumer's single seq check passed before the read, so an overwrite mid-read tore it.
	rec.seq.store(0, std::memory_order_relaxed);
	std::atomic_thread_fence(std::memory_order_release);
	rec.pid = pid;
	rec.level = level;
	rec.millis = millis;
	std::strncpy(rec.text, text, kLogMsgLen - 1);
	rec.text[kLogMsgLen - 1] = '\0';
	std::atomic_thread_fence(std::memory_order_release);
	rec.seq.store(idx + 1, std::memory_order_relaxed); // publish: generation idx is ready
}

// Consumer (host): emit each new record since `cursor` (advanced in place). Skips
// records lost to ring wrap; stops at an in-flight record and retries next call.
template <typename F>
inline void log_ring_drain(LogRing& r, std::uint64_t& cursor, F&& emit)
{
	const std::uint64_t w = r.write_index.load(std::memory_order_acquire);
	if (w <= cursor)
	{
		return;
	}
	const std::uint64_t lo = (w > r.capacity) ? (w - r.capacity) : 0;
	std::uint64_t i = cursor < lo ? lo : cursor; // skip records already overwritten
	LogRecord* recs = log_ring_records(&r);
	for (; i < w; ++i)
	{
		LogRecord& rec = recs[i % r.capacity];
		const std::uint64_t s1 = rec.seq.load(std::memory_order_acquire);
		if (s1 <= i)
		{
			break; // generation i not written yet (in-flight, or being overwritten); retry next call
		}
		if (s1 != i + 1)
		{
			continue; // s1 > i+1: overwritten by a later generation before we got here; lost, skip
		}
		// Seqlock read: copy the record out, then re-check seq. A producer overwriting this slot stores
		// seq 0 before it writes and the new generation after, so any change means our copy may be torn.
		LogRecord snap{};
		snap.pid = rec.pid;
		snap.level = rec.level;
		snap.millis = rec.millis;
		std::memcpy(snap.text, rec.text, kLogMsgLen);
		std::atomic_thread_fence(std::memory_order_acquire);
		if (rec.seq.load(std::memory_order_relaxed) == i + 1)
		{
			emit(snap); // consistent snapshot
		}
		// else: overwritten while we copied -> skip (lost)
	}
	cursor = i;
}

inline std::wstring log_ring_name(unsigned long target_pid)
{
	return std::wstring(kLogRingPrefix) + std::to_wstring(target_pid);
}

} // namespace coop