// TinySafeBoot on libavr — tier 1: pure, idiomatic C++.
//
// A ≤512-byte serial flash bootloader for the ATmega328P boot section,
// reimplementing the TinySafeBoot wire protocol (native-UART fixed-baud
// lineage) on libavr. This variant is written for clarity: well-factored
// functions, no compiler-specific size hacks, no inline assembly. The only
// attributes are the ones the task inherently needs — [[gnu::progmem]] for the
// flash-resident info block and the naked reset entry that stands in for the
// absent C runtime.
//
// The structure follows the hand-written reference: one SRAM page buffer that
// every page transfer shares, separate flash/EEPROM leaf routines (so nothing
// is duplicated by constant propagation), and the polled `unused` interrupt
// posture so every SPM/EEPROM lock folds away.

#include <libavr/libavr.hpp>

#include <avr/io.h> // SP / RAMEND for the crt-free boot entry

using namespace avr::literals;
namespace spm = avr::spm;
namespace ee = avr::eeprom;

using dev = avr::device<{.clock = 16_MHz}>;
using serial_t = dev::uart0<{.baud = 115200_Bd, .max_baud_error = 3_pct}>;
inline constexpr serial_t serial{};

namespace tsb {

// The loader is purely polled — it never enables interrupts — so every SPM and
// EEPROM lock folds to nothing under this posture.
constexpr auto off = avr::irq::guard_policy::unused;

// The two handshake bytes, identical across every TSB host.
constexpr std::uint8_t confirm = '!';
constexpr std::uint8_t request = '?';

// Boot geometry for the ATmega328P 512-byte boot section (BOOTSZ=11). The page
// size, flash and EEPROM extents are the chip database's to know. app_end is
// both the first byte the loader protects and the config page (the LASTPAGE
// holding app-jump vector, timeout and password), one page below the boot
// section.
constexpr std::uint16_t page = spm::page_bytes;
constexpr std::uint16_t boot_bytes = 1024;
constexpr std::uint16_t app_end = spm::flash_bytes - boot_bytes - page;
constexpr std::uint16_t eeprom_end = avr::hw::db.mem.eeprom_size - 1;

// Firmware version stamp: YY*512 + MM*32 + DD, the encoding the host decodes.
constexpr std::uint16_t build_date = 26 * 512 + 7 * 32 + 19;

// The 16-byte device-info block the host reads on activation. Flash-resident so
// it needs no .data (there is no crt to copy it).
// clang-format off
[[gnu::progmem]] constexpr std::uint8_t info[16] = {
    'T', 'S', 'B',
    build_date & 0xFF, build_date >> 8,
    0xF3,             // status byte (native-UART fixed-baud lineage)
    0x1E, 0x95, 0x0F, // ATmega328P signature
    page / 2,         // page size in words
    (app_end / 2) & 0xFF, (app_end / 2) >> 8, // app-flash boundary, words
    eeprom_end & 0xFF, eeprom_end >> 8,
    0xAA, 0xAA, // ATmega processor-type marker (bytes 14 == 15)
};
// clang-format on

// One page staged in SRAM. Scratch that is always filled before it is read, so
// it lives in .noinit — no startup clear (there is no crt to run one) and no
// bytes in .text, which is the only thing the boot-section budget counts.
[[gnu::section(".noinit")]] std::uint8_t buffer[page];

std::uint8_t rx()
{
	for (;;)
		if (auto byte = serial.read())
			return *byte;
}

void tx(std::uint8_t byte)
{
	serial.write(byte);
}

const std::uint8_t *flash_ptr(std::uint16_t addr)
{
	return reinterpret_cast<const std::uint8_t *>(addr);
}

// Stream `count` bytes to the host, from flash (LPM) or from EEPROM.
void send_flash(std::uint16_t addr, std::uint16_t count)
{
	while (count--)
		tx(avr::flash_load(flash_ptr(addr++)));
}

void send_eeprom(std::uint16_t addr, std::uint16_t count)
{
	while (count--)
		tx(ee::read(addr++));
}

// Take one page from the host into the SRAM buffer.
void get_page()
{
	for (std::uint16_t i = 0; i < page; ++i)
		buffer[i] = rx();
}

// Prompt the host with '?' and report whether it answered '!'.
bool request_confirm()
{
	tx(request);
	return rx() == confirm;
}

// Program the SRAM buffer into one already-erased flash page (low byte then
// high, as the SPM word buffer wants).
void write_flash_page(std::uint16_t addr)
{
	for (std::uint16_t i = 0; i < page; i += 2)
		spm::fill<off>(addr + i, static_cast<std::uint16_t>(buffer[i] | (buffer[i + 1] << 8)));
	spm::write_page<off>(addr);
	spm::wait();
}

// Write the SRAM buffer into EEPROM byte by byte.
void write_eeprom_page(std::uint16_t addr)
{
	for (std::uint16_t i = 0; i < page; ++i)
		ee::write<off>(addr + i, buffer[i]);
}

// Run the application: reset vector at 0x0000. Any non-command byte, a wrong
// password, or an idle programmer port lands here.
[[noreturn]] void appjump()
{
	spm::wait(); // make sure any pending SPM finished before handing over
	reinterpret_cast<void (*)()>(0)();
	__builtin_unreachable();
}

// 'f': stream the application flash back, one page per host '!'. Self-terminates
// at the application boundary; the host normally stops earlier with a non-'!'.
void read_flash()
{
	for (std::uint16_t a = 0; a < app_end; a += page) {
		if (rx() != confirm)
			return;
		send_flash(a, page);
	}
}

// 'e': stream EEPROM back, one page per host '!', until the host stops.
void read_eeprom()
{
	for (std::uint16_t a = 0;; a += page) {
		if (rx() != confirm)
			return;
		send_eeprom(a, page);
	}
}

// 'F': erase the whole application first (unwritten pages stay erased), then
// take pages the host offers behind '?'.
void write_flash()
{
	for (std::uint16_t a = 0; a < app_end; a += page) {
		spm::erase_page<off>(a);
		spm::wait();
	}
	for (std::uint16_t a = 0; request_confirm(); a += page) {
		get_page();
		write_flash_page(a);
	}
	spm::rww_enable<off>();
}

// 'E': take pages the host offers behind '?' into EEPROM.
void write_eeprom()
{
	for (std::uint16_t a = 0; request_confirm(); a += page) {
		get_page();
		write_eeprom_page(a);
	}
}

// 'C': replace the config page, then echo it back for the host to verify.
void write_config()
{
	if (!request_confirm())
		return;
	get_page();
	spm::erase_page<off>(app_end);
	spm::wait();
	write_flash_page(app_end);
	spm::rww_enable<off>();
	send_flash(app_end, page);
}

[[noreturn]] void run()
{
	// A bootloader may be entered by a watchdog reset; the reference loader
	// hands straight back to the application in that case rather than run.
	if (avr::hw::reg<"MCUSR">::read() & avr::hw::field<"MCUSR", "WDRF">{}(1).value)
		appjump();

	avr::init<serial_t>();

	// Activation: the host knocks three '@'. With no programmer attached the
	// port stays idle, so a bounded wait boots the application instead of
	// hanging forever.
	std::uint8_t knocks = 0;
	std::uint32_t idle = 4000000;
	while (knocks < 3) {
		if (auto byte = serial.read())
			knocks = *byte == '@' ? knocks + 1 : 0;
		else if (--idle == 0)
			appjump();
	}

	// Password gate: the config page holds it at app_end+3, terminated by 0xff.
	// A blank page (0xff there) means no password. A wrong byte hangs the loader
	// silently, as TSB does.
	for (const std::uint8_t *pw = flash_ptr(app_end + 3); avr::flash_load(pw) != 0xff; ++pw)
		if (rx() != avr::flash_load(pw))
			for (;;) {
			}

	send_flash(reinterpret_cast<std::uint16_t>(&info[0]), sizeof(info));

	for (;;) {
		tx(confirm); // Mainloop ready
		switch (rx()) {
		case 'f':
			read_flash();
			break;
		case 'F':
			write_flash();
			break;
		case 'e':
			read_eeprom();
			break;
		case 'E':
			write_eeprom();
			break;
		case 'c':
			send_flash(app_end, page);
			break;
		case 'C':
			write_config();
			break;
		default:
			appjump(); // 'q' or any other byte runs the application
		}
	}
}

} // namespace tsb

// Reset lands here: BOOTRST vectors to the boot section base and .vectors is
// laid first, so this is the first instruction executed. No crt ran, so set the
// stack pointer before anything is called.
extern "C" [[gnu::naked, gnu::used, gnu::section(".vectors")]] void __boot_entry()
{
	SP = RAMEND;
	tsb::run();
}