#include "bba.h"
#include <string.h>
#include <unistd.h>
#include <ogc/lwp_watchdog.h>

u32 bba_speed = EXI_SPEED32MHZ;   /* nominal name; ~27 MHz in reality */
static int begin(void);
static void end(void);

u32 bba_gap_us = 16;              /* see bba.h -- 0 reproduces the raw failure */

/* Busy-wait, not usleep(): we are holding the EXI lock with CS asserted, so
 * yielding here is both unnecessary and imprecise at these timescales. */
static void gap(void)
{
	u64 t0;
	if (!bba_gap_us) return;
	t0 = gettime();
	while (ticks_to_microsecs(diff_ticks(t0, gettime())) < bba_gap_us) ;
}

/* Real GC<->BBA protocol (docs/exi_protocol_real.md), verified against libogc
 * and Swiss.  Register space is a 4-byte header followed by a data phase that
 * runs until CS -- there is no length field.
 *     read   [0x80, addr>>8, addr&0xFF, 0x00] + N bytes
 *     write  [0xC0, addr>>8, addr&0xFF, 0x00] + N bytes
 * The trailing 0x00 is the device's prefetch window: the address completes a
 * whole byte before the data phase so the register file can fetch in time. */
int begin_raw(void) { return begin(); }
void end_raw(void) { end(); }

static int begin(void)
{
	int tries;
	/* EXI_Lock fails outright (rather than blocking) when another subsystem
	 * holds the channel -- Swiss or a libogc driver may have it briefly.
	 * Retry for ~100 ms before giving up so a transient collision does not
	 * surface as a test failure. */
	for (tries = 0; tries < 100; tries++) {
		if (EXI_Lock(BBA_EXI_CHAN, BBA_EXI_DEV, NULL) > 0) break;
		usleep(1000);
	}
	if (tries == 100) return -1;
	if (EXI_Select(BBA_EXI_CHAN, BBA_EXI_DEV, bba_speed) <= 0) {
		EXI_Unlock(BBA_EXI_CHAN);
		return -2;
	}
	return 0;
}

static void end(void)
{
	EXI_Deselect(BBA_EXI_CHAN);
	EXI_Unlock(BBA_EXI_CHAN);
}

static int put_hdr(u16 addr, int write)
{
	u32 h = ((u32)(write ? 0xC0 : 0x80) << 24) | ((u32)addr << 8);
	if (EXI_Imm(BBA_EXI_CHAN, &h, 4, EXI_WRITE, NULL) <= 0) return -3;
	if (EXI_Sync(BBA_EXI_CHAN) <= 0) return -4;
	return 0;
}

int bba_read(u16 addr, void *buf, u32 len)
{
	int rc;
	if (len < 1) return -10;
	if ((rc = begin()) < 0) return rc;
	if ((rc = put_hdr(addr, 0)) < 0) { end(); return rc; }
	gap();
	memset(buf, 0, len);
	if (EXI_ImmEx(BBA_EXI_CHAN, buf, len, EXI_READ) <= 0) { end(); return -5; }
	end();
	return 0;
}

int bba_write(u16 addr, const void *buf, u32 len)
{
	int rc;
	u8 tmp[64];
	if (len < 1 || len > sizeof(tmp)) return -10;
	memcpy(tmp, buf, len);
	if ((rc = begin()) < 0) return rc;
	if ((rc = put_hdr(addr, 1)) < 0) { end(); return rc; }
	if (EXI_ImmEx(BBA_EXI_CHAN, tmp, len, EXI_WRITE) <= 0) { end(); return -6; }
	end();
	return 0;
}

int bba_devid(u8 id[4])
{
	int rc;
	u16 h = 0x0000;          /* command space, reg 0, read */
	if ((rc = begin()) < 0) return rc;
	if (EXI_Imm(BBA_EXI_CHAN, &h, 2, EXI_WRITE, NULL) <= 0) { end(); return -3; }
	if (EXI_Sync(BBA_EXI_CHAN) <= 0) { end(); return -4; }
	gap();
	memset(id, 0, 4);
	if (EXI_ImmEx(BBA_EXI_CHAN, id, 4, EXI_READ) <= 0) { end(); return -5; }
	end();
	return 0;
}

int bba_r8(u16 addr)
{
	u8 v;
	int rc = bba_read(addr, &v, 1);
	return rc < 0 ? rc : (int)v;
}

int bba_w8(u16 addr, u8 val)
{
	return bba_write(addr, &val, 1);
}

int bba_stream(u16 addr, void *buf, u32 len)
{
	/* Identical to bba_read now: the real protocol always streams until CS,
	 * so there is no separate "DMA" header form. */
	return bba_read(addr, buf, len);
}

int bba_tx_bytes(const u8 *buf, u32 len)
{
	u32 off = 0;

	/* Stream in 64-byte chunks, ONE EXI transaction each.
	 *
	 * Writing byte by byte with bba_w8() costs a full EXI_Lock / EXI_Select /
	 * header / EXI_Deselect / EXI_Unlock PER BYTE -- about sixty transactions
	 * for one 60-byte ACK.  That latency is why ~1/3 of netboot blocks were
	 * retried: the ACK always reached the W5100 (txto=0, TI always set) but
	 * missed the host's timeout.
	 *
	 * ⚠️ BUT THE CHUNK SIZE IS BOUNDED BY tx_bytes, WHICH IS 16 DEEP.
	 * BBARegisterFile silently drops bytes when tx_bytes.w_rdy is low,
	 * because SPI cannot back-pressure.  Within one transaction the bytes
	 * arrive back-to-back at EXI speed -- far faster than TXFrameDrain can
	 * start consuming -- so a 64-byte chunk overran the FIFO, the frame went
	 * out truncated, and the host saw no valid ACK at all ("no ACK after 20
	 * tries").  Byte-by-byte worked only because the per-byte EXI overhead
	 * happened to give the drain time.
	 *
	 * tx_bytes is now 2048 deep -- a whole maximum-size frame -- so the old
	 * 8-byte cap is gone and 64 (bba_write()'s buffer limit) sends a 60-byte
	 * ACK in a SINGLE transaction.  The depth is what makes this safe: SPI
	 * cannot back-pressure, so the FIFO must absorb the entire burst. */
	while (off < len) {
		u32 n = len - off;
		int rc;
		if (n > 64) n = 64;
		rc = bba_write(REG_TXDATA, buf + off, n);
		if (rc < 0) return rc;
		off += n;
	}
	return 0;
}