libbitfury.c

/*
 * Copyright 2014 Con Kolivas
 * Copyright 2013 Andrew Smith
 * Copyright 2013 bitfury
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms of the GNU General Public License as published by the Free
 * Software Foundation; either version 3 of the License, or (at your option)
 * any later version.  See COPYING for more details.
 */

#include "miner.h"
#include "driver-bitfury.h"
#include "libbitfury.h"
#include "sha2.h"

void ms3steps(uint32_t *p)
{
	uint32_t a, b, c, d, e, f, g, h, new_e, new_a;
	int i;

	a = p[0];
	b = p[1];
	c = p[2];
	d = p[3];
	e = p[4];
	f = p[5];
	g = p[6];
	h = p[7];
	for (i = 0; i < 3; i++) {
		new_e = p[i+16] + sha256_k[i] + h + CH(e,f,g) + SHA256_F2(e) + d;
		new_a = p[i+16] + sha256_k[i] + h + CH(e,f,g) + SHA256_F2(e) +
			SHA256_F1(a) + MAJ(a,b,c);
		d = c;
		c = b;
		b = a;
		a = new_a;
		h = g;
		g = f;
		f = e;
		e = new_e;
	}
	p[15] = a;
	p[14] = b;
	p[13] = c;
	p[12] = d;
	p[11] = e;
	p[10] = f;
	p[9] = g;
	p[8] = h;
}

uint32_t decnonce(uint32_t in)
{
	uint32_t out;

	/* First part load */
	out = (in & 0xFF) << 24;
	in >>= 8;

	/* Byte reversal */
	in = (((in & 0xaaaaaaaa) >> 1) | ((in & 0x55555555) << 1));
	in = (((in & 0xcccccccc) >> 2) | ((in & 0x33333333) << 2));
	in = (((in & 0xf0f0f0f0) >> 4) | ((in & 0x0f0f0f0f) << 4));

	out |= (in >> 2) & 0x3FFFFF;

	/* Extraction */
	if (in & 1)
		out |= (1 << 23);
	if (in & 2)
		out |= (1 << 22);

	out -= 0x800004;
	return out;
}

/* Test vectors to calculate (using address-translated loads) */
static unsigned int atrvec[] = {
	0xb0e72d8e, 0x1dc5b862, 0xe9e7c4a6, 0x3050f1f5, 0x8a1a6b7e, 0x7ec384e8, 0x42c1c3fc, 0x8ed158a1, /* MIDSTATE */
	0,0,0,0,0,0,0,0,
	0x8a0bb7b7, 0x33af304f, 0x0b290c1a, 0xf0c4e61f, /* WDATA: hashMerleRoot[7], nTime, nBits, nNonce */
};
static bool atrvec_set;

void bitfury_work_to_payload(struct bitfury_payload *p, struct work *work)
{
	memcpy(p->midstate, work->midstate, 32);
	p->m7 = *(unsigned int *)(work->data + 64);
	p->ntime = *(unsigned int *)(work->data + 68);
	p->nbits = *(unsigned int *)(work->data + 72);
	applog(LOG_INFO, "INFO nonc: %08x bitfury_scanHash MS0: %08x, ", p->nnonce,
	       ((unsigned int *)work->midstate)[0]);
	applog(LOG_INFO, "INFO merkle[7]: %08x, ntime: %08x, nbits: %08x", p->m7,
	       p->ntime, p->nbits);
}

/* Configuration registers - control oscillators and such stuff. PROGRAMMED when
 * magic number matches, UNPROGRAMMED (default) otherwise */
void spi_config_reg(struct bitfury_info *info, int cfgreg, int ena)
{
	static const uint8_t enaconf[4] = { 0xc1, 0x6a, 0x59, 0xe3 };
	static const uint8_t disconf[4] = { 0, 0, 0, 0 };

	if (ena)
		spi_add_data(info, 0x7000 + cfgreg * 32, enaconf, 4);
	else
		spi_add_data(info, 0x7000 + cfgreg * 32, disconf, 4);
}

void spi_set_freq(struct bitfury_info *info)
{
	uint64_t freq;
	const uint8_t *osc6 = (unsigned char *)&freq;

	freq = (1ULL << info->osc6_bits) - 1ULL;
	spi_add_data(info, 0x6000, osc6, 8); /* Program internal on-die slow oscillator frequency */
}

#define FIRST_BASE 61
#define SECOND_BASE 4

void spi_send_conf(struct bitfury_info *info)
{
	const int8_t nf1_counters[16] = { 64, 64, SECOND_BASE, SECOND_BASE+4, SECOND_BASE+2,
		SECOND_BASE+2+16, SECOND_BASE, SECOND_BASE+1, (FIRST_BASE)%65, (FIRST_BASE+1)%65,
		(FIRST_BASE+3)%65, (FIRST_BASE+3+16)%65, (FIRST_BASE+4)%65, (FIRST_BASE+4+4)%65,
		(FIRST_BASE+3+3)%65, (FIRST_BASE+3+1+3)%65 };
	int i;

	for (i = 7; i <= 11; i++)
		spi_config_reg(info, i, 0);
	spi_config_reg(info, 6, 1); /* disable OUTSLK */
	spi_config_reg(info, 4, 1); /* Enable slow oscillator */
	for (i = 1; i <= 3; ++i)
		spi_config_reg(info, i, 0);
	/* Program counters correctly for rounds processing, here it should
	 * start consuming power */
	spi_add_data(info, 0x0100, nf1_counters, 16);
}

void spi_send_init(struct bitfury_info *info)
{
	/* Prepare internal buffers */
	/* PREPARE BUFFERS (INITIAL PROGRAMMING) */
	unsigned int w[16];

	if (!atrvec_set) {
		atrvec_set = true;
		ms3steps(atrvec);
	}
	memset(w, 0, sizeof(w));
	w[3] = 0xffffffff;
	w[4] = 0x80000000;
	w[15] = 0x00000280;
	spi_add_data(info, 0x1000, w, 16 * 4);
	spi_add_data(info, 0x1400, w, 8 * 4);
	memset(w, 0, sizeof(w));
	w[0] = 0x80000000;
	w[7] = 0x100;
	spi_add_data(info, 0x1900, w, 8 * 4); /* Prepare MS and W buffers! */
	spi_add_data(info, 0x3000, atrvec, 19 * 4);
}
void spi_clear_buf(struct bitfury_info *info)
{
	info->spibufsz = 0;
}

void spi_add_buf(struct bitfury_info *info, const void *buf, const int sz)
{
	if (unlikely(info->spibufsz + sz > SPIBUF_SIZE)) {
		applog(LOG_WARNING, "SPI bufsize overflow!");
		return;
	}
	memcpy(&info->spibuf[info->spibufsz], buf, sz);
	info->spibufsz += sz;
}

void spi_add_break(struct bitfury_info *info)
{
	spi_add_buf(info, "\x4", 1);
}

void spi_add_fasync(struct bitfury_info *info, int n)
{
	int i;

	for (i = 0; i < n; i++)
		spi_add_buf(info, "\x5", 1);
}

static void spi_add_buf_reverse(struct bitfury_info *info, const char *buf, const int sz)
{
	int i;

	for (i = 0; i < sz; i++) { // Reverse bit order in each byte!
		unsigned char p = buf[i];

		p = ((p & 0xaa) >> 1) | ((p & 0x55) << 1);
		p = ((p & 0xcc) >> 2) | ((p & 0x33) << 2);
		p = ((p & 0xf0) >> 4) | ((p & 0x0f) << 4);
		info->spibuf[info->spibufsz + i] = p;
	}
	info->spibufsz += sz;
}

void spi_add_data(struct bitfury_info *info, uint16_t addr, const void *buf, int len)
{
	unsigned char otmp[3];

	if (len < 4 || len > 128) {
		applog(LOG_WARNING, "Can't add SPI data size %d", len);
		return;
	}
	len /= 4; /* Strip */
	otmp[0] = (len - 1) | 0xE0;
	otmp[1] = (addr >> 8) & 0xFF;
	otmp[2] = addr & 0xFF;
	spi_add_buf(info, otmp, 3);
	len *= 4;
	spi_add_buf_reverse(info, buf, len);
}

// Bit-banging reset... Each 3 reset cycles reset first chip in chain
bool spi_reset(struct cgpu_info *bitfury, struct bitfury_info *info)
{
	struct mcp_settings *mcp = &info->mcp;
	int r;

	// SCK_OVRRIDE
	mcp->value.pin[NF1_PIN_SCK_OVR] = MCP2210_GPIO_PIN_HIGH;
	mcp->direction.pin[NF1_PIN_SCK_OVR] = MCP2210_GPIO_OUTPUT;
	mcp->designation.pin[NF1_PIN_SCK_OVR] = MCP2210_PIN_GPIO;
	if (!mcp2210_set_gpio_settings(bitfury, mcp))
		return false;

	for (r = 0; r < 16; ++r) {
		char buf[1] = {0x81}; // will send this waveform: - _ _ _ _ _ _ -
		unsigned int length = 1;

		if (!mcp2210_spi_transfer(bitfury, &info->mcp, buf, &length))
			return false;
	}

	// Deactivate override
	mcp->direction.pin[NF1_PIN_SCK_OVR] = MCP2210_GPIO_INPUT;
	if (!mcp2210_set_gpio_settings(bitfury, mcp))
		return false;

	return true;
}

bool mcp_spi_txrx(struct cgpu_info *bitfury, struct bitfury_info *info)
{
	unsigned int length, sendrcv;
	int offset = 0;

	length = info->spibufsz;
	applog(LOG_DEBUG, "%s %d: SPI sending %u bytes total", bitfury->drv->name,
	       bitfury->device_id, length);
	while (length > MCP2210_TRANSFER_MAX) {
		sendrcv = MCP2210_TRANSFER_MAX;
		if (!mcp2210_spi_transfer(bitfury, &info->mcp, info->spibuf + offset, &sendrcv))
			return false;
		if (sendrcv != MCP2210_TRANSFER_MAX) {
			applog(LOG_DEBUG, "%s %d: Send/Receive size mismatch sent %d received %d",
			       bitfury->drv->name, bitfury->device_id, MCP2210_TRANSFER_MAX, sendrcv);
		}
		length -= MCP2210_TRANSFER_MAX;
		offset += MCP2210_TRANSFER_MAX;
	}
	sendrcv = length;
	if (!mcp2210_spi_transfer(bitfury, &info->mcp, info->spibuf + offset, &sendrcv))
		return false;
	if (sendrcv != length) {
		applog(LOG_WARNING, "%s %d: Send/Receive size mismatch sent %d received %d",
		       bitfury->drv->name, bitfury->device_id, length, sendrcv);
		return false;
	}
	return true;
}

#define READ_WRITE_BYTES_SPI0     0x31

bool ftdi_spi_txrx(struct cgpu_info *bitfury, struct bitfury_info *info)
{
	int err, amount, len;
	uint16_t length;
	char buf[1024];

	len = info->spibufsz;
	length = info->spibufsz - 1; //FTDI length is shifted by one 0x0000 = one byte
	buf[0] = READ_WRITE_BYTES_SPI0;
	buf[1] = length & 0x00FF;
	buf[2] = (length & 0xFF00) >> 8;
	memcpy(&buf[3], info->spibuf, info->spibufsz);
	info->spibufsz += 3;
	err = usb_write(bitfury, buf, info->spibufsz, &amount, C_BXM_SPITX);
	if (err || amount != (int)info->spibufsz) {
		applog(LOG_ERR, "%s %d: SPI TX error %d, sent %d of %d", bitfury->drv->name,
		       bitfury->device_id, err, amount, info->spibufsz);
		return false;
	}
	info->spibufsz = len;
	/* We shouldn't even get a timeout error on reads in spi mode */
	err = usb_read(bitfury, info->spibuf, len, &amount, C_BXM_SPIRX);
	if (err || amount != len) {
		applog(LOG_ERR, "%s %d: SPI RX error %d, read %d of %d", bitfury->drv->name,
		       bitfury->device_id, err, amount, info->spibufsz);
		return false;
	}
	amount = usb_buffer_size(bitfury);
	if (amount) {
		applog(LOG_ERR, "%s %d: SPI RX Extra read buffer size %d", bitfury->drv->name,
		       bitfury->device_id, amount);
		usb_buffer_clear(bitfury);
		return false;
	}
	return true;
}

#define BT_OFFSETS 3

bool bitfury_checkresults(struct thr_info *thr, struct work *work, uint32_t nonce)
{
	const uint32_t bf_offsets[] = {-0x800000, 0, -0x400000};
	int i;

	for (i = 0; i < BT_OFFSETS; i++) {
		uint32_t noffset = nonce + bf_offsets[i];

		if (test_nonce(work, noffset)) {
			submit_tested_work(thr, work);
			return true;
		}
	}
	return false;
}

/* Currently really only supports 2 chips, so chip_n can only be 0 or 1 */
bool libbitfury_sendHashData(struct thr_info *thr, struct cgpu_info *bitfury,
			     struct bitfury_info *info, int chip_n)
{
	unsigned newbuf[17];
	unsigned *oldbuf = &info->oldbuf[17 * chip_n];
	struct bitfury_payload *p = &info->payload[chip_n];
	unsigned int localvec[20];

	/* Programming next value */
	memcpy(localvec, p, 20 * 4);
	ms3steps(localvec);

	spi_clear_buf(info);
	spi_add_break(info);
	spi_add_fasync(info, chip_n);
	spi_add_data(info, 0x3000, (void*)localvec, 19 * 4);
	if (!info->spi_txrx(bitfury, info))
		return false;

	memcpy(newbuf, info->spibuf + 4 + chip_n, 17 * 4);

	info->job_switched[chip_n] = newbuf[16] != oldbuf[16];

	if (likely(info->second_run[chip_n])) {
		if (info->job_switched[chip_n]) {
			int i;

			for (i = 0; i < 16; i++) {
				if (oldbuf[i] != newbuf[i] && info->owork[chip_n]) {
					uint32_t nonce; //possible nonce

					nonce = decnonce(newbuf[i]);
					if (bitfury_checkresults(thr, info->owork[chip_n], nonce))
						info->nonces++;
				}
			}
			memcpy(oldbuf, newbuf, 17 * 4);
		}
	} else
		info->second_run[chip_n] = true;

	cgsleep_ms(BITFURY_REFRESH_DELAY);

	return true;
}