// SPDX-License-Identifier: GPL-2.0
// ff-protocol-former.c - a part of driver for RME Fireface series
//
// Copyright (c) 2019 Takashi Sakamoto

#include <linux/delay.h>

#include "ff.h"

#define FORMER_REG_SYNC_STATUS		0x0000801c0000ull
/* For block write request. */
#define FORMER_REG_FETCH_PCM_FRAMES	0x0000801c0000ull
#define FORMER_REG_CLOCK_CONFIG		0x0000801c0004ull

static int parse_clock_bits(u32 data, unsigned int *rate,
			    enum snd_ff_clock_src *src)
{
	static const struct {
		unsigned int rate;
		u32 mask;
	} *rate_entry, rate_entries[] = {
		{  32000, 0x00000002, },
		{  44100, 0x00000000, },
		{  48000, 0x00000006, },
		{  64000, 0x0000000a, },
		{  88200, 0x00000008, },
		{  96000, 0x0000000e, },
		{ 128000, 0x00000012, },
		{ 176400, 0x00000010, },
		{ 192000, 0x00000016, },
	};
	static const struct {
		enum snd_ff_clock_src src;
		u32 mask;
	} *clk_entry, clk_entries[] = {
		{ SND_FF_CLOCK_SRC_ADAT1,	0x00000000, },
		{ SND_FF_CLOCK_SRC_ADAT2,	0x00000400, },
		{ SND_FF_CLOCK_SRC_SPDIF,	0x00000c00, },
		{ SND_FF_CLOCK_SRC_WORD,	0x00001000, },
		{ SND_FF_CLOCK_SRC_LTC,		0x00001800, },
	};
	int i;

	for (i = 0; i < ARRAY_SIZE(rate_entries); ++i) {
		rate_entry = rate_entries + i;
		if ((data & 0x0000001e) == rate_entry->mask) {
			*rate = rate_entry->rate;
			break;
		}
	}
	if (i == ARRAY_SIZE(rate_entries))
		return -EIO;

	if (data & 0x00000001) {
		*src = SND_FF_CLOCK_SRC_INTERNAL;
	} else {
		for (i = 0; i < ARRAY_SIZE(clk_entries); ++i) {
			clk_entry = clk_entries + i;
			if ((data & 0x00001c00) == clk_entry->mask) {
				*src = clk_entry->src;
				break;
			}
		}
		if (i == ARRAY_SIZE(clk_entries))
			return -EIO;
	}

	return 0;
}

static int former_get_clock(struct snd_ff *ff, unsigned int *rate,
			    enum snd_ff_clock_src *src)
{
	__le32 reg;
	u32 data;
	int err;

	err = snd_fw_transaction(ff->unit, TCODE_READ_QUADLET_REQUEST,
				 FORMER_REG_CLOCK_CONFIG, &reg, sizeof(reg), 0);
	if (err < 0)
		return err;
	data = le32_to_cpu(reg);

	return parse_clock_bits(data, rate, src);
}

static int former_switch_fetching_mode(struct snd_ff *ff, bool enable)
{
	unsigned int count;
	__le32 *reg;
	int i;
	int err;

	count = 0;
	for (i = 0; i < SND_FF_STREAM_MODE_COUNT; ++i)
		count = max(count, ff->spec->pcm_playback_channels[i]);

	reg = kcalloc(count, sizeof(__le32), GFP_KERNEL);
	if (!reg)
		return -ENOMEM;

	if (!enable) {
		/*
		 * Each quadlet is corresponding to data channels in a data
		 * blocks in reverse order. Precisely, quadlets for available
		 * data channels should be enabled. Here, I take second best
		 * to fetch PCM frames from all of data channels regardless of
		 * stf.
		 */
		for (i = 0; i < count; ++i)
			reg[i] = cpu_to_le32(0x00000001);
	}

	err = snd_fw_transaction(ff->unit, TCODE_WRITE_BLOCK_REQUEST,
				 FORMER_REG_FETCH_PCM_FRAMES, reg,
				 sizeof(__le32) * count, 0);
	kfree(reg);
	return err;
}

static void dump_clock_config(struct snd_ff *ff, struct snd_info_buffer *buffer)
{
	__le32 reg;
	u32 data;
	unsigned int rate;
	enum snd_ff_clock_src src;
	const char *label;
	int err;

	err = snd_fw_transaction(ff->unit, TCODE_READ_BLOCK_REQUEST,
				 FORMER_REG_CLOCK_CONFIG, &reg, sizeof(reg), 0);
	if (err < 0)
		return;
	data = le32_to_cpu(reg);

	snd_iprintf(buffer, "Output S/PDIF format: %s (Emphasis: %s)\n",
		    (data & 0x00000020) ? "Professional" : "Consumer",
		    (data & 0x00000040) ? "on" : "off");

	snd_iprintf(buffer, "Optical output interface format: %s\n",
		    (data & 0x00000100) ? "S/PDIF" : "ADAT");

	snd_iprintf(buffer, "Word output single speed: %s\n",
		    (data & 0x00002000) ? "on" : "off");

	snd_iprintf(buffer, "S/PDIF input interface: %s\n",
		    (data & 0x00000200) ? "Optical" : "Coaxial");

	err = parse_clock_bits(data, &rate, &src);
	if (err < 0)
		return;
	label = snd_ff_proc_get_clk_label(src);
	if (!label)
		return;

	snd_iprintf(buffer, "Clock configuration: %d %s\n", rate, label);
}

static void dump_sync_status(struct snd_ff *ff, struct snd_info_buffer *buffer)
{
	static const struct {
		char *const label;
		u32 locked_mask;
		u32 synced_mask;
	} *clk_entry, clk_entries[] = {
		{ "WDClk",	0x40000000, 0x20000000, },
		{ "S/PDIF",	0x00080000, 0x00040000, },
		{ "ADAT1",	0x00000400, 0x00001000, },
		{ "ADAT2",	0x00000800, 0x00002000, },
	};
	static const struct {
		char *const label;
		u32 mask;
	} *referred_entry, referred_entries[] = {
		{ "ADAT1",	0x00000000, },
		{ "ADAT2",	0x00400000, },
		{ "S/PDIF",	0x00c00000, },
		{ "WDclk",	0x01000000, },
		{ "TCO",	0x01400000, },
	};
	static const struct {
		unsigned int rate;
		u32 mask;
	} *rate_entry, rate_entries[] = {
		{ 32000,	0x02000000, },
		{ 44100,	0x04000000, },
		{ 48000,	0x06000000, },
		{ 64000,	0x08000000, },
		{ 88200,	0x0a000000, },
		{ 96000,	0x0c000000, },
		{ 128000,	0x0e000000, },
		{ 176400,	0x10000000, },
		{ 192000,	0x12000000, },
	};
	__le32 reg[2];
	u32 data[2];
	int i;
	int err;

	err = snd_fw_transaction(ff->unit, TCODE_READ_BLOCK_REQUEST,
				 FORMER_REG_SYNC_STATUS, reg, sizeof(reg), 0);
	if (err < 0)
		return;
	data[0] = le32_to_cpu(reg[0]);
	data[1] = le32_to_cpu(reg[1]);

	snd_iprintf(buffer, "External source detection:\n");

	for (i = 0; i < ARRAY_SIZE(clk_entries); ++i) {
		const char *state;

		clk_entry = clk_entries + i;
		if (data[0] & clk_entry->locked_mask) {
			if (data[0] & clk_entry->synced_mask)
				state = "sync";
			else
				state = "lock";
		} else {
			state = "none";
		}

		snd_iprintf(buffer, "%s: %s\n", clk_entry->label, state);
	}

	snd_iprintf(buffer, "Referred clock:\n");

	if (data[1] & 0x00000001) {
		snd_iprintf(buffer, "Internal\n");
	} else {
		unsigned int rate;
		const char *label;

		for (i = 0; i < ARRAY_SIZE(referred_entries); ++i) {
			referred_entry = referred_entries + i;
			if ((data[0] & 0x1e0000) == referred_entry->mask) {
				label = referred_entry->label;
				break;
			}
		}
		if (i == ARRAY_SIZE(referred_entries))
			label = "none";

		for (i = 0; i < ARRAY_SIZE(rate_entries); ++i) {
			rate_entry = rate_entries + i;
			if ((data[0] & 0x1e000000) == rate_entry->mask) {
				rate = rate_entry->rate;
				break;
			}
		}
		if (i == ARRAY_SIZE(rate_entries))
			rate = 0;

		snd_iprintf(buffer, "%s %d\n", label, rate);
	}
}

static void former_dump_status(struct snd_ff *ff,
			       struct snd_info_buffer *buffer)
{
	dump_clock_config(ff, buffer);
	dump_sync_status(ff, buffer);
}

static int former_fill_midi_msg(struct snd_ff *ff,
				struct snd_rawmidi_substream *substream,
				unsigned int port)
{
	u8 *buf = (u8 *)ff->msg_buf[port];
	int len;
	int i;

	len = snd_rawmidi_transmit_peek(substream, buf,
					SND_FF_MAXIMIM_MIDI_QUADS);
	if (len <= 0)
		return len;

	// One quadlet includes one byte.
	for (i = len - 1; i >= 0; --i)
		ff->msg_buf[port][i] = cpu_to_le32(buf[i]);
	ff->rx_bytes[port] = len;

	return len;
}

#define FF800_STF		0x0000fc88f000
#define FF800_RX_PACKET_FORMAT	0x0000fc88f004
#define FF800_ALLOC_TX_STREAM	0x0000fc88f008
#define FF800_ISOC_COMM_START	0x0000fc88f00c
#define   FF800_TX_S800_FLAG	0x00000800
#define FF800_ISOC_COMM_STOP	0x0000fc88f010

#define FF800_TX_PACKET_ISOC_CH	0x0000801c0008

static int allocate_tx_resources(struct snd_ff *ff)
{
	__le32 reg;
	unsigned int count;
	unsigned int tx_isoc_channel;
	int err;

	reg = cpu_to_le32(ff->tx_stream.data_block_quadlets);
	err = snd_fw_transaction(ff->unit, TCODE_WRITE_QUADLET_REQUEST,
				 FF800_ALLOC_TX_STREAM, &reg, sizeof(reg), 0);
	if (err < 0)
		return err;

	// Wait till the format of tx packet is available.
	count = 0;
	while (count++ < 10) {
		u32 data;
		err = snd_fw_transaction(ff->unit, TCODE_READ_QUADLET_REQUEST,
				FF800_TX_PACKET_ISOC_CH, &reg, sizeof(reg), 0);
		if (err < 0)
			return err;

		data = le32_to_cpu(reg);
		if (data != 0xffffffff) {
			tx_isoc_channel = data;
			break;
		}

		msleep(50);
	}
	if (count >= 10)
		return -ETIMEDOUT;

	// NOTE: this is a makeshift to start OHCI 1394 IR context in the
	// channel. On the other hand, 'struct fw_iso_resources.allocated' is
	// not true and it's not deallocated at stop.
	ff->tx_resources.channel = tx_isoc_channel;

	return 0;
}

static int ff800_allocate_resources(struct snd_ff *ff, unsigned int rate)
{
	u32 data;
	__le32 reg;
	int err;

	reg = cpu_to_le32(rate);
	err = snd_fw_transaction(ff->unit, TCODE_WRITE_QUADLET_REQUEST,
				 FF800_STF, &reg, sizeof(reg), 0);
	if (err < 0)
		return err;

	// If starting isochronous communication immediately, change of STF has
	// no effect. In this case, the communication runs based on former STF.
	// Let's sleep for a bit.
	msleep(100);

	// Controllers should allocate isochronous resources for rx stream.
	err = fw_iso_resources_allocate(&ff->rx_resources,
				amdtp_stream_get_max_payload(&ff->rx_stream),
				fw_parent_device(ff->unit)->max_speed);
	if (err < 0)
		return err;

	// Set isochronous channel and the number of quadlets of rx packets.
	// This should be done before the allocation of tx resources to avoid
	// periodical noise.
	data = ff->rx_stream.data_block_quadlets << 3;
	data = (data << 8) | ff->rx_resources.channel;
	reg = cpu_to_le32(data);
	err = snd_fw_transaction(ff->unit, TCODE_WRITE_QUADLET_REQUEST,
				 FF800_RX_PACKET_FORMAT, &reg, sizeof(reg), 0);
	if (err < 0)
		return err;

	return allocate_tx_resources(ff);
}

static int ff800_begin_session(struct snd_ff *ff, unsigned int rate)
{
	unsigned int generation = ff->rx_resources.generation;
	__le32 reg;

	if (generation != fw_parent_device(ff->unit)->card->generation) {
		int err = fw_iso_resources_update(&ff->rx_resources);
		if (err < 0)
			return err;
	}

	reg = cpu_to_le32(0x80000000);
	reg |= cpu_to_le32(ff->tx_stream.data_block_quadlets);
	if (fw_parent_device(ff->unit)->max_speed == SCODE_800)
		reg |= cpu_to_le32(FF800_TX_S800_FLAG);
	return snd_fw_transaction(ff->unit, TCODE_WRITE_QUADLET_REQUEST,
				 FF800_ISOC_COMM_START, &reg, sizeof(reg), 0);
}

static void ff800_finish_session(struct snd_ff *ff)
{
	__le32 reg;

	reg = cpu_to_le32(0x80000000);
	snd_fw_transaction(ff->unit, TCODE_WRITE_QUADLET_REQUEST,
			   FF800_ISOC_COMM_STOP, &reg, sizeof(reg), 0);
}

// Fireface 800 doesn't allow drivers to register lower 4 bytes of destination
// address.
// A write transaction to clear registered higher 4 bytes of destination address
// has an effect to suppress asynchronous transaction from device.
static void ff800_handle_midi_msg(struct snd_ff *ff, unsigned int offset,
				  __le32 *buf, size_t length)
{
	int i;

	for (i = 0; i < length / 4; i++) {
		u8 byte = le32_to_cpu(buf[i]) & 0xff;
		struct snd_rawmidi_substream *substream;

		substream = READ_ONCE(ff->tx_midi_substreams[0]);
		if (substream)
			snd_rawmidi_receive(substream, &byte, 1);
	}
}

const struct snd_ff_protocol snd_ff_protocol_ff800 = {
	.handle_midi_msg	= ff800_handle_midi_msg,
	.fill_midi_msg		= former_fill_midi_msg,
	.get_clock		= former_get_clock,
	.switch_fetching_mode	= former_switch_fetching_mode,
	.allocate_resources	= ff800_allocate_resources,
	.begin_session		= ff800_begin_session,
	.finish_session		= ff800_finish_session,
	.dump_status		= former_dump_status,
};

#define FF400_STF		0x000080100500ull
#define FF400_RX_PACKET_FORMAT	0x000080100504ull
#define FF400_ISOC_COMM_START	0x000080100508ull
#define FF400_TX_PACKET_FORMAT	0x00008010050cull
#define FF400_ISOC_COMM_STOP	0x000080100510ull

// Fireface 400 manages isochronous channel number in 3 bit field. Therefore,
// we can allocate between 0 and 7 channel.
static int ff400_allocate_resources(struct snd_ff *ff, unsigned int rate)
{
	__le32 reg;
	enum snd_ff_stream_mode mode;
	int i;
	int err;

	// Check whether the given value is supported or not.
	for (i = 0; i < CIP_SFC_COUNT; i++) {
		if (amdtp_rate_table[i] == rate)
			break;
	}
	if (i >= CIP_SFC_COUNT)
		return -EINVAL;

	// Set the number of data blocks transferred in a second.
	reg = cpu_to_le32(rate);
	err = snd_fw_transaction(ff->unit, TCODE_WRITE_QUADLET_REQUEST,
				 FF400_STF, &reg, sizeof(reg), 0);
	if (err < 0)
		return err;

	msleep(100);

	err = snd_ff_stream_get_multiplier_mode(i, &mode);
	if (err < 0)
		return err;

	// Keep resources for in-stream.
	ff->tx_resources.channels_mask = 0x00000000000000ffuLL;
	err = fw_iso_resources_allocate(&ff->tx_resources,
			amdtp_stream_get_max_payload(&ff->tx_stream),
			fw_parent_device(ff->unit)->max_speed);
	if (err < 0)
		return err;

	// Keep resources for out-stream.
	ff->rx_resources.channels_mask = 0x00000000000000ffuLL;
	err = fw_iso_resources_allocate(&ff->rx_resources,
			amdtp_stream_get_max_payload(&ff->rx_stream),
			fw_parent_device(ff->unit)->max_speed);
	if (err < 0)
		fw_iso_resources_free(&ff->tx_resources);

	return err;
}

static int ff400_begin_session(struct snd_ff *ff, unsigned int rate)
{
	unsigned int generation = ff->rx_resources.generation;
	__le32 reg;
	int err;

	if (generation != fw_parent_device(ff->unit)->card->generation) {
		err = fw_iso_resources_update(&ff->tx_resources);
		if (err < 0)
			return err;

		err = fw_iso_resources_update(&ff->rx_resources);
		if (err < 0)
			return err;
	}

	// Set isochronous channel and the number of quadlets of received
	// packets.
	reg = cpu_to_le32(((ff->rx_stream.data_block_quadlets << 3) << 8) |
			  ff->rx_resources.channel);
	err = snd_fw_transaction(ff->unit, TCODE_WRITE_QUADLET_REQUEST,
				 FF400_RX_PACKET_FORMAT, &reg, sizeof(reg), 0);
	if (err < 0)
		return err;

	// Set isochronous channel and the number of quadlets of transmitted
	// packet.
	// TODO: investigate the purpose of this 0x80.
	reg = cpu_to_le32((0x80 << 24) |
			  (ff->tx_resources.channel << 5) |
			  (ff->tx_stream.data_block_quadlets));
	err = snd_fw_transaction(ff->unit, TCODE_WRITE_QUADLET_REQUEST,
				 FF400_TX_PACKET_FORMAT, &reg, sizeof(reg), 0);
	if (err < 0)
		return err;

	// Allow to transmit packets.
	reg = cpu_to_le32(0x00000001);
	return snd_fw_transaction(ff->unit, TCODE_WRITE_QUADLET_REQUEST,
				 FF400_ISOC_COMM_START, &reg, sizeof(reg), 0);
}

static void ff400_finish_session(struct snd_ff *ff)
{
	__le32 reg;

	reg = cpu_to_le32(0x80000000);
	snd_fw_transaction(ff->unit, TCODE_WRITE_QUADLET_REQUEST,
			   FF400_ISOC_COMM_STOP, &reg, sizeof(reg), 0);
}

// For Fireface 400, lower 4 bytes of destination address is configured by bit
// flag in quadlet register (little endian) at 0x'0000'801'0051c. Drivers can
// select one of 4 options:
//
// bit flags: offset of destination address
//  - 0x04000000: 0x'....'....'0000'0000
//  - 0x08000000: 0x'....'....'0000'0080
//  - 0x10000000: 0x'....'....'0000'0100
//  - 0x20000000: 0x'....'....'0000'0180
//
// Drivers can suppress the device to transfer asynchronous transactions by
// using below 2 bits.
//  - 0x01000000: suppress transmission
//  - 0x02000000: suppress transmission
//
// Actually, the register is write-only and includes the other options such as
// input attenuation. This driver allocates destination address with '0000'0000
// in its lower offset and expects userspace application to configure the
// register for it.
static void ff400_handle_midi_msg(struct snd_ff *ff, unsigned int offset,
				  __le32 *buf, size_t length)
{
	int i;

	for (i = 0; i < length / 4; i++) {
		u32 quad = le32_to_cpu(buf[i]);
		u8 byte;
		unsigned int index;
		struct snd_rawmidi_substream *substream;

		/* Message in first port. */
		/*
		 * This value may represent the index of this unit when the same
		 * units are on the same IEEE 1394 bus. This driver doesn't use
		 * it.
		 */
		index = (quad >> 8) & 0xff;
		if (index > 0) {
			substream = READ_ONCE(ff->tx_midi_substreams[0]);
			if (substream != NULL) {
				byte = quad & 0xff;
				snd_rawmidi_receive(substream, &byte, 1);
			}
		}

		/* Message in second port. */
		index = (quad >> 24) & 0xff;
		if (index > 0) {
			substream = READ_ONCE(ff->tx_midi_substreams[1]);
			if (substream != NULL) {
				byte = (quad >> 16) & 0xff;
				snd_rawmidi_receive(substream, &byte, 1);
			}
		}
	}
}

const struct snd_ff_protocol snd_ff_protocol_ff400 = {
	.handle_midi_msg	= ff400_handle_midi_msg,
	.fill_midi_msg		= former_fill_midi_msg,
	.get_clock		= former_get_clock,
	.switch_fetching_mode	= former_switch_fetching_mode,
	.allocate_resources	= ff400_allocate_resources,
	.begin_session		= ff400_begin_session,
	.finish_session		= ff400_finish_session,
	.dump_status		= former_dump_status,
};