/*
 * File      : mmcsd_core.c
 * This file is part of RT-Thread RTOS
 * COPYRIGHT (C) 2006, RT-Thread Development Team
 *
 * The license and distribution terms for this file may be
 * found in the file LICENSE in this distribution or at
 * http://www.rt-thread.org/license/LICENSE
 *
 * Change Logs:
 * Date           Author		Notes
 * 2011-07-25     weety		first version
 */

#include <rtthread.h>
#include <drivers/mmcsd_core.h>
#include <drivers/sd.h>

#ifndef RT_MMCSD_STACK_SIZE
#define RT_MMCSD_STACK_SIZE 1024
#endif
#ifndef RT_MMCSD_THREAD_PREORITY
#if (RT_THREAD_PRIORITY_MAX == 32)
#define RT_MMCSD_THREAD_PREORITY  0x16
#else
#define RT_MMCSD_THREAD_PREORITY  0x40
#endif
#endif

//static struct rt_semaphore mmcsd_sem;
static struct rt_thread mmcsd_detect_thread;
static rt_uint8_t mmcsd_stack[RT_MMCSD_STACK_SIZE];
static struct rt_mailbox  mmcsd_detect_mb;
static rt_uint32_t mmcsd_detect_mb_pool[4];

void mmcsd_host_lock(struct rt_mmcsd_host *host)
{
	rt_sem_take(&host->bus_lock, RT_WAITING_FOREVER);
}

void mmcsd_host_unlock(struct rt_mmcsd_host *host)
{
	rt_sem_release(&host->bus_lock);
}

void mmcsd_req_complete(struct rt_mmcsd_host *host)
{
	rt_sem_release(&host->sem_ack);
}

void mmcsd_send_request(struct rt_mmcsd_host *host, struct rt_mmcsd_req *req)
{
    req->cmd->err = 0;
	req->cmd->mrq = req;
	if (req->data)
    {   
        req->cmd->data = req->data;
        req->data->err = 0;
    	req->data->mrq = req;
		if (req->stop)
		{
			req->data->stop = req->stop;
			req->stop->err = 0;
			req->stop->mrq = req;
		}    	
   }
	host->ops->request(host, req);
	rt_sem_take(&host->sem_ack, RT_WAITING_FOREVER);
}

rt_int32_t mmcsd_send_cmd(struct rt_mmcsd_host *host, struct rt_mmcsd_cmd *cmd, int retries)
{
	struct rt_mmcsd_req req;

	rt_memset(&req, 0, sizeof(struct rt_mmcsd_req));
	rt_memset(cmd->resp, 0, sizeof(cmd->resp));

	req.cmd = cmd;
	cmd->data = RT_NULL;

	mmcsd_send_request(host, &req);

	return cmd->err;
}

rt_int32_t mmcsd_go_idle(struct rt_mmcsd_host *host)
{
	rt_int32_t err;
	struct rt_mmcsd_cmd cmd;

	if (!controller_is_spi(host)) {
		mmcsd_set_chip_select(host, MMCSD_CS_HIGH);
		mmcsd_delay_ms(1);
	}

	rt_memset(&cmd, 0, sizeof(struct rt_mmcsd_cmd));

	cmd.cmd_code = GO_IDLE_STATE;
	cmd.arg = 0;
	cmd.flags = RESP_SPI_R1 | RESP_NONE | CMD_BC;

	err = mmcsd_send_cmd(host, &cmd, 0);

	mmcsd_delay_ms(1);

	if (!controller_is_spi(host)) 
	{
		mmcsd_set_chip_select(host, MMCSD_CS_IGNORE);
		mmcsd_delay_ms(1);
	}


	return err;
}

rt_int32_t mmcsd_spi_read_ocr(struct rt_mmcsd_host *host, rt_int32_t high_capacity, rt_uint32_t *ocr)
{
	struct rt_mmcsd_cmd cmd;
	rt_int32_t err;

	rt_memset(&cmd, 0, sizeof(struct rt_mmcsd_cmd));

	cmd.cmd_code = SPI_READ_OCR;
	cmd.arg = high_capacity ? (1 << 30) : 0;
	cmd.flags = RESP_SPI_R3;

	err = mmcsd_send_cmd(host, &cmd, 0);

	*ocr = cmd.resp[1];
	return err;
}


rt_int32_t mmcsd_all_get_cid(struct rt_mmcsd_host *host, rt_uint32_t *cid)
{
	rt_int32_t err;
	struct rt_mmcsd_cmd cmd;

	rt_memset(&cmd, 0, sizeof(struct rt_mmcsd_cmd));

	cmd.cmd_code = ALL_SEND_CID;
	cmd.arg = 0;
	cmd.flags = RESP_R2 | CMD_BCR;

	err = mmcsd_send_cmd(host, &cmd, 3);
	if (err)
		return err;

	rt_memcpy(cid, cmd.resp, sizeof(rt_uint32_t) * 4);

	return 0;
}


rt_int32_t mmcsd_get_cid(struct rt_mmcsd_host *host, rt_uint32_t *cid)
{
	rt_int32_t err, i;
	struct rt_mmcsd_req req;
	struct rt_mmcsd_cmd cmd;
	struct rt_mmcsd_data data;
	rt_uint32_t *buf = RT_NULL;

	if (!controller_is_spi(host)) 
	{
		if (!host->card)
			return -RT_ERROR;
		rt_memset(&cmd, 0, sizeof(struct rt_mmcsd_cmd));

		cmd.cmd_code = SEND_CID;
		cmd.arg = host->card->rca << 16;
		cmd.flags = RESP_R2 | CMD_AC;
		err = mmcsd_send_cmd(host, &cmd, 3);
		if (err)
			return err;

		rt_memcpy(cid, cmd.resp, sizeof(rt_uint32_t) * 4);
		return 0;
	}

	buf = (rt_uint32_t *)rt_malloc(16);
	if (!buf) 
	{
		rt_kprintf("allocate memory failed\n");
		return -RT_ENOMEM;
	}

	rt_memset(&req, 0, sizeof(struct rt_mmcsd_req));
	rt_memset(&cmd, 0, sizeof(struct rt_mmcsd_cmd));
	rt_memset(&data, 0, sizeof(struct rt_mmcsd_data));

	req.cmd = &cmd;
	req.data = &data;

	cmd.cmd_code = SEND_CID;
	cmd.arg = 0;

	/* NOTE HACK:  the RESP_SPI_R1 is always correct here, but we
	 * rely on callers to never use this with "native" calls for reading
	 * CSD or CID.  Native versions of those commands use the R2 type,
	 * not R1 plus a data block.
	 */
	cmd.flags = RESP_SPI_R1 | RESP_R1 | CMD_ADTC;

	data.blksize = 16;
	data.blks = 1;
	data.flags = DATA_DIR_READ;
	data.buf = buf;
	/*
	 * The spec states that CSR and CID accesses have a timeout
	 * of 64 clock cycles.
	 */
	data.timeout_ns = 0;
	data.timeout_clks = 64;

	mmcsd_send_request(host, &req);

	if (cmd.err || data.err)
	{
		rt_free(buf);
		return -RT_ERROR;
	}

	for (i = 0;i < 4;i++)
		cid[i] = buf[i];
	rt_free(buf);

	return 0;
}

rt_int32_t mmcsd_get_csd(struct rt_mmcsd_card *card, rt_uint32_t *csd)
{
	rt_int32_t err, i;
	struct rt_mmcsd_req req;
	struct rt_mmcsd_cmd cmd;
	struct rt_mmcsd_data data;
	rt_uint32_t *buf = RT_NULL;

	if (!controller_is_spi(card->host))
	{
		rt_memset(&cmd, 0, sizeof(struct rt_mmcsd_cmd));

		cmd.cmd_code = SEND_CSD;
		cmd.arg = card->rca << 16;
		cmd.flags = RESP_R2 | CMD_AC;
		err = mmcsd_send_cmd(card->host, &cmd, 3);
		if (err)
			return err;

		rt_memcpy(csd, cmd.resp, sizeof(rt_uint32_t) * 4);
		return 0;
	}

	buf = (rt_uint32_t*)rt_malloc(16);
	if (!buf) 
	{
		rt_kprintf("allocate memory failed\n");
		return -RT_ENOMEM;
	}

	rt_memset(&req, 0, sizeof(struct rt_mmcsd_req));
	rt_memset(&cmd, 0, sizeof(struct rt_mmcsd_cmd));
	rt_memset(&data, 0, sizeof(struct rt_mmcsd_data));

	req.cmd = &cmd;
	req.data = &data;

	cmd.cmd_code = SEND_CSD;
	cmd.arg = 0;

	/* NOTE HACK:  the RESP_SPI_R1 is always correct here, but we
	 * rely on callers to never use this with "native" calls for reading
	 * CSD or CID.  Native versions of those commands use the R2 type,
	 * not R1 plus a data block.
	 */
	cmd.flags = RESP_SPI_R1 | RESP_R1 | CMD_ADTC;

	data.blksize = 16;
	data.blks = 1;
	data.flags = DATA_DIR_READ;
	data.buf = buf;

	/*
	 * The spec states that CSR and CID accesses have a timeout
	 * of 64 clock cycles.
	 */
	data.timeout_ns = 0;
	data.timeout_clks = 64;

	mmcsd_send_request(card->host, &req);

	if (cmd.err || data.err)
	{
		rt_free(buf);
		return -RT_ERROR;
	}

	for (i = 0;i < 4;i++)
		csd[i] = buf[i];
	rt_free(buf);

	return 0;
}

static rt_int32_t _mmcsd_select_card(struct rt_mmcsd_host *host, struct rt_mmcsd_card *card)
{
	rt_int32_t err;
	struct rt_mmcsd_cmd cmd;

	rt_memset(&cmd, 0, sizeof(struct rt_mmcsd_cmd));

	cmd.cmd_code = SELECT_CARD;

	if (card) 
	{
		cmd.arg = card->rca << 16;
		cmd.flags = RESP_R1 | CMD_AC;
	} 
	else 
	{
		cmd.arg = 0;
		cmd.flags = RESP_NONE | CMD_AC;
	}

	err = mmcsd_send_cmd(host, &cmd, 3);
	if (err)
		return err;

	return 0;
}

rt_int32_t mmcsd_select_card(struct rt_mmcsd_card *card)
{
	return _mmcsd_select_card(card->host, card);
}

rt_int32_t mmcsd_deselect_cards(struct rt_mmcsd_card *card)
{
	return _mmcsd_select_card(card->host, RT_NULL);
}

rt_int32_t mmcsd_spi_use_crc(struct rt_mmcsd_host *host, rt_int32_t use_crc)
{
	struct rt_mmcsd_cmd cmd;
	rt_int32_t err;

	rt_memset(&cmd, 0, sizeof(struct rt_mmcsd_cmd));

	cmd.cmd_code = SPI_CRC_ON_OFF;
	cmd.flags = RESP_SPI_R1;
	cmd.arg = use_crc;

	err = mmcsd_send_cmd(host, &cmd, 0);
	if (!err)
		host->spi_use_crc = use_crc;
	return err;
}


rt_inline void mmcsd_set_iocfg(struct rt_mmcsd_host *host)
{
	struct rt_mmcsd_io_cfg *io_cfg = &host->io_cfg;

	mmcsd_dbg("clock %uHz busmode %u powermode %u cs %u Vdd %u "
		"width %u \n",
		 io_cfg->clock, io_cfg->bus_mode,
		 io_cfg->power_mode, io_cfg->chip_select, io_cfg->vdd,
		 io_cfg->bus_width);

	host->ops->set_iocfg(host, io_cfg);
}

/*
 * Control chip select pin on a host.
 */
void mmcsd_set_chip_select(struct rt_mmcsd_host *host, rt_int32_t mode)
{
	host->io_cfg.chip_select = mode;
	mmcsd_set_iocfg(host);
}

/*
 * Sets the host clock to the highest possible frequency that
 * is below "hz".
 */
void mmcsd_set_clock(struct rt_mmcsd_host *host, rt_uint32_t clk)
{
	if (clk < host->freq_min)
	{
		rt_kprintf("clock too low\n");
	}

	host->io_cfg.clock = clk;
	mmcsd_set_iocfg(host);
}

/*
 * Change the bus mode (open drain/push-pull) of a host.
 */
void mmcsd_set_bus_mode(struct rt_mmcsd_host *host, rt_uint32_t mode)
{
	host->io_cfg.bus_mode = mode;
	mmcsd_set_iocfg(host);
}

/*
 * Change data bus width of a host.
 */
void mmcsd_set_bus_width(struct rt_mmcsd_host *host, rt_uint32_t width)
{
	host->io_cfg.bus_width = width;
	mmcsd_set_iocfg(host);
}

void mmcsd_set_data_timeout(struct rt_mmcsd_data *data, const struct rt_mmcsd_card *card)
{
	rt_uint32_t mult;

	if (card->card_type == CARD_TYPE_SDIO) 
	{
		data->timeout_ns = 1000000000;	/* SDIO card 1s */
		data->timeout_clks = 0;
		return;
	}

	/*
	 * SD cards use a 100 multiplier rather than 10
	 */
	mult = (card->card_type == CARD_TYPE_SD) ? 100 : 10;

	/*
	 * Scale up the multiplier (and therefore the timeout) by
	 * the r2w factor for writes.
	 */
	if (data->flags & DATA_DIR_WRITE)
		mult <<= card->csd.r2w_factor;

	data->timeout_ns = card->tacc_ns * mult;
	data->timeout_clks = card->tacc_clks * mult;

	/*
	 * SD cards also have an upper limit on the timeout.
	 */
	if (card->card_type == CARD_TYPE_SD) 
	{
		rt_uint32_t timeout_us, limit_us;

		timeout_us = data->timeout_ns / 1000;
		timeout_us += data->timeout_clks * 1000 /
			(card->host->io_cfg.clock / 1000);

		if (data->flags & DATA_DIR_WRITE)
			/*
			 * The limit is really 250 ms, but that is
			 * insufficient for some crappy cards.
			 */
			limit_us = 300000;
		else
			limit_us = 100000;

		/*
		 * SDHC cards always use these fixed values.
		 */
		if (timeout_us > limit_us || card->flags & CARD_FLAG_SDHC) 
		{
			data->timeout_ns = limit_us * 1000;	/* SDHC card fixed 250ms */
			data->timeout_clks = 0;
		}
	}

	if (controller_is_spi(card->host)) 
	{
		if (data->flags & DATA_DIR_WRITE) 
		{
			if (data->timeout_ns < 1000000000)
				data->timeout_ns = 1000000000;	/* 1s */
		} 
		else 
		{
			if (data->timeout_ns < 100000000)
				data->timeout_ns =  100000000;	/* 100ms */
		}
	}
}


/*
 * Mask off any voltages we don't support and select
 * the lowest voltage
 */
rt_uint32_t mmcsd_select_voltage(struct rt_mmcsd_host *host, rt_uint32_t ocr)
{
	int bit;

	ocr &= host->valid_ocr;

	bit = ffs(ocr);
	if (bit) 
	{
		bit -= 1;

		ocr &= 3 << bit;

		host->io_cfg.vdd = bit;
		mmcsd_set_iocfg(host);
	} 
	else 
	{
		rt_kprintf("host doesn't support card's voltages\n");
		ocr = 0;
	}

	return ocr;
}


static void mmcsd_power_up(struct rt_mmcsd_host *host)
{
	int bit = fls(host->valid_ocr) - 1;

	host->io_cfg.vdd = bit;
	if (controller_is_spi(host))
	{
		host->io_cfg.chip_select = MMCSD_CS_HIGH;
		host->io_cfg.bus_mode = MMCSD_BUSMODE_PUSHPULL;
	} 
	else
	{
		host->io_cfg.chip_select = MMCSD_CS_IGNORE;
		host->io_cfg.bus_mode = MMCSD_BUSMODE_OPENDRAIN;
	}
	host->io_cfg.power_mode = MMCSD_POWER_UP;
	host->io_cfg.bus_width = MMCSD_BUS_WIDTH_1;
	mmcsd_set_iocfg(host);

	/*
	 * This delay should be sufficient to allow the power supply
	 * to reach the minimum voltage.
	 */
	mmcsd_delay_ms(10);

	host->io_cfg.clock = host->freq_min;
	host->io_cfg.power_mode = MMCSD_POWER_ON;
	mmcsd_set_iocfg(host);

	/*
	 * This delay must be at least 74 clock sizes, or 1 ms, or the
	 * time required to reach a stable voltage.
	 */
	mmcsd_delay_ms(10);
}

static void mmcsd_power_off(struct rt_mmcsd_host *host)
{
	host->io_cfg.clock = 0;
	host->io_cfg.vdd = 0;
	if (!controller_is_spi(host)) 
	{
		host->io_cfg.bus_mode = MMCSD_BUSMODE_OPENDRAIN;
		host->io_cfg.chip_select = MMCSD_CS_IGNORE;
	}
	host->io_cfg.power_mode = MMCSD_POWER_OFF;
	host->io_cfg.bus_width = MMCSD_BUS_WIDTH_1;
	mmcsd_set_iocfg(host);
}

void mmcsd_change(struct rt_mmcsd_host *host)
{
	rt_mb_send(&mmcsd_detect_mb, (rt_uint32_t)host);
}

void mmcsd_detect(void *param)
{
	struct rt_mmcsd_host *host;
	rt_uint32_t  ocr;
	rt_int32_t  err;

	while (1) 
	{
		if (rt_mb_recv(&mmcsd_detect_mb, (rt_uint32_t*)&host, RT_WAITING_FOREVER) == RT_EOK)
		{
			if (host->card == RT_NULL)
			{
				mmcsd_host_lock(host);
				mmcsd_power_up(host);
				mmcsd_go_idle(host);

				mmcsd_send_if_cond(host, host->valid_ocr);

				err = sdio_io_send_op_cond(host, 0, &ocr);
				if (!err) {
					if (init_sdio(host, ocr))
						mmcsd_power_off(host);
					mmcsd_host_unlock(host);
					continue;
				}

				/*
				 * detect SD card
				 */
				err = mmcsd_send_app_op_cond(host, 0, &ocr);
				if (!err) 
				{
					if (init_sd(host, ocr))
						mmcsd_power_off(host);
					mmcsd_host_unlock(host);
					continue;
				}
				mmcsd_host_unlock(host);
			}
		}
	}
}

struct rt_mmcsd_host *mmcsd_alloc_host(void)
{
	struct rt_mmcsd_host *host;

	host = rt_malloc(sizeof(struct rt_mmcsd_host));
	if (!host) 
	{
		rt_kprintf("alloc host failed\n");
		return RT_NULL;
	}

	rt_memset(host, 0, sizeof(struct rt_mmcsd_host));

	host->max_seg_size = 65535;
	host->max_dma_segs = 1;
	host->max_blk_size = 512;
	host->max_blk_count = 4096;

	rt_sem_init(&host->bus_lock, "sd_bus_lock", 1, RT_IPC_FLAG_FIFO);
	rt_sem_init(&host->sem_ack, "sd_ack", 0, RT_IPC_FLAG_FIFO);

	return host;
}

void mmcsd_free_host(struct rt_mmcsd_host *host)
{
	rt_sem_detach(&host->bus_lock);
	rt_sem_detach(&host->sem_ack);
	rt_free(host);
}

void rt_mmcsd_core_init(void)
{
	rt_err_t ret;

	/* init detect sd cart thread */
	/* init mailbox and create detect sd card thread */
	ret = rt_mb_init(&mmcsd_detect_mb, "mmcsdmb",
		&mmcsd_detect_mb_pool[0], sizeof(mmcsd_detect_mb_pool),
		RT_IPC_FLAG_FIFO);
	RT_ASSERT(ret == RT_EOK);

	ret = rt_thread_init(&mmcsd_detect_thread, "mmcsd_detect", mmcsd_detect, RT_NULL, 
			     &mmcsd_stack[0], RT_MMCSD_STACK_SIZE, RT_MMCSD_THREAD_PREORITY, 20);
	if (ret == RT_EOK) 
	{
		rt_thread_startup(&mmcsd_detect_thread);
	}

	rt_sdio_init();
}