BlueSCSI-v2/lib/BlueSCSI_platform_RP2040/rp2040_sdio.cpp

// Implementation of SDIO communication for RP2040
// Copyright (c) 2022 Rabbit Hole Computing™
// Copyright (c) 2024 Tech by Androda, LLC
//
// The RP2040 official work-in-progress code at
// https://github.com/raspberrypi/pico-extras/tree/master/src/rp2_common/pico_sd_card
// may be useful reference, but this is independent implementation.
//
// For official SDIO specifications, refer to:
// https://www.sdcard.org/downloads/pls/
// "SDIO Physical Layer Simplified Specification Version 8.00"

#include "rp2040_sdio.h"
#include "rp2040_sdio.pio.h"
#include <hardware/pio.h>
#include <hardware/dma.h>
//#include <hardware/gpio.h>
#include <BlueSCSI_platform.h>
#include <BlueSCSI_log.h>
#include "timings_RP2MCU.h"

#define SDIO_PIO pio1
#define SDIO_CMD_SM 0
#define SDIO_DATA_SM 1
#define SDIO_DMA_CH 4
#define SDIO_DMA_CHB 5

#define PIO_INSTR_MASK_REMOVE_DELAY 0xF8FF
#define PIO_INSTR_MASK_GET_DELAY 0x700

#define PIO_INSTR_JMP_MASK 0xE000
#define PIO_INSTR_JMP_ADDR 0x1F

// Maximum number of 512 byte blocks to transfer in one request
#define SDIO_MAX_BLOCKS 256

enum sdio_transfer_state_t { SDIO_IDLE, SDIO_RX, SDIO_TX, SDIO_TX_WAIT_IDLE};

static struct {
    uint32_t pio_cmd_rsp_clk_offset;
    pio_sm_config pio_cfg_cmd_rsp;
    uint32_t pio_data_rx_offset;
    pio_sm_config pio_cfg_data_rx;
    uint32_t pio_data_tx_offset;
    pio_sm_config pio_cfg_data_tx;

    sdio_transfer_state_t transfer_state;
    uint32_t transfer_start_time;
    uint32_t *data_buf;
    uint32_t blocks_done; // Number of blocks transferred so far
    uint32_t total_blocks; // Total number of blocks to transfer
    uint32_t blocks_checksumed; // Number of blocks that have had CRC calculated
    uint32_t checksum_errors; // Number of checksum errors detected
    uint8_t cmdBuf[6];
    // Variables for block writes
    uint64_t next_wr_block_checksum;
    uint32_t end_token_buf[3]; // CRC and end token for write block
    sdio_status_t wr_status;
    uint32_t card_response;

    // Variables for block reads
    // This is used to perform DMA into data buffers and checksum buffers separately.
    struct {
        void * write_addr;
        uint32_t transfer_count;
    } dma_blocks[SDIO_MAX_BLOCKS * 2];
    struct {
        uint32_t top;
        uint32_t bottom;
    } received_checksums[SDIO_MAX_BLOCKS];
} g_sdio;

void rp2040_sdio_dma_irq();

/*******************************************************
 * Checksum algorithms
 *******************************************************/

// Table lookup for calculating CRC-7 checksum that is used in SDIO command packets.
// Usage:
//    uint8_t crc = 0;
//    crc = crc7_table[crc ^ byte];
//    .. repeat for every byte ..
static const uint8_t crc7_table[256] = {
	0x00, 0x12, 0x24, 0x36, 0x48, 0x5a, 0x6c, 0x7e,
    0x90, 0x82, 0xb4, 0xa6, 0xd8, 0xca, 0xfc, 0xee,
	0x32, 0x20, 0x16, 0x04, 0x7a, 0x68, 0x5e, 0x4c,
    0xa2, 0xb0, 0x86, 0x94, 0xea, 0xf8, 0xce, 0xdc,
	0x64, 0x76, 0x40, 0x52, 0x2c, 0x3e, 0x08, 0x1a,
    0xf4, 0xe6, 0xd0, 0xc2, 0xbc, 0xae, 0x98, 0x8a,
	0x56, 0x44, 0x72, 0x60, 0x1e, 0x0c, 0x3a, 0x28,
    0xc6, 0xd4, 0xe2, 0xf0, 0x8e, 0x9c, 0xaa, 0xb8,
	0xc8, 0xda, 0xec, 0xfe, 0x80, 0x92, 0xa4, 0xb6,
    0x58, 0x4a, 0x7c, 0x6e, 0x10, 0x02, 0x34, 0x26,
	0xfa, 0xe8, 0xde, 0xcc, 0xb2, 0xa0, 0x96, 0x84,
    0x6a, 0x78, 0x4e, 0x5c, 0x22, 0x30, 0x06, 0x14,
	0xac, 0xbe, 0x88, 0x9a, 0xe4, 0xf6, 0xc0, 0xd2,
    0x3c, 0x2e, 0x18, 0x0a, 0x74, 0x66, 0x50, 0x42,
	0x9e, 0x8c, 0xba, 0xa8, 0xd6, 0xc4, 0xf2, 0xe0,
    0x0e, 0x1c, 0x2a, 0x38, 0x46, 0x54, 0x62, 0x70,
	0x82, 0x90, 0xa6, 0xb4, 0xca, 0xd8, 0xee, 0xfc,
    0x12, 0x00, 0x36, 0x24, 0x5a, 0x48, 0x7e, 0x6c,
	0xb0, 0xa2, 0x94, 0x86, 0xf8, 0xea, 0xdc, 0xce,
    0x20, 0x32, 0x04, 0x16, 0x68, 0x7a, 0x4c, 0x5e,
	0xe6, 0xf4, 0xc2, 0xd0, 0xae, 0xbc, 0x8a, 0x98,
    0x76, 0x64, 0x52, 0x40, 0x3e, 0x2c, 0x1a, 0x08,
	0xd4, 0xc6, 0xf0, 0xe2, 0x9c, 0x8e, 0xb8, 0xaa,
    0x44, 0x56, 0x60, 0x72, 0x0c, 0x1e, 0x28, 0x3a,
	0x4a, 0x58, 0x6e, 0x7c, 0x02, 0x10, 0x26, 0x34,
    0xda, 0xc8, 0xfe, 0xec, 0x92, 0x80, 0xb6, 0xa4,
	0x78, 0x6a, 0x5c, 0x4e, 0x30, 0x22, 0x14, 0x06,
    0xe8, 0xfa, 0xcc, 0xde, 0xa0, 0xb2, 0x84, 0x96,
	0x2e, 0x3c, 0x0a, 0x18, 0x66, 0x74, 0x42, 0x50,
    0xbe, 0xac, 0x9a, 0x88, 0xf6, 0xe4, 0xd2, 0xc0,
	0x1c, 0x0e, 0x38, 0x2a, 0x54, 0x46, 0x70, 0x62,
    0x8c, 0x9e, 0xa8, 0xba, 0xc4, 0xd6, 0xe0, 0xf2
};

// Calculate the CRC16 checksum for parallel 4 bit lines separately.
// When the SDIO bus operates in 4-bit mode, the CRC16 algorithm
// is applied to each line separately and generates total of
// 4 x 16 = 64 bits of checksum.
__attribute__((optimize("O3")))
uint64_t __not_in_flash_func(sdio_crc16_4bit_checksum)(uint32_t *data, uint32_t num_words)
{
    uint64_t crc = 0;
    uint32_t *end = data + num_words;
    while (data < end)
    {
        for (int unroll = 0; unroll < 4; unroll++)
        {
            // Each 32-bit word contains 8 bits per line.
            // Reverse the bytes because SDIO protocol is big-endian.
            uint32_t data_in = __builtin_bswap32(*data++);

            // Shift out 8 bits for each line
            uint32_t data_out = crc >> 32;
            crc <<= 32;

            // XOR outgoing data to itself with 4 bit delay
            data_out ^= (data_out >> 16);

            // XOR incoming data to outgoing data with 4 bit delay
            data_out ^= (data_in >> 16);

            // XOR outgoing and incoming data to accumulator at each tap
            uint64_t xorred = data_out ^ data_in;
            crc ^= xorred;
            crc ^= xorred << (5 * 4);
            crc ^= xorred << (12 * 4);
        }
    }

    return crc;
}


/*******************************************************
 * Clock Runner
 *******************************************************/
void __not_in_flash_func(cycleSdClock)() {
    pio_sm_exec(SDIO_PIO, SDIO_CMD_SM, pio_encode_nop() | pio_encode_sideset_opt(1, 1) | pio_encode_delay(1));
    pio_sm_exec(SDIO_PIO, SDIO_CMD_SM, pio_encode_nop() | pio_encode_sideset_opt(1, 0) | pio_encode_delay(1));
}

/*******************************************************
 * Status Register Receiver
 *******************************************************/
sdio_status_t __not_in_flash_func(receive_status_register)(uint8_t* sds) {
    rp2040_sdio_rx_start(sds, 1, 64);

    // Wait for the DMA operation to complete, or fail if it took too long
waitagain:
    while (dma_channel_is_busy(SDIO_DMA_CHB) || dma_channel_is_busy(SDIO_DMA_CH))
    {
        if ((uint32_t)(millis() - g_sdio.transfer_start_time) > 2)
        {
            // Reset the state machine program
            dma_channel_abort(SDIO_DMA_CHB);
            pio_sm_set_enabled(SDIO_PIO, SDIO_CMD_SM, false);
            pio_sm_clear_fifos(SDIO_PIO, SDIO_CMD_SM);
            return SDIO_ERR_RESPONSE_TIMEOUT;
        }
    }

    // Assert that both DMA channels are complete
    if(dma_channel_is_busy(SDIO_DMA_CHB) || dma_channel_is_busy(SDIO_DMA_CH)) {
        // Wait failure, go back.
        goto waitagain;
    }

    pio_sm_set_enabled(SDIO_PIO, SDIO_DATA_SM, false);
    g_sdio.transfer_state = SDIO_IDLE;

    return SDIO_OK;
}


/*******************************************************
 * Basic SDIO command execution
 *******************************************************/

static void __not_in_flash_func(sdio_send_command)(uint8_t command, uint32_t arg, uint8_t response_bits)
{
    // if (command != 41 && command != 55) {
    //     log("C: ", (int)command, " A: ", arg);
    // }
    io_wo_8* txFifo = reinterpret_cast<io_wo_8*>(&SDIO_PIO->txf[SDIO_CMD_SM]);

    // Reinitialize the CMD SM
    pio_sm_init(SDIO_PIO, SDIO_CMD_SM, g_sdio.pio_cmd_rsp_clk_offset, &g_sdio.pio_cfg_cmd_rsp);
    pio_sm_set_consecutive_pindirs(SDIO_PIO, SDIO_CMD_SM, SDIO_CLK, 1, true);
    pio_sm_set_consecutive_pindirs(SDIO_PIO, SDIO_CMD_SM, SDIO_CMD, 1, true);
    pio_sm_set_consecutive_pindirs(SDIO_PIO, SDIO_CMD_SM, SDIO_D0, 4, false);

    // Pin direction: output, initial state should be high
    pio_sm_exec(SDIO_PIO, SDIO_CMD_SM, pio_encode_set(pio_pins, 1));
    pio_sm_exec(SDIO_PIO, SDIO_CMD_SM, pio_encode_set(pio_pindirs, 1));

    // Write the number of tx / rx bits to the SM
    *txFifo = 55;  // Write 56 bits total
    pio_sm_exec(SDIO_PIO, SDIO_CMD_SM, pio_encode_out(pio_x, 8));
    *txFifo = response_bits ? response_bits - 1 : 0;    // Bit count to receive
    pio_sm_exec(SDIO_PIO, SDIO_CMD_SM, pio_encode_out(pio_y, 8));
    pio_sm_set_enabled(SDIO_PIO, SDIO_CMD_SM, true);

    // Build the command bytes (commands are 48 bits long)
    g_sdio.cmdBuf[0] = command | 0x40;
    g_sdio.cmdBuf[1] = (uint8_t)(arg >> 24U);
    g_sdio.cmdBuf[2] = (uint8_t)(arg >> 16U);
    g_sdio.cmdBuf[3] = (uint8_t)(arg >> 8U);
    g_sdio.cmdBuf[4] = (uint8_t)arg;

    // Get the SM clocking while we calculate CRCs
    *txFifo = 0XFF;

    // CRC calculation
    uint8_t crc = 0;
    for(uint8_t i = 0; i < 5; i++) {
        crc = crc7_table[crc ^ g_sdio.cmdBuf[i]];
    }
    crc = crc | 0x1;
    g_sdio.cmdBuf[5] = crc;

    dma_channel_config dmacfg = dma_channel_get_default_config(SDIO_DMA_CH);
    channel_config_set_transfer_data_size(&dmacfg, DMA_SIZE_8);
    channel_config_set_read_increment(&dmacfg, true);
    channel_config_set_write_increment(&dmacfg, false);
    channel_config_set_dreq(&dmacfg, pio_get_dreq(SDIO_PIO, SDIO_CMD_SM, true));
    dma_channel_configure(SDIO_DMA_CH, &dmacfg, &SDIO_PIO->txf[SDIO_CMD_SM], &g_sdio.cmdBuf, 6, true);
}

sdio_status_t __not_in_flash_func(rp2040_sdio_command_R1)(uint8_t command, uint32_t arg, uint32_t *response)
{
    uint32_t resp[2];
    if (response) {
        dma_channel_config dmacfg = dma_channel_get_default_config(SDIO_DMA_CHB);
        channel_config_set_transfer_data_size(&dmacfg, DMA_SIZE_8);
        channel_config_set_read_increment(&dmacfg, false);
        channel_config_set_write_increment(&dmacfg, true);
        channel_config_set_dreq(&dmacfg, pio_get_dreq(SDIO_PIO, SDIO_CMD_SM, false));  //6 * 8 = 48 bits
        dma_channel_configure(SDIO_DMA_CHB, &dmacfg, &resp, &SDIO_PIO->rxf[SDIO_CMD_SM], 6, true);
    }

    sdio_send_command(command, arg, response ? 48 : 0);

    uint32_t start = millis();
    if (response)
    {
        // Wait for DMA channel to receive response
        while (dma_channel_is_busy(SDIO_DMA_CHB))
        {
            if ((uint32_t)(millis() - start) > 2)
            {
                if (command != 8) {
                    /*debug*/log("Timeout waiting for response in rp2040_sdio_command_R1(", (int)command, "), ",
                        "PIO PC: ", (int)pio_sm_get_pc(SDIO_PIO, SDIO_CMD_SM) - (int)g_sdio.pio_cmd_rsp_clk_offset,
                        " RXF: ", (int)pio_sm_get_rx_fifo_level(SDIO_PIO, SDIO_CMD_SM),
                        " TXF: ", (int)pio_sm_get_tx_fifo_level(SDIO_PIO, SDIO_CMD_SM));
                }

                // Reset the state machine program
                dma_channel_abort(SDIO_DMA_CHB);
                pio_sm_set_enabled(SDIO_PIO, SDIO_CMD_SM, false);  // Turn off the CMD SM, there was an error
                pio_sm_clear_fifos(SDIO_PIO, SDIO_CMD_SM);
                return SDIO_ERR_RESPONSE_TIMEOUT;
            }
        }
        // Must bswap due to 8 bit segmentation
        resp[0] = __builtin_bswap32(resp[0]);
        resp[1] = __builtin_bswap32(resp[1]) >> 16;
        // debuglog("SDIO R1 response: ", resp0, " ", resp1);

        // Calculate response checksum
        uint8_t crc = 0;
        crc = crc7_table[crc ^ ((resp[0] >> 24) & 0xFF)];
        crc = crc7_table[crc ^ ((resp[0] >> 16) & 0xFF)];
        crc = crc7_table[crc ^ ((resp[0] >>  8) & 0xFF)];
        crc = crc7_table[crc ^ ((resp[0] >>  0) & 0xFF)];
        crc = crc7_table[crc ^ ((resp[1] >>  8) & 0xFF)];

        uint8_t actual_crc = ((resp[1] >> 0) & 0xFE);
        if (crc != actual_crc)
        {
            debuglog("rp2040_sdio_command_R1(", (int)command, "): CRC error, calculated ", crc, " packet has ", actual_crc);
            debuglog("resp[0]:", resp[0], "resp[1]:", resp[1]);
            return SDIO_ERR_RESPONSE_CRC;
        }

        uint8_t response_cmd = ((resp[0] >> 24) & 0xFF);
        if (response_cmd != command && command != 41)
        {
            debuglog("rp2040_sdio_command_R1(", (int)command, "): received reply for ", (int)response_cmd);
            return SDIO_ERR_RESPONSE_CODE;
        }

        *response = ((resp[0] & 0xFFFFFF) << 8) | ((resp[1] >> 8) & 0xFF);
    } else {
        // Wait for CMD SM TX FIFO Stall (all command bits were sent)
        uint32_t tx_stall_flag = 1u << (PIO_FDEBUG_TXSTALL_LSB + SDIO_CMD_SM);
        // Clear the stall marker
        SDIO_PIO->fdebug = tx_stall_flag;
        // Wait for the stall
        while (!(SDIO_PIO->fdebug & tx_stall_flag)) {
            if ((uint32_t)(millis() - start) > 2)
            {
                if (command != 8) {
                    /*debug*/log("Timeout waiting for CMD TX in rp2040_sdio_command_R1(", (int)command, "), ",
                        "PIO PC: ", (int)pio_sm_get_pc(SDIO_PIO, SDIO_CMD_SM) - (int)g_sdio.pio_cmd_rsp_clk_offset,
                        " RXF: ", (int)pio_sm_get_rx_fifo_level(SDIO_PIO, SDIO_CMD_SM),
                        " TXF: ", (int)pio_sm_get_tx_fifo_level(SDIO_PIO, SDIO_CMD_SM));
                }

                // Reset the state machine program
                pio_sm_set_enabled(SDIO_PIO, SDIO_CMD_SM, false);  // Turn off the CMD SM, there was an error
                pio_sm_clear_fifos(SDIO_PIO, SDIO_CMD_SM);
                return SDIO_ERR_RESPONSE_TIMEOUT;
            }
        }
    }

    pio_sm_set_enabled(SDIO_PIO, SDIO_CMD_SM, false);
    pio_sm_clear_fifos(SDIO_PIO, SDIO_CMD_SM);
    return SDIO_OK;
}

sdio_status_t __not_in_flash_func(rp2040_sdio_command_R2)(uint8_t command, uint32_t arg, uint8_t response[16])
{
    // The response is too long to fit in the PIO FIFO, so use DMA to receive it.
    uint32_t response_buf[5];
    dma_channel_config dmacfg = dma_channel_get_default_config(SDIO_DMA_CHB);
    channel_config_set_transfer_data_size(&dmacfg, DMA_SIZE_8);
    channel_config_set_read_increment(&dmacfg, false);
    channel_config_set_write_increment(&dmacfg, true);
    channel_config_set_dreq(&dmacfg, pio_get_dreq(SDIO_PIO, SDIO_CMD_SM, false));          //17 * 8 = 136
    dma_channel_configure(SDIO_DMA_CHB, &dmacfg, &response_buf, &SDIO_PIO->rxf[SDIO_CMD_SM], 17, true);

    sdio_send_command(command, arg, 136);

    uint32_t start = millis();
    while (dma_channel_is_busy(SDIO_DMA_CHB))
    {
        if ((uint32_t)(millis() - start) > 2)
        {
            debuglog("Timeout waiting for response in rp2040_sdio_command_R2(", (int)command, "), ",
                  "PIO PC: ", (int)pio_sm_get_pc(SDIO_PIO, SDIO_CMD_SM) - (int)g_sdio.pio_cmd_rsp_clk_offset,
                  " RXF: ", (int)pio_sm_get_rx_fifo_level(SDIO_PIO, SDIO_CMD_SM),
                  " TXF: ", (int)pio_sm_get_tx_fifo_level(SDIO_PIO, SDIO_CMD_SM));

            // Reset the state machine program
            dma_channel_abort(SDIO_DMA_CHB);
            pio_sm_set_enabled(SDIO_PIO, SDIO_CMD_SM, false);  // Turn off the CMD SM, there was an error
            pio_sm_clear_fifos(SDIO_PIO, SDIO_CMD_SM);
            return SDIO_ERR_RESPONSE_TIMEOUT;
        }
    }

    pio_sm_set_enabled(SDIO_PIO, SDIO_CMD_SM, false);  // Turn off the CMD SM, its job is done
    pio_sm_clear_fifos(SDIO_PIO, SDIO_CMD_SM);
    dma_channel_abort(SDIO_DMA_CHB);

    // Must byte swap because receiving 8-bit chunks instead of 32 bit
    response_buf[0] = __builtin_bswap32(response_buf[0]);
    response_buf[1] = __builtin_bswap32(response_buf[1]);
    response_buf[2] = __builtin_bswap32(response_buf[2]);
    response_buf[3] = __builtin_bswap32(response_buf[3]);
    response_buf[4] = __builtin_bswap32(response_buf[4]) >> 24;

    // Copy the response payload to output buffer
    response[0]  = ((response_buf[0] >> 16) & 0xFF);
    response[1]  = ((response_buf[0] >>  8) & 0xFF);
    response[2]  = ((response_buf[0] >>  0) & 0xFF);
    response[3]  = ((response_buf[1] >> 24) & 0xFF);
    response[4]  = ((response_buf[1] >> 16) & 0xFF);
    response[5]  = ((response_buf[1] >>  8) & 0xFF);
    response[6]  = ((response_buf[1] >>  0) & 0xFF);
    response[7]  = ((response_buf[2] >> 24) & 0xFF);
    response[8]  = ((response_buf[2] >> 16) & 0xFF);
    response[9]  = ((response_buf[2] >>  8) & 0xFF);
    response[10] = ((response_buf[2] >>  0) & 0xFF);
    response[11] = ((response_buf[3] >> 24) & 0xFF);
    response[12] = ((response_buf[3] >> 16) & 0xFF);
    response[13] = ((response_buf[3] >>  8) & 0xFF);
    response[14] = ((response_buf[3] >>  0) & 0xFF);
    response[15] = ((response_buf[4] >>  0) & 0xFF);

    // Calculate checksum of the payload
    uint8_t crc = 0;
    for (int i = 0; i < 15; i++)
    {
        crc = crc7_table[crc ^ response[i]];
    }

    uint8_t actual_crc = response[15] & 0xFE;
    if (crc != actual_crc)
    {
        debuglog("rp2040_sdio_command_R2(", (int)command, "): CRC error, calculated ", crc, " packet has ", actual_crc);
        return SDIO_ERR_RESPONSE_CRC;
    }

    uint8_t response_cmd = ((response_buf[0] >> 24) & 0xFF);
    if (response_cmd != 0x3F)
    {
        debuglog("rp2040_sdio_command_R2(", (int)command, "): Expected reply code 0x3F");
        return SDIO_ERR_RESPONSE_CODE;
    }

    return SDIO_OK;
}


sdio_status_t __not_in_flash_func(rp2040_sdio_command_R3)(uint8_t command, uint32_t arg, uint32_t *response)
{
    uint32_t resp[2];
    dma_channel_config dmacfg = dma_channel_get_default_config(SDIO_DMA_CHB);
    channel_config_set_transfer_data_size(&dmacfg, DMA_SIZE_8);
    channel_config_set_read_increment(&dmacfg, false);
    channel_config_set_write_increment(&dmacfg, true);
    channel_config_set_dreq(&dmacfg, pio_get_dreq(SDIO_PIO, SDIO_CMD_SM, false));  //6 * 8 = 48 bits
    dma_channel_configure(SDIO_DMA_CHB, &dmacfg, &resp, &SDIO_PIO->rxf[SDIO_CMD_SM], 6, true);
        
    sdio_send_command(command, arg, 48);

    // Wait for response
    uint32_t start = millis();
    while (dma_channel_is_busy(SDIO_DMA_CHB))
    {
        if ((uint32_t)(millis() - start) > 2)
        {
            debuglog("Timeout waiting for response in rp2040_sdio_command_R3(", (int)command, "), ",
                  "PIO PC: ", (int)pio_sm_get_pc(SDIO_PIO, SDIO_CMD_SM) - (int)g_sdio.pio_cmd_rsp_clk_offset,
                  " RXF: ", (int)pio_sm_get_rx_fifo_level(SDIO_PIO, SDIO_CMD_SM),
                  " TXF: ", (int)pio_sm_get_tx_fifo_level(SDIO_PIO, SDIO_CMD_SM));

            // Reset the state machine program
            dma_channel_abort(SDIO_DMA_CHB);
            pio_sm_set_enabled(SDIO_PIO, SDIO_CMD_SM, false);  // Turn off the CMD SM, there was an error
            pio_sm_clear_fifos(SDIO_PIO, SDIO_CMD_SM);
            return SDIO_ERR_RESPONSE_TIMEOUT;
        }
    }

    pio_sm_set_enabled(SDIO_PIO, SDIO_CMD_SM, false);  // Turn off the CMD SM, its job is done
    pio_sm_clear_fifos(SDIO_PIO, SDIO_CMD_SM);

    // Must bswap due to 8 bit transfer
    resp[0] = __builtin_bswap32(resp[0]);
    resp[1] = __builtin_bswap32(resp[1]) >> 16;
    
    *response = ((resp[0] & 0xFFFFFF) << 8) | ((resp[1] >> 8) & 0xFF);
    // debuglog("SDIO R3 response: ", resp0, " ", resp1);

    return SDIO_OK;
}

/*******************************************************
 * Data reception from SD card
 *******************************************************/

sdio_status_t __not_in_flash_func(rp2040_sdio_rx_start)(uint8_t *buffer, uint32_t num_blocks, uint32_t block_size)
{
    // Buffer must be aligned
    assert(((uint32_t)buffer & 3) == 0 && num_blocks <= SDIO_MAX_BLOCKS);

    g_sdio.transfer_state = SDIO_RX;
    g_sdio.transfer_start_time = millis();
    g_sdio.data_buf = (uint32_t*)buffer;
    g_sdio.blocks_done = 0;
    g_sdio.total_blocks = num_blocks;
    g_sdio.blocks_checksumed = 0;
    g_sdio.checksum_errors = 0;

    // Create DMA block descriptors to store each block of block_size bytes of data to buffer
    // and then 8 bytes to g_sdio.received_checksums.
    for (int i = 0; i < num_blocks; i++)
    {
        g_sdio.dma_blocks[i * 2].write_addr = buffer + (i * block_size);
        g_sdio.dma_blocks[i * 2].transfer_count = block_size / sizeof(uint32_t);

        g_sdio.dma_blocks[i * 2 + 1].write_addr = &g_sdio.received_checksums[i];
        g_sdio.dma_blocks[i * 2 + 1].transfer_count = 2;
    }
    g_sdio.dma_blocks[num_blocks * 2].write_addr = 0;
    g_sdio.dma_blocks[num_blocks * 2].transfer_count = 0;

    // Configure first DMA channel for reading from the PIO RX fifo
    dma_channel_config dmacfg = dma_channel_get_default_config(SDIO_DMA_CH);
    channel_config_set_transfer_data_size(&dmacfg, DMA_SIZE_32);
    channel_config_set_read_increment(&dmacfg, false);
    channel_config_set_write_increment(&dmacfg, true);
    channel_config_set_dreq(&dmacfg, pio_get_dreq(SDIO_PIO, SDIO_DATA_SM, false));
    channel_config_set_bswap(&dmacfg, true);
    channel_config_set_chain_to(&dmacfg, SDIO_DMA_CHB);
    dma_channel_configure(SDIO_DMA_CH, &dmacfg, 0, &SDIO_PIO->rxf[SDIO_DATA_SM], 0, false);

    // Configure second DMA channel for reconfiguring the first one
    dmacfg = dma_channel_get_default_config(SDIO_DMA_CHB);
    channel_config_set_transfer_data_size(&dmacfg, DMA_SIZE_32);
    channel_config_set_read_increment(&dmacfg, true);
    channel_config_set_write_increment(&dmacfg, true);
    channel_config_set_ring(&dmacfg, true, 3);
    dma_channel_configure(SDIO_DMA_CHB, &dmacfg, &dma_hw->ch[SDIO_DMA_CH].al1_write_addr,
        g_sdio.dma_blocks, 2, false);

    // Initialize PIO state machine
    pio_sm_init(SDIO_PIO, SDIO_DATA_SM, g_sdio.pio_data_rx_offset, &g_sdio.pio_cfg_data_rx);
    pio_sm_set_consecutive_pindirs(SDIO_PIO, SDIO_DATA_SM, SDIO_CLK, 1, true);
    pio_sm_set_consecutive_pindirs(SDIO_PIO, SDIO_DATA_SM, SDIO_D0, 4, false);

    // Write number of nibbles to receive to Y register
    pio_sm_put(SDIO_PIO, SDIO_DATA_SM, (block_size * 2) + 16 - 1);
    pio_sm_exec(SDIO_PIO, SDIO_DATA_SM, pio_encode_out(pio_y, 32));

    // Enable RX FIFO join because we don't need the TX FIFO during transfer.
    // This gives more leeway for the DMA block switching
    SDIO_PIO->sm[SDIO_DATA_SM].shiftctrl |= PIO_SM0_SHIFTCTRL_FJOIN_RX_BITS;

    // Start PIO and DMA
    dma_channel_start(SDIO_DMA_CHB);
    pio_sm_set_enabled(SDIO_PIO, SDIO_DATA_SM, true);

    return SDIO_OK;
}

// Check checksums for received blocks
static void __not_in_flash_func(sdio_verify_rx_checksums)(uint32_t maxcount)
{
    while (g_sdio.blocks_checksumed < g_sdio.blocks_done && maxcount-- > 0)
    {
        // Calculate checksum from received data
        int blockidx = g_sdio.blocks_checksumed++;
        uint64_t checksum = sdio_crc16_4bit_checksum(g_sdio.data_buf + blockidx * SDIO_WORDS_PER_BLOCK,
                                                     SDIO_WORDS_PER_BLOCK);

        // Convert received checksum to little-endian format
        uint32_t top = __builtin_bswap32(g_sdio.received_checksums[blockidx].top);
        uint32_t bottom = __builtin_bswap32(g_sdio.received_checksums[blockidx].bottom);
        uint64_t expected = ((uint64_t)top << 32) | bottom;

        if (checksum != expected)
        {
            g_sdio.checksum_errors++;
            if (g_sdio.checksum_errors == 1)
            {
                log("SDIO checksum error in reception: block ", blockidx,
                      " calculated ", checksum, " expected ", expected);
            }
        }
    }
}

sdio_status_t __not_in_flash_func(rp2040_sdio_rx_poll)(uint32_t *bytes_complete)
{
    // Was everything done when the previous rx_poll() finished?
    if (g_sdio.blocks_done >= g_sdio.total_blocks)
    {
        g_sdio.transfer_state = SDIO_IDLE;
    }
    else
    {
        // Use the idle time to calculate checksums
        sdio_verify_rx_checksums(4);

        // Check how many DMA control blocks have been consumed
        uint32_t dma_ctrl_block_count = (dma_hw->ch[SDIO_DMA_CHB].read_addr - (uint32_t)&g_sdio.dma_blocks);
        dma_ctrl_block_count /= sizeof(g_sdio.dma_blocks[0]);

        // Compute how many complete 512 byte SDIO blocks have been transferred
        // When transfer ends, dma_ctrl_block_count == g_sdio.total_blocks * 2 + 1
        g_sdio.blocks_done = (dma_ctrl_block_count - 1) / 2;

        // NOTE: When all blocks are done, rx_poll() still returns SDIO_BUSY once.
        // This provides a chance to start the SCSI transfer before the last checksums
        // are computed. Any checksum failures can be indicated in SCSI status after
        // the data transfer has finished.
    }

    if (bytes_complete)
    {
        *bytes_complete = g_sdio.blocks_done * SDIO_BLOCK_SIZE;
    }

    if (g_sdio.transfer_state == SDIO_IDLE)
    {
        pio_sm_set_enabled(SDIO_PIO, SDIO_DATA_SM, false);
        // Verify all remaining checksums.
        sdio_verify_rx_checksums(g_sdio.total_blocks);

        if (g_sdio.checksum_errors == 0)
            return SDIO_OK;
        else
            return SDIO_ERR_DATA_CRC;
    }
    else if ((uint32_t)(millis() - g_sdio.transfer_start_time) > 1000)
    {
        debuglog("rp2040_sdio_rx_poll() timeout, "
            "PIO PC: ", (int)pio_sm_get_pc(SDIO_PIO, SDIO_DATA_SM) - (int)g_sdio.pio_data_rx_offset,
            " RXF: ", (int)pio_sm_get_rx_fifo_level(SDIO_PIO, SDIO_DATA_SM),
            " TXF: ", (int)pio_sm_get_tx_fifo_level(SDIO_PIO, SDIO_DATA_SM),
            " DMA CNT: ", dma_hw->ch[SDIO_DMA_CH].al2_transfer_count,
            " BD: ", g_sdio.blocks_done);
        rp2040_sdio_stop();
        return SDIO_ERR_DATA_TIMEOUT;
    }

    return SDIO_BUSY;
}


/*******************************************************
 * Data transmission to SD card
 *******************************************************/

static void __not_in_flash_func(sdio_start_next_block_tx)()
{
    // Initialize PIOs
    pio_sm_init(SDIO_PIO, SDIO_CMD_SM, g_sdio.pio_data_tx_offset, &g_sdio.pio_cfg_data_tx);

    // Re-set the pin direction things here
    pio_sm_set_pins(SDIO_PIO, SDIO_CMD_SM, 0xF);
    pio_sm_set_consecutive_pindirs(SDIO_PIO, SDIO_CMD_SM, SDIO_CLK, 1, true);
    pio_sm_set_consecutive_pindirs(SDIO_PIO, SDIO_CMD_SM, SDIO_D0, 4, true);

    // Configure DMA to send the data block payload (512 bytes)
    dma_channel_config dmacfg = dma_channel_get_default_config(SDIO_DMA_CH);
    channel_config_set_transfer_data_size(&dmacfg, DMA_SIZE_32);
    channel_config_set_read_increment(&dmacfg, true);
    channel_config_set_write_increment(&dmacfg, false);
    channel_config_set_dreq(&dmacfg, pio_get_dreq(SDIO_PIO, SDIO_CMD_SM, true));
    channel_config_set_bswap(&dmacfg, true);
    channel_config_set_chain_to(&dmacfg, SDIO_DMA_CHB);
    dma_channel_configure(SDIO_DMA_CH, &dmacfg,
        &SDIO_PIO->txf[SDIO_CMD_SM], g_sdio.data_buf + g_sdio.blocks_done * SDIO_WORDS_PER_BLOCK,
        SDIO_WORDS_PER_BLOCK, false);

    // Prepare second DMA channel to send the CRC and block end marker
    uint64_t crc = g_sdio.next_wr_block_checksum;
    g_sdio.end_token_buf[0] = (uint32_t)(crc >> 32);
    g_sdio.end_token_buf[1] = (uint32_t)(crc >>  0);
    g_sdio.end_token_buf[2] = 0xFFFFFFFF;
    channel_config_set_bswap(&dmacfg, false);
    dma_channel_configure(SDIO_DMA_CHB, &dmacfg,
        &SDIO_PIO->txf[SDIO_CMD_SM], g_sdio.end_token_buf, 3, false);

    // Enable IRQ to trigger when block is done
    dma_hw->ints1 = 1 << SDIO_DMA_CHB;
    dma_set_irq1_channel_mask_enabled(1 << SDIO_DMA_CHB, 1);

    // Initialize register X with nibble count
    pio_sm_put(SDIO_PIO, SDIO_CMD_SM, 1048);
    pio_sm_exec(SDIO_PIO, SDIO_CMD_SM, pio_encode_out(pio_x, 32));

    // Initialize CRC receiver Y bit count
    pio_sm_put(SDIO_PIO, SDIO_CMD_SM, 7);
    pio_sm_exec(SDIO_PIO, SDIO_CMD_SM, pio_encode_out(pio_y, 32));

    // Initialize pins to output and high
    pio_sm_exec(SDIO_PIO, SDIO_CMD_SM, pio_encode_set(pio_pins, 15));
    pio_sm_exec(SDIO_PIO, SDIO_CMD_SM, pio_encode_set(pio_pindirs, 15));

    // Write start token and start the DMA transfer.
    pio_sm_put(SDIO_PIO, SDIO_CMD_SM, 0xFFFFFFF0);
    dma_channel_start(SDIO_DMA_CH);

    // Start state machine
    pio_set_sm_mask_enabled(SDIO_PIO, (1ul << SDIO_CMD_SM)/* | (1ul << SDIO_DATA_SM)*/, true);
}

static void __not_in_flash_func(sdio_compute_next_tx_checksum)()
{
    assert (g_sdio.blocks_done < g_sdio.total_blocks && g_sdio.blocks_checksumed < g_sdio.total_blocks);
    int blockidx = g_sdio.blocks_checksumed++;
    g_sdio.next_wr_block_checksum = sdio_crc16_4bit_checksum(g_sdio.data_buf + blockidx * SDIO_WORDS_PER_BLOCK,
                                                             SDIO_WORDS_PER_BLOCK);
}

// Start transferring data from memory to SD card
sdio_status_t __not_in_flash_func(rp2040_sdio_tx_start)(const uint8_t *buffer, uint32_t num_blocks)
{
    // Buffer must be aligned
    assert(((uint32_t)buffer & 3) == 0 && num_blocks <= SDIO_MAX_BLOCKS);

    g_sdio.transfer_state = SDIO_TX;
    g_sdio.transfer_start_time = millis();
    g_sdio.data_buf = (uint32_t*)buffer;
    g_sdio.blocks_done = 0;
    g_sdio.total_blocks = num_blocks;
    g_sdio.blocks_checksumed = 0;
    g_sdio.checksum_errors = 0;

    // Compute first block checksum
    sdio_compute_next_tx_checksum();

    // Start first DMA transfer and PIO
    sdio_start_next_block_tx();

    if (g_sdio.blocks_checksumed < g_sdio.total_blocks)
    {
        // Precompute second block checksum
        sdio_compute_next_tx_checksum();
    }

    return SDIO_OK;
}

sdio_status_t __not_in_flash_func(check_sdio_write_response)(uint32_t card_response)
{
    uint8_t wr_status = card_response & 0x1F;
    //  5 = 0b0101 = data accepted  (11100101)
    // 11 = 0b1011 = CRC error      (11101011)
    // 13 = 0b1101 = Write Error    (11101101)

    if (wr_status == 0b101)
    {
        return SDIO_OK;
    }
    else if (wr_status == 0b1011)
    {
        log("SDIO card reports write CRC error, status ", card_response);
        return SDIO_ERR_WRITE_CRC;
    }
    else if (wr_status == 0b1101)
    {
        log("SDIO card reports write failure, status ", card_response);
        return SDIO_ERR_WRITE_FAIL;
    }
    else
    {
        log("SDIO card reports unknown write status ", card_response);
        return SDIO_ERR_WRITE_FAIL;
    }
}

// When a block finishes, this IRQ handler starts the next one
static void __not_in_flash_func(rp2040_sdio_tx_irq)()
{
    dma_hw->ints1 = 1 << SDIO_DMA_CHB;

    if (g_sdio.transfer_state == SDIO_TX)
    {
        if (!dma_channel_is_busy(SDIO_DMA_CH) && !dma_channel_is_busy(SDIO_DMA_CHB))
        {
            // Main data transfer is finished now.
            // When card is ready, PIO will put card response on RX fifo
            g_sdio.transfer_state = SDIO_TX_WAIT_IDLE;
            if (!pio_sm_is_rx_fifo_empty(SDIO_PIO, SDIO_CMD_SM))
            {
                // Card is already idle
                g_sdio.card_response = pio_sm_get(SDIO_PIO, SDIO_CMD_SM);
            }
            else
            {
                // Use DMA to wait for the response
                dma_channel_config dmacfg = dma_channel_get_default_config(SDIO_DMA_CHB);
                channel_config_set_transfer_data_size(&dmacfg, DMA_SIZE_8);
                channel_config_set_read_increment(&dmacfg, false);
                channel_config_set_write_increment(&dmacfg, false);
                channel_config_set_dreq(&dmacfg, pio_get_dreq(SDIO_PIO, SDIO_CMD_SM, false));
                dma_channel_configure(SDIO_DMA_CHB, &dmacfg,
                    &g_sdio.card_response, &SDIO_PIO->rxf[SDIO_CMD_SM], 1, true);
            }
        }
    }

    if (g_sdio.transfer_state == SDIO_TX_WAIT_IDLE)
    {
        if (!dma_channel_is_busy(SDIO_DMA_CHB))
        {
            g_sdio.wr_status = check_sdio_write_response(g_sdio.card_response);

            if (g_sdio.wr_status != SDIO_OK)
            {
                rp2040_sdio_stop();
                return;
            }

            g_sdio.blocks_done++;
            if (g_sdio.blocks_done < g_sdio.total_blocks)
            {
                sdio_start_next_block_tx();
                g_sdio.transfer_state = SDIO_TX;

                if (g_sdio.blocks_checksumed < g_sdio.total_blocks)
                {
                    // Precompute the CRC for next block so that it is ready when
                    // we want to send it.
                    sdio_compute_next_tx_checksum();
                }
            }
            else
            {
                rp2040_sdio_stop();
            }
        }
    }
}

// Check if transmission is complete
sdio_status_t __not_in_flash_func(rp2040_sdio_tx_poll)(uint32_t *bytes_complete)
{
    if (scb_hw->icsr & (0x1FFUL))
    {
        // Verify that IRQ handler gets called even if we are in hardfault handler
        rp2040_sdio_tx_irq();
    }

    if (bytes_complete)
    {
        *bytes_complete = g_sdio.blocks_done * SDIO_BLOCK_SIZE;
    }

    if (g_sdio.transfer_state == SDIO_IDLE)
    {
        rp2040_sdio_stop();
        return g_sdio.wr_status;
    }
    else if ((uint32_t)(millis() - g_sdio.transfer_start_time) > 1000)
    {
        debuglog("rp2040_sdio_tx_poll() timeout, "
            "PIO PC: ", (int)pio_sm_get_pc(SDIO_PIO, SDIO_CMD_SM) - (int)g_sdio.pio_data_tx_offset,
            " RXF: ", (int)pio_sm_get_rx_fifo_level(SDIO_PIO, SDIO_CMD_SM),
            " TXF: ", (int)pio_sm_get_tx_fifo_level(SDIO_PIO, SDIO_CMD_SM),
            " DMA CNT: ", dma_hw->ch[SDIO_DMA_CH].al2_transfer_count);

        rp2040_sdio_stop();
        return SDIO_ERR_DATA_TIMEOUT;
    }

    return SDIO_BUSY;
}

// Force everything to idle state
sdio_status_t __not_in_flash_func(rp2040_sdio_stop)()
{
    dma_channel_abort(SDIO_DMA_CH);
    dma_channel_abort(SDIO_DMA_CHB);
    dma_set_irq1_channel_mask_enabled(1 << SDIO_DMA_CHB, 0);
    pio_set_sm_mask_enabled(SDIO_PIO, (1ul << SDIO_CMD_SM) | (1ul << SDIO_DATA_SM), false);
    g_sdio.transfer_state = SDIO_IDLE;
    return SDIO_OK;
}

void __not_in_flash_func(rp2040_sdio_init)(int clock_divider)
{
    // Mark resources as being in use, unless it has been done already.
    static bool resources_claimed = false;
    if (!resources_claimed)
    {
        pio_sm_claim(SDIO_PIO, SDIO_CMD_SM);
        pio_sm_claim(SDIO_PIO, SDIO_DATA_SM);
        dma_channel_claim(SDIO_DMA_CH);
        dma_channel_claim(SDIO_DMA_CHB);
        resources_claimed = true;
    }

    memset(&g_sdio, 0, sizeof(g_sdio));

    dma_channel_abort(SDIO_DMA_CH);
    dma_channel_abort(SDIO_DMA_CHB);
    pio_sm_set_enabled(SDIO_PIO, SDIO_CMD_SM, false);
    pio_sm_set_enabled(SDIO_PIO, SDIO_DATA_SM, false);

    // Load PIO programs
    pio_clear_instruction_memory(SDIO_PIO);

    // Set pull resistors for all SD data lines
    gpio_set_pulls(SDIO_CLK, true, false);
    gpio_set_pulls(SDIO_CMD, true, false);
    gpio_set_pulls(SDIO_D0, true, false);
    gpio_set_pulls(SDIO_D1, true, false);
    gpio_set_pulls(SDIO_D2, true, false);
    gpio_set_pulls(SDIO_D3, true, false);

    //
    // Command state machine
    //
    uint16_t rw_prg_instr[cmd_rsp_program.length];
    memcpy(rw_prg_instr, cmd_rsp_program_instructions, sizeof(cmd_rsp_program_instructions));
    // If the instruction is a side-set 0 use delay0
    // If the instruction is a side-set 1 use delay1
    // cmd_rsp does not follow a certain pattern, so bluntly set delay values here.
    // NOTE: Any changes to cmd_rsp will require updates here as well
    rw_prg_instr[0] = cmd_rsp_program_instructions[0] | pio_encode_delay(g_bluescsi_timings->sdio.delay0);
    rw_prg_instr[1] = cmd_rsp_program_instructions[1] | pio_encode_delay(g_bluescsi_timings->sdio.delay1);

    rw_prg_instr[2] = cmd_rsp_program_instructions[2] | pio_encode_delay(g_bluescsi_timings->sdio.delay0);
    rw_prg_instr[3] = cmd_rsp_program_instructions[3] | pio_encode_delay(g_bluescsi_timings->sdio.delay1 + 1);

    rw_prg_instr[4] = cmd_rsp_program_instructions[4] | pio_encode_delay(g_bluescsi_timings->sdio.delay0);
    rw_prg_instr[5] = cmd_rsp_program_instructions[5] | pio_encode_delay(g_bluescsi_timings->sdio.delay1);

    rw_prg_instr[8] = cmd_rsp_program_instructions[8] | pio_encode_delay(g_bluescsi_timings->sdio.delay0);
    rw_prg_instr[9] = cmd_rsp_program_instructions[9] | pio_encode_delay(g_bluescsi_timings->sdio.delay1);

    pio_program cmd_rsp_prg_w_delays = {
        rw_prg_instr,
        cmd_rsp_program.length,
        cmd_rsp_program.origin,
        cmd_rsp_program.pio_version
    };

    g_sdio.pio_cmd_rsp_clk_offset = pio_add_program(SDIO_PIO, &cmd_rsp_prg_w_delays);
    g_sdio.pio_cfg_cmd_rsp = pio_cmd_rsp_program_config(g_sdio.pio_cmd_rsp_clk_offset, SDIO_CMD, SDIO_CLK, clock_divider, 0);

    pio_sm_init(SDIO_PIO, SDIO_CMD_SM, g_sdio.pio_cmd_rsp_clk_offset, &g_sdio.pio_cfg_cmd_rsp);
    pio_sm_set_pins(SDIO_PIO, SDIO_CMD_SM, 1);
    pio_sm_set_consecutive_pindirs(SDIO_PIO, SDIO_CMD_SM, SDIO_CLK, 1, true);
    pio_sm_set_consecutive_pindirs(SDIO_PIO, SDIO_CMD_SM, SDIO_CMD, 1, true);
    pio_sm_set_consecutive_pindirs(SDIO_PIO, SDIO_CMD_SM, SDIO_D0, 4, false);

    //
    // Data reception program
    //
    // Clear the temp memory array
    memset(rw_prg_instr, cmd_rsp_program.length, sizeof(rw_prg_instr[0]));
    memcpy(rw_prg_instr, rd_data_w_clock_program_instructions, sizeof(rd_data_w_clock_program_instructions));
    rw_prg_instr[1] = rd_data_w_clock_program_instructions[1] | pio_encode_delay(g_bluescsi_timings->sdio.delay0 + 1);
    rw_prg_instr[2] = rd_data_w_clock_program_instructions[2] | pio_encode_delay(g_bluescsi_timings->sdio.delay1 + 1);
    rw_prg_instr[3] = rd_data_w_clock_program_instructions[3] | pio_encode_delay(g_bluescsi_timings->sdio.delay0);
    rw_prg_instr[4] = rd_data_w_clock_program_instructions[4] | pio_encode_delay(g_bluescsi_timings->sdio.delay1);
    rw_prg_instr[5] = rd_data_w_clock_program_instructions[5] | pio_encode_delay(g_bluescsi_timings->sdio.delay0);

    pio_program rd_data_w_delays = {
        rw_prg_instr,
        rd_data_w_clock_program.length,
        rd_data_w_clock_program.origin,
        rd_data_w_clock_program.pio_version
    };

    g_sdio.pio_data_rx_offset = pio_add_program(SDIO_PIO, &rd_data_w_delays);
    g_sdio.pio_cfg_data_rx = pio_rd_data_w_clock_program_config(g_sdio.pio_data_rx_offset, SDIO_D0, SDIO_CLK, clock_divider);

    //
    // Data transmission program
    //
    // Clear the temp memory array
    memset(rw_prg_instr, cmd_rsp_program.length, sizeof(rw_prg_instr[0]));
    memcpy(rw_prg_instr, tx_data_w_clock_program_instructions, sizeof(tx_data_w_clock_program_instructions));
    rw_prg_instr[0] = tx_data_w_clock_program_instructions[0] | pio_encode_delay(g_bluescsi_timings->sdio.delay0);
    rw_prg_instr[1] = tx_data_w_clock_program_instructions[1] | pio_encode_delay(g_bluescsi_timings->sdio.delay1);
    rw_prg_instr[2] = tx_data_w_clock_program_instructions[2] | pio_encode_delay(g_bluescsi_timings->sdio.delay1);
    rw_prg_instr[3] = tx_data_w_clock_program_instructions[3] | pio_encode_delay(g_bluescsi_timings->sdio.delay1 + 1);
    rw_prg_instr[4] = tx_data_w_clock_program_instructions[4] | pio_encode_delay(g_bluescsi_timings->sdio.delay0 + 1);
    rw_prg_instr[5] = tx_data_w_clock_program_instructions[5] | pio_encode_delay(g_bluescsi_timings->sdio.delay1 + 1);
    rw_prg_instr[6] = tx_data_w_clock_program_instructions[6] | pio_encode_delay(g_bluescsi_timings->sdio.delay0 + 1);

    pio_program tx_data_w_delays = {
        rw_prg_instr,
        tx_data_w_clock_program.length,
        tx_data_w_clock_program.origin,
        tx_data_w_clock_program.pio_version
    };

    g_sdio.pio_data_tx_offset = pio_add_program(SDIO_PIO, &tx_data_w_delays);
    g_sdio.pio_cfg_data_tx = pio_tx_w_clock_program_config(g_sdio.pio_data_tx_offset, SDIO_D0, SDIO_CLK, clock_divider);

    // Disable SDIO pins input synchronizer.
    // This reduces input delay.
    // Because the CLK is driven synchronously to CPU clock,
    // there should be no metastability problems.
    SDIO_PIO->input_sync_bypass |= (1 << SDIO_CLK) | (1 << SDIO_CMD)
                                 | (1 << SDIO_D0) | (1 << SDIO_D1) | (1 << SDIO_D2) | (1 << SDIO_D3);

    // Redirect GPIOs to PIO
    gpio_set_function(SDIO_CMD, GPIO_FUNC_PIO1);
    gpio_set_function(SDIO_CLK, GPIO_FUNC_PIO1);
    gpio_set_function(SDIO_D0, GPIO_FUNC_PIO1);
    gpio_set_function(SDIO_D1, GPIO_FUNC_PIO1);
    gpio_set_function(SDIO_D2, GPIO_FUNC_PIO1);
    gpio_set_function(SDIO_D3, GPIO_FUNC_PIO1);

    // Set up IRQ handler when DMA completes.
    irq_set_exclusive_handler(DMA_IRQ_1, rp2040_sdio_tx_irq);
    irq_set_enabled(DMA_IRQ_1, true);
#if 0
#ifndef ENABLE_AUDIO_OUTPUT
    irq_set_exclusive_handler(DMA_IRQ_1, rp2040_sdio_tx_irq);
#else
    // seem to hit assertion in _exclusive_handler call due to DMA_IRQ_0 being shared?
    // slightly less efficient to do it this way, so investigate further at some point
    irq_add_shared_handler(DMA_IRQ_1, rp2040_sdio_tx_irq, 0xFF);
#endif
    irq_set_enabled(DMA_IRQ_1, true);
#endif
}

void __not_in_flash_func(rp2040_sdio_update_delays)(pio_program program, uint32_t offset, uint16_t additional_delay) {
    //log("Offset:", offset);
    uint16_t instr_to_rewrite;
    uint16_t existing_delay;
    for (int i = 0; i < program.length; i++) {
        instr_to_rewrite = program.instructions[i];
        //log("Old Instr:", i, ":", (uint32_t)instr_to_rewrite);
        if (instr_to_rewrite & PIO_INSTR_MASK_GET_DELAY) {  // If there's a delay, increment it.  Otherwise, leave it alone.
            existing_delay = (instr_to_rewrite & PIO_INSTR_MASK_GET_DELAY) >> 8;
            existing_delay += additional_delay;
            instr_to_rewrite = (instr_to_rewrite & PIO_INSTR_MASK_REMOVE_DELAY) | (existing_delay << 8);

            // Canonicalize JMP addresses
            if ((instr_to_rewrite & PIO_INSTR_JMP_MASK) == 0) {  // Highest three bits are zero on a JMP
                uint32_t jmp_address = instr_to_rewrite & PIO_INSTR_JMP_ADDR;
                jmp_address += offset;
                instr_to_rewrite = (instr_to_rewrite & (~ PIO_INSTR_JMP_ADDR)) | jmp_address;
            }

            //log("New Instr:", i, ":", (uint32_t)instr_to_rewrite);
            SDIO_PIO->instr_mem[offset + i] = instr_to_rewrite;
        }
    }
}

void __not_in_flash_func(rp2040_sdio_delay_increment)(uint16_t additional_delay) {
    /*
    Rewrite in-place every SDIO instruction for all the SDIO programs.
    These additional delay cycles effectively decrease the SDIO clock rate, which can be helpful in electrically noisy environments.
    */
    rp2040_sdio_update_delays(cmd_rsp_program, g_sdio.pio_cmd_rsp_clk_offset, additional_delay);
    rp2040_sdio_update_delays(rd_data_w_clock_program, g_sdio.pio_data_rx_offset, additional_delay);
    rp2040_sdio_update_delays(tx_data_w_clock_program, g_sdio.pio_data_tx_offset, additional_delay);
}