stm32: Add sdio support

Adds sdio support for the stm32f4 to allow for SD card flash updates
without power cycling some boards, e.g. BTT Octopus Pro.

Signed-off-by: H. Gregor Molter <gregor.molter@secretlab.de>
Signed-off-by: Kevin O'Connor <kevin@koconnor.net>

12 files changed, 1050 insertions, 11 deletions

diff --git a/scripts/spi_flash/spi_flash.py b/scripts/spi_flash/spi_flash.py
index db1e3f59..a3231b69 100644
--- a/scripts/spi_flash/spi_flash.py
+++ b/scripts/spi_flash/spi_flash.py
@@ -1,7 +1,8 @@
 #!/usr/bin/env python2
-# Module supporting uploads Klipper firmware to an SD Card via SPI
+# Module supporting uploads Klipper firmware to an SD Card via SPI and SDIO
 #
 # Copyright (C) 2021 Eric Callahan <arksine.code@gmail.com>
+# Copyright (C) 2022 H. Gregor Molter <gregor.molter@secretlab.de>
 #
 # This file may be distributed under the terms of the GNU GPLv3 license.
 import sys
@@ -36,7 +37,7 @@ def output(msg):
     sys.stdout.write("%s" % (msg,))
     sys.stdout.flush()
 
-def calc_crc7(data):
+def calc_crc7(data, with_padding=True):
     # G(x) = x^7 + x^3 + 1
     # Shift left as we are only calculating a 7 bit CRC
     poly = 0b10001001 << 1
@@ -47,6 +48,8 @@ def calc_crc7(data):
             crc = (crc << 1) ^ poly if crc & 0x80 else crc << 1
     # The sdcard protocol likes the crc left justfied with a
     # padded bit
+    if not with_padding:
+        return crc
     return crc | 1
 
 def calc_crc16(data):
@@ -89,11 +92,12 @@ def check_need_convert(board_name, config):
 
 ###########################################################
 #
-# SPI FLash Implementation
+# SPI / SDIO Flash Implementation
 #
 ###########################################################
 
 SPI_OID = 0
+SDIO_OID = 0
 SPI_MODE = 0
 SD_SPI_SPEED = 400000
 # MCU Command Constants
@@ -114,6 +118,20 @@ SW_SPI_BUS_CMD = "spi_set_software_bus oid=%d " \
 SPI_SEND_CMD = "spi_send oid=%c data=%*s"
 SPI_XFER_CMD = "spi_transfer oid=%c data=%*s"
 SPI_XFER_RESPONSE = "spi_transfer_response oid=%c response=%*s"
+SDIO_CFG_CMD = "config_sdio oid=%d blocksize=%u"
+SDIO_BUS_CMD = "sdio_set_bus oid=%d sdio_bus=%s"
+SDIO_SEND_CMD = "sdio_send_command oid=%c cmd=%c argument=%u wait=%c"
+SDIO_SEND_CMD_RESPONSE = "sdio_send_command_response oid=%c error=%c " \
+    "response=%*s"
+SDIO_READ_DATA="sdio_read_data oid=%c cmd=%c argument=%u"
+SDIO_READ_DATA_RESPONSE="sdio_read_data_response oid=%c error=%c read=%u"
+SDIO_WRITE_DATA="sdio_write_data oid=%c cmd=%c argument=%u"
+SDIO_WRITE_DATA_RESPONSE="sdio_write_data_response oid=%c error=%c write=%u"
+SDIO_READ_DATA_BUFFER="sdio_read_data_buffer oid=%c offset=%u len=%c"
+SDIO_READ_DATA_BUFFER_RESPONSE="sdio_read_data_buffer_response oid=%c data=%*s"
+SDIO_WRITE_DATA_BUFFER="sdio_write_data_buffer oid=%c offset=%u data=%*s"
+SDIO_SET_SPEED="sdio_set_speed oid=%c speed=%u"
+
 FINALIZE_CFG_CMD = "finalize_config crc=%d"
 
 class SPIFlashError(Exception):
@@ -135,6 +153,40 @@ class SPIDirect:
     def spi_transfer(self, data):
         return self._spi_transfer_cmd.send([self.oid, data])
 
+class SDIODirect:
+    def __init__(self, ser):
+        self.oid = SDIO_OID
+        self._sdio_send_cmd = mcu.CommandQueryWrapper(
+            ser, SDIO_SEND_CMD, SDIO_SEND_CMD_RESPONSE, self.oid)
+        self._sdio_read_data = mcu.CommandQueryWrapper(
+            ser, SDIO_READ_DATA, SDIO_READ_DATA_RESPONSE, self.oid)
+        self._sdio_write_data = mcu.CommandQueryWrapper(
+            ser, SDIO_WRITE_DATA, SDIO_WRITE_DATA_RESPONSE, self.oid)
+        self._sdio_read_data_buffer = mcu.CommandQueryWrapper(
+            ser, SDIO_READ_DATA_BUFFER, SDIO_READ_DATA_BUFFER_RESPONSE,
+            self.oid)
+        self._sdio_write_data_buffer = mcu.CommandWrapper(ser,
+            SDIO_WRITE_DATA_BUFFER)
+        self._sdio_set_speed = mcu.CommandWrapper(ser, SDIO_SET_SPEED)
+
+    def sdio_send_cmd(self, cmd, argument, wait):
+        return self._sdio_send_cmd.send([self.oid, cmd, argument, wait])
+
+    def sdio_read_data(self, cmd, argument):
+        return self._sdio_read_data.send([self.oid, cmd, argument])
+
+    def sdio_write_data(self, cmd, argument):
+        return self._sdio_write_data.send([self.oid, cmd, argument])
+
+    def sdio_read_data_buffer(self, offset, length=32):
+        return self._sdio_read_data_buffer.send([self.oid, offset, length])
+
+    def sdio_write_data_buffer(self, offset, data):
+        return self._sdio_write_data_buffer.send([self.oid, offset, data])
+
+    def sdio_set_speed(self, speed):
+        return self._sdio_set_speed.send([self.oid, speed])
+
 
 # FatFs Constants. Enums are implemented as lists. The item's index is its value
 DRESULT = ['RES_OK', 'RES_ERROR', 'RES_WRPRT', 'RES_NOTRDY', 'RES_PARERR']
@@ -154,8 +206,11 @@ SECTOR_SIZE = 512
 
 # FAT16/32 File System Support
 class FatFS:
-    def __init__(self, ser):
-        self.sdcard = SDCardSPI(ser)
+    def __init__(self, ser, spi=True):
+        if spi:
+            self.sdcard = SDCardSPI(ser)
+        else:
+            self.sdcard = SDCardSDIO(ser)
         self.disk_status = STA_NO_INIT | STA_NO_DISK
         self.ffi_callbacks = []
         self.ffi_main, self.ffi_lib = fatfs_lib.get_fatfs_ffi()
@@ -429,6 +484,10 @@ class SDCardFile:
 
 SD_COMMANDS = {
     'GO_IDLE_STATE': 0,
+    'ALL_SEND_CID': 2,
+    'SET_REL_ADDR': 3,
+    'SET_BUS_WIDTH': 6,
+    'SEL_DESEL_CARD': 7,
     'SEND_IF_COND': 8,
     'SEND_CSD': 9,
     'SEND_CID': 10,
@@ -785,6 +844,322 @@ class SDCardSPI:
             if err_msgs:
                 raise OSError("\n".join(err_msgs))
 
+class SDCardSDIO:
+    def __init__(self, ser):
+        self.sdio = SDIODirect(ser)
+        self.rca = 0
+        self.reactor = ser.get_reactor()
+        self.enable_crc = True
+        self.mutex = self.reactor.mutex()
+        self.initialized = False
+        self.sd_version = 0
+        self.high_capacity = False
+        self.write_protected = False
+        self.total_sectors = 0
+        self.card_info = collections.OrderedDict()
+
+    def init_sd(self):
+        def check_for_ocr_errors(reg):
+            # returns False if an error flag is set
+            return ((reg[0]&0xFD) | (reg[1]&0xFF) |
+                    (reg[2]&0xE0) | (reg[3]&0x08)) == 0
+        with self.mutex:
+            if self.initialized:
+                return
+            # Send reset command (CMD0)
+            if not self._send_command('GO_IDLE_STATE', 0):
+                raise OSError(
+                    "flash_sdcard: failed to reset SD Card\n"
+                    "Note that older (Version 1.0) SD cards can not be\n"
+                    "hot swapped.  Execute FIRMWARE_RESTART with the card\n"
+                    "inserted for successful initialization.")
+            # Check Voltage Range (CMD8). Only Cards meeting the v2.0 spec
+            # support this. V1.0 cards (and MMC) will return illegal command.
+            check_pattern = 0b1010
+            resp = self._send_command_with_response(
+                'SEND_IF_COND', (1 << 8) | check_pattern)
+            resp = resp.strip(b'\xFF')
+            if len(resp) != 4:
+                # CMD8 is illegal, this is a version 1.0 card
+                self.sd_version = 1
+            else:
+                self.sd_version = 2
+                if not (resp[-2] == 1 and resp[-1] == check_pattern):
+                    raise OSError("flash_sdcard: SD Card not running in a "
+                                    "compatible voltage range")
+            if self.sd_version == 2:
+
+                # Init card and come out of idle (ACMD41)
+                # Version 2 Cards may init before checking the OCR
+                # Allow vor LVDS card with 1.8v, too.
+                resp = self._check_command(lambda x: x[0]>>7 == 1,
+                    'SD_SEND_OP_COND', 0xC1100000, is_app_cmd=True,
+                    ignoreCRC=True)
+                if resp is None:
+                    raise OSError("flash_sdcard: SD Card did not come"
+                                  " out of IDLE after reset")
+            if len(resp) == 4:
+                if self.sd_version == 1:
+                    # Check acceptable volatage range for V1 cards
+                    if resp[1] != 0xFF:
+                        raise OSError("flash_sdcard: card does not support"
+                                      " 3.3v range")
+                elif self.sd_version == 2:
+                    # Determine if this is a high capacity sdcard
+                    if resp[0] & 0x40:
+                        self.high_capacity = True
+            else:
+                raise OSError("flash_sdcard: Invalid OCR Response")
+            if self.sd_version == 1:
+                # Init card and come out of idle (ACMD41)
+                # Version 1 Cards do this after checking the OCR
+                if not self._check_command(0, 'SD_SEND_OP_COND', 0,
+                                           is_app_cmd=True):
+                    raise OSError("flash_sdcard: SD Card did not come"
+                                  " out of IDLE after reset")
+
+            # Read out CID information register
+            self._process_cid_reg()
+
+            # Get card's relative address (RCA)
+            resp = self._send_command_with_response('SET_REL_ADDR', 0)
+            # Check if bits 15:13 have some error set
+            if (resp[-2] & 0xe0) != 0:
+                raise OSError("flash_sdcard: set card's "
+                              "relative address failed")
+            self.rca = resp[0]<<8 | resp[1]
+
+            # Read out CSD information register
+            self._process_csd_reg()
+
+            # Select the current card
+            if not self._check_command(check_for_ocr_errors, 'SEL_DESEL_CARD',
+                self.rca << 16, tries=1):
+                raise OSError("flash_sdcard: failed to select the card")
+
+            # Set SDIO clk speed to approx. 1 MHz
+            self.sdio.sdio_set_speed(1000000)
+
+            if self._check_command(check_for_ocr_errors, 'SET_BLOCKLEN',
+                SECTOR_SIZE, tries=5):
+                self.initialized = True
+            else:
+                raise OSError("flash_sdcard: failed to set block size")
+
+
+    def deinit(self):
+        with self.mutex:
+            if self.initialized:
+                # Reset the SD Card
+                try:
+                    if not self._send_command('GO_IDLE_STATE', 0):
+                        logging.info("flash_sdcard: failed to reset SD Card")
+                except Exception:
+                    logging.exception("Error resetting SD Card")
+            self.initialized = False
+            self.sd_version = 0
+            self.high_capacity = False
+            self.total_sectors = 0
+            self.card_info.clear()
+
+    def _check_command(self, check_func, cmd, args, is_app_cmd=False, tries=15,
+        ignoreCRC=False):
+        func = self._send_app_cmd_with_response if is_app_cmd else \
+            self._send_command_with_response
+        while True:
+            resp, rt = func(cmd, args, get_rt=True, ignoreCRC=ignoreCRC)
+            #logging.info("flash_sdcard: Check cmd %s, response: %s"
+            #             % (cmd, repr(resp)))
+            if resp and check_func(resp):
+                return resp
+            tries -= 1
+            if tries < 1:
+                return None
+            self.reactor.pause(rt + .1)
+
+    def _send_command(self, cmd, args, wait=0):
+        cmd_code = SD_COMMANDS[cmd]
+        argument = 0
+        if isinstance(args, int) or isinstance(args, long):
+            argument = args & 0xFFFFFFFF
+        elif isinstance(args, list) and len(args) == 4:
+            argument = ((args[0] << 24) & 0xFF000000) | \
+                        ((args[1] << 16) & 0x00FF0000) | \
+                        ((args[2] << 8) & 0x0000FF00) | \
+                        ((args[3] << 0) & 0x000000FF)
+        else:
+            raise OSError("flash_sdcard: Invalid SD Card Command argument")
+        params = self.sdio.sdio_send_cmd(cmd_code, argument, wait)
+        #logging.debug(f'_send_command({cmd=}, {args=}, {wait=}) -> '
+        #               'CMD: {cmd_code} ARG: {argument} -> {params=}')
+        if (wait == 0):
+            # Just return the error code if no response was requested
+            return params['error'] == 0
+        return params
+
+    def _send_command_with_response(self, cmd, args, check_error=True,
+        ignoreCRC=False, get_rt=False):
+        # Wait for a short response
+        params = self._send_command(cmd, args, wait=1)
+        response = params['response']
+        if check_error:
+            if params['error'] != 0:
+                if ignoreCRC and params['error'] != 4:
+                    response = []
+        if get_rt:
+            return bytearray(response), params['#receive_time']
+        else:
+            return bytearray(response)
+
+    def _send_app_cmd_with_response(self, cmd, args,
+        ignoreCRC=False, get_rt=False):
+        # CMD55 tells the SD Card that the next command is an
+        # Application Specific Command.
+        self._send_command_with_response('APP_CMD', self.rca << 16)
+        return self._send_command_with_response(
+            cmd, args, ignoreCRC=ignoreCRC, get_rt=get_rt)
+
+    def _process_cid_reg(self):
+        params = self._send_command('ALL_SEND_CID', 0, wait=2)
+        reg = bytearray(params['response'])
+        if reg is None:
+            raise OSError("flash_sdcard: Error reading CID register")
+
+        cid = collections.OrderedDict()
+        cid['manufacturer_id'] = reg[0]
+        cid['oem_id'] = reg[1:3].decode(encoding='ascii', errors='ignore')
+        cid['product_name'] = reg[3:8].decode(
+            encoding='ascii', errors='ignore')
+        cid['product_revision'] = str(reg[8] >> 4 & 0xFF) + "." \
+            + str(reg[8] & 0xFF)
+        cid['serial_number'] = "".join(["%02X" % (c,) for c in reg[9:13]])
+        mfg_year = (((reg[13] & 0xF) << 4) | ((reg[14] >> 4) & 0xF)) + 2000
+        mfg_month = reg[14] & 0xF
+        cid['manufacturing_date'] = "%d/%d" % (mfg_month, mfg_year)
+        crc = calc_crc7(reg[:15], with_padding=False)
+        if crc != reg[15]:
+            raise OSError("flash_sdcard: CID crc mismatch: 0x%02X, recd: 0x%02X"
+                          % (crc, reg[15]))
+        self.card_info.update(cid)
+
+    def _process_csd_reg(self):
+        params = self._send_command('SEND_CSD', self.rca << 16, wait=2)
+        reg = bytearray(params['response'])
+        if reg is None:
+            raise OSError("flash_sdcard: Error reading CSD register")
+        str_capacity = "Invalid"
+        max_capacity = 0
+        csd_type = (reg[0] >> 6) & 0x3
+        if csd_type == 0:
+            # Standard Capacity (CSD Version 1.0)
+            max_block_len = 2**(reg[5] & 0xF)
+            c_size = ((reg[6] & 0x3) << 10) | (reg[7] << 2) | \
+                ((reg[8] >> 6) & 0x3)
+            c_mult = 2**((((reg[9] & 0x3) << 1) | (reg[10] >> 7)) + 2)
+            max_capacity = (c_size + 1) * c_mult * max_block_len
+            str_capacity = "%.1f MiB" % (max_capacity / (1024.0**2))
+        elif csd_type == 1:
+            # High Capacity (CSD Version 2.0)
+            c_size = ((reg[7] & 0x3F) << 16) | (reg[8] << 8) | reg[9]
+            max_capacity = (c_size + 1) * 512 * 1024
+            str_capacity = "%.1f GiB" % (max_capacity / (1024.0**3))
+        else:
+            logging.info("sdcard: Unsupported csd type: %d" % (csd_type))
+        self.write_protected = (reg[14] & 0x30) != 0
+        crc = calc_crc7(reg[:15], with_padding=False)
+        if crc != reg[15]:
+            raise OSError("flash_sdcard: CSD crc mismatch: 0x%02X, recd: 0x%02X"
+                          % (crc, reg[15]))
+        self.card_info['capacity'] = str_capacity
+        self.total_sectors = max_capacity // SECTOR_SIZE
+
+    def print_card_info(self, print_func=logging.info):
+        print_func("\nSD Card Information:")
+        print_func("Version: %.1f" % (self.sd_version))
+        print_func("SDHC/SDXC: %s" % (self.high_capacity))
+        print_func("Write Protected: %s" % (self.write_protected))
+        print_func("Sectors: %d" % (self.total_sectors,))
+        for name, val in self.card_info.items():
+            print_func("%s: %s" % (name, val))
+
+    def read_sector(self, sector):
+        buf = None
+        err_msg = "flash_sdcard: read error, sector %d" % (sector,)
+        with self.mutex:
+            if not 0 <= sector < self.total_sectors:
+                err_msg += " out of range"
+            elif not self.initialized:
+                err_msg += ", SD Card not initialized"
+            else:
+                offset = sector
+                if not self.high_capacity:
+                    offset = sector * SECTOR_SIZE
+
+                params = self.sdio.sdio_read_data(
+                    SD_COMMANDS['READ_SINGLE_BLOCK'], offset)
+                if params['error'] != 0:
+                    raise OSError(
+                        'Read data failed. Error code=%d' %(params['error'],) )
+                if params['read'] != SECTOR_SIZE:
+                    raise OSError(
+                        'Read data failed. Expected %d bytes but got %d.' %
+                        (SECTOR_SIZE, params['read']) )
+
+                buf = bytearray()
+                offset = 0
+                while SECTOR_SIZE-len(buf)>0:
+                    rest = min(SECTOR_SIZE-len(buf), 32)
+                    params = self.sdio.sdio_read_data_buffer(
+                        offset, length=rest)
+                    temp = bytearray(params['data'])
+                    if len(temp) == 0:
+                        raise OSError("Read zero bytes from buffer")
+                    buf += temp
+                    offset += len(temp)
+            if buf is None:
+                raise OSError(err_msg)
+            return buf
+
+    def write_sector(self, sector, data):
+        with self.mutex:
+            if not 0 <= sector < self.total_sectors:
+                raise OSError(
+                    "flash_sdcard: write error, sector number %d invalid"
+                    % (sector))
+            if not self.initialized:
+                raise OSError("flash_sdcard: write error, SD Card not"
+                              " initialized")
+            outbuf = bytearray(data)
+            if len(outbuf) > SECTOR_SIZE:
+                raise OSError("sd_card: Cannot write sector larger"
+                              " than %d bytes"
+                              % (SECTOR_SIZE))
+            elif len(outbuf) < SECTOR_SIZE:
+                outbuf += bytearray([0] * (SECTOR_SIZE - len(outbuf)))
+            offset = sector
+            if not self.high_capacity:
+                offset = sector * SECTOR_SIZE
+
+            CHUNKSIZE = 32
+            for i in range(0, SECTOR_SIZE, CHUNKSIZE):
+                self.sdio.sdio_write_data_buffer(i, outbuf[i:i+CHUNKSIZE])
+            params = self.sdio.sdio_write_data(
+                SD_COMMANDS['WRITE_BLOCK'], offset)
+            if (params['error'] != 0) or (params['write'] != SECTOR_SIZE):
+                raise OSError(
+                    "flash_sdcard: Error writing to sector %d"% (sector,))
+
+            status = self._send_command_with_response(
+                'SEND_STATUS', self.rca << 16)
+            if len(status) != 4:
+                raise OSError("flash_sdcard: Failed to get status response"
+                              " after write: %s" % (repr(status),))
+            if ((status[3]>>1) & 0x0F) != 0:
+                # Bit 12:9 are not "0" (card is in idle)
+                raise OSError("flash_sdcard: Write error."
+                              " Card is not in transfer state: 0x%02X"
+                              % (((status[3]>>1) & 0x0F)))
+
 SDIO_WARNING = """
 This board requires a manual reboot to complete the flash process.
 If the board's bootloader uses SDIO mode for its SDCard, then a full
@@ -896,7 +1271,7 @@ class MCUConnection:
             return True
         return False
 
-    def configure_mcu(self, printfunc=logging.info):
+    def _configure_mcu_spibus(self, printfunc=logging.info):
         # TODO: add commands for buttons?  Or perhaps an endstop?  We
         # just need to be able to query the status of the detect pin
         cs_pin = self.board_config['cs_pin'].upper()
@@ -952,6 +1327,41 @@ class MCUConnection:
             raise SPIFlashError(
                 "Failed to Initialize SD Card. Is it inserted?")
 
+    def _configure_mcu_sdiobus(self, printfunc=logging.info):
+        bus = self.board_config['sdio_bus']
+        bus_enums = self.enumerations.get(
+            'sdio_bus', self.enumerations.get('bus'))
+        pin_enums = self.enumerations.get('pin')
+        if bus not in bus_enums:
+            raise SPIFlashError("Invalid SDIO Bus: %s" % (bus,))
+        bus_cmd = SDIO_BUS_CMD % (SDIO_OID, bus)
+        sdio_cfg_cmd = SDIO_CFG_CMD % (SDIO_OID, SECTOR_SIZE)
+        cfg_cmds = [ALLOC_OIDS_CMD % (1,), sdio_cfg_cmd, bus_cmd]
+        for cmd in cfg_cmds:
+            self._serial.send(cmd)
+        config_crc = zlib.crc32('\n'.join(cfg_cmds).encode()) & 0xffffffff
+        self._serial.send(FINALIZE_CFG_CMD % (config_crc,))
+        config = self.get_mcu_config()
+        if not config["is_config"] or config["is_shutdown"]:
+            raise MCUConfigError("Failed to configure MCU")
+        printfunc("Initializing SD Card and Mounting file system...")
+        self.fatfs = FatFS(self._serial,spi=False)
+        self.reactor.pause(self.reactor.monotonic() + .5)
+        try:
+            self.fatfs.mount(printfunc)
+        except OSError:
+            logging.exception("SD Card Mount Failure")
+            raise SPIFlashError(
+                "Failed to Initialize SD Card. Is it inserted?")
+
+    def configure_mcu(self, printfunc=logging.info):
+        if 'spi_bus' in self.board_config:
+            self._configure_mcu_spibus(printfunc=printfunc)
+        elif 'sdio_bus' in self.board_config:
+            self._configure_mcu_sdiobus(printfunc=printfunc)
+        else:
+            raise SPIFlashError("Unknown bus defined in board_defs.py.")
+
     def sdcard_upload(self):
         output("Uploading Klipper Firmware to SD Card...")
         input_sha = hashlib.sha1()
diff --git a/src/Kconfig b/src/Kconfig
index 936e3410..283639b7 100644
--- a/src/Kconfig
+++ b/src/Kconfig
@@ -115,6 +115,8 @@ config HAVE_GPIO_ADC
     bool
 config HAVE_GPIO_SPI
     bool
+config HAVE_GPIO_SDIO
+    bool
 config HAVE_GPIO_I2C
     bool
 config HAVE_GPIO_HARD_PWM
diff --git a/src/Makefile b/src/Makefile
index f5c32f1a..dc0427c7 100644
--- a/src/Makefile
+++ b/src/Makefile
@@ -5,6 +5,7 @@ src-$(CONFIG_HAVE_GPIO) += initial_pins.c gpiocmds.c stepper.c endstop.c \
     trsync.c
 src-$(CONFIG_HAVE_GPIO_ADC) += adccmds.c
 src-$(CONFIG_HAVE_GPIO_SPI) += spicmds.c thermocouple.c
+src-$(CONFIG_HAVE_GPIO_SDIO) += sdiocmds.c
 src-$(CONFIG_HAVE_GPIO_I2C) += i2ccmds.c
 src-$(CONFIG_HAVE_GPIO_HARD_PWM) += pwmcmds.c
 bb-src-$(CONFIG_HAVE_GPIO_SPI) := spi_software.c sensor_adxl345.c sensor_angle.c
diff --git a/src/sdiocmds.c b/src/sdiocmds.c
new file mode 100644
index 00000000..d9c020b4
--- /dev/null
+++ b/src/sdiocmds.c
@@ -0,0 +1,166 @@
+// Commands for sending messages on an SDIO bus
+//
+// Copyright (C) 2022 H. Gregor Molter <gregor.molter@secretlab.de>
+//
+// This file may be distributed under the terms of the GNU GPLv3 license.
+
+#include <string.h> // memcpy
+#include "board/gpio.h" // gpio_in_setup
+#include "board/sdio.h" // sdio_setup
+#include "basecmd.h" // oid_alloc
+#include "command.h" // DECL_COMMAND
+#include "sched.h" // DECL_SHUTDOWN
+
+struct sdiodev_s {
+    struct sdio_config sdio_config;
+    uint32_t blocksize;
+    uint32_t speed;
+    uint8_t data_buffer[4096];
+};
+
+#define TIMEOUT_MSEC 500
+
+void
+command_config_sdio(uint32_t *args)
+{
+    struct sdiodev_s *sdio = oid_alloc(args[0], command_config_sdio
+                                       , sizeof(*sdio));
+    sdio->blocksize = args[1];
+    sdio->speed = 400000; // Initial speed set to ~400khz
+}
+DECL_COMMAND(command_config_sdio, "config_sdio oid=%c blocksize=%u");
+
+struct sdiodev_s *
+sdiodev_oid_lookup(uint8_t oid)
+{
+    return oid_lookup(oid, command_config_sdio);
+}
+
+void
+command_sdio_set_bus(uint32_t *args)
+{
+    struct sdiodev_s *sdio = sdiodev_oid_lookup(args[0]);
+    sdio->sdio_config = sdio_setup(args[1]);
+}
+DECL_COMMAND(command_sdio_set_bus, "sdio_set_bus oid=%c sdio_bus=%u");
+
+void
+command_sdio_set_speed(uint32_t *args)
+{
+    struct sdiodev_s *sdio = sdiodev_oid_lookup(args[0]);
+    sdio->speed = args[1];
+    sdio_set_speed(sdio->sdio_config, sdio->speed);
+}
+DECL_COMMAND(command_sdio_set_speed, "sdio_set_speed oid=%c speed=%u");
+
+void
+command_sdio_send_command(uint32_t *args)
+{
+    uint8_t oid = args[0];
+    uint8_t cmd = args[1];
+    uint32_t argument = args[2];
+    uint8_t wait = args[3];
+    struct sdiodev_s *sdio = sdiodev_oid_lookup(oid);
+    uint8_t response[16];
+    uint8_t response_len = 0;
+    uint8_t err = sdio_send_command(sdio->sdio_config, cmd, argument, wait
+                                    , response, &response_len);
+    sendf("sdio_send_command_response oid=%c error=%c response=%*s"
+          , oid, err, response_len, response);
+}
+DECL_COMMAND(command_sdio_send_command
+             , "sdio_send_command oid=%c cmd=%c argument=%u wait=%c");
+
+void
+command_sdio_read_data(uint32_t *args)
+{
+    uint8_t oid = args[0];
+    uint8_t cmd = args[1];
+    uint32_t argument = args[2];
+    uint32_t data_len = 0;
+    struct sdiodev_s *sdio = sdiodev_oid_lookup(oid);
+    uint32_t timeout = TIMEOUT_MSEC*sdio->speed/1000;
+    uint8_t err = sdio_prepare_data_transfer(sdio->sdio_config, 1, 1
+                                             , sdio->blocksize, timeout);
+    if (err == 0) {
+        err = sdio_send_command(sdio->sdio_config, cmd, argument
+                                , 1, NULL, NULL);
+        if (err == 0) {
+            data_len = sdio->blocksize;
+            if (data_len <= sizeof(sdio->data_buffer)) {
+                err = sdio_read_data(sdio->sdio_config, sdio->data_buffer
+                                     , 1, sdio->blocksize);
+            } else {
+                data_len = 0;
+            }
+        }
+    }
+    sendf("sdio_read_data_response oid=%c error=%c read=%u"
+          , oid, err, data_len);
+}
+DECL_COMMAND(command_sdio_read_data
+             , "sdio_read_data oid=%c cmd=%c argument=%u");
+
+void
+command_sdio_write_data(uint32_t *args)
+{
+    uint8_t oid = args[0];
+    uint8_t cmd = args[1];
+    uint32_t argument = args[2];
+    uint32_t data_len = 0;
+    struct sdiodev_s *sdio = sdiodev_oid_lookup(oid);
+    uint32_t timeout = TIMEOUT_MSEC*sdio->speed/1000;
+    uint8_t err = sdio_prepare_data_transfer(sdio->sdio_config, 0, 1
+                                             , sdio->blocksize, timeout);
+    if (err == 0) {
+        err = sdio_send_command(sdio->sdio_config, cmd, argument
+                                , 1, NULL, NULL);
+        if (err == 0) {
+            data_len = sdio->blocksize;
+            if (data_len <= sizeof(sdio->data_buffer)) {
+                err = sdio_write_data(sdio->sdio_config, sdio->data_buffer
+                                      , 1, sdio->blocksize);
+            } else {
+                data_len = 0;
+            }
+        }
+    }
+    sendf("sdio_write_data_response oid=%c error=%c write=%u"
+          , oid, err, data_len);
+}
+DECL_COMMAND(command_sdio_write_data
+             , "sdio_write_data oid=%c cmd=%c argument=%u");
+
+void
+command_sdio_read_data_buffer(uint32_t *args)
+{
+    uint8_t oid = args[0];
+    uint32_t offset = args[1];
+    uint8_t len = args[2];
+    struct sdiodev_s *sdio = sdiodev_oid_lookup(oid);
+    uint8_t *buf = &(sdio->data_buffer[offset]);
+
+    if (offset + len > sizeof(sdio->data_buffer)) {
+        len = 0;
+    }
+    sendf("sdio_read_data_buffer_response oid=%c data=%*s", oid, len, buf);
+}
+DECL_COMMAND(command_sdio_read_data_buffer
+             , "sdio_read_data_buffer oid=%c offset=%u len=%c");
+
+void
+command_sdio_write_data_buffer(uint32_t *args)
+{
+    uint8_t oid = args[0];
+    uint32_t offset = args[1];
+    uint8_t write_data_len = args[2];
+    uint8_t *write_data = command_decode_ptr(args[3]);
+    struct sdiodev_s *sdio = sdiodev_oid_lookup(oid);
+    uint8_t *buf = &(sdio->data_buffer[offset]);
+
+    if (offset + write_data_len <= sizeof(sdio->data_buffer)) {
+        memcpy(buf, write_data, write_data_len);
+    }
+}
+DECL_COMMAND(command_sdio_write_data_buffer
+             , "sdio_write_data_buffer oid=%c offset=%u data=%*s");
diff --git a/src/sdiocmds.h b/src/sdiocmds.h
new file mode 100644
index 00000000..24ec9811
--- /dev/null
+++ b/src/sdiocmds.h
@@ -0,0 +1,8 @@
+#ifndef __SDIOCMDS_H
+#define __SDIOCMDS_H
+
+#include <stdint.h> // uint8_t
+
+struct sdiodev_s *sdiodev_oid_lookup(uint8_t oid);
+
+#endif // sdiocmds.h
diff --git a/src/stm32/Kconfig b/src/stm32/Kconfig
index 0597f63d..50db29d2 100644
--- a/src/stm32/Kconfig
+++ b/src/stm32/Kconfig
@@ -9,6 +9,7 @@ config STM32_SELECT
     select HAVE_GPIO_ADC
     select HAVE_GPIO_I2C if !(MACH_STM32F031 || MACH_STM32H7)
     select HAVE_GPIO_SPI if !MACH_STM32F031
+    select HAVE_GPIO_SDIO if MACH_STM32F4
     select HAVE_GPIO_HARD_PWM if MACH_STM32F1 || MACH_STM32F4 || MACH_STM32G0 || MACH_STM32H7
     select HAVE_GPIO_BITBANGING if !MACH_STM32F031
     select HAVE_STRICT_TIMING
diff --git a/src/stm32/Makefile b/src/stm32/Makefile
index 15ea4e2b..390ad8d5 100644
--- a/src/stm32/Makefile
+++ b/src/stm32/Makefile
@@ -64,6 +64,8 @@ src-$(CONFIG_MACH_STM32L4) += stm32/stm32h7_adc.c stm32/stm32f0_i2c.c
 spi-src-y := stm32/spi.c
 spi-src-$(CONFIG_MACH_STM32H7) := stm32/stm32h7_spi.c
 src-$(CONFIG_HAVE_GPIO_SPI) += $(spi-src-y)
+sdio-src-y := stm32/sdio.c
+src-$(CONFIG_HAVE_GPIO_SDIO) += $(sdio-src-y)
 usb-src-$(CONFIG_HAVE_STM32_USBFS) := stm32/usbfs.c
 usb-src-$(CONFIG_HAVE_STM32_USBOTG) := stm32/usbotg.c
 src-$(CONFIG_USBSERIAL) += $(usb-src-y) stm32/chipid.c generic/usb_cdc.c
diff --git a/src/stm32/gpioperiph.c b/src/stm32/gpioperiph.c
index 06cdaa05..ef421c77 100644
--- a/src/stm32/gpioperiph.c
+++ b/src/stm32/gpioperiph.c
@@ -6,7 +6,7 @@
 
 #include "internal.h" // gpio_peripheral
 
-// Set the mode and extended function of a pin
+// Set the mode, extended function and speed of a pin
 void
 gpio_peripheral(uint32_t gpio, uint32_t mode, int pullup)
 {
@@ -16,7 +16,8 @@ gpio_peripheral(uint32_t gpio, uint32_t mode, int pullup)
     gpio_clock_enable(regs);
 
     // Configure GPIO
-    uint32_t mode_bits = mode & 0xf, func = (mode >> 4) & 0xf, od = mode >> 8;
+    uint32_t mode_bits = mode & 0xf, func = (mode >> 4) & 0xf;
+    uint32_t od = (mode >> 8) & 0x1, hs = (mode >> 9) & 0x1;
     uint32_t pup = pullup ? (pullup > 0 ? 1 : 2) : 0;
     uint32_t pos = gpio % 16, af_reg = pos / 8;
     uint32_t af_shift = (pos % 8) * 4, af_msk = 0x0f << af_shift;
@@ -33,6 +34,6 @@ gpio_peripheral(uint32_t gpio, uint32_t mode, int pullup)
     // stm32f4 is ~50Mhz at 40pF
     // stm32g0 is ~30Mhz at 50pF
     // stm32h7 is ~85Mhz at 50pF
-    uint32_t ospeed = CONFIG_MACH_STM32F0 ? 0x01 : 0x02;
+    uint32_t ospeed = hs ? 0x03 : (CONFIG_MACH_STM32F0 ? 0x01 : 0x02);
     regs->OSPEEDR = (regs->OSPEEDR & ~m_msk) | (ospeed << m_shift);
 }
diff --git a/src/stm32/internal.h b/src/stm32/internal.h
index b5971ed7..0422bf65 100644
--- a/src/stm32/internal.h
+++ b/src/stm32/internal.h
@@ -32,6 +32,7 @@ extern GPIO_TypeDef * const digital_regs[];
 #define GPIO_INPUT 0
 #define GPIO_OUTPUT 1
 #define GPIO_OPEN_DRAIN 0x100
+#define GPIO_HIGH_SPEED 0x200
 #define GPIO_FUNCTION(fn) (2 | ((fn) << 4))
 #define GPIO_ANALOG 3
 void gpio_peripheral(uint32_t gpio, uint32_t mode, int pullup);
diff --git a/src/stm32/sdio.c b/src/stm32/sdio.c
new file mode 100644
index 00000000..c64bca38
--- /dev/null
+++ b/src/stm32/sdio.c
@@ -0,0 +1,414 @@
+// SDIO functions on STM32
+//
+// Copyright (C) 2022  H. Gregor Molter <gregor.molter@secretlab.de>
+//
+// This file may be distributed under the terms of the GNU GPLv3 license.
+
+#include "board/io.h" // readb, writeb
+#include "command.h" // shutdown
+#include "sdio.h" // sdio_setup
+#include "internal.h" // gpio_peripheral
+#include "sched.h" // sched_shutdown
+#include "generic/armcm_timer.h" // udelay
+
+struct sdio_info {
+    SDIO_TypeDef *sdio;
+    uint8_t clk_pin, cmd_pin, dat0_pin, dat1_pin, dat2_pin, dat3_pin, function;
+};
+
+enum {
+    SDIO_OK = 0,
+    SDIO_ERROR = 1,
+    SDIO_TIMEOUT = 2,
+    SDIO_CMD_RESPONSE_TIMEOUT = 3,
+    SDIO_CRC_FAIL = 4,
+    SDIO_WRONG_CMD_RESPONSE = 5,
+    SDIO_WRONG_BLOCKSIZE = 6,
+    SDIO_DATA_TIMEOUT = 7,
+    SDIO_READ_OVERRUN = 8,
+    SDIO_WRITE_UNDERRUN = 9,
+    SDIO_NO_DATA_MEMORY = 10,
+};
+
+enum {
+    SDIO_CARDVER_UNKNOWN = 0,
+    SDIO_CARDVER_1X = 1,
+    SDIO_CARDVER_2X = 2
+};
+
+enum {
+    SDIO_CARDTYPE_UNKNOWN = 0,
+    SDIO_CARDTYPE_SDHC_SDXC = 1,
+    SDIO_CARDTYPE_SDSC = 2
+};
+
+enum {
+    SDIO_WAIT_NO_RESPONSE = 0,
+    SDIO_WAIT_SHORT_RESPONSE = 1,
+    SDIO_WAIT_LONG_RESPONSE = 2
+};
+
+// PINS: CLK -> PC12 , CMD -> PD2,
+//       DAT0 -> PC8, DAT1 -> PC9, DAT2 -> PC10, DAT3 -> PC11
+DECL_ENUMERATION("sdio_bus", "sdio", 0);
+DECL_CONSTANT_STR("BUS_PINS_sdio", "PC12,PD2,PC8,PC9,PC10,PC11");
+
+#define SDIO_FUNCTION GPIO_FUNCTION(12)
+
+static const struct sdio_info sdio_bus[] = {
+    { SDIO, GPIO('C', 12), GPIO('D', 2), GPIO('C', 8),
+      GPIO('C', 9), GPIO('C', 10), GPIO('C', 11), SDIO_FUNCTION },
+};
+
+#define SDIO_CLK_FREQ 48000000
+#define SDIO_INIT_CLK 400000
+#define SDIO_MAX_TIMEOUT 500 // Wait for at least 500ms before a timeout occurs
+#define CLKCR_CLEAR_MASK (SDIO_CLKCR_CLKDIV | SDIO_CLKCR_PWRSAV | \
+    SDIO_CLKCR_BYPASS | SDIO_CLKCR_WIDBUS | SDIO_CLKCR_NEGEDGE | \
+    SDIO_CLKCR_HWFC_EN)
+#define DCTRL_CLEAR_MASK (SDIO_DCTRL_DTEN | SDIO_DCTRL_DTDIR | \
+    SDIO_DCTRL_DTMODE | SDIO_DCTRL_DBLOCKSIZE)
+#define CMD_CLEAR_MASK (SDIO_CMD_CMDINDEX | SDIO_CMD_WAITRESP | \
+    SDIO_CMD_WAITINT | SDIO_CMD_WAITPEND | SDIO_CMD_CPSMEN | \
+    SDIO_CMD_SDIOSUSPEND)
+#define SDIO_CLOCK_BYPASS_DISABLE 0
+#define SDIO_CLOCK_EDGE_RISING 0
+#define SDIO_CLOCK_POWER_SAVE_DISABLE 0
+#define SDIO_BUS_WIDE_1B 0
+#define SDIO_BUS_WIDE_4B SDIO_CLKCR_WIDBUS_0
+#define SDIO_HARDWARE_FLOW_CONTROL_DISABLE 0
+#define SDIO_HARDWARE_FLOW_CONTROL_ENABLE SDIO_CLKCR_HWFC_EN
+#define SDIO_CMD_FLAGS (SDIO_STA_CCRCFAIL | SDIO_STA_CTIMEOUT | \
+    SDIO_STA_CMDREND | SDIO_STA_CMDSENT)
+#define SDIO_STATIC_FLAGS (SDIO_STA_CCRCFAIL | SDIO_STA_DCRCFAIL | \
+    SDIO_STA_CTIMEOUT | SDIO_STA_DTIMEOUT | SDIO_STA_TXUNDERR | \
+    SDIO_STA_RXOVERR | SDIO_STA_CMDREND | SDIO_STA_CMDSENT | \
+    SDIO_STA_DATAEND | SDIO_STA_DBCKEND | SDIO_STA_SDIOIT)
+
+struct sdio_config
+sdio_setup(uint32_t bus)
+{
+    if (bus >= ARRAY_SIZE(sdio_bus))
+        shutdown("Invalid sdio bus");
+
+    // Enable SDIO
+    SDIO_TypeDef *sdio = sdio_bus[bus].sdio;
+    if (!is_enabled_pclock((uint32_t)sdio)) {
+        // Enable clock
+        enable_pclock((uint32_t)sdio);
+        // Initialize pins
+        gpio_peripheral(
+            sdio_bus[bus].dat0_pin, sdio_bus[bus].function|GPIO_HIGH_SPEED, 1);
+        gpio_peripheral(
+            sdio_bus[bus].dat1_pin, sdio_bus[bus].function|GPIO_HIGH_SPEED, 1);
+        gpio_peripheral(
+            sdio_bus[bus].dat2_pin, sdio_bus[bus].function|GPIO_HIGH_SPEED, 1);
+        gpio_peripheral(
+            sdio_bus[bus].dat3_pin, sdio_bus[bus].function|GPIO_HIGH_SPEED, 1);
+        gpio_peripheral(
+            sdio_bus[bus].cmd_pin, sdio_bus[bus].function|GPIO_HIGH_SPEED, 1);
+        gpio_peripheral(
+            sdio_bus[bus].clk_pin, sdio_bus[bus].function|GPIO_HIGH_SPEED, 1);
+    }
+
+    struct sdio_config sdio_config = { .sdio = sdio };
+
+    // Setup SDIO with 1 bit width first and slow clock ~400 kHz
+    sdio_set_speed(sdio_config, SDIO_INIT_CLK);
+
+    // Disable clk
+    CLEAR_BIT(sdio->CLKCR, SDIO_CLKCR_CLKEN);
+    // Set power state to _on_
+    sdio->POWER = SDIO_POWER_PWRCTRL;
+    // Wait for 2ms (standard: at least 1ms) to settle
+    udelay(2000);
+    // Enable Clk
+    SET_BIT(sdio->CLKCR, SDIO_CLKCR_CLKEN);
+
+    return sdio_config;
+}
+
+uint32_t
+sdio_get_cmd_error(struct sdio_config sdio, uint32_t flags)
+{
+    SDIO_TypeDef *regs = sdio.sdio;
+
+    // wait for a timeout (max. SDIO_MAX_TIMEOUT) in msec.
+    // 8 cycles is the instruction cycles for the loop below.
+    uint32_t sta;
+    uint32_t count = SDIO_MAX_TIMEOUT * (SystemCoreClock / 8U / 1000U);
+    do {
+        if (count-- == 0) {
+            return SDIO_TIMEOUT;
+        }
+        sta = regs->STA;
+    } while ((sta & flags) == 0 || (sta & SDIO_STA_CMDACT) != 0);
+
+    regs->ICR = SDIO_CMD_FLAGS;
+
+    if (sta & SDIO_STA_CTIMEOUT) {
+        return SDIO_CMD_RESPONSE_TIMEOUT;
+    }
+    if (sta & SDIO_STA_CCRCFAIL) {
+        return SDIO_CRC_FAIL;
+    }
+
+    return SDIO_OK;
+}
+
+uint8_t
+sdio_send_command(struct sdio_config sdio_config, uint8_t cmd,
+    uint32_t argument, uint8_t wait, uint8_t *response_data,
+    uint8_t *response_data_len)
+{
+    SDIO_TypeDef *sdio = sdio_config.sdio;
+    uint32_t wait_flags = 0; // valid for SDIO_WAIT_NO_RESPONSE
+    uint32_t sta_flags = (wait == SDIO_WAIT_NO_RESPONSE) ?
+        SDIO_STA_CMDSENT : SDIO_STA_CCRCFAIL | SDIO_STA_CMDREND | \
+        SDIO_STA_CTIMEOUT;
+
+    if (wait == SDIO_WAIT_SHORT_RESPONSE) {
+        wait_flags = SDIO_CMD_WAITRESP_0;
+    } else if (wait == SDIO_WAIT_LONG_RESPONSE) {
+        wait_flags = SDIO_CMD_WAITRESP;
+    }
+
+    // Step 1: Send command and argument
+    // CMD and State Machine enabled.
+    // Wait for response like specified by wait_flags.
+    uint32_t cmdreg = (cmd & 0x3F) | wait_flags | SDIO_CMD_CPSMEN;
+    sdio->ARG = argument;
+    MODIFY_REG(sdio->CMD, CMD_CLEAR_MASK, cmdreg);
+
+    // Step 2: Wait until response
+    // wait for a timeout (max. SDIO_MAX_TIMEOUT) in msec.
+    // 8 cycles is the instruction cycles for the loop below.
+    uint32_t sta;
+    uint32_t count = SDIO_MAX_TIMEOUT * (SystemCoreClock / 8U / 1000U);
+    do {
+        if (count-- == 0) {
+            return SDIO_TIMEOUT;
+        }
+        sta = sdio->STA;
+    } while ((sta & sta_flags) == 0 || (sta & SDIO_STA_CMDACT) != 0);
+
+    sdio->ICR = SDIO_CMD_FLAGS;
+
+    // Step 3: Store response_data and check for short and long responses
+    // timeout and crc.
+    if (response_data != NULL) {
+        if (wait == SDIO_WAIT_SHORT_RESPONSE) {
+            response_data[0] = (uint8_t) ((sdio->RESP1 >> 24) & 0xFF);
+            response_data[1] = (uint8_t) ((sdio->RESP1 >> 16) & 0xFF);
+            response_data[2] = (uint8_t) ((sdio->RESP1 >> 8) & 0xFF);
+            response_data[3] = (uint8_t) ((sdio->RESP1) & 0xFF);
+            *response_data_len = 4;
+        } else if (wait == SDIO_WAIT_LONG_RESPONSE) {
+            // TODO Inverse?
+            response_data[0] =  (uint8_t) ((sdio->RESP1 >> 24) & 0xFF);
+            response_data[1] =  (uint8_t) ((sdio->RESP1 >> 16) & 0xFF);
+            response_data[2] =  (uint8_t) ((sdio->RESP1 >> 8) & 0xFF);
+            response_data[3] =  (uint8_t) ((sdio->RESP1) & 0xFF);
+            response_data[4] =  (uint8_t) ((sdio->RESP2 >> 24) & 0xFF);
+            response_data[5] =  (uint8_t) ((sdio->RESP2 >> 16) & 0xFF);
+            response_data[6] =  (uint8_t) ((sdio->RESP2 >> 8) & 0xFF);
+            response_data[7] =  (uint8_t) ((sdio->RESP2) & 0xFF);
+            response_data[8] =  (uint8_t) ((sdio->RESP3 >> 24) & 0xFF);
+            response_data[9] =  (uint8_t) ((sdio->RESP3 >> 16) & 0xFF);
+            response_data[10] = (uint8_t) ((sdio->RESP3 >> 8) & 0xFF);
+            response_data[11] = (uint8_t) ((sdio->RESP3) & 0xFF);
+            response_data[12] = (uint8_t) ((sdio->RESP4 >> 24) & 0xFF);
+            response_data[13] = (uint8_t) ((sdio->RESP4 >> 16) & 0xFF);
+            response_data[14] = (uint8_t) ((sdio->RESP4 >> 8) & 0xFF);
+            response_data[15] = (uint8_t) ((sdio->RESP4) & 0xFF);
+            *response_data_len = 16;
+        }
+    }
+
+    if (wait != SDIO_WAIT_NO_RESPONSE) {
+        // CTIMEOUT and CCRCFAIL check only for short or long responses.
+        if (sta & SDIO_STA_CTIMEOUT) {
+            return SDIO_CMD_RESPONSE_TIMEOUT;
+        }
+        if (sta & SDIO_STA_CCRCFAIL) {
+            return SDIO_CRC_FAIL;
+        }
+    }
+
+    // Step 4: For a short response check the response cmd field, too.
+    if (wait == SDIO_WAIT_SHORT_RESPONSE) {
+        if (sdio->RESPCMD != cmd) {
+            return SDIO_WRONG_CMD_RESPONSE;
+        }
+    }
+
+    return SDIO_OK;
+}
+
+uint32_t
+sdio_get_dctrl_blocksize(uint32_t value)
+{
+    switch(value) {
+        case 1:     return 0U;
+        case 2:     return SDIO_DCTRL_DBLOCKSIZE_0;
+        case 4:     return SDIO_DCTRL_DBLOCKSIZE_1;
+        case 8:     return SDIO_DCTRL_DBLOCKSIZE_1|SDIO_DCTRL_DBLOCKSIZE_0;
+        case 16:    return SDIO_DCTRL_DBLOCKSIZE_2;
+        case 32:    return SDIO_DCTRL_DBLOCKSIZE_2|SDIO_DCTRL_DBLOCKSIZE_0;
+        case 64:    return SDIO_DCTRL_DBLOCKSIZE_2|SDIO_DCTRL_DBLOCKSIZE_1;
+        case 128:   return SDIO_DCTRL_DBLOCKSIZE_2|SDIO_DCTRL_DBLOCKSIZE_1| \
+            SDIO_DCTRL_DBLOCKSIZE_0;
+        case 256:   return SDIO_DCTRL_DBLOCKSIZE_3;
+        case 512:   return SDIO_DCTRL_DBLOCKSIZE_3|SDIO_DCTRL_DBLOCKSIZE_0;
+        case 1024:  return SDIO_DCTRL_DBLOCKSIZE_3|SDIO_DCTRL_DBLOCKSIZE_1;
+        case 2048:  return SDIO_DCTRL_DBLOCKSIZE_3|SDIO_DCTRL_DBLOCKSIZE_1| \
+            SDIO_DCTRL_DBLOCKSIZE_0;
+        case 4096:  return SDIO_DCTRL_DBLOCKSIZE_3|SDIO_DCTRL_DBLOCKSIZE_2;
+        case 8192:  return SDIO_DCTRL_DBLOCKSIZE_3|SDIO_DCTRL_DBLOCKSIZE_2| \
+            SDIO_DCTRL_DBLOCKSIZE_0;
+        case 16384: return SDIO_DCTRL_DBLOCKSIZE_3|SDIO_DCTRL_DBLOCKSIZE_2| \
+            SDIO_DCTRL_DBLOCKSIZE_1;
+    }
+    return SDIO_DCTRL_DBLOCKSIZE_Msk;
+}
+
+uint8_t
+sdio_prepare_data_transfer(struct sdio_config sdio_config, uint8_t read,
+    uint32_t numblocks, uint32_t blocksize, uint32_t timeout)
+{
+    SDIO_TypeDef *sdio = sdio_config.sdio;
+    uint32_t dctrl_blocksize = sdio_get_dctrl_blocksize(blocksize);
+    uint32_t reg = dctrl_blocksize | ((read > 0) ?
+        SDIO_DCTRL_DTDIR : 0U) | SDIO_DCTRL_DTEN;
+
+    if (dctrl_blocksize == SDIO_DCTRL_DBLOCKSIZE_Msk)
+        return SDIO_WRONG_BLOCKSIZE;
+
+    sdio->DCTRL = 0;
+    sdio->DTIMER = timeout;
+    sdio->DLEN = numblocks*blocksize;
+    MODIFY_REG(sdio->DCTRL, DCTRL_CLEAR_MASK, reg);
+    return SDIO_OK;
+}
+
+uint8_t
+sdio_read_data(struct sdio_config sdio_config, uint8_t *data,
+    uint32_t numblocks, uint32_t blocksize)
+{
+    // Read data by polling
+    SDIO_TypeDef *sdio = sdio_config.sdio;
+    uint32_t data_remaining = numblocks*blocksize;
+    uint8_t *buf = data;
+
+    if (data == NULL) {
+        return SDIO_NO_DATA_MEMORY;
+    }
+
+    while ((sdio->STA & (SDIO_STA_RXOVERR | SDIO_STA_DCRCFAIL |
+        SDIO_STA_DTIMEOUT | SDIO_STA_DATAEND)) == 0) {
+        if ((sdio->STA & SDIO_STA_RXDAVL) != 0) {
+            uint32_t tmp = sdio->FIFO;
+            for (uint8_t i=0; (i<4) && (data_remaining>0); i++) {
+                *buf = (uint8_t)(tmp & 0xFF);
+                buf++;
+                data_remaining--;
+                tmp >>= 8U;
+            }
+        }
+    }
+
+    uint32_t sta = sdio->STA;
+    sdio->ICR = SDIO_STATIC_FLAGS;
+
+    if (sta & SDIO_STA_DTIMEOUT) {
+        return SDIO_DATA_TIMEOUT;
+    }
+    if (sta & SDIO_STA_DCRCFAIL) {
+        return SDIO_CRC_FAIL;
+    }
+    if (sta & SDIO_STA_RXOVERR) {
+        return SDIO_READ_OVERRUN;
+    }
+
+    // Empty FIFO and clear flags again
+    while (((sdio->STA & SDIO_STA_RXDAVL) != 0) && (data_remaining > 0)) {
+        uint32_t tmp = sdio->FIFO;
+        for (uint8_t i=0; (i<4) && (data_remaining>0); i++) {
+            *buf = (uint8_t)(tmp & 0xFF);
+            buf++;
+            data_remaining--;
+            tmp >>= 8U;
+        }
+    }
+
+    sdio->ICR = SDIO_STATIC_FLAGS;
+    return SDIO_OK;
+}
+
+uint8_t
+sdio_write_data(struct sdio_config sdio_config, uint8_t *data,
+    uint32_t numblocks, uint32_t blocksize)
+{
+    // Write data by polling
+    SDIO_TypeDef *sdio = sdio_config.sdio;
+    uint32_t data_remaining = numblocks*blocksize;
+    uint8_t *buf = data;
+
+    if (data == NULL) {
+        return SDIO_NO_DATA_MEMORY;
+    }
+
+    while ((sdio->STA & (SDIO_STA_TXUNDERR | SDIO_STA_DCRCFAIL |
+        SDIO_STA_DTIMEOUT | SDIO_STA_DATAEND)) == 0) {
+        if ((sdio->STA & SDIO_STA_TXFIFOF) == 0) {
+            uint32_t tmp = 0;
+            for (uint8_t i=0; (i<4) && (data_remaining>0); i++) {
+                tmp |= ((uint32_t)(*buf) << (i<<3));
+                buf++;
+                data_remaining--;
+            }
+            sdio->FIFO = tmp;
+        }
+    }
+
+    uint32_t sta = sdio->STA;
+    sdio->ICR = SDIO_STATIC_FLAGS;
+
+    if (sta & SDIO_STA_DTIMEOUT) {
+        return SDIO_DATA_TIMEOUT;
+    }
+    if (sta & SDIO_STA_DCRCFAIL) {
+        return SDIO_CRC_FAIL;
+    }
+    if (sta & SDIO_STA_TXUNDERR) {
+        return SDIO_WRITE_UNDERRUN;
+    }
+
+    return SDIO_OK;
+}
+
+void
+sdio_send_cmd(struct sdio_config sdio_config, uint8_t cmd, uint32_t argument,
+    uint8_t wait)
+{
+    SDIO_TypeDef *sdio = sdio_config.sdio;
+
+    sdio->ARG = argument;
+    //CMD and State Machine enabled. Wait for response like specified.
+    uint32_t cmdreg = (cmd & 0x3F) | ((wait & 0xC0)) | SDIO_CMD_CPSMEN;
+
+    MODIFY_REG(sdio->CMD, CMD_CLEAR_MASK, cmdreg);
+}
+
+void
+sdio_set_speed(struct sdio_config sdio_config, uint32_t speed)
+{
+    SDIO_TypeDef *sdio = sdio_config.sdio;
+
+    uint8_t clkdiv = (SDIO_CLK_FREQ/speed)-2;
+
+    uint32_t sdio_confreg = SDIO_CLOCK_EDGE_RISING | \
+        SDIO_CLOCK_BYPASS_DISABLE | SDIO_CLOCK_POWER_SAVE_DISABLE | \
+        SDIO_BUS_WIDE_1B | SDIO_HARDWARE_FLOW_CONTROL_DISABLE | (clkdiv & 0xFF);
+    MODIFY_REG(sdio->CLKCR, CLKCR_CLEAR_MASK, sdio_confreg);
+}
diff --git a/src/stm32/sdio.h b/src/stm32/sdio.h
new file mode 100644
index 00000000..eaf6596f
--- /dev/null
+++ b/src/stm32/sdio.h
@@ -0,0 +1,25 @@
+#ifndef __STM32_SDIO_H
+#define __STM32_SDIO_H
+
+#include <stdint.h> // uint32_t
+
+struct sdio_config {
+    void *sdio;
+};
+struct sdio_config sdio_setup(uint32_t bus);
+void sdio_send_cmd(struct sdio_config sdio, uint8_t cmd, uint32_t argument
+                   , uint8_t wait);
+uint8_t sdio_send_command(struct sdio_config sdio_config, uint8_t cmd
+                          , uint32_t argument, uint8_t wait
+                          , uint8_t *response_data
+                          , uint8_t *response_data_len);
+uint8_t sdio_prepare_data_transfer(struct sdio_config sdio_config, uint8_t read
+                                   , uint32_t numblocks, uint32_t blocksize
+                                   , uint32_t timeout);
+uint8_t sdio_read_data(struct sdio_config sdio_config, uint8_t *data
+                       , uint32_t numblocks, uint32_t blocksize);
+uint8_t sdio_write_data(struct sdio_config sdio_config, uint8_t *data
+                        , uint32_t numblocks, uint32_t blocksize);
+void sdio_set_speed(struct sdio_config sdio_config, uint32_t speed);
+
+#endif // sdio.h
diff --git a/src/stm32/stm32f4.c b/src/stm32/stm32f4.c
index a2060d76..fd3eb0ba 100644
--- a/src/stm32/stm32f4.c
+++ b/src/stm32/stm32f4.c
@@ -138,8 +138,8 @@ enable_clock_stm32f446(void)
     while (!(PWR->CSR & PWR_CSR_ODSWRDY))
         ;
 
-    // Enable 48Mhz USB clock
-    if (CONFIG_USB) {
+    // Enable 48Mhz USB clock for USB or for SDIO
+    if (CONFIG_USB || CONFIG_HAVE_GPIO_SDIO) {
         uint32_t ref = (CONFIG_STM32_CLOCK_REF_INTERNAL
                         ? 16000000 : CONFIG_CLOCK_REF_FREQ);
         uint32_t plls_base = 2000000, plls_freq = FREQ_USB * 4;
@@ -153,6 +153,14 @@ enable_clock_stm32f446(void)
             ;
 
         RCC->DCKCFGR2 = RCC_DCKCFGR2_CK48MSEL;
+    } else {
+        // Reset value just in case the booloader modified the default value
+        RCC->DCKCFGR2 = 0;
+    }
+
+    // Set SDIO clk to PLL48CLK
+    if (CONFIG_HAVE_GPIO_SDIO) {
+        MODIFY_REG(RCC->DCKCFGR2, RCC_DCKCFGR2_SDIOSEL, 0);
     }
 #endif
 }