SDIO 4-Bit mode not working / DAT2 pulled down by ESP32

[olell]

Checklist

• Checked the issue tracker for similar issues to ensure this is not a duplicate.
• Provided a clear description of your suggestion.
• Included any relevant context or examples.

Issue or Suggestion Description

I'm currently porting esp-hosted to a custom board using a Rockchip RV1106G3 as hosting SoC and an ESP32-WROOM-32. In 1-Bit mode the SDIO communication between the ESP32 and the SoC works fine except the low data throughput. Currently I'm trying to switch to 4-Bit mode and encountered a problem with the DAT2 line.

The connections on the board are kept short, below 30mm, referencing to a solid ground plane and very similar length. I will attach a schematic diagram. According to the documentation I added 51kΩ pull up resistors on CMD & DAT0..3.

I detected the following behavior: All lines are working correctly in 4-Bit mode except DAT2, there I can measure a voltage of around 1.5V which stays permanently. Therefore I guess the initialization of the driver fails. I put both chips in reset state to test the influence of those chips on the line:

Chip in Reset	Voltage on DAT2
ESP32	3.3V
None	1.5V
RV1106G3	0V

from that I concluded, that the ESP32 is pulling the line down with a similar resistor as the SoC is pulling it up. My conclusion is that it might have something to do with the MTDI bootstrapping pin.

But I burned the efuses of the ESP32 so that the MTDI strapping pin is disabled:

Vdd fuses:
XPD_SDIO_REG (BLOCK0)                              read for XPD_SDIO_REG                              = True R/W (0b1)
XPD_SDIO_TIEH (BLOCK0)                             If XPD_SDIO_FORCE & XPD_SDIO_REG                   = 3.3V R/W (0b1)
XPD_SDIO_FORCE (BLOCK0)                            Ignore MTDI pin (GPIO12) for VDD_SDIO on reset     = True R/W (0b1)

Is there anything in the firmware or configuration that I might have overseen or can you maybe help me figure out why the ESP32 is pulling that line down. Or is it maybe something completely different. I'm beginning to despair because I can't find the error :/

# dmesg | grep -C4 esp32

[    5.713461] g_cdc gadget.0: CDC Composite Gadget, version: King Kamehameha Day 2008
[    5.713476] g_cdc gadget.0: g_cdc ready
[    5.728879] g_cdc gadget.0: suspend
[    5.857192] usb0: eth_open
[    5.966190] esp32_sdio: loading out-of-tree module taints kernel.
[    7.232466] gs_open: ttyGS0 (b05e7824,e6a6bf49)
[    7.241472] gs_close: ttyGS0 (b05e7824,e6a6bf49) ...
[    7.241550] gs_close: ttyGS0 (b05e7824,e6a6bf49) done!
[    7.243320] gs_open: ttyGS0 (2c3e93e5,99650811)
--
[    8.691486] mmc2: queuing unknown CIS tuple 0x01 [d9 01 ff] (3 bytes)
[    8.699221] mmc2: queuing unknown CIS tuple 0x1a [01 01 00 02 07] (5 bytes)
[    8.702663] mmc2: queuing unknown CIS tuple 0x1b [c1 41 30 30 ff ff ff ff] (8 bytes)
[    8.705858] mmc2: new SDIO card at address 0001
[    8.706685] esp32_sdio: esp_probe: ESP network device detected
[    8.707328] esp32_sdio: probe of mmc2:0001:1 failed with error -84
[    8.707811] esp32_sdio: probe of mmc2:0001:2 failed with error -22
[    9.844665] g_cdc gadget.0: packet filter 0e
[    9.844720] g_cdc gadget.0: ecm req21.43 v000e i0000 l0
[    9.903971] g_cdc gadget.0: packet filter 0e
[    9.904017] g_cdc gadget.0: ecm req21.43 v000e i0000 l0

# cat /sys/kernel/debug/mmc2/ios

clock:          25000000 Hz
vdd:            21 (3.3 ~ 3.4 V)
bus mode:       2 (push-pull)
chip select:    0 (don't care)
power mode:     2 (on)
bus width:      2 (4 bits)
timing spec:    0 (legacy)
signal voltage: 0 (3.30 V)
driver type:    0 (driver type B)

Hardware Setup / Schematic

The bright green lines are the SDIO lines

Linux device tree section

Everything actually works if I set the bus-width to 1. But the throughput is very poor then.

&sdio {
	status = "okay";
	bus-width = <4>;
	num-slots = <1>;
	clock-frequency = <50000000>;
	clock-freq-min-max = <200000 50000000>;
	supports-highspeed;
	supports-sdio;
	broken-cd;
	disable-wp;

	pinctrl-names = "default", "idle";
	pinctrl-0 = <&sdmmc1m1_cmd &sdmmc1m1_clk &sdmmc1m1_bus4>;
	pinctrl-1 = <&sdmmc1m1_idle_pins>;
};

The esp-hosted serial output after start

rst:0x1 (POWERON_RESET),boot:0x1b (SPI_FAST_FLASH_BOOT)
configsip: 0, SPIWP:0xee
clk_drv:0x00,q_drv:0x00,d_drv:0x00,cs0_drv:0x00,hd_drv:0x00,wp_drv:0x00
mode:DIO, clock div:2
load:0x3fff0030,len:6276
load:0x40078000,len:15716
load:0x40080400,len:4
ho 8 tail 4 room 4
load:0x40080404,len:3572
entry 0x4008062c
I (30) boot: ESP-IDF 67c1de1e-dirty 2nd stage bootloader
I (30) boot: compile time Apr 21 2025 20:09:59
I (30) boot: Multicore bootloader
I (32) boot: chip revision: v1.0
I (35) boot.esp32: SPI Speed      : 40MHz
I (39) boot.esp32: SPI Mode       : DIO
I (42) boot.esp32: SPI Flash Size : 4MB
I (46) boot: Enabling RNG early entropy source...
I (50) boot: Partition Table:
I (53) boot: ## Label            Usage          Type ST Offset   Length
I (59) boot:  0 nvs              WiFi data        01 02 00009000 00004000
I (66) boot:  1 otadata          OTA data         01 00 0000d000 00002000
I (72) boot:  2 phy_init         RF data          01 01 0000f000 00001000
I (79) boot:  3 factory          factory app      00 00 00010000 00100000
I (85) boot:  4 ota_0            OTA app          00 10 00110000 00100000
I (92) boot:  5 ota_1            OTA app          00 11 00210000 00100000
I (98) boot: End of partition table
I (101) boot: Defaulting to factory image
I (105) esp_image: segment 0: paddr=00010020 vaddr=3f400020 size=19e18h (106008) map
I (149) esp_image: segment 1: paddr=00029e40 vaddr=3ff80000 size=0001ch (    28) load
I (149) esp_image: segment 2: paddr=00029e64 vaddr=3ffbdb60 size=05d60h ( 23904) load
I (162) esp_image: segment 3: paddr=0002fbcc vaddr=40080000 size=0044ch (  1100) load
I (162) esp_image: segment 4: paddr=00030020 vaddr=400d0020 size=92b9ch (600988) map
I (373) esp_image: segment 5: paddr=000c2bc4 vaddr=4008044c size=1e7ech (124908) load
I (437) boot: Loaded app from partition at offset 0x10000
I (437) boot: Disabling RNG early entropy source...
I (448) cpu_start: Multicore app
I (456) cpu_start: Pro cpu start user code
I (456) cpu_start: cpu freq: 240000000 Hz
I (456) app_init: Application information:
I (456) app_init: Project name:     network_adapter
I (461) app_init: App version:      NG-1.0.3.0.12
I (465) app_init: Compile time:     Apr 21 2025 20:10:08
I (470) app_init: ELF file SHA256:  b05617139...
I (474) app_init: ESP-IDF:          67c1de1e-dirty
I (479) efuse_init: Min chip rev:     v0.0
I (483) efuse_init: Max chip rev:     v3.99
I (487) efuse_init: Chip rev:         v1.0
I (491) heap_init: Initializing. RAM available for dynamic allocation:
I (497) heap_init: At 3FFAFF10 len 000000F0 (0 KiB): DRAM
I (502) heap_init: At 3FFB7CD8 len 00000328 (0 KiB): DRAM
I (507) heap_init: At 3FFB9A20 len 00004108 (16 KiB): DRAM
I (512) heap_init: At 3FFD2490 len 0000DB70 (54 KiB): DRAM
I (518) heap_init: At 3FFE0440 len 00003AE0 (14 KiB): D/IRAM
I (523) heap_init: At 3FFE4350 len 0001BCB0 (111 KiB): D/IRAM
I (528) heap_init: At 4009EC38 len 000013C8 (4 KiB): IRAM
I (535) spi_flash: detected chip: generic
I (537) spi_flash: flash io: dio
I (541) coexist: coex firmware version: 49a8cdc
I (546) main_task: Started on CPU0
I (549) main_task: Calling app_main()
I (551) stats: *********************************************************************
I (558) stats:                 ESP-Hosted Firmware version :: NG-1.0.3.0.12
I (567) stats:                 Transport used :: SDIO only
I (574) stats: *********************************************************************
I (582) FW_MAIN: Supported features are:
I (585) FW_MAIN: - WLAN over SDIO
I (588) FW_BT: - BT/BLE
I (591) FW_BT:    - HCI Over SDIO
I (594) FW_BT:    - BT/BLE dual mode
I (597) FW_MAIN: Capabilities: 0x1d
I (610) wifi:wifi driver task: 3ffd5ea4, prio:23, stack:6656, core=1
I (613) wifi:wifi firmware version: 87d6f3c
I (613) wifi:wifi certification version: v7.0
I (614) wifi:config NVS flash: disabled
I (618) wifi:config nano formatting: disabled
I (622) wifi:Init data frame dynamic rx buffer num: 40
I (627) wifi:Init static rx mgmt buffer num: 5
I (631) wifi:Init management short buffer num: 32
I (635) wifi:Init dynamic tx buffer num: 64
I (639) wifi:Init static rx buffer size: 1600
I (644) wifi:Init static rx buffer num: 16
I (647) wifi:Init dynamic rx buffer num: 40
I (652) wifi_init: rx ba win: 32
I (654) wifi_init: accept mbox: 6
I (657) wifi_init: tcpip mbox: 32
I (660) wifi_init: udp mbox: 6
I (663) wifi_init: tcp mbox: 6
I (666) wifi_init: tcp tx win: 5760
I (669) wifi_init: tcp rx win: 5760
I (672) wifi_init: tcp mss: 1440
I (675) wifi_init: WiFi IRAM OP enabled
I (679) wifi_init: WiFi RX IRAM OP enabled
I (683) BTDM_INIT: BT controller compile version [20b459d]
I (689) BTDM_INIT: Bluetooth MAC: e0:e2:e6:d6:42:1e
I (695) phy_init: phy_version 4840,02e0d70,Sep  2 2024,19:39:07
I (1075) FW_MAIN: ESP Bluetooth MAC addr: e0-e2-e6-d6-42-1e

I (1075) FW_SDIO_SLAVE: Using SDIO interface
I (2077) gpio: GPIO[22]| InputEn: 0| OutputEn: 1| OpenDrain: 0| Pullup: 0| Pulldown: 0| Intr:0
I (491) heap_init: Initializing. RAM available for dynamic allocation:
I (497) heap_init: At 3FFAFF10 len 000000F0 (0 KiB): DRAM
I (502) heap_init: At 3FFB7CD8 len 00000328 (0 KiB): DRAM
I (507) heap_init: At 3FFB9A20 len 00004108 (16 KiB): DRAM
I (512) heap_init: At 3FFD2490 len 0000DB70 (54 KiB): DRAM
I (518) heap_init: At 3FFE0440 len 00003AE0 (14 KiB): D/IRAM
I (523) heap_init: At 3FFE4350 len 0001BCB0 (111 KiB): D/IRAM
I (528) heap_init: At 4009EC38 len 000013C8 (4 KiB): IRAM
I (535) spi_flash: detected chip: generic
I (537) spi_flash: flash io: dio
I (541) coexist: coex firmware version: 49a8cdc
I (546) main_task: Started on CPU0
I (549) main_task: Calling app_main()
I (551) stats: *********************************************************************
I (558) stats:                 ESP-Hosted Firmware version :: NG-1.0.3.0.12
I (567) stats:                 Transport used :: SDIO only
I (574) stats: *********************************************************************
I (582) FW_MAIN: Supported features are:
I (585) FW_MAIN: - WLAN over SDIO
I (588) FW_BT: - BT/BLE
I (591) FW_BT:    - HCI Over SDIO
I (594) FW_BT:    - BT/BLE dual mode
I (597) FW_MAIN: Capabilities: 0x1d
I (610) wifi:wifi driver task: 3ffd5ea4, prio:23, stack:6656, core=1
I (613) wifi:wifi firmware version: 87d6f3c
I (613) wifi:wifi certification version: v7.0
I (614) wifi:config NVS flash: disabled
I (618) wifi:config nano formatting: disabled
I (622) wifi:Init data frame dynamic rx buffer num: 40
I (627) wifi:Init static rx mgmt buffer num: 5
I (631) wifi:Init management short buffer num: 32
I (635) wifi:Init dynamic tx buffer num: 64
I (639) wifi:Init static rx buffer size: 1600
I (644) wifi:Init static rx buffer num: 16
I (647) wifi:Init dynamic rx buffer num: 40
I (652) wifi_init: rx ba win: 32
I (654) wifi_init: accept mbox: 6
I (657) wifi_init: tcpip mbox: 32
I (660) wifi_init: udp mbox: 6
I (663) wifi_init: tcp mbox: 6
I (666) wifi_init: tcp tx win: 5760
I (669) wifi_init: tcp rx win: 5760
I (672) wifi_init: tcp mss: 1440
I (675) wifi_init: WiFi IRAM OP enabled
I (679) wifi_init: WiFi RX IRAM OP enabled
I (683) BTDM_INIT: BT controller compile version [20b459d]
I (689) BTDM_INIT: Bluetooth MAC: e0:e2:e6:d6:42:1e
I (695) phy_init: phy_version 4840,02e0d70,Sep  2 2024,19:39:07
I (1075) FW_MAIN: ESP Bluetooth MAC addr: e0-e2-e6-d6-42-1e

I (1075) FW_SDIO_SLAVE: Using SDIO interface
I (2077) gpio: GPIO[22]| InputEn: 0| OutputEn: 1| OpenDrain: 0| Pullup: 0| Pulldown: 0| Intr:0

There is nothing more is happening after this line was printed

[SohKamYung-Espressif]

@olell I believe you have burned the efuses following this instructions, https://docs.espressif.com/projects/esp-idf/en/latest/esp32/api-reference/peripherals/sd_pullup_requirements.html#strapping-conflicts-dat2 ?

In your table, when the RV1106G3 is in reset, is the ESP32 running? And the voltage you see on DAT2 is 0V?

What is the voltage when both chips are in reset state? It should be 3.3V.

@kapilkedawat Can you take a look?

[olell]

Hello,

Yes I burned the efuses according to the instructions you've linked.

Yes, when the SoC was in reset state the ESP was running and vice versa for the ESP in reset state.

Currently I'm not at my workplace and don't have access to the board, but I am pretty sure that the DAT2 line has 3.3V if both chips are in reset state. I will measure it later when I have access to it.

[olell]

I just measured a few more voltages and states. This is my result. To check if there is something wrong with the hardware I also wrote a little script which "blinks" this particular pin to see if it does that and that works quite fine it blinks between 0 and 3.3V.

Here are the results of my measurements:

	No external pull-up	External pull-up 47kΩ
No chip in reset	3.30V	3.3V
RV1106 in reset	1.72V	2.16V
ESP32 in reset	3.30V	3.3V
Both in reset	0V	1.32V

In the setting where both are in reset, if i release the reset state of the ESP32 it shortly drops down to 0.7V until the firmware has started. Maybe the boot loader does something with this pin?

Repeated using pre compiled esp-hosted firmware. Linux host unchanged:

	External pull-up 47kΩ
No chip in reset	3.3V
RV1106 in reset	1.32V
ESP32 in reset	3.3V
Both in reset	1.32V

Difference between those firmware: probably internal pull-up resistor enabled and not

Then I cutted the PCB trace, pull up resistor is on ESP32 side:

External pull-up 47kΩ	ESP32 Side	SoC Side
IC in reset	3.3V	0V
IC not in reset	3.3V	3.3V
IC booting	1V	N/A

[SohKamYung-Espressif]

Can you provide information on the specific ESP32 module you are using?

There is a list of ESP32 related SDIO compatibility issues which is Soc or Module dependent https://docs.espressif.com/projects/esp-idf/en/latest/esp32/api-reference/peripherals/sd_pullup_requirements.html#overview-of-compatibility

[olell]

Yes sure, I am using an ESP32-WROOM-32 module as slave which contains an ESP32-D0WDQ6 and probably a GD25Q32 Flash chip (didn't decap the module to check).

I've read the linked info multiple times and all I figured out was to burn the mentioned MTDI related efuses using the set_flash_voltage 3.3V command.

[mantriyogesh]

Hello @olell ,

I have gone through the thread again, below are some understanding and observations:

1. I also understand that you are unsure of external pull-up value. >10KOhm is okay. We use 51KOhm. In case you wish to get our schematic, please feel free to mail me.

2. You already have done the following:

components/esptool_py/esptool/espefuse.py set_flash_voltage 3.3V

and external pull-ups on CMD, DAT0, DAT1, DAT2 , DAT3 with >=10KOhm.

3. Are you able to flash the esp32-wroom-32 reliably? if so, no need to care for 'download mode not working'

4. In normal working conditions, you observe that you get 3.3V on all the SDIO pins.

If you reached point (3), everything is fine, except one thing, which is SDIO interference.

5. From logs at host,

[    8.707328] esp32_sdio: probe of mmc2:0001:1 failed with error -84

#define EILSEQ 84 /* Illegal byte sequence */

• Possible causes:

  • 5.1 Incorrect SDIO pull-up resistors (esp. on CMD and DAT lines).

  • 5.2 Clock too high or too low.

  • 5.3 Signal integrity problems.

  • 5.4 Mismatch in SDIO firmware settings on the ESP32 side. (Timing tuning explained in porting guide)

  • 5.5 SoC is sufficiently powered and 3.3V on SDIO lines is consistent (in normal SDIO operation)

6. From logs

There is nothing more is happening after this line was printed

Here ESP SDIO slave is waiting to get first communication from host to slave, which would print at ESP:

esp-hosted/esp_hosted_ng/esp/esp_driver/network_adapter/main/app_main.c

Line 611 in d9e607d

ESP_EARLY_LOGI(TAG, "Start Data Path");

which is triggered from host esp driver at

esp-hosted/esp_hosted_ng/host/sdio/esp_sdio.c

Line 762 in d9e607d

generate_slave_intr(context, BIT(ESP_OPEN_DATA_PATH));

6.1 From the logs, I can see the device was tried to probe. Which exact code leg did the Linux driver started to return -84 would be the key

esp-hosted/esp_hosted_fg/host/linux/host_driver/esp32/sdio/esp_sdio.c

Line 722 in d9e607d

static int esp_probe(struct sdio_func *func,