machine.SDCard() fails with OSError: 16 on ESP32-P4 (ESP32_GENERIC_P4 firmware) (Micropython)

[EgeOnderX]

ESP32-P4 machine.SDCard() returns OSError: 16 (works on ESP32 DevKit v2)

Description

I am trying to use an SD card with machine.SDCard() on an ESP32-P4 board running MicroPython.
However, mounting the SD card always fails with OSError: 16.
NOTE : I used the built-in SD card reader of P4
The same SD card and the exact same code work correctly on a regular ESP32 DevKit v2 board.

Because of this, I suspect SDCard support might not yet be fully implemented or configured for the ESP32-P4 port.

Hardware

Board: ESP32-P4 WiFi6 PoE ETH board
MCU: ESP32-P4
SD Card: 16 GB
Filesystem: FAT32

Comparison board (working):

• ESP32 DevKit v2

Firmware

MicroPython firmware:
ESP32_GENERIC_P4

Test Code

import machine
import os

try:
    sd = machine.SDCard(slot=0)
    os.mount(sd, "/sd")
    print("SD OK")
    print(os.listdir("/sd"))

except Exception as e:
    print("SD ERROR:", e)

Result on ESP32-P4

SD ERROR: 16

Testing multiple slots:

Trying slot: 0 -> error: 16
Trying slot: 1 -> error: 16
Trying slot: 2 -> SPI pin values required
Trying slot: 3 -> error: 16

Result on ESP32 DevKit v2

The same SD card (But connected with SPI) and code work correctly and the filesystem mounts without problems.

Questions

• Does the ESP32_GENERIC_P4 firmware include SDMMC support?
• Are specific pin mappings required for ESP32-P4 boards?

Additional Question

Is there any workaround to make SD cards work on ESP32-P4 with MicroPython (for example SPI mode or custom pin configuration)?

If not, what would be the correct way to enable SD support for ESP32-P4?

Any guidance would be greatly appreciated.

(Improved using ChatGPT)

Guidelines

• I have read and understood this category's guidelines before making this post.

[Ayouub-aj]

OSError: 16 is EBUSY — the driver is trying to claim a peripheral that's already in use or not properly released. On the ESP32-P4 this almost certainly means SDMMC support in ESP32_GENERIC_P4 firmware is incomplete or misconfigured for the built-in slot.
A few things to try:

1. Force SPI mode with explicit pins (most reliable workaround right now):

import machine, os sd = machine.SDCard( slot=2, # SPI mode miso=13, mosi=11, sck=12, cs=10 ) os.mount(sd, "/sd") print(os.listdir("/sd"))

Adjust the pin numbers to match your board's pinout — check your ESP32-P4 board schematic for the SD SPI pins.

2. Add a soft reset before initialising:

import machine, os, time machine.soft_reset() # or time.sleep(1) before SDCard()

Sometimes the SDMMC peripheral isn't released cleanly on boot.

4. Check firmware maturity — ESP32-P4 MicroPython support is still relatively new. Check the [MicroPython ESP32 port changelog] (https://github.com/micropython/micropython/blob/master/CHANGELOG.md) and consider building from the latest master branch, as P4 SDMMC fixes may not be in a stable release yet.

5. Report upstream — If SPI mode works but native SDMMC doesn't, this is worth filing as a bug at [micropython/micropython] with (https://github.com/micropython/micropython/issues) test results. The P4 port is actively being developed and your findings would be valuable.

[EgeOnderX]

Hi, thanks for taking the time to respond and for the suggestions 🙂

Unfortunately SPI mode is not really an option in this case. According to the board schematic and the ESP32-P4 documentation, the onboard SD slot is wired directly to the SDMMC peripheral, not SPI. The SPI pins are not routed to the socket, so forcing machine.SDCard(slot=2) with SPI pins is effectively impossible on this board.

I also tried the soft reset suggestion earlier (and a few variations around it). Sadly it didn't change the behavior — still the same OSError: 16. I promise I didn’t miss that easy one 🙂.

Another thing I noticed while testing: the 1.28 builds seem a bit unstable right now. The 1.27 release is currently the most reliable for ESP32-P4 in my testing, but the SD card issue occurs there as well.

So at the moment it seems the situation is basically:

• Built-in slot uses SDMMC only
• SPI workaround cannot be used due to hardware wiring
• soft_reset() does not resolve it
• Both 1.27 (stable) and 1.28 builds show the same issue

Thanks again for the help 👍

[Ayouub-aj]

Since you’ve confirmed the hardware is strictly wired for SDMMC and the standard software resets aren't clearing the OSError: 16, it appears we’ve reached the limits of current firmware support. The 'Busy' error on the P4 usually points to a pending fix needed in the MicroPython C-source for this specific port’s IOMUX or power domain initialization.

Since we’ve narrowed it down to a driver/firmware limitation rather than a configuration error, could you please mark this as 'Answered' so it can help others troubleshooting the P4 and It will also help the community track which peripherals still need active development.

Best of luck with the rest of your project 👍

[EgeOnderX]

Thanks for the help and for taking the time to look into the issue 👍

I appreciate the clarification and the suggestions. Hopefully SDMMC support for the P4 will improve in future firmware updates.

Thanks again for your help!

[projectgus]

FWIW there is a solution to this, see https://github.com/orgs/community/discussions/188887#discussioncomment-17503642

[khizra-javed]

It looks like the issue may be related to the ESP32-P4 MicroPython port rather than the SD card itself. Since the same card and code work correctly on an ESP32 DevKit v2, the OSError: 16 suggests that the SD interface is either not initialized properly or not yet fully supported on the current ESP32_GENERIC_P4 firmware. I would recommend testing with the latest nightly build and checking whether the board's built-in SD card reader is currently supported by the ESP32-P4 MicroPython port. This seems more like a firmware or driver limitation than a hardware fault.

[projectgus]

ESP32-P4 has an internal voltage regulator (LDO) architecture that requires enabling the correct LDO channel before it's possible to access the SD card.

In MicroPython nightly builds (and the next release v1.29) the correct LDO channel is enabled automatically, so SDCard should work normally. (Details).

In MicroPython v1.28 we have support for the ESP32-P4 LDO but it's not automatically integrated with machine.SDCard() yet. You can work around it on MicroPython v1.28 like this:

import machine, esp32, vfs
ldo = esp32.LDO(4, 3300)

# then use the SDCard normally, i.e.
vfs.mount(machine.SDCard(), "/sd")

(The workaround may need to be removed after you update to MicroPython v1.29.)

[EgeOnderX]

Please allow about half an hour for me to test this and reply. Since I had previously concluded that the SD card was not functioning under MicroPython, I had switched the board to a custom C firmware. I will need to re-flash MicroPython to properly run this test and verify the results. (As a side note, I have not yet succeeded in getting the SD card to work under the C environment either, but that is a separate issue.)

[EgeOnderX]

I finally got it working! It turns out the issue was a combination of the MicroPython version and a sneaky hardware power management design on this Waveshare board.

The board uses an Active-LOW MOSFET on GPIO 45 to control power to the SD card slot. If you don't explicitly pull it LOW, the card gets zero power. Also, using MicroPython v1.29 was the key, as older versions didn't map the P4's SDMMC pins properly.

Here is the exact MicroPython code that fixed it for me:

from machine import Pin, SDCard
import time
import os

# 1. Power up the SD Card slot (Active-LOW MOSFET on Waveshare P4)
sd_pwr = Pin(45, Pin.OUT)
sd_pwr.value(0)  # Pulling LOW sends 3.3V to the card
time.sleep_ms(500)  # Wait for voltage to stabilize

# 2. Initialize and Mount
try:
    sd = SDCard()  # Auto-detects the built-in slot on v1.29
    os.mount(sd, '/sd')
    print("Success! Directory contents:", os.listdir('/sd'))
except Exception as e:
    print("Mount failed:", e)

So, in conclusion, it seems the SD card issues have been officially fixed in v1.29! Since the exact same code failed to work on v1.28, the firmware-level updates in the latest version definitely made the difference.

[projectgus]

Thanks for updating @EgeOnderX. For the benefit of anyone who stumbles across this in the future: I've raised a MicroPython issue about handling GPIO45 power automatically in a Waveshare-specific board profile: micropython/micropython#19439

[jlsilicon]

Problem is solved