This is an automated email from the ASF dual-hosted git repository. xiaoxiang781216 pushed a commit to branch master in repository https://gitbox.apache.org/repos/asf/nuttx.git
commit 6de8c13d99aef1712634c5b31184979ec3278826 Author: Matteo Golin <[email protected]> AuthorDate: Mon Jul 13 17:36:24 2026 -0400 docs/raspberrypi-4b: Documentation about the SD card updated Updated the documentation to reflect the fix of buggy SD card behaviour at the sacrifice of slower performance. Signed-off-by: Matteo Golin <[email protected]> --- Documentation/platforms/arm64/bcm2711/index.rst | 24 +++++++++--------------- 1 file changed, 9 insertions(+), 15 deletions(-) diff --git a/Documentation/platforms/arm64/bcm2711/index.rst b/Documentation/platforms/arm64/bcm2711/index.rst index 98307f92caa..f6dcdc0b4c9 100644 --- a/Documentation/platforms/arm64/bcm2711/index.rst +++ b/Documentation/platforms/arm64/bcm2711/index.rst @@ -52,23 +52,17 @@ PCM No The EMMC2 peripheral connects to the microSD card slot on the :doc:`Raspberry Pi 4B </platforms/arm64/bcm2711/boards/raspberrypi-4b/index>`. Currently, it - has been tested using a 32GB Samsung microSD card and it has passed testing - with that. The only quirks are: + has passed testing with a few different uSD cards. The only quirks are: * No card insertion/removal interrupts work, so hotswapping isn't possible - * The :doc:`sdstress </applications/testing/sd_stress/index>` example works - unless the 'number of bytes' option is greater than 1023. I suspect this is - something to do with the FIFO depth being only 1023 bytes. - - However, the 64GB microSD card exhibits very strange behaviour. There are - often data CRC errors that prevent the boot filesystem from mounting. When - that somehow passes (intermittent), running ``ls`` on the filesystem repeatedly - sometimes causes certain files to disappear from the listing (they are not - deleted and appear again on next boot), or causes their filenames to be shown - in all caps. Writing to the card with ``echo`` often fails with data CRC - error or timeout, and then the card is buggy for the remainder of the - session. **It is not recommended to use 64GB cards with this implementation - for the time being.** + * Multi-block transfers must be restricted to 1 block at a time + (``CONFIG_MMCSD_MULTIBLOCK_LIMIT=1``), since the upper-half driver's method + of doing block setup is not immediately compatible with the BCM2711 EMMC + controller. Changes to the common-upperhalf would require extensive + testing, so this performance sacrifice is done for correct behaviour in the + short-term. + + All in all, be aware of issues with the SD card implementation. Supported Boards ================
