SDHC SPI Send Command

RE: SDHC SPI Send Command

dario.sortino — Mon, 17 Apr 2023 15:03:22 GMT

Turns out that there was a gpio forwarder for the uart(s) interfaces that were still active, and were causing troubles because pin-clashing.

I solved by adding the following line to my cpuapp_overlay.

/delete-node/ &{/nrf-gpio-forwarder/uart};

RE: SDHC SPI Send Command

Simonr — Fri, 14 Apr 2023 06:24:01 GMT

If you set CONFIG_BOOTLOADER_MCUBOOT=y in the prj.conf of you application, the mcuboot will get added as a child image and you will have a multi image build (the partition manager will also get enabled and used to partition the flash).

Best regards,

Simon

RE: SDHC SPI Send Command

dario.sortino — Thu, 13 Apr 2023 09:38:53 GMT

I just tried to enable CONFIG_BT in the fat_fs sample and I see the same erroneous behaviour.

About MCUBoot, can you provide to me a reference on how to build with MCUBoot enabled?

RE: SDHC SPI Send Command

Simonr — Wed, 12 Apr 2023 10:32:20 GMT

By multicore, do you mean the multicore sample available here? If yes, do you see the same issue if you let CONFIG_BT enable the net core in the fat_fs sample?

Try to build with MCUBoot enabled instead of both cores, and see if that still works (MCUBoot enables the partition manager).

Best regards,

Simon

RE: SDHC SPI Send Command

dario.sortino — Wed, 05 Apr 2023 08:33:10 GMT

I made some tests:
- I run the fat_fs example in my custom board, using the above _cpuapp.overlay. It works fine

- I use the fat_fs example in my custom board in the multicore example running in the APPCORE. I wrote an empty NETCORE fw. Attached you will find the project. Card is not recognized

This seems absurd to me and it seems that I don't have any margin of making it work in a multicore application

devzone.nordicsemi.com/.../testmulticore.zip

RE: SDHC SPI Send Command

Turbo J — Wed, 22 Mar 2023 08:58:21 GMT

That is a correct response "illegal command + idle state" from an SD card to a CMD5. SD cards don't typically support CMD5.

I've seen cards that need a power cycle when they receive unsupported commands.

RE: SDHC SPI Send Command

dario.sortino — Wed, 22 Mar 2023 08:06:33 GMT

Hi simonr,

for sure. Right now I'm building for nrf5340dk_nrf5340_cpuapp, with the same structure of the multicore sample, using the following overlay for appcore and netcore

devzone.nordicsemi.com/.../0116.nrf5340dk_5F00_nrf5340_5F00_cpunet.overlay devzone.nordicsemi.com/.../6283.nrf5340dk_5F00_nrf5340_5F00_cpuapp.overlay

When I opened the ticket, I was building only for the appcore using the following board devicetree

devzone.nordicsemi.com/.../board.dts

RE: SDHC SPI Send Command

Simonr — Wed, 22 Mar 2023 07:30:57 GMT

Hi can you share some details on what changes you made when changing the sample to also build for the NET core? Can you upload this overlay file so we can take a look?

Best regards,

Simon

RE: SDHC SPI Send Command

dario.sortino — Mon, 20 Mar 2023 13:08:12 GMT

I can imagine it's not an easy and fast task, don't worry.

- I cannot report the SD size anymore

- The SD size never shows in the devicetree.

The card is powered at 3.336V (measured with a multimeter)

One thing I want to mention is that when I was able to retrieve the SD size, I was using a custom dts for my board, that was building only for the APPCORE. Due to the need of building also for the NETCORE, I started using it to build for nrf5340dk and used an overlay file to define my hardware.

RE: SDHC SPI Send Command

Simonr — Mon, 20 Mar 2023 12:30:14 GMT

The only reasons for error -134 (ENOTSUP) should be that the SD card is not recognized (likely due to incorrect setup) or that the SD card you're using is not supported which I don't think is the case with the SD cards you've mentioned using. Unfortunately I will need some time to set up something like this to test on my end, so I can't guarantee getting back to you with a tested solution any time soon. Some things you can check on your end:

Is the SD card reader powered sufficiently, maybe it is set to too low a voltage and thus browns out after reporting the SD size for example?
Is the SD size reported by the actual SD card, or is the size perhaps reported by the device tree?

Without knowing what bytes we should expect return data on it's very hard to gather something from these scope traces, as it only seems to be empty data (FF) from the SD card reader except that 05 value.

Best regards,

Simon

RE: SDHC SPI Send Command

dario.sortino — Fri, 17 Mar 2023 10:45:06 GMT

Same return error also trying with a SanDisk Ultra microSDXC 128Gb.

I don't think it's a problem of the SD I'm using, since as the original post says, I was able at least to retrieve the SD size

RE: SDHC SPI Send Command

dario.sortino — Fri, 17 Mar 2023 08:58:55 GMT

Some investigation led me to understand where it arises the -134 error.

It happens when the system checks the sd card type (either SDIO or SDMMC). The function sd_test_sdio returns -134.

Inside of that, we try to send a sdhc command with opcode 5 and response type 85.

This is the scope trace of that transaction is the following

RE: SDHC SPI Send Command

dario.sortino — Thu, 16 Mar 2023 16:11:09 GMT

This is instead sdhc_request when the SD card is in the slot. I honestly have no idea of what's happening.

RE: SDHC SPI Send Command

dario.sortino — Thu, 16 Mar 2023 14:56:27 GMT

Here you have a scope trace of the SPI pins (SD card not placed) when I call sdhc_request

I know that MISO is stale because there is no SD (I'll try get a scope trace also with the SD in), but how does the rest look?

RE: SDHC SPI Send Command

dario.sortino — Thu, 16 Mar 2023 13:51:44 GMT

I was debugging where the ENOTSUP -134 error is originating, and I can see that it comes from

ret = sdhc_request(card->sdhc, &cmd, NULL); (sd.c)

Apparently, the cmd structure will contain information about the response of the SD card. In my case, the field response of the structure sdhc_command cmd is 0, hence it returns -134.

At this point I need to check if I see something from the logic analyzer when the sdhc_request is sent. I'll post updates soon

RE: SDHC SPI Send Command

dario.sortino — Thu, 16 Mar 2023 13:07:44 GMT

As you can see from the snippet above, i directly print the return value of disk_access_init.

If I keep in the DTS the following part

		mmc {
			compatible = "zephyr,sdmmc-disk";
			status = "okay";
		};

Then it returns ENOTSUP (-134)

I tried different combinations of possible configurations based on what I can find online, but it seems to be something very unclear.

At the moment it seems that it's not a problem of the SD itself, neither of the SPI bus, but an hardware configuration that somehow is not working.

By receiving -22 (EINOTVAL), I think that I miss the link between the const char disk name, and the device registered in the dts/prj.conf

RE: SDHC SPI Send Command

Simonr — Thu, 16 Mar 2023 12:05:50 GMT

Error message -22 means that there is an invalid argument in that function. The disk_access_init() only input parameter should be the Disk name as a const char, so are you sure that is the function returning this error?

Best regards,

Simon

RE: SDHC SPI Send Command

dario.sortino — Wed, 15 Mar 2023 11:17:44 GMT

I call the following function in my main

void mount_disk(){

	do {
		static const char *disk_pdrv = "SD";
		uint64_t memory_size_mb;
		uint32_t block_count;
		uint32_t block_size;

		if (disk_access_init(disk_pdrv) != 0) {
			printk("Storage init ERROR %d!\n",disk_access_init(disk_pdrv));
			break;
		}

		if (disk_access_ioctl(disk_pdrv,
				DISK_IOCTL_GET_SECTOR_COUNT, &block_count)) {
			printk("Unable to get sector count");
			break;
		}
		printk("Block count %u", block_count);

		if (disk_access_ioctl(disk_pdrv,
				DISK_IOCTL_GET_SECTOR_SIZE, &block_size)) {
			printk("Unable to get sector size");
			break;
		}
		printk("Sector size %u\n", block_size);

		memory_size_mb = (uint64_t)block_count * block_size;
		printk("Memory Size(MB) %u\n", (uint32_t)(memory_size_mb >> 20));
	} while (0);

	mp.mnt_point = disk_mount_pt;

	int res = fs_mount(&mp);

	if (res == FR_OK) {
		printk("Disk mounted.\n");
		lsdir(disk_mount_pt);
	}
	else {
		printk("Error mounting disk.\n");
	}
}

The first if(disk_access_init..) returns -22.

If I jump in disk_access_init, I then enter in disk_access_get_di, and I can see that it returns 0

This fails to let me enter in the following if, thus returning -22.

I am quite sure that it is a configuration problem of the SPI device, just because in my first tests I could at least print the Memory Size. Now I encounter this problem for every spi-max-freq that I set (1Mhz, 4Mhz, 8Mhz, 16Mhz)

I'm planning to attach some logic analyzer probes asap, but I'm expecting to see no data being exchanged because of above.

I tried to get rid of the mmc definition in the dts, but that didn't work.

About the proj.conf, here there are the related config

CONFIG_DISK_DRIVER_SDMMC=y
CONFIG_DISK_ACCESS=y
CONFIG_FILE_SYSTEM=y
CONFIG_FAT_FILESYSTEM_ELM=y
CONFIG_SPI=y

Is there anything that I'm missing? Is there anything that can create conflict?

RE: SDHC SPI Send Command

Simonr — Tue, 14 Mar 2023 14:04:28 GMT

[quote user="dario.sortino"]I now encounter the storage init error[/quote]

What error exactly is this? What function returns the error, and what error code do you see? Did you try to set the frequency even lower (1MHz for example) to see if you still see the storage init ERROR with that frequency setting as well. If possible, can you get a scope trace of the SPI pins to see that the expected data is output on them, and what exactly happens before and as the device resets.

Best regards,

Simon

RE: SDHC SPI Send Command

dario.sortino — Fri, 10 Mar 2023 11:19:27 GMT

I updated the overlay as it follows:

&spi1 {
	compatible = "nordic,nrf-spim";
	status = "okay";
	cs-gpios = < &gpio0 7 GPIO_ACTIVE_LOW >;
	pinctrl-0 = <&spi1_default>;
	pinctrl-names = "default";
	sdhc0: sdhc@0 {
		compatible = "zephyr,sdhc-spi-slot";
		reg = <0>;
		status = "okay";
		label = "SDHC0";
		mmc {
			compatible = "zephyr,sdmmc-disk";
			status = "okay";
		};
		spi-max-frequency = <16000000>;		
	};	
};

Using pinctrl as:

spi1_default: spi1_default{
	group1 {
		psels = <NRF_PSEL(SPIM_SCK, 0, 8)>,
			<NRF_PSEL(SPIM_MISO, 0, 10)>,
			<NRF_PSEL(SPIM_MOSI, 0, 9)>;
	};

};

With this setting (16MHz), I now encounter the storage init error, so there is something odd.

I do not have a uSD slot that I can plug into the devkit, so It would be prone to errors to unsolder my socket and hand soldering all the connections to the dk pins.

Should I init in some way the SPI in my code? Even tough in the example there is not any kind of initialization

RE: SDHC SPI Send Command

Simonr — Thu, 02 Mar 2023 10:17:30 GMT

Hmm, that's very strange, a few more things you can try:

Have you tested this on a Development kit yet. That way we can narrow this down to whether it's a HW or SW issue at least.
You can also try to set the frequency even lower (1MHz for example) to see if you still see the storage init ERROR with that frequency setting as well.
Another thing I noticed is that you're using a deprecated way of setting your pins on. Please try using the pin control to set the SPI pins. You can see how it can be done in this SPI example by my colleague Torbjørn.
If possible, it could also be helpful to get a scope trace of the SPI pins to see that the expected data is output on them, and what exactly happens before and as the device resets.

Best regards,

Simon

RE: SDHC SPI Send Command

dario.sortino — Wed, 01 Mar 2023 10:59:48 GMT

Hi Simonr.

I tried to lower the SPI freq to 8M, and this time I cannot even see the size of the SD. If you take into consideration the fat_fs example, with 8M I enter this control

if (disk_access_init(disk_pdrv) != 0) {

printk("Storage init ERROR!\n");

break;

}

While with 16M I successfully skip that and while debugging it ends up to the first sdhc_spi_send_cmd that it encounters deep down to the function fs_mount inside functions, showing no errors but rebooting the MCU.

My SDK is NCS 2.2.0.

RE: SDHC SPI Send Command

Simonr — Wed, 01 Mar 2023 10:22:44 GMT

I see that you've set the SPI max frequency to 16MHz, but the SPI1 instance in the nRF5340 does not support higher data rates than 8Mbps, so can you try lowering the spi-max-frequency down to 8000000?

What SDK version are you using for development? When debugging, do you see any kinds of error codes as when the application moves into fs_mount(), or where does the debugging session end up?

Best regards,
Simon