[BUG] ESP32C6 SPI Read Issue

[FelipeMdeO]

Description / Steps to reproduce the issue

Hello All,

I am porting MFRC522 rfid card reader to esp32 c6. During port test I am facing issue related SPI message read.

1 - Add breakpoint in mfrc522_getfwversion() function, - I did minor change in code, not necessary:
file: mfrc522.c
line 1160

2 - Run the code until it hits the breakpoint

3 - In my case I start logic analyze and hit "next" in the gdb.
Looking syslog msgs and results in logic Analyze We can see that the rfid IC answer 0xB2, but spi drive can't get this value.

Look images below:

On which OS does this issue occur?

[Linux]

What is the version of your OS?

Ubuntu 24

NuttX Version

master

Issue Architecture

[risc-v]

Issue Area

[Drivers]

Verification

• I have verified before submitting the report.

[acassis]

@fdcavalcanti PTAL

[FelipeMdeO]

Hello @fdcavalcanti

I added a breakpoint in:
esp_spi.c:759

Reading:
trans.rcv_buffer and trans.send_buffer it always have same value, maybe this behavior explain how the example is working.

Steps:
1 - add breakpoint in mfrc522_getfwversion function.
2 - add brakpoint in esp_spi.c esp_spi_poll_send function

Running code and get trans.* value, you will see:

Note: 0xee is expected value:
( 0x80 | (MFRC522_VERSION_REG & 0x7e)

[fdcavalcanti]

Hello @FelipeMdeO, let's figure this out.

We'll start making sure there's no hardware issues.

• Is it an Espressif dev-kit or a custom board?
• Which pins are you using with SPI?
• Any pull-ups on the circuit?
• Which SPI mode are you working with? just noticed it is already inside the lock function.

[FelipeMdeO]

Hello @fdcavalcanti

1 -
I am using esp32c6 dev kit m and rfid-rc522 rf card read:

2 -
SPI2 CS - 15
SPI2 CLK - 6
SPI2 MOSI - 7
SPI2 MISO - 2

3 - No, direct connection

I don't know if this the right way to share, but I did changes in fork from nuttx master:
https://github.com/FelipeMdeO/nuttx.git
branch: esp32-mfrc522-support

Below, my defconfig:

defconfig.zip

[fdcavalcanti]

I see no issue on C6 driver side.
Please use NuttX main branch and try again. We need to be on common ground regarding the source code.
Let me know if anything different happens.

[FelipeMdeO]

Hello @fdcavalcanti,

How do you test in your side?
I will try do the same thing on my side.

[FelipeMdeO]

Hello @fdcavalcanti.

I tested using spi example and known slave device, in this situation it worked properly. So the issue is not in the drive.
I will investigate it better, sorry for false bug.

[fdcavalcanti]

No problem! Good luck.

[FelipeMdeO]

Hello @fdcavalcanti ,

I am reopen this issue, I will explain.

My issue happens when I tried use MFRC522 driver, this driver uses SPI_SEND function:

Using SPI_SEND function I cannot get read data properly.

The ESP SPI example uses SPI_EXCHANGE function. When we request once a byte, it send using exchange with nwords=1.

To do a proof, I changed the spi driver to esp_spi_send call esp_spi_exchange instead of esp_spi_poll_send

In this way I was able to read MFRF522 FW Version.

@fdcavalcanti can you check SPI_SEND in your side, please ?

[FelipeMdeO]

Hello,

I was clear in my last explanation, do you need more information?

[fdcavalcanti]

Hi Felipe. I don't have this hardware to test.
But could you please try this inside esp_spi_poll_send:

------------------ arch/risc-v/src/common/espressif/esp_spi.c ------------------
index e84b22b336..4988f17aa3 100644
@@ -750,8 +750,9 @@ static uint32_t esp_spi_poll_send(struct esp_spi_priv_s *priv, uint32_t wd)
   trans.line_mode.cmd_lines = 0;
   priv->ctx->trans_config = trans;
 
-  spi_ll_set_mosi_bitlen(priv->ctx->hw, priv->nbits);
+  // spi_ll_set_mosi_bitlen(priv->ctx->hw, priv->nbits);
   spi_hal_prepare_data(priv->ctx, priv->dev_cfg, &trans);
+  spi_hal_setup_trans(priv->ctx, priv->dev_cfg, &trans);
   spi_hal_user_start(priv->ctx);

Make sure the rest of the file is the same as the main branch.

[FelipeMdeO]

Hello @fdcavalcanti,

Sorry, not yet.
I put side by side comparation between SPI Exchange and SPI Send result.
Note that the SPI Send return strange data, like memory address, this happens with current master code too.
Maybe you can try use any spi slave device, if it uses SPI Send in the driver.

[acassis]

Hi Felipe. I don't have this hardware to test. But could you please try this inside esp_spi_poll_send:

------------------ arch/risc-v/src/common/espressif/esp_spi.c ------------------
index e84b22b336..4988f17aa3 100644
@@ -750,8 +750,9 @@ static uint32_t esp_spi_poll_send(struct esp_spi_priv_s *priv, uint32_t wd)
   trans.line_mode.cmd_lines = 0;
   priv->ctx->trans_config = trans;
 
-  spi_ll_set_mosi_bitlen(priv->ctx->hw, priv->nbits);
+  // spi_ll_set_mosi_bitlen(priv->ctx->hw, priv->nbits);
   spi_hal_prepare_data(priv->ctx, priv->dev_cfg, &trans);
+  spi_hal_setup_trans(priv->ctx, priv->dev_cfg, &trans);
   spi_hal_user_start(priv->ctx);

Make sure the rest of the file is the same as the main branch.

@FelipeMdeO I think you can use the spi tool to so this test. Just do the ordinary loop test, but using send instead of exchange command.

[FelipeMdeO]

Thank you for you suggestion @acassis, very nice.

I did changes in spi_transfer.c to use SPI_SEND instead SPI_EXCHANGE in the spi tool.
Transfer is ok, so I don't know whats happen when I am try send/read data to MFRC device.

I will take a look better during next days.

[acassis]

Actually the message was to Filipe Cavalcanti that doesn't have the board:
@fdcavalcanti (since you don't have the board) I think you can use the spi tool to so this test. Just do the ordinary loop test, but using send instead of exchange command.