NCS v1.6.1 with ENC28J60 crashes with MPU fault

[newUser]

Hi,

It seems that we are experiencing exactly the same as what is described in this topic by @kienasv. That topic seems closed for replies, and the solution has never been posted.
For us, the problems started after updating from NCS v1.5.1 to NCS v1.6.1. In the other topic, the problem seems to exist in 1.5.0 already.

The code crashes after a few iterations in eth_enc28j60_read_mem(). In the same place as in the other topic. (See image)

We have tried CONFIG_STACK_SENTINEL=y and CONFIG_HW_STACK_PROTECTION=y, but no stack overflows have been detected.
Also increasing CONFIG_ETH_ENC28J60_RX_THREAD_STACK_SIZE by a lot did not help.

Also the ethernet logging does not show us anything..

[Simon]

Could you describe how to reproduce this, or upload your project (or a simplified project that demonstrates the issue) as a zipped file?

[newUser]

Our setup is one with a nRF52840 connected to an ENC28J60. When data is coming in on the ethernet chip, it crashes on the read_mem function using SPI.

Also, could we make the ticket private if we are to share code? Do you have nRF52840 hardware with the ENC28J60?

Thanks

[Simon]

Sorry, I was a little quick to ask, I don't have an ENC28J60 at hand at the moment, so you don't need to share the project. I'll try to get a hold of one if needed.

If you use addr2line, what line do you find? The same as in MPU Fault when running central coded bl...?

Also could you share the Fail log? Set CONFIG_THREAD_NAME=y so we can see what thread caused the fault.

[newUser]

No problem!

Indeed, it's the same as in the other topic.

[Simon]

If you got the same fail log as the customer in MPU Fault when running central coded bl..., the issue occurs because you're trying to write to the address 0x30a03 ("MMFAR Address: 0x30a03"), which is below the RAM memory address 0x20000000. So p_cb->evt.xfer_desc.p_rx_buffer points to 0x30a03 for some reason.

Can you try to set CONFIG_ASSERT=y in the prj.conf and see if you get any hints on what the issue may be.

Best regards,

Simon