Random NRF_FAULT_ID_APP_MEMACC errors

[Martin Ziel]

I've tried this with S132 and S140 7.2.0 and NRF SDK 17 on an NRF52832. I get random NRF_FAULT_ID_APP_MEMACC errors, that seem to be triggered by long running functions (though it might be unrelated). The RAM is configured correctly, according to RAM start value accquired from nrf_sdh_ble_default_cfg_set. The info parameter in the app_error_fault_handler is 0. Is there another way to figure out where the error originates from?

[Vidar Berg]

Hi,

I guess you may have checked out the documunetion on this already(NRF_FAULT_ID_APP_MEMACC), but this assertion with the info pointer set to NULL indicates that there has been a RAM access violation made by the app. That is, the app has written something to the RAM region reserved to Softdevice (Memory layout). Please try to see if the call stack might reveal where this RAM access violation happened in your app.

Here's an example where I intentionally triggered the assert from the battery_level_update() callback. As you can see, the call stack confirms the assertion occured right after the battery_level_update() call in this case.

[Martin Ziel]

Call Stack:

app_error_fault_handler@0x000344cc

??@0x00024d08 (Unknown Source:0)

The call stack points me towards a code section in the Softdevice (which goes from 0x0 to 0x27000)

I already used the IRQ fix suggested in the release notes to no avail.

[Vidar Berg]

You need to halt the CPU before it reaches the app_error_fault_handler() function in your app. Please place a breakpoint in the MBR's hardfault handler as shown in the screenshot I uploaded. And I would not recommend using the s140 for testing on 52832 since it is not supported.