Unexpected branch address rewrite when using XIP

RE: Unexpected branch address rewrite when using XIP

Einar Thorsrud — Tue, 24 Oct 2023 06:58:46 GMT

Hi Torsten,

I am glad to hear you found the issue. Thank you for letting us know and explaining what had happened.

Einar

RE: Unexpected branch address rewrite when using XIP

Torsten Robitzki — Tue, 24 Oct 2023 06:19:58 GMT

Good Morning Einar,

I did not tried a second board, because the issue was so reproducible. But now, I've tried a second board (V2.0.1), installed the firmware, ran it and issued nrfjprog --readregs. When issuing that command quite shortly after starting the firmware, I see different PCs. When I wait for a little bit longer (>15s), I see the PC at 0x000003B6, which is according to my lis file the HardFault_Handler.

The 15s seem a little bit long to me, so I've attached a debugger and this time, the Hardfault was caused by something else. And the call stack came from with main(), and all addresses on the call stack are looking reasonable. So, with the board, the problem is not reproducible.

So, I've switched back to the V3.0.1 board. I again, got reproducible a Hardfault during the initialization of the C/C++ runtime. But this time the call stack was different from the original problem report.

I've soldered small wires to the QSPI interface to debug the flash memory. That wires where attached to a Saleae and after I've disconnected the Saleae, the problem was gone.

So, it looks like this was caused by me, attaching a logic analyzer. I'm sorry for wasting your time.

Thank you very much,

Torsten

RE: Unexpected branch address rewrite when using XIP

Einar Thorsrud — Mon, 23 Oct 2023 14:11:12 GMT

Hi Torsten,

I see. I will test more tomorrow, but I wonder if you see this on multiple boards? The reason I ask is that for now I only tested running for a few seconds, and then reading out the CPU registers with "nrfjprog --readregs" I don't see a hard fault. Can you share your output from "nrfjprog --readregs"?

RE: Unexpected branch address rewrite when using XIP

Torsten Robitzki — Mon, 23 Oct 2023 13:39:11 GMT

Hi Einar,

cool, thanks! If everything would work as expected, you should for example be able to set a breakpoint in main() and after a reset, you should set a debugger waiting there.

But currently, the initialization of the C/C++ runtime does not execute as expected (see above) and a hard fault is caused by the jump to the unexpected address 0x122400da. If you are able to reproduce this hard fault, you have reproduced the issue.

Maybe you can have a look at the initialization of the QSPI peripheral and maybe you can shed some light on why the jump instruction is redirected to that address.

Thanks in advance

Torsten

RE: Unexpected branch address rewrite when using XIP

Einar Thorsrud — Mon, 23 Oct 2023 13:32:25 GMT

Hi Torsten,

I am sorry for the delay, I was not able to look into your project until now. I am able to build your application according to the instructions and program a nRF52840 DK with it. How should I test in order to reproduce the issue? And how should it normally function?

RE: Unexpected branch address rewrite when using XIP

Torsten Robitzki — Wed, 18 Oct 2023 09:43:28 GMT

Yes, it is open source: https://github.com/TorstenRobitzki/bluetoe/tree/feature/scheduled_radio2/tests/scheduled_radio

I've tried to add a short description as how to build the firmware (tester.elf). The project uses cmake/make/gcc for the build. One additional dependency is Catch2 (https://github.com/catchorg/Catch2). As I'm currently evaluating different test framework, the path to Catch2 is hard coded in CMakeLists.txt:41 and has to be adjusted accordingly.

Thank you very much for your effort. Please let me know, when I can assist or if you want me to provide binaries.

RE: Unexpected branch address rewrite when using XIP

Einar Thorsrud — Wed, 18 Oct 2023 08:56:22 GMT

Hi Torsten,

That is a good point. Can you share the project you run on the DK along with instructions on how to reproduce it?

RE: Unexpected branch address rewrite when using XIP

Torsten Robitzki — Mon, 16 Oct 2023 12:12:46 GMT

Hi Einar,

yes, I've stumbled over that too, but I would read it as, there is a problem when executing code out of the internal flash or RAM, that reads from the external flash. In my example, I'm purely executing from external flash.

best regards,

Torsten

RE: Unexpected branch address rewrite when using XIP

Einar Thorsrud — Mon, 16 Oct 2023 12:01:23 GMT

Hi Torsten,

It seems like you are seeing erratum 216, which unfortunately limits the usability of XIP on the nRF52840.

Einar