Group address message crashes all nodes?!

RE: Group address message crashes all nodes?!

JVKran — Fri, 06 Nov 2020 08:26:31 GMT

Hi Mttrinh,

Okay, I'm glad to let you know that it's working as it's supposed to be. None of the above described behaviour occurs anymore. Had to do a rewrite from the bottom up to determine the cause.

As far as the origin of the problem, I've got some suspicions. In the end, the SDK's didn't turn out to be the problem; everything is working with the most recent (according to the example modified) SDK's. Am very happy with that.

I've found the following things to cause said behaviour:

The mesh_stack_init_params being defined in a local context resulting in the destruction when the function returns. As is the case when using an object oriented approach and initializing the mesh stack in the constructor.
Using the Idle Handler method instead of the Bearer Event Handler as described in the example. I should note however, that there are a lot of tasks running in our program. It's very plausible the Idle Handler works fine with less demanding tasks.

To be honest, I think using the SDK with the SDK for Mesh running on FreeRTOS being written in C++ didn't help either. Since it isn't really supported, a mistake or incompatibility can easily slip in...

All in all I'm glad the problem is solved; it was quite a large one as you can tell. Neither of the ways out (abandoning C++ or FreeRTOS) was an option to me, so it really had to work in the end.

Anyway, thanks for your support and have a great day!

Kind regards,

Jochem

RE: Group address message crashes all nodes?!

Mttrinh — Wed, 04 Nov 2020 13:19:21 GMT

It hasn't been clear what is causing the issue but might be something with FreeRTOS and SDK versions like you said. Keep me updated :)

RE: Group address message crashes all nodes?!

JVKran — Wed, 04 Nov 2020 12:49:07 GMT

Hi Mttrinh,

I went back to the first commit I've done and can now say with 100% certainty it has always been this way; I just didn't notice it. Hence, I've come to the conclusion that the FreeRTOS example on GitHub doesn't work with my SDK versions.

I'll try to revert to V4.1.0 and V16.0.0 to see if the issue is resolved. Gonna take some time though... Have to redo all changes to the SDK's and name conflicts between FreeRTOS and the SDK for Mesh.

I'll let you know how it goes and if it turns out to be functional.

Kind regards,

Jochem

RE: Group address message crashes all nodes?!

JVKran — Mon, 02 Nov 2020 12:57:05 GMT

Yep, found it out. Defining the "DEBUG_NRF" or "DEBUG_NRF_USER" macro resulting in the definition of "configASSERT" in FreeRTOSConfig.h results in the "ASSERTION FAILED at :0". When said macro isn't defined, the nodes crash.

Below the part in FreeRTOSConfig.h it's all about.

#if defined(DEBUG_NRF) || defined(DEBUG_NRF_USER)
#include "nrf_mesh_assert.h"
#define configASSERT( x )                                               NRF_MESH_ASSERT(x)
#endif

It's clear that something is wrong, but what? Has anyone got an idea? Now that I think of it, it's also impossible to subscribe to addresses. The app says "Not a subscribe model" when trying to do so. Could it have to do something with that?

Kind regards,

Jochem

RE: Group address message crashes all nodes?!

JVKran — Mon, 02 Nov 2020 12:14:17 GMT

Hi Mttrinh,

I'm glad to let you know I can provision the nodes again; reverted a couple of commits... The problem as described in the opening post is still the same though.

Kind regards,

Jochem

RE: Group address message crashes all nodes?!

JVKran — Mon, 02 Nov 2020 10:49:15 GMT

Hi Mttrinh,

Now I can't even get the composition data or ttl after adding a newly provisioned node... I'm absolutely clueless... The same errors occur.

Changing to SoftDevice 'S140_nrf52_7.0.1' doesn't change anything. Maybe it's important to note that the provisioning through the app does never work and I always have to do it manually which is, impossible now.

Kind regards,

Jochem

RE: Group address message crashes all nodes?!

JVKran — Mon, 02 Nov 2020 10:13:52 GMT

Hi Mttrinh,

After enabling the configASSERT in FreeRTOS, I managed to produce output in the terminal. This is what's printed in the terminal of the server with unicast address 0x0002 when the client transmits to 0xFFFF.

<info> app: RX: [aop: 0x00C1]

<info> app: RX: Msg
<info> app: Led is now turned on.
<info> app: TX: [aop: 0x00C4]

<info> app: TX: Msg
<error> app: ASSERTION FAILED at :0

And this is what the debugger shows.

Program
 received signal SIGTRAP, Trace/breakpoint trap.
0x00052aa4 in app_error_fault_handler (id=id@entry=16386, pc=324690, info=info@entry=536902224) at C:/Users/kraan/Downloads/clean/nRF5_SDK_17.0.2_d674dde/components/libraries/util/app_error_weak.c:100
100	    NRF_BREAKPOINT_COND;

Hence, it appears to be a weak app error. A quick look in the Q&A shows 'ASSERTION FAILED at :0' is most likely caused by calling 'NRF_MESH_ASSERT(false)'. Is that correct? It isn't anywhere in code written by me.

Kind regards,

Jochem

RE: Group address message crashes all nodes?!

JVKran — Mon, 02 Nov 2020 09:28:36 GMT

Hi Mttrinh,

I don't know really. It just stops... Nothing is printed in the terminal. I'll look into it right now though!

As for your second question; I'm using V4.2.0 from the SDK for Mesh and V17.0.2 from the nRF5 SDK.

Kind regards,

Jochem

RE: Group address message crashes all nodes?!

Mttrinh — Mon, 02 Nov 2020 09:20:54 GMT

Hi,

Could you give some more details on what you mean by your node "crashing"? Is it a hardfault? app_error_handler? or other resets?

Also, can you provide which SDK versions you are using?