Mesh DFU Relay Problem

[Panda]

Hi ,

     I have tested two boarrds(PCA10040) with  same application id and higher application version,These work  well done.Then I want to test mesh DFU Relay function,However ,the target board does not upgrade.

    Prepare: Name two boards, Board A which is connected to pc,acts as source ,Board B which is the targect to upgrade using mesh  .

   The follow is what i do:

   (1) To generate Device Page with config  "application_id": 1,"application_version": 1, .Using it to program Board  A  as mesh dfu guide .

   (2)To generate Device Page with config  "application_id": 2,"application_version": 1, .Using it to program Board  B  as mesh dfu guide .

 （3)To generate DFU file with command  :nrfutil dfu genpkg --application ... using config  "application_id": 2,"application_version": 2, 

   (4)Then ,Baord A connect PC with USB serial. Board B is just supplied power .Using command: nrfutil --verbose dfu serial -pkg ... to start mesh dfu.

   The result is:

     At the first few seconds ,both boards  go to dfu with LED 0and LED2 be lighted, A few mnutes later ,Baords B ends dfu with LED 0and Led1 be lighted . and Board A is still going with dos  windows shows rate of progress.About an hour later ,the dfu transport for Baord A is done,Even I still wait for another nearly one hour, However Board B is still not be upgarded.

  I check the Board B's final ending RTT log:

......

0> <t: 16666007>, nrf_mesh_dfu.c, 324, Write complete (0x2000FE90)
0> <t: 16666010>, nrf_mesh_dfu.c, 333, Flash idle.
0> <t: 16688463>, nrf_mesh_dfu.c, 528, RADIO TX! SLOT 6, count 3, interval: exponential, handle: FFFC
0> <t: 16688479>, nrf_mesh_dfu.c, 324, Write complete (0x2000FE90)
0> <t: 16688483>, nrf_mesh_dfu.c, 333, Flash idle.
0> <t: 16700264>, nrf_mesh_dfu.c, 383, Abort event. Reason: 0x3
0> <t: 16700267>, main.c, 175, Mesh DFU End !

    The Abort Reason:0x3 is found out in file dfu_types_mesh.h  DFU_END_ERROR_PACKET_LOSS  .  So the relay dfu fail  may be packet loss over RF radio.How  to repair this ? who can give me some advice ? Thanks you.

Best Regards,

Panda.

[Hung Bui]

Hi Panda, 

Could you post the full log on the B device ? Also please add a breakpoint in the firmware and check if fw_updated_event_is_for_me() return true.

Please test with a smaller application so the testing time will be shorter, the blinky application for example. 

I don't see why sending applicaction with the different id to the one connected to PC causing this issue. The issue  DFU_END_ERROR_PACKET_LOSS   happens when the segments missing are >64 segments. 

We need to know if the issue can be reproduced every time or it only happens from time to time and if it happens on very image or on some particular images. 

[Panda]

Hi Hung,

  The below is  Device B's full log :

0> <t: 903695>, nrf_mesh_dfu.c, 390, New firmware!
0> <t: 903698>, main.c, 154, Requesting DFU transfer with bank at 0x0004C000
0> <t: 903701>, nrf_mesh_dfu.c, 528, RADIO TX! SLOT 0, count 255, interval: periodic, handle: FFFD
0> <t: 903705>, nrf_mesh_dfu.c, 534, Killing a TX slot prematurely (repeats done: 3).
0> <t: 903709>, nrf_mesh_dfu.c, 561, SERIAL TX!
0> <t: 968745>, nrf_mesh_dfu.c, 528, RADIO TX! SLOT 0, count 255, interval: periodic, handle: FFFD
0> <t: 968749>, nrf_mesh_dfu.c, 534, Killing a TX slot prematurely (repeats done: 0).
0> <t: 968753>, nrf_mesh_dfu.c, 561, SERIAL TX!
0> <t: 1035468>, nrf_mesh_dfu.c, 430, DFU start
0> <t: 1035471>, main.c, 170, Mesh DFU Start...
0> <t: 1035474>, nrf_mesh_dfu.c, 528, RADIO TX! SLOT 1, count 6, interval: exponential, handle: FFFC
0> <t: 1041587>, nrf_mesh_dfu.c, 329, Erase complete (0x4C000)
0> <t: 1041590>, nrf_mesh_dfu.c, 333, Flash idle.
0> <t: 1101978>, nrf_mesh_dfu.c, 528, RADIO TX! SLOT 2, count 3, interval: exponential, handle: FFFC
0> <t: 1101995>, nrf_mesh_dfu.c, 324, Write complete (0x2000FE90)
0> <t: 1101998>, nrf_mesh_dfu.c, 333, Flash idle.

......

0> <t: 6693194>, nrf_mesh_dfu.c, 528, RADIO TX! SLOT 7, count 3, interval: exponential, handle: FFFC
0> <t: 6693211>, nrf_mesh_dfu.c, 324, Write complete (0x2000FE90)
0> <t: 6693214>, nrf_mesh_dfu.c, 333, Flash idle.
0> <t: 6710839>, nrf_mesh_dfu.c, 528, RADIO TX! SLOT 1, count 3, interval: exponential, handle: FFFC
0> <t: 6710856>, nrf_mesh_dfu.c, 324, Write complete (0x2000FE90)
0> <t: 6710859>, nrf_mesh_dfu.c, 333, Flash idle.
0> <t: 6726942>, nrf_mesh_dfu.c, 528, RADIO TX! SLOT 2, count 3, interval: exponential, handle: FFFC
0> <t: 6726958>, nrf_mesh_dfu.c, 324, Write complete (0x2000FE90)
0> <t: 6726962>, nrf_mesh_dfu.c, 333, Flash idle.
0> <t: 6743185>, nrf_mesh_dfu.c, 383, Abort event. Reason: 0x3
0> <t: 6743189>, main.c, 175, Mesh DFU End !

    From the log ,Device B receives the dfu packets.The application size is:147372 bytes. The Device A connected to PC can relay the whole process.

  The Device B's issue happens every time.

   My project use nRF SDK:15.3.0 ,Mesh SDK: 3.2.0 .

  

 If there are any  advice to repair this issue ,Please tell me .

Best Regards,

Panda

  

  

[Hung Bui]

Hi Panda,
 
Good news, we have a workaround prototype that you can test. We are still continue testing but it seems to work. 

As explained earlier it was an issue that the serial node doesn't re-transmit the missing segment. The fix here involved modification on nrfutil, bootloader and the serial dfu app. 
Please make sure you rebuild nrfutil after the patch. 

Attached is the .patch file and the patched file that I was using in the test. Please let me know the result on your side.

patch.zip

[rmarques]

Hello, our team had this problem some odd months ago with nrf52840 - post.

Can you guys compile a bootloader for this chip with the same fix please? We implemented our own recovery routines on the app side but we feel it's clunky and it fails sometimes.

[Hung Bui]

Hi, 

Please implement the .patch file for the bootloader and re-compile it. 
To build the bootloader from command line do this: 

cmake -G Ninja .. -DSDK_ROOT=[path to your SDK] -DCMAKE_BUILD_TYPE=MinSizeRel -DBUILD_BOOTLOADER=ON

[Hung Bui]

Hi Marques, 
If you still have trouble with the nRF52840, please use the patch I provided here. It's for SDK v4.0. The command to generate bootloader file for the nRF52840 is: 

cmake -G Ninja .. -DSDK_ROOT=[path to your nRF5 SDK] -DCMAKE_BUILD_TYPE=MinSizeRel -DBUILD_BOOTLOADER=ON -DPLATFORM=nrf52840_xxAA -DSOFTDEVICE=s140_7.0.1

Please make sure you recompile the nrfutil using the patch as well. 

patchnRF52840.zip

[Panda]

Hi Hung,

   sorry to reply later.I have seen the patch .I am not familiar with cmake.Could you tell me how to build bootloader with camke in detail?

   I install camke using the source https://cmake.org/download .Then I input "cmake''in dos command line ,it works.Next I input"  cmake -G Ninja .. -DSDK_ROOT=[E:/Project/Mesh_v3.2.0/nrf5_SDK_for_Mesh_v3.2.0_src/mesh/bootloader] -DCMAKE_BUILD_TYPE=MinSizeRel -DBUILD_BOOTLOADER=ON"  with your patch files replace .Error shows "The source directory "C:/Users" does not appear to contain CMakeLists.txt."

   the another method is use CMake 3.17.0 gui, I have tried but also failed.

   I notice under the mesh sdk file ,there is a CMake file.Maybe it could be used?

  Waitting for your help .on the same time,i will learn something about cmake.Thank you.

Best Regards,

Panda

   

[Hung Bui]

Hi Panda, 
Please follow the instruction on how to build the mesh stack with cmake here: https://infocenter.nordicsemi.com/topic/com.nordic.infocenter.meshsdk.v3.2.0/md_doc_getting_started_how_to_build.html?cp=7_6_0_1_2

You need to call the script for Cmake inside a subfolder (for example "build" folder) inside the mesh folder. 

[Panda]

Hi Hung,

      Do you mean the bootloader can also be builded by SES? just create a booeloader project using SES ,then get the hex ? I will raed he source link you gave above.Thanks

[Hung Bui]

No, there would be no SES project for the bootloader in the SDK. 
But we do have an unofficial SES project here: 
https://devzone.nordicsemi.com/f/nordic-q-a/51534/stepping-through-nrf-mesh-bootloader-code-in-segger-embedded-studio

[Panda]

Hi Hung,

 Could you give  .emProject and flash_placement.xml for nrf52832 using sdk nRF5_SDK_15.3.0_59ac345 and nrf5_SDK_for_Mesh_v3.2.0_src ?I will try to tansplant since the builded cannot be programed. Thanks.

Best Regards,

Panda

[Hung Bui]

Hi Panda, 

Have you tried to compile using cmake ? Please follow the instruction on how to build mesh stack and let me know which issue you have.