PCA10056 v3.0.1 cannot run the matter example of NCS2.6.0

[zero]

Dear support team,
I now have three PCA10056 v3.0.1 versions of nRF52840-DK. After connecting the development board to the PC, I downloaded the \v2.6.0\nrf\samples\matter\light_switch example. I opened the serial port to view the log. There is no log information, and LED1 does not flash. , indicating that the matter example is not running; download the \v2.5.2\nrf\samples\matter\light_switch example, open the serial port to view the log, there is matter log output, and LED1 will flash, indicating that the matter example has been run; 3 PCA10056 v3.0.1 phenomena consistent. Please tell me why the matter example of ncs2.6.0 cannot be run on the PCA10056 v3.0.1 development board? Please help analyze the cause and give a solution.
I found an nRF52840-DK with PCA10056 v0.9.2, which can run the matter example of ncs2.6.0.

BR,

Zero

[Hieu]

Hi Zero,

I don't have a nRF52840 DK v3.0.1 with me at the moment and could not reproduce your observation with my v3.0.0.

I will try to get a v3.0.1 and test tomorrow.

To rule out any possible differences, could you please try with this hex file?
matter_light_switch_260_nrf52840dk_nrf52840.hex
It is my compilation of the Matter Light Switch sample on NCS v2.6.0.

BR,

Hieu

---

Please be informed that our team is currently having reduced availability due to the Easter holiday. There will be some delays in our response during this time, but we will respond to you at the earliest opportunity. Our apologies for the inconvenience.

[zero]

Hi Hieu，

I burned the hex you compiled onto the PCA10056 v3.0.1 board, but the phenomenon is the same and it cannot run.

At the same time, I asked my colleague to borrow a v3.0.0 52840-DK, which can indeed run the matter examples of ncs2.6.0. I also verified that the v3.0.2 DK cannot run the matter examples of ncs2.6.0.

BR,

Zero

[Hieu]

Hi Zero,

I have been consulting internally and work off of the pointers I received but haven't gotten anything meaningful to talk about yet. This isn't straightforward, as I have feared...

I'm very sorry for the lack of good news.

I will keep trying and keep you updated.

BR,

Hieu

[zero]

Hi Hieu,

Maybe my test is helpful for you, I see through debug that in dev board v3.0.0 qspi initialization returns 0 and can read jedec-id = [C2 28 17];  But it is in development board v3.0.1/v3.0.2 that QSPI initialization returns 19 and reads jedec-id = [00 00 00];  You can analyze it in this direction, and I don't know how to analyze it.

I also did the following tests:
1. Connect the v3.0.1/v3.0.2 52840-DK to the pc
2. cmd enter nrfjprog --qspieraseall;
3. Download the ncs2.6.0 matter example.
Result: v3.0.1/v3.0.2 52840-DK works fine with the matter example.
Note that all ncs2.6.0 matter examples cannot be run on 52840-DK. The version of dk that I test can run may be because I have operated on external flash before.

BR,

Zero

[Hieu]

Hi Zero,

I also found some relevant internal records yesterday and found that nrfjprog --qspieraseall fixes the issue to me.

I planned to talk with those records' authors to fully understand the technicalities before answering you, but it seems you also reached the same findings.

I would like to confirm your observation. After the Matter sample work once, does it ever fail again?

Also, if you still have a v3.0.1/v3.0.2 DK that you haven't "fixed," can you confirm if flashing a sample on NCS v2.4.x or v2.5.0 also "fixes" it?

On my end I currently cannot find a DK that I can reproduce the issue again after fixing it once.

BR,

Hieu

[zero]

Hi Hieu,

Hieu said:

I would like to confirm your observation. After the Matter sample work once, does it ever fail again?

After nrfjprog --qspieraseall is repaired, the matter example can always be run.

Hieu said:

Also, if you still have a v3.0.1/v3.0.2 DK that you haven't "fixed," can you confirm if flashing a sample on NCS v2.4.x or v2.5.0 also "fixes" it?

I tried running the ncs2.5.2 matter example using the unfixed v3.0.1/v3.0.2 DK and then ran the ncs2.6.0 matter example. The problem still exists.

BR,

Zero

[Hieu]

Hi Zero,

What about 2.4.x and 2.5.0? This nuance is important for our current problem.

BR,

Hieu

[zero]

Hi Hieu,

Hieu said:

What about 2.4.x and 2.5.0? This nuance is important for our current problem.

I don't have the v3.0.1/v3.0.2 DK anymore that can't run the "'matter" example

BR,

Zero

[Hieu]

Hi Zero,

We have recorded this issue. For the time being, is this "fix" satisfactory for you? We will continue to investigate and work on a better fix, but your insight could help us with prioritization.

If you see a large impact, could you please also start conversation with your Nordic Sales contact?

BR,

Hieu

[zero]

Hi Hieu,

This solution is currently acceptable. But please really fix this problem as soon as possible.

BR,

Zero

[Hieu]

Hi Zero,

A fix was made in the Zephyr upstream: https://github.com/zephyrproject-rtos/zephyr/pull/71373.

and cherry picked into NCS: https://github.com/nrfconnect/sdk-nrf/pull/14812.

It should be fixed in the next release. I cannot comment on when that is, however.

You could try to cherry pick the fix as well.
The issue can be reproduced on a factory condition DK, not necessarily the 3.0.1 or newer, if you want to verify the fix.
I have not verified this, but to "unfix" a DK to be susceptible to the issue again, you can use this overlay:

&mx25r64 {
	writeoc = "pp";
	readoc = "fastread";
};

 

BR,

Hieu

[zero]

Hi Hieu,

I tested it and added: &mx25r64 {
writeoc = "pp";
readoc = "fastread";
};
52840-DK cannot run the matter example downloaded from ncs2.6.0. You can reproduce the problem and then use nrfjprog --qspieraseall.
But when the 52840-DK is reset after powering off and on, it cannot run. What difference does erasing bring to the flash itself? Or does Brand new DK mean there is something different about mx25r64 that causes this problem?

BR,

Zero