Problem about file create on nRF52840

RE: Problem about file create on nRF52840

Hung Bui — Thu, 06 Aug 2020 13:37:17 GMT

Hi Blance,

I'm glad that you found a solution. We are still waiting for the developer to be back to have a look at the issue. We will give you an update when we hear from them.

RE: Problem about file create on nRF52840

Blance — Tue, 04 Aug 2020 02:36:48 GMT

Hi Hung:

I have fond https://devzone.nordicsemi.com/f/nordic-q-a/56609/fatfs-corruption

and try set qspi writeback_mode = false(default :true),Problem seems solved

success : 2293

fail : 0

But I don't understand the function of writeback_mode, and I can't determine whether not using it will cause other problems. Can you help confirm?

RE: Problem about file create on nRF52840

Hung Bui — Mon, 03 Aug 2020 15:17:01 GMT

Hi Blance,

Could you provide the code you used for automatic testing ?
I tried to test here with your previous code again but the result is different now that I don't see the issue any more.
But as you mentioned the issue rate is low at ~1/60. But it's strange that last time on the first trial I got the issue showed in the screenshot.
Have you tried "nrfjprog --qspierase" before testing ? It's to make sure the qspi is fully erased before we test.

RE: Problem about file create on nRF52840

Blance — Wed, 29 Jul 2020 10:19:56 GMT

Hello everyone:

Can anyone provide a program to change the usbd_msc example to use SPI to control PCA10056 Flash?
I think it will be useful

RE: Problem about file create on nRF52840

Blance — Wed, 29 Jul 2020 04:03:22 GMT

Hello everyone:

I use usbd_msc.7z to continue the test
I found that changing the FAT memory to nrf_block_dev_ram and retesting, everything seems to be normal

This strengthens my opinion that there is a problem with QSPI controlling FLASH
Or incompatible with FAT bit definition

RE: Problem about file create on nRF52840

Blance — Tue, 28 Jul 2020 03:49:57 GMT

Hi All:

I tried to modify the experiment and change 400 to 40. The test result is still that about 60 files will have one file corruption. The preliminary conclusion has nothing to do with the size of the files, but with the number of files.

RE: Problem about file create on nRF52840

Blance — Fri, 24 Jul 2020 06:40:11 GMT

Hi All:

Add the test to continuously write files to the root directory, the content is also 0x00~0xFF loop,
4k bytes are written once,
After writing 400 times, whether the read data content meets the 0x00~0xFF cycle,

if yes, the number of successes is +1, otherwise the number of failures is +1

Create a new file after the comparison is over, if the memory is insufficient, delete the oldest file

Performing this test on my personal hardware can still reproduce the abnormal file problem

format setting FAT12 4096 Bytes

three device test at same time,and stop test when get fail,finally I get same result

"success: 62, fail: 1"

I think there may be incompatibility between QSPI and FATFS
The address needs to be aligned to 4 Bytes. This requirement is suspicious
Can someone provide an interface that does not need to be aligned with 4 Bytes?

RE: Problem about file create on nRF52840

Blance — Thu, 23 Jul 2020 12:28:01 GMT

Hi Hung,

Thanks for your reply, this problem is easy to reproduce, but few people pay attention?

In addition, tests were conducted, and if the file is not written to the folder, but is changed to the root directory, the possibility of a problem is much lower.

RE: Problem about file create on nRF52840

Hung Bui — Tue, 21 Jul 2020 12:22:39 GMT

Thanks Blance.

I can see the issue now. We will need to investigate on how this happened. Unfortunately our developers on QSPI and USB are on vacation so we would need some time to look into this.

RE: Problem about file create on nRF52840

Blance — Mon, 20 Jul 2020 09:34:31 GMT

Hi Hung:

my file content is loop on 0x00 to 0xFF

and look on your address 0x3000 ~ 0x31FF

that's the issue,

RE: Problem about file create on nRF52840

Hung Bui — Mon, 20 Jul 2020 09:20:20 GMT

Hi Blance,
I tried your example here on a PCA10059 but I don't see the issue.
Attached is the file.
Have you tried to test on another board /another QSPI flash chip ?

devzone.nordicsemi.com/.../0041C6.TXT