MSC exception

RE: MSC exception

dongxie — Thu, 07 Dec 2023 14:53:36 GMT

I have encountered the same problem, and there has been significant improvement after applying 1134.nrf_block_dev_qspi.c, but there is still data corruption.Looking forward to solving it

RE: MSC exception

Einar Thorsrud — Tue, 24 Oct 2023 07:18:34 GMT

I see. Please let me know if you find a way to reproduce the issue, as we will not be able to make progress on this on our end without it.

RE: MSC exception

Jason — Mon, 23 Oct 2023 02:30:11 GMT

Hi,

unfortunately, I have not found a way to consistently reproduce this issue after the latest patch was applied.

RE: MSC exception

Einar Thorsrud — Fri, 20 Oct 2023 14:08:53 GMT

Hi,

That is significant numbers. However, we have not been able to identify any potential cause for this, so we need a way to reproduce the issue in order to debug this issue further. Have you had any success in finding way to consistently reproduce it?

RE: MSC exception

Jason — Thu, 19 Oct 2023 07:48:32 GMT

Hi Einar,

I got the latest data from our server.

We have two products(let's say product A and product B) that use the nrf52840 and the same SDK, these two products have similar functions. Product A is developed by ourselves, and product B is developed by the 3rd party company.

In the last 90 Days, we have 310 different Users with this problem with product A. With product B we have 1241 different Users with this problem.

RE: MSC exception

Einar Thorsrud — Mon, 07 Aug 2023 10:55:21 GMT

Hi,

I see. Can you quantify how bit this issue is with the latest patch (failure rate)? I do not have any updates from our side. We need some way to reproduce in order to debug and understand the issue.

RE: MSC exception

Jason — Thu, 27 Jul 2023 05:49:34 GMT

Hi,

I can not find a way to reproduce so far. and We have less claims from the end-user since the latest patch but it is not fixed completely. I would like to know if you have any updates on this issue. Thanks!

RE: MSC exception

Einar Thorsrud — Fri, 10 Mar 2023 08:20:19 GMT

Thanks for the info. It will be difficult to make progress on this without a way to reproduce, so if you are able to find a way, then please let us know. Any other additional information you are able to gather regarding how/under which conditions this happen can also be useful.

RE: MSC exception

Jason — Thu, 09 Mar 2023 04:05:19 GMT

unfortunately, I am not able to reproduce this, also have no additional logs.

for the flow "open file->append data(one FIT message)->close file", I added more details.

user starts the training

->"create file->write the FIT header at beginning of file(the header includes the total length of FIT data)->close file"

-> "open file->append data(one FIT message)->close file" every second

user saves the training

->"open file->rewind file and update the FIT header->close file->rename file"

RE: MSC exception

Einar Thorsrud — Wed, 08 Mar 2023 11:46:40 GMT

I understand that this is not a good situation. I cannot promise any results at the moment, but we will review the QSPI block device module further to see if we can find more corner cases that are not properly handled or other issues.

Are you able to reproduce this somehow? Or do you have some additional logs or other information that can help us understand more how this happened?

RE: MSC exception

Jason — Tue, 07 Mar 2023 02:33:38 GMT

we don't have exact data for this issue, just got a few claims from the user sometimes. and I do not think all of the users will feedback after they encounter this issue. But if the issue can not be fixed completely, that means there is always a negative comment on our product. this product is a bike computer. You can imagine how disappointed the user was when they got the corrupted training data after more than an hour of training for the ride.

RE: MSC exception

Einar Thorsrud — Mon, 06 Mar 2023 09:41:00 GMT

Hi,

Thanks for the information. We will look into it. This may be a difficult issue though, so it could take some time.

In the mean time can you quantify how often you see data corruption now (with the latest fixes in nrf_block_dev_qspi.c) compared to before? And how big is the issue now?

RE: MSC exception

Jason — Sun, 05 Mar 2023 03:24:37 GMT

[quote userid="7377" url="~/f/nordic-q-a/73389/msc-exception/413205"]Have you confirmed that this corruption happened after applying the fix (meaning that the devices in the field were updated before it occurred)?[/quote]

yes, because the FIT file also logs the FW version, we can see the FW version is the latest version.

[quote userid="7377" url="~/f/nordic-q-a/73389/msc-exception/413205"]I see. Can you say more about this data (that is presented here) is stored? And can you upload the corrupt raw file here and explain in detail how you can see that it is corrupt (in case we can se a pattern or similar).[/quote]

we log the data every second: "open file->append data->close file", which means log one FIT message every second, the FIT file is a bin file, and it can be decoded to a text file by the PC tool if the file is integrity, if the decoder encounters a corrupted FIT message, then terminal decoding. please see this link FIT to get more details about the FIT Protocol.

I just deleted the corrupted messages(here only one message was corrupted) and then the FIT file can be decoded completely. I located the corrupted place and found the following 7 messages are strange, you can see the timestamp and other data are the same as the previous messages, which means this part of the data was overwritten by the old data.

as I mentioned, "open file->append data(one FIT message)->close file", so from the view of the application level, it is impossible to cause such a problem.

I attached the raw data below, (I copied the corrupted raw data to the txt file from the raw bin file)

below is the raw file.

devzone.nordicsemi.com/.../raw.zip

RE: MSC exception

Einar Thorsrud — Fri, 03 Mar 2023 10:18:04 GMT

Hi Jason,

Have you confirmed that this corruption happened after applying the fix (meaning that the devices in the field were updated before it occurred)?

I see. Can you say more about this data (that is presented here) is stored? And can you upload the corrupt raw file here and explain in detail how you can see that it is corrupt (in case we can se a pattern or similar).

Also, can you say to what extent you are seeing data corruption now with the fixes compared to without it?

Br,

Einar

RE: MSC exception

Jason — Thu, 02 Mar 2023 10:02:50 GMT

Hi，

Unfortunately, we still received a few claims from customers there is a chance that files become corrupt.

this happens in the FW with the latest patch, for sure it also happened in the FW without the latest patch before.

Our product is a bike computer, and it logs the ride live data in FIT format. I recovered one of the corrupted FIT files and attached the corrupted part below. you can see that part of the data was overwritten by the old data. The data inside the red box is the timestamp.

If I understand correctly, there is a cached page in the QSPI block device module, this issue may due to the dirty cached page was not clear somehow.

RE: MSC exception

Einar Thorsrud — Thu, 12 Jan 2023 12:42:53 GMT

Hi,

It is good to hear you are no longer seeing the data corruption. We will continue to look into the formatting issue. I will update here when we have some news.

RE: MSC exception

Jason — Wed, 11 Jan 2023 08:21:16 GMT

Hi Einar，

I tested the patch on the nRF52840 DK by the below method, the patch seems to work well and can not reproduce the corrupted folder issue so far. but the MS Windows still pop up the message "There is a problem with this driver, Scan the driver now and fix it" after I created a new file in the mass storage.

I think it is due to the format function because there is no issue if I format the mass storage via MS Windows.

[quote userid="85932" url="~/f/nordic-q-a/73389/msc-exception/320156"]

I have found a way to reproduce this issue, I tested it on the below project that I attached before.

Jason said:
Please replace the nrf_block_dev_qspi.c of the SDK, I use RTT for logging output and test on the nRF5_SDK_17.0.2.

2626.Project.zip

1. press button 3 to format the mass storage

2. press button 1 to create an ACTIVI folder and TOTALS folder

3. connect the USB to the PC, and create a new file as below

[/quote]

RE: MSC exception

Jason — Tue, 10 Jan 2023 01:30:17 GMT

Hi Eniar,

Thanks for your and your team's effort. It is really good news. We will apply the fix to our project and keep observing.

RE: MSC exception

Einar Thorsrud — Thu, 05 Jan 2023 11:55:59 GMT

Hi,

We have improved the testing and found some issues in the QSPI block device module that may lead to corrupt data. These have been fixed in the attached devzone.nordicsemi.com/.../1134.nrf_5F00_block_5F00_dev_5F00_qspi.c which is for nRF5 SDK 17.x.x. (Replace the existing file with the same name under /components/libraries/block_dev/qspi/).

I am sorry this has taken so long.

RE: MSC exception

Einar Thorsrud — Thu, 16 Jun 2022 06:28:45 GMT

Hi,

Unfortunately not. The team was looking into it, but have not concluded. I will talk to the team and let you know if they have anything to share, but it does not seem like it.

RE: MSC exception

Jason — Wed, 15 Jun 2022 01:02:20 GMT

Hi Einar

Is there any good news on this issue so far?

RE: MSC exception

Einar Thorsrud — Wed, 03 Nov 2021 16:49:20 GMT

Hi,

This has turned out to be a tricky problem. R&D is still looking into it, and it has not been forgotten. Unfortunately I do not have any new information to share yet, though.

RE: MSC exception

Jason — Wed, 03 Nov 2021 04:25:59 GMT

Hi Einar

Is there any progress on this issue?

RE: MSC exception

Einar Thorsrud — Thu, 05 Aug 2021 08:42:52 GMT

Hi,

[quote user="greyorbit"]Yes, it seems so[/quote]

Good. Then we see the same.

[quote user="greyorbit"]Do you mean ejecting the drive like the below picture? if yes, I am not sure, because we always unplug the USB directly.[/quote]

Yes, that is what I mean.

RE: MSC exception

Jason — Thu, 05 Aug 2021 07:27:24 GMT

[quote userid="7377" url="~/f/nordic-q-a/73389/msc-exception/323439#323439"]Corruption only seems to happen after writing to the USB drive from the PC?[/quote]

Yes, it seems so, and it also may happen after formating the mass storage via the f_mkfs function(It seems to make it more likely to happen after formating).

[quote userid="7377" url="~/f/nordic-q-a/73389/msc-exception/323439#323439"]Corruption never seems to happen if properly ejecting the drive before disconnecting it?[/quote]

Do you mean ejecting the drive like the below picture? if yes, I am not sure, because we always unplug the USB directly.