WinCE6 TFAT hive register gets reset sometimes

Hi all,

I have a probem with our CE6 devices.

Im using TFAT, Stratad and hive-based registry and sometimes the registry
gets reset back to default (ROM registry).
I read about the registry in PB help and it sais that if ROM registry get
changed the hive's are invalid and CE recreates the hive. But the ROM is
untouched.

The filesystem seems OK. I have not seems any filesystem problem.

Are there other situations when CE decides to recover the hive?

Its very hard to reproduce and it seems that some units have a bigger
frequency of faults then others. These units have been used for years thou
and ive have run flash and RAM tests without any errors.


Any suggestions? :)


Best regards

foolblah
Tue Apr 29 05:40:02 PDT 2008

I have had similar problems with TFAT on CE 5.0 PXA270 w/2x256Mb P30 flash.
I too found it unreliable to reproduce, but I was able to debug to the point
of finding that something with the filesystem caused the hive(s) to become
corrupt which either caused them to be restored from ROM or cause the boot
process to hang.

Did anything with your hardware or target change, has this always been a
problem, or did it just start occuring out of the blue?

Can you tell us a little about your platform? How often is the target cold
booted, does the target typically get totally powered down, or does it go
into a power-save mode and then restore where power is always present?

From the testing I have done, I would not advise anyone to use the FMD
driver, especially if the target is subject to having power removed at any
time.

FWIW, I am running the latest QFEs for CE 5.0.

"Bjorn Hall" wrote:

> Hi all,
>
> I have a probem with our CE6 devices.
>
> Im using TFAT, Stratad and hive-based registry and sometimes the registry
> gets reset back to default (ROM registry).
> I read about the registry in PB help and it sais that if ROM registry get
> changed the hive's are invalid and CE recreates the hive. But the ROM is
> untouched.
>
> The filesystem seems OK. I have not seems any filesystem problem.
>
> Are there other situations when CE decides to recover the hive?
>
> Its very hard to reproduce and it seems that some units have a bigger
> frequency of faults then others. These units have been used for years thou
> and ive have run flash and RAM tests without any errors.
>
>
> Any suggestions? :)
>
>
> Best regards

BjornHall
Tue Apr 29 05:49:00 PDT 2008

No hardware changes. We have been running WinCE4.2 on those devices for a
couple of years but with Ipsm.

Our hardware is PXA270 based with 32MB Intel J3 flashes and 64MB RAM.

The problem has been known since we ported to CE6 but have been so rare that
I, at first, ignored it. But with thousands of units out in public we fear it
will be a big problem.

I feel that the TFAT and stratad that comes with CE6 is much more stable
then before.

Our platform goes into deep-sleep when powered down.

The problem occurs even if no register or FS writes are done.

foolblah
Tue Apr 29 06:13:03 PDT 2008

You are positive that the filesystem is indeed running TFAT? For TFAT to be
enabled a format needs to take place. An easy way to check is to download a
debug image to the target from PB and do a search in the debug output for
"TFAT". You should see something like "TFAT enabled = TRUE".

I will tell you that iPSM is a much more stable driver than FMD. I have
just spent over a week doing stability tests on the latest FMD with QFE
updates for CE 5.0 against iPSM 4.0 and iPSM wins hands down. I wrote a test
app that is launched at CE boot that reads/writes/verifies randomly sized
files to the file system and the target's power is cut via hardware. The FMD
cannot handle this abuse but iPSM has had no problems with it.

"Bjorn Hall" wrote:

> No hardware changes. We have been running WinCE4.2 on those devices for a
> couple of years but with Ipsm.
>
> Our hardware is PXA270 based with 32MB Intel J3 flashes and 64MB RAM.
>
> The problem has been known since we ported to CE6 but have been so rare that
> I, at first, ignored it. But with thousands of units out in public we fear it
> will be a big problem.
>
> I feel that the TFAT and stratad that comes with CE6 is much more stable
> then before.
>
> Our platform goes into deep-sleep when powered down.
>
> The problem occurs even if no register or FS writes are done.

BjornHall
Tue Apr 29 07:08:05 PDT 2008

I think that is true pre CE6. Ive run a test with our hardware hooked up to a
plc and automatically rebooted the terminal. After 250000+ reboots the TFAT
is still allright.

Yes, I am sure that it's formated as TFAT.

I don't think its TFAT related, but I can of course be wrong :)

I would like to know why CE decides that the hive is invalid and recreates
it. I can't find the source code for it and I think microsoft have not
released it, is this correct?