Strange USB thumb drive behavior - not detected after an error

[Michael Bate]

Can anyone help me with this problem?

We have a very simple configuration with a EWF-protected Compact Flash
for the OS and our app (the C: drive) and a writable USB thumb drive.
The CF has an IDE connection. USB is also used for keyboard, mouse, and
touch screen.

The thumb drive will be in an enclosed case and is not customer
serviceable (field engineers may replace it).

The thumb drive is FAT formatted and has plenty of space for our needs.

Sometimes, for no reason that I can figure out, we get a "Delayed Write
Failure" when shutting down the system. On another occasion, the OS
simply did not shut down correctly. After a couple of minutes of viewing
the display "Windows is Shutting Down" (usually this takes about 20
seconds) I powered off the computer.

When either of these happens, the thumb drive is not detected by the
BIOS on the next boot. The computer has been fully powered off and
restarted. The BIOS normally displays a message that it has seen the
thumb drive and even displays information about it (that it is a
"Cruzer"). But on these occasions, it reported nothing and, when the OS
was booted up, "My Computer" displayed no thumb drive. Repeated reboots
failed to solve the problem.

The only way to correct this problem appears to be to pull out the thumb
drive and reinsert it. Then everything works fine.

Our applicaation closes the file on exit. It does not "unmount" the
drive which, I am told, requires use of undocumented Windows API functions.

Thanks in advance for any help.

Michael Bate

 

[Ralph A. Schmid, dk5ras]

The only way to correct this problem appears to be to pull out the thumb
drive and reinsert it. Then everything works fine.

Interl 855GME chipset? This chipset has a hardware bug that shows up
when more then one USB2.0 devices are connected.

Ralph.

 

[KM]

Michael,

I have seen similar problems on one of the boards here. It appeared to be more or less hardware problem (a bug of the USB controller
on that board). Interesting to note here that the bug was repro under XPe and WinCE (we had images of both OSes for that x86 based
device).

That doesn't mean you are running into the same issue but was worth to mention.

To unmount the drive you are NOT required to use an undocumented PI. There is documented API and tools you can use. E.g.,
DeviceIoControl(FSCTL_LOCK_VOLUME), DeviceIoControl(FSCTL_DISMOUNT_VOLUME).
You can grab some code here: http://msdn2.microsoft.com/en-us/library/aa460084.aspx

Another trick. You can do a "chkdsk /X <drive>" and it will force the volume to remount before scanning.

 

[Michael Bate]

Ralph,

Thanks.

We are using an Advantech 9587 with Intel 852GM GMCH/ 6300ESB Chipset.

Hopefully this chipset does not have the same problems as the 855ME.
I'm not an expert on these chipsets and how they differ.

Thanks again,
Michael

 

[Ralph A. Schmid, dk5ras]

Ralph,

Thanks.

We are using an Advantech 9587 with Intel 852GM GMCH/ 6300ESB Chipset.

Hopefully this chipset does not have the same problems as the 855ME.
I'm not an expert on these chipsets and how they differ.

Is the 82801DB or 82801DBM chip present on these boards? Also known as
ICH4... Then I am afraid you have the problem

Exactly for this reason we wanted to go to 945 GM, and now there are
rumours that delivery of this chipset will be stopped soon, due to
licensing problems for some video copy protection routines no one in
the industrial apps range ever uses Does anyone here know more
abiut this?

Thanks again,
Michael

Ralph.

 

[Ralph A. Schmid, dk5ras]

Is the 82801DB or 82801DBM chip present on these boards? Also known as
ICH4... Then I am afraid you have the problem

I have found out that really the 82801DB/ICH4 IC is used in the 852GM
chipset. So there definetely is a hardware issue inside the the chip,
causing influences between different USB2.0 devices on different
poorts; a thing that must not happen, under no circumstances. Intel
says, "ask your hardware supplier", and they deny any responsibility,
they even do not confirm that there is a problem.

Their policy really sucks. Would be another thing if we would buy
those ICs in 100000 counts; but our few dozens of mainboards per year
just seem not to care at all :-(

Ralph.