USB drives sometimes hang after a failed sector

[mscotgrove]

When creating a disk image, typically of a disk on it's last legs, I
get lots of failed sectors. Occasionally, and especially on SATA
drives the whole USB - drive interface locks up. The only solution is
a power off reset. I would love to find a software solution around
this problem. Although, not perfect, when the SATA drive is connected
to PC main controller, the failure rate is reduced.

The use of USB interface and extenal box is highly prefered for
logistic reasons, and the type of USB bridge cannot be sepecifed.

Areas I am looking at in my software is what to do after the USB
returns a timeout error. My interface is using SPTI and a
DeviceIoControl command.

I have tried a SCSIOP_SEND_DIAGNOSTIC command, but this is not
recognised by the USB drive/bridge. I do a Test Unit Ready.


Am I correct, there is no way to do in software the equivalent of
poweroff. Just unplugging the USB cable does not help. I therefore
want to try and trap a condition that can cause the system to hang,
before it hangs. Therefore any suggestion on Error Recovery will be
appreciated. Main use is on XP SP3, but same problem happens on
Vista.

Thanks

Michael

 

[mscotgrove]

Linux does try pretty hard to reset the USB interface in these
cases, often successful. Maybe you are using a substandard OS?

Arno- Hide quoted text -

- Show quoted text -

Someone asking and an Irish man how to get somewhere, was told 'I
wouldn't start from here'.

Linux may be far better, but it is not the solution I want.

Michael

 

[Rod Speed]

Someone asking and an Irish man how to get somewhere, was told 'I
wouldn't start from here'.

Linux may be far better, but it is not the solution I want.

But you may well be able to see how it puts the boot into the USB drive and do that yourself.

 

[Arno Wagner]

Someone asking and an Irish man how to get somewhere, was told 'I
wouldn't start from here'.

Linux may be far better, but it is not the solution I want.

Sorry, I should have said this clearer: Linux may be able to do some
things for you, namely help you find out whether what you want is
possible. If Linux can do it, there may be a way to get Windows to
do it too. If Linux fails as well, I would say chances are really
slim. And Linux will give you a lot of verbose log entries when
it encounters the problem that may also help to diagnose the
issue. Of course if you do not want to use Linux at all, you
will have to go without a second measurement. Vista is not really
different enough.

In addition, if this works under Linux, you can see in the logs
what and maybe the driver sources what it did to get it working
again.

Arno