In Linux it's absolutely forbidden that an I/O address used by a driver
is accessed by another running driver. After an access each driver will
leave an I/O device in a state, unknown to anybody else, that it assumes
not to be modified when it accesses it the next time.

So you need to deactivate (unload) one driver before activating another.

As said already a hundred times, when accessing the DID hardware you
either need to _use_ the driver provided by Linux _or_ not use it at
all: unload it or (maybe better for someone thinking along the lines
you do) not use Linux at all.

-Michael

> Is there a way to temporarily shutdown the running IDE interface and
> resume it once the diagnostics is done?

Compile it as a module in the kernel and unload this module (or don't load
it at all).
Beware: your system cannot be on an IDE disk in that case.

Nutty wrote:

>
> Hello to all,
>
> I am doing diagnostic of HDD on the linux 2.6. The challenge we are
> facing is the synchronization with the existing ide driver. The kernel
> module that we have written directly accesses the ide controller
> registers. The restriction is that we cannot make any changes in the
> currently running ide driver. But how to make sure that while my module
> is writing into a particular sector, the existing driver does not
> interfere?
> Is there a way to temporarily shutdown the running IDE interface and
> resume it once the diagnostics is done?
>
> Thanks & Regards,
> -Nutty

Others have said it - don't do it. Unload the driver first.

If you absolutely must, you _may_ be able to catch
an opportune moment when the IDE driver is idle
and disable interrupts on all processors.

You then wait to make sure that any running DMA transfer
is finished.

At this point, you should be able to access the registers,
provided you can restore the state afterwards exactly.
Don't even think of writing to an active filesystem.

Don't do it.

Kind regards,

Iwo

Iwo,

Has this anything to do with your previous posts?

If so, as previously pointed out IDEs do badblock management
by themselves so doing it again in software is completely
pointless. All that should be required is an IDE device status
and the current driver will already be doing this.


---------------

I m a beginner in the Linux platform. Since last two years,i was
working on Embedded C,C++ on the keil compiler, on the 89C51RD2. I have

also worked on the WinCE 5.0 Platform Builder with EVC++.
According to my new project, i would like to write the diagnostic
application for the PC type device (32 bit processor) i.e i have to
test the functionality of each and every hardware like CPU,HDD,Serial
Ports,RAID,etc.The client has given the procedure to carry out each
test.
As per the requirement the procedure to test HDD should be


# Check HDD
1. Validates the hard-disk device by writing a pattern onto an unused
sector and compares with the expected pattern.
2. Checks all the sectors (HDD surface test) and report bad sectors.
3. Retains the original sector contents.

Can anyone please help me to start the application.

Please guide me...




Iwo Mergler wrote:

> Nutty wrote:
>
> >
> > Hello to all,
> >
> > I am doing diagnostic of HDD on the linux 2.6. The challenge we are
> > facing is the synchronization with the existing ide driver. The kernel
> > module that we have written directly accesses the ide controller
> > registers. The restriction is that we cannot make any changes in the
> > currently running ide driver. But how to make sure that while my module
> > is writing into a particular sector, the existing driver does not
> > interfere?
> > Is there a way to temporarily shutdown the running IDE interface and
> > resume it once the diagnostics is done?
> >
> > Thanks & Regards,
> > -Nutty