On 2006-04-07, mcharon@gmail.com wrote:

> we are trying write diagnostics application in user space. the
> criteria for the diagnostics is that the testing should be
> done in user space as much as possible. and we need to run the
> application on top of linux. cpu is powerPC.

> memory test. how do you implement memory test from user
> space.

I don't think you do.

> If it's done in user space, there are only 2 options. The
> first option is using mmap but with mmap option, how do you
> know which area of memories are availble for testig, and which
> ones are used by other processes and kernels. The other
> option is to test memory using the virtual address returned by
> malloc. But this way, when you have a failure, how do you
> know which physical memory location is faulty? In both cases,
> do we need to be concerned with paging?
>
>
> 2. Do you also use mmap to access PCI device, pretty much the
> same way that you use mmap to access physical memory?

Yes.

> 3. For certain hardware such as FE, the buffer address
> required by the hardware has to be in physical address form.

I have a special "test driver" for the PCI device that
allocates non-paged test buffer and then allows that to be
mmaped.

> From user space, malloc returns virtual address. How do you
> convert this to physical address so that it can be used by the
> hardware.

The device driver has to do that.

> Also, how do you ensure that no pagin occurs.

No idea what you're asking there.

> 4. last but not least. Interrupt. I think you mentioned this
> to me when we met.

Huh? Where did we meet?

> Do we implement kernel driver to handle hardware interrupt?

Yes.

> This driver will dispatch interrupt to the user app??

Yes.

--
Grant Edwards grante Yow! Oh, I get it!! "The
at BEACH goes on", huh,
visi.com SONNY??