Re: [uml-devel] compile errors with 2.4.19-9: `DMA_MODE_READ' undeclared

[Brian J. M. <cdc...@in...>]

On Mon, Oct 07, 2002 at 10:26:54AM -0400, William Stearns wrote:
> Good morning, Brian,

'Morning William (and Jeff),

> 	I haven't gotten to coffee cup #2 yet, so it's highly likely I'm=20
> being more dense than usual, so my apologies in advance.

Nope, 'tis I that was being dense, or rather perhaps made an unsafe
assumption based on prior experience with UML.

> 	It looks like you're trying to compile support for the localtalk=20
> card into a UML kernel.

I didn't look that closely, but I will defer to your deduction.  I had
incorrectly assumed it was part of the appletalk stack.  But upon
reflection, of course, it's need/want for DMA resources makes it a
hardware driver.

> With a few exceptions (iomem driver, virtual=20
> usb, virtual nic, etc.) UML doesn't address hardware.

Right.  I knew that.  I was being dense in assuming it was part of the
protocol stack and not a driver, but for a reason...

On Mon, Oct 07, 2002 at 10:52:31AM -0500, Jeff Dike wrote:
> c12...@in... said:
> > So, how is ltpc.c supposed to get a definition for DMA_MODE_READ when
> > it's not defined for the "um" architecture?=20
>=20
> ltpc.c is a hardware driver.  Why do you think it has any chance of runni=
ng
> in UML?

I didn't.  My past experience with UML had led me to believe that
during kernel configuration, UML did a good job of weeding out
hardware devices (heck, the .config is 25% the size of my disto's
=2Econfig!).  It seems like this one did not get weeded out.  Any idea
why?  What part of the UML kernel config process does the pruning of
hardware devices for .config?  Give me a pointer and I will go fix it
and bring the fix back here.

b.

--=20
Brian J. Murrell