On Tue, May 03, 2022 at 01:08:04PM +0000, David Laight wrote:
> From: Linus Walleij
> > Sent: 01 May 2022 22:56
> > 
> > On Fri, Apr 29, 2022 at 5:37 PM William Breathitt Gray
> > <william.gray@linaro.org> wrote:
> > > On Fri, Apr 29, 2022 at 04:46:00PM +0200, Niklas Schnelle wrote:
> > 
> > > > Good question. As far as I can see most (all?) of these have "select
> > > > ISA_BUS_API" which is "def_bool ISA". Now "config ISA" seems to
> > > > currently be repeated in architectures and doesn't have an explicit
> > > > HAS_IOPORT dependency (it maybe should have one). But it does only make
> > > > sense on architectures with HAS_IOPORT set.
> > >
> > > There is such a thing as ISA DMA, but you'll still need to initialize
> > > the device via the IO Port bus first, so perhaps setting HAS_IOPORT for
> > > "config ISA" is the right thing to do: all ISA devices are expected to
> > > communicate in some way via ioport.
> > 
> > Adding that dependency seems like the right solution to me.
> 
> I think it all depends on what HAS_IOPORT is meant to mean and
> how portable kernel binaries need to be.
> 
> x86 is (probably) the only architecture that actually has 'in'
> and 'out' instructions - but that doesn't mean that some other
> cpu (and I mean cpu+pcb not architecture) have the ability to
> generate 'IO' bus cycles on a specific physical bus.
> 
> While the obvious case is a physical address window that generates
> PCI(e) IO cycles from normal memory cycles it isn't the only one.
> 
> I've used sparc cpu systems that have pcmcia card slots.
> These are pretty much ISA and the drivers might expect to
> access port 0x300 (etc) - certainly that would be right on x86.
> 
> In this case is isn't so much that the ISA_BUS depends on support
> for in/out but that presence of the ISA bus provides the required
> in/out support.

That's true, it does seem somewhat backwards to have a depends on line
when the bus is really just providing the support for devices that want
to use it rather than requiring it. Do you think a HAVE_IOPORT line
should be added independently for each driver instead of adding it to
ISA_BUS?

> Now, maybe, the drivers should be using some ioremap variant and
> then calling ioread8() rather than directly calling inb().
> But that seems orthogonal to this changeset.
> 
> 	David
> 
> -
> Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
> Registration No: 1397386 (Wales)

Using ioremap() does have the benefit of making it easier to reuse the
code for some of these PC104 drivers with their PCI device variants; the
ioread8() calls and such can stay the same and we just initialize to the
proper address during probe. I plan to look into this in the future
then.

William Breathitt Gray