Re: [Qemu-devel] [RFC v4 04/12] isa: Allow to un-assign I/O ports

[Andreas Färber]

Am 09.06.2011 um 17:03 schrieb Markus Armbruster:

> Andreas Färber <address@hidden> writes:
>
> > Signed-off-by: Andreas Färber <address@hidden>
> > ---
> > hw/isa-bus.c |   14 ++++++++++++++
> > hw/isa.h     |    1 +
> > 2 files changed, 15 insertions(+), 0 deletions(-)
> >
> > diff --git a/hw/isa-bus.c b/hw/isa-bus.c
> > index d258932..1f64673 100644
> > --- a/hw/isa-bus.c
> > +++ b/hw/isa-bus.c
> > @@ -105,6 +105,20 @@ void isa_init_ioport(ISADevice *dev, uint16_t  
> > ioport)
> >     isa_init_ioport_range(dev, ioport, 1);
> > }
> >
> > +void isa_discard_ioport_range(ISADevice *dev, uint16_t start,  
> > uint16_t length)
> > +{
> > +    int i, j;
> > +    for (i = 0; i < dev->nioports; i++) {
> > +        if (dev->ioports[i] == start) {
> > +            for (j = 0; j < dev->nioports - i; j++) {
> > +                dev->ioports[i + j] = dev->ioports[i + length + j];
> > +            }
> > +            dev->nioports -= length;
> > +            break;
> > +        }
> > +    }
> > +}
> > +
>
> "discard"?  It's the opposite of isa_init_ioport_range(), name should
> make that obvious.  "uninit"?

"uninit" felt wrong.

> Note: this only affects the I/O-port bookkeeping within ISADevice, not
> the actual I/O port handler registration.  Any use must be accompanied
> by a matching handler de-registration.  Just like any use of
> isa_init_ioport_range() must be accompanied by matching
> register_ioport_FOO()s.  You can thank Gleb for this mess.

Right, I didn't spot any wrong usage though.

So what about cleaning up the mess and doing  
isa_[un]assign_ioport_range(), wrapping the ioport.c functions?
The overhead of calling it up to six times for the different widths  
and read/write would seem negligible as a startup cost. And for  
pc87312 we don't seriously have to care about performance imo.

Andreas