Adam Belay wrote:

> Here's the patch.  Any testing would be appreciated.

Tested with two more ISA-Pnp soundcards, ES1868 and OPTi 82c933, and
their ALSA drivers, snd-es18xx and snd-opti93x, and both work the same
as they do without the patch (not quite right that is, but nothing to do
with PnP). Also tested with ISA-PnP IDE (on the ES1868), ISA-PnP NE2000
(RTL8019), and ISA-PnP modem. All fine.

Minor point about the patch itself though. In pnp.h, you do:

> -#define pnp_port_valid(dev,bar) (pnp_port_flags((dev),(bar)) &
> IORESOURCE_IO)
> +#define pnp_port_valid(dev,bar) \ +
> ((pnp_port_flags((dev),(bar)) & IORESOURCE_IO) && \ +
> !(pnp_port_flags((dev),(bar)) & IORESOURCE_UNSET))

and the same for mem,irq,dma. It seems you could roll these two tests
into one with:

#define pnp_port_valid(dev,bar) \
((pnp_port_flags((dev),(bar)) & (IORESOURCE_IO | IORESOURCE_UNSET)) ==
IORESOURCE_IO)

Basically just an optimisation I guess (just checked and gcc doesn't do
this itself) but this also stops the macro arguments from being accessed
more than once.

One more point, in isapnp/core.c:isapnp_set_resources()

> -	for (tmp = 0; tmp < PNP_MAX_PORT && res->port_resource[tmp].flags & IORESOURCE_IO; tmp++)
> +	for (tmp = 0; tmp < PNP_MAX_PORT && !(res->port_resource[tmp].flags & IORESOURCE_UNSET); tmp++)

and again same for mem,irq,dma. That is, it goes from only checking
IORESOURCE_<TYPE> to only checking IORESOURCE_UNSET. Also checking for
the type does sound like a valid sanity check, so would it be better to
also check both flags here? Ie:

for (tmp = 0; tmp < PNP_MAX_PORT && (res->port_resource[tmp].flags &
(IORESOURCE_IO | IORESOURCE_UNSET)) == IORESOURCE_IO; tmp++)

Incremental patch attached, in case you agree. Compiled, booted and tested.

Rene.