:>If your Ethernet card is PCI, set Plug-N-Plag OS to NO and be sure you
:>have enough IRQs for all your PCI cards. If your Ethernet card is ISA,
:>reserve it's IRQ in the BIOS to prevent it from being passed out to a PCI
:>card. Also it your ISA card has plug-n-play option, turn it off and
:>manually set the Address, IRQ, etc yourself.
: PnP is mandatory for PCI devices in the sense that something must allocate
: and reserve IO ports and IRQ levels for each device. Win9x do that on
: its own, bot some other OS relies on the BIOS to do this.
You are horribly and irredeemedly confused. I'll correct you for
posterity and dejanews. PnP ONLY refers to ISA devices. It is a
protocol that attempts to repair the lack of a builtin resource
allocation scheme in the ISA bus spec. Isa+PnP = "win PCI", or
PCI (minus the different addressing) with the resource allocation
done in software or firmware instead of hardware.
So PnP is certainly NOT mandatory for PCI devices! It's got nothing to
do with them. What allocates and reserves irq's and ioports on the PCI
bus IS the pci controller, bus, and devices, acting together. The bios
can restrict and guide these choices - you can select irq by location,
negotiation, and so on on the bios screen. What's more, the pci bus has
it's own interrupts, and they can be mapped in any way you like to the
(ISA) PnP devices can have resources programmed either at bios time or
later. In the latter case, it is the O/S that does it, talking to the
PnP controller. Both Linux and Win95 ARE pnP O/S,s in that they can do
this. Selecting PnP o/s = yes in the bios is the RIGHT Thing to do,
because it means that the bios won't try and program the devices, but
will leave that to the o/s, which is us, and is what we want.
In some rare cases it might be advantageous to permit "PnP o/s = no"
in order to prevent the bios putting another device on a an irq that you
intend later to allocate to a pnp device in Linux. But I can think of
no real concrete example of this. PnP devices like all ISA devices
connect directly to the i86 interrupt lines so it is important to
configure them. An unconfigured card may be physically latched to an
irq line you are trying to use. That's the point in favour of "pnp
o/s = no" in the bios. With "= yes", by the time linux gets to try
and reconfigure the offender, it may be too late to prevent it having
masked another card during the bootup detection sequence. Linux'
detection sequence is before pnp configuration (unless you have pnp
built in to the kernel, which I do not recommend) so pnp cannot play a
role. But that should be irrelevant if you allocated resources
correctly in the bios.
: So if you disable this in BIOS you might see that some PCI devices won't
: work with linux, but works fine with win9x.
I believe the apparent confusion is connected with the explanation