The Linux Plug-and-Play-HOWTO
 David S.Lawyer    <mailto:[email protected]>
 v0.05, August 1999

 Help with understanding and dealing with the complex Plug-and-Play
 issue.  How to get your Linux system to support Plug-and-Play.
 ______________________________________________________________________

 Table of Contents

 1. Introduction

    1.1 Copyright, Trademarks, Disclaimer, & Credits
       1.1.1 Copyright
       1.1.2 Trademarks
       1.1.3 Disclaimer
    1.2 Future Plans; You Can Help
    1.3 New Versions of this HOWTO

 2. What PnP Should Do: Allocate "Resources"

    2.1 What is Plug-and-Play (PnP)?
    2.2 How a Computer Finds Devices (and conversely)
    2.3 I/O Addresses, etc.
    2.4 IRQs --Overview
    2.5 DMA Channels
    2.6 Memory Ranges
    2.7 "Resources" to both Device and Driver
    2.8 The Problem
    2.9 PnP Finds Devices Plugged Into Serial Ports

 3. The Plug-and-Play (PnP) Solution

    3.1 Introduction to PnP
    3.2 How It Works (simplified)
    3.3 Starting Up the PC
    3.4 Buses
    3.5 Linux Needs to Cope Better with PnP

 4. Configuring a PnP BIOS

    4.1 Do you have a PnP operating system?
       4.1.1 Interoperability with Windows9x
    4.2 How are resources to be controlled?
    4.3 Reset the configuration?

 5. How to Deal with PnP Cards

    5.1 Introduction to Dealing with PnP Cards
    5.2 Disable PnP ?
    5.3 BIOS Configures PnP
       5.3.1 Intro to Using the BIOS to Configure PnP
       5.3.2 The BIOS's ESCD Database
       5.3.3 Using Windows to set the ESCD
       5.3.4 Adding a New Device (under Linux or Windows)
    5.4 Isapnp (part of isapnptools)
    5.5 PCI Utilities
    5.6 Patch the Kernel to Make Linux PnP
    5.7 Windows Configures
    5.8 Device Driver Configures
    5.9 PnP Software/Documents

 6. Tell the Driver the Configuration

    6.1 Introduction
    6.2 Serial Port Driver: setserial
    6.3 Sound Card Drivers
       6.3.1 OSS-Lite
       6.3.2 OSS (Open Sound System) and ALSA

 7. What Is My Current Configuration?

    7.1 How Are My Device Drivers Configured?
    7.2 How Are My Hardware Devices Configured?

 8. Appendix
    8.1 Addresses
       8.1.1 ISA Bus Configuration Address (Read-Port etc.)
       8.1.2 Address ranges
       8.1.3 Address space
       8.1.4 Range Check (ISA Testing for IO Address Conflicts)
       8.1.5 Communicating Directly via Memory
    8.2 Interrupts --Details
    8.3 PCI Interrupts
    8.4 Isolation

 ______________________________________________________________________

 1.  Introduction

 1.1.  Copyright, Trademarks, Disclaimer, & Credits

 1.1.1.  Copyright

 Copyright (c) 1998, 1999 by David S. Lawyer. Please freely copy and
 distribute (sell or give away) this document.  For corrections and
 minor changes contact the maintainer.  Otherwise you may create
 derivative works and distribute them provided you:

 1. Discuss it with the maintainer (if there is one).  2. Put the
 derivative work at the mirrored LDP Internet site (or the like) for
 free downloading.  3. License the work in the spirit of this license
 or use GPL.  4. Give due credit to previous authors and major
 contributors.

 1.1.2.  Trademarks

 If certain words are trademarks, the context should make it clear to
 whom they belong.  For example "MS Windows" (or just "Windows")
 implies that "Windows" belongs to Microsoft.

 1.1.3.  Disclaimer

 Much of the info in this HOWTO was obtained from the Internet,
 implications in books that may be obsolete, etc.  While I haven't
 intentionally tried to mislead you, there are likely a number of
 errors in this document.  Please let me know about them.  Since this
 is free documentation, it should be obvious that neither I nor
 previous authors can be held legally responsible for any errors.

 1.2.  Future Plans; You Can Help

 Please let me know of any errors in facts, opinions, logic, spelling,
 grammar, clarity, links, etc.  But first, if the date is over a month
 old, check to see that you have the latest version.  Please send me
 any info that you think belongs in this document.

 I haven't studied in detail neither isapnptools nor David Howells'
 patches to the kernel (but I plan to).  Nor do I fully understand how
 PnP is configured by the BIOS (it depends on which BIOS) nor how
 Windows9x updates the ESCD.  Thus this HOWTO is still incomplete and
 may be inaccurate (let me know where I'm wrong).  In this HOWTO I've
 sometimes used ??  to indicate that I don't really know the answer.

 1.3.  New Versions of this HOWTO

 New versions of the Plug-and-Play-HOWTO should appear every month or
 so and will be available to browse and/or download at LDP mirror
 sites.  For a list of mirror sites see:
 <http://metalab.unc.edu/LDP/mirrors.html>.  Various formats are
 available.  If you only want to quickly check the date of the latest
 version look at:  <http://metalab.unc.edu/LDP/HOWTO/Plug-and-Play-
 HOWTO.html>.

 2.  What PnP Should Do: Allocate "Resources"

 2.1.  What is Plug-and-Play (PnP)?

 Oversimplified, Plug-and-Play automatically tells the software (device
 drivers) where to find various pieces of hardware (devices) such as
 modems, network cards, sound cards, etc.  Plug-and-Play's task is to
 match up physical devices with the software (device drivers) that
 operates them and to establish channels of communication between each
 device and its driver.  In order to achieve this, PnP allocates the
 following "resources" to both drivers and hardware: I/O addresses,
 IRQ's, DMA channels (ISA bus only), and memory regions.  If you don't
 understand what these 4 items are read the following subsections.
 Once these resources have been assigned (and if the correct driver is
 installed), the names for such devices in the /dev directory are ready
 to use.

 This PnP assignment of certain resources is sometimes called
 "configuring" but it is only a low level type of configuring.  Even
 with PnP fully utilized, much configuring of devices is done by other
 than PnP.  For example, for modem configuration an "init string" is
 sent to the modem over the I/0 address "channel".  This "init string"
 has nothing to do with PnP although the "channel" used to send it to
 the modem was allocated by PnP.  Setting the speed (and many other
 parameters) of a serial port is done by sending messages to the device
 driver from programs run by the user (often automatically boot-time).
 This configuring also has nothing to do with PnP.  Thus when talking
 about PnP, "resources" means only a limited subset of resources and
 "configuring" means only a certain type of configuring.

 2.2.  How a Computer Finds Devices (and conversely)

 A computer consists of a CPU/processor to do the computing and memory
 to store programs and data.  In addition, there are a number of
 devices such as various kinds of disk-drives, a video card, a
 keyboard, network cards, modem cards, sound cards, serial and parallel
 ports, etc.  There is also a power supply to provide electric energy,
 various buses on a motherboard to connect the devices to the CPU, and
 a case to put all this into.

 In olden days most all devices had their own plug-in cards (printed
 circuit boards).  Today, in addition to plug-in cards, many "devices"
 are small chips permanently mounted on the "motherboard".  Cards which
 plug into the motherboard may contain more than one device.  Memory
 chips are also sometimes considered to be devices but are not plug-
 and-play in the sense used in this HOWTO.

 For the computer system to work right, each device must be under the
 control of its "device driver".  This is software which is a part of
 the operating system (or a module) and runs on the CPU.  Device
 drivers are associated with "special files" in the /dev directory
 although they are not really files.  They have names such as hda1
 (first partition on hard drive a), ttyS0 (the first serial port), eth1
 (the second ethernet card), etc.  To make matters more complicated,
 the particular device driver selected, say for eth1, will depend on
 the type of ethernet card you have.  Thus eth1 can't just be assigned
 to any ethernet driver.  It must be assigned to a certain driver that
 will work for the type of ethernet card you have installed.  To
 control a device, the CPU (under the control of the device driver)
 sends commands (and data) to and reads info from the various devices.
 In order to do this each device driver must know the address of the
 device it controls.  Knowing such an address is equivalent to setting
 up a communication channel, even though the physical "channel" is
 actually the data bus inside the PC which is shared with almost
 everything else.

 The communication channel is actually a little more complex than
 described above.  An "address" is actually a range of addresses and
 there is a reverse part of the channel (known as interrupts) which
 allows devices to send an urgent "help" request to their device
 driver.

 2.3.  I/O Addresses, etc.

 PC's have 3 address spaces: I/O, main memory, and configuration (only
 on the PCI bus).  All of these 3 types of addresses share the same
 address bus inside the PC.  But the voltage on certain dedicated wires
 tells all devices which "space" an address is in: I/O, main memory, or
 configuration.  See ``Addresses'' for more details.  Devices were
 normally located in I/O address space although today they may use
 space in main memory.  An I/0 address is sometimes just called "I/O",
 "IO", "i/o" or "io".  The term "I/O port" also used.  There are two
 main steps to allocate the I/O addresses (or other resources such as
 interrupts):

 1. Set the I/O address, etc. on the card (in one of its registers)

 2. Let its device driver know what this I/O address, etc. is

 The two step process above is something like the two part problem of
 finding someone's house number on a street.  You must obtain (and
 write down) the house number and someone must install a number on the
 front of the house so that it may be found.  In computers, the device
 driver must obtain the address and the device hardware must get the
 same address set in one of its registers.  Both of these must be done,
 but some people make the mistake of doing only one of these and then
 wonder why the computer can't find the device.  For example, they will
 use "setserial" to assign an address to a serial port without
 realizing that this only tells the driver the address.  It doesn't set
 the address in the serial port hardware itself.  If you told the
 driver the wrong address, you're in trouble.

 Another obvious requirement is that the I/O address must be set on the
 card before the device driver tries to use this address.  Since device
 drivers often start up soon after you start the computer, they
 sometimes try to access a card (to see if it's there, etc.) before the
 address has been set in the card by a PnP configuration program.  Then
 you see an error message that they can't find the card even though
 it's there (but doesn't yet have an address).

 What was said in the last 2 paragraphs regarding I/O addresses applies
 with equal force to other resources: ``IRQs --Overview'', ``DMA
 Channels'', and ``Memory Regions''.  What theses are will be explained
 in the next 3 sections.

 2.4.  IRQs --Overview

 After reading this you may read ``Interrupts --Details'' for some more
 details.  The following is intentionally oversimplified:  Besides the
 address, there is also an interrupt number to deal with (such as IRQ
 5).  It's called an IRQ (Interrupt ReQuest) number.  We already
 mentioned above that the device driver must know the address of a card
 in order to be able to communicate with it.  But what about
 communication in the opposite direction?  Suppose the the device needs
 to tell its device driver something immediately?  For example, the
 device may have just received a lot of bytes destined for main memory
 and the device needs to call its driver to fetch these bytes at once
 and transfer them from the device's nearly full buffer into main
 memory.

 How should the device call for help?  It can't use the main data bus
 since it's likely already in use.  Instead it puts a voltage on an
 dedicated interrupt wire (part of the bus) which is often reserved for
 that device alone.  This signal is called an interrupt.  There are the
 equivalent of 16 such wires in a PC and each wire leads (indirectly)
 to a certain device driver.  Each wire has a unique IRQ (Interrupt
 ReQuest) number.  The device must put its interrupt on the correct
 wire and the device driver must listen for the interrupt on the
 correct wire.  Which wire it's put on is determined by the IRQ number
 stored in the device.  This same IRQ number must be known to the
 device driver so that the device driver knows which IRQ line to listen
 to.

 Once the device driver gets the interrupt (a call for help) it must
 find out why the interrupt was issued and take appropriate action to
 service the interrupt.  On the ISA bus each device needs its own
 unique IRQ number.  For the PCI bus and other special cases the
 sharing of IRQs is allowed.

 2.5.  DMA Channels

 DMA channels are only for the ISA bus.  DMA stands for "Direct Memory
 Access".  This is where a device is allowed to take over the main
 computer bus from the CPU and transfer bytes directly to main memory.
 Normally the CPU would make such a transfer in a two step process: 1.
 reading from the I/O memory space of the device and putting these
 bytes into the CPU itself  2. writing these bytes from the CPU to main
 memory.  With DMA it's usually a one step process of sending the bytes
 directly from the device to memory.  The device must have such
 capabilities built into its hardware and thus not all devices can do
 DMA.  While DMA is going on the CPU can't do too much since the main
 bus is being used by The DMA transfer.

 The PCI bus doesn't really have any DMA but instead it has something
 even better: bus mastering.  It works something like DMA and is
 sometimes called DMA (for example, hard disk drives that call
 themselves "UltraDMA").  It allows devices to temporarily become bus
 masters and to transfer bytes almost like the bus master was the CPU.
 It doesn't use any channel numbers since the organization of the PCI
 bus is such that the PCI hardware knows which device is currently the
 bus master and which device is requesting to become a bus master.
 Thus there is no allocation of DMA channels for the PCI bus.

 When a device on the ISA bus wants to do DMA it issues a DMA-request
 using dedicated DMA request wires much like an interrupt request.  DMA
 actually could have been handled by using interrupts but this would
 introduce some delays so it's faster to do it by having a special type
 of interrupt known as a DMA-request.  Like interrupts, DMA-request are
 numbered so as to identify which device is making the request.  This
 number is called a DMA-channel.  Since DMA transfers all use the main
 bus (and only one can run at a time) they all actually use the same
 channel but the "DMA channel" number serves to identify who is using
 the "channel".  Hardware registers exist on the motherboard which
 store the current status of each "channel".  Thus in order to issue a
 DMA-request, the device must know its DMA-channel number which must be
 stored in a register on the physical device.

 2.6.  Memory Ranges

 Some devices are assigned address space in main memory.  It's often
 "shared memory" or "memory-mapped I/O".  Sometimes it's ROM memory on
 the device.  When discussing PnP resources it's often just called
 "memory".  Such a device might also use I/O address space.

 When you plug in such a card, you are in effect also plugging in a
 memory module for main memory.  This memory can either be ROM (Read
 Only Memory) or shared memory.  Shared memory is shared between the
 device and the CPU (running the device driver).  This memory can serve
 as a means of direct data "transfer" between the device and main
 memory.  It's not really a transfer since the device puts data into
 its own memory on its card which also happens to be in main memory.
 Both the card and the device driver need to know where it is.  The
 memory address are likely to be very high so that they do not conflict
 with the lower addresses of the memory chips in your computer.

 ROM is different.  It is likely a program (perhaps a device driver)
 which will be used with the device.  Hopefully, it may work with Linux
 and not just Windows ??  It may need to be shadowed which means that
 it is copied to your main memory chips in order to run faster.  Once
 it's shadowed it's no longer "read only".

 2.7.  "Resources" to both Device and Driver

 Thus device drivers must be "attached" in some way to the hardware
 they control.  This is done by supplying "resources" (I/O, Memory,
 IRQ's, DMA's) to both the physical device and the device driver
 software.  For example, a serial port uses only 2 (out of 4 possible)
 resources: an IRQ and an I/O address.  Both of these values must be
 supplied to the device driver and the physical device.  The driver
 (and its device) is also given a name in the /dev directory (such as
 ttyS1).  The address and IRQ number is stored by the physical device
 in registers on the card (or in a chip on the motherboard).  For the
 case of jumpers, this info is always stored in the device hardware (on
 the card, etc.).  But for the case of PnP, the register data is
 usually lost when the PC is powered down (turned off) so that the
 resource data must be supplied to each device anew each time the PC is
 powered on.

 2.8.  The Problem

 The architecture of the PC provides only a limited number of IRQ's,
 DMA channels, I/O address, and memory regions.  If there were only
 several devices and they all had standardized resource (such as unique
 I/O addresses and IRQ numbers) there would be no problem of attaching
 device drivers to devices.  Each device would have a fixed resources
 which would not conflict with any other device on your computer.  No
 two devices would have the same addresses, there would be no IRQ
 conflicts, etc.  Each driver would be programmed with the unique
 addresses, IRQ, etc. hard-coded into the program.  Life would be
 simple.

 But it's not.  Not only are there so many different devices today that
 conflicts are frequent, but one sometimes needs to have more than one
 of the same type of device.  For example, one may want to have a few
 different disk-drives, a few serial ports, etc.  For these reasons
 devices need to have some flexibility so that they can be set to
 whatever address, IRQ, etc. is needed to avoid conflicts.  But some
 IRQ's and addresses are pretty standard such as the ones for the clock
 and keyboard.  These don't need such flexibility.

 Besides the problem of conflicting allocation of resources, there is a
 problem of making a mistake in telling the device driver what the
 resources are.  For example, suppose that you enter IRQ 4 in a
 configuration file when the device is actually set at IRQ 5.  This is
 another type of resource allocation error.

 The allocation of resources, if done correctly, establishes channels
 of communication between physical hardware and their device drivers.
 For example, if a certain I/O address range (resource) is allocated to
 both a device driver and a piece of hardware, then this has
 established a one-way communication channel between them.  The driver
 may send commands and info to the device.  It's actually a little more
 than one-way since the driver may get information from the device by
 reading it's registers.  But the device can't initiate any
 communication this way.  To initiate the device needs an IRQ in order
 to create a two-way communication channel where both the driver and
 the device can initiate communication.

 2.9.  PnP Finds Devices Plugged Into Serial Ports

 External devices that connect to the serial port via a cable (such as
 external modems) can also be called Plug-and-Play.  Since only the
 serial port itself needs resources (an IRQ and I/O address) there are
 no resources to allocate to such plug-in devices.  Thus PnP is not
 really needed for them.  Even so, there is a PnP specification for
 such external serial devices.

 A PnP operating system will find such an external device and read its
 model number, etc.  Then it may be able to find a device driver for it
 so that you don't have to tell an application program that you have a
 certain device on say /dev/ttyS1.  Since you should be able to
 manually inform your application program (via a configuration file,
 etc.) what serial port the device is on (and possibly what model
 number it is) you should not really need this "serial port" feature of
 PnP.

 3.  The Plug-and-Play (PnP) Solution

 3.1.  Introduction to PnP

 The term Plug-and-Play (PnP) has various meanings.  In the broad sense
 it is just auto-configuration where one just plugs in a device and it
 configures itself.  In the sense used in this HOWTO, the configuration
 is only that of configuring PnP resources and letting the device
 drivers know about it.  In a more narrow sense it just setting
 resources in the hardware devices.  It may also mean the PnP
 specifications which (among other things) specify how PnP resource
 data is to be read and written to devices (often cards) on the ISA
 bus.  The standard PCI (and not PnP) specifications do the same for
 the PCI bus.

 PnP matches up devices with their device drivers and specifies their
 communication channels.  On the ISA bus before Plug-and-Play the
 resources were set in hardware devices by jumpers.  Software drivers
 were assigned resources by configuration files (or the like) or by
 probing the for the device at addresses where it's expected to reside.
 The PCI bus was PnP-like from the beginning so it was trivial to
 implement PnP for this bus.  Since the PCI bus specifications don't
 use the term PnP it's not clear whether or not the PCI bus should be
 called PnP (but it supports in hardware what today is called PnP).

 3.2.  How It Works (simplified)

 Here's an oversimplified view of how PnP works.  The PnP configuration
 program (perhaps a program in the BIOS) finds all PnP devices and asks
 each what resources it needs.  Then it checks what resources (IRQs,
 etc.) it has to give away.  Of course if it has reserved resources
 used by non-PnP (legacy) devices (if it knows about them) it doesn't
 give these away.  Then it uses some criteria (not specified by PnP
 specifications) to give out the resources so that there are no
 conflicts and so that all devices get what they need (if possible).
 It then tells each physical device what resources are assigned to it
 and the devices set themselves up to use only the assigned resources.
 Then the device drivers somehow find out what resources their devices
 use and are thus able to communicate effectively with the devices they
 control.

 For example, suppose a card needs one interrupt (IRQ number) and 1 MB
 of shared memory.  The PnP program reads this request from the card.
 It then assigns the card IRQ5 and 1 MB of memory addresses space,
 starting at address 0xe9000000.  It's not always this simple as the
 card may specify that it can only use certain IRQ numbers (ISA only)
 or that the 1 MB of memory must lie within a certain range of
 addresses.  The details are different for the PCI and ISA buses with
 more complexity on the ISA bus.

 There are some shortcuts that PnP software may use.  One is to keep
 track of how it assigned resources at the last configuration (when the
 computer was last used) and reuse this.   Windows9x and PnP BIOSs do
 this but standard Linux doesn't.  Windows9x stores this info in its
 "Registry" and a PnP BIOS stores it in non-volatile memory in your PC
 (known as ESCD; see ``The BIOS's ESCD Database'').

 Under Linux it's each device for itself and there is no centralized
 registry of resource assignments.  Some device drivers store the last
 configuration they used and use it next time the computer is powered
 on.  They implicitly assume that the hardware will be somehow
 configured with the same resources.

 If the device hardware remembered their previous configuration, then
 there wouldn't be any hardware to configure at the next boot-time, but
 they seem to forget their configuration when the power is turned off.
 Some devices contain a default configuration (but not necessarily the
 last one used).  Thus a PnP configuration program needs to be run each
 time the PC is powered on.  Also, if a new device has been added, then
 it too needs to be configured.  Allocating resources to this new
 device might involve taking some resources away from an existing
 device and assigning the existing device alternative resources that it
 can use instead.

 3.3.  Starting Up the PC

 When the PC is first turned on the BIOS chip runs its program to get
 the computer started (the first step is to check out the hardware).
 If the operating system is stored on the hard-drive (as it normally
 is) then the BIOS must know about the hard-drive.  If the hard-drive
 is PnP then the BIOS may use PnP methods to find it.  Also, in order
 to permit the user to manually configure the BIOS's CMOS and respond
 to error messages when the computer starts up, a screen (video card)
 and keyboard are also required.  Thus the BIOS must PnP-configure
 these devices on its own.
 Once the BIOS has identified the hard-drive, the video card, and the
 keyboard it is ready to start booting (loading the operating system
 into memory from the hard-disk).  If you've told the BIOS that you a
 have a PnP operating system (PnP OS), it should start booting the PC
 as above and let the operating system finish the PnP configuring.
 Otherwise, a PnP-BIOS will (prior to booting) likely try to do the
 rest of the PnP configuring of devices (but not their drivers).

 3.4.  Buses

 ISA is the old bus of the old IBM PC's while PCI is a newer and faster
 bus from Intel.  The PCI bus was designed for what is today called
 PnP.  It makes it easy (as compared to the ISA bus) to find out how
 PnP resources have been assigned to hardware devices.  To see what has
 happened look at the /proc/pci "file" (/proc/bus/pnp/devices for
 kernel 2.2+), the boot-up messages on your display (use shift-PageUp
 to back up), or use PCI Utilities (for kernel 2.2+).

 For the ISA bus there is a real problem with implementing PnP since no
 one had PnP in mind when the ISA bus was designed and there are almost
 no I/O addresses available for PnP to use for sending configuration
 info to physical device.  As a result, the way PnP was shoehorned onto
 the ISA bus is very complicated.  A whole book has been written about
 it.  See ``PnP Book''.  Among other things, it requires that each PnP
 device be assigned a temporary "handle" by the PnP program so that one
 may address it for PnP configuring.  Assigning these "handles" is call
 "isolation".  See ``Isolation'' for the complex details.

 Eventually, the ISA bus should become extinct.  When it does, PnP will
 be easier since it will be easy to find out how the BIOS has
 configured the hardware.  There will still be the need to match up
 device drivers with devices and also a need to configure devices that
 are added when the PC is up and running.  These needs would be
 satisfied if Linux was a PnP operating system.

 3.5.  Linux Needs to Cope Better with PnP

 PnP (for the ISA bus) was invented by Compaq, Intel, and Phoenix.
 Microsoft has been a leading promoter of it.  Linux would have been
 better off if PnP had never been "invented".  Eventually the ISA bus
 will have become extinct and the PnP-like PCI bus will prevail so that
 we will have in effect gotten an easy-to-implement PnP.  But like it
 or not, most all new ISA hardware today is PnP and Linux has no choice
 but to deal effectively with PnP.  But standard Linux (as of early
 1999) makes dealing with PnP complicated (especially on the ISA bus)
 while the purpose of PnP was to make it simple.

 In a sense, Linux is already somewhat PnP for the PCI bus.  When the
 PC starts up you may note from the messages on the screen that some
 Linux device drivers often find their hardware devices (and the
 resources the BIOS has assigned them).  But there are situations that
 a PnP operating system could handle better: 1. A shortage of resources
 2. More than one driver for a physical device  3. An activated driver
 which can't find it's physical device 4. Hot installation of a device
 (docking, etc.)

 Linux users should not need to delve into the details of PnP to
 configure ISA PnP devices as they now need to.  One solution would be
 a standardized version of the Linux kernel that supports Plug-and-Play
 on the ISA, PCI, and other buses.  A patch to the kernel has been
 written although most drivers don't support it.  It's not part of
 standard Linux.  See ``Patch Kernel''.

 4.  Configuring a PnP BIOS

 When the computer is first turned on, the BIOS runs before the
 operating system is loaded.  Newer BIOSs are PnP and will configure
 some or all of the PnP devices.  For most PnP BIOSs there is no way to
 disable PnP so you have to live with it.  Here are some of the choices
 which may exist in your BIOS's CMOS menu:

 �  ``Do you have a PnP operating  system?''

 �  ``How are resources to be controlled?''

 �  ``Reset the configuration?''

 4.1.  Do you have a PnP operating system?

 If you say yes, then the PnP BIOS will PnP-configure the hard-drive,
 video card, and keyboard to make the system bootable.  But it will
 leave it up to the operating system to finish the configuration job.
 It may do an ``Isolation'' on the ISA bus leaving the devices disabled
 but ready to be configured by the operating system.  For Linux you
 should probably tell it that you don't have a PnP operating system.
 If you don't do this, the BIOS might leave the ISA devices it hasn't
 configured in a disabled state ??  Also PCI devices might not get
 configured ??

 If you tell the BIOS you don't have a PnP OS, then the BIOS will do
 the configuring itself.  Unless you have added new PnP devices, it
 should use the configuration which it has stored in its non-volatile
 memory (ESCD).  See ``The BIOS's ESCD Database'' If the last session
 on your computer was with Linux, then there should be no change in
 configuration.  See ``BIOS Configures PnP''.  But if the last session
 was with Windows9x (which are PnP) then Windows could have set up the
 configuration differently and possibly saved some of it in the ESCD.
 See ``Using Windows to set ESCD''.  If you are using the isapnp or PCI
 Utilities program(s) to do configuring, they will run after the BIOS
 runs and change things the way you told them to.

 4.1.1.  Interoperability with Windows9x

 If you are running Windows on the same PC, how do you answer the
 question: Do you have a PnP OS?  Normally (and truthfully) you would
 say no for standard Linux and yes for Windows9x.  But it's a lot of
 bother to have to set up the CMOS menu manually each time you want to
 switch OSs.  One solution is set the CMOS for no PnP OS, including
 when you start Windows.  One might expect that Windows would be able
 to handle this situation where it is presented hardware that has been
 fully configured by the BIOS.  In addition, one might expect that even
 if Windows didn't realize that the hardware was already configured, it
 would do the configuration itself and then work OK.  But it doesn't
 seem to work this way.  It seems that Windows may just tell its device
 drivers what has been stored in the Windows' Registry.  But the actual
 hardware configuration (done by the BIOS) is what was stored in the
 ESCD and may not be the same as the Registry => trouble.

 One way (the only way??) to try to get the Registry and the ESCD the
 same is to install (or reinstall) Windows when the BIOS is set for
 "not a PnP OS".  This should present Windows with system configured by
 the BIOS (except for the devices drivers).  If this configuration is
 without conflicts, Windows will hopefully leave it alone and save it
 in it's Registry.  If this works for you (and this is the latest
 version of this HOWTO), let me know as I only have one report of this
 working out OK.
 Another thing you might try if there is only a problem with one device
 is to tell Windows you are removing the device (perhaps hide the
 device driver).  Then restart the PC with "not a PnP OS" and install
 the device under Windows.  I have no idea if this works or not.  Let
 me know if it does and what you did (only if this is the latest
 version of this HOWTO).

 4.2.  How are resources to be controlled?

 This may involve just deciding how to allocate IRQ and DMA resources.
 If set to "auto", the BIOS will do the allocation.  If set to manual,
 you manually reserve some IRQ's for use on "legacy" (non-pnp) cards.
 The BIOS may or may not otherwise know about your legacy cards.  The
 BIOS will only know about your legacy cards if you ran ICU (or the
 like) under Windows to tell the BIOS about them.  If the BIOS knows
 about them, then try using "auto".  If it doesn't know about them then
 manually reserve the IRQ's needed for the legacy ISA cards and let the
 rest be for the BIOS PnP to allocate.

 4.3.  Reset the configuration?

 This will erase the BIOSs ESCD data-base of how your PnP devices
 should be configured as well as the list of how legacy (non-PnP)
 devices are configured.  Never do this unless you are convinced that
 this data-base is wrong and needs to be remade.  It was stated
 somewhere that you should do this only if you can't get your computer
 to boot.  If the BIOS loses the data on legacy devices, then you'll
 need to run ICA again under DOS/Windows to reestablish this data.

 5.  How to Deal with PnP Cards

 5.1.  Introduction to Dealing with PnP Cards

 Today most all new internal boards (cards) are Plug-and-Play (PnP).
 Although some software exists in Linux to handle PnP, it is not always
 easy to use.  There are 6 different methods listed below to cope with
 PnP (but some may not be feasible in your situation).  Which one(s)
 you should use depends on your goals.  What may be most expedient to
 do now may not be the easiest and best in the long run.  A seemingly
 simple way is to do nothing and just let a PnP-BIOS configure it but
 then you may need to do some exploring to to find out what the BIOS
 has done.  A comparison of these methods needs to be written by
 someone who has tried them all.  You may need to use more than one
 method to do the job.

 �  ``Disable PnP''  by jumpers or DOS/Windows software (but many cards
    can't do this)

 �  ``BIOS Configures PnP'' (For the PCI bus you only need a PCI BIOS,
    otherwise you need a PnP BIOS)

 �  ``Isapnp'' is a program you can always use to configure PnP devices
    on the ISA bus only

 �  ``PCI Utilities'' is for configuring the PCI bus

 �  ``Windows Configures'' and then you boot Linux from within
    Windows/DOS.  Use as a last resort

 �  ``Patch Kernel'' to transform Linux into a PnP operating system

 �  ``Device Driver Configures'' but few do

 Any of the above will set the resources in the hardware.  But only the
 last two should tell device driver what it's done.  Only the last one
 definitely tells the driver (since it is the driver).  How the driver
 gets informed depends on the driver and you may need to do something
 to inform it.  See ``Tell the Driver the Configuration''

 5.2.  Disable PnP ?

 Many devices are PnP only with no option for disabling PnP.  But for
 some, you may be able to disable PnP by jumpers or by running a
 Windows program that comes with the device (jumperless configuration).
 This will avoid the often complicated task of configuring PnP.  Don't
 forget to tell the BIOS that these resources are reserved.  There are
 also some reasons why you might not want to disable PnP:

 1. If you have MS Windows on the same machine, then you may want to
    allow PnP to configure devices differently under Windows from what
    it does under Linux.

 2. The range of selection for IRQ numbers (or port addresses) etc.
    may be quite limited unless you use PnP.

 3. You might have a Linux device driver that uses PnP methods to
    search for the device it controls.

 4. If you need to change the configuration in the future, it may be
    easier to do this if it's PnP (no setting of jumpers or running a
    Dos/Windows program).

 5. You may have (or will have) other PnP devices that need configuring
    so that you'll need to provide for (or learn about) PnP anyway.

    Once configured as non-PnP devices, they can't be configured by PnP
    software or the BIOS (until you move jumpers and/or use the
    Dos/Windows configuration software again).

 5.3.  BIOS Configures PnP

 5.3.1.  Intro to Using the BIOS to Configure PnP

 If you have a PnP BIOS, it can configure the hardware.  This means
 that your BIOS reads the resource requirements of all devices and
 configures them (allocates resources to them).  It is a substitute for
 a PnP OS except that the BIOS doesn't match up the devices with their
 drivers nor tell the drivers how it has done the configuring.  It
 should give preference to using the configuration it has stored in its
 non-volatile memory (ESCD).

 Your BIOS must support such configuring but there have been cases
 where it doesn't do it correctly or completely.  An advantage of using
 the BIOS is that it's simple since in most cases there is nothing to
 set up (except to tell the BIOS's CMOS menu it's not a PnP OS).  While
 some device drivers may be able to automatically detect what the BIOS
 has done, in some cases you'll need to determine it (not always easy).
 See ``What Is My Current Configuration?'' Another advantage is that
 the BIOS does its work before Linux starts so that all the resources
 are ready to be used (and found) by the device drivers that start up
 later.

 According to MS it's only optional (not required) that a PnP BIOS be
 able to PnP-configure the devices (without help from MS Windows).  But
 it seems that most of the ones made after 1996 ?? or so can do it.  We
 should send them thank-you notes if they do it right.  They configure
 both the PCI and ISA buses, but it has been claimed that some older
 BIOSs can only do the PCI.  To try to find out more about your BIOS,
 look on the Web.  Please don't ask me as I don't have data on this.
 The details of the BIOS that you would like to know about may be hard
 to find (or not available).  Some BIOSs may have minimal PnP
 capabilities and try to turn over the difficult parts of the
 configuration task to Window utilities.  If this happens you'll either
 have to find another method (such as isapnptools) or try to set up the
 ESCD database if the BIOS has one.  See the next section.

 5.3.2.  The BIOS's ESCD Database

 The BIOS's maintains a non-volatile database containing a PnP-
 configuration that it will try to use.  It's called the ESCD (Extended
 System Configuration Data).  Again, the provision of ESCD is optional
 but most PnP-BIOSs have it.  The ESCD not only stores the resource-
 configuration of PnP devices but also stores configuration information
 of non-PnP devices (and marks them as such) so as to avoid conflicts.
 The ESCD data is usually saved on a chip and remains intact when the
 power is off, but sometimes it's put on a hard-drive??

 The ESCD is intended to hold the last used configuration, but if you
 use a program such as Linux's isapnp or pci utilities (which doesn't
 update the ESCD) then the ESCD will not know about what isapnp has
 set.  A good PnP OS might update the ESCD so you can use it later on
 for a non-PnP OS (like standard Linux).  Windows may do this in some
 cases.  See ``Using Windows to set ESCD''.

 To use what's set in ESCD be sure you've set "Not a PnP OS" or the
 like in the BIOS.  Then each time the BIOS starts up (before the Linux
 OS is loaded) it should configure things this way.  If the BIOS
 detects a new PnP card which is not in the ESCD, then it must then
 allocate resources to the card and update the ESCD.  It may even have
 to change the resources assigned to existing PnP cards and modify ESCD
 accordingly.

 If devices saved their last configuration in their hardware, hardware
 configuring wouldn't be needed each time you start your PC.  But it
 doesn't work this way.  So all the ESCD data needs to be kept correct
 if you use the BIOS for PnP.  There are some BIOSs that don't have an
 ESCD but do have some non-volatile memory to store info on which
 resources have been reserved for use by non-PnP cards.  Many BIOSs
 have both.

 5.3.3.  Using Windows to set the ESCD

 If the BIOS doesn't set up the ESCD the way you want it (or the way it
 should be) then it would be nice to have a Linux utility to set the
 ESCD.  As of early 1999 there isn't any.  Thus one may resort to
 attempting to use Windows (if you have it on the same PC) to do this.

 There are two ways to use Windows to try to set/modify the ESCD.  One
 way is to use the ICU utility designed for DOS or Windows 3.x.  It
 should also work OK for Windows 9x/2k ??  Another way is to set up
 devices manually ("forced") under Windows 9x/2k so that Windows will
 put this info into the ESCD when Windows is shut down normally.  If
 devices are configured automatically by Windows (without the user
 telling it to "change setting") the setting will probably not make it
 into the ESCD ??  Of course Windows may well decide on its own to
 configure the same as what is set in the ESCD so they could wind up
 being the same by coincidence.

 Windows 9x are PnP operating systems and automatically PnP-configure
 devices.  They maintain their own PnP-database deep down in the
 Registry (stored in binary Windows files).  There is also a lot of
 other configuration stuff in the Registry besides PnP-resources.
 There is both a current PnP resource configuration in memory and
 another (perhaps about the same) stored on the hard disk.  To look at
 (the one in memory?) in Windows98 you use the Device Manager.

 In Windows98 There are 2 ways to get to the Device Manager: 1. Control
 Panel --> System Properties --> Device Manager.  2. My Computer -->
 Properties --> Device Manager.  Then in Device Manager you select a
 device (sometimes a multi-step process if there are a few devices of
 the same class).  Then click on Properties and then on Resources.  To
 attempt to change the resource configuration manually, uncheck "Use
 automatic settings" and then click on  Change Settings.  It may not
 let you change it.  Also, since Windows may assign IRQs differently
 than Linux, what is a conflict under Windows may not be a conflict
 under Linux and conversely.  When you "force" a change of resources in
 Windows, it should put your change into the ESCD (provided you exit
 Windows normally).

 5.3.4.  Adding a New Device (under Linux or Windows)

 If you add a new PnP device and have the BIOS set to "not a PnP OS",
 then the BIOS should automatically configure it and store the
 configuration in ESCD.  If it's a non-PnP legacy device (or one made
 that way by jumpers, etc.) then there are a few options to handle it.

 You may be able to tell the BIOS directly (via the CMOS setup menus)
 that certain resources it uses (such as IRQs) are reserved and are not
 to be allocated by PnP.  This does not put this info into the ESCD.
 But there may be a BIOS menu selection as to whether or not to have
 these CMOS choices override what may be in the ESCD in case of
 conflict.  Another method is to run ICU under DOS/Windows.  Still
 another is to install it manually under Windows 9x/2k.  Since this
 configuration is "forced" Windows should update the ESCD when you shut
 down the PC.

 5.4.  Isapnp (part of isapnptools)

 Unfortunately, much of the documentation for isapnp is still difficult
 to understand unless you know the basics of PnP.  This HOWTO should
 help you understand it as well the FAQ that comes with it.  isapnp is
 only for PnP devices on the ISA bus (non-PCI).  Running the Linux
 program "isapnp" at boot-time will configure such devices to the
 resource values you set in /etc/isapnp.conf.  Its possible to create
 this configuration file automatically but you then must edit it
 manually to chose between various options.  With isapnp, a device
 driver which is part of the kernel may run too early before isapnp has
 set the address, etc. in the hardware.  This results in the device
 driver not being able to find the device.  The driver trys the right
 address but the address hasn't been set yet in the hardware.

 If your Linux distribution automatically installed isapnptools, isapnp
 may already be running at startup.  In this case, all you need to do
 is to edit /etc/isapnp.conf per "man isapnp.conf".  Note that this is
 like manually configuring PnP since you make the decisions as to how
 to configure as you edit the configuration file.  You can use the
 program "pnpdump" to help create the configuration file.  If you use
 "isapnp" like this and have a PnP BIOS, you should probably tell the
 BIOS (when you set it up) that you don't have a PnP OS since you still
 want the BIOS to configure the PCI devices.  While the BIOS may also
 configure the ISA devices, isapnp will redo it.

 5.5.  PCI Utilities

 The new package PCI Utilities (= pciutils, incorrectly called
 "pcitools"), should let you manually PnP-configure the PCI bus.
 "lspci" lists resources while "setpci" sets resource allocations in
 the hardware devices.

 5.6.  Patch the Kernel to Make Linux PnP

 David Howells has created a patch to do this called "Linux Kernel
 Configuration/Resource Manager" (sometimes called Hardware
 Configuration Manager).  The patch may not be against the most recent
 kernel.  The resulting kernel is is claimed to be stable but bugs have
 been reported.  It includes documentation such as serial.txt to show
 how to deal with the serial port.  It provides "files" in the /proc
 tree so that you can see what is going on and can echo commands into
 one of these files for custom configuration.  One problem is that most
 device drivers don't know about it so that you still may have to use
 the traditional configuration files, etc. for configuration.  The
 webpage for it is  <http://www.astarte.free-online.co.uk>

 5.7.  Windows Configures

 If you have Windows9x (or 2k) on the same PC, then just start Windows
 and let it configure PnP.  Then start Linux from Windows (or DOS).  It
 been reported that Windows erased the IRQs from PCI devices registers.
 Then Linux complained that it found a zero IRQ.  Thus you may not be
 able to use this method.

 5.8.  Device Driver Configures

 A few device drivers will use PnP methods to set the resources in the
 hardware but only for the device that they control.  Since the driver
 has done the configuring, it obviously knows the configuration and
 there is no need for you to tell it this info.

 The problem with this is twofold.  It's difficult to incorporate all
 of this into the driver, and the driver may grab resources that are
 needed by other devices.  It does make it easy for the user but a PnP
 Linux kernel might be better.  See ``Linux Needs to Cope Better with
 PnP''

 5.9.  PnP Software/Documents

 �  Isapnptools homepage <http://www.roestock.demon.co.uk/isapnptools/>

 �  Patch to  make the Linux kernel PnP <http://www.astarte.free-
    online.co.uk>

 �  PnP driver project <http://www.io.com/~cdb/mirrors/lpsg/pnp-
    linux.html>

 �  PnP Specs. from Microsoft
    <http://www.microsoft.com/hwdev/respec/pnpspecs.htm>

 �  Book: PCI System Architecture, 3rd ed. by Tom Shanley +, MindShare
    1995.  Covers PnP-like features on the PCI bus.

 �  Book: Plug and Play System Architecture, by Tom Shanley, Mind Share
    1995.  Details of PnP on the ISA bus.  Only a terse overview of PnP
    on the PCI bus.
 6.  Tell the Driver the Configuration

 6.1.  Introduction

 Just how this is done depends upon the driver.  Some drivers have more
 than one way to find out how their physical device is configured.  At
 one extreme is the case where you must hard-code the resources into
 the kernel and recompile.  At the other extreme, the driver does
 everything automatically and you have nothing to do.  It may even set
 the resources in the hardware using PnP methods.

 In the middle are cases where you run a program to give the resource
 info to the driver or put the info in a file.  In some cases the
 driver may probe for the device at addresses where it suspects the
 device resides.  It may then try to test various IRQs to see which one
 works.   It may or may not automatically do this.  In other cases the
 driver may use PnP methods to find the device and how the resources
 have been set, but will not actually set them.  It may also look in
 some of the files in the /proc directory.

 One may need to give the resources as a parameter to the kernel to to
 a loadable module.  See /usr/lib/modules_help/descr.gz for a list of
 possible parameters.  The module to load is listed in /etc/modules
 along with its parameters.  In some other case the resources may be
 given as parameters to the kernel.  These are put into the lilo.conf
 file as append="...".   Then the lilo program must be run to save this
 in the kernel boot code.

 While there is great non-uniformity about how drivers find out about
 resources, the end goal is the same.  There are so many different
 hardware devices and drivers for them that you may need to look at
 documentation for your driver to find out how it finds out about
 resources and what you need to do to insure that it gets the info it
 needs.  Some brief info on a few drivers is presented in the following
 section.

 6.2.  Serial Port Driver: setserial

 For the standard serial port driver (not for multiport cards) you use
 setserial to configure the driver.  It is often run from a start-up
 file.  In newer versions there is a /etc/serial.conf file that you
 "edit" by simply using the setserial command in the normal way and
 what you set using setserial is saved in the serial.conf configuration
 file.  The serial.conf file should be consulted when the setserial
 command runs from a start-up file.  Your distribution may or may not
 set this up for you.

 There are two different ways to use setserial depending on the options
 you give it.  One way is used to manually tell the driver the
 configuration.  The other way is to probe at a given address and
 report if a serial port exists there.  It can also probe this address
 and try to detect what IRQ is used for this port.  The driver runs
 something like setserial at start-up but it doesn't probe for IRQs, it
 just assigns the "standard" IRQ which may be wrong.  It does probe for
 the existence of a port.  See Serial-HOWTO for more details.

 6.3.  Sound Card Drivers

 6.3.1.  OSS-Lite

 You must give the IO, IRQ, and DMA as parameters to a module or
 compile them into the kernel.  But some PCI cards will get
 automatically detected (likely by looking at /proc/pci or the like).
 RedHat supplies a program "sndconfig" which detects ISA PnP cards and
 automatically sets up the modules for loading with the detected
 resources.

 6.3.2.  OSS (Open Sound System) and ALSA

 These will detect the card by PnP methods and then select the
 appropriate driver and load it.  It will also set the resources on an
 ISA-PnP card.  You may need to manually intervene to avoid conflicts.
 For the ALSA driver, support for ISA-PnP is optional and you may use
 isapnp tools if you want to.

 7.  What Is My Current Configuration?

 Here "configuration" means the assignment of PnP resources (addresses,
 IRQs, and DMAs).  There are two parts to this question for each
 device.  Each part should have the same answer.

 1. What is the configuration of the device driver software?  I.e.:
    What does the driver think the hardware configuration is?

 2. What configuration (if any) is set in the device hardware?

 Of course the configuration of the device hardware and its driver
 should be the same (and it normally is).  But if things are not
 working right, there may be a difference.  This means the the driver
 has incorrect information about the actually configuration of the
 hardware.  This spells trouble.  If the software you use doesn't
 adequately tell you what's wrong (or automatically configure it
 correctly) then you need to investigate how your hardware devices and
 their drivers are configured.  While Linux device drivers should "tell
 all" in some cases it's not easy to determine what has been set in the
 hardware.

 Another problem is that when you view configuration messages on the
 screen, it's sometimes not clear whether the reported configuration is
 that of the device driver, the device hardware, or both.  If the
 device driver is assigned a configuration and then checks the hardware
 out to see if it's configured the same, then the configuration
 reported by the driver should be that of both the hardware and the
 driver.

 But some drivers don't do this may accept a configuration that doesn't
 check out.  For example, "setserial" will accept a configuration that
 doesn't check out (even if you've told it to probe for resources).
 Thus "setserial" may only be telling you the configuration of the
 driver and not the hardware.

 Some info on configuration may be obtained from the messages from the
 BIOSs and Linux that appear on the screen when you first start the
 computer.  After all the  messages have flashed by, type shift-PageUp
 to scroll back to them.  Typing "dmesg" at any time to the shell
 prompt will show only the Linux kernel messages and miss some of the
 most important ones (including ones from the BIOS).  The messages from
 Linux may sometimes only show what the device driver thinks the
 configuration is, perhaps as told it via an incorrect configuration
 file.  The BIOS messages will show the actual hardware configuration
 at that time, but a PnP OS, isapnp, or pci utilities, may have changed
 it since then.

 7.1.  How Are My Device Drivers Configured?

 There may be a programs you can run from the command line (such as
 "setserial" for serial ports) to determine this.   The /proc directory
 tree is useful.  /proc/ioports shows the I/O addresses that the
 drivers use (or try if it's wrong).  They might not be set this way in
 hardware.

 /proc/interrupts shows only interrupts currently in use and many that
 have been allocated to drivers don't show at all since they're not
 currently being used.  For example, even though your floppy drive has
 a floppy disk in it and is ready to use, the interrupt for it will not
 show unless its in use.  Again, just because an interrupt shows up
 here doesn't mean that it exists in the hardware.  A clue that it
 doesn't exist in hardware will be if it shows that 0 interrupts have
 been issued by this interrupt.  Even if it shows some interrupts have
 been issued, it may mean that this interrupt doesn't exist on that
 device does exist on some other device which is not in use, but which
 somehow has issued an interrupt or two.  As of kernel 2.2 the /proc
 tree has changed.

 7.2.  How Are My Hardware Devices Configured?

 It's easy to find out what resources have been assigned to devices on
 the PCI bus: Either use the "lspci" command or for kernel <: 2.2: look
 at /proc/pci; for kernel 2.2+: /proc/bus/pci/devices.  For the ISA bus
 you may try running pnpdump --dumpregs but it's not a sure thing.  The
 results may be difficult to decipher.  Don't confuse the read-port
 address which pnpdump "trys" (and finds something there) with the I/O
 address of the found device.  They are not the same.

 Messages from the BIOS at boot-time tell you how the hardware
 configuration was then.  If you rely on the BIOS for configuring, then
 it should still be the same.  Messages from Linux may be from drivers
 that have checked to see that the hardware is there (and possibly
 checked the IRQ and DMA).  Of course, if the device works fine, then
 it's likely configured the same as the driver.

 8.  Appendix

 8.1.  Addresses

 There are three types of addresses: main memory addresses, I/O
 addresses and configuration addresses.  On the PCI bus, configuration
 addresses constitute a separate address space just like I/O addresses
 do.  Except for the complicated case of ISA configuration addresses,
 whether or not an address on the bus is a memory address, I/O address,
 or configuration address depends only on the voltage on other wires
 (traces) of the bus.

 8.1.1.  ISA Bus Configuration Address (Read-Port etc.)

 For the ISA bus, there is technically no configuration address space,
 but there is a special way for the CPU to access PnP configuration
 registers on the PnP cards.  For this purpose 3 @ I/O addresses are
 allocated.  This is not 3 addresses for each card but 3 addresses
 shared by all cards.

 These 3 addresses are named read-port, write-port, and address-port.
 Each port is just one byte in size.  Each PnP card has many
 configuration registers so that just 3 addresses are not even
 sufficient for these registers on a single card.  To communicate with
 a certain card, a specially-assigned card number (handle) is sent to
 all cards at the write-port address.  After that only that the only
 card still listening is the card with this handle.  Then the address
 of the configuration register (of that card) is sent to the address-
 port (of all cards --but only one is listening).  Next communication
 takes place with one configuration register on that card by either
 doing a read on the read-port or a write on the write-port.

 The write-port is always at A79 and the address-port is always at 279
 (hex).  But the read-port is not fixed but is set by the configuration
 software at some address that will supposedly not conflict with any
 other ISA card.  If there is a conflict, it will change the address.
 All PnP cards get "programmed" with this address.  Thus if you use say
 isapnp to set or check configuration data it must determine this read-
 port address.

 8.1.2.  Address ranges

 The term "address" is sometimes used in this document to mean a
 contiguous range of addresses.  Since addresses are given in bytes, a
 single address only contains one byte but I/O (and main memory)
 addresses need more than this.  So a range of say 8 bytes is often
 used for I/O address while the range for main memory addresses
 allocated to a device is much larger.  For a serial port (an I/O
 device) it's sufficient to give the starting I/O address of the device
 (such as 3F8) since it's well known that the range of addresses for
 serial port is only 8 bytes.  The starting address is known as the
 "base address".

 8.1.3.  Address space

 For ISA, to access both I/O and (main) memory address "spaces" the
 same address bus is used (the wires used for the address are shared).
 How does the device know whether or not an address which appears on
 the address bus is a memory address or I/O address?  Well, there are 4
 dedicated wires on the bus that convey this information and more.  If
 a certain one of these 4 wires is asserted, it says that the CPU wants
 to read from an I/O address, and the main memory ignores the address
 on the bus.  The other 3 wires serve similar purposes.  In summary:
 Read and write wires exist for both main memory and I/O addresses (4
 wires in all).

 For the PCI bus it's the same basic idea also using 4 wires but it's
 done a little differently.  Instead of only one or the four wires
 being asserted, a binary number is put on the wires (16 different
 possibilities).  Thus more info may be conveyed.   Four of these 16
 numbers serve the I/O and memory spaces as in the above paragraph.  In
 addition there is also configuration address space which uses up two
 more numbers.  Ten extra numbers are left over for other purposes.

 8.1.4.  Range Check (ISA Testing for IO Address Conflicts)

 On the ISA bus, there's a method built into each PnP card for checking
 that there are no other cards that use the same address.  If two or
 more cards use the same IO address, neither card is likely to work
 right (if at all).  Good PnP software should assign resources so as to
 avoid this conflict, but even in this case a legacy card might be
 lurking somewhere with the same address.

 The test works by a card putting a test number it's own IO registers.
 Then the PnP software reads it and verifies that it reads the same
 test number.  If not, something is wrong (such as another card with
 the same address.  It repeats the same test with another test number.
 Since it actually checks the range of IO addresses assigned to the
 card, it's called a "range check".  It could be better called an
 address-conflict test.  If there is an address conflict you get an
 error message and need to resolve it yourself.

 8.1.5.  Communicating Directly via Memory

 Traditionally, most I/O devices used only I/O memory to communicate
 with the CPU.  For example, the serial port does this.  The device
 driver, running on the CPU would read and write data to/from the I/O
 address space and main memory.  A faster way would be for the device
 itself to put the data directly into main memory.  One way to do this
 is by using ``DMA Channels'' or bus mastering.  Another way is to
 allocate some space in main memory to the device.  This way the device
 reads and writes directly to main memory without having to bother with
 DMA or bus mastering.  Such a device may also use IO addresses.

 8.2.  Interrupts --Details

 Interrupts convey a lot of information but only indirectly.  The
 interrupt signal (a voltage on a wire) just tells a chip called the
 interrupt controller that a certain device needs attention.  The
 interrupt controller then signals the CPU.  The CPU find the driver
 for this device and runs a part of it known as an "interrupt service
 routine" (or "interrupt handler").  This "routine" tries to find out
 what has happened and then deals with the problem such a transferring
 bytes from (or to) the device.   This program (routine) can easily
 find out what has happened since the device has registers at addresses
 known to the the driver software (provided the IRQ number and the I/O
 address of the device has been set correctly).  These registers
 contain status information about the device .  The software reads the
 contents of these registers and by inspecting the contents, finds out
 what happened, and takes appropriate action..

 Thus each device driver needs to know what interrupt number (IRQ) to
 listen to.  On the PCI bus (and for the serial ports on the ISA bus
 starting with Kernel 2.2) it's possible for two (or more) devices to
 share the same IRQ number.  When such an interrupt is issued, the CPU
 runs all interrupt service routines for all devices using that
 interrupt.  The first thing the first service routine does is to check
 to see if an interrupt actually happened for its device.  If there was
 no interrupt (false alarm) it likely will exit and the next service
 routine starts, etc.

 8.3.  PCI Interrupts

 PCI interrupts are different but since they are normally mapped to
 IRQ's they behave in about the same way.  A major difference is that
 PCI interrupts may be shared.  For example IRQ5 may be shared between
 two PCI devices.  This sharing ability is automatic: you don't need
 special hardware or software.  There have been some reports of
 situations where such sharing didn't work, but it's likely due to a
 defect in the device driver software.  All device drivers for PCI
 devices are supposed to provide for interrupt sharing.  Note that you
 can't share the same interrupt between the PCI and ISA bus.  However,
 illegal sharing will work provided the devices which are in conflict
 are not in use at the same time.  "In use" here means that a program
 is running which "opened" the device in it's C programming code.

 You may need to know some of the details of the PCI interrupt system
 in order to set up the BIOS's CMOS or to set jumpers on old PCI cards.
 Each PCI card has 4 possible interrupts: INTA#, INTB#, INTC#, INTD#.
 Thus for a 7-slot system there could be 7 x 4 = 28  different
 interrupt lines.  But the specs permit a fewer number of interrupt
 lines.  This is not too restrictive since interrupts may be shared.
 Many PCI buses seem to be made with only 4 interrupt lines.  Call
 these lines (wires or traces) W, X, Y, Z.  Suppose we designate the B
 interrupt from slot 3 as interrupt 3B.  Then wire W could be used to
 share interrupts 1A, 2B, 3C, 4D, 5A, 6B, 7C.  This is done by
 physically connecting wire W to wires 1A, 2B, etc.  Likewise wire X
 could be connected to wires 1B, 2C, 3D, 4A, 5B, 6C, 7D.  Then on
 startup, the BIOS maps the X, W, Y, Z to IRQ's.  After that it writes
 the IRQ that each device is mapped to into a hardware register in each
 device.  Then and anything interrogating the device can find out what
 IRQ it uses.

 The above mentioned wires X, W, Y, Z  are labeled per PCI specs as
 INTA#, INTB#, INTC# and INTD#.  This official PCI notation is
 confusing since now INTA# has 2 possible meanings depending on whether
 we are talking about a slot or the PCI bus.  For example, if 3C is
 mapped to X then we say that INTC# of slot 3 is cabled to INTA# (X) of
 the PCI bus.  Confusing notation.

 There's another requirement also.  A PCI slot must use the lower
 interrupt letters first.  Thus if a slot only uses one interrupt, it
 must be INTA#.  If it uses 2 interrupts they must be INTA# and INTB#,
 etc.  A card in a slot may have up to 8 devices on it but there are
 only 4 PCI interrupts for it.  This is OK since interrupts may be
 shared so that each of the 8 devices (if they exist) can have an
 interrupt.  The PCI interrupt letter of a device is often fixed and
 hardwired into the device.

 The BIOS assigns IRQs (interrupts) so as to avoid conflicts with the
 IRQs it knows about on the ISA bus.  Sometimes in the CMOS BIOS menu
 one may assign IRQs to PCI cards (but it's not simple as explained
 above).  There's a situation where Windows zeroed out all the IRQ
 numbers in the PCI cards after the IRQ mappings had been set.  Then
 someone running Windows booted Linux from Windows with the result that
 Linux only found only incorrect IRQs of zero.

 You might reason that since the PCI is using IRQ's (ISA bus) it might
 be slow, etc.  Not really.  The ISA Interrupt Controller Chip(s) has a
 direct interrupt wire going to the CPU so it can get immediate
 attention.  While signals on the ISA address and data buses need to go
 thru the PCI bus to get to the CPU, the IRQ interrupt signals go there
 almost directly.

 8.4.  Isolation

 This is only for the ISA bus.  Isolation is a complex method of
 assigning a temporary handle (id number or Card Select Number = CSN)
 to each PnP device on the ISA bus.  Since there are more efficient
 (but more complex) ways to do this, some might claim that it's a
 simple method.  Only one write address is used for PnP writes to all
 PnP devices so that writing to this address goes to all PnP device
 that are listening.  This write address is used to send (assign) a
 unique handle to each PnP device.  To assign this handle requires that
 only one device be listening when the handle is sent (written) to this
 common address.  All PnP devices have a unique serial number which
 they use for the process of isolation.  Doing isolation is something
 like a game.  It's done using the equivalent of just one common bus
 wire connecting all PnP devices and the isolation program.

 For the first round of the "game" all PnP devices listen on this wire
 and send out simultaneously a sequence of bits to the wire.  The
 allowed bits are either a 1 (positive voltage) or an "open 0" of no
 voltage (open circuit or tri-state).  Each PnP device just starts to
 sequentially send out its serial number, bit-by-bit, starting with the
 high-order bit, on this wire.  If any device sends a 1, a 1 will be
 heard on the wire by all other devices.  If all devices send an "open
 0" nothing will be heard on the wire.  The object is to eliminate (by
 the end of this first round) all but highest serial number device.
 "Eliminate" means to cease to listen anymore to the write address that
 all devices still in the game are still listening to.  This is also
 called "dropping out".  (Note that all serial numbers are of the same
 length.)

 First consider only the high order bit of the serial number which is
 put on the wire first by all devices which have no handle yet.  If any
 PnP device sends out a 0 (open 0) but hears a 1, this means that some
 other PnP device has a higher serial number, so it temporarily drops
 out of this round and doesn't listen anymore until the round is
 finished (when a handle is assigned to the winner: the highest serial
 number).  Now the devices still in the game all have the same leading
 digit (a 1) so we may strip off this digit and consider only the
 resulting "stripped serial number" for future participation in this
 round.  Then go to the start of this paragraph and repeat until the
 entire serial number has been examined for each device (see below for
 the all-0 case).

 Thus it's clear that the highest serial number will not be eliminated
 from the game.  But what happens if the leading digits (of the
 possibly stripped serial numbers) are all 0?  In this case an "open 0"
 is sent on the line and all participants stay in the game.  If they
 all have a leading 0 then this is a tie and the 0's are stripped off
 just like the 1's were in the above paragraph.  The game then
 continues as the next digit (of the serial number) is sent out.

 At the end of the round (after the low-order bit of the serial number
 has been sent out by whatever participants remain) only one PnP device
 with the highest serial number remains.  It then gets assigned a
 handle and drops out of the game permanently.  Then all the dropouts
 from the last round (that don't have a handle yet) reenter the game
 and a new round begins with one less participant.  Eventually, all PnP
 devices are assigned handles.  It's easy to prove that this algorithm
 works.

 Once all handles are assigned, they are used to address each PnP
 device and send it a configuration as well as to read configuration
 info from the PnP device.  Note that these handles are only used for
 PnP configuration and are not used for normal communication with the
 PnP device.  When the computer starts up, all of the handles are lost
 so that a PnP BIOS usually does the isolation process again each time
 you start your PC.

 END OF Plug-and-Play-HOWTO