hpr-knowledge-base/hpr_transcripts/hpr0135.txt

Episode: 135
Title: HPR0135: LPI Ceritification Part 5 PCI Cards
Source: https://hub.hackerpublicradio.org/ccdn.php?filename=/eps/hpr0135/hpr0135.mp3
Transcribed: 2025-10-07 12:15:35

---


Hello everybody, my name is Kevin Fallon and welcome to this episode of hacker
public radio. Today is a continuation on the LPI certification. Today we're
going to be covering PC cards. As I suggested previously we're going to be
changing our tech slightly so we're going to move from using the IBM
documentation which I've heard some comments about that seems to be a bit out of
date and also I was never really comfortable with the fact that it was
licensed correctly. So we're moving to some documentation produced by
leading edge business solutions as part of their Linux training program.
You can find them as www.ledge.co.za. Their documentation is released
under the GNU-free documentation license, a link which will be
also included in the show notes. So that means I am free to read word for word
in this chapter. They also have at the end of every chapter a quiz
questions so I was thinking here to be a good topic to put at the beginning of
the podcast. So first of all what is the purpose of an interrupt
an iO address and a DMA channel. What happens when there is a resource
conflict between two devices and under what circumstances can two devices
share the same IRQ value? All these questions and more will be answered in the
upcoming podcast. Okay topic one does one or one
dot five setting up PCI expansion cards. At the end of this you should be able to
configure various cards for the various expansion slots. You should know the
difference between ISA and PCI cards with respect to configuration issues.
This objective includes correctly setting IRQs, DMAs and IO ports for the cards,
especially to avoid conflicts between devices. And this will include
using the ISA PMP if there is a ISA Plum Play device in your computer.
A nice thing that they have in this documentation is they have key
files, terms and utilities. They include their directory proc DMA
which is the direct memory addressing channel assigned to devices.
Proc interrupts which is the hardware interrupts assigned to devices.
Proc iO ports which is the iO ports input output ports assigned to devices.
Proc PCI which is the PCI bus information as shown by the command
LSPCI. PCI dump which dumps possible ISA Plum Play information.
ISA PMP which tells ISA Plum Play devices what port to use
and LSPCI which shows PCI bus information. Section five dot one bus architectures.
Now we've covered quite a lot of this stuff before but as a result of
switching documentation we're going to be covering again it's no harm to
do a little bit of revision. A bus is physically a set of parallel
conductors that connect the components of the computer. The components of
the computer use the bus to communicate by sending electrical signals to
each other. The components that are connected by the bus include the CPU,
the system memory, RAM RAM and peripherals, disk interfaces, display
adapters, interrupts and controllers, bus resources.
The following bus concepts are used both by the ISA and PCI bus
and iO address. To specify the specific components on a bus
the CPU uses its iO address. This means that the system and the bus cannot share iO addresses.
To read your life from a particular location and memory the CPU
harder uses memory addresses. The ISA bus separates the concept of memory
addresses from iO addresses although there is not a huge difference between these.
When a peripheral requires CPU attention it issues an interrupt request.
The CPU in response to the request stops what it's doing and speaks to the relevant peripheral
usually to retrieve data or to send data to it. Buses generally have a number of
interrupt request lines which enable the kernel to determine which peripheral
caused the interrupt. If two peripherals share an interrupt request line it is
possible that they will trigger an interrupt request simultaneously.
The kernel must respond to both devices for proper operation.
Shared interrupts work when interrupts are level triggered for example on a PCI bus.
It is not possible when interrupts are edge triggered for example on a ISA bus.
Direct memory access. Using a DMA channel allows a peripheral to copy data
directly to in and out of computer memory. This means that the CPU is not
intimately involved with the data transfer once it is in motion.
Use of the DMA channel allows a device to perform direct memory access.
It is generally not possible to share DMA channels between peripherals.
5.3 Bus conflict resolution. The Linux kernel is responsible for managing
the assignment of IRQs. Interrupt request lines. I-O ports and DMA channels.
The kernel will not assign a resource to two drivers unless it is shareable.
That's a level triggered interrupt on a PCI bus.
A number of resource conflicts can occur on a bus for example.
I-O address conflict. Either one or both of the devices will fail to work.
Even if the one of the devices is misconfigured and no kernel margin has been loaded first,
it may respond to its I-O address on cause problems.
Another resource conflict might be shared IRQs.
Sharing IRQs with level trigger interrupts can cause conflicts if the kernel
only serves as one of the devices and not both.
This is not likely with new kernels but it can happen.
Sharing IRQs with edge triggered interrupts can cause one or both
of the devices to fail. If a kernel module for a given device is not loaded,
the device will never generate interrupts and will call exist with the other device.
Another point to note is DMA channels cannot be shared.
The kernel will refuse to assign DMA channels to two devices.
The device for which the module is loaded first will work.
To resolve conflicts for resources you need to reconfigure the devices so that they do not conflict.
All cards can be configured by changing jumpers on the card.
ISA PMP can change the resources assigned for plug and click cards.
Some PC biases contain this facility too.
Often the biases include facilities for changing the interrupts
and memory locations of built-in devices.
IRQ conflicts between devices on the PCI boss and ISA boss can be resolved
by telling the bias to reserve certain IRQ lines for the ISA boss.
I-O ports do not conflict since the PCI boss uses an
independent range of port addresses.
Section 5.4 PCI card configuration.
The peripheral component interconnect PCI boss
includes resource assignment as parts of its specification.
To make a PCI card work you merely need to load the correct kernel module.
To determine which kernel module will work you can consult the output of LSPCI command.
The module is then loaded using the command mod probed module name.
If there are options required for the modules you can enter them in the file
slash egcmodules.conf and run a DEP-MOD-A.
LSPCI lists PCI devices.
Each device on the PCI boss is connected to a specific slot
and this identifies itself by a manufacturer and device number.
Typing in man space LSPCI gives us the man page for the PCI utilities command
and it says LSPCI is a utility for displaying information
about all PCI bosses in the system.
Each device on the PCI boss is plugged into a specific slot
and identified by a manufacturer and device number.
Happily, LSPCI can display the names of the devices
but cross-referencing them against slash user slash share PCI.ids file
which contains a list of vendors and their products.
That might be in a different location in your distribution.
For example, Antobuntu, it's slash user slash share slash miss slash PCI.ids.
Even if the vendor ID is not listed, LSPCI will tell you what type of device it is.
To find the components on the PCI boss you can look in the file slash proc slash PCI
or use the LSPCI command.
Just type in LSPCI, we'll dump out a list of PCI numbers and what they are.
When I do this, I see that 0, 0, 0, 0 is the whole storage.
For example, 0, 2 colon 0, 0.0 is the ethernet controller.
You can get more specific information on a particular component
by typing LSPCI minus S and then the ID number, for example, 0, 2 colon 0, 0.0 minus V for the boss.
And we see here more information about it.
The IO port that's used, the memory is at where the IRQ that it's using.
Section 5.5, ISA card configuration.
To change the settings that an ISA card is using generally requires changing the
jumpers on the card itself.
The section following is one example how to change the settings on an ISA card
which supports ISA plug and play.
With few exceptions, the kernel module supporting ISA cards have to be configured
with specific parameters.
Most drivers are capable of detecting intro values,
so the critical values to assign are the IO addresses.
The best way to assign values is to read the relevant documentation
and set the jumpers according to the free resources of your PC.
You can sometimes just guess.
Here's an example of loading a kernel module for an NE2000 ethernet adapter.
The NE2000 driver requires the 839 old driver.
So on the printout, the two in-smod space is 390.0
and it gives back no feedback.
To do an in-smod space, any .0 and that gives
they are back no such device or address.
And the hint is in-smod errors can be caused by incorrect modular parameters,
including invalid IO or IRQ parameters.
You may find more information in the syslog or the output from D message.
So that didn't work.
Let's guess an IO address.
Now they put in the command in-smodspace-any.all
space-io equals 0x220 and it gives no such device or address.
So we learn that network card is not at 0x22.all
and they now go and do a little loop command to find out
everyone that is is and then at the end they do an LS mod.
I would like to hear some feedback from people
as to how relevant ISA stuff is to the API certification
exams.
Section 5.6.
ISA plug-and-play devices.
A plug-and-play device is very similar to a card which is configured with
jumpers, except that the jumpers are missing.
ISA PMP devices are automatically configured by the bias or their operating system.
Before the ISA PMP standard was devised, many hardware manufacturers
configured their hardware IO addresses using special MS-DOS-based utility
programs. The ISA PMP program replaces this functionality.
To configure a PMP device, use PMP dump redirected to slash
ETC-forward slash ISA PMP.conf to determine possible configuration settings for the file.
Address that file and remove comments,
optionally change the parameters from the defaults.
Then run ISA PMP-space-forward slash ETC-forward slash
ISA PMP.conf to set the parameters the device actually uses
and then load the kernel module for the device.
Your kernel modules include a module named ISA-PMP,
which automatically configures ISA PMP cards.
You can configure misbehaving and on-supported cards manually by writing
to forward slash proc, forward slash ISA PMP,
although using PMP dump and ISA PMP may prove simpler.
Section 5.7 kernel interface commands.
The following kernel interface files in slash proc,
display the assigned BOS resources,
and that is forward slash proc, forward slash the file interrupts.
For each CPU, proc interrupts shows an interrupt,
which are in use by the kernel, and the number of time each interrupt has been triggered.
The file forward slash proc, forward slash IO ports,
shows the IO ports, which are in use by the kernel drivers,
and are displayed by the IO ports interface.
Similarly, the command cache space forward slash
proc, forward slash IO mem provides memory mapping of the IO allocations.
That brings us to the end of this section.
We had a quiz at the beginning.
Question number 1 was the purpose of an interrupt.
An interrupt indicates that a peripheral requires CPU attention.
Could be reading or writing.
And what's the purpose of an IO address?
The answer is an IO address can be used to edit by the BOS
to identify a particular peripheral or a component of the peripheral.
On what's the purpose of a DNA channel?
A DNA channel is used for coordinating data transfers between peripherals and memory.
This frees the CPU to do more productive work.
What happens when there is a resource conflict between two devices?
And the answer is either one or both devices will not work.
It is possible that one of the devices may function unreliably.
And under what circumstances can two devices share the same IRQ value?
And the answer is on a BOS which uses level triggered interrupts.
It is possible to share interrupt values.
On the ISA BOS and on the EISA BOS interrupt sharing is not possible.
On the PCI BOS interrupt sharing is the norm.
And just to give you a summary of what the O'Reilly book says in relation to this.
It describes setting up different PCI cards.
It gives the example of three areas, the jumper,
era, the non-volatile era and the modern era.
It goes through the CAT PROC interrupts.
It also refers to CAT PROC DMA, CAT PROC IO ports.
And it says that on the exam you should be aware of the default resource assignments listed.
You should also know how to examine a running Linux systems resource assignment
using the PROC file system.
It makes reference to table 3.1 which has TTYS0 come on as IO address of 3 F8 of IRQ4.
However, there says you don't need to memorize all the device settings in PCI architecture
but you should be ready to answer specific questions regarding some of the more common ones
such as setting ports and such.
Again, somebody might be able to advise us as to whether that's something that we would need
to remember or not for the exam.
That brings me to the end of this section.
I hope you like the change in format.
It's a lot drier probably but what it should allow us to do also is that if there are other
presenters they can simply take a section from this course where and just follow up through.
I got this information from the lot of Linux, one of the guys on the lot of Linux links,
IRC channel, which is a nice place to hang out.
And they've also pointed me to a CD which I'll also include on the show notes.
Okay, thank you very much for listening, ladies and gentlemen.
The following time we will be talking about device configuration if I don't do something else
in the meantime.
But for me, have a good evening and keep packing.
Thank you very much for joining us on the Macro Public Radio.
HBR is sponsored by Carol.net so head on over to C-A-R-O dot N-E-C for all of our team.
Thank you very much for joining us on the Macro Public Radio.