Re: Host bridge config (Galileo)

[Mitch Kahn <mitch@demoworks.com>]

Hi Paul,

First, Galileo would have been glad to help you with this
one. Our support address is: support@galileoT.com.

At 05:35 PM 4/23/97 +0000, Paul Slade wrote:
>Help,
>
>We are using a Galileo Technology 'Galileo 4' board. This consists of 
>an R4600 MIPS processor connected to the PCI bus by a Galilieo 
>GT-64010A Host Bridge. This appears as Class 0x060400 Host Bridge PCI 
>device.
>
>The PCI configuration specifies a type 0 header, with 5 BASE 
>ADDRESSES defined of various sizes.
>
>My problem is that most BIOS's that I have tried seem to ignore the 
>device at boot up and do not configure the BASE ADDRESS registers. On 
>BIOS's that print up which devices have been found, it is not listed.
>

This happens on <20% of the BIOSs that we've seen, however, newer
BIOSs do seeem to have this problem more often (newer Pentium boards.)

>When Windows 95 boots up it sees the card but since all the BASE 
[stuff deleted]
>
>Does anyone know why the device gets ignored by the BIOS? Is there 
>some special significance that BIOS's assume for Host Bridges? Is 
>there any work-around?
>

In my opinion, the BIOS should not be doing this. Some BIOS
writer somewhere *assumed* the only host bridge would be the
Pentium's and decided to *ignore* any other (why? anyone
from BIOSland want to explain?)

The "workaround" is to change the class code of the board. The GT-64010 is
a host bridge, by default, but this can be changed by software.
You will need to change the initialization code on Galileo-4 to
do this. If you are using the supplied monitor, then we (Galileo)
will need to provide you with a new version (it will be a couple
of weeks, our IDT monitor guy's wife just had a baby yesterday!)

We actually found this problem first with Galileo-9, our R4640
board. The GT-64011 device has a PCI header autoload feature
that we used to work around this BIOS error on this board. A
new version of the monitor for that board will be released
soon as well. Check our website regularly. Also contact
you local rep to sign-up for automated tech support (Eurodis
Bytech).

By the way, a safe class code to use is for generic memory
device.

>Also, under Windows 95, the board is not detected at all if it is the 
>other side of a PCI to PCI bridge, though other devices do get 
>detected. Is this due to some problems that have been fixed in the 
>OSR2 release that I have heard about?
>

Probably another BIOS error. Changing the class code will likely
fix this.

WL<