On Wed, 01 Jan 2003 18:09:41 +0100 (CET), Daniela Engert wrote:
>>Or should I work through all unassigned bridges in the whole making: 1st bridge ->
>bus 1. Bridge with NIC
to
>>bus 2. 2nd bridge on bus 0 -> bus 3 and bridge with soundcard to bus 4?
>Now, this is ok. Do a depth-first tree walk.
Like I thought.
Well, I never knew, because I don't have a multiple PCI2PCI bridge system here and I
couldn't find any logic
on enumeration on the net especially no information about the subordinate bus number.
The only Linux
sources that I found set the subordinate bus number to the assigned bus number at
*any* time, which is
wrong. I also found some mails (one or 2 even from Linus, lol) that discussed
enumeration on systems with
multiple PCI2PCI bridges behind each other and that the Linux kernel would have
problems with such cases.
>>As far as I understood the 2nd solution seems to be the right one.
>>Also I would really like to leave all assigned bridges (by BIOS) as they are. Is
>this okay?
>I think fiddling with BIOS assigned bus numbers may make the system
>fail sooner or later. Errors should be corrected, of course.
Well, I think I will only implement configuring, not reconfiguration. If the BIOS does
bad things, especially that
low level, messing around it could make it even worse. That's why GCONFIG
configurations are also capable
of having the "do not touch" flag, which means that the BIOS settings will be
preserved at any time. So if this
interferes with configurations for other devices, those will fail. BIOS will get first
try, so at least systems that are
already 100% configured by BIOS won't get messed up or will only get messed up, when
the user wishes to.
;-))
PCI-Bus assignment will be "hardcoded" (which means won't be user adjustable) and I
will just change the
method to the one mentioned above.
cu, Kiewitz
-----------
To unsubscribe yourself from this list, send the following message
to [EMAIL PROTECTED]
unsubscribe acpi-os2
end
