[Soekris] Y adaptor for PCI interface

[Andrea Cristofanini -- [GedamEurope]]

Thanks Bernard for depth explenantion

I'm tring to save money in our project.
I'm trying to use the Y adaptor for plug into same  board 2 BRI (1 bri 
35 euros.)
I know that there are 2 BRI from beronet or someone else but the price 
is more that 1 BRI plus Soekris board ....
So i know also clock problem and interrupt problem very well,
BTW  i will just test, if does not work i will use redundand board with 
right sw .

I let you know ASAP i get this Y adaptor and post my test result.
I hope to have a lot of luck !!!!!
Best Regards Andrea
Bernd Walter ha scritto:
> On Wed, Mar 14, 2007 at 02:46:26PM +0100, Raphaël Jacquot wrote:
>   
>> Andrea Cristofanini -- [GedamEurope] a écrit :
>>     
>>> Dear all
>>> i'm looking forward to find out a Y PCI adaptor for Soekrsi net4801
>>>
>>> So exactly i'm looking for an adaptor that let me plug into soekris  
>>> more that one PCI interface.
>>> Does this exist ?
>>> Pls anyone have with info can drop me an email to andrea at gedameurope.com
>>>
>>> Besta Regards Andrea
>>> Gedam  Group.
>>>       
>> I think you can't do that.
>>     
>
> There are many aspects why this would not work.
>
> Ever PCI slot has a unique line selecting this slot.
> This line is physically attached to any of the AD* line accoring
> to the slot number.
> A splitter card can't know which of them is already in use by onboard
> components or other slot headers - normaly they ship with a small card
> stealing the signal (together with INT*) from another slot header.
> Theoretically you can steal lines from the Mini-PCI.
>
> Interrupts are the next aspect - the BIOS can't tell the OS about
> how interrupt lines are connected to PCI slot numbers it doesn't know
> about.
>
> Another aspect is line length.
> PCI is very strict about line lenght and extending the lines is always
> risky.
> Obviously violating PCI specs in this area isn't that critical, as this
> point is although true for other riser-cards contructs that the
> motherboard vendor don't support with their board layout.
>
> clock line is the next problem point.
> Each slot has a different clock line, since they can be loaded with
> a sine chip only.
> This is typically a limited resource.
> Normally the clock comes from the host side of a given PCI bus and
> you use an internal clock drivera with just a hand full output lines.
> You usually can use an external clock driver with more output lines,
> but all components on a given PCI bus must use their clock signal from
> the same source.
> Unless there are free clock lines you have to rewire onboard components
> as well, or violate PCI specs by attaching multiple chips to the same
> line - as all cheap riser cards do, so it isn't that problematic in
> praxis.
>
> All in all a cheap riser may work given some software hacks, but it
> will very likely not work out of the box.
>
> A real working splitter/riser card would have to use a PCI-PCI bridge.
> This way you create a new PCI bus with new addressing, new clock,
> a well specified interrupt routing, etc...
> However those cards are rare.
>
>