VIA EPIA-M and two port riser?

Anyone know what's involved in getting Via's two port PCI riser (which
supposedly makes use of special features in the VIA BIOS) working under
linuxbios? As it stands I don't get an IRQ assigned to whatever's in the
top slot, so I'm assuming it's not working as of now.

Is it as simple as adding the missing pci_assign_irqs for device 0x13?

[Marc Nicholas]

Interesting....does the riser have a bridge chip on it or something? I was 
under the impression they were passive, but have never actually used one.

-marc

On Wed, 15 Sep 2004 [e-mail] wrote:
> Anyone know what's involved in getting Via's two port PCI riser (which
> supposedly makes use of special features in the VIA BIOS) working under
> linuxbios? As it stands I don't get an IRQ assigned to whatever's in the
> top slot, so I'm assuming it's not working as of now.
>
> Is it as simple as adding the missing pci_assign_irqs for device 0x13?
> _______________________________________________
> Linuxbios mailing list
> [e-mail]
> http://www.clustermatic.org/mailman/listinfo/linuxbios

On Wed, Sep 15, 2004 at 10:19:44AM -0400, Marc Nicholas wrote:
> Interesting....does the riser have a bridge chip on it or something? I was 
> under the impression they were passive, but have never actually used one.

Nope, no bridge chip. Just a handful of jumpers.

[Marc Nicholas]

Hmmmmmm...I don't really understand why you'd need to do anything special 
for a passive riser. But I suppose it's possible that LinuxBIOS only 
expects one PCI device on that target?

-marc

On Thu, 16 Sep 2004 [e-mail] wrote:
> On Wed, Sep 15, 2004 at 10:19:44AM -0400, Marc Nicholas wrote:
> > Interesting....does the riser have a bridge chip on it or something? I was
> > under the impression they were passive, but have never actually used one.
>
> Nope, no bridge chip. Just a handful of jumpers.
> _______________________________________________
> Linuxbios mailing list
> [e-mail]
> http://www.clustermatic.org/mailman/listinfo/linuxbios

Marc Nicholas  writes:
> Hmmmmmm...I don't really understand why you'd need to do anything special for a
> passive riser. But I suppose it's possible that LinuxBIOS only expects one PCI
> device on that target?

There likely needs to be something so we can get the proper associate
of device select to irq.

Eric

VIA EPIA-M and two port riser? - Solved?

On Wed, Sep 15, 2004 at 10:47:21PM +0900, [e-mail] wrote:
> Is it as simple as adding the missing pci_assign_irqs for device 0x13?

Anyone who's curious, this would seem to have done it.

Sorry I don't have values for B, C and D.

On an unrelated note, anyone know which magic numbers to tweak in order to bump
an EPIA-M up to 256MB RAM under linuxbios 1?

Index: mainboard.c
===================================================================
--- mainboard.c (revision 742)
+++ mainboard.c (working copy)
@@ -8,6 +8,7 @@
 static const unsigned char usbIrqs[4] = { 11, 10, 12, 5 };
 static const unsigned char enetIrqs[4] = { 11, 5, 10, 12 };
 static const unsigned char slotIrqs[4] = { 10, 12, 5, 11 };
+static const unsigned char slot2Irqs[4] = { 11, 0, 0, 0 };
 static const unsigned char firewireIrqs[4] = {10, 12, 5, 11 };
 static const unsigned char vt8235Irqs[4] = { 5,10, 12, 11 };
 static const unsigned char vgaIrqs[4] = { 11, 5, 12, 10 };
@@ -58,6 +59,10 @@
        printk_info("setting vga\n");
        pci_assign_irqs(1, 0x00, vgaIrqs);

+       // PCI Riser slot
+       printk_info("setting pci riser slot\n");
+       pci_assign_irqs(0, 0x13, slot2Irqs);
+
        // PCI slot
        printk_info("setting pci slot\n");
        pci_assign_irqs(0, 0x14, slotIrqs);

VIA EPIA-M and two port riser? - Solved?

[Stefan Reinauer]

* [e-mail]  [040916 05:14]:
> On Wed, Sep 15, 2004 at 10:47:21PM +0900, [e-mail] wrote:
> > Is it as simple as adding the missing pci_assign_irqs for device 0x13?
>
> Anyone who's curious, this would seem to have done it.
>
> Sorry I don't have values for B, C and D.

They are often cyclic.

In case anyone can explain this in more detail: How tight are these
interrupt numbers connected to the real hardware. In a system with 
lots of local and IOi APICs the wiring should be purely virtual anyways,
should it not?
> static const unsigned char usbIrqs[4] = { 11, 10, 12, 5 };
>  static const unsigned char enetIrqs[4] = { 11, 5, 10, 12 };
>  static const unsigned char slotIrqs[4] = { 10, 12, 5, 11 };
> +static const unsigned char slot2Irqs[4] = { 11, 0, 0, 0 };
>  static const unsigned char firewireIrqs[4] = {10, 12, 5, 11 };
>  static const unsigned char vt8235Irqs[4] = { 5,10, 12, 11 };
>  static const unsigned char vgaIrqs[4] = { 11, 5, 12, 10 };

Stefan