[Soekris] net45xx interrupt routing?

Bernd Walter soekris-tech at cicely.de
Sat Aug 2 09:46:19 UTC 2003


On Fri, Aug 01, 2003 at 11:06:48PM -0700, Trevor Talbot wrote:
> On Friday, Aug 1, 2003, at 18:50 US/Pacific, Bernd Walter wrote:
> 
> >On Fri, Aug 01, 2003 at 03:30:24PM -0700, Trevor Talbot wrote:
> >>>I'm looking for info on how the physical interrupt lines are 
> >>>connected.  E.g. on the net4501, the CF controller gets GP10, and 
> >>>the three ethernet controllers get PCI C, B, and A.  What about the 
> >>>mini and full PCI slots, and how does the net4521 differ?
> >
> >PCI chips always get INTA first, but INTA can be differently routed 
> >for different slot IDs.
> 
> That's why I wanted the memory dump from BIOS.  It shows how the Elan's 
> PIC is routing the physical lines.

Here is it from a 4501 (comBIOS ver. 1.15):
> db fffefd30
0000:FFFEFD30  04 0D 06 00 00 00 00 00 00 00 00 00 00 00 00 00  ................
0000:FFFEFD40  00 00 00 03 00 00 08 00 00 00 00 00 00 00 00 00  ................
0000:FFFEFD50  00 00 00 00 00 00 00 00 00 05 09 00 20 00 20 00  ............ . .
0000:FFFEFD60  20 00 20 00 20 00 20 00 20 00 20 00 20 00 20 00   . . . . . . . .
0000:FFFEFD70  00 00 00 00 01 00 20 00 20 00 20 00 20 00 20 00  ...... . . . . .
0000:FFFEFD80  00 00 FF FF FF FF 00 00 00 00 00 00 00 00 20 00  ..****........ .
0000:FFFEFD90  00 00 00 00 20 00 20 00 00 00 00 00 20 00 20 00  .... . ..... . .
0000:FFFEFDA0  00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00  ................

> >The miniPCI and the classical slot seem to be identicaly routed for at 
> >least INTA (routed to IRQ10) so they always share.
> 
> Given the other evidence, this suggests they're both on the PIC's INTD 
> line.
> 
> >Also the classical slot INTB end with IRQ10 - no idea on INTC, because 
> >the BIOS forgets to initialise function 2.
> >The onboard Ethernet interfaces get different ones (IRQ 5, 9, 11).
> >I would asume the 4521 is identic to the 4501 for the common parts.
> 
> Boot messages that have been posted from net4521s show the cardbus 
> controller to be a multifunction device -- and it appears to get only 
> one interrupt.  Since there's one ethernet missing (compared to the 
> net4501), either both it and the miniPCI slot are routed to INTD, 
> giving INTC and INTB to the two ethernet controllers, or the miniPCI 
> gets INTD to itself and the cardbus controller ends up on INTC.  I 
> won't know until I see a dump of the PIC registers (or someone digs up 
> a schematic).

Sorry - I have no 4521 handy.

> >>Here's how anyone with a bit of time can help me out: in the BIOS, 
> >>enter "db fffefd30" and show me the results.  I want data from a 
> >>net4501 with stuff in the mini and full PCI slots, as well as a 
> >>net4521, again with the slots in use.  It can be one slot or all of 
> >>them, I just want to see one of each.  I'd also like to know what IRQ 
> >>each device (slot) ends up on, which is typically shown in the OS 
> >>boot messages.
> >
> >Here are the important lines from FreeBSD:
> >hifn0 mem 0xa0001000-0xa0001fff,0xa0000000-0xa0000fff irq 10 at device 
> >16.0 on pci0
> >ohci0: <NEC uPD 9210 USB controller> mem 0xa0000000-0xa0000fff irq 10 
> >at device 17.0 on pci0
> >ohci1: <NEC uPD 9210 USB controller> mem 0xa0001000-0xa0001fff irq 10 
> >at device 17.1 on pci0
> 
> These got mapped into the same space?  That doesn't look right...

Ignore the mem ranges - hifn and usb cards were no plugged in at the
same boot - the messages were concatenated from different saved logs.
But I remember the IRQs were the same when I had both in.

> >ehci0: <NEC uPD 720100 USB 2.0 controller> at device 17.2 on pci0
> 
> Curious; I wonder why this wasn't initialized.

Me too...

-- 
B.Walter                   BWCT                http://www.bwct.de
ticso at bwct.de                                  info at bwct.de




More information about the Soekris-tech mailing list