Hi Geert,
> The definition of  NUM_ATARI_SOURCES will need to be changed again;
> we'd best hardcode it to 140:
> For one, allowing for 152 VME interrupts will exceed the boundaries of
> the allocated bitmap (i.e. trash other memory).
> We won't need the full 200 either (I seem to have got the EtherNAT
> source numbers wrong in my earlier patch).
>
>   
See attached - the 3.3.0-ethernat patches implement switching over to 
the mainstream 91Cx driver (card detected on my Falcon, link brought up 
but not detected by the driver due to hardware problems). The 
3.3.0-atari-cleanup-num-irqs.diff sets the number of interrupt sources 
to 141 on Atari, setting the number of VME sources back to 16 so we 
don't accidentially trash memory contiguous with the free_vme_irq bitmap.

For some reason, I have to set NUM_ATARI_SOURCES to 141 to actually get 
interrupt source 140 to register. Off-by-one error somewhere in the 
generic interrupt code I'd think, but I've not been able to spot it. 
(Just increasing the number of interrupts that are registered by 
atari_init_IRQ results in 'unexpected interrupt from 112' until the cows 
come home. And yes, it happens in ARAnyM, too.)

That pretty much wraps it up for EtherNEC and EtherNAT - EtherNEC is 
fully tested by me, EtherNAT needs testing on Christian's Falcon or by 
someone else with fully functional hardware. Next project : SCSI, SCC 
serial, or what?...

Cheers,

  Michael