On Wed, Mar 21, 2001 at 12:48:41AM -0000, eeh@netbsd.org wrote:
> 
> 	I saw this before sbus streaming cache was fixed by adding
> 	BUS_DMA_COHERENT flags to bus_dmamap_load in hme.c.
> 	I think iommu.c changes from BUS_DMA_COHERENT to BUS_DMA_STREAMING
> 	broke the driver(s). The fix should be straightforward
> 	replacing those defines in all sbus-related drivers to new define.
> 	The patch for hme.c is below.
> 
> 		Andrey
> 
> This is actually exactly the opposite of what you
> want.  BUS_DMA_STREAMING == !BUS_DMA_COHERENT.
> 

-current works fine without this(I mean as-is).
Don't know when I'd be able to test it on my friend's u30,
people scream on sudden reboot - I've got to wait. 

Meanwhile what could be checked is if hme got interrupts, what is
needed is to set up breakpoint in hme_rint: 
break into ddb Stop-A and  'break hme_rint'
Use 'c' to continue.

Ideally your server and client should be on separate segment or
you can get too many interrupts. If everything is all right there
you should get interrupt in sync with rarp response packet.


Regards,
	Andrey