> Is there any way to prevent soli overflows during disk accesses with high
> speed serial connections? It's annoying having to slow V.34 down to 19.2K,
> as it is actually faster than 28.8, because of all the silo overflows.
				^^^^
				you mean 38.4k?


That's interesting - with my GVP A530 Accel/SCSI system at 38.4k, disk
access merely bring all serial access to a dead stop until a second or
two after the access stops.  Of course, I am using CSLIP, so the protocol 
takes care of the corrupted packets before they could create an overflow - 
but the root of the problem is the serial driver.



> So, has anyone written a GvpPatch for NetBSD?


There is a patch for NetBSD 1.0 that claimed to give much better 
serial performance, but I don't know if it was incorporated into 
-current.  When I tried it, the kernel didn't work on my machine...

I understand the x86 NetBSD serial driver is at least as bad or worse 
than the Amiga driver, even though the x86s have a much better UART, so 
we might consider ourselves lucky!



---
Jeff Thieleke
thieleke@icaen.uiowa.edu
http://www.icaen.uiowa.edu/~thieleke