On Sun, 19 Jan 1997, Ben Strawson wrote:

> 
> I've tried kernel 4922 on my machine (SA110, RiscBSD on 420Mb IDE, Cumana
> SCSI (not used for RiscBSD), 24Mb RAM etc.)
> 
> The floppy drive does not work still (I had the same problems as everyone
> else reported).

Ok I need some clarification here. I know of two problems that have been
reported.

1. The floppy is not detected at probe time
2. The floppy reports lots of errors / disk corruption

Neither of these problems effect my machine ;-)

The floppy not being detected is an odd one. Neil (think it was him) had
this problem for a while when he replaces his floppy drive. When he
swapped the floppy drive again the problem disappears which made it look
like it was related to particular floppy drives.

Any more info etc would be helpful.

> The com driver seems to be much better now, - I get 0 ibuf overflows, 0
> uart failures, but still some silo overflows (as expected).  As a result,
> ppp performance under RiscBSD is much better (I was using RiscOS for
> downloading stuff, simply because it was much faster).

Yep I have noticed a diff in my xfer rates as well ;-) I am looking a t
ways to minimise the unmnber of SILO overflow. Basically I need to reduce
interrupt latency etc. Also triggering serial interrupts when there are
less bytes in the SILO may help.

> Out of interest, the kernel is detecting my SCSI card and drive on startup
> (reports one drive on csc0 as I would expect).  However, I am currently
> using *only* the IDE disc for RiscBSD, so not sure how much further it
> would get if I was running RiscBSD off the SCSI drive.

Well technically if RiscBSD were installed on the SCSI disk you could boot
straight from it.
Now that the interrupt problems with the cumana and powertec drivers are
fixed I expect xfer rates to be improving soon make the drivers a
reasonable alternative to IDE.

Cheers,
				Mark