> 
> Seeing a lot of these on my SS2 recently:
> 
> dma0: xfer (-1) > req (7)
> esp0: !TC [intr c, stat 87, step 4] prevphase 3, resid 400
> [...]
> Can I assume that the "req (7)" is DMA-enabled target selection?  Sort of
> a small transfer, yes?
> 

Yes, you see these if the target selection is interrupted by another
device re-selecting. The dma routine is computing the residual count
by also taking in consideration the bytes in the ESP fifo, but those
fifo bytes no longer belong to the current transfer but are the
re-selection data of the other target.

Similarly, the early termination of the original selection causes
confusion in the "!TC" warning.

> Ideas?  The machine stays up, and is stable... should I be worrying?  :-)

Both messages are spurious. The check for negative dma count should probably
be dropped from dma.c, or perhaps be changed to recognize the selection
phase.

-pk