After upgrading to 1.5Q, I'm seeing something similar:

Jan  4 11:53:59 alchemy /netbsd: dmaintr: DMA xfer of zero xferred 65509
Jan  4 11:53:59 alchemy /netbsd: esp0: !TC on DATA XFER [intr 10, stat 83, step
4
] prevphase 0, resid 41f0
Jan  4 11:53:59 alchemy /netbsd: dmaintr: DMA xfer of zero xferred 65493
Jan  4 11:53:59 alchemy /netbsd: esp0: !TC on DATA XFER [intr 10, stat 83, step
4
] prevphase 0, resid 3fe0
Jan  4 11:53:59 alchemy /netbsd: dmaintr: DMA xfer of zero xferred 65493
Jan  4 11:53:59 alchemy /netbsd: esp0: !TC on DATA XFER [intr 10, stat 83, step
4
] prevphase 0, resid 3be0

It's probably something with the tagged queuing changes as Allen mentioned.
I'll try the changes he suggests and see what happens.  If it matters at all,
the relevant lines from the messages file are:

Jan  4 11:42:52 alchemy /netbsd: esp0 at obio0 (quick): address 0x812000:
NCR53C9
6, 16MHz, SCSI ID 7
Jan  4 11:42:52 alchemy /netbsd: scsibus0 at esp0: 8 targets, 8 luns per target
Jan  4 11:42:52 alchemy /netbsd: esp0: target 0: sync negotiation
Jan  4 11:42:52 alchemy /netbsd: sd0 at scsibus0 target 0 lun 0: <SEAGATE,
ST3457
3N, 6244> SCSI2 0/direct fixed
Jan  4 11:42:52 alchemy /netbsd: sd0(esp0:0:0): async mode
Jan  4 11:42:52 alchemy /netbsd: sd0: 4340 MB, 7501 cyl, 5 head, 237 sec, 512
byt
es/sect x 8888924 sectors
Jan  4 11:42:52 alchemy /netbsd: esp0: target 2: sync negotiation
Jan  4 11:42:52 alchemy /netbsd: cd0 at scsibus0 target 2 lun 0: <MATSHITA,
CD-RO
M CR-8024, 2.0e> SCSI2 5/cdrom removable
Jan  4 11:42:52 alchemy /netbsd: probe(esp0:2:1): max sync rate 4.03MB/s
Jan  4 11:42:52 alchemy /netbsd: esp0: target 6: sync negotiation
Jan  4 11:42:52 alchemy /netbsd: st0 at scsibus0 target 6 lun 0: <CONNER, CTMS
3
200, 7.15> SCSI2 1/sequential removable
Jan  4 11:42:52 alchemy /netbsd: st0: st0(esp0:6:0): max sync rate 3.57MB/s
Jan  4 11:42:52 alchemy /netbsd: drive empty

Steve Brown

Allen Briggs wrote:

> > I just compiled a 1.5Q kernel, Dec 29 sources, and now I get:
> > esp0 at obio0: address 0x83f000: NCR53C96, 25MHz, SCSI ID 7
>
> OK, so you're not using "quick" support.  The changes to esp.c are
> not going to affect this at all (or shouldn't anyway).  Much more
> likely is that the "tagged queuing" support that eeh added is
> allergic to your system for some reason.  Can you try turning off
> the tagged queuing by setting ncr53c9x_notag = 1 in sys/dev/ic/ncr53c9x.c
> ?  If you don't have DEBUG defined, move the declaration and use
> of ncr53c9x_notag outside of the #ifdef DEBUG.  If that doesn't
> help (or actually, even if it does), I'd like to see what the output
> looks like if you boot with ncr53c9x_debug set to 0xffff.
>
> This is the first I've heard of this...
>
> -allen
>
> --
>  Allen Briggs                     briggs@wasabisystems.com
>  http://www.wasabisystems.com/    Quality NetBSD CDs, Sales, Support, Service