"scsi bus continuously busy"

[Warner Losh]

On Sat, May 2, 2020 at 11:22 AM Alan Perry via cctalk <cctalk at classiccmp.org>
wrote:

> Courtesy of a Raspberry Pi serving as the ND server, I am now able to
> load SunOS 3.5 over the network onto my 3/260 and it is now coming up
> into the OS. I am now seeing this error:
>
>  >sc0 at vme24d16 200000 vec 0x40
>  >sd0 at sc0 slave 0
>  >si0:  sc_cmd:  scsi bus continuously busy
>  >sc0:  resetting scsi bus
>  >sd1 at sc0 slave 1
>  >si0:  sc_cmd:  scsi bus continuously busy
>  >sc0:  resetting scsi bus
>
> The SCSI controller is the "Sun 2" SCSI card. I saw some corrosion-ish
> crap on the board and cleaned it off. It is SCSI, so, of course, I
> played with termination. No change in behavior.
>
> Is this likely to be a controller board problem or a device problem?
>
> Are these boards picky about SCSI devices?
>
> *Any other suggestions?*


This comes from code:
                /* wait for scsi bus to become free */
                for (j = 0; j < SI_WAIT_COUNT/4; j++) {
                        if ((SBC_RD.cbsr & SBC_CBSR_BSY) == 0)
                                goto SI_ARB_SEL_FREE;
...
                }
                printf("si%d:  si_arb_sel: scsi bus continuously busy\n",
                        SINUM(c));

So the busy bit is stuck on and there's no reselect that happens. The chip
in question is NCR 5380 SBC. This is a phase issue on the bus. After 10s
the busy bit in this controller didn't clear. I didn't look it up, but I'll
wager that's the BSY bit on the bus not clearing. This suggests an
electrical issue (maybe termination, maybe not, so not surprised it didn't
help since this is a signal line, not a data line).

Does this happen with no devices on the bus?

Warner