Hi,

with recent kernels I frequently see things like this:


NetBSD/mac68k (se30) (tty00)

login: ncrscsi0 : Resetting SCSI-bus (Timeout waiting for phase-change)
ncrscsi0 : Resetting SCSI-bus (Timeout waiting for phase-change)
ncrscsi0 : Resetting SCSI-bus (Timeout waiting for phase-change)
ncrscsi0 : Resetting SCSI-bus (Timeout waiting for phase-change)
ncrscsi0 : Resetting SCSI-bus (Timeout waiting for phase-change)
ncrscsi0 : Resetting SCSI-bus (Timeout waiting for phase-change)
ncrscsi0 : Resetting SCSI-bus (Timeout waiting for phase-change)
ncrscsi0 : Resetting SCSI-bus (Timeout waiting for phase-change)
ncrscsi0 : Resetting SCSI-bus (Timeout waiting for phase-change)
ncrscsi0 : Resetting SCSI-bus (Timeout waiting for phase-change)
panic: union_lookup returning . (0x97ceb0) not same as startdir (0x794600)
Stopped at      _Debugger+0x6:  unlk    a6
db>


Configuration:
        -current kernel as of 15 May

        ncrscsi0        at obio?                # SCSI NCR 5380
        #ncr96scsi0     at obio?                # SCSI NCR 53C96
        #sbc0           at obio?                # MI NCR 5380 SCSI Bus
Controller

        /dev/sd2g mounted at /usr/src (Fujitsu MO)
        /var/obj union-mounted on top of /usr/src

The Fujitsu MO spins down when it is not accessed for more than half an
hour. When you wake it up it spins up -- this is transparent to the host or
at least meant to be. The sleep/wake up mechanism works just fine under
MacOS and it used to work under older kernels.

>From the sound that the MO makes I suppose it does not like being reset
while spinning up: it restarts the spin-up just to be hit by the next scsi
bus reset.

Any ideas anyone? Any timeout parameters I could play with?


	hauke

---
"It's never straight up and down"     (DEVO)