netbsd-5 nfs deadlock?

[Hauke Fath <hauke%Espresso.Rhein-Neckar.DE@localhost>]

Hi,

after updating a netbsd-5/i386 nfs server from Februarish netbsd-5, I see
frequent (as in: daily) deadlocks of the nfs subsystem.

The previous build had been flawless, with uptime > 150d, so the current
issue - which is new to me, "usually" machines with kernel locking issues
tend to lock up completely - feels like a regression. I'd rather it locked
up, too - it has a hardware watchdog, but that does not trigger here.

The machine has an LSI MegaRAID scsi controller. Sometimes (not always) it
logs an "amr0: bad status (not active:0x040)" in the context of the
deadlock. Would that be cause, or effect, of the deadlock?

Rebooting the machine hangs while unmounting the disks, so I have to break
into the debugger and issue a "reboot 0x04". The nfsd threads (?) are all
stuck in tstile, save one which is in uvn_fp2. Anything I should look for
while I am there?

Any help is appreciated,

hauke


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