Crash with new bind/host cmds on -CURRENT (poss. runq_remove?)

[Chris Hedley]

Hi all,

Hopefully I'm not covering old ground with this one, I had a quick look 
through the list and couldn't find anything.  I've been having a few 
problems with either bind9 or the programs which call it via localhost 
which cause the system to panic.  I'm afraid I seem to have neglected the 
debugging symbols, but the function stack is as follows:

#0  0xc04ecd72 in doadump ()
#1  0xc04ed2dd in boot ()
#2  0xc04ed657 in panic ()
#3  0xc0451d2d in db_panic ()
#4  0xc0451cc2 in db_command ()
#5  0xc0451d8e in db_command_loop ()
#6  0xc0453abc in db_trap ()
#7  0xc0506886 in kdb_trap ()
#8  0xc067b039 in trap_fatal ()
#9  0xc067a78d in trap ()
#10 0xc0668c6a in calltrap ()
#11 0x00000018 in ?? ()
#12 0x00000010 in ?? ()
#13 0x00000010 in ?? ()
#14 0x00000000 in ?? ()
#15 0x000000d8 in ?? ()
#16 0xe267dc50 in ?? ()
#17 0xe267dc30 in ?? ()
#18 0x0000001a in ?? ()
#19 0x00000000 in ?? ()
#20 0xc26e77d0 in ?? ()
#21 0xc0712b88 in kseq_cpu ()
#22 0x0000000c in ?? ()
#23 0x00000002 in ?? ()
#24 0xc04ff658 in runq_remove ()
#25 0xc04fdf55 in sched_prio ()
#26 0xc050e4f8 in propagate_priority ()
#27 0xc050ed4e in turnstile_wait ()
#28 0xc04e46ac in _mtx_lock_sleep ()
#29 0xc05630e2 in swi_net ()
#30 0xc04d828a in ithread_loop ()
#31 0xc04d72e9 in fork_exit ()
#32 0xc0668ccc in fork_trampoline ()

and the running threads appear to be for the "host" command.  This panic 
was reproduced by booting it into single user, mounting root & user 
read-only and var as read-write, running dumpon, ifconfig lo0, starting 
syslogd & named and then running "host korma" (where korma is defined in 
one of my master zone files, so it's not going outside my network).  I 
have a copy of the kernel dump lying around, although at 1GB I doubt if 
anyone would thank me for including it as an attachment.  :)

I've been having problems with this for approximately a fortnight with 
systems compiled for both pentium and amd64 (both running on an amd64 
[hardware config below] although I'm still compiling it for pentium by 
preference for the time being), but the kernel I compiled on 21st 
September (cvsup date unknown, but probably no more than a few days prior 
to that at most) seems quite happy.  I haven't included the config file 
(but will on request) to save space, but it's a fairly standard SMP 
config, the only oddity I can think of offhand is that I use gpt 
partitioning.

The hardware is a Tyan S2882 dual mb with a single Opteron 248 (in spite 
of the SMP kernel... 2nd CPU when I can afford it!), 1GB registered ECC 
RAM, 2410SA disc controller, Adaptec 39160 SCSI with just a CDROM 
connected.

I guess since I haven't seen anyone else comment, it must be something 
that I'm doing, but I dunno what.  Is there any other information I should 
include?

Cheers,

Chris.

PS I also found an obscure bug in tr if anyone's interested, although that 
should probably go in its own email.