Subject: hp380 hangs with high load...
To: None <>
From: Jason Thorpe <>
List: port-hp300
Date: 09/11/1994 16:57:40

Maybe someone has an idea about this...

I have a 380 that just became the CS Dept.'s anon ftp server.  It's 
running a fairly cpu-intensive ftpd (wu's).  What happens is when there 
are about 5 connections the machine's load goes gonzo (these things are 
chewing up roughly 17% of the cpu each - when one drops out, the rest 
pick up the slack...)  Well, after a while of this, I guess the machine 
decides it wants to soon as the drive light begins blinking, 
the machine hangs.  It's not like the kernel hangs, though, because you 
can telnet to the machine and it will connect, and you can ping the 
machine and it will respond.  However, no daemon will fork, and hitting 
return on the console (since getty just restarts on the console with a 
NULL login:) causes the console to hang - like getty can't restart.  This 
leads me to believe that the kernel is having a problem somewhere inside 
of it can't actually create the new process.  To further 
demonstrate this, I can type in a bogus login/password on the console (so 
that login(1) is running, not getty).  If the machine `hangs' before 
login(1) times out, the console still responds, but ^D to get getty back 
hangs the console.

This particular 380 (santiam) has 32meg of memory.  I cannot reproduce 
this behavior on a machine with less memory (even down to 8meg).  I do 
not have a machine with more at this time, but I will be getting a 433 
with 64meg fairly soon.  Interestingly enough, I have seen the same 
behavior on a 370, also with 32meg.

The machine is running sources sup'd roughly 2 weeks ago.  (I have a 433, 
helix, running newer, but I haven't had the opportunity to upgrade 
santiam yet.)  My kernel config follows the sig...

If anybody has any ideas, I'd really appreciate it...

PS...I thought I saw that mycroft made some changes to pmap on the hp300, 
and the i386 too, I believe.  I also notice that the sup scan hasn't gone off 
yet.  Can one of the core folk either run the sup scan or e-mail me those 
changes to I can try them out?  They may be totally unrelated, but I'd be 
interested in giving them a shot, at least...


Jason R. Thorpe                         754-1554
Systems Administrator             CSWest Room 12                     737-5567
Oregon State University

-----cut here-----
# The CS Dept's new FTP server - SANTIAM
machine		"hp300"

cpu		"HP380"		# includes 425t, 425s, 433s
ident		SANTIAM

# Need to set locally
timezone	8 dst
maxusers	32

# Standard system options
options		SWAPPAGER		# swap pager (anonymous and swap space)
options		VNODEPAGER		# vnode pager (mapped files)
options		DEVPAGER		# device pager (mapped devices)
options		"COMPAT_43"		# compatibility with 4.3BSD binaries
options		"COMPAT_44"		# compatibility with 4.4BSD binaries
options		"COMPAT_09"		# compatibility with NetBSD 0.9
options		SYSVSHM			# System V shared memory
options		SYSVSEM			# System V semaphores
options		SYSVMSG			# System V message queues
options		KTRACE			# system call tracing support
options		"NKMEMCLUSTERS=1024"	# 4K pages in kernel malloc pool

# Filesystem options
options		FIFO		# POSIX fifo support (in all filesystems)
options		FFS,QUOTA	# fast filesystem with user and group quotas
options		NFSCLIENT	# Sun NFS-compatible client
options		NFSSERVER	# Sun NFS-compatible server
options		UNION		# Union filesystem
options		KERNFS		# kernel data-structure filesystem
options		FDESC		# user file descriptor filesystem
options		PROCFS		# /proc filesystem

# Networking options
options		INET		# Internet protocols
options		"TCP_COMPAT_42"	# compatibility with 4.2BSD TCP/IP
options		GATEWAY		# IP packet forwarding
options		MULTICAST	# IP Multicast (redundant)
options		MROUTING	# Multicast routing support

# Options for HP hardware
options		FPSP		# floating point emulation for MC68040

# To support the double-autoconf slow hpib disks - LOCAL!!!!! - thorpej
options		SPAM

config		netbsd root on rd0 swap on rd0b

master		hpib0	at scode7
master		hpib1	at scode?
master		hpib2	at scode?
master		hpib3	at scode?
disk		rd0	at hpib? slave 0
disk		rd1	at hpib? slave 1
disk		rd2	at hpib? slave 2
disk		rd3	at hpib? slave 3
disk		rd4	at hpib? slave ?
disk		rd5	at hpib? slave ?
disk		rd6	at hpib? slave ?
tape		ct0	at hpib0 slave ?

master		scsi0	at scode?
master		scsi1	at scode?
master		scsi2	at scode?
master		scsi3	at scode?
disk		sd0	at scsi0 slave 2
disk		sd1	at scsi? slave ?
disk		sd2	at scsi? slave ?
disk		sd3	at scsi? slave ?
disk		sd4	at scsi? slave ?
disk		sd5	at scsi? slave ?
disk		sd6	at scsi? slave ?
disk		sd7	at scsi? slave ?
disk		sd8	at scsi? slave ?
disk		sd9	at scsi? slave ?
disk		sd10	at scsi? slave ?
tape		st0	at scsi? slave ?
tape		st1	at scsi? slave ?

device		dca0	at scode9 flags 1

device		le0	at scode?

device		grf0	at scode?

pseudo-device	pty		128	# pseudo ptys
pseudo-device	bpfilter	16	# packet filter ports
pseudo-device	ite
pseudo-device	loop
pseudo-device	ether

pseudo-device	ccd0 on rd2e and rd3e interleave 8192