FreeS/WAN 0.8 as a SG hard crashes Linux kernel, whee.

  Fun time this morning:

	  Hard crash on west.  First it griped about being out of
	memory for init or some such, then I got a screen full of
	"Aiee: scheduling in interrupt 001246d1\n" and it was so hard
	crashed that Ctl-Alt-Del would not wake it up!  The hardware
	reset worked just fine.  I guess it's time to look into them
	hardware watchdog boards...
	  At the time of the crash I was flood pinging from sunrise to
	sunset waiting for the replay counter to let packets from east
	to west (west had been rebooted while east had seen a few
	minutes of flood pings previously).  The machine had been
	rebooted in the last few minutes and had run nothing wose then
	xterm or ipsec_systat.
	  Humm, I wonder if what crashed it might have been syslog
	queueing up some ipsec related  message for the log file?
	  Bingo, got it to happen again!  The console looked something
	like this:
		swap_in:
		Out of memory for update.
		swap_in:
		Out of memory for klogd.
		swap_in:
		Out of memory for syslogd.
		swap_in:
		Out of memory for bash.
		swap_in:
		Couldn't get a fre page.....
		Kernel Panic: skput:over: 0018b68c:150
	and then it hard crashed again.  Whee.
	  I was doing the same thing, a flood ping from sunrise to
	sunset but this time I rebooted east insted of west.  I will
	try it again to see if I really know the way to casue this and
	can see any reason.
	  Bingo again, third time.  I shoud have put in a sync in my
	memory status dumping script.
	  Forth time is a charm.  I was even able to grap a cut &
	paste of the very last output of the status dumping script I
	wrote.
	  It really is the kernel and not any of the various
	processes as I had a rolling ps and /proc/meminfo running and
	these are what changed (start, last):

	        total:    used:    free:  shared: buffers:  cached:
	Mem:  15118336 14745600   372736  6402048  5906432  4337664
	Swap: 133885952        0 133885952
	MemTotal:     14764 kB
	MemFree:        364 kB
	MemShared:     6252 kB
	Buffers:       5768 kB
	Cached:        4236 kB
	SwapTotal:   130748 kB
	SwapFree:    130748 kB


	        total:    used:    free:  shared: buffers:  cached:
	Mem:  15118336 14888960   229376   339968   233472   294912
	Swap: 133885952  1994752 131891200
	MemTotal:     14764 kB
	MemFree:        224 kB  <--!
	MemShared:      332 kB  <--!
	Buffers:        228 kB  <--!
	Cached:         288 kB  <--!
	SwapTotal:   130748 kB
	SwapFree:    128800 kB

  This shows up a major problem with Linux, no way to dump a memory core to see what went wrong in extrem failures.   Would someone in a free country (ie not the USS of A) mind seeing if they can figure this one out?

		||ugh Daniel
		hugh@toad.com
			Systems Testing & Project mis-Management
			The Linux FreeS/WAN Project
			
http://www.xs4all.nl/~freeswan