Subject: Re: Consistent crashes with linux emulation
To: None <current-users@netbsd.org>
From: Steve Rumble <rumble@ephemeral.org>
List: current-users
Date: 11/05/2003 21:03:48
I believe to have made some progress into determining the
problem. My system appears stable without the following two
sysctl settings:

	net.inet.ip.anonportmin=16384
	net.inet.ip.anonportmax=32767

which I've used for a year or so in order to netboot SGI machines
with broken prom code. 

Additionally, the machine does not actually panic. By running
Firebird over ssh on another client and watching console, the
machine just hangs. I can drop into ddb and no programs are
running, but using 'next' it appears to just loop through:
	Xdoreti
	Xspllower
	tcp_input
	ip_input

It does this over and over, and the machine is effectively hung.
I can only break into ddb, not switch ttys or anything else. sync
or reboot hangs the box hard.

Does anyone have any ideas? Does this ring a bell with some
fairly recent changes that could cause this? I've been using the
same anonport parameters for quite a while and have only recently
been bitten by this. 

I'll keep plugging away for a bit to make sure that reverting to the
defaults produces a stable system before filing a PR.

Steve

On Sun, Nov 02, 2003 at 08:12:04PM -0500, Steve Rumble wrote:
> Ever since I sync'd back up to current a month or so ago (from
> ~1.6W), I've been able to crash my i386 laptop consistently with
> linux emulated applications. Opera causes a lock on extremely
> rare occasions, java will at times, especially when using a GUI
> application while running xmms, and Firebird will do so
> consistently and with very little effort.
> 
> I don't see anything interesting in gnats nor on the lists. Have
> anybody else experienced similar stability problems? Each crash
> has occurred within X, so I can't get to ddb. I've tried booting
> over serial, but I couldn't access the debugger there either. I'm
> not even sure where to begin with this. I don't believe that this
> is due to any sort of flux as it's been happening for weeks
> across many re-syncs. 
> 
> Any insights or suggestions as to what I should do to begin
> tracking this down would be appreciated.
> 
> Cheers,
> Steve