Subject: Re: softdeps + 2.0_RC4 = crash
To: Jukka Marin <>
From: Brian Buhrow <>
List: netbsd-users
Date: 11/08/2004 10:52:51
	Hello.  I've been running with softdeps enabled on many 2.0_BETA
machines for many months now, and nary a problem.  However, I wonder if
your large swap partition might be responsible for your panics.  I have a
machine which is fairly busy, which has 256MB of RAM, and 2GB of swap
configured.  After it has been running for about a week, it experiences
some event, I'm not sure what, which causes it to spew the following
messages thousands of times for about 2 minutes before it settles down and
continues running without a whimper for months and months:

warning: resource shortage: 1 pages of swap lost
warning: resource shortage: 1 pages of swap lost

	Upon further examination of this message, I found the actual message
in /usr/src/sys/uvm/uvm_swap.c.  The message seems to indicate that there
isn't enough kernel memory to index all of the swap pages, and so it dumps
some of them, making them inaccessible.
	I wonder, because your partition is so large, if, rather than whining,
your kernel is merely panicing when it can't index the swap partition, and,
because softdeps excercise the buffer cache heavily, if you aren't running
into this bug, thinking it is a softdep problem, rather than an inability
for you to manage all of that swap?