Re: linux-ipsec: Small Memory leek detected in KLIPS setup/teardown

  You asked to know where memory is leaking, here is what I have seen. Since when a userspace process is finished all of it's memory _should_ be reclaimed I presume that the losses are in kernel space.

  Using the /proc/meminfo as a metric, memory first comes out of 'MemFree'. When that gets down to about 640KB memory starts comming out of 'Cached' and a bit out of 'Buffers'.   Then 'Buffers' gets hit big time, then 'MemShared' starts taking hits as well.
  Lastly if you have not turned it off (like I did for sevral of these tests) 'SwapFree' starts loosing it when the others are all below about 400KB.

  By the time one reaches the swap space the system is crawling, this is quite the failure mode and one that maybe the Linux VM people should study. There is plenty of swap at this point but the system is totaly hosed in any case and the disk is running full at bandwidth.

  Oh bad, a reboot command sends one into never never land, ghods I hope this will not mess up Watchdogs! The system still forward packets, and some command line stuff works, but the system does not reboot. Worse, a three fingered salute does not get it's attention in this state! Linux 2.0 does not fail gracefully when it is out of VM!

  In my current state I do not grok how to run a test that only has spi or spigrp commands. If there was a way to un-spigrp I could do it, the way things are I am not shure what you want.

  So, there are still leaks. Whee.

		||ugh Daniel
		hugh@toad.com

			Systems Testing & Project mis-Management
			The Linux FreeS/WAN Project
			
http://www.xs4all.nl/~freeswan