[Bug 1186 (linux mm)] Changed - 2.4.1: OOM killer doesn't trigger

[bugzilla-daemon@distro.conectiva.com.br]

http://distro.conectiva.com.br/bugzilla/show_bug.cgi?id=1186





------- Additional Comments From ingo.oeser@informatik.tu-chemnitz.de  2001-03-21 17:07 -------
In ac19 and ac20 this Bug does still trigger.
Scenario:
  A Cyrix III 500 with 64MB
  cramfs over loop file of 10MB size
  3x ramfs with limit of 16M (but not used too much)
  16MB ext2 / (which also holds the loop file)
  20G of other ext2 partitions

System runs iptables over ipchains module doing forwarding and filtering.
Also running dhcpd and an eth0 (8139C, 3c590C-TX both the same results) and lo.

script to trigger:

nc localhost 7 </dev/zero >/dev/null &
nc localhost 7 </dev/zero >/dev/null &
nc localhost 7 </dev/zero >/dev/null &
nc localhost 7 </dev/zero >/dev/null &
vmstat 1

And watch it die in some seconds.

Also tested with ftp to this machine and smb to this machine.

Adding another 64MB of RAM solved all problems.

Using only 10Mbit-Networking also solved it.

So it only shows up with 100MBit connections and maybe higher.

I saw around up to 5000 cs and usally 3000 cs in vmstat with only 2000 ints.

Adding the 64MB of RAM showed only some hundreds of cs.

Can provide even more details on demand.
-
Linux-mm-bugs:  bugzilla list for the Linux-MM subsystem
Archive:        http://mail.nl.linux.org/linux-mm-bugs/
Web site:       http://www.linux.eu.org/Linux-MM/
Development:    linux-mm@kvack.org