I had a similar issue with 2.6.22.9, but as I had a proprietary nvidia
module loaded, I didn't report it. X was not enabled, though.
At this moment, the machine was spawning quite a bit of bash / awk etc.
processes with large variables (50 MB or so), and used memory and CPU a lot.
Normally, it's my desktop machine, and it's rarely on for more than ~12
hours, but this time, I left it on for a couple of days.
After this happened, these bash / awk processes died. After I restarted
the script again, I lost ssh access to the machine, and I saw no more
entries in the syslog. The machine was pingable though, and had it's
network sockets still open (I could telnet to ssh port).
I used SysRq to reboot the machine.
Oct 3 10:14:09 tomek kernel: Eeek! page_mapcount(page) went negative! (-1)
Oct 3 10:14:09 tomek kernel: page pfn = 13aa
Oct 3 10:14:10 tomek kernel: page->flags = 40000400
Oct 3 10:14:10 tomek kernel: page->count = 1
Oct 3 10:14:10 tomek kernel: page->mapping = 00000000
Oct 3 10:14:10 tomek kernel: vma->vm_ops = 0x0
Oct 3 10:14:10 tomek kernel: ------------[ cut here ]------------
Oct 3 10:14:10 tomek syslogd: /dev/tty12: Interrupted system call
Oct 3 10:14:10 tomek kernel: kernel BUG at mm/rmap.c:628!
Oct 3 10:14:10 tomek kernel: invalid opcode: 0000 [#1]
Oct 3 10:14:10 tomek kernel: PREEMPT
Oct 3 10:14:10 tomek kernel: Modules linked in: nvidia(P) iptable_nat
nf_nat ipt_ULOG ipt_recent af_packet nf_conntrack_ipv4 xt_state
nf_conntrack nfnetli
nk ipt_REJECT xt_tcpudp iptable_filter ip_tables snd_seq_dummy x_tables
snd_seq_oss snd_seq_midi_event snd_seq usblp loop dm_mod video thermal
sbs fan cont
ainer dock battery ac floppy cpufreq_conservative cpufreq_powersave
processor snd_pcm_oss snd_mixer_oss snd_via82xx snd_ac97_codec ac97_bus
snd_pcm snd_tim
er snd_page_alloc snd_mpu401_uart snd_rawmidi snd_seq_device snd
soundcore ehci_hcd i2c_viapro i2c_core via_rhine uhci_hcd tsdev evdev
usbcore via_agp agpg
art 8139cp 813...
