Re: High load which eventually leads to lockup

[Greg KH <gregkh@linuxfoundation.org>]

On Mon, Aug 19, 2013 at 09:00:46AM +0200, Joakim Tjernlund wrote:
> 
> -----Greg KH <gregkh@linuxfoundation.org> wrote: -----
> > 
> > On Sun, Aug 18, 2013 at 04:32:23PM +0200, Joakim Tjernlund wrote: > The last week I have had 4
> > lockups which required power on/off. > Before getting there I noticed that the machine was
> > getting slow. >  > top reported high load(5-10) but there was no process consuming CPU except >
> > for migration/0 which were spicing 100% on and off.  > Ping times went up with a factor of 40
> > too. >  > Eventually I got a few entries in the kernel log: > Aug 16 12:40:51 gentoo-jocke
> > kernel: Modules linked in: nfnetlink_log nfnetlink bluetooth rfkill sg isofs ipt_MASQUERADE
> > iptable_nat nf_nat_ipv4 nf_nat nf_conntrack_ipv4 nf_defrag_ipv4 xt_conntrack nf_conntrack
> > ipt_REJECT xt_CHECKSUM iptable_mangle xt_tcpudp ip6table_filter ip6_tables iptable_filter
> > ip_tables ebtables x_tables autofs4 nfsd dm_crypt vboxnetadp(O) vboxnetflt(O) vboxdrv(O) usbhid
> > dm_mod kvm_intel kvm snd_hda_codec_hdmi aesni_intel ablk_helper cryptd xts lrw gf128mul
> > snd_hda_codec_realtek ehci_pci ehci_hcd usbcore snd_hda_intel snd_hda_codec snd_hwdep snd_pcm
> > snd_page_alloc snd_timer snd microcode usb_common radeon e1000e ttm firmware_class ptp pps_core
> > pcspkr > Aug 16 12:40:51 gentoo-jocke kernel: CPU 0  > Aug 16 12:40:51 gentoo-jocke kernel: Pid:
> > 2421, comm: X Tainted: G           O 3.9.11 #1 Hewlett-Packard HP Compaq 8200 Elite CMT PC/1494
> > The virtual box drivers are a huge mess.  Seriously, I really don't even know how the things are
> > able to work at all (virtual table pointers from userspace through to the kernel?)  I spent a
> > bunch of time to try to clean them up and get them into mergable state, but without the ability
> > to also fix up the userspace side, that ended up being a dead-end.  Can you duplicate the problem
> > without these modules loaded?  I'd blame them for any problems you might ever have, given how
> > crazy they are.  thanks,  greg k-h 
> 
> ok you don't trust them even if I haven't used VirtualBox between failures?

They are using C++ within the kernel, and passing C++ objects from user
to kernelspace.  Do you trust that?  :)

greg k-h