[Debian] fatal "Cannot allocate memory" when doing "cat /proc/self/mountinfo"

[Corin Langosch]

Hi Hubert,

Am 11.01.2012 10:17, schrieb Hubert Krause:
> I can reproduce the above error message on my Sytem. The system hang
> until the erromessage apears. I am running squeeze too and use the
> latest Debian OpenVZ Kernel. On our site we use puppet inside openvz
> containers too, but no selinux. Did a facter call lead to discover
> this bug? Best regards, Hubert

In fact I'm not using selinux, but when puppet (facter) tries to detect 
if selinux is active it opens /proc/self/mountpoint. So running facter 
or doing a "cat /proc/self/mountpoint" shows the problem. The memory 
flushing does not seem to happen everytime, but most of the time. To 
reproduce I filled the page cache (read/ copy any huge file on the 
host), checked the memory usage with "free" on the host, did the cat or 
facter run inside the guest and checked the memory again on the host 
with "free". Page cache almost went to 0, swap went up to almost 100% 
(before 0) and free memory went up to around 90%.

I tried on a system using centos 6 and it works fine. I also tried the 
kernel und openvz userspace tools from proxmox 
(http://download.proxmox.com/debian/dists/lenny/pve/binary-amd64/) with 
my squeeze system, they work fine too. So there must definitely be a 
serious bug in the debian kernels, which should be fixed asap :)

Corin