You are not logged in.
Pages: 1
Hey guys!
I'm using an Arch box as VPN-server and router at home. My private subnet is 10.5.5.0/24 on tap0. My WAN connection is eth0. Iptables looks like this:
-t nat -A POSTROUTING -s 10.5.5.0/24 -o eth0 -j MASQUERADE
The setup works just fine. In the past, I was able to do `cat /proc/net/ip_conntrack` to get a list of all active connections on the NAT. After the last kernel (2.6.37-ARCH #1 SMP PREEMPT Sat Jan 29 19:40:04) it looks like this:
[root@Taurine ~]# cat /proc/net/ip_conntrack
cat: /proc/net/ip_conntrack: No space left on device
Mount shows the following:
proc on /proc type proc (rw,nosuid,nodev,noexec,relatime)
The NAT table is not full. I can still cat any other file but this one from /proc... I have been looking for a solution for this problem for the last 2 days. Does anyone know what may be the cause of this?
Offline
Interesting one...!
Does dmesg or /var/log/messages reveal anything relevant?
Are you familiar with our Forum Rules, and How To Ask Questions The Smart Way?
BlueHackers // fscanary // resticctl
Offline
Yep, I fully agree that 2.6.37 has broken ip_conntrack. I've seen that 2 days ago (reported by Tarqi - a user of the vuurmuur firewall that I maintain) and looking at the vuurmuur code it's really an issue with /proc/net/ip_conntrack (no space left on device).
By doing a 'conntrack -F' I can empty the conntrak tables but as soon as I use again the network it fills up and becomes full (I guess) directly.
@nullvoid: have you already filed a bug for this ? if not I will do it.
Anyway, it's something to be checked upstream.
@fuckawi: in dmesg I can see plenty of vuurmuur messages (I can't post them right now as I'm at work) but it's things like vrmr: DROP no SYN...
Last edited by dobedo (2011-02-10 08:05:59)
Offline
Not much:
[root@Taurine ~]# cat /var/log/messages.log | grep -i conntrack
Feb 9 18:26:59 Taurine kernel: nf_conntrack version 0.5.0 (15901 buckets, 63604 max)
Feb 9 18:38:21 Taurine kernel: nf_conntrack version 0.5.0 (15901 buckets, 63604 max)
Feb 9 18:40:06 Taurine kernel: nf_conntrack version 0.5.0 (15901 buckets, 63604 max)
[root@Taurine ~]# cat /proc/sys/net/ipv4/netfilter/ip_conntrack_max
63604
[root@Taurine ~]# cat /proc/sys/net/ipv4/netfilter/ip_conntrack_buckets
16384
Edit: @dobedo: I haven't filled any bug report, please go ahead and do it :) Thanks for the answer btw
Last edited by nullvoid (2011-02-10 08:04:59)
Offline
Ok, will do tonight when I have correct info to provide.
Offline
I wonder if it's upstream or Arch specific... What is the config on your systems? I'm still on 2.6.36 and I've got this:
$ zgrep CONNTRACK /proc/config.gz
CONFIG_NF_CONNTRACK=m
CONFIG_NF_CONNTRACK_MARK=y
CONFIG_NF_CONNTRACK_SECMARK=y
CONFIG_NF_CONNTRACK_ZONES=y
CONFIG_NF_CONNTRACK_EVENTS=y
CONFIG_NF_CONNTRACK_AMANDA=m
CONFIG_NF_CONNTRACK_FTP=m
CONFIG_NF_CONNTRACK_H323=m
CONFIG_NF_CONNTRACK_IRC=m
CONFIG_NF_CONNTRACK_NETBIOS_NS=m
CONFIG_NF_CONNTRACK_PPTP=m
CONFIG_NF_CONNTRACK_SANE=m
CONFIG_NF_CONNTRACK_SIP=m
CONFIG_NF_CONNTRACK_TFTP=m
CONFIG_NETFILTER_XT_MATCH_CONNTRACK=m
CONFIG_NF_CONNTRACK_IPV4=m
CONFIG_NF_CONNTRACK_PROC_COMPAT=y
CONFIG_NF_CONNTRACK_IPV6=m
Are you familiar with our Forum Rules, and How To Ask Questions The Smart Way?
BlueHackers // fscanary // resticctl
Offline
I've got exactly the same on 2.6.37
Last edited by dobedo (2011-02-10 11:59:00)
Offline
Looks exactly the same to me. Running 2.6.37.
Offline
Confirmed. No NAT, just a simple workstation. In case someone can read russian, here is a link to Ubuntu forums: http://forum.ubuntu.ru/index.php?topic=123722.0
However, those people never bothered to file a bug report...
Last edited by Leonid.I (2011-02-10 14:52:16)
Arch Linux is more than just GNU/Linux -- it's an adventure
pkill -9 systemd
Offline
Ha, yes I google_translated that one. There is also that one which, I think, is the same problem: http://osdir.com/ml/debian-bugs-dist/20 … 01478.html as I have also that error with conntrack.
Anyway, I still have to file a bug report...but I'm still at work for few hours.
If anyone wants to do it for me, please feel free.
Offline
I've got exactly the same on 2.6.37
Must be upstream then I reckon... That's a pretty big bug to sneak through!
Are you familiar with our Forum Rules, and How To Ask Questions The Smart Way?
BlueHackers // fscanary // resticctl
Offline
archlinux follow-up: https://bugs.archlinux.org/task/22847
upstream: https://bugzilla.kernel.org/show_bug.cgi?id=28862
Note: I faced a niiiice trace in dmesg: WARNING: at mm/page_alloc.c:1990 __alloc_pages_nodemask+0x4be/0x600(), etc... see the report upstream for more info.
Don't hesitate to suggest info to better report the case upstream as I'm not used to that.
Cheers
Offline
Pages: 1