You are not logged in.
Hi all,
after a lot of years with Debian, I am trying this awesome linux distribution and seem I will continue with this
I have installed Arch Linux about 2 weeks ago with systemd as default. All worked fine, but lately, after some updates, I have noticed that some web sites does't show anymore (for example facebook.com). Here some infos:
Firefox neither Chromium can show the content with "connecting" status looping.
I can ping affected sites.
I have a router with MTU set to 1492 (tryed also with 1472).
With Windows in dual booth with my Arch Linux and with the other pc with debian connected to the same router, I can connect so it is not a router problem.
DNS is set to the router IP and the DNS set into the router are those retrieved by my internet provider ( I tryed also with google DNS 8.8.8.8 and 4.4.8.8 ).
# uname -a
Linux HAL 3.6.5-1-ARCH #1 SMP PREEMPT Wed Oct 31 20:57:39 CET 2012 x86_64 GNU/Linux
(also with 3.6.4 there was the same issue)
This is my systemd entry
# cat /etc/network.d/ethernet-static
CONNECTION='ethernet'
DESCRIPTION='HAL static ethernet connection'
INTERFACE='eth0'
IP='static'
ADDR='192.168.1.2'
ROUTES=('192.168.0.0/24 via 192.168.1.2')
GATEWAY='192.168.1.1'
DNS=('192.168.1.1')
I googled around for 2 days without any clue, please anything I can try to fix this ?
I have noticed now that I can browse this forum but cannot submit this message. Maybe a java issue ?
thanks for your time
Marco
Offline
perhaps, you sould use another DNS servers other than your router's ip (Gateway)
by the way, Google's servers are 8.8.8.8 and 8.8.4.4 not 4.4.8.8
I'm, flawlessly, using OpenDNS's servers '208.67.222.222 208.67.220.220
Offline
I have a router with MTU set to 1492 (tryed also with 1472).
Umm, why?
Offline
thanks for your answer
@clickit: yes I used those Google's DNS, I mispelled that . Also with OpenDNS servers I have the same issue. I tryed to set those in resolv.conf and also in the router configuration. By the way, the router configuration, works with windows and the other pc.
@Gusar: I am not sure, but if I remember well, to try to have no fragmented packets that can cause transfer to stall, it is necessary to lower the MTU. But that 1492 value is the default for my router and I think for PPOE connection.
I really don't have any clue on what can cause this problem and tryed that MTU value
Offline
If you don't have a reason to set the MTU to 1492, then don't. Set it to 1500, that's the size of an ethernet frame. It's exactly *because* you've set it to 1492 that you have the problem.
Offline
I never touched that 1492 value in my router, it is its default value and if I try to change, it tell me that a value greater then 1493 could not be entered.
Talking about MTU I tried to play with it in my eth0 card. I never touched that also before. Doing:
# ip link show|grep eth0
2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP mode DEFAULT qlen 1000
shows that MTU is 1500, which is the right value. I tried lowering it down with:
# ip link set eth0 mtu 1492
but the same issue happens. With MTU set to 1472 all is back to normal ! All sites I cannot see before, are now displayed !
Don't know if this issue could be caused only by MTU and I cannot explain why now all is fine now, I never had this problem before.
Also in the other pc, where all is working, the "ip link show" command, tells me that eth0 uses 1500 MTU.
thanks for your time
Last edited by lildeimos (2012-11-05 21:23:46)
Offline
Is your router a dedicated device or a Linux box you put together yourself? When my router was a Linux box, I had this problem, all machines in the LAN needed to set their MTU to 1492. Then I learned about this iptables rule, and voila, no configuration required anymore on LAN machines, things worked even if they used the default MTU of 1500:
-A FORWARD -p tcp -m tcp --tcp-flags SYN,RST SYN -j TCPMSS --clamp-mss-to-pmtu
A dedicated ADSL router should automatically be doing the equivalent of that.
Offline
My router is a Netgear R6300 with a separate D-Link as adsl modem. Some times ago, the router got a firmware upgrade. Can't tell if I had my problem by that time. But this doesn't explain why the other pc (and also this booted into Windows) has no problems.
I don't know if changing the eth0 MTU is the right fix of the problem, maybe some packages update. So should I mark this post as solved ?
thanks again
Offline
For MTU, see FAQ.
I use, with an ADSL Ethernet router:
ifconfig eth0 192.168.1.2 up netmask 255.255.255.0
ip link set eth0 up mtu 1492
Edit: Removed false alarm about failing to connect to http://forums.freebsd.org/
Last edited by brebs (2015-05-03 20:43:53)
Offline
Is your router a dedicated device or a Linux box you put together yourself? When my router was a Linux box, I had this problem, all machines in the LAN needed to set their MTU to 1492. Then I learned about this iptables rule, and voila, no configuration required anymore on LAN machines, things worked even if they used the default MTU of 1500:
-A FORWARD -p tcp -m tcp --tcp-flags SYN,RST SYN -j TCPMSS --clamp-mss-to-pmtu
A dedicated ADSL router should automatically be doing the equivalent of that.
When I try to add the rule to iptables I get the following error:
iptables: No chain/target/match by that name.
How can I add the TCPMSS target to iptables?
**Turns out I upgraded my kernel and did not reboot, that's why the xt_tcpmss module could not load..
Anyway the mtu clamping rule solved the problem.
Last edited by FNC (2012-11-07 13:48:33)
Offline