[SOLVED] static ip, first network access slow startup

monojp · 2014-05-26 13:50:13

Hi,

I've configured my network adapter with a static IP and a systemd file described like in https://wiki.archlinux.org/index.php/Ne … figuration
I noticed that after setting my IP and default gateway that there is a latency after which I am able to use my network. It also occurs if I manually do the relevant commands like

ip link set dev eth0 up
ip addr add <ip>/<snm> broadcast <broadcast> dev eth0
ip route add default via <gateway>

after executing this (or simply doing a restart of the systemd unit) and pinging e.g. 8.8.8.8 I have like 3-5 seconds of silence after suddenly networking starts to work:

ping 8.8.8.8
PING 8.8.8.8 (8.8.8.8) 56(84) bytes of data.
From <my ip> icmp_seq=1 Destination Host Unreachable
From <my ip> icmp_seq=2 Destination Host Unreachable
..
64 bytes from 8.8.8.8: icmp_seq=19 ttl=49 time=21.3 ms
64 bytes from 8.8.8.8: icmp_seq=20 ttl=49 time=20.7 ms

At first I thought this may be to a slow / bad DNS server, but it occurs everytime on startup even when using IPs directly.
Does anybody know why there is this strange slow startup or is it just normal? Or where could I start looking next to debug this?

EDIT: problem was my local network. admin says that a slow firewall, switch or antivirus server/proxy is the fault

Last edited by monojp (2014-06-03 12:45:54)

Rexilion · 2014-05-26 14:12:20

Maybe there is still some dhcp server running in the background? Did you check dmesg?

monojp · 2014-05-26 14:41:18

thx for your response. I have dhcpcd and dhclient not even installed. relevant dmesg messages:

[    2.238147] e1000e: Intel(R) PRO/1000 Network Driver - 2.3.2-k
[    2.238150] e1000e: Copyright(c) 1999 - 2013 Intel Corporation.
[    2.238419] e1000e 0000:00:19.0: Interrupt Throttling Rate (ints/sec) set to dynamic conservative mode
[    2.238447] e1000e 0000:00:19.0: irq 45 for MSI/MSI-X
[    2.437997] e1000e 0000:00:19.0 eth0: registered PHC clock
[    2.438001] e1000e 0000:00:19.0 eth0: (PCI Express:2.5GT/s:Width x1) <mac addr>
[    2.438004] e1000e 0000:00:19.0 eth0: Intel(R) PRO/1000 Network Connection
[    2.438033] e1000e 0000:00:19.0 eth0: MAC: 10, PHY: 11, PBA No: 0100FF-0FF
[    2.906886] e1000e 0000:00:19.0: irq 45 for MSI/MSI-X
[    3.008003] e1000e 0000:00:19.0: irq 45 for MSI/MSI-X
[    4.521252] e1000e: eth0 NIC Link is Up 100 Mbps Full Duplex, Flow Control: None
[    4.521259] e1000e 0000:00:19.0 eth0: 10/100 speed: disabling TSO

maybe those disabling TSO messages could relate to something. I will look further into them..

lspci reports that I have the following network card:
Ethernet controller: Intel Corporation 82579LM Gigabit Network Connection (rev 04)

my kernel is: 3.14.4-2-ck
I will also try it with the vanilla kernel instead of -ck EDIT: it's the same there (3.14.4-1-ARCH #1) with enabled ipv6 stack (I have it disabled in my -ck kernel)
It takes almost 20 seconds after "systemctl restart network@eth0.service" for the network to be available again (the restart of the unit and the execution of the ip commands go instantly). inbetween there is no additional output in dmesg or the systemd journal

I think I'm going to try the same tests in a fresh arch live environment to see what happens there..

Last edited by monojp (2014-05-26 15:23:12)

Rexilion · 2014-05-26 15:22:36

It's kind of odd that the irq message is showing up 3 times. Also notice the delay between the third last and second last message (odd).

monojp · 2014-05-26 15:26:31

the multiple occurences are odd, right. I currently don't know where the delay came from, but the first messages could be from booting and the rest from manually restarting the unit. an already idling system gives the following response after restarting the systemd unit (on the vanilly kernel with ipv6 not disabled):

[ 1238.119717] e1000e 0000:00:19.0: irq 44 for MSI/MSI-X
[ 1238.222043] e1000e 0000:00:19.0: irq 44 for MSI/MSI-X
[ 1238.222368] IPv6: ADDRCONF(NETDEV_UP): eth0: link is not ready
[ 1239.863624] e1000e: eth0 NIC Link is Up 100 Mbps Full Duplex, Flow Control: None
[ 1239.863631] e1000e 0000:00:19.0 eth0: 10/100 speed: disabling TSO
[ 1239.863667] IPv6: ADDRCONF(NETDEV_CHANGE): eth0: link becomes ready

that's just 1.7 seconds for all, but an open ping session gives me "Destination Host Unreachable message) for roughly 20 seconds.. (dig is also just hanging)

maybe the e1000e 2.3.2-k driver is just that slow on startup.. I'm compiling and testing a newer version right know EDIT: it behaves exactly the same with version "3.0.4-NAPI"

EDIT
I think it's (just) a problem with my gateway or a local switch, because I can see arp-requests for it's ip in wireshark while the network seems to "hang". I'm asking the administrator

EDIT
yeah it was. admin said that a firewall, switch or antivirus server/proxy is slow. it also happened in a fresh live usb system and I was able to track it down via wireshark. thx for your suggestions

Last edited by monojp (2014-06-03 12:46:15)

Arch Linux

#1 2014-05-26 13:50:13

[SOLVED] static ip, first network access slow startup

#2 2014-05-26 14:12:20

Re: [SOLVED] static ip, first network access slow startup

#3 2014-05-26 14:41:18

Re: [SOLVED] static ip, first network access slow startup

#4 2014-05-26 15:22:36

Re: [SOLVED] static ip, first network access slow startup

#5 2014-05-26 15:26:31

Re: [SOLVED] static ip, first network access slow startup

Board footer