help me debug strange network connectivity error

hello,

i need help with a strange network error. where to start… to put it straigt: as with all strange bugs it doesn’t happen always, only affects certain kind of things and has unusual surroundings… and the strangest thing: at my place (i’m at my parents now) everything works fine.

ok: once a day, typically when my PC runs for about 10h+, internet becomes slow. i’m able to ping everything i want without problems. just web sites, email, ftp, ssh and so on become slow. so i went wireshark it. turns out the TCP connection looks like this:

MyPc: SYN en.wikipedia.org
wiki: SYN, ACK
MyPc: RST

so a connection is made, it’s established with a TCP handshake and then my pc RSTs the whole thing. this goes on and on and on until something is received. however, this takes like 2 minutes per website.

i just can’t put my finger on anything. i’ve got 2 network cars (1 onboard) and switched. same. i tried a wlan usb stick and the same happens. i even went ahead and removed the extra network card and nothing. restarting network doesn’t help. reinserting the kernel modules doesn’t help. ONLY A REBOOT HELPS. so i can only assume there’s something fishy with the TCP stack in the kernel. however, i dunno how to debug this. any ideas?

[while my pc doesn’t connect, others - like my tablet - do]

unusual surroundings:
i’m at the country, so there’s no cable internet. so my folks here set up a wireless to the next village, which of course has cable, so theres kinda fast internet. works since 2-3 years and i’ve never had this problem with 11.X before.

opensuse 12.3 (updated yesterday, same with 12.2)
Kernel 3.7.10-1.1-desktop (same with 3.4)
Nvidia tainted… (same with noveou, tried yesterday)
enlightenment 0.17.1 (same with Xfce)
using ifup

Network:
VT6105/VT6106S [Rhine-III] (driver via_rhine)
RTL8111/8168B PCI Express ONBOARD (driver r8169) [currently used]
802.11 n WLAN (driver rt2800usb)

custom kernel settings:
sysctl -w net.ipv4.tcp_frto=1
sysctl -w net.ipv4.tcp_frto_response=2
sysctl -w net.ipv4.tcp_low_latency=1
echo “westwood” > /proc/sys/net/ipv4/tcp_congestion_control

> it doesn’t happen always. . . at my place (i’m at my parents now)
> everything works fine … typically when my PC runs for about 10h+,
> internet … ping…without problems. just web sites, email, ftp,
> ssh and so on become slow… my folks here set up a wireless to
> the next village, which of course has cable, so theres kinda fast
> internet . . . never had this problem with 11.X before.

note: i’m no network guru–this may all be a waste of time:

so, i’ve selected above some of your important points, and thinking
about those alone, and even though you had none of these troubles
with openSUSE 11.x, i still lean toward thinking it is not your
computer, nor its software…

now, of course i could be wrong, but everything you say points me
toward congestion on the network somewhere after your computer…that
is, on the wireless link to the next village or once into the cable
system…

practically without exception both wi-fi and cable speeds and
bandwidth vary greatly according to the number of users trying to
share a pipe with a fixed max throughput size…that is, the more
on the net the slower will be the service to every user…but,
pings will remain stable and reliable because the packages sent back
and forth are so small, pulsed and relatively infrequent…

and, the “about 10h+” also point outside your machine because cable
operators/ISPs have the ability to, and often do throttle back
users at arbitrary times or cumulative data transfer…and, your
“ONLY A REBOOT HELPS” would tend to support this, because a network
restart or kernel module R&R takes few seconds while the reboot a
little longer and maybe the difference in disconnect time is enough
to cause the cable/ISP to reset/restart their clock/count of data flow…

you might notice if your IP changes over a reboot, but not a network
restart…an IP change will most always reset their measurments…

on the other hand, if it is a software or software setup problem i
would really like to hear about it, because i have a similar problem
here: if i leave the DSL modem powered constantly for more than about
24 hours–the net turns to thick cold syrup–until i power down and
then restart DSL modem and router for about 30 seconds or so…(i’ve
decided my problem lies in my ISP, but i’d like to learn different)


dd
openSUSE®, the “German Engineered Automobile” of operating systems!
http://tinyurl.com/DD-Caveat