bizarre network problem - can't connect to IPs (solved)

Lightgazer · n00b Joined: 24 May 2003 Posts: 7

Setting up my new Athlon-64 box, I've run into an incredibly bizarre problem: every time I load my ethernet driver, there's certain IP addresses that I simply cannot connect to in any way (DNS resolution works fine, though). Any program that attempts to connect to these addresses fails with the error "connect: Invalid argument."

The odd thing is that the particular addresses are somewhat random - some addresses (forums.gentoo.org) always fail, some addresses (www.google.com) always succeed, and some (open-systems.ufl.edu) sometimes fail and sometimes succeed. Reloading my ethernet driver sometimes changes which addresses fail and succeed.

After puzzling this over for a while, I noticed that my laptop also exhibits this same problem, but my server does not (all three are on my home network behind the same router). Even more puzzling, neither my desktop nor my laptop have this problem when I boot into Windows - all addresses are reachable in Windows. And even MORE puzzling is that my laptop does not have this problem at other locations - from my university, all addresses are reachable for my laptop, even in Linux. My systems are as follows (two dual boot, one is Gentoo only)

Desktop: Gentoo AMD64 / Windows XP Pro (32-bit), nforce4 ethernet
Server: Gentoo x86, Realtek 8139 ethernet
Laptop: Gentoo x86 / Windows 2000, Intel eepro100 ethernet

At first I was speculating that it was due to the piece of crap router I bought from Gateway, which would explain why my laptop worked outside of my home network, but if so, how come one of my Linux boxes works correctly at home? And why are there no problems for any of my boxes in Windows? I can't see any point in common between the desktop and laptop that's not shared by the server...

So confused... :cry:

Lightgazer · n00b Joined: 24 May 2003 Posts: 7

I forgot to mention that my previous desktop box, a Pentium 4 with an Intel eepro100 card that dual-booted Gentoo x86 and Windows XP Pro, never had any problems like this.

gnuageux · Veteran Joined: 17 Apr 2004 Posts: 1201

I agree with your initial guess, the router. When you have these connectivity problems lynx -head -source https://forums.gentoo.org returns an error but host forums.gentoo.org works? Do you have statically assigned ip addresses or do you lease them from your router, provider, ?. Also when you start encountering the problem are you able to reach your gateway?
_________________
The realOTW: http://forums.realotw.org/index.php

Registered Linux user#364538

Lightgazer · n00b Joined: 24 May 2003 Posts: 7

The problems occur regardless of whether I'm using static IP or DHCP - I've tried both. Your example about lynx and host is correct, though I used ping and tracepath for the test access and dig (from bind-tools) for the test DNS lookup - I did it on yahoo.com, another address that didn't work. ping and tracepath (and even Firefox, if run from an xterm) print the same error, "connect: Invalid argument" but dig reported completely correct DNS info on yahoo.com. Also, wget correctly displays the IP address with any site, even if it errors afterwards the same way that ping and tracepath did.

I haven't looked to see if I can still reach my ISP's default gateway, but I assume that I can because even while I can't reach some sites, I can reach other sites - I can always run searches on google and look at google caches regardless of what other problems I have. I also can always reach my router and everything on my LAN.

One difficulty is that I do not know when this issue started. I changed ISPs about a month ago, and about six months before that, got this router. However, since my old P4 desktop and my server always worked, I have no idea if this problem occurred before the ISP change, because I never used my laptop at home. The only reason I noticed the problem on my laptop is because I was doing an emerge at home instead of at school due to getting a new wireless card and needing ndiswrapper.

So while I want to believe it's the router, I just don't understand why a router problem would only cause problems on Linux on an Athlon-64 and a Pentium 4, but not a Athlon-XP or a different Pentium 4, and why network card chipset doesn't seem to correlate with the problem either (both a working and a non-working one used Intel eepro100). And why Windows always works fine.

Lightgazer · n00b Joined: 24 May 2003 Posts: 7

At the risk of going a little out of my league, I looked up further what exactly "connect" is, and found out it's a system call into the Linux kernel. It seems to be returning EINVAL for "bad" IP's and working fine with "good" IP's. So that means the problem is originating in the kernel, or possibly glibc. Thus I believe the problem is either in glibc, the kernel, or an external source like the router or ISP. I just don't know which.

Lightgazer · n00b Joined: 24 May 2003 Posts: 7

Ahahaha... I don't know why this didn't occur to me earlier, but I connected my Athlon-64 directly to my ADSL modem (bypassing the router) and set that up... and everything works fine. So the problem is in fact the router, though I don't understand why. Regardless, I won't be sad to see it go, its user interface is pretty horrid. Sorry for the useless posts... any good recommendations on an 802.11-g DSL router?