cinc

Okay guys... help?


The machine:

Sun Ultra 10, no add-in cards

Built-in 10/100 ethernet (more on that later)

512MB RAM (in case anyone thought it might have 32MB)


The network:

GOOD CABLES! (trust me on this one)

Actiontec GT701-WG, AKA. "A really crappy DSL modem". DHCP server running. Firewall and all the fancy crap is off, a few non-standard ports are forwarded to my main system.

3Com SuperStack II Dual Speed Hub 500

Linksys WRT54G between the DSL modem and the 3Com hub because the modem doesn't like the hub very much. All the fancy stuff has been turned off and the Linksys is acting only as a ethernet hub/wireless bridge.

Five other computers running either WinXP, OSX, or Ubuntu 9.04 and using DHCP. No problems of any kind with these.


The horror:

I can ping most sites but some don't return (microsoft, ebay, etc.) but everything resolves correctly. I'm working under the assumption that the sites which don't return pings are doing this on purpose. Most of the time sites which I CAN ping from the command line give me a "connection refused" or "timed out" error from a browser. Some sites work intermittently and others don't work at all; sometimes I get nothing but Debian's site . I'm by no means an expert in the world of Linux but I have general troubleshooting skills and enough sense to read. Here's what I've done, from the beginning:

Ubuntu 9.04 - Can't find the CD drive it's loading itself from

Gentoo ?.? (the newest one) - Same as Ubuntu

Debian 5.0.1 - Someone broke the kernel and it won't start X on the Ultra 10. I haven't tried 5.0.2 because I'm getting tired of burning CDs and running out of faith.

Debian 4.0R8 - Installed! But "connection refused" problem described above

FreeBSD 7.2? (the newest one) - Good download & good burn. Couldn't find some install files.

Aurora 2.0 - Installed. Runs fine. Same "connection refused" problem. This is the distro I'm currently running.

Installed the following network cards simultaneously: 3Com 905B-TXNM, Linksys LNE100TX, Netgear FA311 (rev. A2), and some unknown card with Intel S82557 chip. All but the Netgear were detected but the Intel card couldn't grab an IP address from the DHCP server. The 3Com and Linksys cards recieved addresses just fine but had the same "connection refused" problem. I've tried ethernet cables plugged into both the 3Com hub and the Linksys hub; same problem. Therefore, the issue is almost certainly not hardware.

Pulled all of the network cards I'd added and configured IP manually with an address within my subnet but outside the DHCP server's assignment range (guaranteed to be an unused IP). Same problem as before. IP is still configured manually.

Ran a few more searches. The only useful idea I found dealt with setting "echo 0 > /proc/sys/net/ipv4/tcp_ecn" in sysctl.conf (something to do with ECN). Sorry I can't post links, but the poster says it's not a bug. To my way of thinking software which doesn't work with the majority of available hardware is buggy. I tried it but same problem as always.


Current ifconfig result after managing to load google:
eth0
Link encap:Ethernet HWaddr [I don't want to type all that]
inet addr:192.168.0.231 Bcast:192.168.0.255 Mask:255.255.255.0
inet6 addr: fe80::a00::20ff::fec1::7717/64 Scope:Link
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:493 errors:0 dropped:0 overruns:0 frame:0
TX packets: 402 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:1000
RX bytes:287459 TX bytes:46234
Interrupt:224 Base address:0xf000

lo
Link encap:Local Loopback
inet addr:127.0.0.1 Mask:255.0.0.0
inet6 addr: ::1/128 Scope:Host
UP LOOPBACK RUNNING MTU:16436 Metric:1
RX packets:1686 errors:0 dropped:0 overruns:0 frame:0
TX packets:1686 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:0
RX bytes:1828055 TX bytes:1828055


Results of "route -n":
Kernel IP routing table
Destination____Gateway_____Genmask______Flags__Met ric__Ref__Use__Iface
192.168.0.0____0.0.0.0______255.255.255.0__U_____0 ______0___0____eth0
169.254.0.0____0.0.0.0______255.255.0.0____U_____0 ______0___0____eth0
0.0.0.0________192.168.0.1__0.0.0.0________UG____0 ______0___0____eth0


As I said, I don't know a bunch about Linux but I'm not prone to falling on the ground and squealing for help until I think I've run out of other places to look for solutions. I've reached that point and am therefore calling upon the knowledge of the 'net. If anyone needs more information I'm quite sure I can oblige and if anyone has suggestions other than slamming my fingers in a door or banging my head against a wall I'd love to hear 'em. Aww heck, if anyone actually DOES think finger-slamming or head-banging might work I'll listen to that too.
-Chris

Lazydog

OK, now that you have given all that how about you show use what you are doing.

• What are you typing on the command line to ping a host and what host?
• When connecting how are you connecting? Web, Telnet, SSH?
• When trying to connect to the host are you doing it from the command line or from within a program, i.e., using FF to web to the host.

__________________

Regards
Robert

Linux
The adventure of a life time.

Linux User #296285
Get Counted

cinc

Ack. I knew I'd forget something, and thanks for the reply . Right now I'm doing most things from X and only using the command line for diagnostic purposes...

Let's see if I can get around the forum 'can't post links yet' limitation. To ping I'm typing "ping www{whatever}com" (obviously with dots in place of the brackets) where 'whatever' is 'google' or 'ebay' or 'youtube' or 'opera' or any other site I can't get to. Some sites I use as standards because they're horrifically complicated and help identify general connection problems which wouldn't show up loading simpler sites like google or craigslist. As I said, all the websites I've tried will resolve to IP correctly and the only ones I can't get a return ping from are the sites which I suspect are worried about being pinged to death. IP resolution and ping returns are being checked with a computer which is running XP and which has no network problems.

The load failures are coming from Firefox (in the case of Aurora) or both Epiphany and Iceweasel (in Debian 4.0) so no telnet/SSH/command line. And to reiterate, occasionally google and the less complicated sites will load in the browser, though never more than one or two of them at the same sitting. The complex sites I use for testing (microsoft, youtube, ebay, etc.) have never loaded in the browser. My install of Aurora doesn't have Lynx and trying it hadn't occurred to me during the Debian period so I can't help you there at the moment.

And yes, along with the other ethernet cables I've tried I did try several from computers with known-good internet access . Since I've made my own cables for some years now I've managed to run into all sorts of weirdness... but not this time.
-Chris

Lazydog

OK, it sounds like your problem is just with the Linux setup and this is normally because something is not setup correctly. Lets do some thing to see how you are configured.

Please provide the following:

• Are you using DHCP or Static IP's
• What is the default GW for the linux box
• Post the out put from ifconfig
• Post the output from /etc/resolv.conf
• How is the linux box connected to the network

We will start there and work forward.

__________________

Regards
Robert

Linux
The adventure of a life time.

Linux User #296285
Get Counted

cinc

Hi again. In my initial post I made note of my attempts to use both DHCP and static IP. DHCP is being served from my DSL modem (192.168.0.1) and when I tried a static IP I used the modem's IP for the gateway and DNS address. Bear in mind that when I'm setting the static IP configuration I'm using the graphical network configuration utility and not editing config files directly (mainly because I don't much know what I'm doing in Linux config files). When it was set to use DHCP it grabbed the modem's IP for gateway and DNS and also grabbed the IP of one of my ISP's DNS servers which it added as the secondary DNS. The problem remains the same whether I let it use DHCP or I configure it myself and it's currently still configured statically.

The output from 'ifconfig' is also in my initial post and I haven't changed anything since then.

The only entry in /etc/resolv.conf is "nameserver 192.168.0.1"

I also generally described my physical connections (and the different physical configurations I've tried) in my first post, but to reiterate: The Ultra 10 is connected to the 3Com SuperStack hub, the 3Com hub is connected to the Linksys router (with all of the router functions turned off; it's acting only as a wireless bridge and a hub), the Linksys router is connected to the DSL modem. I've tried it hooked up directly to the Linksys router but that made no difference and I've tried known-good cables. At the time that I had the four ethernet cards installed (two of which the system could use) I had five total cables coming from the machine, some connected to the 3Com hub and some connected to the Linksys router. Since nothing I've tried on the physical side made any difference I'm currently back to just the built-in ethernet connected to the 3Com hub as described above (the same physical configuration as all the other computers on my network) and configured with a static IP, also as described above.
-Chris

HROAdmin26

So you are running the Sparc version of these Linux distro's? I don't see an Ubuntu 9.04 built for Sparc on their main download page? Is it available somewhere else?

Sun Ultra 10 Specs

If you are getting successful connections, then failed connections, you *most likely* have a HW problem. Software (unless you change it) will do the same thing again and again. In this scenario, it's far more likely a physical problem is occurring. HW problems can be internal to the Sun box as well.

You listed several distro's - many of which use different ways to setup networking. The one you're currently using (Aurora?) is not a mainstream distro and you would get more help using something more people are familiar with.

cinc

Yep, the Sparc version of Ubuntu 9.04 is available at cdimage{ubuntu}com/ports/releases/jaunty/release/ (still can't post links so replace the brackets with dots) and is under the "Alternate install CD" section along with some other interesting ports. I also tried the Sparc port of 8.10 but no luck there either.

I've done all I can think of to isolate a possible hardware problem (see what I've written about multiple ethernet cards) and nothing changes no matter what the physical configuration. Having seen all sorts of weird harware problems over the years that was the first thing I suspected, but if you can think of anything I haven't done I'd love to hear it. I even tried putting the Ultra's current static IP on the DMZ in my DSL modem (to open ALL ports) and still had the same problem. At this point the only thing which seems logical to me is a software configuration problem in Linux itself; everything outside the Ultra seems fine and since I've tried multiple different e-net cards it doesn't seem like hardware inside the system is causing problems. Unless, of course, the PCI drivers Linux is using are weird for some reason.

The differences in software network implementations (and possibly different driver sets) is one of the reasons I tried so many distros. Since every distro I've been able to try (effectively, Debian and Aurora) has behaved the same I'm starting to think the problem might be kernel-based (etc.) rather than based in software which is specific to a certain distro. Aurora 2.0 is actually a Sparc port of Fedora Core 3 (since Redhat itself dropped Sparc support) and so it's a bit more mainstream than it seems. While it IS a little old I didn't think that'd be a bad thing since I'm working with old hardware. For now it looks like I'm stuck using either that distro, Debian 4.0, or giving up and installing Solaris 10 (which is unusably slow on that hardware).
-Chris

HROAdmin26

If I *really* wanted to used this HW, I would install Solaris 9 (or maybe an older openSolaris) in order to confirm this as a HW/SW issue. Solaris 8 will run fine, 9 will be "a little slow," and 10 is impossible. This is assuming as well that you disable the CDE desktop and just use command line.

cinc

Got tired of rewriting things which were already in my initial post, decided that nobody here was interested in helping, and already installed Solaris 7 (11/99). Everything works as it should... hadn't I already mentioned that it wasn't a hardware problem?

Conclusion: "Linux" doesn't work. It didn't work twelve years ago when I started playing with it, it didn't work four years ago when I gave up on it the last time, and it doesn't work now. Maybe in another decade or so...

But that doesn't excuse your silly unhelpful attitude or borderline illiteracy. I do believe I'll be finding another forum and I really do hope you don't do tech support for a living.

HROAdmin26

Re-read your previous posts genius - none of them show this machine ever in a working state. Take your word for it?

Having run Linux on Alpha, Intel, IA64, and Sparc hardware, I *know* it works. I guess it's an amazing thing that Amazon, Google, Ebay, etc. have somehow made a business decision to use something that "doesn't work."

You also seem to be under the misguided perception that someone "owes" you something - what a joke.

If you weren't so clueless, you would be checking name resolution when connections fail vs. when they connect, using traceroute/ping, and collecting packet captures with tcpdump/wireshark.