> hi mikol, collision is nothing to do with the segmentation fault error...
> collision is on layer 1 (physical layer) while segmentation fault is on
> layer 7 (application layer) of the osi layers... counter-strike version 1.3
> is pretty much stable as far as i know.. 
> > How do I resolve collisions?
>
> convert your hub into *switch*.

In my previous workplace, we resolve collisions partly through trial and 
error. We've been able to stop clients that cause collision by re-crimping 
the connectors. We've found that some collisions are caused by improperly 
crimped connectors, the ones where the connector's gold teeth doesn't quite 
sink into the wire. We were using a cheap (P 700) crimp tool then. It was 
hard to get a good crimp. We would strike the tool with a hammer for added 
pressure. Sometimes it did sink the gold teeth quite well, sometimes the gold 
teeth cut through the wire, but the teeth got skewed.

So we bought a pressure-multiplying (is this the right term?) Siemon crimp 
tool (P 8,000) and re-crimped all connectors. We got good results with this 
tool. The gold teeth sinks deep and really cuts through the wire. The gold 
teeth are likewise straight and not skewed. This improved the connections 
considerably.

We've also found that some collision-causing clients have disconnected/faulty 
strands. The wires are Alcatel UTP CAT5. Some of the wires that we didn't 
enclosed in PVC or GI pipes got chewed by rats. The UTP wires must have 
tasted like Pringles to them. We replaced these wires with good ones and 
insulated them well. This also stopped the collision.

Now I know that aside from improper crimping, loose connectors, proximity to 
power lines, and faulty wiring, the only step left to resolve collisions is 
to get a switch. Thanks.

Before I forget, let me squeeze in this question so I don't pass the blame 
again to the OSI layers if I run into problems.

Is it okay to coil patch cords? (Rick Moen must be laughing right now with my 
English) We sometimes wind patch cords so it's neat, tidy, and doesn't sag. 
I've heard that coiling patch cords is bad. The coil becomes a rough 
electromagnet, causing interference and mucking up the signal inside the 
patch cord. Is this true?

> my main suspect with your problem
> is that you had a bad ram in which there is a portion in your ram that
> probably at the higher address where the bad memory occured which causes of
> segmentation fault everytime counter-strike program is trying to write in
> it... to see if my suspect is correct try to replace a temporary good ram
> and observe.

I enabled memory-parity checking in the BIOS. Will this correctly check the 
memory for bad addresses? My BIOS once flagged a faulty memory module, and 
wouldn't proceed with the boot-up process, so I replaced it with a good one 
and it's been fine ever since. The memory in my Counter-Strike server runs 
fine and doesn't get flagged by the BIOS. Maybe I should download those 
lm-sensors from the net to make sure.

The server seldom segfaults (<-- sounds like a limerick) and runs fine most of 
the time. I couldn't predict when it will segfault again.

Am I correct in assuming that the server crashes because collisions forces it 
to recalculate more than usual to compensate for the increased latency of 
clients and to accurately reconcile clients' hits? This increase in 
calculations goes way out of hand for the server to handle, so it segfaults. 
Is this correct? I've also read that the Counter-Strike 1.4 update has a bug 
fix that makes the server default to 32MB heapsize. Is this bug-fix related 
to my problem? I don't know what a heapsize is. Please pardon me for my 
ignorance. I'm just an accountancy graduate, but I'd like to know more.

By the way, I run the server with sv_clienttrace variable set to zero. 
Counter-Strike internet servers sets this value to 1 to compensate for 
internet-connection lag. I 
disable sv_clienttrace to prevent the telescope cheat (getting shot even if 
you're not in the center crosshairs of their telescopic sight), but I have a 
hunch it adds extra work to the server. With sv_clienttrace set to zero, the 
server has to send extra packets to confirm that indeed a hit has occurred 
from the time the bullet left the client to the time it hits another client. 
The server compensates for the lack of error-resolution with the UDP 
protocol, and the collisions adds more overhead to the error-resolution 
process. Is this a correct assumption? 

I might send my query to Eric Smith of Valve Software later, but I guess the 
Linux gurus in this mailing list wouldn't mind sharing their most-esteemed 
expert opinions. Thanks in advance for more helpful explanations.

mikol
_
Philippine Linux Users Group. Web site and archives at http://plug.linux.org.ph
To leave: send "unsubscribe" in the body to [EMAIL PROTECTED]

To subscribe to the Linux Newbies' List: send "subscribe" in the body to 
[EMAIL PROTECTED]

Reply via email to