Matt’s Blog

I searched around a bit, but couldn’t find an init.d script for Ventrilo. I found a few, but they all either sucked, or didn’t work. So I picked the best one for a template and improved it. Save this file to /etc/init.d/ventrilo and use the command “service ventrilo start” to start the server.

#!/bin/sh
# Ventrilo script init by Matthew Kemmerer
# For ventrilo server v3.x

VENTPATH=/usr/bin/ventrilo

case “$1″ in
start)
echo “Starting Ventrilo Server.”
cd $VENTPATH
./ventrilo_srv -d
usleep 500000
renice -5 `cat $VENTPATH/ventrilo_srv.pid`
echo “Ventrilo Server Started”
;;
stop)
echo “Shutting Down Ventrilo Server.”
cd $VENTPATH
kill `cat $VENTPATH/ventrilo_srv.pid`
echo “Ventrilo Server Is Now Down”
;;
restart)
echo “Restarting Ventrilo Server…”
cd $VENTPATH
kill `cat $VENTPATH/ventrilo_srv.pid`
usleep 500000
service ventrilo start
echo “Ventrilo Server Restarted”
;;
status)
cd $VENTPATH
./ventrilo_status -c1 -t127.0.0.1
;;
*)
echo “Usage: $0 {start|stop|restart}”
exit 1
esac
exit 0

a lot of great inventions come from people just filling a need. I have a need that I think would be a great invention to add to synergy. I’ve seen people describe synergy as a cross between vnc and a kvm switch. this idea kinda adds more vnc to the mix.

my current setup is 2 computers, one linux and one windows. each computer has dual monitors. sometimes when I’m doing some programming on windows, I find myself really wanting a third screen to put up some code or document or something to look off while I do my work on the other 2 windows screens. usually, what I want to put up on the 3rd screen is something that can only be done on windows, so my 2 linux screens go to waste. the easiest solution for me is to grab my windows laptop and use that as a 5th screen. now it’s getting out of hand. I can only fit 3 screens on my desk as it is. vnc somehow has the answer.

here’s my idea. it’s a bit complicated to implement, so it probably wont be seen for a long time, if ever, but it would be an amazing piece of software. start with some kind of virtual screen driver. something that will add a virtual monitor to the system without actual hardware, but instead outputs only to a vnc server of sorts. the opposite system then connects to and displays that vnc server. it multiplies the number of screens you can have on your system, kinda like what a real kvm switch would do. if you have 2 screens and 2 systems, one screen on each, you could use hot keys to switch between dual monitors on each system without a kvm switch. with my current setup, I could turn my main linux screen into a 3rd windows screen, or the other way around. go from 4 to 6 or 8 screens without more monitors.

is there some way to possibly set this up manually? it would need some kind of vnc server that acts like a virtual video driver. they have something like this in virtualization servers already. I can vnc into the virtual systems running on my server, and they have a virtual display driver. on one of my virtual systems, the driver is called Cirrus Logic 5446 Compatible Graphics Adapter. could I add something like that driver to a system that’s not virtual and doesn’t have a host os to talk to. thanks for taking the time to read my thoughts.

EDIT, 30 minutes later: It was so hard to describe my idea, I didn’t think I’d even know what to search for, but I found a solution!! Check out this software http://www.zoneos.com/zonescreen.htm It installs a ZoneScreen Virtual Display Driver, and shows up in the display properties as another screen. It’s hidden of course. There’s no real display adapter, so no place to plug in a real monitor. It comes with a modified TightVNC server and client to run. I installed the driver and set up the 3rd screen and run the server on my windows xp system, then run the client on my fedora linux system with wine. Go to full screen, and it looks like I have 3 monitors on my windows system! More screen space for me to do my work. I love it!

I was using someone elses wireless router for a while, but they took it to use, so I needed one. I got a used Linksys WRT610N. It supports dual band wireless N, which I think is the fastest connection available right now. Not that I need it, I don’t have anything that supports N. The reason I got it, it has a 4 port gigabit switch build in. 2 devices in one, so I don’t have to spend more money in the future on a gigabit switch. The backbone of my network is extremely fast, but has a limited number of gigabit ports, which were all in use by servers and desktops. Setup only took a few minutes. First, I disable most of the router features, DHCP, NAT, and UPNP. This turns the wireless router into a wireless access point. Those features I disabled would only interfere with the existing network I’m plugging it into. Especially DHCP. If left enabled, it will only confuse the other computers and devices on the network, and loss of connectivity is very possible for devices not even using wifi. Next I set up the wireless secutiry, which is always the biggest pain of the whole thing. Once that’s done, it’s safe to plug into the network and start using it. The wireless access point plugs into the netgear managed switch stack, which plugs into the virtualization server running the router software, which finally plugs into the cable modem. All gigabit. Roughly 10 times faster than a standard home network.

Next, something a lot cheaper than a second hand wireless router. This camping stove is actually made out of a soda can. It weights only 11 grams, a few grams less than an empty soda can. You can use 91% rubbing alcohol or heet gas antifreeze. Both of which can be found at any gas station. 1oz of fuel lasts about 8 minutes, but only takes about 5 minutes to boil 2 cups of water. It fits nicely inside my small camping pot, and a full meal can be made in a few minutes with less than a shot glass of fuel. You’d think heet would smell bad and maybe make the food taste bad. Actually, when lit, heet has less of a smell than rubbing alcohol. I don’t smell much of anything after it burns out. Alcohol has a yellow/orange flame. Heet has a blue flame like pictured.

In my search for the best way to communicate between processes (seperate exe’s), I’ve come across some interesting techniques. I’d like to highlight some of the features and downfalls of what I’ve found for future reference. This quest is just about over, as I’m finishing up my code for this project.

Sockets – The first method that comes to mind is to use a socket. TCP could be used for a stateful connection, or UDP to send information blindly and wait for a response. Localhost can be used for both sending and binding, which avoids hacks from over the network or the internet, but firewall exceptions will still need to be handled on some systems. None of this is very hard to do, but the down side of using a socket is the added overhead on the system. A single port or 2 may not sound like much, but TCP/UDP are complex protocols, which leads me to believe there’s a better way.

Memory Map – A memory map is when an isolated chunk of memory is shared and accessible by multiple processes. I’ve had a prototype project of this working, and seems stable enough to use in the right situation, but timing is a price you’ll have to pay. App A can send information to the shared memory space, but app B doesn’t know of its existence until a timer is triggered to check it. If the receiving app isn’t actively checking the memory space for incoming data, the data will be stuck there clogging the mailbox, so to speak. In this case, your only 2 options are to overwrite the contents, or keep waiting.

Subclassing – Subclassing is when a program inserts its own function in before the normal event handling. Let me try to explain this a little better. Say you click on a window, or type something into a text box. Those events get sent from the Windows OS into the program. Those events can be intercepted by subclassing, sorta like a man in the middle. As it sounds, there is a lot of power that can be harnessed using this method. Along with that power comes a lot of danger. If implemented improperly, your program can crash, or even the IDE (development environment) if you’re running it from there. I’ve found some great code by Paul Caton, on PlanetSourceCode.com, which doesn’t crash the IDE when stopping or pausing the project. I still have yet to see how it performs as part of my full Grow Control project, but I have high hopes. If it fails to meet my expectations, I may have to try moving the subclassing into a DLL or OCX to try preventing crashes while developing.

The second part of subclassing is SendMessage and WM_SETTEXT. This simulates sending an event to the app with a string. This can be captured by the subclass proc function before it makes it to the form. This is easy enough to do with a single line of code. The hard part was grabbing the string back from the captured event. I’ve uploaded my code to PlanetSourceCode, and below is the trick to getting the job done.

l = lstrlen(lParam)
If l Then
Call CopyMemory(sString, lParam, l)
PointerToString = StrConv(sString, vbUnicode)
End If

Mutex – The subclassing method above is what I plan to use, but I think this is worth mentioning too. Mutex stands for mutual exclusion. It’s an api for creating a named lock. Only one mutex can be created with a unique name. (they can be shared, but that’s not what we’re going for here, so nevermind that.) It can’t send strings or commands for communication, but it is very useful for managing the number of instances of an app are allowed to run. Most of the code examples I found on the internet are missing a critical part. You must first run the ReleaseMutex api before CloseHandle. If the handle is only closed and ReleaseMutex isn’t used, the Mutex lock remains until the apps process is ended. This works out fine if you’re only intending to use it to allow a single instance, but it poses a problem if another app is trying to test the lock and not leaving it enforced. Use this api properly and it will prove to be extremely useful.

Handle = CreateMutex(0&, 1&, Name)
If Handle <> 0 Then
If WaitForSingleObject(Handle, 0) = 0 Then
MsgBox Handle
ReleaseMutex Handle
End If
CloseHandle Handle
End If

No more residential internet for me. I now have commercial cable internet. I have a new Motorola SurfBoard 6121. So far I’ve figured out that the modem supports gigabit, and it runs open source linux. I have gigabit ethernet from my desktop computer, into the non-blocking network switches, then from there into the server which runs a linux router, then gigabit from the router built into the server, right into the cable modem. So my computer is pretty much connected into the modem at gigabit speed. This is the first time I’ve ever gotten the actual speed I’m paying for! The old cable internet was 15/3MB download/upload. I now have 8/2 and get better speed. I actually get 1000kbs download and 250kbs upload. Ping rates are about the same as before. 30ms to google.com and 20ms to mattsoft.net . Still waiting for the static ip’s to be set up. By plugging in an unknown mac address to the isp’s network, it gives me a residential ip, and testing shows incoming ports are still blocked using this ip. Once the static ip’s are set, there shouldn’t be any blocked ports anymore! I can use port 80 for a web server, 53 for a name server, email server, game servers, streaming servers.. whatever I want to set up will no longer be blocked. More news on the web server move tomorrow once I’m static.

this is a quick reference guide for ssh public keys

public key - goes on the remote machine that you’ll be connecting to

• ~/.ssh/authorized_keys
• chmod 700 .ssh
• chmod 644 authorized_keys

private key - goes on your local system that you’ll be connecting from

• ~/.ssh/id_rsa
• chmod 700 .ssh
• chmod 600 id_rsa

generate a key:

ssh-keygen -t rsa

this will create 2 files

1. id_rsa – this file can be left on the local machine
2. id_rsa.pub – this file can be copied to the remote machine’s authorized_keys file

this line will install the public key into the remote system automatically

ssh-copy-id -i ~/.ssh/id_rsa.pub username@host

imagine this. you’re paying your internet provider every month for them to provide your internet access, and netflix for their streaming service to the wii or pc or whatever. you go on hulu.com, justin.tv, youtube.com… all the normal sources for your internet entertainment. then a day comes along where your isp (the one you’re paying good money for) sends you a letter saying “you use the internet to much and we’re shutting you down!” bull shit. I pay for the fastest broadband connection I can get in my area. I pay good money for it! who the hell are they to dictate what I can and cannot do on my internet connection! if I was breaking laws, I would have been shut down for that reason, but purely for bandwidth usage? money hungry corporate ass holes.

so here’s my current plan.. commercial internet has finally come down in price a little, and is somewhat reasonable. it costs about twice as much money for a slightly slower connection.. I’ll be on the same exact shared cable internet network, but without their crappy restrictions and blocked ports. it costs them $0 to set up my static ip, unblock the ports, and allow me to use the bandwidth I want, but they make double from me.

they want me to use less bandwidth? ok.. so now I’m downloading many gigs of backups from my webserver, and switching to an account with the same isp that will give me what should be truly “unlimited bandwidth”. I think they failed at that point, but they’re getting more money… so many reasons to call bull shit on them.

I’m installing centos 6 on my server, like I said in my previous post. I was working on it last night and some bad things happened. I did get samba working. all it needed was 2 reboots… but anyway, I had an ext3 partition for /files then my virtual disks and iso’s and mounted hard drives and everything will go under that. during one of my reboots, the partition got corrupted somehow. it was still there but came up unknown and lost the ext3 format. that’s fine, there wasn’t much on it yet anyway. what went so horribly wrong was the drive letters decided to reorder themselves, putting my main boot drive at /dev/sdd, instead of /dev/sda. this changed the letter for every drive I have installed in the system. here’s how to fix it using a UUID (universally unique identifier). first type this command:

ls -l /dev/disk/by-uuid

this will list the UUID for each partition on each drive. in your /etc/fstab file, normally you would type /dev/sda1 to identify the drive and partition. instead, here is an example line using UUID:

UUID=73209d0d-0a63-4553-8c87-1c93428c44b8 /files    ext3    defaults        1 2

now, if the drive letters change, the UUID will still be the same and be updated to point to the same drive.

here’s another quick tip to mounting a windows or samba share to your filesystem:

//hostname/sharename/          /media/sharename            cifs    username=user,password=pass   0 0

that line goes in /etc/fstab then just run the mount command to finish:

mount /media/sharename

I’m installing the new centos 6 on my server this week and I came across a useful command, I thought I’d post it. I’ll post more about the server next.

I was trying to run “yum install qemu” and it was giving me an error “Transaction Check Error:” and a package name. I think what’s happening is the package was installed by rpm and yum wont install over top the old one, so have to remove the old rpm first. here’s how:

rpm -e –nodeps –allmatches qemu

on a side note.. centos no longer comes with xen… sad, I liked it. I started working on installing xen manually, then I changed my mind and went the easier route, so I don’t have to deal with upgrade problems later down the road. something called kvm is the new build in virtualization for centos. still working on installing it. hope all goes well. networking still looks like a complicated pain in the ass, but it does look much simpler then it was with xen! no more custom network bridge scripts I’m hoping, but we’ll see.

EDIT 10/8/2011 – I’ve been running into a problem when I do a yum update. I was messing with a beta version of java at one point and one of the updates conflicted with it. couldn’t find any information anywhere on the specific error, so here’s how I fixed it. “rpm -qa \*java\*” shows me a list of installed rpm’s. then I ran “rpm -e –nodeps <packagename>”. I also had to “yum remove flash-plugin” for some reason, but yum is now updating as I type this. next I hope to install flash-plugin again and hope minecraft still works.

I have some computer venting to do. it saddens me deeply when people don’t listen to me. crappy hardware will give you crappy performance. that, and it’s never a good idea to run a serious server on a residential router. I have to back brazil firewall 100% or more. I’d almost place my life in its hands.

those home routers have crappy hardware to start with. 300-500mhz cpu, and very little ram. grab a decade old computer and clean it up, seriously. those old 500mhz amd’s can almost run without a fan, and they were the leading performing cpu at their time. I loved having one in my desktop at one time. give it 64mb ram or something. that’s even many times more then it needs, but it’s good to have extra no matter what the case. find a 1 or 2gb hard drive and install brazil firewall. that old system is the same, if not better hardware then a standard router.

honestly, I don’t know what the limits are for a linux router like this. I’ve read stories of people running an entire city on one in an emergency, hardware failure. I think it was based in anchorage alaska. I remember it took them until the next day to get parts for the real routers. see if you can find the story online. I’ve read other stories/reviews of people testing this router with OC speed trunks. OC lines are the fiber optic cables that make up the internets network. they are the internet. if you can easily set up a router on an ordinary desktop and have the performance of an OC-192, that’s some serious power to be considering.

the first year I had a brazilfw router set up, I had it go down twice. what happened was, it was an old pentium 133 and the cpu fan died both times. I don’t know how many hours that router ran without a fan before it finally locked up. that system had a 1gb hard drive in it that would load into a ram drive and shut down after it booted. no hard drive spinning, a small cpu fan, and I wired the power supply fan to run on 5 volts, so this system was whisper quiet. now days, I run a more complex setup. I have the same brazilfw software running, but it’s running virtually, with no actual hardware. I love my setup, and I’ll never go back to a linksys. my cable modem plugs directly into a quad core 9000 something.. gigabit from there into a stack of netgear managed switches! this thing handles anything I try to give it. the only limitation is my cable modem. I’ll upgrade to fiber when it’s available.

care to know what I’ve seen it handle on my own network? I’ve had lan parties, torrent server runs constantly, minecraft server, ventrilo, web server and database, 3tb+ file server, pbx, and anything else I can think of, all runs on the same hardware at the same time. downloading torrents and iso’s, 3 hd streams, music streaming, playing video games, browsing the web, and using voip. can do it all at the same time, without any of them breaking up.

I think I’m done with my rant for now. my point, I’ve seen what it can do, and I know it works well, but zero people listen to me.