Basically LTSP as quoted from wikipedia “Linux Terminal Server Project (LTSP) is a free and open source add-on package for Linux that allows many people to simultaneously use the same computer. Applications run on the server with a terminal known as a thin client (also known as an X terminal) handling input and output. Generally, terminals are low-powered, lack a hard disk and are quieter than desktop computers because they do not have any moving parts.”

LTSP facilities to use a powerful machine shared by many users from different terminals or thin clients. Thin client machines actually don’t do any processing rather than setting up a minimal system to run an X server. Basic principle of an LTSP system are as follows.

There will be a server machine which is having considerable ram, processing power and attached to a network. The client machines are low end machine with few megabytes of ram, low processing power and attached to the same network over LAN through ethernet card. Ethernet cards come with a special chip socket. We can actually flash a chip / ROM containing a minimal OS. We flash a minimal OS in it and call it PXE ( Preboot Execution Environment).
This ROM will setup a Linux kernel and an initial ramdisk atmost of 10MB. Thus the minimal OS will boot and detect the required devices, like ethernet card. The it loads nfs kernel module and establishes network connection through DHCP and remote root filesystem is mounted via NFS as read only. It places symbolic links to the core parts of root filesystem structure like /bin , /sbin, /lib, /usr and other directories are created in the ram and copies from the readonly directory. The system switches the root to the new root filesystem, sets up a minimal system that can run X. Now it remote login to the server machine using XDMCP.

I am implementing a similar version of LTSP with wireless support. The limitation of Wireless card is that it does not have a facility like PXE to boot the minimal OS. So we have to look at some other ways to boot the minimal OS. So I decided to use either a CDRom or USB pendrive to make an equivalent setup like PXE. I am basing Pardus 2009 for my project though Pardus 2009.1 has recently released.

I have been hacking around Linux distros over 4-5 years. My first tryst with hacking a distro was with Knoppix in 2005.
My first task was to hack around the Pardus initial ramdisk scripts. Basically to boot a Linux OS, we need Linux kernel and a filesystem to support it. We use a bootloader like GRUB, Lilo, Syslinux. GRUB is the most advanced one among them now. I will be using syslinux as my client bootloader since we need something very light weight.

A custom compiled kernel or native Pardus 2009 kernel ( 2.6.30-123), an initrd file to act as base filesystem for the WLTSP client. My aim is to implement wireless support and reinventing LTSP with Pardus. I was more enthusiastic to see how it would work such that a client boots from a 10MB ROM and runs a full fledged OS. I wrote a small script to decompress and compress the initrd.img file.

#!/bin/bash

if [ "$1" == "extract" ];
then
	mkdir fsroot;
	cp initrd.img fsroot/
	cd fsroot
	mv initrd.img initrd.gz
	gunzip initrd.gz
	cpio -i < initrd
	rm initrd ;

elif [ "$1" == "bake" ];
then
	cd fsroot;
	find ./ | cpio -H newc -o > ../initrd.cpio
	cd ..
	gzip -c initrd.cpio > initrd.img
	rm initrd.cpio
	rm -rf fsroot
fi

I customized the init script to add few more features, I cut off some of the unnecessary things from the script. Basically an initrd.img file consists of a Busybox filesystem. Busybox is a commonly used initial ramdisk base. There are a lot of interesting things about Busybox. Busybox looks a small real Root filesystem as most GNU/Linux operating systems has. Firing an ls inside initrd would look like this.

slynux@slynux-laptop:~/miniproject/initrd_bake/fsroot$ ls
bin  bootsplash  dev  etc  init  lib  newroot  proc  sbin  sys

/bin and /sbin contain lots of binaries. There are many common utilities that comes along it including ifconfi, grep, sed, awk etc. I thought they are similar to real elf files, like they are individual binaries. Interestingly, in Busybox there is only one elf file. All other binaries are hardlinked to the same It actually checks for argv[0] and finds its name and does its function. That is, suppose if i rename /bin/ls to /bin/ifconfig. It will work as ifconfig rather than showing ls ouput. It is done to keep the system really small. You can also put your own binaries to the busybox environment. But it should be noted that dynamically linked binaries won’t work. You have to compile as static. For that, use -static option for gcc while compiling.

The initrd that came with pardus had no wireless extensions since no kernel modules required to fire up the drivers were present. I looked to to /lib/modules/2.6.30-123 and went through .ko files, modprobe.dep, modprobe.alias. The device probing and corresponding module loading is handled by coolplug, an excellent program written by Pardus developers. Kudos to them. I went to my Pardus 2009 installation, looked into lsmod and figured out some of the required modules and their dependencies. I hacked the initrd again and added the dependencies, required kernel modules. I booted from the minimal system I had. I am using intel wireless card, so I tried to load manually modprobe iwl3945. It worked fine, but no wlan0 interface comes up. I looked into dmesg and found some errors related to crypt auth etc. I troubled me a little to figure out what was happening around. Finally I understood that some more kernel modules are required to make it work. The culprit were “arc4 and ecb”. I added those modules. The wlan0 interface came up. Next task was to port some of the wireless utilities to busybox environment. I decided to use klcc, which klibc compiler to make it small and simple. I tried to port iwconfig first. I failed to do it since it required lots of changes to make it suit for klibc. After trying out for some days, i left it and went through new set of wireless utils called iw. After few tweaks with Makefile, I got iw compiled with klcc. But the problem is that I couldn’t make it static. I fired up ldd to check the dependencies and found that it is dependent on lib-nl (netlink lib). So in order to make it static, lib-nl is also made to be static as libnl.a. So went tweaking libnl Makefile also. Finally i got iw as static bin having huge size around 1MB. It was the first time I was encountering with iw. A funny thing happened, I found iw being very experimental so it was not working as I expected :P. I went back to iwconfig and tweaked the Makefiles to get a static version. I placed iwconfig and iwlist in /bin of initrd and executed a shell to check the wireless. Everything went fine. But when I used ifconfig to set IP or making the wireless interface up, it hangs up. I got clueless. After going through enough googleing and mailing lists, I learned that it was problem with firmware loading. Devices like wireless cards require firmware to work, when they are first used, the kernel will look for a specific location, /sys$DEVPATH/data for firmware data. It is to be made available at request time. Usually, it is handled by Linux hotplug daemon. Currently in most distros, udev is the device node creation and management backend. udev handles the firmware loading. In busybox environment, there is no udev available, so I have to manually do the firmware loading somehow. /proc/sys/kernel/hotplug can be manupulated to run a hotplug application so that, when the kernel does some requests it executes the file that is set as hotplug with some environment variables like $SUBSYSTEM,$ACTION,$FIRMWARE.

So I setup a script to handle the firmware loading to act as firmware loader.

if [ "$SUBSYSTEM/$ACTION" == "firmware/add" ]; then
    [ -e "/sys$DEVPATH/loading" ] || exit 1
    DIR=/lib/firmware
    [ -e "$DIR/$FIRMWARE" ] || exit 1
    echo 1 > /sys$DEVPATH/loading

    cat "$DIR/$FIRMWARE" > /sys$DEVPATH/data
    echo 0 > /sys$DEVPATH/loading
    exit 0
fi

passed it path to /proc/sys/kernel/hotplug

# echo /bin/hotplug > /proc/sys/kernel/hotplug

Required fimware are placed in /lib/firmware. For iwl, its iwlwifi-3945-1.ucode.

Now everything went fine. iwconfig and ifconfig tuned well and wireless connectivity is up.

I setup another machine with Pardus-2009 to act as the server. Installed NFS server.

Next thing to be done is to get connectivity from client, mount an NFS share from the server. Build the newroot and switch to it. I needed to build a small and custom version of Pardus that just requires to run an X server. Pardusman, my GSOC-2009 project came for the help. Oh. I didn’t upload any videos of Pardusman yet ? I will do it in a couple of days. Using pardusman, I built a small iso having size around 130 MB. I extracted the pardus.img squashfs and exported via NFS share. The client could now mount the share using mount -o nfs, nolock, ro 192.168.1.100:/opt/pardusroot /newroot/mnt/client. Now some of the required directories to be writable are copied from the ro filesystem. The init script is setup to handle all things. I added a little configuration parser so that init script will read a .conf file and perform the actions like wifi ip, nfs server etc.

After a couple of days of effort i could setup the minimal system with X server support. I used XDMCP to remote login. So the remote login worked pretty well. Now the client can login to a beatiful KDE 4 desktop

It wasn’t that slow as I expected through wifi. It was working considerably well. Now it does everything from the server. Now as the final task, I need to forward the sound and device discovery. Since it logs into the server, once you play some music, it plays it at server. No sound comes out of client. Similarly devices plugged in the client are not shown, it shows the storage devices from the server only. Pulse audio came to help for sound forwarding. Pulse server is an awesome sound server that can be used to reroute sound. I also added pulseaudio server to the base pardus.img. Pulseaudio is very advanced so that it can run in session mode and system wide mode. We need system wide. Enough configuration tweaks are to be done in /etc/pulseaudio/system.pa to allow the tcp via authentication over a network. In the server side, /etc/X11/ a script is added to setup the PULSE_SERVER env variable to IP address from $DISPLAY. Now the X will be started by passing the environment variable PULSE_SERVER=client_ip:port. Hence sound will be rerouted.

Next and final thing to be done was forwarding devices ~/Drives directory is created in user home directories. lbuscd and ltspfs came to help. ltspfs is a fuse extension like sshfs which and remotely mount a filesystem. lbuscd can handle a pseudo fstab so that it can listen and mount,umount devices as it comes. lbuscd is a daemon that listens to /tmp/lbus.fifo, which is a first in first out file. A udev rule is setup such that once devices like [sh]d[a-z] or [sh]d[a-z][0-9] appears their dev path is notified to another script so that it can update the fifo. The lbusd which listens to it handles the mounting and umounting. ltspfs updates the availability of devices in ~/Drives and synchronises the changes.

Finally the 11MB Client is ready it boots to a full fledged KDE 4 powered desktop in a couple of minutes. I check out the free -m and found that it is taking atmost 150 MB ram usage for caching all the remote readonly filesystem.

Right now I am tweaking it more to make it lightweight. Probably using desktop like XFCE4 will scale down the memory usage to very minimal level. KDE is a heavy desktop environment.

I will push the WLTSP code to github soon.

Happy Hacking.

This is a post which was tagged in draft from the day I’m back from NITC FOSSMeet. But due to hectic life around college and end of the semester, I couldn’t post it.

From FOSSMeet@NITC 2010

From FOSSMeet@NITC 2010

From FOSSMeet@NITC 2010

From FOSSMeet@NITC 2010

From FOSSMeet@NITC 2010

From FOSSMeet@NITC 2010

From FOSSMeet@NITC 2010

FOSSMeet @ NITC was a nice experience meet all the FOSS folks whom we used to meet online frequently. It was a good chance to meet all of them in real time and have fun. One of the surprizes during the event was Praveen’s birthday. His friends and FOSS community surprised him announcing it.

Binny VA and I conducted a split session Web frameworks - workshop. Binny gave a short tutorial on CakePHP and Ruby on Rails. I conducted a short kickstart workshop on Django development.

Blender workshop was one of the highlight of FOSSMeet this time. Many participants got much enthusiastic about it.

Every FOSSMeet looks at having a few new contributors at the end of the day.

See you all folks at FOSSMeet next year

UPDATE:There is a Django development tutorial written by me published in Linux For You Magazine, March 2010.

I will be conducting a basic Django Web development workshop on Saturday 6th Feb. Check out the schedule from the website. [link]

I was being invited to the college being a Google Summer of Coder 2009 to share about my experience on FOSS and how to get started. I spoke to the students about how I started using the Revolution OS ‘GNU/Linux’, how does the community work and all. The students were very new to the GNU/Linux and they hadn’t been exposed to any GNU/Linux technology stuff before. They were really enthusiastic and excited hearing about it. After my talk on how to get started working with open source projects and how it benefits them, they shot me with numerous questions. They were really enthusiastic about it. After the formal session, we had another one hour question and answers session around the auditorium.

I never had seen this much of questions from any of my audiences before. I felt very glad about it. Felt like a mass transformation would happen around TocH students to GNU/Linux. After the session I had been in conversation with the Jayakumar sir, The head of the Department, Computer Science and Engineering. He told that they were trying to build a FOSS cell around TocH and transforming atleast the computer science labs to GNU/Linux.

It was really nice experience to inspire students to Free Software.

I just analyzed the http requests from that page and automated using the curl utility.

Here is the script it feel, it would be useful for many.

#!/bin/bash

customer=CNXXXX
password=XXXXXX

if [ -n "`ifconfig | grep -A 1  wlan0 | grep "inet addr"`" ];
then

curl https://mwcp-ekm-04.adlkerala.com:8001 --data "auth_user=$customer&auth_pass=$password&accept=Login" &> /dev/null

echo "Connected :)"

while true
do
	sleep 300
	curl https://mwcp-ekm-04.adlkerala.com:8001 --data "alive=y&un=$customer" &> /dev/null ;

done

else

echo "Wifi not connected";

fi

Copy down the above script in a text file and rename to internet.sh

slynux@slynux-laptop:~$ chmod a+x  internet.sh
slynux@slynux-laptop:~$ ./internet.sh
Connected  

This script has an initial check for ethernet interface. Here it is given Wifi interface, wlan0. If you use a LAN for connection change wlan0 to eth0 and the message as well

Happy New year to all

The 2009 has end. When I look back to the year 2009, lots of events and moments scroll around my mind. It was a very unique year in my life till now. Till 2009, I was very rigid with my thinking and activities around. I was categorized to be a geek and I lived around that circle. But there happened lots of changes and the environment was totally different in the year 2009. I watched more around the world rather than sitting in front of an electronic box most of the time. i started to become more social. I received more new friends, lots of fun, lots of serious stuff and lot more.
Only one word suits better for that year with respect to my life, its very “Dynamic”. The year brought to me a lot of changes, lots of life learnings, lots of visions. It was really awesome to keep in memory.

2010 started off with nice traits. We had a lot of fun with friends around. I am doing absolutely nothing these days. But I love the very calm and peaceful environment around.

But more techie savvy stuffs are in my workshop TODO, it includes Android, Linux kernel and more fun.

Once again, wishing you all a very Happy new year 2010.

It is interesting to see that around 50 out of 100 entreprenuers are from India.

It is glad that Ram Kumar is chosen to be one of the future 100 entrepreneurs in Asia. Dreamajax has been growing fast in the recent couple of months. They have recently launched WSNIndia, A platform for Wireless Sensor research based on Zigbee protocol. During the time I first visited Dreamajax during last july, it was a small company with 25 people around. Now the employee number has grown around 100 and they are expanding in more cities.

http://www.ted.com/talks/pattie_maes_demos_the_sixth_sense.html

http://www.pranavmistry.com/projects/sixthsense

Once after I have completed watching this video, a relevent question came to my mind. Why does these kind of innovative research never happens in Indian universities ?

Currently I am the official Web admin at Model Engineering College. Its been a long time since I am elected to be the web admin, but I haven’t been doing any work regarding that being busy with Summer of Coding.

Our official college website http://www.mec.ac.in is a well designed site with a template system and plain html static files + php in the backend. But we don’t have any CMS used to maintain the website. But once you look at the website, it would seem like some CMS is running backend. It the successful design which gives such an impression. Kudos to the Alumnis of MEC who spent lots of sleepless night coding the MEC website and the template. Currently we give edit access to different departments and staff via a web based file manager. We were using phpfm, which is dead project and of consists of numerous bugs. Recent days, I have noticed a good amount of attacks towards the website. I was pretty sure that the phpfm would be the vulnerability made use by the attackers. They injected base64 encrypted shellcode into all the php scripts. Once they are executed, it made copies of html and php files with extension filename.infected.

I removed the phpfm and replaced it with File Thingie.

File Thingies seems to be a easy to use and fine web file manager. It supports multiple user accounts restricted to subdirectories other than giving access to the entire website.

MEC Webserver hosts not only http://www.mec.ac.in, but also several other college websites. So it handles large amount of data. Once some attack happens, it is really hard to recover the data since some proper backup techniques are used. So I have decided to go with decisio to install subversion in the webserver and make all the changes occuring to the website to be written through subversion version control only. It would help to maintain a changelog of all changes which are made in the website/webserver. Once we need some rollback or recovery, within seconds we can rollback to the prevous states. Also it enables to implement several users. It would keep track of users made changes to the website, and what are the changes. It can maintain a history of all changes happening to the website. Once new webadmins come, they can easily trace the history from the changelog and he would be able to learn a lot from the changelog also. Like how some problem is identified, how are they rectified, how is an update made etc.

Anyway I am going to implement the version control system. I haven’t decided whether to go with git or subversion.

These days I have been hacking with my Android phone. It is a great device to play with. Yesterday, I got enough time and I couldn’t stop myself in flashing the default OS with a customised android ROM.

Android is an Open Source Operating System. Hence anyone who loves to play with it can modify the OS and release their on firmware builds. Currently two types of HTC G1 are available. They are Google Dev phone1 and retail phones available with T-Mobile in US. Developer phone is a fully free(dom) phone and is for experimental purposes by build. Other ones are meant for users. Hence they vendor locks the Root access and restricts the users freedom to play with its internals.

Luckly no software is bugfree Android hackers made use of some bugs to exploit and get the root access on the phones. Several rooting methods can be found at xda-developers forum. The methods varies according to the firmware version.

Android also has got GNU/Linux distros like flavour effect these days. You can hear the names Cyanogen, Huykuro, JacX ROMS in the community. These are popular android hackers who releases their own improved android builds.

I used a simple method to root my phone. It is recovery image flashing method.? I downloaded the Recovery Flasher package and installed it. http://zenthought.org/system/files/asset/2/flashrec-1.1.1-20090908.apk

Ran it. and flashed the the recovery image with new cyanogen’s recovery image.

The phone rebooted and I got a bootloader like menu. (Recalls GRUB ).

So the phone got Rooted. It exploited some bug and could managed to get ROOT access. Now that I can install any custom ROM on it.

I prefered Cyanogen’s ROM. Get it from http://www.cyanogenmod.com/

I connected the USB cable and mounted the sdcard. Copied? the firmware image I downloaded from Cyangen’s site and placed it in root of SD Card and renamed it to update.zip

From the recovery flasher I rebooted to recovery mode. Then I selected apply sdcard:update.zip option. I didnt wipe out the ROM to factory settings. The installation went fine. I did a reboot.

Here comes the panic. The phone became stuck at T-mobile G1 splash screen nothing in progress. I waited for around 15 mins. I got tensed whether my phone will get bricked. If the Flashing of the ROM gets corrupted somehow, the phone will get useless like a ‘Brick’.

I went to IRC channel, #android-root and the folks at community helped me out to recover the stuff. God I am saved

The problem was that I didnt wipe out to factory settings. I removed the battery manually. After holding the Home Button, inserted the battery. It straight away went to recovery mode.

Did Alt+w to wipe out. After wiping out, I went for flashing the firmware. It went fine.

It took a couple of minutes to boot for the initial time.

Android has an great SDK with a set of utilities ADB is such a great utility. Download the SDK from here, http://developer.android.com/sdk. I connected the USB datacable and ran the command

slynux@slynux-laptop:~/android-sdk/tools# ./adb logcat | head
* daemon not running. starting it now *
* daemon started successfully *
W/System.err( 1150): ?? ?at com.rechild.advancedtaskkiller.AdvancedTaskKiller.getRunningProcess(AdvancedTaskKiller.java:207)
W/System.err( 1150): ?? ?at com.rechild.advancedtaskkiller.AdvancedTaskKiller$5.run(AdvancedTaskKiller.java:157)
W/System.err( 1150): ?? ?at java.lang.Thread.run(Thread.java:1060)
I/AdMob SDK( 1150): AdMob SDK version is 20090331-ANDROID-cc0d740c1b8c3da5
D/AdMob SDK( 1150): Publisher ID read from AndroidManifest.xml is a14a6ac7fd9f387
I/AdMob SDK( 1150): Publisher ID set to a14a6ac7fd9f387
D/dalvikvm( 1150): GC(2) freed 2368 objects / 164776 bytes in 148ms
I/AdMob SDK( 1150): The user ID is 490BCF661AA5AB2645C71F147068194C

It printed out the background stuffs going on on android. When I did logcat during the first boot after flashing Cyanogen’s ROM I could see some applications installs were going on. That is why it took lots of time for the boot.

We can also get the Root shell on our development machine using,

root@slynux-laptop:~android-sdk/tools# ./adb shell
#
# ls
cache                 init.goldfish.rc      root
data                  init.rc               sbin
default.prop          init.sapphire.rc      sdcard
dev                   init.trout.rc         sqlite_stmt_journals
etc                   lib                   sys
init                  proc                  system
# cat /proc/cpuinfo
Processor	: ARMv6-compatible processor rev 2 (v6l)
BogoMIPS	: 245.36
Features	: swp half thumb fastmult edsp java
CPU implementer	: 0x41
CPU architecture: 6TEJ
CPU variant	: 0x1
CPU part	: 0xb36
CPU revision	: 2

Hardware	: trout
Revision	: 0080
Serial		: 0000000000000000
# cat /proc/meminfo
MemTotal:          97880 kB
MemFree:            2760 kB
Buffers:             416 kB
Cached:            22584 kB
SwapCached:         1292 kB
Active:            36848 kB
Inactive:          38540 kB
Active(anon):      25188 kB
Inactive(anon):    27736 kB
Active(file):      11660 kB
Inactive(file):    10804 kB
Unevictable:         252 kB
Mlocked:               0 kB
SwapTotal:         24464 kB
SwapFree:          13516 kB
Dirty:                 0 kB
Writeback:             0 kB
AnonPages:         52148 kB
Mapped:            15852 kB
Slab:               6044 kB
SReclaimable:        820 kB
SUnreclaim:         5224 kB
PageTables:         4112 kB
NFS_Unstable:          0 kB
Bounce:                0 kB
WritebackTmp:          0 kB
CommitLimit:       73404 kB
Committed_AS:    1125304 kB
VmallocTotal:     811008 kB
VmallocUsed:       57508 kB
VmallocChunk:     676860 kB

Now I have cyanogen’s ROM running on my phone. It is a very nicely customized rom with good performance. The improvements in the ROM includes Multitouch support, better camera options, Faster applications, 5 desktops etc. Anway I am addicted to Android. Trying out more and more hacks these days.

Thinking of trying, HERO ROM for G1