The Curse of the Gravel Machine

Trudging through college life, we come across several obnoxious myths, stories and legends belonging to our college. One such legend I came across is that my college is a cursed one. The curse is that the moment it stops constructing something new, building a new annexe block, pulling down a perfectly nice place and building another... the moment it stops constructing... It will cease to exist! Just like that. God's WILL.

Obviously its obnoxious and untrue, or so we'd hope as long as we're here. But the construction bug has bitten my college since before the guards were toddlers (or so they say), and you're bound to collect different proportions of sand, cement and pebbles in your slippers depending on which path you take across the hostel campus.

Oh and I'm on a break from blogging, which is an odd statement to make since I blog in intervals of decades anyway. I'm also on a break from nosy-technical-fun-&-games since there are a couple of exciting new projects I and my friends have embarked upon which is taking up all my time.

Cheers
Shashank

Some sad news

Been busy with exams, and work on the latest project which we're submitting for this year's Google Product Prodigy. But this post is not about that.

This post is about the tragic passing of my friend, Ishan Goel, who passed away a few days back. He was just 20. He was a colleague in the Sun Ambassador programme, i.e. he was my colleague at the Sun Open Source club we were founding here at college, and worked very hard with me in making the Software Freedom Day (September 2008) a success, and also a member of our last year's Google Product Prodigy team (which I covered in great detail in a series of blog posts previously), which we worked for close to 6 months together. A brilliant and immensely creative fellow, and a "Linux Helpline". This is not an obituary, or any formal note of sort. It's just hard to absorb this. My hand still tries to dail his number whenever a problem with my Linux system occurs.

~Shashank

Hello World! iPodLinux - Java style!

October's been a good month for me, and better for my iPod. iPodLinux.org is back online and they've released a couple of pretty neat tools, and updated builds. Loader 2 (the new bootloader) is great - no more messing around with complicated wheel/key combinations to select the boot OS. Loader 2 gives you a GRUB-like list of installed OS images on the iPod disk, and can read kernel images from ext2 and ext3 partitions, as well as FAT32 (W95). So this effectively removes the need to re-create the firmware each time a kernel update takes place. All you need to do is use ipodpatcher (which originally was used for installing Rockbox) and install Loader 2, which will now get merged with the Apple Firmware.

The Podzilla interface has been improved too, and lots of annoying little problems have been ironed out mostly. The text-input is (or atleast, feels) WAY better now, and I can actually input something meaningful in a meaningful amount of time. Here is the text file I "typed" (with awkward scrolling and tapping) on my iPodLinux's PodWrite with a "Scroll with Prediction" text-input mode:

first.txt : "This isn't bad at all. The prediction algo is pretty good as far as this text input session is concerned. Took me 10 minutes to input this on the iPod, but I'll improve I guess."

Then I finally got my first Java program running on the iPod! After encountering wierd errors and battling nasty iPod text-input mechanisms for months, things have finally come to a closure. The latest build of podzilla, makes things easier, and thus enabled me to run Java programs using the K Virtual Machine (kvm) by Sun Microsystems. I got hold of the kvm binary from some place, but I don't think its available on the Sun website anymore, which is pretty odd. But this is how I got my "Hello world" in Java working on the iPod:

1. Install iPodLinux by using the latest Podzilla and loader.
2. Download kvm from here. Put the kvm binary into your iPod's /bin folder and make sure it has execute permission.
3. Write your java code, but keep it simple. KVM just supports basic I/O, data types, Vector, Stack and a couple of other simple things. No graphics, gui etc. For e.g.:

Test.java:
public class Test {
public static void main(String[] args) {
System.out.println("Hello World!");
}
}

4. Then compile the code to the JDK 1.4 format:

#javac -source 1.4 -target 1.4 Test.java

5. Finally, create a script file in gedit or vim called, say, Test.sh:

Test.sh:
kvm Test

6. Save the script file in the same folder as the Test.class file, and give both Test.sh and Test.class execute permissions:

#chmod +x Test.sh
#chmod +x Test.class

7. Put all these files in some folder in the iPod's Linux partition. I put it in "/home/shashank/java" (on the iPod partition, not your own system! Also avoid spaces in your folder path.). Then boot into iPodLinux and open the folder where you placed the files using the File Browser, and scroll to "Test.sh", and keep the "Select" button of the iPod pressed until the context menu shows up. Then select "Execute" and select "Read output". Voila! You should get the two words: "Hello World!" on your screen!

No cross-compilation, no recompiling of gcc to work with ARM7 processors (most iPods run on ARM7 processors, thus C code compiled on your PC will not work on the iPod). Life is easier with this if you want to do simple programming. Maybe if someone wrote an extended Virtual Machine, with more libraries, then it would make this a pretty solid platform.

BTW, the above steps have been written assuming you have a Linux/UNIX machine. If you have Windows, a VMWare or xVM VirtualBox image of some Linux distro on your Windows machine would make life tonnes easier compared to trying to use Windows' limited functionality with everything in life in general.

Cheers
Shashank

Possible loophole with dd command in multi-user Linux servers

Possible loophole with dd command in multi-user Linux servers

I didn't think this would be a big deal, but I noticed that several remote logging Linux servers permit users to have an unlimited disk quota, which can easily be exploited using a command like:

~$dd if=/dev/zero of=null.bin

It doesn't require the user to 'sudo'. I checked it out in one of the remote access servers that need to be logged in to using Telnet, and I executed this command and was surprised to see it being permitted and having a free and unchecked run. Effectively, /dev/zero is a virtual disk that can generate a series of 'zero' bytes, and the 'dd' command will copy from the input file (i.e. /dev/zero) to the output file (null.bin here) and will not terminate until it reaches the End of File of /dev/zero. And because I didn't give it any count or bytes limit (which can be done using the bs=xxx and count=xx parameter), and since it never reaches the end when reading from /dev/zero, it will effectively create a massively huge file within seconds, and fill up an entire normal-sized hard-disk in about a minute and a half! I tried it out and in less than 7 seconds, I had created a 962 MB null.bin in my home folder. And then I tried on a virtual machine (Ubuntu 8.04 Hardy Heron running in VMWare on Windows) I had created, with a fixed disk-space of 3 GB, and in about 10-15 seconds my 'dd' command died with a "No space left on disk" error message.

I'm not sure if its really a loophole or not, but it certainly made my home directory expand to 1.2 GB unchecked within seconds in a remote telnet login to a Redhat Server. If it is a loophole, I think it could be solved by assigning a strict disk-quota to each user, encompassing the total areas where he can write to.

~Shashank

How to boot from USB in xVM VirtualBox

I've looked around the internet a lot and mostly didn't find any working solution apart from trying to make an image/virtualDisk from a raw physical disk (USB). I was basically experimenting with writing simple bootloaders in assembly language, and wanted to rapidly test and make changes to the code without rebooting my system again and again to try booting. The obvious solution is to use a Virtualization tool like VMWare or xVM VirtualBox by Sun (the latter is free), but the trouble is that neither of them support booting from pen-drives and I have long forgotten the existence of floppy drives, and have thus no way of directly writing to a disk and virtual booting straight-away, without creating an new image of the disk each time I make changes to it. So I decided to write to my USB pen-drive directly and try booting from it. It works when I boot the entire computer with it, but VMWare and VirtualBox do not have the option of booting from a USB device.

So messing around xVM VirtualBox 2.0.2 36488 for Windows yesterday, I was pretty surprised to find my USB Pen-Drive listed under the field of Floppies! That is,

a) Create a new Virtual "Machine"
b) Click on the "Floppy" link in the main screen (right panel)
c) Enable "Mount Floppy Drive", and from the drop-down list in "Host Floppy Device", select your USB drive's Drive Letter (haven't tried it in Linux, I guess you'll have to select the correct device like /dev/sdxx)

That's it. Try starting the Virtual Machine now.

Its quite wierd, and I'm still wondering why my USB drive gets listed under floppy, but it works for me! Neither VMWare Workstation, nor xVM VirtualBox permit booting directly from the USB otherwise.

Cheers
Shashank

Of the Elusive Breakfast Syndrome

I really have a problem, or rather had one. It is an uninteresting topic for anyone but me, but I haven't managed to have breakfast on any weekend this ENTIRE semester i.e. in the past 4 months. I have never managed to wake up before the closing hours (9 am) of the hostel mess (often I directly wake up for evening snacks on weekends, let alone breakfast and lunch). I have even attempted, several times, to stay awake the entire night, till the golden moment when the clock struck 6:45 am, and each time some odd conspiracy of Nature conspired against me. Once I watched a movie till 6:35 am, and then pondered over how to pass the intermediate 10 minutes, and then continued pondering as I next regained conciousness at 10:30 am. Then once I managed to sleep off on the floor of my room, beside my door, while trying to open it to leave for the mess, again at 6:35 am.

Today, however, I grabbed Fate by the wrist and bent it to my will! I left my room at 6:30 am, as I was sure I wouldn't manage to survive any longer after watching Lord of the Rings III (for the 5th time), and also figured that I probably wouldn't sleep off in the corridor beside the mess, though I'm not sure I'd put that past my capabilities. Anyhow, here I was - unshaven, unkempt, with a haggardly-desperate look, sitting outside the stairs of the mess waiting for it to open. Then I tried walking around the area a bit, and I discovered I didn't like the cold air too much, and again returned to my patient look-out post. And then, the Gates of Destiny opened! And I had wonderful breakfast, and Tea! One of the greatest battles of will and destiny in my life had been fought, and I had emerged the victor - a fed victor at that. :-)

Shashank
PS: I haven't gone loopy. Food breaks the best of men, I'm just a mere, humble mortal.

DMOZ Local Directories and BatchDMake

Today morning I finally settled in to perform a long-pending weekend house-cleaning job, except that the cleaning work was electronic, i.e. my computer. It is a common scenario to find your files so badly scattered around the hard-disk, inspite of some organization scheme that we all implement. Years ago, I saw a nice scheme Roshan, a friend of mine, implemented. He created directories and sub-directories based on topics and categories, like "Science", "Computers", etc, and Science would contain "Physics", "Chemistry" ... while Computers would contain "Programming", "Internet" and so on. It was a complete hierarchial structure, and I've been using it on my systems too ever since I saw it there.

But, its hard to maintain my directory, as creating such a structure is a rather tough task. You could either create an entire directory structure before-hand, with empty folders waiting to be filled, or you could create new folders as and when you require, in the correct category of course. The latter option obviously seems more efficient, as you don't have an empty mammoth directory structure, but a dynamic and to-your-taste structure. But after several years of trying out both, I have found that the first approach is much better. When you are, say, saving a file from the Internet, you often don't have the patience to ponder over the hierarchial relationship between "Geese Hunting" and "Duck Hunting", and whether you should create a new directory for each, or just club them into one single directory called "Bird Hunting". This inherent impatience of people when using files, as understandable and natural it is, is THE reason why files are scattered and chaotic. Thus I have concluded that people need an existing order to follow, that is as extensive and accomodating as it can be, but also flexible in order to allow people to delete and modify and create new topics and directories.

With this long discourse on the philosophy of file management in mind, I sat about thinking! :-D

This is what I came up with: Why not use an existing MASSIVE topic-wise directory structure, which has been developed and nurtured for about a decade by tens of thousands of people across the world? That directory structure is the DMOZ Open Directory Project (http://www.dmoz.org) by the Mozilla Foundation (of Firefox fame). It forms the backbone of Google Directory and several other places on the internet, and is a collaborative and open effort at building efficient web directories. My interest, however, is not in the links to websites that is provides, but the topic-based organizational structure. And as luck has it, they offer a downloadable plain-text file containing the ENTIRE folder structure. Perfect. Its available at:
http://rdf.dmoz.org/rdf/categories.txt (57.4 MB)

Then I wrote a simple tool in Java called "BatchDMake", which will read a given text-file line-by-line and create the folders listed in it. Each folder path is listed in a separate line. For example, this is from the DMOZ dump:

Arts
Arts/Movies
Arts/Music
Computers
Computers/Artificial Intelligence
Computers/Games

The syntax for operating it is:

java BatchDMake [filename] [root folder]

The filename parameter is obvious, the root folder is the folder name relative to the current working directory in which it should create the folders listed in the text file. For e.g. using the downloaded DMOZ categories file:

java BatchDMake categories.txt dmozRoot

The source code for BatchDMake is available here, and the executable .class file is here.

I just created the ENTIRE folder structure of DMOZ on my hard-disk, it has about 770,000 directories & sub directories. A lot of them are useless ones, so I'll go about pruning the tree to come up with the most useful structure for normal users. Will post that one soon. In the meanwhile, you could try the DMOZ structure and use BatchDMake to create it on your system!

Cheers!
Shashank
PS: I even found the place I'm currently in!
root\Regional\Asia\India\Tamil_Nadu\Districts\Vellore
This is like exploring an uncharted world in your hard-disk! :-P