December 2008 (#157):

• Mailbag
• Talkback
• 2-Cent Tips
• News Bytes, by Deividson Luiz Okopnik and Howard Dyckoff
• Init Call Mechanism in the Linux Kernel, by Andrew Murray
• Keymap Blues in Ubuntu's Text Console, by Anonymous
• Book Review: CUPS Administrative Guide, by Deividson Luiz Okopnik
• Away Mission: A Vision for Embedded Linux, by Howard Dyckoff
• Jetty: The Twelve Year Journey to Market Maturity, by Adam Lieber
• Joey's Notes: Basic Samba Configuration, by Joey Prestia
• HelpDex, by Shane Collinge
• XKCD, by Randall Munroe
• The Linux Launderette

Mailbag

This month's answers created by:

[ Ben Okopnik, Kapil Hari Paranjape, Neil Youngman, Rick Moen, Suramya Tomar, Thomas Adam ]
...and you, our readers!

---

Gazette Matters

---

list bug???

Rick Moen [rick at linuxmafia.com]

Tue, 25 Nov 2008 08:41:08 -0800

Quoting Michael Makuch (linuxgazette@makuch.org):

[You wrote to tag-owner@lists.linuxgazette.net. I've decided to CC this response to TAG.]

> Curious. I am subscribed to TAG with the address
> linuxgazette@makuch.org.
> 
> However, I accidentally sent a message to TAG just now from
> mike8@makuch.org, yet the message went through to the list.
> 
> There seems to be a problem with the list if email from any ole
> unregistered email address can send to it...

{sigh} No, it's a deliberate list configuration. You're absolutely right that that particular list configuration is archaic and essentially never seen (otherwise) on a usable mailing list in this decade, because of the continually rising spam problem. TAG has always been set up to permit non-subscriber posting because of its unique intended role: It's supposed to permit any reader of _Linux Gazette_ to address the mailing list to converse with The Answer Gang. Doing so of course makes it very vulnerable to spam.

You will probably have noticed that some spam does get through my MTA and mailing list software defences. I (as listadmin and MTA owner) then manually analyse ASAP any spam that does so, and take immediate action to ensure that no further mail from that (pick one; whatever seems most effective and least disruptive of legitimate mail) SMTP HELO string, or sending address, or sending domain, or IP will be accepted in the future. That is of course a no-win strategy with unavoidable collateral damage, but it's the best I can contrive under the circumstances.

TAG is the only mailing list with that configuration I know that's not utterly overwhelmed with spam and thereby made useless, so I must be doing something right. But the possibility of it, too, getting overwhelmed is always present.

[ Thread continues here (2 messages/2.91kB) ]

---

Our Mailbag

---

Tessboxes

Jimmy O'Regan [joregan at gmail.com]

Thu, 13 Nov 2008 15:37:14 +0000

Tessboxes is a viewer/editor for tesseract boxfiles (used when creating training data for tesseract). http://www.lbreyer.com/tessboxes.html

On the topic of tesseract; there's also http://code.google.com/p/tesseractindic/ which is a version of tesseract adapted to Indian languages.

cuneiform-linux 0.5 was released recently, too; I'll see if I can find the announcement. It now supports hocr.

---

Mandrake v9.1 on a Dell CPi D300XT

EJB [ejb at intergate.com]

Wed, 19 Nov 2008 13:02:14 -0500

Folks;

Several years ago, on a whim while at an OfficMax store I came across a sale going on regarding Mandrake v7.0 Linux and dis-satisfaction with MS (and the lower sales price) gave me reason enough to purchase the Linux package ... from the view of an owner/operator of several units - best thing I think I ever did. Couldn't believe that Linux software went on as smoothly as it did - and it worked - well mostly it worked but it did it well enough for me to keep an eye on the Linux OS which were available.

Recently I pulled an old Dell Latitude CPi D330XT out of "storage", deciding to keep it as a back up and fully operational unit and once again I've come across an old problem, one I never did solve in the past.

What I have is the following:

. A Dell D300XT laptop with a 10GB HD and 256kRam

. Mandrake v9.1 Linux and MS 98SE for Operating Systems

With disk partitioning, I've set or better, limited the HD space to just 2GB, with Mandrake v9.1 getting the rest of the available HD space. Yes, it does work/fit and very nicely at that but ...

PROBLEM I: SOUND - how do you get Linux to recognize the built-in sound card/chip which is available with these old Dell machines? At present,there is no sound at all and there is little doubt here but that and for want of a better term, "it's there" and it's just a case of finding out how to activate it.

PROBLEM II: PPP - It's timing out. I can get the dialer to dial in to the ISP but there's no data flow going on and the PPP times out. How do I get this software to "click" and connect?

PROBLEM III: SCANNER - I have an old (but very reliable V300 Microtech scanner but I can't get the Linux software system to recognize that the scanner is there. Win98SE will recognize it and operate it and there's no doubt here but that Linux will do it to - if I find and "set all the switches" in the proper manner.

I'm open to any and all ideas and suggestions.

Thanks for your time an interset.

ejb

[ Thread continues here (3 messages/13.27kB) ]

---

once-thought-safe-wpa-wi-fi-encryption-cracked

Chris Bannister [mockingbird at earthlight.co.nz]

Sun, 9 Nov 2008 02:32:23 +1300

Hi,

FYI

http://www.itworld.com/security/57285/once-thought-safe-wpa-wi-fi-encryption-cracked

-- 
Chris.
======
I contend that we are both atheists. I just believe in one fewer god
than you do. When you understand why you dismiss all the other
possible gods, you will understand why I dismiss yours.
                                           -- Stephen F Roberts

[ Thread continues here (2 messages/1.66kB) ]

---

How to get a random file name from a directory

Suramya Tomar [security at suramya.com]

Tue, 25 Nov 2008 00:40:30 +0530

Hi Everyone,

I am working on a collage creation script in bash and am using find to get a list of all jpg files in a given directory. However find always returns the image names in the same order so the collage tends to have images from the same photo set i.e. at a given time it will have the most of the images from a particular subfolder.

What I want to do is get a list of random image file names from the system and then use it to create the collage, so that the collage looks more varied.

Here's what I have done so far to get a random file name:

   awk "NR==$(($RANDOM % $(wc -l fi.txt| awk '{print $1}')))" fi.txt

where fi.txt was created using

   find . -iname "*.jpg" -true | sed 's/ /\\ /g' >fi.txt

Now, this way works but I have to create a temp file which I want to avoid. Is there some way of getting find to return results in a random order? I tried searching the web but didn't find any useful results.

I am attaching the current version of the Collage creation script with this email to give you an idea of how it works. This version is without randomization of the filenames.

Thanks in advance for the help.

- Suramya

PS: Please do give your feedback/suggestions on how the script can be improved in addition to a solution

[ ... ]

[ Thread continues here (28 messages/36.11kB) ]

---

adding key combination in inittab file

J.Bakshi [j.bakshi at icmail.net]

Sat, 1 Nov 2008 11:10:12 +0530

Hi list,

Hope you all are well.

I am trying to modify the /etc/inittab file to add more key combination. One of the reasons is to shut down the system. I have added this line in the inittab

~~~~~~~~~~~~~~~~~~~~~~~~~~~`
#shut down wit ctrl+alt+end
ce:12345:ctrlaltend:/sbin/halt
~~~~~~~~~~~~~~~~~~~~~~

but "init q" reports

	:ctrlaltend:   unknown action field

Could any one kindly point out my mistake ? I like to define some more key combination like ctrl+alt+< ; ctrl+alt+s etc to call commands. How can I define these key combination ?

Thanks

[ Thread continues here (5 messages/4.31kB) ]

---

seeking suggestion about lftp

J.Bakshi [j.bakshi at icmail.net]

Sat, 22 Nov 2008 22:42:10 +0530

Hello,

Hope you all are well.

I have designed a script which compress some files and store those files in a folder named by date like 10-11-2008 ( for 10th Nov 2008 ) and uploads it in a remote ftp server through lftp. the script runs every day. As the FTP space is little after every 15 or so days the FTP space qouata is full. I like to include a section in the script which will check the folders in the ftp server through lftp and deletes the folders which already 10 days matured.

Can any one suggest me how to do this throug lftp ?

thanks a lot

[ Thread continues here (10 messages/13.97kB) ]

---

LWN article: Wikipedia transition from GFDL to CC BY-SA

Rick Moen [rick at linuxmafia.com]

Fri, 21 Nov 2008 12:18:55 -0800

Of possible interest to LG in that it's another notable episode in the perennial "What's a good free licence for non-software content?" debate.

----- Forwarded message from Rick Moen <rick@linuxmafia.com> -----

Date: Wed, 5 Nov 2008 22:28:28 -0800
From: Rick Moen <rick@linuxmafia.com>
To: discuss@en.tldp.org
Subject: LWN article: Wikipedia transition from GFDL to CC BY-SA

Those of you who are Linux Weekly News subscribers, please see: http://lwn.net/Articles/305892/

GFDL 1.3: Wikipedia's exit permit by Jonathan Corbet November 5, 2008

[...] Members of the Wikipedia project have wanted to move away from the GFDL for some time. [...] The presence of the "or any later version" language allows Wikipedia content to be distributed under the terms of later versions of the GFDL with no need to seek permission from individual contributors. Surprisingly, the Wikimedia Foundation has managed to get the Free Software Foundation to cooperate in the use of the "or any later version" permission to carry out an interesting legal hack. [...]

In other words, GFDL-licensed sites like Wikipedia have a special, nine-month window in which they can relicense their content to the Creative Commons attribution-sharealike license. This works because (1) moving to version 1.3 of the license is allowed under the "or any later version" terms, and (2) relicensing to CC-BY-SA is allowed by GFDL 1.3.

LWN articles become available to the general public after eight days, so non-subscribers will be able to read the story starting Nov. 13.

----- End forwarded message -----

[ Thread continues here (4 messages/9.76kB) ]

---

Apertium en-gl (English-Galician) and eu-es (Basque-Spanish) released

Jimmy O'Regan [joregan at gmail.com]

Thu, 20 Nov 2008 16:51:42 +0000

"The first version (0.5.1) of apertium-en-gl, the Apertium-based English-Galician translator, has just been released. It has been developed by the Galician company imaxin|software, one of the companies in the OpenTrad consortium, with collaboration from the universities of Vigo and Santiago de Compostela, using preexisting data developed by groups from many different institutions. Apertium-en-gl has been developed with the support of the Consellar�a de Innovaci�n e Industria of the Galician Government (Xunta) and the Ministry of Industry, Tourism and Commerce of the Spanish Government"

http://sourceforge.net/forum/forum.php?forum_id=889264

The Apertium project has released linguistic data for translating from Basque to Spanish on this free/open-source machine translation platform. The translator is tuned to help Spanish speakers read Basque newspaper articles but does not aim at generating post-editable output. There are a number of identified problems which may be easily addressed to make it more usable. Developers welcome! Try it online at http://www.erdaratu.eu

https://sourceforge.net/forum/forum.php?forum_id=886821

---

Patent lreversal on "business practice" patents

Ben Okopnik [ben at linuxgazette.net]

Sun, 9 Nov 2008 09:23:47 -0500

[Forwarded from the RISKS digest]

John Oram, Microsoft has a problem, *IT Examiner*, 31 Oct 2008

Much of the patent portfolio of some of the world's biggest software companies has become worthless overnight, thanks to a ruling yesterday by the US patent court.

The US Court of Appeals for the Federal Circuit (CAFC) in Washington DC has decided that in the future, instead of automatically granting a patent for a business practice, there will be a specific testing procedure to determine how patentable is that process.

The decision is a nearly complete reversal of the court's controversial State Street Bank judgment of 1998, which started the stampede for patenting business practices.

http://www.itexaminer.com/us-court-throws-out-most-software-patents.aspx http://www.groklaw.net/pdf/07-1130.pdf

-- 
* Ben Okopnik * Editor-in-Chief, Linux Gazette * http://LinuxGazette.NET *

---

[OT] OpenLibrary 'scan-on-demand'

Jimmy O'Regan [joregan at gmail.com]

Tue, 4 Nov 2008 01:25:25 +0000

[Totally off-topic, but we all like to read, right?]

The OpenLibrary is a project from the Internet Archive; in short, they're building a Google Books-type catalogue of titles, integrated with the ability to search the text of those titles that have been made available through the Internet Archive. The coolest thing they have, though, is a 'scan on demand' feature for any book available at Boston Public Library -- if the book hasn't already been scanned, a librarian will put it in the scanner for you: http://openlibrary.org/bpl

---

Talkback: Discuss this article with The Answer Gang

Talkback

Talkback:124/smith.html

[ In reference to "Build a Six-headed, Six-user Linux System" in LG#124 ]

Ben Okopnik [ben at linuxgazette.net]

Fri, 28 Nov 2008 07:44:50 -0500

----- Forwarded message from "D. Osborn" <dgosborn@hotmail.com> -----

Security can be addressed by using something like puppy linux. Once you get your config worked out you can burn the CD and boot from it. Any compromises can be fixed with a nightly reboot. giving the users qemu or vmware sessions can isolate them and I think you can lock them into those sessions for troubleshooting the other problems try first creating just one change or another, extending the number of possible devices (a-f), or adding extra keyboard, or mouse, etc. one at a time to see when the problem shows up. Any news on this? I'm thinking of using it in my call center if I can work out the bugs. regards dale

----- End forwarded message -----

-- 
* Ben Okopnik * Editor-in-Chief, Linux Gazette * http://LinuxGazette.NET *

[ Thread continues here (2 messages/2.18kB) ]

---

Talkback:156/dellomodarme.html

[ In reference to "A (not so) short overview of the Geographic Information System GRASS" in LG#156 ]

Thomas Adam [thomas.adam22 at gmail.com]

Tue, 4 Nov 2008 22:12:17 +0000

I was pleased to see in November's LG an article on GRASS. I've been playing with this for years, and recently bought the third edition of the book:

"OPEN SOURCE GIS: A GRASS GIS Approach"

Which is briliant, in my opinion. I don't use GRASS for anything fancy, but it helps me plot local geological data I collect in my spare time.

Nice article. Good to see it in LG.

-- Thomas Adam

"It was the cruelest game I've ever played and it's played inside my head." -- "Hush The Warmth", Gorky's Zygotic Mynci.

---

Talkback:156/lg_talkback.html

[ In reference to "Talkback" in LG#156 ]

Jimmy O'Regan [joregan at gmail.com]

Sun, 2 Nov 2008 11:36:44 +0000

Talkback: [LG 86] 2c Tips #20

> This is clearly the wrong Talkback link, but after some serious
> searching of the LG archives, I can't find the source anywhere.

It's here: http://linuxgazette.net/149/misc/lg/alien_tongues.html with this note "This originally had a subject line that allegedly referred to an old LG (issue 86!) tips."

Sorry for the confusion!

---

Talkback: Discuss this article with The Answer Gang

2-Cent Tips

2-cent tip - KDE 4.1 on 10 year old PC

oscar laycock [oscar.laycock08 at googlemail.com]

Mon, 17 Nov 2008 14:53:22 +0000

I thought you might be interested to know that the latest KDE 4.1 runs surprisingly well on my 10 year old PC. It is only a 350mhz Pentium 2 with 128mb of swap. I built KDE myself - you just run "cmake" instead of configure. I have also done some Qt programming, though sometime I have to increase the swap by 200 or 300 meg.

I simplified Linux from Scratch (LFS). I run a 1998 version of Red Hat Linux(!) alongside my new stuff in /usr/local. I hardcode in a rpath to the new libraries when compiling new programs.

I would encourage everyone to try LFS (www.linuxfromscratch.org). I only knew a little Unix and some C when I started. It is fun, you learn a lot, and get a great sense of achievement.

---

2-Cent Tip: LG search plugin

Ben Okopnik [ben at linuxgazette.net]

Wed, 19 Nov 2008 20:22:34 -0500

If you use Firefox, you're probably addicted to its oh-so-nifty little search bar (how did we ever live without those?) What you may not know, though, is that you can add your own entries to that pull-down list - all you need is a properly-formatted search URL, a bit of XML and maybe a Base64-encoded icon. So, for those of you who want to search LG in the easiest way possible, here it is; save it as '~/.mozilla/firefox/<profile>/searchplugins/lg.xml', restart Firefox, and enjoy!

[ ... ]

[ Thread continues here (1 message/3.60kB) ]

---

2-cent Tip: command line calendar with cal

Mulyadi Santosa [mulyadi.santosa at gmail.com]

Tue, 11 Nov 2008 00:51:50 +0700

A quick way to show current month's calendar is by executing "cal" command. By default, it will show current month calendar.

Should you need to show calendar of certain month and year, simply type it as parameter. For example:

$ cal 1 1979

will show you calendar of January 1979

regards,

Mulyadi.

[ Thread continues here (9 messages/9.48kB) ]

---

2-cent Tip: To solve memory allocation problem in Cacti

Mulyadi Santosa [mulyadi.santosa at gmail.com]

Sun, 23 Nov 2008 10:20:48 +0700

This kind of problem doesn't manifest in explicit message. What you see is just graphs that disappear. But it will become clear when you run the poller directly like this:

$ sudo -u cacti /usr/bin/php /var/www/cacti/poller.php

Allowed memory size of 8388608 bytes exhausted (tried to allocate 12 bytes)

At first, one might think it's something related to ulimit and rush to change stack size limit for example. But that's not the problem. The problem lies in php CLI (Command Line Interface) which tries to allocate memory in stack in order to prepare data to be shown in graph tree and so on.

In Ubuntu, try to edit this line in /etc/php5/cli/php.ini:

memory_limit = 8M

8 Megabyte might be too low especially you poll hundreds of devices, so try to rise the number. From some reference, it is declared that 128 M is the maximum size of memory PHP CLI can allocate in stack.

regards,

Mulyadi.

---

2-cent Tip: File versioning/backup

Ben Okopnik [ben at linuxgazette.net]

Sat, 1 Nov 2008 23:32:43 -0400

[For some reason, I thought I'd submitted this ages ago; searching LG shows that I was apparently wrong.]

There are lots of times - e.g., when you're editing a critical file, or want to keep multiple versions of a script that you're developing - that you want to take a snapshot of your current directory, or at least of some of the files within it. I've been using the following script for the past several years, and it's proven to be flexible, powerful, and fully sufficient for the task.

Note that you can also define a ".backup" file within the directory (this should contain the names of the files to be backed up, one per line) so you can run "backup" without any arguments and still have it snapshoot the right files. Enjoy!

#!/bin/sh
# Created by Ben Okopnik on Wed Jan 26 21:48:05 EST 2005
 
name="`/usr/bin/printf $0|/bin/sed 's|.*/||'`"
 
usage () {
cat <<!
Usage:
	$name -h
	$name [-a] [file[s]]
	
	-h	Show this help
	-a	Back up all specified files unconditionally
	
	The list of files to be backed up can be defined either in a
	.backup file in the current directory, on the command line, or as a
	combination of the two. The files will be saved in a "backup"
	subdirectory, with a timestamp extension appended to their
	filenames.
 
	If the latest archived copy of a specified file does not differ
	from the current version, the file will not be backed up unless the
	"-a" option is used.
 
!
}
 
case "$1" in
	-h) usage; exit ;;
	-a) do_all=1; shift ;;
	-n) no_file=1; shift ;;
esac
 
 
backup () {
	# Create a backup directory if necessary
	[ -d "backup" ] || {
		[ -e "backup" ] && { 
			echo "Error: $PWD contains a file called 'backup': can't create a 'backup' directory!"
			exit
		}
		echo -n "Creating 'backup' directory... "
		mkdir backup
		[ "$?" -eq 0 ] && echo "Success!" || { echo "Failed - exiting."; exit; }
	}
 
	for file in "$@"
	do
		[ -f "$file" ] || { printf "$file not found!\n"; continue; }
		# Find the last backup for each file
		latest="`ls backup/$file.20[0-9]* 2>/dev/null|tail -1`"
		[ -z "$latest" ] && do_all=1
		if [ "$do_all" = "1" ] || [ -n "`diff -q $file $latest`" ]
		then
			cp -i "$file" "backup/$file.`date +'%Y%m%d%H%M%S'`"
			saved="$saved $file"
		else
			nodiff="$nodiff $file"
		fi
	done
 
	[ -n "$saved" ] && echo "Backed up: $saved"
	[ -n "$nodiff" ] && echo "NOT backed up (identical files in archive): $nodiff"
}
 
[ "$no_file" = 1 ] || { [ -f ".backup" ] && set "$@ `cat .backup`"; }
 
if [ -n "`echo $*|grep '[^ ]'`" ]
then
	backup $@
else
	echo "No files specified and no ./.backup file found; exiting."
	exit 1
fi

-- 
* Ben Okopnik * Editor-in-Chief, Linux Gazette * http://LinuxGazette.NET *

[ Thread continues here (5 messages/8.30kB) ]

---

Talkback: Discuss this article with The Answer Gang

News Bytes

By Deividson Luiz Okopnik and Howard Dyckoff

Contents: News in General Conferences and Events Distro News Software and Product News

Selected and Edited by Deividson Okopnik

Please submit your News Bytes items in plain text; other formats may be rejected without reading. [You have been warned!] A one- or two-paragraph summary plus a URL has a much higher chance of being published than an entire press release. Submit items to bytes@linuxgazette.net.

---

Talkback: Discuss this article with The Answer Gang

---

Deividson was born in União da Vitória, PR, Brazil, on 14/04/1984. He became interested in computing when he was still a kid, and started to code when he was 12 years old. He is a graduate in Information Systems and is finishing his specialization in Networks and Web Development. He codes in several languages, including C/C++/C#, PHP, Visual Basic, Object Pascal and others.

Deividson works in Porto União's Town Hall as a Computer Technician, and specializes in Web and Desktop system development, and Database/Network Maintenance.

---

Howard Dyckoff is a long term IT professional with primary experience at Fortune 100 and 200 firms. Before his IT career, he worked for Aviation Week and Space Technology magazine and before that used to edit SkyCom, a newsletter for astronomers and rocketeers. He hails from the Republic of Brooklyn [and Polytechnic Institute] and now, after several trips to Himalayan mountain tops, resides in the SF Bay Area with a large book collection and several pet rocks.

Howard maintains the Technology-Events blog at blogspot.com from which he contributes the Events listing for Linux Gazette. Visit the blog to preview some of the next month's NewsBytes Events.

Init Call Mechanism in the Linux Kernel

By Andrew Murray

Introduction

The Linux kernel has for a long time (at least since v2.1.23) contained a clever and well-optimised mechanism for calling initialisation code in drivers. It's clever because its functionality is largely abstracted from the driver developer, and it's well-optimised because after initialisation, memory containing the initialisation code is released. This article explores how this mechanism works.

Typical Usage

We'll start by seeing how driver developers make use of this functionality; the following code comes from linux-2.6.27.6/drivers/net/smc911x.c and is the driver for a common Ethernet chipset (smc911x).

2206: static int __init smc911xinit(void)
2207: {
2208:     return platform_driver_register(&smc911x_driver);
2209: }
...
2216: module_init(smc911x_init);

The smc911xinit function can be considered as the entry point into the driver; of particular interest is the __init macro and the static declaration. The __init macro is used to describe the function as only being required during initialisation time: once initialisation is performed, the kernel will remove this function and release its memory. The module_init macro is used to tell the kernel where the initialisation entry point to the module lives, i.e. what function to call at 'start of day'. In a typical driver, you will often see many functions marked with the __init macro - these are used for initialisation - and a single module_init declaration.

Even though we are expecting the kernel to call smc911x_init at 'start of day', we have marked it as static - but that's OK (later, we will see how the function is called). This is a particular strength of the init call mechanism: it reduces the amount of public symbols and reduces the coupling between driver modules and other parts of the kernel.

The optimisation provided by the init call mechanism also provides a means for recovering memory used by the initialisation data. Such data can be 'tagged' with the __initdata macro.

With the above code in place, at an appropriate time during start-up, the kernel will call the smc911xinit function, and once it has been executed its memory will be released. You can see this in the output from kernel (e.g. dmesg); for example, an x86 machine may print the following:

Freeing unused kernel memory: 386k freed

This line is telling us that 386k of memory that previously contained initialisation code and data has now been freed.

Under the Hood

OK - So we've seen how the mechanism is used. Now, let's take a closer look and see how it works under the hood. A quick 'grep' reveals that the __init macro is defined in include/linux/init.h:

43: #define __init      __section(.init.text) __cold

The __section and __cold macros are defined in the various include/linux/compiler*.h files:

compiler.h: 182: #define __section(S)  __attribue__ ((__section__(#S)))
compiler-gcc4.h: #define __cold        __attribue__ ((cold))

And when we expand it out we get:

#define __init __attribute__((__section__(".init.text"))) __attribute__ ((cold))

Thus, when the __init macro is used, a number of GCC attributes are added to the function declaration - in the case of a different compiler, the compiler.h file will ensure the macros expand out to whatever is necessary for the relevant compiler. The cold attribute is a relatively new GCC attribute and has existed since GCC4.3 - its purpose is to mark the function as one that is rarely used, which results in the compiler optimising the function for size instead of speed. What we are really interested here is the 'section' attribute. The __init macro uses this attribute to inform the compiler to put the text for this function in a special section named ".init.text". The purpose here is to put all initialisation functions in a single ELF section such that the entire section can be removed after initialisation has been performed.

So what does module_init do? Its exact functionality depends on whether the module in question is built-in or compiled as a loadable module. For the purposes of this article, we'll be looking at the built-in modules. Back to include/linux/init.h:

259: #define module_init(x) __initcall(x);
204: #define __initcall(fn) device_initcall(fn)
199: #define device_initcall __define_initcall("6", fn, 6)
169: #define __define_initcall(level, fn, id) \
170:            static initcall_t __initcall_##fn##id __used \
171:            __attribute__ ((__section__(".initcall" level ".init"))) = fn

Yet another load of macros that result in even more GCC attributes being defined!

#define module_init(x) static initcall_t __initcall_x6 __used \
                       __attribute__ ((__section(".initcall6.init"))) = x;

And for clarity, let's expand our module_init macro as used in our ethernet driver:

static initcall_t __initcall_smc911x_init6 __used \
                  __attribute__ ((__section(".initcall6.init"))) = smc911x_init;

As you can see, module_init in the context of a built-in driver results in declaring a function pointer with a unique name to our point of entry. In addition, the macro ensures the function pointer is located in a special section of the ELF - we'll see why shortly.

So at present, we have ensured that all our initialisation code and data is stored in the .init.text section, and that each module has a function pointer for its point of entry - which has a unique name and is also stored in a special section of the resulting ELF. In addition, during link time the include/asm-generic/vmlinux.lds.h and arch/*/kernel/vmlinux.lds.S scripts ensure that some labels/symbols surround the start and end of these sections. I.e. __early_initcall_end and __initcall_end mark the start and end of the function pointers and __init_begin and __init_end mark the start and end of the .init.text section.

Finally we are in a position to see how these functions get called and how they are eventually freed. The function do_initcalls in init/main.c is called during kernel startup. This is shown below:

749: static void __init do_initcalls(void)
750: {
751:      initcall_t *call;
752:
753:      for (call = __early_initcall_end; call < __initcall_end; call++)
754:           do_one_initcall(*call);
755:

The purpose of this loop is to execute each of the init functions as set up by the module_init macros. This is achieved with a simple 'for' loop and a function pointer. Initially, the function pointer is pointed to the label at the start of our function pointer's ELF section, and is incremented (by the size of a function pointer (sizeof(initcall_t *)) until the end of the ELF section is reached. For each step, the pointer is invoked and the init function is thus executed.

Once initialisation is complete, a function found in the architecture- specific code named free_initmem is used to release the memory pages taken up by the initialisation functions and data. The exact nature of the function varies between architectures.

Summary

In a nutshell, the kernel makes clever use of macros and GCC attributes to ensure that initialisation functions and pointers to them are stored in unique sections of the ELF. Initialisation code at kernel startup then iterates through these function pointers and executes them in turn. Finally, once all init code has been executed, the entire ELF section (.init.text) is freed for re-use! The best part of this mechanism is that the provided macros completely hide its underlying complexity, thus leaving more time for driver developers to focus on the job at hand.

Further Reading

The best way to fully understand parts of the Linux kernel is to browse the source code - and that's exactly how I wrote this article. I did, however, make extensive use of the Linux Cross Reference - this site and many like it allow you to explore the source code and easily find out where functions are called and defined.

For more information on GCC attributes, read the GCC online documentation - in particular see section 5.2.7 Declaring Attributes of Functions.

Talkback: Discuss this article with The Answer Gang

---

Andrew Murray is an embedded systems engineer at MPC Data Limited - one of the UK's leading systems integrator. His day to day role fulfils his passion for learning and provides him with a wide range of experiences including embedded Linux such as driver and kernel development, embedded applications development and even Windows driver development. Working on a wide range of projects has allowed Andrew to explore a wide range of technologies such as the inter-workings of PCI Express and High Definition (HD) audio.

Prior to his employment, Andrew graduated in 2007 from the University of Wales, Aberystwyth with a Masters degree in Software Engineering (MEng). His final year dissertation involved the creation of a 'black-box' for a sail plane glider that would assist in the automated marking of aerobatic gliding competitions. Making use of MEMS sensors, barometers, magnetometers and GPS along with Kalman filtering - the device was able to successfully record not only position but orientation in an aerobatic environment.

Being a member of the Institute of Engineering and Technology, curator of http://www.embedded-bits.co.uk and (currently) a 'one-time' kernel contributor, Andrew continually tries to contribute to the community more and more whenever possible.

Keymap Blues in Ubuntu's Text Console

By Anonymous

Today's news is that, after years of cosying up to SUSE/openSUSE, I have decided to move unambiguously to Ubuntu. KDE did it to me. KDE on plasmoids gives me the jinx, and GNOME on openSUSE is an innocent abroad. On to Ubuntu 8.10.

Moving to Ubuntu is work - at least if you like the text console and spend time in it. You have to install lots of packages before Ubuntu's text-mode environment becomes kind of comfortable for administrative tasks and leisure programming. The 8.10 repositories don't even offer Midnight Commander, right now. Also, when you think you are done, there are still surprises waiting around the corner. Take, for instance, the keymap. (Warning: We are talking about the text-mode keymap, no X11 involved. I do not intend to repeat this every second sentence.)

1. Where are the keymaps?

Where are Ubuntu's text-mode keymaps for languages and special keyboard layouts like Dvorak? Out there in the wilderness of distros, you will find a collection of keymaps below /usr/share/keymaps/ or /usr/share/kbd/. A past version of Fedora I was playing with, two years ago, had them in /etc/keymaps. Where are they in Ubuntu?

Good question. Go ahead and look for them. Poor reader, I see you coming back with tail nicely folded between hind legs. How come? They are not there! They are nowhere! They are behind Locked Door Number X!

A few reminiscences. In the late 90s, the maintainer of the kbd package, which was the basis for keymaps and keymap handling in symbiosis with the Linux kernel, appeared to be feebly motivated. A competing project, called console-tools, sprang up and was promptly embraced by Debian, while Red Hat and SUSE stuck to kbd. However, kbd gathered momentum and released several new versions in the early 2000s, while console-tools stagnated and became moribund. It is fair to say that the project is abandoned, while kbd has just been blessed with the energy of a new maintainer. Debian and Ubuntu are now moving back to kbd, and facing annoyances galore with compatibility issues.

For Ubuntu, the approach is simple: they do not care about the text console, so they ignore, as far as possible, the trouble surrounding keymaps. There is one and only one keymap in Ubuntu 8.10: it is called boottime.kmap.gz, and is found in /etc/console-setup.

Needless to say, those keymaps must exist, since they are console configuration options -- but they are in the archives in binary form, not accessible for tinkering and individual customisation. Not accessible for fun.

2. Why would you want to change the keymap?

Surely: Because you are entitled to your idiosyncrasies (I mean preferences), and that's a strong enough reason. But there are also a couple of things that, soberly considered, call for intervention.

2.1 Some definitions

Sorry, terminology pain ahead: 'Keymap' normally indicates a file where the key assignments get defined. The assignments can refer to plain keys, but can also refer to modified keys, e.g., <ctrl> or <left>. In the keymap file, a set of assignments for given modifiers is also called a keymap, so we get a keymap for <ctrl>, a keymap for <alt> and so on. Now, hold yourself tight - because Ubuntu defines 64 keymaps. All possible modifier combinations are there. Do not ask me why odd combinations with 3 or 4 or 5 modifiers are there, since they cannot possibly be typed as long as humans have only two hands. But they are there, and that's the end of it. Compare with openSUSE or Fedora, where you will find only single and double modifiers from the <shift>-<ctrl>-<alt>-<altgr> set - and not even all of them. That makes sense.

Perhaps this is a marginal issue, but let me point out that Ubuntu's keymap is 20 times larger than openSUSE's and Fedora's, and therefore must have some impact in memory. I never noticed any difference while using openSUSE and Ubuntu, but the effect must be there.

2.2 Read this thread...

https://bugs.launchpad.net/Ubuntu/+source/console-data/+bug/279973

...and you will realize that pressing <printscreen> in the Ubuntu console produces a Control_backslash, AKA char 28. The bug report above refers to 8.04, but the "feature" is still there in 8.10: character 28 has vandalizing tendencies in some applications, and so work may be lost -- and it was, indeed, as the thread points out. Control_backslash can still be produced with <ctrl><printscreen>, no objection to that, since it would be intentional. But pressing just plain <printscreen> inadvertently is just too likely to happen.

2.3 The US default keymap and the default keymap for a couple of languages I have checked (German, French, Spanish, Italian, Russian) define strings for variables F1 - F20. They also have variables past F20 (F21 and up), but without any strings defined for them. This last oddity is a problem everywhere, but here we are concerned with differences between Ubuntu and mainstream, i.e., between the legacy approach of console-tools versus kbd. Distros based on kbd are the overwhelming majority. Contrast them with Ubuntu, with regard to definitions for <shift><f1> and up:

                    kbd             console-tools
                    (openSUSE)      (Ubuntu)

    <shift><f1>     F13             F11
    <shift><f2>     F14             F12
    <shift><f3>     F15             F13
    <shift><f4>     F16             F14
    <shift><f5>     F17             F15
    <shift><f6>     F18             F16
    <shift><f7>     F19             F17
    <shift><f8>     F20             F18
    <shift><f9>     F21 (void)      F19
    <shift><f10>    F22 (void)      F20
    <shift><f11>    F23 (void)      F21 (void)
    <shift><f12>    F24 (void)      F22 (void)

My first remark, here, would be that the shifted <f9>, <f10>, <f11>, <f12> are not usable in openSUSE and others, because F21 and up is not defined; it is just the empty string to which no application can associate a function. For the same reason, shifted <f11> and <f12> are not available in Ubuntu.

However, owing to that fatal Debian decision to switch to console-tools ten years ago, what we have now is evil: a text-mode application that binds a function to, say, <shift><f4> will have the function executed either under openSUSE, Fedora and co., or under Debian and Ubuntu. It is not possible to have both.

Therefore, here is a gentle request to Ubuntu, Canonical, and Mark Shuttleworth: it is time to correct those assignments, and accept the kbd package's defaults for the modified function keys.

On top of that, please give yourself a push and introduce strings for F21, F22 and up. What's the point of assigning these variables to modified function keys, if the variables deliver only the empty string?

With these two mildly reformist measures, 2-3 dozen modified function keys would instantly become available to text-mode applications. As an example, Ubuntu installs the "nano" editor by default. Why is it that "nano" uses only the plain function keys, and resorts to abstruse measures like using <alt><underscore> to extend the set of keybindings? It is because the modified function keys are not coherently defined across distros. It's long past time to overcome that, all the more so since the modified function keys would be international by nature. Just look for <alt><underscore> on non-US keyboards.

2.4. The Tao of SAK

The Secure Attention Key (SAK) is not a physical key; it is a feature that may or may not have an assignment compiled into the kernel. If it is compiled into the kernel, it is triggered by <alt><printscreen><k>. However, whatever the kernel compile options, SAK can also be assigned to any other key of your choice. It kills all processes in the current console, and logs you out. That's practical when you are playing around with some application and manage to get a totally unresponsive keyboard: as long as the kernel is alive, SAK will always respond.

The 8.10 kernel has built-in SAK -- but a buggy SAK:

http://ubuntuforums.org/showthread.php?p=6169912

Pressing SAK gives you a kernel panic, and only switching the power off will help. Kind of odd, isn't it? The built-in SAK cannot be disabled, and so users can just crash the system to their heart's content. Didn't Ubuntu specialize in security?

The remedy here is not editing the keymap; it is leaving out the compile option for SAK (SysReq). Users will still be able to set up SAK in the keymap, if they so wish - of course, after the SAK bug has been squashed.

(The SAK bug, by the way, has been proven to be hardware-specific, see http://article.gmane.org/gmane.linux.ubuntu.user/165993/match=sak+producing+kernel+panic.)

3. How do you modify the Ubuntu's keymap?

First, prepare a text file with the keymap of your liking; you could do that by extracting boottime.kmap from boottime.kmap.gz and modifying it at your leisure. However, if your editing includes reducing those 64 keymaps to sensible size (say, 10), you will be busy for one full hour, at least. It is easier to copy and edit the keymap from any Fedora, openSUSE, or Mandriva to which you have access. Avoid typos; any little error will cause the keymap to be refused. (A CRLF end-of-line is an error; only LF is accepted.) Not required but useful would be a distinctive name, e.g., myown.kmap.

When you are through:

    loadkeys -s myown.kmap

The catch is that you would need to issue this command every time you boot. To have that done automatically, you compress myown.kmap and move it as boottime.kmap.gz to /etc/console-setup. Surprise, it does not work; Ubuntu wants a MD5 hash for the file, and it wants it precisely in /etc/default/console-setup. So, you think you are clever, run md5sum on the new boottime.kmap.gz, and insert the hash as appropriate. Surprise - it still does not work. With or without the correct MD5 hash at the correct place, Ubuntu is determined to not let you muck around with the boot-time keymap.

At that moment, it's like a late Western, post-modern: You do not confront your enemy face-to-face so that the quicker and better and nobler of the two wins. No, you shoot your enemy from behind. Safe and effective.

Among Ubuntu's (run-level independent) boot-time scripts, the last one in execution order is called /etc/init.d/console-screen.kbd.sh. Here, you replace the very first lines of code:

if type setupcon >/dev/null 2>&1; then
    exit 0
fi

with

if type setupcon >/dev/null 2>&1; then
    loadkeys -s /etc/console-setup/myown.kmap
    exit 0
fi

...after placing myown.kmap at the indicated location. It works like a charm. Ubuntu even obliges with a message that it is loading that keymap of yours.

4. Where do we go from here?

Who knows? However, surely the legacy of console-tools has to be overcome, and more consistency has to be achieved for text-mode keymaps across distros.

Technically, the issue is trivial: assign strings to the keys with a functional slant. These are keys that normally do not insert anything; usually, they trigger functions. We are talking about <f1> to <f12>, <insert> to <pagedown>, the arrow keys, <printscreen>, <scrolllock>, <break>. They - plain or in combination with modifiers - don't have to be different across distros or across languages. However, to make them consistent, you need consensus.

And that seems to be a problem.

Talkback: Discuss this article with The Answer Gang

---

A. N. Onymous has been writing for LG since the early days - generally by sneaking in at night and leaving a variety of articles on the Editor's desk. A man (woman?) of mystery, claiming no credit and hiding in darkness... probably something to do with large amounts of treasure in an ancient Mayan temple and a beautiful dark-eyed woman with a snake tattoo winding down from her left hip. Or maybe he just treasures his privacy. In any case, we're grateful for his contributions.
-- Editor, Linux Gazette

Book Review: CUPS Administrative Guide

By Deividson Luiz Okopnik

CAG.RVW 20081130
================
%A Ankur Shah
%C Birmingham, UK
%D 2008
%G ISBN-13: 978-1-84719-258-5
%I Packt Publishing Ltd.
%P 225 pages
%T "CUPS Administrative Guide"
%O http://packtpub.com

---

"Is there so much to administer on CUPS that it requires a 225-page book?" That was the very first question that crossed my mind when I decided to review this book. After reading it, and in reading this review, you may notice, as I did, that there's more than enough material.

The book starts with a quick overview of CUPS features, and some of the history of Linux printing. Then, chapter 2 shows you how to do simple CUPS installation and basic printer configuration.

Chapter 3 guides you through all basic steps of managing your printers, including adding and removing printers and managing print jobs via the command line. Then, it shows you how to use the Web interface to do the same management tasks - in other words, basic CUPS administration.

Starting with Chapter 4, this book starts to shine: it details several advanced CUPS features - and I really mean "details"! In Chapter 4, "Managing Multiple Printers at a Time": the book explains how such management gets done, what the printer classes are, the advantages and disadvantages of using them, how to create configurations, how to configure using both the command-line and Web interfaces, when to use and when not to use various features, and so on.

Other topics covered include management of the CUPS server, details of the CUPS configuration files, client setup including Windows clients over Samba, quotas, server monitoring, per-user print filtering, security, and much more.

The book explains those extra features with a level of detail that really impressed me: yes, there really is enough good material on CUPS administration to fill a 225-page book, with no room left over for padding.

Concluding the review, the book offers a lot of information. More than just information, it gives you the knowledge of someone who seems to have been administrating CUPS for quite some time. Really valuable.

More information about the book can be found in this page: http://www.packtpub.com/printing-with-cups-common-unix-printing-system/book

Talkback: Discuss this article with The Answer Gang

---

Deividson was born in União da Vitória, PR, Brazil, on 14/04/1984. He became interested in computing when he was still a kid, and started to code when he was 12 years old. He is a graduate in Information Systems and is finishing his specialization in Networks and Web Development. He codes in several languages, including C/C++/C#, PHP, Visual Basic, Object Pascal and others.

Deividson works in Porto União's Town Hall as a Computer Technician, and specializes in Web and Desktop system development, and Database/Network Maintenance.

Away Mission: A Vision for Embedded Linux

By Howard Dyckoff

Away Mission: A Vision for Embedded Linux

On the mark for its organization and on target for the current hot topics in the Embedded Linux Community, October's M-Vision was a gem in the small conference category. The "M" stands for MontaVista, a company committed to developing Linux kernels and tools for embedded and mobile devices. This conference, MontaVista Vision, was much more focused and satisfying than the competing mobile and embedded tracks at last August's LinuxWorld conference.

MVison in general

This was an excellent location for a small conference - the SF Palace Hotel entrance is just about 100-150 feet from the entrance to the regional 'subway' (BART). There are multiple bus and municipal rail lines that converge on the area, and it's only a 10-12 minute walk from the ferries.

The session presentations were not available on-line before the conference and there was at first a hint that they would not be posted later, a mistake cleared up by the first afternoon: most of these were on a 1GB USB drive (not included in the conference bag) that had to be requested in the afternoon in the registration area. There were 19 presentations on the USB stick and about 11 missing. The organizers said that all presentations would be posted, and most were indeed up after a week.
http://www.mvista.com/download/topic.php?t=18

Conference perk: a good roller bag with detachable, separate shoulder bag. The main bag was not padded for carrying laptops, however. Well, it would have been nice to be able to use it for every occasion. There were additional perks in the small expo area, but the best was the 2X binoculars at one of the MontaVista podiums.

Of the breakout sessions, only one room had enchained power cords -- this one had a session with hands-on exercises. My own 15-foot extension cord was invaluable in connecting to the few outlets in rooms I attended sessions in, allowing others to share the same power connection.

There was less competition for outlets as most participants took hand-written notes. This is largely due, I think, to MV's reliability in depositing slides and videos on-line after their events. There also were a few product managers and marketeers in the audience.

On the food front, there was a good full breakfast - it was eggs, potatoes, meats, fruit, cereals - and there was a decent box lunch with a bag of gourmet chips that could be carried away.

While there was soda and a tea/coffee service at the breaks, there were no snacks in the first afternoon. A small amount of breakfast pastry was available during the AM keynote break. Cookies and pretzels were put out for the afternoon break on the second day.

For the evening, there was a party with a SF theme and ethnic food stations. There also was an awesome dessert bar, and Nintendo Wii stations and computer arcade rides for fun. Considering that the regular conference cost was only $595.00 (USD) - $350 for customers - this was a very good value.

Besides good conference swag and edibles, there were a good set of keynotes and good mix of sessions on 5 separate tracks:

• Working with Open Source
• High Availability
• Bringing up HW
• Linux on Freescale
• Monte Vista How-to Sessions

M-Vision Keynotes

The MVison opening keynote was by Rusty Harris, CEO of MontaVista. Harris called for more support for open source and explicitly more contributions by members of the embedded community and more events like M-Vision. "It's tough to work together when we can't find each other," he told the attendees.

"The embedded Linux community is growing faster and is building more smart devices than ever, and it shows no signs of slowing," said Rusty Harris. "Generating this growth during today's economic uncertainty shows how strongly committed embedded engineers are to getting the most out of open source."

Harris said that the embedded community is driving Linux now that there is more uniformity and standardization. But he also spoke of a need to counteract the PC and Windows bias of developers and OEMs now that "Linux is the most viable model of open source on the planet!"

He said MV would work more closely with semiconductor companies and pledged MV resources to work with attendees to build a better embedded community.

The second keynote, "Linux for Internet Devices" was presented by Pete Kronowitt from Intel's Open Source Technology Center. He discussed the growth of Linux in mobile Internet devices (MIDs) and presented Intel's view, noting that the embedded market has grown rapidly since 2000.

"Now there are about about 1.4 billion of us attached to the Internet -- but there are another 4-5 billion who will soon be attached. They reside in underdeveloped countries. And affordability will be key to driving the next 2 billion users and probably the 2 billion users after that."

The number of people using mobile devices to get to the Internet is expected to triple from 2007 to 2012, and triple again in another 4-5 years. Kronowitt said the smart phone market is exploding and Linux is growing even faster: over 30% of units in 2012 will be Linux-based.

A Forrester survey in February of 2008 asked potential buyers of a mobile Internet device about the main deterring factor:

• 60% of respondents say phone screens are too small to use online
• 24% say laptops are too big and heavy to carry every where
• 38% say they would pay up to $20 a month for wireless broadband service

So this helps to define what Intel thinks of as the 3 pillars needed for a good mobile Internet experience:

• The right ultra mobile size and performance with long battery life
• Internet compatibility and SW consistency - with video especially
• Always-connected, high-speed wireless

He also gave a plug for Intel's Atom CPU, saying it was the "smallest piece of silicon we've launched to date with our most advanced (45nm) technology."

Intel launched Moblin.org last year with the goal of developing a lean FOSS SW stack to operate everywhere. It also has an application framework in development. The Moblin community is seeking to lower component count and costs with optimized memory footprints.

But something happened along the way: the netbook exploded on the scene! This has changed the game for device OEMs and consumers and expanded Moblin's base rapidly. Currently, Moblin is a Linux-based software platform for building applications that run on devices based on Intel Atom CPU.

Kronowitt said that Intel isn't doing any advertising for Moblin.org: "Code speaks louder than ads." He also hinted there would be new Moblin announcements in November.

Slides and video from this session are at www.mvista.com/download/author.php?a=66.

In another keynote, Deepak Saxena - a principal in the OLPC project - addressed "Linux Power Management Challenges for the OLPC and Beyond".

At first Saxena summarized vision and history of the XO and its current status. He noted it was built for both the tropics and deserts and also for rough handling. It needed a new model of networking for remote locations. The XO was also designed for collaboration up front. It also had a goal of 10-20 hours runtime -- doing something all the while.

Saxena pointed out that the key element in power control is that the graphics engine is not directly linked to the CPU, as in conventional notebooks - the frame buffer is driven from the display controller. The wireless chip also works even if the CPU is in sleep mode.

See the Saxena video at http://www.mvista.com/download/author.php?a=26.

State of the Linux Kernel

Jonathan Corbet, the editor at LWN.net, gave the second day keynote. Corbet reviewed the huge on-going effort that goes into a Linux release and ways to contribute. He put up a slide showing the wide diversity of of industry kernel contributors. Red Hat, IBM, Novell, and Intel are standouts.

Who contributes?

from 2.6.23 -> 2.6.27-rc5

(Non-affiliated)         19%
Red Hat                  12%
IBM                       7%
unknown                   6%
Novell                    6%
Intel                     5%
Parallels                 2%
Oracle                    2%
linutronix                2%
consultants               2%
Movial                    2%
SGI                       1%
academia                  1%
Analog Devices            1%
Renasas Tech              1%
Freescale                 1%
MontaVista                1%
Fujitsu                   1%
Google                    1%
Astaro                    1%

Corbet also spoke about new Linux file systems, specifically ext4 as a descendant of ext3. This adds journal checksums and also lifts size limits from files and filesystems. He also discussed btrfs, which has a faster fsck, full checksumming, subvolumes and snapshots.

Corbet said there was lots of on-going work in drivers and the kernel and also in userland space. He also asked why a project hasn't been started to port DTrace to Linux from Solaris.

The slide deck from his presentation and related videos are available here: http://www.mvista.com/download/author.php?a=8

Fast Boot Session

One session I'd recommend is "Fast boot: Tips & Techniques for Improving Linux Start-up Time" by Christopher Hallinan, who contributed to the u-boot boot loader.

A typical embedded system contains a bootloader and kernel, both of which are typically configured with many useful default features that may or may not be important for a given product requirement. This presentation shows how to significantly reduce boot time for embedded systems.

Hallinan saw Intel's 5-second boot demo at the recent IDF conference and wanted to see how far he could go on his own. He set this up as a personal project to see what was do-able in a short time. How short? How about 5-7 person-days over a month?

Hallinan said he thought that Intel spent over 40 man-weeks on its effort. And he got within 10% of Intel's 4.5-second boot. Not bad at all.

Hallinan based his efforts in part on suggestions from Tim Bird of the Linux CE forum, who had written a paper on reducing boot times (http://kernel.org/doc/ols/2004/ols2004v1-pages-79-88.pdf).

One key delay noted there is ide_init, which can take over 3 seconds at boot time, even in modern Linux kernels. Bird noted that "The call sequence underneath ide_init() shows that a large number of calls are made to the routine ide_delay_50ms()...." Many embedded devices use SSD storage so this can be stripped out of many embedded kernels.

Hallinan used a Freescale MPC8349 Emitx board and U-Boot loader. This setup had 256 MB of RAM and 16 MB of flash with a 533 MHz clock, plus serial and USB and net ports.

With a 2.6.25 kernel with typical configuration he had a boot time of 1:15 to cmd prompt, with ~5 seconds for U-Boot alone.

A full description of the tech sessions is here: http://www.mvista.com/vision/index.php?p=sessions#bordug

And all the presentations are here: http://www.mvista.com/download/topic.php?t=18

MontaVista also held a technical bootcamp on embedded Linux drivers the day before M-Vision, pricing it at only about $100 more for M-Vision conference attendees.

The tutorial instructor was Mike Anderson, chief scientist of the PTR Group, an experienced instructor on embedded development. Hands-on labs were included covering driver entry points, blocking I/O, moving between user and kernel space, driver debugging, and other topics.

Talkback: Discuss this article with The Answer Gang

---

Howard Dyckoff is a long term IT professional with primary experience at Fortune 100 and 200 firms. Before his IT career, he worked for Aviation Week and Space Technology magazine and before that used to edit SkyCom, a newsletter for astronomers and rocketeers. He hails from the Republic of Brooklyn [and Polytechnic Institute] and now, after several trips to Himalayan mountain tops, resides in the SF Bay Area with a large book collection and several pet rocks.

Howard maintains the Technology-Events blog at blogspot.com from which he contributes the Events listing for Linux Gazette. Visit the blog to preview some of the next month's NewsBytes Events.

Jetty: The Twelve Year Journey to Market Maturity

By Adam Lieber

Normally, I'd redirect content like this (advertising, to be frank) to our NewsBytes team, to be brutally stripped down and turned into a minimal press release [1] - but somehow, perhaps because it seemed a little more sincere than the average product puffery, or perhaps because it's Jetty's birthday, or maybe because it's the holiday season here in the US and I've read a few too many O'Henry stories when I was young - I stopped just short of it. Heck, maybe publishing this will encourage other open source projects to wax lyrical about their particular baby and explain to us all exactly why their hunk of code is so amazingly wonderful.

Sometimes (especially right after I accidentally discover a piece of FOSS software that does exactly what I've desperately needed a program to do, possibly for years), I think that we as a community take this "we don't do $200M advertising campaigns with naked cuties and catchy jingles" attitude too far; there's nothing wrong with at least a few catchy tunes and a naked cutie or two - especially when done by enthusiastic volunteers. Even Linus Torvalds has been known to describe the goal of Linux as "World domination. Fast. [...] and scantily clad females, of course. Who cares if it's below zero outside." So, why not start somewhere?...

[sigh] I'm rambling again. Perhaps I've been proofreading these articles a little too long and it's time for a shot of Nassau Royale. But my point about the cuties - with or without some jammin' music - remains; any FOSS project that can play a jingle and show some wiggle (well, OK, it's also got to be a good project, but there's LOTS of those) is sure to be wildly popular, and I'm officially pinning my hopes to seeing it in the near future. Legacy software won't have a leg to stand on.

---

Jetty was written as a lightweight Java web and application server and in many ways it has been ahead of its time. Jetty with Cometd handles two-way communications via a method known as long-polling which allows servers to efficiently handle requests in an environment where constant or long-lived communication is needed between the server and the client.

Each month, Jetty grows in number of users and market share. With the rise of rich Internet applications and the need for increased embeddability, Jetty has become increasingly popular and is on track to become the leading Web and application server. To celebrate its 12th year, here are 12 things that are interesting about Jetty:

1. Jetty was originally written as an issue tracker application

1995 – America was in a media frenzy over the OJ Simpson murder trial, Microsoft had released Windows 95, and at Mort Bay Consulting Limited in Australia, Greg Wilkins created what was possibly the first Java web application, an issue tracking system served over HTTP from a Java server. It soon became apparent that there was a lot more interest in the HTTP server within the issue tracking application than in the application itself. By 1996, the HTTP server was spun out as an early form of Jetty.

2. Jetty was originally named MBServler

Named for Mort Bay, an area of Sydney, the first itineration of Jetty was called Mort Bay SERVLet servER, or MBServler for short. The leading figures behind Jetty, Greg Wilkins and Jan Bartel, were not entirely happy with this. “We changed it to Jetty pretty quickly,” says Wilkins. The Mort Bay logo is an image of Sydney Harbour which includes a small jetty. As the word jetty started with a J, it was picked as the new name for the Java web server.

3. Jetty has frequently broken tests that try to benchmark it

Jetty breaks most benchmarking tools, as it is designed to scale to many thousands of simultaneous requests and connections. Most HTTP benchmarking tools do not support the asynchronous features needed to scale to such levels and thus bottleneck at a few hundred or a thousand connections. Jetty now includes an asynchronous HTTP client that was written initially just to be able to test tens of thousands of simultaneous requests and connections.

4. More people are using Jetty than you might think

Jetty is used by a huge number of brands and applications. In fact, many people use Jetty and don’t even realize it. Eclipse, BEA WebLogic Event Server, Apache Geronimo, Zimbra, IBM Tivoli Netview, Sybase EA Server, IGN.com and Chess.com all use Jetty behind the scenes. In fact, the first WiFi access provided at Starbucks was handled by a proxy based on Jetty. Apache Maven also works great with Jetty. And there are many more – a full list can be found at http://docs.codehaus.org/display/JETTY/Jetty+Powered

5. Hundreds of thousands of domains are active on Jetty

Netcraft is a company that provides research and analysis on numerous aspects of the internet. Netcraft makes it possible to see how many instances of Jetty are directly linked to the Internet, as opposed to behind a firewall. According to the company, there are hundreds of thousands of domains active on Jetty – and this is very likely to be a conservative estimate.

6. Jetty was designed backwards

One of the reasons that Jetty is so prevalent is that it was designed to be a good software component, and it happens to also be a software container. Jetty embeds so well that it is often completely hidden, while other Web servers are primarily software containers and not designed to be embedded within other applications. While they can be embedded in applications, the applications more often than not take on the shape of the Webserver rather than vice versa.

7. Jetty proves that size matters

Jetty’s small size makes it perfect for providing Web services to all applications – even those on handheld devices. Because of its small footprint, Jetty leaves more memory and cache free to be allocated to running the application rather than running the server.

8. Jetty is the first server for Google Android

Jetty will be the first server available for Google’s Android mobile platform, in the form of i-Jetty. i-Jetty will allow users to turn their phone into a server, and broadcast and share videos and photos on their phones. Users will also have the ability to manage their phones from their computer desktop. Essentially, i-Jetty and Android will give mobile phone users much more control and flexibility over their devices.

9. Jetty is open source

That Jetty is open source is no secret – it is, in fact, its calling card and the people behind it believe passionately that open source methods are the best way to produce quality software that is well targeted to users' needs. There is no better way for various people to collaborate, both in development, and in discussion of requirements. It was the right decision, as the number of Jetty users and developers today shows.

10. Jetty is an important component in Eclipse

Jetty handles the help system in the open source Eclipse software platform. Eclipse projects are hosted by the Eclipse Foundation, a non-profit organization whose members include IBM, Oracle, Intel, and Nokia.

11. Jetty is a record breaker

Earlier this year, Yahoo broke the record for the fastest sort of a terabyte of random data. The record was achieved by one of Yahoo’s Apache Hadoop clusters, in which Jetty was a crucial component. The new record stands at 209 seconds, compared to the previous record of 297 seconds.

12. Jetty is supported by Webtide

The Jetty community is based at http://www.mortbay.org. However, the founders of Mort Bay realized that a successful open source project needs commercial input and support, as a member of the community, to work with "the" community. So webtide.com was founded to provide commercial support and development provider for Jetty.

Jetty has evolved a great deal over the years, from a poorly named bug tracker to the flexible software it is today. It keeps growing, breaking records and benchmarks and moving into interesting new areas such as Web 2.0, gaming and mobile apps. It is poised to keep growing, because the community continues to thrive, with up to 20 people developing it at any given moment. The future is not just secure for Jetty, it’s positively buoyant.

---

[1]

"You had to admire the way perfectly innocent words were mugged, ravished,
stripped of all true meaning and decency, and then sent to walk the gutter
for Reacher Gilt, although 'synergistically' had probably been a whore from
the start."
 -- Terry Pratchett, "Going Postal"

Talkback: Discuss this article with The Answer Gang

---

Adam Lieber is the CEO of Webtide. Adam was a co-founder of Gluecode Software, one of the early open source software companies and delivered solutions to numerous industries and served as Gluecode's representative on OASIS. After Gluecode's acquisition by IBM, Adam ran worldwide sales for open source middleware for IBM.

Prior to Gluecode, Adam was at the IT-focused venture capital fund Mission Ventures.

Adam received his A.B. in Economics concentration, with Computer Science from Princeton University.

Joey's Notes: Basic Samba Configuration

By Joey Prestia

Samba is a program that runs on Linux and allows seamless interaction between the Linux and Windows operating systems. By using Samba, your users can be using Windows and actually access their data on a Linux server. This benefits the administrator - in the form of ease of administration, along with the inherent security of Linux - while also benefitting the user through familiarity. This month's article covers basic setup of a Samba server for use as a user home directory server, and configuration for a simple group collaboration share.

"Microsoft's file access protocol [is] called the Com