All BrfrtiiPtf

Cr'- lade.1, Jain oenler rnanogemeni salulians arteri □ lull ange of security features across ils entire product line of corso:e =e*ver power rrantigemer.! KVhl, biofinelric scanner and nelwarl' mcjiwjgwrant With 55.-2, IF' Reefing, strong authentication event 'oçging and dala logging, Cyclades can maVei your network inlo o secure heovywHighl contender in the date center *world

For ¡1 FREE white püiwi on (MA ocntcr security, pLçj^u vieil us it wvn^-cycliiflK OTm/çcctfitywp www^cyclade&conyija CyclädGS

1 HÜH rs-r.riric-s - ulna n cyEladdE.Hirn FvrvYwh: ■1 wrth . InUH

¡»ilïflcir, IjUllUn ^ UMlUK II itH Ulirou l lII fuilli id ItUI lUrlLN -.^Hù'. I IL J iftHIULr vifril Ii :

/usr/s rc/li nux/CREDlTS

Linux kernel hackers at the 2003 Kernel Developers' Summit in Ontario, Canada, photographed by Richard Guy Briggs and submitted by Robert Love.

milter-sender: milter-sender milter-sender is a Sendmail milter plugin that connects back to the sender of a message as it's being received. Basically, if the sender doesn't exist on the sender's alleged primary mail server, the message is almost certainly spam and is rejected. This plugin has reduced my spam significantly. And, although the arms race isn't over, I always can add more milter plugins. I'm now running milter-sender and SpamAssassin via milter. Requires: Berkeley DB, Sendmail with milter, libsnert and glibc.

LJ Index—January 2004

1. Number of changes to the Linux 2.6 kernel in development from 2002 through mid-September 2003: 12,479

2. Number of merges in the kernel, not counting apply-by-proxy patches: 2,386

3. Approximate average number of daily merges: 9

4. Number of individual developers working on the kernel: 554

5. Percentage of all changes done by the top 10% of kernel developers: 78

6. Percentage of all changes done by the top 20% of kernel developers: 90

7. Linux percentage of the paid server operating environment (SOE) category in 2002: 23.1

8. Microsoft percentage of the paid SOE category in 2002: 55.1

9. Millions of new licenses shipped in the SOE environment in 2002: 5.7

10. Linux percentage share of the client operating environment (COE) category in 2002: 2.8

11. Macintosh percentage share of the COE category in 2002: 2.9

12. Percentage share of the COE category in 2002 for other OSes: 0.5

13. Microsoft percentage share of the COE category in 2002: 93.8

14. Number of operating systems other than Linux and Windows that grew in revenue in 2002: 0


1-6: Linus Torvalds 7-14: International Data Corp.

They Said It

There are a lot of reasons for the penguin.

I was bitten by a penguin. And it's a true story. It's funny, because there are a lot of Web sites about the penguin. There's The History of Tux ( and things like that. And some of these Web sites have some of my explanation. And they almost universally say, "It's a great story, but it's not true."—that I was bitten by a penguin.

It's true! I was bitten by a penguin!

I mean, really! Take it from me! I'm wounded!

Okay, so he wasn't six foot tall.


We're very aggressive around migrating to Linux; the target is to buy no new UNIX kit Until one year ago most mainstream applications were on UNIX. The back-office systems are on main-

frames, but everything else was on Sun.

—LEHMAN BROTHERS CIO EMEA RUSSELL ( 2d7f2e75fecd8bbfc2256d7800317032)

Anything touching Linux.


You Decide-..Linux ready on every board lanPMC-4Gx

Quad part Gigabit Ethernet capper or fiber PMC adapte

© 2 or A a uto-nec| abating 10/100/1D00 Ethernet ports

I Fiber version supports long an<i short reach

© Anfto sensing speed, simpler/duplex, How control, and detection of poia^ly and cable renglhs on copper version

Channelized quad part T1/E1/J1 PCI WAN adapter e 12B channels (D50) & Line build outs from DSX to DS1 0 Supports D4, SF, ESFr SLC-96, JT-E7D4 framing C Selectable clock options

6U CPCI processing blade with dual PTMC

9 333 MHz PowejPC CPU

G PlCMG 2.1 & Packet Switching Backplane enabled t Dual 10/100 Ethernet ports includesTirmeSys Linux GPL Software Developer's Kit 0 Drivers for T1/E1, T3/E3, HSSI, V.35, X.21 4 other SEE PMCs

WAN * LAW » Storage * Carriers » Custom Flexibility on demEind 92S-355-2QGD [email protected] www.-sbEH.ccnn diff -u: What's New in Kernel Development

Intel has come out with some event notification patches that provide profiling hooks at process termination, in addition to the already existing hooks provided at process creation. Intel's work should be useful for enhancing such profiling tools as Oprofile, Perfmon and VTune. Although adding hooks like these potentially can provide a way for proprietary software vendors to bypass the GPL licensing restrictions, it does not appear that Intel's work suffers from that problem. Frederic Rossi also has been working on his own unrelated event notification code, called asynchronous event mechanism (AEM). AEM allows general-purpose user-space applications to register for system events by providing their own sets of event handlers [LJ, July 2003]. The main goal for this is to be able to pass data quickly from the kernel back to user space. Although Intel and AEM's work are not directly related, they reveal an interesting trend in Linux development toward greater and greater communication between the layers of the system.

The DSI Development team has produced a standalone module, called digsig, that checks the embedded digital signature of an executable program before running it. If the file has been modified, digsig prevents its invocation. A wide variety of applications could make use of this—for everything from worm protection to digital rights management. Unlike Trusted Computing's attestation feature, digsig's verification process remains under the control of the machine owner. It is not possible, for example, for remote systems to rely on the digsig module on your system to verify that your computer is running particular versions of particular software that those remote systems have approved.

Apparently, all AMD Athlon processors suffer from a bug that can cause prefetch instructions to receive memory management faults for certain addresses under certain conditions. This is similar to the infamous f00f Pentium bug, in that both are bugs within the hardware itself that the OS simply must work around. The good news is that, in most cases, the bug doesn't do any actual harm. Triggering the bug is rare, which may account for why it's gone undetected for so long. Andi Kleen has been working on a Linux workaround for 2.4 and 2.6, and the AMD folks have affirmed that his workaround will be forward-compatible with future Athlons that have the bug corrected in hardware. However, according to Andi, some Athlons, including the K7, may never be fixed, and so his workaround may be needed on new systems for quite a long time.

SGI has released a GPLed driver for its Altix serial con sole, but in the copyright notice, SGI explicitly reserves the rights to any patents used in that code. Although the Free Software Foundation has confirmed to SGI that this is not technically a violation of the GPL, it could result in a situation in which code from a GPLed project (like the kernel) would not be allowed to be used in other GPLed projects. The question then becomes, should the code be allowed into the kernel under those conditions? Alan Cox has decided to investigate the situation with both SGI and the FSF to confirm that the FSF really doesn't consider it a violation of the GPL and to see if SGI is willing to modify its position.

Chris Rivera and Robert Love have released the slabtop standalone utility, a tool to monitor the allocation and freeing of memory in the cache. The kernel maintains a cache of allocated memory that it disburses to processes as needed. But sometimes this memory needs to be freed, and the slab is the way that such freeing is accomplished. The slabtop tool monitors each slab cache, presenting data for that slab in real time, and it also concurrently updates a set of global statistics on all slab activity on the system.

Matthew Wilcox has begun separating kernel headers into two categories, those that should be visible to user-space sources and those that shouldn't. It's always a problem when a user application comes to rely on the wrong kernel headers. As the kernel grows and changes, the headers may change as well, and the user software may break. To recover from this, a strict separation must be maintained. Matthew's work helps make this easier. He also managed to inspire other folks, such as Erik Andersen, to start cleaning up some type naming conventions, in preparation for 2.6.

A new division of kernel code has been initiated, similar to the arch/ directory. But instead of dividing code on an architecture-by-architecture basis, the new division works on a compiler-by-compiler basis. Ever since Intel's ecc compiler became the second compiler (after GCC) that could build the Linux kernel successfully, subtle conflicts have crept in, causing developers to start submitting patches specifically to handle particularities of one compiler or the other. After some recent debate over support for in-line assembly, which GCC supports but ecc does not, Linus Torvalds decided to create per-compiler directories to house all patches relating to particular compilers. Sam Ravnborg and David Mosberger apparently have decided to lead that effort, and they already have submitted patches to initiate the transition.

They Said It

It is very very hard to find people who don't flame and are calm and rational— and have good taste. I mean it's like...give me one honest man. It doesn't happen...too much. And at the same time, when it happens, it matters a lot. Just a few of these people make a huge difference.


In the early days, Linux was under the wire without the chief information officer's (CIO) knowledge. It was the technical decision for things such as a file and print server, with cost or reliability as the driver. Now we have CIO adoption.


A slide rule is a precision instrument. A good one will last a long time, in fact, for a lifetime unless you abuse it. You've got to look ahead for 10, 20, 30 or even 40 years if you want to be sure you are getting your money's worth.


Rackspace — Managed Hosting backed by Fanatical Support."

Servers, data centers and bandwidth are not the key to hosting enterprise class Web sites and Web applications. At Rackspace, we believe hosting is a service, not just technology.

Fanatical Support is our philosophy, our credo. It reflects our desire to bring responsiveness and value to everything we do for our customers. You will experience Fanatical Support from the moment we answer the phone and you begin to interact with our employees.

Fanatical Support has made Rackspace the fastest-growing hosting company in the world. Call today to experience the difference with Fanatical Support at Rackspace.

Thanks for honoring us with the rackspace

T READERS' 2003 Linux Journal « choice Readers' Choice Award for MIIIMl "Favorite Web-Hosting Service'


1.888.571.8976 or visit us at

Was this article helpful?

0 0

Post a comment