LJ Archive


diff -u: What's New in Kernel Development

Zack Brown

Issue #175, November 2008

PowerPC maintainership gradually will migrate from Paul Mackerras to Benjamin Herrenschmidt. The reason for taking time on the transition is because PowerPC is a very big, very active project with trees, branches and processes of its own. Paul's estimate is that six months should be enough time to do the transition. But in the meantime, Benjamin will be doing a sudden two-week takeover of the project while Paul is on vacation. This involves setting up a new git tree and relevant branches, and starting to accept patches from the large group of contributors. It's not clear whether success in the two-week period will shorten the overall transition, but it certainly will be a good test run.

OProfile also probably is changing maintainership. Because this involves both a kernel and user-space code set, there are several issues. Robert Richter will be taking over from Philippe Elie as maintainer of the kernel portion. It's also unclear whether Philippe is onboard with this change, so it may turn out that OProfile has some deeper communication issues between the developers. The user-space portion of OProfile also seems to be changing hands, though not as formally as with kernel code.

The kmemcheck code now has an official maintainer—or actually two official maintainers. Vegard Nossum and Pekka Enberg will share maintainership. The kmemcheck patch is a cool little debugging tool that logs whenever memory is read that had not been written previously. Clearly, there's a bug if we're trying to read data from a location to which we never wrote.

The ZFS filesystem may start to have more of a presence in the official kernel tree. Sun has released a read-only version of the filesystem under the GPL. This is pretty cool, but not as cool as if Sun had released the whole thing. A read-write version could be useful to anyone, but the read-only version will be useful really only to people who've been using Solaris. In its full form, ZFS is a very interesting filesystem—it can handle multiple exabytes of storage, pools multiple block devices seamlessly together, checksums everything and snapshots the history of all data changes on the filesystem. Alan Cox speculates that ZFS may be one of the only things keeping Solaris alive as an operating system, and that this is why Sun doesn't want to GPL the full code. Even if a read-write version were open sourced though, there still are patents covering parts of ZFS, and those patents currently are being fiercely litigated by Sun and NetApp. Even if the source code were fully available, the problem of getting permission to use the patents might be too thorny to overcome. In the meantime, Kevin Winchester has volunteered to port the read-only code into Linux, and he'll almost certainly have help—or at least advice—from Christoph Hellwig. Ricardo Correia also is attempting to rewrite ZFS from scratch as a FUSE-based filesystem. Ricardo's work already has made great strides, to the point where Patrick Draper, for one, has been able to use it for all his data. However, anyone trying it out, he cautions, should be sure to back stuff up.

Karsten Keil submitted mISDN, a driver intended to replace the I4L architecture for passive ISDN cards. The new architecture is apparently a big improvement, as folks like Tillman Schmidt, who maintain the old I4L drivers, are very excited to get started using it.

The touchscreen driver continues to graft new and exciting touchscreen hardware onto its list of supported devices. Alastair Bridgewater recently added support for the eGalax touchscreen used in the HP tx1305us tablet PC as well as in other devices. Part of the problem with supporting all these different pieces of hardware is that the driver has to detect them by probing their various behaviors. This can lead to some very subtle weirdness, as Alastair discovered when his first patch submission was rejected on the grounds that it didn't programmatically detect the eGalax hardware.

David Altobelli has submitted code supporting the HP iLO/iLO2 management processor, used by system administrators to access servers remotely instead of at the console.

Michael Buesch has written a driver supporting the GPIO pins on all Brooktree 8xx chips. GPIO (General-Purpose Input/Output) pins are used for both input and output, depending on how they're configured. Michael's code supports that configuration. After a bit of trouble locating the appropriate maintainer, Andrew Morton finally pointed Michael to David Brownell, who could receive the patch.

Lifting the Fog from Cloud Computing

Shawn Powers

Issue #175, November 2008

Back in August 2008, at LinuxWorld in San Francisco, the big buzzword was “Cloud Computing”. It's a neat concept, but after a week of hearing folks talk about “in the cloud”, I was about at the end of my rope. To add insult to injury, it seemed that the San Francisco fog confused many folks, and “Cloud Computing” started to be used synonymously with “Grid Computing”, “Clustered Virtualization” and “My Company Is Cool”.

For clarity's sake, I thought a brief vocabulary lesson was in order. Cloud computing is indeed a viable, exciting idea—but it helps if we all know what we're talking about.

The idea behind cloud computing is that services, not servers, are offered to the end user. If people need a Web server, they buy Web services from the “cloud”, and have no idea what is actually offering them the serving. The “cloud” essentially hides the server infrastructure from the client, and ideally scales on the fly and so on. Much of the confusion in terminology happens, because the cloud of services almost always is powered by a grid of computers in the background. Cloud computing itself, however, is just the abstraction of services away from servers themselves.

The advantage is that a vendor can offer more reliable, diverse and scalable services to a user without the cost of dedicating hardware to each user. This allows for more graceful temporary spikes (Slashdot, Digg and so on), while not letting servers sit idle during low times. Because the back end is transparent to the user, those actual grids of computers in the background can be geographically diverse, and oftentimes virtualized for easy migration, all without any end-user interaction. Ideally, it offers a reliable “service” to the end user, at a lower cost, and gives vendors flexibility in the back end, so they can manage servers in the most efficient way possible.

Most people don't realize that cloud computing ultimately is shared hosting. Vendors avoid terms like “shared hosting”, because that implies multiple people sharing a single computer. By its strictest definition, however, cloud computing certainly could be run from a single back-end server. With current scalability and virtualization technologies, vendors have much more robust ways to serve to the “cloud”, and the traditional hangups with shared hosting are largely eliminated. Still, it's important to understand what cloud computing really is, so you don't get fooled into buying more or less than what you truly need.

  • What sort of back-end servers are you running?

  • Do you have the ability to fail over to a secondary data center behind the cloud, transparent to me?

  • How do you differ from traditional shared hosting? (This one should spark some heated retorts!)

  • How well do you scale, and how does pricing work for occasional spikes?

LJ Index, November 2008

1. Percentage of employees in Florida who go to work when they're sick because they need the money: 44

2. Percentage of employees in Ohio who go to work when they're sick because they need the money: 50

3. Percentage of Delta Airlines domestic fleet that will have Internet access for passengers by mid-2009: 100

4. Exceeded average Internet-generated IP traffic in terabytes per day in 2007: 9,000

5. Expected exceeded average Internet-generated IP traffic in terabytes per day in 2012: 21,000

6. Price in dollars per megabyte of wireless texting: 1,000

7. Price in dollars per megabyte of wireless voice: 1

8. Price in dollars per megabyte of wireline voice: .1

9. Price in dollars per megabyte of residential Internet: .01

10. Price in dollars per megabyte of backbone Internet: .0001

11. Percentage of the Top 50 Most Reliable Hosting Company sites that run on Windows: 18

12. Percentage of the Top 50 Most Reliable Hosting Company sites that run on Linux: 50

13. Percentage of the Top 10 Most Reliable Hosting Company sites that run on Linux: 60

14. Percentage of the Top 5 Most Reliable Hosting Company sites that run on Linux: 80

15. Percentage of the Top 3 Most Reliable Hosting Company sites that run on Linux: 100

16. Maximum thousands of Ultra-Mobile PC (UMPC) shipments in 2007: 500

17. Projected millions of UMPCs to ship in 2012: 9

18. Millions of ASUS Eee PCs sold in the first half of 2008: 1.7

19. Total 2007 Linux ecosystem spending in billions of dollars: 21

20. Projected 2011 Linux ecosystem spending in billions of dollars: 49

1, 2: NPR

3: Delta Airlines

4, 5: “Managing Proliferating Traffic Growth”, Pieter Poll, PhD, Chief Technology Officer, Qwest

6–10: “Network neutrality, search neutrality, and the never-ending conflict between efficiency and fairness in markets”, by Andrew Odlyzko, Digital Technology Center, University of Minnesota

11–15: Netcraft.com report covering July 2008

16–18: IDC, via Investors Business Daily

19, 20: “The Role of Linux Servers and Commercial Workloads”, by IDC for the Linux Foundation

Keeping the Kernel Klean

Doc Searls

Issue #175, November 2008

Operating systems drive devices. Linux is driven by open-source imperatives. So, naturally, Linux's kernel developers have a problem with closed-source kernel modules. And, just as naturally, they've hacked up a statement they hope will discourage the closed and encourage the open.

On his blog, Greg Kroah-Hartman explained, “As part of the Linux Foundation Technical board...we wanted to do something that could be seen as a general 'public statement' about them that is easy to understand and point to when people have questions”. Here it is:

Position Statement on Linux Kernel Modules, June 2008

We, the undersigned Linux kernel developers, consider any closed-source Linux kernel module or driver to be harmful and undesirable. We have repeatedly found them to be detrimental to Linux users, businesses and the greater Linux ecosystem. Such modules negate the openness, stability, flexibility and maintainability of the Linux development model and shut their users off from the expertise of the Linux community. Vendors that provide closed-source kernel modules force their customers to give up key Linux advantages or choose new vendors. Therefore, in order to take full advantage of the cost savings and shared support benefits open source has to offer, we urge vendors to adopt a policy of supporting their customers on Linux with open-source kernel code.

We speak only for ourselves, and not for any company we might work for today, have in the past or will in the future.

Below that are 176 names.

The Linux Foundation has a slightly broader statement:

The Linux Foundation recommends that hardware manufacturers provide open-source kernel modules. The open-source nature of Linux is intrinsic to its success. We encourage manufacturers to work with the kernel community to provide open-source kernel modules in order to enable their users and themselves to take advantage of the considerable benefits that Linux makes possible. We agree with the Linux kernel developers that vendors who provide closed-source kernel modules force their customers to give up these key Linux advantages. We urge all vendors to adopt a policy of supporting their customers on Linux with open-source kernel modules.

Either way the message is clear.

They Said It

The slow adoption of Vista among businesses and budget-conscious CIOs, coupled with the proven success of a new type of Microsoft-free PC in every region, provides an extraordinary window of opportunity for Linux....We'll work to unlock the desktop to save our customers money and give freedom of choice by offering this industry-leading solution.

—Kevin Cavanaugh, Vice President for IBM Lotus Software, blogs.zdnet.com/open-source/?p=2754

Linux should stop copying Windows circa 2001 and rather look at what Apple is doing these days around usability and design. I understand that to gain acceptance for new software it makes it easier for users if you mimic the behavior of the old software, but at some time, you need to step out and innovate in the user interface.

This “innovation” might just be driving maximum consistency in look and feel. I want to at least feel that the system is held together by well-engineered common design principles and APIs rather than aging string and bubble gum.

Open-source projects like WordPress have attracted excellent graphic designers to build themes and skins. We need more graphic designers involved with open source. We need to build more software so that others can make it look great.

There are certainly many in the general Open Source community who understand this and are working toward this goal.

—Bob Sutor, VP Open Source and Standards, IBM, www.sutor.com/newsite/blog-open/?p=2455


Here's another neutral, open network: the electricity grid. It's transparent, open, anybody can do anything on it. As long as you know the protocols, you can plug any technology in to it. We can imagine a world where when you plugged something in, the network asked, “Is it a Panasonic or a Sony TV you're plugging in?” “Is it radio or television?” “Is it pay TV or free TV?” And then, depending upon the answer to these questions, allocate the resource according to that information. It's possible to imagine an electricity grid like that, but would it be better?

—Lawrence Lessig, testifying to the FCC at a Stanford hearing in April 2008, www.lessig.org/blog/2008/04/testifying_fcc_stanford.html

Well, the e2e advocates are essentially arguing that end-to-end in engineering is the equivalent of the perfect, competitive market that economists know and love....But in fact, that's not the way the real world works. It's neither the economist nirvana of perfect competition nor is it the engineers' nirvana of e2e. It doesn't work that way.

—Gerlald Faulhaber, then the chief economist at the FCC, during a Stanford panel on e2e (end-to-end) in December 2000, cyberlaw.stanford.edu/e2e/papers/e2e.panel5.pdf

YOU Are LinuxJournal.com

Katherine Druckman

Issue #175, November 2008

Yes, that's right, LinuxJournal.com is all about you. We want to know what makes you happy, and we want to share it with all our readers. You may have noticed something a little different with our Web articles—you get to rate them. What better way to let us know what works and give other readers the benefit of your experience? So, feel free to judge, rate, applaud, rant or anything else that strikes you. Did I mention, it's all about you?

One thing we already know is that you are fans of OpenOffice.org. Be sure to check out our regular tutorial articles on-line, and you might learn a new trick or two. Whether you need to unlock the mysteries of line spacing in OpenOffice.org Writer or understand Calc functions, the answers are there. You also may happen upon a new piece of software that will quite literally broaden your horizons. If you missed Mike Diehl's article “Exploring Space with Celestia”, go check it out at www.linuxjournal.com/content/exploring-space-celestia. It is one of my personal favorites.

I hope to see YOU around the Web!

What They're Using: Alan Robertson's Eee PC Tote-Bag Hack

Doc Searls

Issue #175, November 2008

I ran into Alan Robertson at LinuxWorld in August 2008, at a keynote given by an IBM executive. Alan works for IBM, where his title is Senior Software Engineer, Business Resilience. That's biz-speak for Alan's leading work on High-Availability Linux. It was while talking about the latter that Alan let slip that he uses an ASUS Eee PC.

Alan Robertson

I was fascinated by the extreme polarity in scale between Linux-HA—with its heartbeat-enabled death-of-node detection, cluster management and other mission-critical industrial-grade uses—and the hot little portable. When we got to a table where he could show off the machine, Alan pulled out of his knapsack what looked like a colorful striped purse. It was an accessory bag Alan had hacked to hold the Eee PC as if it were nothing more than an extra battery. It was way cool.

Eee PC Bag Hack

“The deal is...it's supposed to be just for cables that you carry with your regular laptop, but it happens to fit the Eee PC”, he said. That's because the bag comes with three pockets on each side, and Alan removed the two lines of stitching on one side, giving it one pocket for the Eee PC and three for its accessories. The result is an all-in-one case that's custom-hacked for the Eee PC, which fits snugly inside the de-stitched side, across from its relatively tiny AC power supply and whatever other small stuff you'd like to carry with it.

Alan's Eee PC has the new 9-inch screen, 20GB of solid-state storage and stock Xandros distro. Beyond its dimensional virtues, Alan likes the Wi-Fi signal-finding utility, because “it shows you the DHCP negotiation” and “all that stuff you need to know” if something isn't working—which it too often doesn't with Wi-Fi. He also showed off the Eee PC's multitouch, which I hadn't seen working before outside of an iPhone. And, he noted how well Skype is integrated with Linux. “Skype already knows about the Webcam...so if I were video-conferencing you, it would just come up and work.”

Alan's main laptop for work is still a Lenovo ThinkPad 360p. He notes that the airline seat power supply for that one weighs more than the Eee PC does by itself. Meanwhile, his Eee PC is a nice little all-in-four (pockets) portability solution.

If you're interested in Alan's bag hack, you want BuiltNY's “charger bag”. It comes in four color themes: brown/mint green, wood-grain black/slate, black/powder blue, and Alan's version, called stripe #7/lava. BuiltNY retails the bag on its Web site for $25, but at the time of this writing, Amazon sells it for $13.97. Either way, it's in alignment with the Eee PC's own downscale cost.

LJ Archive