KernelTrap Interviews Andrea Arcangeli

An anonymous reader writes "Andrea Arcangeli completely rewrote the 2.4 Linux kernel virtual memory subsystem several years ago, a surprising event during the evolution of a stable kernel series. A very intelligent 27-year-old from Italy, Andrea spoke with KernelTrap in great detail about the past, present and future of his Linux kernel efforts. An interesting interview ."

Comments

Re:Excellent

[Anonymous Coward]

Andrea derives from the ancient greek andros which "means" man so Andrea should be a male name everywhere!

I think he likes "free as in beer" more then...

[Anonymous Coward]

"free as in rights" That whole Bitkeeper talk got me thinking that maybe this guy is a little on the cheap side. Doesn't both him its binary, or closed source and would use it if it was freeware. Buddy why don't you take all that money your saviing from living with your parents and buy the damn thing.

Although, the license to use bitkeeper is really anti-competitive and agree when he said, "if no open source project could ever beat bitkeeper in the long run, Larry wouldn't need this weird licence in the first place."

Maybe Andrea would fix PCIPCI DMA for us?

[Anonymous Coward]

On modern CompactPCI hardware I should be able to DMA between (for instance) SCSI and a DSP array on the PCI bus.

Currently this isn't possible. For our application we had to make a very nasty hack to our SCSI driver.

What we would have liked to do would have been something like:

Get physical address from libpci

mmap region to user process

open a rawio device

read()/write() the rawio device to-from the mmap()'d buffers to effect DMA transfers directly to the PCI device.

In our case (a streaming media server) we have no need whatsoever to bring the disk buffers into main memory.

Programmed transfers from main memory to PCI memory are expensive.

Re:Amazing considering no education above high sch

[Anonymous Coward]

Maybe it was a professor named Armstrong. Googling on "A degree will get you an interview" (with the quotes), produces one result: a pdf slide show titled "Study Skills and Analytical Thinking".

Slide 12 says "Is it critical to learn in university? 1) A good question. Most students really just want a degree not an education. 2) Remember that a degree will get you an interview and maybe a job but what you learn determines whether you are successful."

Of course, being something of a autodidact myself - I have degrees, but learned much more on my own - I have to say "duh!". But certainly, as I've moved around and met more people, I've known plenty who seem to judge ideas more by the credentials of the speaker than the contents of the speach.

Re:Nice to see some recognition for kernel devs.

[tanguyr]

The subject of this interview works for, and is paid by, SUSE - which belongs to Novell: When I started working for SUSE I was still attending University but over time I got more interested about the work I was doing on linux, plus I could make a living thanks to linux while I didn't get any money by studying more years at University ;).

Your pov is a pretty common misconception: big business profiting from the work of hundreds of thousands of idealistic but naive developpers. The truth of the matter is that big business wants - needs - enterprise features in open source software, and you're going to get there a heck of a lot faster by paying somebody to develop them than you are waiting for some guy to decide that he personally *really needs* to support 32 parallel processors.

Re:Amazing considering no education above high sch

[KnacTheMife]

"It always amazes me when people, without formal education, can accomplish so much." I'm not as amazed because I've always held the opinion that passion and determination are two of the primary ingredients of success and/or accomplishment.

Andrea likes TCPA

[swillden]

I found this bit really interesting (and insightful, actually, more on that below):

Actually in my spare time I had an idea of one revolutionary and ambitious project I can build on top of the trusted computing capable hardware (that project has nothing to do with linux by the way, but for it to run on linux too, linux would need to provide some basic trusted computing support), that's something I wanted to build for a long time but it has never been feasible until they added the trusted computing to the hardware and they filled the gap to make my idea possible, so I'm quite happy about these new hardware features (despite clearly they can be misused for some annoying things too).

I bring it up because this is so contrary to the common opinion on /., which is that TCPA is unabashedly evil and has no utility. Andrea is obviously one very smart guy, and a person who feels the need to have complete control over his machine, but who likes TCPA in spite of the risk of misuse. Contradiction?

The fact is that TCPA *is* an extremely useful and valuable technology for systems that require a high degree of security. It's not clear to me that the average home PC benefits from it, but it's very valuable for cheap, high-performance key management systems and cryptographic accelerators, systems that contain valuable data (like many businessmen's laptops), and systems at critical points in network infrastructure. I'm sure there are other valuable, and non rights-eroding, applications as well.

In my work as a designer and developer of high-security systems, I'm extremely excited about the fact that we can now buy low-end computing equipment that has TCP hardware. It enables so much. The next step is TCP hardware that is tamper-resistant, or even tamper-reactive, but still cheap. For now, really high-security systems still require something better, but TCPA can fill the niche between systems that require serious security and those that can get by with purely software-based security (or no security, which is fine for the majority of desktops and laptops).

To be clear, DRM is a bad idea, in general. The business applications (self-destructing documents, confidential documents that cannot be printed) do have potential utility, but I doubt they're worth the complexity they'll create. And Palladium aka NGSCB aka whatever-it's-called-today is an unquestionably evil notion, focused on removing the ability of people to control their own hardware, in an effort to allow a couple of declining business models to prop one another up.

IMO, what geek activists need to focus on is not killing the development of tools like TCPA, but rather on legal and social means of ensuring our rights.

Tools are not evil. Only users are evil.

Re:Andrea likes TCPA

[swillden]

I don't see what hardware can do that software cannot.

Hardware can keep keys and policies secure. Although general-purpose computing hardware is perfectly capable of performing cryptographic operations and executing security decisions, there is no way to prevent an attacker from getting hold of the keys used or subverting the decisions made.

As an example, suppose you're a bank, and you need to encrypt transactions that you exchange with other banks. These transactions are worth hundreds of millions of dollars per day. If he encryption keys were to be revealed to an attacker, he could bribe someone at the telco that manages the leased data lines and modify the data as it moves across the line. Or, more subtly, if he could just read the data as it flows by, he could learn who is paying money to whom, and then exploit this information.

So, how do you secure those encryption keys? If you put them in any sort of general-purpose computer, an employee with access to the machine can get them out. In fact, if you ever "put them in" anywhere, that implies that at some point they were "out", where people could potentially see them (and copy them!). And with security issues of this sort there is no single individual anywhere that is sufficiently trustworthy... even if you can find someone who you know is perfectly honest, you have no way to be sure someone can't torture the information out of them or otherwise coerce them.

The answer is secure hardware. You generate the keys within secure hardware that is running special, non-modifiable software that will refuse to *ever* divulge those keys, except in a secure fashion (encrypted with key encryption keys that meet certain criteria, or in parts, etc.). In this way it's possible for the keys to be created, managed, used and destroyed without any possibility that anyone can ever get hold of them.

That's one example, there are many others. How to encrypt the data on your laptop hard drive securely, for example.

Canadian schools cover all this

[IncohereD]

Andrea did some cool stuff. How come our high schools don't teach us how to:
# Live off the land

If your high school is in a (sub/)urban area, how useful is this? Or possible?

# Modify our cars

My high school had auto courses (my school was in Southern Ontario, Canada)

# Hack computers

I took 4 or 5 computer courses in high school. Basic application use, programming in Turing and C++, graphics in 3D Studio, large projects.

# Understand personal finance

I believe there were business courses along these lines, I didn't take any, though. I think this was also covered in the general stream of math.

# Write contracts
# Defend ourselves in court

We had grade 12 and 13 law. I'm not sure how much time was given to contracts (I only took half of the grade 12 course), but we even went on a visit to court in both mandatory grade 10 history and optional grade 12 law.

# Defend ourselves physically

In grade 9 gym we covered wrestling. The upper gym courses covered more defense.

# Handle a gun safely

Not really a big deal in Canada. I know a lot of people who learned this in cadets or the reserves, though.

# Change our government for the better

History (Canadian, Ancient, American, 20th Century), Politics, Law...all covered this. Your comment about understanding the past rather than the present is missing the point entirely. But we did always talk about current elections in classes, even in elementary school.

# Think critically
In general, people don't need to know how to calculate the area under a curve. But everyone needs to know how to think critically and not be manipulated.

Doesn't calculating the area under a curve require critical thinking?? Regardless, Calculus wasn't taught till grade 13, and anything past grade 10 math was optional. If you're school taught you anything at all you probably learned to think critically. Didn't you have to write essays? Solve problems? Every single one of our grade 13 classes had an "independent study unit" which we had to do something on our own, requiring critical thinking.

We even covered media bias in our english classes (I didn't take the full english media class, but we did cover it in the required grade 13 english class). We took stories and advertisements from different newspapers and looked at their political bias. Then we watched some Chomsky videos.

As far as I can tell most Americans seem to need to move to Canada to actually get their American values.

My how he has come along in 8 years...

[ZombieEngineer]

A bit of background history for the parport driver. This original started by two guys (Philip Blundell and Tim Waugh) blabbering about treating the parallel port as bus. There was something like ten e-mails for a couple of days while most of the regulars were scratching their heads.

The parallel port ZIP drive maintainer asked them to provide a function prototype of this thing that they were talking about, of them (Phil/Tim) quickly whipped up a rough 50 line C header file which was turned into a working parport driver + parport enabled ZIP and printer driver (removing the infamous "printer-on-fire" message in the process). There were bugs in the parport driver (it was the first pass but you could print and use the ZIP drive together which was something that previously could not be done) but Phil/Tim/Andrea quickly pounced on the driver and straighten it out. Some of the routines for supporting NatSemi and SMC chipsets are there due to the ZIP drive maintainer not being able to use EPP mode on his Dell desktop.

When Andrea first appeared on the parallel port scene he was lacking a little confidence (appologising for his poor english which was far better than my italian :-) but once he got his feet wet with kernel hacking there was nothing stopping him.

Unfortunately I dropped out of the parallel port group around 2000 due to work commitments (linux hacking was one of those phases that I went through).

I congratulate Andrea on where his life has taken him.

ZombieEngineer

Formerly-the-hacker-who-maintained-linux-zip-drive rs.

Re:Surprise

[TyrranzzX]

It's kinda wierd, the old countries like the middle east, it's custom for an entire family to live in the same house generation after generation. I had a teacher who's family lived in a house for 800 years. The boon to that system is that the grandparents and great grandparents get to see their kids and educate them about their childhood every day. Tradition stays and sticks.

Then in Europe, there's different expectations. Either you move out when you marry or you move out when you have to financially. Depends on the culture.

Here in america, it's usually you goto school for 18 years, then college and at college either you stay at home or at a dorm. Then, you move into your own place. There are people who are smart enough to move out when they're 18 or 19. Hell, back in the early 1900's it was customary for kids to stay at home until they married.

The only advantage to the american way is freedom from tradition. I kinda think that it's bad that way myself, considering the fact we've got girls who are pregnant as young as 14 or 15 around here.