after my laptop I tried to get 2.6.24 running on my desktop machine; several attempts (Debian kernels, self-compiled ones with different options) all led to the same results: a machine that feels rather unresponsive & where music has sporadic dropouts.

some searching on the intarweb showed that the new CFS scheduler (cf. Documentation/sched-design-CFS.txt) (at least the enabled-by-default CONFIG_FAIR_USER_SCHED flavour) might have latency problems, in itself or in combination with the ondemand CPU frequency scaling governor or with nice'd processes; after turning off all CONFIG_FAIR_* options in .config, recompiling & rebooting the machine seems to behave normally.

maybe I'm missing some great new features but actually I prefer a working desktop computer :)

update 1: approximately half an hour after the original blog entry I had my first "ion sorbet", i.e. a half-frozen ion3 (the applications are still running, I can switch to a tty but I cannot change/close/whatever ion3's frames). at the moment I have no idea what's going (wr)on(g) here ...

update 2: two more "ion sorbets" today yet; let's see if a new kernel (CONFIG_FAIR_USER_SCHED seems to be fixed in 2.6.24.1) & a new ion3 version help.

update 3: still the same problems (& inbetween some glitches with switching from X to a tty and back; I hope I solved them with some kernel config options); ATM I suspect ion3 to be the culprit; when it hangs it eats almost all the CPU, & strace'ing the pid gives only SIGALRM messages.

update 4: maybe I fixed the problem; by changing from ion3's deprecated statusbar_external.lua to the newer statusd_exec.lua (I just wanted to get rid of the error when starting ion3 ...).

One of my favorite radio stations, KEXP, based in Seattle is coming to NYC by parenting with a local station. Not much will change except that we will have the incredible John in the Morning for some of his time and he will, according to their website, scout out local talent:

Does this mean that John in the Morning is leaving Seattle?
No. Starting in June 2008, John will split time between Seattle and New York, broadcasting his show live from both locations throughout the year. While he s in New York, he ll be searching for new bands and discovering new music that The Morning Show listeners will be the first to hear.

The politics and language of judging code. And speaking of peer review, Scott and Samir of Decoding Liberation have posted a blog post about the problems of peer review in the sciences, making some good points and opening what I hope will be an on-going discussion. I have a lot to add from the perspective of the social sciences and humanities, but that will have to wait for another day.

The problem with the Internet is that it never goes to sleep. And if you study the Internet as I do, that means spending a lot of time starting at the screen tagging, copying, pasting, downloading, watching, and assessing. I have taken a step or two back away as I have moved from life as a postdoc to working full time but ever since Anonymous declared war on Scientology last month, I have been spending more time online again (more than I want to at least), tracking and following some of their developments, including spending time on some of their IRC channels, downloading lots and lots of videos as well as all sorts of great flyers. Well today there was an IRL action loosely coordinated by Anonymous to protest the CoS and I have to say, I am simultaneously surprised and unsurprised at what I think was a pretty vigorous response and turnout. I was not too surprised as the attack against and criticism of Scientology is one of the older wars spurred on in part by the net and net ethics and it always seems to crop of from time to time. But I think the depth of smack talking I witnessed on IRC, (often clearly youthful in nature as this example reveals—>, [mizzanon] I will have my full license in approx. 3-4 weeks, as soon as I do, i will be driving around my entire region with tons of
anti-propaganda) led me to believe that they could not pull of this global day of action. But they did and perhaps with the exception of New Zealand (where there are probably better things to do anyway), they seemed to garner a fair bit of participants. It will be interesting to see how long they will continue to generate this type of support. The next day of action will be around the middle of March.

After my shameless beg, a number of people emailed me (thanks -- sorry about the lack of comments) to tell me that I should really be using awesome, and I am. (Other popular suggestions were wmii and dwm.) I'm well happy with it so far, so thanks!

I don't know why this is appears to be a problem but AFAICT there is no particular impediment in talking to delegates at conferences, no matter what the gender.
Maybe it is just the way that such discussions are begun or the tone being used or maybe it is just that my RealLife(tm) world involves a radically different environment.
Various stats claim that IT is massively dominated by male workers.
Many stats show that pharmacy is massively dominated by female staff.
Maybe it is just an indication that a degree in a course other than IT is more of a benefit than would otherwise appear.
As a single guy, the pharmacy world is little different to IT as far as conversations with colleagues is concerned. To me, that is how it should be.
Maybe I'm just too old to be deemed "a problem".

Now internally based on Unicode, with a new Xft-capable font backend, D-Bus integration, built-in EasyPG, XEmbed goodness, etc. And legions of bugs waiting to bite you. Get it from here as usual.

We also have a (completely fabricated) screenshot of the new font backend, for your viewing pleasure:

My Apple Powerbook 5,4 just booted for the first time using GRUB, with no manual intervention, from Apple chime to GDM prompt. This is a great milestone for GRUB2 on powerpc-ieee1275 and OpenFirmware, as until now, multiple problems in the loader would drop you to OF console straight away, although other PowerPC hardware with other firmware implementations did manage to work. Recent fixes by Pavel in the core and some other minor fixes for the userland utils have taken grub2 to the point where it is usable on most PowerPC hardware. On Apple, the only minor issue remaining is grub-probe insisting on (hd) not being a valid device name, so for now I have to trick it into believing it's really (hd1). In parallel, many other GRUB2 improvements haven't stopped hitting CVS in the last months, which have seen how new contributors joined grub-devel and helped GRUB2 get the great momentum it's enjoying right now. Vesa, Robert and Bean have been really active lately, and have fixed long standing issues or written a lot of new code. One of the features GRUB2 acquired recently was image loading for background images. Much more powerful than the implementation in GRUB Legacy, GRUB2 can now read images in multiple formats, can handle up to 24bit colour and render a menu in arbitrary resolutions. The menu can now show UTF-8, and the Debian package will configure a pretty theme that matches the rest of the system if desktop-base is installed:

GRUB2 speaking UTF-8 Catalan Although I'm not sure if GRUB2 is completely up-to-par with GRUB Legacy on i386/amd64, it seems the tricky bits, like video, LVM, RAID and the standard filesystems are supported and working. What GRUB2 needs now, in order to finally replace the aging and upstream-lacking GRUB Legacy you probably have installed, is massive testing. Debian has traditionally been a testbed for GRUB Legacy patches, and is also the platform where GRUB2 is being more widely tested. Having GRUB2 included in lenny's debian-installer would be a great step forward, and by the looks, I think we're well on time to manage this. Replacing GRUB Legacy with GRUB2 is trivial. On PCs, just install the grub-pc package. You'll be offered to keep GRUB Legacy, but with an added menu entry to chainload GRUB2. If you're worried that GRUB2 might fail on your hardware, accept this, and try to load GRUB2 from GRUB. If it works, you then know you can get rid of GRUB Legacy completely and keep GRUB2 in the MBR. On PowerPC-based Macs, you'll have to work around the small issue I mentioned above. Install the grub-ieee1275 package. You also need a very recent powerpc-ibm-utils package, which was just uploaded to unstable. Mount your bootstrap partition, probably /dev/hda2 in /boot/grub, and generate a device.map file with grub-mkdevicemap. Check the contents. If your first device name lacks a drive number such as (hd), it's probably correct, although this will make things fail later. Change it to (hd0) for now. As grub-install relies on grub-probe, you'll have to generate your grub image by hand. Copy all .mod files in /usr/lib/grub/powerpc-ieee1275 to your bootstrap partition, and generate a core.img:

root@powerpc:/boot/grub# grub-mkimage -d . -o /boot/grub/core.img *.mod
root@powerpc:/boot/grub# update-grub
root@powerpc:/boot/grub# nvsetenv boot-device hd:2,core.img

The generated grub.cfg will have references to (hd0,X), which you'll have to correct back to (hd,X) if necessary for your OpenFirmware. After this, you are probably ready to reboot, cross two fingers and get a warm "Welcome to GNU GRUB!" message at boot, which will then be followed by a standard GRUB menu, but on your nice PowerPC box. Unfortunately, the eye candy in the screenshot above isn't available yet in this platform, as it lacks VESA. Does someone in the audience want to contribute a video driver for powerpc? :)

• RISE Security vs ASUS Eee PC
  "Easy to learn, Easy to work, Easy to root."
• Niall's technique for sniffing browser history to customize pages
  I'm still not sure if this is really neat, or really evil.
• Kevin Kelly has written a thoughtful post about how to...
  "When copies are super abundant, they become worthless. When copies are super abundant, stuff which can't be copied becomes scarce and valuable."
• Knee-Deep in the Dead
  More Microhoo drama.
• Mad Driving Skills

Joachim, the key is probably to implement patch support in srcinst and actually implement something nice which I think has nice applications.

It might be nice to combine srcinst with debian-xcontrol to obtain also customizable dependencies (e.g. no arts support).

I guess you'll be the best person to take over srcinst since:

1. John Goerzen doesn't want to/can't continue maintaining it
2. is written in Haskell and you know Haskell
3. source is available for grabs from http://darcs.complete.org/srcinst/

The basic idea is to have another command[1] wrap aptitude/apt-get/srcinst and make sure it does the proper job, depending on the local configuration.


[1] let's say, the debtoo command - srcinst should remain the tool for clear 'from source installs'

Cannot help but laugh and share.
I've been triaging and trying to reproduce some oldish bugs on pkg-perl's packages. Some bugs are no longer there, some have to be forwarded upstream, and so on. Usual tasks, yes.
Until I stumbled with #406227. I just have to laugh and share! Hope nobody feels ashamed - The bug is the result of different people coding maybe under pressure and with quite different mindsets :)
For some reason I fail to understand, the submitter's test case (rot13 implemented over a HTTP proxy) is invoked in the report as ./rot13 2>/dev/null. Of course, when trying to debug a bug report, the first thing to do is not to ignore STDERR. So, off goes the 2>/dev/null. What happens next? 0 gwolf@mosca[2]/tmp$ perl ./rot13 &
[1] 4394
0 gwolf@mosca[3]/tmp$ GET -p http://localhost:8080/ http://www.debian.org/
Can't locate object method "filter" via package "UGGC::Cebkl::ObqlSvygre::fvzcyr=UNFU(0k604160)" (perhaps you forgot to load "UGGC::Cebkl::ObqlSvygre::fvzcyr=UNFU(0k604160)"?) at /usr/share/perl5/HTTP/Proxy/FilterStack.pm line 126.
500 EOF when chunk header expected WTF... Well, at least the program name gives me a clue... Lets try to "decrypt" the error message... gwolf@mosca[4]/tmp$ echo 'UGGC::Cebkl::ObqlSvygre::fvzcyr=UNFU(0k604160)' rot13
HTTP::Proxy::BodyFilter::simple=HASH(0x604160) hrm... How comes the filter is filtering its own code and only then refusing to find itself!? Ok, time to open up the manpage - Remember, I'm only group-maintaining this pacakge, I am not yet at all familiar with it! Ok, so the core of the filter is when the submitter states: my $proxy = new HTTP::Proxy();
$proxy->push_filter(response => new HTTP::Proxy::BodyFilter::simple(sub tr/a-zA-z/n-za-mN-ZA-M/; )); While the manpage states it should be invoked as: my $filter = HTTP::Proxy::BodyFilter::simple->new( sub $ $_[1] =~ s/foo/bar/g; );
$proxy->push_filter( response => $filter ); Of course, once looking at it, the answer is simple: The submitter left out which element to act on in the anonymous function body - The $ $_[1] =~ part. Adding it makes gur svygre jbex nf rkcrpgrq... Err, sorry - makes the filter work as expected. Now, bonus points: For the non-Perlers out ther: How come we get the namespace translated as well? Oh, that's very simple: In Perl, as in Python (and concievably other languages I'm unaware of), the object is passed to any of its methods as the first argument. Functions in Perl receive their arguments via @_ (read: the default array). And, of course, the tr (regex-based transliteration) takes by default the first thing it sees - the object itself. And what happens when you apply a (string-oriented) regex to an object? Of course, it gets stringified - which, by default, in Perl means converting it to the closest possible description: "a hash reference blessed as an object of the class such-and-such at this memory location". That string gets worked on, and we get UGGC::Cebkl::ObqlSvygre::fvzcyr=UNFU(0k604160). This proxy does not die on very-very-short web pages, where the whole content fits on one iteration of the code (although it does not work correctly - the text remains unaltered, of course, as it was not worked on), but if the request spans several chunks, the second time the filter is called, it will be... just gibberish. Oh, and what about the extra $ (...) around $_[1]? Oh, simple: The string is passed as a scalar reference, so it can be modified in place. Yes, it's the Perl way of pass-by-reference instead of pass-by-value (the default behaviour): Of course the parameter is only passed as a value. Only that the value is incidentally a reference - but who cares? ;-) Anyway... Many oddities. I would implement the module in a completely different way, and it looks quite backwardish in my book. But then again, TIMTOWTDI.

Few days ago, I was looking at my gmail mails. As usual, there was a couple of ads on the right hand side of the screen (they never really bothered me before). But that day, there was something quite different: an advertisement for Microsoft under the term "Embedded Debian" ! Well, I can understand that Microsoft wants to advertise for the keywords "Embedded Debian". After all, if they do provide good embedded OS services, that is only fair. But to do so with a title "Embedded Debian" does not really qualify as fair use of the Debian trademark ! The trouble is that this is something hard to prove, as it appears seldom and apparently randomly. When I google for the terms Embedded Debian, I come up around every second time with nearly the same ad, but taking to www.microsoftembedded-de2.com. This domain does not appear to be connected to Microsoft anyhow (see its whois entry). Some other people tried and never saw the ad, as I did when I looked from another computer (a Windows one, funnily). The www.windowsembeddedsolutions.com link which is depicted above did only show up in my gmail account so far. However, that domain is owned by Microsoft, so they can't deny this comes from them. Why, are we scaring them so much ??

Fun round-up of web calendars on BBC Click this weekend. Google, Kiko and 30boxes get mentions. I'm mainly using my new phone but reviewer Rob was quite scathing about phone calendars and didn't bother to say how well the different sites integrate with handhelds. Feedback by email was invited... Are there good interoperable web calendars? Any as free software? Talking of BBC being on a different calendar to everyone else, BBC belatedly announces that iPlayer-for-Mac will come at some point in 2008. Sadly, the DG is still repeating the "only for Sony TVs"-like single-platform explanation and still not mentioning a (DRM-free?) GNU/Linux service.

I am 10 000 metres above sea level, on my way from Melbourne to Wellington. I am looking back at a very enjoyable week of conferencing, with LCA 2008 ending yesterday, followed by today s Open Day. The purpose of this final day is to invite the general public to learn about open-source. Individual projects present their work at booths and field questions by bypassers. Jacinta Richardson and the other organisers and helpers of the Open Day have done an amazing job. The place was buzzing and the selection of projects broad and interesting, even to me. Two talks and a series of lightning talks, as well as catered food for everyone rounded it off. I will try to have this event in mind as we organise a similar event in Buenos Aires after the forthcoming DebConf8 in Argentina. I had a splendid time at the conference and probably can t thank Donna Benjamin and her army of mignons enough for organising it. Compared to the other open-source conferences I previously attended, this one was the most professional. Good job, everyone!!! Here are the highlights:

• I opened the Debian mini-conf with the state of the project address.
• My second presentation on that day was on my packaging workflow (slides), including a live demonstration, which was received well even though I should have prepared it.
• At the keysigning, I (again) used my Transnational Republic identity card and reported on it in another post. I was pleased by the level of attention this got, and how many people approached me during the week to discuss the issue.
• I really enjoyed the talk by Adaora Oniya on communication challenges in Tuesday s LinuxChix mini-conf. Among other interesting points, she listed traits that are typical to males and females and stated how none is 100% on either side. Glancing over the adjectives, I started to wonder about my feminine nature.
• Later that afternoon, a bunch of us sat down to find similarities in packaging workflows of the Debian and Fedora projects.
• Bruce Schneier s keynote titled Reconceptualising Security didn t teach me anything new, but that was to be expected since I am a regular reader of his Cryptogram newsletter. He s a good speaker, but I think his talk would have benefitted from a few, simple slides to illustrate some of his points.
• A few of us met on Wednesday morning for a few games of the Asian board game Go. Paul Wayper and I played two non-competitive games, in which we talked about each others moves and strategies. I think I learnt quite a bit during that session.
• Jeff Waugh s mini-conf talk titled Disintermediating Distributions as well as Stormy Peters keynote Would you do it again for free? , as well as a few other sets of slides, taught me that I really need to work on my own presentation style and get rid of bullet points once and for all.
• The Penguin dinner, traditionally a catered event for conference attendees, took place at the Melbourne Night Market this year. Everyone got food tickets and went off to the dozens of stalls to get food and drinks. This was an excellent idea to deal with 600 people.
• I was very glad to have attended Jason White s presentation on braille and text-to-speech interfaces on Thursday.
• My presentation on [netconf](http://netconf.alioth.debian.org] was well attended and the audience had plenty of good questions. I was able to answer most of them with reference to the design of the netconf core, which is a good omen. The idea of on-demand dialing went into my notes. I screwed up once when I went off on a tangent while answering a question about firewalling and then forgot the original question, but my listeners showed mercy.
• Luke Kanies presentation of Puppet convinced me to allocate more time to this system administration tool. He did a very good job fielding questions and managing time, but he should try to slow down a bit.

Apart from the busy programme, I particularly enjoyed the hallway track , which is usually the reason why I attend these events. I really ought to practice remembering names and faces a bit better. I am not paying enough tribute to this week with this report, but I shall conclude it regardless. Unless something very unexpected comes up, I will attend next year s LCA in Tasmania. NP: Porcupine Tree: Futile

Those who program C are used to seeing function / subroutine definitions in header files. Not so in Fortran, there people use the C pre-processor to put Header files in subroutine definitions. Don't try this at home, folks. Again, names changed to protect the afflicted ...

subroutine  funmame( param1,
	      param2, param3,
#ifdef FOO32
	     foo, bar,
#endif
	      param7, param8, 
#include "basarg.h"
	     )

Remarkably, basarg.h included:

	C Comment.
	param5, param6,
#ifdef SOMETHING
        param10, param20,
#endif

There were in fact dozens of parameters in basarg.h.. And even a higher.h that included basarg.h but had more parameters ... Tags programming, bizarre, work, fortran

Jaldhar Vyas points to the rogues gallery of oil producing states and argues that Neoconservatism is good, well at least it gets rid of the dictators such as Saddam Hussein while it grabs the oil. As he ightly points out, we are researching new alternative energy sources, and just as importantly (or even more importantly) energy conservation. But we have been doing this for decades. Sooner or later we have to realize that the "Oil Crisis" isn't due to a shortage of energy sources: its due to deliberate politics to make us dependent on oil. Peak Oil is "only" a crisis for us consumers, dependent on oil to drive our cars, heat our houses, and supply our food. For oil companies, it means (1) they are shipping oil at a greater rate than ever before, (2) at higher demand than ever before, and (3) in circumstances where pesky politics like OPEC can't interfere and try and moderate the price down (I'm not arguing OPEC or cartels are good; its just that its long-term 'maximize the benefit for the oil producing nations' strategy interferes with oil companies shorter-term profit maximization). For the Oil Companies, its a good thing. In this context we have to see our current dependence on oil as the result of deliberate policy by the oil companies. The small filip that USanian governments have done towards alternative energy, such as funding Hydrogen fuel-cell research, nuclear power, etc. are so long-term, and so small in scale, that they have to be seen as deliberately not solving the problem. Thinking that research into alternative sources is good is gpoing to save us (disclaimer: as a scientist I think research is a good thing :-)) is missing the real problem. Its politics: the oil companies have simply been better at politics than the strugging alternatives; we need to recognize this, and fight for alternatives. When oil prices reach $100/barrel, and we struggle to pay rising food and energy bills, its because other have deliberately caused the crisis. Oil has often been called a Curse because of the effects 'Oil Wealth' have caused so many countries. They are not co-incidentally dictatorships; Saddam Hussein remained in power in Iraq for so long because it benefitted the West (see the weapons supplies to Iraq in the so-called "Iran-Contra" affair, for example). Ditto Saudi Arabia, remaining 'friends' of the USA despite its ties to 911. Overthrowing Saddam Hussein was about consolidating even more power in the Gulf than America had previously, not democracy. What was planned for Iraq was democracy in name only: all real power was already neutralised, with decisions to US benefit cemented in place before the parliament met: non-negotiable contracts already signed giving US companies control. This is straight imperialism, and cannot be considered democracy or a good thing. Tags environment, oil, politics

So a few folks left some interesting comments in response to my link to the article on the rise of the alpha-girl based on the research of Harvard psychologist, Dan Kindlon. My response to both Joe and Karl is that it is worthy to lower the barriers to entry not because girls will change the cultural ecology of geekdom in positive ways (though they may) or because geekdom is inherently “omg totally awesome!!!!!!!!!!!!!!!!!!!!” but for a much more mundane reason: it may allow girls to follow interests and eventually discover a passion. I think it is a great idea if more women were involved in geekdom and open source development not because it is inherently great but because I am sure there a lot of women out there who would enjoy it (and by extension, I am sure a lot who would not). By highlighting this article I am not advocating some forced hacker-female-labor-migration-policy but pointing to the fact that social conditions outside of geekdom play a role in ensuring more participation among women. But let me explain a few things before I argue my position a little further. First a little clarification, especially now that I have read the article. By alpha the author does not mean dominant, aggressive, and totally jerky, which is what might immediately come to mind when you read/hear alpha (especially if one knows any alpha-geeks or alpha-lawyers for that matter, who are much worse, imho). And I certainly don’t want any more alpha-anything (alpha-women, alpha-dogs, alpha-cats, alpha-hamsters ) given our world is already chock-full of alpha-jerks. By using the term, the author basically means confidence, which apparently is on the rise among girls, leaving them much more likely to be go-getters, undaunted by explicit and implicit barriers and obstacles and thus more likely then ever to enter arenas that have been thought of as traditionally male or in fact totally dominated by males. If this trend is indeed in place, I think it is great and I hope that this eventually translates into more girls/women populating traditionally male areas whether it is neurology, movie directing, hacking, or surfing. What I appreciate about the article is that his research can help us question the idea that women are naturally averse to competition due to their estrogenic hormones while boys, brimming with testosterone, just love it and exude it. I never identified with that conventional well-worn script, in part because I guess I am fairly confident and somewhat competitive. Being in academia for the last 10 years, I have witnessed a lot of really confident women that have helped inspire me and keep me going, especially when times got rough. What is fascinating about the article (and by extension probably his book) is that gender parity and equity and changes in the psychological makeup of women have not changed overnight but have taken a much longer time to settle in place. We are only now bearing the fruits of structural and educational changes first instituted decades ago and that tackled some serious forms of discrimination. As a result, we are seeing girls and women donning a deeper confidence that may help them participate with more success in the arenas they want to whether it be sports or in the workplaces (though there are certainly still major barriers and issues, which are addressed in the article). So why would someone want to participate in the world of free software and hacking in particular? My first answer is there may be any number of reasons why anyone, female or male or transgender or whatever, may want to do so for the love of technology, to feel an intense belonging to a community who share your passions and who work together to create something with a lot of value, to enjoy the challenge of learning, to spread freedom to every corner of the earth, etc… It is not that geekdom is inherently fun and exciting but that it will be interesting to some slice of the population men or women (or other) and what is the harm in lowering the barriers to entry, especially if it brings enjoyment and frankly a lot of economic security too? There are a number of girls/women who already find it worthwhile and I am sure many more who would. And the point is not to create some policy to make sure that we shuttle women into hackerdom but as a society we should equip them with the necessary psychological tools so that if they think this is worthwhile, and discover that they love it, etc, they will dare to venture in there and more importantly, stay if they want to. In terms of Karls point that a lot of people are professional geeks in part because they d have trouble being anything else; like being gay, it s not a choice, I think that is far too narrow of an assessment, not to mention an outdated caricature of geeks. While there is certainly a class of socially awkward geeks (if that is what you mean??), I would say they are in the minority though they may certainly stand out precisely because they are the odd ducks and because the stereotype is so entrenched. Many geeks I have met, while they may pretty darn focused on geekdom, also have full and rich lives/personalities that cannot be easily collapsed into one immutable personality type. Yes they may be obsessed with tech but aren’t doctors, academics, musicians, lawyers sort of fanatical too? I spend like my whole week working on academic stuff (part of necessity, part out of love). And at least hacking has way better conferences and economic perks, which may help explain why people stay Finally, I think you assume a little to strongly that we do what we do because we have a pre-formed existing desired to do so. While I think this is the case for many things (and I knew the minute I learned about anthropology, I wanted to be one, which was odd but it proved to be correct), I think desire is also formed as much through experience and hence the importance of exposure to different worlds and experiences. I know that there are many things I could have never imagined I would have loved–karoake and sailing are two things that come to mind—until I tried them about both took some degree of courage. In the later case, it took a lot of guts to move onto some ship at the age of 18 instead of going to college and I am so glad I did. This is an instance where confidence and an initial curiosity led to discovering a love and passion I never knew I had. So if desire can be formed and not just expressed, I think it is key to make sure people have all sorts of opportunities to cultivate the passions they never knew they had.

Using the “ulimit” controls over process resource use it is possible to limit RAM for processes and to limit the number of processes per UID. The problem is that this often is only good for accidental problems not dealing with malicious acts. For a multi-user machine each user needs to be allowed to have two processes to be able to do anything (IE the shell and a command that they execute). A more practical limit is five processes for a single shell session (one or two background jobs, a foreground job where one process pipes data to another, and the shell). But even five processes is rather small (a single Unix pipe can have more than that). A shell server probably needs a limit of 20 processes per user if each user will have the possibility of running multiple logins. For running the occasional memory intensive process such as GCC the per-process memory limit needs to be at least 20M, if the user was to compile big C++ programs then 100M may be needed (I’ve seen a G++ process use more than 90M of memory when compiling a KDE source file). This means that a single user who can launch 20 processes which can each use 20M of memory could use 400M of memory, if they have each process write to a pages in a random order then 400M of RAM would be essentially occupied by that user. If a shell server had 512M of RAM (which until quite recently was considered a lot of memory - the first multi-user Linux machine I ran on the net had 4M of RAM) then 400M of that could be consumed by a single hostile user. Leaving 100M for the real users might make the machine unusable. Note that the “hostile user” category also encompasses someone who gets fooled by the “here’s a cool program you should run” trick (which is common in universities). I put my first SE Linux Play Machine [1] on the net in the middle of 2002 and immediately faced problems with DOS attacks. I think that the machine had 128M of RAM and because the concept was new (and SE Linux itself was new and mysterious) many people wanted to login. Having 20 shell users logged in at one time was not uncommon, so a limit of 50 processes for users was minimal. Given that GCC was a necessary part of the service (users wanted to compile their own programs to test various aspects of SE Linux) the memory limit per process had to be high. The point of the Play Machine was to demonstrate that “root” was restricted by SE Linux such that even if all Unix access control methods failed then SE Linux would still control access (with the caveat that a kernel bug still makes you lose). So as all users logged into the same account (root) the process limit had to be adequate to handle all their needs, 50 processes was effectively the bare minimum. 50 processes with 5M of memory each is more than enough to cause a machine with 128M of RAM to swap to death. One thing to note is that root owned system processes count towards the ulimit for user processes as SE Linux does not have any resource usage controls. The aim of the SE Linux project is access control not protection against covert channels [2]. This makes it a little harder to restrict things as the number of processes run by daemons such as Postfix varies a little over time so the limits have to be a little higher to compensate, while Postfix is run with no limits the processes that it creates apply to the global limit when determining whether user processes can call fork(). So it was essentially impossible to implement any resource limits on my Play Machine that would prevent a DOS. I changed the MOTD (message of the day - displayed at login time) to inform people that a DOS attack is the wrong thing to do. I implemented some resource limits but didn’t seriously expect them to help much (the machine was DOSed daily). Recently I had a user of my Play Machine accidentally DOS it and ask whether I should install any resource limits. After considering the issue I realised that I can actually do so in a useful manner nowadays. My latest Play Machine is a Xen DomU which I have now assigned 300M of RAM, I have configured the limit for root processes to be 45, as the system and my login comprise about 30 processes that leaves 15 for unprivileged (user_r) logins. Of recent times my Play Machine hasn’t been getting a lot of interest, having two people logged in at the same time is unusual so 15 processes should be plenty. Each process is limited to 20M of memory so overflowing the 300M of RAM should take a moderate amount of effort. Recently I intentionally have not used swap space on that machine to save on noise when there’s a DOS attack (on the assumption that the DOS attack would succeed regardless of the amount of swap). Now that I have put resource limits in place I have installed 400M of swap space. A hostile user can easily prevent other unprivileged users from logging in by keeping enough long-running processes active - but they could achieve the same goal by having a program kill users shells as soon as they login (which a few people did in the early days). But it should not be trivial for them to prevent me from logging in via a simple memory or process DOS attack. Update: It was email discussion with Cam McKenzie that prompted this blog psot.

• [1] http://www.coker.com.au/selinux/play.html


• [2] http://en.wikipedia.org/wiki/Covert_channels

Share This

Here's a question that has bugged me for a while, and to which I've never found a satisfactory answer. How do people handle mailing list moderation when they have spam filters active? For instance, I get a lot of spam in the moderation queues for the Apache mailing lists I moderate. Sometimes gmail flags it as spam, and I can't quite say it's wrong... I mean the body of the email that's help up for moderation is often full of spam words. But the moderation message itself isn't really spam. I'm afraid of confusing the filter if I let it through. I guess this is where an open source system might be better, as you could simply put a rule in saying "if it's a moderation email, don't even bother looking at the keywords". Gmail's awfully convenient though... anyone else out there with the same problem?

I am not quite up to listening to this myself, but this interview with author Terry Pratchet about hisearly on-set of alzheimer’s disease seems interesting and it is good to see people talk about it frankly.