Slashdot Mirror
Rik van Riel on Kernels, VMs, and Linux
Andrea Scrimieri writes "
An Interesting
interview with Rik van Riel, the kernel developer, in which he talks
about the Linux's VM, particurarly about his own implementation (which was
recently adopted in Alan Cox's tree). With some controversy towards Linus
Torvalds.
"
>> (which was recently adopted in Alan Cox's tree).
As I understand it, the Rik VM is what we started the 2.4 series with.
The Andrea VM was adopted in 2.4.10 amidst much controvery, and Alan has kept
the Rik VM as a part in the -ac kernels.
25% Funny, 25% Insightful, 25% Informative, 25% Troll
I'm wondering why both VM's can't be included in a distro and allowing the end user to select the one he/she wishes to compile into the kernel? Are the two implementations THAT mutually exclusive?
BTW, this kind of bashing between the high priests of Linux is not good. You can bet your bottom dollar that MS is going to use this conflict to fuel their propaganda machine, saying Linux is a fractious OS run by a bunch of young upstarts who can't agree on anything.
In the end they will lay their freedom at our feet and say to us, Make us your slaves, but feed us. - Fyodor Dostoyevsky
I'm not so sure I agree with him -- if you want to make a dent in the market shares of Solaris and NT/2000/XP you have to keep up with their innovations (Async-I/O, better SMP, etc.). As a user of Linux as our OS of choice for our database and web servers I am feeling a lot of pressure to switch to Solaris because of their better handling of higher-load environments (OLTP databases, web servers, etc.). If Solaris wasn't so damn expensive we'd probably be using SunFire 280's. So I'm pleading to keep up with the big dogs so that I can be reassured that Linux has what it takes (it's handling things fine now but as he said in the article, everyone needs more RAM, CPU, etc.).
Thanks,
--
Matt
I think it was an excellent decision of Linus to remove Rik's VM from the mainline kernel. If not for technical reasons then for political reasons.
Rik's VM obviously needed to be fixed and/or tuned, but apparently lacked the necessary attention from Rik. If Linus had not removed the VM, it would probably have been the situation for a while. Instead we now have TWO VM's which are rather stable and Rik working full speed to make his VM the best.
Competition is good! Which VM will be the best for the future will be determined by Survival Of The Fittest(tm)
It can be argued though, that it was not the right time during 2.4, but Andreas VM seemed to stabilise rather quick with the high level of attention to the problem. Sometimes it takes drastic measures to get results...
I have a lot of respect for Rik van Riel, but I think that Linus made a good decision to "cut bait" on his VM implementation for 2.4.
It was not that Rik's ideas were bad, it was just that their complexity and implementation were going to take too long - they should have been hashed out in 2.3 instead of 2.4.10.
I'm looking forward to having Rik prove his reverse mapping technology implementation in 2.5.
May the best ideas ultimately win, and may the giants of the kernel not take offense at each other. It would be a real shame if something stupid like Linus' lossy source code control system put off Rik so much the Linux community at large lost his wonderful contributions.
Here's to hoping that Linus gets more sensitive in some cases, and that Rik gets less sensitive in some cases.
"Provided by the management for your protection."
Yea, but then they'd have to call it "Coxux" and they'd never get that past the censors.
---
Darn right I'm not signing my name to this post.
Yes, such a thing would be appreciated. It's all very well developing linux in order to "improve itself" but one can take the "stuff the users" approach too far.
The problem is that there isn't a decent multi-patch versioning system out there: how would you tell CVS you wanted to store versions of files pertaining to 2.5.2-mjc and 2.4.13-ac1 and 2.4.18 and then a set of files for Rik's VM? Then how on earth would you pull out the set of files that constitutes a `linus+mjc' tree, or a `linus+ac' tree, from what you've stored?
~Tim
--
Rushing on down to the circle of the turn
Open Source's biggest PR dilema is this sort of argument.
Make no mistake, every company has developers that do this. There's two differences in the Open Source world: 1) you can't just fire an Open Source developer who won't "play ball" with management's edict 2) it's usually public.
These are actually both really good things. The fact that you can't silence someone leads to repeated analysis of a problem. OSS' biggest benefit is that it brings massive peer review to bare not just on the code, but on the process.
The fact that it's public feeds into that, and is equally good.
The problem is PR. The Linux kernel is starting to look like anarchy to non-developers. I suggest that the process works, so we should all take a deep breath and leave it be. However, we all need to take the front lines on PR. Spin is all-important. This is not a "spat" or a "fight", this is "parallel development" and "peer review". The joy of this kind of spin is that, unlike most spin, it's TRUE! This guy is pissed at Linus. Linus has dumped his code. Yet, the two of them keep working hard to meet their customers' demands and producing what they feel is the best possible product.
Please, don't foster the idea that we're a bunch of anarchists producing code that's any less functional than the rest of industry, because quite the opposite is true.
I strongly feel that honesty wins in the end, because people aren't stupid. No one believes that IBM or Microsoft is one happy camp singing "we are the world."
It's great there is a lot of attention on the VM and intense effort to make it better. I have no doubt linux and Rik are professionals and have no problems putting politics aside to get the job done. That is after all part of being a professional. Rik makes some good argument and given enough time and money he'll build the VM of his design. Will it matter 10 years from now? Most likely not. Development will continue and linux will get better. Butting heads is part of the fun, because without conflict people tend to stagnate.
You really cannot expect these people to read all their email all the time, so patches and bugfixes get lost and may need to be resent various times before they get noticed.
Add to that the fact that many of the people writing these patches are also extremely busy and may not get around to resending the patch all the time (I know I don't).
The solution here would be to have the patch re-sent automatically as long as it still works ok with the latest kernel version ... this can all be checked automatically.
cvs co -r 2.5.2
:)
# patch mjc-1
cvs tag -b 2.5.2-mjc
cvs tag mjc-1
cvs commit
# elsewhere/when
cvs co -r 2.5.2-mjc
# patch mjc-2
cvs tag mjc-2
cvs commit
cvs co -r 2.4.13
#patch ac1
cvs tag -b 2.4.13-ac
cvs tag ac-1
cvs commit
# elsewhere/when
cvs co -r 2.4.13-ac
# patch 2.4.13-ac2
cvs tag ac-2
cvs commit
# assuming that Rik's VM patches are independent
cvs co 2.5.2
cvs co rvr-VM
# patch rvr-VM
# or, maintain Rik's VM patches as their own
# files:
# cvs co rvr-VM
# cvs update # forces merge
cvs tag -b 2.5.2-rvr-VM
cvs tag rvr-VM-1
cvs commit
# elsewhere
cvs co 2.5.2-rvr-VM
Why wouldn't something like this work? You could even wrap everything up in a nice GUI if you wanted to.
-_Quinn
Reality Maintenance Group, Silver City Construction Co., Ltd.
An honest environment -- such as fostered by "free" software -- is both good and bad. On one hand, I (as a programmer) am comforted to read the kernel mailing list and other resources that let me know exactly what is happening with my tools. I don't need to wonder what's happening with "free" software -- and this is more comforting to an engineer like myself than is the closed-door, silence-is-golden, hide-the-bugs policy of a Microsoft.
On the other hand: Show this interview to an MIS manager who need 24/7/365 reliability, and she is going to be very nervous about deploying a Linux-based solution. You can talk until you're blue in the face about reliable distros and the open road to sofwtare quality -- what the MIS/corporate person sees is chaos and feels a lack of COMFORT .
"Out of sight, out of mind" is a philosophy many people adhere to, especially when dealing with complex issues they can not or do not want to grasp. From waste storage in Nevada to the the war in Afghanistan, most people lack the time and initiative to understand what is really happening; they go on appearances and marketing, and ignore complex and disturbing facts.
Technology is no different. The MIS manager doesn't want to hear about VM conflicts or file system bugs or different kernels -- such issues are beyond their capability and desire to understand. Buying Microsoft is (or was, until recently) comforting, because no one ever saw the internal debates and code battles and what-not that any development team expresses. Even recent security disclosures about WinXP are unlikely to shake the faithful -- but those same people will run in fear from the blunt honesty of Linux.
Ignorance may be bliss, but it can also get you killed. I know people whose lives depend on cars, but they have no knowledge of how to check the oil. Most MIS managers simply want to drive software; if it looks good (like a Jeep Liberty), they don't pay attention to whether it is safe (the Liberty performs poorly on crash tests).
I doubt, however, we're going to change human nature -- and I'd rather have spirited debate and even some nasty contention if it means that people are striving to make Linux the "best" it can be.
All about me
Rik is an extremely bright (and likeable) guy, but his adherence to the OOM killer concept is disappointing. I've seen a lot of dumb ideas gain currency in the computing community or some part of it; OOM killer is the dumbest. If your process was allowed to exist in the first place, it should not be killed by the VM system. The worst that should happen is that it gets suspended with all of its pages taken away. If that doesn't free up any memory then neither would killing it (modulo some metadata - read on). If there are other processes waiting for the one that's suspended, then eventually they'll go to sleep, their pages will be released, and the suspended process will wake up - which won't happen if you killed it. There are only two differences between the two approaches:
The usual whine from OOM-killer advocates is that you can still get into a situation where all of that retained metadata clogs up the system and essential system functions can't allocate pages. However, that's preventable too. All you need to do is preallocate a special pool of memory that's only available for use by those essential system processes - either individually or collectively. The size of that pool and the exact details of how it gets allocated (e.g. which processes are considered essential) could be treated as site-specific tuning parameters. The same idea can then be further generalized to allow definition of multiple private pools, creating a semi-hard barrier between different sets of tasks running on the system (if you want one; the default pool is still there otherwise). This actually fits in very nicely with other things like processor affinity and NUMA-friendly VM, which I know because I once worked on a kernel that had all of these features.
In short, there's no need for the OOM killer. Plenty of systems, many of which handle extreme VM load much better than Linux, have been implemented without such a crock. Rik contends that a lot of people make suggestions without actually understanding the problem, and he's right, but I also submit that sometimes he also rejects suggestions from people who do know what they're talking about. This row has been hoed before, and Rik's smart enough that he should know to avoid the NIH syndrome that afflicts so many of the other Linux kernel heavyweights.
Slashdot - News for Herds. Stuff that Splatters.
It started out with Rik's VM in the kernel, since it was a promising new development. However, once it was in Linus's kernel, the fact that Rik's development style was not compatible with Linus's source control style because an issue, because the VM wasn't getting updated in Linus's tree.
So Linus switches to the other VM, which is based more on the original. This means that Rik can do his development without dealing with Linus and the Linus tree can have an up-to-date VM. When Rik's is to the point where he's really happy with it and he doesn't think he'll have to make a lot of patches (and it does all the things he wants), it will probably go back in.
Since then, Rik and Linus have figured out (hopefully) how their interaction failed to work, and what Rik has to say along with his patches to make Linus know they're worth looking at. It turns out that it is possible to automate this process, such that a script will send the patches when appropriate, with the right assurances of freshness (having actually tested them, of course).
Linus wants to be able to ignore any patch that isn't for the part he's thinking about at the time (e.g., non-block-i/o patches around the beginning of 2.5). When it becomes interesting again, however, the original patch may not be right any more. Having not looked at the patch at the time when it was sent, Linus can't determine whether it is still good, since the author may have found bugs, and he doesn't know exactly what the patch was supposed to do. He wants the author to make any updates needed and resend it. It may be, of course, that the patch doesn't need to be changed, and the author doesn't have a new and better patch, but Linus can't tell unless the author sends it again with a note that it's still good.
So Rik's patchbot will test whether the patch still applies and still works, and has not been replaced by a new version, and then will send it again until Linus actually looks at it. This seems to me like a good plan, since it doesn't require Linus to test everyone's old patches and have a complicated mail system. And Linus won't accidentally apply the wrong version of a patch or be unable to find a patch.
Somebody has to speak up against Linus. Linus is not a god. The man makes mistakes. And over the last view years it becomes increasingly a problem that "Linus doesn't scale".
Linus however continues to develop the kernel pretty much the same way he started doing it ten years ago. And not many people think that's a problem. Rik does (AFAIK). And I tend to agree with Rik: the current system just isn't working very well. It's not very bad, but it certainly isn't optimal, IMHO.
However, remaining silent doesn't solve the problem. Somebody has to speak up.
This is your sig. There are thousands more, but this one is yours.
A VM is basicly a small thing: a list of pages, every page has a set of properties and an interface on top of that to get things done with the pages (claim/free/mark dirty etc). I wrote one on an MSX2 in 1986 for having 256KB roms in 128K ram + 128K vidram (and 32K disk :)). Of course, modern OS-es need a VM that can take decisions, is scalable on different hardware, and can handle the requests fast.
A lot of research has been done on virtual memory and the managercode for this type of memory. Also a lot of different types of VM's are implemented in different OS-es, all with pro's and con's in different situations. It's therefor not hard to dig in and get the knowledge you need.
F.e.: the rmap stuff is a nobrainer. If you let the VM handle every request to share/allocate a mempage, that VM can keep a set of pid's per page. IIRC NT's VM (VMM32) does this. That the current VM in Linux doesn't already have this feature is beyond me.
Never underestimate the relief of true separation of Religion and State.
The problem is that there isn't a decent multi-patch versioning system out there
Uh, yes there are. Perforce, aide-de-camp, bitkeeper, and others all do this just fine. I haven't used squeak much, but I think this is also how the built-in version control in their smalltalk image works as well. Every change management system that uses changesets works pretty much exactly this way.
CVS basically sucks, which is why some people are trying to replace it. It only gets used because it is popular and free, not because it is technically superior. The only thing it is better than is RCS/SCCS. Every other possible solution is no worse, and usually much better, than CVS.
I don't want to get marked down or flamed for trolling or anything, because I am not:
Thanks for posting this quote. The more I read in to this stuff, it often times seems as though Linux has an immature attitude, and often acts like a baby. Ignoring patches because you have a disagreement from someone is just plain immature. I can see how frustrated I would get working with Linus. He still acts like it is his little baby project, and that is just not the case anymore. Thosands of developers are working on it, and this kind of attitude by Linus would just turn people off from the project.
These kind of snide remarks by Linus are not needed, and if I was Rik, I would tell Linus to fuck off and put my talent to use somewhere else. Linus, act a little more mature.
Moon Macrosystems. Sun's biggest competitor.
RTI - Read The Interview.
...Rik's repeated attacks on Linus will certainly not move the operating system forward.
... Yes, though I guess I have to add that I have a lot of respect for Linus. He is a very user unfriendly source management system, but he is also very honest about it.
Rik was interviewed in order to get insight into how he thinks/sees things, no? So if he doesn't like the way Linus does things, is he not at liberty to say so? (also, see quote below about still having respect for Linus in spite of their disagreements/conflicts)
Rik's behavior really isn't funny... It speaks volumes about Rik's emotional maturity or more accurately his lack thereof.
Rik Say:
With Linus out of the way, I can make a good VM. I no longer have to worry about what Linus likes or doesn't like.
I don't quite think that qualifies as immature - granted, there is a lot of conflict going on, but they still have respect for each other, even if Rik doesn't like to work with him, and there's not really anything showstopping about it. The VM situation wasn't pretty, but it's being resolved.
There's a lot of back-and-forth discussion, not only on the VM, but on the feature (un)freeze of 2.4/2.5, and on how Linus is a lousy patch control system. But maybe that's not the most important thing here.
Way back when, the purpose of a development kernel was to feed things in to a stable kernel tree. Now part of the problem has to be that Linus started 2.4 way before 2.3.X was ready for it, but it looks like history is repeating itself. 2.4 isn't all that stable, even now, but Linux is happily accepting lots of new goodies to play with in 2.5.
Something is not working right here. Is Linus less demanding of quality now, since he's willing for somebody else to come in and fix up the allegedly stable kernel tree? Or is he accepting too many things to allow a development tree to stabilize?
I suspect it's a combination of too much stuff and too big a kernel. Instead of the heady days of 2-3 kernels per week in the development tree, and the stable tree gets another kernel every week or two, now we have a development kernel every week or two and a stable patch every month or three. And the kernel size is 10x bigger than in the 1.0 days.
Look how long it took the USB stuff to filter through the development into the stable tree.
It seems obvious the Linus Linux development process is not scaling. I'm not sure what the answer will turn out to be, but it may be some combination of the following:
(1) More "boutique" kernels like Alan Cox's ac series, feeding into the "stable development" kernels that Linus has been generating.
(2) More formal check-in methods, a la CVS commit. This may take some developer training in how to use CVS -- does anyone want to offer Linus a course and set up a server for him? I bet he'd take a complementary Geek Cruise!
(3) Some kind of more rigorous control in the stable kernel tree. I suppose you could say Redhat and SuSE are doing this informally now; if they start coordinating their efforts, and get IBM involved, the kernel will be incredibly stable. And even more incredibly slow to update.
(4) More beta testers to crack the newer kernels. This is going to get harder, as more of us need to get work done on our Linux boxes. It used to be a hassle when Linux crashed; now it's not acceptable any more!
(5) Better ways for these users to track down problems and report bugs. This last week I heard myself say, "Try rebooting your Linux box and see if the problem goes away." I just don't have the time, energy, knowledge, and skills to deal with lusers' "I've got a problem" whines any more.
(6) Is the quality of kernel patches too low? Do we need to develop some regression tests for the kernel, which a patch would have to pass before it would be accepted? (And how do you do a regression test program of this magnitude without Microsoft's beta testers, AKA customers?)
Anybody want to contribute more ideas to the list? We can spam Linus with them until he agrees!