From the article at Kerneltrap: “This week’s profile interview is with Robert Love who currently maintains the preemptible kernel patches, among other things. He’s been using Linux now for about 7 years, with numerous contributions in the current kernel. All of this is best described in his own words.”
Linux Kernel Hacker Interview: Robert Love
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Read the last paragraph of the interview as well, where the editor used a patch that Robert gave him!
Nice to see Linux to make great advancements too, but it is weird that such a patch is not already in the kernel as default. And if I reffer to the Linus interview we had the other day who said that his kernel is pretty much as preemptive as it would like it to be without many important giant locks around it, makes me wonder. This patch, I think proves that lots can be done to make Linux more responsive.
Great to see that there are such patches (that bring responsiveness to a near BeOS quality), but the big question is when these patches are going to be merged to the main source tree as default (along with the sound latency patches as well)… Can’t wait!
<BR>
See you all in 3 days, back to my house move now.
Linus isn’t including the patch because it is a hack (albeit a good one). It may make things quicker in the short-run (it works well for me
, but in the long run it is just another patch for an underlying problem. It’d be better to find what the problem is and then fix it properly than to simply cover it up. If the patch is included in the official kernel, nobody would be bothered to fix the underlying problem, which can make things very difficult in the future.
Linus is doing well at trying to write a _good_ kernel, not just another product that is rushed out the door. When he said that Linux is as pretty much preemptive as he wants it to be, he was referring to the SMP implementation.
Linux is very powerful, extensible and scalable. Quick hacks can ruin this good design. Things need to be carefully thought-out and planned so that they stay in order in the long run.
Uhm.. where is the actual interview?
There is a link in the article above you know… 😉
Hey, it sounds great… so where to download the patch??
Never mind the previous question. There is link to the patch in the article…
I guess I did notice when I first read it…
First, the patch works great for me. I am very happy that the work is being done.
Second, to Yama, I disagree that the patch is a hack. I follow LKML and Linus has been favorable to it in the past. Don’t let that extremely negative interview (where he seems pretty amicable to it even so) sway you
Solaris has a preemptive kernel. Mach has one. 2000 has a very threaded kernel, not sure if it is fully preemptive. Making the kernel preemptive isn’t a hack to fix a problem, it removes the the problem alltogether! With a preemptive kernel you don’t have the issue of kernel code running to completion, which is the source of the problem. Current solutions like conditional scheduling (low-latency patch) are hacks. What the concern is is not that a preemptible kernel is a hack, but that maybe the latency can be fixed without the patch by rewriting the algorithms. Even so the preempt kernel is a great idea because it is a feature, a new design, that brings with it a behavior that should be standard to begin with.
I am using the patch now and find it GREAT. I originally had problems with the patch on my HIGHMEM machine and Rob Love worked on a fix for me. It is in their now. It is a great patch. I want to see it in 2.5.
Seth
The reason the patch isn’t included isn’t so much the quality of the patch, nor its size, nor its “hackness” (which it really isn’t). The reason is that the patch, though limited in scope, changes something very central (spinlocks in the kernel) and thus interacts with almost every piece of code in the kernel. That’s the reason its 2.5 material.