View previous topic :: View next topic |
Author |
Message |
iago n00b
Joined: 04 Apr 2003 Posts: 16 Location: Rivarolo Canavese, Turin, Italy
|
Posted: Sun Sep 26, 2004 9:15 am Post subject: Preemptive kernel? |
|
|
Is it a good idea to enable preemption in a gentoo-dev-sources-2.6.7-r14 kernel?
has it known problems? Is there some app that woldn't work with it enabled?
Thanks a lot |
|
Back to top |
|
|
Deranger Veteran
Joined: 26 Aug 2004 Posts: 1215
|
Posted: Sun Sep 26, 2004 10:17 am Post subject: |
|
|
I would say, disable it. I've never found any advantages of that. And Con Kolivas (creator of ck-sources patchset) says that it should be disabled. I trust him... |
|
Back to top |
|
|
fb l33t
Joined: 08 Dec 2003 Posts: 636 Location: New Zealand
|
Posted: Sun Sep 26, 2004 10:32 pm Post subject: |
|
|
Don't use it! I tried it and I got into heaps of weird troubles
which were resolved when I removed it. The removal in
itself was not an easy prospect as one of the problem I got
was that gcc was breaking randomly during big compilations,
possibly in combination with some loads (like running KDE). |
|
Back to top |
|
|
Tesseract n00b
Joined: 03 Jun 2004 Posts: 21
|
Posted: Mon Sep 27, 2004 2:37 pm Post subject: |
|
|
Back when ppc was still using ppc-development-sources, there was a message when you emerged which flat-out said that preempt was broken on ppc. gentoo-dev-sources doesn't have that message, but I haven't seen anything in the changelogs since then which would suggest that the problem has been fixed. If it hasn't, then that message really needs to come back, along with a note in the ppc handbook. Many guides to configuring the 2.6 kernel encourage using preempt on desktop machines, so this could easily trip people up. |
|
Back to top |
|
|
spitzwegerich l33t
Joined: 04 Mar 2003 Posts: 697 Location: Lower Bavaria, Central Europe
|
Posted: Mon Sep 27, 2004 5:02 pm Post subject: |
|
|
fb wrote: | Don't use it! I tried it and I got into heaps of weird troubles
which were resolved when I removed it. The removal in
itself was not an easy prospect as one of the problem I got
was that gcc was breaking randomly during big compilations,
possibly in combination with some loads (like running KDE). |
I made similar expreriences on ~2.6.2. Don't know if anything changed in the meantime. _________________ "Work is the curse of the drinking classes."
-Oskar Wilde |
|
Back to top |
|
|
HighOnBonsai Apprentice
Joined: 05 Jan 2004 Posts: 260
|
Posted: Mon Sep 27, 2004 6:02 pm Post subject: |
|
|
I had A LOT OF trouble the last weeks - gcc always aborted during long compilations and I wasn't able to figure out whats wrong. I blamed the memory (called Apple support) and blamed the heat (enabled "cooling kernel modules") and ... but now everything works: I disabled preempt option last week...
Christopher _________________ Are YOU high - on bonsai? |
|
Back to top |
|
|
Hydraulix Guru
Joined: 12 Dec 2003 Posts: 447
|
Posted: Mon Sep 27, 2004 9:50 pm Post subject: |
|
|
Preemptive kernel should be removed from ppc, it's the devil!! I had the same problem that everyone else here did. _________________ It is the fate of operating systems to become free.
- Neal Stephenson |
|
Back to top |
|
|
jeepfr n00b
Joined: 17 Nov 2004 Posts: 3
|
Posted: Fri Nov 19, 2004 1:58 pm Post subject: |
|
|
same here, CONFIG_PREEMPT is responsible of random segfaults under heavy loads (especially long compiles with gcc but no app is safe, i seen gaim, evolution just crash silently)
i was about to blame my memory when i found a post saying some of those were symptoms of CONFIG_PREEMPT on ppc.
made kernels without it... works rock solid. |
|
Back to top |
|
|
reverseninja n00b
Joined: 23 Jul 2003 Posts: 31
|
Posted: Fri Nov 19, 2004 3:24 pm Post subject: |
|
|
I've got preempt in my kernel *2.6.9-r4* and my ibooks been buzzing along just fine doing a gnome 2.8 install. 173 packages in total, and when last I checked this morning before heading to work, it was at 112. So far so good. |
|
Back to top |
|
|
bruda Guru
Joined: 06 May 2004 Posts: 376 Location: Sherbrooke, QC, Canada
|
Posted: Fri Nov 19, 2004 5:31 pm Post subject: |
|
|
You are being lucky, plain and simple. Preemption gives wonderfully strange errors all over the place on PPC, is evil, and should not be attempted. But then hey, if it works for you be happy and enjoy. _________________ Quid latine dictum sit altum videtur |
|
Back to top |
|
|
reverseninja n00b
Joined: 23 Jul 2003 Posts: 31
|
Posted: Fri Nov 19, 2004 7:42 pm Post subject: |
|
|
Haha, quite right I'm enjoying. Though I don't suppose it's a matter of luck, just that the ibook fears if it should fail me, as my wrath would be absolute. |
|
Back to top |
|
|
kazakhan n00b
Joined: 19 Apr 2002 Posts: 12
|
Posted: Sun Nov 21, 2004 1:03 am Post subject: |
|
|
I just upgraded my kernel from 2.4.x to 2.6.7-r14 on my 12" PB, not knowing any better I enabled preempt and all has been fine for the last 24 hrs |
|
Back to top |
|
|
stimuli Apprentice
Joined: 16 Dec 2002 Posts: 292 Location: Dartmouth, NS, Canada
|
Posted: Sun Nov 21, 2004 5:52 am Post subject: |
|
|
I've used preempt for long periods of time as well, but have had many long builds fail mysteriously. I've since recompiled my kernel w/out KP, and have had no such problems.
It's still experimental (and to be honest, I'm not sure there was any sort of quantifiable difference with it enabled, either).
The re-written scheduler in 2.6 kernels is really the show stealer, in terms of responsiveness. |
|
Back to top |
|
|
codergeek42 Bodhisattva
Joined: 05 Apr 2004 Posts: 5142 Location: Anaheim, CA (USA)
|
Posted: Sun Nov 21, 2004 6:06 am Post subject: |
|
|
It was a good thing in the 2.4 kernels, but in 2.6 the scheduler is now quite awesome and so kernel pre-emption seems to bea kludge that has more cons than pros. _________________ ~~ Peter: Programmer, Mathematician, STEM & Free Software Advocate, Enlightened Agent, Transhumanist, Fedora contributor
Who am I? :: EFF & FSF |
|
Back to top |
|
|
BetterUnborn n00b
Joined: 28 Sep 2004 Posts: 57 Location: Germany
|
Posted: Sun Mar 27, 2005 5:18 pm Post subject: |
|
|
I just tried out 2.6.12_rc1, because I saw much fixes about ppc in changelog ... and while playing around a bit with the kernel config, I decided to give preemption a try.
So far, it's really fine for me. My iBook is compiling for about an hour now, and hasn't aborted so far (this was the place where I could first see whether something is wrong with the kernel: compilation processes tended to die unexpectedly with preemption enabled).
On the other hand, I can't see that much of an improvement. Yes, the system is more responsive while the CPU is cooking, I hadn't been able to watch a vcd while compiling stuff before, but without heavy load it's business as usual.
I just wanted to say that preemption seems to be quite stable on ppc right now (even on an unstable kernel ) |
|
Back to top |
|
|
silian87 Advocate
Joined: 06 Oct 2003 Posts: 2318 Location: Treviso, Italy
|
|
Back to top |
|
|
fuoco Guru
Joined: 23 May 2004 Posts: 386 Location: Israel
|
Posted: Fri Oct 21, 2005 11:07 am Post subject: |
|
|
Anyone knows what's the status of preemptive kernel with recent kernels and recent gcc versions (including 4) on ppc ? |
|
Back to top |
|
|
JoseJX Retired Dev
Joined: 28 Apr 2002 Posts: 2774
|
|
Back to top |
|
|
silian87 Advocate
Joined: 06 Oct 2003 Posts: 2318 Location: Treviso, Italy
|
|
Back to top |
|
|
JoseJX Retired Dev
Joined: 28 Apr 2002 Posts: 2774
|
Posted: Fri Oct 21, 2005 2:40 pm Post subject: |
|
|
Yeah, it's been that way for quite a while. I personally don't see a performance difference, maybe the overhead of SMP negates the benefit of preempt. _________________ Gentoo PPC FAQ: http://www.gentoo.org/doc/en/gentoo-ppc-faq.xml |
|
Back to top |
|
|
fuoco Guru
Joined: 23 May 2004 Posts: 386 Location: Israel
|
Posted: Sat Oct 22, 2005 10:42 pm Post subject: |
|
|
I've been using preempt for some time on 2.6.13 kernel, and haven't noticed any problems. Do you suggest that I turn it off ? |
|
Back to top |
|
|
JoseJX Retired Dev
Joined: 28 Apr 2002 Posts: 2774
|
Posted: Sat Oct 22, 2005 11:19 pm Post subject: |
|
|
If you have SMP on, then it's safe to leave on. If you have SMP off and you're just using preempt, then eventually, you'll get random segfaults. Some people have them more often than others, but it'll happen eventually. _________________ Gentoo PPC FAQ: http://www.gentoo.org/doc/en/gentoo-ppc-faq.xml |
|
Back to top |
|
|
AaronPPC Guru
Joined: 29 May 2005 Posts: 522 Location: Tucson, AZ
|
Posted: Sun Oct 23, 2005 12:24 am Post subject: |
|
|
Interesting thread. I thought that pre-emption was a good thing to have on, so I enabled it with SMP. Now I am having second thoughts. _________________ --Aaron |
|
Back to top |
|
|
JoseJX Retired Dev
Joined: 28 Apr 2002 Posts: 2774
|
Posted: Sun Oct 23, 2005 2:00 am Post subject: |
|
|
As long as you have SMP on, it's 100% stable. I've been running it on my PegasosII with SMP+preempt for a few months now with no issue. But *I* personally don't see any real difference between that and no SMP and no preempt. Preempt without SMP still eventually leads to segfaults during builds with the 2.6.14 kernel on my machine, but really only in larger packages like kdelibs. _________________ Gentoo PPC FAQ: http://www.gentoo.org/doc/en/gentoo-ppc-faq.xml |
|
Back to top |
|
|
96140 Retired Dev
Joined: 23 Jan 2005 Posts: 1324
|
Posted: Sun Oct 23, 2005 5:51 am Post subject: |
|
|
JoseJX wrote: | If you have SMP on, then it's safe to leave on. If you have SMP off and you're just using preempt, then eventually, you'll get random segfaults. Some people have them more often than others, but it'll happen eventually. |
Whoa. Now, is this also the case for x86 kernels? (Sorry, wandered in from Intel-land, and am vaguely concerned; I don't have SMP (as I have a uniprocessor system) but always enable pre-empt--including the big kernel lock--in my kernel config.) |
|
Back to top |
|
|
|