RT strikes back!!! 3.0-rc7-rt0

[Scary Hallo]

Hope it will be stable.

[todoesverso]

Great!. I can't wait to add that kernel to GNUGuitarINUX

[brummer]

First build failed, so I reported the bug to Thomas Gleixner at linux-rt-users mailing list, it takes less then a hour and Thomas comes with a fix patch. I'm running now Kernel 3.0.0-rc7-rt0 i686 (32 bit gcc 4.5.2)

Indeed, it feels good to run a rt kernel again after the long offset.

[Capoeira]

[Capoeira]

released: https://lkml.org/lkml/2011/7/22/288

[brummer]

aktuell it's version 3.0-rt3 now,
http://news.gmane.org/gmane.linux.rt.user

Anyhow, there are still some problems, at least on 32 bit. Here 3.0-rt1 <-> rt3 still refuse to run/boot.
but 3.0-rc7-rt0 running stable like a rock.

[trulan]

3.0-rt2 works here on 32 bit. I haven't had time to seriously stress-test it yet (or try -rt3, and -rt1 wouldn't build on 32 bit), but so far it looks good. Except it doesn't always finish shutting down...

[brummer]

Hi trulan

well, here, all rt1-rt3 brings me random freeze sooner or later, without any log output.

greats
hermann

[trulan]

Well, that isn't exactly what I'd call 'stable' then...

But, 3.0-rt5 is working well currently, the failure to shut down that I was seeing on all the previous 3.0-rt*'s seems to have been fixed. Also, the -rt5 patch disables some troublesome config options for now, until tglx and company get a chance to fix them.

[brummer]

Yea, I have seen the announce, but haven’t the time to check it out right now. So you said it runs stable for you, (as far you could say for now )
Will see if I could give it a try today.

[brummer]

Indeed, 3.0-rt5 works here also well so far, .. ..

[bluesscream]

3.0-rt5 runs stable and performant here on 64bit systems, rtirq allows high priority for firewire_ohci. Kudos @ Thomas Gleixner

[brummer]

the biggest problem I have now with the 3.0-rt series is
CONFIG_NOHIGHMEM=y

no highmem is allowed, this leads on my box to only 830MB ram, and in follow of that, it leads to swap, resulting in Xruns.
For now I switch back to 2.6.39 witch gives me , in fact, a better performance then the 3.0-rt series.

[brummer]

lucky I'm :
Peter Zijlstra announced 3.0-rt7

- some raw_spinlock patch from Uwe.
- cpupri rework from Rostedt, something which should greatly benefit
everybody with smp hardware, the more cpus the more benefit.
- a possible fix for some posix timer splat reported by Fernando.
- i386-highmem support because I'd already written the code anyway
so I thought I might as well make it work.

[raboof]

The amount of progress being made is awesome. Hope it'll lead to noticable improvements .