Package Details: linux-rt-docs 4.11.12_rt13-1

Git Clone URL: https://aur.archlinux.org/linux-rt.git (read-only)
Package Base: linux-rt
Description: Kernel hackers manual - HTML documentation that comes with the Linux-rt kernel
Upstream URL: https://www.kernel.org/
Licenses: GPL2
Submitter: schivmeister
Maintainer: jhernberg
Last Packager: jhernberg
Votes: 143
Popularity: 0.485203
First Submitted: 2011-08-09 20:03
Last Updated: 2017-09-07 12:07

Pinned Comments

jhernberg commented on 2017-09-07 12:12

I finally had some time to update to 4.11-rt.

Note that I still have slightly higher kernel scheduling latency than with 4.9-rt. I get cyclictest max values on the order of 150us instead of the normal 80us. Hopefully this is not a problem for any of you..!

If I hear no complaints I'll update linux-rt-lts to 4.9-rt soon.

Latest Comments

Gimmeapill commented on 2017-09-12 07:07

@jhernberg: Some perf issues here, I had to rollback to 4.9.x. I didn't confirm with cyclictest (yet), but I had problems with audio applications GUI responsiveness and stability on my Intel based Ivy bridge notebook. Nothing outstanding in regards to xruns though.

vee commented on 2017-09-10 01:00

@jhernberg thank you for the update to `4.11`.

Did you notice too many warnings during build on this particular version. Didn't really notice this during `4.9` build.

jhernberg commented on 2017-09-09 16:04

@blackhole: I normally run this: "cyclictest -S -m -p98". I suppose it's possible it got fixed in the last patch, or that it's hardware dependent (driver/module). For the time being I'm using a laptop that always had higher max values, so hard to test.

blackhole commented on 2017-09-07 21:40

Just tested here. I don't see differences from the previous kernel.
My test is with tuna and CPU governor set to performance.
cyclictest -t1 -n -p99 -i100 -o10 -v | oscilloscope -s1000 >/dev/null
I have a max value of 11us and most of time between 1-3 us
Maybe your test is different from mine.
Also with a strong test like this
hackbench -l $loop &
cyclictest -l $loop -n -m -Sp98 -i100 -d0
I have these results:
T: 0 (10268) P:98 I:100 C: 19747 Min: 1 Act: 1 Avg: 1 Max: 10
T: 1 (10335) P:98 I:100 C: 19723 Min: 1 Act: 4 Avg: 1 Max: 15
T: 2 (10411) P:98 I:100 C: 19695 Min: 1 Act: 2 Avg: 1 Max: 13
T: 3 (10461) P:98 I:100 C: 19669 Min: 1 Act: 2 Avg: 1 Max: 7
T: 4 (10529) P:98 I:100 C: 19648 Min: 1 Act: 2 Avg: 1 Max: 10
T: 5 (10605) P:98 I:100 C: 19622 Min: 1 Act: 2 Avg: 1 Max: 16
T: 6 (10606) P:98 I:100 C: 19600 Min: 1 Act: 2 Avg: 1 Max: 10
T: 7 (10607) P:98 I:100 C: 19576 Min: 1 Act: 3 Avg: 1 Max: 11

jhernberg commented on 2017-09-07 12:16

@RedTide: Did you try building and installing it with makepkg -cfi?

jhernberg commented on 2017-09-07 12:14

@dvzrv: Sorry for the late reply, real life gets in the way sometimes..

I haven't seen any problems like that, hopefully 4.11-rt fixes it for you..

jhernberg commented on 2017-09-07 12:12

I finally had some time to update to 4.11-rt.

Note that I still have slightly higher kernel scheduling latency than with 4.9-rt. I get cyclictest max values on the order of 150us instead of the normal 80us. Hopefully this is not a problem for any of you..!

If I hear no complaints I'll update linux-rt-lts to 4.9-rt soon.

RedTide commented on 2017-08-24 05:46

I had the same problem as ToxicAvenger, no idea if it's a pamac fault, but after some time the build fails with a no available space on disk error, where the are 15GB available.

dvzrv commented on 2017-08-07 11:28

@jhernberg: I have a weird issue with the current xf86-video-intel on linux-rt. Whenever I do something graphical (starting certain applications), this happens:
[ 71.210650] [drm:intel_cpu_fifo_underrun_irq_handler [i915]] *ERROR* CPU pipe A FIFO underrun
[ 71.211089] [drm:intel_set_pch_fifo_underrun_reporting [i915]] *ERROR* uncleared pch fifo underrun on pch transcoder A
[ 71.211142] [drm:intel_pch_fifo_underrun_irq_handler [i915]] *ERROR* PCH transcoder A FIFO underrun

and weird overlay glitches appear, that make me restart at some point, because they become to annoying.
Let me know, when there's anything to test for 4.11-rt. I'd be happy to help (especially because this bug is making it hard to work with my x220 atm).

jhernberg commented on 2017-08-07 09:38

The solution to the nvidia kernel module build problem is to do something like this: make IGNORE_PREEMPT_RT_PRESENCE=1 SYSSRC=/usr/lib/modules/"${_kernver}/build" module

I've asked a nvidia maintainer and the fix will be in nvidia-dkms-384.59-3, so hopefully we're back to trouble free installation soon!

All comments