Arch Linux User Repository

Is anyone interested in testing the 5.17.y ck kernel? I'd like some broader feedback before I push it out to the AUR and to the repo. Please try it here if interested:

http://repo-ck.com/testing/

You can download a prebuilt package or build yourself.

Please post feedback good or bad to the repo-ck thread to keep it all in one place: https://bbs.archlinux.org/viewtopic.php?pid=2031356

@zerophase I've been having the (arguably) same problem since very long time (see e.g. https://bbs.archlinux.org/viewtopic.php?pid=1966653#p1966653), and that actually forced me to go back to the stock kernel since I never found a solution.

Anyone else having issues getting Docker to start up with the CK kernel?

I've noticed issues with mounting archiso too.

Is anyone interested in testing the 5.17.y ck kernel? I'd like some broader feedback before I push it out to the AUR and to the repo. Please try it here if interested:

http://repo-ck.com/testing/

You can download a prebuilt package or build yourself.

Please post feedback good or bad to the repo-ck thread to keep it all in one place: https://bbs.archlinux.org/viewtopic.php?pid=2031356

It's not MuQSS, but the PKGBUILD now contains CK's hrtimer patch and the recommended 1000 Hz tick rate.

sad news https://www.phoronix.com/scan.php?page=news_item&px=Con-Kolivas-EOL-MuQSS-CK

I'm happy to go along with @air-g4p ! Thank you very much @graysky, and all contributors and commenters.

@graysky - WOW!! Today marks linux-ck's 10th Birthday!

Thank you very much for all your hard work during the past decade, graysky!

Your efforts are much appreciated...

Cheers

Pretty sure, delete the tar ball and re-download :p

@graysky, are you sure about b2sum checksum for 'more-uarches-20210610.tar.gz'? I downloaded it manually and checksum didn't match from PKGBUILD: 30d1df754608bb423cbc99c2097ad521baa091b9a3b39df4bd5c2d50c57eec54d8fa0e4a4a04b847c3d1b87ba682cadc8db45fabeefdc9ad7caaf8e77b96e41a tmp/kernel_compiler_patch-20210610.tar.gz

If the 'https://github.com/graysky2/kernel_compiler_patch' git repository is yours would it be feasible to use signed tags to accompany releases? In such cases signed tag would help others to simply change checksum in PKGBUILD. If it is too troublesome then don't bother with it.

According to this:

https://www.infoq.com/news/2021/01/docker-engine-cgroups-logging/

the current version of docker should work with cgroups v2. So I guess it's some interplay between the CK patchset and cgroups v2 that breaks docker. Sadly, I got no reply from CK on his blog...

@artafinde your suspect was correct. Setting the kernel parameter you suggested makes docker work again with linux-ck. I'm going to report this on CK's blog. Thanks.

I suspect your issues has something to do with cgroups v2. Try to replicate the issue after setting (temporarily) systemd.unified_cgroup_hierarchy=0 as a kernel param which forces the cgroup v1 behavior.

@snack had your issue and also a system freeze with the latest release. First time after almost a year of using CK. Went back to Zen.

@zerophase I'm not sure. For sure it's something inside the CK patchset (since the stock and many other patched kernels work fine), and it's not MuQSS (see e.g. the comment by Steven Barret on CK's blog stating that Liquorix with MuQSS works fine). I'd need to understand if there is some hope for this to be fixed, otherwise I'll have to ditch linux-ck.

@snack It has been a bug for awhile. I tried reporting it about a year ago, when trying to author a custom arch install disk, and got some explanation from a guy that this is not a CK issue. (Wasn't CK)

I cannot start docker containers when using the current (5.12) linux-ck-skylake kernel. I get this error:

$ docker run --rm -it -p 80:80 ckulka/baikal:nginx
docker: Error response from daemon: failed to create shim: OCI runtime create failed: container_linux.go:380: starting container process caused: process_linux.go:545: container init caused: process_linux.go:508: setting cgroup config for procHooks process caused: load program: invalid argument: unknown.

With the stock Arch kernel it works, and seemingly also with other patched kernels like Liquorix (although I didn't try by myself).

I know that it is usually suggested to post to CK's blog for this kind of issue, and in fact I did it but from there I got the advice to post to AUR...

graysky> GCC 11 reorders things and since GCC 11 is in [core], I updated the PKGBUILD.

That was a culprit! I updated PKGBUILD, rebuild the package (linux-ck-sandybridge) and was able to boot with linux-ck-sandybridge kernel.

@graysky, thanks for your suggestion. There is not any problem for me, I had already corrected it when I saw it on nconfig. I mentioned about it just to inform.

@graysky, no it is GCC 10.2

Right, define an updated mirror and update your system :D

@graysky, no it is GCC 10.2

It could also happen on automation level, where the build system is still using old IDs. I'm building for Skylake right now so I can check if it's actually built for Skylake in the end.

Yeah, as I hadn't any issues myself I wanted to suggest "a wrong architecture" source of the problem since I've noticed a change of architecture IDs in the package. For example, I've used 25 for Skylake for quite a long time but in the very last release I see it numbered as 26.

@elektorronikci - Is your system up-to-date? GCC 11 reorders things and since GCC 11 is in [core], I updated the PKGBUILD.

There is a problem in the PKGBUILD subarch configuration. When I select Haswell, it was actually Broadwell in the .config file. That may cause boot problem because of some unimplemented machine codes.

Ok, I posted message on ck's blog.

I suspect an issue with ck patch and GCC11 but I don't understand why newer processors (haswell and skylake specifically) work.

@graysky, thanks I'll do that when I finish testing. I compiled linux-ck-sandybridge 5.12.2, 5.12.3 with the same configuration as 5.12.4 (but without 0002-drm* patch applied) but both kernels didn't boot.

Recommend you post to ck's blog and see what he thinks. I have no insights.

Similar situation here: my notebook with Intel Core i5 2410M can't boot with linux-ck-sandybridge 5.12.4 but works fine with distro linux 5.12.4. I locally built linux-ck package (though without patch 0002-drm...) with gcc 11 and config the same as for linux package. Seems that something affects "old" processors because on other hosts with Intel processors of haswell and skylake family both linux-ck-haswell and linux-ck-skylake 5.12.4 boot fine. I didn't tried 5.12.3 version but 5.12.2 worked. I'll try to build 5.12.2 and 5.12.3 for sandybridge and report results.

Had to downgrade to 5.12.1-1-ck-nehalem because linux-ck-nehalem-5.12.4-1 freezes the system at boot when "Loading initial ramdisk".

Hello

I boot to a black screen with 5.12.3-2 on a Gemini lake CPU. Version 5.12.3-1 worked fine. (i'm using the bin version).

Report/ask on ck's blog: https://ck-hack.blogspot.com/

@coderkun It's probably something to do with Linux-ck I tried mounting the archiso build system for a custom installer, a year ago, and I can't do it on Linux-ck. Works fine on the default. I told CK about, but got no response. I also received message, from someone else, accusing me of causing the problem.

I think we have the same Arch default settings, other than the CK stuff. I don't see why this would be our fault. It's probably a bug in his code.

@coderkun, I already listed that on the forums: https://bbs.archlinux.org/viewtopic.php?pid=1967535#p1967535

I'm still looking for a solution for that, btw :-)

For a week or two I have been getting the following error when trying to start a docker container:

Error response from daemon: OCI runtime create failed: container_linux.go:367: starting container process caused: process_linux.go:495: container init caused: process_linux.go:458: setting cgroup config for procHooks process caused: can't load program: invalid argument: unknown

It work without error on Arch’s default kernel. Is anyone else experience the same and knows what the issue might be?

Thanks!

@graysky: You can remove the hack for xz since 70261 is fixed.

I sure love updating my kernel 4 times in a week, lol

@powerball253- Comments like this serve absolutely no purpose. In general, this package follows the progress of the official kernel which has bumped several times in the past week.[1] Further, no one is forcing you to rebuild a kernel you opted into using in the first place. Please don't spam the comments page with such statements.

1. https://github.com/archlinux/svntogit-packages/commits/packages/linux/trunk

@Kr1ss - Updated in 5.11.11-4, thanks.

@graysky - Currently, building fails in environments with a modified SRCDEST folder. (EDIT/ For instance, mine is set to /tmp/makepkg/srcdest)

linux-ck-5.11.11-3-x86_64-prepare.log :

xz: /startdir/patch-5.11-ck1.xz: No such file or directory
==> ERROR: A failure occurred in prepare().
    Aborting...
==> Removing installed dependencies...
checking dependencies...
[...]

Hence I'd like to suggest a little patch :

diff -U1 old/PKGBUILD new/PKGBUILD
--- old/PKGBUILD    2021-04-02 14:11:24.598889831 +0200
+++ new/PKGBUILD    2021-04-02 14:11:02.978925899 +0200
@@ -107,3 +107,3 @@
     unlink "$srcdir/$_ckpatch.xz"
-    xz -dc "$startdir/$_ckpatch.xz" > "$_ckpatch"
+    xz -dc "$SRCDEST/$_ckpatch.xz" > "$_ckpatch"
   fi

Thank you, cheers !

@graysky: Thanks already using it - noticed the PR from sirlucian :) Thanks!

@Zalgo - 5.11.11-3 should fix this. See the forum thread for discussion: https://bbs.archlinux.org/viewtopic.php?pid=1965531

@artafinde - with some help, the native option now works with Intel and AMD. You will have to select the correct option though:

...
  33. Intel-Native optimizations autodetected by GCC (MNATIVE_INTEL) (NEW)
  34. AMD-Native optimizations autodetected by GCC (MNATIVE_AMD) (NEW)

Thanks for the tip :) If i figure out a way to fix the PKGBUILD that doesn't require manual untarring I'll pose it here

That's odd. Makepkg doesn't seem to be untarring the patch. If you do it manually and add that to the source array, it works as expected.

This currently fails on my system with

Setting config...
sed: can't read ../patch-5.11-ck1: No such file or directory
==> ERROR: A failure occurred in prepare().
    Aborting...

can anyone else reproduce? I'll test in a clean chroot

@artafinde - I need to dig deeper, but I think it will be difficult to assign a native CPU out of some configs.

For example, X86_INTEL_USERCOPY config. I do that now but no other AMD chips are listed by upstream[1].

Further, I believe AMD chips should not be in the X86_P6_NOP config[2] or X86_CMOV config[3].

1. https://github.com/graysky2/kernel_gcc_patch/blob/master/enable_additional_cpu_optimizations_for_gcc_v10.1%2B_kernel_v5.8%2B.patch#L407

2. https://github.com/graysky2/kernel_gcc_patch/blob/master/enable_additional_cpu_optimizations_for_gcc_v10.1%2B_kernel_v5.8%2B.patch#L421

3. https://github.com/graysky2/kernel_gcc_patch/blob/master/enable_additional_cpu_optimizations_for_gcc_v10.1%2B_kernel_v5.8%2B.patch#L438

@graysky checksum for the gcc kernel patch is failing - needs an update.

I noticed the changes in the patch repo and it mentions one should not use native on AMD chips. What's the reasoning behind this since the gcc 10 (and clang for what matters) can detect the cpu arch just fine:

$ gcc -march=native -E -v - </dev/null 2>&1 | grep cc1
 /usr/lib/gcc/x86_64-pc-linux-gnu/10.2.0/cc1 -E -quiet -v - -march=znver2 -mmmx -mno-3dnow -msse -msse2 -msse3 -mssse3 -msse4a -mcx16 -msahf -mmovbe -maes -msha -mpclmul -mpopcnt -mabm -mno-lwp -mfma -mno-fma4 -mno-xop -mbmi -mno-sgx -mbmi2 -mno-pconfig -mwbnoinvd -mno-tbm -mavx -mavx2 -msse4.2 -msse4.1 -mlzcnt -mno-rtm -mno-hle -mrdrnd -mf16c -mfsgsbase -mrdseed -mprfchw -madx -mfxsr -mxsave -mxsaveopt -mno-avx512f -mno-avx512er -mno-avx512cd -mno-avx512pf -mno-prefetchwt1 -mclflushopt -mxsavec -mxsaves -mno-avx512dq -mno-avx512bw -mno-avx512vl -mno-avx512ifma -mno-avx512vbmi -mno-avx5124fmaps -mno-avx5124vnniw -mclwb -mmwaitx -mclzero -mno-pku -mrdpid -mno-gfni -mno-shstk -mno-avx512vbmi2 -mno-avx512vnni -mno-vaes -mno-vpclmulqdq -mno-avx512bitalg -mno-avx512vpopcntdq -mno-movdiri -mno-movdir64b -mno-waitpkg -mno-cldemote -mno-ptwrite -mno-avx512bf16 -mno-enqcmd -mno-avx512vp2intersect --param l1-cache-size=32 --param l1-cache-line-size=64 --param l2-cache-size=512 -mtune=znver2

@simona, AFAIU, graysky keeps config sync with linux package and these two are in linux-zen but not yet in linux.

ashmem_linux and binder_linux missing?

@Krlss, same here.

@graysky, why not use git.kernel.org? Like that 0003-SUNRPC-Fix-NFS-READs-the-start-at-non-page-aligned-o.patch::https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/patch/?id=bad4c6eb5eaa8300e065bd4426727db5141d687d

P.S. I like the beginning of commit hash ;)

That's weird... Github doesn't always serve the exact same patch !

Sometimes when you download it, the object hashes are abbreviated to 13 characters and sometimes to 14.

That's why the b2sum didn't match here, the first time.

https://i.imgur.com/ILo6CqU.png

@graysky pls disregard my last comment. Seems like I had an issue with an unclean chroot.

@Kr1ss - Matches here. It is pulled as a source/not included with package.

@graysky - The checksum of 0003-SUNRPC-Fix-NFS-READs-the-start-at-non-page-aligned-o.patch doesn't match after the latest update. ;)

EDIT the file is missing from the AUR sources.

@artafinde - Yes, it was a typo. It doesn't hurt anything.

Latest update bumbed version to 5.10.13 but forgot to reset to 1 the pkgrel - not a real issue, but just a reminde for next bump.

@graysky thanks for the update but... I think it's missing a $ ($pkgbase instead of just pkgbase).

@Distorted - It is undone in the PKGBUILD. Maintaining is easier if I take the Arch config as-is. See https://aur.archlinux.org/cgit/aur.git/tree/PKGBUILD?h=linux-ck#n133 and else where in the PKGBUILD other Arch specific calls are commented out.

@bryanjhv - sure, 5.10.5-2

Why is CONFIG_DEBUG_INFO=y set? It causes builds to use over 10GB in disk space.

@graysky, is there possibility to change _package-headers' depends array using "$pkgbase" instead of fixed 'linux-ck'? I was building for silvermont (changing $pkgbase to linux-ck-silvermont the way some repositories provide it), and didn't notice it until installing linux-ck-silvermont-headers asked me to also install linux-ck instead of linux-ck-silvermont.

Ignoring the error would make things unpredictable. Just remove "-N" option so patch will try to reverse-apply and re-apply.

@LucidComplex That would make it pass though also in situations where the patch itself fails for a different reason. Not sure if that's what you want to achieve.

prepare() fails when patches have already been applied. I suppose you can append a || true to the end of patch commands in prepare() to address this.

@graysky The checksum for config doesn't match in the recent release.

EDIT/ Thank you fixing this so quickly ! :)

My bad... git push didn't respond for some reason.

I haven't found 5.10.4-2, but as long as linux-headers isn't in the provides array anymore, I'm fine with it.

@scimmia - Please review 5.10.4-2 and comment

If packages require that and aren't building against the 'linux' package, they're poorly packaged and should be fixed there, not hacked around here.

At one time there was a need to have the linux-ck-headers package provide linux-headers. Thought it had to do with other packages requiring for actual linux-headers package which is wrong if a user is running linux-ck not linux.

See the previous link, this should not be providing linux-headers. Packages saying they provide themselves makes no sense, either.

@Scimmia - Context? "Fixed?"

3 updates and the provides has not been fixed. Is there a problem, graysky?

Please see the warning here: https://wiki.archlinux.org/index.php/Kernel/Arch_Build_System#Modifying_the_PKGBUILD

@nTia89 - fixed in 5.9.11-2 https://aur.archlinux.org/cgit/aur.git/commit/?h=linux-ck&id=1ab6b24512b14ba2abd23f6e8774d315eed6931a

If anyone as issues with suspend/poweroff, this is the reason why: https://bugs.archlinux.org/task/68762 Just to share

@nTia89 - My bad. Order of events in the PKGBUILD matter. Fixed in 5.9.1-2

Since last update, there is no more the interactive micro-architecture selection before compilation. Does anyone have the same issue?

LOLs!!! I cannot fucking keep up!

IIRC:

1. First we had a fucked up 'unfuck-xx' patch
2. Then the fucked up unfuck-xx patch got unfucked
3. Then @graysky accidentally fucked up something else
4. Finally, @graysky unfucked the most recent fuck up

The bottom line: @graysky, happy to report that linux-ck 5.7.17-1 built and installed without any fuck ups! Cheers!

Working fine here.

Thx a lot @ooo & @graysky, cheers !

@ooo - Thanks for the fix. I do not have time to build/test. Can someone please verify that 5.7.17-1 works as expected?

@graysky Apparently the error with 5.7.17 was fixed in zen-kernel already, although I didn't test it myself yet.

See the commits from 2020-08-22 at https://github.com/zen-kernel/zen-kernel/commits/5.7/muqss

EDIT: Applying the fix from zen-sources actually gets a bit confusing, so to make this a little simpler, first apply this to patch-5.7-ck1, to make it apply cleanly against 5.7.17: https://pastebin.com/uMMJJq3y

Then apply this to fix the build: https://github.com/zen-kernel/zen-kernel/commit/e7113f2c1d56962d4afabc627ad761c138f5b858

5.7.17 introduces some changes that break the ck1 patch. If someone wants to take a look at that and share a solution, that would be good. I am busy at the moment.

https://ck-hack.blogspot.com/2020/08/more-delays-and-motivation-issues.html

Yes, I fucked it up, should be fixed with https://aur.archlinux.org/cgit/aur.git/commit/?h=linux-ck&id=473e18c7ccdec2e641eab76ffcf93e87ff6140a1

You don't need to rebuild unless you have [testing] enabled or have it enabled in a clean build root btw.

The recent pkgrel bump broke the validity of the config checksum, @graysky. Could you update that pls ?

Thank you !

@graysky - 5.7.14-1-ck compiled and installed without issue. Good work, done here!

@graysky I'm running 5.7.14-1-ck right now, built w/ localmodconfig.

Looks fine so far, thx.

I cannot reboot now so can someone verify that 5.7.14-ck1 is working properly for me? A small adjustment to ck1 was required.

Inexplicably, this I ran into the SHA256 issue on one of my machines but not the other. I diffed the two versions of unfuck-ck1.patch - check it out:

--- unfuck-ck1.patch.works  2020-07-30 18:00:17.955122911 -0700
+++ unfuck-ck1.patch.borked 2020-07-30 18:01:49.000000000 -0700
@@ -8,7 +8,7 @@
  1 file changed, 11 insertions(+)

 diff --git a/kernel/sched/MuQSS.c b/kernel/sched/MuQSS.c
-index 18a9b4a23e44e..c55c158e8b8e4 100644
+index 18a9b4a23e44ea..c55c158e8b8e42 100644
 --- a/kernel/sched/MuQSS.c
 +++ b/kernel/sched/MuQSS.c
 @@ -3246,6 +3246,17 @@ static inline u64 do_task_delta_exec(struct task_struct *p, struct rq *rq)

It seems like https://github.com/ckolivas/linux/commit/0b69e633d6b0b08ae8547dc4099c8c0985019553.patch is inconsistent in how many digits of the Git SHAs it sends.

I think it's just a wild guess (or just my guts), but I think this might be some caching problem when downloading that particular file. Sometimes you might get it from a different server with a different content (even though this is odd, yes), but might explain why sometimes (and how inconsistently) this happens ...

@graysky and @artafinde: Thanks for trying to reproduce on your respective ends.

Inexplicably, the build completed without any PKGBUILD SHA256 modifications on my other laptop, and again running makepkg -sric. I suppose this success is consistent with @artafinde's '50% of the time' finding when running makepkg -sric.

Cheers guys...

@graysky I was able to reproduce @vltr's report by typing 'makepkg -si' instead of 'makepkg' on a fresh git clone.

@graysky that's ODD. I always build after cleaning up the git repository (git clean -dfx) ... and the file had that signature I mentioned (hence my build failed). I cleaned (again), ran makepkg and everything went smoothly this time (with unfuck-ck1-fix-suspend-to-ram.patch signature being, indeed, 961ed94b8d905f1e901cacb08d253c4170af0a25828111b7558d9c874e923558). that's really weird ... sorry about the confusion, whatever it happened.

@vltr - You sure?

% sha256sum unfuck-ck1-fix-suspend-to-ram.patch
961ed94b8d905f1e901cacb08d253c4170af0a25828111b7558d9c874e923558  unfuck-ck1-fix-suspend-to-ram.patch

hi there, @graysky! just one problem with 5.7.10: the unfuck-ck1-fix-suspend-to-ram.patch has a sha256sum of value 961ed94b8d905f1e901cacb08d253c4170af0a25828111b7558d9c874e923558 in the PKGBUILD, which would not pass the validity check since the correct sha256sum should be 9ad11f720d4dd83ddb8876b453f4526ad35f1ff72465213764fb1d206fea0379. apart from that, everything works fine. thanks!

Above all, thanks @graysky for the change! Anyway, /tmp defaults to tmpfs with size=50% but not everybody has so much RAM. By my own, this is the biggest mistake behind that choice...

I suggest that you disable CONFIG_DEBUG_INFO, as it was before 5.7.2. That's what I am doing in my own builds anyway to build with a 4GB /tmp.

graysky: Got my answer: https://bugs.archlinux.org/task/66260

Kr1ss: I think devs would probably reply don't build in tmpfs (and require ~15G of space instead of ~3.5G) but that's a discussion for forums possibly :D - Thanks graysky.

Thx @graysky ! I was only able to build 5.7.2-1 after blowing up /tmp to 20GiB, that's kind of ridiculous. If a user needs debug symbols they can enable them; I think most users wouldn't need'em anyways.

Cheers !

@artafinde - Fuck it, 5.7.2-2-ck1. Change my mind.

@graysky - Agree, especially since later we strip it from the bug symbols. I raised the question in IRC but no answer. I guess forums or a bug might have better luck.

@artafinde - Yes, the debug stuff causes a failure on one of my machines with 16G of memory building in tmpfs (allocating 14.5G to the tmpfs partition on which the build root resides!)

This change was introduced with 5.7.2-1-arch1 here: https://git.archlinux.org/svntogit/packages.git/commit/trunk?h=packages/linux&id=663b08666b269eeeeaafbafaee07fd03389ac8d7

First question is why do we need the debug options?

"If you say Y here the resulting kernel image will include debugging info resulting in a larger kernel image. This adds debug symbols to the kernel and modules (gcc -g), and is needed if you intend to use kernel crashdump or binary object tools like crash, kgdb, LKCD, gdb, etc on the kernel. Say Y here only if you plan to debug the kernel. If unsure, say N."

I am getting the same error with /tmp running out of space as of version 5.7.2-1.

nTia89: The default configuration changed (now we produce some debug info) which we strip on packaging and creation of vmlinuz. See config and change in 5.7.1 (https://aur.archlinux.org/cgit/aur.git/commit/?h=linux-ck&id=b0b10d913f31d37155520286c98ce5282e962060) Config change comes from upstream (arch default config. PS: In my case with 16G /tmp was filled while building without any change in config or PKGBUILD. I suggest to not use tmpfs for now.

Staring with 5.7.1, I cannot compile kernel because I run out of space on my /tmp (tmpfs with size of 6GB). Did it change anything recently?

@as3ii: nvidia-dkms 440.82-2 works for -ck kernel. No idea about bbswitch.

linux-ck 5.7.2 compiled correctly, but bbswitch and nvidia doesn't work (dkms install error)

Error! Bad return status for module build on kernel: 5.7.2-1-ck

linux-ck 5.7.2 compiled successfully and already up and running. Thanks, @graysky!

does not cleanbuild 5.6.16.1 -> 5.7.1-2 tried on two different archlinux hosts (bulldozer und core2)

please fix that

@kwe - Yes, I saw that posted by sir_lucjan. I am reluctant just hack it together without CK's review. See his blog.

Getting the same error. It seems to be just that single function that seems to be missing for the compilation to finish, though.

I just added the function that was introduced in https://gitlab.com/alfredchen/linux-bmq/-/commit/ad9eb862e06f395e846a2f2e084e5b0e7eb9ca4b to the top part of drivers/thermal/cpufreq_cooling.c and it went through.

Not sure if anything else is missing, though, but please don't consider this a proper fix; it might be dangerous, even. I'm going with this solution, for now, because I have accepted the risk.

@graysky yes, exact same error output.

As of 5.7.1-2 for me:

LD .tmp_vmlinux.btf ld: drivers/thermal/cpufreq_cooling.o: in function cpufreq_set_cur_state': /opt/makepkg/build/linux-ck/src/linux-5.7.1/drivers/thermal/cpufreq_cooling.c:458: undefined reference toarch_set_thermal_pressure' BTF .btf.vmlinux.bin.o scripts/link-vmlinux.sh: line 114: 613105 Segmentation fault (core dumped) LLVM_OBJCOPY=${OBJCOPY} ${PAHOLE} -J ${1} LD .tmp_vmlinux.kallsyms1 .btf.vmlinux.bin.o: file not recognized: file format not recognized make: *** [Makefile:1117: vmlinux] Error 1 ==> ERROR: A failure occurred in build().

@graysky

Look at this. I had a similar error using BMQ. Con apparently skipped resync.

https://gitlab.com/alfredchen/linux-bmq/-/issues/6#note_353781623

@graysky yes, first I was going to mention here to add pahole as a dependency but as I was writing it down, you updated the package. Since then, I've been having the error you described.

Is anyone else getting this error:

  GEN     .version
  CHK     include/generated/compile.h
  LD      vmlinux.o
  MODPOST vmlinux.o
  MODINFO modules.builtin.modinfo
  GEN     modules.builtin
  LD      .tmp_vmlinux.btf
ld: drivers/thermal/cpufreq_cooling.o: in function `cpufreq_set_cur_state':
/scratch/linux-ck/src/linux-5.7.1/drivers/thermal/cpufreq_cooling.c:458: undefined reference to `arch_set_thermal_pressure'
  BTF     .btf.vmlinux.bin.o
tag__check_id_drift: subroutine_type id drift, core_id: 1129, btf_type_id: 1127, type_id_off: 0
libbpf: Unsupported BTF_KIND:0
btf_elf__encode: btf__new failed!
free(): double free detected in tcache 2
scripts/link-vmlinux.sh: line 114: 3291862 Aborted                 (core dumped) LLVM_OBJCOPY=${OBJCOPY} ${PAHOLE} -J ${1}
  LD      .tmp_vmlinux.kallsyms1
.btf.vmlinux.bin.o: file not recognized: file format not recognized
make: *** [Makefile:1117: vmlinux] Error 1

@stano00 - No, you are correct. Fixed in 5.6.8-2

Is the sphinx-workaround.patch actually being applied by your PKGBUILD? I didn't see it print while the other patches were applied.

@graysky @kittenlol I am aware of those two lines, since the change happened years(?) ago. Anyway, using #31 architecture and saying 'y' to P6_NOPs question have never given me any issue. Why?

Gets stuck at asking for X86_P6_NOP with _subarch=31

@kittenlol - See commented line 15 and 16.

Gets stuck at asking for X86_P6_NOP with _subarch=31

Fixed!

config checksum don't match

@graysky - `Merge branch '5.5/muqss' into 5.5/master' https://github.com/zen-kernel/zen-kernel/commit/82b455de1e967abf6ec5d75b5702a6d7376ddfc9

@redfang - It's in the output.

...
  25. Intel Skylake (MSKYLAKE)
  26. Intel Skylake X (MSKYLAKEX)
...

what option should i choose for intel skylate cpu (7th gen) during installation

@dwzg - Not out of date/no ck1. See: https://wiki.archlinux.org/index.php/Linux-ck#Release_cycle

@graysky @superice97 thx

@Bambus89 MuQSS uses subtick accounting and tickless expiry, and thus is unaffected by the kernel's tick rate. The only thing that would change by increasing the tick rate from 100 to 1000 would be a slight increase in power consumption and a slight decrease in throughput due to increased overhead. This is why the MuQSS patches automatically set 100Hz as the default. This is talked about in the MuQSS design doc (search for "Tickless expiry"): http://ck.kolivas.org/patches/muqss/sched-MuQSS.txt

@bambus89- ck patchset uses this value

Can you set a Option in pkgconfig for set 1000hz? Why is the kernel set to 100hz and are there benefits for 100hz?

Ah, you were asking for a build in [repo-ck]. I can add on with 5.4.7-1-ck.

@graysky - ah, yes I know I can build my own, was hoping to get a pre-build binary ;-)

Happy to contribute cpu time if that's needed for repo-ck

@trthomps - You can ... it is option 12. See the PKGBUILD.

Would it be possible to add a zen2 build as well? I'm running zen successfully but it looks like amd added additional optimizations for the latest zen 2 arch.

[travis@travis-pc ~]$ gcc -c -Q -march=native --help=target | grep march
  -march=                           znver2
  Known valid arguments for -march= option:

can't add comment a long comment... even if if i code put it in code code

@blazessdd - Can you post a build log and link it here?

i have set _localmodcfg=y but whenever i try to compile modprobed wont show up and i get a kernel with all the modules compile what can i do?

@Althorion -

# Note - the march=native option is unavailable by this method, use the nconfig
# and manually select it.

Tired with clean PKGBUILD and -march=native (option 31). Manual patching works just fine, no idea why.

BTW (unrelated problem, I think), I tried manually patching the sources, removing patching from the PKGBUILD, and when I put some value for _subarch=, it fails at make oldconfig—it tries giving it the value I put in _subarch= (that would be 31), thus entering the never-ending loop of wrong arguments.

Finally, I managed to successfully build the package by manual patching and giving it empty _subarch.

Need more to go on in order to reproduce. Any edits to the PKGBUILD? Which number are you selecting on the gcc patch stage?

No, I manually cloned the package (git clone <https://aur.archlinux.org/linux-ck.git>) and run makepkg -i on it. I tried removing the folder and re-cloning and clearing the build folder (rm -rf /tmp/makepkg/linux-ck), but to no avail.

How are you building? Works fine here. If you're using an AUR helper, DON'T USE AN AUR HELPER!

...
patching file kernel/time/Kconfig
patching file kernel/time/clockevents.c
patching file kernel/time/hrtimer.c
patching file kernel/time/posix-cpu-timers.c
patching file kernel/time/timer.c
patching file kernel/trace/trace_selftest.c
patching file mm/vmscan.c
patching file net/core/pktgen.c
patching file sound/pci/maestro3.c
patching file sound/soc/codecs/rt5631.c
patching file sound/soc/codecs/wm8350.c
patching file sound/soc/codecs/wm8900.c
patching file sound/soc/codecs/wm9713.c
patching file sound/soc/soc-dapm.c
patching file sound/usb/line6/pcm.c
  HOSTCC  scripts/basic/fixdep
  HOSTCC  scripts/kconfig/conf.o
  HOSTCC  scripts/kconfig/confdata.o
...

Version 5.4-3 fails at patching: https://gist.github.com/Althorion/a817b8bd5027cf9344af52963d063951

Last few lines:

patching file mm/vmscan.c
patching file net/core/pktgen.c
patching file sound/pci/maestro3.c
patching file sound/soc/codecs/rt5631.c
patching file sound/soc/codecs/wm8350.c
patching file sound/soc/codecs/wm8900.c
patching file sound/soc/codecs/wm9713.c
patching file sound/soc/soc-dapm.c
patching file sound/usb/line6/pcm.c
?[1m    Aborting...?(B?[m

No idea what pikaur is but if you are using an i686 fork of Arch, see man makepkg to see how to ignore the architecture or modify the PKGBUILD itself.

How can I build this on a 32bit system? I tried it now with pikaur, but I'm getting "linux-ck, linux-ck-headers can't be built on the current arch (pentium4). Supported: x86_64"

I know Arch abandoned 32 bit, but this is a source package so I expected it could be built.

@kyak - Must have missed that when diffing official PKGBUILD vs this. Modifycation will be included in 5.3.11, thanks.

Hi graysky,

Official kernels do now install kernel image to /boot anymore: https://www.archlinux.org/news/new-kernel-packages-and-mkinitcpio-hooks/

Do you plan to follow the same path with linux-ck?

Fixed! @sir_lucjan thank you, I tried again and now it compiles flawless. Maybe the reboot fixed the issue, since I upgraded some packages, including mkinitcpio, systemd and gcc...

mkinitcpio is not involved in the compilation of the kernel at all. It is used only to generate initramfs and it is also possible to do it with dracut.

With latest mkinitcpio-27-2, linux-ck compilation (in real it stops before...) stucks just after patching. No errors printed, just "failed in prepare()"

@toki1990

It's not a linux-ck problem, it's Manjaro. Due to recent updates in Arch Linux (kmod and mkinitcpio) hooks are no longer provided in kernel only in kmod (depmod-hook) and mkinitcpio (mkinitcpio-install and mkinitcpio remove). Unfortunately, at the moment, as a Manjaro user you will not install most of the current AUR kernels and the problem lies solely with your distribution, i.e. Manjaro.

I'm manjaro stable repo user. linux-ck 5.3.8-4 not booting. I'm tried linux-slim 5.3.8-2 custom kernel same problem. Xanmod 5.3.8-1 not doing this. Because problem is this:

"Updating linux-xanmod initcpios... ==> Building image from preset: /etc/mkinitcpio.d/linux-xanmod.preset: 'default' -> -k /boot/vmlinuz-linux-xanmod -c /etc/mkinitcpio.conf -g /boot/initramfs-linux-xanmod.img ==> Starting build: 5.3.8-xanmod6-1-xanmod -> Running build hook: [base] -> Running build hook: [udev] -> Running build hook: [autodetect]...."

Xanmod custom kernel doing this last of compilation. Linux-ck 5.3.8-4 not doing this.

@Superice - Yes, I am ignoring that particular change. For example, the substitution will make _srcver=5.3.8-arch1 in this case whereas this package will be 5.3.8-1-ck which is fine for me.

The pkgver and _srcver variables changed in the main linux package:

pkgver=5.3.8.1
_srcver=${pkgver%.*}-arch${pkgver##*.}

@sir_lucjan - Good catch, thanks. Fixed in 5.3.8-4.

@graysky

As far as I can see, you forgot to change one line.

-  echo "$_kernelname" > localversion.20-pkgname
+  echo "${pkgbase#linux}" > localversion.20-pkgname

@graysky,

It was interesting to see how much 'heat' my wireguard-dkms post generated, and perhaps several of us wireguard-dkms and -ck users have learned some important tips.

Bottom line = tldr version - no action is required provided that: the fou, ip_tunnel and wire modules are found within lsmod and/or modprobed-db list if you run modprobed-db.

Those 3 modules were NOT loaded within MY modprobed-db list, which was quite surprising, given that I have run wireguard-dkms and modprobed-db for a VERY LONG time!

However, once I re-ran: modprobed-db store, it found the three new modules.

Of course, during a subsequent, clean, rebuild of linux-ck, wireguard-dkms was successfully installed.

Thanks for your input graysky, and to the others who offered their comments.

@kwe: OK. I'll try to explain.

1. localmodconfig disables every module, that not exactly listed in provided file

2. the file modprobed.db filling automatically only by modules that are in memory when modprobed-db service runs.

3. fou module did not get into the modprobed.db for some reason

4. But it is needed as indirect dependency for wireguard-dkms

5. when you build kernel with localmodconfig - fou module become disabled

6. And consequent build of wireguard-dkms fails.

As I can see - the most simple way to fix such a problem - is to add fou/fou6 modules into modprobed.db manually.

Or - load them and run modprobed-db.service for the same thing.

Blame me, if I wrong, cause there is more coffee than blood in veins)

@laenco: Either I'm misunderstanding you or I have to disagree.
When digging though lsmod AFTER HAVING USED WIREGUARD (so I assume all the required modules have been loaded), I can't seem to find any specific modules that are NOT in modprobed.db. It seems to me like the part where .config is configured for the kernel compilation process is assembled in a wrong way, missing out on the loaded (!) modules. This doesn't seem to have anything to do with modprobed-db as far as I can tell (?).

@kwe: so it's look like wireguard needs fou module for dkms build/install, but does not load it during routine operations, so modprobed-db did not detect it.

You could try to add fou and fou6 modules to modprobed.db manually and rebuild.

@laenco: Here is my current system state:

$ grep -i fou ~/.config/modprobed.db
$ lsmod | grep -i fou
$ lsmod | grep -i wire
wireguard             233472  0
ip6_udp_tunnel         16384  1 wireguard
udp_tunnel             16384  1 wireguard
$ grep -i tun ~/.config/modprobed.db
ip6_udp_tunnel
tun
udp_tunnel

When I run makepkg to extract the linux-ck src directory, I end up with this .config diff against the current running kernel config:

$ zdiff /proc/config.gz config.last 
1069d1068
< CONFIG_NET_IP_TUNNEL=m
1077,1078c1076
< CONFIG_NET_UDP_TUNNEL=m
< CONFIG_NET_FOU=m
---
> # CONFIG_NET_FOU is not set
1116d1113
< CONFIG_IPV6_FOU=m

@kwe could you check ~/.config/modprobed.db for presence of "fou"/"fou6" modules?

The question is "If localmodconfig is unable to detect fou or is it a modprobed-db detection problem"?.

@graysky I am building linux-ck with modprobed-db, and for some reason, I always end up with those configuration options being disabled. I have to re-enable them manually every single time.

% uname -r
5.3.8-1-ck

% zgrep CONFIG_NET_FOU /proc/config.gz
CONFIG_NET_FOU=m
CONFIG_NET_FOU_IP_TUNNELS=y

Should be there as a module. Building wireguard-dkms on CK works for me:

# pacman -S wireguard-dkms
resolving dependencies...
looking for conflicting packages...

Packages (2) dkms-2.7.1-1  wireguard-dkms-0.0.20191012-1

Total Download Size:   0.24 MiB
Total Installed Size:  1.75 MiB

:: Proceed with installation? [Y/n] 
:: Retrieving packages...
 wireguard-dkms-0.0.20191012-1-x86_64            245.7 KiB  3.43 MiB/s 00:00 [############################################] 100%
(2/2) checking keys in keyring                                               [############################################] 100%
(2/2) checking package integrity                                             [############################################] 100%
(2/2) loading package files                                                  [############################################] 100%
(2/2) checking for file conflicts                                            [############################################] 100%
:: Processing package changes...
(1/2) installing dkms                                                        [############################################] 100%
Optional dependencies for dkms
    linux-headers: build modules against the Arch kernel [installed]
    linux-lts-headers: build modules against the LTS kernel
    linux-zen-headers: build modules against the ZEN kernel
    linux-hardened-headers: build modules against the HARDENED kernel
(2/2) installing wireguard-dkms                                              [############################################] 100%
:: Running post-transaction hooks...
(1/2) Arming ConditionNeedsUpdate...
(2/2) Install DKMS modules
==> dkms install wireguard/0.0.20191012 -k 5.3.7-arch1-2-ARCH
==> dkms install wireguard/0.0.20191012 -k 5.3.8-1-ck

FWIW, I'm also setting that option manually, hence I'd support @air-g4p's request. Thx again for maintaining this btw !

Cheers !

@graysky - wireguard-dkms failed to build on CK with this output:

http://ix.io/20kT

I asked the devs at #wireguard about this failure, and was told I needed to enable: CONFIG_NET_FOU

(FOU = foo over udp)

After enabling that and rebuilding linux-ck --> Success.

Per Jason's (zx2c4), the lead wireguard dev, suggestion, can you please enable CONFIG_NET_FOU by default on linux-ck?

Thanks graysky

@timo - OK... that is likely because your .config got tweaked due to your CC= line. I opened a flyspray shown below. See it for a solution: https://bugs.archlinux.org/task/64315

@graysky What I did was add CC=clang before make in the PKGBUILD to use Clang as compiler.

What worked was using makepkg for compiling till it gets stuck, and then end it with make modules_install and make install from the src folder. Doesn't give a package for pacman, but works.

I've tried linux-pf the same way, though that got stuck at the same place :)

@simona - the recent failure of dkms is due to the fact that there are both a newer version of the toolchain and of the dkms packages in [testing] currently. The advice I have been given is that it always a good idea to build the kernel against [testing] to prevent version mismatch errors when toolchain packages come out of [testing] and into [core].

I need to figure out what the best option is now.

@timo_capa - I am not sure how you go about building in Clang... I am using the official Arch devtools for clean chroot building. Are you patching the kernel source with ck1?

==> dkms install vboxhost/6.0.14_OSE -k 5.3.7-1-ck-skylake Error! Bad return status for module build on kernel: 5.3.7-1-ck-skylake (x86_64)

I'm not sure if this is something you're interested in fixing, but compiling the kernel with Clang gets stuck at Installing boot image.... I'm assuming it's spamming something related to Kconfigs in the background related to GCC, as cancelling the build shows

"install: cannot stat '*'$'\n''* Restart config...'$'\n''*'$'\n''*'$'\n''* GCC plugins'$'\n''*'$'\n''Generate some entropy during boot and runtime (GCC_PLUGIN_LATENT_ENTROPY) [N/y/?] (NEW) ': No such file or directory"

E: Manually compiling a kernel from the tarball from kernel.org works and boots.

For version 5.2.21, I think applying the sed expression sed -i -re 's/(CFLAGS[[:space:]]+)\+=/\1:=/' to the ck stuff prior to patching will avoid a patch failure in tools/objtool/Makefile.

I'm getting DKMS errors when upgrading to linux-ck-ivybridge-5.2.19-1. On step (3/4) Install DKMS modules: Error! Bad return status for module build on kernel: 5.2.19-1-ck-ivybridge (x86_64).

It seems initcpio image can't build with any DKMS modules over ck-ivybridge-5.2.19-1

I'm using nvidia 435.21 dkms and vboxhost 6.0.12 dkms.

In make.logs I see error: incompatible gcc/plugin versions

I'm using gcc (9.1.0-2 updated: 2019-06-25)

Edit: There literally was just an update for gcc a few hours after I posted this. All solved.

It is building successfully now.

Please try 5.2.14-4

The addition of fix_patch-5.2-ck1_for_systemd-detect-virt_containers.patch seems to cause these last 3 lines to be dropped from kernel/sched/MuQSS.c even though they are in patch-5.2-ck1 which causes a build failure.

"#endif / CONFIG_CGROUP_SCHED /"

"#undef CREATE_TRACE_POINTS"

@graysky - It now appears the issue is exclusively related to modprobed-db. My latest build worked well in a clean chroot with ccache enabled, but I used an empty _localmodcfg= statement. Of course, the build time was quite lengthy, but it worked.

Any thoughts on a workaround, or a permanent fix?

Thanks for your help graysky....

@air-g4p - I do not use ccache so I cannot comment. I will say that builds using distcc on 5.2 fail where they worked fine with 5.1. Related? The change in the chroot build might be due to the modprobed-db change I made wherein the old behavior was to call it in recall mode via sudo which was always hacky in a PKGBUILD. The current method reads the database directly which the build scripts might not allow since your homedir is outside of the build root. Try building in the chroot without using the modprobed-db enabled; that will probably work.

@graysky - built without issue with ccache disabled, while running makepkg -sri so what is driving that fail?

Also, why did -ck previously build (many times) without issue in a clean chroot, but now it fails?

Thanks...

@air-g4p - Disable ccache and try again

@graysky,

In a clean folder, and running makepkg -sri the build failed, but in a different way. I got as far as:

-> Applying patch 0004-drm-amdgpu-pin-the-csb-buffer-on-hw-init-for-gfx-v8.patch... patching file drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c -> Setting config... -> Applying enable_additional_cpu_optimizations_for_gcc_v9.1+_kernel_v4.13+.patch ... patching file arch/x86/include/asm/module.h patching file arch/x86/Kconfig.cpu patching file arch/x86/Makefile patching file arch/x86/Makefile_32.cpu HOSTCC scripts/basic/fixdep /usr/bin/ccache: invalid option -- 'W'

then I also saw:

make[1]: [scripts/Makefile.host:90: scripts/basic/fixdep] Error 1 make: [Makefile:508: scripts_basic] Error 2 ==> ERROR: A failure occurred in prepare(). Aborting...

Any ideas? Thanks

@air-g4p - Does it build outside of the chroot? I'm thinking the chroot build script doesn't look outside of that environment and will therefore fail.

@graysky,

I'm seeing several inexplicable errors with the latest build:

When trying to build ck with time extra-x86_64-build in a clean chroot, I get as far as:

Kernel Live Patching (LIVEPATCH) [N/y/?] n

configuration written to .config

-> No modprobed.db data found /startdir/PKGBUILD: line 157: exit1: command not found ==> ERROR: A failure occurred in prepare(). Aborting... ==> ERROR: Build failed, check /var/lib/archbuild/extra-x86_64/user/build

this DESPITE: modprobed-db store showing:

180 modules currently loaded per /proc/modules 180 modules are in /home/user/.config/modprobed.db

AND the correct path in PKGBUILD:

_localmodcfg=/home/user/.config/modprobed.db

Also, please have a look at Line 157 of PKGBUILD.

Did you intend to type exit as opposed to your current exit1?

I have never seen this modprobed error previously.

Ideas are welcome...

I haven't made an ISO/no idea. You could ask on CK's blog why it fails with that mount command. I do not edit beyond the changes his patchset make.

@graysky archiso failing to build appears to be an issue with the CK kernel. No, idea what's different from Arch, as the only thing I changed was the compression algorithm used for initramfs.

Have any idea what might be causing the mount failure with CK?

https://bugs.archlinux.org/task/63508

@graysky This is a forum post I made on the issue. It looks like folders under the work directory are not able to be identified as mount points. If you install archiso, and just run sudo cp -r /usr/share/archiso/configs/releng/ ~/archlive; cd ~/archlive; sudo ./build.sh -v The issue should reproduce itself.

I just happened to decide to try the Arch kernel to see if that resolved the build failure I was receiving, and everything just built successfully. Looks like others have identified this issue in the past, but did not report it. I have a feeling it has something to do with mkarchiso.

https://bbs.archlinux.org/viewtopic.php?id=248473

https://unix.stackexchange.com/questions/460043/how-can-i-successfully-build-an-archiso-image-airootfs-is-not-a-mountpoint

You have provided nothing to go on...

I'm having issues getting archiso to successfully build an image when running the CK kernel. Everything builds fine with the default kernel. Is anyone else having this issue? Could it be from disabling initramfs support, other than LZ4? Is this a CK issue or archiso issue?

@charveey thx but fix pkgbuild please ;D

Will the AMD Llano processors family be added to the kernel?

@rzax666 if you're interested you can just install my package, linux-bootsplash which provides the kernel with framebuffer bootsplash support. I personally run the 5.2.5 ck kernel with the patch you want. I got a hardware issue on my computer so I can't test the latest version (5.2.6)

please add this patch - https://github.com/philmmanjaro/linux-bootsplash

https://gitlab.manjaro.org/packages/core/linux52

Yeah, I think you shouldn't use native in non-interactive mode. See the comments:

Note - the march=native option is unavailable by this method, use the and manually select it.

Since switching from _subarch=20 (MIVYBRIDGE) to _subarch=28 (MNATIVE) in the PKGBUILD, it won't build:

  24. Intel Skylake X (MSKYLAKEX) (NEW)  
  25. Intel Cannon Lake (MCANNONLAKE) (NEW)  
  26. Intel Ice Lake (MICELAKE) (NEW)  
> 27. Generic-x86-64 (GENERIC_CPU)  
  28. Native optimizations autodetected by GCC (MNATIVE) (NEW)  
choice[1-28?]: 28  
Support for P6_NOPs on Intel chips (X86_P6_NOP) [N/y/?] (NEW) 28  
Support for P6_NOPs on Intel chips (X86_P6_NOP) [N/y/?] (NEW) 28  
Support for P6_NOPs on Intel chips (X86_P6_NOP) [N/y/?] (NEW) 28  
Support for P6_NOPs on Intel chips (X86_P6_NOP) [N/y/?] (NEW) 28  
Support for P6_NOPs on Intel chips (X86_P6_NOP) [N/y/?] (NEW) 28  
Support for P6_NOPs on Intel chips (X86_P6_NOP) [N/y/?] (NEW) 28  

The last message keeps looping forever.

When I switch back to the default _subarch= with no value, the process/interaction seems suboptimal:
- it asks me choice[1-28?]:
- I choose 28
- it asks me Support for P6_NOPs on Intel chips (X86_P6_NOP) [N/y/?] (NEW)
- I choose y
- it does the modprobed-db recall (as configured)
- it asks me BT8XX GPIO abuser (GPIO_BT8XX) [N/m/y/?] (NEW)
- I choose n (why does this ask me every time?)
- it asks me Studio Evolution SE6X (SND_SE6X) [N/m/?] (NEW)
- I choose n (again, why does this ask me every time?)
- ==> Sources are ready.

@graysky - Thanks, and understood.

@zerophase - Thanks, what you said made sense, and got me thinking about my chroot. As I had been using makechrootpkg of late to build linux-ck, I completely forgot that command neither cleans, nor updates a chroot.

As soon as I built linux-ck with my typical extra-x86_64-build my chroot got cleaned and updated, which resolved all the DKMS incompatible gcc versions installation issues.

@air-g4p if you can rollback to the last good kernel you should be able to install. The last time I had dkms issues was from updating the kernel multiple times without rebooting. On the second kernel update I usually receive a bunch of errors about the version of the kernel dkms is looking for does not exist. Pretty sure that's what is happening to you too.

I do not use dkms so I wouldn't know where to start with it.

@graysky - Saw these DKMS installation errors today after building -ck on my fully updated system:

(2/6) Install DKMS modules ==> dkms install bbswitch/0.8 -k 5.2.4-1-ck Error! Bad return status for module build on kernel: 5.2.4-1-ck (x86_64) Consult /var/lib/dkms/bbswitch/0.8/build/make.log for more information. ==> dkms install vboxhost/6.0.10_OSE -k 5.2.4-1-ck Error! Bad return status for module build on kernel: 5.2.4-1-ck (x86_64) Consult /var/lib/dkms/vboxhost/6.0.10_OSE/build/make.log for more information. ==> dkms install nvidia/430.34 -k 5.2.4-1-ck Error! Bad return status for module build on kernel: 5.2.4-1-ck (x86_64) Consult /var/lib/dkms/nvidia/430.34/build/make.log for more information. ==> dkms install wireguard/0.0.20190702 -k 5.2.4-1-ck Error! Bad return status for module build on kernel: 5.2.4-1-ck (x86_64) Consult /var/lib/dkms/wireguard/0.0.20190702/build/make.log for more information.

My /build/make.logs show multiple instances of:

cc1: error: incompatible gcc/plugin versions cc1: error: fail to initialize plugin ./scripts/gcc-plugins/structleak_plugin.so

Is there a workaround, or a solution?

@SuperIce - Right you are thanks, fixed.

@graysky The 26 should be changed to 27 in this comment in the PKGBUILD: https://aur.archlinux.org/cgit/aur.git/tree/PKGBUILD?h=linux-ck#n15

great!

@nTia89 - Yes, but was waiting for 5.2.3 to incorporate that and the other 2 patches. See: https://aur.archlinux.org/cgit/aur.git/commit/?h=linux-ck&id=c4e43c48637e

@graysky Don't you follow upstream changes about CONFIG_GCC_PLUGIN_STACKLEAK[1]?

[1] https://git.archlinux.org/svntogit/packages.git/commit/trunk?h=packages/linux&id=8c1be2f6cdd6b3fa1b9d06dfd8791dc580ebe364

5.2 patches are out now: http://ck.kolivas.org/patches/5.0/5.2/5.2-ck1/

@nivekuil It shouldn't. I checked the config, and "CONFIG_USER_NS_UNPRIVILEGED" is enabled.

Also see: https://aur.archlinux.org/cgit/aur.git/tree/config?h=linux-ck#n166

Does this linux-zen bug also apply here? https://bugs.archlinux.org/task/62990

@freenes - if the unpatched kernel doesn't work there is little likelihood that the MuQSS scheduler will fix it. You need to find out what's breaking it.

@graysky @zerophase Thank you for advises. I tried 5.1.8-arch, and updated bios(my laptop model is ThinkPad T430u, bios is updated to Version: H6ETA0WW 2.18 Release Date: 06/01/2018), neither working.

@graysky The version from the aur seems to have that feature disabled by default.

@gee - I try to keep linux-ck as close to the official ARCH kernel as I can. I do not know what negative effect changing that parameter might have for other users. Recommend you build from the AUR yourself modifying it as you see fit.

Hello again,

I'll retry my last request in an easier way if you don't mind graysky.

Would you mind using 'N' for Linux kconfig's RAID6_PQ_BENCHMARK? I have not been able to find a way to do so without altering the kconfig file yet, and since I grab the kernels from your repos, it wouldn't work that way.

Setting it to 'N' save a few seconds on boot time, which I think is quite nice.

@freenestor if that does not work try updating your bios / uefi. I would also post more system specs. As a last, resort you could try doing a clean install just to rule out configs breaking from waiting awhile to upgrade.

I think the issue probably has something to do with the motherboard, though. If nothing works try to use the warranty for a replacement. A new version of the same mobo might just work.

@freenestor - can you boot 5.1.x-arch without the parameter? Rule out ck1 patchset as the cause in other words.

I cannot boot linux-ck 5.1.x without nolapic kernel parameter, and with nolapic parameter the performance is worse. I can boot linux-ck 4.19.x normally. Can anyone give some advises?

@ismet - Something happened with the rsync mirror. Fixing now, please try in 5 min or so.

Hey graysky, pre-compiled linux-ck-zen package cannot be downloaded from repo-ck repositories.

error: failed retrieving file 'linux-ck-zen-5.1.9-2-x86_64.pkg.tar.zst' from repo-ck.com : The requested URL returned error: 404

Shame. Seems like a useful option.

@Althorin - Purely archival.

The PKGBUILD saves config.last for later, but never does anything with it (i.e., giving the option to use one as a basis for compilation). Is there something missing?

@galaxy0419 - Not out of date, https://wiki.archlinux.org/index.php/Linux-ck#Release_cycle

@gatlinnewhouse

I fixed the issue, it seems there was a problem with my makepkg.cfg

trying to compile this now causes my computer to freeze, drop out of the x server and go back to the vt login screen (and my configs automatically log me in) and just sit there, fans spinning inconsistently

@graysky

You could rebase http://ck.kolivas.org/patches/5.0/5.0/5.0-ck1/patches/0002-Fix-Werror-build-failure-in-tools.patch against 5.0.6

CK1 needs to be tweaked to patch into 5.0.6 it seems...

patching file tools/objtool/Makefile
Hunk #1 FAILED at 31.
1 out of 1 hunk FAILED -- saving rejects to file tools/objtool/Makefile.rej

Can someone post to CK's blog?

Hello graysky, I use to patch the local copy of the PKGBUILD to allow the currently running config (/proc/config) to be used as base config so maybe it can be useful to merge it ?

The patchset for 5.0 is live: http://ck.kolivas.org/patches/5.0/5.0/5.0-ck1/

He is just too busy to officially announce: http://ck-hack.blogspot.com/2018/12/linux-420-ck1-muqss-version-0185-for.html?m=1#comment-form

Great.

Thx for researching and fixing so quickly ! PKGREL 2 built fine. (EDIT without changing any configs)

Seems like the ck patchset needs a tweak: http://ck-hack.blogspot.com/2018/12/linux-420-ck1-muqss-version-0185-for.html?showComment=1550063841657#c3680780559065276989

Testing now and will push 4.20.10-2 if successful.

Looks like I could build after setting CONFIG_KVM and CONFIG_KVM_INTEL from <m> to [y]. I'm building for Haswell architecture btw.

Yea it tricked me into thinking it built find until I found 0 packages built from the batch... currently reviewing build log to see wtf.

Yep, me too :

ERROR: "sched_smt_present" [arch/x86/kvm/kvm-intel.ko] undefined!                                                                                                                                      
make[1]: *** [scripts/Makefile.modpost:92: __modpost] Error 1                                                                                                                                          
make: *** [Makefile:1275: modules] Error 2                                                                                                                                                             
==> ERROR: A failure occurred in build().                                                                                                                                                              
    Aborting...                                                                                                                                                                                        
==> Removing installed dependencies...                                                                                                                                                                 
checking dependencies...                                                                                                                                                                               

Package (2)  Old Version  Net Change                                                                                                                                                                   

bc           1.07.1-3      -0.17 MiB                                                                                                                                                                   
inetutils    1.9.4-7       -1.12 MiB                                                                                                                                                                   

Total Removed Size:  1.29 MiB                                                                                                                                                                          

:: Do you want to remove these packages? [Y/n]                                                                                                                                                         
:: Running pre-transaction hooks...                                                                                                                                                                    
(1/1) Removing old entries from the info directory file...                                                                                                                                             
:: Processing package changes...                                                                                                                                                                       
removing inetutils...                                                                                                                                                                                  
removing bc...                                                                                                                                                                                         
:: Running post-transaction hooks...                                                                                                                                                                   
(1/2) Reloading system manager configuration...                                                                                                                                                        
  Skipped: Current root is not booted.                                                                                                                                                                 
(2/2) Arming ConditionNeedsUpdate...                                                                                                                                                                   
==> ERROR: Build failed, check /var/lib/buildroot/x86_64/kriss/build

Is 4.20.10 giving anyone else compile errors?

It probably is out of focus, but it would be really nice to add this patchset: https://lists.ubuntu.com/archives/kernel-team/2019-January/098018.html

My boot takes 2 sec to test various raid6 algorithms, when I don't even use raid... Being able to specify the algorithm to use would shave that time off and be awesome.

@graysky,

Thanks for pointing out my non-Arch standard settings and I forgot there was a new /etc/makepkg.conf.pacnew hanging about. I'm back to standard now.

I see you removed 'subarch 27' from PKGBUILD, and now as expected, linux-ck-4.20.3-1-x86_64.pkg.tar.xz and its headers build without issue.

@air-g4p - As well, your CFLAGS are not the Arch standard... drop -fstack-protector-strong per the new makepkg.conf in pacman-5.1.2-2.

@air-g4p - See the comments in the PKGBUILD and the discussion here in the AUR.

<h1>Note - the march=native option is unavailable by this method, use the nconfig and manually select it.</h1>

In short, I do not know how to allow the native option to get set via the method used for the other discretely defined options.

@graysky,

I'm running an i7 8550U.

Under both /etc/makepkg.conf and ~/chroot/root/etc/makepkg.conf, I have these identical settings:

ARCHITECTURE, COMPILE FLAGS

CARCH="x86_64" CHOST="x86_64-pc-linux-gnu"

-- Compiler and Linker Flags

CPPFLAGS="-D_FORTIFY_SOURCE=2" CFLAGS="-march=native -O2 -pipe -fstack-protector-strong -fno-plt" CXXFLAGS="-march=native -O2 -pipe -fstack-protector-strong -fno-plt" LDFLAGS="-Wl,-O1,--sort-common,--as-needed,-z,relro,-z,now"

-- Make Flags: change this for DistCC/SMP systems

MAKEFLAGS="-j8"

-- Debugging flags

DEBUG_CFLAGS="-g -fvar-tracking-assignments" DEBUG_CXXFLAGS="-g -fvar-tracking-assignments"

Is there anything you see in my setup that would drive:

'The build fails with hundreds of instances of:

Support for P6_NOPs on Intel chips (X86_P6_NOP) [N/y/?] (NEW) 27'

that I mentioned when '_subarch=27' is set?

I normally build linux-ck in a clean chroot without issue, but this infinite loop issue disturbs me.

I remember looking at this a while ago and concluding it isn't possible... prove me wrong :)

Hi graysky,

I can confirm the build loop issue zoopp mentioned last month is still present in linux-ck 4.20.1-1.

With this PKGBUILD option set:

27. Native optimizations autodetected by GCC (MNATIVE)

_subarch=27

The build fails with hundreds of instances of:

Support for P6_NOPs on Intel chips (X86_P6_NOP) [N/y/?] (NEW) 27

It appears to me that the only way out of this loop is crtl-c.

Of course, the build proceeds normally with any other appropriate subarch assigned.

@artafinde - ck1 disables NUMA

@vltr: try enabling NUMA on ck.

@marceliq ohhh ... I haven't seen that. And that's actually a real bummer :(

In patch-4.20-ck1 there is:

@@ -896,6 +913,7 @@ config CGROUP_DEVICE

 config CGROUP_CPUACCT
        bool "Simple CPU accounting controller"
+       depends on !SCHED_MUQSS
        help
          Provides a simple controller for monitoring the
          total CPU consumed by the tasks in a cgroup.

which disable CPUACCT if SCHED_MUQSS is enabled :(

Thanks, @graysky! Yeah, I tried to revert some of the changes - docker stopped working after 4.19.4 or 4.19.5 (from repo-ck), but I just got some time to find the source of the error now. I thought it was CONFIG_CGROUP_CPUACCT at first because of error given by docker:

$ docker run --rm hello-world                        
docker: Error response from daemon: unable to find "cpuacct" in controller set: unknown.

That's just an example. Any other docker command (build, run, etc) would eventually ran into this issue (although the docker daemon seem to work normally).

This is something that doesn't happen in the stock kernel (and I make use of testing repositories as well). Anyway, I reverted not only that flag but some others as well (I actually did the diff myself earlier), but docker is still a no go. I'm not that expert in the kernel to go further and analyze what's wrong, unfortunately. I wanted to share my experience here in case anyone bumps into the same problem.

Would you mind providing me CK's blog so I can post there as well?

Thanks a lot!

EDIT

Just found out the issue, pointed out by @marceliq and then CK's blog and comments regarding this issue as well as the forum post related. Thanks!

@vltr - It would appear the ck1 patchset does this, see here for a diff of the Arch config before and after simply patching with CK1: https://github.com/graysky2/4.20.1_temp/commit/82d2aee40035766d2e225e4b2a3db36faa187273

Recommend you post to CK's blog referencing that diff and your problem. Might be that he did not mean to do that if I am correct.

Hello, @graysky! Just one question: is there any reason why the CONFIG_CGROUP_CPUACCT flag is not set anymore on linux-ck? Thanks a lot!

@saren - sorry about that... fixed now but did not bump the pkgver.

Hi, since 4.20-ck NUMA is disabled regardless of _NUMAdisable. $ numactl -s shows No NUMA support available on this system.

I usually build with native optimisations detected by GCC however there's a problem if I assign anything to the _subarch variable. Early during the build this question pops up:

Support for P6_NOPs on Intel chips (X86_P6_NOP) [N/y/?] (NEW)

.. which is answered with the value of _subarch variable (27 in my case).

Since this isn't an accepted answer, the question loops endlessly.

Not really a big issue but I thought I should mention it.

@graysky:

Same here: Built linux-ck 4.19.8-1 without issue yesterday, and all seems to be running well and as expected since then.

@graysky:

On first look, it seems to work fine for me. I built and booted into this 4.19.8 version a few hours ago.

$ uname -r
4.19.8-1-ck

$ uptime
 04:05:41 up  5:57,  1 user,  load average: 0.32, 1.34, 1.90

I don't see anything unusual happening in the journal. Programs, games and the desktop don't seem to misbehave.

I heard from CK that it is unlikely he will be able to unfuck 4.19-ck1 before the release of 4.20. Fortunately, a user on the blog posted two patches which seem to do just that. I have adapted them to the PKGBUILD for 4.19.7 and for 4.19.8. I have been testing 4.19.7-ck1 and it has been running without issues. I am building 4.19.8-ck1 now. I pushed 4.19.8 to the AUR just now and am requesting some user feedback on stability and performance. I will build the repo set once we some confirmation that all is well. Thanks.

@lukpod - Not out of date, see pinned comment:

Even though 4.19.7 has been released, several upstream changes therein break ck1. Please do not flag out-of-date until Con releases an update.

Even though 4.19.7 has been released, several upstream changes therein break ck1. Please do not flag out-of-date until Con releases an update.

@FarStar - False alarm... I made a mistake in my config when building. No issues with 4.19.6.

Hello graysky,

What kind of problems? I have slowdowns at times but I don't know if it's related.

Anyone else experiencing problems relating to networking under 4.19.6?

I need to clear yay's cache after every major kernel update.

Highly discourage users from AUR helpers. 9 times out of 10, problems with builds are the apfsult of the helper.

@graysky Found the issue. I need to clear yay's cache after every major kernel update.

I don't think this is something with linux-ck but with how you configured/compiled/whatever it. Try the control experiment: build it straight-up and see if it works.

@graysky I think it might be since I've modified my initramfs to the minimal size.

I could have sworn some of the usb subsystems were added directly to the kernel this release. I just don't want to reboot and find I have an unbootable kernel.

Any idea what to do about modules not being found?

I get no such errors

When running mkinitcpio, on 4.19.2, I'm getting errors over ehci_pci, xhci_pci, usbhid modules not being found. Have those modules been renamed?

@graysky: wiki for modprobed-db mention to call the localmodconfig as below:

make LSMOD=$HOME/.config/modprobed.db localmodconfig 

What's the difference of the two methods?

Anyone tried unofficial patch submittied in ck-hack comments? https://jki.io/ck-4.19.patch.bz2

Could be tampered, I haven't tested yet (and not advocating this package use it).

Here's patch-4-18.ck1.xz

https://cloud.disroot.org/s/niPyHzJW9wy7XoQ

Anyone have a mirror of patch-4.18-ck1.xz?

Meh, ck.kolivas.org seems down - can't download patch

I'm getting an error while building this, although it's unclear what it is. All the previously logged patching steps are either skipped or success. Here's just the last bit:

patching file sound/soc/soc-dapm.c patching file sound/usb/line6/pcm.c patching file tools/objtool/Makefile ==> ERROR: A failure occurred in prepare(). Aborting...

I simply clone this AUR link, modify PKGBUILD to "_makeenconfig=y" and "_NUMAdisable=", then "makepkg -sri". Changing "_subarch=" has no effect.

I checked that last Makefile that's patched and it's indeed patched. I think the problem is the next line in the PKGBUILD, "local src"

@graysky https://bugs.archlinux.org/task/60272 - seems to be a known issue, thanks for looking into it. :)

@ELMastro - Do not build with an AUR helper. Use makepkg.

After [code] INSTALL sound/usb/line6/snd-usb-line6.ko INSTALL sound/usb/line6/snd-usb-pod.ko INSTALL sound/usb/line6/snd-usb-podhd.ko INSTALL sound/usb/line6/snd-usb-toneport.ko INSTALL sound/usb/line6/snd-usb-variax.ko INSTALL sound/usb/misc/snd-ua101.ko INSTALL sound/usb/snd-usb-audio.ko INSTALL sound/usb/snd-usbmidi-lib.ko INSTALL sound/usb/usx2y/snd-usb-us122l.ko INSTALL sound/usb/usx2y/snd-usb-usx2y.ko INSTALL virt/lib/irqbypass.ko DEPMOD 4.18.12-1-ck-ck rm: [/code] mine complains it can't remove '/tmp/packerbuild-1001/linux-ck/linux-ck/pkg/linux-ck/usr/lib/modules/4.18.12-1-ck/source' Can someone please confirm I'm the only one with this problem?

@fran - I mirror the Arch official config just allowing CK1 to set the options it wants to. Does your setup work with the official Arch kernel?

@graysky Something goes wrong after the latest kernel update. I use something like this cryptkey=/dev/disk/by-uuid/..:ext4:/keys/mykey as kernel parameter, but it seems the USB is no longer init at boot. When using linux 4.10 it works correctly and the USB contents will be read.

Seems changes have been made in the config (https://aur.archlinux.org/cgit/aur.git/commit/?h=linux-ck&id=b0017a84f42a) - don't know why this happens, do you have an idea? :)

@graysky - Thank you for sending me down the issues/29 path. I learned quite a bit, and now realize this 'problem' is upstream in nature. I'm sure this will not come as a surprise to you, but I'm running gcc version 8.2.1 20180831 (GCC) and the output of: gcc -c -Q -march=native --help=target | grep march is: -march= skylake Now I understand why, which I do appreciate.

@air-g4p - Your question is really for the gcc patchset I maintain. Issue 29 linked below applies but to my knowledge, there hasn't been any gcc changes to support your CPU. If I am wrong, plz post to the issue against the patch: https://github.com/graysky2/kernel_gcc_patch/issues/29

o/ graysky:

I'm running an i7-8550U which Intel describes as belonging to the Kaby Lake R family. This family was launched in Q3 2017.

That family of CPUs lives here: https://ark.intel.com/products/codename/126287/Kaby-Lake-R

Are you planning to add Kaby Lake support within your PKGBUILD or do you suggest I use another one of the other PKGBUILD 'Lakes'?

Thanks

@ntia89 - https://www.kernel.org/minor-changes-to-tarball-release-format.html

oh, thanks... I was missing the change upstream (kernel dev) :-(

This change mirrors that of the upstream Arch kernel and is due to the change of upstream (kernel source) not signing patches any more. You can modify the PKGBUILD to be like the old behavior if you want, but I will maintain as closely as I can how Arch upstream does it.

some releases ago, I used to download 4.X tar (90MB) and a patch-4.X.Y tar (~KB), while now I download 4.X.Y tar, for every new kernel sub-version. Is there a way to revert to a patch-system in order to build the kernel? I have an very very slow internet connection but I bet it will be appreciate for metered-connection users.

What?

Is there an easy way to get back with the kernel patch system?

@graysky Sorry for the noise, the commit has been reverted. :|

@graysky

https://git.archlinux.org/svntogit/packages.git/commit/trunk?h=packages/linux&id=62b80117614b44f4a17921b0d1cb9b9e6d08fa4e

@graysky Thanks for the update! :)

@francoism90 - Was waiting for 4.18.6 but I don't even see an rc for it. I will push 4.18.5-4 with these incorporated shortly.

@graysky Could you please implement the following changes: https://git.archlinux.org/svntogit/packages.git/commit/trunk?h=packages/linux&id=c75a915313f72924fa0a3ed45356f9e0ea488f3b

Should this also be changed? https://git.archlinux.org/svntogit/packages.git/commit/trunk?h=packages/linux&id=df01e3111f453e720762a55d68a2e6de2a25f247

Thanks!

@keren_sky, did you search for 'apm_bigs' in the kernel source? I did and found nothing, but searching for 'apm_bios' gives many results, for example

$ find -name 'apm_bios*'

./include/linux/apm_bios.h ./include/uapi/linux/apm_bios.h

I would assume that the error has nothing to do with 'apm_bios*' files.

P.S. If somehow your kernel source doesn't have 'apm_bios*' files then something wrong on your side.

I see the following: CC [M] drivers/gpu/drm/nouveau/nvkm/engine/disp/dmacgv100.o fixdep: error opening file: include/linux/apm_bigs.h: No such file or directory make[4]: *** [scripts/Makefile.build:318: drivers/gpu/drm/nouveau/nvkm/engine/disp/dmacgv100.o] Error 2

Which pkg provides that file? thanks.

repository http://repo-ck.com/ will be updated soon?

@biosin, did you read your log and other comments? Read the log again and these comments: freenestor commented on 2018-08-28 04:17 (edited on 2018-08-28 04:21 by freenestor)

ltsdw commented on 2018-08-28 05:24 (edited on 2018-08-28 05:24 by ltsdw)

skydrome commented on 2018-08-28 09:09 (edited on 2018-08-28 09:09 by skydrome)

P.S. Sorry, no links.

Edit: My fault, somehow the AUR helper didn't delete everything before rebuilding it. 4.18.5-3 does compile.

Somehow, linux-ck 4.18.5-3 and 4.18.5-1 failed building for me at the same seemingly random spot:

LDS     arch/x86/kernel/vmlinux.lds
  CC [M]  arch/x86/kernel/msr.o
  CC [M]  arch/x86/kernel/cpuid.o
  AR      arch/x86/kernel/built-in.a
  AR      arch/x86/built-in.a
==> ERROR: A failure occurred in build().
    Aborting...

Full log: https://0bin.net/paste/Y1S1HgocIkxQ8yTW#Ak1q9uDS+t9SSgwppr80Kj3T6rXI51EceXDIdJoNtSQ

Thanks for your hard work @graysky! I can also confirm that it's building fine.

Great! Confirming 4.18.5-3 builds fine for me now, you did the same thing I tried by removing the DEPMOD=/doesnt/exist. :) I was just testing it out and for my particular setup there are some regressions (AMDGPU related, I have 4 monitors, tends to be finicky) so I'll hold off for now and stick with 4.17.11-6. Thank you for the quick support!

@zerophase - I don't think I can automate that since there is are 2 prompts...

4.18.5-3 builds, sorry for the fuck-up.

Also received the same issue as @Lessaj

I made changes in pkgbuild to apply muqss to another package kernel, and it build successfully here.

setting _subarch=27 for GCC detecting native cpu features causes the setup script to loop infinitely on Support for P6_NOPs on Intel chips (X86_P6_NOP) [N/y/?] Would it be possible to allow user input here for whether to turn this feature on or not?

I can confirm the updated build 4.18.5-2-ck fails as per @Lessaj. (I am using _subarch=19)

  INSTALL sound/usb/usx2y/snd-usb-usx2y.ko
  INSTALL virt/lib/irqbypass.ko
  DEPMOD  4.18.5-2-ck
'make modules_install' requires /doesnt/exist. Please install it.
This is probably in the kmod package.
make: *** [Makefile:1248: _modinst_post] Error 1
==> ERROR: A failure occurred in package_linux-ck().
    Aborting...
==> ERROR: Makepkg was unable to build linux-ck.

I updated with ck's fix in 4.18.5-2 ... please confirm.

Able to complete the build with the patch mentioned by @skydrone, but fails right at the end when installing modules as part of package(). Comparing with previous PKGBUILD this part is unchanged, not sure why I'm experiencing this error now. Confirmed I have the kmod package.

  INSTALL sound/xen/snd_xen_front.ko
  INSTALL virt/lib/irqbypass.ko
  DEPMOD  4.18.5-1-ck
'make modules_install' requires /doesnt/exist. Please install it.
This is probably in the kmod package.
make: *** [Makefile:1248: _modinst_post] Error 1
==> ERROR: A failure occurred in package_linux-ck().
    Aborting...

@skydrome how can I apply this? Should I edit something in pkgbuild or before/pre pkgbuild?

Hello. When I installed my current 4.17.11 kernel, I made some changes to the config with menuconfig (mostly removed unneeded modules). What is the preferred way to integrate the changes I made into the new version? Do I just load the current config from the menu, or should I modify the PKGBUILD in some way?

https://github.com/ckolivas/linux/commit/5d1737e81a61d1cddfab5b5f4e8a0639761148f6.patch

I'm getting this:

kernel/sched/MuQSS.c: In function ‘sched_tick_remote’: kernel/sched/MuQSS.c:3321:34: error: ‘struct task_struct’ has no member named ‘se’ delta = rq_clock_task(rq) - curr->se.exec_start;

4.18.5 is failing yea :(

I failed buid it on my computer with linux-4.17-ck installed

CC kernel/rcu/rcu_segcblist.o

AR kernel/rcu/built-in.a

CC kernel/sched/MuQSS.o

kernel/sched/MuQSS.c: In function ‘sched_tick_remote’:

kernel/sched/MuQSS.c:3321:34: error: ‘struct task_struct’ has no member named ‘se’

delta = rq_clock_task(rq) - curr->se.exec_start; ^~

make[2]: *** [scripts/Makefile.build:318: kernel/sched/MuQSS.o] Error 1

make[1]: *** [scripts/Makefile.build:558: kernel/sched] Error 2

make: *** [Makefile:1033: kernel] Error 2

==> ERROR: A failure occurred in build(). Aborting...

Is the 4.18.5 failing to build for anyone else?

@graysky https://git.archlinux.org/svntogit/packages.git/commit/trunk?h=packages/linux&id=c75a915313f72924fa0a3ed45356f9e0ea488f3b - could you also enable this in linux-ck? :)

Just a heads up, a 4.18 ck-patchset has just been released: http://ck.kolivas.org/patches/4.0/4.18/4.18-ck1/

@simona

http://ck.kolivas.org/patches/4.0/

Do you see patchset for 4.18?

I hope 4.18.2 will arrive soon...

@Jeffrey4l - I don't see why not. Did you create a ticket for the Arch kernel to do the same? @francoism90 - No idea, email ck.

@graysky Does this mean no patch for 4.18 yet?

Could you enable

CONFIG_CGROUP_CPUACCT=y

in config in default?

This is need for docker stats and systemd-cgtop

@simona:

https://aur.archlinux.org/cgit/aur.git/commit/?h=linux-ck&id=7574ccd5abcc

4.17.11-6 ? what change?

:: installing nvidia-utils (396.51-1) breaks dependency 'nvidia-utils=396.45' required by nvidia-ck-sandybridge

ok :-))) thx.

@simona - See the pinned post from Aug 4th.

stopped to 4.17.11?

Oh it's pkgbase, got it!

Yeah I'm running it, but the changes were made in bluetooth that I have not tested much, and schedutils that I don't use. If you look at the patch you'll see it is just a 3-way manual diff, nothing fancy.

@gee - Thanks, any feedback (assuming you're running that patched code)? I passed CK your proposal for review. To your question, just rename the pkgbase from linux-ck to linux-whatever-you-want

@ sir_lucjan

thank you, but I think I don't understand what to do. I did:

sudo pacman-key --recv-key 38DBBDC86092693E

it replied:

gpg: bad data signature from key 8396F1D05506E82D: Wrong key usage (0x19, 0x2)

gpg: bad data signature from key 20E8A9C77716EB4F: Wrong key usage (0x19, 0x2)

I did also that without realizing "wrong key usage" warning:

sudo pacman-key --lsign-key 38DBBDC86092693E

Now when I repeat above command (--recv-key) it says: Remote key not fetched correctly from keyserver.

I don't know if these info is helpful. If not, I'm sorry...

edit: I found this from a forum and solved: gpg --recv-keys 38DBBDC86092693E

@ sispus

https://bbs.archlinux.org/viewtopic.php?id=191954

I couldn't install it to manjaro xfce. linux-4.17.11.tar ... FAILED (unknown public key 38DBBDC86092693E) ==> ERROR: One or more PGP signatures could not be verified!

@graysky, I've patched ck's patch to build with .14, it was fairly easy, hopefully I made no mistake: https://pastebin.com/WDVt52ZB

Unrelated, but how do I easily get this package to be named similarly to the one I'd get from your repo, in my case -haswell?

@graysky - I'll try thanks

EDIT: Perfect! Working as intended. Thanks a lot! :)

Hi, since last 2 updates I'm unable to boot to GUI, because of nvidia isn't found in kernel modules. I get this error during install:

==> ERROR: module not found: nvidia' ==> ERROR: module not found:nvidia_modeset' ==> ERROR: module not found: nvidia_uvm' ==> ERROR: module not found:nvidia_drm'

I have nvidia-ck and linux-ck-headers installed. Is there anything I'm doing wrong?

GPU is Nvidia GTX960. Before this everything worked flawlessly.

Thanks

@melcor333 - broadcom and nvidia-* have been updated. Please try and reported back. broadcom and nvidia-340xx work for me on my test box.

Yes, the naming schema has changed. I will push some updates shortly.

All - I think I have all the kinks ironed out of the new PKGBUILD structure. We are still unfortunately in the position of ck1 not patching into 4.17.12-current so we seem stuck on 4.17.11. Good news is that 4.18.0 should be released this weekend and CK is likely to work on that rather than doing so to an EOL kernel (4.17.x series).

I just pushed 4.17.11-4 with a few minor tweaks. Builds and runs fine for me. Please d/l and test.

@Saren - Fixed the NUMA issue in -4. Please try.

Hi, my system needs NUMA support, and changing _NUMAdisable=y to _NUMAdisable= still disables NUMA support on 4.17.11-3. I have already compiled the kernel twice and I am sure I have changed that line. It was totally fine before the major PKGBUILD format change.

Yeah if only they provide a checksum and/or signed packages :)

https://github.com/archlinux/linux/releases

artafinde,

That sounds better than my suggestion. :)

Following the new way for the linux package fetches the whole linux repo (bare) with cloning which downloads 2Gb just to checkout HEAD of it (~150Mb). The ideal would beto download from release i.e. https://github.com/archlinux/linux/archive/v4.17.13-arch1.tar.gz

Hey graysky,

Is there any reason you're not using https://github.com/archlinux/linux/tree/v4.17.11-arch1? It contains all of the relevant Arch-specific changes:

https://github.com/archlinux/linux/commits/v4.17.11-arch1

@graysky should the kernel be installing modules as ck-ck now?

@id942701 - Opps, debug helper. Removed no in 4.17.11-3.

PKGBUILD problem: exit in prepare(). Build successful.

I changed my mind... I just commited 4.17.11-2 which just changes to the PKGBUILD to mirror that of Arch upstream. Builds for me. Can some others please test and report back? CK has been contacted about ck1 not patching in to 4.17.12.

@QuartzDragon - I have already merged it too but am don't see much value in using 4.17.11-2 with just the PKGBUILD changes... as I posted on 08-04, ck1 does not patch against 4.17.12.

Hey, graysky,

You may want to check out https://git.archlinux.org/svntogit/packages.git/tree/trunk?h=packages/linux because the kernel PKGBUILD has changed significantly.

I've already merged in the changes locally, and it builds fine.

I am unable to patch into 4.17.12+ due to changes upstream... need to fix before updating the PKGBUILD:

Hunk #1 FAILED at 185.
Hunk #2 succeeded at 473 (offset -1 lines).
1 out of 2 hunks FAILED -- saving rejects to file
kernel/sched/cpufreq_schedutil.c.rej

Small typo "Note that the generic (default) option is 24" but "26. Generic-x86-64 (GENERIC_CPU)"

I'm trying to build 4.14.20 because I have not been able to build a functional version since 4.16.15. Well 4.16 builds fine but it just hangs at boot same with 4.17-ck and vanilla. Only working kernel I got now is latest linux-lts.

You're building 4.16?

I am unable to build any version prior to 4.16 due to the following error message:

pager.c: In function ‘pager_preexec’: pager.c:36:12: error: passing argument 2 to restrict-qualified parameter aliases with argument 4 [-Werror=restrict] select(1, &in, NULL, &in, NULL); ^~~ ~~~ cc1: all warnings being treated as errors mv: cannot stat '/tmp/makepkg/linux-ck/src/linux-4.14/tools/objtool/.pager.o.tmp': No such file or directory make[4]: [/tmp/makepkg/linux-ck/src/linux-4.14/tools/build/Makefile.build:97: /tmp/makepkg/linux-ck/src/linux-4.14/tools/objtool/pager.o] Error 1 make[3]: [Makefile:52: /tmp/makepkg/linux-ck/src/linux-4.14/tools/objtool/libsubcmd-in.o] Error 2 make[2]: [Makefile:54: /tmp/makepkg/linux-ck/src/linux-4.14/tools/objtool/libsubcmd.a] Error 2 make[1]: [Makefile:62: objtool] Error 2 make: *** [Makefile:1637: tools/objtool] Error 2 ==> ERROR: A failure occurred in prepare(). Aborting...

Anyone know what this is about?

I'm the same person with the one on the forums :)

@artafinde Yup, just saw that in the thread. Thanks

@SilverMight: Enable NUMA, recompile and CUDA now works.

Have a few people with the same issue in the thread: https://bbs.archlinux.org/viewtopic.php?id=239174

No idea about CUDA/don't have the hardware. You might try CK's blog or someone might reply to what I think is your bbs thread.

Now it works, thanks. Do you know if there's an issue with CUDA? Doesn't seem to initialize on this kernel, but every other kernel works fine for me

Did you pacman -Syyu? Fine for me % pacman -Ss core2-h repo-ck/linux-ck-core2-headers 4.17.11-1 (ck-core2) Header files and scripts for building modules for Linux-ck-core2 kernel Intel Core2/Newer Xeon optimized.

Some reason my pacman is trying to fetch 4.10-1 instead of 4.11 from repo-ck, even though I have refreshed the lists.. anyone experiencing this?

Also - is anyone else having an issue with CUDA with this kernel?

@ graysky

No.

Can anyone else confirm that 4.17.6 causes errors relating to mounting disks.

Proposed change in PKGBUILD:

Note that the
# generic (default) option is 24.

to ====>

Note that the
# generic (default) option is 26.

On linux-ck 4.17.3-1 makepkg error:

==> Verifying source file signatures with gpg...
    linux-4.17.tar ... FAILED (unknown public key 79BE3E4300411886)
    patch-4.17.3 ... FAILED (unknown public key 38DBBDC86092693E)
==> ERROR: One or more PGP signatures could not be verified!

I asked for a timeline for 4.17-ck1. I will update the [repo-ck] thread when I hear back. Doesn't make too much sense to go though heroic measures to troubleshoot if 4.17-ck1 is just around the corner. I believe 4.16.x will be EOL'ed shortly as well.

CK expects 4.17-ck1 in a few days so it's not worth trying to unfuck 4.16-ck1 to fit into 4.16.17 or 4.16.18 IMO.

Yea, it's confirmed that 4.16.17-1 doesn't play well with ck's patchset. I asked for a timeline for 4.17-ck1. I will update the [repo-ck] thread when I hear back. Doesn't make too much sense to go though heroic measures to troubleshoot if 4.17-ck1 is just around the corner. I believe 4.16.x will be EOL'ed shortly as well.

Same problem as the others, just a black screen during boot, ditto with the pre built pkg from the repo ck.

me too. 4.16.17-2 stucks just after boot selection. I am on broadwell, compiled with native and P6_NOP. Downgraded

@zero - Yes, I am having this problem as well.

Anyone else getting stuck after the boot loader with the last update? Don't believe I have any logs since the system stalls early in boot.

@Saren - Thanks... I ran the test in another terminal but never checked the result. Fixed now.

Fail at prepare() for 4.16.17-1

patching file drivers/acpi/acpi_watchdog.c
Reversed (or previously applied) patch detected!  Skipping patch.
1 out of 1 hunk ignored -- saving rejects to file drivers/acpi/acpi_watchdog.c.rej

@banzay - I like to mirror the official Arch kernel patch-wise @Morgan - I believe that was needed for some reason ... can't remember why but it's harmless as-is.

https://bugzilla.kernel.org/show_bug.cgi?id=198167 these patches will not be included?

Is there a reason for the linux-ck=4.16.10 provide? All packages automatically provide their own name.

@graysky: thanks. Do you mean 4.16.6-2? There is no -3 in the repository.

@sir_l - CK1 includes the nvidia patch (0003) and I bumped linux-rc to 4.16.7 which includes 0004 so I accidentally omitted it.... fixing in 4.16.6-3, thanks.

@graysky

Probably you forgot about other patches:

1. https://git.archlinux.org/svntogit/packages.git/tree/trunk/0003-Partially-revert-swiotlb-remove-various-exports.patch?h=packages/linux/linux

2. https://git.archlinux.org/svntogit/packages.git/tree/trunk/0004-Fix-vboxguest-on-guests-with-more-than-4G-RAM.patch?h=packages/linux (added: 4.16.7-rc1).

@sir_lucjan - Yes, I thought I diffed that out >:( ... fixed now but since I have been building the repo packages, I don't want to bump the pkgver and start all over.

@zebulon - Thanks for reporting. Please try now.

@graysky

You should remove

# http://bugs.archlinux.org/task/9912
  install -Dt "${_builddir}/drivers/media/dvb-core" -m644 drivers/media/dvb-core/*.h

from headers()

@zebulon

UP

Hi, packaging fails with this error:

==> Starting package_linux-ck-headers()... install: cannot stat 'drivers/media/dvb-core/*.h': No such file or directory ==> ERROR: A failure occurred in package_linux-ck-headers(). Aborting...

Or am I doing something wrong? The only option I used is Skylake optimisations, but I do not think this is relevant.

@simona

You should ask Con.

4.16? soon?

Today receiving the key just worked and the packages updated as well without issues. Must have been a network issue elsewhere.

@vp1981: The missing sudo on the second command was just an oversight on my part when I prepared a cleaned up, condensed message for posting here. In reality I ran all commands in a root shell and only added sudo in front for the log here to make it clear that I was running those commands as root.

@graysky

I already installed that packages using 'pacman -U'. I had to say that these are the only packages that not upgraded smoothly, on other two platforms (haswell and sandybridge) all was fine. At first I suspected that it is the 'archlinux-keyring' that cause the problem, I tried to update all other packages but without '-ck-skylake*' and only after that tried to update these two packages using usual 'pacman -Su' but 'pacman' reported something about either broken package or a key, I couldn't remember the details, sorry.

@Dennis

I tried to reinstall the key and went fine (actually pacman-key returned that the key is already installed). Besides, according to the your message the problem with 'gpg': it cannot retrieve the key. You may try again but I'm not sure what causes the failure (may be network problem or gpg misconfiguration, you may want to [re]read https://wiki.archlinux.org/index.php/Pacman/Package_signing). Also I noticed that you issue command like

sudo pacman-key ... && pacman-key ...

without 'sudo' before the second pacman-key. Try to append 'sudo' to both 'pacman-key'.

P.S. I usually do that everyday:

(reflector stuff) && pacman -Scc && pacman -Syy && pacman -Su && pacman -Scc

I'm also having problems updated linux-ck-skylake from repo-ck:

$ sudo pacman -Syyu

:: Starting full system upgrade... resolving dependencies... looking for conflicting packages...

Packages (2) linux-ck-skylake-4.15.12-1 linux-ck-skylake-headers-4.15.12-1

Total Installed Size: 152.68 MiB Net Upgrade Size: 0.05 MiB

:: Proceed with installation? [Y/n] Y (2/2) checking keys in keyring [#######################################################################################] 100% downloading required keys... error: key "88A032865EE46C4C" could not be looked up remotely error: required key missing from keyring error: failed to commit transaction (unexpected error) Errors occurred, no packages were upgraded.

$ sudo pacman-key -r 5EE46C4C && pacman-key --lsign-key 5EE46C4C gpg: keyserver receive failed: Address family not supported by protocol ==> ERROR: Remote key not fetched correctly from keyserver.

Also tried updating archlinux-keyring alone first and reinitialized and repopulated the keyring but I am still getting above errors.

@vp1981 and @Silvermight - Please pacman -Syyu and it should work.

@SilverMight: same here, I checked downloaded files (-ck-skylake and -ck-skylake-headers) and they are correct and 'pacman -U' installs them fine.

Getting error: linux-ck-skylake: missing required signature when trying to update from repo-ck.com... I have the pacman key added and I tried it again, however it still fails to update

@shanti

See:

https://github.com/sirlucjan/lucjan-kernels/blob/master/lucjan-kernels-experimental/v4.15/linux-bfq-sq-mq-haswell/PKGBUILD#L137

Works fine.

@sir_lucjan

thanks I added the lines you provided however I am still unable to apply the bfq-sq-mq patch successfully,

Makefile.rej gives:

[code]--- Makefile +++ Makefile @@ -2,7 +2,7 @@ VERSION = 4 PATCHLEVEL = 15 SUBLEVEL = 11 -EXTRAVERSION = +EXTRAVERSION = NAME = Fearless Coyote

DOCUMENTATION --- Makefile +++ Makefile @@ -2,7 +2,7 @@ VERSION = 4 PATCHLEVEL = 15 SUBLEVEL = 11 -EXTRAVERSION = +EXTRAVERSION =-mq NAME = Fearless Coyote

DOCUMENTATION[/code]

@sir_lucjan

thanks I added the lines you provided however I am still unable to apply the bfq-sq-mq patch successfully,

Makefile.rej gives:

[--- Makefile +++ Makefile @@ -2,7 +2,7 @@ VERSION = 4 PATCHLEVEL = 15 SUBLEVEL = 11 -EXTRAVERSION = +EXTRAVERSION = NAME = Fearless Coyote

# DOCUMENTATION --- Makefile +++ Makefile @@ -2,7 +2,7 @@ VERSION = 4 PATCHLEVEL = 15 SUBLEVEL = 11 -EXTRAVERSION = +EXTRAVERSION =-mq NAME = Fearless Coyote

# DOCUMENTATION]

@Terence:

you do not have the option to enable bfq-mq anymore

you mean the kernel cmdline option?

it should appear as bfq-mq in the list of schedulers

I have felling that that's wrong (may be because now BFQ in mainline). Some time ago I tried to follow closely BFQ development and had impression that new 'BFQ-MQ' is now 'BFQ' in mainline. This line

mq-deadline kyber [bfq] none

tells me that bfq in my case is BFQ-MQ (because all these schedulers are MQ).

@Terence

Nope, I have my own kernel with bfq patchset.

@sir_lucjan yes, you have bfq-mq and bfq I was just saying vp1981 only had bfq. I see you have bfq-mq. Is this from latest linux-ck? Did you have to do somethign to enable it?

@Terence:

bfq != bfq-mq

[lucjan@archlinux ~]$ cat /sys/block/sda/queue/scheduler
bfq [bfq-mq] none

@vp1981, I think so because you do not have the option to enable bfq-mq anymore and it should appear as bfq-mq in the list of schedulers.

$ uname -a

Linux smoon.vl-lomov.ru 4.15.11-1-ck-haswell #1 SMP PREEMPT Tue Mar 20 11:27:59 EDT 2018 x86_64 GNU/Linux

$ cat /proc/cmdline

BOOT_IMAGE=/vmlinuz-linux-ck-haswell root=UUID=162d2744-158f-46b5-b3e8-fdc25739c96a rw systemd.unified_cgroup_hierarchy=1 scsi_mod.use_blk_mq=1 quiet

$ cat /sys/block/sd*/queue/scheduler

mq-deadline kyber [bfq] none mq-deadline kyber [bfq] none mq-deadline kyber [bfq] none mq-deadline kyber [bfq] none mq-deadline kyber [bfq] none

@Terence, is that mean that I don't have BFQ-MQ?

bfq-mq is also missing.

Why I'm I only seeing the following options: $ cat /sys/block/sd*/queue/scheduler noop deadline [cfq]

$ uname -a Linux 4.15.11-1-ck-skylake #1 SMP PREEMPT Tue Mar 20 11:44:14 EDT 2018 x86_64 GNU/Linux

Thanks

@LinX as one of the devs for it I'm glad it does ;)

@Morganamilo thanks, yay works.

@LinX works with me using yay. AUR helpers are not officially supported though. your best bet is makepkg.

I tried pacaur instead of yaourt but it freezes at:

==> Validating source files with sha256sums... linux-4.15.tar.xz ... Passed linux-4.15.tar.sign ... Skipped patch-4.15.11.xz ... Passed patch-4.15.11.sign ... Skipped config ... Passed 60-linux.hook ... Passed 90-linux.hook ... Passed linux.preset ... Passed enable_additional_cpu_optimizations-20180310.tar.gz ... Passed patch-4.15-ck1.xz ... Passed 0001-add-sysctl-to-disallow-unprivileged-CLONE_NEWUSER-by.patch ... Passed 0002-drm-i915-edp-Only-use-the-alternate-fixed-mode-if-it.patch ... Passed ==> Verifying source file signatures with gpg... linux-4.15.tar ... Passed patch-4.15.11 ... Passed

@Morganamilo I run through yaourt -Syua. Month ago there was no such problem

@LinX It wants user input, are you running makepkg through a script that doesn't pass on stdin or something?

I'm stuck on "Support for P6_NOPs". This line appears endlessly: https://i.imgur.com/N8FzBst.gif

@shanti:

### Patch source with BFQ-SQ-MQ
        msg "Fix naming schema in BFQ-SQ-MQ patch"
        sed -i -e "s|SUBLEVEL = 0|SUBLEVEL = 11|g" \
            -i -e "s|EXTRAVERSION = -bfq|EXTRAVERSION =|g" \
            -i -e "s|EXTRAVERSION =-mq|EXTRAVERSION =|g" ../${_bfq_sq_mq_patch}
        msg "Patching source with BFQ-SQ-MQ patches"
        patch -Np1 -i ../${_bfq_sq_mq_patch}

Works fine.

@sir_lucjan

Thank you very much, unfortunately despite my trials, the patch fails to apply. I reverted to 4.14.12-2-ck which seem to be the last linux-ck including bfq-sq scheduler.

So to my understanding, bfq-sq has been removed from ck patchset itself but I couldn't find relevant information on Paolo Valente's website.

Can somebody elaborate why? or have any information whether it may come back or not.

@CarterCox See https://bbs.archlinux.org/viewtopic.php?pid=1488777#p1488777

==> ERROR: One or more PGP signatures could not be verified!

This is what I'm getting. Anyone knows how to solve it?

@shanti

You can add bfq-sq/bfq-mq patchset:

Download:

https://github.com/sirlucjan/kernel-patches/blob/master/4.15/4.15-bfq-sq-mq-git-20180227.patch

As mentioned by someone earlier, bfq-sq scheduler is gone in recent releases. Can someone please explain why such an important feature of this kernel is being omitted or dropped? Thanks

I see several warnings in my log "NOHZ: local_softirq_pending xxx" 4.15.8-1-ck-skylake

The problem is referenced here: https://bugs.archlinux.org/task/57741

Problem should be fixed in systemd 238.0-2.

The system is crashing in the latest version wrt systemd and glibc, I am having the same issue as @QuartzDragon

Anyone else get a systemd PID1-related coredump after the 238 systemd update currently in testing?

It's weird that only the stock Arch kernel worked, while systemd wasn't happy with linux-ck...

Yep, removing this file from the cache and fetching again, worked.

For reference and as a follow up on the touchPad + trackPoint of my pc not being fully enabled by this config, enabling _makenconfig=y option in the PKGBUILD permitted adding this two modules to the build:

Device Drivers -> input device support -> (M) RMI4 SMB Support

Device Drivers -> HID support -> Special HID drivers -> (M) Synaptics RMI4 device support

After a reboot everything was working as before, TITS (that is to say :D) the trackPoint was hardly moving in one direction only.

But adding psmouse.synaptics_intertouch=1 as a boot parameter (advertised in boot log) made the two freshly built modules to be loaded at boot, allowing the trackPoint to be usable without shortcomings !

As a proof, using xinput on X11 the touchPad is recognized in a different way : "SynPS/2 Synaptics TouchPad" now is "Synaptics TM2758-001". In Wayland the name is different but both devices work well too.

@RaskAr Remove the one you have and fetch it again.

Hello graysky, linux-ck 4.15.7-1 fails to build here : enable_additional_cpu_optimizations_for_gcc_v4.9+_kernel_v4.13+.patch ... ÉCHEC

@RaskAr - Not sure what that means... builds fine for me.

Hello graysky, linux-ck 4.15.7-1 fails to build here : enable_additional_cpu_optimizations_for_gcc_v4.9+_kernel_v4.13+.patch ... ÉCHEC

I'm getting this output in my dmesg whenever a VM is running:

BUG: using smp_processor_id() in preemptible [00000000] code: CPU 0/KVM/690
BUG: using smp_processor_id() in preemptible [00000000] code: CPU 1/KVM/691

It seems to be related to CONFIG_PREEMPT, since changing to CONFIG_PREEMPT_NONE=y or CONFIG_PREEMPT_VOLUNTARY=y eliminates the message.

This does not occur on the regular Arch kernel

@RaskAr - I use the ARCH .config as the base for linux-ck making only the changes the ck patchset requires. If this is something shipped with the kernel but silent in the config itself, you can enable it and build with it. If this is an out-of-tree addition, you are free to patch the source and build it.

Hello graysky, the config seems to be missing a builtin config for laptop users :

From log : « kernel: psmouse serio3: synaptics: The touchpad can support a better bus than the too old PS/2 protocol. Make sure MOUSE_PS2_SYNAPTICS_SMBUS and RMI4_SMB are enabled to get a better touchpad experience. »

After checking the config it only misses RMI4, which is the successor of the PS2 PROTOCOL, see http://lkml.iu.edu/hypermail/linux/kernel/1603.2/01731.html Would be great to have it enabled !

bfq-sq is gone again.

-CONFIG_BFQ_SQ_GROUP_IOSCHED=y

PS: Putting "scsi_mod.use_blk_mq=1 " in kernel options is enabling the BFQ MQ scheduler. The selecting it with TLP, works just fine.

PS2: elevator option in the kernel parameters is not recognized.

which processor family should i chose for apollo lake (goldmont) processor Celeron N3450?

Did you try: https://wiki.archlinux.org/index.php/Unofficial_user_repositories/Repo-ck#Selecting_the_correct_CPU_optimized_package

which processor family should i chose for apollo lake (goldmont) processor Celeron N3450?

@nTia89 It had to be from my side, it's working fine now with 4.14.20.

no problem here with .19 and DKMS

Something with 4.14.19 is making DKMS not working for any driver.

@QuartzDragon - I applied the 4.14 patch with -F5 (fuzz-level 5, dangerous; requires manual review of every patch but "shoves" things in when at all possible. It even managed to shove some code right into the middle of a comment!), fixed the rejections, added one or two new compat functions as required due to API/ABI changes between .14 and .15, did some basic research into where things moved around and such and applied those changes.

I've done some kernel work in the past, so I like to think I'm somewhat familiar with the layout of things. It shouldn't take too long for you to pick up where things are, thanks to you're friendly General Regular Expression Parser (grep) .

Basically, checked the diff between 4.14 and 4.15 on the core scheduler to see any abi/api changes, and others as they arose ( I remember there being a change in some of the "compat" files which required updates/additions/rewrites beyond what the original patch provided, there may have been others ).

Then just built, and fixed any issues that arise during compile (these can be due to changing where #ifdef wrappers are in the baseline, where the patch may force things into the wrong section, or a linker issue from a missing function, etc.)

I'd say just dive right in! Try to merge 4.14 to 4.15 yourself, basically make it till you break it, or succeed! Probably a good idea to use a virtualbox image for testing rather than your primary system, or what I do on this laptop I am using to type to you right now, which is running arch in a virtualbox on top of windows 10 (work / warrenty reasons) is this: I take a snapshot prior to development, so any damage that gets done I can just roll back and alls good.

Hope this helps, I'm not too sure on what you are looking for specifically, so I tried to keep it general!

I would say just dig in! Try you m

I mentioned specifically that I didn't know if they were true or not -- I just found a couple articles when searching to see if anyone else had ported it as it was time for me to build a new kernel. I mean no disrespect nor harm toward anybody. @beojan is probably right -- they may have been dated 2007 which I mistook for 2017.

If you've spoken with him and he says he's not -- all the better. But please don't try to make me out to be some sort of demon spreading nonsense -- I specifically said "I don't know if these reports are legitimate or not."

You're free to enjoy the early merge and update I did, or you're free to wait for him to release an official version. I'm still running strong without any issues on my merge, so it seems pretty stable to me.

Again, I apoligize if I've upset anybody, this was not my intent at all.

@kata198 I wonder if the "rumours" you're hearing are really old news from when he stopped maintaining the old ck patchset in 2007.

@kata198- While I applause your efforts to help port the ck patchset, announcing uncited and unconfirmed rumors as pretext is not cool.

CK himself replied to me in a private email:

Yeah bullshit. It'll probably be a week or so though.

Cheers,
Con

On 4 February 2018 at 22:06, John <graysky AT archlinux DOT us> wrote:
> https://wiki.archlinux.org/index.php?title=Talk:Linux-ck&diff=next&oldid=483793
>
> Specifically: "Hey guys, so I've read some rumors that Con Kolivas is taking
> a break / no longer maintaining the ck patchset; that 4.14 was the last
> version for a while. "

@kata198

Where did you hear about these rumours? I'm curious as to why he'd stop maintaining the patchset without notifying everyone, and passing the reins onto someone capable.

Also, out of curiousity, how did you rebase the patches? I'm curious about learning how to do rebasing, myself.

Linux-4.15 (Unofficial)

Hey guys, I've heard rumros that Con Kolivas may no longer be maintaining the ck patchset? I don't know if these reports are legitimate or not, but just in case they are, I went ahead and manually updated the ck patchset to cleanly apply with 4.15. It's been a little bit of time since I've done any sort of active kernel development, but I believe everything is correct here, and I've been running this ported patchset (ported patch-4.14-ck1.xz to work with linux-4.15) for almost 2 full days now without even the slightestnd ev indication of issue, I'm running on an intel i7 in SMP/SMT mode, through virtualbox as my primary laptop/desktop.

I'm posting to share my unofficial merge of the patchset, in the hopes that it may be of benefit to others.

You can find the patch and even a drop-in replacement of the PKGBUILD on stock arch-4.15 package, HERE:

https://github.com/kata198/con-ck-patches

Enjoy!

drivers/gpu/drm/amd/amdgpu/.tmp_uvd_v7_0.o: warning: objtool: uvd_v7_0_ring_emit_fence()+0x45a: return with modified stack frame CC [M] drivers/gpu/drm/amd/amdgpu/amdgpu_vce.o /bin/sh: línea 1: 27516 Violación de segmento (`core' generado) ./tools/objtool/objtool orc generate --no-fp "drivers/gpu/drm/amd/amdgpu/.tmp_amdgpu_vce.o" make[4]: [scripts/Makefile.build:321: drivers/gpu/drm/amd/amdgpu/amdgpu_vce.o] Error 139 make[3]: [scripts/Makefile.build:579: drivers/gpu/drm/amd/amdgpu] Error 2 make[2]: [scripts/Makefile.build:579: drivers/gpu/drm] Error 2 make[1]: [scripts/Makefile.build:579: drivers/gpu] Error 2 make: *** [Makefile:1032: drivers] Error 2 ==> ERROR: Se produjo un fallo en build(). Cancelando...

/bin/sh: línea 1: 5200 Violación de segmento (`core' generado) ./tools/objtool/objtool orc generate --no-fp "drivers/infiniband/hw/mlx4/.tmp_mcg.o" make[4]: [scripts/Makefile.build:321: drivers/infiniband/hw/mlx4/mcg.o] Error 139 make[3]: [scripts/Makefile.build:579: drivers/infiniband/hw/mlx4] Error 2 make[2]: [scripts/Makefile.build:579: drivers/infiniband/hw] Error 2 make[1]: [scripts/Makefile.build:579: drivers/infiniband] Error 2 make: *** [Makefile:1032: drivers] Error 2 ==> ERROR: Se produjo un fallo en build(). Cancelando...

==> Verificando las firmas de las fuentes con gpg... linux-4.14.tar ... HA FALLADO (clave pública desconocida 79BE3E4300411886) patch-4.14.16 ... HA FALLADO (clave pública desconocida 38DBBDC86092693E) ==> ERROR: ¡No se ha podido verificar alguna de las firmas PGP!

sudo pacman-key --recv 38DBBDC86092693E sudo pacman-key --recv 79BE3E4300411886

@AriseEVE

==> Verifying source file signatures with gpg... linux-4.14.tar ... Passed patch-4.14.16 ... Passed

Works for me.

I assume that you have to add the keys to your system (looks like you are missing Linus' and Greg's key)

Following commands should help:

gpg --recv-key 79BE3E4300411886<br> gpg --recv-key 38DBBDC86092693E

==> Verifying source file signatures with gpg... linux-4.14.tar ... FAILED (unknown public key 79BE3E4300411886) patch-4.14.16 ... FAILED (unknown public key 38DBBDC86092693E) ==> ERROR: One or more PGP signatures could not be verified! <br> ==> ERROR: Makepkg was unable to build linux-ck.

Unable to build, key check failed.

@Mthw - Not out-of-date for the same reason as graysky commented on 2018-01-29 09:52

Ah my bad. Wasn't thinking about the patches.

@morganamilo-not out of date until ck1 for 4.15.x is released.

4.14.15-2 works for me as well, thanks

linux-ck-4.14.15-2 builds successfully for me. Thanks for the patch!

I included a patch from loqs that should fix the build error in 4.14.15-2. Please test and give some feedback.

I tried to disable CONFIG_SCHED_MUQSS in defconfig (and regenerated the config file thus enabled FAIR_GROUP_SCHED and CFS_BANDWIDTH automatically), the build was success without error. Hope this info helps.

Hello, just for the record: https://www.spinics.net/lists/linux-block/msg21999.html and corresponding commit is in 4.14.15.

I'm not familiar with MuQSS source but hope that the fix is simple enough and CK make new ck2.

I tried to compile linux-ck but I had the same error.

Look, I think there's been a misunderstanding.

Error seems related to ck-patchset/muqss. I don't use muqss / ck patchset.

Sources 4.14.15 build fine without errors.

Look, I think there's been a misunderstanding.

Error seems related to ck-patchset/muqss. I don't use muqss / ck patchset.

Sources 4.14.15 build fine without errors.

Yes but without MuQSS. Works fine. Seems related to ck-patchset.

...without MuQSS? So if you apply all the patches in the broken out ck1 except for 0001-MuQSS-version-0.162-CPU-scheduler.patch you are able to build without error?

@graysky

Yes but without MuQSS. Works fine.

Seems related to ck-patchset.

@sir_l - you compiled 4.14.15-1 without errors? Looks like you're building a differently named kernel. What are the differences vs linux-ck?

@graysky

Yes, seems related to MuQSS. I've compiled without errors:

Linux version 4.14.15-0.4-bfq-sq-mq-haswell-git (lucjan@archlinux) (gcc version 7.2.1 20171224 (GCC)) #1 SMP PREEMPT Mon Jan 22 17:03:42 CET 2018
Linux archlinux 4.14.15-0.4-bfq-sq-mq-haswell-git #1 SMP PREEMPT Mon Jan 22 17:03:42 CET 2018 x86_64 GNU/Linux

(stable-rc version)

@sir_l - Seems related to MuQSS. CK has been emailed.

Full log: https://gist.github.com/graysky2/b11cc7627168d951d96ceba14d79565c Clipped log:

kernel/sched/MuQSS.c: In function ‘try_to_wake_up’:
kernel/sched/MuQSS.c:1962:3: error: too few arguments to function ‘delayacct_blkio_end’
   delayacct_blkio_end();
   ^~~~~~~~~~~~~~~~~~~
In file included from kernel/sched/MuQSS.c:41:0:
./include/linux/delayacct.h:125:20: note: declared here
 static inline void delayacct_blkio_end(struct task_struct *p)
                    ^~~~~~~~~~~~~~~~~~~
kernel/sched/MuQSS.c: In function ‘try_to_wake_up_local’:
kernel/sched/MuQSS.c:2025:4: error: too few arguments to function ‘delayacct_blkio_end’
    delayacct_blkio_end();
    ^~~~~~~~~~~~~~~~~~~
In file included from kernel/sched/MuQSS.c:41:0:
./include/linux/delayacct.h:125:20: note: declared here
 static inline void delayacct_blkio_end(struct task_struct *p)
                    ^~~~~~~~~~~~~~~~~~~
  CC [M]  sound/drivers/vx/vx_pcm.o
  CC [M]  net/8021q/vlanproc.o
make[2]: *** [scripts/Makefile.build:320: kernel/sched/MuQSS.o] Error 1
make[1]: *** [scripts/Makefile.build:579: kernel/sched] Error 2
make: *** [Makefile:1032: kernel] Error 2

@graysky

Could you paste a log?

All - It seems that 4.14.15-1 is erroring out on the build for me. Need to dig into it further to see why.

@sir_lucjan - Thanks. I removed it but since it's harmless, I won't push a -2 release... it can be effective on the next version.

"chmod +x tools/objtool/sync-check.sh" is not needed anymore.

https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/commit/?h=linux-4.14.y&id=50c1c6cc09dcea608ace4bd258714d01909fed45

could you please append 'make clean' and 'rm -fr Documentation' after 'make mrproper'?

'rm -fr Documentation' saves >2Gbytes on tmpfs and can avoid 'No space left on device' error which happens in low-ram PC.

@enihcam - The make clean would be redundant and I show linux-4.14/Documentation as being only 38M, not 2G.

@QuartzDragon Thanks got it working. I was zcatting config over too early in prepare earlier. (might have been missing switching Retpoline on) For some reason running zcat too early in prepare turns every module on.

@graysky @zerophase

The Retpoline patches for Spectre got backported to the current 4.14.14 release, not before. The Memory Page Isolation patches for Meltdown were earlier, though, yeah.

@zerophase - Yes, since version 4.14.11-1 I believe.

I tried the whole zcat /proc/config.gz > ./.config. It seems to work fine, but it looks like more modules are turned on, and the kernel ballooned up by 51.66 MiBs for me. Did anyone else notice the kernel get larger?

Are the Spectre and Meltdown protections turned on by default?

@graysky No, I do not change anything. But I use the system in Turkish. Could it be a problem?

@cokomoko - Builds find for me. Are you modifying anything when building?

@gustawho - Yes.

Does/will this build include any patches to address Meltdown and/or Spectre?

CC [M] arch/x86/crypto/sha512-mb/sha512_mb.o AS [M] arch/x86/crypto/sha512-mb/sha512_mb_mgr_flush_avx2.o CC [M] arch/x86/crypto/sha512-mb/sha512_mb_mgr_init_avx2.o AS [M] arch/x86/crypto/sha512-mb/sha512_mb_mgr_submit_avx2.o AS [M] arch/x86/crypto/sha512-mb/sha512_x4_avx2.o LD [M] arch/x86/crypto/sha512-mb/sha512-mb.o arch/x86/crypto/sha512-mb/sha512_mb.o: file not recognized: Dosya biçemi tanınmıyor make[3]: [scripts/Makefile.build:560: arch/x86/crypto/sha512-mb/sha512-mb.o] Error 1 make[2]: [scripts/Makefile.build:573: arch/x86/crypto/sha512-mb] Error 2 make[1]: [scripts/Makefile.build:573: arch/x86/crypto] Error 2 make: [Makefile:1032: arch/x86] Error 2

Where's the problem?

Thank you @QuartzDragon.

Maybe @graysky could simple add a step to zcat /proc/config.gz to somewhere, for example /tmp/.config? I thought this should be harmless. In this way, people who want to reuse current config could save the manual step, i.e. just open the menu and load /tmp/.config.

I wouldn't mind a solution for config_current either. I can't always remember what I turned on in the kernel every time.

@enihcam

You can always just keep config.last aside, or just zcat /proc/config.gz manually if you really need to.

@graysky, is there any alternative solutions to using '_use_current'? I have many specialized devices running linux-ck, removal of _use_current causes a lot of troubles in maintaining the kernels.

As long as it stops to ask you for any CONFIG that it's missing it's good to keep. It used to stop to ask, I think what's preventing it is:

yes "" | make config >/dev/null

...I'm beginning to wonder if the _use_current variable is more trouble than it's worth

@eugen-b @Tjuh The parameter is not in the current config unless you've booted with a kernel that contains it so I added it, I believe it's actually getting stuck further down in prepare but commenting that out fixes it because it doesn't overwrite the current config.

  zcat /proc/config.gz > ./.config
  echo "CONFIG_PAGE_TABLE_ISOLATION=y" >> ./.config

Same issue as eugen-b when enabling config_audit=y in the config file. Haven't been able to compile since version 4.14.6-2.

I got endless lines related to CONFIG_PAGE_TABLE_ISOLATION=y after makeconfig -sri

I edited the PKGBUILD with _use_current=n, but it didn't help.

What helped was commenting the following lines in PKGBUILD

  ### Optionally use running kernel's config
  # code originally by nous; http://aur.archlinux.org/packages.php?ID=40191
#  if [ -n "$_use_current" ]; then
#    if [[ -s /proc/config.gz ]]; then
#      msg "Extracting config from /proc/config.gz..."
#      # modprobe configs
#      zcat /proc/config.gz > ./.config
#    else
#      warning "Your kernel was not compiled with IKCONFIG_PROC!"
#      warning "You cannot read the current config!"
#      warning "Aborting!"
#      exit
#    fi
#  fi

@graysky Thanks for making the toggle back.

@presianbg - My bad. Fixed in 4.14.12-3.

@saren - Yes, some recent benchmarking I did on a dual quad showed a pretty big hit so I reasoned that 99+% of users want it disabled. I reenabled the variable toggle.

Hi, happy 2018! Wish you all the best! bfq-sq is missing again... what is going on :?

Hi there. Since https://aur.archlinux.org/cgit/aur.git/diff/PKGBUILD?h=linux-ck&id=2315c7af5630 does that mean I will have to comment out the NUMA disabled code everytime? Thanks.

@sl13kp Disabilita le notifiche, dovresti trovare un link in alto a destra nel riquadro "Package Actions" (e per favore evita di spammare nei commenti).

avete scocciato con le notifiche non mi interessa questo programma

@kurwall

Try deleting your old build files, because the patch applies perfectly fine for me. If you're not using TMPFS for your builds, then your old build files will stick around, causing you issues if you don't delete them every now and then.

The GCC patch needs to be redone, as currently it is not applying properly

Hi grayslake,

SUCCESS! Linux-ck can now be built without issue in a clean chroot.

To make it absolutely clear to anyone who wants to build linux-ck in a clean chroot, grayslake referred to specifying the correct _subarch variable for your system on line 38 of PKGBUILD prior to building.

You, ROCK, grayslake!

Thank you for your assistance.

All the best - and Merry Christmas,

air|g4p

@air-g4p - You need to assign a value to the _subarch variable (line 38) before you try building.

oops: my missing (below) linux-ck 4.14.7-1 build failure link is here:

https://bpaste.net/show/d4da2db9822e

Hi grayslake,

I am sorry to have to inform you that using: extra-x86_64-build

fails with either linux-ck 4.14.6-3 or with linux-ck 4.14.7-1.

I have pasted the complete failed build runs at:

linux-ck 4.14.6-3 (per your suggestion) lives here: https://bpaste.net/show/cad411ec029a

and also with: linux-ck 4.14.7-1, here:

Despite the 'advice' generated just prior to build failure:

Console input/output is redirected. Run 'make oldconfig' to update configuration.

$ make oldconfig make: *** No rule to make target 'oldconfig'. Stop.

and ofc, this also:

$ sudo make oldconfig [sudo] password for <userid>: make: *** No rule to make target 'oldconfig'. Stop.</userid>

Those 'make oldconfig' failures occur irrespective of whether a user is in their correct build directory, or in their chroot directory = /var/lib/archbuild/extra-x86_64/<userid>/build</userid>

Also, carefully note, the build process never stopped, prior to failure, to ask me, which CPU I wanted to build support for. The current linux-ck AURs I have tested above both only auto-select '> 23. Generic-x86-64 (GENERIC_CPU)', then immediately fail.

grayslake - I do appreciate your ongoing support, but I would like to get to the bottom of this issue, sometime 'soon'. If you have specific commands you want me to test, let me know.

All the best,

air|g4p

@artafinde - good idea.

@foi

RTFM

https://wiki.archlinux.org/index.php/GnuPG#Import_a_public_key

@graysky I suggest for now to delete the 24 option (line #38) on the PKGBUILD. If a user wants native detection let him go through makepkg.

Hi!

Failed in install (distro: manjaro gnome):

==> Verifying source file signatures with gpg... linux-4.14.tar ... FAILED (unknown public key 79BE3E4300411886) patch-4.14.7 ... FAILED (unknown public key 38DBBDC86092693E) ==> ERROR: One or more PGP signatures could not be verified! ==> ERROR: Makepkg was unable to build linux-ck.

Hi graysky,

Thank you very much for attention and extremely quick correction of this issue. Highly impressive, indeed!

When I get a chance, I'll try a re-build with the current release, and let you how things went.

All the best,

air|g4p

@artafinde - Short of applying this patch and shipping the corresponding config (which makes building for repo-ck more difficult), I don't know how to cover that use-case.

It might be a bit of a hassle for you, but if you just run makepkg -s outside of the chroot, setup the config to your liking, you can replace the shipped config with config.last and update the checksums before you build in the chroot.

@graysky: if you choose 24 (native) then you are asked again for P6_NOPS which fails to auto-configure with "yes" Support for P6_NOPs on Intel chips (X86_P6_NOP) [N/y/?] (NEW)

@air-g4p - Yes, the failure is due to user intervention being required to select a sub-arch. Please try 4.14.16-3 in which the _subarch variable should take care of this use case.

Hi graysky,

In the event you are not aware, your linux-ck 4.14.6-2 AUR fails to build in a clean chroot, using: extra-x86_64-build.

Your immediately prior release also failed to build using either extra-x86_64-build or makechrootpkg -c -r $CHROOT.

We know for a fact your AUR is broken because when discussing this linux-ck build failure on #archlinux, another experienced AUR user also replicated the same extra-x86_64-build failure on his system.

I have pasted my entire build run for your review at: https://bpaste.net/show/3c5c3d9ea326. That paste URL will remain valid for one month.

Additionally, for those users willing to employ this approach, I can confirm your AUR can be successfully built and installed with: makepkg -si.

I hope you are able to determine, and correct, whichever part(s) of your code is/are causing the build failure, as many of us strongly prefer to build AURs exclusively within a clean chroot environment.

Thank you for providing an otherwise excellent alternative kernel. Your effort is immensely appreciated.

Cheers,

air|g4p

Works OK now. Thanks!

@edu and sha - Please try now.

I encountered the same problem as @eduardoeae.

Your packages (from repo-ck.com) are failing with "missing required signature". I have been using them for months without problem. Also tried re-adding your key (pacman-key -r 5EE46C4C && pacman-key --lsign-key 5EE46C4C) as the site suggests, but still not working.