unofficial linux-ck repo with cpu optimized packages is ONLINE

graysky · 2018-06-22 23:05:13

My test machine is failing to boot into 4.16.17-2-ck .. I am reluctant to index and update the repo. Can others confirm? Use the http interface to repo-ck to manually download the kernels which should be there (pacman will not see them until I index).

loqs · 2018-06-23 00:15:39

If others confirm have a look at https://github.com/cchalpha/linux-gc/co … 72597cbb36 and https://github.com/ckolivas/linux/blob/ … SS.c#L3885 and https://git.kernel.org/pub/scm/linux/ke … 36b4199785

graysky · 2018-06-23 05:38:08

@loqs - It is confirmed... not sure what to do with the 3 references you provided though.

loqs · 2018-06-23 08:53:10

The third link I provided is the possible cause from 4.16.17 if you revert just that commit does 4.16.7 work with ck patchset?
The first link shows how a MuQSS fork was patched for this change in 4.17. The second link shows MuQSS using the old call ie its the same as the code in the first link prepatch.

graysky · 2018-06-23 12:12:34

Reversed the last one and building now; will report back.

Nope: still get the freeze after the grub screen.

Last edited by graysky (2018-06-23 12:21:40)

loqs · 2018-06-23 12:22:59

Can only suggest you bisect between the 4.16.7 and whichever version is your last known good looking for the cause.

graysky · 2018-06-23 13:52:08

Thanks for the help, loqs. Let's see about timing for 4.17-ck1 as I suspect this is a larger problem needing more expert input.

vuudochile · 2018-06-23 18:43:43

Since you are at it, are you going to disable Speck on 4.17?

loqs · 2018-06-23 23:18:47

$ git bisect log 
git bisect start
# good: [f77427a94b30b1f0b31f2364f4203f2ca94d335d] Linux 4.16.16
git bisect good f77427a94b30b1f0b31f2364f4203f2ca94d335d
# bad: [a988ccdea68969e7894398d8953476d2afdaa508] Linux 4.16.17
git bisect bad a988ccdea68969e7894398d8953476d2afdaa508
# good: [5ebca90cc69a918448154b19ae42b30aface83f8] ARM: davinci: board-omapl138-hawk: fix GPIO numbers for MMC/SD lookup
git bisect good 5ebca90cc69a918448154b19ae42b30aface83f8
# bad: [461adb75f2bb196ca57d773d1c4cf8d14b5a2967] agp: uninorth: make two functions static
git bisect bad 461adb75f2bb196ca57d773d1c4cf8d14b5a2967
# bad: [d047fef7eb1321939af74e10a5cb603e20cd19f6] IB/hfi1: Fix memory leak in exception path in get_irq_affinity()
git bisect bad d047fef7eb1321939af74e10a5cb603e20cd19f6
# good: [c3a28e6c556ec735c682d37dc6d047b7259cd126] parisc: drivers.c: Fix section mismatches
git bisect good c3a28e6c556ec735c682d37dc6d047b7259cd126
# bad: [b6e47b0da4563504367396f7740cae79dcd22620] nvme: Set integrity flag for user passthrough commands
git bisect bad b6e47b0da4563504367396f7740cae79dcd22620
# bad: [f41c7444f3583f6cbf20c5fdab599773887c1c6e] arm64: tegra: Make BCM89610 PHY interrupt as active low
git bisect bad f41c7444f3583f6cbf20c5fdab599773887c1c6e
# good: [7a1b71fa0c4b433f8d1c7922812b2e79c9c3f7ee] kthread, sched/wait: Fix kthread_parkme() wait-loop
git bisect good 7a1b71fa0c4b433f8d1c7922812b2e79c9c3f7ee
# bad: [0ad11385532e11cfd541621b3c097f6df33bf467] kthread, sched/wait: Fix kthread_parkme() completion issue
git bisect bad 0ad11385532e11cfd541621b3c097f6df33bf467
# first bad commit: [0ad11385532e11cfd541621b3c097f6df33bf467] kthread, sched/wait: Fix kthread_parkme() completion issue

So revert 0ad11385532e11cfd541621b3c097f6df33bf467 and 7a1b71fa0c4b433f8d1c7922812b2e79c9c3f7ee or or use something like https://github.com/cchalpha/linux-gc/co … 639c47fb31 as well as the previous fix.
Edit:
https://ptpb.pw/fZg-
or revert 0ad11385532e11cfd541621b3c097f6df33bf467 and 7a1b71fa0c4b433f8d1c7922812b2e79c9c3f7ee as well as previously noted e93d112700952f2d4b93e62eb7c16e36b4199785
Edit2:
@graysky the patches works fine on this system does it still have the same issue for you?

Last edited by loqs (2018-06-24 16:27:12)

loqs · 2018-06-24 23:43:19

graysky wrote:

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.

Hours of kernel building for the bisection is hardly heroic just time consuming. Fix generation took under 15 minutes transcription.
Edit:
https://lkml.org/lkml/2018/6/24/315 yes 4.16.18 will be the last upstream release but upstream. Does ck support linux-stable?

Last edited by loqs (2018-06-24 23:51:06)

graysky · 2018-06-25 13:14:54

I heard from CK that all is going well with 4.17-ck1 and that he 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.

loqs · 2018-06-25 13:24:59

I just have a different perspective that as you you could reproduce the issue you should have bisected it as a matter of course as the package maintainer
then the fix was less than 15 minutes or revert that commit less than 1 minute then repeat if needed.
Edit:
I would also suggest you check each future stable patch for changes to kernel/sched/* as those will be the likely cause of issues with MuQSS.

Last edited by loqs (2018-06-25 13:31:38)

graysky · 2018-06-25 13:38:02

Thanks for the perspective, loqs. I am just short on free-time right now so I was thinking more practically about the 4.16 vs 4.17 situation. Thanks for all your insights and support.

graysky · 2018-06-27 14:50:39

I updated linux-ck to use 4.17-ck1 but I am experiencing the same unbootabled issue as with 4.16.17.
@loqs - Are you able to boot? Does the same fix you recommended for 4.16.17 hold true here as well? Find the source here: http://repo-ck.com/PKG_source/next/linu … 7.3.tar.xz

loqs · 2018-06-27 16:07:50

Works for me https://ptpb.pw/5fB6
Edit:
Default options on prompting

Last edited by loqs (2018-06-27 16:08:58)

graysky · 2018-06-27 17:54:33

You omitted the 0003 patch... I will try wo/ it.

graysky · 2018-06-27 18:16:07

Omitting that patch fixed it for me, thanks.

loqs · 2018-06-27 18:32:27

  # https://bugs.archlinux.org/task/56780
  patch -Np1 -i ../0003-ACPI-watchdog-Prefer-iTCO_wdt-always-when-WDAT-table.patch

The patch itself would not be in the diff as it would be an untracked file but it was applied in the PKGBUILD as the diff shows.
Edit:
What does CK make of the your failure to boot? 0003-ACPI-watchdog-Prefer-iTCO_wdt-always-when-WDAT-table.patch is in 4.18-rc2.

Last edited by loqs (2018-06-28 09:28:19)

Ropid · 2018-06-28 10:32:08

4.17.3-ck seems to run fine here for me.

I noticed the CPU load and usage percentage shows rather normal numbers again at idle compared to what was going on with the last few kernel versions. It still seems to be higher numbers than what I can see through turbostat, but it's a good bit closer to what was happening in kernel 4.16.

I run turbostat like this to make it not print a ton of stuff:

sudo turbostat --interval 1 --show CPU,Busy% --quiet

EDIT:

I just tried experimenting with perl one-liners that do a mix of sleeping and a busy loop. It seems to me the CPU usage percentage measurements in htop for a process is now pretty close to correct in 4.17-ck. I tried experimenting like this:

This here is supposed to do 50% CPU usage:

perl -E 'use Time::HiRes qw(time sleep); $t = 0.010; while (1) { $t0 = time; while ($t > time - $t0) {}; sleep $t }'

75% CPU usage:

perl -E 'use Time::HiRes qw(time sleep); $t = 0.010; while (1) { $t0 = time; while ($t * 3 > time - $t0) {}; sleep $t }'

25%:

perl -E 'use Time::HiRes qw(time sleep); $t = 0.010; while (1) { $t0 = time; while ($t / 3 > time - $t0) {}; sleep $t }'

Last edited by Ropid (2018-06-28 10:45:30)

graysky · 2018-06-28 12:01:25

@ropid -recommend that you post to ck’s blog with feedback,

vuudochile · 2018-06-28 19:30:35

vuudochile wrote:

Since you are at it, are you going to disable Speck on 4.17?

In case you didn't see this question, is the NSA's crypto-speck module enabled in -ck-4.17?

graysky · 2018-06-28 19:33:47

@vuudochile - I saw the question but am unwilling to modify the package. Setup a no extract or blacklist: https://bbs.archlinux.org/viewtopic.php?id=237765

vuudochile · 2018-06-28 19:52:06

graysky wrote:

@vuudochile - I saw the question but am unwilling to modify the package. Setup a no extract or blacklist: https://bbs.archlinux.org/viewtopic.php?id=237765

But I thought linux-ck was all about optimizing (modifying the kernel) for efficient use in a DE. Simply the users that -ck appeals to if they were able to blacklist, recompile kernels without modules, or add other ones, they wouldn't use -ck but compile their own kernels.

Endorsing the NSA's proposed/approved cryptography in your kernel is in a way a political statement. Linus made his own, are you?

ewaller · 2018-06-28 21:00:09

vuudochile wrote:

Endorsing the NSA's proposed/approved cryptography in your kernel is in a way a political statement. Linus made his own, are you?

Enough.

https://wiki.archlinux.org/index.php/Co … _hijacking
https://wiki.archlinux.org/index.php/Co … o_trolling
https://wiki.archlinux.org/index.php/Co … cs.2Frants
https://wiki.archlinux.org/index.php/Co … shed.22.29

Tharbad · 2018-07-03 17:32:12

The NVIDIA driver package still supports only v4.16. Is everything OK?

Arch Linux

#3326 2018-06-22 23:05:13

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3327 2018-06-23 00:15:39

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3328 2018-06-23 05:38:08

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3329 2018-06-23 08:53:10

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3330 2018-06-23 12:12:34

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3331 2018-06-23 12:22:59

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3332 2018-06-23 13:52:08

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3333 2018-06-23 18:43:43

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3334 2018-06-23 23:18:47

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3335 2018-06-24 23:43:19

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3336 2018-06-25 13:14:54

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3337 2018-06-25 13:24:59

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3338 2018-06-25 13:38:02

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3339 2018-06-27 14:50:39

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3340 2018-06-27 16:07:50

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3341 2018-06-27 17:54:33

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3342 2018-06-27 18:16:07

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3343 2018-06-27 18:32:27

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3344 2018-06-28 10:32:08

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3345 2018-06-28 12:01:25

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3346 2018-06-28 19:30:35

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3347 2018-06-28 19:33:47

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3348 2018-06-28 19:52:06

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3349 2018-06-28 21:00:09

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

#3350 2018-07-03 17:32:12

Re: unofficial linux-ck repo with cpu optimized packages is ONLINE

Board footer