unofficial linux-ck repo with cpu optimized packages is ONLINE

loqs · 2017-08-30 11:27:21

Please switch back to the 4.12 kernel from repo-ck try the same kernel command line options and see if suspend still works.
If it does you may need patches from https://bugzilla.kernel.org/show_bug.cgi?id=196675 / https://aur.archlinux.org/packages/linux-bfq-mq/ to get multiqueue/bfq working.
Edit:
replace bfq with multiqueue/bfq.

Last edited by loqs (2017-08-30 11:31:45)

zaxdan69 · 2017-08-30 14:07:05

Yes, it still works. But I'm not familiar with kernel building, so I think I will stay with the previous version for now. I just hope that all these can help to fix this issue in next kernel versions.
Thanks for your efforts.

loqs · 2017-08-30 15:33:58

@zaxdan69 no problem glad you have MuQSS and suspend working now even if it means disabling bfq.
@graysky have you considered hosting a series of test kernels on repo-ck to try and pinpoint the cause of some of the issues and possible fixes?

graysky · 2017-08-30 18:32:42

@loqs - No, but engaging users such as yourself can work to this end

loqs · 2017-08-30 19:16:33

With no response from Con Kolivas or Jens Axboe and a worst case estimate of 4.13 being out in under two weeks I do not want to invest effort into a 4.12 investigation.
Hypothetically speaking if I found adding 9 patches from upstream linux-block tree made fixed the issues with 4.13. What would that finding achieve?

nomorewindows · 2017-09-03 01:48:53

Apparently having trouble with graysky signature on recent packages ck-kx.

raggerv8 · 2017-09-03 01:50:56

nomorewindows wrote:

Apparently having trouble with graysky signature on recent packages ck-kx.

Same issue? https://bbs.archlinux.org/viewtopic.php … 7#p1713887

nomorewindows · 2017-09-03 03:17:24

raggerv8 wrote:

nomorewindows wrote:
Apparently having trouble with graysky signature on recent packages ck-kx.
Same issue? https://bbs.archlinux.org/viewtopic.php … 7#p1713887

error: linux-ck-kx:  signature from "graysky (used to sign repo-ck packages) <graysky@archlinux.us>" is invalid
:: File /var/cache/pacman/pkg/linux-ck-kx-4.12.10-1-x86_64.pkg.tar.xz is corrupted (invalid or corrupted package (PGP signature)).
Do you want to delete it? [Y/n]

raggerv8 · 2017-09-03 10:26:43

nomorewindows wrote:

raggerv8 wrote:
nomorewindows wrote:
Apparently having trouble with graysky signature on recent packages ck-kx.
Same issue? https://bbs.archlinux.org/viewtopic.php … 7#p1713887
error: linux-ck-kx:  signature from "graysky (used to sign repo-ck packages) <graysky@archlinux.us>" is invalid
:: File /var/cache/pacman/pkg/linux-ck-kx-4.12.10-1-x86_64.pkg.tar.xz is corrupted (invalid or corrupted package (PGP signature)).
Do you want to delete it? [Y/n]

Do you use powerpill? I only get that with powerpill, not with pacman. It's annoying but apparently something that can be ignored.

nomorewindows · 2017-09-03 16:35:17

raggerv8 wrote:

nomorewindows wrote:
raggerv8 wrote:
Same issue? https://bbs.archlinux.org/viewtopic.php … 7#p1713887
error: linux-ck-kx:  signature from "graysky (used to sign repo-ck packages) <graysky@archlinux.us>" is invalid
:: File /var/cache/pacman/pkg/linux-ck-kx-4.12.10-1-x86_64.pkg.tar.xz is corrupted (invalid or corrupted package (PGP signature)).
Do you want to delete it? [Y/n]
Do you use powerpill? I only get that with powerpill, not with pacman. It's annoying but apparently something that can be ignored.

Using pacman...

nomorewindows · 2017-09-04 16:11:47

I deleted the file and redownloaded and was able to get it to go...
I tried pacman-key but that didn't go anywhere since it wasn't able to get the keyserver...
Don't know what happened, but I got it going now...

aphirst · 2017-09-05 14:37:12

I'm having problems with the integrated SD card reader on my laptop (ThinkPad X230) when running linux-ck-ivybridge, though it works fine under plain linux (-ARCH). The module appears to crash, and pcmanfm seems unable to automount any of the SD card partitions.

linux 4.12.10-1
linux-ck-ivybridge 4.12.10-1

dmesg log at moment of insertion of SD card https://bpaste.net/show/e09a7590f931
result of lsusb and lspci https://bpaste.net/show/d174f7f5e60e

If you'd like me to clarify anything else, just ask.

Last edited by aphirst (2017-09-05 14:38:29)

biosin · 2017-09-05 17:05:42

deleted

Last edited by biosin (2020-03-31 15:03:49)

loqs · 2017-09-05 17:38:23

@aphirst are you doing a like for like test by default the kernel provided by the linux package has CONFIG_SCSI_MQ_DEFAULT=n did you rebuild that kernel or override the option with scsi_mod.use_blk_mq=1?
@graysky the patch I sent to ck is causing an OOPS would be great if you could test and see if it works or OOPS on someone elses system. Still no response to the kernel bug report.

aphirst · 2017-09-05 21:15:37

@loqs No, I'm just comparing between the behaviour of the "linux" kernel in the arch repo, and the "linux-ck-ivybridge" kernel provided in [repo-ck]. Rebuilding kernels to further test this myself is something which I'm not going to be able to do until the weekend at the very earliest, also assuming I don't have anything else to do. The most I could do "soon" would be to also compare with the behaviour of just "linux-ck" also from [repo-ck].

loqs · 2017-09-05 21:41:30

@aphirst you can avoid the rebuild if you pass the kernel parameters instead scsi_mod.use_blk_mq=1 for arch linux or scsi_mod.use_blk_mq=0 elevator=cfq for linux-ck ( you only need to change the parameter set for one of the kernels )
If you can trigger the issue on arch linux then please report it upstream at kernel.org if it is unique to linux-ck report it to ck.

OdinEidolon · 2017-09-10 17:58:24

It seems that with any 4.12.* version my laptop (Dell XPS15 9550, skylake) suffers from hangs and crashes using the CK kernel; vanilla kernel works fine.
Sometimes it will hang on suspend, sometimes on boot, sometimes on login, sometimes on resume... just like that, randomly.
Anyone else has been experiencing this?

Last edited by OdinEidolon (2017-09-10 17:58:40)

kazimir · 2017-09-12 12:09:05

I wish I could suffer hangs and crashes.
With a thinkpad x230 all the 4.12-*-ck kernel I've tried fail to boot and stay stuck at "loading initial ramdisk".

aphirst · 2017-09-16 13:33:14

@loqs I tried running linux-ARCH and linux-CK with and without the options you specified, and got the following results (the OPTIONS line from my /boot/loader/entries/* files):

linux 4.12.13-1

options	cryptdevice=UUID=[redacted]:root root=/dev/mapper/root quiet rw acpi_osi="!Windows 2012"

- WORKS

options	cryptdevice=UUID=[redacted]:root root=/dev/mapper/root quiet rw acpi_osi="!Windows 2012" scsi_mod.use_blk_mq=1

- WORKS

linux-ck-ivybridge 4.12.13-1

options	cryptdevice=UUID=[redacted]:root root=/dev/mapper/root quiet rw acpi_osi="!Windows 2012"

- DOES NOT WORK - DMESG EXCERPT

options	cryptdevice=UUID=[redacted]:root root=/dev/mapper/root quiet rw acpi_osi="!Windows 2012" scsi_mod.use_blk_mq=0 elevator=cfq

- WORKS

Interesting, I didn't get the symmetry you expected I would.

So, either there's some problem with my -ck options line (am I missing something? I was under the impression that since the introduction of the new scheduler, I didn't need to specify anything extra, but maybe that's wrong), or there's a bug of some kind with the -ck build. I'll wait for a little feedback here, then I'll escalate this.

UPDATE #1: I tried adding elevator=bfq to both of the above -ck-ivybridge lines, which did not affect the results.

UPDATE #2: I installed linux-ck (not custom-built for a particular CPU) and repeated the tests - the same occurred as with linux-ck-ivybridge.

Last edited by aphirst (2017-09-16 13:57:46)

loqs · 2017-09-16 18:27:52

aphirst wrote:

I was under the impression that since the introduction of the new scheduler, I didn't need to specify anything extra, but maybe that's wrong),

Without any additional options you get MuQSS + scsi_mod.use_blk_mq=1 you do not get bfq but you get mq-deadline or none as the IO scheduler depending on the devices properties.
You would need a udev rules such as the following to get bfq

ACTION=="add|change", KERNEL=="sd[a-z]",  ATTR{queue/scheduler}="bfq"

However your system seems to fail when scsi_mod.use_blk_mq=1 is enabled similar to zaxdan69 slightly earlier in this thread.

aphirst · 2017-09-16 21:12:54

Ah, so even "elevator=bfq" doesn't work any more? Strange. Well, at least the scsi_mod.use_blk_mq issue seems "known" - is CK aware?

loqs · 2017-09-16 21:49:08

elevator=bfq does not work because in 4.12 CK switched to using the in kernel version of bfq that uses blk_mq but https://git.kernel.org/pub/scm/linux/ke … c0db2c3032
ignores the elevator parameter when the queue type is multiqueue such as scsi_mod.use_blk_mq=1 or the device is NVME (this covers 4.12 ck kernels)
So elevators you can select with the elevator parameter are cfq,noop and deadline only if the queue is not multiqueue
Multiqueue elevators are bfq,kyber,mq-deadline and none and can only be changed using udev or similar and the queue is a multiqueue if you do not specify you get none if the device is NVME or similar otherwise mq-deadline.
I think the closest anyone has come to reporting the issue on CK's blog is http://ck-hack.blogspot.com/2017/08/lin … 7797998186
I do not know if graysky has reported any of the issues from this thread or the AUR package comments to CK.
Edit:
changed linked from this one where CK changes the config so that SCSI_MQ_DEFAULT (scsi_mod.use_blk_mq) is enabled for mq schedulers
https://github.com/ckolivas/linux/commi … 1e22fa1728
to https://git.kernel.org/pub/scm/linux/ke … c0db2c3032

Last edited by loqs (2017-09-16 21:56:07)

pr0dukter · 2017-09-17 10:49:59

how would i rebuild the ck-bulldozer kernel and add in the AMD DC/DAL ?(for hdmi audio)

merg diffs on .configs?

merge pkgbuilds?

can someone set me on the right path?

loqs · 2017-09-17 16:25:19

pr0dukter wrote:

how would i rebuild the ck-bulldozer kernel and add in the AMD DC/DAL ?(for hdmi audio)

You mean https://aur.archlinux.org/packages/linu … aging-git/ or https://aur.archlinux.org/packages/linu … -next-git/

pr0dukter wrote:

merg diffs on .configs?
merge pkgbuilds?

You would have to merge the source trees as well.
Edit:
@graysky off topic but 4.14-rc1 has another case of a symbol being marked GPL only breaking the nvidia driver. Nvidia is aware of the issue https://lkml.org/lkml/2017/9/15/499
other fixes for the nvidia driver appear trivial.

Last edited by loqs (2017-09-17 19:18:04)

pr0dukter · 2017-09-22 05:54:50

loqs wrote:

pr0dukter wrote:
how would i rebuild the ck-bulldozer kernel and add in the AMD DC/DAL ?(for hdmi audio)
You mean https://aur.archlinux.org/packages/linu … aging-git/ or https://aur.archlinux.org/packages/linu … -next-git/
pr0dukter wrote:
merg diffs on .configs?
merge pkgbuilds?
You would have to merge the source trees as well.
other fixes for the nvidia driver appear trivial.

i thought i would have to merge ck-bulldozer 4.12 with linux-amd-staging (because my thinking was it was for kernel 4.12)
i can pull in the absolute latest drm-next to the current bulldozer? don't have to match up versions?

how would i merge source trees in this case? i was thinking i would need to apply all the ck-bulldozer packages to the linux-amd-staging (or drm-next)...

or do i start with ck-bulldozer as base and pull in all the amd-staging?

Arch Linux

#3051 2017-08-30 11:27:21

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

#3052 2017-08-30 14:07:05

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

#3053 2017-08-30 15:33:58

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

#3054 2017-08-30 18:32:42

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

#3055 2017-08-30 19:16:33

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

#3056 2017-09-03 01:48:53

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

#3057 2017-09-03 01:50:56

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

#3058 2017-09-03 03:17:24

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

#3059 2017-09-03 10:26:43

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

#3060 2017-09-03 16:35:17

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

#3061 2017-09-04 16:11:47

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

#3062 2017-09-05 14:37:12

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

#3063 2017-09-05 17:05:42

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

#3064 2017-09-05 17:38:23

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

#3065 2017-09-05 21:15:37

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

#3066 2017-09-05 21:41:30

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

#3067 2017-09-10 17:58:24

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

#3068 2017-09-12 12:09:05

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

#3069 2017-09-16 13:33:14

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

#3070 2017-09-16 18:27:52

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

#3071 2017-09-16 21:12:54

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

#3072 2017-09-16 21:49:08

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

#3073 2017-09-17 10:49:59

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

#3074 2017-09-17 16:25:19

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

#3075 2017-09-22 05:54:50

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

Board footer