unofficial linux-ck repo with cpu optimized packages is ONLINE

loqs · 2018-01-14 16:10:10

What is the output from the following

# pacman -S linux-ck-k10
$ file /boot/vmlinuz-linux-ck-k10
$ stat /boot/vmlinuz-linux-ck-k10
$ pacman -Qkk linux-ck-k10

Edit:
@lano1106 have you compared performance to linux 4.14.6 from the Arch_Linux_Archive?

Last edited by loqs (2018-01-14 19:51:07)

pga256 · 2018-01-15 14:45:27

loqs wrote:

What is the output from the following
# pacman -S linux-ck-k10
$ file /boot/vmlinuz-linux-ck-k10
$ stat /boot/vmlinuz-linux-ck-k10
$ pacman -Qkk linux-ck-k10
Edit:
@lano1106 have you compared performance to linux 4.14.6 from the Arch_Linux_Archive?

# file /boot/vmlinuz-linux-ck-k10
/boot/vmlinuz-linux-ck-k10: Linux kernel x86 boot executable bzImage, version 4.14.13-1-ck-k10 (squishy@waste) #1 SMP PREEMPT Wed Jan 10 16:48:28 EST 2018, RO-rootFS, swap_dev 0x4, Normal VGA
# stat /boot/vmlinuz-linux-ck-k10
  File: /boot/vmlinuz-linux-ck-k10
  Size: 4732688   	Blocks: 9272       IO Block: 4096   regular file
Device: 807h/2055d	Inode: 11          Links: 1
Access: (0644/-rw-r--r--)  Uid: (    0/    root)   Gid: (    0/    root)
Access: 2018-01-15 14:32:07.044699214 +0000
Modify: 2018-01-10 21:50:33.000000000 +0000
Change: 2018-01-15 14:32:06.964697975 +0000
 Birth: -
# pacman -Qkk linux-ck-k10
warning: linux-ck-k10: /usr/lib/modules/4.14.13-1-ck-k10/modules.alias (Modification time mismatch)
warning: linux-ck-k10: /usr/lib/modules/4.14.13-1-ck-k10/modules.alias (Size mismatch)
warning: linux-ck-k10: /usr/lib/modules/4.14.13-1-ck-k10/modules.alias.bin (Modification time mismatch)
warning: linux-ck-k10: /usr/lib/modules/4.14.13-1-ck-k10/modules.alias.bin (Size mismatch)
warning: linux-ck-k10: /usr/lib/modules/4.14.13-1-ck-k10/modules.builtin.bin (Modification time mismatch)
warning: linux-ck-k10: /usr/lib/modules/4.14.13-1-ck-k10/modules.dep (Modification time mismatch)
warning: linux-ck-k10: /usr/lib/modules/4.14.13-1-ck-k10/modules.dep (Size mismatch)
warning: linux-ck-k10: /usr/lib/modules/4.14.13-1-ck-k10/modules.dep.bin (Modification time mismatch)
warning: linux-ck-k10: /usr/lib/modules/4.14.13-1-ck-k10/modules.dep.bin (Size mismatch)
warning: linux-ck-k10: /usr/lib/modules/4.14.13-1-ck-k10/modules.devname (Modification time mismatch)
warning: linux-ck-k10: /usr/lib/modules/4.14.13-1-ck-k10/modules.softdep (Modification time mismatch)
warning: linux-ck-k10: /usr/lib/modules/4.14.13-1-ck-k10/modules.symbols (Modification time mismatch)
warning: linux-ck-k10: /usr/lib/modules/4.14.13-1-ck-k10/modules.symbols (Size mismatch)
warning: linux-ck-k10: /usr/lib/modules/4.14.13-1-ck-k10/modules.symbols.bin (Modification time mismatch)
warning: linux-ck-k10: /usr/lib/modules/4.14.13-1-ck-k10/modules.symbols.bin (Size mismatch)
linux-ck-k10: 4847 total files, 9 altered files

loqs · 2018-01-15 16:06:04

pga256 wrote:

Running uname reports 'Linux' and running uname -r shows the version number as 4.14.11.

# file /boot/vmlinuz-linux-ck-k10
/boot/vmlinuz-linux-ck-k10: Linux kernel x86 boot executable bzImage, version 4.14.13-1-ck-k10 (squishy@waste) #1 SMP PREEMPT Wed Jan 10 16:48:28 EST 2018, RO-rootFS, swap_dev 0x4, Normal VGA

If with linux-ck-k10-4.14.13-1 still installed uname -r still reports 4.14.11 when you are dropped to the rescue shell then the system is booting some other kernel than /boot/vmlinuz-linux-ck-k10.
Edit:
grammar duplicate installed

Last edited by loqs (2018-01-15 16:06:43)

pga256 · 2018-01-16 12:15:13

I have managed to upgrade the kernel properly now. I use syslinux as the bootloader and running a

syslinux-install_update -iam

after the update fixed the issue.
It seems that although the partitions were mounted correctly at the time of upgrade, the bootstrap was pointing to the kernel on an experimental Arch build that I have on another disk. Lesson learned and cheers for the help.

loqs · 2018-01-25 18:07:52

@Graysky 4.14.15 MuQSS patch

diff --git a/kernel/sched/MuQSS.c b/kernel/sched/MuQSS.c
index e84d700709ff..b0be7fcfe41f 100644
--- a/kernel/sched/MuQSS.c
+++ b/kernel/sched/MuQSS.c
@@ -55,6 +55,7 @@
 #include <linux/security.h>
 #include <linux/syscalls.h>
 #include <linux/tick.h>
+#include <linux/version.h>
 
 #include <asm/switch_to.h>
 #include <asm/tlb.h>
@@ -1959,7 +1960,11 @@ try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
 	p->state = TASK_WAKING;
 
 	if (p->in_iowait) {
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 14, 15)
 		delayacct_blkio_end();
+#else
+		delayacct_blkio_end(p);
+#endif
 		atomic_dec(&task_rq(p)->nr_iowait);
 	}
 
@@ -1970,7 +1975,11 @@ try_to_wake_up(struct task_struct *p, unsigned int state, int wake_flags)
 #else /* CONFIG_SMP */
 
 	if (p->in_iowait) {
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 14, 15)
 		delayacct_blkio_end();
+#else
+		delayacct_blkio_end(p);
+#endif
 		atomic_dec(&task_rq(p)->nr_iowait);
 	}
 
@@ -2022,7 +2031,11 @@ static void try_to_wake_up_local(struct task_struct *p)
 
 	if (!task_on_rq_queued(p)) {
 		if (p->in_iowait) {
+#if LINUX_VERSION_CODE < KERNEL_VERSION(4, 14, 15)
 			delayacct_blkio_end();
+#else
+			delayacct_blkio_end(p);
+#endif
 			atomic_dec(&rq->nr_iowait);
 		}
 		ttwu_activate(rq, p);

Edit:
https://github.com/ckolivas/linux/pull/8

Last edited by loqs (2018-01-25 18:24:06)

ooo · 2018-01-25 19:37:07

Thanks for the patch loqs. Applied, built and booted without issues.

graysky · 2018-01-25 19:54:53

Good job, loqs. Let's see what CK says... he hasn't replied to my email yet. Perhaps is into something.

Enverex · 2018-01-29 12:22:04

Just thought I'd try out the linux-ck-broadwell kernel to see how it runs vs the current normal kernel and it didn't go well. Booted normally and everything seemed fine, but CPU usage was all over the place on an idle machine. I was seeing a constantly high CPU usage of between 20 and 80% despite the machine being completely idle (just X loaded and openbox with nothing else). The CPU usage in top was showing as rcu_preempt, nvidia, xorg mostly with a few other random things cropping up, none of which should have been using any amount of CPU power at the time. This didn't seem to go away.

On a normal kernel it's between 0 and 2% whilst idle (with the 2% being caused by top itself).

Any idea what could be going on there?

Here's an image of the issue. It's almost like everything is using exponentially more CPU time than it should.
https://i.imgur.com/61z41gI.png

Here's the stock kernel with the same boot options for comparison.
https://i.imgur.com/rscUy9X.png

Mod Edit - Replaced oversized images with links.
CoC - Pasting pictures and code

Last edited by Slithery (2018-01-31 12:02:52)

Enverex · 2018-01-31 11:55:04

And for the sake of completeness, here's the CK kernel (compiled myself this time, not from the static repo) vs the stock kernel:

== CK Kernel (4.14.15-ck)
Collecting data for 60 seconds...
# NumSamples = 60; Min = 1500.00; Max = 4006.70
# Mean = 3091.651667; Variance = 580503.199497; SD = 761.907606; Median 3356.350000
# each ∎ represents a count of 1
 1500.0000 -  1750.6700 [     3]: ∎∎∎ (5.00%)
 1750.6700 -  2001.3400 [     4]: ∎∎∎∎ (6.67%)
 2001.3400 -  2252.0100 [     3]: ∎∎∎ (5.00%)
 2252.0100 -  2502.6800 [     4]: ∎∎∎∎ (6.67%)
 2502.6800 -  2753.3500 [     6]: ∎∎∎∎∎∎ (10.00%)
 2753.3500 -  3004.0200 [     4]: ∎∎∎∎ (6.67%)
 3004.0200 -  3254.6900 [     3]: ∎∎∎ (5.00%)
 3254.6900 -  3505.3600 [     7]: ∎∎∎∎∎∎∎ (11.67%)
 3505.3600 -  3756.0300 [    15]: ∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎ (25.00%)
 3756.0300 -  4006.7000 [    11]: ∎∎∎∎∎∎∎∎∎∎∎ (18.33%)

== Stock Kernel (4.14.15)
Collecting data for 60 seconds...
# NumSamples = 60; Min = 800.10; Max = 3968.60
# Mean = 950.550000; Variance = 293674.428167; SD = 541.917363; Median 801.200000
# each ∎ represents a count of 1
  800.1000 -  1116.9500 [    56]: ∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎ (93.33%)
 1116.9500 -  1433.8000 [     0]:  (0.00%)
 1433.8000 -  1750.6500 [     1]: ∎ (1.67%)
 1750.6500 -  2067.5000 [     0]:  (0.00%)
 2067.5000 -  2384.3500 [     1]: ∎ (1.67%)
 2384.3500 -  2701.2000 [     0]:  (0.00%)
 2701.2000 -  3018.0500 [     0]:  (0.00%)
 3018.0500 -  3334.9000 [     1]: ∎ (1.67%)
 3334.9000 -  3651.7500 [     0]:  (0.00%)
 3651.7500 -  3968.6000 [     1]: ∎ (1.67%)

As you can see, the increased CPU usage is keeping the processor awake so it's not just a pseudo anomaly.

Last edited by Enverex (2018-01-31 11:55:22)

ooo · 2018-01-31 13:40:56

What program are you using for the cpu frequency statistics? I'd be interested to test this on my system.

IIRC, there's been some issues with load reporting with ck patchset, but i don't know if those are still present. And as you said, since your cpu is running at higher frequency with ck patches, this seems like something more serious.

In any case, I think it would be better to ask Con directly about this issue.

Enverex · 2018-01-31 14:27:48

That's just TOP in the images, the script output below is from graysky's own frequency monitoring script (https://github.com/graysky2/bin/blob/ma … stogram.sh) which reads /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq.

graysky · 2018-01-31 20:29:19

@Enverex - Please post to CK's blog.

Last edited by graysky (2018-01-31 20:29:29)

DamDam · 2018-02-10 21:42:56

Hi,

since some time the packages for linux-ck-piledriver are not updated anymore on repo-ck. Do i miss something, is there any reason for this?

graysky · 2018-02-10 22:31:48

Correct: there is an issue with the piledriver code that prevents it from building.

https://github.com/graysky2/kernel_gcc_patch/issues/30

DamDam · 2018-02-11 07:26:15

@graysky

Thanks for clarification!

I'm not so into the CPU topic but I think I also can use the bulldozer kernel for a FX 8350, or?

graysky · 2018-02-11 12:14:45

DamDam wrote:

I'm not so into the CPU topic but I think I also can use the bulldozer kernel for a FX 8350, or?

This comment in the bug report would suggest so. I cannot test to verify since I don't have the hardware.

inglor · 2018-02-13 10:17:28

Hi all,

I've been using the linux-ck for years without problem on both my FX cpu and the i7. I've upgraded my platform to AMD Threadripper 1950X and I can't boot with linux-ck package. I've tried various build configurations but all the times the boot process is stuck on

Loading Linux linux-ck ...
Loading initial ramdisk ...

So far I'm using the -ARCH kernel which works fine. I've tried compiling the kernel with just the package downloaded from AUR and no changes. I've even tried enabling the NUMA and using the config from -ARCH.

Any suggestions at this point would be appreciated.

john29 · 2018-02-13 11:05:44

packages for linux-ck-piledriver are not updated anymore on repo-ck and cannot update nvidia-ck-piledriver has it depend on nvidia-utils=387.34

Instal the bulldozer kernel on a FX 6350 with cpu CFLAGS="-O2 -march=bdver2 -pipe"
CXXFLAGS="${CFLAGS}"

Last edited by john29 (2018-02-13 17:58:47)

graysky · 2018-02-13 18:59:06

john29 wrote:

packages for linux-ck-piledriver are not updated anymore on repo-ck and cannot update nvidia-ck-piledriver has it depend on nvidia-utils=387.34
Instal the bulldozer kernel on a FX 6350 with cpu CFLAGS="-O2 -march=bdver2 -pipe"
CXXFLAGS="${CFLAGS}"

I am currently hosting any piledriver packages due to this bug which prevents the kernel from building. Your suggestion to modify the CFLAGS will not work. See here.

loqs · 2018-02-13 19:23:17

@inglor what was the last version that did work and the first version you encountered that did not?
Have you tried changing the boot options such as removing quiet if present adding loglevel=7 earlyprintk=efi/earlyprintk=vga (vga for bios boot/efi for efi)

vinnom · 2018-02-15 18:56:40

graysky wrote:

john29 wrote:
packages for linux-ck-piledriver are not updated anymore on repo-ck and cannot update nvidia-ck-piledriver has it depend on nvidia-utils=387.34
Instal the bulldozer kernel on a FX 6350 with cpu CFLAGS="-O2 -march=bdver2 -pipe"
CXXFLAGS="${CFLAGS}"
I am currently hosting any piledriver packages due to this bug which prevents the kernel from building. Your suggestion to modify the CFLAGS will not work. See here.

Today I downloaded linux-ck-4.14.19-1.src.tar.gz and tried to build for piledriver. It worked and I already installed. If needed, I can share the build.

Funny thing is that yesterday I wasn't even being able to build generic kernel, but after @loqs help I could... then I checked here and gave it a try (my gcc version is 7.3.0)

graysky · 2018-02-16 20:11:50

vinnom wrote:

Today I downloaded linux-ck-4.14.19-1.src.tar.gz and tried to build for piledriver. It worked and I already installed. If needed, I can share the build.
Funny thing is that yesterday I wasn't even being able to build generic kernel, but after @loqs help I could... then I checked here and gave it a try (my gcc version is 7.3.0)

Something in 4.14.19 or gcc must have changed to allow it? In any case, I restored the piledriver package set. Files are online now. Thanks for reporting.

john29 · 2018-02-17 00:04:06

Thanks for restoring piledriver package set

john29 · 2018-02-18 06:28:05

Pliedriver kernel not updated again to linux-ck-piledriver-4-14.20-1

graysky · 2018-02-18 13:44:57

john29 wrote:

Pliedriver kernel not updated again to linux-ck-piledriver-4-14.20-1

Fixed. Will transition to the "old" kernel version in a few hours if the 4.15.x series checks out.

Arch Linux

#3251 2018-01-14 16:10:10

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

#3252 2018-01-15 14:45:27

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

#3253 2018-01-15 16:06:04

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

#3254 2018-01-16 12:15:13

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

#3255 2018-01-25 18:07:52

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

#3256 2018-01-25 19:37:07

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

#3257 2018-01-25 19:54:53

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

#3258 2018-01-29 12:22:04

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

#3259 2018-01-31 11:55:04

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

#3260 2018-01-31 13:40:56

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

#3261 2018-01-31 14:27:48

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

#3262 2018-01-31 20:29:19

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

#3263 2018-02-10 21:42:56

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

#3264 2018-02-10 22:31:48

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

#3265 2018-02-11 07:26:15

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

#3266 2018-02-11 12:14:45

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

#3267 2018-02-13 10:17:28

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

#3268 2018-02-13 11:05:44

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

#3269 2018-02-13 18:59:06

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

#3270 2018-02-13 19:23:17

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

#3271 2018-02-15 18:56:40

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

#3272 2018-02-16 20:11:50

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

#3273 2018-02-17 00:04:06

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

#3274 2018-02-18 06:28:05

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

#3275 2018-02-18 13:44:57

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

Board footer