You are not logged in.
Thanks. Enjoy.
CPU-optimized Linux-ck packages @ Repo-ck • AUR packages • Zsh and other configs
Offline
Yes! Works now! Thank you!
Offline
Since the latest update, I'm not getting more than ~2 hrs. of battery life out of linux-ck-sandybridge, compared to the previous ~6 hrs. The wifi interface uses ~ 6 watts of power, while the ethernet interface uses ~13 watts, and core temps run into nearly 60 degrees celsius (while normally being half that much). Arch's vanilla kernel does just fine, though. Any thoughts on troubleshooting this?
Lenovo T520
i7 2620QM Sandybridge quad-core
Intel HD 3000
Offline
Hi, I couldn't see if anyone had this problem so far, but I have some trouble using systemd to start user sessions with the ck-bulldozer kernel as it misses the cpu cgroup.
Is it possible that you add this option to the build? ( CONFIG_CGROUP_CPUACCT=y if I'm not mistaken)
So this is why I can never get the user sessions to work! I always thought that I had something configured all wonky that wouldn't allow this.
Thanks for the insight habarnam! I'm going to compile a Cl kernel myself evening get home and test this out!
BTW, thanks for another perfectly working update graysky (linuck-ck-ivybridge).
Offline
Since the latest update, I'm not getting more than ~2 hrs. of battery life ...
Can you verify that this is true using the 3.7.6-1 kernel from [testing]? 99 % of the time, these regressions are not due to the ck patchset, but are due to something in upstream. I would be shocked if this were any different.
I'm going to compile a Cl kernel myself evening get home and test this out!
You're going to add-back the option that CK specifically disabled???
BTW, thanks for another perfectly working update graysky (linuck-ck-ivybridge).
YW.
Last edited by graysky (2013-02-05 20:06:20)
CPU-optimized Linux-ck packages @ Repo-ck • AUR packages • Zsh and other configs
Offline
I'm getting a (total) power draw of approx. 9-10 W on battery with 3.7.6-1-ck, which is about how it's always been for me, so no problems here…
Last edited by MrCode (2013-02-05 20:19:00)
Offline
You're going to add-back the option that CK specifically disabled???
Yeah, I figure it is worth a try, as I always enjoy tinkering. My thoughts are that since you so graciously offer us a repo with such an awesome selection of optimized kernels, switching back will be no problem. Also, I figure that since it is actually enabled in the official's repo, maybe it is something that will be of no consequence either way? I really have no idea, but I have been wondering for quite some time why I cannot for the life of me get user sessions to work. So if this fixes it, great, if not, well I'll be back to repo-ck.
Edit: Okay, better idea... I am going to try with the normal kernel first, then if that works, then I will try modifying the ck config. I figure I should make sure that it works with the stock kernel before assuming that its inability to function correctly is caused by the kernel.
Edit2: Okay I have confirmed that user sessions work just fine on the official kernel. Though graysky, I am curious if you know of any specific reasoning why this configuration might have been disabled intentionally by ck? I have yet to get to actually testing a modified kernel, but you certainly know more about this patchset than I, so I'm just curious.
Last edited by WonderWoofy (2013-02-06 03:19:17)
Offline
is it okay that with ck-kernel Arch boots slower than with default, or i did smth wrong ?
Offline
Did you enable BFQ?
Offline
anonymous_user, of course
dmesg | grep -i schedu
[ 0.688662] io scheduler noop registered
[ 0.688663] io scheduler deadline registered
[ 0.688682] io scheduler cfq registered
[ 0.688699] io scheduler bfq registered (default)
[ 0.728329] BFS CPU scheduler v0.427 by Con Kolivas.
installed from AUR
Last edited by Perfect Gentleman (2013-02-06 04:59:15)
Offline
Is your boot faster if you disable BFQ? If so, just remember that BFQ focuses more on responsiveness rather than disk throughput.
Offline
i haven't tried to disable BFQ in ck, just booted with default kernel from core repo.
if it's okay, it' okay for me. i was just afraid that i did smth wrong
Offline
Okay, so I cannot turn on CONFIG_CGROUP_CPUACCT without turning off BFS... and that does not seem worthwhile.
Offline
i haven't tried to disable BFQ in ck, just booted with default kernel from core repo.
if it's okay, it' okay for me. i was just afraid that i did smth wrong
For repo users, BFQ is just a module like any other. It is therefore not enabled until the user enables it. Do this:
1) boot into the ck kernel and post the output of 'systemd-analyze'
2) boot into the ck kernel and do the same
CPU-optimized Linux-ck packages @ Repo-ck • AUR packages • Zsh and other configs
Offline
graysky,
1 - default kernel
systemd-analyze
Startup finished in 2332ms (kernel) + 5054ms (userspace) = 7386ms
dmesg | grep -i schedul
[ 0.840572] io scheduler noop registered
[ 0.840573] io scheduler deadline registered
[ 0.840593] io scheduler cfq registered (default)
2 - ck kernel
systemd-analyze
Startup finished in 3644ms (kernel) + 7442ms (userspace) = 11087ms
dmesg | grep -i schedule
[ 0.681531] io scheduler noop registered
[ 0.681532] io scheduler deadline registered
[ 0.681551] io scheduler cfq registered
[ 0.681568] io scheduler bfq registered (default)
[ 0.720797] BFS CPU scheduler v0.427 by Con Kolivas.
Offline
ANOKNUSA wrote:Since the latest update, I'm not getting more than ~2 hrs. of battery life ...
Can you verify that this is true using the 3.7.6-1 kernel from [testing]? 99 % of the time, these regressions are not due to the ck patchset, but are due to something in upstream. I would be shocked if this were any different.
Checking this now. I'm actually on battery power right now with linux-3.7.5, and the power drain exists there as well. Not sure what conditions gave me extra juice yesterday. I'll edit this post to let you know shortly.
EDIT: 3.7.6-1 seems to be doing fine.; I'm looking at ~2 hrs. on 34% charge.
Last edited by ANOKNUSA (2013-02-06 15:21:20)
Offline
Do you think it is possible to have the headers update before the kernel, so that it is possible to use the catalyst-hook package without installing the kernel twice with every kernel release?
To anyone who is not aware of this. The way to do this, is to add linux-ck-sandybridge-headers (or w/e version you use) to the SyncFirst option in pacman.conf.
Offline
@moscar - Interesting... but why is this not needed for the official kernel? In other words, what controls this order for it?
CPU-optimized Linux-ck packages @ Repo-ck • AUR packages • Zsh and other configs
Offline
@moscar - Interesting... but why is this not needed for the official kernel? In other words, what controls this order for it?
I use the catalyst-hook package, which adds linux-headers to SyncFirst. I was not aware of this when I asked about this some days ago. So it is needed for every kernel
I have this line in pacman.conf:
SyncFirst = pacman linux-headers linux-ck-sandybridge-headers
Offline
Hey Graysky, if you don't mind it would be nice if you could build the Piledriver ck-kernel. But if I am the only person who needs/wants it I can also use the generic package.
Offline
if you don't mind it would be nice if you could build the Piledriver ck-kernel.
I don't mind. Built and uploaded the ck-piledriver group. Plz report any issues.
% sudo pacman -S ck-piledriver
:: There are 7 members in group ck-piledriver:
:: Repository repo-ck
1) broadcom-wl-ck-piledriver 2) linux-ck-piledriver 3) linux-ck-piledriver-headers 4) nvidia-304xx-ck-piledriver
5) nvidia-ck-piledriver 6) virtualbox-ck-guest-modules-piledriver 7) virtualbox-ck-host-modules-piledriver
Enter a selection (default=all):
CPU-optimized Linux-ck packages @ Repo-ck • AUR packages • Zsh and other configs
Offline
I don't mind. Built and uploaded the ck-piledriver group. Plz report any issues.
Thank you very much , I will test it tomorrow.
Offline
kernel 3.7.8-1-{ARCH,ck} with laptop-mode-tools fail to restore from suspend some keyboard/mouse input until <CTRL+ALT+BACKSPACE>
As a workaround uninstall laptop-mode-tools.
Last edited by ontobelli (2013-02-15 17:50:51)
Offline
@ontobelli - Since the problem affects both the ARCH and ck kernel, it is not a function of the ck patchset. Plz don't report such things to this thread since it is only for repo related discussion.
CPU-optimized Linux-ck packages @ Repo-ck • AUR packages • Zsh and other configs
Offline
Hi Graysky,
I am using your linux-ck-atom build and I just noticed that CONFIG_X86_GENERIC was enabled which has the following effect to the compilation:
Makefile_32.cpu:cflags-$(CONFIG_X86_GENERIC) += $(call tune,generic,$(call tune,i686))
IMO, this might slightly defeat the purpose of having very targeted binaries. All 32 bits targets are possibly affected by that....
Offline