View previous topic :: View next topic |
Author |
Message |
mi_unixbird Tux's lil' helper
Joined: 24 Jul 2015 Posts: 130
|
Posted: Wed Nov 18, 2015 3:36 am Post subject: |
|
|
Yeah, just copying the default ck-sources ebuild and changing those things as well as changing a compression format from bz2 to xz seems to work. The =ck-sources-4.1.6 also included the BFQ I/O Scheduler though, or at least mine does.
I had to patch that one in manually but with it it seems to work. Not sure if manually patching inside kernel sources is a good idea. The ebuild tekst seems to imply that BFQ scheduler is drawn in but no such thing happens, maybe I needed to change something else for that. It seems to work though. _________________ execctl --path exec filectl --current-directory list |
|
Back to top |
|
|
mi_unixbird Tux's lil' helper
Joined: 24 Jul 2015 Posts: 130
|
Posted: Wed Nov 18, 2015 8:16 pm Post subject: |
|
|
So I switched back to 4.1.6 as my system crashed twice within 24 hours now, before this point it just didn't crash.
But I'm not sure if it's because of this or my switch from ext4 to btrfs on the same day, so I'll be running 4.1.6 with btrfs to see if it continues to crash. _________________ execctl --path exec filectl --current-directory list |
|
Back to top |
|
|
khayyam Watchman
Joined: 07 Jun 2012 Posts: 6227 Location: Room 101
|
Posted: Wed Nov 18, 2015 8:24 pm Post subject: |
|
|
mi_unixbird wrote: | So I switched back to 4.1.6 as my system crashed twice within 24 hours now, before this point it just didn't crash. But I'm not sure if it's because of this or my switch from ext4 to btrfs on the same day, so I'll be running 4.1.6 with btrfs to see if it continues to crash. |
mi_unixbird ... if you look in your kmsg logs you should see the backtrace, and so the component causing the crash.
I've not moved to 4.x so I can't comment, but I generally hold off upgrading to a new major revision until its been better tested.
best ... khay |
|
Back to top |
|
|
ulenrich Veteran
Joined: 10 Oct 2010 Posts: 1483
|
Posted: Thu Nov 19, 2015 12:08 am Post subject: |
|
|
Hey, hey ... it is just too early to blindly assume bfs is the crasher. Linux-4.3.0 is the zero release without fixes and probably is full of bugs.
For me it works a few hours now. |
|
Back to top |
|
|
thunderrd n00b
Joined: 20 Aug 2010 Posts: 59
|
Posted: Sat Dec 12, 2015 1:45 pm Post subject: |
|
|
I'd really like to see a new version get into the tree after seeing this in Con's IRC channel yesterday:
Quote: | * conman (~con@220.240.178.83) has joined
<conman> woohoo improved throughput on next bfs
<entil> conman: you rock!
<conman> I diddled with 3 lines of code for a month
<conman> no shit
<conman> and it finally paid off
<conman> the patch is ridiculously small
<entil> there must be a penus joke somewhere about diddling with something small for a month ;D
<conman>
<conman> http://ck.kolivas.org/patches/bfs/4.0/4.3/Pending/bfs465-466.patch
<conman> will make a ck2 later
<entil> let us know; I'm tight on hard drive space (in the year 2015, but hey, laptop at work) so I'll need to put some effort into compiling this later
<conman>
* ChanServ gives channel operator status to conman
*conman has changed the topic to: http://ck-hack.blogspot.com ; Latest -ck: 4.3-ck2 | http://kernel.kolivas.org | lrzip 0.621 http://lrzip.kolivas.org | latest BFS: 0.466 |
|
|
Back to top |
|
|
khayyam Watchman
Joined: 07 Jun 2012 Posts: 6227 Location: Room 101
|
Posted: Sat Dec 12, 2015 2:53 pm Post subject: |
|
|
thunderrd wrote: | I'd really like to see a new version get into the tree after seeing this in Con's IRC channel yesterday: |
thunderrd ... this is a diff against ck-sources-4.1.6.ebuild for ck-sources-4.3.2.ebuild (using 4.3-ck2) ... untested, but I did build the manifest and so *it should* find the required sources/patches, and merge.
diff ~portdir/sys-kernel/ck-sources/ck-sources-4.1.6.ebuild ck-sources-4.3.2.ebuild: | 17c17
< K_GENPATCHES_VER="10"
---
> K_GENPATCHES_VER="3"
19c19,20
< K_DEBLOB_AVAILABLE="1"
---
> K_DEBLOB_AVAILABLE="0"
> K_KDBUS_AVAILABLE="0"
39c40
< CK_FILE="patch-${K_BRANCH_ID}-ck${CK_VERSION}.bz2"
---
> CK_FILE="patch-${K_BRANCH_ID}-ck${CK_VERSION}.xz" |
... so, not much you'd need to change ... and so you should be able to bump the package in a local overlay.
best ... khay |
|
Back to top |
|
|
thunderrd n00b
Joined: 20 Aug 2010 Posts: 59
|
Posted: Sat Dec 12, 2015 4:04 pm Post subject: |
|
|
Thank you for that.
I will give it a try as soon as I have time, and report back. It's a pity; it's been such a long time since this package has had a steady maintainer. I feel bad asking for a version bump sometimes, I know that everyone is a volunteer. |
|
Back to top |
|
|
spectromas n00b
Joined: 13 Jul 2014 Posts: 50
|
Posted: Sat Dec 12, 2015 4:12 pm Post subject: |
|
|
It does have a steady maintainer doesn't it? The patches themselves haven't been updated since August or something http://ck-hack.blogspot.com/
As far as I remember the ebuild gets updated fairly quickly. |
|
Back to top |
|
|
thunderrd n00b
Joined: 20 Aug 2010 Posts: 59
|
Posted: Sun Dec 13, 2015 7:16 am Post subject: |
|
|
The diff that khayyam provided works fine, builds and boots. There are quite a few new options in .config, though, so be ready.
As for the maintainer, I think yngwin is doing it now, but he's likely busy as there haven't been any updates in portage for two major versions [since 4.1.6, now at 4.3]. I'll see if I can talk to him in #gentoo if he's around. |
|
Back to top |
|
|
khayyam Watchman
Joined: 07 Jun 2012 Posts: 6227 Location: Room 101
|
Posted: Sun Dec 13, 2015 1:35 pm Post subject: |
|
|
thunderrd wrote: | The diff that khayyam provided works fine, builds and boots. There are quite a few new options in .config, though, so be ready. |
thunderrd ... ok, good. I didn't intend to use it myself as I'll probably stick with 3.x for a while ... starting a config from scratch, which I tend to do for major kernel revisions, is not something I have time for right now.
best ... khay |
|
Back to top |
|
|
spectromas n00b
Joined: 13 Jul 2014 Posts: 50
|
Posted: Sun Dec 13, 2015 1:38 pm Post subject: |
|
|
thunderrd wrote: | there haven't been any updates in portage for two major versions [since 4.1.6, now at 4.3]. |
Yes, because there were not any ck patches released. Look at his blog, 4.1.6 was the last version until the recent new one. The guy from pf-sources was maintaining the patch and releasing it for his kernel sources but between 4.1.6 and now there was nothing to update to, it wasn't because the ebuild is unmaintained. |
|
Back to top |
|
|
Chiitoo Administrator
Joined: 28 Feb 2010 Posts: 2718 Location: Here and Away Again
|
Posted: Mon Dec 14, 2015 1:02 pm Post subject: |
|
|
The maintainer, yngwin, is in fact unavailable at this time, as can be seen here: Unavailable Gentoo Developers (devaway)
With that in mind, creating bug reports for version bumps for packages they maintain would probably be a good idea, so that another developer might perhaps jump to it. ^^ _________________ Kindest of regardses. |
|
Back to top |
|
|
mi_unixbird Tux's lil' helper
Joined: 24 Jul 2015 Posts: 130
|
Posted: Tue Dec 15, 2015 5:06 pm Post subject: |
|
|
So I've been running "4.3.2-ck" for a couple of days now without any issues. I think my initial issues were btrfs but you can never really tell. Latency certainly feels better but this might also be because I switched to on-demand cpufreq which I'm pretty sure is a bigger factor all in all.
I just took the 4.1.6-ck ebuild and put K_GENPATCHES_VER=3, changed the extension of the archive from bz to xz and that was it really.
I'm also back on ext4 as btrfs was giving me issues on the 4.1.6-ck kernel as well. _________________ execctl --path exec filectl --current-directory list |
|
Back to top |
|
|
Chiitoo Administrator
Joined: 28 Feb 2010 Posts: 2718 Location: Here and Away Again
|
Posted: Tue Dec 15, 2015 7:13 pm Post subject: ><)))°€ |
|
|
One may want to set CK_VERSION="3" due to “detrimental effect on interactivity” in the earlier version.
Posted today at -ck hacking:
ck at ck-hack.blogspot.com wrote: | Tuesday, 15 December 2015
BFS 467, linux-4.3-ck3
Announcing an updated BFS for linux-4.3 based kernels.
BFS by itself:
4.3-sched-bfs-467.patch
-ck branded linux-4.3-ck3 patches:
4.3-ck3
After my initial enthusiasm regarding the improved throughput on the previous BFS release, I unfortunately had some reports of regressions in interactive behaviour for the first time in a while, both on this forum and through other channels. So for the first time in a while, I've released a -ck3 with yet another updated BFS to address this since I can't stand having a dodgy release out for any extended period.
With this release what I've done is reinstate an old tunable that used to be on my scheduler patches many years ago
/proc/sys/kernel/interactive
This has two settings, 1 for on and 0 for off. By default - of course since this is BFS - it is set to 1. What it does is prioritise latency over throughput in mode 1 and vice versa in mode 0. In addition to addressing the latency issues in the previous kernel, mode 1 actually completely turns off all soft affinity scheduling in the kernel, for the lowest possible latencies all round, so this may be the first kernel with an improvement in latency in a while too.
Bear in mind none of these changes make any difference on uniprocessor kernels so there is no need for UP users to update unless they need the build fixes that came with BFS466+.
Amusingly enough, linux-4.3.2 wouldn't boot for me for unrelated reasons so I'm using 4.3.0-ck3 myself. So if you can't get 4.3.2 to boot, roll back.
Enjoy!
お楽しみください
Posted by ck at 17:51 |
I just built mine (using 4.3.2 sources), but haven't tried booting yet. _________________ Kindest of regardses. |
|
Back to top |
|
|
mi_unixbird Tux's lil' helper
Joined: 24 Jul 2015 Posts: 130
|
Posted: Wed Dec 16, 2015 12:13 am Post subject: |
|
|
No can't do m'lady. I already used my reboot quota for this year on the old update. The e-peen must remain strong. More than two reboots per year and you're basically a Windows and/or Ubuntu user. _________________ execctl --path exec filectl --current-directory list |
|
Back to top |
|
|
ulenrich Veteran
Joined: 10 Oct 2010 Posts: 1483
|
Posted: Wed Dec 16, 2015 12:50 pm Post subject: |
|
|
I want to know about the new 0_or_1 setting of BFS 467, linux-4.3-ck3
Is it possible to change at runtime the behavior of BFS by just
echo "0" >/proc/sys/kernel/interactive
... which I would prefer having to compile quickly some ebuild
and then afterwards
echo "1" >/proc/sys/kernel/interactive
to get my snappy desktop again? |
|
Back to top |
|
|
Zentoo Apprentice
Joined: 18 Nov 2002 Posts: 197 Location: /dev/console
|
Posted: Thu Dec 17, 2015 2:03 pm Post subject: |
|
|
A little quick message to tell that I'm really happy with ck-sources kernel actually.
I'm using ck-sources since 4.04-r1.
I've posted this void message to keep track of this thread.
Thanks to everyone involved in this ebuild ! _________________ ACCEPT_KEYWORDS="~amd64"
USE="-systemd"
Desktop: openbox|picom|ROX-Filer|wbar|window maker dockapps
Hardware: Ryzen 7950X | 64 Gb | Nvidia 3080Ti |
|
Back to top |
|
|
Chiitoo Administrator
Joined: 28 Feb 2010 Posts: 2718 Location: Here and Away Again
|
Posted: Thu Dec 17, 2015 3:07 pm Post subject: ><)))°€ |
|
|
ulenrich wrote: | I want to know about the new 0_or_1 setting of BFS 467, linux-4.3-ck3
Is it possible to change at runtime the behavior of BFS by just
echo "0" >/proc/sys/kernel/interactive
... which I would prefer having to compile quickly some ebuild
and then afterwards
echo "1" >/proc/sys/kernel/interactive
to get my snappy desktop again? |
That's pretty much how I understand it. Setting 0 will make things happen like they do with BFS 466.
I tried flipping it around betwixt 1 and 0, while building Wine and running XCOM: Enemy Unknown. While 0, switching between applications and stuff would become quite noticeably less snappy-like, and I had a few short freezes even. When setting 1, things would seem to become fluid and snappy again.
Not a great test by any means, but it's something, I maybe guess. ^^ _________________ Kindest of regardses. |
|
Back to top |
|
|
ulenrich Veteran
Joined: 10 Oct 2010 Posts: 1483
|
Posted: Thu Dec 17, 2015 6:04 pm Post subject: |
|
|
I have done it:
echo "0">/proc/sys/kernel/interactive
because I am transitioning to gcc-5.3 the whole night, by
emerge gcc; gcc-config set 2; emerge libtool
revdep-rebuild --library 'libstdc\+\+.so.6'
Arghh .... this is a long job compiling 196 C++ related ebuilds
The linux43-bfs467 kernel is rock solid!
I suspect the new C++11 Abi needs more time for compilation than the older C++ language of 99.
Or the gcc-5.3 compiler needs optimization
Or the program source code could be optimized
.... |
|
Back to top |
|
|
Zentoo Apprentice
Joined: 18 Nov 2002 Posts: 197 Location: /dev/console
|
Posted: Fri Dec 18, 2015 8:12 am Post subject: |
|
|
@ulenrich:
Compilation time with gcc-5.3 with new C++11 ABI don't take significantly more time than gcc-4 with C++99ABI:
Code: | Fri Oct 30 02:22:04 2015 >>> media-video/vlc-2.2.1-r1
merge time: 2 minutes and 32 seconds. (gcc4)
Tue Dec 15 09:40:51 2015 >>> media-video/vlc-2.2.1-r1
merge time: 2 minutes and 19 seconds. (gcc5)
Tue Nov 10 03:55:41 2015 >>> app-office/libreoffice-5.0.3.2
merge time: 1 hour, 4 minutes and 58 seconds. (gcc4)
Tue Dec 15 16:57:46 2015 >>> app-office/libreoffice-5.0.3.2
merge time: 1 hour, 7 minutes and 24 seconds. (gcc5) |
But gcc5 compilation take more time than gcc4.
Code: |
Mon Dec 14 20:35:35 2015 >>> sys-devel/gcc-4.9.3
merge time: 12 minutes and 45 seconds.
Mon Dec 14 16:25:00 2015 >>> sys-devel/gcc-5.3.0
merge time: 20 minutes and 46 seconds. |
In fact I've observed similar compilation time once I got recompiled my kernel with gcc 5 and reboot.
And my quick CPU scalability/memory throughput bench show that I got a +2.8 % gain with new gcc5.3 using kernel 4.1.6-ck.
Quick bench: Code: | cat /dev/zero | pv -ar | pbzip2 > /dev/null |
=> 780 MiB/s with gcc4 kernel
=> 800 MiB/s with gcc5 kernel.
So I presume that gcc5 compilation is in reality a bit slower that gcc4 but it's balanced by a better throughput with a gcc5 compiled kernel. _________________ ACCEPT_KEYWORDS="~amd64"
USE="-systemd"
Desktop: openbox|picom|ROX-Filer|wbar|window maker dockapps
Hardware: Ryzen 7950X | 64 Gb | Nvidia 3080Ti |
|
Back to top |
|
|
mi_unixbird Tux's lil' helper
Joined: 24 Jul 2015 Posts: 130
|
Posted: Sat Dec 19, 2015 2:55 am Post subject: |
|
|
4.3.2-ck has crashed a couple of times already on me the last few days, not sure if this is due to the 4.3.2 kernel or the ck version, I'm going to try with ck3 (I had 1) and if that crashes again I'll be back at 4.1.6 I guess. _________________ execctl --path exec filectl --current-directory list |
|
Back to top |
|
|
Chiitoo Administrator
Joined: 28 Feb 2010 Posts: 2718 Location: Here and Away Again
|
Posted: Sun Dec 20, 2015 9:36 pm Post subject: |
|
|
I believe I'm on my third day with 4.3.2 and BFS 467, without a single crash or any other issues (aside from what the quick test I mentioned earlier brought up; I have not played more with the /proc/sys/kernel/interactive setting though, having left it at 1). _________________ Kindest of regardses. |
|
Back to top |
|
|
mi_unixbird Tux's lil' helper
Joined: 24 Jul 2015 Posts: 130
|
Posted: Mon Dec 21, 2015 5:27 pm Post subject: |
|
|
Wel, 4.3.2-ck3 has not crashed on me yet since that last message. So it looks okay.
The crashes are rarely arbitrary though. Like, one of them happened when I was testing some OCaml code and accidentally made a fifo with it with decimal permissions 600 instead of octal and the entire box blew. _________________ execctl --path exec filectl --current-directory list |
|
Back to top |
|
|
thunderrd n00b
Joined: 20 Aug 2010 Posts: 59
|
Posted: Wed Dec 23, 2015 2:42 am Post subject: |
|
|
Agreed, the -ck3 patches are doing a good job for me as well.
Version bump request has been filed today, let's see if someone picks it up:
https://bugs.gentoo.org/show_bug.cgi?id=569262 |
|
Back to top |
|
|
Ant P. Watchman
Joined: 18 Apr 2009 Posts: 6920
|
Posted: Wed Dec 23, 2015 5:25 am Post subject: |
|
|
2¢:
I've been using zen-sources.git's version of the -ck stuff for a long time now, and I've had maybe one or two inexplicable kernel crashes in that time (setting CONFIG_PANIC_TIMEOUT turned out a good idea) and a few failures to compile at all around 4.2. But to be honest the only real instability I've seen all year was the one time I turned blk-mq on. That was... bad. |
|
Back to top |
|
|
|