The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

nous · 2012-02-20 11:29:00

alecmg wrote:

weird, I haven't had problems hibernating
then again I am not sure if I even use compression... where do I check?

After resuming once: dmesg | grep Compressor

See a sample.

lucke · 2012-02-20 11:30:28

About reverting the commit: http://lists.tuxonice.net/pipermail/tux … 07102.html

nous · 2012-02-20 11:40:48

Lockheed wrote:

Good to hear it might work.
If you find it works, will I have to wait till the next pf-kernel release to take advantage of it?
nous wrote:
If you absolutely want to use the latest linux-pf and BFS/LZO, you can easily so: just use your favourite AUR helper to get the PKGBUILD and, when prompted, select the menuconfig option, General setup ---> [ * ] BFS cpu scheduler (deselect it), exit saving configuration.
If I do that, and then install newer linux-pf with working BFS/TOI by copying kernel config from existing kernel, will it be enough just to change this setting back to ON, or does switching it OFF affect some other settings that will need to be reverted?

The PKGBUILD gives the option to use your CURRENT kernel's config through /proc/config.gz, so once you boot with your non-BFS kernel you won't need to enter menuconfig anymore.

Lockheed · 2012-02-20 11:42:48

nous wrote:

The PKGBUILD gives the option to use your CURRENT kernel's config through /proc/config.gz, so once you boot with your non-BFS kernel you won't need to enter menuconfig anymore.

I know, and hence my question. Please, reread it.

nous · 2012-02-20 13:37:54

Lockheed wrote:

If I do that, and then install newer linux-pf with working BFS/TOI by copying kernel config from existing kernel, will it be enough just to change this setting back to ON, or does switching it OFF affect some other settings that will need to be reverted?

nous wrote:

The PKGBUILD gives the option to use your CURRENT kernel's config through /proc/config.gz, so once you boot with your non-BFS kernel you won't need to enter menuconfig anymore.

Lockheed wrote:

I know, and hence my question. Please, reread it.

If you install a newer linux-pf with working BFS/TOI, either precompiled or with default settings, its config obviously will have BFS enabled. The -pf config follows the stock ARCH to the letter, except for BFS, Hz, TOI and Aufs3, which are extras. There's no point in copying over an older config once what you want starts working, that's why I misread your question.

If you still want to do this, the only thing BFS disables is "Automatic process group scheduling" in the General submenu, which CK deems useless/incompatible with his patchset.

Lockheed · 2012-02-20 14:50:37

I tried installing linux-pf with BFS disabled, but my system did not boot. I guess I will wait.

nous · 2012-02-20 21:04:01

Lockheed wrote:

I tried installing linux-pf with BFS disabled, but my system did not boot. I guess I will wait.

My bad. From what I remember, the BFS patch removes entirely the mainline kernel scheduler. That means that when one patches with BFS, one MUST use it or be left with no scheduler at all...

lucke · 2012-02-20 21:05:23

I'm using linux-pf with BFS disabled in the kernel config.

Last edited by lucke (2012-02-20 21:05:53)

nous · 2012-02-20 21:21:13

lucke wrote:

I'm using linux-pf with BFS disabled in the kernel config.

Then my memory has started deteriorating...

Lockheed · 2012-02-20 21:33:38

Have you had a chance to test it after work, nous?

gee · 2012-02-21 05:03:20

Not sure if that is the right way, but I have confirmed that this fixes building aufs3-util:

sed -i "s:__user::g" include/linux/aufs_type.h

it seems we are supposed to use

make headers_install

but I can't figure this one out.

If another aufs3 user confirms this, can you add it to your pkgbuild please?

Lockheed · 2012-02-25 21:57:44

post-factum, nous

Any developments on that matter? Was it something that could help with tux hibernation?

Last edited by Lockheed (2012-02-25 21:58:32)

nous · 2012-02-26 22:03:29

Lockheed wrote:

post-factum, nous
Any developments on that matter? Was it something that could help with tux hibernation?

I tried to find that commit at toi's git but couldn't.

post-factum · 2012-02-27 23:37:57

Lockheed wrote:

Any developments on that matter? Was it something that could help with tux hibernation?

No idea at the moment. I'm still waiting for Nigel's TOI update against last stable tree.

Lockheed · 2012-02-28 08:18:47

I think it would be really useful to add 'localmodconfig' with later option to customise it manually before the compilation starts.

nous · 2012-02-28 20:50:33

Lockheed wrote:

I think it would be really useful to add 'localmodconfig' with later option to customise it manually before the compilation starts.

I think what you're asking is already there.

Lockheed · 2012-03-02 15:31:56

Is there any reason not to use Hz value lower than the default 1000Hz for ph kernel?

I am on a laptop with 2-code processor and thought about switching to 250 or 300Hz (which I think would effectively translate into 500 and 600Hz).

el mariachi · 2012-03-02 17:03:56

300Hz will save you some battery on a laptop (can't remember the source for this, so if I'm wrong do tell), but will be a tad slower

post-factum · 2012-03-08 23:23:34

Here is my call for help with TOI+BFS issue: http://lists.tuxonice.net/pipermail/tux … 07183.html

post-factum · 2012-03-12 21:23:18

Possible temporary fix: http://lists.tuxonice.net/pipermail/tux … 07186.html
Give it a try.

Lockheed · 2012-03-12 21:55:44

post-factum wrote:

Possible temporary fix: http://lists.tuxonice.net/pipermail/tux … 07186.html
Give it a try.

What would be the downside of setting no_multithreaded_io to 1?
EDIT:
Where am I supposed to set it? I looked in kernel config, but found nothing alike.

Last edited by Lockheed (2012-03-12 22:21:36)

nous · 2012-03-12 22:35:34

I can confirm this workaround to work around.

The only downside is a hit on hibernation disk write speed, which is maybe 1 or 2 seconds of real time. I can live with it.
With setting no_multithreaded_io=1:

# dmesg|grep "I/O speed"
[29012.053060] - I/O speed: Write 240 MB/s, Read 346 MB/s.
[29051.157699] - I/O speed: Write 215 MB/s, Read 309 MB/s.

Without (and BFS disabled to allow for lzo compression):

# grep "I/O speed" /var/log/everything.log.?
/var/log/everything.log.1:Mar  5 18:02:13 hyperion kernel: [14644.634405] - I/O speed: Write 457 MB/s, Read 418 MB/s.
/var/log/everything.log.1:Mar  8 10:51:00 hyperion kernel: [13692.637524] - I/O speed: Write 333 MB/s, Read 277 MB/s.
/var/log/everything.log.1:Mar  9 22:03:38 hyperion kernel: [ 6851.629070] - I/O speed: Write 472 MB/s, Read 461 MB/s.
/var/log/everything.log.1:Mar 10 13:38:10 hyperion kernel: [23956.266669] - I/O speed: Write 353 MB/s, Read 357 MB/s.
/var/log/everything.log.1:Mar 11 08:30:34 hyperion kernel: [57086.648825] - I/O speed: Write 378 MB/s, Read 366 MB/s.
/var/log/everything.log.2:Mar  1 01:05:18 hyperion kernel: [ 6545.107878] - I/O speed: Write 345 MB/s, Read 353 MB/s.
/var/log/everything.log.2:Mar  2 10:07:28 hyperion kernel: [ 7372.325599] - I/O speed: Write 439 MB/s, Read 459 MB/s.
/var/log/everything.log.2:Mar  2 20:50:13 hyperion kernel: [ 5272.970279] - I/O speed: Write 468 MB/s, Read 419 MB/s.
/var/log/everything.log.2:Mar  3 18:12:47 hyperion kernel: [ 4986.890391] - I/O speed: Write 347 MB/s, Read 221 MB/s.
/var/log/everything.log.2:Mar  4 10:10:48 hyperion kernel: [28957.404124] - I/O speed: Write 333 MB/s, Read 331 MB/s.

nous · 2012-03-12 22:37:46

post-factum wrote:

Possible temporary fix: http://lists.tuxonice.net/pipermail/tux … 07186.html
Give it a try.

Great work. I had seen that freezing with CFS and thawing with BFS worked, but I never got suspicious.

post-factum · 2012-03-13 09:51:46

nous wrote:

Great work. I had seen that freezing with CFS and thawing with BFS worked, but I never got suspicious.

Time to remove https://aur.archlinux.org/packages.php?ID=40191 ?

Lockheed · 2012-03-14 11:09:08

I am getting this just after GRUB and before Arch starts booting:

[TuxOnIce] No storage allocators have been registered.

which apparently breaks my hibernation. What is this?

Arch Linux

#326 2012-02-20 11:29:00

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#327 2012-02-20 11:30:28

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#328 2012-02-20 11:40:48

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#329 2012-02-20 11:42:48

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#330 2012-02-20 13:37:54

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#331 2012-02-20 14:50:37

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#332 2012-02-20 21:04:01

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#333 2012-02-20 21:05:23

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#334 2012-02-20 21:21:13

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#335 2012-02-20 21:33:38

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#336 2012-02-21 05:03:20

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#337 2012-02-25 21:57:44

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#338 2012-02-26 22:03:29

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#339 2012-02-27 23:37:57

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#340 2012-02-28 08:18:47

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#341 2012-02-28 20:50:33

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#342 2012-03-02 15:31:56

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#343 2012-03-02 17:03:56

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#344 2012-03-08 23:23:34

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#345 2012-03-12 21:23:18

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#346 2012-03-12 21:55:44

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#347 2012-03-12 22:35:34

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#348 2012-03-12 22:37:46

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#349 2012-03-13 09:51:46

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

#350 2012-03-14 11:09:08

Re: The linux-pf thread; BFS/CK, TuxOnIce, BFQ, AUFS3

Board footer