Official thread: "zenmm-sources" - Latest: 2.6.27-rc1-zenmm0

Message

cheater1034 · Post by **cheater1034** » Tue Jul 08, 2008 8:01 am

I got around to making some commits today.

Fixes and some improvements:
1) Fix PPC build failure
2) Fix paravirt.c build failure (x86 build failure)
3) Kexec -v12
4) Some other stuff see for yourself.

mroconnor · Post by **mroconnor** » Tue Jul 08, 2008 3:37 pm

Most recent pull and pull of tp_smapi was this AM and it is running great!!!!

Great work guys.

Has anyone been looking at power consumption on the -zenmm kernels?

mroconnor · Post by **mroconnor** » Tue Jul 08, 2008 5:05 pm

For some reason suspend(TuxOnIce) works too well for me, it never wakes up.

I would post /var/log/hibernate.log but it really shows nothing.

cheater1034 · Post by **cheater1034** » Tue Jul 08, 2008 7:03 pm

mroconnor wrote:For some reason suspend(TuxOnIce) works too well for me, it never wakes up.

I would post /var/log/hibernate.log but it really shows nothing.

There is no tuxonice in zenmm, so maybe it doesn't work at all :-p ?

mroconnor · Post by **mroconnor** » Tue Jul 08, 2008 7:08 pm

That might explain it.....maybe it was the kernel based hibernate combined with the hibernate script.
/me scratches head.

mroconnor · Post by **mroconnor** » Tue Jul 08, 2008 7:20 pm

Ok it is only an issue with

Code: Select all

CONFIG_SUSPEND

Not sure what the issue is though, nothing is logged.

whitesouls · Post by **whitesouls** » Wed Jul 09, 2008 3:00 am

Hi there..thanks for the great kernel.. Any idea why am i getting this?

Code: Select all

SkullCrusherX ~ # emerge -av unionfs
WARNING: repository at /usr/local/portage is missing a repo_name entry

These are the packages that would be merged, in order:

Calculating dependencies... done!
[ebuild  N    ] sys-fs/unionfs-1.5_pre200701042308  USE="acl -debug -nfs" 160 kB 

Total: 1 package (1 new), Size of downloads: 160 kB

Would you like to merge these packages? [Yes/No] y
>>> Verifying ebuild Manifests...

>>> Emerging (1 of 1) sys-fs/unionfs-1.5_pre200701042308 to /
>>> Downloading 'http://ftp.lecl.net/pub/gentoo/distfiles/unionfs-1.5pre-cvs200701042308.tar.gz'
--2008-07-09 10:57:56--  http://ftp.lecl.net/pub/gentoo/distfiles/unionfs-1.5pre-cvs200701042308.tar.gz
Resolving ftp.lecl.net... 124.217.198.26
Connecting to ftp.lecl.net|124.217.198.26|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 163418 (160K) [application/x-gzip]
Saving to: `/usr/portage/distfiles/unionfs-1.5pre-cvs200701042308.tar.gz'

100%[======================================================================================>] 163,418      128K/s   in 1.2s    

2008-07-09 10:57:58 (128 KB/s) - `/usr/portage/distfiles/unionfs-1.5pre-cvs200701042308.tar.gz' saved [163418/163418]

 * unionfs-1.5pre-cvs200701042308.tar.gz RMD160 SHA1 SHA256 size ;-) ...                                                  [ ok ]
 * checking ebuild checksums ;-) ...                                                                                      [ ok ]
 * checking auxfile checksums ;-) ...                                                                                     [ ok ]
 * checking miscfile checksums ;-) ...                                                                                    [ ok ]
 * checking unionfs-1.5pre-cvs200701042308.tar.gz ;-) ...                                                                 [ ok ]
 * Determining the location of the kernel source code
 * Found kernel source directory:
 *     /usr/src/linux
 * Found kernel object directory:
 *     /lib/modules/2.6.26-rc8-zen1/build
fatal: cannot describe '226f29e401141f4f46e121ff31e55fedf6daadbe'
 * Found sources for kernel version:
 *     2.6.26-rc8-zen1
 * 
 * you need the proper kernel version!
 * 
 * kernel: 2.4.x (x>19)    Version: 1.0.14
 * kernel: 2.6.x (x<9)     Version: Not Supported
 * kernel: 2.6.9 - 2.6.15  Version: 1.1.5
 * kernel: 2.6.16          Version: 1.2
 * kernel: 2.6.17          Version: 1.3
 * kernel: 2.6.18          Version: 1.4
 * kernel: 2.6.19          Version: 1.5
 * 
 * ERROR: sys-fs/unionfs-1.5_pre200701042308 failed.
 * Call stack:
 *                            ebuild.sh, line   49:  Called pkg_setup
 *   unionfs-1.5_pre200701042308.ebuild, line   40:  Called die
 * The specific snippet of code:
 *   		die
 *  The die message:
 *   (no error message)
 * 
 * If you need support, post the topmost build error, and the call stack if relevant.
 * A complete build log is located at '/var/log/portage/sys-fs:unionfs-1.5_pre200701042308:20080709-025756.log'.
 * The ebuild environment file is located at '/var/tmp/portage/sys-fs/unionfs-1.5_pre200701042308/temp/die.env'.
 * 

 * Messages for package sys-fs/unionfs-1.5_pre200701042308:

 * 
 * you need the proper kernel version!
 * 
 * 
 * ERROR: sys-fs/unionfs-1.5_pre200701042308 failed.
 * Call stack:
 *                            ebuild.sh, line   49:  Called pkg_setup
 *   unionfs-1.5_pre200701042308.ebuild, line   40:  Called die
 * The specific snippet of code:
 *   		die
 *  The die message:
 *   (no error message)
 * 
 * If you need support, post the topmost build error, and the call stack if relevant.
 * A complete build log is located at '/var/log/portage/sys-fs:unionfs-1.5_pre200701042308:20080709-025756.log'.
 * The ebuild environment file is located at '/var/tmp/portage/sys-fs/unionfs-1.5_pre200701042308/temp/die.env'

Thank you in advance..

cheater1034 · Post by **cheater1034** » Wed Jul 09, 2008 3:38 am

whitesouls wrote: Hi there..thanks for the great kernel.. Any idea why am i getting this?
Code: Select all
SkullCrusherX ~ # emerge -av unionfs
<long emerge>

That's old (2007), unionfs is in the mainline kernel you should just be able to build it in there. (Filesystems -> Layered Filesystem)

whitesouls · Post by **whitesouls** » Wed Jul 09, 2008 3:44 am

oh..sorry..thanks for the tip..

gimpel · Post by **gimpel** » Wed Jul 09, 2008 8:30 am

w00t! I love this kernel...
click
... gcc-4.3.1 compiled kvm.

time to say bye to vbox & co

mroconnor · Post by **mroconnor** » Wed Jul 09, 2008 1:25 pm

I have to agree love this kernel too.

I do have a question with regard to CFQ vs Deadline. I really couldnt tell the difference with normal browsing and email but when I load up the system by doing something liek this cp ~/movies/*.avi . or anything really disk intensive CFQ just slows the whole system to an almost unusable point.

Does anyone have experience with the different I/O schedulers and what has worked best for them on a desktop(laptop in my case)? I wasn't sure if I had maybe implemented to conflicting things in my .config or if it had anything to do with file systems, etc.

Thanks.

tranquilcool · Post by **tranquilcool** » Thu Jul 10, 2008 7:48 am

kernel panics with the last pulls. worked until the first time the CFQ tunables
were inserted.
would love to reset to that point.

EDIT:

after last git pull kernel has been consistent. no panics.

kernelOfTruth · Post by **kernelOfTruth** » Thu Jul 10, 2008 7:53 pm

Merge branch 'compcache'

great job !
thanks a lot, guys; I'm appreciating this kernel more & more (very very stable & fast)

for those don't realizing the "stableness" compared to regular zen-sources:

try compiling your kernel with make -j 25 and keep on working sometime afterwards with the box + additionally hearing webradio during compilation

the difference: with regular zen-sources the kernel is practically dead (i/o error, nothing works anymore), with zenmm: "keep on trucking"

some info on compcache (for those who don't know what it is):

Hi All,

(sending to lkml since I didn't get any reply at linux-mm).

This implements a RAM based block device which acts as swap disk.
Pages swapped to this disk are compressed and stored in memory itself.
This allows more applications to fit in given amount of memory. This is
especially useful for embedded devices, OLPC and small desktops
(aka virtual machines).

Project home: http://code.google.com/p/compcache/

It consists of following components:
- compcache.ko: Creates RAM based block device
- tlsf.ko: Two Level Segregate Fit (TLSF) allocator
- LZO de/compressor: (Already in mainline)

Project home contains some performance numbers for TLSF and LZO.
For general desktop use, this is giving *significant* performance gain
under memory pressure. For now, it has been tested only on x86.

Thanks,
Nitin

source: http://lwn.net/Articles/274653/

cheater1034 · Post by **cheater1034** » Fri Jul 11, 2008 3:45 am

drivers/ide/ide-disk.c
Anyone notice this failing to build? it's due to disk protection on hdaps <patch>, i wrote a patch for it and it all seems to be well now so if you stopped using zenmm for this reason without reporting the problem, it's now fixed in git.

Here's the diff:
http://nopaste.com/p/aGpthaJx2

KoT wrote:great job !
thanks a lot, guys; I'm appreciating this kernel more & more (very very stable & fast)

for those don't realizing the "stableness" compared to regular zen-sources:

Fast = yes, i wouldn't consider it stable by any means, because it's -mm, and we havent tested it on any architectures besides x86 and x86_64, and we had an oops reported (but he needs to try the latest git pull, i dont think he'll have it ne more)

By fastest, i mean, i have reiser4 w/ lzo and was using iosched=AS, with CONFIG_BOOST_PRIVILEGED_TASKS=y, and LL-Desktop settings, and i ran cpu-burn-in, emerge -e system, emerge -e world, emerge -uD world, emerge boost, make (on a kernel build).

To top it off, i was also listening to a song, and playing a video while browsing the web in FF3. I had no lag at all, on anything. Before, when i would change tab in firefox when doing something else, or make a new tab in gnome-terminal (yes, im also running gnome 2.22), it would lag a bit, but under all that load it was all instantaneous (I was in shock)

This was just from today, it's going to get a version bump in a little bit from me.

kernel panics with the last pulls. worked until the first time the CFQ tunables
were inserted.
would love to reset to that point.

EDIT:

after last git pull kernel has been consistent. no panics.

can i see the git log on your current pull, i want to see the changes there. (i know zen-tunables and cfs-boost were tinkered with for awhile, and they are finally in good working order, i just want to see which one of the 2 was bad)

I have to agree love this kernel too.

I do have a question with regard to CFQ vs Deadline. I really couldnt tell the difference with normal browsing and email but when I load up the system by doing something liek this cp ~/movies/*.avi . or anything really disk intensive CFQ just slows the whole system to an almost unusable point.

Does anyone have experience with the different I/O schedulers and what has worked best for them on a desktop(laptop in my case)? I wasn't sure if I had maybe implemented to conflicting things in my .config or if it had anything to do with file systems, etc.

Thanks.

It depends on the FS as well. For example, on reiserfs or reiser4 i would never use cfq/bfq (they just don't work right, i notice significant change). Though CFQ is massively different than AS or Deadline. I generally bounce back and forth between AS/Deadline, right now i'm on AS and loving it.

Although, if your / is on ext*, then CFQ is generally the best choice from what i've found.

cheater1034 · Post by **cheater1034** » Fri Jul 11, 2008 5:11 am

[VERSION Bump]
2.6.26-rc8-zenmm2 "L-L-L-Like a Lollipop"
INFO: It was time for a version bump, LOTS of changes since zenmm1 was first put on the mirror, I suggest upgrading for this one (although it's just git, so git fetch; git pull).
Not only does it now work on PPC systems and paravirt.c build is fixed, tp_smapi has also been reported working, and CFS-Boost and the zen-tunables add a NOTICABLE performance increase on desktop systems. (I myself am using the CFS boost, and have the pre-defined LL Desktop settings.

Also compcache, etc. A great deal of the commits below came from today, and i know over in #zen-sources people love how it's running (including myself), i'll talk to rmh about releasing it because i'm in love with it lol.

Code: Select all

commit fe71f1a80a0600f1146a08ef9240850277bf1430
Author: Brandon Berhent <cheater1034@gmail.com>
Date:   Fri Jul 11 00:56:22 2008 -0400

    [VERSION] I couldnt find zenmm version info anywhere, pulling zenmmversion gave me rc5
    Oh well, version bump rc8-zenmm2

commit 14a50a501812b30e8f5d2922adc54a2ebe8c1911
Merge: b9c4c26... df6b519...
Author: Brandon Berhent <cheater1034@gmail.com>
Date:   Thu Jul 10 20:23:02 2008 -0400

    Merge branch 'lkml'

commit df6b519bcb51c31b3b6d514bd23ef53a6422d60b
Author: Brandon Berhent <cheater1034@gmail.com>
Date:   Thu Jul 10 20:20:45 2008 -0400

    Revert "[PATCH] pxamci: fix byte aligned DMA transfers"
    
    This reverts commit 12e85ba3aa9817793dfd12a451b687559b8130a0.

commit 70b0748592942740ce444f349b4ae77870b270e5
Author: Brandon Berhent <cheater1034@gmail.com>
Date:   Thu Jul 10 20:19:42 2008 -0400

    Revert "ov7670: clean up ov7670_read semantics"
    
    This reverts commit 779406b6eb5d6f7631f1dbb441d52b87b57e0ba4.

commit 32fe5879a47de74816e09317fc05777674f1bc4c
Author: Brandon Berhent <cheater1034@gmail.com>
Date:   Thu Jul 10 20:19:20 2008 -0400

    Revert "[Kexec] kexec-jump v12"
    
    This reverts commit c4d9a6675f7e8325d765a029e433e67342507594.

commit b9c4c2683e9ed57c5821c4126588f2771d61a1b2
Merge: d2fff43... 2240dfb...
Author: Brandon Berhent <cheater1034@gmail.com>
Date:   Thu Jul 10 20:07:31 2008 -0400

    Merge branch 'zen-tunables'
commit 2240dfb6eb817a372b1a0c0662aff946bcd7890b
Author: Brandon Berhent <cheater1034@gmail.com>
Date:   Thu Jul 10 20:04:22 2008 -0400

    OOPS: Clean up *.orig and *.rej

commit e2dff9a3a27bef9354636833dc2b11591fc2882d
Author: Brandon Berhent <cheater1034@gmail.com>
Date:   Thu Jul 10 20:02:27 2008 -0400

    CFS-Boost: Re-nice privileged tasks to give them more cpu power.
    This is especially useful for desktop systems running X.

commit d2fff43b998a59c1f7dcd1dad619423b92c6269a
Merge: 3bcb3e2... 15e6bc0...
Author: Brandon Berhent <cheater1034@gmail.com>
Date:   Thu Jul 10 18:08:22 2008 -0400

    Merge branch 'lkml'

commit 4fe455cf7bdd3e42389611aaad4de5d48768026b
Author: Brandon Berhent <cheater1034@gmail.com>
Date:   Thu Jul 10 17:11:55 2008 -0400

    Delete un-used kconfig options

commit 3bcb3e2266c4d252788da1174910cc53b4fd683b
Merge: 44309a8... a9082da...
Author: Brandon Berhent <cheater1034@gmail.com>
Date:   Thu Jul 10 17:05:52 2008 -0400

    Merge branch 'tp_smapi'

commit a9082da3adfcfc1249d410a05ce23805d0e9fad1
Author: Brandon Berhent <cheater1034@gmail.com>
Date:   Thu Jul 10 17:04:16 2008 -0400

    Why did i return 0?
    Fix it :-p

commit cf9424acb115d9e811eac5d91d2540a929a710cf
Author: Brandon Berhent <cheater1034@gmail.com>
Date:   Thu Jul 10 16:52:31 2008 -0400

    Fix "add disk protection for hdaps"

commit 44309a8d3dc0f299cb241bec3c51c9117fa32fa7
Merge: a7a0e05... 10f6727...
Author: Ryan Hope <rmh3093@gmail.com>
Date:   Wed Jul 9 22:27:01 2008 -0400

    Merge branch 'zen-tunables'
commit b6e301fee454b03b7b1ade12363f2904fe986e84
Merge: 194f984... 9221a45...
Author: Ryan Hope <rmh3093@gmail.com>
Date:   Wed Jul 9 19:50:25 2008 -0400

    Merge branch 'compcache'

commit f8e959bd9a0e6d8b34ebf33ae698a868efe02ee5
Author: Brandon Berhent <cheater1034@gmail.com>
Date:   Tue Jul 8 01:14:59 2008 -0400

    Revert s390-build-fixes.c
    ptep_modify* should not be converted to a macro.

    This fixes build of paravirt.c on x86 systems (maybe other archs?)

commit 944adac3a606a877bfeaebeacb3ad86c4fc63778
Author: Brandon Berhent <cheater1034@gmail.com>
Date:   Tue Jul 8 01:01:53 2008 -0400

    PPC: Fix build on powerpc systems

dusanc · Post by **dusanc** » Fri Jul 11, 2008 5:56 am

cheater1034 wrote:...
It depends on the FS as well. For example, on reiserfs or reiser4 i would never use cfq/bfq (they just don't work right, i notice significant change). Though CFQ is massively different than AS or Deadline. I generally bounce back and forth between AS/Deadline, right now i'm on AS and loving it.

Although, if your / is on ext*, then CFQ is generally the best choice from what i've found.

Can you give more info on the reiser4+CFQ issue? I'm using that combo now and can't reproduce any problems in heavy multitasking. everything is working fine.

cheater1034 · Post by **cheater1034** » Fri Jul 11, 2008 6:58 am

dusanc wrote:
cheater1034 wrote:...
It depends on the FS as well. For example, on reiserfs or reiser4 i would never use cfq/bfq (they just don't work right, i notice significant change). Though CFQ is massively different than AS or Deadline. I generally bounce back and forth between AS/Deadline, right now i'm on AS and loving it.

Although, if your / is on ext*, then CFQ is generally the best choice from what i've found.
Can you give more info on the reiser4+CFQ issue? I'm using that combo now and can't reproduce any problems in heavy multitasking. everything is working fine.

Well, there's a bug (http://bugzilla.kernel.org/show_bug.cgi?id=10746), and i haven't used CFQ in a long time (always usually deadline, but right now i'm on AS), but a couple people i've talked to had suffered. I talked to THoff after he installed conrad (reiser4), and he couldn't figure out why his system was so laggy in X, he said it was unbearable, but he slapped in elevator=deadline in grub.conf and he said it made a worlds of difference.

Apparently it's only on certain hardware setups, i havent used cfq in probably a year + however, but Thoff's experience was just when i put up the conrad guide (2.6.26-rc6-zen1)

kicior · Post by **kicior** » Fri Jul 11, 2008 1:57 pm

I have got some error installing nvidia drivers on rc8-zenmm2:

Code: Select all

/portage/var/tmp/portage/x11-drivers/nvidia-drivers-177.13/work/NVIDIA-Linux-x86_64-177.13-pkg2/usr/src/nv/os-interface.c:703: error: implicit declaration of function 'kill_proc'
make[3]: *** [/portage/var/tmp/portage/x11-drivers/nvidia-drivers-177.13/work/NVIDIA-Linux-x86_64-177.13-pkg2/usr/src/nv/os-interface.o] Error 1
make[2]: *** [_module_/portage/var/tmp/portage/x11-drivers/nvidia-drivers-177.13/work/NVIDIA-Linux-x86_64-177.13-pkg2/usr/src/nv] Error 2
NVIDIA: left KBUILD.
nvidia.ko failed to build!
make[1]: *** [module] Error 1
make: *** [module] Error 2

Ebuild is from zen-overlay. Can anybody help?

tranquilcool · Post by **tranquilcool** » Fri Jul 11, 2008 2:11 pm

kicior wrote:I have got some error installing nvidia drivers on rc8-zenmm2:

Code: Select all

/portage/var/tmp/portage/x11-drivers/nvidia-drivers-177.13/work/NVIDIA-Linux-x86_64-177.13-pkg2/usr/src/nv/os-interface.c:703: error: implicit declaration of function 'kill_proc'
make[3]: *** [/portage/var/tmp/portage/x11-drivers/nvidia-drivers-177.13/work/NVIDIA-Linux-x86_64-177.13-pkg2/usr/src/nv/os-interface.o] Error 1
make[2]: *** [_module_/portage/var/tmp/portage/x11-drivers/nvidia-drivers-177.13/work/NVIDIA-Linux-x86_64-177.13-pkg2/usr/src/nv] Error 2
NVIDIA: left KBUILD.
nvidia.ko failed to build!
make[1]: *** [module] Error 1
make: *** [module] Error 2

Ebuild is from zen-overlay. Can anybody help?

sed -i -e 's:return kill_proc:return kill_pid:g' usr/src/nv/os-interface.c
then it will be ok.

kicior · Post by **kicior** » Fri Jul 11, 2008 3:00 pm

Thank you very much - now nvidia compiled but i have kernel panic just after hitting enter in grub.
Here is photo with message: http://hektor.umcs.lublin.pl/~leszeks/panic.jpg

tranquilcool · Post by **tranquilcool** » Fri Jul 11, 2008 3:04 pm

kicior wrote:Thank you very much - now nvidia compiled but i have kernel panic just after hitting enter in grub.
Here is photo with message: http://hektor.umcs.lublin.pl/~leszeks/panic.jpg

well that happened to me several times.
then i 'make clean' compile use 'elevator=deadline' and now kernel is consistent.
i didn't have to recompile nvidia-drivers.

kicior · Post by **kicior** » Fri Jul 11, 2008 3:08 pm

tranquilcool wrote:
kicior wrote:well that happened to me several times.
then i 'make clean' compile use 'elevator=deadline' and now kernel is consistent.
i didn't have to recompile nvidia-drivers.
You mean just
Code: Select all
make clean ; make oldconfig
?
And I'm now using deadline.

tranquilcool · Post by **tranquilcool** » Fri Jul 11, 2008 3:11 pm

kicior wrote:
tranquilcool wrote:
kicior wrote:well that happened to me several times.
then i 'make clean' compile use 'elevator=deadline' and now kernel is consistent.
i didn't have to recompile nvidia-drivers.
You mean just
Code: Select all
make clean ; make oldconfig
?
And I'm now using deadline.
did make oldconfig, make clean, make && make modules_install.
then copy the bzIimage and System-map to boot.

as a matter of fact i still don't know why kernel panicked and then suddenly became consistent.

ebrsc · Post by **ebrsc** » Fri Jul 11, 2008 3:19 pm

hi, latest pull 1 hour ago. here's what dmesg likes to spit:

ACPI: Video Device [VGA] (multi-head: yes rom: no post: no)
proc_dir_entry 'VGA' already registered
Pid: 2806, comm: modprobe Not tainted 2.6.26-mmz #1
[<c01cdf08>] proc_register+0x158/0x180
[<c01ce103>] proc_mkdir_mode+0x33/0x50
[<f8c6a415>] acpi_video_bus_add+0x1b8/0xb5b [video]
[<c01d739d>] sysfs_create_link+0x8d/0x160
[<c02b9b1c>] acpi_device_probe+0x3a/0x7a
[<c02cf426>] driver_probe_device+0x86/0x1a0
[<c02cf5b1>] __driver_attach+0x71/0x80
[<c02b9a84>] acpi_device_remove+0x0/0x5e
[<c02ced54>] bus_for_each_dev+0x44/0x70
[<c02b9a84>] acpi_device_remove+0x0/0x5e
[<c02cf2b6>] driver_attach+0x16/0x20
[<c02cf540>] __driver_attach+0x0/0x80
[<c02ce6e7>] bus_add_driver+0x1a7/0x220
[<c02b9a84>] acpi_device_remove+0x0/0x5e
[<c02cf74c>] driver_register+0x5c/0x130
[<f8c5d02f>] acpi_video_init+0x2f/0x4d [video]
[<c014f378>] sys_init_module+0x168/0x1cf0
[<c01c92bf>] proc_reg_open+0xef/0x140
[<c02b9db8>] acpi_bus_register_driver+0x0/0x38
[<c0103d7d>] sysenter_past_esp+0x6a/0x91
=======================

should i fear pain and sorrow ?

TIA

kicior · Post by **kicior** » Fri Jul 11, 2008 4:19 pm

tranquilcool wrote:did make oldconfig, make clean, make && make modules_install.
then copy the bzIimage and System-map to boot.

It didn't help - still kernel panic.