2.6.12-rc2-love1 (Multiple CPU Schedulers)

[fallow]

2.6.12-rc2-love1 release info
There is no name for this release , but We can naming it "a release"

Now some infos about the release , everybody like the infos

Please don't report any bugs to gentoo-devs or bugs.gentoo.org.
We are making patchset for x86 . We will be not able to help You with x86_64

If You want support or jusk ask about something - please join #love-sources on freenode or just
post in the love-sources thread in "Unsupported Software" forum on forums.gentoo.org
check also love-sources.org

Short info about main features .

This release is based on Peter Williams PlugSched4.0
I also made a genetic nickscheduler patch for this version of plugsched.
So now You can choose a CPU Scheduler at booting moment via kernel BOOTPARAMS line
In Love-Features menu You can choose the default CPU Scheduler.

if You will add "cpusched=nicksched" the kernel will boot with nickscheduler or genetic nickscheduler
depending on selection in menuconfig ( genetic nickscheduler on/off etc. )

available options are :
"cpusched=nicksched" - nickscheduler or genetic nickscheduler
"cpusched=staircase" - staircase 10.7 cpu scheduler
"cpusched=ingosched" - ingosched - Ingo Molnar CPU scheduler from vanilla
"cpusched=zaphod" - P.Williams Zaphod CPU scheduler
"cpusched=spa_no_frills" - P.Williams Single Priority Array CPU Scheduler

available IO Schedulers are :
- deadline
- no-op
- anticipatory or genetic anticipatory ( menuconfig selection )
- CFQv2 with TIMESLICED DESIGN and fixes from 2.6.12-rc2-mm1

You can choose default IO Scheduler from Love-Features or via kernel BOOTPARAMS line
If You will add "elevator=cfq" the kernel will bot with cfq io scheduler
You can also switch it at runtime for example via script from /love_src/!extras/sched_change_script
of via /sys/block/hdX/scheduler/queue interface.

possible option are : cfq , deadline , noop, anticipatory

If You are using genetic nicksched or anticiapatory You can watch /proc/genetic/***
!IMPORTANT! now all interfaces to cpu scheduler are moved from /proc to /sys/cpusched
so base_time_slice and genetic_bonus for nicksched or interactive and compute mode for staircase
are in /sys/cpusched/scheduler_name.

example .torsmorc theme for watching genetic nicksched fluctuations is in love_src/!extras dir.

If You have selected Genetic Library and Genetic NickScheduler in Love-Features menu.
You can set some additional Genetic NickScheduler options :

a) Base_Timeslice range ( 4 - 32 - std ) --->
You can select the range of base_timeslice for nickscheduler : 4-16,4-32,4-64,4-128.
b) [*] Genetic Bonus Calculation (NEW) [/color] [/b]
This is adding simple genetic parameter to the bonus calculations.
c) [*] Give a bonus for a long timeslice (NEW) [/color] [/b]
When no real waiting occurs the timeslice should be large to avoid overhead.
d) [*] MAGNITUDE (NEW) [/color] [/b]
Base Timeslice will be generated based on multiply of 4 only in 8-32 range.
This can give better results for rating function.

If You have selected "normal" NickScheduler You can select base timeslice from 16,32,64,128

You can also select VM_MAPPED for Mapped_Watermarks3 beetwen 33,66,95.

In this kernel is also KEXEC. You can run another (or this) kernel without reboot now.
For using it , please emerge kexec-tools and search the net for some how-to
On the forums or google is many of it

Ok now is the time for patchlist.
In love-sources are around ~140 small patches. Patches are grouped and cuted for less reading.
All patches You will see in broken-out file (link at the end of this post ).

Code: Select all

[main part]
plugsched-4.0-for-2.6.12-rc2.patch - P.Williams/Con Kolivas pluggable scheduler patch.
gen4nicksched_003_plugsched302mm3_to_0.04 - my old version of gen4nicksched-0.03 for plugsched.
  modified by hand up-to-version 0.04 (latest).
plugsched4-staircase10.7.diff - New Staircase for PlugSched from Con :) thx.

genetic-lib-2.6.10-gl4.patch - J.Moilanen's Genetic Lib 0.4
genetic-as-sched-2.6.10-gl4.patch - J.Moilanen's Genetic Anticipatory IO Sched.
genetic-io-sched-2.6.10-gl4.patch - Gen Lib and IO Scheds "connection".
2.6.11-love2-usb-oops-fix1.diff - Genetic Anticipatory USB deplug. oops fix.
2.6.12-rc2-xbox1.bz2 - a set of XBOX patches for love , maintained by FireBurn.
CFQv2-with-timesliced-design-and-fixes-all - all patches for CFQ from 2.6.12-rc2-mm1.

[bk] - all bk patches from 2.6.12-rc2-mm1
bk-acpi.patch
bk-drm-via.patch
bk-drm.patch
bk-jfs.patch
bk-netdev.patch - new version of hostap and orinoco 
bk-ntfs.patch
bk-scsi.patch

[other nice features]
reiser4-from-2.6.12-rc2-mm1 - latest reiser4 code 
fbsplash-0.9.2-r1-2.6.12-rc1-bk5.patch - Spock's fbsplash with small fix.
vesafb-tng-0.9-rc6-r2-2.6.12-rc1.patch - Spock's vesa-tng.
exec-shield-nx-2.6.11-bk10-B1 - security system , http://people.redhat.com/mingo/exec-shield/
rt-lsm.patch - Real Time Linux Security Module.
KEXEC , KDUMP , CRASHDUMP from 2.6.12-rc2-mm1 - for booting new kernel without reboot.
fuse-all-patches - FileSystem in User Space - from 2.6.12-rc2-mm1
inotify-022-for-2612-rc1-mm4.patch - inotify from 2.6.12-rc2-mm1 ;)
1g_lowmem1_i386.diff - for big mem users hehe  from CK:)
IT821X_AC9_2.6.10-gentoo-r5.patch - it821X driver , I know that somebody likes it.
lirc-2.6.9-20041024-px1.patch - lirc with fixes sir.
shfs-0.35.patch.bz2 - shfs filesystem.
squashfs2.1-patch - squashfs filesystem.
usbhid-readd-kconfig-r1.patch - nice Krejler's stuff.
acpi-dsdt-initrd-patch-v0.7d-2.6.9.patch - acpi extension for user dsdt.
cddvd-cmdfilter-drop.patch - non-root cdr user hack from CK.
config-nr-tty-devices.patch - as in name.
cpad-2.6_v1.1.patch - toshiba pad driver.
lufs-0.9.7-2.6.0-test9.patch.bz2 - lufs with fixes compilation as module.
mapped_watermark3.diff - Con's MWIII
nvidia6111-6629-compact2.diff - a set of nvidia fixes from ck.
atxp1-0.6.patch - for i2c supprot for some motherboards.
inter-module-get fix for ATI users.
love-sources-logo : http://vivid.dat.pl/fallow/pub/love-sources/love-logo/love1.png
many-of-x86-and-x86_64-fixes from 2.6.12-rc2-mm1 - as in name ;)
and some other fixes from mm like : 
 acpi-video-pointer-size-fix.patch
 bug-fix-in-usbdevfs.patch
 ext2-corruption-regression-between-269-and-2610.patch
 fix-agp-code-to-work-again-with-non-pci-agp-bridges.patch
 reiserfs-use-null-instead-of-0.patch
 serport-oops-fix.patch
 vmlinux-fix-physical-addrs.patch

[additional wi-fi drivers]
2.6.12-rc1-love1-sk98lin-709-patch.bz2
madwifi-cvs 
hostap and orinoco new version are in bk-netdev.patch from mm

*** last part - versions req ***

1) NVIDIA users
NVIDIA : minimum nvidia-kernel is a 6629-r4 from portage

2) ATI users
ATI : will work or not , I fixed inter_module_get issue and made a ebuild for the driver
it is compiled fine on my box , but I cant do more testing - I have only NVIDIA CARD
propably the same issues as with previous 2.6.12-rc1-love1
ebuild for 8.12.10 :
http://vivid.dat.pl/fallow/pub/love-sou ... t1.tar.bz2

3) fbsplash users
min version of splashutils - 1.1.9.2

where this is available ?

[ -> my location ] :
main dir : http://vivid.dat.pl/fallow/pub/index.ph ... rc2-love1/

patch : http://vivid.dat.pl/fallow/pub/love-sou ... -love1.bz2
info : http://vivid.dat.pl/fallow/pub/love-sou ... /notes.txt
ebuild : http://vivid.dat.pl/fallow/pub/love-sou ... -love1.bz2
test ati ebuild : http://vivid.dat.pl/fallow/pub/love-sou ... t1.tar.bz2
split out : http://vivid.dat.pl/fallow/pub/love-sou ... st.tar.bz2
THX to Milu for the space on the server
[ -> sourceforge.net ] :
http://sourceforge.net/projects/love-sources

cheers

[smg]

Nice work fallow.
Sadly nvidia 77* doesn't work properly with -rc2.

Bye.

[sobers_2002]

man, again the 6629-r4 and above!! life sucks with older graphics cards......seems like unles
i have some budget expansions, i won't be in love again

[fallow]

it is working here fine with 6629-r4 .
also it is working with 7*** series .

I have problems with 7*** and renderaacel + composite but this is on every kernel and it is my geforce2gts . I must upgrade gfx card to some new for using 7***.

cheers.

[Poe]

Yeap, I yesterday compiled latest version with staircase. All works very well, havent 'cuting' mouse and system how I had with older staircase.

[mirko_3]

Compiled, and ati works! Thanks. I'm on nickcase right now, but I'll try staircase soon...

EDIT: never mind, it doesn't work... still, it's not really important for me, and the kernel seems nice (on staircase now..); I still can't judge on responsiveness, that'll have to wait a bit.
BTW, what's that single array scheduler? I've never seen it before, is it for servers or something?

[fallow]

no problem , good to hear that also ati works

cheers.

[Legoguy]

man, again the 6629-r4 and above!! life sucks with older graphics cards......seems like unles
i have some budget expansions, i won't be in love again

The new 7174 driver restores support for TNT2 and similar cards.. the old P2 350mhz behind me has a TNT2 and is happily using that version. Although from the looks of other posts, that might not work on this kernel if you want renderaccell and Composite. no biggie really if you don't use those things anyway.

[fallow]

7*** renderaccel badness on nvidia cards isnt related to any kernel.
it is just nvidia 7*** series fault.

cheers.

[NiLuJe]

Working fine here, except that cedega segfault immediatly after lanching something ... ;?

(Forgot to try with wine & cxoffice .. :s)

I'll try with a vanilla -rc2 asap

[superstoned]

7*** renderaccel badness on nvidia cards isnt related to any kernel.
it is just nvidia 7*** series fault.

cheers.

yes, this really sucks. I expected the second 7xxx to fix it (7174). It didn't.

[sobers_2002]

Code: Select all

Unpacking patch-2.6.12-rc2.bz2 to /var/tmp/portage/love-sources-2.6.12_rc2-r1/work

bzip2: Compressed file ends unexpectedly;
        perhaps it is corrupted?  *Possible* reason follows.
bzip2: Success
        Input file = /usr/portage/distfiles/patch-2.6.12-rc2.bz2, output file = (stdout)

It is possible that the compressed file(s) have become corrupted.
You can use the -tvv option to test integrity of such files.

You can use the `bzip2recover' program to attempt to recover
data from undamaged sections of corrupted files.


!!! ERROR: sys-kernel/love-sources-2.6.12_rc2-r1 failed.
!!! Function unpack, Line 384, Exitcode 2
!!! failure unpacking patch-2.6.12-rc2.bz2

whyz this?? an btw fallow the ebuild link is wrong in ur post

[fallow]

ebuild is ok , post is ok .

You downloaded patch-2.6.12-rc2.bz2 is INCOMPLETE.
delete it and download again.

cheers.

[NiLuJe]

About cedega's segfault : It's working fine with a vanilla 2.6.12-rc2 kernel .. Wine & CrossOver are working fine.

[Evil Dark Archon]

slightly annoying warning (changed include/asm/processor.h to include/asm-x86_64/processor.h to reflect the fact that this only affects, afaik, x86_64), this warning pops up a lot because just about every .c file includes include/asm/processor.h

Code: Select all

In file included from include/linux/spinlock.h:16,
                 from include/linux/percpu.h:3,
                 from include/linux/kernel_stat.h:8,
                 from mm/memory.c:41:
include/asm-x86_64/processor.h:168:1: warning: "TASK_SIZE" redefined
include/asm-x86_64/processor.h:164:1: warning: this is the location of the previous definition

*EDIT* compilation error (affects only x86_64 afaik)

Code: Select all

  CC      arch/x86_64/pci/../../i386/pci/direct.o
In file included from include/linux/prefetch.h:14,
                 from include/linux/list.h:7,
                 from include/linux/pci.h:478,
                 from arch/x86_64/pci/../../i386/pci/direct.c:5:
include/asm/processor.h:168:1: warning: "TASK_SIZE" redefined
include/asm/processor.h:164:1: warning: this is the location of the previous definition
arch/x86_64/pci/../../i386/pci/direct.c:7:32: mach_pci_blacklist.h: No such file or directory
arch/x86_64/pci/../../i386/pci/direct.c: In function `pci_conf1_read':
arch/x86_64/pci/../../i386/pci/direct.c:25: warning: implicit declaration of function `mach_pci_is_blacklisted'
make[1]: *** [arch/x86_64/pci/../../i386/pci/direct.o] Error 1

the include statement in question should read #include <asm-i386/mach-default/mach_pci_blacklist.h>

*EDIT 2* another error (x86_64)

Code: Select all

  LD      vmlinux
arch/x86_64/mm/built-in.o(.text+0x2440): In function `arch_align_stack':
: multiple definition of `arch_align_stack'
arch/x86_64/kernel/built-in.o(.text+0x1250): first defined here
make: *** [vmlinux] Error 1

[Sith_Happens]

Why don't you name it "For Her Pleasure...".

[darklegion]

Your patchset gave me a kernel panic with a message related to ld.so and an undefined reference after it.I would post the exact message but syslog did not get it.

[vipernicus]

Wow, everything works great, I am very happy with this release!

[fallow]

what the world strange is

for some errors , please test vanilla before

Im feeling a little that maybe Im done with making these patchsets..

cheers.

[gimpel]

Working fine here, except that cedega segfault immediatly after lanching something ... ;?

(Forgot to try with wine & cxoffice .. :s)

I'll try with a vanilla -rc2 asap

no problems here, just played sc4

yeah! this release is like it rocks! the sched overkill

currently using "ck-setup" (staircase/cfq) on a reiser4 system \\o o// seems like problems with staircase+reiser4 are gone, still bit higher cpu as with nicksched for exp when copying files, but much better than last time i tried staircase

[darklegion]

what the world strange is

for some errors , please test vanilla before

Im feeling a little that maybe Im done with making these patchsets..

cheers.

I wasn't being a smartarse,fallow.I think its makes sense that I report the errors that I see just in case it is an easy fix rather than something non-trivial.I'm actually pretty happy with rc1-love1 but wanted to see if the new staircase was an improvement at all.

[fallow]

of course please report evetything You have . Trivials are easy to fix , but with some other types , just please test vanilla also . of example usb oops etc...

Im going to think a little about the sense of making this kind of patchsets...

cheers.

[NiLuJe]

Gimpel > Wooops! ^^ I'll try without changing cflags then ..

[smg]

let's boycott nvidia and use ati *scnr*

[TheCoop]

what does each cpu sched do? Whats the differences between them?