2.6.11-love2 (usb & genetic-as deplug. oops fix)

[fallow]

2.6.11-love2 aka "The Time is Now" release info

patchlist/info/everythins about it :

--In short--

Code: Select all

*) CPU Scheduler : Nick`s Scheduler (Nick Piggin) with J.Moilanen`s Genetic Alg. implementation port.
also default BASE_TIMESLICE selectable from menuconfig. Included multiprocessor fixes.
(genetic nicksched improvements in thanks to M.Hennings) 

*) IO Scheduler(s) : switchable at runtime,default selectable from menuconfig or bootparams
   also anticipatotory io scheduler with genetic algor. implementation 
   and timesliced CFQ v2.

*) VM : Con`s MW3. with VM_MAPPED settigns selectable from menuconfig.

-> fix of X86_64 scheduling while atomic from love1 (in thanks to Nick Piggin )
-> fix of X86_64 mach_blacklisted from love1
-> new genetic nicksched version 0.02 with adaptability impr.
-> new hostap driver
-> old (working) toshiba acpi
-> new inotify
-> cfq-timeslice fix from ck
-> security (exploit) fix for 2.6.11 (2.6.11.2)
-> new sk98lin driver
-> xfs kconfig error fix
-> Kconfig interface to chmp
-> test version of nvidiafb driver (vesafb,nvidiafb both works)
-> more mm fixes

--full info--

Code: Select all

--patches--
[-base-]
love-menuc.diff - love menuconfig part.
  [*] Show timing information on printks                   
  [*] Genetic Library                                     
  [*]   Genetic Anticipatory I/O scheduler                 
  [*] [cpu_sched] Genetic NickScheduler [experimental]      
        [cpu_sched] Genetic NickScheduler options  --->     
      [cpu sched] NickSched Base Timeslice = (64-love-sources std) --->
      [mwIII] VM_MAPPED = 66 ---> 
      [io sched] IO Schedulers to compile  --->                
      [io sched] Default IO Scheduler (Deadline)  --->         
  [*] /proc/mm support (SKAS patch)                            
  [ ] 1Gb Low Memory Support (from CK)                         
  [*] Enable seccomp to safely compute untrusted bytecode      
      Kernel internal timer frequency (1000)  --->           

//gen4nicksched0.02_test1 is my port of J.Moilanen Genetic Alg implementation in 
//P.Williams Zaphod CPU Scheduler.It`s based on Genetic Lib 0.4
//at now it`s BETTER to use default Genetic NickSched options.
//adaptability improvements by M.Hennings
//addition Genetic NickSched options are : 
  [ ] [testing!] Genetic Bonus Calculation [not_reco/very_exp]     
  [ ] [testing!] Rounding of Sched_BaseTimeslice [not_reco/very_exp] 
//First is for activate additional genetic parameter in bonus calculations
//Can provides some interactivity loss of not :) -> dangerous :)
//Second for rounding of base_timeslice to range 4,8,16,32,64,128,256 only
//So, for a "little less" fluctuation range,If You will have a little less "genetics"
// /proc/mm is from SKAS patchsets from CKO
// SECCOMP : This kernel feature is useful for number crunching applications
//that may need to compute untrusted bytecode during their execution.
// via /proc/<pid>/seccomp,
//NickScheduler BASE_TIMESLICE selectable beetwenn 256,128,64,32
//VM_MAPPED selectable beetwen 95,66,33

patch-2.6.11.2.diff - security fix for 2.6.11 from kernel.org
2.6.11-nicksched.gz - latest NickSheduler including multiprocessor fixes.
gen4nicksched-0.02_test1.diff - my port of J.Moilanen Genetic Impl in Zaphod
adaptability improvements by M.Hennings 
x86_64_atomic_fix1.diff - scheduling while atomic fix on x86_64 - thanks to N.Piggin
genetic-as-sched-2.6.10-gl4.patch - J.Moilanen`s genetic anticipatory io sched
genetic-io-sched-2.6.10-gl4.patch - J.Moilanen`s "as to io connection "
genetic-lib-2.6.10-gl4.patch - J.Moilanen`s Genetic Lib 0.4
cfq-ts-21.diff - J.Axboe`s Timesliced CFQv2 from CK
cfq-ts21-fix.diff - fixes from CK.
reiser4-NEW-code-from-mm - latest NEW reiser4 code. /thx Gimpel for update/
reiser4_replay_stats.patch.bz2 - a reiser4 fix from CKO
Software Suspend 2.1.8 - Software Suspend 2 ;)
fbsplash-0.9.1-r2-2.6.11-rc4.patch - Spock`s great splash tool.
vesafb-tng-0.9-rc6-2.6.11-rc1.patch - Spock`s great vesafb tool.
bk-ntfs.patch - bk ntfs update
fuse-filesystem-ALL.patch - Filesystem in UserSpace patches from mm
mapped_watermark3.diff - NEW version of Con`s MW.
1g_lowmem1_i386.diff - for 1GB lowmem from CK
lirc-2.6.9-20041024-px1_FIXED_SIR.patch - lirc with fixed SIR
lufs-0.9.7-2.6.0-test9-FIXED_MODULE_COMPILATION.patch.bz2 - LUFS 
IT821X_AC9_2.6.10-gentoo-r5.patch - I know that somebody need`s it :)
390-slabtune.txt - slabtune from MJB 
acerhk.patch - acer keyboard support
cddvd-cmdfilter-drop.patch - non-root cd hack from CK
cflags-selection.patch - OneOfOne`s tool , extended by Troll
chmp_last_lkml.diff - thanks Krejler
config-nr-tty-devices.patch - :) 
config_hz.diff - from CK
gcloop-2.6-20040527.patch - gcloop :)
inotify-0.20-rml-2.6.11-mm1-3.patch - NEW inotify for GNOME users :)
nvidia_6111-6629_compat2.diff - pack of nvidia fixes from CK .
orinoco-0.15rc2-for-2.6.11.patch - by RMH3093 - Thanks :)
Real-TIME-lsm.patch - RealTIme Security Module :)
SECCOMP.patch - SECCOMP info at the top :)
bk-libata.diff - somebody likes it :)
skas-2.6.11-v8-rc2-fix.patch - SKAS fix from CKO
skas-2.6.11-v8-rc2.patch.bz2 - SKAS UML from CKO
shfs-0.35.patch.bz2 - shfs :)
squashfs2.1-patch - new squashfs :)
cpad-2.6_v1.1.patch - thx RMH3093
daconfig-2.2.0-r1.patch - DaMouse`s nice tool :P :)
toshiba_fixes.patch - thx RMH3093
xbox-sources-2.6.10-20050102.patch - all patches from XBOX-sources
acpi-dsdt-initrd-patch-v0.7d-2.6.9.patch - acpi extension.
alps1- thx Lovechild
usbhid-readd-kconfig.patch - thanks Krejler
hostap-linux-0.3.7-2.6.2.patch
//remember to chose only card You have
//for example will be conflict beetwen PLX9052 and Prism2.5 PCI
//must be compiled only one of them.
sk98lin_v7.09_2.6.11_patch
xfs Kconfig fix - thanks Hirion

[mm`s part]
nvidiafb-add-update-framebuffer-driver-for-nvidia-chipsets.patch
copy_pte_range-latency-fix.patch
irqpoll.patch
properly-share-process-and-session-keyrings-with-clone_thread.patch
put-newly-registered-shrinkers-at-the-tail-of-the-list.patch
run-softirqs-on-proper-processor-on-offline.patch
serialize-access-to-ide-devices.patch
setup_per_zone_lowmem_reserve-oops-fix.patch
simpler-topdown-mmap-layout-allocator.patch
smpbootc-cleanups.patch
sys_setpriority-euid-semantics-fix.patch
tty-output-lossage-fix.patch
vm-pageout-throttling.patch
add-compiler-gcc4h.patch
bk-usb.patch
blockdev-fixes-race-between-mount-umount-tidy.patch
blockdev-fixes-race-between-mount-umount.patch
add-timing-information-to-printk-messages.patch
allow-admin-to-enable-only-some-of-the-magic-sysrq-functions.patch
annotate-proc-pid-maps-with--markers.patch
areca-raid-linux-scsi-driver-fix.patch
areca-raid-linux-scsi-driver.patch
fix-iounmap-and-a-pageattr-memleak-x86-and-x86-64.patch
invalidate-range-of-pages-after-direct-io-write.patch
make-tree_lock-an-rwlock.patch

--proc interfaces---
/proc/genetic/nick-sched - genetic NickScheduler
/proc/genetic/as-ioscheduler - genetic as io scheduler 
/proc/sys/vm - mwIII
/proc/sys/kernel/base_timeslice - NickSched Base TimeSlice - 
/proc/sys/kernel/genetic_bonus - genetic bonus for nicksched if it`s enabled

You can check also selected options from love-sources menu via dmesg and this message during boot.

Code: Select all

/logo by rmh3093/
(...) 
[14.985006] ACPI: Looking for DSDT in initrd... not found.
[14.997794] ACPI: setting ELCR to 0200 (from 0e20)
[14.999331] |--------------------------------------------------------------------------|
[14.999364] |  __    _____  _  _  ____      ___  _____  __  __  ____   ___  ____  ___  |
[14.999395] | (  )  (  _  )( \/ )( ___)___ / __)(  _  )(  )(  )(  _ \ / __)( ___)/ __) |
[14.999426] |  )(__  )(_)(  \  /  )__)(___)\__ \ )(_)(  )(__)(  )   /( (__  )__) \__ \ |
[14.999457] | (____)(_____)  \/  (____)    (___/(_____)(______)(_)\_) \___)(____)(___/ |
[14.999488] |                                                                          |
[14.999519] | Version : 2.6.11-love2                          ,o8o, ,o8o,              |
[14.999550] | made by fallow for Love-Community             ,888888,888888,            |
[14.999581] | Have a nice day ;)                            888888888888888            |
[14.999612] |                                               888888888888888            |
[14.999643] | NickScheduler BaseTimeslice = 64              `8888888888888'            |
[14.999674] | VM_MAPPED = 95                                  `888888888'              |
[14.999705] |                                                   `88888'                |
[14.999736] | GENETIC NICKSCHED  ON ! :)                          `8'                  |
[14.999767] |--------------------------------------------------------------------------|
[15.000056] checking if image is initramfs... it is
[15.027949] Freeing initrd memory: 258k freed
(...)

* [nvidia] Tested with nvidia-kernel-1.0.6629-r3/r4 ebuild from portage and AGPGART
* [ati] Should proper work with latest ebuilds from portage

Just give a try to Deadline Imho is very good
If You have oops with usb deplugging ,
[EDIT]
I received a fix for this from J.Moilanen - please test it : http://vivid.dat.pl/fallow/pub/love-sou ... -fix1.diff
[/EDIT]
Just don`t enable genetic anticipatory (in menuconfig)(is off by default) , genetic nicksched is safe - You can still use other IO schedulers and "normal"/ regular anticipatory with genetic nicksched cpu scheduler without a risk.



Also update from gen4nicksched 0.02 to 0.03 is availabe here : http://vivid.dat.pl/fallow/pub/love-sou ... to-03.diff . You can now select range of Base_Timeslice . Std in love2 was a 4-256 , If You want better interactivity - choose 4-128 for example and thanks to RMH3093 for the idea of additional selection


available at http://sourceforge.net/project/showfile ... _id=311614
check also http://www.love-sources.org
for support please join #love-sources on freenode
also here : http://vivid.dat.pl/fallow/pub/index.ph ... .11-love2/ [/b]

alone very pre version of port of J.Moilanen`s Genetic Alg impl. in Zaphod for Nicksched is here : http://vivid.dat.pl/fallow/pub/index.ph ... nicksched/

have a nice day;)

[DrWoland]

Is the USB unplug problem fixed? Guess I'll find out soon enough.

[ashtophet]

Compiled and booted. Using Genetic Nicksed (with both testing options) & anticipatory i/o sched w/ genetic lib. Previous release was amazing, didn't suffered any problem, great performance. Hope this one is better...

Thanks a lot, fallow... lovely job!

[DrWoland]

Is the USB unplug problem fixed? Guess I'll find out soon enough.

So uh, yeah, unplugging a USB storage device definitely causes a hardlock. It ouputs a bunch of lines I don't understand, nor can I copy them since it's locked, nor do they appear in /var/log/messages, but I did notice that the first 3 lines in what appeared to be a chain of events contained the word "genetic". I went in, turned off everything that said "genetic" in Love Options, rebooted - works perfect. Am I missing out on a lot?

[Sith_Happens]

Why don't you just set a different I/O scheduler for your usb device in /sys? Or set a different I/O scheduler globally at boot?

Code: Select all

*) IO Scheduler(s) : switchable at runtime,default selectable from menuconfig or bootparams
   also anticipatotory io scheduler with genetic algor. implementation
   and timesliced CFQ v2. 

Just use the regular anticipatory io sheduler for your usb storage device and you should be fine. That sounds like a serious bug though that the maintainers of the anticipatory io scheduler with genetic algorithm implemenation need to know about.

[DrWoland]

Why don't you just set a different I/O scheduler for your usb device in /sys? Or set a different I/O scheduler globally at boot?

Code: Select all

*) IO Scheduler(s) : switchable at runtime,default selectable from menuconfig or bootparams
   also anticipatotory io scheduler with genetic algor. implementation
   and timesliced CFQ v2. 

Yeah, I set it to Deadline I believe. I wasn't sure which genetic option was to blame, so I got rid of anything that said Genetic. I'm not very familiar with it - in fact I know nothing about it. I didn't switch to Love for it, so I'm not too worried. I'll do some reading on it later and perhaps try it out.

Fallow, do you want me to recompile another copy with Genetic enabled and then type out the error it dumps? I can do that, anything for Love

[Sith_Happens]

I'm having a problem digesting the ebuild, who wants to be the first to tell me what I did wrong

Code: Select all

root@BlueBox love-sources # ebuild  /usr/local/portage/sys-kernel/love-sources/love-sources-2.6.11-r2.ebuild digest
!!! aux_get(): ebuild path for 'sys-kernel/love-sources-2.6.11-r2' not specified:
!!!            None
!!! aux_get(): ebuild path for 'sys-kernel/love-sources-2.6.11-r2' not specified:
!!!            None
doebuild(): aux_get() error reading sys-kernel/love-sources-2.6.11-r2; aborting.

[DrWoland]

I'm having a problem digesting the ebuild, who wants to be the first to tell me what I did wrong

Code: Select all

root@BlueBox love-sources # ebuild  /usr/local/portage/sys-kernel/love-sources/love-sources-2.6.11-r2.ebuild digest
!!! aux_get(): ebuild path for 'sys-kernel/love-sources-2.6.11-r2' not specified:
!!!            None
!!! aux_get(): ebuild path for 'sys-kernel/love-sources-2.6.11-r2' not specified:
!!!            None
doebuild(): aux_get() error reading sys-kernel/love-sources-2.6.11-r2; aborting.

Did you place the ebuild in the folder?

[Sith_Happens]

Nevermind, I had a typo in my /etc/make.conf, I didn't check it carefully enough.

[feld]

I also came to the conclusion that Genetic is causing the USB drive problems.

I love your sources so i'm just turning off Genetic stuff until this is resolved.


Later!



-Feld

[Sith_Happens]

I'm having some trouble loading the fglrx module (ati-drivers). I compiled love-sources with the same config I used for my other kernel, only when I boot with love-sources, I can't modprobe fglrx. It gives me this error:

Code: Select all

FATAL: Error inserting fglrx (/lib/modules/2.6.11-love2/video/fglrx.ko): Operation not permitted

fallow said in his release notes that the ATi drivers in portage should work with love-sources. Anybody have any idea what I'm doing wrong? I tried downgrading from version 8.10.19 to 8.8.25 of the ATi drivers and running modules-update, but it didn't help.

[arch4nge1]

I too would like to confirm that unplugging USB devices (specifically USB storage) causes kernel panic with Genetic stuff turned on. Currently I have the Genetic Library compiled in, but nothing else. I'm guessing it does nothing without enabling the Genetic schedulers (Anticipatory and Nicksched).

Now running without Genetic schedulers, USB storage working smoothly.

Cheers

[fallow]

at now :

If You have oops with usb deplugging , Just don`t enable genetic anticipatory (in menuconfig)(is off by default) , genetic nicksched is safe - You can still use other IO schedulers and "normal"/ regular anticipatory . look to the previous release - this release is a some kind of fixes from love1 updates, is a less new stufff here.

I prefers genetick nicksched + deadline for ex
in this release has also most fresh cfq2 .

That sounds like a serious bug though that the maintainers of the anticipatory io scheduler with genetic algorithm implemenation need to know about.

Yeah - I`ll send a mail to J.Moilanen about this today.

btw.tanks for reporting.

Compiled and booted. Using Genetic Nicksed (with both testing options) & anticipatory i/o sched w/ genetic lib. Previous release was amazing, didn't suffered any problem, great performance. Hope this one is better..
Thanks a lot, fallow... lovely job!

thanks for nice words :)
btw.Propably You can get better genetic results when You turn out the "Rounding" option from the genetic nicksched menu .

cheers.

[Sith_Happens]

Nevermind, I fixed my problem. I just had to remove the DRM manager from my kernel config. Funny, it worked fine in gentoo-dev-sources.

[Master One]

Another butt-kicking-love-release. There are definitely no better sources for use with Reiser4.

Code: Select all

Reiser4 1.0.4 + love on genetic nicksched & deadline = slipstream-speed

[DrWoland]

Another butt-kicking-love-release. There are definitely no better sources for use with Reiser4.

Code: Select all

Reiser4 1.0.4 + love on nicksched & deadline = slipstream-speed

/me concurs. I'm loving it so far, no problems when not using genetic scheduler either.

[Master One]

no problems when not using genetic scheduler either.

You don't have to miss out on genetic nicksched, just disable genetic as and use deadline as i/o sched, then your USB problem should not occure.

[fallow]

thanks people it seems to be not so bad

I was really disappointed because of this usb bug in this release. I think many bugs from previous (love1) are fixed now ... is still this damn usb-unpluging.

Will be great if somebody will send me shot or maybe some text , EIP etc. -> fallow@neostrada.pl
I`m going to send this problem review to maintainer of Genetic Lib and Genetic Anticipatory.

Maybe we also will try to fix it - if I could have oops text

I can also prepare a patch to reverse genetic anticipatory implementation from love2.

however - it seems that using of Genetic Nicksched with CFQ,Deadline,Regular/Normal Anticipatory don`t gives problems

and we decided also to make a longer testing period , before any next love sources release will be official.
at this stage any help or bugs reporting on the irc will be very useful

cheers.

[ExZombie]

Great job fallow!
I can finally get back to using love sources after being "stranded" with nitro for quite a while .
Since I am getting those USB oopses I'll just stick with deadline as you have recommended. But I'd like to know, how does
deadline perform with multiple heavy I/O operations at once?

Also, has anyone tried benchmarking genetic schedulers versus normal schedulers (both CPU and I/O)?
They feel more responsive, but I'd be interesting to see some numbers. I lack the time to do proper tests.

[DrWoland]

thanks people it seems to be not so bad

I was really disappointed because of this usb bug in this release. I think many bugs from previous (love1) are fixed now ... is still this damn usb-unpluging.

Will be great if somebody will send me shot or maybe some text , EIP etc. -> fallow@neostrada.pl
I`m going to send this problem review to maintainer of Genetic Lib and Genetic Anticipatory.

Maybe we also will try to fix it - if I could have oops text

I can also prepare a patch to reverse genetic anticipatory implementation from love2.

however - it seems that using of Genetic Nicksched with CFQ,Deadline,Regular/Normal Anticipatory don`t gives problems

and we decided also to make a longer testing period , before any next love sources release will be official.
at this stage any help or bugs reporting on the irc will be very useful

cheers.

I'll recompile it and take a photo. Give me a few minutes.

Here it is in all its glory: http://www.drwoland.com/usboops.jpg
(Huge screenshot so it's easy to read)

[sofcik]

HI. I'm just going to test new love . Anyway could you tell me what is new nvidiafb driver ? Is this better then vesafb-tng ?

[TheCoop]

whats the differences between deadline and cfq? why are ppl using deadline instead of cfq?

[yoshi314]

i have a couple of problems with love-sources
-----------------------------------------------------
i'm not really good at this stuff, but i tried love-sources 2.6.11-r1 and on some heavier compilations (those lasting longer then 25min) my system would go into complete halt. CPU was extremely hot at that moment (i cannot tell you exactly, but the very air inside of the box was at least 40C, which is absolutely rare for me). it happened numerous times on love-sources. gentoo-dev-sources work all right all the time, however.

i'm not an overclocker. i don't do things like that. i have an 1.7Ghz celeron (pentium4-compatible cpu)

i did not use clock modulation features in kernel config, the config was the same as in my gentoo-dev-sources 2.6.10 kernel. i only enabled love-specific features.

so,umm, what did i do wrong?

oh, by the way - i cannot get framebuffer to work with love-sources, but it works on 2.6.10-r6 with the same setup. (i mean links -driver fb complains that device /dev/fb0 does not exist on love-sources, but it works with gentoo-dev-2.6.10-r6 (and older dev-2.6 as well)). (i'm using udev btw.)

[ashtophet]

btw.Propably You can get better genetic results when You turn out the "Rounding" option from the genetic nicksched menu

done. Let's reboot!

[Master One]

whats the differences between deadline and cfq? why are ppl using deadline instead of cfq?

dealine is supposed to work better with Reiser4, it's small and fast, see the description in menuconfig.