vivid-sources

[travisau]

I know its a little past its prime now but I thought I should share this with you guys. I chose to use vivid on my mythtv system a while back because it included lirc and I thought it was cool:) Anyway I compiled up a 2.6.10-rc2-vv_e2 on a P4 HT2.8. Since then I've been recording 100s of gigs of shows (ivtv kernel module), playing them over the network, as well as through the connected TV (nvidia out), transcoding, and commercial flagging. The system has been incredibly stable for quite some time now as you can see.

myth root # ps -ef | grep mythbackend
root 19237 7412 0 2004 tty2 00:00:51 mythbackend (thats right mythbackend up solid since last year!:)
myth root # uptime
14:38:58 up 36 days, 14:03, 5 users, load average: 2.26, 2.38, 2.28
myth root # uname -a
Linux myth 2.6.10-rc2-vv_e2 #2 SMP Mon Nov 29 05:21:31 EST 2004 i686 Intel(R) Pentium(R) 4 CPU 2.80GHz GenuineIntel GNU/Linux

I didn't really expect this kernel to be this stable...but stable it is! I'm in the process right now of upgrading one of my desktop machines to the latest vve. If its as good as this one has been I will be WELL pleased. Many thanks to the devels who work on this kernel (fallow!)

[fallow]

I know its a little past its prime now but I thought I should share this with you guys. I chose to use vivid on my mythtv system a while back because it included lirc and I thought it was cool:) Anyway I compiled up a 2.6.10-rc2-vv_e2 on a P4 HT2.8. Since then I've been recording 100s of gigs of shows (ivtv kernel module), playing them over the network, as well as through the connected TV (nvidia out), transcoding, and commercial flagging. The system has been incredibly stable for quite some time now as you can see.

myth root # ps -ef | grep mythbackend
root 19237 7412 0 2004 tty2 00:00:51 mythbackend (thats right mythbackend up solid since last year!:)
myth root # uptime
14:38:58 up 36 days, 14:03, 5 users, load average: 2.26, 2.38, 2.28
myth root # uname -a
Linux myth 2.6.10-rc2-vv_e2 #2 SMP Mon Nov 29 05:21:31 EST 2004 i686 Intel(R) Pentium(R) 4 CPU 2.80GHz GenuineIntel GNU/Linux

I didn't really expect this kernel to be this stable...but stable it is! I'm in the process right now of upgrading one of my desktop machines to the latest vve. If its as good as this one has been I will be WELL pleased. Many thanks to the devels who work on this kernel (fallow!)

wow I`m really happy !, thanks for good words
2.6.10-rc2-vv_e2 is a good release.
IMHO 2.6.11-rc1 is not so good , I`m going to do release based on 2.6.11-rc2 tommorow ,so maybe You wait some time I hope that vanilla rc2 will be better than rc1 .
also , I will reduce number of patches - some clean up for trash non-required / useless stuff like moving qsort from xfs to libs.

also in my planes is to add Ingo`s Real Time Preempting patchset to the next release - so will be interesing

cheers

[fallow]

I have some dillema

what version to do better
I have some issues with staircase 10.5 and reiser4 ( not only vivid, everything stair10.5+r4 kernel , I tested vanilla+s10.5+r4 also ).
My mouse pointer freezes for 1-2 seconds with high I/O usage - for ex. copying kernel sources.
I talk on #ck . Con told me that is reiser4 related problem and not sure about able to fix it, maybe changig kernel reiser4 prior can do something but not sure.
Everything is ok with O(1)...now I`m little confused

as I think options are :

Code: Select all

1) 2.6.11-rc2 (nvidia works with some patch) with Staircase etc. ( but can be reiser4 + staircase troubles with high i/o ) as base
2) 2.6.11-rc2 O(1) based + Real Time Preemption and mm fixes for O(1) maybe also with Ingo new RT limits http://people.redhat.com/mingo/rt-limit-patches/ + other stuff
3) 2.6.10 + plugscheds but ..hm , 2.6.10 ...going to the past is not best way I think is 2.6.11-rc2 now :) (Syncing plugsched to 2.6.11-rc2 is possible but is too much work for me now, I think Peter Williams will do it in some near future )

I also agree that reiser4 is high i/o usage filesystem ,more than reiser3 with maybe no noticeable benefits in compare to the reiser3 hmm . objectivly this is good to think , maybe to return to reiser3 hehe because goodies is less than costs of use reiser4.

however maybe option 2 as base will be better , or maybe 1

cheers

[Genn]

I've been having that problem with the mouse. I thought it was a scheduler related issue.

[fallow]

OK - after many of tests , as fact as now , is better to have reiser4 ( whole system - me ) with O(1) than Staircase 10.5 - no freezes , no slow-downs etc.

I am doing new vivid release based on 2.6.10-rc2-bk4 + O(1) with Ingo`s Real Time Preemption and RT_LIMITS , with full reiser4 system it`s playing pretty well , no freezes , lags etc.
Of coruse genetic anticiapatory as well instead of not so good imho cfq2 ts patches ( not so good at the time , many of people has this opinion ) , however vivid release with push on good working reiser4 system will be tommorow

cheers.

[rmh3093]

hey I when I was using the staircase10.5 from ck's patches I was experiencing some random freezes (I have reiser4 also on my whole system) anyway when I use staircase10.5 from the plugsched patch it works great in my opinion, no freezes yet

[gimpel]

I am doing new vivid release based on 2.6.10-rc2-bk4

..thanks for keeping reiser4 as a priority! i really appretiate that, and am gladly looking forward to testing your new release. the last one stil runs and runs and runs....

[fallow]

ok after tests for good working whole reiser4 system - new release with priority for this.
cosmetic change : name from vv_e to vivid as in the past.
Main Mission Objectives

-->> Patches selection for good working reiser4
-->> Return to the Ingo O(1) CPU Scheduler intead of Staircase ( slow downs / lags with reiser4 ) with RT_Limits patch
-->> Ingo Real Time Preemption patchset
-->> reiser4 from mm with CKO fixes
-->> Genetic Anticipatory IO Scheduler based on Genetic Lib ver 0.3
-->> fbsplash,vesa-tng of course
-->> additional fs`s like lufs,supermount or FUSE
-->> Con`s mwII

Code: Select all

--vivid menu--
  Preemption Mode (Preemptible Kernel (Low-Latency Desktop))  --->   
   [ ] Thread Softirqs                                          
   [ ] Thread Hardirqs                                         
   [ ] Old-Style Big Kernel Lock                                
  Default io scheduler ((Genetic) Anticipatory io scheduler)  --->  
  [*] Genetic Library                                          
  [*]   Genetic Anticipatory I/O scheduler (EXPERIMENTAL)      

4 Levels of Preemption thanks to Ingo`s RTP patchset :         
  -- No Forced Preemption (Server)         
  -- Voluntary Kernel Preemption (Desktop)  
  -- Preemptible Kernel (Low-Latency Desktop)  
  -- Complete Preemption (Real-Time)                   
Thread Softirqs and Hardirqs are disabled by default because some people has problems with it.Default IO Scheduler is Genetic Anticipatory based on J.Moilanen Genetic Lib v0.3

--bk-tree--
bk-acpi.patch
bk-input.patch
bk-netdev.patch
bk-ntfs.patch
bk-usb.patch
bk-xfs.patch
--base--
realtime-preempt-2.6.11-rc2-V0.7.37-01
rt-limit-2.6.11-rc2-D8
genetic-as-sched-gl3.patch
genetic-io-sched-gl3.patch
genetic-lib-gl3.patch
realtime_lsm_module.patch
scheduler_change_script (made by Genn)
--additional fs--
reiser4 from 2.6.11-rc2-mm1
reiser4_repacker_stats.patch.bz2 from CKO
reiser4_replay_stats.patch.bz2 from CKO
cachefs from 2.6.11-rc2-mm1
fuse (filesystem in userspace) from 2.6.11-rc2-mm1
lufs (fixed compilation as module)
squashfs2.0
supermount-ng208 from CK
--all other patches--
1g_lowmem1_i386.diff from CK
390-slabtune.txt from MJB
500-sched_tunables.txt from MJB 
acerhk.patch 
madwifi-cvs
cddvd-cmdfilter-drop.patch from CK
cflags-selection.patch 
chmp-r5-FULL.patch
config-nr-tty-devices.patch
config_hz.diff from CK
configurable-delay-before-mounting-root-device.patch
daconfig-2.1.1.patch
detect-atomic-counter-underflows.patch
fbsplash-0.9.1-r1-2.6.11-rc1-bk8.patch
fix-64-bit-problems-in-cachefs.patch
fix-cachefs-barrier-handling-and-other-kernel-discrepancies.patch
fix-race-between-the-nmi-code-and-the-cmos-clock.patch
inotify.patch
lirc-2.6.9-20041024-px1.patch
make-tree_lock-an-rwlock.patch
mwII.diff
nvidia_6111-6629_compat2.diff
pio_font_fix1.diff
vesafb-tng-0.9-rc6-2.6.11-rc1.patch
--/proc--
/proc/sys/sched - additional CPU Scheduler settings (MJB)
/proc/genetic - for genetic anticipatory IO scheduler 
/proc/sys/vm - for VM tune (mwII from CK)

available at sourceforge.net :
https://sourceforge.net/projects/vivid-sources/
or :
https://sourceforge.net/project/showfil ... _id=300933


Tested with whole reiser4 system and nvidia-kernel 6629-r3 from portage via AGPGART , previous ebuilds before r3 doesn`t work with 2.6.11-rc2 kernel(s) , req fix - from r3 ebuild.

of course ,feedback , ideas , concepts etc. -> welcome
have a nice day
cheers.

[gimpel]

YAY! new vivid!!

just compiling!

some small issues on iptables tftp protocol support when it is built in...

CC net/ipv4/netfilter/ipt_TCPMSS.o
LD net/ipv4/netfilter/built-in.o
net/ipv4/netfilter/ip_nat_tftp.o(.bss+0x0): multiple definition of `ip_nat_tftp_hook'
net/ipv4/netfilter/ip_conntrack_tftp.o(.bss+0x0): first defined here
make[3]: *** [net/ipv4/netfilter/built-in.o] Fehler 1
make[2]: *** [net/ipv4/netfilter] Fehler 2
make[1]: *** [net/ipv4] Fehler 2
make: *** [net] Fehler 2

while compiling tftp proto as module:
a small feature request would be to include alan cox' ITE821x IDE PCI raid controller driver. i took the patch from http://forums.gentoo.org/viewtopic.php? ... 60#1994660 (applied without rejects)
it's quite a piece of standard hardware meanwile... so it would be nice to see that in other patchsets than -ac too as it is a bit of pain compiling the original ones from iteusa.com (they even suck)... alan's just work fine... but it's a really small one cause it's no problem to apply it for those who need it.

ok, building tftp proto as module worked it seems... reboot

(some minutes later..)

uname -a
Linux SiRiUS.home 2.6.11-rc2-vivid1 #1 Sun Jan 30 04:28:03 CET 2005 i686 AMD Athlon(tm) XP 1800+ AuthenticAMD GNU/Linux

first -rc2 patchset with working fbsplash.. woot! nvidia.. reiser4.. working!

thx once more so far fallow!

[predatorfreak]

Great to see a new vivid release, I haven't been very active with dark-sources due to the fact that no new 2.6.10 holes have been found and there haven't been many things that i believe i should merge. At least nitro and vivid are keeping the world happy . o and 2.6.10-dark5 due out sooner or later, with varstacks (thanks to Tiger683) and a few other patches such as slabtune and irqbal-fast. Anyway, keep up the great work fallow.

[fallow]

thx :) nice

gimpel : no problem , Your patches will be in next release
also sorry for netfilter error , I try to fix it too in ..next ril hehe

maybe some mm based release next ?

predatorfreak : nice, yes it is , mjb patches are great - schedtunable etc. and " more sources = more ways to fun "

cheers.

[chickaroo]

Yay!
i really love your kernels fallow, your a real genius.
love you *flirts*

[gimpel]

hmm... i've been too tired to check if really everything works fine with the new release, but i'd better should have done so...

cause i found my root filled up with a really extremely large /var/log/messages

here is all klogd output related to that

blabla edited out... i should more often look at that picture: http://www.chaotix.orcon.net.nz/useGoogleNoob.gif

[edit]
got it: http://www.webservertalk.com/archive242 ... 92015.html
and: http://www.webservertalk.com/archive242 ... 92021.html

^^ fixed in -bk7 or 8 it seems

[fallow]

Yay!
i really love your kernels fallow, your a real genius.
love you *flirts*

hehehehe :)

hmm... i've been too tired to check if really everything works fine with the new release, but i'd better should have done so...
cause i found my root filled up with a really extremely large /var/log/messages
here is all klogd output related to that
blabla edited out... i should more often look at that picture: http://www.chaotix.orcon.net.nz/useGoogleNoob.gif
[edit]
got it: http://www.webservertalk.com/archive242 ... 92015.html
and: http://www.webservertalk.com/archive242 ... 92021.html
^^ fixed in -bk5 it seems

hmmm :)wow .
thx for announcing it here...

I personally don`t have this issue (pppd and adsl usb modem)

so really it seems that it`s fixed in bk .

I try to fix it with the current release or do partial bk sync ( I think will be many of rejects with bk-trees or with RTP )

however , I try to fix it today/tommorow

cheers

[gimpel]

I try to fix it with the current release or do partial bk sync ( I think will be many of rejects with bk-trees or with RTP )

however , I try to fix it today/tommorow Smile

wow, i'd love you for that!!
but no need to do too much work on it! i also can wait your next release!
..and i seem to be the only one having that issue...

[Tiger683]

Disable debugging preemptible kernel and sleep inside spinlock debugging in Kconfig and recompile,
this should fix this....

cheers

T

EDIT:
Most of the spinlocks got replaces with semaphores(mutex?) so the proper thing is to turn on the mutex/blabla debugging.... in Ingos rt patch that is....

[gimpel]

thx for the hint!
hmm... i don't have any debugging enabled. perhaps i just got you wrong

i got
CONFIG_DEBUG_BUGVERBOSE=y which could explain that massive debug output, as i got some BUG entries in the logs
but "default !EMBEDDED"
no way to turn that off...

when i enable DEBUG_KERNEL is see that
# CONFIG_DEBUG_SPINLOCK_SLEEP is not set

if i search for DEBUG_PREEMPT in menuconfig:
DEBUG_PREEMPT [=n]
well because
PREEMPT [=n]
as there is
PREEMPT_VOLUNTARY [=y]

debug preempt deps on preempt

so there seems to be some kind of bug in ppp/net handling which causes a BUG signal that gets verbose in messages :/

[chickaroo]

it seems like there's more than one bug in that log. do you have thread softIRQ's on? i think that might still be a bit buggy.

[fallow]

ok I reprocuce this error . 2.6.11-rc2-vivid1 is bases on bk0 so clearly rc2 without bk.

http://lkml.org/lkml/2005/1/24/345
so it is fixed in bk

I`m now on 2.6.11-rc2-mm2 with fixed nvidia and fixed modules loading , so I don`t know what to do in first hehe , try to sync bk first or finish mm release .. no problems with softirq here.

cheers.

[Genkaku]

Nice kernel, but... i have to use driver "nv" in /etc/X11/xorg.conf to start X server. When i try "nvidia" i just get a black screen.
Yes, I know this kernel require agpgart, so i included it as a module, but it's no use.
My X is now slow, i see it for example when maximizing windows, etc.

Code: Select all

Section "Device"
    Identifier  "GeForce 2 MX400"
    Driver      "nv"

    Option	"NoLogo"	"true"
#    Option	"AGPMode"	"2"
#    Option	"AGPFastWrites"	"true"
	Option "UseInternalAGPGART" "yes"
#    Option	"EnablePageFlip" "true"
    Option 	"RenderAccel"	"true"
    Option	"Render"	"true"

Maybe I could manipulate the Option "AGPMode" section ?
Ummm... I remember that didn't worked.

But it's a great kernel.

[gimpel]

ok I reprocuce this error . 2.6.11-rc2-vivid1 is bases on bk0 so clearly rc2 without bk.

http://lkml.org/lkml/2005/1/24/345
so it is fixed in bk

I`m now on 2.6.11-rc2-mm2 with fixed nvidia and fixed modules loading , so I don`t know what to do in first hehe , try to sync bk first or finish mm release .. no problems with softirq here.

cheers.

just concentrate on your next release!
yesterday i tried a pure -rc2 with reiser4 and genetic (horribly failed on fbsplash, once more i realized the real great work you kernel-patch guys do!) and got the same errors, so it's not rt or something. currently i'm on a -rc2-bk8 with exactly the same setup, no problems here! even runs like a charme! but i'm really very interested in rt to do some audio stuff! i totally failed to patch in rt in -bk8...it just didn't compile

and: wow! a fixed -mm2 sounds good too! i tried -mm2 too, but the modules..heh!

[fallow]

new full mm ( 2.6.11-rc2-mm2 ) based release ( no any dmesg problems ).

for people want to use Kexec-tools , ebuild is here : http://bugs.gentoo.org/show_bug.cgi?id=50313

New feature in O(1) is Con`s ISO policy implementation. Now O(1) users can run programs in ISO mode like Staircase users. ISO means Isochronous (class for unprivileged soft RT scheduling).
( Con`s announce of it : http://ck.kolivas.org/patches/SCHED_ISO ... nounce.txt ). I think it`s good for jackd users for ex. and for all ohters who like to use ISO mode on Staircase before ( like me hehe )
To run some program with ISO mode You need :

Code: Select all

[ebuild   R   ] sys-apps/schedtool-1.2.4  

and to run something :

Code: Select all

schedtool -I -e something_program 

Main Mission Objectives

-->> mm2 modules support fix "for normal using" hehe
-->> good working reiser4 with CKO fixes ( I have whole system on it )
-->> O(1) CPU Scheduler with Con`s ISO policy implementation
-->> Genetic Anticipatory IO Scheduler based on Genetic Lib ver 0.3
-->> fbsplash,vesa-tng of course (both works - I`m using it all the time )
-->> all stuff like Kexec,FUSE,CacheFS and many others is in mm2 http://www.kernel.org/pub/linux/kernel/ ... nounce.txt
-->> Con`s mwII for "VM managment"

patchlist & info :

Code: Select all

--patches--

(r) - reversed agp changes for working nvidia
(r) - reversed some of vt cleanups for working fbsplash
(r) - reversed scheduler RT_Limits , apllied ISO class/policy (Con Kolivas)

lkml_2611rc2mm2_modules1.fix - kernel modules support fix for mm2 from lkml.
lkml_2611rc2mm2_xfs1.fix - KConfig fix for xfs (linker error)
mmvivid1_menu.diff - vivid menu in menuconfig
fbsplash-0.9.1-r1-2.6.11-rc1-bk8.patch - no comment also ;)
vesafb-tng-0.9-rc6-2.6.11-rc1.patch - no comment also ;)
reiser4_repacker_stats.patch.bz2 - 1st fix from CKO
reiser4_replay_stats.patch.bz2 - 2nd fix from CKO
genetic-as-sched-2.6.10-gl3.patch - gen.alg.implementation in anticipatory io sched
genetic-io-sched-2.6.10-gl3.patch - connection to genetic lib
genetic-lib-2.6.10-gl3.patch - genetic lib v0.3
sched-isochronous-class-for-unprivileged-soft-rt-scheduling.patch - as in name (mm1)
sched-account-rt_tasks-as-iso_ticks.patch - as in name (mm1)
mwII.diff - Con`s virtual mem "managment" ;)
nvidia_6111-6629_compat2.diff - pack of nvidia fixes.
squashfs2.0-patch - squash fs :)
supermount-ng208-2611rc1.diff - supermount from CK.
390-slabtune.txt - slab.c tunes from MJB.
500-sched_tunables.txt - O(1) scheduler tunables from MJB (via /proc)
acerhk.patch - acer keyboard driver.
acpi-dsdt-initrd-patch-v0.7d - http://acpi.sourceforge.net/dsdt/index.php
atxp1-0.3.patch - dynamic CPU speed with nforce 2
cddvd-cmdfilter-drop.patch - non root hack from ck.
cflags-selection.patch - tool for easy kernel cflags changing.
chmp-r5-FULL.patch - configurable hid mouse polling.
config-nr-tty-devices.patch - for setting number of tty`s.
config_hz.diff - internal kernel timer setting.
madwifi-cvs - wireless madwifi driver.
lirc-2.6.9-20041024-px1.patch - lirc driver
lufs-0.9.7-2.6.0-test9.patch.bz2 - fixed lufs.
pio_font_fix1.diff - pio font error with fbsplash fix.
realtime_lsm_module.patch - RT security extension
scheduler_change_script - for easy io scheduler changing script ( Genn )
--proc interfaces---
/proc/genetic - genetic as io scheduler 
/proc/sys/sched - O(1) cpu scheduler 
/proc/sys/vm - mwII

available at sourceforge.net :
https://sourceforge.net/projects/vivid-sources/
or :
http://sourceforge.net/project/showfile ... _id=301391


Tested with whole reiser4 system and nvidia-kernel 6629-r3 from portage via AGPGART , previous ebuilds before r3 doesn`t work with 2.6.11-rc2 kernel(s) , req fix - from r3 ebuild.

of course ,feedback , ideas , concepts etc. -> welcome
have a nice day

[fallow]

Nice kernel, but... i have to use driver "nv" in /etc/X11/xorg.conf to start X server. When i try "nvidia" i just get a black screen.
Yes, I know this kernel require agpgart, so i included it as a module, but it's no use.
(cut cut)
But it's a great kernel.

Did You load agpgart module before start ?
I have compiled agpgart static, my xorg.conf config is similar to Your.but I have some time ago problem with 6629 and black screen on all of new kernels. (agpgart or nvagp always I`ve got black screen) , I configured my kernel .config from "zero" with base config from Troll ( He had working 6629 ) and since this - it`s working good. If You want I will send You my xorg.conf, kernel .config etc.

cheers

[Genkaku]

Yup, i loaded agpgart module before start
I tryed also without it.

So... could you send me your configs ?
My mail: genkaku@interia.pl
Hyhy, thanks.

[babykart]

the 2.6.11-rc2-mmvivid1 works fine for me with HyperThreading, reiser4, nvidia... (better than the 2.6..11-rc2-vivid1...)...
thx for the job fallow


PS - where is my "1Gb Low Mem" ....