2.6.18-rad1 "My Rad" | 50% mm, all ck, nosched, stable!

Message

cheater1034 · Post by **cheater1034** » Sat Sep 30, 2006 10:29 pm

2.6.18-rad1 "My Rad"
STABLE RELEASE
----
This is a one-of-a-kind patchset , I did do extreme stability testing on this, and put some extreme work to bring this to you, it has several features, and is not based on one thing alone, but based on several things, making it more independent.
----
IMPORTANT:
Brand new patchset, builds with allyesconfig, allmodconfig, and allnoconfig, my best guess is it's stable! And one of a kind, I've never seen a patchset quite like this before

----
Featurelist
- Orignates off of 2.6.18 vanilla
- 2.6.18-git13
- NoSched
- close to 50% or more of the 2.6.18-mm2 patchset, performance, and stability
- Entire patchset for thinkpads, except suspend2.
- All of 2.6.18-ck1
- Staircase Tunables
- Reiser4 2.6.18-mm2
- fixed config_nr_tty_devices
- Entire RCU features from -mm, and lockless readside, etc.
- RCU Preempt patch set.
- Netfilter layer7
- mm tree includes acx1xx, ecryptfs, etc.
- vesa-tng fixed correctly*
----
Patchlist:

Code: Select all

#git
patch-2.6.18-git12.bz2
git12-git13.patch

2.6-sony_acpi4_and_resume.patch			| Sony acpi4

# NoSched patches
NoSched-01-Sched.c				| NoSched: cleanup sched.c
NoSched-02-Makefile				| NoSched: do makefile
NoSched-03-Kconfig				| NoSched: do Kconfig
NoSched-04-Ingosched-2.6.18-rad1		| NoSched: implement ingosched
NoSched-05-Staircase16.2-base			| NoSched: implement basic staircase
NoSched-06-Staircase16.2-interactive		| NoSched: Staircase int. tunable
NoSched-07-Staircase16.2-compute		| NoSched: Staircase comp. tunable
NoSched-08-Staircase16.2-range			| NoSched: Staircase range
NoSched-09-Staircase16.2-headers_ifdefs		| NoSched: do ifdefs
NoSched-10-Staircase16.2-iso-4.6		| NoSched: implement sched_iso
NoSched-11-Staircase-idleprio-1.11		| NoSched: implement idleprio
NoSched-12-Max_Delay				| NoSched: implement max_delay stat
NoSched-13-Viper-Tunes				| NoSched: implement staircase tunables
# Scheduler patches (no descriptions)
lower-migation-thread-stop-machine-prio.patch	
sched-add-above-background-load-function.patch
sched-cleanup-remove-task_t-convert-to-struct-task_struct-prefetch.patch
sched-cleanup-sched_group-cpu_power-setup.patch
sched-dont-print-migration-cost-when-only-1-cpu.patch
sched-fixing-wrong-comment-for-find_idlest_cpu.patch
sched-force-sbin-init-off-isolated-cpus.patch
sched-introduce-child-field-in-sched_domain.patch
sched-remove-unnecessary-sched-group-allocations.patch
sched2-sched-domain-sysctl.patch
scheduler-numa-aware-placement-of-sched_group_allnodes.patch

# Thinkpad Patches
acpi-removable_bay-for-2.6.18.patch		| removable bay thinkpad feature
ibm_acpi-bay-for-2.6.18.patch			| acpi bay
ibm_acpi-fan-for-2.6.18.patch			| acpi extended fancontrol
ibm_acpi-thermal-for-2.6.18.patch		| acpi thermal control
input-unknown_keycodes-for-2.6.18.patch		| ibm keycodes
libata-shutdown-for-2.6.18.patch		| libata shutdown
tp_smapi-0.3.0-for-2.6.18.patch			| Implement thinkpad SMAPI
speedstep-centrino-undervolt-0.2.6-for-2.6.18.patch | Linux-PHC

# -mm patches 
all-rcu-mm-patches_also-fix_and_upgrade-rcupreempt.patch | Enhance the RCU, and upgrade rcu preempt patch
git-drm.patch					| Implement the git-drm tree
mm2-big-rollup-1.patch				| Big -mm rollups -> tons of patches
mm2-big-rollup-2.patch				
mm2-big-rollup-3.patch
mm2-big-rollup-4.patch
mm2-big-rollup-5.patch
mm2-big-rollup-6.patch

# NR_TTY (now fixed for splashutils)
config-nr-tty-devices-rev5.patch		| configure tty devices in /dev, fixed for splashutils

# -tiny patches
cpu-support.patch				| selectable extendable cpu support
movsl-mask.patch				| fixes:
mtrr-cpusupport.patch

# damouse patches
daconfig-2.3.1.patch				| damouse's cool script

# spock patches
fbsplash-0.9.2-r5.patch				| gensplash
vesa-tng-module-fix.patch			| fix vesa-tng
vesafb-tng-1.0-rc2.patch			| vesafb-tng

# -ck patches
cfq-ioprio_inherit_rt_class_all.patch		| no descriptions:
cfq-iso_idleprio_ionice-staircase-only.patch
kconfig-expose_vmsplit_option.patch
mm-background_scan-2.patch
mm-convert_swappiness_to_mapped.patch
mm-decrease_minimum_dirty_ratio.patch
mm-filesize_dependent_lru_cache_add.patch
mm-idleprio_prio-staircase-only.patch
mm-kswapd_inherit_prio-1.patch
mm-lots_watermark.patch
mm-prio_dependent_scan-1.patch
mm-set_zero_dirty_ratio.patch
track-mutexes-1.patch

# netfilter layer 7			
netfilter-layer7-2.6.18.patch

# optimized default kconfig settings
no-optima-defs.patch

# swap prefetching
mm-implement-swap-prefetching.patch
swap_prefetch-vs-zoned-counters.patch
numa-add-zone_to_nid-function-swap_prefetch.patch
reduce-max_nr_zones-swap_prefetch-remove-incorrect-use-of-zone_highmem.patch
zvc-support-nr_slab_reclaimable--nr_slab_unreclaimable-swap_prefetch.patch

# preemptible rcu
rcu-preempt-2.6.18.patch			| preempt rcu option, and implementation

# realtime-lsm
realtime-lsm.patch

# reiser4
reiser4-all-2.6.18-mm2.patch			| latest reiser4

# squashfs
squashfs-3.1-r2_working_port.patch		| port of squashfs

#usb
usb-mouse-polling.patch
usbxchange5.patch

tag						| tag rad

----
Downloads:
EBUILD
PATCH
NOTES

Good luck

vipernicus · Post by **vipernicus** » Sat Sep 30, 2006 11:58 pm

You might want to try modularizing your big-mm-rollups according to features. Maybe, "fixes", "rcu", "sched", "io", etc.

That way others can benefit more from your broken-out.

p4r0l3 · Post by **p4r0l3** » Sun Oct 01, 2006 4:21 am

Congrats on the new patchset!

Definately looks great cheater. I'm compiling right now....hopefully I won't get the reiser4 problems I've been having with 2.6.18 but I'll let ya know

mbar · Post by **mbar** » Sun Oct 01, 2006 6:15 am

OMG a new series

bigbob73 · Post by **bigbob73** » Sun Oct 01, 2006 11:50 am

nvidia-drivers bombs on me...

open_wr: /usr/src/linux-2.6.18-rad1/astest7214.out
open_wr: /usr/src/linux-2.6.18-rad1/astest7218.out
open_wr: /usr/src/linux-2.6.18-rad1/astest7225.out
open_wr: /usr/src/linux-2.6.18-rad1/astest7229.out

masterdriverz · Post by **masterdriverz** » Sun Oct 01, 2006 12:23 pm

bigbob73 wrote:nvidia-drivers bombs on me...

open_wr: /usr/src/linux-2.6.18-rad1/astest7214.out
open_wr: /usr/src/linux-2.6.18-rad1/astest7218.out
open_wr: /usr/src/linux-2.6.18-rad1/astest7225.out
open_wr: /usr/src/linux-2.6.18-rad1/astest7229.out

I'm guessing thats a sandbox error. Try with FEATURES="-sandbox -usersandbox".

n0rbi666 · Post by **n0rbi666** » Sun Oct 01, 2006 12:54 pm

Wow, new patchset !

but ... with 2.6.18-rad1, and with 2.6.18-no1 mozilla-firefox-2.0_rc1 doesn't work for me (firefox-bin exited with non-zero status (1) )
here is my .config ...

vipernicus · Post by **vipernicus** » Sun Oct 01, 2006 1:21 pm

n0rbi666 wrote:Wow, new patchset !
but ... with 2.6.18-rad1, and with 2.6.18-no1 mozilla-firefox-2.0_rc1 doesn't work for me (firefox-bin exited with non-zero status (1) )
here is my .config ...

At this point, try Gentoo-Sources and see if Firefox works.

n0rbi666 · Post by **n0rbi666** » Sun Oct 01, 2006 3:11 pm

vipernicus - with gentoo-sources-2.6.18 and config from 2.6.18-rad1 firefox works ok.

.t. · Post by **.t.** » Sun Oct 01, 2006 4:53 pm

Plus, ieee80211_crypt_wep is broken; or the crypto API - as the WEP module fails using it. gentoo- and git- and beyond- sources all work fine. I can't use your sources unless I can't get on the internet, although I'd love to see any performance differences.

Kollin · Post by **Kollin** » Sun Oct 01, 2006 5:27 pm

Smooches dear Cheater - this is first 2.6.18 kernel that plays encrypted dvds on my system

wrc1944 · Post by **wrc1944** » Sun Oct 01, 2006 5:27 pm

Compiles and runs perfectly (first try) on my SATA main box!

I did notice on the boot screen 10+ lines fly by with IIRC something about alsactrl, but lsmod shows all my sound modules loaded as usual. Nothing about this in my log files.

Dmesg also has normal complete output. All in all, 2.6.18-rad1 seems great so far for my day-to-day usage.

Will try rad1 later today on my ide box out in my home office, which is out of necessity wireless, where no1 and mm1 would compile and boot, but not compile my madwifi-ng drivers.

vipernicus · Post by **vipernicus** » Sun Oct 01, 2006 5:54 pm

wrc1944 wrote:Compiles and runs perfectly (first try) on my SATA main box!

I did notice on the boot screen 10+ lines fly by with IIRC something about alsactrl, but lsmod shows all my sound modules loaded as usual. Nothing about this in my log files.

Dmesg also has normal complete output. All in all, 2.6.18-rad1 seems great so far for my day-to-day usage.

Will try rad1 later today on my ide box out in my home office, which is out of necessity wireless, where no1 and mm1 would compile and boot, but not compile my madwifi-ng drivers.

Try running alsaconf again.

maudib · Post by **maudib** » Sun Oct 01, 2006 6:18 pm

n0rbi666 wrote: but ... with 2.6.18-rad1, and with 2.6.18-no1 mozilla-firefox-2.0_rc1 doesn't work for me (firefox-bin exited with non-zero status (1) )

exact same thing here..works in -ck1

wrc1944 · Post by **wrc1944** » Sun Oct 01, 2006 7:26 pm

On my ide wireless box, rad1 compiles and boots fine, but madwifi still fails with same warnings/errors as 2.6.18-no1, mm1, and viper1.

rc4-no1, rc4-viper1, and 2.6.18-ck1 wireless work fine.

I've tried minimal C and CXX flags, unsetting LDFLAGS, and using:

env MAKEOPTS="-j1" emerge -1 madwifi-ng

and CFLAGS="-w" which is suppose to supress the "warnings treated as errors" effects, but apparently doesn't.

If the problems were in madwifi-ng, seems like it wouldn't work with ck1 either.

Has anyone gotten madwifi-ng to compile on no1, mm1, mm2, viper1, or rad1? Guess I'll just use ck1 for now on my wireless box. I can post my errors to pastebin and link, if anyone's interested.

klixon · Post by **klixon** » Sun Oct 01, 2006 9:15 pm

concerning the sandbox violations when emerging modules, I've come up with a patch to 2.6.18-mm1 (and probably this kernel as well, since it's the same symptoms

).
see bug #149307. The patch is supplied as an attachment there...

hope this helps, JP

.t. · Post by **.t.** » Sun Oct 01, 2006 9:57 pm

For your madwifi and my ipw2200 problems: the problem is due to ieee80211_crypt_wep not being able to access the crypto API, so cannot do any encrypting. Perhaps a way to test if this is indeed the problem would be to try and use an open network and if it works, we know where the problem lies.

Kollin · Post by **Kollin** » Sun Oct 01, 2006 10:26 pm

klixon wrote:concerning the sandbox violations when emerging modules, I've come up with a patch to 2.6.18-mm1 (and probably this kernel as well, since it's the same symptoms ).
see bug #149307. The patch is supplied as an attachment there...

hope this helps, JP

Confirming here - patch or

Code: Select all

FEATURES=-sandbox emerge x11-drm

P.S. This is nvidia bug, but i have ati rage128 video card and it`s the same problem

wrc1944 · Post by **wrc1944** » Sun Oct 01, 2006 10:57 pm

klixon's a genius! Many thanks!

FEATURES="-sandbox" emerge =net-wireless/madwifi-ng-0.9.2

works for me. Now rad1 works perfectly on my wireless office box.

I've been going round and round editing kernel .h files, madwifi-ng source files, and various config options, when if I wasn't so ignorant I would have paid more attention to:

--------------------------- ACCESS VIOLATION SUMMARY ---------------------------
LOG FILE = "/var/log/sandbox/sandbox-net-wireless_-_madwifi-ng-0.9.2-11518.log"

open_wr: /home/wrc/kern/linux-2.6.18-rad1/astest11655.out
open_wr: /home/wrc/kern/linux-2.6.18-rad1/astest11659.out
open_wr: /home/wrc/kern/linux-2.6.18-rad1/astest11663.out
open_wr: /home/wrc/kern/linux-2.6.18-rad1/astest11667.out
open_wr: /home/wrc/kern/linux-2.6.18-rad1/astest11814.out
open_wr: /home/wrc/kern/linux-2.6.18-rad1/astest11818.out
open_wr: /home/wrc/kern/linux-2.6.18-rad1/astest11822.out
open_wr: /home/wrc/kern/linux-2.6.18-rad1/astest11826.out

and figured it out myself, instead of following all the other blind alleys, like:

Building modules, stage 2.
MODPOST 2 modules
WARNING: "ath_rate_setupxtxdesc" [/var/tmp/portage/madwifi-ng-0.9.2/work/madwifi-0.9.2/ath/ath_pci.ko] undefined!
WARNING: "ath_rate_newstate" [/var/tmp/portage/madwifi-ng-0.9.2/work/madwifi-0.9.2/ath/ath_pci.ko] undefined!
WARNING: "ieee80211_find_txnode" [/var/tmp/portage/madwifi-ng-0.9.2/work/madwifi-0.9.2/ath/ath_pci.ko] undefined! MANY MORE LINES OF THESE

AND:

cc1: warnings being treated as errors
/var/tmp/portage/madwifi-ng-0.9.2/work/madwifi-0.9.2/net80211/ieee80211_crypto_ccmp.c: In function 'rijndael_encrypt':
/var/tmp/portage/madwifi-ng-0.9.2/work/madwifi-0.9.2/net80211/ieee80211_crypto_ccmp.c:310: warning: 'crypto_cipher_encrypt' is deprecated (declared at include/linux/crypto.h:842)
make[2]: *** [/var/tmp/portage/madwifi-ng-0.9.2/work/madwifi-0.9.2/net80211/ieee80211_crypto_ccmp.o] Error 1
make[1]: *** [_module_/var/tmp/portage/madwifi-ng-0.9.2/work/madwifi-0.9.2/net80211] Error 2

EDIT: Forgot to mention, all my problems were on an open network (no wep), so doesn't sound like that's the source of the problem.
---------------------------------------------

broch · Post by **broch** » Mon Oct 02, 2006 1:03 am

maudib wrote:
n0rbi666 wrote: but ... with 2.6.18-rad1, and with 2.6.18-no1 mozilla-firefox-2.0_rc1 doesn't work for me (firefox-bin exited with non-zero status (1) )
exact same thing here..works in -ck1

same problem (FF 1.5.0.7) + thunderbird (1.5.0.7) will not start either (quits without any error messages)

Kollin · Post by **Kollin** » Mon Oct 02, 2006 8:56 am

I have FF 1.5.0.7 and it`s working perfectly

broch · Post by **broch** » Mon Oct 02, 2006 1:23 pm

In my case FF works with vanilla 2.6.18 and -ck, but not with -rad1. removing user setting does not help.

vipernicus · Post by **vipernicus** » Mon Oct 02, 2006 1:44 pm

broch wrote:In my case FF works with vanilla 2.6.18 and -ck, but not with -rad1. removing user setting does not help.

What about 'emerge mm-sources'? Does Firefox work with plain mm-sources?

DocterD · Post by **DocterD** » Mon Oct 02, 2006 1:55 pm

Could you add next time Dynticks?

http://www.tglx.de/projects/hrtimers/2.6.18/

broch · Post by **broch** » Mon Oct 02, 2006 2:24 pm

vipernicus wrote:
broch wrote:In my case FF works with vanilla 2.6.18 and -ck, but not with -rad1. removing user setting does not help.
What about 'emerge mm-sources'? Does Firefox work with plain mm-sources?

sorry, can't do

running suse.
However, -rad1 also "kills" cisco vpn.
I am using ieee80211-1.2.15 and ipw3945-1.1.0
Everything compiles fine and network (not tunel) works, however vpnclient complains about not being able to use ipw3945.
I would eventually install FF from sources, but vpn is quite critical.
Again all the above works under vanilla or-ck.