2.6.18-viper1: Mutatis Mutandis

Message

vipernicus · Post by **vipernicus** » Sat Sep 23, 2006 12:25 am

Viper Sources is pretty much my playground for new patches.

The main goal of this patchset is pretty much the same as E-Mission Sources: to
offer latest features, decrease latencies, decrease overhead, and improve interactivity. The only difference is experimental versions of everything. I boot test each release and run for an extended amount of time.

/* Notes */
Alot of changes in this version, I dropped Lockless, RCU-Preempt, and Adaptive Readahead (for now) in favor
of playing around with the new Compressed Pagecache. http://linuxcompressed.sourceforge.net/

The LinuxCompressed Site wrote:Compressed caching is the introduction of a new level into the virtual memory hierarchy. Specifically, RAM is used to store both an uncompressed cache of pages in their 'natural' encoding, and a compressed cache of pages in some compressed format. By using RAM to store some number of compressed pages, the effective size of RAM is increased, and so the number of page faults that must be serviced by very slow hard disks is decreased. Our aim is to improve system performance. When that is not possible, our goal is to introduce none (or a minimum overhead) when compressed caching is enabled in the system.

It's not enabled by default. Read: http://linuxcompressed.sourceforge.net/help/index.html

Example of new E-Mission Tunables (All but Custom are preconfigured options):

Code: Select all

E-Mission Tunables  ---> 
( ) Server
( ) File Server
( ) Compute Server
( ) Gaming
( ) Desktop
( ) Low Latency Desktop
( ) Custom

Custom Menu:

Code: Select all

E-Mission Kernel Tunables (Custom)  --->
(6) Minimum Round Robin Timeslice (in milliseconds) (NEW)
(19) Factor For Default Value (NEW)
(1) Interactive CPU Scheduling (NEW)
(0) Compute CPU Scheduling (NEW)
(80) Percentage CPU For SCHED_ISO Tasks (NEW)
(66) Percentage RAM Filled With Mapped Pages (NEW)
(1) Hard Mapped Limit (NEW)
(1) Tail Large Files (NEW)
(0) Dirty Ratio (NEW)

2.6.18-viper1: Mutatis Mutandis

Code: Select all

/* Patch List */ 
00-genpatches-rollup-2.6.18.patch
01-patch-2.6.18-ck1
02-iosched-rollup-2.6.18-rc7-mm1.patch
03-patch-ccache-alpha-007-2.6.18
04-gregkh-all-2.6.18.patch
05-sleazy-fpu-feature.patch
06-vectorize-aio-rollup.patch
07-reiser4-for-2.6.18-rc4-mm2.patch
08-unionfs-latest.patch
09-realtime-lsm-2.6.18.patch
11-cpu-support-rollup-2.6.18.patch
12-inode-diet-rollup-2.6.18-rc7-mm1.patch
13-inode-diet-reiser4-2.6.18-rc7-mm1.patch
14-inode-diet-squashfs-3.0.patch
15-inode-diet-unionfs-latest.patch
16-linux-phc-0.2.6-2.6.18.patch
17-build-fixes.patch
18-ati-fix.patch
19-lxdialog-2.6.18-rc7-mm1.patch
20-daconfig-2.3.1-2.6.18.patch
21-emission-tunables-4-2.6.18-ck0.patch
22-reiser4-update-2.6.18-rc7-mm1.patch
24-mm-rollup-2.6.18-rc7-mm1.patch
25-ntp-rollup-2.6.18-rc5-mm1.patch
26-ro-bind-mount-2.6.18-rc7-mm1.patch
28-non-libata-driver-for-jmicron-devices.patch
29-usb-mouse-polling-sysctl-2.6.18.patch
31-optims-from-mm-2.6.18-rc7-mm1.patch
32-fixes-from-mm-2.6.18-rc7-mm1.patch
34-name-tag-2.6.18-viper1.patch

Enjoy!

Patch:
http://vipernicus.evolution-mission.org ... viper1.bz2

Ebuild:
http://vipernicus.evolution-mission.org ... -r1.ebuild

Broken-Out:
http://vipernicus.evolution-mission.org ... ut.tar.bz2

Broken-Out (Directory):
http://vipernicus.evolution-mission.org ... 1/patches/

Making it work:

Ebuild Method:

Code: Select all

mkdir -p /usr/local/portage/sys-kernel/viper-sources/files
cd /usr/local/portage/sys-kernel/viper-sources/
wget http://vipernicus.evolution-mission.org/patches/2.6.18/viper1/viper-sources-2.6.18-r1.ebuild
cd /usr/local/portage/sys-kernel/viper-sources/files/
wget http://distfiles.evolution-mission.org/scripts/emission-tunables 
cd /usr/local/portage/sys-kernel/viper-sources/
ebuild viper-sources-2.6.18-r1.ebuild digest
USE="symlink" emerge viper-sources
cd /usr/src/linux
make MENUCONFIG_COLOR=blackbg menuconfig

Manual Method:

Code: Select all

cd /usr/src
rm linux
wget http://www.kernel.org/pub/linux/kernel/v2.6/linux-2.6.18.tar.bz2
tar -xvjf linux-2.6.18.tar.bz2
mv linux-2.6.18 linux-2.6.18-viper1
ln -s linux-2.6.18-viper1 linux
cd linux
wget http://vipernicus.evolution-mission.org/patches/2.6.18/viper1/linux-2.6.18-viper1.bz2
bzcat linux-2.6.18-viper1.bz2 | patch -p1
make MENUCONFIG_COLOR=blackbg menuconfig

Nesl247 has created a script to switch the settings while running the current kernel and permanently through sysctl.conf

E-Mission Tunables Script 2.2.1:

http://distfiles.evolution-mission.org/ ... n-tunables

How-to Install:

Code: Select all

wget http://distfiles.evolution-mission.org/scripts/emission-tunables --output-document=/usr/bin/emission-tunables
chmod +x /usr/bin/emission-tunables

How-to Use (Choose one of the following options):

Code: Select all

emission-tunables compute-server
emission-tunables desktop
emission-tunables file-server
emission-tunables gaming
emission-tunables ll-desktop
emission-tunables server
emission-tunables custom

This lets you switch between modes while running kernel.

For all support, requests, patches, please check the Emission-Sources Bugzilla Tracker
and check Emission-Sources Forums for new releases!

vipernicus · Post by **vipernicus** » Sat Sep 23, 2006 1:42 am

Suspend2 removed from patch. If you d/led it before 20:41 CDT, try again.

dbcoder · Post by **dbcoder** » Sat Sep 23, 2006 4:34 am

I've been running for several hours rock solid. Nice and snappy too

. Good job!!!

Kirktis · Post by **Kirktis** » Sat Sep 23, 2006 7:34 am

For reasons I have not yet determined, building inevitably fails when I tell it I don't want the Rio Karma drivers... I'll post back when I discover more.

IvanMajhen · Post by **IvanMajhen** » Sat Sep 23, 2006 11:09 am

Why did you remove Suspend2? It is working here.
EDIT: Ok, I saw that there is still suspend2 as separate patch.

What are good values for Compressed Pagecache? The page dosn't say much. Will it be added to script? I added it to sysctl.conf

bigbob73 · Post by **bigbob73** » Sat Sep 23, 2006 11:31 am

Kirktis wrote:For reasons I have not yet determined, building inevitably fails when I tell it I don't want the Rio Karma drivers... I'll post back when I discover more.

getting the same here...
CC [M] drivers/usb/storage/usb.o
In file included from drivers/usb/storage/usb.c:179:
drivers/usb/storage/unusual_devs.h:221: error: 'US_PR_KARMA' undeclared here (not in a function)
drivers/usb/storage/unusual_devs.h:221: error: 'rio_karma_init' undeclared here (not in a function)
make[3]: *** [drivers/usb/storage/usb.o] Error 1
make[2]: *** [drivers/usb/storage] Error 2
make[1]: *** [drivers/usb] Error 2
make: *** [drivers] Error 2

vipernicus · Post by **vipernicus** » Sat Sep 23, 2006 12:54 pm

Alright, I fixed the patch again, hopefully this will do it.

GregKH's Rio Karma patch was incomplete.

Lather, rinse, repeat.

Code: Select all

rm /usr/portage/distfiles/linux-2.6.18-viper1.bz2
emerge viper-sources

wrc1944 · Post by **wrc1944** » Sat Sep 23, 2006 2:38 pm

Isn't this new Emission tunables script similar to (or a "linux" version of) the old win95/98 registry tweak where you can add the options as posted on the axcel216 website to the Control Panel -> (double)-click System applet -> Performance tab -> File System button -> Hard Disk tab -> "Typical role of this computer" box?

http://www.mdgx.com/newtip16.htm
scroll down about halfway, and look at:
MAX CACHE SPEED

And/or similar to the 3rd party "Cacheman" windows utility? http://www.outertech.com/?_charisma_page=product&id=2

I'm also curious about one thing:
If you have lots of ram, and never go into swap at all, just how relevant is the Compressed Pagecache going to be? I guess effectively "adding" ram is good, but if you never use swap anyway, wouldn't any overhead caused by Compressed Pagecache be counter-productive?

In other words, if one never uses swap, wouldn't dropping Lockless, RCU-Preempt, and Adaptive Readahead be less effective than adding Compressed Pagecache, which on some systems might be essentially useless?

tranquilcool · Post by **tranquilcool** » Sat Sep 23, 2006 4:06 pm

nvidia-drivers wont build.

any help?

dR0PS · Post by **dR0PS** » Sat Sep 23, 2006 4:14 pm

same prob here, cannot help except detailed output; checked with nvidia-drivers-1.0.8762
cannot check for newer bins because of other reasons

Code: Select all

  i686-pc-linux-gnu-gcc -Wp,-MD,/var/tmp/portage/nvidia-drivers-1.0.8762-r1/work/NVIDIA-Linux-x86-1.0-8762-pkg1/usr/src/nv/.nv-i2c.o.d  -nostdinc -isystem /usr/lib/gcc/i686-pc-linux-gnu/4.1.1/include -D__KERNEL__ -Iinclude  -include include/linux/autoconf.h -Wall -Wundef -Wstrict-prototypes -Wno-trigraphs -fno-strict-aliasing -fno-common -Os -pipe -msoft-float -mpreferred-stack-boundary=2  -march=athlon  -ffreestanding -Iinclude/asm-i386/mach-default -fomit-frame-pointer  -fno-stack-protector -Wdeclaration-after-statement -Wno-pointer-sign -I/var/tmp/portage/nvidia-drivers-1.0.8762-r1/work/NVIDIA-Linux-x86-1.0-8762-pkg1/usr/src/nv -Wall -Wimplicit -Wreturn-type -Wswitch -Wformat -Wchar-subscripts -Wparentheses   -Wno-multichar  -Werror -O -fno-common -MD    -Wno-cast-qual -Wno-error -D_LOOSE_KERNEL_NAMES -D__KERNEL__ -DMODULE  -DNTRM -DNVRM -DDYNAMIC_SLI  -DNV_MAJOR_VERSION=1 -DNV_MINOR_VERSION=0 -DNV_PATCHLEVEL=8762  -UDEBUG -U_DEBUG -DNDEBUG -DNV_SIGNAL_STRUCT_RLIM -DNV_MULTIPLE_BRIDGE_AGPGART_PRESENT -DNV_PCI_GET_CLASS_PRESENT -DNV_SYSCTL_MAX_MAP_COUNT_PRESENT -DNV_PM_MESSAGE_T_PRESENT -DNV_PCI_CHOOSE_STATE_PRESENT -DNV_VM_INSERT_PAGE_PRESENT -DNV_REMAP_PFN_RANGE_PRESENT -DNV_CHANGE_PAGE_ATTR_PRESENT -DNV_VMAP_4_PRESENT  -DMODULE -D"KBUILD_STR(s)=#s" -D"KBUILD_BASENAME=KBUILD_STR(nv_i2c)"  -D"KBUILD_MODNAME=KBUILD_STR(nvidia)" -c -o /var/tmp/portage/nvidia-drivers-1.0.8762-r1/work/NVIDIA-Linux-x86-1.0-8762-pkg1/usr/src/nv/.tmp_nv-i2c.o /var/tmp/portage/nvidia-drivers-1.0.8762-r1/work/NVIDIA-Linux-x86-1.0-8762-pkg1/usr/src/nv/nv-i2c.c
/var/tmp/portage/nvidia-drivers-1.0.8762-r1/work/NVIDIA-Linux-x86-1.0-8762-pkg1/usr/src/nv/nv-i2c.c:26: error: unknown field ‘slave_send’ specified in initializer
/var/tmp/portage/nvidia-drivers-1.0.8762-r1/work/NVIDIA-Linux-x86-1.0-8762-pkg1/usr/src/nv/nv-i2c.c:27: error: unknown field ‘slave_recv’ specified in initializer
make[3]: *** [/var/tmp/portage/nvidia-drivers-1.0.8762-r1/work/NVIDIA-Linux-x86-1.0-8762-pkg1/usr/src/nv/nv-i2c.o] Error 1
make[2]: *** [_module_/var/tmp/portage/nvidia-drivers-1.0.8762-r1/work/NVIDIA-Linux-x86-1.0-8762-pkg1/usr/src/nv] Error 2
NVIDIA: left KBUILD.
nvidia.ko failed to build!
make[1]: *** [mdl] Error 1
make: *** [module] Error 2

dR0PS

wrc1944 · Post by **wrc1944** » Sat Sep 23, 2006 4:15 pm

Just finished compiling, and rebooting- works like a charm!

I selected low-latency desktop in the Emission tunables make xconfig section.

Can anyone give us a brief explanation/rundown of the theory and rationale invovled for choosing the values assigned to each of the different usages?

Example of new E-Mission Tunables (All but Custom are preconfigured options):
Code:
E-Mission Tunables --->
( ) Server
( ) File Server
( ) Compute Server
( ) Gaming
( ) Desktop
( ) Low Latency Desktop
( ) Custom

Custom Menu:
Code:
E-Mission Kernel Tunables (Custom) --->
(6) Minimum Round Robin Timeslice (in milliseconds) (NEW)
(19) Factor For Default Value (NEW)
(1) Interactive CPU Scheduling (NEW)
(0) Compute CPU Scheduling (NEW)
(80) Percentage CPU For SCHED_ISO Tasks (NEW)
(66) Percentage RAM Filled With Mapped Pages (NEW)
(1) Hard Mapped Limit (NEW)
(1) Tail Large Files (NEW)
(0) Dirty Ratio (NEW)

vipernicus · Post by **vipernicus** » Sat Sep 23, 2006 5:19 pm

wrc1944 wrote:Just finished compiling, and rebooting- works like a charm!

I selected low-latency desktop in the Emission tunables make xconfig section.

Can anyone give us a brief explanation/rundown of the theory and rationale invovled for choosing the values assigned to each of the different usages?

Example of new E-Mission Tunables (All but Custom are preconfigured options):
Code:
E-Mission Tunables --->
( ) Server
( ) File Server
( ) Compute Server
( ) Gaming
( ) Desktop
( ) Low Latency Desktop
( ) Custom

Custom Menu:
Code:
E-Mission Kernel Tunables (Custom) --->
(6) Minimum Round Robin Timeslice (in milliseconds) (NEW)
(19) Factor For Default Value (NEW)
(1) Interactive CPU Scheduling (NEW)
(0) Compute CPU Scheduling (NEW)
(80) Percentage CPU For SCHED_ISO Tasks (NEW)
(66) Percentage RAM Filled With Mapped Pages (NEW)
(1) Hard Mapped Limit (NEW)
(1) Tail Large Files (NEW)
(0) Dirty Ratio (NEW)

On Con Kolivas' site, he recommends those different numbers for the different situations. I also looked into what values he used for ckserver.

Server - CK Server
File Server - Tuned like server, but vm.mapped = 0, since a File Server should be using memory more for file cache and not for applications.
Compute Server - Tuned like file server, but with Compute mode turned on.
Gaming - Tuned like server, but with Compute mode turned on.
Desktop - Default CK
Low Latency Desktop - Old Nitro Recommendation
Custom - defaults are set to Desktop settings.

tranquilcool · Post by **tranquilcool** » Sat Sep 23, 2006 5:27 pm

SENAME=KBUILD_STR(nv_i2c)" -D"KBUILD_MODNAME=KBUILD_STR(nvidia)" -c -o /var/tmp/portage/nvidia-drivers-1.0.8774/work/NVIDIA-Linux-x86-1.0-8774-pkg1/usr/src/nv/.tmp_nv-i2c.o /var/tmp/portage/nvidia-drivers-1.0.8774/work/NVIDIA-Linux-x86-1.0-8774-pkg1/usr/src/nv/nv-i2c.c
/var/tmp/portage/nvidia-drivers-1.0.8774/work/NVIDIA-Linux-x86-1.0-8774-pkg1/usr/src/nv/nv-i2c.c:26: error: unknown field âslave_sendâ specified in initializer
/var/tmp/portage/nvidia-drivers-1.0.8774/work/NVIDIA-Linux-x86-1.0-8774-pkg1/usr/src/nv/nv-i2c.c:27: error: unknown field âslave_recvâ specified in initializer
make[3]: *** [/var/tmp/portage/nvidia-drivers-1.0.8774/work/NVIDIA-Linux-x86-1.0-8774-pkg1/usr/src/nv/nv-i2c.o] Error 1
make[2]: *** [_module_/var/tmp/portage/nvidia-drivers-1.0.8774/work/NVIDIA-Linux-x86-1.0-8774-pkg1/usr/src/nv] Error 2
NVIDIA: left KBUILD.
nvidia.ko failed to build!
make[1]: *** [mdl] Error 1
make: *** [module] Error 2

these are nvidia-errors

tranquilcool · Post by **tranquilcool** » Sat Sep 23, 2006 5:32 pm

this seems to solve the nvidia-drivers compile problem

remove lines 26 and 27 from

/var/tmp/portage/nvidia-drivers-1.0.8774/work/NVIDIA-Linux-x86-1.0-8774-pkg1/usr/src/nv/nv-i2c.c

wrc1944 · Post by **wrc1944** » Sat Sep 23, 2006 5:40 pm

When I compiled, I looked at the values for each setting, and decided low-latency desktop would suit me best, but not knowing much about what the values really mean, maybe that's incorrect.

I recall back in my old win98 days, and even before the axcel216 registry tweak came out, most tweaking sites recommended setting windows to "network server" for best performance, even for desktop machines. Seemed to help some, as I recall, but I was REALLY ignorant in those days, as opposed to just "somewhat ignorant" now, so maybe it didn't really do much, if anything.

Maybe I'll try some of the other settings in the future. Am I correct in thinking you just run "emission-tunables whatever" as root, and the new settings are automatically adopted? Does the /usr/src/linux symlink necessarily need to be pointing to the running kernel for this to work? I'm assuming no reboot is needed- correct?

tranquilcool · Post by **tranquilcool** » Sat Sep 23, 2006 5:42 pm

can't insert module bt878. so no /dev/video0
kernel simply hangs

any help?

richard_ablitt · Post by **richard_ablitt** » Sat Sep 23, 2006 5:56 pm

The Beta nvidia drivers seem to work (using the installer, I've not tried the ebuild).

ejs76 · Post by **ejs76** » Sat Sep 23, 2006 6:40 pm

The Beta nvidia drivers seem to work (using the installer, I've not tried the ebuild).

Nvidia beta drivers, installed with ebuild, work just fine here...

Kollin · Post by **Kollin** » Sat Sep 23, 2006 8:17 pm

getting the same here...
CC [M] drivers/usb/storage/usb.o
In file included from drivers/usb/storage/usb.c:179:
drivers/usb/storage/unusual_devs.h:221: error: 'US_PR_KARMA' undeclared here (not in a function)
drivers/usb/storage/unusual_devs.h:221: error: 'rio_karma_init' undeclared here (not in a function)
make[3]: *** [drivers/usb/storage/usb.o] Error 1
make[2]: *** [drivers/usb/storage] Error 2
make[1]: *** [drivers/usb] Error 2
make: *** [drivers] Error 2

Same with the new patch

vipernicus · Post by **vipernicus** » Sat Sep 23, 2006 10:20 pm

Kollin wrote:getting the same here...
CC [M] drivers/usb/storage/usb.o
In file included from drivers/usb/storage/usb.c:179:
drivers/usb/storage/unusual_devs.h:221: error: 'US_PR_KARMA' undeclared here (not in a function)
drivers/usb/storage/unusual_devs.h:221: error: 'rio_karma_init' undeclared here (not in a function)
make[3]: *** [drivers/usb/storage/usb.o] Error 1
make[2]: *** [drivers/usb/storage] Error 2
make[1]: *** [drivers/usb] Error 2
make: *** [drivers] Error 2

Same with the new patch

Did you remove the old one from /usr/portage/distfiles ?

Kollin · Post by **Kollin** » Sat Sep 23, 2006 10:32 pm

vipernicus wrote:
Kollin wrote:getting the same here...
CC [M] drivers/usb/storage/usb.o
In file included from drivers/usb/storage/usb.c:179:
drivers/usb/storage/unusual_devs.h:221: error: 'US_PR_KARMA' undeclared here (not in a function)
drivers/usb/storage/unusual_devs.h:221: error: 'rio_karma_init' undeclared here (not in a function)
make[3]: *** [drivers/usb/storage/usb.o] Error 1
make[2]: *** [drivers/usb/storage] Error 2
make[1]: *** [drivers/usb] Error 2
make: *** [drivers] Error 2

Same with the new patch
Did you remove the old one from /usr/portage/distfiles ?

Yes i did it!

I solved this by enabling something in usb storage called "karma bla bla"

CONFIG_USB_STORAGE_KARMA=y

EDIT: Compillation successful - next step is booting

seren · Post by **seren** » Sun Sep 24, 2006 3:00 am

Awsome patchset works like a charm. one question is how can i see if compressed caching is working? although the kernel is really snappy.

thanks

EAD · Post by **EAD** » Sun Sep 24, 2006 6:59 am

When will this great patch will make it way to the Emission patch?

Kirktis · Post by **Kirktis** » Sun Sep 24, 2006 7:04 am

Rebuilt, seems to be working fine. Will run some intense apps overnight to make sure nothing breaks on me

vipernicus · Post by **vipernicus** » Sun Sep 24, 2006 10:36 am

EAD wrote:When will this great patch will make it way to the Emission patch?

E-Mission won't be using compcache, and it will come out whenever we get a bug worked out of a cvs build of grsecurity.