Official thread: "zen-sources" - Part 8

Message

cyberdemon · Post by **cyberdemon** » Sat Oct 10, 2009 9:37 am

What a pool!!
I forgot that VMware is using Fusion MPT driver..
I got the right module name by running lspci -v for Fusion MPT,
I'm recompiling a kernel now.

Thanks for your help, KernelOfTruth!!

cyberdemon · Post by **cyberdemon** » Sat Oct 10, 2009 9:50 am

Compiling done, running so flawlessly..!
Thanks again! What a pool I'm...

mantoo · Post by **mantoo** » Sat Oct 10, 2009 10:54 am

Saundersx wrote:One thing I have noticed since upgrading to 2.6.31-zen2

hedgehog / $ modprobe it87
FATAL: Error inserting it87 (/lib/modules/2.6.31-zen2/kernel/drivers/hwmon/it87.ko): Device or resource busy

Anyone else having this?

ok, back again

1st workaround:

Code: Select all

pnpacpi=off

as kernelparameter, which i found here

2nd, and maybe better workaround:

Code: Select all

acpi_enforce_resources=no

as kernel parameter, as suggested here

even switching to atk0110 as sensordriver was suggested. dont know which one of the approaches fits best, as i couldn´t test´em.

~~~~~~~~~
edit:
and some additional related stuff talking abount lm_sensors and patches/ebuilds here

~~~~~~~~~
my mobo is a asus p5v2d-mx btw

~~~~~~~~~
edit2:
and

Code: Select all

acpi_enforce_resources=lax

should work too

~~~~~~~~~
3rd and last edit:

one more time i should´ve trust a bit more in this powerful forum, as the problem is handled in that thread...

Post by **asturm** » Sat Oct 10, 2009 11:57 am

cheater1034 wrote:should be fixed now anyway

please confirm (im waiting on someone else's confirmation too)

confirmed, works again!

cheater1034 · Post by **cheater1034** » Sat Oct 10, 2009 3:55 pm

Those with 2.6.32 bfs problems, try now please

Saundersx · Post by **Saundersx** » Sat Oct 10, 2009 9:19 pm

mantoo wrote:
Saundersx wrote:One thing I have noticed since upgrading to 2.6.31-zen2

hedgehog / $ modprobe it87
FATAL: Error inserting it87 (/lib/modules/2.6.31-zen2/kernel/drivers/hwmon/it87.ko): Device or resource busy

Anyone else having this?
ok, back again

1st workaround:
Code: Select all
pnpacpi=off
as kernelparameter, which i found here

2nd, and maybe better workaround:
Code: Select all
acpi_enforce_resources=no
as kernel parameter, as suggested here

even switching to atk0110 as sensordriver was suggested. dont know which one of the approaches fits best, as i couldn´t test´em.

~~~~~~~~~
edit:
and some additional related stuff talking abount lm_sensors and patches/ebuilds here

~~~~~~~~~
my mobo is a asus p5v2d-mx btw

~~~~~~~~~
edit2:
and
Code: Select all
acpi_enforce_resources=lax
should work too

~~~~~~~~~
3rd and last edit:

one more time i should´ve trust a bit more in this powerful forum, as the problem is handled in that thread...

I did some reading as well. Now using "acpi_enforce_resources=lax" and it seems to work fine.

cheater1034 · Post by **cheater1034** » Sun Oct 11, 2009 5:10 am

Hey guys, bunch of great new things

1. BFQ is the default i/o scheduler in ALL the trees (2.6.30/2.6.31, unstable)
- Test it, fabio is interested in both bug reports and any benchmarks/results!!!!! so give them to me and i can forward them
2. new 2.6.31 patch released in the past few days (new one will follow shortly due to alsa and drm fixes in git)
3. new 2.6.30 patch released (includes bfs 303 and 2.6.30.9)
4. new 2.6.32-rc3-zen1 patch released (YES! it's very solid and builds with the allmodconfig A OK, reiser4 is broken however)

BFS/BFQ make zen more geared for desktop usage than ever before!

Plus, now is the time to try zen if you haven't, it's growing in popularity - a few devs want their patches in to get tested and forwarded results/bug reports - so do it! I know people use it on several distributions now (varying from momongo to sourcemage to arch to debian to fedora to ubuntu to sidux to gentoo to exherbo to yoper) - quite impressive!

cheater1034 · Post by **cheater1034** » Sun Oct 11, 2009 5:47 am

Hey guys, PLEASE vouch for zen on bugs.gentoo.org for inclusion into portage:
http://bugs.gentoo.org/show_bug.cgi?id=288512

I'm sure the more interest it shows the more likely it will get into gentoo (which many people would want, because they would get ebuilds again)

tranquilcool · Post by **tranquilcool** » Sun Oct 11, 2009 9:25 am

LD .tmp_vmlinux1
kernel/built-in.o: In function `__mutex_lock_interruptible_slowpath':
mutex.c:(.sched.text+0x14db): undefined reference to `__schedule'
kernel/built-in.o: In function `__mutex_lock_slowpath':
mutex.c:(.sched.text+0x16d6): undefined reference to `__schedule'
kernel/built-in.o: In function `__mutex_lock_killable_slowpath':
mutex.c:(.sched.text+0x1842): undefined reference to `__schedule'
make: *** [.tmp_vmlinux1] Error 1

latest zen 2.6.30-zen6 with bfs/bfq checkout errors.
help please.

rahulthewall · Post by **rahulthewall** » Sun Oct 11, 2009 9:29 am

cheater1034 wrote:Hey guys, PLEASE vouch for zen on bugs.gentoo.org for inclusion into portage:
http://bugs.gentoo.org/show_bug.cgi?id=288512

I'm sure the more interest it shows the more likely it will get into gentoo (which many people would want, because they would get ebuilds again)

In the bug you mention that you are looking at including TuxOnIce back in to zen-sources. Is that true?

Ant P. · Post by **Ant P.** » Sun Oct 11, 2009 2:30 pm

Will compcache ever work with BFS?

If not I might just go out and buy two extra RAM sticks instead

ponciarello · Post by **ponciarello** » Sun Oct 11, 2009 3:06 pm

I think it works

http://code.google.com/p/compcache/wiki ... ndUsingNew

you need to download and compile code to get rzscontrol binary for activating swap before mounting that. load module (by default it creates /dev/ramzsawp0) then

Code: Select all

rzscontrol /dev/ramzswap0 --init
swapon -p 1 /dev/ramzswap0

DigitalCorpus · Post by **DigitalCorpus** » Sun Oct 11, 2009 4:56 pm

cheater1034 wrote:Hey guys, bunch of great new things

1. BFQ is the default i/o scheduler in ALL the trees (2.6.30/2.6.31, unstable)
- Test it, fabio is interested in both bug reports and any benchmarks/results!!!!! so give them to me and i can forward them
2. new 2.6.31 patch released in the past few days (new one will follow shortly due to alsa and drm fixes in git)
3. new 2.6.30 patch released (includes bfs 303 and 2.6.30.9)
4. new 2.6.32-rc3-zen1 patch released (YES! it's very solid and builds with the allmodconfig A OK, reiser4 is broken however)

BFS/BFQ make zen more geared for desktop usage than ever before!

Plus, now is the time to try zen if you haven't, it's growing in popularity - a few devs want their patches in to get tested and forwarded results/bug reports - so do it! I know people use it on several distributions now (varying from momongo to sourcemage to arch to debian to fedora to ubuntu to sidux to gentoo to exherbo to yoper) - quite impressive!

Ahhh man! Reiser4 is broken? GRRRR! I have the Btrfs guys telling me to bump to 2.6.32 and that zen-sources has the latest btrfs pulled from git. I'm looking to test how the little panic once disk is 90% full works out, but my root is on R4

. cheater1034, could you please post when this gets fixed? I'm extremely anxious....

Okay so I just updated zen-stable and I see

Code: Select all

 * [new branch]      btrfs      -> origin/btrfs

What version or git pull is this of btrfs?

kernelOfTruth · Post by **kernelOfTruth** » Sun Oct 11, 2009 5:06 pm

DigitalCorpus wrote:
cheater1034 wrote:Hey guys, bunch of great new things

1. BFQ is the default i/o scheduler in ALL the trees (2.6.30/2.6.31, unstable)
- Test it, fabio is interested in both bug reports and any benchmarks/results!!!!! so give them to me and i can forward them
2. new 2.6.31 patch released in the past few days (new one will follow shortly due to alsa and drm fixes in git)
3. new 2.6.30 patch released (includes bfs 303 and 2.6.30.9)
4. new 2.6.32-rc3-zen1 patch released (YES! it's very solid and builds with the allmodconfig A OK, reiser4 is broken however)

BFS/BFQ make zen more geared for desktop usage than ever before!

Plus, now is the time to try zen if you haven't, it's growing in popularity - a few devs want their patches in to get tested and forwarded results/bug reports - so do it! I know people use it on several distributions now (varying from momongo to sourcemage to arch to debian to fedora to ubuntu to sidux to gentoo to exherbo to yoper) - quite impressive!
Ahhh man! Reiser4 is broken? GRRRR! I have the Btrfs guys telling me to bump to 2.6.32 and that zen-sources has the latest btrfs pulled from git. I'm looking to test how the little panic once disk is 90% full works out, but my root is on R4 . cheater1034, could you please post when this gets fixed? I'm extremely anxious....

Okay so I just updated zen-stable and I see
Code: Select all
 * [new branch]      btrfs      -> origin/btrfs
What version or git pull is this of btrfs?

DigitalCorpus try to revert reiser4 from mmotm and replace it with the one from 2.6.31 (it worked at least with 2.6.31-zen2 for me - I manually removed it

)

nero37 · Post by **nero37** » Sun Oct 11, 2009 5:25 pm

tranquilcool wrote:LD .tmp_vmlinux1
kernel/built-in.o: In function `__mutex_lock_interruptible_slowpath':
mutex.c:(.sched.text+0x14db): undefined reference to `__schedule'
kernel/built-in.o: In function `__mutex_lock_slowpath':
mutex.c:(.sched.text+0x16d6): undefined reference to `__schedule'
kernel/built-in.o: In function `__mutex_lock_killable_slowpath':
mutex.c:(.sched.text+0x1842): undefined reference to `__schedule'
make: *** [.tmp_vmlinux1] Error 1

latest zen 2.6.30-zen6 with bfs/bfq checkout errors.

I'm seeing this too.

Jupiter1TX · Post by **Jupiter1TX** » Sun Oct 11, 2009 5:29 pm

cheater1034 wrote:Hey guys, PLEASE vouch for zen on bugs.gentoo.org for inclusion into portage:
http://bugs.gentoo.org/show_bug.cgi?id=288512

I'm sure the more interest it shows the more likely it will get into gentoo (which many people would want, because they would get ebuilds again)

+1

Latest PTS test results 2.6.32-rc3-zen2+BFS/BFQ
http://global.phoronix-test-suite.com/i ... 3788-24286

Ant_P wrote:Will compcache ever work with BFS?

If not I might just go out and buy two extra RAM sticks instead

Am using a 2Gb ramdisk for portage temp and ccache and i must say it makes
a HUGE difference as seen in this before and after.

Code: Select all

vger ~ # genlop -t glibc
 * sys-libs/glibc

     Tue Sep 22 15:35:07 2009 >>> sys-libs/glibc-2.10.1
       merge time: 1 hour, 58 minutes and 26 seconds.

     Tue Sep 22 18:15:39 2009 >>> sys-libs/glibc-2.10.1
       merge time: 7 minutes and 39 seconds.

kernelOfTruth · Post by **kernelOfTruth** » Sun Oct 11, 2009 5:35 pm

Jupiter1TX wrote: Am using a 2Gb ramdisk for portage temp and ccache and i must say it makes
a HUGE difference as seen in this before and after.
Code: Select all
vger ~ # genlop -t glibc
 * sys-libs/glibc

     Tue Sep 22 15:35:07 2009 >>> sys-libs/glibc-2.10.1
       merge time: 1 hour, 58 minutes and 26 seconds.

     Tue Sep 22 18:15:39 2009 >>> sys-libs/glibc-2.10.1
       merge time: 7 minutes and 39 seconds.

you most be doing something wrong:

Mon May 18 20:16:19 2009 >>> sys-libs/glibc-2.10.1
merge time: 32 minutes and 50 seconds.

Mon Aug 10 12:52:50 2009 >>> sys-libs/glibc-2.10.1
merge time: 23 minutes and 8 seconds.

Mon Aug 10 14:10:42 2009 >>> sys-libs/glibc-2.10.1
merge time: 25 minutes and 20 seconds.

without ccache and tmpfs only on and E6600 (Core2Duo @stock clock, 2.4 GHz)

try to raise MAKEOPTS at least to -j5 that should make it even faster

Jupiter1TX · Post by **Jupiter1TX** » Sun Oct 11, 2009 6:07 pm

kernelOfTruth wrote:
Jupiter1TX wrote: Am using a 2Gb ramdisk for portage temp and ccache and i must say it makes
a HUGE difference as seen in this before and after.
Code: Select all
vger ~ # genlop -t glibc
 * sys-libs/glibc

     Tue Sep 22 15:35:07 2009 >>> sys-libs/glibc-2.10.1
       merge time: 1 hour, 58 minutes and 26 seconds.

     Tue Sep 22 18:15:39 2009 >>> sys-libs/glibc-2.10.1
       merge time: 7 minutes and 39 seconds.
you most be doing something wrong:

Mon May 18 20:16:19 2009 >>> sys-libs/glibc-2.10.1
merge time: 32 minutes and 50 seconds.

Mon Aug 10 12:52:50 2009 >>> sys-libs/glibc-2.10.1
merge time: 23 minutes and 8 seconds.

Mon Aug 10 14:10:42 2009 >>> sys-libs/glibc-2.10.1
merge time: 25 minutes and 20 seconds.
without ccache and tmpfs only on and E6600 (Core2Duo @stock clock, 2.4 GHz)

try to raise MAKEOPTS at least to -j5 that should make it even faster

Am using -j8. In the first emerge i was booted from the NEO cd.
I agree with you though it took to long. It would be nice to know
what i am doing wrong. If anyone can suggest better CFLAGS/LDFLAGS
settings for Core i7 920 please do. Here is what i am using now.

Code: Select all

CFLAGS="-O2 -march=native -msse4 -pipe"
LDFLAGS="-Wl,--hash-style=gnu -Wl,--sort-common -Wl,--as-needed"

DigitalCorpus · Post by **DigitalCorpus** » Sun Oct 11, 2009 6:14 pm

How much RAM do you have?
What filesystem is /usr/portage residing?
What is your root filesystem?
What I/O scheduler do you use?

-msse4 should be implied by -march=native
Your LDFLAGS and CFLAGS are sane.

Jupiter1TX · Post by **Jupiter1TX** » Sun Oct 11, 2009 6:20 pm

DigitalCorpus wrote:How much RAM do you have?
What filesystem is /usr/portage residing?
What is your root filesystem?
What I/O scheduler do you use?

-msse4 should be implied by -march=native
Your LDFLAGS and CFLAGS are sane.

6Gb of ram
/usr/portage and / are ext4
linux-2.6.32-zen2-BFS+BFQ

This pc is just for fun/performance. Thanks...

DigitalCorpus · Post by **DigitalCorpus** » Sun Oct 11, 2009 6:43 pm

what are the outputs of

Code: Select all

hdparm -t /dev/hd* or sd*

for each or your hard drives?

Jupiter1TX · Post by **Jupiter1TX** » Sun Oct 11, 2009 6:47 pm

DigitalCorpus wrote:what are the outputs of
Code: Select all
hdparm -t /dev/hd* or sd*
for each or your hard drives?

heh, yea not very fast

Code: Select all

vger ~ # hdparm -t /dev/sda

/dev/sda:
 Timing buffered disk reads:  226 MB in  3.02 seconds =  74.72 MB/sec

DigitalCorpus · Post by **DigitalCorpus** » Sun Oct 11, 2009 6:59 pm

Those are the basic for what will slow you down. Unless Ext4 suffers major degradation from fragmentation, I'm not sure why you have that much of a slow down. Try one more thing for me please:

Code: Select all

echo 1 > /proc/sys/vm/drop_caches
time emerge -pev @system @world

On a Reiser4 partition I get about 16 seconds

Jupiter1TX · Post by **Jupiter1TX** » Sun Oct 11, 2009 7:08 pm

DigitalCorpus wrote:Those are the basic for what will slow you down. Unless Ext4 suffers major degradation from fragmentation, I'm not sure why you have that much of a slow down. Try one more thing for me please:
Code: Select all
echo 1 > /proc/sys/vm/drop_caches
time emerge -pev @system @world
On a Reiser4 partition I get about 16 seconds

Total: 893 packages (7 upgrades, 2 downgrades, 884 reinstalls), Size of downloads: 1,155,829 kB
Portage tree and overlays:
[0] /usr/portage
[1] /usr/portage/local/layman/redspot
[2] /usr/portage/local/layman/desktop-effects

real 0m49.875s
user 0m11.446s
sys 0m0.224s

UPDATE:
After taking care of the '7 upgrades, 2 downgrades'

Code: Select all

Total: 893 packages (893 reinstalls), Size of downloads: 1,140,966 kB
Portage tree and overlays:
 [0] /usr/portage
 [1] /usr/portage/local/layman/redspot
 [2] /usr/portage/local/layman/desktop-effects

real    0m11.388s
user    0m11.128s
sys     0m0.127s

DigitalCorpus · Post by **DigitalCorpus** » Sun Oct 11, 2009 7:11 pm

I don't see anythign that would indicate that slow down you mentioned. You did say that time was when you booted to a CD instead of off your hard disk. If you compile glibc now, do you still get fairly lengthy times?