Official thread: "zen-sources"

Message

RobertDavid · Post by **RobertDavid** » Wed Jan 02, 2008 6:40 pm

Waninkoko wrote:That's pretty strange... You said that you tried to remove the tree and do a "git clone"?

Yes,
for example if I use git radeonhd drivers it works ok, can do anything like git checkout, pull etc, w/o downloading all the sources.

But zen-sources never. I just pulled only few times but it is anoying, (for me and for your server).. So I will try to clean it once more an git clone...

EDIT: it looks like this:

Code: Select all

[root@robert-server zen-sources]# git checkout
error: packfile .git/objects/pack/pack-152ef39e31380ece87b0067ef90657b7ad91059f.pack does not match index
error: packfile .git/objects/pack/pack-152ef39e31380ece87b0067ef90657b7ad91059f.pack cannot be accessed
error: packfile .git/objects/pack/pack-152ef39e31380ece87b0067ef90657b7ad91059f.pack does not match index
error: packfile .git/objects/pack/pack-152ef39e31380ece87b0067ef90657b7ad91059f.pack cannot be accessed
fatal: failed to unpack tree object d0c2e54a01a68a46d71443f0e881e527f96e30a5

gigs94 · Post by **gigs94** » Wed Jan 02, 2008 6:49 pm

bigbob73 wrote:Getting this error on boot...

NOHZ local_softirq_pending 02

then it freezes. Using the same config from 2.6.23-kamakaze. Looks to be a usb issue, but I can't nail it down.

I just changed my system to preempt and got this error. When I changed to using threaded soft/hard irqs, the system comes up and it appears fine.

Strowi · Post by **Strowi** » Wed Jan 02, 2008 7:58 pm

hi,

i am using "zen-sources-2.6.24_rc6-r0" on my ibm-x31 notebook. Works pretty good, except i got an error with the cpu-governors, no matter which one i try, it always falls back to performance-governor because of
[code]{ondemand,conservative} governor failed, too long transition latency of HW, fallback to performance governor[/code]

It worked before, with the kamikaze- or gentoo-sources, any ideas what might be the problem? I can provide dmesg and kernel-config if needed...

kernelOfTruth · Post by **kernelOfTruth** » Wed Jan 02, 2008 8:17 pm

thanks to Veldrin, eatnumber1, deno

disabling fair group scheduler & setting Hz to 100 Hz did the trick - no weird behavior so far

broch · Post by **broch** » Wed Jan 02, 2008 9:04 pm

kernelOfTruth wrote:it must be dddddddddependent on the scheduler (just compiling some stuff niced to +15 / 15

and you see the loads of d's I justtttttttt wrote (and the t's)

this happens with core2 duos (laptop & computer) and centriiiiiiiiiiiiiiiiiiiiiiiiiino (dothaaaaaaaaaaaaaan) cpu's

-> x86 and amd64 / 64 bit kernels, it's also independent if I usee ati-drivers or nvidia-drivers, preemption & soft-irq, hard-irq, whatever - is enabled ...

this all began wwwwwwwwwwwith usage of the cfs scheduler ...

... I'm hoping that ingo knows a fix foooooor that / improves the scheduler that ttttttttthis will stop in the near future ...

I have
HP dv5000 dual core T2500 1GB RAM 5400RPM SATA
enabled
CONFIG_ZEN_LL_DESKTOP=y
CONFIG_FAIR_GROUP_SCHED=y
CONFIG_FAIR_USER_SCHED=y
CONFIG_HZ_864=y
CONFIG_DEFAULT_AS=y

however, with cfs if I am trying to play CD in AmaroK my touchpad is not usable (super slow). Fix?
turning on permanently
IO_support = 3 (32-bit w/sync)
unmaskirq = 1 (on)
for cd/dvd recorder and now playing CD in AmaroK does not affect touchpad.

this simply points out to the fact that though CFS may not be at fault it may expose some issues with system configuration that did not affect you before.

Of course CFS is still buggy (and in fact it is), but the resolution related to your problem may not be in cfs.

kernelOfTruth · Post by **kernelOfTruth** » Wed Jan 02, 2008 10:35 pm

@broch:

it in fact was cfs

try to disable fair group scheduling & setting Hz to 100 Hz, that might increase you battery runtime & responsiveness, also try to use softirq & hardirq stuff (realtime linux features)

a newer firmware for you drive might help, or maybe there's a faulty cable ?

are you using pure libata ? (no old atapi / ide drivers)

Mr. Tao · Post by **Mr. Tao** » Wed Jan 02, 2008 10:38 pm

I can’t set any of Security options to be compiled as modules like I used to do with former kenel configs. As a result, realtime doesn’t work as it has to be compiled as module.

Post by **asturm** » Wed Jan 02, 2008 11:43 pm

kernelOfTruth wrote:also try to use softirq & hardirq stuff (realtime linux features)

Anyone using it with irqbalance (great tool, btw) or are those conflicting?

kernelOfTruth · Post by **kernelOfTruth** » Thu Jan 03, 2008 12:34 am

genstorm wrote:
kernelOfTruth wrote:also try to use softirq & hardirq stuff (realtime linux features)
Anyone using it with irqbalance (great tool, btw) or are those conflicting?

isn't irqbalance useful not until you have more than 2 cores (>=4 / >= 3 cores) ?

update:

IRQ load balancing isn't worthwhile until you have more than one socket, or more than two CPU cores.

ah, ok, so it might be useful, I'll give it a try

update2:

I see it's obligatory for good interactivity

If you are running a dual-core system (SMP), you *need* to install sys-apps/irqbalance, and set it to run on startup via its initscript - it makes a world of difference in system responsiveness.

see

*raises hand* just having started using it, I'm curious how responsibility will change

bigbob73 · Post by **bigbob73** » Thu Jan 03, 2008 11:46 am

gigs94 wrote:
bigbob73 wrote:Getting this error on boot...

NOHZ local_softirq_pending 02

then it freezes. Using the same config from 2.6.23-kamakaze. Looks to be a usb issue, but I can't nail it down.
I just changed my system to preempt and got this error. When I changed to using threaded soft/hard irqs, the system comes up and it appears fine.

Thanks, I'll give that one a try.

Waninkoko · Post by **Waninkoko** » Thu Jan 03, 2008 12:57 pm

Mr. Tao wrote:I can’t set any of Security options to be compiled as modules like I used to do with former kenel configs. As a result, realtime doesn’t work as it has to be compiled as module.

That's normal. This changed in .24 kernels.

gimpel · Post by **gimpel** » Thu Jan 03, 2008 3:38 pm

Waninkoko wrote:
Mr. Tao wrote:I can’t set any of Security options to be compiled as modules like I used to do with former kenel configs. As a result, realtime doesn’t work as it has to be compiled as module.
That's normal. This changed in .24 kernels.

Therefore use PAM RLIMITS
http://www.gentoo.org/proj/en/desktop/s ... altime.xml -> pam
http://proaudio.tuxfamily.org/wiki/inde ... el#RLIMITS

vacula · Post by **vacula** » Thu Jan 03, 2008 10:04 pm

Hello!

Confirmation - I have the same problem as kernelOfTruth with CFS.

I could not load my kde on Core2Duo laptop untill setting CONFIG_FAIR_GROUP_SCHED=n.

Also I have following in dmesg, however wifi works fine for me:

Code: Select all

Clocksource tsc unstable (delta = -760850190 ns)
wlan0: Initial auth_alg=0
wlan0: authenticate with AP 00:17:9a:21:f3:1e
wlan0: RX authentication from 00:17:9a:21:f3:1e (alg=0 transaction=2 status=0)
wlan0: authenticated
wlan0: associate with AP 00:17:9a:21:f3:1e
wlan0: RX AssocResp from 00:17:9a:21:f3:1e (capab=0x411 status=0 aid=1)
wlan0: associated
wlan0: WMM queue=2 aci=0 acm=0 aifs=3 cWmin=15 cWmax=1023 burst=0
wlan0: WMM queue=3 aci=1 acm=0 aifs=7 cWmin=15 cWmax=1023 burst=0
wlan0: WMM queue=1 aci=2 acm=0 aifs=2 cWmin=7 cWmax=15 burst=30
wlan0: WMM queue=0 aci=3 acm=0 aifs=2 cWmin=3 cWmax=7 burst=15
WARNING: at net/mac80211/rx.c:1672 __ieee80211_rx()
Pid: 6358, comm: X Not tainted 2.6.24-rc6-zen0-g39471294 #18

Call Trace:
 <IRQ>  [<ffffffff880b42ab>] :iwl4965:iwl4965_handle_data_packet+0x272/0x2a6
 [<ffffffff880973d6>] :mac80211:__ieee80211_rx+0x3f6/0xbb6
 [<ffffffff805abd1e>] scsi_end_request+0xbb/0xc9
 [<ffffffff880aa053>] :iwl4965:iwl4965_rx_queue_restock+0xb3/0x12e
 [<ffffffff8808a075>] :mac80211:ieee80211_tasklet_handler+0x5b/0xdc
 [<ffffffff8023c422>] tasklet_action+0x54/0x97
 [<ffffffff8023be27>] ___do_softirq+0x60/0x142
 [<ffffffff880b389b>] :iwl4965:iwl4965_isr+0x79/0xaa
 [<ffffffff8023c52e>] __do_softirq+0x32/0x45
 [<ffffffff8020d1ac>] call_softirq+0x1c/0x28
 [<ffffffff8020ee4c>] do_softirq+0x2c/0x7d
 [<ffffffff8023c2e2>] irq_exit+0x3f/0x82
 [<ffffffff8020f0ac>] do_IRQ+0xb9/0xd9
 [<ffffffff8020c551>] ret_from_intr+0x0/0xa
 <EOI>

RobertDavid · Post by **RobertDavid** » Thu Jan 03, 2008 11:03 pm

RobertDavid wrote:
Waninkoko wrote:That's pretty strange... You said that you tried to remove the tree and do a "git clone"?
Yes,
for example if I use git radeonhd drivers it works ok, can do anything like git checkout, pull etc, w/o downloading all the sources.

But zen-sources never. I just pulled only few times but it is anoying, (for me and for your server).. So I will try to clean it once more an git clone...

EDIT: it looks like this:
Code: Select all
[root@robert-server zen-sources]# git checkout
error: packfile .git/objects/pack/pack-152ef39e31380ece87b0067ef90657b7ad91059f.pack does not match index
error: packfile .git/objects/pack/pack-152ef39e31380ece87b0067ef90657b7ad91059f.pack cannot be accessed
error: packfile .git/objects/pack/pack-152ef39e31380ece87b0067ef90657b7ad91059f.pack does not match index
error: packfile .git/objects/pack/pack-152ef39e31380ece87b0067ef90657b7ad91059f.pack cannot be accessed
fatal: failed to unpack tree object d0c2e54a01a68a46d71443f0e881e527f96e30a5

Maybe I have found the bug:

I was using /usr/src as experimental ext4dev partition with mballoc and delalloc

After cleand git-clone every think works ok, but after remount apeard the problem with unpacking (every time). So I have disabled delalloc and mballoc, git-clone, try, remount, etc and now it works ok also after remount or reboot..

broch · Post by **broch** » Fri Jan 04, 2008 1:23 am

vacula wrote:Hello!

Confirmation - I have the same problem as kernelOfTruth with CFS.

I could not load my kde on Core2Duo laptop untill setting CONFIG_FAIR_GROUP_SCHED=n.

Also I have following in dmesg, however wifi works fine for me:

Code: Select all

Clocksource tsc unstable (delta = -760850190 ns)
wlan0: Initial auth_alg=0
wlan0: authenticate with AP 00:17:9a:21:f3:1e
wlan0: RX authentication from 00:17:9a:21:f3:1e (alg=0 transaction=2 status=0)
wlan0: authenticated
wlan0: associate with AP 00:17:9a:21:f3:1e
wlan0: RX AssocResp from 00:17:9a:21:f3:1e (capab=0x411 status=0 aid=1)
wlan0: associated
wlan0: WMM queue=2 aci=0 acm=0 aifs=3 cWmin=15 cWmax=1023 burst=0
wlan0: WMM queue=3 aci=1 acm=0 aifs=7 cWmin=15 cWmax=1023 burst=0
wlan0: WMM queue=1 aci=2 acm=0 aifs=2 cWmin=7 cWmax=15 burst=30
wlan0: WMM queue=0 aci=3 acm=0 aifs=2 cWmin=3 cWmax=7 burst=15
WARNING: at net/mac80211/rx.c:1672 __ieee80211_rx()
Pid: 6358, comm: X Not tainted 2.6.24-rc6-zen0-g39471294 #18

Call Trace:
 <IRQ>  [<ffffffff880b42ab>] :iwl4965:iwl4965_handle_data_packet+0x272/0x2a6
 [<ffffffff880973d6>] :mac80211:__ieee80211_rx+0x3f6/0xbb6
 [<ffffffff805abd1e>] scsi_end_request+0xbb/0xc9
 [<ffffffff880aa053>] :iwl4965:iwl4965_rx_queue_restock+0xb3/0x12e
 [<ffffffff8808a075>] :mac80211:ieee80211_tasklet_handler+0x5b/0xdc
 [<ffffffff8023c422>] tasklet_action+0x54/0x97
 [<ffffffff8023be27>] ___do_softirq+0x60/0x142
 [<ffffffff880b389b>] :iwl4965:iwl4965_isr+0x79/0xaa
 [<ffffffff8023c52e>] __do_softirq+0x32/0x45
 [<ffffffff8020d1ac>] call_softirq+0x1c/0x28
 [<ffffffff8020ee4c>] do_softirq+0x2c/0x7d
 [<ffffffff8023c2e2>] irq_exit+0x3f/0x82
 [<ffffffff8020f0ac>] do_IRQ+0xb9/0xd9
 [<ffffffff8020c551>] ret_from_intr+0x0/0xa
 <EOI>

out of curiosity: do you have set
CONFIG_NO_HZ=y

vacula · Post by **vacula** » Fri Jan 04, 2008 8:33 am

broch wrote:
vacula wrote:Hello!

Confirmation - I have the same problem as kernelOfTruth with CFS.

I could not load my kde on Core2Duo laptop untill setting CONFIG_FAIR_GROUP_SCHED=n.
...
[/code]
out of curiosity: do you have set
CONFIG_NO_HZ=y

yes:

# grep _HZ .config
CONFIG_NO_HZ=y
# CONFIG_HZ_100 is not set
# CONFIG_HZ_216 is not set
# CONFIG_HZ_250_NODEFAULT is not set
CONFIG_HZ_300=y
# CONFIG_HZ_432 is not set
# CONFIG_HZ_864 is not set
# CONFIG_HZ_1000 is not set
# CONFIG_HZ_1500 is not set
# CONFIG_HZ_2000 is not set
# CONFIG_HZ_3000 is not set
# CONFIG_HZ_4000 is not set
# CONFIG_HZ_5000 is not set
# CONFIG_HZ_7500 is not set
# CONFIG_HZ_10000 is not set
CONFIG_HZ=300

broch · Post by **broch** » Fri Jan 04, 2008 2:00 pm

well on lklm was a discussion about NO_HZ breaking a lot of things. Not to mention that savings on battery aren't big. Devs seemed to be cautious about current NO_HZ.

Turn NO_HZ off try cfs again, hopefully this will help

Waninkoko · Post by **Waninkoko** » Fri Jan 04, 2008 2:37 pm

RobertDavid wrote:
RobertDavid wrote:
Waninkoko wrote:That's pretty strange... You said that you tried to remove the tree and do a "git clone"?
Yes,
for example if I use git radeonhd drivers it works ok, can do anything like git checkout, pull etc, w/o downloading all the sources.

But zen-sources never. I just pulled only few times but it is anoying, (for me and for your server).. So I will try to clean it once more an git clone...

EDIT: it looks like this:
Code: Select all
[root@robert-server zen-sources]# git checkout
error: packfile .git/objects/pack/pack-152ef39e31380ece87b0067ef90657b7ad91059f.pack does not match index
error: packfile .git/objects/pack/pack-152ef39e31380ece87b0067ef90657b7ad91059f.pack cannot be accessed
error: packfile .git/objects/pack/pack-152ef39e31380ece87b0067ef90657b7ad91059f.pack does not match index
error: packfile .git/objects/pack/pack-152ef39e31380ece87b0067ef90657b7ad91059f.pack cannot be accessed
fatal: failed to unpack tree object d0c2e54a01a68a46d71443f0e881e527f96e30a5
Maybe I have found the bug:

I was using /usr/src as experimental ext4dev partition with mballoc and delalloc

After cleand git-clone every think works ok, but after remount apeard the problem with unpacking (every time). So I have disabled delalloc and mballoc, git-clone, try, remount, etc and now it works ok also after remount or reboot..

delalloc normally works fine but mballoc is VEEEEEEERY experimental.

RobertDavid · Post by **RobertDavid** » Fri Jan 04, 2008 7:08 pm

Waninkoko wrote: delalloc normally works fine but mballoc is VEEEEEEERY experimental.

thanks for the info, I will enable only delalloc

ilikenwf · Post by **ilikenwf** » Sat Jan 05, 2008 2:48 am

Another error with madwifi:

Code: Select all

CC [M]  drivers/net/wireless/libertas/if_usb.o
CC [M]  drivers/net/wireless/madwifi/ath/if_ath_pci.o In file included from drivers/net/wireless/madwifi/ath/if_ath_pci.c:341: /usr/src/zen-sources/drivers/net/wireless/madwifi/ath/../release.h:38:24: error: svnversion.h: No such file or directory   
LD [M]  drivers/net/wireless/libertas/libertas.o make[6]: *** [drivers/net/wireless/madwifi/ath/if_ath_pci.o] 
Error 1 make[5]: *** [drivers/net/wireless/madwifi/ath] Error 2

hirakendu · Post by **hirakendu** » Sat Jan 05, 2008 3:09 am

Was about to mention that yesterday. Just copy that file svnversion.h from an older snapshot or madwifi's subversion repo to /usr/src/zen-sources/drivers/net/wireless/madwifi/svnversion.h .

ilikenwf · Post by **ilikenwf** » Sat Jan 05, 2008 5:30 am

hirakendu wrote:Was about to mention that yesterday. Just copy that file svnversion.h from an older snapshot or madwifi's subversion repo to /usr/src/zen-sources/drivers/net/wireless/madwifi/svnversion.h .

I can't find a copy with any code in it...

hirakendu · Post by **hirakendu** » Sat Jan 05, 2008 6:43 am

Hmm. It anyway just contains one line

. So create the file /usr/src/zen-sources/drivers/net/wireless/madwifi/svnversion.h manually using your favorite editor with the content

#define SVNVERSION "svn r3073"

(may replace it with current number, but doesn't matter, just a version).

Waninkoko · Post by **Waninkoko** » Sat Jan 05, 2008 12:31 pm

Fixed

zxy · Post by **zxy** » Sat Jan 05, 2008 3:37 pm

Wanikoko, now you can add one more post and you'll have 1000th.

Official thread: "zen-sources"

Security options as modules

Re: Security options as modules

Re: Security options as modules

unresponsive system

Re: unresponsive system

Re: unresponsive system