| View previous topic :: View next topic |
| Author |
Message |
drescherjm
Advocate
Joined: 05 Jun 2004
Posts: 2790
Location: Pittsburgh, PA, USA
|
 Posted: Mon Sep 06, 2010 10:39 pm Post subject: |
 |
|
| stuNNed2 wrote: | | if you don't mind me asking what frontend app were you using for the cpu frequency scaling? |
I have not used a frontend application in 3 to 4 years. Just set the kernel governor to ondemand and let the kernel handle everything.
_________________
John
My gentoo overlay
Instructons for overlay |
|
| Back to top |
|
 |
drescherjm
Advocate
Joined: 05 Jun 2004
Posts: 2790
Location: Pittsburgh, PA, USA
|
| Posted: Mon Sep 06, 2010 10:41 pm Post subject: |
|
|
| zx2c4 wrote: | At least partially... There's still some slowness, but overall it seems improved.
Can someone else try this and report back? |
That did not work for me. Also disabling speedstep and c1x in the bios did not work.
_________________
John
My gentoo overlay
Instructons for overlay |
|
| Back to top |
|
|
zx2c4
Developer
Joined: 09 Jun 2005
Posts: 177
|
| Posted: Tue Sep 07, 2010 1:02 am Post subject: |
|
|
| stuNNed2 wrote: | | if you don't mind me asking what frontend app were you using for the cpu frequency scaling? |
The problem started on boot, before any programs were loaded. Both for performance governer and userspace governer. |
|
| Back to top |
|
|
drescherjm
Advocate
Joined: 05 Jun 2004
Posts: 2790
Location: Pittsburgh, PA, USA
|
|
| Back to top |
|
|
zx2c4
Developer
Joined: 09 Jun 2005
Posts: 177
|
| Posted: Wed Sep 08, 2010 5:26 pm Post subject: |
|
|
Disabling cpufreq in fact did not fix the issue.  |
|
| Back to top |
|
|
optiluca
Guru
Joined: 16 Jan 2006
Posts: 551
Location: Rivergaro, Italy
|
| Posted: Wed Sep 08, 2010 6:38 pm Post subject: |
|
|
Hi all. I was suffering the random soft lockups at boot time, and I replaced my kernel with zen sources, which is working just fine. The only obvious changes I made from stock gentoo kernel was to enable tuxonice/BFS/BFQ/SQLB
In any case, find my .config here:
http://pastebin.com/qj8ZBpWb
Hope that works for someone 
_________________
# "Hmm, sounds like your system froze up."
# "I don't know why. It's about 80 degrees in here!"
http://www.rinkworks.com/stupid/cs_mincing.shtml |
|
| Back to top |
|
|
Naib
Watchman
Joined: 21 May 2004
Posts: 6069
Location: Removed by Neddy
|
| Posted: Wed Sep 08, 2010 7:16 pm Post subject: |
|
|
| optiluca wrote: | Hi all. I was suffering the random soft lockups at boot time, and I replaced my kernel with zen sources, which is working just fine. The only obvious changes I made from stock gentoo kernel was to enable tuxonice/BFS/BFQ/SQLB
In any case, find my .config here:
http://pastebin.com/qj8ZBpWb
Hope that works for someone |
I changed to the CK sources and it seems alot smoother. Every now and again there is a slight reduced responsiveness (this actually corresponds to a change in a core's freq - conky showing 4cores FTW)
_________________
#define HelloWorld int
#define Int main()
#define Return printf
#define Print return
#include <stdio>
HelloWorld Int {
Return("Hello, world!\n");
Print 0; |
|
| Back to top |
|
|
drescherjm
Advocate
Joined: 05 Jun 2004
Posts: 2790
Location: Pittsburgh, PA, USA
|
| Posted: Fri Sep 10, 2010 12:17 am Post subject: |
|
|
I believe with the help of Nix on the following thread I have sloved this:
http://lkml.org/lkml/2010/9/9/541
If I add the following kernel parameter I have no slowness in a ssh session (and no 200+ second boots): clocksource=acpi_pm
| Code: | [ 25.301748] r8169 0000:08:00.0: eth0: link up
[ 25.301758] r8169 0000:08:00.0: eth0: link up
[ 25.510863] Bridge firewalling registered
[ 25.518425] device eth0 entered promiscuous mode
[ 25.523103] device vbox0 entered promiscuous mode
[ 25.524055] br0: new device vbox0 does not support netpoll (disabling)
[ 25.526106] br0: port 2(vbox0) entering forwarding state
[ 25.526111] br0: port 2(vbox0) entering forwarding state
[ 25.526115] br0: port 1(eth0) entering forwarding state
[ 25.526119] br0: port 1(eth0) entering forwarding state
[ 30.554156] ip_tables: (C) 2000-2006 Netfilter Core Team
[ 30.646698] nf_conntrack version 0.5.0 (16384 buckets, 65536 max)
[ 30.646916] CONFIG_NF_CT_ACCT is deprecated and will be removed soon. Please use
[ 30.646919] nf_conntrack.acct=1 kernel parameter, acct=1 nf_conntrack module option or
[ 30.646922] sysctl net.netfilter.nf_conntrack_acct=1 to enable it.
[ 34.129901] NFSD: Using /var/lib/nfs/v4recovery as the NFSv4 state recovery directory
[ 34.130029] NFSD: starting 90-second grace period
[ 35.630049] nvidia: version magic '2.6.35-zen2-sleep SMP mod_unload ' should be '2.6.35-zen2-sleep SMP preempt mod_unload ' |
_________________
John
My gentoo overlay
Instructons for overlay |
|
| Back to top |
|
|
Genewb
Apprentice
Joined: 09 Jan 2007
Posts: 165
|
| Posted: Fri Sep 10, 2010 10:57 am Post subject: |
|
|
Replying to this thread as it seems to have surpassed the other:
I tried using apci_pm and the problem seemed to have been solved initially, but once X started it was apparent that it'd just been improved somewhat. Passing hpet=verbose to the kernel worked wonders though, problem solved.
_________________
I don't give a darn about "experience", just functional copyleft software. |
|
| Back to top |
|
|
drescherjm
Advocate
Joined: 05 Jun 2004
Posts: 2790
Location: Pittsburgh, PA, USA
|
| Posted: Fri Sep 10, 2010 12:59 pm Post subject: |
|
|
I am downgrading it from solved to improved to do more testing. Also I need to test in X windows. I will look into the Passing hpet=verbose and see if that changes things.. There now is some developer activity on the lkml about this so hopefully soon workarounds will not be needed..
_________________
John
My gentoo overlay
Instructons for overlay |
|
| Back to top |
|
|
zx2c4
Developer
Joined: 09 Jun 2005
Posts: 177
|
| Posted: Fri Sep 10, 2010 5:32 pm Post subject: |
|
|
| Code: | | cd /usr/src/linux && curl 'http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=patch;h=30a564be9d9554c168a654eddc2165869cc0d7bf' | patch -R -p1 |
|
|
| Back to top |
|
|
drescherjm
Advocate
Joined: 05 Jun 2004
Posts: 2790
Location: Pittsburgh, PA, USA
|
| Posted: Fri Sep 10, 2010 5:48 pm Post subject: |
|
|
So you are removing the commit that Nix bisected and believes was the cause. Keep us informed..
_________________
John
My gentoo overlay
Instructons for overlay |
|
| Back to top |
|
|
zx2c4
Developer
Joined: 09 Jun 2005
Posts: 177
|
| Posted: Fri Sep 10, 2010 6:38 pm Post subject: |
|
|
So far so good!
But someone else should try as well to confirm. |
|
| Back to top |
|
|
drescherjm
Advocate
Joined: 05 Jun 2004
Posts: 2790
Location: Pittsburgh, PA, USA
|
|
| Back to top |
|
|
zx2c4
Developer
Joined: 09 Jun 2005
Posts: 177
|
| Posted: Sat Sep 11, 2010 6:11 am Post subject: |
|
|
| Code: | zx2c4@ZX2C4-Laptop ~ $ cat /sys/devices/system/clocksource/clocksource0/current_clocksource
tsc
|
my dmesg has some things about hpet |
|
| Back to top |
|
|
Naib
Watchman
Joined: 21 May 2004
Posts: 6069
Location: Removed by Neddy
|
| Posted: Sat Sep 11, 2010 10:10 am Post subject: |
|
|
| Code: |
HPET: 8 timers in total, 5 timers will be used for per-cpu timer
hpet0: at MMIO 0xfed00000, IRQs 2, 8, 24, 25, 26, 27, 28, 0
hpet0: 8 comparators, 64-bit 14.318180 MHz counter
hpet: hpet2 irq 24 for MSI
hpet: hpet3 irq 25 for MSI
hpet: hpet4 irq 26 for MSI
hpet: hpet5 irq 27 for MSI
hpet: hpet6 irq 28 for MSI
Switching to clocksource tsc
|
from dmesg
yet later on:
CE: hpet increased min_delta_ns to 7500 nsec
_________________
#define HelloWorld int
#define Int main()
#define Return printf
#define Print return
#include <stdio>
HelloWorld Int {
Return("Hello, world!\n");
Print 0; |
|
| Back to top |
|
|
drescherjm
Advocate
Joined: 05 Jun 2004
Posts: 2790
Location: Pittsburgh, PA, USA
|
| Posted: Sat Sep 11, 2010 2:45 pm Post subject: |
|
|
This is what I have for acpi_pm clocksource.
| Code: | jmd1 linux # cat /sys/devices/system/clocksource/clocksource0/current_clocksource
acpi_pm
jmd1 linux # dmesg | grep hpet
[ 0.971338] hpet clockevent registered
[ 1.376607] rtc0: alarms up to one month, 242 bytes nvram, hpet irqs
[ 10.858552] CE: hpet increased min_delta_ns to 7500 nsec
[ 1006.469327] CE: hpet increased min_delta_ns to 11250 nsec
jmd1 linux #
|
Right now I am using 2.6.35-zen-r2 without reversing the commit (reversing did not apply cleanly on zen):
| Code: | jmd1 linux # uname -a
Linux jmd1 2.6.35-zen2-acpi_pm #8 ZEN SMP PREEMPT Thu Sep 9 20:47:54 EDT 2010 x86_64 Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz GenuineIntel GNU/Linux
jmd1 linux #
|
It still seems to be working almost as good as 2.6.34. I say almost because I have seen a little random seconds of slowdown via ssh. I have not used it in X long enough to see any bad behavior. Also KVM seems to be working fine also. I have had the most use in that.
_________________
John
My gentoo overlay
Instructons for overlay |
|
| Back to top |
|
|
drescherjm
Advocate
Joined: 05 Jun 2004
Posts: 2790
Location: Pittsburgh, PA, USA
|
| Posted: Tue Sep 14, 2010 10:25 pm Post subject: |
|
|
I just got the email from the kernel mailing list for the thread that describes this issue:
| Code: | FYI, we got a reasonable explanation now what's going on in hardware
land. Patch is in the works. |
_________________
John
My gentoo overlay
Instructons for overlay |
|
| Back to top |
|
|
zx2c4
Developer
Joined: 09 Jun 2005
Posts: 177
|
| Posted: Tue Sep 14, 2010 10:28 pm Post subject: |
|
|
| Hurray! |
|
| Back to top |
|
|
drescherjm
Advocate
Joined: 05 Jun 2004
Posts: 2790
Location: Pittsburgh, PA, USA
|
|
| Back to top |
|
|
Naib
Watchman
Joined: 21 May 2004
Posts: 6069
Location: Removed by Neddy
|
| Posted: Wed Sep 15, 2010 10:48 am Post subject: |
|
|
interesting.
by ati do they mean a crossfire mobo? cause i have intel cpu and nv card.
also they talk about u32 and not casting. my bios has an option for 64 or 32bit hpet. i have it set to 64b at the moment i might try as 32b as well as this patch.
also any word if this patch is making into 2.6.36?
might be worth a gentoo bug and ask this patch to go into the gentoo patchset
_________________
#define HelloWorld int
#define Int main()
#define Return printf
#define Print return
#include <stdio>
HelloWorld Int {
Return("Hello, world!\n");
Print 0; |
|
| Back to top |
|
|
drescherjm
Advocate
Joined: 05 Jun 2004
Posts: 2790
Location: Pittsburgh, PA, USA
|
| Posted: Wed Sep 15, 2010 11:18 am Post subject: |
|
|
| Quote: | | by ati do they mean a crossfire mobo? cause i have intel cpu and nv card. |
A patch in 2.6.35 to fix a a bug with the high resolution timer on ati/amd motherboards broke the high resolution timer on intel boards under certain conditions.
| Quote: | | also they talk about u32 and not casting. my bios has an option for 64 or 32bit hpet. i have it set to 64b at the moment i might try as 32b as well as this patch. |
I believe they were talking about the C source code of the patch. Specifically a redundant statement that converted the output to unsigned 32 bit integer when it already was a 32 bit unsigned integer.
| Quote: | | also any word if this patch is making into 2.6.36? |
This has been pushed to the -tip tree which is described by the following:
| Quote: | | The -tip tree: maintenance and integration tree for scheduler, timer, tracing, genirq, x86 and other subsystems. |
I expect this to hit the mainline git (for 2.6.36) within the day or so.
_________________
John
My gentoo overlay
Instructons for overlay |
|
| Back to top |
|
|
mengzhao
n00b
Joined: 18 Mar 2004
Posts: 13
|
| Posted: Wed Sep 15, 2010 1:11 pm Post subject: |
|
|
| I think it is the sata driver problem, the disk become not responsive intermitently. |
|
| Back to top |
|
|
drescherjm
Advocate
Joined: 05 Jun 2004
Posts: 2790
Location: Pittsburgh, PA, USA
|
| Posted: Wed Sep 15, 2010 1:30 pm Post subject: |
|
|
Read the last page of this thread. A bug has been found in the kernel and it is caused by a change in the code for the system high resolution timer.
BTW, Just now there is a new patch in the works that does not have an extra read penalty but seems to fix the ati/amd bug. I have volunteered to test this on my i7 machine tonight..
_________________
John
My gentoo overlay
Instructons for overlay |
|
| Back to top |
|
|
depontius
Advocate
Joined: 05 May 2004
Posts: 3525
|
| Posted: Wed Sep 15, 2010 2:08 pm Post subject: |
|
|
Will this patch get backported into the "stable" kernel series, or should CoreI7 users stay <2.6.35 until 2.6.36 is out. (I've heard that this hits 2.6.35+.)
I'm a new CoreI7 user, installing a Thinkpad W510 now.
_________________
.sigs waste space and bandwidth |
|
| Back to top |
|
|
|
Kernel & Hardware
