Skip to contentSkip to conversation

2.6.22 - Suspend and Weird Load Averages

DISCUSSION Parag Warudkar 3 participants 7 messages Jul 18, 2007 – Jul 26, 2007

lore
Share & tools
lore
#0 lore raw

For the first time since 1997 I have a machine (Mac Mini) running a stock kernel which I am suspending/resuming for 2 days now without any problems. Amongst the various hardware and kernel versions that I've used so far this combination is the first to suspend/resume without issues!

But one thing that annoys me a bit is that the load averages spike up quite a bit after resume although it doesn't feel like the machine is loaded -

Before Resume -

parag@parag-desktop:~$ uptime
09:35:14 up  5:18,  2 users,  load average: 0.08, 0.87, 1.00
After Resume -
parag@parag-desktop:~$ uptime
09:36:37 up  5:18,  2 users,  load average: 12.53, 3.83, 1.98
parag@parag-desktop:~$

Is this a known problem? Worth fixing?

Thanks
Parag

#1 lore raw
Quoted 11 lines · Parag Warudkar

For the first time since 1997 I have a machine (Mac Mini) running a stock kernel which I am suspending/resuming for 2 days now without any problems. Amongst the various hardware and kernel versions that I've used so far this combination is the first to suspend/resume without issues!

But one thing that annoys me a bit is that the load averages spike up quite a bit after resume although it doesn't feel like the machine is loaded -

Before Resume -

Quoted 8 lines
parag@parag-desktop:~$ uptime
 09:35:14 up  5:18,  2 users,  load average: 0.08, 0.87, 1.00
After Resume -
parag@parag-desktop:~$ uptime
 09:36:37 up  5:18,  2 users,  load average: 12.53, 3.83, 1.98
parag@parag-desktop:~$

Is this a known problem?

I think so.

This probably is due to the freezing of tasks. Namely, the frozen tasks are in TASK_UNINTERRUPTIBLE and add to the load average.

Quoted 1 line

Worth fixing?

Cartainly.

Greetings,
Rafael

Signature · 1 line
"Premature optimization is the root of all evil." - Donald Knuth
#2 lore raw
Quoted 21 lines · Rafael J. Wysocki
Quoted 11 lines · Parag Warudkar

For the first time since 1997 I have a machine (Mac Mini) running a stock kernel which I am suspending/resuming for 2 days now without any problems. Amongst the various hardware and kernel versions that I've used so far this combination is the first to suspend/resume without issues!

But one thing that annoys me a bit is that the load averages spike up quite a bit after resume although it doesn't feel like the machine is loaded -

Before Resume -

Quoted 8 lines
parag@parag-desktop:~$ uptime
 09:35:14 up  5:18,  2 users,  load average: 0.08, 0.87, 1.00
After Resume -
parag@parag-desktop:~$ uptime
 09:36:37 up  5:18,  2 users,  load average: 12.53, 3.83, 1.98
parag@parag-desktop:~$

Is this a known problem?

Quoted 8 lines

I think so.

This probably is due to the freezing of tasks. Namely, the frozen tasks are in TASK_UNINTERRUPTIBLE and add to the load average.

Quoted 1 line

Worth fixing?

Cartainly.

s/Cartainly/Certainly/

#3 lore raw
In reply to Parag Warudkar

Hi!

Quoted 21 lines

For the first time since 1997 I have a machine (Mac Mini) running a stock kernel which I am suspending/resuming for 2 days now without any problems. Amongst the various hardware and kernel versions that I've used so far this combination is the first to suspend/resume without issues!

But one thing that annoys me a bit is that the load averages spike up quite a bit after resume although it doesn't feel like the machine is loaded -

Before Resume -

parag@parag-desktop:~$ uptime
09:35:14 up  5:18,  2 users,  load average: 0.08, 0.87, 
1.00
After Resume -
parag@parag-desktop:~$ uptime
09:36:37 up  5:18,  2 users,  load average: 12.53, 3.83, 
1.98
parag@parag-desktop:~$
Quoted 1 line

Is this a known problem? Worth fixing?

Known, and Rafael/Nigel have patches to fix that.

OTOH I'm not sure it is a problem -- lets just say that suspend is hard work and therefore it raises load avg?
Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html
#4 lore raw

Hi Pavel!

Quoted 1 line · Pavel Machek

Known, and Rafael/Nigel have patches to fix that.

Thanks for the info - Rafael/Nigel do you have any patches for me to try/modify - I am just curious to see what approach the patches take to fix this - I had started thinking about TASK_SUSPENDED which would be equivalent to TASK_UNINTERRUPTIBLE for all purposes except load calculation (not counted in nr_uninterruptible) but somehow I think there might be a less intrusive / more correct way to do this.

Quoted 4 lines

OTOH I'm not sure it is a problem -- lets just say that suspend is hard work and therefore it raises load avg?
Pavel

Hehe, nope - that would be inefficient suspend for me! Any way I understand it's purely a cosmetic problem but if we could be cosmetically correct to report the right uptime after suspend sans the sleep time (unlike other OSes which include sleep time when reporting it) I think we should also fix this one to report correct load averages.

Cheers

Parag

#5 lore raw
Quoted 8 lines · Parag Warudkar

Hi Pavel!

Quoted 1 line · Pavel Machek

Known, and Rafael/Nigel have patches to fix that.

Thanks for the info - Rafael/Nigel do you have any patches for me to try/modify -

Well, I don't.

Quoted 5 lines

I am just curious to see what approach the patches take
to fix this - I had started thinking about TASK_SUSPENDED which would be equivalent to TASK_UNINTERRUPTIBLE for all purposes except load calculation (not counted in nr_uninterruptible) but somehow I think there might be a less intrusive / more correct way to do this.

We wanted to save the pre-suspend load averages and restore them during the resume.

Greetings,
Rafael

Signature · 1 line
"Premature optimization is the root of all evil." - Donald Knuth
#6 lore raw
Quoted 11 lines · Rafael J. Wysocki
Quoted 8 lines · Parag Warudkar

Hi Pavel!

Quoted 1 line · Pavel Machek

Known, and Rafael/Nigel have patches to fix that.

Thanks for the info - Rafael/Nigel do you have any patches for me to try/modify -

Well, I don't.

Okay, someone did, sorry for confusion. It should be in linux-pm archives.
Pavel

Signature · 2 lines
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html