[procps] Re: C-States handling - new switch?
- From: Jim Warner <james.warner@xxxxxxxxxxx>
- To: procps@xxxxxxxxxxxxx
- Date: Wed, 1 Feb 2012 04:45:10 -0600
On Feb 1, 2012, at 4:18 AM, Jaromir Capik wrote:
>> I'm not sure if that is the one that is incorporated.
>> Could you try a 3.3.x top on your system to see if the problem goes
>> away?
>
> Hi guys.
>
> Craig,
> The results sent by me were collected using the 3.3.2 version.
> I've checked the source and the patch is identical to the
> corresponding part of the procps-ng code.
> ...
> I'd like to do some tests here in order to learn how
> it behaves.
Hi Guys,
Yep, that was the patch I was referring to but the database approach sounds
unreasonable.
We might be in search of an unnecessary solution. Here's what I've discovered
so far.
When the full compliment of cpus is online, the 1st /proc/stat line (the
summary line) exactly mirrors the sum total of each individual cpu for all of
the state categories. However, when one or more cpus is taken offline, then
the summary /proc/stat line exceeds the total of individual cpus.
So the assumption that very few or tics on a tickless system could distort
summary display may not hold true. Wouldn't that make the perceived problem go
away?
Anyway, I've got more research to do...
Regards,
Jim
p.s. I've got a patch to make top tolerant of the loss of a cpu. As it stands
now he exits with a failed /proc/stat read.
Other related posts:
