Re: getloadavg and proc filesystem

[Andrew Miehs <andrew@2sheds.de>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Dear List,

I finally got around to asking the linux kernel mailing list about
reading from the proc filesystem, and reading /proc/stat

These are the two most interesting comments to note:

On Dec 5, 2005, at 4:58 PM, Bernd Eckenfels wrote:
...
> the files in proc are all to some degree atomic to some not. You  
> will have
> to look into the source for the specific file to see the details. For
> smaller files if they fit into a single 4k page you have good chances.
> Bigger files can cause all kind of trouble. (most often you miss new
> records)
>
> In case of stat the readings should be consistent within itself  
> since all
> cpu lines fit into a single page.
...

and secondly regarding exactly on how to read /proc/stat


On Dec 7, 2005, at 4:12 PM, Rik van Riel wrote:
> On Wed, 7 Dec 2005, Andrew Miehs wrote:
>
> > After reading your mail http://mail.nl.linux.org/linux-mm/2002-05/ 
> > msg00099.html regarding the implementation of iowait in /proc/ 
> > stat, I was wondering what the best method was to calculate iowait  
> > statistics.
> >
> > According to point 3...
> >
> > > 3) iowait time is counted separately from user/system/idle and
> > >   can overlap with either system or idle (when no process is
> > >   running the system can still be busy processing interrupts)
> >
> > is it correct to assume that
> >   user + system + idle + nice + interrupts + iowait > 100% cputime
> > ?
>
> No, the total = 100%



Regards

Andrew
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (Darwin)

iD8DBQFD+PrdW126qUNSzvURAn0gAJ9NtunOccmQHTaM1eDNaSx5jX89VACdHYUp
zRRsuNmtNPZfecdRJ5NwJ4o=
=arTx
-----END PGP SIGNATURE-----