I'm trying to use the sysctl function call to get a process table. It kind of works, but p_pctcpu is always 0, even for processes which are consuming a lot of cpu time, such as backupd.
First, does sysctl work? It must since that is how ps is implemented. Can I see the source code to ps somewhere?
Second, there are lots of references to something called 'struct proc' but it isn't defined anywhere. There is an extern_proc structure, but no proc structure. mystifying. How does it work?
First, does sysctl work? It must since that is how ps is implemented. Can I see the source code to ps somewhere?
Second, there are lots of references to something called 'struct proc' but it isn't defined anywhere. There is an extern_proc structure, but no proc structure. mystifying. How does it work?
posted by:
|
SF Bay Area
|
-
Re: struct proc and sysctl(3)
Tue, December 11, 2007 - 7:24 PMI don't know if it'll help much, but the FreeBSD source is all online, and in particular, the source for that version of "ps" is here:
www.freebsd.org/cgi/cvsweb...src/bin/ps/
HTH - Good luck!
Dana -
-
Re: struct proc and sysctl(3)
Tue, December 11, 2007 - 8:08 PM
Dana, Hoco ...
You can download the source to the Darwin kernel and the command line utils -- pretty much the whole *nix part of OSX -- yourself from Apple, free of charge.
Regards,
John
Falling You - exploring the beauty of voice and sound
www.fallingyou.com
-
-
Re: struct proc and sysctl(3)
Wed, December 12, 2007 - 9:03 AMI actually independently discovered this, but thanks. I finally found the ps program in the adv_cmds archive. I dove in and found that I still don't know what a struct proc is, but it doesn't seem to matter. There is a kinfo_proc that is what is really being used.
The source to ps was enlightening. It seems that there is a member called p_pctcpu which is supposed to be the percentage used. It must be vestigal because it is always 0. I instrumented the ps program and recompiled it and found that even there it is 0. In OS X (and apparently in BSD) the cpu usage is calculated by adding up the cpu usage of the threads in a process. A chunk of code that makes use of some low-level mach functions for which I can find no docs; however, I was able to learn what I needed and use the same approach to get the cpu usage I wanted.
Having access to the darwin source was absolutely essential -- I never would have figured it out without it.
-
-
