> I haven't dug into the complete details of the > calculation, but the top entry is the RSS as > calculated for output via ps(1) via /proc and is the > RSS for just that process. > > The second line is the RSS calculated as being in use > throughout the system for the user in question, and > as such may include some shared memory areas the > first calculation does not.
Well...in his example (prstat -a -u smmsp) NPROC in the second section was 1, which makes it a little less clear what's happening. And no SysV shared memory segments were present at all. It's the usual case (absent zones or incoming to the submission port) that there's only one process running with that particular user id. I looked at those vs pmap -x for the same PID on a Solaris 10 system, and the total RSS very roughly matched the 2nd section of the prstat output, while the values for just the executable itself (minus shared libs, anon, stack, and heap) very roughly matched the 1st section. Maybe that's what you mean? Anyway, short of reading the prstat code (wasn't super obvious when I glanced _very_ quickly at it), comparing to pmap -x seems a reasonable thing to do... This message posted from opensolaris.org _______________________________________________ opensolaris-discuss mailing list opensolaris-discuss@opensolaris.org
