Hi, folks.
Well, I got far enough along on the mmap()/symtab approach of pstat() to
realize that the cure would be worse than the symptom. The biggest pain was
that most of what the /proc server would examine could become unmapped
asynchronously. So as the poor server is trying to make forward progress
bits and pieces can be going away. The /proc server thus has to catch
events and structure his scan so that the event handler can longjmp() out to
a point where the server can realize that a structure went away while he was
looking at it.
If it sounds bad on paper, just imagine the code I was writing....
So, let's just stick with pstat() and make it as short and sweet (on the
kernel side) as possible. I'll apply Dave's patches as soon as he makes a
final judgement on the walk-the-list versus list-of-pids approach.
Andy
Received on Tue Dec 6 08:32:29 1994
This archive was generated by hypermail 2.1.8 : Thu Sep 22 2005 - 15:12:10 PDT