On Fri, Nov 11, 2005 at 08:26:08AM -0800, John Liang
wrote:> one more question on this. Is there a one to one mapping between
> a user mutex and a kernel mutex? I am not sure when a lockstat
> provider should be used.
If you''re only interested in user-land locking primitives, you
shouldn''t
need to use lockstat at all. The one case which is a slight exception is
for process robust mutexes. You can see this bug for more details:
http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6286326
Arguably, however, this is actually a bug in the implementation of
process robust mutexes:
http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6296770
Adam
--
Adam Leventhal, Solaris Kernel Development http://blogs.sun.com/ahl