On Sat, Oct 27, 2001 at 07:45:41PM +0100, Lars Heidieker wrote:

 > Yes using the ref count seems quite good to me as is it makes the difference.
 > The trouble with making the difference by looking at vtext is that only the 
 > text section are mapped vtext and that the datasection of an executable are 
 > mapped like "any other filemapping" or are anons if they have been cow.
 > I think there has been some discussion about the fact that libraries are 
 > not acounted as vtext at all (in the current implementation).

Currently, only images that are exec'd are marked VTEXT.  It might make
sense to mark a vnode as VTEXT when a PROT_EXEC mmap'ing happens or when
a PROT_EXEC mprotect'ing happens.

-- 
        -- Jason R. Thorpe <thorpej@wasabisystems.com>