Random processes hanging in unkillable state in -BETA6

[Brian Fundakowski Feldman]

On Thu, Sep 30, 2004 at 11:22:15AM -0700, Julian Elischer wrote:
> 
> 
> Damian Gerow wrote:
> 
> >Thus spake Damian Gerow (dgerow at afflictions.org) [30/09/04 00:32]:
> >: Thus spake Julian Elischer (julian at elischer.org) [29/09/04 18:28]:
> >: : It actually needs a *arg to be a valid ctor..
> >: : 
> >: : -ksegrp_init(void *mem, int size, int flags)
> >: : +ksegrp_ctor(void *mem, int size, void *arg, int flags)
> >: 
> >: Okay, change made, kernel recompiled, installed, and booted into.  But 
> >I'm
> >: still seeing the hangs from firefox, dig, and ogle.
> >
> >And the next ddb output is up, as per your previous instructions, at
> ><http://www.afflictions.org/~dgerow/ddb2.out>.
> >
> 
> oh ok.
> I see the garbage values are still there.  they look like flags of some 
> sort. Something must
> be writing to that location.. I'll look for an overwrite.
> ..

I don't see anything wrong with how kse groups are allocated.  If it's
for a new proc, fork clears/copies appropriately, and kse the same.
Where do you see that you'd need a constructor that clears it all so
heavy-handedly?  If anything, put the "zero" clearing part in the
constructor, and remove it from everywhere else, and things should be
more clear...

-- 
Brian Fundakowski Feldman                           \'[ FreeBSD ]''''''''''\
  <> green at FreeBSD.org                               \  The Power to Serve! \
 Opinions expressed are my own.                       \,,,,,,,,,,,,,,,,,,,,,,\