RHEL6 init.d script (was Re: Initscript Debian Wheezy)

Jeremy Allison jra at samba.org
Tue Sep 11 17:05:56 MDT 2012

On Wed, Sep 12, 2012 at 12:09:36AM +0200, Stefan (metze) Metzmacher wrote:
> It might not happen, but it's at least very confusing
> and might cause problems in future.

Ok, *now* you're making sense :-). Thanks for admitting
the real problem :-).

> After the destructor returns (*pp_state)->pp_self_ref points
> to invalid memory currently.

But the destructor only fires in one of 2 cases.


                /* we want to keep the session */
                tevent_req_nterror(req, status);

The tevent_req_nterror will free the req and then
also the state.


Shutdown where state->session is being
freed outside of the sessionsetup being
shut down. As the state destructor only
ever calls TALLOC_FREE(state->session)
and the pp_self_ref_destructor has set
(*pp_state)->session = NULL; if the
in-flight session was freed first, then
it doesn't matter (*pp_state)->pp_self_ref
is pointing to freed memory, no one is
ever going to touch it again.

So the current code is completely safe
as written.

But I'll add nulling out (*pp_state)->pp_self_ref
in the destructor to make it explicit no one
is going to access this again. 


More information about the samba-technical mailing list