[PATCH] Make loadparm more common

Volker Lendecke Volker.Lendecke at SerNet.DE
Thu Apr 3 05:54:11 MDT 2014

On Thu, Apr 03, 2014 at 01:45:47PM +0200, David Disseldorp wrote:
> > > > No, I'm not willing to add any more talloc_tos() to this area of the
> > > > code.  Almost all the odd unexpected failures caused by this patch set
> > > > were due to new talloc_tos() calls, because not all callers had a
> > > > talloc_stackframe().    
> > 
> > Well, those is a bug then, and the developer mode to panic
> > in that case was introduced so that we can fix the callers up.
> > 
> > I think adding talloc_tos() is way better than adding talloc(NULL,...)
> > since this way we can at least easily spot the leaks.
> I disagree. IMO tallocations on the null context are much more readable
> and debugable. For talloc_tos() tallocations one needs to consider
> whether a stackframe is around, and when the next garbage collection
> could take place.

Just assume in all routines except external interfaces that
a talloc stackframe is around.

There is no garbage collection, freeing the stackframe is
fully under your control if you add an explicit
talloc_stackframe() to your function.

If you don't do an explicit talloc_stackframe() in your
function, assume that the memory will not be around very
quickly after you returned from your function.

talloc_tos() does not liberate you from talloc_free'ing your
memory. I consider leaking on talloc_tos() bad style. It is
purely a safety measure against accidential memory leaks, in
particular in error return paths. That's the basic
difference: It's bad style versus a potentially devastating
memory leak.

I believe that talloc_tos() has contributed big ways to the
fact that memory leaks are almost gone from the Samba
buglist. Not fully, true, but memleaks are almost no issue

Hope that clarifies things a bit.

With best regards,

Volker Lendecke

SerNet GmbH, Bahnhofsallee 1b, 37081 Göttingen
phone: +49-551-370000-0, fax: +49-551-370000-9
AG Göttingen, HRB 2816, GF: Dr. Johannes Loxen
http://www.sernet.de, mailto:kontakt at sernet.de

More information about the samba-technical mailing list