patch set for #90289
nmav at redhat.com
Wed Jun 24 07:13:30 PDT 2015
On Wed, 2015-06-24 at 14:53 +0100, David Woodhouse wrote:
> On Wed, 2015-06-24 at 15:50 +0200, Nikos Mavrogiannopoulos wrote:
> > On Wed, 2015-06-24 at 14:41 +0100, David Woodhouse wrote:
> > > > A better/more precise fix (replacement for 0002-Do-not-finalize
> > > > -modules-created-in-another-pid.patch) is attached.
> > > Hm, what if the parent exits but the child remains. Doesn't the
> > > parent
> > > deinitialise the functions that the child still needs? Don't you
> > > need
> > > some kind of refcount? Not that I know how you can achieve that,
> > > without locking in the atfork child handler.
> > Correct, that seems like a source of many troubles. I think a ref
> > counter for that would complicate things a lot. Would it make sense
> > not
> > to release it at all? It will get released on the last user exits.
I think that makes sense short term, to avoid more crashes with the
current libffi. I've put my recommended patch set which fixes p11-kit
It would be nice if Stef you could provide a comment and/or merge them.
> Or maybe we should just file a libffi bug and expect libffi to take
> care of the fork handling if it's going to do things this way?
More information about the p11-glue