----- Original Message -----
> On 07/07/16 02:31, Dave Brolley wrote:
> > On 07/05/2016 11:05 PM, Nathan Scott wrote:
> >> ... and it looks like we are seeing a sockaddr that is (still) completely
> >> zeroed after we accept on the fd in pmcd/client.c AcceptNewClient. The
> >> attached patch seems to tidy it up for me ... whaddya think Dave? Are we
> >> likely to see other places where this happens, I wonder?
> >>
> > I would think that it could happen for any call to __pmAccept().
> >
> > In your patch, based on __pmCheckAcceptedAddress() requiring that the
> > family be set in the address, you set it before calling __pmAccept(). If
> > we're going to require that the family be provided to
> > __pmCheckAcceptedAddress() via __pmAccept(), then we should probably
> > bite the bullet and enforce that by making the family a 4th parameter to
> > __pmAccept() and have __pmAccept set it, when needed.
> >
> > Dave
> >
>
> I don't think the change needs to be that intrusive. I have a simpler
> approach that I've committed in my tree (patch attached).
Yep, lets go with that one.
> This addresses the Mac OS X issues and passes all the -g pmcd QA tests
> on both Mac OS X and Ubuntu 16.04 and now has good air time on a handful
> of other QA hosts.
>
Looking good on Fedora & RHEL here too, as you'd expect - thanks Ken.
cheers.
--
Nathan
|