Re: [Qemu-devel] [6216] Add a model string to VLANClientState (Mark McLoughlin)

[Mark McLoughlin <markmc@redhat.com>]

On Wed, 2009-01-07 at 20:23 +0100, Stefan Weil wrote:
> Anthony Liguori schrieb:
> > Revision: 6216
> >           http://svn.sv.gnu.org/viewvc/?view=rev&root=qemu&revision=6216
> > Author:   aliguori
> > Date:     2009-01-07 17:42:25 +0000 (Wed, 07 Jan 2009)
> >
> > Log Message:
> > -----------
> > Add a model string to VLANClientState (Mark McLoughlin)
> >
> > Don't lose track of what type/model a vlan client is so that we can
> > e.g. assign a global per-model id to clients.
> >
> > The entire patch is basically a tedious excercise in making sure the
> > type/model string gets propagated down to qemu_new_vlan_client().
> >
> > Signed-off-by: Mark McLoughlin <markmc@redhat.com>
> > Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
...
> > Modified: trunk/net.c
> > ===================================================================
> > --- trunk/net.c	2009-01-07 17:40:15 UTC (rev 6215)
> > +++ trunk/net.c	2009-01-07 17:42:25 UTC (rev 6216)
> > @@ -297,6 +297,7 @@
> >  #endif
> >  
> >  VLANClientState *qemu_new_vlan_client(VLANState *vlan,
> > +                                      const char *model,
> >                                        IOReadHandler *fd_read,
> >                                        IOCanRWHandler *fd_can_read,
> >                                        void *opaque)
> > @@ -305,6 +306,7 @@
> >      vc = qemu_mallocz(sizeof(VLANClientState));
> >      if (!vc)
> >          return NULL;
> > +    vc->model = strdup(model);
> >   
> 
> Some targets don't force a value for model, so it can be NULL.
> This raises a SIGSEGV crash. Should targets set model, or is
> it better to catch this case here?

Ouch.

It'd be much nicer to make sure we always get a model string here rather
than having names default to e.g. "nic.1".

So, I went and fixed up all the targets to make sure that nd->model is
always set.

That sucked, though, so I went through a couple of iterations of
cleaning all this stuff up and settled on the patches that follow this
mail.

Cheers,
Mark.