Re: [Qemu-devel] [PATCHv6 3/5] fw_cfg: move assert() and linking of fw_cfg device to the machine into instance_init()

[Igor Mammedov]

On Wed, 28 Jun 2017 16:21:40 +0200
Laszlo Ersek <address@hidden> wrote:

> On 06/28/17 16:12, Igor Mammedov wrote:
> > On Wed, 28 Jun 2017 08:09:35 +0100
> > Mark Cave-Ayland <address@hidden> wrote:  
> 
> >> 1) Alter fw_cfg_find() to use object_resolve_path_type("", TYPE_FW_CFG,
> >> NULL)  
> > I'd make use of the 3rd argument &ambiguous and assert on it  
> 
> I vaguely recall playing with "ambiguous" in find_vmgenid_dev(), but it
> didn't work as I expected (or, it wasn't necessary to use). Not sure
> about the details, I only remember that, when calling
> object_resolve_path_type() like above, from within a realize function,
> the object under realization is already considered existent, so if we're
> realizing the second (or later) instance of the class,
> object_resolve_path_type() will return NULL (regardless of "ambiguous").
> If we're realizing the very first instance, then
> object_resolve_path_type() will return non-NULL.
> 
> I don't mind "ambiguous" if it can be made work fine, just thought that
> I'd add this tidbit.


looking at object_resolve_partial_path()

        if (ambiguous && *ambiguous) {                                          
 
            return NULL;                                                        
 
        }

might make object_resolve_partial_path() return non NULL if ambiguous is NULL


object_resolve_partial_path(,,, NULL)
p0 - p00 - fw0
    |    |
    |      fw1
    |
     fw2

p00 returns NULL &&  only fw2 is found without knowing at all about p00 failure


> 
> Thanks,
> Laszlo