Re: [Qemu-devel] [PATCH] qapi: add explicit null to string input and output visitors

[Greg Kurz <groug@kaod.org>]

[-- Attachment #1: Type: text/plain, Size: 4105 bytes --]

On Mon, 21 Nov 2016 10:32:50 -0600
Eric Blake <eblake@redhat.com> wrote:

> On 11/19/2016 02:28 AM, Greg Kurz wrote:
> > This may be used for deprecated object properties that are kept for
> > backwards compatibility.
> > 
> > Signed-off-by: Greg Kurz <groug@kaod.org>
> > ---
> > 
> > This is enough to fix David's patch:
> > 
> > [RFCv2 10/12] pseries: Move CPU compatibility property to machine  
> 
> Do we need both input and output visitors, or is the problem only caused
> on an output visit?
> 

David's patch unifies the get/set property info callbacks into a single
getset_compat_deprecated() function which calls visit_type_null(). So,
even if the operation is a nop for input visitors, we need it anyway,
otherwise visit_type_null() dereferences a NULL pointer.

> > 
> > Messag-Id: <1479248275-18889-11-git-send-email-david@gibson.dropbear.id.au>
> > 
> >  qapi/string-input-visitor.c  |    5 +++++
> >  qapi/string-output-visitor.c |   14 ++++++++++++++
> >  2 files changed, 19 insertions(+)
> > 
> > diff --git a/qapi/string-input-visitor.c b/qapi/string-input-visitor.c
> > index 8dfa5612522b..4a378d7b67e6 100644
> > --- a/qapi/string-input-visitor.c
> > +++ b/qapi/string-input-visitor.c
> > @@ -314,6 +314,10 @@ static void parse_type_number(Visitor *v, const char *name, double *obj,
> >      *obj = val;
> >  }
> >  
> > +static void parse_type_null(Visitor *v, const char *name, Error **errp)
> > +{
> > +}  
> 
> In particular, this always succeeds, even though we don't have a
> symmetric way to guarantee that the user can specifically request NULL
> input (the qobject visitor checks for the JSON keyword NULL, but that's
> different from string parsing), so I'm a bit worried that it is not
> doing what we want. If you omit this hunk, does the problem with David's
> patch still go away with just the output visitor change?
> 

What we want here is to simply ignore any string the user would pass.

As explained above, this hunk is needed as long as David's patch calls
visit_type_null() for both input and output. Otherwise, it should have
separate functions: get_compat_deprecated() which calls visit_type_null()
and set_compat_deprecated() which does nothing.

> > +
> >  static void parse_optional(Visitor *v, const char *name, bool *present)
> >  {
> >      StringInputVisitor *siv = to_siv(v);
> > @@ -348,6 +352,7 @@ Visitor *string_input_visitor_new(const char *str)
> >      v->visitor.type_bool = parse_type_bool;
> >      v->visitor.type_str = parse_type_str;
> >      v->visitor.type_number = parse_type_number;
> > +    v->visitor.type_null = parse_type_null;
> >      v->visitor.start_list = start_list;
> >      v->visitor.next_list = next_list;
> >      v->visitor.end_list = end_list;
> > diff --git a/qapi/string-output-visitor.c b/qapi/string-output-visitor.c
> > index 94ac8211d144..5ec5352ca87c 100644
> > --- a/qapi/string-output-visitor.c
> > +++ b/qapi/string-output-visitor.c
> > @@ -266,6 +266,19 @@ static void print_type_number(Visitor *v, const char *name, double *obj,
> >      string_output_set(sov, g_strdup_printf("%f", *obj));
> >  }
> >  
> > +static void print_type_null(Visitor *v, const char *name, Error **errp)
> > +{
> > +    StringOutputVisitor *sov = to_sov(v);
> > +    char *out;
> > +
> > +    if (sov->human) {
> > +        out = g_strdup("<null>");
> > +    } else {
> > +        out = g_strdup("");
> > +    }
> > +    string_output_set(sov, out);
> > +}
> > +
> >  static void
> >  start_list(Visitor *v, const char *name, GenericList **list, size_t size,
> >             Error **errp)
> > @@ -351,6 +364,7 @@ Visitor *string_output_visitor_new(bool human, char **result)
> >      v->visitor.type_bool = print_type_bool;
> >      v->visitor.type_str = print_type_str;
> >      v->visitor.type_number = print_type_number;
> > +    v->visitor.type_null = print_type_null;
> >      v->visitor.start_list = start_list;
> >      v->visitor.next_list = next_list;
> >      v->visitor.end_list = end_list;
> > 
> > 
> >   
> 


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 181 bytes --]