Re: [Qemu-devel] [PATCH 6/6] chardev: convert file backend to realize

[Anthony Liguori <aliguori@us.ibm.com>]

Paolo Bonzini <pbonzini@redhat.com> writes:

> Il 15/10/2012 21:34, Anthony Liguori ha scritto:
>> +static char *chardev_file_get_path(Object *obj, Error **errp)
>> +{
>> +    CharDriverState *chr = CHARDEV(obj);
>> +    FDCharDriver *s = chr->opaque;
>> +
>> +    return s->path ? g_strdup(s->path) : g_strdup("");
>> +}
>> +
>> +static void chardev_file_set_path(Object *obj, const char *value, Error **errp)
>> +{
>> +    CharDriverState *chr = CHARDEV(obj);
>> +    FDCharDriver *s = chr->opaque;
>> +
>> +    if (chr->realized) {
>> +        error_set(errp, QERR_PERMISSION_DENIED);
>> +        return;
>> +    }
>> +
>> +    if (s->path) {
>> +        g_free(s->path);
>> +    }
>> +
>> +    s->path = g_strdup(value);
>> +}
>> +
>> +static void chardev_file_initfn(Object *obj)
>> +{
>> +    CharDriverState *chr = CHARDEV(obj);
>> +
>> +    object_property_add_str(obj, "path", chardev_file_get_path, chardev_file_set_path, NULL);
>> +#ifndef _WIN32
>> +    chr->opaque = CHARDEV_FILE(obj);
>> +#endif
>> +}
>> +
>
> IMHO this really really calls for pushing static properties and realized
> up to Object...

I agree actually.  But this doesn't add anything to Object, this just
adds a new set of properties that are implemented entirely in terms of
the existing property infrastructure.

What I don't want to do though is push the notion of "realize" to
Object.  Instead, we should have a mechanism to "lock" read/write
properties so they no longer are mutable.  This should not affect the
object's state but rather the individual property state.

IOW, we shouldn't be adding anything to Object to do this.

Regards,

Anthony Liguori

>
> Paolo