[ovs-dev] [PATCH 8/9] vconn: Reimplement in terms of the "stream" abstraction.

[Justin Pettit]

On Dec 21, 2009, at 1:15 PM, Ben Pfaff wrote:

> 
> +
> +/* Creates a new vconn that will send and receive data on a stream named 'name'
> + * and stores a pointer to the vconn in '*vconnp'.
> + *
> + * Returns 0 if successful, otherwise a positive errno value.  (The current
> + * implementation never fails.) */

Is this true?  It seems like stream_open can fail, which will return an error.

> +static int
> +vconn_stream_open(const char *name_, char *suffix UNUSED,
> +                  struct vconn **vconnp)
> +{
> +    struct stream *stream;
> +    char *name;
> +    int error;
> +
> +    if (!strncmp(name_, "tcp:", 4) && count_fields(name_) < 3) {
> +        name = xasprintf("%s:%d", name_, OFP_TCP_PORT);
> +    } else if (!strncmp(name_, "ssl:", 4) && count_fields(name_) < 3) {
> +        name = xasprintf("%s:%d", name_, OFP_SSL_PORT);
> +    } else {
> +        name = xstrdup(name_);
> +    }
> +    error = stream_open(name, &stream);
> +    free(name);
> +
> +    if (error && error != EAGAIN) {
> +        return error;
> +    }
> +
> +    *vconnp = vconn_stream_new(stream, error);
>     return 0;
> }

Otherwise, from my quick read-through, it seemed reasonable.

--Justin