Re: [Qemu-devel] [PATCH 1/2] block: Add bs->node_name to hold the name of a bs node of the bs graph.

[Eric Blake <eblake@redhat.com>]

[-- Attachment #1: Type: text/plain, Size: 2247 bytes --]

On 11/07/2013 08:01 AM, Benoît Canet wrote:
> Add the minimum of code to prepare the followings patches.
> 
> If no node_name is provided to bdrv_new the bs->node_name is set to "undefined".
> This will allow to have some default string to communicate in QMP and HMP.
> This also make "undefined" a reserved string for bs->node_name.
> 
> Signed-off-by: Benoit Canet <benoit@irqsave.net>
> ---

> +    /* if node name is given store it in bs and insert bs in the graph bs list
> +     * note: undefined is a reserved node name
> +     */
> +    if (node_name &&
> +        node_name[0] != '\0' &&
> +        strcmp(node_name, "undefined")) {
> +        pstrcpy(bs->node_name, sizeof(bs->node_name), node_name);
> +        QTAILQ_INSERT_TAIL(&graph_bdrv_states, bs, node_list);
> +    /* else set the bs node name to undefined for QMP and HMP */
> +    } else {
> +        sprintf(bs->node_name, "undefined");

strcpy() is more efficient than sprintf() with no % in the format string.


>  
> +/* This function is to find a node in the bs graph */
> +BlockDriverState *bdrv_find_node(const char *node_name)
> +{
> +    BlockDriverState *bs;
> +

Should this function assert that node_name is not "undefined"?


> +++ b/include/block/block_int.h
> @@ -297,11 +297,18 @@ struct BlockDriverState {
>      BlockdevOnError on_read_error, on_write_error;
>      bool iostatus_enabled;
>      BlockDeviceIoStatus iostatus;
> +
> +    /* the following member give a name to every node on the BlockDriverState

s/give/gives/

> +     * graph.
> +     */
> +    char node_name[32];
> +    QTAILQ_ENTRY(BlockDriverState) node_list;
> +    /* Device name is the name associated with the "drive" the guest see */

s/see/sees/

>      char device_name[32];
> +    QTAILQ_ENTRY(BlockDriverState) device_list;

Maybe I missed it, but is there code that ensures there are no duplicate
node names (other than the magic "undefined")?

Seems like it's moving in the right direction, although I'm not sure
it's worth applying this until we know the qapi for working with node
names makes sense.

-- 
Eric Blake   eblake redhat com    +1-919-301-3266
Libvirt virtualization library http://libvirt.org


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 621 bytes --]