Re: [Qemu-devel] [PATCH 02/11] Add migrate-set-capabilities and query-migrate-capabilities

[Orit Wasserman]

On 08/06/2012 07:39 PM, Eric Blake wrote:
> On 08/06/2012 10:28 AM, Orit Wasserman wrote:
> 
>>> That is, BOTH commands end up iterating over a list of caps, and output
>>> identical information in the case where caps exist of 'name: state' for
>>> each capability.
>>>
> 
>> The information is different:
>> the first:
>> MigrationCapabilityStatusList *
>> qmp_query_migrate_supported_capabilities(Error **errp)
>> {
>>     MigrationCapabilityStatusList *caps_list = g_malloc0(sizeof(*caps_list));
>>
>>     caps_list->value = g_malloc(sizeof(*caps_list->value));
>>     caps_list->value->capability = MIGRATION_CAPABILITY_XBZRLE;
>>     caps_list->value->state = true;
>>     caps_list->next = NULL;
>>
>>     return caps_list;
>> }
>>
>> the second:
>> MigrationCapabilityStatusList *
>> qmp_query_migrate_supported_capabilities(Error **errp)
> 
> I think you meant this for the second:
> 
> +MigrationCapabilityStatusList *qmp_query_migrate_capabilities(Error **errp)
> +{
> +    MigrationCapabilityStatusList *head = NULL;
> +    MigrationCapabilityStatusList *caps;
> +    MigrationState *s = migrate_get_current();
> +    int i;
> +
> +    for (i = 0; i < MIGRATION_CAPABILITY_MAX; i++) {
> +        if (head == NULL) {
> +            head = g_malloc0(sizeof(*caps));
> +            caps = head;
> +        } else {
> +            caps->next = g_malloc0(sizeof(*caps));
> +            caps = caps->next;
> +        }
> +        caps->value =
> +            g_malloc(sizeof(*caps->value));
> +        caps->value->capability = i;
> +        caps->value->state = s->enabled_capabilities[i];
> +    }
> 
>> you can look at it as 64bit support one is to know if the processor supports 
>> 64 bit
>> the other to know if the OS uses 64 bit
> 
> Okay, so the QMP code is currently different (one outputs a list where
> every entry in the list is hard-coded to true, the other outputs a list
> where each entry reflects the current state).  But that still doesn't
> address my concern that you don't need two QMP commands.
> 
> Merely listing 'xbzrle' in the list is enough to know that 'this
> particular qemu knows how to do xbzrle', and the user is free to ignore
> the additional information of whether it is actually in use at the time
> if all the application cared about was determining the set of known
> capabilities.
> 
> Given your argument with 64-bit processing, the same reasoning applies -
> ask for a list of capabilities, and the result will either omit '64bit'
> (the capability is not possible), include '64bit:false' (the capability
> is known by the emulator, but not in use by the guest), or include
> '64bit:true' (the capability is both known and in-use).  A user that
> only cares about querying supported capabilities will check whether
> '64bit' is in the list, and throw away the information about whether it
> is on or off (and since the QMP command currently returns a hard-coded
> true, that information is already being discarded via your
> query-migrate-supported-capabilities command).  That is, your
> implementation for query-migrate-capabilities is sufficient to satisfy
> both class of users, without needing query-migrate-supported-capabilities.
> 
As you represent the user (libvirt) you know the best.
I will remove supported-capabilities commands.

Orit