[Libvirt-cim] [PATCH 2 of 3] Add input support to DefneSystem()

[Dan Smith]

KR> +static bool default_input_device(struct domain *domain)
KR> +{
KR> +        free(domain->dev_input);
KR> +        domain->dev_input = calloc(1, sizeof(*domain->dev_input));
KR> +        if (domain->dev_input == NULL) {
KR> +                CU_DEBUG("Failed to allocate default input device");
KR> +                return false;
KR> +        }
KR> +
KR> +        domain->dev_input->dev.input.type = strdup("mouse");
KR> +
KR> +        if (domain->type == DOMAIN_LXC) {
KR> +                domain->dev_input->dev.input.bus = strdup("usb");

I don't think we should have a default for LXC.  The fact that the LXC
driver even allows that in the domain definition is a bug, IMHO.
If/when that gets fixed, this code will break us.

KR> +static const char *input_rasd_to_vdev(CMPIInstance *inst,
KR> +                                      struct virt_device *dev)
KR> +{
KR> +        CMPIObjectPath *op = NULL;
KR> +        const char *val;
KR> +        const char *msg;
KR> +        char *type = NULL;
KR> +        char *bus = NULL;
KR> +        char *vtype = NULL;
KR> +
KR> +        if (cu_get_str_prop(inst, "ResourceSubType", &val) != CMPI_RC_OK) {
KR> +                msg = "InputRASD ResourceSubType field not valid";
KR> +                goto out;
KR> +        }
KR> +
KR> +        dev->dev.input.type = strdup(val);

For this,

KR> +
KR> +        op = CMGetObjectPath(inst, NULL);
KR> +        if (op == NULL) {
KR> +                CU_DEBUG("Unable to determine class of InputRASD");
KR> +                return NULL;
KR> +        }
KR> +
KR> +        if ((cu_get_str_prop(inst, "Caption", &val) != CMPI_RC_OK) || 
KR> +           (get_input_type_bus_from_cap(val, &type, &bus, &vtype) != 1)) {
KR> +                if (STREQC(dev->dev.input.type, "mouse"))
KR> +                        dev->dev.input.bus = strdup("ps2");
KR> +                else if (STREQC(dev->dev.input.type, "tablet"))
KR> +                        dev->dev.input.bus = strdup("usb");
KR> +                else {
KR> +                        msg = "Invalid value for ResourceSubType in InputRASD";
KR> +                        goto out;
KR> +                }
KR> +        } else
KR> +                dev->dev.input.bus = strdup(bus);

...and this, we should make sure that an invalid input type properly
rejects with the libvirt error when we try to define it.  Now that we
have that infrastructure in place, it would be nice to make sure that
if we don't do input checking here, a reasonable error makes it back
to the user.

-- 
Dan Smith
IBM Linux Technology Center
Open Hypervisor Team
email: danms at us.ibm.com