On Fri, 2018-03-02 at 07:34 -0500, Laine Stump wrote:
On 02/21/2018 09:14 AM, Andrea Bolognani wrote: [...]
+static int +qemuDomainDeviceDefValidateControllerPCI(const virDomainControllerDef *cont, + const virDomainDef *def, + virQEMUCapsPtr qemuCaps) + +{ + const virDomainPCIControllerOpts *pciopts = &cont->opts.pciopts; + const char *model = virDomainControllerModelPCITypeToString(cont->model); + const char *modelName = virDomainControllerPCIModelNameTypeToString(pciopts->modelName); + + if (!model) { + virReportError(VIR_ERR_INTERNAL_ERROR, + _("Unknown virDomainControllerModelPCI value: %d"), + cont->model); + return -1; + } + if (!modelName) { + virReportError(VIR_ERR_INTERNAL_ERROR, + _("Unknown virDomainControllerPCIModelName value: %d"), + pciopts->modelName); + return -1; + }
(meant to send this before, but kept forgetting...)
1) I thought modelName wasn't set for pci-root. Doesn't the above cause a validation error in that case? (too early in the day, haven't tried it)
The default value is _MODEL_NAME_NONE aka zero, which is still part of the enumeration, so virDomainControllerPCIModelNameTypeToString() won't return NULL and no error will be raised. For pSeries guests, it will be set to _MODEL_NAME_SPAPR_PCI_HOST_BRIDGE so once again no problem there.
2) danpb made a nice new function to standardize/simplify errors of the above type: virReportEnumRangeError().
His efforts on switch normalization and me rebasing this series happened pretty much at the same time; more specifically, the function you're talking about was introduced in 3b1020ac805e, while my series is based on the earlier f565321b26df. I guess this means another rebase! Yay! \o/ -- Andrea Bolognani / Red Hat / Virtualization