Re: [libvirt] [PATCH v2 01/42] conf: add enum constants for default controller models
On Mon, Feb 19, 2018 at 01:33:23PM -0500, John Ferlan wrote:
On 02/15/2018 11:43 AM, Daniel P. Berrangé wrote:
> The controller model is slightly unusual in that the default value is
> -1, not 0. As a result the default value is not covered by any of the
> existing enum cases. This in turn means that any switch() statements
> that think they have covered all cases, will in fact not match the
> default value at all. In the qemuDomainDeviceCalculatePCIConnectFlags()
> method this has caused a serious mistake where we fallthough from the
> SCSI controller case, to the VirtioSerial controller case, and from
> the USB controller case to the IDE controller case.
>
> By adding explicit enum constant starting at -1, we can ensure switches
> remember to handle the default case.
>
> Signed-off-by: Daniel P. Berrangé <berrange(a)redhat.com>
> ---
> src/conf/domain_addr.c | 5 +++++
> src/conf/domain_conf.c | 1 +
> src/conf/domain_conf.h | 4 ++++
> src/qemu/qemu_command.c | 17 ++++++++++++++---
> src/qemu/qemu_domain.c | 10 +++++++++-
> src/qemu/qemu_domain_address.c | 22 ++++++++++++++++++++++
> src/vbox/vbox_common.c | 13 +++++++++----
> 7 files changed, 64 insertions(+), 8 deletions(-)
>
There's a few places where in the code where model is compared against
-1 that could perhaps use the VIR_DOMAIN_CONTROLLER_MODEL_*_DEFAULT (I
used 'model.*=.*-1' in the Find this egrep pattern in cscope).
At least one of them (below) caused a Coverity complaint.
So should any of those be altered or should we just live with the fact
that using/knowing -1 is the 'default' model?
Probably best to use the constant rather than literal -1.
> diff --git a/src/conf/domain_addr.c b/src/conf/domain_addr.c
> index 6422682391..df19c6be1a 100644
> --- a/src/conf/domain_addr.c
> +++ b/src/conf/domain_addr.c
[...]
> @@ -1746,6 +1750,7 @@
virDomainUSBAddressControllerModelToPorts(virDomainControllerDefPtr cont)
> return cont->opts.usbopts.ports;
> return 8;
>
> + case VIR_DOMAIN_CONTROLLER_MODEL_USB_DEFAULT:
Coverity points out that because at the top of this function, we have:
if (model == -1)
model = VIR_DOMAIN_CONTROLLER_MODEL_USB_PIIX3_UHCI;
this particular condition cannot be met - IOW: DEADCODE
Since the code isn't changing cont->model - we could just remove the
@model local and move the DEFAULT into the return 2 leaving some sort of
comment (perhaps) that DEFAULT == PIIX3_UHCI?
I'll just move the DEFAULT: case to be a fallthrough to the PIIX3_UHCI
case instead, and delete that if (...).
> @@ -540,6 +550,16 @@
qemuDomainDeviceCalculatePCIConnectFlags(virDomainDeviceDefPtr dev,
>
> case VIR_DOMAIN_CONTROLLER_TYPE_SCSI:
> switch ((virDomainControllerModelSCSI) cont->model) {
> + case VIR_DOMAIN_CONTROLLER_MODEL_SCSI_DEFAULT:
> + /* XXX it isn't clear that this is the right
> + * thing to do here. We probably ought to
> + * return 0, but that breaks test suite right
> + * now because we call this method before we
> + * have turned the _DEFAULT case into a
> + * specific choice
> + */
> + return pciFlags;
> +
I think it was Laine that asked at one time about re-ordering things -
/me taking a brief look caused my head to spin ;-) and I think this is a
fine alternative (at least for now) until someone gets the gumption to
attempt to fix it.
Laine suggested changing code so that it assigned USB addresses before
assigning PCI addresses, but that does not work. We unfortunately don't
assign default controller models until much later.
> case VIR_DOMAIN_CONTROLLER_MODEL_SCSI_VIRTIO_SCSI:
> return virtioFlags;
[...]
For what's here w/ the issue Coverity noted handled... Adding setting
the model to *_DEFAULT is an add on if it's done...
Regards,
Daniel
--
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org -o- https://fstop138.berrange.com :|
|: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|