On Wed, Jan 17, 2024 at 09:02:12 -0500, Andrea Bolognani wrote:
> On Wed, Jan 17, 2024 at 01:18:55PM +0100, Peter Krempa wrote:
> > On Wed, Jan 17, 2024 at 10:54:39 +0100, Andrea Bolognani wrote:
> > > All of these are either a subset of other tests, or provide
> > > coverage for scenarios that are not really possible: for all
> > > versions of QEMU that we support, the virt machine type has a
> > > hard dependency on the generic PCIe controller, which means
> > > that we will never need to fall back to virtio-mmio.
> > >
> > > - /* test default config if pcie bus is not available */
> > > - DO_TEST_FULL("aarch64-virt-virtio", "-MMIO.aarch64.latest",
> > > - ARG_CAPS_ARCH, "aarch64",
> > > - ARG_CAPS_VER, "latest",
> > > - ARG_QEMU_CAPS_DEL,
> > > - QEMU_CAPS_OBJECT_GPEX, QEMU_CAPS_DEVICE_PCI_BRIDGE,
> > > - QEMU_CAPS_DEVICE_DMI_TO_PCI_BRIDGE, QEMU_CAPS_DEVICE_IOH3420,
> > > - QEMU_CAPS_LAST, ARG_END);
> > >
> > > - /* The generic pcie bridge emulation device can be compiled out of qemu. */
> > > - DO_TEST_CAPS_ARCH_LATEST_FULL("riscv64-virt", "riscv64",
> > > - ARG_QEMU_CAPS_DEL,
> > > - QEMU_CAPS_OBJECT_GPEX,
> > > - QEMU_CAPS_LAST);
> >
> > I'm not sure about riscv64, but I'm fairly sure we do support non-virt
> > arm machines and it seems that GPEX can be compiled out if you compile
> > out that machine.
>
> You're right in saying that it's possible to obtain a GPEX-less
> aarch64 QEMU binary by compiling out the virt machine type.
>
> However, the only way in which we actually use the capability outside
> of the test suite is to decide whether the virt machine is capable of
> PCI, for which the answer is always "yes".
>
> The remaining usages, in the test suite, are there to confirm that
> the virt machine type will fall back to virtio-mmio is PCI is not
> available. But that's no longer a scenario that can present itself,
> as demonstrated by the fact that we have to go way out of our way to
> trick libvirt into behaving that way, so why have coverage?
>
> > Arguably that would be an insane configuration and might not be worth
> > worrying about, but we had tests for it. Would it make sense to switch
> > to a different machine type?
>
> The original tests were introduced at a time when we still supported
> versions of QEMU where the virt machine type wasn't PCI capable. They
> made sense at the time, but that's no longer the case today.
>
> With regards to other machine types, we already have some coverage
> showing how libvirt handles the absence of PCI support: specifically
> arm-vexpressa9-* and disk-arm-virtio-sd. That seems good enough to
> me.
Reviewed-by: Peter Krempa <pkrempa@xxxxxxxxxx>
_______________________________________________
Devel mailing list -- devel@xxxxxxxxxxxxxxxxx
To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxx
