Hi Daniel,
On 5/12/25 10:13, Daniel P. Berrangé wrote:
On Mon, May 12, 2025 at 01:36:33PM +0100, Gustavo Romero wrote:
Add a functional test, aarch64_hotplug_pci, to exercise PCI hotplug and
hot-unplug on arm64. Currently, the aarch64 'virt' machine uses the PCIe
native controller and does not support ACPI-based hotplugging. However,
since support is planned, this test sets 'acpi=force' and specifies an
EDK2 firmware image in advance. This is harmless and prepares for future
ACPI support.
Based on what's written here, it sounds like 'acpi=force' will result
in this test case changing from exercising PCI hotplug, to exercising
ACPI hotplug at some arbitrary point in the future. This isn't what
I'd call a harmless future change.
Yeah, your reading is correct.
If the 'virt' machine ever switches from PCI to ACPI hotplug, this
will have to be tied to a machine type version. As such QEMU will
need to support /both/ PCI and ACPI hotplug for a long period of
time, and by implication we would need to be testing both.
I'd think this test should be written to exclusively test the current
PCI hotplug, and at a future date when ACPI hotplug is introduced,
then a 2nd test method can be addded, albeit with a lot of common
code shared.
I have to agree :) I removed acpi=force in v3. I'll add the other
test variants as the right time comes. Thanks.
Cheers.
Gustavo
