On 2024/12/28 5:36, Philippe Mathieu-Daudé wrote:
> On 23/12/24 23:16, Phil Dennis-Jordan wrote:
>
>> Known issues:
>> - Keyboard and mouse/tablet input is laggy. The reason for this is
>> that macOS's XHCI driver seems to expect interrupter mapping to
>> be disabled when MSI/MSI-X is unavailable. I have found a
>> workaround but discovered a bunch of other XHCI spec non-compliance
>> in the process, so I'm fixing all of those in a separate patch
>> set.
>
> Should we wait for your XHCI fixes series to get reviewed first
> so I can base this series on top (removing this comment) when merging?
This series no longer cleanly applies due to patches merged early so
please rebase it.
I've already done this locally, but I need to incorporate some of Phil M-D's feedback, I don't think there's much point posting a v17 to the list before those points are resolved.
I need to set up an aarch64 Linux host to test the build with --enable-kvm --enable-tcg to disprove or confirm my worry that
select ARM_GICV3 if !KVM
would deselect ARM_GICV3 even though it's needed for the TCG accelerator. And if that suspicion turns out to be correct, I'll need to find a good alternative.
select ARM_GICV3 if TCG || HVF || NVMM || WHPX || XEN_EMU || XEN
seems… awkward.
It will be nice if you also:
1. rebase this vmapple series onto the xHC series and
2. move "[PATCH v3 6/6] hw/vmapple: XHCI controller's interrupt mapping
workaround for macOS" into this vmapple series
xHC changes that are not specific to vmapple may be merged early in this
way. I will be also able to test all vmapple code at once.
I'm going to combine the two series and put the generic (standards compliant) XHCI patches first. The final XHCI patch has been squashed into the final vmapple patch.
Please use Based-on: tag when you post a series that based on another.
It will create a Git tree that I can use for testing at:
https://patchew.org/QEMU/
docs/devel/submitting-a-patch.rst describes the details.
Thanks!
Regards,
Akihiko Odaki