Re: [PATCH libvirt v1] tests: add capabilities for QEMU 6.0.0 on s390x

On 5/13/21 11:53 AM, Andrea Bolognani wrote: > On Thu, May 13, 2021 at 11:39:24AM +0200, Michal Prívozník wrote: >> On 5/12/21 6:11 PM, Andrea Bolognani wrote: >>> Overall looks reasonable, but comparing the computed capabilities >>> with those for QEMU 5.2 highlights a couple of changes that I'm not >>> so sure about, specifically >>> >>> --- tests/qemucapabilitiesdata/caps_5.2.0.s390x.xml 2021-05-12 >>> 10:52:29.826021415 +0200 >>> +++ tests/qemucapabilitiesdata/caps_6.0.0.s390x.xml 2021-05-12 >>> 17:39:47.778445212 +0200 >>> @@ -87,7 +87,6 @@ >>> <flag name='disk-share-rw'/> >>> <flag name='iscsi.password-secret'/> >>> <flag name='dump-completed'/> >>> - <flag name='virtio-gpu-ccw'/> >>> @@ -115,7 +114,6 @@ >>> <flag name='iothread.poll-max-ns'/> >>> <flag name='egl-headless.rendernode'/> >>> <flag name='memory-backend-file.align'/> >>> - <flag name='memory-backend-file.pmem'/> >>> >>> Because of the former I would expect the s390x-ccw-graphics xml2argv >>> test to fail, but it looks like we don't do a good job at validating >>> the <video> element and that QEMU_CAPS_DEVICE_VIRTIO_GPU_CCW in >>> particular is completely unused. >>> >>> As for the latter there doesn't seem to be any test coverage outside >>> of x86_64, so perhaps it would be a good idea to look into that? >> pmem was reported by QEMU even if it was later denied. This was fixed by: >> >> https://gitlab.com/qemu-project/qemu/-/commit/def835f0da0d153b397071e6bb8... >> >> qemu.git $ git describe --contains def835f0da0d153b397071e6bb8f2b46f51f96b4 >> v6.0.0-rc0~77^2 > Okay for the pmem part - it was reported as available on s390x even > though it probably never worked. > >> And it's the same story with virtio-gpu-ccw: >> >> https://gitlab.com/qemu-project/qemu/-/commit/adcf33a504de29feb720736051d... >> >> qemu.git $ git describe --contains adcf33a504de29feb720736051dc32889314c9e6 >> v6.0.0-rc1~5^2~1 > But virtio-gpu-ccw surely should be reported as available on s390x? > Perhaps I'm missing something obvious in the commit you pointed me > to and I should grab another coffee :) > I admit, I don't know. I thought that virtio-gpu was also misleadingly reported as supported. But as you pointed out it's not the case. Shalini, can you chime in?

Michal