Re: [PATCH v4 0/4] avocado-qemu: New SMMUv3 and intel IOMMU tests

[Willian Rampazzo]

On Mon, Jul 5, 2021 at 6:20 PM Philippe Mathieu-Daudé <philmd@redhat.com> wrote:
>
> On 7/5/21 11:10 PM, Willian Rampazzo wrote:
> > Hi Eric,
> >
> > On Mon, Jul 5, 2021 at 4:55 AM Eric Auger <eric.auger@redhat.com> wrote:
> >>
> >> Hi Wainer,
> >>
> >> On 7/1/21 1:22 AM, Wainer dos Santos Moschetta wrote:
> >>> Hi,
> >>>
> >>> On 6/29/21 5:17 PM, Eric Auger wrote:
> >>>> Hi Cleber, all,
> >>>>
> >>>> On 6/29/21 4:36 PM, Eric Auger wrote:
> >>>>> This series adds ARM SMMU and Intel IOMMU functional
> >>>>> tests using Fedora cloud-init images.
> >>>>>
> >>>>> ARM SMMU tests feature guests with and without RIL
> >>>>> (range invalidation support) using respectively fedora 33
> >>>>> and 31.  For each, we test the protection of virtio-net-pci
> >>>>> and virtio-block-pci devices. Also strict=no and passthrough
> >>>>> modes are tested. So there is a total of 6 tests.
> >>>>>
> >>>>> The series applies on top of Cleber's series:
> >>>>> - [PATCH 0/3] Acceptance Tests: support choosing specific
> >>>>>
> >>>>> Note:
> >>>>> - SMMU tests 2, 3, 5, 6 (resp. test_smmu_noril_passthrough and
> >>>>> test_smmu_noril_nostrict) pass but the log reports:
> >>>>> "WARN: Test passed but there were warnings during execution."
> >>>>> This seems due to the lack of hash when fetching the kernel and
> >>>>> initrd through fetch_asset():
> >>>>> WARNI| No hash provided. Cannot check the asset file integrity.
> >>>> I wanted to emphasize that point and wondered how we could fix that
> >>>> issue. Looks a pity the tests get tagged as WARN due to a lack of sha1.
> >>>> Any advice?
> >>>
> >>> As Willian mentioned somewhere, to supress the WARN you can pass the
> >>> kernel and initrd checksums (sha1) to the fetch_asset() method.
> >>>
> >>> Below is an draft implementation. It would need to fill out the
> >>> remaining checksums and adjust the `smmu.py` tests.
> >>>
> >>> - Wainer
> >>>
> >>> ----
> >>>
> >>> diff --git a/tests/acceptance/avocado_qemu/__init__.py
> >>> b/tests/acceptance/avocado_qemu/__init__.py
> >>> index 00eb0bfcc8..83637e2654 100644
> >>> --- a/tests/acceptance/avocado_qemu/__init__.py
> >>> +++ b/tests/acceptance/avocado_qemu/__init__.py
> >>> @@ -312,6 +312,8 @@ class LinuxDistro:
> >>>                  {'checksum':
> >>> 'e3c1b309d9203604922d6e255c2c5d098a309c2d46215d8fc026954f3c5c27a0',
> >>>                  'pxeboot_url':
> >>> "https://archives.fedoraproject.org/pub/archive/fedora/";
> >>> "linux/releases/31/Everything/x86_64/os/images/pxeboot/",
> >>> +                'pxeboot_initrd_chksum':
> >>> 'dd0340a1b39bd28f88532babd4581c67649ec5b1',
> >>> +                'pxeboot_vmlinuz_chksum':
> >>> '5b6f6876e1b5bda314f93893271da0d5777b1f3c',
> >> where did you get the checksum? I don't see any at the URL? Did you
> >> generate it yourself?
> >
> > It is possible to use the hash you generate from the downloaded file.
> >
> > While I was reviewing this series, I thought it makes more sense to
> > have Wainer's path applied first and then have your changes. I did
> > this here, with the addition of myu suggestions in the series:
> > https://gitlab.com/willianrampazzo/qemu/-/commits/test_eric_auger_v5.
>
> Off-list review is a bit unhandy (in particular when asked on the list).
>
> Why don't you post your improvements as v5? I don't think Eric will be
> offended: this is the opposite, you are helping him to get his patches
> merged ;)

Oh, I did review each of his patches in the list and also already made
the changes to speed up the process :)

He mentioned today to me that his series is still depending on one
from Cleber that was not merged yet, so we need to wait for that.

>
> > Feel free to pick it and resend a new version.
> >
> > Wainer, check if you agree with the changes to your patch and ack it.
> >
> > Regards,
>