Re: [PULL v3 00/27] Block patches

[Stefan Hajnoczi]

On Mon, Feb 08, 2021 at 11:02:57AM +0100, Philippe Mathieu-Daudé wrote:
> On 2/8/21 10:27 AM, Stefan Hajnoczi wrote:
> > On Sat, Feb 06, 2021 at 05:03:20PM +0000, Peter Maydell wrote:
> >> On Fri, 5 Feb 2021 at 22:53, Peter Maydell <peter.maydell@linaro.org> 
> >> wrote:
> >>>
> >>> On Fri, 5 Feb 2021 at 16:45, Stefan Hajnoczi <stefanha@redhat.com> wrote:
> >>>>
> >>>> The following changes since commit 
> >>>> e2c5093c993ef646e4e28f7aa78429853bcc06ac:
> >>>>
> >>>>   iotests: 30: drop from auto group (and effectively from make check) 
> >>>> (2021-02-05 15:16:13 +0000)
> >>>>
> >>>> are available in the Git repository at:
> >>>>
> >>>>   https://gitlab.com/stefanha/qemu.git tags/block-pull-request
> >>>>
> >>>> for you to fetch changes up to b07011f375bda3319cf72eee7cb18d310078387b:
> >>>>
> >>>>   docs: fix Parallels Image "dirty bitmap" section (2021-02-05 16:36:36 
> >>>> +0000)
> >>>>
> >>>> ----------------------------------------------------------------
> >>>> Pull request
> >>>>
> >>>> v3:
> >>>>  * Replace {0} array initialization with {} to make clang happy [Peter]
> >>>>
> >>>> ----------------------------------------------------------------
> >>>
> >>>
> >>> Fails 'make check' on s390x host:
> >>
> >> I gave this a rerun to check it was reproducible (it is) and realised
> >> I missed what looks like an important line in the log. As usual,
> >> trying to disentangle which lines of a parallel make check correspond
> >> to the failure is pretty tricky, but the lines
> >>  Type 'remote-pcihost' is missing its parent 'pcie-host-bridge'
> >>
> >> are probably the proximate causes of the assertion failures.
> >>
> >> MALLOC_PERTURB_=${MALLOC_PERTURB_:-$(( ${RANDOM:-0} % 255 + 1))}
> >> QTEST_QEMU_IMG=./qemu-img
> >> G_TEST_DBUS_DAEMON=/home/ubuntu/qemu/tests/dbus-vmstate-daemon.sh
> >> QTEST_QEMU_BINARY=./qemu-system-rx tests/qtest/qos-test --tap -k
> >> PASS 45 qtest-rx/qmp-cmd-test /rx/qmp/query-memory-size-summary
> >> SKIP
> >> MALLOC_PERTURB_=${MALLOC_PERTURB_:-$(( ${RANDOM:-0} % 255 + 1))}
> >> QTEST_QEMU_IMG=./qemu-img
> >> G_TEST_DBUS_DAEMON=/home/ubuntu/qemu/tests/dbus-vmstate-daemon.sh
> >> QTEST_QEMU_BINARY=./qemu-system-s390x tests/qtest/pxe-test --tap -k
> >> PASS 46 qtest-rx/qmp-cmd-test /rx/qmp/query-memory-devices
> >> Type 'remote-pcihost' is missing its parent 'pcie-host-bridge'
> >> PASS 47 qtest-rx/qmp-cmd-test /rx/qmp/query-replay
> >> PASS 48 qtest-rx/qmp-cmd-test /rx/qmp/query-yank
> >> PASS 49 qtest-rx/qmp-cmd-test /rx/qmp/query-name
> >> PASS 50 qtest-rx/qmp-cmd-test /rx/qmp/query-iothreads
> >> PASS 51 qtest-rx/qmp-cmd-test /rx/qmp/query-fdsets
> >> PASS 52 qtest-rx/qmp-cmd-test /rx/qmp/query-command-line-options
> >> PASS 53 qtest-rx/qmp-cmd-test /rx/qmp/query-acpi-ospm-status
> >> PASS 54 qtest-rx/qmp-cmd-test /rx/qmp/object-add-failure-modes
> >> MALLOC_PERTURB_=${MALLOC_PERTURB_:-$(( ${RANDOM:-0} % 255 + 1))}
> >> QTEST_QEMU_IMG=./qemu-img
> >> G_TEST_DBUS_DAEMON=/home/ubuntu/qemu/tests/dbus-vmstate-daemon.sh
> >> QTEST_QEMU_BINARY=./qemu-system-s390x tests/qtest/test-netfilter --tap
> >> -k
> >> Type 'remote-pcihost' is missing its parent 'pcie-host-bridge'
> >> socket_accept failed: Resource temporarily unavailable
> >> socket_accept failed: Resource temporarily unavailable
> >> **
> >> ERROR:../../tests/qtest/libqtest.c:308:qtest_init_without_qmp_handshake:
> >> assertion failed: (s->fd >= 0 && s->qmp_fd >= 0)
> >> **
> >> ERROR:../../tests/qtest/libqtest.c:308:qtest_init_without_qmp_handshake:
> >> assertion failed: (s->fd >= 0 && s->qmp_fd >= 0)
> >> ../../tests/qtest/libqtest.c:181: kill_qemu() detected QEMU death from
> >> signal 6 (Aborted) (core dumped)
> >> ../../tests/qtest/libqtest.c:181: kill_qemu() detected QEMU death from
> >> signal 6 (Aborted) (core dumped)
> >> ERROR qtest-s390x/pxe-test - Bail out!
> >> ERROR:../../tests/qtest/libqtest.c:308:qtest_init_without_qmp_handshake:
> >> assertion failed: (s->fd >= 0 && s->qmp_fd >= 0)
> >> ERROR qtest-s390x/test-netfilter - Bail out!
> >> ERROR:../../tests/qtest/libqtest.c:308:qtest_init_without_qmp_handshake:
> >> assertion failed: (s->fd >= 0 && s->qmp_fd >= 0)
> >> Makefile.mtest:3113: recipe for target 'run-test-387' failed
> >> make: *** [run-test-387] Error 1
> >> make: *** Waiting for unfinished jobs....
> >> Makefile.mtest:3121: recipe for target 'run-test-388' failed
> > 
> > Hi Elena and Jag,
> > Please take a look at this QOM failure. I guess remote-pcihost is being
> > built but pcie-host-bridge is missing from the s390x-softmmu target.
> 
> Fix suggested here:
> https://www.mail-archive.com/qemu-block@nongnu.org/msg80536.html
> 
> But beside the fix what would be better is to restrict this feature
> where it makes sense (we are having hard time building/testing all
> features, better enable new ones where they are used).
> 
> Would it be enough to enable this feature on X86 hosts/targets for
> mainstream CI?

Trying to check if I understand correctly:

Instead of writing configure/meson rules that enable the feature
whenever the dependencies are satisfied (KVM and PCI), explicitly enable
it on x86 targets only. The rationale is that it's not being used and
hasn't been tested on non-x86 targets, so it's not really supported
there yet.

?

I think that approach makes sense too. We just need to make a clear
distinction in the configure/meson rules which conditionals are true
dependencies and which indicate support. In other words x86 isn't
actually a dependency, it's just a condition expressing that the
architecture is supported and others aren't yet. That way people will be
able to modify the logic in the future without accidentally dropping a
true dependency condition, etc.

Stefan

signature.asc
Description: PGP signature