On Tue, Dec 03, 2024 at 10:38:26AM +0100, Philippe Mathieu-Daudé wrote:
> On 3/12/24 10:26, Philippe Mathieu-Daudé wrote:
> > On 3/12/24 10:18, Daniel P. Berrangé wrote:
> > > On Tue, Dec 03, 2024 at 10:10:35AM +0100, Philippe Mathieu-Daudé wrote:
> > > > When testing with a HVF-only binary, we get:
> > > >
> > > > 3/12 qemu:func-quick+func-aarch64 / func-aarch64-
> > > > version ERROR
> > > > 0.29s exit status 1
> > > > stderr:
> > > > Traceback (most recent call last):
> > > > File "tests/functional/test_version.py", line 22, in
> > > > test_qmp_human_info_version
> > > > self.vm.launch()
> > > > File "machine/machine.py", line 461, in launch
> > > > raise VMLaunchFailure(
> > > > qemu.machine.machine.VMLaunchFailure: ConnectError: Failed to
> > > > establish session: EOFError
> > > > Exit code: 1
> > > > Command: build/qemu-system-aarch64 -display none -vga
> > > > none - chardev socket,id=mon,fd=5 -mon chardev=mon,mode=control
> > > > -machine none -nodefaults
> > > > Output: qemu-system-aarch64: No accelerator selected and
> > > > no default accelerator available
> > > >
> > > > Explicit the QTest accelerator to be able to run the HMP command.
> > > >
> > > > Signed-off-by: Philippe Mathieu-Daudé <phi...@linaro.org>
> > > > ---
> > > > tests/functional/test_version.py | 1 +
> > > > 1 file changed, 1 insertion(+)
> > > >
> > > > diff --git a/tests/functional/test_version.py
> > > > b/tests/functional/ test_version.py
> > > > index 3ab3b67f7e3..d3da796991f 100755
> > > > --- a/tests/functional/test_version.py
> > > > +++ b/tests/functional/test_version.py
> > > > @@ -18,6 +18,7 @@ class Version(QemuSystemTest):
> > > > def test_qmp_human_info_version(self):
> > > > self.set_machine('none')
> > > > + self.vm.add_args('-accel', 'qtest')
> > >
> > > IMHO this is wrong. The functional tests are there to test the
> > > real functional behaviour under an actual accelerator not qtest.
> >
> > It works using '-accel hvf'. The issue is:
> >
> > "No accelerator selected and no default accelerator available"
> >
> > So we should select HVF over QTest by default? I tend to not
> > enforce any default because we always get troubles with them,
> > what is today's default is unlikely tomorrow's one.
>
> So by using:
>
> -- >8 --
> diff --git a/system/vl.c b/system/vl.c
> index 54998fdbc7e..2f855d83fbb 100644
> --- a/system/vl.c
> +++ b/system/vl.c
> @@ -2362,6 +2362,7 @@ static void configure_accelerators(const char
> *progname)
> /* Select the default accelerator */
> bool have_tcg = accel_find("tcg");
> bool have_kvm = accel_find("kvm");
> + bool have_hvf = accel_find("hvf");
>
> if (have_tcg && have_kvm) {
> if (g_str_has_suffix(progname, "kvm")) {
> @@ -2374,6 +2375,8 @@ static void configure_accelerators(const char
> *progname)
> accelerators = "kvm";
> } else if (have_tcg) {
> accelerators = "tcg";
> + } else if (have_hvf) {
> + accelerators = "hvf";
> } else {
> error_report("No accelerator selected and"
> " no default accelerator available");
>
> ---
>
> All test suites pass on my HVF-only build directory. If this is
> OK with you then this is also OK for me.
If all the functional tests pass with HVF that is a good stamp of approval
for the quality & usefulness of HVF, and should justify enabling it by
default IMHO.
I might even suggest that we should flip to rank HVF above TCG, on the
principal that HW accelerator is preferrable when available.
I don't think we slip this into 9.2 at this stage in -rcNN, but perhaps
do this early in 10.0 and propose for stable (TCG preferred over HVF),
then flip the TCG & HVF ordering to prefer HVF for remainder of 10.x
and the future
With regards,
Daniel
--
|: https://berrange.com -o- https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org -o- https://fstop138.berrange.com :|
|: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|
