On Mon, Jan 20, 2025 at 04:43:10PM +0000, Alexandru Elisei wrote:
> kvmtool, unless told otherwise, will do its best to make sure that a kernel
> successfully boots in a virtual machine. Among things like automatically
> creating a rootfs, it also adds extra parameters to the kernel command
> line. This is actively harmful to kvm-unit-tests, because some tests parse
> the kernel command line and they will fail if they encounter the options
> added by kvmtool.
>
> Fortunately for us, kvmtool commit 5613ae26b998 ("Add --nodefaults command
> line argument") addded the --nodefaults kvmtool parameter which disables
added
> all the implicit virtual machine configuration that cannot be disabled by
> using other parameters, like modifying the kernel command line. Always use
> --nodefaults to allow a test to run.
>
> kvmtool can be too verbose when running a virtual machine, and this is
> controlled with parameters. Add those to the default kvmtool command line
> to reduce this verbosity to a minimum.
>
> Before:
>
> $ vm run arm/selftest.flat --cpus 2 --mem 256 --params "setup smp=2 mem=256"
> Info: # lkvm run -k arm/selftest.flat -m 256 -c 2 --name guest-5035
> Unknown subtest
>
> EXIT: STATUS=127
> Warning: KVM compatibility warning.
> virtio-9p device was not detected.
> While you have requested a virtio-9p device, the guest kernel did not initialize it.
> Please make sure that the guest kernel was compiled with CONFIG_NET_9P_VIRTIO=y enabled in .config.
> Warning: KVM compatibility warning.
> virtio-net device was not detected.
> While you have requested a virtio-net device, the guest kernel did not initialize it.
> Please make sure that the guest kernel was compiled with CONFIG_VIRTIO_NET=y enabled in .config.
> Info: KVM session ended normally.
>
> After:
>
> $ vm run arm/selftest.flat --nodefaults --network mode=none --loglevel=warning --cpus 2 --mem 256 --params "setup smp=2 mem=256"
On riscv I've noticed that with --nodefaults if I don't add parameters
with --params then kvmtool segfaults. I have to add --params "" to
avoid it. Does that also happen on arm? Anyway, that's something we
should fix in kvmtool rather than workaround it here.
> PASS: selftest: setup: smp: number of CPUs matches expectation
> INFO: selftest: setup: smp: found 2 CPUs
> PASS: selftest: setup: mem: memory size matches expectation
> INFO: selftest: setup: mem: found 256 MB
> SUMMARY: 2 tests
>
> EXIT: STATUS=1
>
> Note that KVMTOOL_DEFAULT_OPTS can be overwritten by an environment
> variable with the same name, but it's not documented in the help string for
> run_tests.sh. This has been done on purpose, since overwritting
> KVMTOOL_DEFAULT_OPTS should only be necessary for debugging or development
> purposes.
>
> Signed-off-by: Alexandru Elisei <alexandru.elisei@xxxxxxx>
> ---
> scripts/runtime.bash | 13 ++++++++++++-
> 1 file changed, 12 insertions(+), 1 deletion(-)
>
> diff --git a/scripts/runtime.bash b/scripts/runtime.bash
> index 55d58eef9c7c..abfd1e67b2ef 100644
> --- a/scripts/runtime.bash
> +++ b/scripts/runtime.bash
> @@ -2,6 +2,17 @@
> : "${MAX_SMP:=$(getconf _NPROCESSORS_ONLN)}"
> : "${TIMEOUT:=90s}"
>
> +# The following parameters are enabled by default when running a test with
> +# kvmtool:
> +# --nodefaults: suppress VM configuration that cannot be disabled otherwise
> +# (like modifying the supplied kernel command line). Tests that
> +# use the command line will fail without this parameter.
> +# --network mode=none: do not create a network device. kvmtool tries to help the
> +# user by automatically create one, and then prints a warning
> +# when the VM terminates if the device hasn't been initialized.
> +# --loglevel=warning: reduce verbosity
> +: "${KVMTOOL_DEFAULT_OPTS:="--nodefaults --network mode=none --loglevel=warning"}"
> +
> PASS() { echo -ne "\e[32mPASS\e[0m"; }
> SKIP() { echo -ne "\e[33mSKIP\e[0m"; }
> FAIL() { echo -ne "\e[31mFAIL\e[0m"; }
> @@ -103,7 +114,7 @@ function run()
> opts="-smp $smp $qemu_opts"
> ;;
> kvmtool)
> - opts="--cpus $smp $kvmtool_opts"
> + opts="$KVMTOOL_DEFAULT_OPTS --cpus $smp $kvmtool_opts"
> ;;
> esac
>
> --
> 2.47.1
>
Otherwise,
Reviewed-by: Andrew Jones <andrew.jones@xxxxxxxxx>
