> On Jul 12, 2021, at 10:07 AM, Andrew Jones <drjones@redhat.com> wrote:
> 
> On Mon, Jul 12, 2021 at 05:51:55PM +0100, Andre Przywara wrote:
>> On Fri,  2 Jul 2021 17:31:18 +0100
>> Alexandru Elisei <alexandru.elisei@arm.com> wrote:
>> 
>> Hi,
>> 
>>> The arm64 tests can be run under kvmtool, which doesn't emulate a
>>> chr-testdev device. In preparation for adding run script support for
>>> kvmtool, print the test exit status so the scripts can pick it up and
>>> correctly mark the test as pass or fail.
>>> 
>>> Signed-off-by: Alexandru Elisei <alexandru.elisei@arm.com>
>>> ---
>>> lib/chr-testdev.h |  1 +
>>> lib/arm/io.c      | 10 +++++++++-
>>> lib/chr-testdev.c |  5 +++++
>>> 3 files changed, 15 insertions(+), 1 deletion(-)
>>> 
>>> diff --git a/lib/chr-testdev.h b/lib/chr-testdev.h
>>> index ffd9a851aa9b..09b4b424670e 100644
>>> --- a/lib/chr-testdev.h
>>> +++ b/lib/chr-testdev.h
>>> @@ -11,4 +11,5 @@
>>>  */
>>> extern void chr_testdev_init(void);
>>> extern void chr_testdev_exit(int code);
>>> +extern bool chr_testdev_available(void);
>>> #endif
>>> diff --git a/lib/arm/io.c b/lib/arm/io.c
>>> index 343e10822263..9e62b571a91b 100644
>>> --- a/lib/arm/io.c
>>> +++ b/lib/arm/io.c
>>> @@ -125,7 +125,15 @@ extern void halt(int code);
>>> 
>>> void exit(int code)
>>> {
>>> -	chr_testdev_exit(code);
>>> +	if (chr_testdev_available()) {
>>> +		chr_testdev_exit(code);
>>> +	} else {
>>> +		/*
>>> +		 * Print the test return code in the format used by chr-testdev
>>> +		 * so the runner script can parse it.
>>> +		 */
>>> +		printf("\nEXIT: STATUS=%d\n", ((code) << 1) | 1);
>> 
>> It's more me being clueless here rather than a problem, but where does
>> this "EXIT: STATUS" line come from? In lib/chr-testdev.c I see "%dq",
>> so it this coming from QEMU (but I couldn't find it in there)?
>> 
>> But anyways the patch looks good and matches what PPC and s390 do.
> 
> I invented the 'EXIT: STATUS' format for PPC, which didn't/doesn't have an
> exit code testdev. Now that it has also been adopted by s390 I guess we've
> got a kvm-unit-tests standard to follow for arm :-)

I was unaware of this “standard” and I mistakenly used a different format
for x86, in case someone wants to fix it. [1]

[1] https://gitlab.com/kvm-unit-tests/kvm-unit-tests/-/commit/5747945371b47c51cb16187a26111d06f58f06b2