Re: [kvm-unit-tests PATCH v2 3/4] s390x: uv-guest: remove duplicated checks

[Claudio Imbrenda <imbrenda@linux.ibm.com>]

On Thu,  3 Feb 2022 09:19:34 +0000
Steffen Eiden <seiden@linux.ibm.com> wrote:

> Removing some tests which are done at other points in the code
> implicitly.

I'm not sure I like all of this

> 
> In lib/s390x/uc.c#setup_uv(void) the rc of the qui result is verified
> using asserts.
> The whole test is fenced by lib/s390x/uc.c#os_is_guest(void) that

do you mean "lib/s390x/uv.c#uv_os_is_guest(void)" ?

> checks if SET and REMOVE SHARED is present.
> 
> Signed-off-by: Steffen Eiden <seiden@linux.ibm.com>
> Reviewed-by: Janosch Frank <frankja@linux.ibm.com>
> ---
>  s390x/uv-guest.c | 22 +++++++---------------
>  1 file changed, 7 insertions(+), 15 deletions(-)
> 
> diff --git a/s390x/uv-guest.c b/s390x/uv-guest.c
> index 44ad2154..97ae4687 100644
> --- a/s390x/uv-guest.c
> +++ b/s390x/uv-guest.c
> @@ -69,23 +69,15 @@ static void test_query(void)
>  	cc = uv_call(0, (u64)&uvcb);
>  	report(cc == 1 && uvcb.header.rc == UVC_RC_INV_LEN, "length");
>  
> -	uvcb.header.len = sizeof(uvcb);
> -	cc = uv_call(0, (u64)&uvcb);
> -	report((!cc && uvcb.header.rc == UVC_RC_EXECUTED) ||
> -	       (cc == 1 && uvcb.header.rc == 0x100),
> -		"successful query");
> -

ok fair enough, an unsuccessful query would have caused an assert in
the setup code, but I don't think it hurts, and I think it would be
nice to have for completeness.

>  	/*
> -	 * These bits have been introduced with the very first
> -	 * Ultravisor version and are expected to always be available
> -	 * because they are basic building blocks.
> +	 * BIT_UVC_CMD_QUI, BIT_UVC_CMD_SET_SHARED_ACCESS and
> +	 * BIT_UVC_CMD_SET_SHARED_ACCESS are always present as they

I think you meant BIT_UVC_CMD_REMOVE_SHARED_ACCESS here ?

> +	 * have been introduced with the first Ultravisor version.
> +	 * However, we only need to check for QUI as
> +	 * SET/REMOVE SHARED are used to fence this test to be only
> +	 * executed by protected guests.

also, what happens if only one of the two bits is set? (which is very
wrong). In that scenario, I would like this test to fail, not skip.
this means that we can't rely on uv_os_is_guest to decide whether to
skip this test.

>  	 */
> -	report(test_bit_inv(BIT_UVC_CMD_QUI, &uvcb.inst_calls_list[0]),
> -	       "query indicated");
> -	report(test_bit_inv(BIT_UVC_CMD_SET_SHARED_ACCESS, &uvcb.inst_calls_list[0]),
> -	       "share indicated");
> -	report(test_bit_inv(BIT_UVC_CMD_REMOVE_SHARED_ACCESS, &uvcb.inst_calls_list[0]),
> -	       "unshare indicated");
> +	report(uv_query_test_call(BIT_UVC_CMD_QUI), "query indicated");
>  	report_prefix_pop();
>  }
>