Re: [PATCH] testsuite/smokey/psostests: Avoid timeouts in over virtual environments

On Mon, 2026-05-04 at 09:33 +0200, Jan Kiszka wrote:
> From: Jan Kiszka <[email protected]>
> 
> This was seen over QEMU:
> 
> 2026-05-03T09:42:36 [2] at mq-3.c:31
> 2026-05-03T09:42:36    0\"071.079| BUG in __traceobj_assert_failed(): [TSKA] trace assertion failed:
> 2026-05-03T09:42:36               mq-3.c:32 => \"ret == 0\"
> 2026-05-03T09:42:36 [1] at mq-3.c:27
> 2026-05-03T09:42:36 test psostests_mq3 failed: 256
> 
> It's most likely that we ran into a timeout here due to QEMU not being
> able to deliver the messages to all receivers in time. Just wait
> infinitely in such cases to avoid spurious testsuite errors.
> 
> Signed-off-by: Jan Kiszka <[email protected]>
> ---
> 
> Was not worth to bother the AI in this case.
> 
>  testsuite/smokey/psostests/mq-3.c | 10 ++++++++--
>  1 file changed, 8 insertions(+), 2 deletions(-)
> 
> diff --git a/testsuite/smokey/psostests/mq-3.c b/testsuite/smokey/psostests/mq-3.c
> index 1171a70544..4b6f2dcf35 100644
> --- a/testsuite/smokey/psostests/mq-3.c
> +++ b/testsuite/smokey/psostests/mq-3.c
> @@ -1,6 +1,8 @@
>  // SPDX-License-Identifier: GPL-2.0
> +#include <stdbool.h>
>  #include <stdio.h>
>  #include <stdlib.h>
> +#include <string.h>
>  #include <copperplate/traceobj.h>
>  #include <psos/psos.h>
>  
> @@ -12,6 +14,8 @@ static int tseq[] = {
>  
>  static u_long tidA, tidB, qid;
>  
> +static bool on_vm;
> +
>  static void task_A(u_long a0, u_long a1, u_long a2, u_long a3)
>  {
>  	u_long msgbuf[4];
> @@ -27,7 +31,7 @@ static void task_A(u_long a0, u_long a1, u_long a2, u_long a3)
>  	traceobj_mark(&trobj, 1);
>  
>  	for (n = 0; n < 3; n++) {
> -		ret = q_receive(qid, Q_WAIT, 10, msgbuf);
> +		ret = q_receive(qid, Q_WAIT, on_vm ? 0 : 10, msgbuf);
>  		traceobj_mark(&trobj, 2);
>  		traceobj_assert(&trobj, ret == SUCCESS);
>  		traceobj_assert(&trobj, msgbuf[0] == n + 1);
> @@ -56,7 +60,7 @@ static void task_B(u_long a0, u_long a1, u_long a2, u_long a3)
>  	traceobj_mark(&trobj, 4);
>  
>  	for (n = 0; n < 3; n++) {
> -		ret = q_receive(qid, Q_WAIT, 10, msgbuf);
> +		ret = q_receive(qid, Q_WAIT, on_vm ? 0 : 10, msgbuf);
>  		traceobj_mark(&trobj, 5);
>  		traceobj_assert(&trobj, ret == SUCCESS);
>  		traceobj_assert(&trobj, msgbuf[0] == n + 1);
> @@ -75,6 +79,8 @@ int main(int argc, char *const argv[])
>  	u_long args[] = { 1, 2, 3, 4 }, msgbuf[4], count;
>  	int ret, n;
>  
> +	on_vm = argc > 1 && strcmp(argv[1], "--vm") == 0;
> +

That is the synchronization point to the other thread / patch. I'm quite
sure this never worked as expected. Looking at psostest.c, --vm is never
set.

The answer to the question in the other thread is probably: Yes, there
were issues with this test on VMs.

Let me try to fix the cmdline argument forwarding.

>  	traceobj_init(&trobj, argv[0], sizeof(tseq) / sizeof(int));
>  
>  	ret = q_create("QUEUE", Q_NOLIMIT, 0, &qid);
> -- 
> 2.47.3