Re: [LTP] [PATCH v1] recvmsg01: Refactor using new LTP API

[Petr Vorel <pvorel@suse.cz>]

Hi Wei,

	{
	PF_INET, SOCK_STREAM, 0, iov, 1, (void *)buf, sizeof(buf),
		    &msgdat, -1, (struct sockaddr *)&from, -1, -1,
		    EINVAL, setup1, cleanup1, "invalid socket length"},
.flags is -1 here, it's unsigned int, please use UINT_MAX (the same value, but
more obvious).

> -	for (lc = 0; TEST_LOOPING(lc); ++lc) {
> -		tst_count = 0;
> -		for (testno = 0; testno < TST_TOTAL; ++testno) {
> -			if ((tst_kvercmp(3, 17, 0) < 0)
> -			    && (tdat[testno].flags & MSG_ERRQUEUE)
> -			    && (tdat[testno].type & SOCK_STREAM)) {
> -				tst_resm(TCONF, "skip MSG_ERRQUEUE test, "
> -						"it's supported from 3.17");
> -				continue;
> -			}
...

> +	if ((tst_kvercmp(3, 17, 0) < 0)
> +			&& (tc->flags & MSG_ERRQUEUE)
> +			&& (tc->type & SOCK_STREAM)) {
> +		tst_res(TCONF, "skip MSG_ERRQUEUE test, "
> +				"it's supported from 3.17");

The reason for the MSG_ERRQUEUE on SOCK_STREAM was described in
e5fd512d8 ("skip MSG_ERRQUEUE+tcp test on kernels < 3.17")

The limitation for MSG_ERRQUEUE is only on SOCK_STREAM, but because MSG_ERRQUEUE
is tested only on SOCK_STREAM, it should be enough to check just MSG_ERRQUEUE.
Also, UINT_MAX & MSG_ERRQUEUE is always true (in "invalid socket length") and I
don't think that test should be skipped.

The check could be:

	if (tc->flags != UINT_MAX && tc->flags & MSG_ERRQUEUE
		&& tst_kvercmp(3, 17, 0) < 0) {
		tst_res(TCONF, "MSG_ERRQUEUE support for TCP has been added in 3.17");
	}

Kind regards,
Petr

-- 
Mailing list info: https://lists.linux.it/listinfo/ltp