sys/linux: assorted socket_inet fixes

[a-nogikh]

See individual commits.

[a-nogikh]

The tests are failing with:

2025/09/12 13:01:21 [FATAL] generating resource type from GroupType with 3 elements

Looks like the syz_create_resource approach does not make syz-trace2syz happy:

syzkaller/tools/syz-trace2syz/proggen/proggen.go

Lines 439 to 448 in e2beed9

	case *parser.GroupType:
	if len(a.Elems) == 1 {
	// For example: 5028 ioctl(3, SIOCSPGRP, [0]) = 0
	// last argument is a pointer to a resource. Strace will output a pointer to
	// a number x as [x].
	res := prog.MakeResultArg(syzType, dir, nil, syzType.Default())
	ctx.returnCache.cache(syzType, a.Elems[0], res)
	return res
	}
	log.Fatalf("generating resource type from GroupType with %d elements", len(a.Elems))

We could adjust the tests and silence the error, but it means the tool becomes (even more) broken.

Cc @dvyukov

[dvyukov]

We could adjust the tests and silence the error, but it means the tool becomes (even more) broken.

Adjust/disable tests. To the best of my knowledge nobody never used this tool.
It will be broken in such ways always by design. It should have not used strace in the first place.

[a-nogikh]

Dropped the tests, now all CI checks pass.

The only concern is that bind$inet is mentioned > 700 times in our corpus, so syzbot will have to rediscover a lot of its current corpus programs, but maybe it's not that bad.

[dvyukov]

sys/linux: use strinoz in setsockopt

stringnoz

[dvyukov]

You also had some seed that uses these calls. Do you want to add it?

[a-nogikh]

You also had some seed that uses these calls. Do you want to add it?

Not yet, still not sure if the seed brings any actual value (it doesn't help discover the crash during fuzzing, not sure about it giving extra coverage yet).

[a-nogikh]

@dvyukov PTAL