Re: [PATCH] RAW6: Do not allow set IPV6_CHECKSUM for ICMPv6 socket

[Wang Chen <wangchen@cn.fujitsu.com>]

David Miller said the following on 2008-4-18 19:09:
> From: Wang Chen <wangchen@cn.fujitsu.com>
> Date: Fri, 18 Apr 2008 18:32:46 +0800
> 
>> As RFC3542 mentions: An attempt to set IPV6_CHECKSUM for an
>> ICMPv6 socket will fail.
>> Add a check for that in do_rawv6_setsockopt().
>>
>> Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
> 
> Enforcing these kinds of things we've allowed for so many years is
> rather pointless, especially if it breaks real applications.  Which I
> believe it does in this case.  The standards simply do not matter.
> 
> I seem to recall that traceroute6 in iputils does exactly this.
> 

Enforcing this can forbid disabling checksum for icmpv6 socket.
The real applications, such as ping6 & traceroute6, are just doing
what is default in kernel.
Why not remove the RFC-breaking code from applications?

See the source:
kernel->rawv6_init_sk()
---
	case IPPROTO_ICMPV6:
		rp->checksum = 1;
		rp->offset   = 2;
		break;
---

ping6->main()
---
	csum_offset = 2;
	sz_opt = sizeof(int);

	err = setsockopt(icmp_sock, SOL_RAW, IPV6_CHECKSUM, &csum_offset, sz_opt);
	if (err < 0) {
		perror("setsockopt(RAW_CHECKSUM)");
		exit(2);
	}
---

traceroute6->main()
---
	on = 2;
	if (setsockopt(icmp_sock, SOL_RAW, IPV6_CHECKSUM, &on, sizeof(on)) < 0) {
		perror("setsockopt(RAW_CHECKSUM)");
		exit(2);
	}
---