ipv6: bring NLM_DONE out to a separate recv() again

Commit under Fixes optimized the number of recv() calls
needed during RTM_GETROUTE dumps, but we got multiple
reports of applications hanging on recv() calls.
Applications expect that a route dump will be terminated
with a recv() reading an individual NLM_DONE message.

Coalescing NLM_DONE is perfectly legal in netlink,
but even tho reporters fixed the code in respective
projects, chances are it will take time for those
applications to get updated. So revert to old behavior
(for now)?

This is an IPv6 version of commit 460b0d3 ("inet: bring
NLM_DONE out to a separate recv() again").

Reported-by: Maciej Żenczykowski <zenczykowski@gmail.com>
Link: https://lore.kernel.org/all/CANP3RGc1RG71oPEBXNx_WZFP9AyphJefdO4paczN92n__ds4ow@mail.gmail.com
Reported-by: Stefano Brivio <sbrivio@redhat.com>
Link: https://lore.kernel.org/all/20240315124808.033ff58d@elisabeth
Reported-by: Ilya Maximets <i.maximets@ovn.org>
Link: https://lore.kernel.org/all/02b50aae-f0e9-47a4-8365-a977a85975d3@ovn.org
Fixes: 5fc6832 ("ipv6: remove RTNL protection from inet6_dump_fib()")
Tested-by: Ilya Maximets <i.maximets@ovn.org>
Link: https://lore.kernel.org/r/20240618193914.561782-1-kuba@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net/ipv6/ip6_fib.c

@@ -2514,7 +2514,8 @@ int __init fib6_init(void)
 		goto out_kmem_cache_create;
 
 	ret = rtnl_register_module(THIS_MODULE, PF_INET6, RTM_GETROUTE, NULL,
-				   inet6_dump_fib, RTNL_FLAG_DUMP_UNLOCKED);
+				   inet6_dump_fib, RTNL_FLAG_DUMP_UNLOCKED |
+				   RTNL_FLAG_DUMP_SPLIT_NLM_DONE);
 	if (ret)
 		goto out_unregister_subsys;