ares_getnameinfo not working where getnameinfo does

[Asmod4n]

I have a strange issue, running the following code works without an issue.

struct sockaddr_storage ss = {0};
struct sockaddr_in *sa_in = (struct sockaddr_in *) &ss;
inet_pton(AF_INET, "185.199.111.153", &(sa_in->sin_addr));
ss.ss_family = AF_INET;
char host[1024];
char service[20];
getnameinfo((struct sockaddr*)&ss, sizeof ss, host, sizeof host, service, sizeof service, 0);
printf("Host: %s\n", host);

when i am replacing getnameinfo with ares_getnameinfo i am getting the error DNS server does not implement requested operation (ARES_ENOTIMP)

I am running kali linux (rolling release) under wsl 2 on a windows 11 pc, my DNS Server is dnsmasq with unbound with root servers configured. Everything is updated to the latest software.

ares_getaddrinfo works just fine, i have not set any ares_init options except ARES_OPT_SOCK_STATE_CB and its required fields.

[bradh352]

Your sizeof ss is technically wrong. The Linux manpage for getaddrinfo says:

The sockaddr structure sa should point to either a sockaddr_in, sockaddr_in6 or sockaddr_dl structure (for IPv4, IPv6 or link-layer respectively) that is salen bytes long.

Interestingly the OpenGroup docs don't even mention salen:
https://pubs.opengroup.org/onlinepubs/009604599/functions/getnameinfo.html

And even more interestingly. FreeBSD goes into further detail which states on FreeBSD, what you are doing is valid:
https://man.freebsd.org/cgi/man.cgi?query=getnameinfo&sektion=3&format=html

So, at least as it stands now, for AF_INET it should be struct sockaddr_in in size, but yours is struct sockaddr_storage in size. I wouldn't be surprised if this wasn't portable to other systems with your usage (not just c-ares).

c-ares has a sanity check to verify the exact size of the expected data structure is provided, however we might be convinced that we should loosen that up and follow what FreeBSD documents in particular.

[Asmod4n]

Ah, just giving it the correct salen worked.
I've seen my usage in quite a few manpages and code examples after IPv6 got introduced and didn't have an issue with it yet.

[bradh352]

went ahead and loosened up the validation since FreeBSD at least documents it, should be fixed in 48e8cd2