IPv6 sockaddr issues

[Russ Allbery]

Jeffrey M Vinocur <jeff at litech.org> writes:

> We're using getaddrinfo(3) in the IPv6 case.  It returns (by writing
> into an argument "struct addrinfo **r") a struct addrinfo, defined as:

>        struct addrinfo {
>            int     ai_flags;
>            int     ai_family;
>            int     ai_socktype;
>            int     ai_protocol;
>            size_t  ai_addrlen;
>            struct sockaddr *ai_addr;
>            char   *ai_canonname;
>            struct addrinfo *ai_next;
>        };

> I want to copy the contents of r->ai_addr into local storage before
> calling freeaddrinfo(r).  But despite the fact that -- on my platform --
> sizeof(struct sockaddr) is 16, it turns out that both r->ai_addrlen and
> SA_LEN(r->ai_addr) are 28.  Now, I don't really know anything about this
> sockaddr mess, but I'm quite sure that 28 bytes can't fit into a 16-byte
> struct.

Yeah, this is a common problem with struct sockaddr.  It's a really poorly
defined interface and it breaks some of the standard C behavior because
it's often implemented with a struct for which sizeof is smaller than the
actual size of the allocated memory.

> The solution is to use struct sockaddr_storage, from what I understand,
> but obviously we can't depend on that being around on non-IPv6
> platforms.

Can we fake it on platforms that don't have it?  I'd rather go that way if
we can, since it seems to be a saner interface.

Other than that, your solution seems fine to me.

-- 
Russ Allbery (rra at stanford.edu)             <http://www.eyrie.org/~eagle/>

    Please send questions to the list rather than mailing me directly.
     <http://www.eyrie.org/~eagle/faqs/questions.html> explains why.