I wanted to write a blog post but got side-tracked. In the meantime, here is something I learned:
Suppose you have a script for which an fd is a unix socket. How do you call getpeername(2) on this socket from the cli to get the remotely bound socket?
On Debian, there is a package called tcputils which has a getpeername binary. It only works on INET sockets though. The netpipes Debian package also used to have a getpeername binary (23 years ago); but now it is in the getsockname binary to prevent conflicting with the previous package. But getsockname(2) is not the call I want, I already know the local socket name. You can then use exec -a getpeername getsockname to get the behaviour you want. But unfortunately netpipes' getpeername will not print the content of a peer name which is an abstract socket, it will print (abstract) instead.
So, back to square one. By now I could already have written a simple C program (many times over) to do what I want and called it a day. But I like weird constraints, so here I am. Is there any interpreter on a default Debian install that could let me do what I want. Tough luck: there is no python. But there is perl, so this is how you'd do it in Perl
#!/usr/bin/env perl
use IO::Socket::UNIX;
my $sock = IO::Socket::UNIX->new_from_fd(0, "r+");
print $sock->peerpath()
And how do you test it works? socat has a convenient bind= option on which you can pass an abstract unix socket name if it start with @. Launching the program with a unix socket as stdin is left as an exercise to the reader.
The blogpost was supposed to be shorter than this post, but we'll see, if it materializes.
rom1v
bit101
Joseph Lorenzo Hall, PhD
Adam Williamson 
LWN.net