> There was a sort of discussion raised at euro-coders mailing list recently
> which related to the differencies in implementation of recvfrom() call in
> linux socket API and the rest of Unixes (Solaris and two BSD clones (BSDi
> and FreeBSD were tested).
>
> We played with a sort of self-made udp echo daemon, which basically was
> listening to the udp packets on certain port and was sending them back to
> the same IP:port-no. The curious thing is that if the packet is spoofed, and
> no port is being listened on remote machine (and icmp port unreachable goes
> back) linux returns -1 for next recvfrom with herrno(?) set to REFUSED,
> while BSD/Solaris just ignore this.
>
> The question is:
> is there any paper/RFC/FYI which puts a standard on such things, or this is
> basically `the matter of taste' of OS developers, and thus just should being
> watched carefully, while developing multi-platform applications?
The Linux behaviour is strictly correct according to RFC1122.
BSD only passes the error when the socket is connect()ed [understandable
because it is otherwise hard to find out who caused the error].
Linux 2.2 added a new API to fix it: IP_RECVERR, which allows the
application to reliably process asynchronous network errors.
My new network man pages snapshot includes a discussion of the issue
in ip.4 and udp.4, see ftp://zero.aec.at/pub/ak/netman-19990409.tgz or
available via the samba anonymous CVS server (/vger/netman/)
-Andi
-- This is like TV. I don't like TV.- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/