connect(2) behavior with unreacheable hosts

[barney at pit.databus.com (Barney Wolff)]

On Fri, Apr 11, 2003 at 01:14:15PM -0400, BelletJr_(_at_)_aol_(_dot_)_com wrote:
> >> Why does not connect(2) return any error when trying to connect to a host
> >> unreachable because of an infinite loop in the routes? No time-out occurs 
> and
> >> the value 0 is returned by connect(2).
> >> My test was done with TCP/IPv4.

Well, after actually looking at this, I believe it's a bug, in both
current and stable.  The following patch appears to fix it:

--- ip_input.c.orig     Wed Apr  9 14:07:16 2003
+++ ip_input.c  Fri Apr 11 17:54:11 2003
@@ -1696,7 +1696,7 @@
        0,              EMSGSIZE,       EHOSTDOWN,      EHOSTUNREACH,
        EHOSTUNREACH,   EHOSTUNREACH,   ECONNREFUSED,   ECONNREFUSED,
        EMSGSIZE,       EHOSTUNREACH,   0,              0,
-       0,              0,              0,              0,
+       0,              0,              EHOSTUNREACH,   0,
        ENOPROTOOPT,    ECONNREFUSED
 };
  
I'll file a PR.
By the way, the reason programs appear to do nothing after hitting the
problem is that they get SIGPIPE when trying to write on the socket.

-- 
Barney Wolff         http://www.databus.com/bwresume.pdf
I'm available by contract or FT, in the NYC metro area or via the 'Net.