if /etc/inetd.conf turns on the time/tcp service, then connections are left in a CLOSE_WAIT state. For me, a handful of Solaris and Linux clients will cause the server's inetd to reach its fd limit in a few days. RHL 6.2, kernel-2.2.14-12 (686, smp), inetd-0.16-4
*** Bug 14687 has been marked as a duplicate of this bug. ***
*** Bug 12779 has been marked as a duplicate of this bug. ***
heh. --- netkit-base-0.16/inetd/inetd.c~ Wed Nov 24 06:31:53 1999 +++ netkit-base-0.16/inetd/inetd.c Thu Aug 3 20:25:17 2000 @@ -485,12 +485,12 @@ FD_CLR(sep->se_fd, &allsock); nsock--; } - if (!sep->se_wait && sep->se_socktype == SOCK_STREAM) - close(ctrl); } else { sep->se_bi->bi_fn(ctrl, sep); } + if (ctrl != sep->se_fd) + close(ctrl); } static
*** This bug has been marked as a duplicate of 11548 ***