rlinetd is designed to replace the BSD inetd. Feature-wise, it is a
proper superset of the BSD inetd, including rpc support and both udp
and tcp services. In addition, it supports a number of resource
limiting features, including full resource limits a la setrlimit(2),
renicing, chroot and limited per-service instances. To round out the
features stolen from xinetd, it also supports binding to individual
interfaces on request, and configurable logging.
This program is designed to replace inetd. It contains a number of
things that seemed like a good idea at the time (and still do). The
BUGS file lists the obstacles encountered on the way to implementing
the things listed in TODO.
rlinetd is brought to you by the number 23, the letter V, and the
GNU General Public License.
Mikolaj J. Habryn
-*- text -*-
* NOTE: This information predates all of the elfcap stuff in the
kernel. I have no idea how things will work in combination with it. If
you do, drop me a note :)
These are capability inheriting rules (ie, what happens across an
pI' = pI
pP' = fP | (fI & pI)
pE' = pP' & fE
Currently (as of 2.2.3)
On exec, fI and fE is always 0 to a non-root p
-*- text -*-
There are minor gotchas with converting from inetd. At present, initgroups(3)
is not called by default - you must specify the initgroups directive for
this to happen.
The wait functionality is handled by the 'wait yes' directive in the
configuration file. This directive automatically limits max number of
instances to 1.
If you are configuring a udp service (or any other service w