inetd for no-MMU in trunk

Aleksey Demakov a.demakov at pengutronix.de
Thu Mar 20 09:35:50 PDT 2008


Hi all,

Reviewing inetd in the trunk I found that the changes
for no-MMU support IMHO go in wrong direction.

Out of the three builtins that require fork the first two
were converted to exec an external process and the
third was disabled altogether.

As far as I understand fork is only needed for builtins
that work over a stream socket. For datagram sockets
fork is never executed (in the trunk version) so there
is no point to disable the builtin for it.

Also I think that execing a process from within a builtin
kind of defeats the purpose of the builtin.

All in all I believe that either of these options should
make more sense than the current approach:

1. completely disable all builtins that require fork
on no-MMU systems

2. disable them only to work over stream sockets and
enable for datagram sockets.

I tried to implement the second approach for 1.9.1.
Please take a look at the attached patch. I did not
test this code yet. I'll make the patch against the
trunk and test it if there is interest in this approach.

Regards,
Aleksey
-------------- next part --------------
A non-text attachment was scrubbed...
Name: busybox-1.9.1-inetd_nommu.patch
Type: text/x-patch
Size: 2289 bytes
Desc: not available
Url : http://busybox.net/lists/busybox/attachments/20080320/93bdce2b/attachment.bin 


More information about the busybox mailing list