If the kernel does not have the "Packet socket" and "Socket
filtering" options enabled, pump will core dump. I haven't
looked in the source, but I would guess it's getting an
error return from a socket() call, and the error checking is
done incorrectly or completely ignored.
dhclient does the right thing, and exits gracefully while
saying what kernel features should be present. I would
never have made networking work without using dhclient (part
of the isc dhcp distribution), since pump wouldn't tell me
I have verified that compiling a kernel with the Packet Socket and
Socket Filtering options turned off, pump will dump core when trying
to initialize the network device. I have assigned this to a developer
for further review.
*** Bug 4294 has been marked as a duplicate of this bug. ***
pump dumps core when running with my own compiled
I could attach the core dump, if there only was a way to
attach it. (I guess that would qualify as a bug/enhancement
request for bugzilla).
------- Additional Comments From email@example.com 08/02/99 08:48 -------
Submit the core dump in the same method as you would submit a patch.
Check the FAQ for the proper subject line for submitting a patch and
it will get attached to your bug automatically.
After spending several hours trying to get my eth1 up with pump (DHCP via cable
modem) I found out that my packet socket (AF_PACKET) was configured as module in
kernel (tried versions 2.2.5-15 and 2.2.5-22 on RH6.0). After changing the
packet socket option to be built-in feature, pump started working.
Jay, please build a kernel that exhibts this problem. It should be easy enough
Recompiling the kernel this way in 6.2, pump fails gracefully:
[root@localhost /root]# /sbin/pump -i eth0 --status
[root@localhost /root]# uname -a
Linux localhost.localdomain 2.2.14-5.0 #1 Mon Apr 17 15:58:22 EDT 2000 i586