Teresa Noviello wrote: > We have two ELCOM Eurocom E148 hosts, with Slack 9.2 (host with name > "rt_linux-1") and Slack 9.1 (host with name "rt_linux-2"). > We have installed on the two host kernel 2.6.15, rtai 3.3 and rtnet (latest > snapshot). > We tried the configuration Master/Slave first with an hub connecting > "rt_linux-1" and "rt_linux-2" and then with a switch. The result is the > same. The LAN is not connected to Internet. > > We have configured "rt_linux-1" as a master and this is the configuration > file: > ... (looks ok) > > We followed the instructions in the file README of /usr/local/rtnet to load > the modules and to unload the ethernet's module (eepro100) before doing > > "/usr/local/rtnet/sbin/rtnet start" > > in the two hosts. > > In the master, we have the message > "Waiting for all slaves" and nothing else. We must do CTRL+C to stop the > process. In this host, the following > > rtnet stop > > gives error and a Segmentation fault. I can't write you the error because > the process do loop on a message:
I'm picking a standard answer first: please check if there is no IRQ conflict between the realtime NIC and some Linux device (see /proc/interrupts). This can have ugly side effects, up to crashes/lock-ups. > > RTDM: info > IRQ: 11, intrcnt: 21228 > rtnet: unregistering rteth0 deferred -refcount=2 > .... > rtnet: unregistering rteth0 deferred -refcount=2 > rtnet: unregistering rteth0 deferred -refcount=2 > rtnet: unregistering rteth0 deferred -refcount=2 > .... > > and i must shutdown the machine turning off the power. Another reason for the crash (not the stuck setup) might be a problem in the RTDM layer. If there is a kernel oops, this would help to nail it down. You could catch it by redirecting the kernel console to a serial port which is read by another machine (kernel boot parameter: console=ttyS0,115200N8) > > In the slave, we have the message > "Searching for master" and nothing else. When we do on this host > > rtnet stop > > we have no errors. > > What we do wrong?? Does the kernel output contain any warnings, errors, etc. on one of the stations? If you already have a hub between the stations, it would be good to capture the traffic on the wire with a third box using Ethereal, e.g. Then we can see, what is submitted and what are the reactions. Additionally, you can check the states of RTcfg in /proc/rtnet/rtcfg/ on both node to see if there is any change after the master starts up or the slave is attached. Jan
signature.asc
Description: OpenPGP digital signature
