Bernhard Breinbauer wrote: > Am Mittwoch, 7. Juni 2006 09:45 schrieb Jan Kiszka: >> Bernhard Breinbauer wrote: >>> Hello list! >>> >>> I am evaluating the performance of RTnet under several load conditions. >>> The (so called) network consists of two cross-connected PC's with RTL8169 >>> gigabit NIC's. >>> Starting RTnet with the delivered start script is working fine, but I >>> would like to avoid using a RTcfg server. So I tried to produce a >>> procedure for a statically configured RTnet network (mostly extracted >>> from the start script): 1) loading modules like: rtnet.ko, rtipv4.ko,.... >>> 2) `rtifconfig rtlo up 127.0.0.1` >>> 3) loading modules: rtmac.ko and tdma.ko >>> 4) `rtifconfig rteth0 up 10.0.0.1` >>> 5) update routing table with rtroute, to provide a route to the other PC >> Stop. Check the routing tables on both sides after this step (How do you >> set the routes? Via rtroute solicite?). Also try if rtping works. If >> not, some link problem or some issue with the driver likely exists. > > I use on host 1 (with IPaddr1): > rtroute add IPaddr2 MACaddr2 dev rteth0 > and on host 2 vice versa. > The routing table contains the route to the other host after that. > Should rtping work at this point? Without starting master and slave and > defining slots? > Well tested it. rtping *does* work. > >>> 6) run `tdmacfg rteth0 server 10000` on the master >>> 7) run `tdmacfg rteth0 slave` on the slave >>> 8) run `tdmacfg slot 0 2000` on the slave (does not return, because it >>> starts calibration as stated in the docu) >>> 9) run `tdmacfg slot 0 0` on the master (does finish nearly instantly) > > Continuing the procedure. Setting the slot does not return.
Ahh -- let me guess: you are using RTAI, aren't you? Then you got caught in the infamous "Who should start the timer?" trap. In normal setups, rtcfg.ko does this by default when being loaded. In your case, you have to tell tdma.ko to take over: insmod tdma.ko start_timer=1. > >> Unrelated to this problem: Why not using RTcfg? > > I would like to produce an encapsulated system with as less as possible > predictable and not predictable interference. > I know I can disable the heartbeats and set the config redistribution time to > an insane high value, but i would prefer to avoid rtcfg packets completly. > But it remains my fallback solution :-) Ok, I understand. Then you will likely also not make of use of the tunnelling device RTmac provides. This keeps your network "clean". 8) Jan
signature.asc
Description: OpenPGP digital signature
_______________________________________________ RTnet-users mailing list RTnet-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/rtnet-users
