Date: Fri, 23 Jun 2000 22:08:12 -0500 From: Carroll Kong <damascus@home.com> To: freebsd-atm@FreeBSD.ORG Subject: ATM Switch? Message-ID: <4.2.2.20000623214426.0267d2f0@email.eden.rutgers.edu>
next in thread | raw e-mail | index | archive | help
Hey guys. I need to create an ATM Switch which is controllable remotely. I was thinking about using a control socket that handles control commands and dynamically creating a "routing table" of VCI's -> to other VCI's. I need to use PVCs all the way. I took a look at Charles D. Cranor's web site and his driver. At first it seemed like things were very promising, until I realized, we just got a Fore 200E card, not a midway based card. I also tried to do a search for the efficient networks or adaptec cards, and for some odd reason, I cannot find either one! As my search continued, I realized that HARP does have a socket API. However, I am somewhat new to the realm of ATM and I am not familiar with doing routing in C. From what my books tell me, I would have to use a raw socket call to the kernel and pass a message like RTM_ADD plus the sockstructure. (In this case, the sockaddr_atm of HARP). Cranor notes that his driver does not require any special new programs to interface with the ATM's controls. It seems like HARP does... ala "atm X Y Z". This leads me to believe, that I cannot control my FreeBSD box like an ATM Switch as easily with HARP? Also I noted that HARP mentioned it does not support UNI->multipoint. Now, is this a uni->multipoint? VCI-100 -> VCI-301; VCI-100 -> VCI-302; Both going simultaneously? Logic would dictate that it is indeed a UNI->multipoint. :( I would also like it if the VCI-301 could say "hey... switch me to VCI-101 now"... VCI-101->VCI-301; I was hoping originally to do this with creating two sockaddr_atms and piping the two together. But the documentation on the site says To support PVC sockets, there is a new address format (T_ATM_PVC_ADDR) and address definition (Atm_addr_pvc). Since there is no actual signalling involved in setting up a PVC, a PVC socket connection only defines the local socket-to-pvc connection - the remainder of the virtual circuit through the ATM network to the remote endpoint must be defined independent of the local socket creation. PVC socket connections are only allowed via the connect() system call - listen()/accept() sockets cannot be supported. Also, since there are no circuit parameters signalled, most of the setsockopt() options are silently ignored. I know I sound like I am asking a lot, I just want to know if it is feasible and a basic idea. I am a decent programmer, I can hopefully figure out the rest if pointed the way. Thanks a lot guys! -Carroll Kong To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-atm" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4.2.2.20000623214426.0267d2f0>