Hi Petr,
I am not sure if the small patch helps you but you may want to try it out (version 0.8.0). It is a part of much larger patch, so I am not realy sure if it works alone (it should), you probably would not want the whole thing.It uses a linked lists of protocol handlers (just like normal Linux). There is also a check for the interface where the frame came from. Good luck, Petr Grillinger
Mmh, I'm currently considering if adding at least multiple protocols per hash key would actually cost that much overhead. Likely not noticable. The per-interface registration could be made optional (CONFIG_RTNET_PER_IF_PROTO or so) as this does introduce additional checks to the critical path and is a non-common feature. Would you be interested in having these features in the next official release, Petr?
Ok, now I'm sorry to pick on your patches once again ;), but there is a double kfree_rtskb hidden: by convention (ok, that's nowhere explicitely documented...), the responsibility for a skb passes to the protocol handler if there is any called. If I read your patch correctly, you always release the skb at the end of the stack manager loop. This should trigger oopses quite quickly!
Jan ------------------------------------------------------- This SF.Net email is sponsored by Yahoo. Introducing Yahoo! Search Developer Network - Create apps using Yahoo! Search APIs Find out how you can build Yahoo! directly into your own Applications - visit http://developer.yahoo.net/?fr=offad-ysdn-ostg-q22005 _______________________________________________ RTnet-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/rtnet-users
