We are preparing a new version of the Real Time Ethernet module (0.2B) to be released sometime in the next week or so. New to this version: 1) Local SKB cache. All SKB's are allocated from the kernel at init_module time, and free'd during cleanup_module. 2) Input packet ring buffer. (No locking or overflow protection yet) 3) Much more error checking and nice stats printout at cleanup_module time. Still using the tulip driver as an example, with 2 new functions added: alloc_skb and free_skb, which are set using the SIOCDEVPRIVATE ioctl interface (along with mark_bh and netif_rx). Upon examining other drivers written by Donald Becker, it appears that making the changes to these drivers would be fairly trivial. Again, we have been using the above interface for a week or so with no apparent problems relating to the ethernet code. Would it be the recommendation of this group that the interrupt routine be hooked on the real-time side, as opposed to letting the Linux side handle it? We have not been having problems with the current implementation, which does not do this, but are interested in other thoughts. In its current incarnation, it could be used by any kernel module, not just real time modules, but it is unclear whether this is of any value. Thanks for your time. --- Michael M. Morrison Chief Technical Officer Hyperion Technologies, Inc. --- [rtl] --- To unsubscribe: echo "unsubscribe rtl" | mail [EMAIL PROTECTED] OR echo "unsubscribe rtl <Your_email>" | mail [EMAIL PROTECTED] ---- For more information on Real-Time Linux see: http://www.rtlinux.org/~rtlinux/
