[...] > It seems a shame to me that the architecture for this > new feature stops > at the IP/driver connection and doesn't include the > API level. > > -- > James Carlson 42.703N 71.076W > <carls...@workingcode.com>
Particularly since the information is already available - the UD MTU has to be already known, and the per-IB-connection MTU is presumably discovered as part of address resolution (RFC 4755, paragraph 5.1). A new ioctl, that made those available (if the link layer supported it) or failed (otherwise) leaving the application to decide by some other means, doesn't seem as if it would be unduly difficult, nor (unlike path MTU discovery) would it require any additional network overhead. The only question being if there's some precedent in another implementation for what the API should look like. I'm having trouble thinking up a likely google/bing query for that... -- This message posted from opensolaris.org _______________________________________________ opensolaris-arc mailing list opensolaris-arc@opensolaris.org
