On Sat, 20 Aug 2005, David S. Miller wrote: > From: Christoph Lameter <[EMAIL PROTECTED]> > Date: Sat, 20 Aug 2005 21:16:16 -0700 (PDT) > > It does not exist today AFAIK. The hope of such a solution will prevent > > the inclusion of TOE technology that exists today? > What you say isn't exactly true, as Lenoid Grossman and others are > working on LRO schemes for the XFrame-II chips. Read his posts > in this thread.
Yes I have used LSO before (its been awhile I hope I get this right) and while it mostly works its not strictly following TCP since the TCP flow control does not occur between packets. Similar issues may plague LRO (reading the paper from the OLS). I'd rather have some stateful logic between packets as expected by the TCP protocol rather than sending a sequence of messages in brute force out to the net. But if the network card cannot do any better then lets do at least LSO/TSO. Chelsio NICs support LSO and I have no doubt that there would not be an issue with implementing LRO if need be. LSO and LRO are like TOE in that they take elements of TCP away from the nature. LSO/LRO limit that to TCP flow control playing a bit with the logic of TCP to give the illusion of being stateles. > So it isn't hope. People are working on this and it's very real. Its half way thing and likely a bigger problem than getting a few TOE hooks into the tcp stack and maintaining them. I bet the tricks that we hack into the TCP/IP stack for LSO and for LRO will turn out to be more difficult to maintain than the proposed TOE hooks. > At least the XFrame folks, such as Lenoid, such be largely commended > for actually pursuing alternate schemes instead of sticking their > heads in the sand and just accepting TOE "the only" solution. Are you sure that these ideas have broad support? Isnt this simply a rationalization to justify that their network technology has not been able to keep up with emerging offload technologies? As far as I can tell the mainstream of the industry seems to be moving to TOE, seeing LSO as an intermediate implementation on the way to full TCP offload. - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html