On 2011/06/23 17:53, Rechner-Tester wrote: > >However there are other changes that should be mentioned. > >Rx checksum offload support was added, > > Is this usefully at all on the VT6105M? > There was a very interesting analysis of the offloading capabilities > of this chip by Willy Tarreau for the Linux module. Have a look at > <http://lkml.org/lkml/2007/12/30/242> for details.
Interesting indeed. I can certainly think of reasons to favour throughput, but I can also see good reasons to favour reduced cpu use (and know for sure of people who will be seeing a significant number of <1000 byte packets on vr). > >and the longword-alignment > >mentioned in BUGS is no longer required on the current chips, > > That's good to know. > > >it seems simplest to remove BUGS completely but we could > >alternatively keep it and discuss which chips are affected. > > I feel uncomfortable dismissing some Information which could still be > useful for someone. Since I can not help with the technical details > about the affected chips, I suggest to simply change the first > sentence as following: I don't consider it a bug in the driver, it's just what you have to do to support the chip. c.f. rl(4) where there is similar requirement about longword align buffers however there the "driver abuses the fact that the cluster buffer pool is allocated at system startup time in a contiguous region starting at a page boundary" which clearly needs a bit more of a warning.