On Tue, May 22, 2007 at 11:51:30PM -0400, Kumba wrote: > Didn't test on 2.6.22 (yet), but 2.6.21.1 works: That's the only one I care about for purposes of upstream patch submission anyway. > Btw, If we wanted to protect meth from the speculative execution issues of > the R10000 processor, what's the right way for that? I believe IP28 used a > special type of buffer for protecting Seeq from the DMA wonkiness that > occurs, but I got the indication that Meth would need a different approach. I don't see why the standard approach would fail for meth. Ralf
