On Fri, Feb 14, 2003 at 01:21:18PM +1100, Andrew Clausen wrote: > On Fri, Feb 14, 2003 at 02:58:35AM +0100, Ralf Baechle wrote: > > > I'm not sure what the best solution is, but this makes things work: > > > > And is guaranateed to crawl. flush_cache_all() is a no-op for the R10k. > > Right-o. But who cares? Apparently it almost never gets called. > > Also, I couldn't find any documentation on what callers of > flush_cache_all() expect... Documentation/cachetlb.txt - but it takes alot of reading between the lines that is reading of the generic mm code to understand that document. Ralf
