On Thu, Jun 05, 2003 at 03:16:41PM +0530, Ashish anand wrote: > I have seen it a common practice to assume low RAM as source > of good parity data and use it to fill the caches with good parity > data in firmware during elementary hardware initialisation process. > > why it is always safe to assume that low RAM contains good parity data .? > Is it always true..? > this question came in picture after I got cacheparity error sometimes. For the general case it's not safe. Some systems need memory to be initialized by writing to it first because before the parity or ECC bits may have an undefined state. The safe way to initalize the caches is using the Index_Store_Tag_D etc. cacheops but of course that requires knowledge about the particular processor's exactly cache architecture. Ralf
