On Monday 09 November 2009, Øyvind Harboe wrote: > > My issue with that patch is that it should not be needed! > > > > If there is no mmu() method ... that's sufficient to > > know that has_mmu() == false. Etc. > > So all CPUs that have an MMU should implement this fn? > > OK.
Yes. > That would be just about all targets... Most non-microcontroller targets. > >> What does virt2phys translation mean when no MMU > >> is present or MMU is disabled? > > > > Simplest answer: it's an identity mapping. If there > > is *ever* a question, treat a no-MMU target as if it's > > an MMU-enabled target with a permanently installed > > memory mapping equating virtual and physical addresses. > > Why should virt2phys not fail on CPUs with no MMU? Because we want to avoid needing complete parallel stacks for the MMU and no-MMU cases ... and that means not creating needless failure modes. > Are we sure that *all* higher level code would do the right > thing when we treat virt2phys as an identity mapping? > > Clearly the answer to that question is obvious, it just > makes me wonder... All non-buggy code does the right thing, by definition. As you say, obvious: an identity mapping could be installed into an MMU, ergo that "higher level" code must handle it. Q.E.D. > Minimally we have to document that virt2phys MUST > implement *only* a successful identity mapping when no > MMU is supported for that target. That function can be trivially defaulted; no need to document it for target implementors. _______________________________________________ Openocd-development mailing list Openocd-development@lists.berlios.de https://lists.berlios.de/mailman/listinfo/openocd-development
