On Mon, 2013-02-25 at 06:42 -0800, Dave Hansen wrote: > On 02/24/2013 01:28 PM, Peter Hurley wrote: > > Now that the alloc_remap() has been/is being removed, is most/all of > > this being reverted? > > I _believe_ alloc_remap() is the only case where we actually remapped > low memory. However, there is still other code that does __pa() > translations for percpu areas: per_cpu_ptr_to_phys(). I _think_ it's > still theoretically possible to get some percpu data in the vmalloc() area. > > > So in short, my questions are: > > 1) is the slow_virt_to_phys() necessary anymore? Ah, yep. Thanks for pointing out per_cpu_ptr_to_phys(). > kvm_vcpu_arch has a > > struct pvclock_vcpu_time_info hv_clock; > > and I believe I mistook the two 'hv_clock's for each other. However, > this doesn't hurt anything, and the performance difference is probably > horribly tiny. Ok. It was confusing because the fixmap of that same phys memblock done by pvclock was broken and I couldn't understand why the hvclock memblock needed to be looked-up per cpu. Mystery solved. Regards, Peter Hurley -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>