On 10/21/2018 06:57 PM, Ingo Molnar wrote: > Does the CPU _ever_ look look at the PFN if the page is !_PAGE_PRESENT, > for example speculatively? If yes then what is the recommended value for > the pfn - zero perhaps?
I'll never say never. :) For L1TF[1], we know the CPU did exactly this; it ignored the _PAGE_PRESENT bit when fetching data from the L1. That's what is worked around with the gunk in arch/x86/include/asm/pgtable-invert.h. I think Andi plugged the code in here at a low enough level in the page table manipulation that pageattr.c should inherit it without doing anything explicit. But, Sai, you might want to double-check this. 1. https://www.intel.com/content/www/us/en/architecture-and-technology/l1tf.html