Hi Linus, On Mon, Apr 16, 2018 at 08:32:46AM -0700, Linus Torvalds wrote: > Also, it would be nice to have performance numbers, Here are some numbers I gathered for Ingo on v2 of the patch-set: https://marc.info/?l=linux-kernel&m=151844711432661&w=2 I don't think they significantly changed, but I can measure them again if needed. Besides those, what other numbers/tests are you interested in? > and check that the global page issues are at least fixed on 32-bit > too, since we screwed that up on x86-64 initially. > > On x86-32, the global pages are likely a bigger deal since there's no PCID. Okay, I verify if there are any global bits left in the page-tables. According to the PTDUMP_X86 the cpu_entry_area is mapped with G=1 (which should be fine?) and another 4M range in the kernel mapping. I need to check what that is. Thanks, Joerg > > Linus
