CC: kbuild-...@lists.01.org BCC: l...@intel.com CC: linux-ker...@vger.kernel.org TO: David Hildenbrand <da...@redhat.com> CC: Andrew Morton <a...@linux-foundation.org> CC: Linux Memory Management List <linux...@kvack.org>
tree: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master head: 34e047aa16c0123bbae8e2f6df33e5ecc1f56601 commit: 445fcf7c721450dd1d4ec6c217b3c6a932602a44 mm/memory_hotplug: memory group aware "auto-movable" online policy date: 6 months ago :::::: branch date: 12 hours ago :::::: commit date: 6 months ago compiler: powerpc64-linux-gcc (GCC) 11.2.0 If you fix the issue, kindly add following tag as appropriate Reported-by: kernel test robot <l...@intel.com> cppcheck possible warnings: (new ones prefixed by >>, may not real problems) >> mm/memory_hotplug.c:860:16: warning: Local variable online_pages shadows >> outer function [shadowFunction] unsigned long online_pages = 0, max_pages, end_pfn; ^ mm/memory_hotplug.c:1024:11: note: Shadowed declaration int __ref online_pages(unsigned long pfn, unsigned long nr_pages, ^ mm/memory_hotplug.c:860:16: note: Shadow variable unsigned long online_pages = 0, max_pages, end_pfn; ^ vim +860 mm/memory_hotplug.c c246a213f5bad68 Michal Hocko 2017-07-06 808 e83a437faa625ef David Hildenbrand 2021-09-07 809 /* e83a437faa625ef David Hildenbrand 2021-09-07 810 * Determine to which zone to online memory dynamically based on user e83a437faa625ef David Hildenbrand 2021-09-07 811 * configuration and system stats. We care about the following ratio: e83a437faa625ef David Hildenbrand 2021-09-07 812 * e83a437faa625ef David Hildenbrand 2021-09-07 813 * MOVABLE : KERNEL e83a437faa625ef David Hildenbrand 2021-09-07 814 * e83a437faa625ef David Hildenbrand 2021-09-07 815 * Whereby MOVABLE is memory in ZONE_MOVABLE and KERNEL is memory in e83a437faa625ef David Hildenbrand 2021-09-07 816 * one of the kernel zones. CMA pages inside one of the kernel zones really e83a437faa625ef David Hildenbrand 2021-09-07 817 * behaves like ZONE_MOVABLE, so we treat them accordingly. e83a437faa625ef David Hildenbrand 2021-09-07 818 * e83a437faa625ef David Hildenbrand 2021-09-07 819 * We don't allow for hotplugged memory in a KERNEL zone to increase the e83a437faa625ef David Hildenbrand 2021-09-07 820 * amount of MOVABLE memory we can have, so we end up with: e83a437faa625ef David Hildenbrand 2021-09-07 821 * e83a437faa625ef David Hildenbrand 2021-09-07 822 * MOVABLE : KERNEL_EARLY e83a437faa625ef David Hildenbrand 2021-09-07 823 * e83a437faa625ef David Hildenbrand 2021-09-07 824 * Whereby KERNEL_EARLY is memory in one of the kernel zones, available sinze e83a437faa625ef David Hildenbrand 2021-09-07 825 * boot. We base our calculation on KERNEL_EARLY internally, because: e83a437faa625ef David Hildenbrand 2021-09-07 826 * e83a437faa625ef David Hildenbrand 2021-09-07 827 * a) Hotplugged memory in one of the kernel zones can sometimes still get e83a437faa625ef David Hildenbrand 2021-09-07 828 * hotunplugged, especially when hot(un)plugging individual memory blocks. e83a437faa625ef David Hildenbrand 2021-09-07 829 * There is no coordination across memory devices, therefore "automatic" e83a437faa625ef David Hildenbrand 2021-09-07 830 * hotunplugging, as implemented in hypervisors, could result in zone e83a437faa625ef David Hildenbrand 2021-09-07 831 * imbalances. e83a437faa625ef David Hildenbrand 2021-09-07 832 * b) Early/boot memory in one of the kernel zones can usually not get e83a437faa625ef David Hildenbrand 2021-09-07 833 * hotunplugged again (e.g., no firmware interface to unplug, fragmented e83a437faa625ef David Hildenbrand 2021-09-07 834 * with unmovable allocations). While there are corner cases where it might e83a437faa625ef David Hildenbrand 2021-09-07 835 * still work, it is barely relevant in practice. e83a437faa625ef David Hildenbrand 2021-09-07 836 * e83a437faa625ef David Hildenbrand 2021-09-07 837 * We rely on "present pages" instead of "managed pages", as the latter is e83a437faa625ef David Hildenbrand 2021-09-07 838 * highly unreliable and dynamic in virtualized environments, and does not e83a437faa625ef David Hildenbrand 2021-09-07 839 * consider boot time allocations. For example, memory ballooning adjusts the e83a437faa625ef David Hildenbrand 2021-09-07 840 * managed pages when inflating/deflating the balloon, and balloon compaction e83a437faa625ef David Hildenbrand 2021-09-07 841 * can even migrate inflated pages between zones. e83a437faa625ef David Hildenbrand 2021-09-07 842 * e83a437faa625ef David Hildenbrand 2021-09-07 843 * Using "present pages" is better but some things to keep in mind are: e83a437faa625ef David Hildenbrand 2021-09-07 844 * e83a437faa625ef David Hildenbrand 2021-09-07 845 * a) Some memblock allocations, such as for the crashkernel area, are e83a437faa625ef David Hildenbrand 2021-09-07 846 * effectively unused by the kernel, yet they account to "present pages". e83a437faa625ef David Hildenbrand 2021-09-07 847 * Fortunately, these allocations are comparatively small in relevant setups e83a437faa625ef David Hildenbrand 2021-09-07 848 * (e.g., fraction of system memory). e83a437faa625ef David Hildenbrand 2021-09-07 849 * b) Some hotplugged memory blocks in virtualized environments, esecially e83a437faa625ef David Hildenbrand 2021-09-07 850 * hotplugged by virtio-mem, look like they are completely present, however, e83a437faa625ef David Hildenbrand 2021-09-07 851 * only parts of the memory block are actually currently usable. e83a437faa625ef David Hildenbrand 2021-09-07 852 * "present pages" is an upper limit that can get reached at runtime. As e83a437faa625ef David Hildenbrand 2021-09-07 853 * we base our calculations on KERNEL_EARLY, this is not an issue. e83a437faa625ef David Hildenbrand 2021-09-07 854 */ 445fcf7c721450d David Hildenbrand 2021-09-07 855 static struct zone *auto_movable_zone_for_pfn(int nid, 445fcf7c721450d David Hildenbrand 2021-09-07 856 struct memory_group *group, 445fcf7c721450d David Hildenbrand 2021-09-07 857 unsigned long pfn, e83a437faa625ef David Hildenbrand 2021-09-07 858 unsigned long nr_pages) e83a437faa625ef David Hildenbrand 2021-09-07 859 { 445fcf7c721450d David Hildenbrand 2021-09-07 @860 unsigned long online_pages = 0, max_pages, end_pfn; 445fcf7c721450d David Hildenbrand 2021-09-07 861 struct page *page; 445fcf7c721450d David Hildenbrand 2021-09-07 862 e83a437faa625ef David Hildenbrand 2021-09-07 863 if (!auto_movable_ratio) e83a437faa625ef David Hildenbrand 2021-09-07 864 goto kernel_zone; e83a437faa625ef David Hildenbrand 2021-09-07 865 445fcf7c721450d David Hildenbrand 2021-09-07 866 if (group && !group->is_dynamic) { 445fcf7c721450d David Hildenbrand 2021-09-07 867 max_pages = group->s.max_pages; 445fcf7c721450d David Hildenbrand 2021-09-07 868 online_pages = group->present_movable_pages; 445fcf7c721450d David Hildenbrand 2021-09-07 869 445fcf7c721450d David Hildenbrand 2021-09-07 870 /* If anything is !MOVABLE online the rest !MOVABLE. */ 445fcf7c721450d David Hildenbrand 2021-09-07 871 if (group->present_kernel_pages) 445fcf7c721450d David Hildenbrand 2021-09-07 872 goto kernel_zone; 445fcf7c721450d David Hildenbrand 2021-09-07 873 } else if (!group || group->d.unit_pages == nr_pages) { 445fcf7c721450d David Hildenbrand 2021-09-07 874 max_pages = nr_pages; 445fcf7c721450d David Hildenbrand 2021-09-07 875 } else { 445fcf7c721450d David Hildenbrand 2021-09-07 876 max_pages = group->d.unit_pages; 445fcf7c721450d David Hildenbrand 2021-09-07 877 /* 445fcf7c721450d David Hildenbrand 2021-09-07 878 * Take a look at all online sections in the current unit. 445fcf7c721450d David Hildenbrand 2021-09-07 879 * We can safely assume that all pages within a section belong 445fcf7c721450d David Hildenbrand 2021-09-07 880 * to the same zone, because dynamic memory groups only deal 445fcf7c721450d David Hildenbrand 2021-09-07 881 * with hotplugged memory. 445fcf7c721450d David Hildenbrand 2021-09-07 882 */ 445fcf7c721450d David Hildenbrand 2021-09-07 883 pfn = ALIGN_DOWN(pfn, group->d.unit_pages); 445fcf7c721450d David Hildenbrand 2021-09-07 884 end_pfn = pfn + group->d.unit_pages; 445fcf7c721450d David Hildenbrand 2021-09-07 885 for (; pfn < end_pfn; pfn += PAGES_PER_SECTION) { 445fcf7c721450d David Hildenbrand 2021-09-07 886 page = pfn_to_online_page(pfn); 445fcf7c721450d David Hildenbrand 2021-09-07 887 if (!page) 445fcf7c721450d David Hildenbrand 2021-09-07 888 continue; 445fcf7c721450d David Hildenbrand 2021-09-07 889 /* If anything is !MOVABLE online the rest !MOVABLE. */ 445fcf7c721450d David Hildenbrand 2021-09-07 890 if (page_zonenum(page) != ZONE_MOVABLE) 445fcf7c721450d David Hildenbrand 2021-09-07 891 goto kernel_zone; 445fcf7c721450d David Hildenbrand 2021-09-07 892 online_pages += PAGES_PER_SECTION; 445fcf7c721450d David Hildenbrand 2021-09-07 893 } 445fcf7c721450d David Hildenbrand 2021-09-07 894 } 445fcf7c721450d David Hildenbrand 2021-09-07 895 445fcf7c721450d David Hildenbrand 2021-09-07 896 /* 445fcf7c721450d David Hildenbrand 2021-09-07 897 * Online MOVABLE if we could *currently* online all remaining parts 445fcf7c721450d David Hildenbrand 2021-09-07 898 * MOVABLE. We expect to (add+) online them immediately next, so if 445fcf7c721450d David Hildenbrand 2021-09-07 899 * nobody interferes, all will be MOVABLE if possible. 445fcf7c721450d David Hildenbrand 2021-09-07 900 */ 445fcf7c721450d David Hildenbrand 2021-09-07 901 nr_pages = max_pages - online_pages; e83a437faa625ef David Hildenbrand 2021-09-07 902 if (!auto_movable_can_online_movable(NUMA_NO_NODE, nr_pages)) e83a437faa625ef David Hildenbrand 2021-09-07 903 goto kernel_zone; e83a437faa625ef David Hildenbrand 2021-09-07 904 --- 0-DAY CI Kernel Test Service https://lists.01.org/hyperkitty/list/kbuild-...@lists.01.org _______________________________________________ kbuild mailing list -- kbuild@lists.01.org To unsubscribe send an email to kbuild-le...@lists.01.org
