From: Cannon Matthews <cannonmatthews@xxxxxxxxxx> Subject: mm/hugetlb.c: don't zero 1GiB bootmem pages When using 1GiB pages during early boot, use the new memblock_virt_alloc_try_nid_raw() to allocate memory without zeroing it. Zeroing out hundreds or thousands of GiB in a single core memset() call is very slow, and can make early boot last upwards of 20-30 minutes on multi TiB machines. The memory does not need to be zero'd as the hugetlb pages are always zero'd on page fault. Tested: Booted with ~3800 1G pages, and it booted successfully in roughly the same amount of time as with 0, as opposed to the 25+ minutes it would take before. Link: http://lkml.kernel.org/r/20180711213313.92481-1-cannonmatthews@xxxxxxxxxx Signed-off-by: Cannon Matthews <cannonmatthews@xxxxxxxxxx> Acked-by: Mike Kravetz <mike.kravetz@xxxxxxxxxx> Acked-by: Michal Hocko <mhocko@xxxxxxxx> Cc: Andres Lagar-Cavilla <andreslc@xxxxxxxxxx> Cc: Peter Feiner <pfeiner@xxxxxxxxxx> Cc: David Matlack <dmatlack@xxxxxxxxxx> Cc: Greg Thelen <gthelen@xxxxxxxxxx> Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx> --- mm/hugetlb.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) --- a/mm/hugetlb.c~mm-hugetlb-dont-zero-1gib-bootmem-pages +++ a/mm/hugetlb.c @@ -2101,7 +2101,7 @@ int __alloc_bootmem_huge_page(struct hst for_each_node_mask_to_alloc(h, nr_nodes, node, &node_states[N_MEMORY]) { void *addr; - addr = memblock_virt_alloc_try_nid_nopanic( + addr = memblock_virt_alloc_try_nid_raw( huge_page_size(h), huge_page_size(h), 0, BOOTMEM_ALLOC_ACCESSIBLE, node); if (addr) { @@ -2119,6 +2119,7 @@ int __alloc_bootmem_huge_page(struct hst found: BUG_ON(!IS_ALIGNED(virt_to_phys(m), huge_page_size(h))); /* Put them into a private list first because mem_map is not up yet */ + INIT_LIST_HEAD(&m->list); list_add(&m->list, &huge_boot_pages); m->hstate = h; return 1; _
