x86_64: prepare shared mm/init.c
authorThomas Gleixner <tglx@linutronix.de>
Thu, 11 Oct 2007 09:16:11 +0000 (11:16 +0200)
committerThomas Gleixner <tglx@linutronix.de>
Thu, 11 Oct 2007 09:16:11 +0000 (11:16 +0200)
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
arch/x86_64/mm/Makefile
arch/x86_64/mm/init.c [deleted file]
arch/x86_64/mm/init_64.c [new file with mode: 0644]

index c400b19837856b449af917fc540227591b2f41f6..1e2ccf009d4c399f780087ecd08228a633f9abbb 100644 (file)
@@ -2,7 +2,7 @@
 # Makefile for the linux x86_64-specific parts of the memory manager.
 #
 
-obj-y   := init.o fault.o ioremap_64.o extable.o pageattr.o mmap.o
+obj-y   := init_64.o fault.o ioremap_64.o extable.o pageattr.o mmap.o
 obj-$(CONFIG_HUGETLB_PAGE) += hugetlbpage.o
 obj-$(CONFIG_NUMA) += numa.o
 obj-$(CONFIG_K8_NUMA) += k8topology.o
diff --git a/arch/x86_64/mm/init.c b/arch/x86_64/mm/init.c
deleted file mode 100644 (file)
index 458893b..0000000
+++ /dev/null
@@ -1,750 +0,0 @@
-/*
- *  linux/arch/x86_64/mm/init.c
- *
- *  Copyright (C) 1995  Linus Torvalds
- *  Copyright (C) 2000  Pavel Machek <pavel@suse.cz>
- *  Copyright (C) 2002,2003 Andi Kleen <ak@suse.de>
- */
-
-#include <linux/signal.h>
-#include <linux/sched.h>
-#include <linux/kernel.h>
-#include <linux/errno.h>
-#include <linux/string.h>
-#include <linux/types.h>
-#include <linux/ptrace.h>
-#include <linux/mman.h>
-#include <linux/mm.h>
-#include <linux/swap.h>
-#include <linux/smp.h>
-#include <linux/init.h>
-#include <linux/pagemap.h>
-#include <linux/bootmem.h>
-#include <linux/proc_fs.h>
-#include <linux/pci.h>
-#include <linux/pfn.h>
-#include <linux/poison.h>
-#include <linux/dma-mapping.h>
-#include <linux/module.h>
-#include <linux/memory_hotplug.h>
-#include <linux/nmi.h>
-
-#include <asm/processor.h>
-#include <asm/system.h>
-#include <asm/uaccess.h>
-#include <asm/pgtable.h>
-#include <asm/pgalloc.h>
-#include <asm/dma.h>
-#include <asm/fixmap.h>
-#include <asm/e820.h>
-#include <asm/apic.h>
-#include <asm/tlb.h>
-#include <asm/mmu_context.h>
-#include <asm/proto.h>
-#include <asm/smp.h>
-#include <asm/sections.h>
-
-#ifndef Dprintk
-#define Dprintk(x...)
-#endif
-
-const struct dma_mapping_ops* dma_ops;
-EXPORT_SYMBOL(dma_ops);
-
-static unsigned long dma_reserve __initdata;
-
-DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
-
-/*
- * NOTE: pagetable_init alloc all the fixmap pagetables contiguous on the
- * physical space so we can cache the place of the first one and move
- * around without checking the pgd every time.
- */
-
-void show_mem(void)
-{
-       long i, total = 0, reserved = 0;
-       long shared = 0, cached = 0;
-       pg_data_t *pgdat;
-       struct page *page;
-
-       printk(KERN_INFO "Mem-info:\n");
-       show_free_areas();
-       printk(KERN_INFO "Free swap:       %6ldkB\n", nr_swap_pages<<(PAGE_SHIFT-10));
-
-       for_each_online_pgdat(pgdat) {
-               for (i = 0; i < pgdat->node_spanned_pages; ++i) {
-                       /* this loop can take a while with 256 GB and 4k pages
-                          so update the NMI watchdog */
-                       if (unlikely(i % MAX_ORDER_NR_PAGES == 0)) {
-                               touch_nmi_watchdog();
-                       }
-                       if (!pfn_valid(pgdat->node_start_pfn + i))
-                               continue;
-                       page = pfn_to_page(pgdat->node_start_pfn + i);
-                       total++;
-                       if (PageReserved(page))
-                               reserved++;
-                       else if (PageSwapCache(page))
-                               cached++;
-                       else if (page_count(page))
-                               shared += page_count(page) - 1;
-               }
-       }
-       printk(KERN_INFO "%lu pages of RAM\n", total);
-       printk(KERN_INFO "%lu reserved pages\n",reserved);
-       printk(KERN_INFO "%lu pages shared\n",shared);
-       printk(KERN_INFO "%lu pages swap cached\n",cached);
-}
-
-int after_bootmem;
-
-static __init void *spp_getpage(void)
-{ 
-       void *ptr;
-       if (after_bootmem)
-               ptr = (void *) get_zeroed_page(GFP_ATOMIC); 
-       else
-               ptr = alloc_bootmem_pages(PAGE_SIZE);
-       if (!ptr || ((unsigned long)ptr & ~PAGE_MASK))
-               panic("set_pte_phys: cannot allocate page data %s\n", after_bootmem?"after bootmem":"");
-
-       Dprintk("spp_getpage %p\n", ptr);
-       return ptr;
-} 
-
-static __init void set_pte_phys(unsigned long vaddr,
-                        unsigned long phys, pgprot_t prot)
-{
-       pgd_t *pgd;
-       pud_t *pud;
-       pmd_t *pmd;
-       pte_t *pte, new_pte;
-
-       Dprintk("set_pte_phys %lx to %lx\n", vaddr, phys);
-
-       pgd = pgd_offset_k(vaddr);
-       if (pgd_none(*pgd)) {
-               printk("PGD FIXMAP MISSING, it should be setup in head.S!\n");
-               return;
-       }
-       pud = pud_offset(pgd, vaddr);
-       if (pud_none(*pud)) {
-               pmd = (pmd_t *) spp_getpage(); 
-               set_pud(pud, __pud(__pa(pmd) | _KERNPG_TABLE | _PAGE_USER));
-               if (pmd != pmd_offset(pud, 0)) {
-                       printk("PAGETABLE BUG #01! %p <-> %p\n", pmd, pmd_offset(pud,0));
-                       return;
-               }
-       }
-       pmd = pmd_offset(pud, vaddr);
-       if (pmd_none(*pmd)) {
-               pte = (pte_t *) spp_getpage();
-               set_pmd(pmd, __pmd(__pa(pte) | _KERNPG_TABLE | _PAGE_USER));
-               if (pte != pte_offset_kernel(pmd, 0)) {
-                       printk("PAGETABLE BUG #02!\n");
-                       return;
-               }
-       }
-       new_pte = pfn_pte(phys >> PAGE_SHIFT, prot);
-
-       pte = pte_offset_kernel(pmd, vaddr);
-       if (!pte_none(*pte) &&
-           pte_val(*pte) != (pte_val(new_pte) & __supported_pte_mask))
-               pte_ERROR(*pte);
-       set_pte(pte, new_pte);
-
-       /*
-        * It's enough to flush this one mapping.
-        * (PGE mappings get flushed as well)
-        */
-       __flush_tlb_one(vaddr);
-}
-
-/* NOTE: this is meant to be run only at boot */
-void __init 
-__set_fixmap (enum fixed_addresses idx, unsigned long phys, pgprot_t prot)
-{
-       unsigned long address = __fix_to_virt(idx);
-
-       if (idx >= __end_of_fixed_addresses) {
-               printk("Invalid __set_fixmap\n");
-               return;
-       }
-       set_pte_phys(address, phys, prot);
-}
-
-unsigned long __meminitdata table_start, table_end;
-
-static __meminit void *alloc_low_page(unsigned long *phys)
-{ 
-       unsigned long pfn = table_end++;
-       void *adr;
-
-       if (after_bootmem) {
-               adr = (void *)get_zeroed_page(GFP_ATOMIC);
-               *phys = __pa(adr);
-               return adr;
-       }
-
-       if (pfn >= end_pfn) 
-               panic("alloc_low_page: ran out of memory"); 
-
-       adr = early_ioremap(pfn * PAGE_SIZE, PAGE_SIZE);
-       memset(adr, 0, PAGE_SIZE);
-       *phys  = pfn * PAGE_SIZE;
-       return adr;
-}
-
-static __meminit void unmap_low_page(void *adr)
-{ 
-
-       if (after_bootmem)
-               return;
-
-       early_iounmap(adr, PAGE_SIZE);
-} 
-
-/* Must run before zap_low_mappings */
-__meminit void *early_ioremap(unsigned long addr, unsigned long size)
-{
-       unsigned long vaddr;
-       pmd_t *pmd, *last_pmd;
-       int i, pmds;
-
-       pmds = ((addr & ~PMD_MASK) + size + ~PMD_MASK) / PMD_SIZE;
-       vaddr = __START_KERNEL_map;
-       pmd = level2_kernel_pgt;
-       last_pmd = level2_kernel_pgt + PTRS_PER_PMD - 1;
-       for (; pmd <= last_pmd; pmd++, vaddr += PMD_SIZE) {
-               for (i = 0; i < pmds; i++) {
-                       if (pmd_present(pmd[i]))
-                               goto next;
-               }
-               vaddr += addr & ~PMD_MASK;
-               addr &= PMD_MASK;
-               for (i = 0; i < pmds; i++, addr += PMD_SIZE)
-                       set_pmd(pmd + i,__pmd(addr | _KERNPG_TABLE | _PAGE_PSE));
-               __flush_tlb();
-               return (void *)vaddr;
-       next:
-               ;
-       }
-       printk("early_ioremap(0x%lx, %lu) failed\n", addr, size);
-       return NULL;
-}
-
-/* To avoid virtual aliases later */
-__meminit void early_iounmap(void *addr, unsigned long size)
-{
-       unsigned long vaddr;
-       pmd_t *pmd;
-       int i, pmds;
-
-       vaddr = (unsigned long)addr;
-       pmds = ((vaddr & ~PMD_MASK) + size + ~PMD_MASK) / PMD_SIZE;
-       pmd = level2_kernel_pgt + pmd_index(vaddr);
-       for (i = 0; i < pmds; i++)
-               pmd_clear(pmd + i);
-       __flush_tlb();
-}
-
-static void __meminit
-phys_pmd_init(pmd_t *pmd_page, unsigned long address, unsigned long end)
-{
-       int i = pmd_index(address);
-
-       for (; i < PTRS_PER_PMD; i++, address += PMD_SIZE) {
-               unsigned long entry;
-               pmd_t *pmd = pmd_page + pmd_index(address);
-
-               if (address >= end) {
-                       if (!after_bootmem)
-                               for (; i < PTRS_PER_PMD; i++, pmd++)
-                                       set_pmd(pmd, __pmd(0));
-                       break;
-               }
-
-               if (pmd_val(*pmd))
-                       continue;
-
-               entry = _PAGE_NX|_PAGE_PSE|_KERNPG_TABLE|_PAGE_GLOBAL|address;
-               entry &= __supported_pte_mask;
-               set_pmd(pmd, __pmd(entry));
-       }
-}
-
-static void __meminit
-phys_pmd_update(pud_t *pud, unsigned long address, unsigned long end)
-{
-       pmd_t *pmd = pmd_offset(pud,0);
-       spin_lock(&init_mm.page_table_lock);
-       phys_pmd_init(pmd, address, end);
-       spin_unlock(&init_mm.page_table_lock);
-       __flush_tlb_all();
-}
-
-static void __meminit phys_pud_init(pud_t *pud_page, unsigned long addr, unsigned long end)
-{ 
-       int i = pud_index(addr);
-
-
-       for (; i < PTRS_PER_PUD; i++, addr = (addr & PUD_MASK) + PUD_SIZE ) {
-               unsigned long pmd_phys;
-               pud_t *pud = pud_page + pud_index(addr);
-               pmd_t *pmd;
-
-               if (addr >= end)
-                       break;
-
-               if (!after_bootmem && !e820_any_mapped(addr,addr+PUD_SIZE,0)) {
-                       set_pud(pud, __pud(0)); 
-                       continue;
-               } 
-
-               if (pud_val(*pud)) {
-                       phys_pmd_update(pud, addr, end);
-                       continue;
-               }
-
-               pmd = alloc_low_page(&pmd_phys);
-               spin_lock(&init_mm.page_table_lock);
-               set_pud(pud, __pud(pmd_phys | _KERNPG_TABLE));
-               phys_pmd_init(pmd, addr, end);
-               spin_unlock(&init_mm.page_table_lock);
-               unmap_low_page(pmd);
-       }
-       __flush_tlb();
-} 
-
-static void __init find_early_table_space(unsigned long end)
-{
-       unsigned long puds, pmds, tables, start;
-
-       puds = (end + PUD_SIZE - 1) >> PUD_SHIFT;
-       pmds = (end + PMD_SIZE - 1) >> PMD_SHIFT;
-       tables = round_up(puds * sizeof(pud_t), PAGE_SIZE) +
-                round_up(pmds * sizeof(pmd_t), PAGE_SIZE);
-
-       /* RED-PEN putting page tables only on node 0 could
-          cause a hotspot and fill up ZONE_DMA. The page tables
-          need roughly 0.5KB per GB. */
-       start = 0x8000;
-       table_start = find_e820_area(start, end, tables);
-       if (table_start == -1UL)
-               panic("Cannot find space for the kernel page tables");
-
-       table_start >>= PAGE_SHIFT;
-       table_end = table_start;
-
-       early_printk("kernel direct mapping tables up to %lx @ %lx-%lx\n",
-               end, table_start << PAGE_SHIFT,
-               (table_start << PAGE_SHIFT) + tables);
-}
-
-/* Setup the direct mapping of the physical memory at PAGE_OFFSET.
-   This runs before bootmem is initialized and gets pages directly from the 
-   physical memory. To access them they are temporarily mapped. */
-void __meminit init_memory_mapping(unsigned long start, unsigned long end)
-{ 
-       unsigned long next; 
-
-       Dprintk("init_memory_mapping\n");
-
-       /* 
-        * Find space for the kernel direct mapping tables.
-        * Later we should allocate these tables in the local node of the memory
-        * mapped.  Unfortunately this is done currently before the nodes are 
-        * discovered.
-        */
-       if (!after_bootmem)
-               find_early_table_space(end);
-
-       start = (unsigned long)__va(start);
-       end = (unsigned long)__va(end);
-
-       for (; start < end; start = next) {
-               unsigned long pud_phys; 
-               pgd_t *pgd = pgd_offset_k(start);
-               pud_t *pud;
-
-               if (after_bootmem)
-                       pud = pud_offset(pgd, start & PGDIR_MASK);
-               else
-                       pud = alloc_low_page(&pud_phys);
-
-               next = start + PGDIR_SIZE;
-               if (next > end) 
-                       next = end; 
-               phys_pud_init(pud, __pa(start), __pa(next));
-               if (!after_bootmem)
-                       set_pgd(pgd_offset_k(start), mk_kernel_pgd(pud_phys));
-               unmap_low_page(pud);
-       } 
-
-       if (!after_bootmem)
-               mmu_cr4_features = read_cr4();
-       __flush_tlb_all();
-}
-
-#ifndef CONFIG_NUMA
-void __init paging_init(void)
-{
-       unsigned long max_zone_pfns[MAX_NR_ZONES];
-       memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
-       max_zone_pfns[ZONE_DMA] = MAX_DMA_PFN;
-       max_zone_pfns[ZONE_DMA32] = MAX_DMA32_PFN;
-       max_zone_pfns[ZONE_NORMAL] = end_pfn;
-
-       memory_present(0, 0, end_pfn);
-       sparse_init();
-       free_area_init_nodes(max_zone_pfns);
-}
-#endif
-
-/* Unmap a kernel mapping if it exists. This is useful to avoid prefetches
-   from the CPU leading to inconsistent cache lines. address and size
-   must be aligned to 2MB boundaries. 
-   Does nothing when the mapping doesn't exist. */
-void __init clear_kernel_mapping(unsigned long address, unsigned long size) 
-{
-       unsigned long end = address + size;
-
-       BUG_ON(address & ~LARGE_PAGE_MASK);
-       BUG_ON(size & ~LARGE_PAGE_MASK); 
-       
-       for (; address < end; address += LARGE_PAGE_SIZE) { 
-               pgd_t *pgd = pgd_offset_k(address);
-               pud_t *pud;
-               pmd_t *pmd;
-               if (pgd_none(*pgd))
-                       continue;
-               pud = pud_offset(pgd, address);
-               if (pud_none(*pud))
-                       continue; 
-               pmd = pmd_offset(pud, address);
-               if (!pmd || pmd_none(*pmd))
-                       continue; 
-               if (0 == (pmd_val(*pmd) & _PAGE_PSE)) { 
-                       /* Could handle this, but it should not happen currently. */
-                       printk(KERN_ERR 
-              "clear_kernel_mapping: mapping has been split. will leak memory\n"); 
-                       pmd_ERROR(*pmd); 
-               }
-               set_pmd(pmd, __pmd(0));                 
-       }
-       __flush_tlb_all();
-} 
-
-/*
- * Memory hotplug specific functions
- */
-void online_page(struct page *page)
-{
-       ClearPageReserved(page);
-       init_page_count(page);
-       __free_page(page);
-       totalram_pages++;
-       num_physpages++;
-}
-
-#ifdef CONFIG_MEMORY_HOTPLUG
-/*
- * Memory is added always to NORMAL zone. This means you will never get
- * additional DMA/DMA32 memory.
- */
-int arch_add_memory(int nid, u64 start, u64 size)
-{
-       struct pglist_data *pgdat = NODE_DATA(nid);
-       struct zone *zone = pgdat->node_zones + ZONE_NORMAL;
-       unsigned long start_pfn = start >> PAGE_SHIFT;
-       unsigned long nr_pages = size >> PAGE_SHIFT;
-       int ret;
-
-       init_memory_mapping(start, (start + size -1));
-
-       ret = __add_pages(zone, start_pfn, nr_pages);
-       if (ret)
-               goto error;
-
-       return ret;
-error:
-       printk("%s: Problem encountered in __add_pages!\n", __func__);
-       return ret;
-}
-EXPORT_SYMBOL_GPL(arch_add_memory);
-
-int remove_memory(u64 start, u64 size)
-{
-       return -EINVAL;
-}
-EXPORT_SYMBOL_GPL(remove_memory);
-
-#if !defined(CONFIG_ACPI_NUMA) && defined(CONFIG_NUMA)
-int memory_add_physaddr_to_nid(u64 start)
-{
-       return 0;
-}
-EXPORT_SYMBOL_GPL(memory_add_physaddr_to_nid);
-#endif
-
-#endif /* CONFIG_MEMORY_HOTPLUG */
-
-#ifdef CONFIG_MEMORY_HOTPLUG_RESERVE
-/*
- * Memory Hotadd without sparsemem. The mem_maps have been allocated in advance,
- * just online the pages.
- */
-int __add_pages(struct zone *z, unsigned long start_pfn, unsigned long nr_pages)
-{
-       int err = -EIO;
-       unsigned long pfn;
-       unsigned long total = 0, mem = 0;
-       for (pfn = start_pfn; pfn < start_pfn + nr_pages; pfn++) {
-               if (pfn_valid(pfn)) {
-                       online_page(pfn_to_page(pfn));
-                       err = 0;
-                       mem++;
-               }
-               total++;
-       }
-       if (!err) {
-               z->spanned_pages += total;
-               z->present_pages += mem;
-               z->zone_pgdat->node_spanned_pages += total;
-               z->zone_pgdat->node_present_pages += mem;
-       }
-       return err;
-}
-#endif
-
-static struct kcore_list kcore_mem, kcore_vmalloc, kcore_kernel, kcore_modules,
-                        kcore_vsyscall;
-
-void __init mem_init(void)
-{
-       long codesize, reservedpages, datasize, initsize;
-
-       pci_iommu_alloc();
-
-       /* clear the zero-page */
-       memset(empty_zero_page, 0, PAGE_SIZE);
-
-       reservedpages = 0;
-
-       /* this will put all low memory onto the freelists */
-#ifdef CONFIG_NUMA
-       totalram_pages = numa_free_all_bootmem();
-#else
-       totalram_pages = free_all_bootmem();
-#endif
-       reservedpages = end_pfn - totalram_pages -
-                                       absent_pages_in_range(0, end_pfn);
-
-       after_bootmem = 1;
-
-       codesize =  (unsigned long) &_etext - (unsigned long) &_text;
-       datasize =  (unsigned long) &_edata - (unsigned long) &_etext;
-       initsize =  (unsigned long) &__init_end - (unsigned long) &__init_begin;
-
-       /* Register memory areas for /proc/kcore */
-       kclist_add(&kcore_mem, __va(0), max_low_pfn << PAGE_SHIFT); 
-       kclist_add(&kcore_vmalloc, (void *)VMALLOC_START, 
-                  VMALLOC_END-VMALLOC_START);
-       kclist_add(&kcore_kernel, &_stext, _end - _stext);
-       kclist_add(&kcore_modules, (void *)MODULES_VADDR, MODULES_LEN);
-       kclist_add(&kcore_vsyscall, (void *)VSYSCALL_START, 
-                                VSYSCALL_END - VSYSCALL_START);
-
-       printk("Memory: %luk/%luk available (%ldk kernel code, %ldk reserved, %ldk data, %ldk init)\n",
-               (unsigned long) nr_free_pages() << (PAGE_SHIFT-10),
-               end_pfn << (PAGE_SHIFT-10),
-               codesize >> 10,
-               reservedpages << (PAGE_SHIFT-10),
-               datasize >> 10,
-               initsize >> 10);
-}
-
-void free_init_pages(char *what, unsigned long begin, unsigned long end)
-{
-       unsigned long addr;
-
-       if (begin >= end)
-               return;
-
-       printk(KERN_INFO "Freeing %s: %luk freed\n", what, (end - begin) >> 10);
-       for (addr = begin; addr < end; addr += PAGE_SIZE) {
-               ClearPageReserved(virt_to_page(addr));
-               init_page_count(virt_to_page(addr));
-               memset((void *)(addr & ~(PAGE_SIZE-1)),
-                       POISON_FREE_INITMEM, PAGE_SIZE);
-               if (addr >= __START_KERNEL_map)
-                       change_page_attr_addr(addr, 1, __pgprot(0));
-               free_page(addr);
-               totalram_pages++;
-       }
-       if (addr > __START_KERNEL_map)
-               global_flush_tlb();
-}
-
-void free_initmem(void)
-{
-       free_init_pages("unused kernel memory",
-                       (unsigned long)(&__init_begin),
-                       (unsigned long)(&__init_end));
-}
-
-#ifdef CONFIG_DEBUG_RODATA
-
-void mark_rodata_ro(void)
-{
-       unsigned long start = (unsigned long)_stext, end;
-
-#ifdef CONFIG_HOTPLUG_CPU
-       /* It must still be possible to apply SMP alternatives. */
-       if (num_possible_cpus() > 1)
-               start = (unsigned long)_etext;
-#endif
-
-#ifdef CONFIG_KPROBES
-       start = (unsigned long)__start_rodata;
-#endif
-       
-       end = (unsigned long)__end_rodata;
-       start = (start + PAGE_SIZE - 1) & PAGE_MASK;
-       end &= PAGE_MASK;
-       if (end <= start)
-               return;
-
-       change_page_attr_addr(start, (end - start) >> PAGE_SHIFT, PAGE_KERNEL_RO);
-
-       printk(KERN_INFO "Write protecting the kernel read-only data: %luk\n",
-              (end - start) >> 10);
-
-       /*
-        * change_page_attr_addr() requires a global_flush_tlb() call after it.
-        * We do this after the printk so that if something went wrong in the
-        * change, the printk gets out at least to give a better debug hint
-        * of who is the culprit.
-        */
-       global_flush_tlb();
-}
-#endif
-
-#ifdef CONFIG_BLK_DEV_INITRD
-void free_initrd_mem(unsigned long start, unsigned long end)
-{
-       free_init_pages("initrd memory", start, end);
-}
-#endif
-
-void __init reserve_bootmem_generic(unsigned long phys, unsigned len) 
-{ 
-#ifdef CONFIG_NUMA
-       int nid = phys_to_nid(phys);
-#endif
-       unsigned long pfn = phys >> PAGE_SHIFT;
-       if (pfn >= end_pfn) {
-               /* This can happen with kdump kernels when accessing firmware
-                  tables. */
-               if (pfn < end_pfn_map)
-                       return;
-               printk(KERN_ERR "reserve_bootmem: illegal reserve %lx %u\n",
-                               phys, len);
-               return;
-       }
-
-       /* Should check here against the e820 map to avoid double free */
-#ifdef CONFIG_NUMA
-       reserve_bootmem_node(NODE_DATA(nid), phys, len);
-#else                  
-       reserve_bootmem(phys, len);    
-#endif
-       if (phys+len <= MAX_DMA_PFN*PAGE_SIZE) {
-               dma_reserve += len / PAGE_SIZE;
-               set_dma_reserve(dma_reserve);
-       }
-}
-
-int kern_addr_valid(unsigned long addr) 
-{ 
-       unsigned long above = ((long)addr) >> __VIRTUAL_MASK_SHIFT;
-       pgd_t *pgd;
-       pud_t *pud;
-       pmd_t *pmd;
-       pte_t *pte;
-
-       if (above != 0 && above != -1UL)
-               return 0; 
-       
-       pgd = pgd_offset_k(addr);
-       if (pgd_none(*pgd))
-               return 0;
-
-       pud = pud_offset(pgd, addr);
-       if (pud_none(*pud))
-               return 0; 
-
-       pmd = pmd_offset(pud, addr);
-       if (pmd_none(*pmd))
-               return 0;
-       if (pmd_large(*pmd))
-               return pfn_valid(pmd_pfn(*pmd));
-
-       pte = pte_offset_kernel(pmd, addr);
-       if (pte_none(*pte))
-               return 0;
-       return pfn_valid(pte_pfn(*pte));
-}
-
-/* A pseudo VMA to allow ptrace access for the vsyscall page.  This only
-   covers the 64bit vsyscall page now. 32bit has a real VMA now and does
-   not need special handling anymore. */
-
-static struct vm_area_struct gate_vma = {
-       .vm_start = VSYSCALL_START,
-       .vm_end = VSYSCALL_START + (VSYSCALL_MAPPED_PAGES << PAGE_SHIFT),
-       .vm_page_prot = PAGE_READONLY_EXEC,
-       .vm_flags = VM_READ | VM_EXEC
-};
-
-struct vm_area_struct *get_gate_vma(struct task_struct *tsk)
-{
-#ifdef CONFIG_IA32_EMULATION
-       if (test_tsk_thread_flag(tsk, TIF_IA32))
-               return NULL;
-#endif
-       return &gate_vma;
-}
-
-int in_gate_area(struct task_struct *task, unsigned long addr)
-{
-       struct vm_area_struct *vma = get_gate_vma(task);
-       if (!vma)
-               return 0;
-       return (addr >= vma->vm_start) && (addr < vma->vm_end);
-}
-
-/* Use this when you have no reliable task/vma, typically from interrupt
- * context.  It is less reliable than using the task's vma and may give
- * false positives.
- */
-int in_gate_area_no_task(unsigned long addr)
-{
-       return (addr >= VSYSCALL_START) && (addr < VSYSCALL_END);
-}
-
-void * __init alloc_bootmem_high_node(pg_data_t *pgdat, unsigned long size)
-{
-       return __alloc_bootmem_core(pgdat->bdata, size,
-                       SMP_CACHE_BYTES, (4UL*1024*1024*1024), 0);
-}
-
-const char *arch_vma_name(struct vm_area_struct *vma)
-{
-       if (vma->vm_mm && vma->vm_start == (long)vma->vm_mm->context.vdso)
-               return "[vdso]";
-       if (vma == &gate_vma)
-               return "[vsyscall]";
-       return NULL;
-}
diff --git a/arch/x86_64/mm/init_64.c b/arch/x86_64/mm/init_64.c
new file mode 100644 (file)
index 0000000..458893b
--- /dev/null
@@ -0,0 +1,750 @@
+/*
+ *  linux/arch/x86_64/mm/init.c
+ *
+ *  Copyright (C) 1995  Linus Torvalds
+ *  Copyright (C) 2000  Pavel Machek <pavel@suse.cz>
+ *  Copyright (C) 2002,2003 Andi Kleen <ak@suse.de>
+ */
+
+#include <linux/signal.h>
+#include <linux/sched.h>
+#include <linux/kernel.h>
+#include <linux/errno.h>
+#include <linux/string.h>
+#include <linux/types.h>
+#include <linux/ptrace.h>
+#include <linux/mman.h>
+#include <linux/mm.h>
+#include <linux/swap.h>
+#include <linux/smp.h>
+#include <linux/init.h>
+#include <linux/pagemap.h>
+#include <linux/bootmem.h>
+#include <linux/proc_fs.h>
+#include <linux/pci.h>
+#include <linux/pfn.h>
+#include <linux/poison.h>
+#include <linux/dma-mapping.h>
+#include <linux/module.h>
+#include <linux/memory_hotplug.h>
+#include <linux/nmi.h>
+
+#include <asm/processor.h>
+#include <asm/system.h>
+#include <asm/uaccess.h>
+#include <asm/pgtable.h>
+#include <asm/pgalloc.h>
+#include <asm/dma.h>
+#include <asm/fixmap.h>
+#include <asm/e820.h>
+#include <asm/apic.h>
+#include <asm/tlb.h>
+#include <asm/mmu_context.h>
+#include <asm/proto.h>
+#include <asm/smp.h>
+#include <asm/sections.h>
+
+#ifndef Dprintk
+#define Dprintk(x...)
+#endif
+
+const struct dma_mapping_ops* dma_ops;
+EXPORT_SYMBOL(dma_ops);
+
+static unsigned long dma_reserve __initdata;
+
+DEFINE_PER_CPU(struct mmu_gather, mmu_gathers);
+
+/*
+ * NOTE: pagetable_init alloc all the fixmap pagetables contiguous on the
+ * physical space so we can cache the place of the first one and move
+ * around without checking the pgd every time.
+ */
+
+void show_mem(void)
+{
+       long i, total = 0, reserved = 0;
+       long shared = 0, cached = 0;
+       pg_data_t *pgdat;
+       struct page *page;
+
+       printk(KERN_INFO "Mem-info:\n");
+       show_free_areas();
+       printk(KERN_INFO "Free swap:       %6ldkB\n", nr_swap_pages<<(PAGE_SHIFT-10));
+
+       for_each_online_pgdat(pgdat) {
+               for (i = 0; i < pgdat->node_spanned_pages; ++i) {
+                       /* this loop can take a while with 256 GB and 4k pages
+                          so update the NMI watchdog */
+                       if (unlikely(i % MAX_ORDER_NR_PAGES == 0)) {
+                               touch_nmi_watchdog();
+                       }
+                       if (!pfn_valid(pgdat->node_start_pfn + i))
+                               continue;
+                       page = pfn_to_page(pgdat->node_start_pfn + i);
+                       total++;
+                       if (PageReserved(page))
+                               reserved++;
+                       else if (PageSwapCache(page))
+                               cached++;
+                       else if (page_count(page))
+                               shared += page_count(page) - 1;
+               }
+       }
+       printk(KERN_INFO "%lu pages of RAM\n", total);
+       printk(KERN_INFO "%lu reserved pages\n",reserved);
+       printk(KERN_INFO "%lu pages shared\n",shared);
+       printk(KERN_INFO "%lu pages swap cached\n",cached);
+}
+
+int after_bootmem;
+
+static __init void *spp_getpage(void)
+{ 
+       void *ptr;
+       if (after_bootmem)
+               ptr = (void *) get_zeroed_page(GFP_ATOMIC); 
+       else
+               ptr = alloc_bootmem_pages(PAGE_SIZE);
+       if (!ptr || ((unsigned long)ptr & ~PAGE_MASK))
+               panic("set_pte_phys: cannot allocate page data %s\n", after_bootmem?"after bootmem":"");
+
+       Dprintk("spp_getpage %p\n", ptr);
+       return ptr;
+} 
+
+static __init void set_pte_phys(unsigned long vaddr,
+                        unsigned long phys, pgprot_t prot)
+{
+       pgd_t *pgd;
+       pud_t *pud;
+       pmd_t *pmd;
+       pte_t *pte, new_pte;
+
+       Dprintk("set_pte_phys %lx to %lx\n", vaddr, phys);
+
+       pgd = pgd_offset_k(vaddr);
+       if (pgd_none(*pgd)) {
+               printk("PGD FIXMAP MISSING, it should be setup in head.S!\n");
+               return;
+       }
+       pud = pud_offset(pgd, vaddr);
+       if (pud_none(*pud)) {
+               pmd = (pmd_t *) spp_getpage(); 
+               set_pud(pud, __pud(__pa(pmd) | _KERNPG_TABLE | _PAGE_USER));
+               if (pmd != pmd_offset(pud, 0)) {
+                       printk("PAGETABLE BUG #01! %p <-> %p\n", pmd, pmd_offset(pud,0));
+                       return;
+               }
+       }
+       pmd = pmd_offset(pud, vaddr);
+       if (pmd_none(*pmd)) {
+               pte = (pte_t *) spp_getpage();
+               set_pmd(pmd, __pmd(__pa(pte) | _KERNPG_TABLE | _PAGE_USER));
+               if (pte != pte_offset_kernel(pmd, 0)) {
+                       printk("PAGETABLE BUG #02!\n");
+                       return;
+               }
+       }
+       new_pte = pfn_pte(phys >> PAGE_SHIFT, prot);
+
+       pte = pte_offset_kernel(pmd, vaddr);
+       if (!pte_none(*pte) &&
+           pte_val(*pte) != (pte_val(new_pte) & __supported_pte_mask))
+               pte_ERROR(*pte);
+       set_pte(pte, new_pte);
+
+       /*
+        * It's enough to flush this one mapping.
+        * (PGE mappings get flushed as well)
+        */
+       __flush_tlb_one(vaddr);
+}
+
+/* NOTE: this is meant to be run only at boot */
+void __init 
+__set_fixmap (enum fixed_addresses idx, unsigned long phys, pgprot_t prot)
+{
+       unsigned long address = __fix_to_virt(idx);
+
+       if (idx >= __end_of_fixed_addresses) {
+               printk("Invalid __set_fixmap\n");
+               return;
+       }
+       set_pte_phys(address, phys, prot);
+}
+
+unsigned long __meminitdata table_start, table_end;
+
+static __meminit void *alloc_low_page(unsigned long *phys)
+{ 
+       unsigned long pfn = table_end++;
+       void *adr;
+
+       if (after_bootmem) {
+               adr = (void *)get_zeroed_page(GFP_ATOMIC);
+               *phys = __pa(adr);
+               return adr;
+       }
+
+       if (pfn >= end_pfn) 
+               panic("alloc_low_page: ran out of memory"); 
+
+       adr = early_ioremap(pfn * PAGE_SIZE, PAGE_SIZE);
+       memset(adr, 0, PAGE_SIZE);
+       *phys  = pfn * PAGE_SIZE;
+       return adr;
+}
+
+static __meminit void unmap_low_page(void *adr)
+{ 
+
+       if (after_bootmem)
+               return;
+
+       early_iounmap(adr, PAGE_SIZE);
+} 
+
+/* Must run before zap_low_mappings */
+__meminit void *early_ioremap(unsigned long addr, unsigned long size)
+{
+       unsigned long vaddr;
+       pmd_t *pmd, *last_pmd;
+       int i, pmds;
+
+       pmds = ((addr & ~PMD_MASK) + size + ~PMD_MASK) / PMD_SIZE;
+       vaddr = __START_KERNEL_map;
+       pmd = level2_kernel_pgt;
+       last_pmd = level2_kernel_pgt + PTRS_PER_PMD - 1;
+       for (; pmd <= last_pmd; pmd++, vaddr += PMD_SIZE) {
+               for (i = 0; i < pmds; i++) {
+                       if (pmd_present(pmd[i]))
+                               goto next;
+               }
+               vaddr += addr & ~PMD_MASK;
+               addr &= PMD_MASK;
+               for (i = 0; i < pmds; i++, addr += PMD_SIZE)
+                       set_pmd(pmd + i,__pmd(addr | _KERNPG_TABLE | _PAGE_PSE));
+               __flush_tlb();
+               return (void *)vaddr;
+       next:
+               ;
+       }
+       printk("early_ioremap(0x%lx, %lu) failed\n", addr, size);
+       return NULL;
+}
+
+/* To avoid virtual aliases later */
+__meminit void early_iounmap(void *addr, unsigned long size)
+{
+       unsigned long vaddr;
+       pmd_t *pmd;
+       int i, pmds;
+
+       vaddr = (unsigned long)addr;
+       pmds = ((vaddr & ~PMD_MASK) + size + ~PMD_MASK) / PMD_SIZE;
+       pmd = level2_kernel_pgt + pmd_index(vaddr);
+       for (i = 0; i < pmds; i++)
+               pmd_clear(pmd + i);
+       __flush_tlb();
+}
+
+static void __meminit
+phys_pmd_init(pmd_t *pmd_page, unsigned long address, unsigned long end)
+{
+       int i = pmd_index(address);
+
+       for (; i < PTRS_PER_PMD; i++, address += PMD_SIZE) {
+               unsigned long entry;
+               pmd_t *pmd = pmd_page + pmd_index(address);
+
+               if (address >= end) {
+                       if (!after_bootmem)
+                               for (; i < PTRS_PER_PMD; i++, pmd++)
+                                       set_pmd(pmd, __pmd(0));
+                       break;
+               }
+
+               if (pmd_val(*pmd))
+                       continue;
+
+               entry = _PAGE_NX|_PAGE_PSE|_KERNPG_TABLE|_PAGE_GLOBAL|address;
+               entry &= __supported_pte_mask;
+               set_pmd(pmd, __pmd(entry));
+       }
+}
+
+static void __meminit
+phys_pmd_update(pud_t *pud, unsigned long address, unsigned long end)
+{
+       pmd_t *pmd = pmd_offset(pud,0);
+       spin_lock(&init_mm.page_table_lock);
+       phys_pmd_init(pmd, address, end);
+       spin_unlock(&init_mm.page_table_lock);
+       __flush_tlb_all();
+}
+
+static void __meminit phys_pud_init(pud_t *pud_page, unsigned long addr, unsigned long end)
+{ 
+       int i = pud_index(addr);
+
+
+       for (; i < PTRS_PER_PUD; i++, addr = (addr & PUD_MASK) + PUD_SIZE ) {
+               unsigned long pmd_phys;
+               pud_t *pud = pud_page + pud_index(addr);
+               pmd_t *pmd;
+
+               if (addr >= end)
+                       break;
+
+               if (!after_bootmem && !e820_any_mapped(addr,addr+PUD_SIZE,0)) {
+                       set_pud(pud, __pud(0)); 
+                       continue;
+               } 
+
+               if (pud_val(*pud)) {
+                       phys_pmd_update(pud, addr, end);
+                       continue;
+               }
+
+               pmd = alloc_low_page(&pmd_phys);
+               spin_lock(&init_mm.page_table_lock);
+               set_pud(pud, __pud(pmd_phys | _KERNPG_TABLE));
+               phys_pmd_init(pmd, addr, end);
+               spin_unlock(&init_mm.page_table_lock);
+               unmap_low_page(pmd);
+       }
+       __flush_tlb();
+} 
+
+static void __init find_early_table_space(unsigned long end)
+{
+       unsigned long puds, pmds, tables, start;
+
+       puds = (end + PUD_SIZE - 1) >> PUD_SHIFT;
+       pmds = (end + PMD_SIZE - 1) >> PMD_SHIFT;
+       tables = round_up(puds * sizeof(pud_t), PAGE_SIZE) +
+                round_up(pmds * sizeof(pmd_t), PAGE_SIZE);
+
+       /* RED-PEN putting page tables only on node 0 could
+          cause a hotspot and fill up ZONE_DMA. The page tables
+          need roughly 0.5KB per GB. */
+       start = 0x8000;
+       table_start = find_e820_area(start, end, tables);
+       if (table_start == -1UL)
+               panic("Cannot find space for the kernel page tables");
+
+       table_start >>= PAGE_SHIFT;
+       table_end = table_start;
+
+       early_printk("kernel direct mapping tables up to %lx @ %lx-%lx\n",
+               end, table_start << PAGE_SHIFT,
+               (table_start << PAGE_SHIFT) + tables);
+}
+
+/* Setup the direct mapping of the physical memory at PAGE_OFFSET.
+   This runs before bootmem is initialized and gets pages directly from the 
+   physical memory. To access them they are temporarily mapped. */
+void __meminit init_memory_mapping(unsigned long start, unsigned long end)
+{ 
+       unsigned long next; 
+
+       Dprintk("init_memory_mapping\n");
+
+       /* 
+        * Find space for the kernel direct mapping tables.
+        * Later we should allocate these tables in the local node of the memory
+        * mapped.  Unfortunately this is done currently before the nodes are 
+        * discovered.
+        */
+       if (!after_bootmem)
+               find_early_table_space(end);
+
+       start = (unsigned long)__va(start);
+       end = (unsigned long)__va(end);
+
+       for (; start < end; start = next) {
+               unsigned long pud_phys; 
+               pgd_t *pgd = pgd_offset_k(start);
+               pud_t *pud;
+
+               if (after_bootmem)
+                       pud = pud_offset(pgd, start & PGDIR_MASK);
+               else
+                       pud = alloc_low_page(&pud_phys);
+
+               next = start + PGDIR_SIZE;
+               if (next > end) 
+                       next = end; 
+               phys_pud_init(pud, __pa(start), __pa(next));
+               if (!after_bootmem)
+                       set_pgd(pgd_offset_k(start), mk_kernel_pgd(pud_phys));
+               unmap_low_page(pud);
+       } 
+
+       if (!after_bootmem)
+               mmu_cr4_features = read_cr4();
+       __flush_tlb_all();
+}
+
+#ifndef CONFIG_NUMA
+void __init paging_init(void)
+{
+       unsigned long max_zone_pfns[MAX_NR_ZONES];
+       memset(max_zone_pfns, 0, sizeof(max_zone_pfns));
+       max_zone_pfns[ZONE_DMA] = MAX_DMA_PFN;
+       max_zone_pfns[ZONE_DMA32] = MAX_DMA32_PFN;
+       max_zone_pfns[ZONE_NORMAL] = end_pfn;
+
+       memory_present(0, 0, end_pfn);
+       sparse_init();
+       free_area_init_nodes(max_zone_pfns);
+}
+#endif
+
+/* Unmap a kernel mapping if it exists. This is useful to avoid prefetches
+   from the CPU leading to inconsistent cache lines. address and size
+   must be aligned to 2MB boundaries. 
+   Does nothing when the mapping doesn't exist. */
+void __init clear_kernel_mapping(unsigned long address, unsigned long size) 
+{
+       unsigned long end = address + size;
+
+       BUG_ON(address & ~LARGE_PAGE_MASK);
+       BUG_ON(size & ~LARGE_PAGE_MASK); 
+       
+       for (; address < end; address += LARGE_PAGE_SIZE) { 
+               pgd_t *pgd = pgd_offset_k(address);
+               pud_t *pud;
+               pmd_t *pmd;
+               if (pgd_none(*pgd))
+                       continue;
+               pud = pud_offset(pgd, address);
+               if (pud_none(*pud))
+                       continue; 
+               pmd = pmd_offset(pud, address);
+               if (!pmd || pmd_none(*pmd))
+                       continue; 
+               if (0 == (pmd_val(*pmd) & _PAGE_PSE)) { 
+                       /* Could handle this, but it should not happen currently. */
+                       printk(KERN_ERR 
+              "clear_kernel_mapping: mapping has been split. will leak memory\n"); 
+                       pmd_ERROR(*pmd); 
+               }
+               set_pmd(pmd, __pmd(0));                 
+       }
+       __flush_tlb_all();
+} 
+
+/*
+ * Memory hotplug specific functions
+ */
+void online_page(struct page *page)
+{
+       ClearPageReserved(page);
+       init_page_count(page);
+       __free_page(page);
+       totalram_pages++;
+       num_physpages++;
+}
+
+#ifdef CONFIG_MEMORY_HOTPLUG
+/*
+ * Memory is added always to NORMAL zone. This means you will never get
+ * additional DMA/DMA32 memory.
+ */
+int arch_add_memory(int nid, u64 start, u64 size)
+{
+       struct pglist_data *pgdat = NODE_DATA(nid);
+       struct zone *zone = pgdat->node_zones + ZONE_NORMAL;
+       unsigned long start_pfn = start >> PAGE_SHIFT;
+       unsigned long nr_pages = size >> PAGE_SHIFT;
+       int ret;
+
+       init_memory_mapping(start, (start + size -1));
+
+       ret = __add_pages(zone, start_pfn, nr_pages);
+       if (ret)
+               goto error;
+
+       return ret;
+error:
+       printk("%s: Problem encountered in __add_pages!\n", __func__);
+       return ret;
+}
+EXPORT_SYMBOL_GPL(arch_add_memory);
+
+int remove_memory(u64 start, u64 size)
+{
+       return -EINVAL;
+}
+EXPORT_SYMBOL_GPL(remove_memory);
+
+#if !defined(CONFIG_ACPI_NUMA) && defined(CONFIG_NUMA)
+int memory_add_physaddr_to_nid(u64 start)
+{
+       return 0;
+}
+EXPORT_SYMBOL_GPL(memory_add_physaddr_to_nid);
+#endif
+
+#endif /* CONFIG_MEMORY_HOTPLUG */
+
+#ifdef CONFIG_MEMORY_HOTPLUG_RESERVE
+/*
+ * Memory Hotadd without sparsemem. The mem_maps have been allocated in advance,
+ * just online the pages.
+ */
+int __add_pages(struct zone *z, unsigned long start_pfn, unsigned long nr_pages)
+{
+       int err = -EIO;
+       unsigned long pfn;
+       unsigned long total = 0, mem = 0;
+       for (pfn = start_pfn; pfn < start_pfn + nr_pages; pfn++) {
+               if (pfn_valid(pfn)) {
+                       online_page(pfn_to_page(pfn));
+                       err = 0;
+                       mem++;
+               }
+               total++;
+       }
+       if (!err) {
+               z->spanned_pages += total;
+               z->present_pages += mem;
+               z->zone_pgdat->node_spanned_pages += total;
+               z->zone_pgdat->node_present_pages += mem;
+       }
+       return err;
+}
+#endif
+
+static struct kcore_list kcore_mem, kcore_vmalloc, kcore_kernel, kcore_modules,
+                        kcore_vsyscall;
+
+void __init mem_init(void)
+{
+       long codesize, reservedpages, datasize, initsize;
+
+       pci_iommu_alloc();
+
+       /* clear the zero-page */
+       memset(empty_zero_page, 0, PAGE_SIZE);
+
+       reservedpages = 0;
+
+       /* this will put all low memory onto the freelists */
+#ifdef CONFIG_NUMA
+       totalram_pages = numa_free_all_bootmem();
+#else
+       totalram_pages = free_all_bootmem();
+#endif
+       reservedpages = end_pfn - totalram_pages -
+                                       absent_pages_in_range(0, end_pfn);
+
+       after_bootmem = 1;
+
+       codesize =  (unsigned long) &_etext - (unsigned long) &_text;
+       datasize =  (unsigned long) &_edata - (unsigned long) &_etext;
+       initsize =  (unsigned long) &__init_end - (unsigned long) &__init_begin;
+
+       /* Register memory areas for /proc/kcore */
+       kclist_add(&kcore_mem, __va(0), max_low_pfn << PAGE_SHIFT); 
+       kclist_add(&kcore_vmalloc, (void *)VMALLOC_START, 
+                  VMALLOC_END-VMALLOC_START);
+       kclist_add(&kcore_kernel, &_stext, _end - _stext);
+       kclist_add(&kcore_modules, (void *)MODULES_VADDR, MODULES_LEN);
+       kclist_add(&kcore_vsyscall, (void *)VSYSCALL_START, 
+                                VSYSCALL_END - VSYSCALL_START);
+
+       printk("Memory: %luk/%luk available (%ldk kernel code, %ldk reserved, %ldk data, %ldk init)\n",
+               (unsigned long) nr_free_pages() << (PAGE_SHIFT-10),
+               end_pfn << (PAGE_SHIFT-10),
+               codesize >> 10,
+               reservedpages << (PAGE_SHIFT-10),
+               datasize >> 10,
+               initsize >> 10);
+}
+
+void free_init_pages(char *what, unsigned long begin, unsigned long end)
+{
+       unsigned long addr;
+
+       if (begin >= end)
+               return;
+
+       printk(KERN_INFO "Freeing %s: %luk freed\n", what, (end - begin) >> 10);
+       for (addr = begin; addr < end; addr += PAGE_SIZE) {
+               ClearPageReserved(virt_to_page(addr));
+               init_page_count(virt_to_page(addr));
+               memset((void *)(addr & ~(PAGE_SIZE-1)),
+                       POISON_FREE_INITMEM, PAGE_SIZE);
+               if (addr >= __START_KERNEL_map)
+                       change_page_attr_addr(addr, 1, __pgprot(0));
+               free_page(addr);
+               totalram_pages++;
+       }
+       if (addr > __START_KERNEL_map)
+               global_flush_tlb();
+}
+
+void free_initmem(void)
+{
+       free_init_pages("unused kernel memory",
+                       (unsigned long)(&__init_begin),
+                       (unsigned long)(&__init_end));
+}
+
+#ifdef CONFIG_DEBUG_RODATA
+
+void mark_rodata_ro(void)
+{
+       unsigned long start = (unsigned long)_stext, end;
+
+#ifdef CONFIG_HOTPLUG_CPU
+       /* It must still be possible to apply SMP alternatives. */
+       if (num_possible_cpus() > 1)
+               start = (unsigned long)_etext;
+#endif
+
+#ifdef CONFIG_KPROBES
+       start = (unsigned long)__start_rodata;
+#endif
+       
+       end = (unsigned long)__end_rodata;
+       start = (start + PAGE_SIZE - 1) & PAGE_MASK;
+       end &= PAGE_MASK;
+       if (end <= start)
+               return;
+
+       change_page_attr_addr(start, (end - start) >> PAGE_SHIFT, PAGE_KERNEL_RO);
+
+       printk(KERN_INFO "Write protecting the kernel read-only data: %luk\n",
+              (end - start) >> 10);
+
+       /*
+        * change_page_attr_addr() requires a global_flush_tlb() call after it.
+        * We do this after the printk so that if something went wrong in the
+        * change, the printk gets out at least to give a better debug hint
+        * of who is the culprit.
+        */
+       global_flush_tlb();
+}
+#endif
+
+#ifdef CONFIG_BLK_DEV_INITRD
+void free_initrd_mem(unsigned long start, unsigned long end)
+{
+       free_init_pages("initrd memory", start, end);
+}
+#endif
+
+void __init reserve_bootmem_generic(unsigned long phys, unsigned len) 
+{ 
+#ifdef CONFIG_NUMA
+       int nid = phys_to_nid(phys);
+#endif
+       unsigned long pfn = phys >> PAGE_SHIFT;
+       if (pfn >= end_pfn) {
+               /* This can happen with kdump kernels when accessing firmware
+                  tables. */
+               if (pfn < end_pfn_map)
+                       return;
+               printk(KERN_ERR "reserve_bootmem: illegal reserve %lx %u\n",
+                               phys, len);
+               return;
+       }
+
+       /* Should check here against the e820 map to avoid double free */
+#ifdef CONFIG_NUMA
+       reserve_bootmem_node(NODE_DATA(nid), phys, len);
+#else                  
+       reserve_bootmem(phys, len);    
+#endif
+       if (phys+len <= MAX_DMA_PFN*PAGE_SIZE) {
+               dma_reserve += len / PAGE_SIZE;
+               set_dma_reserve(dma_reserve);
+       }
+}
+
+int kern_addr_valid(unsigned long addr) 
+{ 
+       unsigned long above = ((long)addr) >> __VIRTUAL_MASK_SHIFT;
+       pgd_t *pgd;
+       pud_t *pud;
+       pmd_t *pmd;
+       pte_t *pte;
+
+       if (above != 0 && above != -1UL)
+               return 0; 
+       
+       pgd = pgd_offset_k(addr);
+       if (pgd_none(*pgd))
+               return 0;
+
+       pud = pud_offset(pgd, addr);
+       if (pud_none(*pud))
+               return 0; 
+
+       pmd = pmd_offset(pud, addr);
+       if (pmd_none(*pmd))
+               return 0;
+       if (pmd_large(*pmd))
+               return pfn_valid(pmd_pfn(*pmd));
+
+       pte = pte_offset_kernel(pmd, addr);
+       if (pte_none(*pte))
+               return 0;
+       return pfn_valid(pte_pfn(*pte));
+}
+
+/* A pseudo VMA to allow ptrace access for the vsyscall page.  This only
+   covers the 64bit vsyscall page now. 32bit has a real VMA now and does
+   not need special handling anymore. */
+
+static struct vm_area_struct gate_vma = {
+       .vm_start = VSYSCALL_START,
+       .vm_end = VSYSCALL_START + (VSYSCALL_MAPPED_PAGES << PAGE_SHIFT),
+       .vm_page_prot = PAGE_READONLY_EXEC,
+       .vm_flags = VM_READ | VM_EXEC
+};
+
+struct vm_area_struct *get_gate_vma(struct task_struct *tsk)
+{
+#ifdef CONFIG_IA32_EMULATION
+       if (test_tsk_thread_flag(tsk, TIF_IA32))
+               return NULL;
+#endif
+       return &gate_vma;
+}
+
+int in_gate_area(struct task_struct *task, unsigned long addr)
+{
+       struct vm_area_struct *vma = get_gate_vma(task);
+       if (!vma)
+               return 0;
+       return (addr >= vma->vm_start) && (addr < vma->vm_end);
+}
+
+/* Use this when you have no reliable task/vma, typically from interrupt
+ * context.  It is less reliable than using the task's vma and may give
+ * false positives.
+ */
+int in_gate_area_no_task(unsigned long addr)
+{
+       return (addr >= VSYSCALL_START) && (addr < VSYSCALL_END);
+}
+
+void * __init alloc_bootmem_high_node(pg_data_t *pgdat, unsigned long size)
+{
+       return __alloc_bootmem_core(pgdat->bdata, size,
+                       SMP_CACHE_BYTES, (4UL*1024*1024*1024), 0);
+}
+
+const char *arch_vma_name(struct vm_area_struct *vma)
+{
+       if (vma->vm_mm && vma->vm_start == (long)vma->vm_mm->context.vdso)
+               return "[vdso]";
+       if (vma == &gate_vma)
+               return "[vsyscall]";
+       return NULL;
+}