LXR linux/arch/powerpc/mm/init

   1/*
   2 *  PowerPC version
   3 *    Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
   4 *
   5 *  Modifications by Paul Mackerras (PowerMac) (paulus@cs.anu.edu.au)
   6 *  and Cort Dougan (PReP) (cort@cs.nmt.edu)
   7 *    Copyright (C) 1996 Paul Mackerras
   8 *
   9 *  Derived from "arch/i386/mm/init.c"
  10 *    Copyright (C) 1991, 1992, 1993, 1994  Linus Torvalds
  11 *
  12 *  Dave Engebretsen <engebret@us.ibm.com>
  13 *      Rework for PPC64 port.
  14 *
  15 *  This program is free software; you can redistribute it and/or
  16 *  modify it under the terms of the GNU General Public License
  17 *  as published by the Free Software Foundation; either version
  18 *  2 of the License, or (at your option) any later version.
  19 *
  20 */
  21
  22#undef DEBUG
  23
  24#include <linux/signal.h>
  25#include <linux/sched.h>
  26#include <linux/kernel.h>
  27#include <linux/errno.h>
  28#include <linux/string.h>
  29#include <linux/types.h>
  30#include <linux/mman.h>
  31#include <linux/mm.h>
  32#include <linux/swap.h>
  33#include <linux/stddef.h>
  34#include <linux/vmalloc.h>
  35#include <linux/init.h>
  36#include <linux/delay.h>
  37#include <linux/bootmem.h>
  38#include <linux/highmem.h>
  39#include <linux/idr.h>
  40#include <linux/nodemask.h>
  41#include <linux/module.h>
  42#include <linux/poison.h>
  43#include <linux/lmb.h>
  44
  45#include <asm/pgalloc.h>
  46#include <asm/page.h>
  47#include <asm/prom.h>
  48#include <asm/rtas.h>
  49#include <asm/io.h>
  50#include <asm/mmu_context.h>
  51#include <asm/pgtable.h>
  52#include <asm/mmu.h>
  53#include <asm/uaccess.h>
  54#include <asm/smp.h>
  55#include <asm/machdep.h>
  56#include <asm/tlb.h>
  57#include <asm/eeh.h>
  58#include <asm/processor.h>
  59#include <asm/mmzone.h>
  60#include <asm/cputable.h>
  61#include <asm/sections.h>
  62#include <asm/system.h>
  63#include <asm/iommu.h>
  64#include <asm/abs_addr.h>
  65#include <asm/vdso.h>
  66
  67#include "mmu_decl.h"
  68
  69#ifdef CONFIG_PPC_STD_MMU_64
  70#if PGTABLE_RANGE > USER_VSID_RANGE
  71#warning Limited user VSID range means pagetable space is wasted
  72#endif
  73
  74#if (TASK_SIZE_USER64 < PGTABLE_RANGE) && (TASK_SIZE_USER64 < USER_VSID_RANGE)
  75#warning TASK_SIZE is smaller than it needs to be.
  76#endif
  77#endif /* CONFIG_PPC_STD_MMU_64 */
  78
  79phys_addr_t memstart_addr = ~0;
  80phys_addr_t kernstart_addr;
  81
  82void free_initmem(void)
  83{
  84        unsigned long addr;
  85
  86        addr = (unsigned long)__init_begin;
  87        for (; addr < (unsigned long)__init_end; addr += PAGE_SIZE) {
  88                memset((void *)addr, POISON_FREE_INITMEM, PAGE_SIZE);
  89                ClearPageReserved(virt_to_page(addr));
  90                init_page_count(virt_to_page(addr));
  91                free_page(addr);
  92                totalram_pages++;
  93        }
  94        printk ("Freeing unused kernel memory: %luk freed\n",
  95                ((unsigned long)__init_end - (unsigned long)__init_begin) >> 10);
  96}
  97
  98#ifdef CONFIG_BLK_DEV_INITRD
  99void free_initrd_mem(unsigned long start, unsigned long end)
 100{
 101        if (start < end)
 102                printk ("Freeing initrd memory: %ldk freed\n", (end - start) >> 10);
 103        for (; start < end; start += PAGE_SIZE) {
 104                ClearPageReserved(virt_to_page(start));
 105                init_page_count(virt_to_page(start));
 106                free_page(start);
 107                totalram_pages++;
 108        }
 109}
 110#endif
 111
 112static void pgd_ctor(void *addr)
 113{
 114        memset(addr, 0, PGD_TABLE_SIZE);
 115}
 116
 117static void pmd_ctor(void *addr)
 118{
 119        memset(addr, 0, PMD_TABLE_SIZE);
 120}
 121
 122static const unsigned int pgtable_cache_size[2] = {
 123        PGD_TABLE_SIZE, PMD_TABLE_SIZE
 124};
 125static const char *pgtable_cache_name[ARRAY_SIZE(pgtable_cache_size)] = {
 126#ifdef CONFIG_PPC_64K_PAGES
 127        "pgd_cache", "pmd_cache",
 128#else
 129        "pgd_cache", "pud_pmd_cache",
 130#endif /* CONFIG_PPC_64K_PAGES */
 131};
 132
 133#ifdef CONFIG_HUGETLB_PAGE
 134/* Hugepages need an extra cache per hugepagesize, initialized in
 135 * hugetlbpage.c.  We can't put into the tables above, because HPAGE_SHIFT
 136 * is not compile time constant. */
 137struct kmem_cache *pgtable_cache[ARRAY_SIZE(pgtable_cache_size)+MMU_PAGE_COUNT];
 138#else
 139struct kmem_cache *pgtable_cache[ARRAY_SIZE(pgtable_cache_size)];
 140#endif
 141
 142void pgtable_cache_init(void)
 143{
 144        pgtable_cache[0] = kmem_cache_create(pgtable_cache_name[0], PGD_TABLE_SIZE, PGD_TABLE_SIZE, SLAB_PANIC, pgd_ctor);
 145        pgtable_cache[1] = kmem_cache_create(pgtable_cache_name[1], PMD_TABLE_SIZE, PMD_TABLE_SIZE, SLAB_PANIC, pmd_ctor);
 146}
 147
 148#ifdef CONFIG_SPARSEMEM_VMEMMAP
 149/*
 150 * Given an address within the vmemmap, determine the pfn of the page that
 151 * represents the start of the section it is within.  Note that we have to
 152 * do this by hand as the proffered address may not be correctly aligned.
 153 * Subtraction of non-aligned pointers produces undefined results.
 154 */
 155static unsigned long __meminit vmemmap_section_start(unsigned long page)
 156{
 157        unsigned long offset = page - ((unsigned long)(vmemmap));
 158
 159        /* Return the pfn of the start of the section. */
 160        return (offset / sizeof(struct page)) & PAGE_SECTION_MASK;
 161}
 162
 163/*
 164 * Check if this vmemmap page is already initialised.  If any section
 165 * which overlaps this vmemmap page is initialised then this page is
 166 * initialised already.
 167 */
 168static int __meminit vmemmap_populated(unsigned long start, int page_size)
 169{
 170        unsigned long end = start + page_size;
 171
 172        for (; start < end; start += (PAGES_PER_SECTION * sizeof(struct page)))
 173                if (pfn_valid(vmemmap_section_start(start)))
 174                        return 1;
 175
 176        return 0;
 177}
 178
 179/* On hash-based CPUs, the vmemmap is bolted in the hash table.
 180 *
 181 * On Book3E CPUs, the vmemmap is currently mapped in the top half of
 182 * the vmalloc space using normal page tables, though the size of
 183 * pages encoded in the PTEs can be different
 184 */
 185
 186#ifdef CONFIG_PPC_BOOK3E
 187static void __meminit vmemmap_create_mapping(unsigned long start,
 188                                             unsigned long page_size,
 189                                             unsigned long phys)
 190{
 191        /* Create a PTE encoding without page size */
 192        unsigned long i, flags = _PAGE_PRESENT | _PAGE_ACCESSED |
 193                _PAGE_KERNEL_RW;
 194
 195        /* PTEs only contain page size encodings up to 32M */
 196        BUG_ON(mmu_psize_defs[mmu_vmemmap_psize].enc > 0xf);
 197
 198        /* Encode the size in the PTE */
 199        flags |= mmu_psize_defs[mmu_vmemmap_psize].enc << 8;
 200
 201        /* For each PTE for that area, map things. Note that we don't
 202         * increment phys because all PTEs are of the large size and
 203         * thus must have the low bits clear
 204         */
 205        for (i = 0; i < page_size; i += PAGE_SIZE)
 206                BUG_ON(map_kernel_page(start + i, phys, flags));
 207}
 208#else /* CONFIG_PPC_BOOK3E */
 209static void __meminit vmemmap_create_mapping(unsigned long start,
 210                                             unsigned long page_size,
 211                                             unsigned long phys)
 212{
 213        int  mapped = htab_bolt_mapping(start, start + page_size, phys,
 214                                        PAGE_KERNEL, mmu_vmemmap_psize,
 215                                        mmu_kernel_ssize);
 216        BUG_ON(mapped < 0);
 217}
 218#endif /* CONFIG_PPC_BOOK3E */
 219
 220int __meminit vmemmap_populate(struct page *start_page,
 221                               unsigned long nr_pages, int node)
 222{
 223        unsigned long start = (unsigned long)start_page;
 224        unsigned long end = (unsigned long)(start_page + nr_pages);
 225        unsigned long page_size = 1 << mmu_psize_defs[mmu_vmemmap_psize].shift;
 226
 227        /* Align to the page size of the linear mapping. */
 228        start = _ALIGN_DOWN(start, page_size);
 229
 230        pr_debug("vmemmap_populate page %p, %ld pages, node %d\n",
 231                 start_page, nr_pages, node);
 232        pr_debug(" -> map %lx..%lx\n", start, end);
 233
 234        for (; start < end; start += page_size) {
 235                void *p;
 236
 237                if (vmemmap_populated(start, page_size))
 238                        continue;
 239
 240                p = vmemmap_alloc_block(page_size, node);
 241                if (!p)
 242                        return -ENOMEM;
 243
 244                pr_debug("      * %016lx..%016lx allocated at %p\n",
 245                         start, start + page_size, p);
 246
 247                vmemmap_create_mapping(start, page_size, __pa(p));
 248        }
 249
 250        return 0;
 251}
 252#endif /* CONFIG_SPARSEMEM_VMEMMAP */
 253