mm/mremap.c

   1 /*
   2  *      linux/mm/remap.c
   3  *
   4  *      (C) Copyright 1996 Linus Torvalds
   5  */
   6
   7 #include <linux/slab.h>
   8 #include <linux/smp_lock.h>
   9 #include <linux/shm.h>
  10 #include <linux/mman.h>
  11 #include <linux/swap.h>
  12
  13 #include <asm/uaccess.h>
  14 #include <asm/pgtable.h>
  15
  16 extern int vm_enough_memory(long pages);
  17
  18 static inline pte_t *get_one_pte(struct mm_struct *mm, unsigned long addr)
  19 {
  20         pgd_t * pgd;
  21         pmd_t * pmd;
  22         pte_t * pte = NULL;
  23
  24         pgd = pgd_offset(mm, addr);
  25         if (pgd_none(*pgd))
  26                 goto end;
  27         if (pgd_bad(*pgd)) {
  28                 printk("move_one_page: bad source pgd (%08lx)\n", pgd_val(*pgd));
  29                 pgd_clear(pgd);
  30                 goto end;
  31         }
  32
  33         pmd = pmd_offset(pgd, addr);
  34         if (pmd_none(*pmd))
  35                 goto end;
  36         if (pmd_bad(*pmd)) {
  37                 printk("move_one_page: bad source pmd (%08lx)\n", pmd_val(*pmd));
  38                 pmd_clear(pmd);
  39                 goto end;
  40         }
  41
  42         pte = pte_offset(pmd, addr);
  43         if (pte_none(*pte))
  44                 pte = NULL;
  45 end:
  46         return pte;
  47 }
  48
  49 static inline pte_t *alloc_one_pte(struct mm_struct *mm, unsigned long addr)
  50 {
  51         pmd_t * pmd;
  52         pte_t * pte = NULL;
  53
  54         pmd = pmd_alloc(pgd_offset(mm, addr), addr);
  55         if (pmd)
  56                 pte = pte_alloc(pmd, addr);
  57         return pte;
  58 }
  59
  60 static inline int copy_one_pte(struct mm_struct *mm, pte_t * src, pte_t * dst)
  61 {
  62         int error = 0;
  63         pte_t pte;
  64
  65         spin_lock(&mm->page_table_lock);
  66         pte = *src;
  67         if (!pte_none(pte)) {
  68                 error++;
  69                 if (dst) {
  70                         pte_clear(src);
  71                         set_pte(dst, pte);
  72                         error--;
  73                 }
  74         }
  75         spin_unlock(&mm->page_table_lock);
  76         return error;
  77 }
  78
  79 static int move_one_page(struct mm_struct *mm, unsigned long old_addr, unsigned long new_addr)
  80 {
  81         int error = 0;
  82         pte_t * src;
  83
  84         src = get_one_pte(mm, old_addr);
  85         if (src)
  86                 error = copy_one_pte(mm, src, alloc_one_pte(mm, new_addr));
  87         return error;
  88 }
  89
  90 static int move_page_tables(struct mm_struct * mm,
  91         unsigned long new_addr, unsigned long old_addr, unsigned long len)
  92 {
  93         unsigned long offset = len;
  94
  95         flush_cache_range(mm, old_addr, old_addr + len);
  96         flush_tlb_range(mm, old_addr, old_addr + len);
  97
  98         /*
  99          * This is not the clever way to do this, but we're taking the
 100          * easy way out on the assumption that most remappings will be
 101          * only a few pages.. This also makes error recovery easier.
 102          */
 103         while (offset) {
 104                 offset -= PAGE_SIZE;
 105                 if (move_one_page(mm, old_addr + offset, new_addr + offset))
 106                         goto oops_we_failed;
 107         }
 108         return 0;
 109
 110         /*
 111          * Ok, the move failed because we didn't have enough pages for
 112          * the new page table tree. This is unlikely, but we have to
 113          * take the possibility into account. In that case we just move
 114          * all the pages back (this will work, because we still have
 115          * the old page tables)
 116          */
 117 oops_we_failed:
 118         flush_cache_range(mm, new_addr, new_addr + len);
 119         while ((offset += PAGE_SIZE) < len)
 120                 move_one_page(mm, new_addr + offset, old_addr + offset);
 121         zap_page_range(mm, new_addr, new_addr + len);
 122         flush_tlb_range(mm, new_addr, new_addr + len);
 123         return -1;
 124 }
 125
 126 static inline unsigned long move_vma(struct vm_area_struct * vma,
 127         unsigned long addr, unsigned long old_len, unsigned long new_len)
 128 {
 129         struct vm_area_struct * new_vma;
 130
 131         new_vma = kmem_cache_alloc(vm_area_cachep, SLAB_KERNEL);
 132         if (new_vma) {
 133                 unsigned long new_addr = get_unmapped_area(addr, new_len);
 134
 135                 if (new_addr && !move_page_tables(current->mm, new_addr, addr, old_len)) {
 136                         *new_vma = *vma;
 137                         new_vma->vm_start = new_addr;
 138                         new_vma->vm_end = new_addr+new_len;
 139                         new_vma->vm_offset = vma->vm_offset + (addr - vma->vm_start);
 140                         if (new_vma->vm_file)
 141                                 get_file(new_vma->vm_file);
 142                         if (new_vma->vm_ops && new_vma->vm_ops->open)
 143                                 new_vma->vm_ops->open(new_vma);
 144                         insert_vm_struct(current->mm, new_vma);
 145                         merge_segments(current->mm, new_vma->vm_start, new_vma->vm_end);
 146                         do_munmap(addr, old_len);
 147                         current->mm->total_vm += new_len >> PAGE_SHIFT;
 148                         if (new_vma->vm_flags & VM_LOCKED) {
 149                                 current->mm->locked_vm += new_len >> PAGE_SHIFT;
 150                                 make_pages_present(new_vma->vm_start,
 151                                                    new_vma->vm_end);
 152                         }
 153                         return new_addr;
 154                 }
 155                 kmem_cache_free(vm_area_cachep, new_vma);
 156         }
 157         return -ENOMEM;
 158 }
 159
 160 /*
 161  * Expand (or shrink) an existing mapping, potentially moving it at the
 162  * same time (controlled by the MREMAP_MAYMOVE flag and available VM space)
 163  */
 164 asmlinkage unsigned long sys_mremap(unsigned long addr,
 165         unsigned long old_len, unsigned long new_len,
 166         unsigned long flags)
 167 {
 168         struct vm_area_struct *vma;
 169         unsigned long ret = -EINVAL;
 170
 171         down(&current->mm->mmap_sem);
 172         if (addr & ~PAGE_MASK)
 173                 goto out;
 174         old_len = PAGE_ALIGN(old_len);
 175         new_len = PAGE_ALIGN(new_len);
 176
 177         /*
 178          * Always allow a shrinking remap: that just unmaps
 179          * the unnecessary pages..
 180          */
 181         ret = addr;
 182         if (old_len >= new_len) {
 183                 do_munmap(addr+new_len, old_len - new_len);
 184                 goto out;
 185         }
 186
 187         /*
 188          * Ok, we need to grow..
 189          */
 190         ret = -EFAULT;
 191         vma = find_vma(current->mm, addr);
 192         if (!vma || vma->vm_start > addr)
 193                 goto out;
 194         /* We can't remap across vm area boundaries */
 195         if (old_len > vma->vm_end - addr)
 196                 goto out;
 197         if (vma->vm_flags & VM_LOCKED) {
 198                 unsigned long locked = current->mm->locked_vm << PAGE_SHIFT;
 199                 locked += new_len - old_len;
 200                 ret = -EAGAIN;
 201                 if (locked > current->rlim[RLIMIT_MEMLOCK].rlim_cur)
 202                         goto out;
 203         }
 204         ret = -ENOMEM;
 205         if ((current->mm->total_vm << PAGE_SHIFT) + (new_len - old_len)
 206             > current->rlim[RLIMIT_AS].rlim_cur)
 207                 goto out;
 208         /* Private writable mapping? Check memory availability.. */
 209         if ((vma->vm_flags & (VM_SHARED | VM_WRITE)) == VM_WRITE &&
 210             !(flags & MAP_NORESERVE)                             &&
 211             !vm_enough_memory((new_len - old_len) >> PAGE_SHIFT))
 212                 goto out;
 213
 214         /* old_len exactly to the end of the area.. */
 215         if (old_len == vma->vm_end - addr &&
 216             (old_len != new_len || !(flags & MREMAP_MAYMOVE))) {
 217                 unsigned long max_addr = TASK_SIZE;
 218                 if (vma->vm_next)
 219                         max_addr = vma->vm_next->vm_start;
 220                 /* can we just expand the current mapping? */
 221                 if (max_addr - addr >= new_len) {
 222                         int pages = (new_len - old_len) >> PAGE_SHIFT;
 223                         vma->vm_end = addr + new_len;
 224                         current->mm->total_vm += pages;
 225                         if (vma->vm_flags & VM_LOCKED) {
 226                                 current->mm->locked_vm += pages;
 227                                 make_pages_present(addr + old_len,
 228                                                    addr + new_len);
 229                         }
 230                         ret = addr;
 231                         goto out;
 232                 }
 233         }
 234
 235         /*
 236          * We weren't able to just expand or shrink the area,
 237          * we need to create a new one and move it..
 238          */
 239         if (flags & MREMAP_MAYMOVE)
 240                 ret = move_vma(vma, addr, old_len, new_len);
 241         else
 242                 ret = -ENOMEM;
 243 out:
 244         up(&current->mm->mmap_sem);
 245         return ret;
 246 }