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/pgalloc.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                 pgd_ERROR(*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                 pmd_ERROR(*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         if (!pte_none(*src)) {
  67                 pte = ptep_get_and_clear(src);
  68                 if (!dst) {
  69                         /* No dest?  We must put it back. */
  70                         dst = src;
  71                         error++;
  72                 }
  73                 set_pte(dst, pte);
  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
  97         /*
  98          * This is not the clever way to do this, but we're taking the
  99          * easy way out on the assumption that most remappings will be
 100          * only a few pages.. This also makes error recovery easier.
 101          */
 102         while (offset) {
 103                 offset -= PAGE_SIZE;
 104                 if (move_one_page(mm, old_addr + offset, new_addr + offset))
 105                         goto oops_we_failed;
 106         }
 107         flush_tlb_range(mm, old_addr, old_addr + len);
 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, 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         unsigned long new_addr)
 129 {
 130         struct vm_area_struct * new_vma;
 131
 132         new_vma = kmem_cache_alloc(vm_area_cachep, SLAB_KERNEL);
 133         if (new_vma) {
 134                 if (!move_page_tables(current->mm, new_addr, addr, old_len)) {
 135                         *new_vma = *vma;
 136                         new_vma->vm_start = new_addr;
 137                         new_vma->vm_end = new_addr+new_len;
 138                         new_vma->vm_pgoff += (addr - vma->vm_start) >> PAGE_SHIFT;
 139                         new_vma->vm_raend = 0;
 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                         lock_vma_mappings(vma);
 145                         spin_lock(&current->mm->page_table_lock);
 146                         __insert_vm_struct(current->mm, new_vma);
 147                         merge_segments(current->mm, new_vma->vm_start, new_vma->vm_end);
 148                         spin_unlock(&current->mm->page_table_lock);
 149                         unlock_vma_mappings(vma);
 150                         do_munmap(current->mm, addr, old_len);
 151                         current->mm->total_vm += new_len >> PAGE_SHIFT;
 152                         if (new_vma->vm_flags & VM_LOCKED) {
 153                                 current->mm->locked_vm += new_len >> PAGE_SHIFT;
 154                                 make_pages_present(new_vma->vm_start,
 155                                                    new_vma->vm_end);
 156                         }
 157                         return new_addr;
 158                 }
 159                 kmem_cache_free(vm_area_cachep, new_vma);
 160         }
 161         return -ENOMEM;
 162 }
 163
 164 /*
 165  * Expand (or shrink) an existing mapping, potentially moving it at the
 166  * same time (controlled by the MREMAP_MAYMOVE flag and available VM space)
 167  *
 168  * MREMAP_FIXED option added 5-Dec-1999 by Benjamin LaHaise
 169  * This option implies MREMAP_MAYMOVE.
 170  */
 171 unsigned long do_mremap(unsigned long addr,
 172         unsigned long old_len, unsigned long new_len,
 173         unsigned long flags, unsigned long new_addr)
 174 {
 175         struct vm_area_struct *vma;
 176         unsigned long ret = -EINVAL;
 177
 178         if (flags & ~(MREMAP_FIXED | MREMAP_MAYMOVE))
 179                 goto out;
 180
 181         if (addr & ~PAGE_MASK)
 182                 goto out;
 183
 184         old_len = PAGE_ALIGN(old_len);
 185         new_len = PAGE_ALIGN(new_len);
 186
 187         /* new_addr is only valid if MREMAP_FIXED is specified */
 188         if (flags & MREMAP_FIXED) {
 189                 if (new_addr & ~PAGE_MASK)
 190                         goto out;
 191                 if (!(flags & MREMAP_MAYMOVE))
 192                         goto out;
 193
 194                 if (new_len > TASK_SIZE || new_addr > TASK_SIZE - new_len)
 195                         goto out;
 196
 197                 /* Check if the location we're moving into overlaps the
 198                  * old location at all, and fail if it does.
 199                  */
 200                 if ((new_addr <= addr) && (new_addr+new_len) > addr)
 201                         goto out;
 202
 203                 if ((addr <= new_addr) && (addr+old_len) > new_addr)
 204                         goto out;
 205
 206                 do_munmap(current->mm, new_addr, new_len);
 207         }
 208
 209         /*
 210          * Always allow a shrinking remap: that just unmaps
 211          * the unnecessary pages..
 212          */
 213         ret = addr;
 214         if (old_len >= new_len) {
 215                 do_munmap(current->mm, addr+new_len, old_len - new_len);
 216                 if (!(flags & MREMAP_FIXED) || (new_addr == addr))
 217                         goto out;
 218         }
 219
 220         /*
 221          * Ok, we need to grow..  or relocate.
 222          */
 223         ret = -EFAULT;
 224         vma = find_vma(current->mm, addr);
 225         if (!vma || vma->vm_start > addr)
 226                 goto out;
 227         /* We can't remap across vm area boundaries */
 228         if (old_len > vma->vm_end - addr)
 229                 goto out;
 230         if (vma->vm_flags & VM_DONTEXPAND) {
 231                 if (new_len > old_len)
 232                         goto out;
 233         }
 234         if (vma->vm_flags & VM_LOCKED) {
 235                 unsigned long locked = current->mm->locked_vm << PAGE_SHIFT;
 236                 locked += new_len - old_len;
 237                 ret = -EAGAIN;
 238                 if (locked > current->rlim[RLIMIT_MEMLOCK].rlim_cur)
 239                         goto out;
 240         }
 241         ret = -ENOMEM;
 242         if ((current->mm->total_vm << PAGE_SHIFT) + (new_len - old_len)
 243             > current->rlim[RLIMIT_AS].rlim_cur)
 244                 goto out;
 245         /* Private writable mapping? Check memory availability.. */
 246         if ((vma->vm_flags & (VM_SHARED | VM_WRITE)) == VM_WRITE &&
 247             !(flags & MAP_NORESERVE)                             &&
 248             !vm_enough_memory((new_len - old_len) >> PAGE_SHIFT))
 249                 goto out;
 250
 251         /* old_len exactly to the end of the area..
 252          * And we're not relocating the area.
 253          */
 254         if (old_len == vma->vm_end - addr &&
 255             !((flags & MREMAP_FIXED) && (addr != new_addr)) &&
 256             (old_len != new_len || !(flags & MREMAP_MAYMOVE))) {
 257                 unsigned long max_addr = TASK_SIZE;
 258                 if (vma->vm_next)
 259                         max_addr = vma->vm_next->vm_start;
 260                 /* can we just expand the current mapping? */
 261                 if (max_addr - addr >= new_len) {
 262                         int pages = (new_len - old_len) >> PAGE_SHIFT;
 263                         spin_lock(&vma->vm_mm->page_table_lock);
 264                         vma->vm_end = addr + new_len;
 265                         spin_unlock(&vma->vm_mm->page_table_lock);
 266                         current->mm->total_vm += pages;
 267                         if (vma->vm_flags & VM_LOCKED) {
 268                                 current->mm->locked_vm += pages;
 269                                 make_pages_present(addr + old_len,
 270                                                    addr + new_len);
 271                         }
 272                         ret = addr;
 273                         goto out;
 274                 }
 275         }
 276
 277         /*
 278          * We weren't able to just expand or shrink the area,
 279          * we need to create a new one and move it..
 280          */
 281         ret = -ENOMEM;
 282         if (flags & MREMAP_MAYMOVE) {
 283                 if (!(flags & MREMAP_FIXED)) {
 284                         new_addr = get_unmapped_area(0, new_len);
 285                         if (!new_addr)
 286                                 goto out;
 287                 }
 288                 ret = move_vma(vma, addr, old_len, new_len, new_addr);
 289         }
 290 out:
 291         return ret;
 292 }
 293
 294 asmlinkage unsigned long sys_mremap(unsigned long addr,
 295         unsigned long old_len, unsigned long new_len,
 296         unsigned long flags, unsigned long new_addr)
 297 {
 298         unsigned long ret;
 299
 300         down(&current->mm->mmap_sem);
 301         ret = do_mremap(addr, old_len, new_len, flags, new_addr);
 302         up(&current->mm->mmap_sem);
 303         return ret;
 304 }