mm/swapfile.c

   1 /*
   2  *  linux/mm/swapfile.c
   3  *
   4  *  Copyright (C) 1991, 1992, 1993, 1994  Linus Torvalds
   5  *  Swap reorganised 29.12.95, Stephen Tweedie
   6  */
   7
   8 #include <linux/malloc.h>
   9 #include <linux/smp_lock.h>
  10 #include <linux/kernel_stat.h>
  11 #include <linux/swap.h>
  12 #include <linux/swapctl.h>
  13 #include <linux/blkdev.h> /* for blk_size */
  14 #include <linux/vmalloc.h>
  15 #include <linux/pagemap.h>
  16 #include <linux/shm.h>
  17
  18 #include <asm/pgtable.h>
  19
  20 spinlock_t swaplock = SPIN_LOCK_UNLOCKED;
  21 unsigned int nr_swapfiles = 0;
  22
  23 struct swap_list_t swap_list = {-1, -1};
  24
  25 struct swap_info_struct swap_info[MAX_SWAPFILES];
  26
  27 #define SWAPFILE_CLUSTER 256
  28
  29 static inline int scan_swap_map(struct swap_info_struct *si, unsigned short count)
  30 {
  31         unsigned long offset;
  32         /*
  33          * We try to cluster swap pages by allocating them
  34          * sequentially in swap.  Once we've allocated
  35          * SWAPFILE_CLUSTER pages this way, however, we resort to
  36          * first-free allocation, starting a new cluster.  This
  37          * prevents us from scattering swap pages all over the entire
  38          * swap partition, so that we reduce overall disk seek times
  39          * between swap pages.  -- sct */
  40         if (si->cluster_nr) {
  41                 while (si->cluster_next <= si->highest_bit) {
  42                         offset = si->cluster_next++;
  43                         if (si->swap_map[offset])
  44                                 continue;
  45                         si->cluster_nr--;
  46                         goto got_page;
  47                 }
  48         }
  49         si->cluster_nr = SWAPFILE_CLUSTER;
  50
  51         /* try to find an empty (even not aligned) cluster. */
  52         offset = si->lowest_bit;
  53  check_next_cluster:
  54         if (offset+SWAPFILE_CLUSTER-1 <= si->highest_bit)
  55         {
  56                 int nr;
  57                 for (nr = offset; nr < offset+SWAPFILE_CLUSTER; nr++)
  58                         if (si->swap_map[nr])
  59                         {
  60                                 offset = nr+1;
  61                                 goto check_next_cluster;
  62                         }
  63                 /* We found a completly empty cluster, so start
  64                  * using it.
  65                  */
  66                 goto got_page;
  67         }
  68         /* No luck, so now go finegrined as usual. -Andrea */
  69         for (offset = si->lowest_bit; offset <= si->highest_bit ; offset++) {
  70                 if (si->swap_map[offset])
  71                         continue;
  72         got_page:
  73                 if (offset == si->lowest_bit)
  74                         si->lowest_bit++;
  75                 if (offset == si->highest_bit)
  76                         si->highest_bit--;
  77                 si->swap_map[offset] = count;
  78                 nr_swap_pages--;
  79                 si->cluster_next = offset+1;
  80                 return offset;
  81         }
  82         return 0;
  83 }
  84
  85 swp_entry_t __get_swap_page(unsigned short count)
  86 {
  87         struct swap_info_struct * p;
  88         unsigned long offset;
  89         swp_entry_t entry;
  90         int type, wrapped = 0;
  91
  92         entry.val = 0;  /* Out of memory */
  93         if (count >= SWAP_MAP_MAX)
  94                 goto bad_count;
  95         swap_list_lock();
  96         type = swap_list.next;
  97         if (type < 0)
  98                 goto out;
  99         if (nr_swap_pages == 0)
 100                 goto out;
 101
 102         while (1) {
 103                 p = &swap_info[type];
 104                 if ((p->flags & SWP_WRITEOK) == SWP_WRITEOK) {
 105                         swap_device_lock(p);
 106                         offset = scan_swap_map(p, count);
 107                         swap_device_unlock(p);
 108                         if (offset) {
 109                                 entry = SWP_ENTRY(type,offset);
 110                                 type = swap_info[type].next;
 111                                 if (type < 0 ||
 112                                         p->prio != swap_info[type].prio) {
 113                                                 swap_list.next = swap_list.head;
 114                                 } else {
 115                                         swap_list.next = type;
 116                                 }
 117                                 goto out;
 118                         }
 119                 }
 120                 type = p->next;
 121                 if (!wrapped) {
 122                         if (type < 0 || p->prio != swap_info[type].prio) {
 123                                 type = swap_list.head;
 124                                 wrapped = 1;
 125                         }
 126                 } else
 127                         if (type < 0)
 128                                 goto out;       /* out of swap space */
 129         }
 130 out:
 131         swap_list_unlock();
 132         return entry;
 133
 134 bad_count:
 135         printk(KERN_ERR "get_swap_page: bad count %hd from %p\n",
 136                count, __builtin_return_address(0));
 137         goto out;
 138 }
 139
 140
 141 /*
 142  * Caller has made sure that the swapdevice corresponding to entry
 143  * is still around or has not been recycled.
 144  */
 145 void __swap_free(swp_entry_t entry, unsigned short count)
 146 {
 147         struct swap_info_struct * p;
 148         unsigned long offset, type;
 149
 150         if (!entry.val)
 151                 goto out;
 152
 153         type = SWP_TYPE(entry);
 154         if (type >= nr_swapfiles)
 155                 goto bad_nofile;
 156         p = & swap_info[type];
 157         if (!(p->flags & SWP_USED))
 158                 goto bad_device;
 159         offset = SWP_OFFSET(entry);
 160         if (offset >= p->max)
 161                 goto bad_offset;
 162         if (!p->swap_map[offset])
 163                 goto bad_free;
 164         swap_list_lock();
 165         if (p->prio > swap_info[swap_list.next].prio)
 166                 swap_list.next = type;
 167         swap_device_lock(p);
 168         if (p->swap_map[offset] < SWAP_MAP_MAX) {
 169                 if (p->swap_map[offset] < count)
 170                         goto bad_count;
 171                 if (!(p->swap_map[offset] -= count)) {
 172                         if (offset < p->lowest_bit)
 173                                 p->lowest_bit = offset;
 174                         if (offset > p->highest_bit)
 175                                 p->highest_bit = offset;
 176                         nr_swap_pages++;
 177                 }
 178         }
 179         swap_device_unlock(p);
 180         swap_list_unlock();
 181 out:
 182         return;
 183
 184 bad_nofile:
 185         printk("swap_free: Trying to free nonexistent swap-page\n");
 186         goto out;
 187 bad_device:
 188         printk("swap_free: Trying to free swap from unused swap-device\n");
 189         goto out;
 190 bad_offset:
 191         printk("swap_free: offset exceeds max\n");
 192         goto out;
 193 bad_free:
 194         printk("VM: Bad swap entry %08lx\n", entry.val);
 195         goto out;
 196 bad_count:
 197         swap_device_unlock(p);
 198         swap_list_unlock();
 199         printk(KERN_ERR "VM: Bad count %hd current count %hd\n", count, p->swap_map[offset]);
 200         goto out;
 201 }
 202
 203 /* needs the big kernel lock */
 204 swp_entry_t acquire_swap_entry(struct page *page)
 205 {
 206         struct swap_info_struct * p;
 207         unsigned long offset, type;
 208         swp_entry_t entry;
 209
 210         if (!test_bit(PG_swap_entry, &page->flags))
 211                 goto new_swap_entry;
 212
 213         /* We have the old entry in the page offset still */
 214         if (!page->index)
 215                 goto new_swap_entry;
 216         entry.val = page->index;
 217         type = SWP_TYPE(entry);
 218         if (type >= nr_swapfiles)
 219                 goto new_swap_entry;
 220         p = type + swap_info;
 221         if ((p->flags & SWP_WRITEOK) != SWP_WRITEOK)
 222                 goto new_swap_entry;
 223         offset = SWP_OFFSET(entry);
 224         if (offset >= p->max)
 225                 goto new_swap_entry;
 226         /* Has it been re-used for something else? */
 227         swap_list_lock();
 228         swap_device_lock(p);
 229         if (p->swap_map[offset])
 230                 goto unlock_new_swap_entry;
 231
 232         /* We're cool, we can just use the old one */
 233         p->swap_map[offset] = 1;
 234         swap_device_unlock(p);
 235         nr_swap_pages--;
 236         swap_list_unlock();
 237         return entry;
 238
 239 unlock_new_swap_entry:
 240         swap_device_unlock(p);
 241         swap_list_unlock();
 242 new_swap_entry:
 243         return get_swap_page();
 244 }
 245
 246 /*
 247  * The swap entry has been read in advance, and we return 1 to indicate
 248  * that the page has been used or is no longer needed.
 249  *
 250  * Always set the resulting pte to be nowrite (the same as COW pages
 251  * after one process has exited).  We don't know just how many PTEs will
 252  * share this swap entry, so be cautious and let do_wp_page work out
 253  * what to do if a write is requested later.
 254  */
 255 static inline void unuse_pte(struct vm_area_struct * vma, unsigned long address,
 256         pte_t *dir, swp_entry_t entry, struct page* page)
 257 {
 258         pte_t pte = *dir;
 259
 260         if (pte_none(pte))
 261                 return;
 262         if (pte_present(pte)) {
 263                 /* If this entry is swap-cached, then page must already
 264                    hold the right address for any copies in physical
 265                    memory */
 266                 if (pte_page(pte) != page)
 267                         return;
 268                 /* We will be removing the swap cache in a moment, so... */
 269                 set_pte(dir, pte_mkdirty(pte));
 270                 return;
 271         }
 272         if (pte_val(pte) != entry.val)
 273                 return;
 274         set_pte(dir, pte_mkdirty(mk_pte(page, vma->vm_page_prot)));
 275         swap_free(entry);
 276         get_page(page);
 277         ++vma->vm_mm->rss;
 278 }
 279
 280 static inline void unuse_pmd(struct vm_area_struct * vma, pmd_t *dir,
 281         unsigned long address, unsigned long size, unsigned long offset,
 282         swp_entry_t entry, struct page* page)
 283 {
 284         pte_t * pte;
 285         unsigned long end;
 286
 287         if (pmd_none(*dir))
 288                 return;
 289         if (pmd_bad(*dir)) {
 290                 pmd_ERROR(*dir);
 291                 pmd_clear(dir);
 292                 return;
 293         }
 294         pte = pte_offset(dir, address);
 295         offset += address & PMD_MASK;
 296         address &= ~PMD_MASK;
 297         end = address + size;
 298         if (end > PMD_SIZE)
 299                 end = PMD_SIZE;
 300         do {
 301                 unuse_pte(vma, offset+address-vma->vm_start, pte, entry, page);
 302                 address += PAGE_SIZE;
 303                 pte++;
 304         } while (address && (address < end));
 305 }
 306
 307 static inline void unuse_pgd(struct vm_area_struct * vma, pgd_t *dir,
 308         unsigned long address, unsigned long size,
 309         swp_entry_t entry, struct page* page)
 310 {
 311         pmd_t * pmd;
 312         unsigned long offset, end;
 313
 314         if (pgd_none(*dir))
 315                 return;
 316         if (pgd_bad(*dir)) {
 317                 pgd_ERROR(*dir);
 318                 pgd_clear(dir);
 319                 return;
 320         }
 321         pmd = pmd_offset(dir, address);
 322         offset = address & PGDIR_MASK;
 323         address &= ~PGDIR_MASK;
 324         end = address + size;
 325         if (end > PGDIR_SIZE)
 326                 end = PGDIR_SIZE;
 327         if (address >= end)
 328                 BUG();
 329         do {
 330                 unuse_pmd(vma, pmd, address, end - address, offset, entry,
 331                           page);
 332                 address = (address + PMD_SIZE) & PMD_MASK;
 333                 pmd++;
 334         } while (address && (address < end));
 335 }
 336
 337 static void unuse_vma(struct vm_area_struct * vma, pgd_t *pgdir,
 338                         swp_entry_t entry, struct page* page)
 339 {
 340         unsigned long start = vma->vm_start, end = vma->vm_end;
 341
 342         if (start >= end)
 343                 BUG();
 344         do {
 345                 unuse_pgd(vma, pgdir, start, end - start, entry, page);
 346                 start = (start + PGDIR_SIZE) & PGDIR_MASK;
 347                 pgdir++;
 348         } while (start && (start < end));
 349 }
 350
 351 static void unuse_process(struct mm_struct * mm,
 352                         swp_entry_t entry, struct page* page)
 353 {
 354         struct vm_area_struct* vma;
 355
 356         /*
 357          * Go through process' page directory.
 358          */
 359         if (!mm)
 360                 return;
 361         for (vma = mm->mmap; vma; vma = vma->vm_next) {
 362                 pgd_t * pgd = pgd_offset(mm, vma->vm_start);
 363                 unuse_vma(vma, pgd, entry, page);
 364         }
 365         return;
 366 }
 367
 368 /*
 369  * We completely avoid races by reading each swap page in advance,
 370  * and then search for the process using it.  All the necessary
 371  * page table adjustments can then be made atomically.
 372  */
 373 static int try_to_unuse(unsigned int type)
 374 {
 375         struct swap_info_struct * si = &swap_info[type];
 376         struct task_struct *p;
 377         struct page *page;
 378         swp_entry_t entry;
 379         int i;
 380
 381         while (1) {
 382                 /*
 383                  * Find a swap page in use and read it in.
 384                  */
 385                 swap_device_lock(si);
 386                 for (i = 1; i < si->max ; i++) {
 387                         if (si->swap_map[i] > 0 && si->swap_map[i] != SWAP_MAP_BAD) {
 388                                 /*
 389                                  * Prevent swaphandle from being completely
 390                                  * unused by swap_free while we are trying
 391                                  * to read in the page - this prevents warning
 392                                  * messages from rw_swap_page_base.
 393                                  */
 394                                 if (si->swap_map[i] != SWAP_MAP_MAX)
 395                                         si->swap_map[i]++;
 396                                 swap_device_unlock(si);
 397                                 goto found_entry;
 398                         }
 399                 }
 400                 swap_device_unlock(si);
 401                 break;
 402
 403         found_entry:
 404                 entry = SWP_ENTRY(type, i);
 405
 406                 /* Get a page for the entry, using the existing swap
 407                    cache page if there is one.  Otherwise, get a clean
 408                    page and read the swap into it. */
 409                 page = read_swap_cache(entry);
 410                 if (!page) {
 411                         swap_free(entry);
 412                         return -ENOMEM;
 413                 }
 414                 read_lock(&tasklist_lock);
 415                 for_each_task(p)
 416                         unuse_process(p->mm, entry, page);
 417                 read_unlock(&tasklist_lock);
 418                 shm_unuse(entry, page);
 419                 /* Now get rid of the extra reference to the temporary
 420                    page we've been using. */
 421                 if (PageSwapCache(page))
 422                         delete_from_swap_cache(page);
 423                 __free_page(page);
 424                 /*
 425                  * Check for and clear any overflowed swap map counts.
 426                  */
 427                 swap_free(entry);
 428                 swap_list_lock();
 429                 swap_device_lock(si);
 430                 if (si->swap_map[i] > 0) {
 431                         if (si->swap_map[i] != SWAP_MAP_MAX)
 432                                 printk("VM: Undead swap entry %08lx\n",
 433                                                                 entry.val);
 434                         nr_swap_pages++;
 435                         si->swap_map[i] = 0;
 436                 }
 437                 swap_device_unlock(si);
 438                 swap_list_unlock();
 439         }
 440         return 0;
 441 }
 442
 443 asmlinkage long sys_swapoff(const char * specialfile)
 444 {
 445         struct swap_info_struct * p = NULL;
 446         struct dentry * dentry;
 447         int i, type, prev;
 448         int err;
 449
 450         if (!capable(CAP_SYS_ADMIN))
 451                 return -EPERM;
 452
 453         lock_kernel();
 454         dentry = namei(specialfile);
 455         err = PTR_ERR(dentry);
 456         if (IS_ERR(dentry))
 457                 goto out;
 458
 459         prev = -1;
 460         swap_list_lock();
 461         for (type = swap_list.head; type >= 0; type = swap_info[type].next) {
 462                 p = swap_info + type;
 463                 if ((p->flags & SWP_WRITEOK) == SWP_WRITEOK) {
 464                         if (p->swap_file) {
 465                                 if (p->swap_file == dentry)
 466                                   break;
 467                         } else {
 468                                 if (S_ISBLK(dentry->d_inode->i_mode)
 469                                     && (p->swap_device == dentry->d_inode->i_rdev))
 470                                   break;
 471                         }
 472                 }
 473                 prev = type;
 474         }
 475         err = -EINVAL;
 476         if (type < 0) {
 477                 swap_list_unlock();
 478                 goto out_dput;
 479         }
 480
 481         if (prev < 0) {
 482                 swap_list.head = p->next;
 483         } else {
 484                 swap_info[prev].next = p->next;
 485         }
 486         if (type == swap_list.next) {
 487                 /* just pick something that's safe... */
 488                 swap_list.next = swap_list.head;
 489         }
 490         nr_swap_pages -= p->pages;
 491         swap_list_unlock();
 492         p->flags = SWP_USED;
 493         err = try_to_unuse(type);
 494         if (err) {
 495                 /* re-insert swap space back into swap_list */
 496                 swap_list_lock();
 497                 for (prev = -1, i = swap_list.head; i >= 0; prev = i, i = swap_info[i].next)
 498                         if (p->prio >= swap_info[i].prio)
 499                                 break;
 500                 p->next = i;
 501                 if (prev < 0)
 502                         swap_list.head = swap_list.next = p - swap_info;
 503                 else
 504                         swap_info[prev].next = p - swap_info;
 505                 nr_swap_pages += p->pages;
 506                 swap_list_unlock();
 507                 p->flags = SWP_WRITEOK;
 508                 goto out_dput;
 509         }
 510         if (p->swap_device)
 511                 blkdev_put(dentry->d_inode->i_bdev, BDEV_SWAP);
 512         dput(dentry);
 513
 514         dentry = p->swap_file;
 515         p->swap_file = NULL;
 516         p->swap_device = 0;
 517         vfree(p->swap_map);
 518         p->swap_map = NULL;
 519         p->flags = 0;
 520         err = 0;
 521
 522 out_dput:
 523         dput(dentry);
 524 out:
 525         unlock_kernel();
 526         return err;
 527 }
 528
 529 int get_swaparea_info(char *buf)
 530 {
 531         char * page = (char *) __get_free_page(GFP_KERNEL);
 532         struct swap_info_struct *ptr = swap_info;
 533         int i, j, len = 0, usedswap;
 534
 535         if (!page)
 536                 return -ENOMEM;
 537
 538         len += sprintf(buf, "Filename\t\t\tType\t\tSize\tUsed\tPriority\n");
 539         for (i = 0 ; i < nr_swapfiles ; i++, ptr++) {
 540                 if (ptr->flags & SWP_USED) {
 541                         char * path = d_path(ptr->swap_file, page, PAGE_SIZE);
 542
 543                         len += sprintf(buf + len, "%-31s ", path);
 544
 545                         if (!ptr->swap_device)
 546                                 len += sprintf(buf + len, "file\t\t");
 547                         else
 548                                 len += sprintf(buf + len, "partition\t");
 549
 550                         usedswap = 0;
 551                         for (j = 0; j < ptr->max; ++j)
 552                                 switch (ptr->swap_map[j]) {
 553                                         case SWAP_MAP_BAD:
 554                                         case 0:
 555                                                 continue;
 556                                         default:
 557                                                 usedswap++;
 558                                 }
 559                         len += sprintf(buf + len, "%d\t%d\t%d\n", ptr->pages << (PAGE_SHIFT - 10),
 560                                 usedswap << (PAGE_SHIFT - 10), ptr->prio);
 561                 }
 562         }
 563         free_page((unsigned long) page);
 564         return len;
 565 }
 566
 567 int is_swap_partition(kdev_t dev) {
 568         struct swap_info_struct *ptr = swap_info;
 569         int i;
 570
 571         for (i = 0 ; i < nr_swapfiles ; i++, ptr++) {
 572                 if (ptr->flags & SWP_USED)
 573                         if (ptr->swap_device == dev)
 574                                 return 1;
 575         }
 576         return 0;
 577 }
 578
 579 /*
 580  * Written 01/25/92 by Simmule Turner, heavily changed by Linus.
 581  *
 582  * The swapon system call
 583  */
 584 asmlinkage long sys_swapon(const char * specialfile, int swap_flags)
 585 {
 586         struct swap_info_struct * p;
 587         struct dentry * swap_dentry;
 588         unsigned int type;
 589         int i, j, prev;
 590         int error;
 591         static int least_priority = 0;
 592         union swap_header *swap_header = 0;
 593         int swap_header_version;
 594         int nr_good_pages = 0;
 595         unsigned long maxpages;
 596         int swapfilesize;
 597         struct block_device *bdev = NULL;
 598
 599         if (!capable(CAP_SYS_ADMIN))
 600                 return -EPERM;
 601         lock_kernel();
 602         p = swap_info;
 603         for (type = 0 ; type < nr_swapfiles ; type++,p++)
 604                 if (!(p->flags & SWP_USED))
 605                         break;
 606         error = -EPERM;
 607         if (type >= MAX_SWAPFILES)
 608                 goto out;
 609         if (type >= nr_swapfiles)
 610                 nr_swapfiles = type+1;
 611         p->flags = SWP_USED;
 612         p->swap_file = NULL;
 613         p->swap_device = 0;
 614         p->swap_map = NULL;
 615         p->lowest_bit = 0;
 616         p->highest_bit = 0;
 617         p->cluster_nr = 0;
 618         p->sdev_lock = SPIN_LOCK_UNLOCKED;
 619         p->max = 1;
 620         p->next = -1;
 621         if (swap_flags & SWAP_FLAG_PREFER) {
 622                 p->prio =
 623                   (swap_flags & SWAP_FLAG_PRIO_MASK)>>SWAP_FLAG_PRIO_SHIFT;
 624         } else {
 625                 p->prio = --least_priority;
 626         }
 627         swap_dentry = namei(specialfile);
 628         error = PTR_ERR(swap_dentry);
 629         if (IS_ERR(swap_dentry))
 630                 goto bad_swap_2;
 631
 632         p->swap_file = swap_dentry;
 633         error = -EINVAL;
 634
 635         if (S_ISBLK(swap_dentry->d_inode->i_mode)) {
 636                 kdev_t dev = swap_dentry->d_inode->i_rdev;
 637                 struct block_device_operations *bdops;
 638
 639                 p->swap_device = dev;
 640                 set_blocksize(dev, PAGE_SIZE);
 641
 642                 bdev = swap_dentry->d_inode->i_bdev;
 643                 bdops = devfs_get_ops ( devfs_get_handle_from_inode
 644                                         (swap_dentry->d_inode) );
 645                 if (bdops) bdev->bd_op = bdops;
 646
 647                 error = blkdev_get(bdev, FMODE_READ|FMODE_WRITE, 0, BDEV_SWAP);
 648                 if (error)
 649                         goto bad_swap_2;
 650                 set_blocksize(dev, PAGE_SIZE);
 651                 error = -ENODEV;
 652                 if (!dev || (blk_size[MAJOR(dev)] &&
 653                      !blk_size[MAJOR(dev)][MINOR(dev)]))
 654                         goto bad_swap;
 655                 error = -EBUSY;
 656                 for (i = 0 ; i < nr_swapfiles ; i++) {
 657                         if (i == type)
 658                                 continue;
 659                         if (dev == swap_info[i].swap_device)
 660                                 goto bad_swap;
 661                 }
 662                 swapfilesize = 0;
 663                 if (blk_size[MAJOR(dev)])
 664                         swapfilesize = blk_size[MAJOR(dev)][MINOR(dev)]
 665                                 >> (PAGE_SHIFT - 10);
 666         } else if (S_ISREG(swap_dentry->d_inode->i_mode)) {
 667                 error = -EBUSY;
 668                 for (i = 0 ; i < nr_swapfiles ; i++) {
 669                         if (i == type || !swap_info[i].swap_file)
 670                                 continue;
 671                         if (swap_dentry->d_inode == swap_info[i].swap_file->d_inode)
 672                                 goto bad_swap;
 673                 }
 674                 swapfilesize = swap_dentry->d_inode->i_size >> PAGE_SHIFT;
 675         } else
 676                 goto bad_swap;
 677
 678         swap_header = (void *) __get_free_page(GFP_USER);
 679         if (!swap_header) {
 680                 printk("Unable to start swapping: out of memory :-)\n");
 681                 error = -ENOMEM;
 682                 goto bad_swap;
 683         }
 684
 685         lock_page(mem_map + MAP_NR(swap_header));
 686         rw_swap_page_nolock(READ, SWP_ENTRY(type,0), (char *) swap_header, 1);
 687
 688         if (!memcmp("SWAP-SPACE",swap_header->magic.magic,10))
 689                 swap_header_version = 1;
 690         else if (!memcmp("SWAPSPACE2",swap_header->magic.magic,10))
 691                 swap_header_version = 2;
 692         else {
 693                 printk("Unable to find swap-space signature\n");
 694                 error = -EINVAL;
 695                 goto bad_swap;
 696         }
 697
 698         switch (swap_header_version) {
 699         case 1:
 700                 memset(((char *) swap_header)+PAGE_SIZE-10,0,10);
 701                 j = 0;
 702                 p->lowest_bit = 0;
 703                 p->highest_bit = 0;
 704                 for (i = 1 ; i < 8*PAGE_SIZE ; i++) {
 705                         if (test_bit(i,(char *) swap_header)) {
 706                                 if (!p->lowest_bit)
 707                                         p->lowest_bit = i;
 708                                 p->highest_bit = i;
 709                                 p->max = i+1;
 710                                 j++;
 711                         }
 712                 }
 713                 nr_good_pages = j;
 714                 p->swap_map = vmalloc(p->max * sizeof(short));
 715                 if (!p->swap_map) {
 716                         error = -ENOMEM;
 717                         goto bad_swap;
 718                 }
 719                 for (i = 1 ; i < p->max ; i++) {
 720                         if (test_bit(i,(char *) swap_header))
 721                                 p->swap_map[i] = 0;
 722                         else
 723                                 p->swap_map[i] = SWAP_MAP_BAD;
 724                 }
 725                 break;
 726
 727         case 2:
 728                 /* Check the swap header's sub-version and the size of
 729                    the swap file and bad block lists */
 730                 if (swap_header->info.version != 1) {
 731                         printk(KERN_WARNING
 732                                "Unable to handle swap header version %d\n",
 733                                swap_header->info.version);
 734                         error = -EINVAL;
 735                         goto bad_swap;
 736                 }
 737
 738                 p->lowest_bit  = 1;
 739                 p->highest_bit = swap_header->info.last_page - 1;
 740                 p->max         = swap_header->info.last_page;
 741
 742                 maxpages = SWP_OFFSET(SWP_ENTRY(0,~0UL));
 743                 if (p->max >= maxpages)
 744                         p->max = maxpages-1;
 745
 746                 error = -EINVAL;
 747                 if (swap_header->info.nr_badpages > MAX_SWAP_BADPAGES)
 748                         goto bad_swap;
 749
 750                 /* OK, set up the swap map and apply the bad block list */
 751                 if (!(p->swap_map = vmalloc (p->max * sizeof(short)))) {
 752                         error = -ENOMEM;
 753                         goto bad_swap;
 754                 }
 755
 756                 error = 0;
 757                 memset(p->swap_map, 0, p->max * sizeof(short));
 758                 for (i=0; i<swap_header->info.nr_badpages; i++) {
 759                         int page = swap_header->info.badpages[i];
 760                         if (page <= 0 || page >= swap_header->info.last_page)
 761                                 error = -EINVAL;
 762                         else
 763                                 p->swap_map[page] = SWAP_MAP_BAD;
 764                 }
 765                 nr_good_pages = swap_header->info.last_page -
 766                                 swap_header->info.nr_badpages -
 767                                 1 /* header page */;
 768                 if (error)
 769                         goto bad_swap;
 770         }
 771
 772         if (swapfilesize && p->max > swapfilesize) {
 773                 printk(KERN_WARNING
 774                        "Swap area shorter than signature indicates\n");
 775                 error = -EINVAL;
 776                 goto bad_swap;
 777         }
 778         if (!nr_good_pages) {
 779                 printk(KERN_WARNING "Empty swap-file\n");
 780                 error = -EINVAL;
 781                 goto bad_swap;
 782         }
 783         p->swap_map[0] = SWAP_MAP_BAD;
 784         p->flags = SWP_WRITEOK;
 785         p->pages = nr_good_pages;
 786         swap_list_lock();
 787         nr_swap_pages += nr_good_pages;
 788         printk(KERN_INFO "Adding Swap: %dk swap-space (priority %d)\n",
 789                nr_good_pages<<(PAGE_SHIFT-10), p->prio);
 790
 791         /* insert swap space into swap_list: */
 792         prev = -1;
 793         for (i = swap_list.head; i >= 0; i = swap_info[i].next) {
 794                 if (p->prio >= swap_info[i].prio) {
 795                         break;
 796                 }
 797                 prev = i;
 798         }
 799         p->next = i;
 800         if (prev < 0) {
 801                 swap_list.head = swap_list.next = p - swap_info;
 802         } else {
 803                 swap_info[prev].next = p - swap_info;
 804         }
 805         swap_list_unlock();
 806         error = 0;
 807         goto out;
 808 bad_swap:
 809         if (bdev)
 810                 blkdev_put(bdev, BDEV_SWAP);
 811 bad_swap_2:
 812         if (p->swap_map)
 813                 vfree(p->swap_map);
 814         dput(p->swap_file);
 815         p->swap_device = 0;
 816         p->swap_file = NULL;
 817         p->swap_map = NULL;
 818         p->flags = 0;
 819         if (!(swap_flags & SWAP_FLAG_PREFER))
 820                 ++least_priority;
 821 out:
 822         if (swap_header)
 823                 free_page((long) swap_header);
 824         unlock_kernel();
 825         return error;
 826 }
 827
 828 void si_swapinfo(struct sysinfo *val)
 829 {
 830         unsigned int i;
 831         unsigned long freeswap = 0;
 832         unsigned long totalswap = 0;
 833
 834         for (i = 0; i < nr_swapfiles; i++) {
 835                 unsigned int j;
 836                 if ((swap_info[i].flags & SWP_WRITEOK) != SWP_WRITEOK)
 837                         continue;
 838                 for (j = 0; j < swap_info[i].max; ++j) {
 839                         switch (swap_info[i].swap_map[j]) {
 840                                 case SWAP_MAP_BAD:
 841                                         continue;
 842                                 case 0:
 843                                         freeswap++;
 844                                 default:
 845                                         totalswap++;
 846                         }
 847                 }
 848         }
 849         val->freeswap = freeswap;
 850         val->totalswap = totalswap;
 851         return;
 852 }
 853
 854 /*
 855  * Verify that a swap entry is valid and increment its swap map count.
 856  * Kernel_lock is held, which guarantees existance of swap device.
 857  *
 858  * Note: if swap_map[] reaches SWAP_MAP_MAX the entries are treated as
 859  * "permanent", but will be reclaimed by the next swapoff.
 860  */
 861 int swap_duplicate(swp_entry_t entry)
 862 {
 863         struct swap_info_struct * p;
 864         unsigned long offset, type;
 865         int result = 0;
 866
 867         /* Swap entry 0 is illegal */
 868         if (!entry.val)
 869                 goto out;
 870         type = SWP_TYPE(entry);
 871         if (type >= nr_swapfiles)
 872                 goto bad_file;
 873         p = type + swap_info;
 874         offset = SWP_OFFSET(entry);
 875         if (offset >= p->max)
 876                 goto bad_offset;
 877         if (!p->swap_map[offset])
 878                 goto bad_unused;
 879         /*
 880          * Entry is valid, so increment the map count.
 881          */
 882         swap_device_lock(p);
 883         if (p->swap_map[offset] < SWAP_MAP_MAX)
 884                 p->swap_map[offset]++;
 885         else {
 886                 static int overflow = 0;
 887                 if (overflow++ < 5)
 888                         printk("VM: swap entry overflow\n");
 889                 p->swap_map[offset] = SWAP_MAP_MAX;
 890         }
 891         swap_device_unlock(p);
 892         result = 1;
 893 out:
 894         return result;
 895
 896 bad_file:
 897         printk("Bad swap file entry %08lx\n", entry.val);
 898         goto out;
 899 bad_offset:
 900         printk("Bad swap offset entry %08lx\n", entry.val);
 901         goto out;
 902 bad_unused:
 903         printk("Unused swap offset entry in swap_dup %08lx\n", entry.val);
 904         goto out;
 905 }
 906
 907 /*
 908  * Page lock needs to be held in all cases to prevent races with
 909  * swap file deletion.
 910  */
 911 int swap_count(struct page *page)
 912 {
 913         struct swap_info_struct * p;
 914         unsigned long offset, type;
 915         swp_entry_t entry;
 916         int retval = 0;
 917
 918         entry.val = page->index;
 919         if (!entry.val)
 920                 goto bad_entry;
 921         type = SWP_TYPE(entry);
 922         if (type >= nr_swapfiles)
 923                 goto bad_file;
 924         p = type + swap_info;
 925         offset = SWP_OFFSET(entry);
 926         if (offset >= p->max)
 927                 goto bad_offset;
 928         if (!p->swap_map[offset])
 929                 goto bad_unused;
 930         retval = p->swap_map[offset];
 931 out:
 932         return retval;
 933
 934 bad_entry:
 935         printk(KERN_ERR "swap_count: null entry!\n");
 936         goto out;
 937 bad_file:
 938         printk("Bad swap file entry %08lx\n", entry.val);
 939         goto out;
 940 bad_offset:
 941         printk("Bad swap offset entry %08lx\n", entry.val);
 942         goto out;
 943 bad_unused:
 944         printk("Unused swap offset entry in swap_count %08lx\n", entry.val);
 945         goto out;
 946 }
 947
 948 /*
 949  * Kernel_lock protects against swap device deletion.
 950  */
 951 void get_swaphandle_info(swp_entry_t entry, unsigned long *offset,
 952                         kdev_t *dev, struct inode **swapf)
 953 {
 954         unsigned long type;
 955         struct swap_info_struct *p;
 956
 957         type = SWP_TYPE(entry);
 958         if (type >= nr_swapfiles) {
 959                 printk("Internal error: bad swap-device\n");
 960                 return;
 961         }
 962
 963         p = &swap_info[type];
 964         *offset = SWP_OFFSET(entry);
 965         if (*offset >= p->max) {
 966                 printk("rw_swap_page: weirdness\n");
 967                 return;
 968         }
 969         if (p->swap_map && !p->swap_map[*offset]) {
 970                 printk("VM: Bad swap entry %08lx\n", entry.val);
 971                 return;
 972         }
 973         if (!(p->flags & SWP_USED)) {
 974                 printk(KERN_ERR "rw_swap_page: "
 975                         "Trying to swap to unused swap-device\n");
 976                 return;
 977         }
 978
 979         if (p->swap_device) {
 980                 *dev = p->swap_device;
 981         } else if (p->swap_file) {
 982                 *swapf = p->swap_file->d_inode;
 983         } else {
 984                 printk(KERN_ERR "rw_swap_page: no swap file or device\n");
 985         }
 986         return;
 987 }
 988
 989 /*
 990  * Kernel_lock protects against swap device deletion. Grab an extra
 991  * reference on the swaphandle so that it dos not become unused.
 992  */
 993 int valid_swaphandles(swp_entry_t entry, unsigned long *offset)
 994 {
 995         int ret = 0, i = 1 << page_cluster;
 996         unsigned long toff;
 997         struct swap_info_struct *swapdev = SWP_TYPE(entry) + swap_info;
 998
 999         *offset = SWP_OFFSET(entry);
1000         toff = *offset = (*offset >> page_cluster) << page_cluster;
1001
1002         swap_device_lock(swapdev);
1003         do {
1004                 /* Don't read-ahead past the end of the swap area */
1005                 if (toff >= swapdev->max)
1006                         break;
1007                 /* Don't read in bad or busy pages */
1008                 if (!swapdev->swap_map[toff])
1009                         break;
1010                 if (swapdev->swap_map[toff] == SWAP_MAP_BAD)
1011                         break;
1012                 swapdev->swap_map[toff]++;
1013                 toff++;
1014                 ret++;
1015         } while (--i);
1016         swap_device_unlock(swapdev);
1017         return ret;
1018 }