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;
  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 /*
 204  * The swap entry has been read in advance, and we return 1 to indicate
 205  * that the page has been used or is no longer needed.
 206  *
 207  * Always set the resulting pte to be nowrite (the same as COW pages
 208  * after one process has exited).  We don't know just how many PTEs will
 209  * share this swap entry, so be cautious and let do_wp_page work out
 210  * what to do if a write is requested later.
 211  */
 212 static inline void unuse_pte(struct vm_area_struct * vma, unsigned long address,
 213         pte_t *dir, swp_entry_t entry, struct page* page)
 214 {
 215         pte_t pte = *dir;
 216
 217         if (pte_none(pte))
 218                 return;
 219         if (pte_present(pte)) {
 220                 /* If this entry is swap-cached, then page must already
 221                    hold the right address for any copies in physical
 222                    memory */
 223                 if (pte_page(pte) != page)
 224                         return;
 225                 /* We will be removing the swap cache in a moment, so... */
 226                 set_pte(dir, pte_mkdirty(pte));
 227                 return;
 228         }
 229         if (pte_val(pte) != entry.val)
 230                 return;
 231         set_pte(dir, pte_mkdirty(mk_pte(page, vma->vm_page_prot)));
 232         swap_free(entry);
 233         get_page(page);
 234         ++vma->vm_mm->rss;
 235 }
 236
 237 static inline void unuse_pmd(struct vm_area_struct * vma, pmd_t *dir,
 238         unsigned long address, unsigned long size, unsigned long offset,
 239         swp_entry_t entry, struct page* page)
 240 {
 241         pte_t * pte;
 242         unsigned long end;
 243
 244         if (pmd_none(*dir))
 245                 return;
 246         if (pmd_bad(*dir)) {
 247                 pmd_ERROR(*dir);
 248                 pmd_clear(dir);
 249                 return;
 250         }
 251         pte = pte_offset(dir, address);
 252         offset += address & PMD_MASK;
 253         address &= ~PMD_MASK;
 254         end = address + size;
 255         if (end > PMD_SIZE)
 256                 end = PMD_SIZE;
 257         do {
 258                 unuse_pte(vma, offset+address-vma->vm_start, pte, entry, page);
 259                 address += PAGE_SIZE;
 260                 pte++;
 261         } while (address && (address < end));
 262 }
 263
 264 static inline void unuse_pgd(struct vm_area_struct * vma, pgd_t *dir,
 265         unsigned long address, unsigned long size,
 266         swp_entry_t entry, struct page* page)
 267 {
 268         pmd_t * pmd;
 269         unsigned long offset, end;
 270
 271         if (pgd_none(*dir))
 272                 return;
 273         if (pgd_bad(*dir)) {
 274                 pgd_ERROR(*dir);
 275                 pgd_clear(dir);
 276                 return;
 277         }
 278         pmd = pmd_offset(dir, address);
 279         offset = address & PGDIR_MASK;
 280         address &= ~PGDIR_MASK;
 281         end = address + size;
 282         if (end > PGDIR_SIZE)
 283                 end = PGDIR_SIZE;
 284         if (address >= end)
 285                 BUG();
 286         do {
 287                 unuse_pmd(vma, pmd, address, end - address, offset, entry,
 288                           page);
 289                 address = (address + PMD_SIZE) & PMD_MASK;
 290                 pmd++;
 291         } while (address && (address < end));
 292 }
 293
 294 static void unuse_vma(struct vm_area_struct * vma, pgd_t *pgdir,
 295                         swp_entry_t entry, struct page* page)
 296 {
 297         unsigned long start = vma->vm_start, end = vma->vm_end;
 298
 299         if (start >= end)
 300                 BUG();
 301         do {
 302                 unuse_pgd(vma, pgdir, start, end - start, entry, page);
 303                 start = (start + PGDIR_SIZE) & PGDIR_MASK;
 304                 pgdir++;
 305         } while (start && (start < end));
 306 }
 307
 308 static void unuse_process(struct mm_struct * mm,
 309                         swp_entry_t entry, struct page* page)
 310 {
 311         struct vm_area_struct* vma;
 312
 313         /*
 314          * Go through process' page directory.
 315          */
 316         if (!mm)
 317                 return;
 318         for (vma = mm->mmap; vma; vma = vma->vm_next) {
 319                 pgd_t * pgd = pgd_offset(mm, vma->vm_start);
 320                 unuse_vma(vma, pgd, entry, page);
 321         }
 322         return;
 323 }
 324
 325 /*
 326  * We completely avoid races by reading each swap page in advance,
 327  * and then search for the process using it.  All the necessary
 328  * page table adjustments can then be made atomically.
 329  */
 330 static int try_to_unuse(unsigned int type)
 331 {
 332         struct swap_info_struct * si = &swap_info[type];
 333         struct task_struct *p;
 334         struct page *page;
 335         swp_entry_t entry;
 336         int i;
 337
 338         while (1) {
 339                 /*
 340                  * Find a swap page in use and read it in.
 341                  */
 342                 swap_device_lock(si);
 343                 for (i = 1; i < si->max ; i++) {
 344                         if (si->swap_map[i] > 0 && si->swap_map[i] != SWAP_MAP_BAD) {
 345                                 /*
 346                                  * Prevent swaphandle from being completely
 347                                  * unused by swap_free while we are trying
 348                                  * to read in the page - this prevents warning
 349                                  * messages from rw_swap_page_base.
 350                                  */
 351                                 if (si->swap_map[i] != SWAP_MAP_MAX)
 352                                         si->swap_map[i]++;
 353                                 swap_device_unlock(si);
 354                                 goto found_entry;
 355                         }
 356                 }
 357                 swap_device_unlock(si);
 358                 break;
 359
 360         found_entry:
 361                 entry = SWP_ENTRY(type, i);
 362
 363                 /* Get a page for the entry, using the existing swap
 364                    cache page if there is one.  Otherwise, get a clean
 365                    page and read the swap into it. */
 366                 page = read_swap_cache(entry);
 367                 if (!page) {
 368                         swap_free(entry);
 369                         return -ENOMEM;
 370                 }
 371                 read_lock(&tasklist_lock);
 372                 for_each_task(p)
 373                         unuse_process(p->mm, entry, page);
 374                 read_unlock(&tasklist_lock);
 375                 shm_unuse(entry, page);
 376                 /* Now get rid of the extra reference to the temporary
 377                    page we've been using. */
 378                 if (PageSwapCache(page))
 379                         delete_from_swap_cache(page);
 380                 page_cache_release(page);
 381                 /*
 382                  * Check for and clear any overflowed swap map counts.
 383                  */
 384                 swap_free(entry);
 385                 swap_list_lock();
 386                 swap_device_lock(si);
 387                 if (si->swap_map[i] > 0) {
 388                         if (si->swap_map[i] != SWAP_MAP_MAX)
 389                                 printk("VM: Undead swap entry %08lx\n",
 390                                                                 entry.val);
 391                         nr_swap_pages++;
 392                         si->swap_map[i] = 0;
 393                 }
 394                 swap_device_unlock(si);
 395                 swap_list_unlock();
 396         }
 397         return 0;
 398 }
 399
 400 asmlinkage long sys_swapoff(const char * specialfile)
 401 {
 402         struct swap_info_struct * p = NULL;
 403         struct nameidata nd;
 404         int i, type, prev;
 405         int err;
 406
 407         if (!capable(CAP_SYS_ADMIN))
 408                 return -EPERM;
 409
 410         err = user_path_walk(specialfile, &nd);
 411         if (err)
 412                 goto out;
 413
 414         lock_kernel();
 415         prev = -1;
 416         swap_list_lock();
 417         for (type = swap_list.head; type >= 0; type = swap_info[type].next) {
 418                 p = swap_info + type;
 419                 if ((p->flags & SWP_WRITEOK) == SWP_WRITEOK) {
 420                         if (p->swap_file) {
 421                                 if (p->swap_file == nd.dentry)
 422                                   break;
 423                         } else {
 424                                 if (S_ISBLK(nd.dentry->d_inode->i_mode)
 425                                     && (p->swap_device == nd.dentry->d_inode->i_rdev))
 426                                   break;
 427                         }
 428                 }
 429                 prev = type;
 430         }
 431         err = -EINVAL;
 432         if (type < 0) {
 433                 swap_list_unlock();
 434                 goto out_dput;
 435         }
 436
 437         if (prev < 0) {
 438                 swap_list.head = p->next;
 439         } else {
 440                 swap_info[prev].next = p->next;
 441         }
 442         if (type == swap_list.next) {
 443                 /* just pick something that's safe... */
 444                 swap_list.next = swap_list.head;
 445         }
 446         nr_swap_pages -= p->pages;
 447         swap_list_unlock();
 448         p->flags = SWP_USED;
 449         err = try_to_unuse(type);
 450         if (err) {
 451                 /* re-insert swap space back into swap_list */
 452                 swap_list_lock();
 453                 for (prev = -1, i = swap_list.head; i >= 0; prev = i, i = swap_info[i].next)
 454                         if (p->prio >= swap_info[i].prio)
 455                                 break;
 456                 p->next = i;
 457                 if (prev < 0)
 458                         swap_list.head = swap_list.next = p - swap_info;
 459                 else
 460                         swap_info[prev].next = p - swap_info;
 461                 nr_swap_pages += p->pages;
 462                 swap_list_unlock();
 463                 p->flags = SWP_WRITEOK;
 464                 goto out_dput;
 465         }
 466         if (p->swap_device)
 467                 blkdev_put(nd.dentry->d_inode->i_bdev, BDEV_SWAP);
 468         path_release(&nd);
 469
 470         nd.dentry = p->swap_file;
 471         p->swap_file = NULL;
 472         nd.mnt = p->swap_vfsmnt;
 473         p->swap_vfsmnt = NULL;
 474         p->swap_device = 0;
 475         vfree(p->swap_map);
 476         p->swap_map = NULL;
 477         p->flags = 0;
 478         err = 0;
 479
 480 out_dput:
 481         unlock_kernel();
 482         path_release(&nd);
 483 out:
 484         return err;
 485 }
 486
 487 int get_swaparea_info(char *buf)
 488 {
 489         char * page = (char *) __get_free_page(GFP_KERNEL);
 490         struct swap_info_struct *ptr = swap_info;
 491         int i, j, len = 0, usedswap;
 492
 493         if (!page)
 494                 return -ENOMEM;
 495
 496         len += sprintf(buf, "Filename\t\t\tType\t\tSize\tUsed\tPriority\n");
 497         for (i = 0 ; i < nr_swapfiles ; i++, ptr++) {
 498                 if (ptr->flags & SWP_USED) {
 499                         char * path = d_path(ptr->swap_file, ptr->swap_vfsmnt,
 500                                                 page, PAGE_SIZE);
 501
 502                         len += sprintf(buf + len, "%-31s ", path);
 503
 504                         if (!ptr->swap_device)
 505                                 len += sprintf(buf + len, "file\t\t");
 506                         else
 507                                 len += sprintf(buf + len, "partition\t");
 508
 509                         usedswap = 0;
 510                         for (j = 0; j < ptr->max; ++j)
 511                                 switch (ptr->swap_map[j]) {
 512                                         case SWAP_MAP_BAD:
 513                                         case 0:
 514                                                 continue;
 515                                         default:
 516                                                 usedswap++;
 517                                 }
 518                         len += sprintf(buf + len, "%d\t%d\t%d\n", ptr->pages << (PAGE_SHIFT - 10),
 519                                 usedswap << (PAGE_SHIFT - 10), ptr->prio);
 520                 }
 521         }
 522         free_page((unsigned long) page);
 523         return len;
 524 }
 525
 526 int is_swap_partition(kdev_t dev) {
 527         struct swap_info_struct *ptr = swap_info;
 528         int i;
 529
 530         for (i = 0 ; i < nr_swapfiles ; i++, ptr++) {
 531                 if (ptr->flags & SWP_USED)
 532                         if (ptr->swap_device == dev)
 533                                 return 1;
 534         }
 535         return 0;
 536 }
 537
 538 /*
 539  * Written 01/25/92 by Simmule Turner, heavily changed by Linus.
 540  *
 541  * The swapon system call
 542  */
 543 asmlinkage long sys_swapon(const char * specialfile, int swap_flags)
 544 {
 545         struct swap_info_struct * p;
 546         struct nameidata nd;
 547         struct inode * swap_inode;
 548         unsigned int type;
 549         int i, j, prev;
 550         int error;
 551         static int least_priority = 0;
 552         union swap_header *swap_header = 0;
 553         int swap_header_version;
 554         int nr_good_pages = 0;
 555         unsigned long maxpages;
 556         int swapfilesize;
 557         struct block_device *bdev = NULL;
 558
 559         if (!capable(CAP_SYS_ADMIN))
 560                 return -EPERM;
 561         lock_kernel();
 562         p = swap_info;
 563         for (type = 0 ; type < nr_swapfiles ; type++,p++)
 564                 if (!(p->flags & SWP_USED))
 565                         break;
 566         error = -EPERM;
 567         if (type >= MAX_SWAPFILES)
 568                 goto out;
 569         if (type >= nr_swapfiles)
 570                 nr_swapfiles = type+1;
 571         p->flags = SWP_USED;
 572         p->swap_file = NULL;
 573         p->swap_vfsmnt = NULL;
 574         p->swap_device = 0;
 575         p->swap_map = NULL;
 576         p->lowest_bit = 0;
 577         p->highest_bit = 0;
 578         p->cluster_nr = 0;
 579         p->sdev_lock = SPIN_LOCK_UNLOCKED;
 580         p->max = 1;
 581         p->next = -1;
 582         if (swap_flags & SWAP_FLAG_PREFER) {
 583                 p->prio =
 584                   (swap_flags & SWAP_FLAG_PRIO_MASK)>>SWAP_FLAG_PRIO_SHIFT;
 585         } else {
 586                 p->prio = --least_priority;
 587         }
 588         error = user_path_walk(specialfile, &nd);
 589         if (error)
 590                 goto bad_swap_2;
 591
 592         p->swap_file = nd.dentry;
 593         p->swap_vfsmnt = nd.mnt;
 594         swap_inode = nd.dentry->d_inode;
 595         error = -EINVAL;
 596
 597         if (S_ISBLK(swap_inode->i_mode)) {
 598                 kdev_t dev = swap_inode->i_rdev;
 599                 struct block_device_operations *bdops;
 600
 601                 p->swap_device = dev;
 602                 set_blocksize(dev, PAGE_SIZE);
 603
 604                 bdev = swap_inode->i_bdev;
 605                 bdops = devfs_get_ops(devfs_get_handle_from_inode(swap_inode));
 606                 if (bdops) bdev->bd_op = bdops;
 607
 608                 error = blkdev_get(bdev, FMODE_READ|FMODE_WRITE, 0, BDEV_SWAP);
 609                 if (error)
 610                         goto bad_swap_2;
 611                 set_blocksize(dev, PAGE_SIZE);
 612                 error = -ENODEV;
 613                 if (!dev || (blk_size[MAJOR(dev)] &&
 614                      !blk_size[MAJOR(dev)][MINOR(dev)]))
 615                         goto bad_swap;
 616                 error = -EBUSY;
 617                 for (i = 0 ; i < nr_swapfiles ; i++) {
 618                         if (i == type)
 619                                 continue;
 620                         if (dev == swap_info[i].swap_device)
 621                                 goto bad_swap;
 622                 }
 623                 swapfilesize = 0;
 624                 if (blk_size[MAJOR(dev)])
 625                         swapfilesize = blk_size[MAJOR(dev)][MINOR(dev)]
 626                                 >> (PAGE_SHIFT - 10);
 627         } else if (S_ISREG(swap_inode->i_mode)) {
 628                 error = -EBUSY;
 629                 for (i = 0 ; i < nr_swapfiles ; i++) {
 630                         if (i == type || !swap_info[i].swap_file)
 631                                 continue;
 632                         if (swap_inode == swap_info[i].swap_file->d_inode)
 633                                 goto bad_swap;
 634                 }
 635                 swapfilesize = swap_inode->i_size >> PAGE_SHIFT;
 636         } else
 637                 goto bad_swap;
 638
 639         swap_header = (void *) __get_free_page(GFP_USER);
 640         if (!swap_header) {
 641                 printk("Unable to start swapping: out of memory :-)\n");
 642                 error = -ENOMEM;
 643                 goto bad_swap;
 644         }
 645
 646         lock_page(mem_map + MAP_NR(swap_header));
 647         rw_swap_page_nolock(READ, SWP_ENTRY(type,0), (char *) swap_header, 1);
 648
 649         if (!memcmp("SWAP-SPACE",swap_header->magic.magic,10))
 650                 swap_header_version = 1;
 651         else if (!memcmp("SWAPSPACE2",swap_header->magic.magic,10))
 652                 swap_header_version = 2;
 653         else {
 654                 printk("Unable to find swap-space signature\n");
 655                 error = -EINVAL;
 656                 goto bad_swap;
 657         }
 658
 659         switch (swap_header_version) {
 660         case 1:
 661                 memset(((char *) swap_header)+PAGE_SIZE-10,0,10);
 662                 j = 0;
 663                 p->lowest_bit = 0;
 664                 p->highest_bit = 0;
 665                 for (i = 1 ; i < 8*PAGE_SIZE ; i++) {
 666                         if (test_bit(i,(char *) swap_header)) {
 667                                 if (!p->lowest_bit)
 668                                         p->lowest_bit = i;
 669                                 p->highest_bit = i;
 670                                 p->max = i+1;
 671                                 j++;
 672                         }
 673                 }
 674                 nr_good_pages = j;
 675                 p->swap_map = vmalloc(p->max * sizeof(short));
 676                 if (!p->swap_map) {
 677                         error = -ENOMEM;
 678                         goto bad_swap;
 679                 }
 680                 for (i = 1 ; i < p->max ; i++) {
 681                         if (test_bit(i,(char *) swap_header))
 682                                 p->swap_map[i] = 0;
 683                         else
 684                                 p->swap_map[i] = SWAP_MAP_BAD;
 685                 }
 686                 break;
 687
 688         case 2:
 689                 /* Check the swap header's sub-version and the size of
 690                    the swap file and bad block lists */
 691                 if (swap_header->info.version != 1) {
 692                         printk(KERN_WARNING
 693                                "Unable to handle swap header version %d\n",
 694                                swap_header->info.version);
 695                         error = -EINVAL;
 696                         goto bad_swap;
 697                 }
 698
 699                 p->lowest_bit  = 1;
 700                 p->highest_bit = swap_header->info.last_page - 1;
 701                 p->max         = swap_header->info.last_page;
 702
 703                 maxpages = SWP_OFFSET(SWP_ENTRY(0,~0UL));
 704                 if (p->max >= maxpages)
 705                         p->max = maxpages-1;
 706
 707                 error = -EINVAL;
 708                 if (swap_header->info.nr_badpages > MAX_SWAP_BADPAGES)
 709                         goto bad_swap;
 710
 711                 /* OK, set up the swap map and apply the bad block list */
 712                 if (!(p->swap_map = vmalloc (p->max * sizeof(short)))) {
 713                         error = -ENOMEM;
 714                         goto bad_swap;
 715                 }
 716
 717                 error = 0;
 718                 memset(p->swap_map, 0, p->max * sizeof(short));
 719                 for (i=0; i<swap_header->info.nr_badpages; i++) {
 720                         int page = swap_header->info.badpages[i];
 721                         if (page <= 0 || page >= swap_header->info.last_page)
 722                                 error = -EINVAL;
 723                         else
 724                                 p->swap_map[page] = SWAP_MAP_BAD;
 725                 }
 726                 nr_good_pages = swap_header->info.last_page -
 727                                 swap_header->info.nr_badpages -
 728                                 1 /* header page */;
 729                 if (error)
 730                         goto bad_swap;
 731         }
 732
 733         if (swapfilesize && p->max > swapfilesize) {
 734                 printk(KERN_WARNING
 735                        "Swap area shorter than signature indicates\n");
 736                 error = -EINVAL;
 737                 goto bad_swap;
 738         }
 739         if (!nr_good_pages) {
 740                 printk(KERN_WARNING "Empty swap-file\n");
 741                 error = -EINVAL;
 742                 goto bad_swap;
 743         }
 744         p->swap_map[0] = SWAP_MAP_BAD;
 745         p->flags = SWP_WRITEOK;
 746         p->pages = nr_good_pages;
 747         swap_list_lock();
 748         nr_swap_pages += nr_good_pages;
 749         printk(KERN_INFO "Adding Swap: %dk swap-space (priority %d)\n",
 750                nr_good_pages<<(PAGE_SHIFT-10), p->prio);
 751
 752         /* insert swap space into swap_list: */
 753         prev = -1;
 754         for (i = swap_list.head; i >= 0; i = swap_info[i].next) {
 755                 if (p->prio >= swap_info[i].prio) {
 756                         break;
 757                 }
 758                 prev = i;
 759         }
 760         p->next = i;
 761         if (prev < 0) {
 762                 swap_list.head = swap_list.next = p - swap_info;
 763         } else {
 764                 swap_info[prev].next = p - swap_info;
 765         }
 766         swap_list_unlock();
 767         error = 0;
 768         goto out;
 769 bad_swap:
 770         if (bdev)
 771                 blkdev_put(bdev, BDEV_SWAP);
 772 bad_swap_2:
 773         if (p->swap_map)
 774                 vfree(p->swap_map);
 775         nd.mnt = p->swap_vfsmnt;
 776         nd.dentry = p->swap_file;
 777         p->swap_device = 0;
 778         p->swap_file = NULL;
 779         p->swap_vfsmnt = NULL;
 780         p->swap_map = NULL;
 781         p->flags = 0;
 782         if (!(swap_flags & SWAP_FLAG_PREFER))
 783                 ++least_priority;
 784         path_release(&nd);
 785 out:
 786         if (swap_header)
 787                 free_page((long) swap_header);
 788         unlock_kernel();
 789         return error;
 790 }
 791
 792 void si_swapinfo(struct sysinfo *val)
 793 {
 794         unsigned int i;
 795         unsigned long freeswap = 0;
 796         unsigned long totalswap = 0;
 797
 798         for (i = 0; i < nr_swapfiles; i++) {
 799                 unsigned int j;
 800                 if ((swap_info[i].flags & SWP_WRITEOK) != SWP_WRITEOK)
 801                         continue;
 802                 for (j = 0; j < swap_info[i].max; ++j) {
 803                         switch (swap_info[i].swap_map[j]) {
 804                                 case SWAP_MAP_BAD:
 805                                         continue;
 806                                 case 0:
 807                                         freeswap++;
 808                                 default:
 809                                         totalswap++;
 810                         }
 811                 }
 812         }
 813         val->freeswap = freeswap;
 814         val->totalswap = totalswap;
 815         return;
 816 }
 817
 818 /*
 819  * Verify that a swap entry is valid and increment its swap map count.
 820  * Kernel_lock is held, which guarantees existance of swap device.
 821  *
 822  * Note: if swap_map[] reaches SWAP_MAP_MAX the entries are treated as
 823  * "permanent", but will be reclaimed by the next swapoff.
 824  */
 825 int swap_duplicate(swp_entry_t entry)
 826 {
 827         struct swap_info_struct * p;
 828         unsigned long offset, type;
 829         int result = 0;
 830
 831         /* Swap entry 0 is illegal */
 832         if (!entry.val)
 833                 goto out;
 834         type = SWP_TYPE(entry);
 835         if (type >= nr_swapfiles)
 836                 goto bad_file;
 837         p = type + swap_info;
 838         offset = SWP_OFFSET(entry);
 839         if (offset >= p->max)
 840                 goto bad_offset;
 841         if (!p->swap_map[offset])
 842                 goto bad_unused;
 843         /*
 844          * Entry is valid, so increment the map count.
 845          */
 846         swap_device_lock(p);
 847         if (p->swap_map[offset] < SWAP_MAP_MAX)
 848                 p->swap_map[offset]++;
 849         else {
 850                 static int overflow = 0;
 851                 if (overflow++ < 5)
 852                         printk("VM: swap entry overflow\n");
 853                 p->swap_map[offset] = SWAP_MAP_MAX;
 854         }
 855         swap_device_unlock(p);
 856         result = 1;
 857 out:
 858         return result;
 859
 860 bad_file:
 861         printk("Bad swap file entry %08lx\n", entry.val);
 862         goto out;
 863 bad_offset:
 864         printk("Bad swap offset entry %08lx\n", entry.val);
 865         goto out;
 866 bad_unused:
 867         printk("Unused swap offset entry in swap_dup %08lx\n", entry.val);
 868         goto out;
 869 }
 870
 871 /*
 872  * Page lock needs to be held in all cases to prevent races with
 873  * swap file deletion.
 874  */
 875 int swap_count(struct page *page)
 876 {
 877         struct swap_info_struct * p;
 878         unsigned long offset, type;
 879         swp_entry_t entry;
 880         int retval = 0;
 881
 882         entry.val = page->index;
 883         if (!entry.val)
 884                 goto bad_entry;
 885         type = SWP_TYPE(entry);
 886         if (type >= nr_swapfiles)
 887                 goto bad_file;
 888         p = type + swap_info;
 889         offset = SWP_OFFSET(entry);
 890         if (offset >= p->max)
 891                 goto bad_offset;
 892         if (!p->swap_map[offset])
 893                 goto bad_unused;
 894         retval = p->swap_map[offset];
 895 out:
 896         return retval;
 897
 898 bad_entry:
 899         printk(KERN_ERR "swap_count: null entry!\n");
 900         goto out;
 901 bad_file:
 902         printk("Bad swap file entry %08lx\n", entry.val);
 903         goto out;
 904 bad_offset:
 905         printk("Bad swap offset entry %08lx\n", entry.val);
 906         goto out;
 907 bad_unused:
 908         printk("Unused swap offset entry in swap_count %08lx\n", entry.val);
 909         goto out;
 910 }
 911
 912 /*
 913  * Kernel_lock protects against swap device deletion.
 914  */
 915 void get_swaphandle_info(swp_entry_t entry, unsigned long *offset,
 916                         kdev_t *dev, struct inode **swapf)
 917 {
 918         unsigned long type;
 919         struct swap_info_struct *p;
 920
 921         type = SWP_TYPE(entry);
 922         if (type >= nr_swapfiles) {
 923                 printk("Internal error: bad swap-device\n");
 924                 return;
 925         }
 926
 927         p = &swap_info[type];
 928         *offset = SWP_OFFSET(entry);
 929         if (*offset >= p->max) {
 930                 printk("rw_swap_page: weirdness\n");
 931                 return;
 932         }
 933         if (p->swap_map && !p->swap_map[*offset]) {
 934                 printk("VM: Bad swap entry %08lx\n", entry.val);
 935                 return;
 936         }
 937         if (!(p->flags & SWP_USED)) {
 938                 printk(KERN_ERR "rw_swap_page: "
 939                         "Trying to swap to unused swap-device\n");
 940                 return;
 941         }
 942
 943         if (p->swap_device) {
 944                 *dev = p->swap_device;
 945         } else if (p->swap_file) {
 946                 *swapf = p->swap_file->d_inode;
 947         } else {
 948                 printk(KERN_ERR "rw_swap_page: no swap file or device\n");
 949         }
 950         return;
 951 }
 952
 953 /*
 954  * Kernel_lock protects against swap device deletion. Grab an extra
 955  * reference on the swaphandle so that it dos not become unused.
 956  */
 957 int valid_swaphandles(swp_entry_t entry, unsigned long *offset)
 958 {
 959         int ret = 0, i = 1 << page_cluster;
 960         unsigned long toff;
 961         struct swap_info_struct *swapdev = SWP_TYPE(entry) + swap_info;
 962
 963         *offset = SWP_OFFSET(entry);
 964         toff = *offset = (*offset >> page_cluster) << page_cluster;
 965
 966         swap_device_lock(swapdev);
 967         do {
 968                 /* Don't read-ahead past the end of the swap area */
 969                 if (toff >= swapdev->max)
 970                         break;
 971                 /* Don't read in bad or busy pages */
 972                 if (!swapdev->swap_map[toff])
 973                         break;
 974                 if (swapdev->swap_map[toff] == SWAP_MAP_BAD)
 975                         break;
 976                 swapdev->swap_map[toff]++;
 977                 toff++;
 978                 ret++;
 979         } while (--i);
 980         swap_device_unlock(swapdev);
 981         return ret;
 982 }