fs/squashfs/cache.c

   1 /*
   2  * Squashfs - a compressed read only filesystem for Linux
   3  *
   4  * Copyright (c) 2002, 2003, 2004, 2005, 2006, 2007, 2008
   5  * Phillip Lougher <phillip@squashfs.org.uk>
   6  *
   7  * This program is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU General Public License
   9  * as published by the Free Software Foundation; either version 2,
  10  * or (at your option) any later version.
  11  *
  12  * This program is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15  * GNU General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU General Public License
  18  * along with this program; if not, write to the Free Software
  19  * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  20  *
  21  * cache.c
  22  */
  23
  24 /*
  25  * Blocks in Squashfs are compressed.  To avoid repeatedly decompressing
  26  * recently accessed data Squashfs uses two small metadata and fragment caches.
  27  *
  28  * This file implements a generic cache implementation used for both caches,
  29  * plus functions layered ontop of the generic cache implementation to
  30  * access the metadata and fragment caches.
  31  *
  32  * To avoid out of memory and fragmentation issues with vmalloc the cache
  33  * uses sequences of kmalloced PAGE_CACHE_SIZE buffers.
  34  *
  35  * It should be noted that the cache is not used for file datablocks, these
  36  * are decompressed and cached in the page-cache in the normal way.  The
  37  * cache is only used to temporarily cache fragment and metadata blocks
  38  * which have been read as as a result of a metadata (i.e. inode or
  39  * directory) or fragment access.  Because metadata and fragments are packed
  40  * together into blocks (to gain greater compression) the read of a particular
  41  * piece of metadata or fragment will retrieve other metadata/fragments which
  42  * have been packed with it, these because of locality-of-reference may be read
  43  * in the near future. Temporarily caching them ensures they are available for
  44  * near future access without requiring an additional read and decompress.
  45  */
  46
  47 #include <linux/fs.h>
  48 #include <linux/vfs.h>
  49 #include <linux/slab.h>
  50 #include <linux/vmalloc.h>
  51 #include <linux/sched.h>
  52 #include <linux/spinlock.h>
  53 #include <linux/wait.h>
  54 #include <linux/pagemap.h>
  55
  56 #include "squashfs_fs.h"
  57 #include "squashfs_fs_sb.h"
  58 #include "squashfs.h"
  59
  60 /*
  61  * Look-up block in cache, and increment usage count.  If not in cache, read
  62  * and decompress it from disk.
  63  */
  64 struct squashfs_cache_entry *squashfs_cache_get(struct super_block *sb,
  65         struct squashfs_cache *cache, u64 block, int length)
  66 {
  67         int i, n;
  68         struct squashfs_cache_entry *entry;
  69
  70         spin_lock(&cache->lock);
  71
  72         while (1) {
  73                 for (i = cache->curr_blk, n = 0; n < cache->entries; n++) {
  74                         if (cache->entry[i].block == block) {
  75                                 cache->curr_blk = i;
  76                                 break;
  77                         }
  78                         i = (i + 1) % cache->entries;
  79                 }
  80
  81                 if (n == cache->entries) {
  82                         /*
  83                          * Block not in cache, if all cache entries are used
  84                          * go to sleep waiting for one to become available.
  85                          */
  86                         if (cache->unused == 0) {
  87                                 cache->num_waiters++;
  88                                 spin_unlock(&cache->lock);
  89                                 wait_event(cache->wait_queue, cache->unused);
  90                                 spin_lock(&cache->lock);
  91                                 cache->num_waiters--;
  92                                 continue;
  93                         }
  94
  95                         /*
  96                          * At least one unused cache entry.  A simple
  97                          * round-robin strategy is used to choose the entry to
  98                          * be evicted from the cache.
  99                          */
 100                         i = cache->next_blk;
 101                         for (n = 0; n < cache->entries; n++) {
 102                                 if (cache->entry[i].refcount == 0)
 103                                         break;
 104                                 i = (i + 1) % cache->entries;
 105                         }
 106
 107                         cache->next_blk = (i + 1) % cache->entries;
 108                         entry = &cache->entry[i];
 109
 110                         /*
 111                          * Initialise chosen cache entry, and fill it in from
 112                          * disk.
 113                          */
 114                         cache->unused--;
 115                         entry->block = block;
 116                         entry->refcount = 1;
 117                         entry->pending = 1;
 118                         entry->num_waiters = 0;
 119                         entry->error = 0;
 120                         spin_unlock(&cache->lock);
 121
 122                         entry->length = squashfs_read_data(sb, entry->data,
 123                                 block, length, &entry->next_index,
 124                                 cache->block_size, cache->pages);
 125
 126                         spin_lock(&cache->lock);
 127
 128                         if (entry->length < 0)
 129                                 entry->error = entry->length;
 130
 131                         entry->pending = 0;
 132
 133                         /*
 134                          * While filling this entry one or more other processes
 135                          * have looked it up in the cache, and have slept
 136                          * waiting for it to become available.
 137                          */
 138                         if (entry->num_waiters) {
 139                                 spin_unlock(&cache->lock);
 140                                 wake_up_all(&entry->wait_queue);
 141                         } else
 142                                 spin_unlock(&cache->lock);
 143
 144                         goto out;
 145                 }
 146
 147                 /*
 148                  * Block already in cache.  Increment refcount so it doesn't
 149                  * get reused until we're finished with it, if it was
 150                  * previously unused there's one less cache entry available
 151                  * for reuse.
 152                  */
 153                 entry = &cache->entry[i];
 154                 if (entry->refcount == 0)
 155                         cache->unused--;
 156                 entry->refcount++;
 157
 158                 /*
 159                  * If the entry is currently being filled in by another process
 160                  * go to sleep waiting for it to become available.
 161                  */
 162                 if (entry->pending) {
 163                         entry->num_waiters++;
 164                         spin_unlock(&cache->lock);
 165                         wait_event(entry->wait_queue, !entry->pending);
 166                 } else
 167                         spin_unlock(&cache->lock);
 168
 169                 goto out;
 170         }
 171
 172 out:
 173         TRACE("Got %s %d, start block %lld, refcount %d, error %d\n",
 174                 cache->name, i, entry->block, entry->refcount, entry->error);
 175
 176         if (entry->error)
 177                 ERROR("Unable to read %s cache entry [%llx]\n", cache->name,
 178                                                         block);
 179         return entry;
 180 }
 181
 182
 183 /*
 184  * Release cache entry, once usage count is zero it can be reused.
 185  */
 186 void squashfs_cache_put(struct squashfs_cache_entry *entry)
 187 {
 188         struct squashfs_cache *cache = entry->cache;
 189
 190         spin_lock(&cache->lock);
 191         entry->refcount--;
 192         if (entry->refcount == 0) {
 193                 cache->unused++;
 194                 /*
 195                  * If there's any processes waiting for a block to become
 196                  * available, wake one up.
 197                  */
 198                 if (cache->num_waiters) {
 199                         spin_unlock(&cache->lock);
 200                         wake_up(&cache->wait_queue);
 201                         return;
 202                 }
 203         }
 204         spin_unlock(&cache->lock);
 205 }
 206
 207 /*
 208  * Delete cache reclaiming all kmalloced buffers.
 209  */
 210 void squashfs_cache_delete(struct squashfs_cache *cache)
 211 {
 212         int i, j;
 213
 214         if (cache == NULL)
 215                 return;
 216
 217         for (i = 0; i < cache->entries; i++) {
 218                 if (cache->entry[i].data) {
 219                         for (j = 0; j < cache->pages; j++)
 220                                 kfree(cache->entry[i].data[j]);
 221                         kfree(cache->entry[i].data);
 222                 }
 223         }
 224
 225         kfree(cache->entry);
 226         kfree(cache);
 227 }
 228
 229
 230 /*
 231  * Initialise cache allocating the specified number of entries, each of
 232  * size block_size.  To avoid vmalloc fragmentation issues each entry
 233  * is allocated as a sequence of kmalloced PAGE_CACHE_SIZE buffers.
 234  */
 235 struct squashfs_cache *squashfs_cache_init(char *name, int entries,
 236         int block_size)
 237 {
 238         int i, j;
 239         struct squashfs_cache *cache = kzalloc(sizeof(*cache), GFP_KERNEL);
 240
 241         if (cache == NULL) {
 242                 ERROR("Failed to allocate %s cache\n", name);
 243                 return NULL;
 244         }
 245
 246         cache->entry = kcalloc(entries, sizeof(*(cache->entry)), GFP_KERNEL);
 247         if (cache->entry == NULL) {
 248                 ERROR("Failed to allocate %s cache\n", name);
 249                 goto cleanup;
 250         }
 251
 252         cache->curr_blk = 0;
 253         cache->next_blk = 0;
 254         cache->unused = entries;
 255         cache->entries = entries;
 256         cache->block_size = block_size;
 257         cache->pages = block_size >> PAGE_CACHE_SHIFT;
 258         cache->pages = cache->pages ? cache->pages : 1;
 259         cache->name = name;
 260         cache->num_waiters = 0;
 261         spin_lock_init(&cache->lock);
 262         init_waitqueue_head(&cache->wait_queue);
 263
 264         for (i = 0; i < entries; i++) {
 265                 struct squashfs_cache_entry *entry = &cache->entry[i];
 266
 267                 init_waitqueue_head(&cache->entry[i].wait_queue);
 268                 entry->cache = cache;
 269                 entry->block = SQUASHFS_INVALID_BLK;
 270                 entry->data = kcalloc(cache->pages, sizeof(void *), GFP_KERNEL);
 271                 if (entry->data == NULL) {
 272                         ERROR("Failed to allocate %s cache entry\n", name);
 273                         goto cleanup;
 274                 }
 275
 276                 for (j = 0; j < cache->pages; j++) {
 277                         entry->data[j] = kmalloc(PAGE_CACHE_SIZE, GFP_KERNEL);
 278                         if (entry->data[j] == NULL) {
 279                                 ERROR("Failed to allocate %s buffer\n", name);
 280                                 goto cleanup;
 281                         }
 282                 }
 283         }
 284
 285         return cache;
 286
 287 cleanup:
 288         squashfs_cache_delete(cache);
 289         return NULL;
 290 }
 291
 292
 293 /*
 294  * Copy up to length bytes from cache entry to buffer starting at offset bytes
 295  * into the cache entry.  If there's not length bytes then copy the number of
 296  * bytes available.  In all cases return the number of bytes copied.
 297  */
 298 int squashfs_copy_data(void *buffer, struct squashfs_cache_entry *entry,
 299                 int offset, int length)
 300 {
 301         int remaining = length;
 302
 303         if (length == 0)
 304                 return 0;
 305         else if (buffer == NULL)
 306                 return min(length, entry->length - offset);
 307
 308         while (offset < entry->length) {
 309                 void *buff = entry->data[offset / PAGE_CACHE_SIZE]
 310                                 + (offset % PAGE_CACHE_SIZE);
 311                 int bytes = min_t(int, entry->length - offset,
 312                                 PAGE_CACHE_SIZE - (offset % PAGE_CACHE_SIZE));
 313
 314                 if (bytes >= remaining) {
 315                         memcpy(buffer, buff, remaining);
 316                         remaining = 0;
 317                         break;
 318                 }
 319
 320                 memcpy(buffer, buff, bytes);
 321                 buffer += bytes;
 322                 remaining -= bytes;
 323                 offset += bytes;
 324         }
 325
 326         return length - remaining;
 327 }
 328
 329
 330 /*
 331  * Read length bytes from metadata position <block, offset> (block is the
 332  * start of the compressed block on disk, and offset is the offset into
 333  * the block once decompressed).  Data is packed into consecutive blocks,
 334  * and length bytes may require reading more than one block.
 335  */
 336 int squashfs_read_metadata(struct super_block *sb, void *buffer,
 337                 u64 *block, int *offset, int length)
 338 {
 339         struct squashfs_sb_info *msblk = sb->s_fs_info;
 340         int bytes, res = length;
 341         struct squashfs_cache_entry *entry;
 342
 343         TRACE("Entered squashfs_read_metadata [%llx:%x]\n", *block, *offset);
 344
 345         while (length) {
 346                 entry = squashfs_cache_get(sb, msblk->block_cache, *block, 0);
 347                 if (entry->error) {
 348                         res = entry->error;
 349                         goto error;
 350                 } else if (*offset >= entry->length) {
 351                         res = -EIO;
 352                         goto error;
 353                 }
 354
 355                 bytes = squashfs_copy_data(buffer, entry, *offset, length);
 356                 if (buffer)
 357                         buffer += bytes;
 358                 length -= bytes;
 359                 *offset += bytes;
 360
 361                 if (*offset == entry->length) {
 362                         *block = entry->next_index;
 363                         *offset = 0;
 364                 }
 365
 366                 squashfs_cache_put(entry);
 367         }
 368
 369         return res;
 370
 371 error:
 372         squashfs_cache_put(entry);
 373         return res;
 374 }
 375
 376
 377 /*
 378  * Look-up in the fragmment cache the fragment located at <start_block> in the
 379  * filesystem.  If necessary read and decompress it from disk.
 380  */
 381 struct squashfs_cache_entry *squashfs_get_fragment(struct super_block *sb,
 382                                 u64 start_block, int length)
 383 {
 384         struct squashfs_sb_info *msblk = sb->s_fs_info;
 385
 386         return squashfs_cache_get(sb, msblk->fragment_cache, start_block,
 387                 length);
 388 }
 389
 390
 391 /*
 392  * Read and decompress the datablock located at <start_block> in the
 393  * filesystem.  The cache is used here to avoid duplicating locking and
 394  * read/decompress code.
 395  */
 396 struct squashfs_cache_entry *squashfs_get_datablock(struct super_block *sb,
 397                                 u64 start_block, int length)
 398 {
 399         struct squashfs_sb_info *msblk = sb->s_fs_info;
 400
 401         return squashfs_cache_get(sb, msblk->read_page, start_block, length);
 402 }
 403
 404
 405 /*
 406  * Read a filesystem table (uncompressed sequence of bytes) from disk
 407  */
 408 void *squashfs_read_table(struct super_block *sb, u64 block, int length)
 409 {
 410         int pages = (length + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
 411         int i, res;
 412         void *table, *buffer, **data;
 413
 414         table = buffer = kmalloc(length, GFP_KERNEL);
 415         if (table == NULL)
 416                 return ERR_PTR(-ENOMEM);
 417
 418         data = kcalloc(pages, sizeof(void *), GFP_KERNEL);
 419         if (data == NULL) {
 420                 res = -ENOMEM;
 421                 goto failed;
 422         }
 423
 424         for (i = 0; i < pages; i++, buffer += PAGE_CACHE_SIZE)
 425                 data[i] = buffer;
 426
 427         res = squashfs_read_data(sb, data, block, length |
 428                 SQUASHFS_COMPRESSED_BIT_BLOCK, NULL, length, pages);
 429
 430         kfree(data);
 431
 432         if (res < 0)
 433                 goto failed;
 434
 435         return table;
 436
 437 failed:
 438         kfree(table);
 439         return ERR_PTR(res);
 440 }