fs/overlayfs/readdir.c

   1 /*
   2  *
   3  * Copyright (C) 2011 Novell Inc.
   4  *
   5  * This program is free software; you can redistribute it and/or modify it
   6  * under the terms of the GNU General Public License version 2 as published by
   7  * the Free Software Foundation.
   8  */
   9
  10 #include <linux/fs.h>
  11 #include <linux/slab.h>
  12 #include <linux/namei.h>
  13 #include <linux/file.h>
  14 #include <linux/xattr.h>
  15 #include <linux/rbtree.h>
  16 #include <linux/security.h>
  17 #include <linux/cred.h>
  18 #include "overlayfs.h"
  19
  20 struct ovl_cache_entry {
  21         unsigned int len;
  22         unsigned int type;
  23         u64 ino;
  24         struct list_head l_node;
  25         struct rb_node node;
  26         bool is_whiteout;
  27         char name[];
  28 };
  29
  30 struct ovl_dir_cache {
  31         long refcount;
  32         u64 version;
  33         struct list_head entries;
  34 };
  35
  36 struct ovl_readdir_data {
  37         struct dir_context ctx;
  38         bool is_merge;
  39         struct rb_root root;
  40         struct list_head *list;
  41         struct list_head middle;
  42         struct dentry *dir;
  43         int count;
  44         int err;
  45 };
  46
  47 struct ovl_dir_file {
  48         bool is_real;
  49         bool is_upper;
  50         struct ovl_dir_cache *cache;
  51         struct list_head *cursor;
  52         struct file *realfile;
  53         struct file *upperfile;
  54 };
  55
  56 static struct ovl_cache_entry *ovl_cache_entry_from_node(struct rb_node *n)
  57 {
  58         return container_of(n, struct ovl_cache_entry, node);
  59 }
  60
  61 static struct ovl_cache_entry *ovl_cache_entry_find(struct rb_root *root,
  62                                                     const char *name, int len)
  63 {
  64         struct rb_node *node = root->rb_node;
  65         int cmp;
  66
  67         while (node) {
  68                 struct ovl_cache_entry *p = ovl_cache_entry_from_node(node);
  69
  70                 cmp = strncmp(name, p->name, len);
  71                 if (cmp > 0)
  72                         node = p->node.rb_right;
  73                 else if (cmp < 0 || len < p->len)
  74                         node = p->node.rb_left;
  75                 else
  76                         return p;
  77         }
  78
  79         return NULL;
  80 }
  81
  82 static struct ovl_cache_entry *ovl_cache_entry_new(struct dentry *dir,
  83                                                    const char *name, int len,
  84                                                    u64 ino, unsigned int d_type)
  85 {
  86         struct ovl_cache_entry *p;
  87         size_t size = offsetof(struct ovl_cache_entry, name[len + 1]);
  88
  89         p = kmalloc(size, GFP_KERNEL);
  90         if (!p)
  91                 return NULL;
  92
  93         memcpy(p->name, name, len);
  94         p->name[len] = '\0';
  95         p->len = len;
  96         p->type = d_type;
  97         p->ino = ino;
  98         p->is_whiteout = false;
  99
 100         if (d_type == DT_CHR) {
 101                 struct dentry *dentry;
 102                 const struct cred *old_cred;
 103                 struct cred *override_cred;
 104
 105                 override_cred = prepare_creds();
 106                 if (!override_cred) {
 107                         kfree(p);
 108                         return NULL;
 109                 }
 110
 111                 /*
 112                  * CAP_DAC_OVERRIDE for lookup
 113                  */
 114                 cap_raise(override_cred->cap_effective, CAP_DAC_OVERRIDE);
 115                 old_cred = override_creds(override_cred);
 116
 117                 dentry = lookup_one_len(name, dir, len);
 118                 if (!IS_ERR(dentry)) {
 119                         p->is_whiteout = ovl_is_whiteout(dentry);
 120                         dput(dentry);
 121                 }
 122                 revert_creds(old_cred);
 123                 put_cred(override_cred);
 124         }
 125         return p;
 126 }
 127
 128 static int ovl_cache_entry_add_rb(struct ovl_readdir_data *rdd,
 129                                   const char *name, int len, u64 ino,
 130                                   unsigned int d_type)
 131 {
 132         struct rb_node **newp = &rdd->root.rb_node;
 133         struct rb_node *parent = NULL;
 134         struct ovl_cache_entry *p;
 135
 136         while (*newp) {
 137                 int cmp;
 138                 struct ovl_cache_entry *tmp;
 139
 140                 parent = *newp;
 141                 tmp = ovl_cache_entry_from_node(*newp);
 142                 cmp = strncmp(name, tmp->name, len);
 143                 if (cmp > 0)
 144                         newp = &tmp->node.rb_right;
 145                 else if (cmp < 0 || len < tmp->len)
 146                         newp = &tmp->node.rb_left;
 147                 else
 148                         return 0;
 149         }
 150
 151         p = ovl_cache_entry_new(rdd->dir, name, len, ino, d_type);
 152         if (p == NULL)
 153                 return -ENOMEM;
 154
 155         list_add_tail(&p->l_node, rdd->list);
 156         rb_link_node(&p->node, parent, newp);
 157         rb_insert_color(&p->node, &rdd->root);
 158
 159         return 0;
 160 }
 161
 162 static int ovl_fill_lower(struct ovl_readdir_data *rdd,
 163                           const char *name, int namelen,
 164                           loff_t offset, u64 ino, unsigned int d_type)
 165 {
 166         struct ovl_cache_entry *p;
 167
 168         p = ovl_cache_entry_find(&rdd->root, name, namelen);
 169         if (p) {
 170                 list_move_tail(&p->l_node, &rdd->middle);
 171         } else {
 172                 p = ovl_cache_entry_new(rdd->dir, name, namelen, ino, d_type);
 173                 if (p == NULL)
 174                         rdd->err = -ENOMEM;
 175                 else
 176                         list_add_tail(&p->l_node, &rdd->middle);
 177         }
 178
 179         return rdd->err;
 180 }
 181
 182 void ovl_cache_free(struct list_head *list)
 183 {
 184         struct ovl_cache_entry *p;
 185         struct ovl_cache_entry *n;
 186
 187         list_for_each_entry_safe(p, n, list, l_node)
 188                 kfree(p);
 189
 190         INIT_LIST_HEAD(list);
 191 }
 192
 193 static void ovl_cache_put(struct ovl_dir_file *od, struct dentry *dentry)
 194 {
 195         struct ovl_dir_cache *cache = od->cache;
 196
 197         WARN_ON(cache->refcount <= 0);
 198         cache->refcount--;
 199         if (!cache->refcount) {
 200                 if (ovl_dir_cache(dentry) == cache)
 201                         ovl_set_dir_cache(dentry, NULL);
 202
 203                 ovl_cache_free(&cache->entries);
 204                 kfree(cache);
 205         }
 206 }
 207
 208 static int ovl_fill_merge(struct dir_context *ctx, const char *name,
 209                           int namelen, loff_t offset, u64 ino,
 210                           unsigned int d_type)
 211 {
 212         struct ovl_readdir_data *rdd =
 213                 container_of(ctx, struct ovl_readdir_data, ctx);
 214
 215         rdd->count++;
 216         if (!rdd->is_merge)
 217                 return ovl_cache_entry_add_rb(rdd, name, namelen, ino, d_type);
 218         else
 219                 return ovl_fill_lower(rdd, name, namelen, offset, ino, d_type);
 220 }
 221
 222 static inline int ovl_dir_read(struct path *realpath,
 223                                struct ovl_readdir_data *rdd)
 224 {
 225         struct file *realfile;
 226         int err;
 227
 228         realfile = ovl_path_open(realpath, O_RDONLY | O_DIRECTORY);
 229         if (IS_ERR(realfile))
 230                 return PTR_ERR(realfile);
 231
 232         rdd->dir = realpath->dentry;
 233         rdd->ctx.pos = 0;
 234         do {
 235                 rdd->count = 0;
 236                 rdd->err = 0;
 237                 err = iterate_dir(realfile, &rdd->ctx);
 238                 if (err >= 0)
 239                         err = rdd->err;
 240         } while (!err && rdd->count);
 241         fput(realfile);
 242
 243         return err;
 244 }
 245
 246 static void ovl_dir_reset(struct file *file)
 247 {
 248         struct ovl_dir_file *od = file->private_data;
 249         struct ovl_dir_cache *cache = od->cache;
 250         struct dentry *dentry = file->f_path.dentry;
 251         enum ovl_path_type type = ovl_path_type(dentry);
 252
 253         if (cache && ovl_dentry_version_get(dentry) != cache->version) {
 254                 ovl_cache_put(od, dentry);
 255                 od->cache = NULL;
 256                 od->cursor = NULL;
 257         }
 258         WARN_ON(!od->is_real && !OVL_TYPE_MERGE(type));
 259         if (od->is_real && OVL_TYPE_MERGE(type))
 260                 od->is_real = false;
 261 }
 262
 263 static int ovl_dir_read_merged(struct dentry *dentry, struct list_head *list)
 264 {
 265         int err;
 266         struct path realpath;
 267         struct ovl_readdir_data rdd = {
 268                 .ctx.actor = ovl_fill_merge,
 269                 .list = list,
 270                 .root = RB_ROOT,
 271                 .is_merge = false,
 272         };
 273         int idx, next;
 274
 275         for (idx = 0; idx != -1; idx = next) {
 276                 next = ovl_path_next(idx, dentry, &realpath);
 277
 278                 if (next != -1) {
 279                         err = ovl_dir_read(&realpath, &rdd);
 280                         if (err)
 281                                 break;
 282                 } else {
 283                         /*
 284                          * Insert lowest layer entries before upper ones, this
 285                          * allows offsets to be reasonably constant
 286                          */
 287                         list_add(&rdd.middle, rdd.list);
 288                         rdd.is_merge = true;
 289                         err = ovl_dir_read(&realpath, &rdd);
 290                         list_del(&rdd.middle);
 291                 }
 292         }
 293         return err;
 294 }
 295
 296 static void ovl_seek_cursor(struct ovl_dir_file *od, loff_t pos)
 297 {
 298         struct list_head *p;
 299         loff_t off = 0;
 300
 301         list_for_each(p, &od->cache->entries) {
 302                 if (off >= pos)
 303                         break;
 304                 off++;
 305         }
 306         /* Cursor is safe since the cache is stable */
 307         od->cursor = p;
 308 }
 309
 310 static struct ovl_dir_cache *ovl_cache_get(struct dentry *dentry)
 311 {
 312         int res;
 313         struct ovl_dir_cache *cache;
 314
 315         cache = ovl_dir_cache(dentry);
 316         if (cache && ovl_dentry_version_get(dentry) == cache->version) {
 317                 cache->refcount++;
 318                 return cache;
 319         }
 320         ovl_set_dir_cache(dentry, NULL);
 321
 322         cache = kzalloc(sizeof(struct ovl_dir_cache), GFP_KERNEL);
 323         if (!cache)
 324                 return ERR_PTR(-ENOMEM);
 325
 326         cache->refcount = 1;
 327         INIT_LIST_HEAD(&cache->entries);
 328
 329         res = ovl_dir_read_merged(dentry, &cache->entries);
 330         if (res) {
 331                 ovl_cache_free(&cache->entries);
 332                 kfree(cache);
 333                 return ERR_PTR(res);
 334         }
 335
 336         cache->version = ovl_dentry_version_get(dentry);
 337         ovl_set_dir_cache(dentry, cache);
 338
 339         return cache;
 340 }
 341
 342 static int ovl_iterate(struct file *file, struct dir_context *ctx)
 343 {
 344         struct ovl_dir_file *od = file->private_data;
 345         struct dentry *dentry = file->f_path.dentry;
 346         struct ovl_cache_entry *p;
 347
 348         if (!ctx->pos)
 349                 ovl_dir_reset(file);
 350
 351         if (od->is_real)
 352                 return iterate_dir(od->realfile, ctx);
 353
 354         if (!od->cache) {
 355                 struct ovl_dir_cache *cache;
 356
 357                 cache = ovl_cache_get(dentry);
 358                 if (IS_ERR(cache))
 359                         return PTR_ERR(cache);
 360
 361                 od->cache = cache;
 362                 ovl_seek_cursor(od, ctx->pos);
 363         }
 364
 365         while (od->cursor != &od->cache->entries) {
 366                 p = list_entry(od->cursor, struct ovl_cache_entry, l_node);
 367                 if (!p->is_whiteout)
 368                         if (!dir_emit(ctx, p->name, p->len, p->ino, p->type))
 369                                 break;
 370                 od->cursor = p->l_node.next;
 371                 ctx->pos++;
 372         }
 373         return 0;
 374 }
 375
 376 static loff_t ovl_dir_llseek(struct file *file, loff_t offset, int origin)
 377 {
 378         loff_t res;
 379         struct ovl_dir_file *od = file->private_data;
 380
 381         mutex_lock(&file_inode(file)->i_mutex);
 382         if (!file->f_pos)
 383                 ovl_dir_reset(file);
 384
 385         if (od->is_real) {
 386                 res = vfs_llseek(od->realfile, offset, origin);
 387                 file->f_pos = od->realfile->f_pos;
 388         } else {
 389                 res = -EINVAL;
 390
 391                 switch (origin) {
 392                 case SEEK_CUR:
 393                         offset += file->f_pos;
 394                         break;
 395                 case SEEK_SET:
 396                         break;
 397                 default:
 398                         goto out_unlock;
 399                 }
 400                 if (offset < 0)
 401                         goto out_unlock;
 402
 403                 if (offset != file->f_pos) {
 404                         file->f_pos = offset;
 405                         if (od->cache)
 406                                 ovl_seek_cursor(od, offset);
 407                 }
 408                 res = offset;
 409         }
 410 out_unlock:
 411         mutex_unlock(&file_inode(file)->i_mutex);
 412
 413         return res;
 414 }
 415
 416 static int ovl_dir_fsync(struct file *file, loff_t start, loff_t end,
 417                          int datasync)
 418 {
 419         struct ovl_dir_file *od = file->private_data;
 420         struct dentry *dentry = file->f_path.dentry;
 421         struct file *realfile = od->realfile;
 422
 423         /*
 424          * Need to check if we started out being a lower dir, but got copied up
 425          */
 426         if (!od->is_upper && OVL_TYPE_UPPER(ovl_path_type(dentry))) {
 427                 struct inode *inode = file_inode(file);
 428
 429                 realfile = lockless_dereference(od->upperfile);
 430                 if (!realfile) {
 431                         struct path upperpath;
 432
 433                         ovl_path_upper(dentry, &upperpath);
 434                         realfile = ovl_path_open(&upperpath, O_RDONLY);
 435                         smp_mb__before_spinlock();
 436                         mutex_lock(&inode->i_mutex);
 437                         if (!od->upperfile) {
 438                                 if (IS_ERR(realfile)) {
 439                                         mutex_unlock(&inode->i_mutex);
 440                                         return PTR_ERR(realfile);
 441                                 }
 442                                 od->upperfile = realfile;
 443                         } else {
 444                                 /* somebody has beaten us to it */
 445                                 if (!IS_ERR(realfile))
 446                                         fput(realfile);
 447                                 realfile = od->upperfile;
 448                         }
 449                         mutex_unlock(&inode->i_mutex);
 450                 }
 451         }
 452
 453         return vfs_fsync_range(realfile, start, end, datasync);
 454 }
 455
 456 static int ovl_dir_release(struct inode *inode, struct file *file)
 457 {
 458         struct ovl_dir_file *od = file->private_data;
 459
 460         if (od->cache) {
 461                 mutex_lock(&inode->i_mutex);
 462                 ovl_cache_put(od, file->f_path.dentry);
 463                 mutex_unlock(&inode->i_mutex);
 464         }
 465         fput(od->realfile);
 466         if (od->upperfile)
 467                 fput(od->upperfile);
 468         kfree(od);
 469
 470         return 0;
 471 }
 472
 473 static int ovl_dir_open(struct inode *inode, struct file *file)
 474 {
 475         struct path realpath;
 476         struct file *realfile;
 477         struct ovl_dir_file *od;
 478         enum ovl_path_type type;
 479
 480         od = kzalloc(sizeof(struct ovl_dir_file), GFP_KERNEL);
 481         if (!od)
 482                 return -ENOMEM;
 483
 484         type = ovl_path_real(file->f_path.dentry, &realpath);
 485         realfile = ovl_path_open(&realpath, file->f_flags);
 486         if (IS_ERR(realfile)) {
 487                 kfree(od);
 488                 return PTR_ERR(realfile);
 489         }
 490         od->realfile = realfile;
 491         od->is_real = !OVL_TYPE_MERGE(type);
 492         od->is_upper = OVL_TYPE_UPPER(type);
 493         file->private_data = od;
 494
 495         return 0;
 496 }
 497
 498 const struct file_operations ovl_dir_operations = {
 499         .read           = generic_read_dir,
 500         .open           = ovl_dir_open,
 501         .iterate        = ovl_iterate,
 502         .llseek         = ovl_dir_llseek,
 503         .fsync          = ovl_dir_fsync,
 504         .release        = ovl_dir_release,
 505 };
 506
 507 int ovl_check_empty_dir(struct dentry *dentry, struct list_head *list)
 508 {
 509         int err;
 510         struct ovl_cache_entry *p;
 511
 512         err = ovl_dir_read_merged(dentry, list);
 513         if (err)
 514                 return err;
 515
 516         err = 0;
 517
 518         list_for_each_entry(p, list, l_node) {
 519                 if (p->is_whiteout)
 520                         continue;
 521
 522                 if (p->name[0] == '.') {
 523                         if (p->len == 1)
 524                                 continue;
 525                         if (p->len == 2 && p->name[1] == '.')
 526                                 continue;
 527                 }
 528                 err = -ENOTEMPTY;
 529                 break;
 530         }
 531
 532         return err;
 533 }
 534
 535 void ovl_cleanup_whiteouts(struct dentry *upper, struct list_head *list)
 536 {
 537         struct ovl_cache_entry *p;
 538
 539         mutex_lock_nested(&upper->d_inode->i_mutex, I_MUTEX_CHILD);
 540         list_for_each_entry(p, list, l_node) {
 541                 struct dentry *dentry;
 542
 543                 if (!p->is_whiteout)
 544                         continue;
 545
 546                 dentry = lookup_one_len(p->name, upper, p->len);
 547                 if (IS_ERR(dentry)) {
 548                         pr_err("overlayfs: lookup '%s/%.*s' failed (%i)\n",
 549                                upper->d_name.name, p->len, p->name,
 550                                (int) PTR_ERR(dentry));
 551                         continue;
 552                 }
 553                 ovl_cleanup(upper->d_inode, dentry);
 554                 dput(dentry);
 555         }
 556         mutex_unlock(&upper->d_inode->i_mutex);
 557 }