fs/reiserfs/xattr.c

   1 /*
   2  * linux/fs/reiserfs/xattr.c
   3  *
   4  * Copyright (c) 2002 by Jeff Mahoney, <jeffm@suse.com>
   5  *
   6  */
   7
   8 /*
   9  * In order to implement EA/ACLs in a clean, backwards compatible manner,
  10  * they are implemented as files in a "private" directory.
  11  * Each EA is in it's own file, with the directory layout like so (/ is assumed
  12  * to be relative to fs root). Inside the /.reiserfs_priv/xattrs directory,
  13  * directories named using the capital-hex form of the objectid and
  14  * generation number are used. Inside each directory are individual files
  15  * named with the name of the extended attribute.
  16  *
  17  * So, for objectid 12648430, we could have:
  18  * /.reiserfs_priv/xattrs/C0FFEE.0/system.posix_acl_access
  19  * /.reiserfs_priv/xattrs/C0FFEE.0/system.posix_acl_default
  20  * /.reiserfs_priv/xattrs/C0FFEE.0/user.Content-Type
  21  * .. or similar.
  22  *
  23  * The file contents are the text of the EA. The size is known based on the
  24  * stat data describing the file.
  25  *
  26  * In the case of system.posix_acl_access and system.posix_acl_default, since
  27  * these are special cases for filesystem ACLs, they are interpreted by the
  28  * kernel, in addition, they are negatively and positively cached and attached
  29  * to the inode so that unnecessary lookups are avoided.
  30  */
  31
  32 #include <linux/reiserfs_fs.h>
  33 #include <linux/dcache.h>
  34 #include <linux/namei.h>
  35 #include <linux/errno.h>
  36 #include <linux/fs.h>
  37 #include <linux/file.h>
  38 #include <linux/pagemap.h>
  39 #include <linux/xattr.h>
  40 #include <linux/reiserfs_xattr.h>
  41 #include <linux/reiserfs_acl.h>
  42 #include <linux/mbcache.h>
  43 #include <asm/uaccess.h>
  44 #include <asm/checksum.h>
  45 #include <linux/smp_lock.h>
  46 #include <linux/stat.h>
  47 #include <asm/semaphore.h>
  48
  49 #define FL_READONLY 128
  50 #define FL_DIR_SEM_HELD 256
  51 #define PRIVROOT_NAME ".reiserfs_priv"
  52 #define XAROOT_NAME   "xattrs"
  53
  54 static struct reiserfs_xattr_handler *find_xattr_handler_prefix(const char
  55                                                                 *prefix);
  56
  57 static struct dentry *create_xa_root(struct super_block *sb)
  58 {
  59         struct dentry *privroot = dget(REISERFS_SB(sb)->priv_root);
  60         struct dentry *xaroot;
  61
  62         /* This needs to be created at mount-time */
  63         if (!privroot)
  64                 return ERR_PTR(-EOPNOTSUPP);
  65
  66         xaroot = lookup_one_len(XAROOT_NAME, privroot, strlen(XAROOT_NAME));
  67         if (IS_ERR(xaroot)) {
  68                 goto out;
  69         } else if (!xaroot->d_inode) {
  70                 int err;
  71                 down(&privroot->d_inode->i_sem);
  72                 err =
  73                     privroot->d_inode->i_op->mkdir(privroot->d_inode, xaroot,
  74                                                    0700);
  75                 up(&privroot->d_inode->i_sem);
  76
  77                 if (err) {
  78                         dput(xaroot);
  79                         dput(privroot);
  80                         return ERR_PTR(err);
  81                 }
  82                 REISERFS_SB(sb)->xattr_root = dget(xaroot);
  83         }
  84
  85       out:
  86         dput(privroot);
  87         return xaroot;
  88 }
  89
  90 /* This will return a dentry, or error, refering to the xa root directory.
  91  * If the xa root doesn't exist yet, the dentry will be returned without
  92  * an associated inode. This dentry can be used with ->mkdir to create
  93  * the xa directory. */
  94 static struct dentry *__get_xa_root(struct super_block *s)
  95 {
  96         struct dentry *privroot = dget(REISERFS_SB(s)->priv_root);
  97         struct dentry *xaroot = NULL;
  98
  99         if (IS_ERR(privroot) || !privroot)
 100                 return privroot;
 101
 102         xaroot = lookup_one_len(XAROOT_NAME, privroot, strlen(XAROOT_NAME));
 103         if (IS_ERR(xaroot)) {
 104                 goto out;
 105         } else if (!xaroot->d_inode) {
 106                 dput(xaroot);
 107                 xaroot = NULL;
 108                 goto out;
 109         }
 110
 111         REISERFS_SB(s)->xattr_root = dget(xaroot);
 112
 113       out:
 114         dput(privroot);
 115         return xaroot;
 116 }
 117
 118 /* Returns the dentry (or NULL) referring to the root of the extended
 119  * attribute directory tree. If it has already been retreived, it is used.
 120  * Otherwise, we attempt to retreive it from disk. It may also return
 121  * a pointer-encoded error.
 122  */
 123 static inline struct dentry *get_xa_root(struct super_block *s)
 124 {
 125         struct dentry *dentry = dget(REISERFS_SB(s)->xattr_root);
 126
 127         if (!dentry)
 128                 dentry = __get_xa_root(s);
 129
 130         return dentry;
 131 }
 132
 133 /* Opens the directory corresponding to the inode's extended attribute store.
 134  * If flags allow, the tree to the directory may be created. If creation is
 135  * prohibited, -ENODATA is returned. */
 136 static struct dentry *open_xa_dir(const struct inode *inode, int flags)
 137 {
 138         struct dentry *xaroot, *xadir;
 139         char namebuf[17];
 140
 141         xaroot = get_xa_root(inode->i_sb);
 142         if (IS_ERR(xaroot)) {
 143                 return xaroot;
 144         } else if (!xaroot) {
 145                 if (flags == 0 || flags & XATTR_CREATE) {
 146                         xaroot = create_xa_root(inode->i_sb);
 147                         if (IS_ERR(xaroot))
 148                                 return xaroot;
 149                 }
 150                 if (!xaroot)
 151                         return ERR_PTR(-ENODATA);
 152         }
 153
 154         /* ok, we have xaroot open */
 155
 156         snprintf(namebuf, sizeof(namebuf), "%X.%X",
 157                  le32_to_cpu(INODE_PKEY(inode)->k_objectid),
 158                  inode->i_generation);
 159         xadir = lookup_one_len(namebuf, xaroot, strlen(namebuf));
 160         if (IS_ERR(xadir)) {
 161                 dput(xaroot);
 162                 return xadir;
 163         }
 164
 165         if (!xadir->d_inode) {
 166                 int err;
 167                 if (flags == 0 || flags & XATTR_CREATE) {
 168                         /* Although there is nothing else trying to create this directory,
 169                          * another directory with the same hash may be created, so we need
 170                          * to protect against that */
 171                         err =
 172                             xaroot->d_inode->i_op->mkdir(xaroot->d_inode, xadir,
 173                                                          0700);
 174                         if (err) {
 175                                 dput(xaroot);
 176                                 dput(xadir);
 177                                 return ERR_PTR(err);
 178                         }
 179                 }
 180                 if (!xadir->d_inode) {
 181                         dput(xaroot);
 182                         dput(xadir);
 183                         return ERR_PTR(-ENODATA);
 184                 }
 185         }
 186
 187         dput(xaroot);
 188         return xadir;
 189 }
 190
 191 /* Returns a dentry corresponding to a specific extended attribute file
 192  * for the inode. If flags allow, the file is created. Otherwise, a
 193  * valid or negative dentry, or an error is returned. */
 194 static struct dentry *get_xa_file_dentry(const struct inode *inode,
 195                                          const char *name, int flags)
 196 {
 197         struct dentry *xadir, *xafile;
 198         int err = 0;
 199
 200         xadir = open_xa_dir(inode, flags);
 201         if (IS_ERR(xadir)) {
 202                 return ERR_PTR(PTR_ERR(xadir));
 203         } else if (xadir && !xadir->d_inode) {
 204                 dput(xadir);
 205                 return ERR_PTR(-ENODATA);
 206         }
 207
 208         xafile = lookup_one_len(name, xadir, strlen(name));
 209         if (IS_ERR(xafile)) {
 210                 dput(xadir);
 211                 return ERR_PTR(PTR_ERR(xafile));
 212         }
 213
 214         if (xafile->d_inode) {  /* file exists */
 215                 if (flags & XATTR_CREATE) {
 216                         err = -EEXIST;
 217                         dput(xafile);
 218                         goto out;
 219                 }
 220         } else if (flags & XATTR_REPLACE || flags & FL_READONLY) {
 221                 goto out;
 222         } else {
 223                 /* inode->i_sem is down, so nothing else can try to create
 224                  * the same xattr */
 225                 err = xadir->d_inode->i_op->create(xadir->d_inode, xafile,
 226                                                    0700 | S_IFREG, NULL);
 227
 228                 if (err) {
 229                         dput(xafile);
 230                         goto out;
 231                 }
 232         }
 233
 234       out:
 235         dput(xadir);
 236         if (err)
 237                 xafile = ERR_PTR(err);
 238         return xafile;
 239 }
 240
 241 /* Opens a file pointer to the attribute associated with inode */
 242 static struct file *open_xa_file(const struct inode *inode, const char *name,
 243                                  int flags)
 244 {
 245         struct dentry *xafile;
 246         struct file *fp;
 247
 248         xafile = get_xa_file_dentry(inode, name, flags);
 249         if (IS_ERR(xafile))
 250                 return ERR_PTR(PTR_ERR(xafile));
 251         else if (!xafile->d_inode) {
 252                 dput(xafile);
 253                 return ERR_PTR(-ENODATA);
 254         }
 255
 256         fp = dentry_open(xafile, NULL, O_RDWR);
 257         /* dentry_open dputs the dentry if it fails */
 258
 259         return fp;
 260 }
 261
 262 /*
 263  * this is very similar to fs/reiserfs/dir.c:reiserfs_readdir, but
 264  * we need to drop the path before calling the filldir struct.  That
 265  * would be a big performance hit to the non-xattr case, so I've copied
 266  * the whole thing for now. --clm
 267  *
 268  * the big difference is that I go backwards through the directory,
 269  * and don't mess with f->f_pos, but the idea is the same.  Do some
 270  * action on each and every entry in the directory.
 271  *
 272  * we're called with i_sem held, so there are no worries about the directory
 273  * changing underneath us.
 274  */
 275 static int __xattr_readdir(struct file *filp, void *dirent, filldir_t filldir)
 276 {
 277         struct inode *inode = filp->f_dentry->d_inode;
 278         struct cpu_key pos_key; /* key of current position in the directory (key of directory entry) */
 279         INITIALIZE_PATH(path_to_entry);
 280         struct buffer_head *bh;
 281         int entry_num;
 282         struct item_head *ih, tmp_ih;
 283         int search_res;
 284         char *local_buf;
 285         loff_t next_pos;
 286         char small_buf[32];     /* avoid kmalloc if we can */
 287         struct reiserfs_de_head *deh;
 288         int d_reclen;
 289         char *d_name;
 290         off_t d_off;
 291         ino_t d_ino;
 292         struct reiserfs_dir_entry de;
 293
 294         /* form key for search the next directory entry using f_pos field of
 295            file structure */
 296         next_pos = max_reiserfs_offset(inode);
 297
 298         while (1) {
 299               research:
 300                 if (next_pos <= DOT_DOT_OFFSET)
 301                         break;
 302                 make_cpu_key(&pos_key, inode, next_pos, TYPE_DIRENTRY, 3);
 303
 304                 search_res =
 305                     search_by_entry_key(inode->i_sb, &pos_key, &path_to_entry,
 306                                         &de);
 307                 if (search_res == IO_ERROR) {
 308                         // FIXME: we could just skip part of directory which could
 309                         // not be read
 310                         pathrelse(&path_to_entry);
 311                         return -EIO;
 312                 }
 313
 314                 if (search_res == NAME_NOT_FOUND)
 315                         de.de_entry_num--;
 316
 317                 set_de_name_and_namelen(&de);
 318                 entry_num = de.de_entry_num;
 319                 deh = &(de.de_deh[entry_num]);
 320
 321                 bh = de.de_bh;
 322                 ih = de.de_ih;
 323
 324                 if (!is_direntry_le_ih(ih)) {
 325                         reiserfs_warning(inode->i_sb, "not direntry %h", ih);
 326                         break;
 327                 }
 328                 copy_item_head(&tmp_ih, ih);
 329
 330                 /* we must have found item, that is item of this directory, */
 331                 RFALSE(COMP_SHORT_KEYS(&(ih->ih_key), &pos_key),
 332                        "vs-9000: found item %h does not match to dir we readdir %K",
 333                        ih, &pos_key);
 334
 335                 if (deh_offset(deh) <= DOT_DOT_OFFSET) {
 336                         break;
 337                 }
 338
 339                 /* look for the previous entry in the directory */
 340                 next_pos = deh_offset(deh) - 1;
 341
 342                 if (!de_visible(deh))
 343                         /* it is hidden entry */
 344                         continue;
 345
 346                 d_reclen = entry_length(bh, ih, entry_num);
 347                 d_name = B_I_DEH_ENTRY_FILE_NAME(bh, ih, deh);
 348                 d_off = deh_offset(deh);
 349                 d_ino = deh_objectid(deh);
 350
 351                 if (!d_name[d_reclen - 1])
 352                         d_reclen = strlen(d_name);
 353
 354                 if (d_reclen > REISERFS_MAX_NAME(inode->i_sb->s_blocksize)) {
 355                         /* too big to send back to VFS */
 356                         continue;
 357                 }
 358
 359                 /* Ignore the .reiserfs_priv entry */
 360                 if (reiserfs_xattrs(inode->i_sb) &&
 361                     !old_format_only(inode->i_sb) &&
 362                     deh_objectid(deh) ==
 363                     le32_to_cpu(INODE_PKEY
 364                                 (REISERFS_SB(inode->i_sb)->priv_root->d_inode)->
 365                                 k_objectid))
 366                         continue;
 367
 368                 if (d_reclen <= 32) {
 369                         local_buf = small_buf;
 370                 } else {
 371                         local_buf =
 372                             reiserfs_kmalloc(d_reclen, GFP_NOFS, inode->i_sb);
 373                         if (!local_buf) {
 374                                 pathrelse(&path_to_entry);
 375                                 return -ENOMEM;
 376                         }
 377                         if (item_moved(&tmp_ih, &path_to_entry)) {
 378                                 reiserfs_kfree(local_buf, d_reclen,
 379                                                inode->i_sb);
 380
 381                                 /* sigh, must retry.  Do this same offset again */
 382                                 next_pos = d_off;
 383                                 goto research;
 384                         }
 385                 }
 386
 387                 // Note, that we copy name to user space via temporary
 388                 // buffer (local_buf) because filldir will block if
 389                 // user space buffer is swapped out. At that time
 390                 // entry can move to somewhere else
 391                 memcpy(local_buf, d_name, d_reclen);
 392
 393                 /* the filldir function might need to start transactions,
 394                  * or do who knows what.  Release the path now that we've
 395                  * copied all the important stuff out of the deh
 396                  */
 397                 pathrelse(&path_to_entry);
 398
 399                 if (filldir(dirent, local_buf, d_reclen, d_off, d_ino,
 400                             DT_UNKNOWN) < 0) {
 401                         if (local_buf != small_buf) {
 402                                 reiserfs_kfree(local_buf, d_reclen,
 403                                                inode->i_sb);
 404                         }
 405                         goto end;
 406                 }
 407                 if (local_buf != small_buf) {
 408                         reiserfs_kfree(local_buf, d_reclen, inode->i_sb);
 409                 }
 410         }                       /* while */
 411
 412       end:
 413         pathrelse(&path_to_entry);
 414         return 0;
 415 }
 416
 417 /*
 418  * this could be done with dedicated readdir ops for the xattr files,
 419  * but I want to get something working asap
 420  * this is stolen from vfs_readdir
 421  *
 422  */
 423 static
 424 int xattr_readdir(struct file *file, filldir_t filler, void *buf)
 425 {
 426         struct inode *inode = file->f_dentry->d_inode;
 427         int res = -ENOTDIR;
 428         if (!file->f_op || !file->f_op->readdir)
 429                 goto out;
 430         down(&inode->i_sem);
 431 //        down(&inode->i_zombie);
 432         res = -ENOENT;
 433         if (!IS_DEADDIR(inode)) {
 434                 lock_kernel();
 435                 res = __xattr_readdir(file, buf, filler);
 436                 unlock_kernel();
 437         }
 438 //        up(&inode->i_zombie);
 439         up(&inode->i_sem);
 440       out:
 441         return res;
 442 }
 443
 444 /* Internal operations on file data */
 445 static inline void reiserfs_put_page(struct page *page)
 446 {
 447         kunmap(page);
 448         page_cache_release(page);
 449 }
 450
 451 static struct page *reiserfs_get_page(struct inode *dir, unsigned long n)
 452 {
 453         struct address_space *mapping = dir->i_mapping;
 454         struct page *page;
 455         /* We can deadlock if we try to free dentries,
 456            and an unlink/rmdir has just occured - GFP_NOFS avoids this */
 457         mapping->flags = (mapping->flags & ~__GFP_BITS_MASK) | GFP_NOFS;
 458         page = read_cache_page(mapping, n,
 459                                (filler_t *) mapping->a_ops->readpage, NULL);
 460         if (!IS_ERR(page)) {
 461                 wait_on_page_locked(page);
 462                 kmap(page);
 463                 if (!PageUptodate(page))
 464                         goto fail;
 465
 466                 if (PageError(page))
 467                         goto fail;
 468         }
 469         return page;
 470
 471       fail:
 472         reiserfs_put_page(page);
 473         return ERR_PTR(-EIO);
 474 }
 475
 476 static inline __u32 xattr_hash(const char *msg, int len)
 477 {
 478         return csum_partial(msg, len, 0);
 479 }
 480
 481 /* Generic extended attribute operations that can be used by xa plugins */
 482
 483 /*
 484  * inode->i_sem: down
 485  */
 486 int
 487 reiserfs_xattr_set(struct inode *inode, const char *name, const void *buffer,
 488                    size_t buffer_size, int flags)
 489 {
 490         int err = 0;
 491         struct file *fp;
 492         struct page *page;
 493         char *data;
 494         struct address_space *mapping;
 495         size_t file_pos = 0;
 496         size_t buffer_pos = 0;
 497         struct inode *xinode;
 498         struct iattr newattrs;
 499         __u32 xahash = 0;
 500
 501         if (IS_RDONLY(inode))
 502                 return -EROFS;
 503
 504         if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
 505                 return -EPERM;
 506
 507         if (get_inode_sd_version(inode) == STAT_DATA_V1)
 508                 return -EOPNOTSUPP;
 509
 510         /* Empty xattrs are ok, they're just empty files, no hash */
 511         if (buffer && buffer_size)
 512                 xahash = xattr_hash(buffer, buffer_size);
 513
 514       open_file:
 515         fp = open_xa_file(inode, name, flags);
 516         if (IS_ERR(fp)) {
 517                 err = PTR_ERR(fp);
 518                 goto out;
 519         }
 520
 521         xinode = fp->f_dentry->d_inode;
 522         REISERFS_I(inode)->i_flags |= i_has_xattr_dir;
 523
 524         /* we need to copy it off.. */
 525         if (xinode->i_nlink > 1) {
 526                 fput(fp);
 527                 err = reiserfs_xattr_del(inode, name);
 528                 if (err < 0)
 529                         goto out;
 530                 /* We just killed the old one, we're not replacing anymore */
 531                 if (flags & XATTR_REPLACE)
 532                         flags &= ~XATTR_REPLACE;
 533                 goto open_file;
 534         }
 535
 536         /* Resize it so we're ok to write there */
 537         newattrs.ia_size = buffer_size;
 538         newattrs.ia_valid = ATTR_SIZE | ATTR_CTIME;
 539         down(&xinode->i_sem);
 540         err = notify_change(fp->f_dentry, &newattrs);
 541         if (err)
 542                 goto out_filp;
 543
 544         mapping = xinode->i_mapping;
 545         while (buffer_pos < buffer_size || buffer_pos == 0) {
 546                 size_t chunk;
 547                 size_t skip = 0;
 548                 size_t page_offset = (file_pos & (PAGE_CACHE_SIZE - 1));
 549                 if (buffer_size - buffer_pos > PAGE_CACHE_SIZE)
 550                         chunk = PAGE_CACHE_SIZE;
 551                 else
 552                         chunk = buffer_size - buffer_pos;
 553
 554                 page = reiserfs_get_page(xinode, file_pos >> PAGE_CACHE_SHIFT);
 555                 if (IS_ERR(page)) {
 556                         err = PTR_ERR(page);
 557                         goto out_filp;
 558                 }
 559
 560                 lock_page(page);
 561                 data = page_address(page);
 562
 563                 if (file_pos == 0) {
 564                         struct reiserfs_xattr_header *rxh;
 565                         skip = file_pos = sizeof(struct reiserfs_xattr_header);
 566                         if (chunk + skip > PAGE_CACHE_SIZE)
 567                                 chunk = PAGE_CACHE_SIZE - skip;
 568                         rxh = (struct reiserfs_xattr_header *)data;
 569                         rxh->h_magic = cpu_to_le32(REISERFS_XATTR_MAGIC);
 570                         rxh->h_hash = cpu_to_le32(xahash);
 571                 }
 572
 573                 err = mapping->a_ops->prepare_write(fp, page, page_offset,
 574                                                     page_offset + chunk + skip);
 575                 if (!err) {
 576                         if (buffer)
 577                                 memcpy(data + skip, buffer + buffer_pos, chunk);
 578                         err =
 579                             mapping->a_ops->commit_write(fp, page, page_offset,
 580                                                          page_offset + chunk +
 581                                                          skip);
 582                 }
 583                 unlock_page(page);
 584                 reiserfs_put_page(page);
 585                 buffer_pos += chunk;
 586                 file_pos += chunk;
 587                 skip = 0;
 588                 if (err || buffer_size == 0 || !buffer)
 589                         break;
 590         }
 591
 592         /* We can't mark the inode dirty if it's not hashed. This is the case
 593          * when we're inheriting the default ACL. If we dirty it, the inode
 594          * gets marked dirty, but won't (ever) make it onto the dirty list until
 595          * it's synced explicitly to clear I_DIRTY. This is bad. */
 596         if (!hlist_unhashed(&inode->i_hash)) {
 597                 inode->i_ctime = CURRENT_TIME_SEC;
 598                 mark_inode_dirty(inode);
 599         }
 600
 601       out_filp:
 602         up(&xinode->i_sem);
 603         fput(fp);
 604
 605       out:
 606         return err;
 607 }
 608
 609 /*
 610  * inode->i_sem: down
 611  */
 612 int
 613 reiserfs_xattr_get(const struct inode *inode, const char *name, void *buffer,
 614                    size_t buffer_size)
 615 {
 616         ssize_t err = 0;
 617         struct file *fp;
 618         size_t isize;
 619         size_t file_pos = 0;
 620         size_t buffer_pos = 0;
 621         struct page *page;
 622         struct inode *xinode;
 623         __u32 hash = 0;
 624
 625         if (name == NULL)
 626                 return -EINVAL;
 627
 628         /* We can't have xattrs attached to v1 items since they don't have
 629          * generation numbers */
 630         if (get_inode_sd_version(inode) == STAT_DATA_V1)
 631                 return -EOPNOTSUPP;
 632
 633         fp = open_xa_file(inode, name, FL_READONLY);
 634         if (IS_ERR(fp)) {
 635                 err = PTR_ERR(fp);
 636                 goto out;
 637         }
 638
 639         xinode = fp->f_dentry->d_inode;
 640         isize = xinode->i_size;
 641         REISERFS_I(inode)->i_flags |= i_has_xattr_dir;
 642
 643         /* Just return the size needed */
 644         if (buffer == NULL) {
 645                 err = isize - sizeof(struct reiserfs_xattr_header);
 646                 goto out_dput;
 647         }
 648
 649         if (buffer_size < isize - sizeof(struct reiserfs_xattr_header)) {
 650                 err = -ERANGE;
 651                 goto out_dput;
 652         }
 653
 654         while (file_pos < isize) {
 655                 size_t chunk;
 656                 char *data;
 657                 size_t skip = 0;
 658                 if (isize - file_pos > PAGE_CACHE_SIZE)
 659                         chunk = PAGE_CACHE_SIZE;
 660                 else
 661                         chunk = isize - file_pos;
 662
 663                 page = reiserfs_get_page(xinode, file_pos >> PAGE_CACHE_SHIFT);
 664                 if (IS_ERR(page)) {
 665                         err = PTR_ERR(page);
 666                         goto out_dput;
 667                 }
 668
 669                 lock_page(page);
 670                 data = page_address(page);
 671                 if (file_pos == 0) {
 672                         struct reiserfs_xattr_header *rxh =
 673                             (struct reiserfs_xattr_header *)data;
 674                         skip = file_pos = sizeof(struct reiserfs_xattr_header);
 675                         chunk -= skip;
 676                         /* Magic doesn't match up.. */
 677                         if (rxh->h_magic != cpu_to_le32(REISERFS_XATTR_MAGIC)) {
 678                                 unlock_page(page);
 679                                 reiserfs_put_page(page);
 680                                 reiserfs_warning(inode->i_sb,
 681                                                  "Invalid magic for xattr (%s) "
 682                                                  "associated with %k", name,
 683                                                  INODE_PKEY(inode));
 684                                 err = -EIO;
 685                                 goto out_dput;
 686                         }
 687                         hash = le32_to_cpu(rxh->h_hash);
 688                 }
 689                 memcpy(buffer + buffer_pos, data + skip, chunk);
 690                 unlock_page(page);
 691                 reiserfs_put_page(page);
 692                 file_pos += chunk;
 693                 buffer_pos += chunk;
 694                 skip = 0;
 695         }
 696         err = isize - sizeof(struct reiserfs_xattr_header);
 697
 698         if (xattr_hash(buffer, isize - sizeof(struct reiserfs_xattr_header)) !=
 699             hash) {
 700                 reiserfs_warning(inode->i_sb,
 701                                  "Invalid hash for xattr (%s) associated "
 702                                  "with %k", name, INODE_PKEY(inode));
 703                 err = -EIO;
 704         }
 705
 706       out_dput:
 707         fput(fp);
 708
 709       out:
 710         return err;
 711 }
 712
 713 static int
 714 __reiserfs_xattr_del(struct dentry *xadir, const char *name, int namelen)
 715 {
 716         struct dentry *dentry;
 717         struct inode *dir = xadir->d_inode;
 718         int err = 0;
 719
 720         dentry = lookup_one_len(name, xadir, namelen);
 721         if (IS_ERR(dentry)) {
 722                 err = PTR_ERR(dentry);
 723                 goto out;
 724         } else if (!dentry->d_inode) {
 725                 err = -ENODATA;
 726                 goto out_file;
 727         }
 728
 729         /* Skip directories.. */
 730         if (S_ISDIR(dentry->d_inode->i_mode))
 731                 goto out_file;
 732
 733         if (!is_reiserfs_priv_object(dentry->d_inode)) {
 734                 reiserfs_warning(dir->i_sb, "OID %08x [%.*s/%.*s] doesn't have "
 735                                  "priv flag set [parent is %sset].",
 736                                  le32_to_cpu(INODE_PKEY(dentry->d_inode)->
 737                                              k_objectid), xadir->d_name.len,
 738                                  xadir->d_name.name, namelen, name,
 739                                  is_reiserfs_priv_object(xadir->
 740                                                          d_inode) ? "" :
 741                                  "not ");
 742                 dput(dentry);
 743                 return -EIO;
 744         }
 745
 746         err = dir->i_op->unlink(dir, dentry);
 747         if (!err)
 748                 d_delete(dentry);
 749
 750       out_file:
 751         dput(dentry);
 752
 753       out:
 754         return err;
 755 }
 756
 757 int reiserfs_xattr_del(struct inode *inode, const char *name)
 758 {
 759         struct dentry *dir;
 760         int err;
 761
 762         if (IS_RDONLY(inode))
 763                 return -EROFS;
 764
 765         dir = open_xa_dir(inode, FL_READONLY);
 766         if (IS_ERR(dir)) {
 767                 err = PTR_ERR(dir);
 768                 goto out;
 769         }
 770
 771         err = __reiserfs_xattr_del(dir, name, strlen(name));
 772         dput(dir);
 773
 774         if (!err) {
 775                 inode->i_ctime = CURRENT_TIME_SEC;
 776                 mark_inode_dirty(inode);
 777         }
 778
 779       out:
 780         return err;
 781 }
 782
 783 /* The following are side effects of other operations that aren't explicitly
 784  * modifying extended attributes. This includes operations such as permissions
 785  * or ownership changes, object deletions, etc. */
 786
 787 static int
 788 reiserfs_delete_xattrs_filler(void *buf, const char *name, int namelen,
 789                               loff_t offset, ino_t ino, unsigned int d_type)
 790 {
 791         struct dentry *xadir = (struct dentry *)buf;
 792
 793         return __reiserfs_xattr_del(xadir, name, namelen);
 794
 795 }
 796
 797 /* This is called w/ inode->i_sem downed */
 798 int reiserfs_delete_xattrs(struct inode *inode)
 799 {
 800         struct file *fp;
 801         struct dentry *dir, *root;
 802         int err = 0;
 803
 804         /* Skip out, an xattr has no xattrs associated with it */
 805         if (is_reiserfs_priv_object(inode) ||
 806             get_inode_sd_version(inode) == STAT_DATA_V1 ||
 807             !reiserfs_xattrs(inode->i_sb)) {
 808                 return 0;
 809         }
 810         reiserfs_read_lock_xattrs(inode->i_sb);
 811         dir = open_xa_dir(inode, FL_READONLY);
 812         reiserfs_read_unlock_xattrs(inode->i_sb);
 813         if (IS_ERR(dir)) {
 814                 err = PTR_ERR(dir);
 815                 goto out;
 816         } else if (!dir->d_inode) {
 817                 dput(dir);
 818                 return 0;
 819         }
 820
 821         fp = dentry_open(dir, NULL, O_RDWR);
 822         if (IS_ERR(fp)) {
 823                 err = PTR_ERR(fp);
 824                 /* dentry_open dputs the dentry if it fails */
 825                 goto out;
 826         }
 827
 828         lock_kernel();
 829         err = xattr_readdir(fp, reiserfs_delete_xattrs_filler, dir);
 830         if (err) {
 831                 unlock_kernel();
 832                 goto out_dir;
 833         }
 834
 835         /* Leftovers besides . and .. -- that's not good. */
 836         if (dir->d_inode->i_nlink <= 2) {
 837                 root = get_xa_root(inode->i_sb);
 838                 reiserfs_write_lock_xattrs(inode->i_sb);
 839                 err = vfs_rmdir(root->d_inode, dir);
 840                 reiserfs_write_unlock_xattrs(inode->i_sb);
 841                 dput(root);
 842         } else {
 843                 reiserfs_warning(inode->i_sb,
 844                                  "Couldn't remove all entries in directory");
 845         }
 846         unlock_kernel();
 847
 848       out_dir:
 849         fput(fp);
 850
 851       out:
 852         if (!err)
 853                 REISERFS_I(inode)->i_flags =
 854                     REISERFS_I(inode)->i_flags & ~i_has_xattr_dir;
 855         return err;
 856 }
 857
 858 struct reiserfs_chown_buf {
 859         struct inode *inode;
 860         struct dentry *xadir;
 861         struct iattr *attrs;
 862 };
 863
 864 /* XXX: If there is a better way to do this, I'd love to hear about it */
 865 static int
 866 reiserfs_chown_xattrs_filler(void *buf, const char *name, int namelen,
 867                              loff_t offset, ino_t ino, unsigned int d_type)
 868 {
 869         struct reiserfs_chown_buf *chown_buf = (struct reiserfs_chown_buf *)buf;
 870         struct dentry *xafile, *xadir = chown_buf->xadir;
 871         struct iattr *attrs = chown_buf->attrs;
 872         int err = 0;
 873
 874         xafile = lookup_one_len(name, xadir, namelen);
 875         if (IS_ERR(xafile))
 876                 return PTR_ERR(xafile);
 877         else if (!xafile->d_inode) {
 878                 dput(xafile);
 879                 return -ENODATA;
 880         }
 881
 882         if (!S_ISDIR(xafile->d_inode->i_mode))
 883                 err = notify_change(xafile, attrs);
 884         dput(xafile);
 885
 886         return err;
 887 }
 888
 889 int reiserfs_chown_xattrs(struct inode *inode, struct iattr *attrs)
 890 {
 891         struct file *fp;
 892         struct dentry *dir;
 893         int err = 0;
 894         struct reiserfs_chown_buf buf;
 895         unsigned int ia_valid = attrs->ia_valid;
 896
 897         /* Skip out, an xattr has no xattrs associated with it */
 898         if (is_reiserfs_priv_object(inode) ||
 899             get_inode_sd_version(inode) == STAT_DATA_V1 ||
 900             !reiserfs_xattrs(inode->i_sb)) {
 901                 return 0;
 902         }
 903         reiserfs_read_lock_xattrs(inode->i_sb);
 904         dir = open_xa_dir(inode, FL_READONLY);
 905         reiserfs_read_unlock_xattrs(inode->i_sb);
 906         if (IS_ERR(dir)) {
 907                 if (PTR_ERR(dir) != -ENODATA)
 908                         err = PTR_ERR(dir);
 909                 goto out;
 910         } else if (!dir->d_inode) {
 911                 dput(dir);
 912                 goto out;
 913         }
 914
 915         fp = dentry_open(dir, NULL, O_RDWR);
 916         if (IS_ERR(fp)) {
 917                 err = PTR_ERR(fp);
 918                 /* dentry_open dputs the dentry if it fails */
 919                 goto out;
 920         }
 921
 922         lock_kernel();
 923
 924         attrs->ia_valid &= (ATTR_UID | ATTR_GID | ATTR_CTIME);
 925         buf.xadir = dir;
 926         buf.attrs = attrs;
 927         buf.inode = inode;
 928
 929         err = xattr_readdir(fp, reiserfs_chown_xattrs_filler, &buf);
 930         if (err) {
 931                 unlock_kernel();
 932                 goto out_dir;
 933         }
 934
 935         err = notify_change(dir, attrs);
 936         unlock_kernel();
 937
 938       out_dir:
 939         fput(fp);
 940
 941       out:
 942         attrs->ia_valid = ia_valid;
 943         return err;
 944 }
 945
 946 /* Actual operations that are exported to VFS-land */
 947
 948 /*
 949  * Inode operation getxattr()
 950  * Preliminary locking: we down dentry->d_inode->i_sem
 951  */
 952 ssize_t
 953 reiserfs_getxattr(struct dentry * dentry, const char *name, void *buffer,
 954                   size_t size)
 955 {
 956         struct reiserfs_xattr_handler *xah = find_xattr_handler_prefix(name);
 957         int err;
 958
 959         if (!xah || !reiserfs_xattrs(dentry->d_sb) ||
 960             get_inode_sd_version(dentry->d_inode) == STAT_DATA_V1)
 961                 return -EOPNOTSUPP;
 962
 963         reiserfs_read_lock_xattr_i(dentry->d_inode);
 964         reiserfs_read_lock_xattrs(dentry->d_sb);
 965         err = xah->get(dentry->d_inode, name, buffer, size);
 966         reiserfs_read_unlock_xattrs(dentry->d_sb);
 967         reiserfs_read_unlock_xattr_i(dentry->d_inode);
 968         return err;
 969 }
 970
 971 /*
 972  * Inode operation setxattr()
 973  *
 974  * dentry->d_inode->i_sem down
 975  */
 976 int
 977 reiserfs_setxattr(struct dentry *dentry, const char *name, const void *value,
 978                   size_t size, int flags)
 979 {
 980         struct reiserfs_xattr_handler *xah = find_xattr_handler_prefix(name);
 981         int err;
 982         int lock;
 983
 984         if (!xah || !reiserfs_xattrs(dentry->d_sb) ||
 985             get_inode_sd_version(dentry->d_inode) == STAT_DATA_V1)
 986                 return -EOPNOTSUPP;
 987
 988         if (IS_RDONLY(dentry->d_inode))
 989                 return -EROFS;
 990
 991         if (IS_IMMUTABLE(dentry->d_inode) || IS_APPEND(dentry->d_inode))
 992                 return -EROFS;
 993
 994         reiserfs_write_lock_xattr_i(dentry->d_inode);
 995         lock = !has_xattr_dir(dentry->d_inode);
 996         if (lock)
 997                 reiserfs_write_lock_xattrs(dentry->d_sb);
 998         else
 999                 reiserfs_read_lock_xattrs(dentry->d_sb);
1000         err = xah->set(dentry->d_inode, name, value, size, flags);
1001         if (lock)
1002                 reiserfs_write_unlock_xattrs(dentry->d_sb);
1003         else
1004                 reiserfs_read_unlock_xattrs(dentry->d_sb);
1005         reiserfs_write_unlock_xattr_i(dentry->d_inode);
1006         return err;
1007 }
1008
1009 /*
1010  * Inode operation removexattr()
1011  *
1012  * dentry->d_inode->i_sem down
1013  */
1014 int reiserfs_removexattr(struct dentry *dentry, const char *name)
1015 {
1016         int err;
1017         struct reiserfs_xattr_handler *xah = find_xattr_handler_prefix(name);
1018
1019         if (!xah || !reiserfs_xattrs(dentry->d_sb) ||
1020             get_inode_sd_version(dentry->d_inode) == STAT_DATA_V1)
1021                 return -EOPNOTSUPP;
1022
1023         if (IS_RDONLY(dentry->d_inode))
1024                 return -EROFS;
1025
1026         if (IS_IMMUTABLE(dentry->d_inode) || IS_APPEND(dentry->d_inode))
1027                 return -EPERM;
1028
1029         reiserfs_write_lock_xattr_i(dentry->d_inode);
1030         reiserfs_read_lock_xattrs(dentry->d_sb);
1031
1032         /* Deletion pre-operation */
1033         if (xah->del) {
1034                 err = xah->del(dentry->d_inode, name);
1035                 if (err)
1036                         goto out;
1037         }
1038
1039         err = reiserfs_xattr_del(dentry->d_inode, name);
1040
1041         dentry->d_inode->i_ctime = CURRENT_TIME_SEC;
1042         mark_inode_dirty(dentry->d_inode);
1043
1044       out:
1045         reiserfs_read_unlock_xattrs(dentry->d_sb);
1046         reiserfs_write_unlock_xattr_i(dentry->d_inode);
1047         return err;
1048 }
1049
1050 /* This is what filldir will use:
1051  * r_pos will always contain the amount of space required for the entire
1052  * list. If r_pos becomes larger than r_size, we need more space and we
1053  * return an error indicating this. If r_pos is less than r_size, then we've
1054  * filled the buffer successfully and we return success */
1055 struct reiserfs_listxattr_buf {
1056         int r_pos;
1057         int r_size;
1058         char *r_buf;
1059         struct inode *r_inode;
1060 };
1061
1062 static int
1063 reiserfs_listxattr_filler(void *buf, const char *name, int namelen,
1064                           loff_t offset, ino_t ino, unsigned int d_type)
1065 {
1066         struct reiserfs_listxattr_buf *b = (struct reiserfs_listxattr_buf *)buf;
1067         int len = 0;
1068         if (name[0] != '.'
1069             || (namelen != 1 && (name[1] != '.' || namelen != 2))) {
1070                 struct reiserfs_xattr_handler *xah =
1071                     find_xattr_handler_prefix(name);
1072                 if (!xah)
1073                         return 0;       /* Unsupported xattr name, skip it */
1074
1075                 /* We call ->list() twice because the operation isn't required to just
1076                  * return the name back - we want to make sure we have enough space */
1077                 len += xah->list(b->r_inode, name, namelen, NULL);
1078
1079                 if (len) {
1080                         if (b->r_pos + len + 1 <= b->r_size) {
1081                                 char *p = b->r_buf + b->r_pos;
1082                                 p += xah->list(b->r_inode, name, namelen, p);
1083                                 *p++ = '\0';
1084                         }
1085                         b->r_pos += len + 1;
1086                 }
1087         }
1088
1089         return 0;
1090 }
1091
1092 /*
1093  * Inode operation listxattr()
1094  *
1095  * Preliminary locking: we down dentry->d_inode->i_sem
1096  */
1097 ssize_t reiserfs_listxattr(struct dentry * dentry, char *buffer, size_t size)
1098 {
1099         struct file *fp;
1100         struct dentry *dir;
1101         int err = 0;
1102         struct reiserfs_listxattr_buf buf;
1103
1104         if (!dentry->d_inode)
1105                 return -EINVAL;
1106
1107         if (!reiserfs_xattrs(dentry->d_sb) ||
1108             get_inode_sd_version(dentry->d_inode) == STAT_DATA_V1)
1109                 return -EOPNOTSUPP;
1110
1111         reiserfs_read_lock_xattr_i(dentry->d_inode);
1112         reiserfs_read_lock_xattrs(dentry->d_sb);
1113         dir = open_xa_dir(dentry->d_inode, FL_READONLY);
1114         reiserfs_read_unlock_xattrs(dentry->d_sb);
1115         if (IS_ERR(dir)) {
1116                 err = PTR_ERR(dir);
1117                 if (err == -ENODATA)
1118                         err = 0;        /* Not an error if there aren't any xattrs */
1119                 goto out;
1120         }
1121
1122         fp = dentry_open(dir, NULL, O_RDWR);
1123         if (IS_ERR(fp)) {
1124                 err = PTR_ERR(fp);
1125                 /* dentry_open dputs the dentry if it fails */
1126                 goto out;
1127         }
1128
1129         buf.r_buf = buffer;
1130         buf.r_size = buffer ? size : 0;
1131         buf.r_pos = 0;
1132         buf.r_inode = dentry->d_inode;
1133
1134         REISERFS_I(dentry->d_inode)->i_flags |= i_has_xattr_dir;
1135
1136         err = xattr_readdir(fp, reiserfs_listxattr_filler, &buf);
1137         if (err)
1138                 goto out_dir;
1139
1140         if (buf.r_pos > buf.r_size && buffer != NULL)
1141                 err = -ERANGE;
1142         else
1143                 err = buf.r_pos;
1144
1145       out_dir:
1146         fput(fp);
1147
1148       out:
1149         reiserfs_read_unlock_xattr_i(dentry->d_inode);
1150         return err;
1151 }
1152
1153 /* This is the implementation for the xattr plugin infrastructure */
1154 static struct list_head xattr_handlers = LIST_HEAD_INIT(xattr_handlers);
1155 static DEFINE_RWLOCK(handler_lock);
1156
1157 static struct reiserfs_xattr_handler *find_xattr_handler_prefix(const char
1158                                                                 *prefix)
1159 {
1160         struct reiserfs_xattr_handler *xah = NULL;
1161         struct list_head *p;
1162
1163         read_lock(&handler_lock);
1164         list_for_each(p, &xattr_handlers) {
1165                 xah = list_entry(p, struct reiserfs_xattr_handler, handlers);
1166                 if (strncmp(xah->prefix, prefix, strlen(xah->prefix)) == 0)
1167                         break;
1168                 xah = NULL;
1169         }
1170
1171         read_unlock(&handler_lock);
1172         return xah;
1173 }
1174
1175 static void __unregister_handlers(void)
1176 {
1177         struct reiserfs_xattr_handler *xah;
1178         struct list_head *p, *tmp;
1179
1180         list_for_each_safe(p, tmp, &xattr_handlers) {
1181                 xah = list_entry(p, struct reiserfs_xattr_handler, handlers);
1182                 if (xah->exit)
1183                         xah->exit();
1184
1185                 list_del_init(p);
1186         }
1187         INIT_LIST_HEAD(&xattr_handlers);
1188 }
1189
1190 int __init reiserfs_xattr_register_handlers(void)
1191 {
1192         int err = 0;
1193         struct reiserfs_xattr_handler *xah;
1194         struct list_head *p;
1195
1196         write_lock(&handler_lock);
1197
1198         /* If we're already initialized, nothing to do */
1199         if (!list_empty(&xattr_handlers)) {
1200                 write_unlock(&handler_lock);
1201                 return 0;
1202         }
1203
1204         /* Add the handlers */
1205         list_add_tail(&user_handler.handlers, &xattr_handlers);
1206         list_add_tail(&trusted_handler.handlers, &xattr_handlers);
1207 #ifdef CONFIG_REISERFS_FS_SECURITY
1208         list_add_tail(&security_handler.handlers, &xattr_handlers);
1209 #endif
1210 #ifdef CONFIG_REISERFS_FS_POSIX_ACL
1211         list_add_tail(&posix_acl_access_handler.handlers, &xattr_handlers);
1212         list_add_tail(&posix_acl_default_handler.handlers, &xattr_handlers);
1213 #endif
1214
1215         /* Run initializers, if available */
1216         list_for_each(p, &xattr_handlers) {
1217                 xah = list_entry(p, struct reiserfs_xattr_handler, handlers);
1218                 if (xah->init) {
1219                         err = xah->init();
1220                         if (err) {
1221                                 list_del_init(p);
1222                                 break;
1223                         }
1224                 }
1225         }
1226
1227         /* Clean up other handlers, if any failed */
1228         if (err)
1229                 __unregister_handlers();
1230
1231         write_unlock(&handler_lock);
1232         return err;
1233 }
1234
1235 void reiserfs_xattr_unregister_handlers(void)
1236 {
1237         write_lock(&handler_lock);
1238         __unregister_handlers();
1239         write_unlock(&handler_lock);
1240 }
1241
1242 /* This will catch lookups from the fs root to .reiserfs_priv */
1243 static int
1244 xattr_lookup_poison(struct dentry *dentry, struct qstr *q1, struct qstr *name)
1245 {
1246         struct dentry *priv_root = REISERFS_SB(dentry->d_sb)->priv_root;
1247         if (name->len == priv_root->d_name.len &&
1248             name->hash == priv_root->d_name.hash &&
1249             !memcmp(name->name, priv_root->d_name.name, name->len)) {
1250                 return -ENOENT;
1251         } else if (q1->len == name->len &&
1252                    !memcmp(q1->name, name->name, name->len))
1253                 return 0;
1254         return 1;
1255 }
1256
1257 static struct dentry_operations xattr_lookup_poison_ops = {
1258         .d_compare = xattr_lookup_poison,
1259 };
1260
1261 /* We need to take a copy of the mount flags since things like
1262  * MS_RDONLY don't get set until *after* we're called.
1263  * mount_flags != mount_options */
1264 int reiserfs_xattr_init(struct super_block *s, int mount_flags)
1265 {
1266         int err = 0;
1267
1268         /* We need generation numbers to ensure that the oid mapping is correct
1269          * v3.5 filesystems don't have them. */
1270         if (!old_format_only(s)) {
1271                 set_bit(REISERFS_XATTRS, &(REISERFS_SB(s)->s_mount_opt));
1272         } else if (reiserfs_xattrs_optional(s)) {
1273                 /* Old format filesystem, but optional xattrs have been enabled
1274                  * at mount time. Error out. */
1275                 reiserfs_warning(s, "xattrs/ACLs not supported on pre v3.6 "
1276                                  "format filesystem. Failing mount.");
1277                 err = -EOPNOTSUPP;
1278                 goto error;
1279         } else {
1280                 /* Old format filesystem, but no optional xattrs have been enabled. This
1281                  * means we silently disable xattrs on the filesystem. */
1282                 clear_bit(REISERFS_XATTRS, &(REISERFS_SB(s)->s_mount_opt));
1283         }
1284
1285         /* If we don't have the privroot located yet - go find it */
1286         if (reiserfs_xattrs(s) && !REISERFS_SB(s)->priv_root) {
1287                 struct dentry *dentry;
1288                 dentry = lookup_one_len(PRIVROOT_NAME, s->s_root,
1289                                         strlen(PRIVROOT_NAME));
1290                 if (!IS_ERR(dentry)) {
1291                         if (!(mount_flags & MS_RDONLY) && !dentry->d_inode) {
1292                                 struct inode *inode = dentry->d_parent->d_inode;
1293                                 down(&inode->i_sem);
1294                                 err = inode->i_op->mkdir(inode, dentry, 0700);
1295                                 up(&inode->i_sem);
1296                                 if (err) {
1297                                         dput(dentry);
1298                                         dentry = NULL;
1299                                 }
1300
1301                                 if (dentry && dentry->d_inode)
1302                                         reiserfs_warning(s,
1303                                                          "Created %s on %s - reserved for "
1304                                                          "xattr storage.",
1305                                                          PRIVROOT_NAME,
1306                                                          reiserfs_bdevname
1307                                                          (inode->i_sb));
1308                         } else if (!dentry->d_inode) {
1309                                 dput(dentry);
1310                                 dentry = NULL;
1311                         }
1312                 } else
1313                         err = PTR_ERR(dentry);
1314
1315                 if (!err && dentry) {
1316                         s->s_root->d_op = &xattr_lookup_poison_ops;
1317                         reiserfs_mark_inode_private(dentry->d_inode);
1318                         REISERFS_SB(s)->priv_root = dentry;
1319                 } else if (!(mount_flags & MS_RDONLY)) {        /* xattrs are unavailable */
1320                         /* If we're read-only it just means that the dir hasn't been
1321                          * created. Not an error -- just no xattrs on the fs. We'll
1322                          * check again if we go read-write */
1323                         reiserfs_warning(s, "xattrs/ACLs enabled and couldn't "
1324                                          "find/create .reiserfs_priv. Failing mount.");
1325                         err = -EOPNOTSUPP;
1326                 }
1327         }
1328
1329       error:
1330         /* This is only nonzero if there was an error initializing the xattr
1331          * directory or if there is a condition where we don't support them. */
1332         if (err) {
1333                 clear_bit(REISERFS_XATTRS, &(REISERFS_SB(s)->s_mount_opt));
1334                 clear_bit(REISERFS_XATTRS_USER, &(REISERFS_SB(s)->s_mount_opt));
1335                 clear_bit(REISERFS_POSIXACL, &(REISERFS_SB(s)->s_mount_opt));
1336         }
1337
1338         /* The super_block MS_POSIXACL must mirror the (no)acl mount option. */
1339         s->s_flags = s->s_flags & ~MS_POSIXACL;
1340         if (reiserfs_posixacl(s))
1341                 s->s_flags |= MS_POSIXACL;
1342
1343         return err;
1344 }
1345
1346 static int
1347 __reiserfs_permission(struct inode *inode, int mask, struct nameidata *nd,
1348                       int need_lock)
1349 {
1350         umode_t mode = inode->i_mode;
1351
1352         if (mask & MAY_WRITE) {
1353                 /*
1354                  * Nobody gets write access to a read-only fs.
1355                  */
1356                 if (IS_RDONLY(inode) &&
1357                     (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode)))
1358                         return -EROFS;
1359
1360                 /*
1361                  * Nobody gets write access to an immutable file.
1362                  */
1363                 if (IS_IMMUTABLE(inode))
1364                         return -EACCES;
1365         }
1366
1367         /* We don't do permission checks on the internal objects.
1368          * Permissions are determined by the "owning" object. */
1369         if (is_reiserfs_priv_object(inode))
1370                 return 0;
1371
1372         if (current->fsuid == inode->i_uid) {
1373                 mode >>= 6;
1374 #ifdef CONFIG_REISERFS_FS_POSIX_ACL
1375         } else if (reiserfs_posixacl(inode->i_sb) &&
1376                    get_inode_sd_version(inode) != STAT_DATA_V1) {
1377                 struct posix_acl *acl;
1378
1379                 /* ACL can't contain additional permissions if
1380                    the ACL_MASK entry is 0 */
1381                 if (!(mode & S_IRWXG))
1382                         goto check_groups;
1383
1384                 if (need_lock) {
1385                         reiserfs_read_lock_xattr_i(inode);
1386                         reiserfs_read_lock_xattrs(inode->i_sb);
1387                 }
1388                 acl = reiserfs_get_acl(inode, ACL_TYPE_ACCESS);
1389                 if (need_lock) {
1390                         reiserfs_read_unlock_xattrs(inode->i_sb);
1391                         reiserfs_read_unlock_xattr_i(inode);
1392                 }
1393                 if (IS_ERR(acl)) {
1394                         if (PTR_ERR(acl) == -ENODATA)
1395                                 goto check_groups;
1396                         return PTR_ERR(acl);
1397                 }
1398
1399                 if (acl) {
1400                         int err = posix_acl_permission(inode, acl, mask);
1401                         posix_acl_release(acl);
1402                         if (err == -EACCES) {
1403                                 goto check_capabilities;
1404                         }
1405                         return err;
1406                 } else {
1407                         goto check_groups;
1408                 }
1409 #endif
1410         } else {
1411               check_groups:
1412                 if (in_group_p(inode->i_gid))
1413                         mode >>= 3;
1414         }
1415
1416         /*
1417          * If the DACs are ok we don't need any capability check.
1418          */
1419         if (((mode & mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) == mask))
1420                 return 0;
1421
1422       check_capabilities:
1423         /*
1424          * Read/write DACs are always overridable.
1425          * Executable DACs are overridable if at least one exec bit is set.
1426          */
1427         if (!(mask & MAY_EXEC) ||
1428             (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode))
1429                 if (capable(CAP_DAC_OVERRIDE))
1430                         return 0;
1431
1432         /*
1433          * Searching includes executable on directories, else just read.
1434          */
1435         if (mask == MAY_READ || (S_ISDIR(inode->i_mode) && !(mask & MAY_WRITE)))
1436                 if (capable(CAP_DAC_READ_SEARCH))
1437                         return 0;
1438
1439         return -EACCES;
1440 }
1441
1442 int reiserfs_permission(struct inode *inode, int mask, struct nameidata *nd)
1443 {
1444         return __reiserfs_permission(inode, mask, nd, 1);
1445 }
1446
1447 int
1448 reiserfs_permission_locked(struct inode *inode, int mask, struct nameidata *nd)
1449 {
1450         return __reiserfs_permission(inode, mask, nd, 0);
1451 }