fs/xattr.c

   1 /*
   2   File: fs/xattr.c
   3
   4   Extended attribute handling.
   5
   6   Copyright (C) 2001 by Andreas Gruenbacher <a.gruenbacher@computer.org>
   7   Copyright (C) 2001 SGI - Silicon Graphics, Inc <linux-xfs@oss.sgi.com>
   8   Copyright (c) 2004 Red Hat, Inc., James Morris <jmorris@redhat.com>
   9  */
  10 #include <linux/fs.h>
  11 #include <linux/slab.h>
  12 #include <linux/file.h>
  13 #include <linux/xattr.h>
  14 #include <linux/mount.h>
  15 #include <linux/namei.h>
  16 #include <linux/security.h>
  17 #include <linux/evm.h>
  18 #include <linux/syscalls.h>
  19 #include <linux/export.h>
  20 #include <linux/fsnotify.h>
  21 #include <linux/audit.h>
  22 #include <linux/vmalloc.h>
  23 #include <linux/posix_acl_xattr.h>
  24
  25 #include <asm/uaccess.h>
  26
  27 static const char *
  28 strcmp_prefix(const char *a, const char *a_prefix)
  29 {
  30         while (*a_prefix && *a == *a_prefix) {
  31                 a++;
  32                 a_prefix++;
  33         }
  34         return *a_prefix ? NULL : a;
  35 }
  36
  37 /*
  38  * In order to implement different sets of xattr operations for each xattr
  39  * prefix, a filesystem should create a null-terminated array of struct
  40  * xattr_handler (one for each prefix) and hang a pointer to it off of the
  41  * s_xattr field of the superblock.
  42  */
  43 #define for_each_xattr_handler(handlers, handler)               \
  44         if (handlers)                                           \
  45                 for ((handler) = *(handlers)++;                 \
  46                         (handler) != NULL;                      \
  47                         (handler) = *(handlers)++)
  48
  49 /*
  50  * Find the xattr_handler with the matching prefix.
  51  */
  52 static const struct xattr_handler *
  53 xattr_resolve_name(struct inode *inode, const char **name)
  54 {
  55         const struct xattr_handler **handlers = inode->i_sb->s_xattr;
  56         const struct xattr_handler *handler;
  57
  58         if (!(inode->i_opflags & IOP_XATTR)) {
  59                 if (unlikely(is_bad_inode(inode)))
  60                         return ERR_PTR(-EIO);
  61                 return ERR_PTR(-EOPNOTSUPP);
  62         }
  63         for_each_xattr_handler(handlers, handler) {
  64                 const char *n;
  65
  66                 n = strcmp_prefix(*name, xattr_prefix(handler));
  67                 if (n) {
  68                         if (!handler->prefix ^ !*n) {
  69                                 if (*n)
  70                                         continue;
  71                                 return ERR_PTR(-EINVAL);
  72                         }
  73                         *name = n;
  74                         return handler;
  75                 }
  76         }
  77         return ERR_PTR(-EOPNOTSUPP);
  78 }
  79
  80 /*
  81  * Check permissions for extended attribute access.  This is a bit complicated
  82  * because different namespaces have very different rules.
  83  */
  84 static int
  85 xattr_permission(struct inode *inode, const char *name, int mask)
  86 {
  87         /*
  88          * We can never set or remove an extended attribute on a read-only
  89          * filesystem  or on an immutable / append-only inode.
  90          */
  91         if (mask & MAY_WRITE) {
  92                 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
  93                         return -EPERM;
  94                 /*
  95                  * Updating an xattr will likely cause i_uid and i_gid
  96                  * to be writen back improperly if their true value is
  97                  * unknown to the vfs.
  98                  */
  99                 if (HAS_UNMAPPED_ID(inode))
 100                         return -EPERM;
 101         }
 102
 103         /*
 104          * No restriction for security.* and system.* from the VFS.  Decision
 105          * on these is left to the underlying filesystem / security module.
 106          */
 107         if (!strncmp(name, XATTR_SECURITY_PREFIX, XATTR_SECURITY_PREFIX_LEN) ||
 108             !strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN))
 109                 return 0;
 110
 111         /*
 112          * The trusted.* namespace can only be accessed by privileged users.
 113          */
 114         if (!strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN)) {
 115                 if (!capable(CAP_SYS_ADMIN))
 116                         return (mask & MAY_WRITE) ? -EPERM : -ENODATA;
 117                 return 0;
 118         }
 119
 120         /*
 121          * In the user.* namespace, only regular files and directories can have
 122          * extended attributes. For sticky directories, only the owner and
 123          * privileged users can write attributes.
 124          */
 125         if (!strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN)) {
 126                 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
 127                         return (mask & MAY_WRITE) ? -EPERM : -ENODATA;
 128                 if (S_ISDIR(inode->i_mode) && (inode->i_mode & S_ISVTX) &&
 129                     (mask & MAY_WRITE) && !inode_owner_or_capable(inode))
 130                         return -EPERM;
 131         }
 132
 133         return inode_permission(inode, mask);
 134 }
 135
 136 int
 137 __vfs_setxattr(struct dentry *dentry, struct inode *inode, const char *name,
 138                const void *value, size_t size, int flags)
 139 {
 140         const struct xattr_handler *handler;
 141
 142         handler = xattr_resolve_name(inode, &name);
 143         if (IS_ERR(handler))
 144                 return PTR_ERR(handler);
 145         if (!handler->set)
 146                 return -EOPNOTSUPP;
 147         if (size == 0)
 148                 value = "";  /* empty EA, do not remove */
 149         return handler->set(handler, dentry, inode, name, value, size, flags);
 150 }
 151 EXPORT_SYMBOL(__vfs_setxattr);
 152
 153 /**
 154  *  __vfs_setxattr_noperm - perform setxattr operation without performing
 155  *  permission checks.
 156  *
 157  *  @dentry - object to perform setxattr on
 158  *  @name - xattr name to set
 159  *  @value - value to set @name to
 160  *  @size - size of @value
 161  *  @flags - flags to pass into filesystem operations
 162  *
 163  *  returns the result of the internal setxattr or setsecurity operations.
 164  *
 165  *  This function requires the caller to lock the inode's i_mutex before it
 166  *  is executed. It also assumes that the caller will make the appropriate
 167  *  permission checks.
 168  */
 169 int __vfs_setxattr_noperm(struct dentry *dentry, const char *name,
 170                 const void *value, size_t size, int flags)
 171 {
 172         struct inode *inode = dentry->d_inode;
 173         int error = -EOPNOTSUPP;
 174         int issec = !strncmp(name, XATTR_SECURITY_PREFIX,
 175                                    XATTR_SECURITY_PREFIX_LEN);
 176
 177         if (issec)
 178                 inode->i_flags &= ~S_NOSEC;
 179         if (inode->i_opflags & IOP_XATTR) {
 180                 error = __vfs_setxattr(dentry, inode, name, value, size, flags);
 181                 if (!error) {
 182                         fsnotify_xattr(dentry);
 183                         security_inode_post_setxattr(dentry, name, value,
 184                                                      size, flags);
 185                 }
 186         } else if (issec) {
 187                 const char *suffix = name + XATTR_SECURITY_PREFIX_LEN;
 188
 189                 if (unlikely(is_bad_inode(inode)))
 190                         return -EIO;
 191                 error = security_inode_setsecurity(inode, suffix, value,
 192                                                    size, flags);
 193                 if (!error)
 194                         fsnotify_xattr(dentry);
 195         }
 196
 197         return error;
 198 }
 199
 200
 201 int
 202 vfs_setxattr(struct dentry *dentry, const char *name, const void *value,
 203                 size_t size, int flags)
 204 {
 205         struct inode *inode = dentry->d_inode;
 206         int error;
 207
 208         error = xattr_permission(inode, name, MAY_WRITE);
 209         if (error)
 210                 return error;
 211
 212         inode_lock(inode);
 213         error = security_inode_setxattr(dentry, name, value, size, flags);
 214         if (error)
 215                 goto out;
 216
 217         error = __vfs_setxattr_noperm(dentry, name, value, size, flags);
 218
 219 out:
 220         inode_unlock(inode);
 221         return error;
 222 }
 223 EXPORT_SYMBOL_GPL(vfs_setxattr);
 224
 225 ssize_t
 226 xattr_getsecurity(struct inode *inode, const char *name, void *value,
 227                         size_t size)
 228 {
 229         void *buffer = NULL;
 230         ssize_t len;
 231
 232         if (!value || !size) {
 233                 len = security_inode_getsecurity(inode, name, &buffer, false);
 234                 goto out_noalloc;
 235         }
 236
 237         len = security_inode_getsecurity(inode, name, &buffer, true);
 238         if (len < 0)
 239                 return len;
 240         if (size < len) {
 241                 len = -ERANGE;
 242                 goto out;
 243         }
 244         memcpy(value, buffer, len);
 245 out:
 246         security_release_secctx(buffer, len);
 247 out_noalloc:
 248         return len;
 249 }
 250 EXPORT_SYMBOL_GPL(xattr_getsecurity);
 251
 252 /*
 253  * vfs_getxattr_alloc - allocate memory, if necessary, before calling getxattr
 254  *
 255  * Allocate memory, if not already allocated, or re-allocate correct size,
 256  * before retrieving the extended attribute.
 257  *
 258  * Returns the result of alloc, if failed, or the getxattr operation.
 259  */
 260 ssize_t
 261 vfs_getxattr_alloc(struct dentry *dentry, const char *name, char **xattr_value,
 262                    size_t xattr_size, gfp_t flags)
 263 {
 264         const struct xattr_handler *handler;
 265         struct inode *inode = dentry->d_inode;
 266         char *value = *xattr_value;
 267         int error;
 268
 269         error = xattr_permission(inode, name, MAY_READ);
 270         if (error)
 271                 return error;
 272
 273         handler = xattr_resolve_name(inode, &name);
 274         if (IS_ERR(handler))
 275                 return PTR_ERR(handler);
 276         if (!handler->get)
 277                 return -EOPNOTSUPP;
 278         error = handler->get(handler, dentry, inode, name, NULL, 0);
 279         if (error < 0)
 280                 return error;
 281
 282         if (!value || (error > xattr_size)) {
 283                 value = krealloc(*xattr_value, error + 1, flags);
 284                 if (!value)
 285                         return -ENOMEM;
 286                 memset(value, 0, error + 1);
 287         }
 288
 289         error = handler->get(handler, dentry, inode, name, value, error);
 290         *xattr_value = value;
 291         return error;
 292 }
 293
 294 ssize_t
 295 __vfs_getxattr(struct dentry *dentry, struct inode *inode, const char *name,
 296                void *value, size_t size)
 297 {
 298         const struct xattr_handler *handler;
 299
 300         handler = xattr_resolve_name(inode, &name);
 301         if (IS_ERR(handler))
 302                 return PTR_ERR(handler);
 303         if (!handler->get)
 304                 return -EOPNOTSUPP;
 305         return handler->get(handler, dentry, inode, name, value, size);
 306 }
 307 EXPORT_SYMBOL(__vfs_getxattr);
 308
 309 ssize_t
 310 vfs_getxattr(struct dentry *dentry, const char *name, void *value, size_t size)
 311 {
 312         struct inode *inode = dentry->d_inode;
 313         int error;
 314
 315         error = xattr_permission(inode, name, MAY_READ);
 316         if (error)
 317                 return error;
 318
 319         error = security_inode_getxattr(dentry, name);
 320         if (error)
 321                 return error;
 322
 323         if (!strncmp(name, XATTR_SECURITY_PREFIX,
 324                                 XATTR_SECURITY_PREFIX_LEN)) {
 325                 const char *suffix = name + XATTR_SECURITY_PREFIX_LEN;
 326                 int ret = xattr_getsecurity(inode, suffix, value, size);
 327                 /*
 328                  * Only overwrite the return value if a security module
 329                  * is actually active.
 330                  */
 331                 if (ret == -EOPNOTSUPP)
 332                         goto nolsm;
 333                 return ret;
 334         }
 335 nolsm:
 336         return __vfs_getxattr(dentry, inode, name, value, size);
 337 }
 338 EXPORT_SYMBOL_GPL(vfs_getxattr);
 339
 340 ssize_t
 341 vfs_listxattr(struct dentry *dentry, char *list, size_t size)
 342 {
 343         struct inode *inode = d_inode(dentry);
 344         ssize_t error;
 345
 346         error = security_inode_listxattr(dentry);
 347         if (error)
 348                 return error;
 349         if (inode->i_op->listxattr && (inode->i_opflags & IOP_XATTR)) {
 350                 error = -EOPNOTSUPP;
 351                 error = inode->i_op->listxattr(dentry, list, size);
 352         } else {
 353                 error = security_inode_listsecurity(inode, list, size);
 354                 if (size && error > size)
 355                         error = -ERANGE;
 356         }
 357         return error;
 358 }
 359 EXPORT_SYMBOL_GPL(vfs_listxattr);
 360
 361 int
 362 __vfs_removexattr(struct dentry *dentry, const char *name)
 363 {
 364         struct inode *inode = d_inode(dentry);
 365         const struct xattr_handler *handler;
 366
 367         handler = xattr_resolve_name(inode, &name);
 368         if (IS_ERR(handler))
 369                 return PTR_ERR(handler);
 370         if (!handler->set)
 371                 return -EOPNOTSUPP;
 372         return handler->set(handler, dentry, inode, name, NULL, 0, XATTR_REPLACE);
 373 }
 374 EXPORT_SYMBOL(__vfs_removexattr);
 375
 376 int
 377 vfs_removexattr(struct dentry *dentry, const char *name)
 378 {
 379         struct inode *inode = dentry->d_inode;
 380         int error;
 381
 382         error = xattr_permission(inode, name, MAY_WRITE);
 383         if (error)
 384                 return error;
 385
 386         inode_lock(inode);
 387         error = security_inode_removexattr(dentry, name);
 388         if (error)
 389                 goto out;
 390
 391         error = __vfs_removexattr(dentry, name);
 392
 393         if (!error) {
 394                 fsnotify_xattr(dentry);
 395                 evm_inode_post_removexattr(dentry, name);
 396         }
 397
 398 out:
 399         inode_unlock(inode);
 400         return error;
 401 }
 402 EXPORT_SYMBOL_GPL(vfs_removexattr);
 403
 404
 405 /*
 406  * Extended attribute SET operations
 407  */
 408 static long
 409 setxattr(struct dentry *d, const char __user *name, const void __user *value,
 410          size_t size, int flags)
 411 {
 412         int error;
 413         void *kvalue = NULL;
 414         char kname[XATTR_NAME_MAX + 1];
 415
 416         if (flags & ~(XATTR_CREATE|XATTR_REPLACE))
 417                 return -EINVAL;
 418
 419         error = strncpy_from_user(kname, name, sizeof(kname));
 420         if (error == 0 || error == sizeof(kname))
 421                 error = -ERANGE;
 422         if (error < 0)
 423                 return error;
 424
 425         if (size) {
 426                 if (size > XATTR_SIZE_MAX)
 427                         return -E2BIG;
 428                 kvalue = kmalloc(size, GFP_KERNEL | __GFP_NOWARN);
 429                 if (!kvalue) {
 430                         kvalue = vmalloc(size);
 431                         if (!kvalue)
 432                                 return -ENOMEM;
 433                 }
 434                 if (copy_from_user(kvalue, value, size)) {
 435                         error = -EFAULT;
 436                         goto out;
 437                 }
 438                 if ((strcmp(kname, XATTR_NAME_POSIX_ACL_ACCESS) == 0) ||
 439                     (strcmp(kname, XATTR_NAME_POSIX_ACL_DEFAULT) == 0))
 440                         posix_acl_fix_xattr_from_user(kvalue, size);
 441         }
 442
 443         error = vfs_setxattr(d, kname, kvalue, size, flags);
 444 out:
 445         kvfree(kvalue);
 446
 447         return error;
 448 }
 449
 450 static int path_setxattr(const char __user *pathname,
 451                          const char __user *name, const void __user *value,
 452                          size_t size, int flags, unsigned int lookup_flags)
 453 {
 454         struct path path;
 455         int error;
 456 retry:
 457         error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
 458         if (error)
 459                 return error;
 460         error = mnt_want_write(path.mnt);
 461         if (!error) {
 462                 error = setxattr(path.dentry, name, value, size, flags);
 463                 mnt_drop_write(path.mnt);
 464         }
 465         path_put(&path);
 466         if (retry_estale(error, lookup_flags)) {
 467                 lookup_flags |= LOOKUP_REVAL;
 468                 goto retry;
 469         }
 470         return error;
 471 }
 472
 473 SYSCALL_DEFINE5(setxattr, const char __user *, pathname,
 474                 const char __user *, name, const void __user *, value,
 475                 size_t, size, int, flags)
 476 {
 477         return path_setxattr(pathname, name, value, size, flags, LOOKUP_FOLLOW);
 478 }
 479
 480 SYSCALL_DEFINE5(lsetxattr, const char __user *, pathname,
 481                 const char __user *, name, const void __user *, value,
 482                 size_t, size, int, flags)
 483 {
 484         return path_setxattr(pathname, name, value, size, flags, 0);
 485 }
 486
 487 SYSCALL_DEFINE5(fsetxattr, int, fd, const char __user *, name,
 488                 const void __user *,value, size_t, size, int, flags)
 489 {
 490         struct fd f = fdget(fd);
 491         int error = -EBADF;
 492
 493         if (!f.file)
 494                 return error;
 495         audit_file(f.file);
 496         error = mnt_want_write_file(f.file);
 497         if (!error) {
 498                 error = setxattr(f.file->f_path.dentry, name, value, size, flags);
 499                 mnt_drop_write_file(f.file);
 500         }
 501         fdput(f);
 502         return error;
 503 }
 504
 505 /*
 506  * Extended attribute GET operations
 507  */
 508 static ssize_t
 509 getxattr(struct dentry *d, const char __user *name, void __user *value,
 510          size_t size)
 511 {
 512         ssize_t error;
 513         void *kvalue = NULL;
 514         char kname[XATTR_NAME_MAX + 1];
 515
 516         error = strncpy_from_user(kname, name, sizeof(kname));
 517         if (error == 0 || error == sizeof(kname))
 518                 error = -ERANGE;
 519         if (error < 0)
 520                 return error;
 521
 522         if (size) {
 523                 if (size > XATTR_SIZE_MAX)
 524                         size = XATTR_SIZE_MAX;
 525                 kvalue = kzalloc(size, GFP_KERNEL | __GFP_NOWARN);
 526                 if (!kvalue) {
 527                         kvalue = vmalloc(size);
 528                         if (!kvalue)
 529                                 return -ENOMEM;
 530                 }
 531         }
 532
 533         error = vfs_getxattr(d, kname, kvalue, size);
 534         if (error > 0) {
 535                 if ((strcmp(kname, XATTR_NAME_POSIX_ACL_ACCESS) == 0) ||
 536                     (strcmp(kname, XATTR_NAME_POSIX_ACL_DEFAULT) == 0))
 537                         posix_acl_fix_xattr_to_user(kvalue, size);
 538                 if (size && copy_to_user(value, kvalue, error))
 539                         error = -EFAULT;
 540         } else if (error == -ERANGE && size >= XATTR_SIZE_MAX) {
 541                 /* The file system tried to returned a value bigger
 542                    than XATTR_SIZE_MAX bytes. Not possible. */
 543                 error = -E2BIG;
 544         }
 545
 546         kvfree(kvalue);
 547
 548         return error;
 549 }
 550
 551 static ssize_t path_getxattr(const char __user *pathname,
 552                              const char __user *name, void __user *value,
 553                              size_t size, unsigned int lookup_flags)
 554 {
 555         struct path path;
 556         ssize_t error;
 557 retry:
 558         error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
 559         if (error)
 560                 return error;
 561         error = getxattr(path.dentry, name, value, size);
 562         path_put(&path);
 563         if (retry_estale(error, lookup_flags)) {
 564                 lookup_flags |= LOOKUP_REVAL;
 565                 goto retry;
 566         }
 567         return error;
 568 }
 569
 570 SYSCALL_DEFINE4(getxattr, const char __user *, pathname,
 571                 const char __user *, name, void __user *, value, size_t, size)
 572 {
 573         return path_getxattr(pathname, name, value, size, LOOKUP_FOLLOW);
 574 }
 575
 576 SYSCALL_DEFINE4(lgetxattr, const char __user *, pathname,
 577                 const char __user *, name, void __user *, value, size_t, size)
 578 {
 579         return path_getxattr(pathname, name, value, size, 0);
 580 }
 581
 582 SYSCALL_DEFINE4(fgetxattr, int, fd, const char __user *, name,
 583                 void __user *, value, size_t, size)
 584 {
 585         struct fd f = fdget(fd);
 586         ssize_t error = -EBADF;
 587
 588         if (!f.file)
 589                 return error;
 590         audit_file(f.file);
 591         error = getxattr(f.file->f_path.dentry, name, value, size);
 592         fdput(f);
 593         return error;
 594 }
 595
 596 /*
 597  * Extended attribute LIST operations
 598  */
 599 static ssize_t
 600 listxattr(struct dentry *d, char __user *list, size_t size)
 601 {
 602         ssize_t error;
 603         char *klist = NULL;
 604
 605         if (size) {
 606                 if (size > XATTR_LIST_MAX)
 607                         size = XATTR_LIST_MAX;
 608                 klist = kmalloc(size, __GFP_NOWARN | GFP_KERNEL);
 609                 if (!klist) {
 610                         klist = vmalloc(size);
 611                         if (!klist)
 612                                 return -ENOMEM;
 613                 }
 614         }
 615
 616         error = vfs_listxattr(d, klist, size);
 617         if (error > 0) {
 618                 if (size && copy_to_user(list, klist, error))
 619                         error = -EFAULT;
 620         } else if (error == -ERANGE && size >= XATTR_LIST_MAX) {
 621                 /* The file system tried to returned a list bigger
 622                    than XATTR_LIST_MAX bytes. Not possible. */
 623                 error = -E2BIG;
 624         }
 625
 626         kvfree(klist);
 627
 628         return error;
 629 }
 630
 631 static ssize_t path_listxattr(const char __user *pathname, char __user *list,
 632                               size_t size, unsigned int lookup_flags)
 633 {
 634         struct path path;
 635         ssize_t error;
 636 retry:
 637         error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
 638         if (error)
 639                 return error;
 640         error = listxattr(path.dentry, list, size);
 641         path_put(&path);
 642         if (retry_estale(error, lookup_flags)) {
 643                 lookup_flags |= LOOKUP_REVAL;
 644                 goto retry;
 645         }
 646         return error;
 647 }
 648
 649 SYSCALL_DEFINE3(listxattr, const char __user *, pathname, char __user *, list,
 650                 size_t, size)
 651 {
 652         return path_listxattr(pathname, list, size, LOOKUP_FOLLOW);
 653 }
 654
 655 SYSCALL_DEFINE3(llistxattr, const char __user *, pathname, char __user *, list,
 656                 size_t, size)
 657 {
 658         return path_listxattr(pathname, list, size, 0);
 659 }
 660
 661 SYSCALL_DEFINE3(flistxattr, int, fd, char __user *, list, size_t, size)
 662 {
 663         struct fd f = fdget(fd);
 664         ssize_t error = -EBADF;
 665
 666         if (!f.file)
 667                 return error;
 668         audit_file(f.file);
 669         error = listxattr(f.file->f_path.dentry, list, size);
 670         fdput(f);
 671         return error;
 672 }
 673
 674 /*
 675  * Extended attribute REMOVE operations
 676  */
 677 static long
 678 removexattr(struct dentry *d, const char __user *name)
 679 {
 680         int error;
 681         char kname[XATTR_NAME_MAX + 1];
 682
 683         error = strncpy_from_user(kname, name, sizeof(kname));
 684         if (error == 0 || error == sizeof(kname))
 685                 error = -ERANGE;
 686         if (error < 0)
 687                 return error;
 688
 689         return vfs_removexattr(d, kname);
 690 }
 691
 692 static int path_removexattr(const char __user *pathname,
 693                             const char __user *name, unsigned int lookup_flags)
 694 {
 695         struct path path;
 696         int error;
 697 retry:
 698         error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
 699         if (error)
 700                 return error;
 701         error = mnt_want_write(path.mnt);
 702         if (!error) {
 703                 error = removexattr(path.dentry, name);
 704                 mnt_drop_write(path.mnt);
 705         }
 706         path_put(&path);
 707         if (retry_estale(error, lookup_flags)) {
 708                 lookup_flags |= LOOKUP_REVAL;
 709                 goto retry;
 710         }
 711         return error;
 712 }
 713
 714 SYSCALL_DEFINE2(removexattr, const char __user *, pathname,
 715                 const char __user *, name)
 716 {
 717         return path_removexattr(pathname, name, LOOKUP_FOLLOW);
 718 }
 719
 720 SYSCALL_DEFINE2(lremovexattr, const char __user *, pathname,
 721                 const char __user *, name)
 722 {
 723         return path_removexattr(pathname, name, 0);
 724 }
 725
 726 SYSCALL_DEFINE2(fremovexattr, int, fd, const char __user *, name)
 727 {
 728         struct fd f = fdget(fd);
 729         int error = -EBADF;
 730
 731         if (!f.file)
 732                 return error;
 733         audit_file(f.file);
 734         error = mnt_want_write_file(f.file);
 735         if (!error) {
 736                 error = removexattr(f.file->f_path.dentry, name);
 737                 mnt_drop_write_file(f.file);
 738         }
 739         fdput(f);
 740         return error;
 741 }
 742
 743 /*
 744  * Combine the results of the list() operation from every xattr_handler in the
 745  * list.
 746  */
 747 ssize_t
 748 generic_listxattr(struct dentry *dentry, char *buffer, size_t buffer_size)
 749 {
 750         const struct xattr_handler *handler, **handlers = dentry->d_sb->s_xattr;
 751         unsigned int size = 0;
 752
 753         if (!buffer) {
 754                 for_each_xattr_handler(handlers, handler) {
 755                         if (!handler->name ||
 756                             (handler->list && !handler->list(dentry)))
 757                                 continue;
 758                         size += strlen(handler->name) + 1;
 759                 }
 760         } else {
 761                 char *buf = buffer;
 762                 size_t len;
 763
 764                 for_each_xattr_handler(handlers, handler) {
 765                         if (!handler->name ||
 766                             (handler->list && !handler->list(dentry)))
 767                                 continue;
 768                         len = strlen(handler->name);
 769                         if (len + 1 > buffer_size)
 770                                 return -ERANGE;
 771                         memcpy(buf, handler->name, len + 1);
 772                         buf += len + 1;
 773                         buffer_size -= len + 1;
 774                 }
 775                 size = buf - buffer;
 776         }
 777         return size;
 778 }
 779 EXPORT_SYMBOL(generic_listxattr);
 780
 781 /**
 782  * xattr_full_name  -  Compute full attribute name from suffix
 783  *
 784  * @handler:    handler of the xattr_handler operation
 785  * @name:       name passed to the xattr_handler operation
 786  *
 787  * The get and set xattr handler operations are called with the remainder of
 788  * the attribute name after skipping the handler's prefix: for example, "foo"
 789  * is passed to the get operation of a handler with prefix "user." to get
 790  * attribute "user.foo".  The full name is still "there" in the name though.
 791  *
 792  * Note: the list xattr handler operation when called from the vfs is passed a
 793  * NULL name; some file systems use this operation internally, with varying
 794  * semantics.
 795  */
 796 const char *xattr_full_name(const struct xattr_handler *handler,
 797                             const char *name)
 798 {
 799         size_t prefix_len = strlen(xattr_prefix(handler));
 800
 801         return name - prefix_len;
 802 }
 803 EXPORT_SYMBOL(xattr_full_name);
 804
 805 /*
 806  * Allocate new xattr and copy in the value; but leave the name to callers.
 807  */
 808 struct simple_xattr *simple_xattr_alloc(const void *value, size_t size)
 809 {
 810         struct simple_xattr *new_xattr;
 811         size_t len;
 812
 813         /* wrap around? */
 814         len = sizeof(*new_xattr) + size;
 815         if (len < sizeof(*new_xattr))
 816                 return NULL;
 817
 818         new_xattr = kmalloc(len, GFP_KERNEL);
 819         if (!new_xattr)
 820                 return NULL;
 821
 822         new_xattr->size = size;
 823         memcpy(new_xattr->value, value, size);
 824         return new_xattr;
 825 }
 826
 827 /*
 828  * xattr GET operation for in-memory/pseudo filesystems
 829  */
 830 int simple_xattr_get(struct simple_xattrs *xattrs, const char *name,
 831                      void *buffer, size_t size)
 832 {
 833         struct simple_xattr *xattr;
 834         int ret = -ENODATA;
 835
 836         spin_lock(&xattrs->lock);
 837         list_for_each_entry(xattr, &xattrs->head, list) {
 838                 if (strcmp(name, xattr->name))
 839                         continue;
 840
 841                 ret = xattr->size;
 842                 if (buffer) {
 843                         if (size < xattr->size)
 844                                 ret = -ERANGE;
 845                         else
 846                                 memcpy(buffer, xattr->value, xattr->size);
 847                 }
 848                 break;
 849         }
 850         spin_unlock(&xattrs->lock);
 851         return ret;
 852 }
 853
 854 /**
 855  * simple_xattr_set - xattr SET operation for in-memory/pseudo filesystems
 856  * @xattrs: target simple_xattr list
 857  * @name: name of the extended attribute
 858  * @value: value of the xattr. If %NULL, will remove the attribute.
 859  * @size: size of the new xattr
 860  * @flags: %XATTR_{CREATE|REPLACE}
 861  *
 862  * %XATTR_CREATE is set, the xattr shouldn't exist already; otherwise fails
 863  * with -EEXIST.  If %XATTR_REPLACE is set, the xattr should exist;
 864  * otherwise, fails with -ENODATA.
 865  *
 866  * Returns 0 on success, -errno on failure.
 867  */
 868 int simple_xattr_set(struct simple_xattrs *xattrs, const char *name,
 869                      const void *value, size_t size, int flags)
 870 {
 871         struct simple_xattr *xattr;
 872         struct simple_xattr *new_xattr = NULL;
 873         int err = 0;
 874
 875         /* value == NULL means remove */
 876         if (value) {
 877                 new_xattr = simple_xattr_alloc(value, size);
 878                 if (!new_xattr)
 879                         return -ENOMEM;
 880
 881                 new_xattr->name = kstrdup(name, GFP_KERNEL);
 882                 if (!new_xattr->name) {
 883                         kfree(new_xattr);
 884                         return -ENOMEM;
 885                 }
 886         }
 887
 888         spin_lock(&xattrs->lock);
 889         list_for_each_entry(xattr, &xattrs->head, list) {
 890                 if (!strcmp(name, xattr->name)) {
 891                         if (flags & XATTR_CREATE) {
 892                                 xattr = new_xattr;
 893                                 err = -EEXIST;
 894                         } else if (new_xattr) {
 895                                 list_replace(&xattr->list, &new_xattr->list);
 896                         } else {
 897                                 list_del(&xattr->list);
 898                         }
 899                         goto out;
 900                 }
 901         }
 902         if (flags & XATTR_REPLACE) {
 903                 xattr = new_xattr;
 904                 err = -ENODATA;
 905         } else {
 906                 list_add(&new_xattr->list, &xattrs->head);
 907                 xattr = NULL;
 908         }
 909 out:
 910         spin_unlock(&xattrs->lock);
 911         if (xattr) {
 912                 kfree(xattr->name);
 913                 kfree(xattr);
 914         }
 915         return err;
 916
 917 }
 918
 919 static bool xattr_is_trusted(const char *name)
 920 {
 921         return !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN);
 922 }
 923
 924 static int xattr_list_one(char **buffer, ssize_t *remaining_size,
 925                           const char *name)
 926 {
 927         size_t len = strlen(name) + 1;
 928         if (*buffer) {
 929                 if (*remaining_size < len)
 930                         return -ERANGE;
 931                 memcpy(*buffer, name, len);
 932                 *buffer += len;
 933         }
 934         *remaining_size -= len;
 935         return 0;
 936 }
 937
 938 /*
 939  * xattr LIST operation for in-memory/pseudo filesystems
 940  */
 941 ssize_t simple_xattr_list(struct inode *inode, struct simple_xattrs *xattrs,
 942                           char *buffer, size_t size)
 943 {
 944         bool trusted = capable(CAP_SYS_ADMIN);
 945         struct simple_xattr *xattr;
 946         ssize_t remaining_size = size;
 947         int err = 0;
 948
 949 #ifdef CONFIG_FS_POSIX_ACL
 950         if (inode->i_acl) {
 951                 err = xattr_list_one(&buffer, &remaining_size,
 952                                      XATTR_NAME_POSIX_ACL_ACCESS);
 953                 if (err)
 954                         return err;
 955         }
 956         if (inode->i_default_acl) {
 957                 err = xattr_list_one(&buffer, &remaining_size,
 958                                      XATTR_NAME_POSIX_ACL_DEFAULT);
 959                 if (err)
 960                         return err;
 961         }
 962 #endif
 963
 964         spin_lock(&xattrs->lock);
 965         list_for_each_entry(xattr, &xattrs->head, list) {
 966                 /* skip "trusted." attributes for unprivileged callers */
 967                 if (!trusted && xattr_is_trusted(xattr->name))
 968                         continue;
 969
 970                 err = xattr_list_one(&buffer, &remaining_size, xattr->name);
 971                 if (err)
 972                         break;
 973         }
 974         spin_unlock(&xattrs->lock);
 975
 976         return err ? err : size - remaining_size;
 977 }
 978
 979 /*
 980  * Adds an extended attribute to the list
 981  */
 982 void simple_xattr_list_add(struct simple_xattrs *xattrs,
 983                            struct simple_xattr *new_xattr)
 984 {
 985         spin_lock(&xattrs->lock);
 986         list_add(&new_xattr->list, &xattrs->head);
 987         spin_unlock(&xattrs->lock);
 988 }