fs/xattr.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2 /*
   3   File: fs/xattr.c
   4
   5   Extended attribute handling.
   6
   7   Copyright (C) 2001 by Andreas Gruenbacher <a.gruenbacher@computer.org>
   8   Copyright (C) 2001 SGI - Silicon Graphics, Inc <linux-xfs@oss.sgi.com>
   9   Copyright (c) 2004 Red Hat, Inc., James Morris <jmorris@redhat.com>
  10  */
  11 #include <linux/fs.h>
  12 #include <linux/slab.h>
  13 #include <linux/file.h>
  14 #include <linux/xattr.h>
  15 #include <linux/mount.h>
  16 #include <linux/namei.h>
  17 #include <linux/security.h>
  18 #include <linux/evm.h>
  19 #include <linux/syscalls.h>
  20 #include <linux/export.h>
  21 #include <linux/fsnotify.h>
  22 #include <linux/audit.h>
  23 #include <linux/vmalloc.h>
  24 #include <linux/posix_acl_xattr.h>
  25
  26 #include <linux/uaccess.h>
  27
  28 static const char *
  29 strcmp_prefix(const char *a, const char *a_prefix)
  30 {
  31         while (*a_prefix && *a == *a_prefix) {
  32                 a++;
  33                 a_prefix++;
  34         }
  35         return *a_prefix ? NULL : a;
  36 }
  37
  38 /*
  39  * In order to implement different sets of xattr operations for each xattr
  40  * prefix, a filesystem should create a null-terminated array of struct
  41  * xattr_handler (one for each prefix) and hang a pointer to it off of the
  42  * s_xattr field of the superblock.
  43  */
  44 #define for_each_xattr_handler(handlers, handler)               \
  45         if (handlers)                                           \
  46                 for ((handler) = *(handlers)++;                 \
  47                         (handler) != NULL;                      \
  48                         (handler) = *(handlers)++)
  49
  50 /*
  51  * Find the xattr_handler with the matching prefix.
  52  */
  53 static const struct xattr_handler *
  54 xattr_resolve_name(struct inode *inode, const char **name)
  55 {
  56         const struct xattr_handler **handlers = inode->i_sb->s_xattr;
  57         const struct xattr_handler *handler;
  58
  59         if (!(inode->i_opflags & IOP_XATTR)) {
  60                 if (unlikely(is_bad_inode(inode)))
  61                         return ERR_PTR(-EIO);
  62                 return ERR_PTR(-EOPNOTSUPP);
  63         }
  64         for_each_xattr_handler(handlers, handler) {
  65                 const char *n;
  66
  67                 n = strcmp_prefix(*name, xattr_prefix(handler));
  68                 if (n) {
  69                         if (!handler->prefix ^ !*n) {
  70                                 if (*n)
  71                                         continue;
  72                                 return ERR_PTR(-EINVAL);
  73                         }
  74                         *name = n;
  75                         return handler;
  76                 }
  77         }
  78         return ERR_PTR(-EOPNOTSUPP);
  79 }
  80
  81 /*
  82  * Check permissions for extended attribute access.  This is a bit complicated
  83  * because different namespaces have very different rules.
  84  */
  85 static int
  86 xattr_permission(struct user_namespace *mnt_userns, struct inode *inode,
  87                  const char *name, int mask)
  88 {
  89         /*
  90          * We can never set or remove an extended attribute on a read-only
  91          * filesystem  or on an immutable / append-only inode.
  92          */
  93         if (mask & MAY_WRITE) {
  94                 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
  95                         return -EPERM;
  96                 /*
  97                  * Updating an xattr will likely cause i_uid and i_gid
  98                  * to be writen back improperly if their true value is
  99                  * unknown to the vfs.
 100                  */
 101                 if (HAS_UNMAPPED_ID(mnt_userns, inode))
 102                         return -EPERM;
 103         }
 104
 105         /*
 106          * No restriction for security.* and system.* from the VFS.  Decision
 107          * on these is left to the underlying filesystem / security module.
 108          */
 109         if (!strncmp(name, XATTR_SECURITY_PREFIX, XATTR_SECURITY_PREFIX_LEN) ||
 110             !strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN))
 111                 return 0;
 112
 113         /*
 114          * The trusted.* namespace can only be accessed by privileged users.
 115          */
 116         if (!strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN)) {
 117                 if (!capable(CAP_SYS_ADMIN))
 118                         return (mask & MAY_WRITE) ? -EPERM : -ENODATA;
 119                 return 0;
 120         }
 121
 122         /*
 123          * In the user.* namespace, only regular files and directories can have
 124          * extended attributes. For sticky directories, only the owner and
 125          * privileged users can write attributes.
 126          */
 127         if (!strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN)) {
 128                 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
 129                         return (mask & MAY_WRITE) ? -EPERM : -ENODATA;
 130                 if (S_ISDIR(inode->i_mode) && (inode->i_mode & S_ISVTX) &&
 131                     (mask & MAY_WRITE) &&
 132                     !inode_owner_or_capable(mnt_userns, inode))
 133                         return -EPERM;
 134         }
 135
 136         return inode_permission(mnt_userns, inode, mask);
 137 }
 138
 139 /*
 140  * Look for any handler that deals with the specified namespace.
 141  */
 142 int
 143 xattr_supported_namespace(struct inode *inode, const char *prefix)
 144 {
 145         const struct xattr_handler **handlers = inode->i_sb->s_xattr;
 146         const struct xattr_handler *handler;
 147         size_t preflen;
 148
 149         if (!(inode->i_opflags & IOP_XATTR)) {
 150                 if (unlikely(is_bad_inode(inode)))
 151                         return -EIO;
 152                 return -EOPNOTSUPP;
 153         }
 154
 155         preflen = strlen(prefix);
 156
 157         for_each_xattr_handler(handlers, handler) {
 158                 if (!strncmp(xattr_prefix(handler), prefix, preflen))
 159                         return 0;
 160         }
 161
 162         return -EOPNOTSUPP;
 163 }
 164 EXPORT_SYMBOL(xattr_supported_namespace);
 165
 166 int
 167 __vfs_setxattr(struct user_namespace *mnt_userns, struct dentry *dentry,
 168                struct inode *inode, const char *name, const void *value,
 169                size_t size, int flags)
 170 {
 171         const struct xattr_handler *handler;
 172
 173         handler = xattr_resolve_name(inode, &name);
 174         if (IS_ERR(handler))
 175                 return PTR_ERR(handler);
 176         if (!handler->set)
 177                 return -EOPNOTSUPP;
 178         if (size == 0)
 179                 value = "";  /* empty EA, do not remove */
 180         return handler->set(handler, mnt_userns, dentry, inode, name, value,
 181                             size, flags);
 182 }
 183 EXPORT_SYMBOL(__vfs_setxattr);
 184
 185 /**
 186  *  __vfs_setxattr_noperm - perform setxattr operation without performing
 187  *  permission checks.
 188  *
 189  *  @mnt_userns: user namespace of the mount the inode was found from
 190  *  @dentry: object to perform setxattr on
 191  *  @name: xattr name to set
 192  *  @value: value to set @name to
 193  *  @size: size of @value
 194  *  @flags: flags to pass into filesystem operations
 195  *
 196  *  returns the result of the internal setxattr or setsecurity operations.
 197  *
 198  *  This function requires the caller to lock the inode's i_mutex before it
 199  *  is executed. It also assumes that the caller will make the appropriate
 200  *  permission checks.
 201  */
 202 int __vfs_setxattr_noperm(struct user_namespace *mnt_userns,
 203                           struct dentry *dentry, const char *name,
 204                           const void *value, size_t size, int flags)
 205 {
 206         struct inode *inode = dentry->d_inode;
 207         int error = -EAGAIN;
 208         int issec = !strncmp(name, XATTR_SECURITY_PREFIX,
 209                                    XATTR_SECURITY_PREFIX_LEN);
 210
 211         if (issec)
 212                 inode->i_flags &= ~S_NOSEC;
 213         if (inode->i_opflags & IOP_XATTR) {
 214                 error = __vfs_setxattr(mnt_userns, dentry, inode, name, value,
 215                                        size, flags);
 216                 if (!error) {
 217                         fsnotify_xattr(dentry);
 218                         security_inode_post_setxattr(dentry, name, value,
 219                                                      size, flags);
 220                 }
 221         } else {
 222                 if (unlikely(is_bad_inode(inode)))
 223                         return -EIO;
 224         }
 225         if (error == -EAGAIN) {
 226                 error = -EOPNOTSUPP;
 227
 228                 if (issec) {
 229                         const char *suffix = name + XATTR_SECURITY_PREFIX_LEN;
 230
 231                         error = security_inode_setsecurity(inode, suffix, value,
 232                                                            size, flags);
 233                         if (!error)
 234                                 fsnotify_xattr(dentry);
 235                 }
 236         }
 237
 238         return error;
 239 }
 240
 241 /**
 242  * __vfs_setxattr_locked - set an extended attribute while holding the inode
 243  * lock
 244  *
 245  *  @mnt_userns: user namespace of the mount of the target inode
 246  *  @dentry: object to perform setxattr on
 247  *  @name: xattr name to set
 248  *  @value: value to set @name to
 249  *  @size: size of @value
 250  *  @flags: flags to pass into filesystem operations
 251  *  @delegated_inode: on return, will contain an inode pointer that
 252  *  a delegation was broken on, NULL if none.
 253  */
 254 int
 255 __vfs_setxattr_locked(struct user_namespace *mnt_userns, struct dentry *dentry,
 256                       const char *name, const void *value, size_t size,
 257                       int flags, struct inode **delegated_inode)
 258 {
 259         struct inode *inode = dentry->d_inode;
 260         int error;
 261
 262         error = xattr_permission(mnt_userns, inode, name, MAY_WRITE);
 263         if (error)
 264                 return error;
 265
 266         error = security_inode_setxattr(mnt_userns, dentry, name, value, size,
 267                                         flags);
 268         if (error)
 269                 goto out;
 270
 271         error = try_break_deleg(inode, delegated_inode);
 272         if (error)
 273                 goto out;
 274
 275         error = __vfs_setxattr_noperm(mnt_userns, dentry, name, value,
 276                                       size, flags);
 277
 278 out:
 279         return error;
 280 }
 281 EXPORT_SYMBOL_GPL(__vfs_setxattr_locked);
 282
 283 int
 284 vfs_setxattr(struct user_namespace *mnt_userns, struct dentry *dentry,
 285              const char *name, const void *value, size_t size, int flags)
 286 {
 287         struct inode *inode = dentry->d_inode;
 288         struct inode *delegated_inode = NULL;
 289         const void  *orig_value = value;
 290         int error;
 291
 292         if (size && strcmp(name, XATTR_NAME_CAPS) == 0) {
 293                 error = cap_convert_nscap(mnt_userns, dentry, &value, size);
 294                 if (error < 0)
 295                         return error;
 296                 size = error;
 297         }
 298
 299 retry_deleg:
 300         inode_lock(inode);
 301         error = __vfs_setxattr_locked(mnt_userns, dentry, name, value, size,
 302                                       flags, &delegated_inode);
 303         inode_unlock(inode);
 304
 305         if (delegated_inode) {
 306                 error = break_deleg_wait(&delegated_inode);
 307                 if (!error)
 308                         goto retry_deleg;
 309         }
 310         if (value != orig_value)
 311                 kfree(value);
 312
 313         return error;
 314 }
 315 EXPORT_SYMBOL_GPL(vfs_setxattr);
 316
 317 static ssize_t
 318 xattr_getsecurity(struct user_namespace *mnt_userns, struct inode *inode,
 319                   const char *name, void *value, size_t size)
 320 {
 321         void *buffer = NULL;
 322         ssize_t len;
 323
 324         if (!value || !size) {
 325                 len = security_inode_getsecurity(mnt_userns, inode, name,
 326                                                  &buffer, false);
 327                 goto out_noalloc;
 328         }
 329
 330         len = security_inode_getsecurity(mnt_userns, inode, name, &buffer,
 331                                          true);
 332         if (len < 0)
 333                 return len;
 334         if (size < len) {
 335                 len = -ERANGE;
 336                 goto out;
 337         }
 338         memcpy(value, buffer, len);
 339 out:
 340         kfree(buffer);
 341 out_noalloc:
 342         return len;
 343 }
 344
 345 /*
 346  * vfs_getxattr_alloc - allocate memory, if necessary, before calling getxattr
 347  *
 348  * Allocate memory, if not already allocated, or re-allocate correct size,
 349  * before retrieving the extended attribute.
 350  *
 351  * Returns the result of alloc, if failed, or the getxattr operation.
 352  */
 353 ssize_t
 354 vfs_getxattr_alloc(struct user_namespace *mnt_userns, struct dentry *dentry,
 355                    const char *name, char **xattr_value, size_t xattr_size,
 356                    gfp_t flags)
 357 {
 358         const struct xattr_handler *handler;
 359         struct inode *inode = dentry->d_inode;
 360         char *value = *xattr_value;
 361         int error;
 362
 363         error = xattr_permission(mnt_userns, inode, name, MAY_READ);
 364         if (error)
 365                 return error;
 366
 367         handler = xattr_resolve_name(inode, &name);
 368         if (IS_ERR(handler))
 369                 return PTR_ERR(handler);
 370         if (!handler->get)
 371                 return -EOPNOTSUPP;
 372         error = handler->get(handler, dentry, inode, name, NULL, 0);
 373         if (error < 0)
 374                 return error;
 375
 376         if (!value || (error > xattr_size)) {
 377                 value = krealloc(*xattr_value, error + 1, flags);
 378                 if (!value)
 379                         return -ENOMEM;
 380                 memset(value, 0, error + 1);
 381         }
 382
 383         error = handler->get(handler, dentry, inode, name, value, error);
 384         *xattr_value = value;
 385         return error;
 386 }
 387
 388 ssize_t
 389 __vfs_getxattr(struct dentry *dentry, struct inode *inode, const char *name,
 390                void *value, size_t size)
 391 {
 392         const struct xattr_handler *handler;
 393
 394         handler = xattr_resolve_name(inode, &name);
 395         if (IS_ERR(handler))
 396                 return PTR_ERR(handler);
 397         if (!handler->get)
 398                 return -EOPNOTSUPP;
 399         return handler->get(handler, dentry, inode, name, value, size);
 400 }
 401 EXPORT_SYMBOL(__vfs_getxattr);
 402
 403 ssize_t
 404 vfs_getxattr(struct user_namespace *mnt_userns, struct dentry *dentry,
 405              const char *name, void *value, size_t size)
 406 {
 407         struct inode *inode = dentry->d_inode;
 408         int error;
 409
 410         error = xattr_permission(mnt_userns, inode, name, MAY_READ);
 411         if (error)
 412                 return error;
 413
 414         error = security_inode_getxattr(dentry, name);
 415         if (error)
 416                 return error;
 417
 418         if (!strncmp(name, XATTR_SECURITY_PREFIX,
 419                                 XATTR_SECURITY_PREFIX_LEN)) {
 420                 const char *suffix = name + XATTR_SECURITY_PREFIX_LEN;
 421                 int ret = xattr_getsecurity(mnt_userns, inode, suffix, value,
 422                                             size);
 423                 /*
 424                  * Only overwrite the return value if a security module
 425                  * is actually active.
 426                  */
 427                 if (ret == -EOPNOTSUPP)
 428                         goto nolsm;
 429                 return ret;
 430         }
 431 nolsm:
 432         return __vfs_getxattr(dentry, inode, name, value, size);
 433 }
 434 EXPORT_SYMBOL_GPL(vfs_getxattr);
 435
 436 ssize_t
 437 vfs_listxattr(struct dentry *dentry, char *list, size_t size)
 438 {
 439         struct inode *inode = d_inode(dentry);
 440         ssize_t error;
 441
 442         error = security_inode_listxattr(dentry);
 443         if (error)
 444                 return error;
 445         if (inode->i_op->listxattr && (inode->i_opflags & IOP_XATTR)) {
 446                 error = inode->i_op->listxattr(dentry, list, size);
 447         } else {
 448                 error = security_inode_listsecurity(inode, list, size);
 449                 if (size && error > size)
 450                         error = -ERANGE;
 451         }
 452         return error;
 453 }
 454 EXPORT_SYMBOL_GPL(vfs_listxattr);
 455
 456 int
 457 __vfs_removexattr(struct user_namespace *mnt_userns, struct dentry *dentry,
 458                   const char *name)
 459 {
 460         struct inode *inode = d_inode(dentry);
 461         const struct xattr_handler *handler;
 462
 463         handler = xattr_resolve_name(inode, &name);
 464         if (IS_ERR(handler))
 465                 return PTR_ERR(handler);
 466         if (!handler->set)
 467                 return -EOPNOTSUPP;
 468         return handler->set(handler, mnt_userns, dentry, inode, name, NULL, 0,
 469                             XATTR_REPLACE);
 470 }
 471 EXPORT_SYMBOL(__vfs_removexattr);
 472
 473 /**
 474  * __vfs_removexattr_locked - set an extended attribute while holding the inode
 475  * lock
 476  *
 477  *  @mnt_userns: user namespace of the mount of the target inode
 478  *  @dentry: object to perform setxattr on
 479  *  @name: name of xattr to remove
 480  *  @delegated_inode: on return, will contain an inode pointer that
 481  *  a delegation was broken on, NULL if none.
 482  */
 483 int
 484 __vfs_removexattr_locked(struct user_namespace *mnt_userns,
 485                          struct dentry *dentry, const char *name,
 486                          struct inode **delegated_inode)
 487 {
 488         struct inode *inode = dentry->d_inode;
 489         int error;
 490
 491         error = xattr_permission(mnt_userns, inode, name, MAY_WRITE);
 492         if (error)
 493                 return error;
 494
 495         error = security_inode_removexattr(mnt_userns, dentry, name);
 496         if (error)
 497                 goto out;
 498
 499         error = try_break_deleg(inode, delegated_inode);
 500         if (error)
 501                 goto out;
 502
 503         error = __vfs_removexattr(mnt_userns, dentry, name);
 504
 505         if (!error) {
 506                 fsnotify_xattr(dentry);
 507                 evm_inode_post_removexattr(dentry, name);
 508         }
 509
 510 out:
 511         return error;
 512 }
 513 EXPORT_SYMBOL_GPL(__vfs_removexattr_locked);
 514
 515 int
 516 vfs_removexattr(struct user_namespace *mnt_userns, struct dentry *dentry,
 517                 const char *name)
 518 {
 519         struct inode *inode = dentry->d_inode;
 520         struct inode *delegated_inode = NULL;
 521         int error;
 522
 523 retry_deleg:
 524         inode_lock(inode);
 525         error = __vfs_removexattr_locked(mnt_userns, dentry,
 526                                          name, &delegated_inode);
 527         inode_unlock(inode);
 528
 529         if (delegated_inode) {
 530                 error = break_deleg_wait(&delegated_inode);
 531                 if (!error)
 532                         goto retry_deleg;
 533         }
 534
 535         return error;
 536 }
 537 EXPORT_SYMBOL_GPL(vfs_removexattr);
 538
 539 /*
 540  * Extended attribute SET operations
 541  */
 542 static long
 543 setxattr(struct user_namespace *mnt_userns, struct dentry *d,
 544          const char __user *name, const void __user *value, size_t size,
 545          int flags)
 546 {
 547         int error;
 548         void *kvalue = NULL;
 549         char kname[XATTR_NAME_MAX + 1];
 550
 551         if (flags & ~(XATTR_CREATE|XATTR_REPLACE))
 552                 return -EINVAL;
 553
 554         error = strncpy_from_user(kname, name, sizeof(kname));
 555         if (error == 0 || error == sizeof(kname))
 556                 error = -ERANGE;
 557         if (error < 0)
 558                 return error;
 559
 560         if (size) {
 561                 if (size > XATTR_SIZE_MAX)
 562                         return -E2BIG;
 563                 kvalue = kvmalloc(size, GFP_KERNEL);
 564                 if (!kvalue)
 565                         return -ENOMEM;
 566                 if (copy_from_user(kvalue, value, size)) {
 567                         error = -EFAULT;
 568                         goto out;
 569                 }
 570                 if ((strcmp(kname, XATTR_NAME_POSIX_ACL_ACCESS) == 0) ||
 571                     (strcmp(kname, XATTR_NAME_POSIX_ACL_DEFAULT) == 0))
 572                         posix_acl_fix_xattr_from_user(mnt_userns, kvalue, size);
 573         }
 574
 575         error = vfs_setxattr(mnt_userns, d, kname, kvalue, size, flags);
 576 out:
 577         kvfree(kvalue);
 578
 579         return error;
 580 }
 581
 582 static int path_setxattr(const char __user *pathname,
 583                          const char __user *name, const void __user *value,
 584                          size_t size, int flags, unsigned int lookup_flags)
 585 {
 586         struct path path;
 587         int error;
 588
 589 retry:
 590         error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
 591         if (error)
 592                 return error;
 593         error = mnt_want_write(path.mnt);
 594         if (!error) {
 595                 error = setxattr(mnt_user_ns(path.mnt), path.dentry, name,
 596                                  value, size, flags);
 597                 mnt_drop_write(path.mnt);
 598         }
 599         path_put(&path);
 600         if (retry_estale(error, lookup_flags)) {
 601                 lookup_flags |= LOOKUP_REVAL;
 602                 goto retry;
 603         }
 604         return error;
 605 }
 606
 607 SYSCALL_DEFINE5(setxattr, const char __user *, pathname,
 608                 const char __user *, name, const void __user *, value,
 609                 size_t, size, int, flags)
 610 {
 611         return path_setxattr(pathname, name, value, size, flags, LOOKUP_FOLLOW);
 612 }
 613
 614 SYSCALL_DEFINE5(lsetxattr, const char __user *, pathname,
 615                 const char __user *, name, const void __user *, value,
 616                 size_t, size, int, flags)
 617 {
 618         return path_setxattr(pathname, name, value, size, flags, 0);
 619 }
 620
 621 SYSCALL_DEFINE5(fsetxattr, int, fd, const char __user *, name,
 622                 const void __user *,value, size_t, size, int, flags)
 623 {
 624         struct fd f = fdget(fd);
 625         int error = -EBADF;
 626
 627         if (!f.file)
 628                 return error;
 629         audit_file(f.file);
 630         error = mnt_want_write_file(f.file);
 631         if (!error) {
 632                 error = setxattr(file_mnt_user_ns(f.file),
 633                                  f.file->f_path.dentry, name,
 634                                  value, size, flags);
 635                 mnt_drop_write_file(f.file);
 636         }
 637         fdput(f);
 638         return error;
 639 }
 640
 641 /*
 642  * Extended attribute GET operations
 643  */
 644 static ssize_t
 645 getxattr(struct user_namespace *mnt_userns, struct dentry *d,
 646          const char __user *name, void __user *value, size_t size)
 647 {
 648         ssize_t error;
 649         void *kvalue = NULL;
 650         char kname[XATTR_NAME_MAX + 1];
 651
 652         error = strncpy_from_user(kname, name, sizeof(kname));
 653         if (error == 0 || error == sizeof(kname))
 654                 error = -ERANGE;
 655         if (error < 0)
 656                 return error;
 657
 658         if (size) {
 659                 if (size > XATTR_SIZE_MAX)
 660                         size = XATTR_SIZE_MAX;
 661                 kvalue = kvzalloc(size, GFP_KERNEL);
 662                 if (!kvalue)
 663                         return -ENOMEM;
 664         }
 665
 666         error = vfs_getxattr(mnt_userns, d, kname, kvalue, size);
 667         if (error > 0) {
 668                 if ((strcmp(kname, XATTR_NAME_POSIX_ACL_ACCESS) == 0) ||
 669                     (strcmp(kname, XATTR_NAME_POSIX_ACL_DEFAULT) == 0))
 670                         posix_acl_fix_xattr_to_user(mnt_userns, kvalue, error);
 671                 if (size && copy_to_user(value, kvalue, error))
 672                         error = -EFAULT;
 673         } else if (error == -ERANGE && size >= XATTR_SIZE_MAX) {
 674                 /* The file system tried to returned a value bigger
 675                    than XATTR_SIZE_MAX bytes. Not possible. */
 676                 error = -E2BIG;
 677         }
 678
 679         kvfree(kvalue);
 680
 681         return error;
 682 }
 683
 684 static ssize_t path_getxattr(const char __user *pathname,
 685                              const char __user *name, void __user *value,
 686                              size_t size, unsigned int lookup_flags)
 687 {
 688         struct path path;
 689         ssize_t error;
 690 retry:
 691         error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
 692         if (error)
 693                 return error;
 694         error = getxattr(mnt_user_ns(path.mnt), path.dentry, name, value, size);
 695         path_put(&path);
 696         if (retry_estale(error, lookup_flags)) {
 697                 lookup_flags |= LOOKUP_REVAL;
 698                 goto retry;
 699         }
 700         return error;
 701 }
 702
 703 SYSCALL_DEFINE4(getxattr, const char __user *, pathname,
 704                 const char __user *, name, void __user *, value, size_t, size)
 705 {
 706         return path_getxattr(pathname, name, value, size, LOOKUP_FOLLOW);
 707 }
 708
 709 SYSCALL_DEFINE4(lgetxattr, const char __user *, pathname,
 710                 const char __user *, name, void __user *, value, size_t, size)
 711 {
 712         return path_getxattr(pathname, name, value, size, 0);
 713 }
 714
 715 SYSCALL_DEFINE4(fgetxattr, int, fd, const char __user *, name,
 716                 void __user *, value, size_t, size)
 717 {
 718         struct fd f = fdget(fd);
 719         ssize_t error = -EBADF;
 720
 721         if (!f.file)
 722                 return error;
 723         audit_file(f.file);
 724         error = getxattr(file_mnt_user_ns(f.file), f.file->f_path.dentry,
 725                          name, value, size);
 726         fdput(f);
 727         return error;
 728 }
 729
 730 /*
 731  * Extended attribute LIST operations
 732  */
 733 static ssize_t
 734 listxattr(struct dentry *d, char __user *list, size_t size)
 735 {
 736         ssize_t error;
 737         char *klist = NULL;
 738
 739         if (size) {
 740                 if (size > XATTR_LIST_MAX)
 741                         size = XATTR_LIST_MAX;
 742                 klist = kvmalloc(size, GFP_KERNEL);
 743                 if (!klist)
 744                         return -ENOMEM;
 745         }
 746
 747         error = vfs_listxattr(d, klist, size);
 748         if (error > 0) {
 749                 if (size && copy_to_user(list, klist, error))
 750                         error = -EFAULT;
 751         } else if (error == -ERANGE && size >= XATTR_LIST_MAX) {
 752                 /* The file system tried to returned a list bigger
 753                    than XATTR_LIST_MAX bytes. Not possible. */
 754                 error = -E2BIG;
 755         }
 756
 757         kvfree(klist);
 758
 759         return error;
 760 }
 761
 762 static ssize_t path_listxattr(const char __user *pathname, char __user *list,
 763                               size_t size, unsigned int lookup_flags)
 764 {
 765         struct path path;
 766         ssize_t error;
 767 retry:
 768         error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
 769         if (error)
 770                 return error;
 771         error = listxattr(path.dentry, list, size);
 772         path_put(&path);
 773         if (retry_estale(error, lookup_flags)) {
 774                 lookup_flags |= LOOKUP_REVAL;
 775                 goto retry;
 776         }
 777         return error;
 778 }
 779
 780 SYSCALL_DEFINE3(listxattr, const char __user *, pathname, char __user *, list,
 781                 size_t, size)
 782 {
 783         return path_listxattr(pathname, list, size, LOOKUP_FOLLOW);
 784 }
 785
 786 SYSCALL_DEFINE3(llistxattr, const char __user *, pathname, char __user *, list,
 787                 size_t, size)
 788 {
 789         return path_listxattr(pathname, list, size, 0);
 790 }
 791
 792 SYSCALL_DEFINE3(flistxattr, int, fd, char __user *, list, size_t, size)
 793 {
 794         struct fd f = fdget(fd);
 795         ssize_t error = -EBADF;
 796
 797         if (!f.file)
 798                 return error;
 799         audit_file(f.file);
 800         error = listxattr(f.file->f_path.dentry, list, size);
 801         fdput(f);
 802         return error;
 803 }
 804
 805 /*
 806  * Extended attribute REMOVE operations
 807  */
 808 static long
 809 removexattr(struct user_namespace *mnt_userns, struct dentry *d,
 810             const char __user *name)
 811 {
 812         int error;
 813         char kname[XATTR_NAME_MAX + 1];
 814
 815         error = strncpy_from_user(kname, name, sizeof(kname));
 816         if (error == 0 || error == sizeof(kname))
 817                 error = -ERANGE;
 818         if (error < 0)
 819                 return error;
 820
 821         return vfs_removexattr(mnt_userns, d, kname);
 822 }
 823
 824 static int path_removexattr(const char __user *pathname,
 825                             const char __user *name, unsigned int lookup_flags)
 826 {
 827         struct path path;
 828         int error;
 829 retry:
 830         error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
 831         if (error)
 832                 return error;
 833         error = mnt_want_write(path.mnt);
 834         if (!error) {
 835                 error = removexattr(mnt_user_ns(path.mnt), path.dentry, name);
 836                 mnt_drop_write(path.mnt);
 837         }
 838         path_put(&path);
 839         if (retry_estale(error, lookup_flags)) {
 840                 lookup_flags |= LOOKUP_REVAL;
 841                 goto retry;
 842         }
 843         return error;
 844 }
 845
 846 SYSCALL_DEFINE2(removexattr, const char __user *, pathname,
 847                 const char __user *, name)
 848 {
 849         return path_removexattr(pathname, name, LOOKUP_FOLLOW);
 850 }
 851
 852 SYSCALL_DEFINE2(lremovexattr, const char __user *, pathname,
 853                 const char __user *, name)
 854 {
 855         return path_removexattr(pathname, name, 0);
 856 }
 857
 858 SYSCALL_DEFINE2(fremovexattr, int, fd, const char __user *, name)
 859 {
 860         struct fd f = fdget(fd);
 861         int error = -EBADF;
 862
 863         if (!f.file)
 864                 return error;
 865         audit_file(f.file);
 866         error = mnt_want_write_file(f.file);
 867         if (!error) {
 868                 error = removexattr(file_mnt_user_ns(f.file),
 869                                     f.file->f_path.dentry, name);
 870                 mnt_drop_write_file(f.file);
 871         }
 872         fdput(f);
 873         return error;
 874 }
 875
 876 /*
 877  * Combine the results of the list() operation from every xattr_handler in the
 878  * list.
 879  */
 880 ssize_t
 881 generic_listxattr(struct dentry *dentry, char *buffer, size_t buffer_size)
 882 {
 883         const struct xattr_handler *handler, **handlers = dentry->d_sb->s_xattr;
 884         unsigned int size = 0;
 885
 886         if (!buffer) {
 887                 for_each_xattr_handler(handlers, handler) {
 888                         if (!handler->name ||
 889                             (handler->list && !handler->list(dentry)))
 890                                 continue;
 891                         size += strlen(handler->name) + 1;
 892                 }
 893         } else {
 894                 char *buf = buffer;
 895                 size_t len;
 896
 897                 for_each_xattr_handler(handlers, handler) {
 898                         if (!handler->name ||
 899                             (handler->list && !handler->list(dentry)))
 900                                 continue;
 901                         len = strlen(handler->name);
 902                         if (len + 1 > buffer_size)
 903                                 return -ERANGE;
 904                         memcpy(buf, handler->name, len + 1);
 905                         buf += len + 1;
 906                         buffer_size -= len + 1;
 907                 }
 908                 size = buf - buffer;
 909         }
 910         return size;
 911 }
 912 EXPORT_SYMBOL(generic_listxattr);
 913
 914 /**
 915  * xattr_full_name  -  Compute full attribute name from suffix
 916  *
 917  * @handler:    handler of the xattr_handler operation
 918  * @name:       name passed to the xattr_handler operation
 919  *
 920  * The get and set xattr handler operations are called with the remainder of
 921  * the attribute name after skipping the handler's prefix: for example, "foo"
 922  * is passed to the get operation of a handler with prefix "user." to get
 923  * attribute "user.foo".  The full name is still "there" in the name though.
 924  *
 925  * Note: the list xattr handler operation when called from the vfs is passed a
 926  * NULL name; some file systems use this operation internally, with varying
 927  * semantics.
 928  */
 929 const char *xattr_full_name(const struct xattr_handler *handler,
 930                             const char *name)
 931 {
 932         size_t prefix_len = strlen(xattr_prefix(handler));
 933
 934         return name - prefix_len;
 935 }
 936 EXPORT_SYMBOL(xattr_full_name);
 937
 938 /*
 939  * Allocate new xattr and copy in the value; but leave the name to callers.
 940  */
 941 struct simple_xattr *simple_xattr_alloc(const void *value, size_t size)
 942 {
 943         struct simple_xattr *new_xattr;
 944         size_t len;
 945
 946         /* wrap around? */
 947         len = sizeof(*new_xattr) + size;
 948         if (len < sizeof(*new_xattr))
 949                 return NULL;
 950
 951         new_xattr = kvmalloc(len, GFP_KERNEL);
 952         if (!new_xattr)
 953                 return NULL;
 954
 955         new_xattr->size = size;
 956         memcpy(new_xattr->value, value, size);
 957         return new_xattr;
 958 }
 959
 960 /*
 961  * xattr GET operation for in-memory/pseudo filesystems
 962  */
 963 int simple_xattr_get(struct simple_xattrs *xattrs, const char *name,
 964                      void *buffer, size_t size)
 965 {
 966         struct simple_xattr *xattr;
 967         int ret = -ENODATA;
 968
 969         spin_lock(&xattrs->lock);
 970         list_for_each_entry(xattr, &xattrs->head, list) {
 971                 if (strcmp(name, xattr->name))
 972                         continue;
 973
 974                 ret = xattr->size;
 975                 if (buffer) {
 976                         if (size < xattr->size)
 977                                 ret = -ERANGE;
 978                         else
 979                                 memcpy(buffer, xattr->value, xattr->size);
 980                 }
 981                 break;
 982         }
 983         spin_unlock(&xattrs->lock);
 984         return ret;
 985 }
 986
 987 /**
 988  * simple_xattr_set - xattr SET operation for in-memory/pseudo filesystems
 989  * @xattrs: target simple_xattr list
 990  * @name: name of the extended attribute
 991  * @value: value of the xattr. If %NULL, will remove the attribute.
 992  * @size: size of the new xattr
 993  * @flags: %XATTR_{CREATE|REPLACE}
 994  * @removed_size: returns size of the removed xattr, -1 if none removed
 995  *
 996  * %XATTR_CREATE is set, the xattr shouldn't exist already; otherwise fails
 997  * with -EEXIST.  If %XATTR_REPLACE is set, the xattr should exist;
 998  * otherwise, fails with -ENODATA.
 999  *
1000  * Returns 0 on success, -errno on failure.
1001  */
1002 int simple_xattr_set(struct simple_xattrs *xattrs, const char *name,
1003                      const void *value, size_t size, int flags,
1004                      ssize_t *removed_size)
1005 {
1006         struct simple_xattr *xattr;
1007         struct simple_xattr *new_xattr = NULL;
1008         int err = 0;
1009
1010         if (removed_size)
1011                 *removed_size = -1;
1012
1013         /* value == NULL means remove */
1014         if (value) {
1015                 new_xattr = simple_xattr_alloc(value, size);
1016                 if (!new_xattr)
1017                         return -ENOMEM;
1018
1019                 new_xattr->name = kstrdup(name, GFP_KERNEL);
1020                 if (!new_xattr->name) {
1021                         kvfree(new_xattr);
1022                         return -ENOMEM;
1023                 }
1024         }
1025
1026         spin_lock(&xattrs->lock);
1027         list_for_each_entry(xattr, &xattrs->head, list) {
1028                 if (!strcmp(name, xattr->name)) {
1029                         if (flags & XATTR_CREATE) {
1030                                 xattr = new_xattr;
1031                                 err = -EEXIST;
1032                         } else if (new_xattr) {
1033                                 list_replace(&xattr->list, &new_xattr->list);
1034                                 if (removed_size)
1035                                         *removed_size = xattr->size;
1036                         } else {
1037                                 list_del(&xattr->list);
1038                                 if (removed_size)
1039                                         *removed_size = xattr->size;
1040                         }
1041                         goto out;
1042                 }
1043         }
1044         if (flags & XATTR_REPLACE) {
1045                 xattr = new_xattr;
1046                 err = -ENODATA;
1047         } else {
1048                 list_add(&new_xattr->list, &xattrs->head);
1049                 xattr = NULL;
1050         }
1051 out:
1052         spin_unlock(&xattrs->lock);
1053         if (xattr) {
1054                 kfree(xattr->name);
1055                 kvfree(xattr);
1056         }
1057         return err;
1058
1059 }
1060
1061 static bool xattr_is_trusted(const char *name)
1062 {
1063         return !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN);
1064 }
1065
1066 static int xattr_list_one(char **buffer, ssize_t *remaining_size,
1067                           const char *name)
1068 {
1069         size_t len = strlen(name) + 1;
1070         if (*buffer) {
1071                 if (*remaining_size < len)
1072                         return -ERANGE;
1073                 memcpy(*buffer, name, len);
1074                 *buffer += len;
1075         }
1076         *remaining_size -= len;
1077         return 0;
1078 }
1079
1080 /*
1081  * xattr LIST operation for in-memory/pseudo filesystems
1082  */
1083 ssize_t simple_xattr_list(struct inode *inode, struct simple_xattrs *xattrs,
1084                           char *buffer, size_t size)
1085 {
1086         bool trusted = capable(CAP_SYS_ADMIN);
1087         struct simple_xattr *xattr;
1088         ssize_t remaining_size = size;
1089         int err = 0;
1090
1091 #ifdef CONFIG_FS_POSIX_ACL
1092         if (IS_POSIXACL(inode)) {
1093                 if (inode->i_acl) {
1094                         err = xattr_list_one(&buffer, &remaining_size,
1095                                              XATTR_NAME_POSIX_ACL_ACCESS);
1096                         if (err)
1097                                 return err;
1098                 }
1099                 if (inode->i_default_acl) {
1100                         err = xattr_list_one(&buffer, &remaining_size,
1101                                              XATTR_NAME_POSIX_ACL_DEFAULT);
1102                         if (err)
1103                                 return err;
1104                 }
1105         }
1106 #endif
1107
1108         spin_lock(&xattrs->lock);
1109         list_for_each_entry(xattr, &xattrs->head, list) {
1110                 /* skip "trusted." attributes for unprivileged callers */
1111                 if (!trusted && xattr_is_trusted(xattr->name))
1112                         continue;
1113
1114                 err = xattr_list_one(&buffer, &remaining_size, xattr->name);
1115                 if (err)
1116                         break;
1117         }
1118         spin_unlock(&xattrs->lock);
1119
1120         return err ? err : size - remaining_size;
1121 }
1122
1123 /*
1124  * Adds an extended attribute to the list
1125  */
1126 void simple_xattr_list_add(struct simple_xattrs *xattrs,
1127                            struct simple_xattr *new_xattr)
1128 {
1129         spin_lock(&xattrs->lock);
1130         list_add(&new_xattr->list, &xattrs->head);
1131         spin_unlock(&xattrs->lock);
1132 }