Linux Audio

Check our new training course

Loading...
v6.8
   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/filelock.h>
  13#include <linux/slab.h>
  14#include <linux/file.h>
  15#include <linux/xattr.h>
  16#include <linux/mount.h>
  17#include <linux/namei.h>
  18#include <linux/security.h>
  19#include <linux/evm.h>
  20#include <linux/syscalls.h>
  21#include <linux/export.h>
  22#include <linux/fsnotify.h>
  23#include <linux/audit.h>
  24#include <linux/vmalloc.h>
  25#include <linux/posix_acl_xattr.h>
  26
  27#include <linux/uaccess.h>
  28
  29#include "internal.h"
  30
  31static const char *
  32strcmp_prefix(const char *a, const char *a_prefix)
  33{
  34	while (*a_prefix && *a == *a_prefix) {
  35		a++;
  36		a_prefix++;
  37	}
  38	return *a_prefix ? NULL : a;
  39}
  40
  41/*
  42 * In order to implement different sets of xattr operations for each xattr
  43 * prefix, a filesystem should create a null-terminated array of struct
  44 * xattr_handler (one for each prefix) and hang a pointer to it off of the
  45 * s_xattr field of the superblock.
  46 */
  47#define for_each_xattr_handler(handlers, handler)		\
  48	if (handlers)						\
  49		for ((handler) = *(handlers)++;			\
  50			(handler) != NULL;			\
  51			(handler) = *(handlers)++)
  52
  53/*
  54 * Find the xattr_handler with the matching prefix.
  55 */
  56static const struct xattr_handler *
  57xattr_resolve_name(struct inode *inode, const char **name)
  58{
  59	const struct xattr_handler * const *handlers = inode->i_sb->s_xattr;
  60	const struct xattr_handler *handler;
  61
  62	if (!(inode->i_opflags & IOP_XATTR)) {
  63		if (unlikely(is_bad_inode(inode)))
  64			return ERR_PTR(-EIO);
  65		return ERR_PTR(-EOPNOTSUPP);
  66	}
  67	for_each_xattr_handler(handlers, handler) {
  68		const char *n;
  69
  70		n = strcmp_prefix(*name, xattr_prefix(handler));
  71		if (n) {
  72			if (!handler->prefix ^ !*n) {
  73				if (*n)
  74					continue;
  75				return ERR_PTR(-EINVAL);
  76			}
  77			*name = n;
  78			return handler;
  79		}
  80	}
  81	return ERR_PTR(-EOPNOTSUPP);
  82}
  83
  84/**
  85 * may_write_xattr - check whether inode allows writing xattr
  86 * @idmap: idmap of the mount the inode was found from
  87 * @inode: the inode on which to set an xattr
  88 *
  89 * Check whether the inode allows writing xattrs. Specifically, we can never
  90 * set or remove an extended attribute on a read-only filesystem  or on an
  91 * immutable / append-only inode.
  92 *
  93 * We also need to ensure that the inode has a mapping in the mount to
  94 * not risk writing back invalid i_{g,u}id values.
  95 *
  96 * Return: On success zero is returned. On error a negative errno is returned.
  97 */
  98int may_write_xattr(struct mnt_idmap *idmap, struct inode *inode)
  99{
 100	if (IS_IMMUTABLE(inode))
 101		return -EPERM;
 102	if (IS_APPEND(inode))
 103		return -EPERM;
 104	if (HAS_UNMAPPED_ID(idmap, inode))
 105		return -EPERM;
 106	return 0;
 107}
 108
 109/*
 110 * Check permissions for extended attribute access.  This is a bit complicated
 111 * because different namespaces have very different rules.
 112 */
 113static int
 114xattr_permission(struct mnt_idmap *idmap, struct inode *inode,
 115		 const char *name, int mask)
 116{
 117	if (mask & MAY_WRITE) {
 118		int ret;
 119
 120		ret = may_write_xattr(idmap, inode);
 121		if (ret)
 122			return ret;
 123	}
 124
 125	/*
 126	 * No restriction for security.* and system.* from the VFS.  Decision
 127	 * on these is left to the underlying filesystem / security module.
 128	 */
 129	if (!strncmp(name, XATTR_SECURITY_PREFIX, XATTR_SECURITY_PREFIX_LEN) ||
 130	    !strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN))
 131		return 0;
 132
 133	/*
 134	 * The trusted.* namespace can only be accessed by privileged users.
 135	 */
 136	if (!strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN)) {
 137		if (!capable(CAP_SYS_ADMIN))
 138			return (mask & MAY_WRITE) ? -EPERM : -ENODATA;
 139		return 0;
 140	}
 141
 142	/*
 143	 * In the user.* namespace, only regular files and directories can have
 144	 * extended attributes. For sticky directories, only the owner and
 145	 * privileged users can write attributes.
 146	 */
 147	if (!strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN)) {
 148		if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
 149			return (mask & MAY_WRITE) ? -EPERM : -ENODATA;
 150		if (S_ISDIR(inode->i_mode) && (inode->i_mode & S_ISVTX) &&
 151		    (mask & MAY_WRITE) &&
 152		    !inode_owner_or_capable(idmap, inode))
 153			return -EPERM;
 154	}
 155
 156	return inode_permission(idmap, inode, mask);
 157}
 158
 159/*
 160 * Look for any handler that deals with the specified namespace.
 161 */
 162int
 163xattr_supports_user_prefix(struct inode *inode)
 164{
 165	const struct xattr_handler * const *handlers = inode->i_sb->s_xattr;
 166	const struct xattr_handler *handler;
 167
 168	if (!(inode->i_opflags & IOP_XATTR)) {
 169		if (unlikely(is_bad_inode(inode)))
 170			return -EIO;
 171		return -EOPNOTSUPP;
 172	}
 173
 174	for_each_xattr_handler(handlers, handler) {
 175		if (!strncmp(xattr_prefix(handler), XATTR_USER_PREFIX,
 176			     XATTR_USER_PREFIX_LEN))
 177			return 0;
 178	}
 179
 180	return -EOPNOTSUPP;
 181}
 182EXPORT_SYMBOL(xattr_supports_user_prefix);
 183
 184int
 185__vfs_setxattr(struct mnt_idmap *idmap, struct dentry *dentry,
 186	       struct inode *inode, const char *name, const void *value,
 187	       size_t size, int flags)
 188{
 189	const struct xattr_handler *handler;
 190
 191	if (is_posix_acl_xattr(name))
 192		return -EOPNOTSUPP;
 193
 194	handler = xattr_resolve_name(inode, &name);
 195	if (IS_ERR(handler))
 196		return PTR_ERR(handler);
 197	if (!handler->set)
 198		return -EOPNOTSUPP;
 199	if (size == 0)
 200		value = "";  /* empty EA, do not remove */
 201	return handler->set(handler, idmap, dentry, inode, name, value,
 202			    size, flags);
 203}
 204EXPORT_SYMBOL(__vfs_setxattr);
 205
 206/**
 207 *  __vfs_setxattr_noperm - perform setxattr operation without performing
 208 *  permission checks.
 209 *
 210 *  @idmap: idmap of the mount the inode was found from
 211 *  @dentry: object to perform setxattr on
 212 *  @name: xattr name to set
 213 *  @value: value to set @name to
 214 *  @size: size of @value
 215 *  @flags: flags to pass into filesystem operations
 216 *
 217 *  returns the result of the internal setxattr or setsecurity operations.
 218 *
 219 *  This function requires the caller to lock the inode's i_mutex before it
 220 *  is executed. It also assumes that the caller will make the appropriate
 221 *  permission checks.
 222 */
 223int __vfs_setxattr_noperm(struct mnt_idmap *idmap,
 224			  struct dentry *dentry, const char *name,
 225			  const void *value, size_t size, int flags)
 226{
 227	struct inode *inode = dentry->d_inode;
 228	int error = -EAGAIN;
 229	int issec = !strncmp(name, XATTR_SECURITY_PREFIX,
 230				   XATTR_SECURITY_PREFIX_LEN);
 231
 232	if (issec)
 233		inode->i_flags &= ~S_NOSEC;
 234	if (inode->i_opflags & IOP_XATTR) {
 235		error = __vfs_setxattr(idmap, dentry, inode, name, value,
 236				       size, flags);
 237		if (!error) {
 238			fsnotify_xattr(dentry);
 239			security_inode_post_setxattr(dentry, name, value,
 240						     size, flags);
 241		}
 242	} else {
 243		if (unlikely(is_bad_inode(inode)))
 244			return -EIO;
 245	}
 246	if (error == -EAGAIN) {
 247		error = -EOPNOTSUPP;
 248
 249		if (issec) {
 250			const char *suffix = name + XATTR_SECURITY_PREFIX_LEN;
 251
 252			error = security_inode_setsecurity(inode, suffix, value,
 253							   size, flags);
 254			if (!error)
 255				fsnotify_xattr(dentry);
 256		}
 257	}
 258
 259	return error;
 260}
 261
 262/**
 263 * __vfs_setxattr_locked - set an extended attribute while holding the inode
 264 * lock
 265 *
 266 *  @idmap: idmap of the mount of the target inode
 267 *  @dentry: object to perform setxattr on
 268 *  @name: xattr name to set
 269 *  @value: value to set @name to
 270 *  @size: size of @value
 271 *  @flags: flags to pass into filesystem operations
 272 *  @delegated_inode: on return, will contain an inode pointer that
 273 *  a delegation was broken on, NULL if none.
 274 */
 275int
 276__vfs_setxattr_locked(struct mnt_idmap *idmap, struct dentry *dentry,
 277		      const char *name, const void *value, size_t size,
 278		      int flags, struct inode **delegated_inode)
 279{
 280	struct inode *inode = dentry->d_inode;
 281	int error;
 282
 283	error = xattr_permission(idmap, inode, name, MAY_WRITE);
 284	if (error)
 285		return error;
 286
 287	error = security_inode_setxattr(idmap, dentry, name, value, size,
 288					flags);
 289	if (error)
 290		goto out;
 291
 292	error = try_break_deleg(inode, delegated_inode);
 293	if (error)
 294		goto out;
 295
 296	error = __vfs_setxattr_noperm(idmap, dentry, name, value,
 297				      size, flags);
 298
 299out:
 300	return error;
 301}
 302EXPORT_SYMBOL_GPL(__vfs_setxattr_locked);
 303
 304int
 305vfs_setxattr(struct mnt_idmap *idmap, struct dentry *dentry,
 306	     const char *name, const void *value, size_t size, int flags)
 307{
 308	struct inode *inode = dentry->d_inode;
 309	struct inode *delegated_inode = NULL;
 310	const void  *orig_value = value;
 311	int error;
 312
 313	if (size && strcmp(name, XATTR_NAME_CAPS) == 0) {
 314		error = cap_convert_nscap(idmap, dentry, &value, size);
 315		if (error < 0)
 316			return error;
 317		size = error;
 318	}
 319
 320retry_deleg:
 321	inode_lock(inode);
 322	error = __vfs_setxattr_locked(idmap, dentry, name, value, size,
 323				      flags, &delegated_inode);
 324	inode_unlock(inode);
 325
 326	if (delegated_inode) {
 327		error = break_deleg_wait(&delegated_inode);
 328		if (!error)
 329			goto retry_deleg;
 330	}
 331	if (value != orig_value)
 332		kfree(value);
 333
 334	return error;
 335}
 336EXPORT_SYMBOL_GPL(vfs_setxattr);
 337
 338static ssize_t
 339xattr_getsecurity(struct mnt_idmap *idmap, struct inode *inode,
 340		  const char *name, void *value, size_t size)
 341{
 342	void *buffer = NULL;
 343	ssize_t len;
 344
 345	if (!value || !size) {
 346		len = security_inode_getsecurity(idmap, inode, name,
 347						 &buffer, false);
 348		goto out_noalloc;
 349	}
 350
 351	len = security_inode_getsecurity(idmap, inode, name, &buffer,
 352					 true);
 353	if (len < 0)
 354		return len;
 355	if (size < len) {
 356		len = -ERANGE;
 357		goto out;
 358	}
 359	memcpy(value, buffer, len);
 360out:
 361	kfree(buffer);
 362out_noalloc:
 363	return len;
 364}
 365
 366/*
 367 * vfs_getxattr_alloc - allocate memory, if necessary, before calling getxattr
 368 *
 369 * Allocate memory, if not already allocated, or re-allocate correct size,
 370 * before retrieving the extended attribute.  The xattr value buffer should
 371 * always be freed by the caller, even on error.
 372 *
 373 * Returns the result of alloc, if failed, or the getxattr operation.
 374 */
 375int
 376vfs_getxattr_alloc(struct mnt_idmap *idmap, struct dentry *dentry,
 377		   const char *name, char **xattr_value, size_t xattr_size,
 378		   gfp_t flags)
 379{
 380	const struct xattr_handler *handler;
 381	struct inode *inode = dentry->d_inode;
 382	char *value = *xattr_value;
 383	int error;
 384
 385	error = xattr_permission(idmap, inode, name, MAY_READ);
 386	if (error)
 387		return error;
 388
 389	handler = xattr_resolve_name(inode, &name);
 390	if (IS_ERR(handler))
 391		return PTR_ERR(handler);
 392	if (!handler->get)
 393		return -EOPNOTSUPP;
 394	error = handler->get(handler, dentry, inode, name, NULL, 0);
 395	if (error < 0)
 396		return error;
 397
 398	if (!value || (error > xattr_size)) {
 399		value = krealloc(*xattr_value, error + 1, flags);
 400		if (!value)
 401			return -ENOMEM;
 402		memset(value, 0, error + 1);
 403	}
 404
 405	error = handler->get(handler, dentry, inode, name, value, error);
 406	*xattr_value = value;
 407	return error;
 408}
 409
 410ssize_t
 411__vfs_getxattr(struct dentry *dentry, struct inode *inode, const char *name,
 412	       void *value, size_t size)
 413{
 414	const struct xattr_handler *handler;
 415
 416	if (is_posix_acl_xattr(name))
 417		return -EOPNOTSUPP;
 418
 419	handler = xattr_resolve_name(inode, &name);
 420	if (IS_ERR(handler))
 421		return PTR_ERR(handler);
 422	if (!handler->get)
 423		return -EOPNOTSUPP;
 424	return handler->get(handler, dentry, inode, name, value, size);
 425}
 426EXPORT_SYMBOL(__vfs_getxattr);
 427
 428ssize_t
 429vfs_getxattr(struct mnt_idmap *idmap, struct dentry *dentry,
 430	     const char *name, void *value, size_t size)
 431{
 432	struct inode *inode = dentry->d_inode;
 433	int error;
 434
 435	error = xattr_permission(idmap, inode, name, MAY_READ);
 436	if (error)
 437		return error;
 438
 439	error = security_inode_getxattr(dentry, name);
 440	if (error)
 441		return error;
 442
 443	if (!strncmp(name, XATTR_SECURITY_PREFIX,
 444				XATTR_SECURITY_PREFIX_LEN)) {
 445		const char *suffix = name + XATTR_SECURITY_PREFIX_LEN;
 446		int ret = xattr_getsecurity(idmap, inode, suffix, value,
 447					    size);
 448		/*
 449		 * Only overwrite the return value if a security module
 450		 * is actually active.
 451		 */
 452		if (ret == -EOPNOTSUPP)
 453			goto nolsm;
 454		return ret;
 455	}
 456nolsm:
 457	return __vfs_getxattr(dentry, inode, name, value, size);
 458}
 459EXPORT_SYMBOL_GPL(vfs_getxattr);
 460
 461/**
 462 * vfs_listxattr - retrieve \0 separated list of xattr names
 463 * @dentry: the dentry from whose inode the xattr names are retrieved
 464 * @list: buffer to store xattr names into
 465 * @size: size of the buffer
 466 *
 467 * This function returns the names of all xattrs associated with the
 468 * inode of @dentry.
 469 *
 470 * Note, for legacy reasons the vfs_listxattr() function lists POSIX
 471 * ACLs as well. Since POSIX ACLs are decoupled from IOP_XATTR the
 472 * vfs_listxattr() function doesn't check for this flag since a
 473 * filesystem could implement POSIX ACLs without implementing any other
 474 * xattrs.
 475 *
 476 * However, since all codepaths that remove IOP_XATTR also assign of
 477 * inode operations that either don't implement or implement a stub
 478 * ->listxattr() operation.
 479 *
 480 * Return: On success, the size of the buffer that was used. On error a
 481 *         negative error code.
 482 */
 483ssize_t
 484vfs_listxattr(struct dentry *dentry, char *list, size_t size)
 485{
 486	struct inode *inode = d_inode(dentry);
 487	ssize_t error;
 488
 489	error = security_inode_listxattr(dentry);
 490	if (error)
 491		return error;
 492
 493	if (inode->i_op->listxattr) {
 494		error = inode->i_op->listxattr(dentry, list, size);
 495	} else {
 496		error = security_inode_listsecurity(inode, list, size);
 497		if (size && error > size)
 498			error = -ERANGE;
 499	}
 500	return error;
 501}
 502EXPORT_SYMBOL_GPL(vfs_listxattr);
 503
 504int
 505__vfs_removexattr(struct mnt_idmap *idmap, struct dentry *dentry,
 506		  const char *name)
 507{
 508	struct inode *inode = d_inode(dentry);
 509	const struct xattr_handler *handler;
 510
 511	if (is_posix_acl_xattr(name))
 512		return -EOPNOTSUPP;
 513
 514	handler = xattr_resolve_name(inode, &name);
 515	if (IS_ERR(handler))
 516		return PTR_ERR(handler);
 517	if (!handler->set)
 518		return -EOPNOTSUPP;
 519	return handler->set(handler, idmap, dentry, inode, name, NULL, 0,
 520			    XATTR_REPLACE);
 521}
 522EXPORT_SYMBOL(__vfs_removexattr);
 523
 524/**
 525 * __vfs_removexattr_locked - set an extended attribute while holding the inode
 526 * lock
 527 *
 528 *  @idmap: idmap of the mount of the target inode
 529 *  @dentry: object to perform setxattr on
 530 *  @name: name of xattr to remove
 531 *  @delegated_inode: on return, will contain an inode pointer that
 532 *  a delegation was broken on, NULL if none.
 533 */
 534int
 535__vfs_removexattr_locked(struct mnt_idmap *idmap,
 536			 struct dentry *dentry, const char *name,
 537			 struct inode **delegated_inode)
 538{
 539	struct inode *inode = dentry->d_inode;
 540	int error;
 541
 542	error = xattr_permission(idmap, inode, name, MAY_WRITE);
 543	if (error)
 544		return error;
 545
 546	error = security_inode_removexattr(idmap, dentry, name);
 547	if (error)
 548		goto out;
 549
 550	error = try_break_deleg(inode, delegated_inode);
 551	if (error)
 552		goto out;
 553
 554	error = __vfs_removexattr(idmap, dentry, name);
 
 
 555
 556	if (!error) {
 557		fsnotify_xattr(dentry);
 558		evm_inode_post_removexattr(dentry, name);
 559	}
 560
 561out:
 562	return error;
 563}
 564EXPORT_SYMBOL_GPL(__vfs_removexattr_locked);
 565
 566int
 567vfs_removexattr(struct mnt_idmap *idmap, struct dentry *dentry,
 568		const char *name)
 569{
 570	struct inode *inode = dentry->d_inode;
 571	struct inode *delegated_inode = NULL;
 572	int error;
 573
 574retry_deleg:
 575	inode_lock(inode);
 576	error = __vfs_removexattr_locked(idmap, dentry,
 577					 name, &delegated_inode);
 578	inode_unlock(inode);
 579
 580	if (delegated_inode) {
 581		error = break_deleg_wait(&delegated_inode);
 582		if (!error)
 583			goto retry_deleg;
 584	}
 585
 586	return error;
 587}
 588EXPORT_SYMBOL_GPL(vfs_removexattr);
 589
 
 
 
 
 
 
 
 
 
 
 
 590/*
 591 * Extended attribute SET operations
 592 */
 593
 594int setxattr_copy(const char __user *name, struct xattr_ctx *ctx)
 595{
 596	int error;
 597
 598	if (ctx->flags & ~(XATTR_CREATE|XATTR_REPLACE))
 599		return -EINVAL;
 600
 601	error = strncpy_from_user(ctx->kname->name, name,
 602				sizeof(ctx->kname->name));
 603	if (error == 0 || error == sizeof(ctx->kname->name))
 604		return  -ERANGE;
 605	if (error < 0)
 606		return error;
 607
 608	error = 0;
 609	if (ctx->size) {
 610		if (ctx->size > XATTR_SIZE_MAX)
 611			return -E2BIG;
 612
 613		ctx->kvalue = vmemdup_user(ctx->cvalue, ctx->size);
 614		if (IS_ERR(ctx->kvalue)) {
 615			error = PTR_ERR(ctx->kvalue);
 616			ctx->kvalue = NULL;
 617		}
 618	}
 619
 620	return error;
 621}
 622
 623int do_setxattr(struct mnt_idmap *idmap, struct dentry *dentry,
 624		struct xattr_ctx *ctx)
 625{
 626	if (is_posix_acl_xattr(ctx->kname->name))
 627		return do_set_acl(idmap, dentry, ctx->kname->name,
 628				  ctx->kvalue, ctx->size);
 629
 630	return vfs_setxattr(idmap, dentry, ctx->kname->name,
 631			ctx->kvalue, ctx->size, ctx->flags);
 632}
 633
 634static long
 635setxattr(struct mnt_idmap *idmap, struct dentry *d,
 636	const char __user *name, const void __user *value, size_t size,
 637	int flags)
 638{
 639	struct xattr_name kname;
 640	struct xattr_ctx ctx = {
 641		.cvalue   = value,
 642		.kvalue   = NULL,
 643		.size     = size,
 644		.kname    = &kname,
 645		.flags    = flags,
 646	};
 647	int error;
 648
 649	error = setxattr_copy(name, &ctx);
 650	if (error)
 651		return error;
 652
 653	error = do_setxattr(idmap, d, &ctx);
 654
 655	kvfree(ctx.kvalue);
 
 
 656	return error;
 657}
 658
 659static int path_setxattr(const char __user *pathname,
 660			 const char __user *name, const void __user *value,
 661			 size_t size, int flags, unsigned int lookup_flags)
 662{
 663	struct path path;
 664	int error;
 665
 666retry:
 667	error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
 668	if (error)
 669		return error;
 670	error = mnt_want_write(path.mnt);
 671	if (!error) {
 672		error = setxattr(mnt_idmap(path.mnt), path.dentry, name,
 673				 value, size, flags);
 674		mnt_drop_write(path.mnt);
 675	}
 676	path_put(&path);
 677	if (retry_estale(error, lookup_flags)) {
 678		lookup_flags |= LOOKUP_REVAL;
 679		goto retry;
 680	}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 681	return error;
 682}
 683
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 684SYSCALL_DEFINE5(setxattr, const char __user *, pathname,
 685		const char __user *, name, const void __user *, value,
 686		size_t, size, int, flags)
 687{
 688	return path_setxattr(pathname, name, value, size, flags, LOOKUP_FOLLOW);
 689}
 690
 691SYSCALL_DEFINE5(lsetxattr, const char __user *, pathname,
 692		const char __user *, name, const void __user *, value,
 693		size_t, size, int, flags)
 694{
 695	return path_setxattr(pathname, name, value, size, flags, 0);
 
 696}
 697
 698SYSCALL_DEFINE5(fsetxattr, int, fd, const char __user *, name,
 699		const void __user *,value, size_t, size, int, flags)
 700{
 701	struct fd f = fdget(fd);
 702	int error = -EBADF;
 703
 704	if (!f.file)
 705		return error;
 706	audit_file(f.file);
 707	error = mnt_want_write_file(f.file);
 708	if (!error) {
 709		error = setxattr(file_mnt_idmap(f.file),
 710				 f.file->f_path.dentry, name,
 711				 value, size, flags);
 712		mnt_drop_write_file(f.file);
 713	}
 714	fdput(f);
 715	return error;
 716}
 717
 718/*
 719 * Extended attribute GET operations
 720 */
 721ssize_t
 722do_getxattr(struct mnt_idmap *idmap, struct dentry *d,
 723	struct xattr_ctx *ctx)
 724{
 725	ssize_t error;
 726	char *kname = ctx->kname->name;
 
 727
 728	if (ctx->size) {
 729		if (ctx->size > XATTR_SIZE_MAX)
 730			ctx->size = XATTR_SIZE_MAX;
 731		ctx->kvalue = kvzalloc(ctx->size, GFP_KERNEL);
 732		if (!ctx->kvalue)
 733			return -ENOMEM;
 734	}
 735
 736	if (is_posix_acl_xattr(ctx->kname->name))
 737		error = do_get_acl(idmap, d, kname, ctx->kvalue, ctx->size);
 738	else
 739		error = vfs_getxattr(idmap, d, kname, ctx->kvalue, ctx->size);
 740	if (error > 0) {
 741		if (ctx->size && copy_to_user(ctx->value, ctx->kvalue, error))
 742			error = -EFAULT;
 743	} else if (error == -ERANGE && ctx->size >= XATTR_SIZE_MAX) {
 744		/* The file system tried to returned a value bigger
 745		   than XATTR_SIZE_MAX bytes. Not possible. */
 746		error = -E2BIG;
 747	}
 748
 
 749	return error;
 750}
 751
 752static ssize_t
 753getxattr(struct mnt_idmap *idmap, struct dentry *d,
 754	 const char __user *name, void __user *value, size_t size)
 755{
 
 
 
 
 
 
 
 
 
 756	ssize_t error;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 757	struct xattr_name kname;
 758	struct xattr_ctx ctx = {
 759		.value    = value,
 760		.kvalue   = NULL,
 761		.size     = size,
 762		.kname    = &kname,
 763		.flags    = 0,
 764	};
 
 
 765
 766	error = strncpy_from_user(kname.name, name, sizeof(kname.name));
 767	if (error == 0 || error == sizeof(kname.name))
 768		error = -ERANGE;
 769	if (error < 0)
 770		return error;
 771
 772	error =  do_getxattr(idmap, d, &ctx);
 
 
 773
 774	kvfree(ctx.kvalue);
 775	return error;
 
 
 
 
 
 
 
 
 
 
 776}
 777
 778static ssize_t path_getxattr(const char __user *pathname,
 779			     const char __user *name, void __user *value,
 780			     size_t size, unsigned int lookup_flags)
 781{
 782	struct path path;
 783	ssize_t error;
 784retry:
 785	error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
 
 
 
 
 
 
 
 
 786	if (error)
 787		return error;
 788	error = getxattr(mnt_idmap(path.mnt), path.dentry, name, value, size);
 789	path_put(&path);
 790	if (retry_estale(error, lookup_flags)) {
 791		lookup_flags |= LOOKUP_REVAL;
 792		goto retry;
 793	}
 794	return error;
 795}
 796
 797SYSCALL_DEFINE4(getxattr, const char __user *, pathname,
 798		const char __user *, name, void __user *, value, size_t, size)
 799{
 800	return path_getxattr(pathname, name, value, size, LOOKUP_FOLLOW);
 801}
 802
 803SYSCALL_DEFINE4(lgetxattr, const char __user *, pathname,
 804		const char __user *, name, void __user *, value, size_t, size)
 805{
 806	return path_getxattr(pathname, name, value, size, 0);
 
 807}
 808
 809SYSCALL_DEFINE4(fgetxattr, int, fd, const char __user *, name,
 810		void __user *, value, size_t, size)
 811{
 812	struct fd f = fdget(fd);
 813	ssize_t error = -EBADF;
 814
 815	if (!f.file)
 816		return error;
 817	audit_file(f.file);
 818	error = getxattr(file_mnt_idmap(f.file), f.file->f_path.dentry,
 819			 name, value, size);
 820	fdput(f);
 821	return error;
 822}
 823
 824/*
 825 * Extended attribute LIST operations
 826 */
 827static ssize_t
 828listxattr(struct dentry *d, char __user *list, size_t size)
 829{
 830	ssize_t error;
 831	char *klist = NULL;
 832
 833	if (size) {
 834		if (size > XATTR_LIST_MAX)
 835			size = XATTR_LIST_MAX;
 836		klist = kvmalloc(size, GFP_KERNEL);
 837		if (!klist)
 838			return -ENOMEM;
 839	}
 840
 841	error = vfs_listxattr(d, klist, size);
 842	if (error > 0) {
 843		if (size && copy_to_user(list, klist, error))
 844			error = -EFAULT;
 845	} else if (error == -ERANGE && size >= XATTR_LIST_MAX) {
 846		/* The file system tried to returned a list bigger
 847		   than XATTR_LIST_MAX bytes. Not possible. */
 848		error = -E2BIG;
 849	}
 850
 851	kvfree(klist);
 852
 853	return error;
 854}
 855
 856static ssize_t path_listxattr(const char __user *pathname, char __user *list,
 857			      size_t size, unsigned int lookup_flags)
 
 
 
 
 
 
 
 
 
 
 858{
 859	struct path path;
 860	ssize_t error;
 861retry:
 862	error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
 863	if (error)
 864		return error;
 865	error = listxattr(path.dentry, list, size);
 866	path_put(&path);
 867	if (retry_estale(error, lookup_flags)) {
 868		lookup_flags |= LOOKUP_REVAL;
 869		goto retry;
 870	}
 
 
 871	return error;
 872}
 873
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 874SYSCALL_DEFINE3(listxattr, const char __user *, pathname, char __user *, list,
 875		size_t, size)
 876{
 877	return path_listxattr(pathname, list, size, LOOKUP_FOLLOW);
 878}
 879
 880SYSCALL_DEFINE3(llistxattr, const char __user *, pathname, char __user *, list,
 881		size_t, size)
 882{
 883	return path_listxattr(pathname, list, size, 0);
 884}
 885
 886SYSCALL_DEFINE3(flistxattr, int, fd, char __user *, list, size_t, size)
 887{
 888	struct fd f = fdget(fd);
 889	ssize_t error = -EBADF;
 890
 891	if (!f.file)
 892		return error;
 893	audit_file(f.file);
 894	error = listxattr(f.file->f_path.dentry, list, size);
 895	fdput(f);
 896	return error;
 897}
 898
 899/*
 900 * Extended attribute REMOVE operations
 901 */
 902static long
 903removexattr(struct mnt_idmap *idmap, struct dentry *d,
 904	    const char __user *name)
 905{
 906	int error;
 907	char kname[XATTR_NAME_MAX + 1];
 908
 909	error = strncpy_from_user(kname, name, sizeof(kname));
 910	if (error == 0 || error == sizeof(kname))
 911		error = -ERANGE;
 912	if (error < 0)
 913		return error;
 914
 915	if (is_posix_acl_xattr(kname))
 916		return vfs_remove_acl(idmap, d, kname);
 
 917
 918	return vfs_removexattr(idmap, d, kname);
 
 
 
 
 
 
 919}
 920
 921static int path_removexattr(const char __user *pathname,
 922			    const char __user *name, unsigned int lookup_flags)
 
 923{
 924	struct path path;
 925	int error;
 
 926retry:
 927	error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
 928	if (error)
 929		return error;
 930	error = mnt_want_write(path.mnt);
 931	if (!error) {
 932		error = removexattr(mnt_idmap(path.mnt), path.dentry, name);
 933		mnt_drop_write(path.mnt);
 934	}
 935	path_put(&path);
 936	if (retry_estale(error, lookup_flags)) {
 937		lookup_flags |= LOOKUP_REVAL;
 938		goto retry;
 939	}
 
 
 940	return error;
 941}
 942
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 943SYSCALL_DEFINE2(removexattr, const char __user *, pathname,
 944		const char __user *, name)
 945{
 946	return path_removexattr(pathname, name, LOOKUP_FOLLOW);
 947}
 948
 949SYSCALL_DEFINE2(lremovexattr, const char __user *, pathname,
 950		const char __user *, name)
 951{
 952	return path_removexattr(pathname, name, 0);
 953}
 954
 955SYSCALL_DEFINE2(fremovexattr, int, fd, const char __user *, name)
 956{
 957	struct fd f = fdget(fd);
 958	int error = -EBADF;
 959
 960	if (!f.file)
 961		return error;
 962	audit_file(f.file);
 963	error = mnt_want_write_file(f.file);
 964	if (!error) {
 965		error = removexattr(file_mnt_idmap(f.file),
 966				    f.file->f_path.dentry, name);
 967		mnt_drop_write_file(f.file);
 968	}
 969	fdput(f);
 970	return error;
 971}
 972
 973int xattr_list_one(char **buffer, ssize_t *remaining_size, const char *name)
 974{
 975	size_t len;
 976
 977	len = strlen(name) + 1;
 978	if (*buffer) {
 979		if (*remaining_size < len)
 980			return -ERANGE;
 981		memcpy(*buffer, name, len);
 982		*buffer += len;
 983	}
 984	*remaining_size -= len;
 985	return 0;
 986}
 987
 988/**
 989 * generic_listxattr - run through a dentry's xattr list() operations
 990 * @dentry: dentry to list the xattrs
 991 * @buffer: result buffer
 992 * @buffer_size: size of @buffer
 993 *
 994 * Combine the results of the list() operation from every xattr_handler in the
 995 * xattr_handler stack.
 996 *
 997 * Note that this will not include the entries for POSIX ACLs.
 998 */
 999ssize_t
1000generic_listxattr(struct dentry *dentry, char *buffer, size_t buffer_size)
1001{
1002	const struct xattr_handler *handler, * const *handlers = dentry->d_sb->s_xattr;
1003	ssize_t remaining_size = buffer_size;
1004	int err = 0;
1005
1006	for_each_xattr_handler(handlers, handler) {
 
 
1007		if (!handler->name || (handler->list && !handler->list(dentry)))
1008			continue;
1009		err = xattr_list_one(&buffer, &remaining_size, handler->name);
1010		if (err)
1011			return err;
1012	}
1013
1014	return err ? err : buffer_size - remaining_size;
1015}
1016EXPORT_SYMBOL(generic_listxattr);
1017
1018/**
1019 * xattr_full_name  -  Compute full attribute name from suffix
1020 *
1021 * @handler:	handler of the xattr_handler operation
1022 * @name:	name passed to the xattr_handler operation
1023 *
1024 * The get and set xattr handler operations are called with the remainder of
1025 * the attribute name after skipping the handler's prefix: for example, "foo"
1026 * is passed to the get operation of a handler with prefix "user." to get
1027 * attribute "user.foo".  The full name is still "there" in the name though.
1028 *
1029 * Note: the list xattr handler operation when called from the vfs is passed a
1030 * NULL name; some file systems use this operation internally, with varying
1031 * semantics.
1032 */
1033const char *xattr_full_name(const struct xattr_handler *handler,
1034			    const char *name)
1035{
1036	size_t prefix_len = strlen(xattr_prefix(handler));
1037
1038	return name - prefix_len;
1039}
1040EXPORT_SYMBOL(xattr_full_name);
1041
1042/**
1043 * simple_xattr_space - estimate the memory used by a simple xattr
1044 * @name: the full name of the xattr
1045 * @size: the size of its value
1046 *
1047 * This takes no account of how much larger the two slab objects actually are:
1048 * that would depend on the slab implementation, when what is required is a
1049 * deterministic number, which grows with name length and size and quantity.
1050 *
1051 * Return: The approximate number of bytes of memory used by such an xattr.
1052 */
1053size_t simple_xattr_space(const char *name, size_t size)
1054{
1055	/*
1056	 * Use "40" instead of sizeof(struct simple_xattr), to return the
1057	 * same result on 32-bit and 64-bit, and even if simple_xattr grows.
1058	 */
1059	return 40 + size + strlen(name);
1060}
1061
1062/**
1063 * simple_xattr_free - free an xattr object
1064 * @xattr: the xattr object
1065 *
1066 * Free the xattr object. Can handle @xattr being NULL.
1067 */
1068void simple_xattr_free(struct simple_xattr *xattr)
1069{
1070	if (xattr)
1071		kfree(xattr->name);
1072	kvfree(xattr);
1073}
1074
1075/**
1076 * simple_xattr_alloc - allocate new xattr object
1077 * @value: value of the xattr object
1078 * @size: size of @value
1079 *
1080 * Allocate a new xattr object and initialize respective members. The caller is
1081 * responsible for handling the name of the xattr.
1082 *
1083 * Return: On success a new xattr object is returned. On failure NULL is
1084 * returned.
1085 */
1086struct simple_xattr *simple_xattr_alloc(const void *value, size_t size)
1087{
1088	struct simple_xattr *new_xattr;
1089	size_t len;
1090
1091	/* wrap around? */
1092	len = sizeof(*new_xattr) + size;
1093	if (len < sizeof(*new_xattr))
1094		return NULL;
1095
1096	new_xattr = kvmalloc(len, GFP_KERNEL_ACCOUNT);
1097	if (!new_xattr)
1098		return NULL;
1099
1100	new_xattr->size = size;
1101	memcpy(new_xattr->value, value, size);
1102	return new_xattr;
1103}
1104
1105/**
1106 * rbtree_simple_xattr_cmp - compare xattr name with current rbtree xattr entry
1107 * @key: xattr name
1108 * @node: current node
1109 *
1110 * Compare the xattr name with the xattr name attached to @node in the rbtree.
1111 *
1112 * Return: Negative value if continuing left, positive if continuing right, 0
1113 * if the xattr attached to @node matches @key.
1114 */
1115static int rbtree_simple_xattr_cmp(const void *key, const struct rb_node *node)
1116{
1117	const char *xattr_name = key;
1118	const struct simple_xattr *xattr;
1119
1120	xattr = rb_entry(node, struct simple_xattr, rb_node);
1121	return strcmp(xattr->name, xattr_name);
1122}
1123
1124/**
1125 * rbtree_simple_xattr_node_cmp - compare two xattr rbtree nodes
1126 * @new_node: new node
1127 * @node: current node
1128 *
1129 * Compare the xattr attached to @new_node with the xattr attached to @node.
1130 *
1131 * Return: Negative value if continuing left, positive if continuing right, 0
1132 * if the xattr attached to @new_node matches the xattr attached to @node.
1133 */
1134static int rbtree_simple_xattr_node_cmp(struct rb_node *new_node,
1135					const struct rb_node *node)
1136{
1137	struct simple_xattr *xattr;
1138	xattr = rb_entry(new_node, struct simple_xattr, rb_node);
1139	return rbtree_simple_xattr_cmp(xattr->name, node);
1140}
1141
1142/**
1143 * simple_xattr_get - get an xattr object
1144 * @xattrs: the header of the xattr object
1145 * @name: the name of the xattr to retrieve
1146 * @buffer: the buffer to store the value into
1147 * @size: the size of @buffer
1148 *
1149 * Try to find and retrieve the xattr object associated with @name.
1150 * If @buffer is provided store the value of @xattr in @buffer
1151 * otherwise just return the length. The size of @buffer is limited
1152 * to XATTR_SIZE_MAX which currently is 65536.
1153 *
1154 * Return: On success the length of the xattr value is returned. On error a
1155 * negative error code is returned.
1156 */
1157int simple_xattr_get(struct simple_xattrs *xattrs, const char *name,
1158		     void *buffer, size_t size)
1159{
1160	struct simple_xattr *xattr = NULL;
1161	struct rb_node *rbp;
1162	int ret = -ENODATA;
1163
1164	read_lock(&xattrs->lock);
1165	rbp = rb_find(name, &xattrs->rb_root, rbtree_simple_xattr_cmp);
1166	if (rbp) {
1167		xattr = rb_entry(rbp, struct simple_xattr, rb_node);
1168		ret = xattr->size;
1169		if (buffer) {
1170			if (size < xattr->size)
1171				ret = -ERANGE;
1172			else
1173				memcpy(buffer, xattr->value, xattr->size);
1174		}
1175	}
1176	read_unlock(&xattrs->lock);
1177	return ret;
1178}
1179
1180/**
1181 * simple_xattr_set - set an xattr object
1182 * @xattrs: the header of the xattr object
1183 * @name: the name of the xattr to retrieve
1184 * @value: the value to store along the xattr
1185 * @size: the size of @value
1186 * @flags: the flags determining how to set the xattr
1187 *
1188 * Set a new xattr object.
1189 * If @value is passed a new xattr object will be allocated. If XATTR_REPLACE
1190 * is specified in @flags a matching xattr object for @name must already exist.
1191 * If it does it will be replaced with the new xattr object. If it doesn't we
1192 * fail. If XATTR_CREATE is specified and a matching xattr does already exist
1193 * we fail. If it doesn't we create a new xattr. If @flags is zero we simply
1194 * insert the new xattr replacing any existing one.
1195 *
1196 * If @value is empty and a matching xattr object is found we delete it if
1197 * XATTR_REPLACE is specified in @flags or @flags is zero.
1198 *
1199 * If @value is empty and no matching xattr object for @name is found we do
1200 * nothing if XATTR_CREATE is specified in @flags or @flags is zero. For
1201 * XATTR_REPLACE we fail as mentioned above.
1202 *
1203 * Return: On success, the removed or replaced xattr is returned, to be freed
1204 * by the caller; or NULL if none. On failure a negative error code is returned.
1205 */
1206struct simple_xattr *simple_xattr_set(struct simple_xattrs *xattrs,
1207				      const char *name, const void *value,
1208				      size_t size, int flags)
1209{
1210	struct simple_xattr *old_xattr = NULL, *new_xattr = NULL;
1211	struct rb_node *parent = NULL, **rbp;
1212	int err = 0, ret;
1213
1214	/* value == NULL means remove */
1215	if (value) {
1216		new_xattr = simple_xattr_alloc(value, size);
1217		if (!new_xattr)
1218			return ERR_PTR(-ENOMEM);
1219
1220		new_xattr->name = kstrdup(name, GFP_KERNEL_ACCOUNT);
1221		if (!new_xattr->name) {
1222			simple_xattr_free(new_xattr);
1223			return ERR_PTR(-ENOMEM);
1224		}
1225	}
1226
1227	write_lock(&xattrs->lock);
1228	rbp = &xattrs->rb_root.rb_node;
1229	while (*rbp) {
1230		parent = *rbp;
1231		ret = rbtree_simple_xattr_cmp(name, *rbp);
1232		if (ret < 0)
1233			rbp = &(*rbp)->rb_left;
1234		else if (ret > 0)
1235			rbp = &(*rbp)->rb_right;
1236		else
1237			old_xattr = rb_entry(*rbp, struct simple_xattr, rb_node);
1238		if (old_xattr)
1239			break;
1240	}
1241
1242	if (old_xattr) {
1243		/* Fail if XATTR_CREATE is requested and the xattr exists. */
1244		if (flags & XATTR_CREATE) {
1245			err = -EEXIST;
1246			goto out_unlock;
1247		}
1248
1249		if (new_xattr)
1250			rb_replace_node(&old_xattr->rb_node,
1251					&new_xattr->rb_node, &xattrs->rb_root);
1252		else
1253			rb_erase(&old_xattr->rb_node, &xattrs->rb_root);
1254	} else {
1255		/* Fail if XATTR_REPLACE is requested but no xattr is found. */
1256		if (flags & XATTR_REPLACE) {
1257			err = -ENODATA;
1258			goto out_unlock;
1259		}
1260
1261		/*
1262		 * If XATTR_CREATE or no flags are specified together with a
1263		 * new value simply insert it.
1264		 */
1265		if (new_xattr) {
1266			rb_link_node(&new_xattr->rb_node, parent, rbp);
1267			rb_insert_color(&new_xattr->rb_node, &xattrs->rb_root);
1268		}
1269
1270		/*
1271		 * If XATTR_CREATE or no flags are specified and neither an
1272		 * old or new xattr exist then we don't need to do anything.
1273		 */
1274	}
1275
1276out_unlock:
1277	write_unlock(&xattrs->lock);
1278	if (!err)
1279		return old_xattr;
1280	simple_xattr_free(new_xattr);
1281	return ERR_PTR(err);
1282}
1283
1284static bool xattr_is_trusted(const char *name)
1285{
1286	return !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN);
1287}
1288
1289/**
1290 * simple_xattr_list - list all xattr objects
1291 * @inode: inode from which to get the xattrs
1292 * @xattrs: the header of the xattr object
1293 * @buffer: the buffer to store all xattrs into
1294 * @size: the size of @buffer
1295 *
1296 * List all xattrs associated with @inode. If @buffer is NULL we returned
1297 * the required size of the buffer. If @buffer is provided we store the
1298 * xattrs value into it provided it is big enough.
1299 *
1300 * Note, the number of xattr names that can be listed with listxattr(2) is
1301 * limited to XATTR_LIST_MAX aka 65536 bytes. If a larger buffer is passed
1302 * then vfs_listxattr() caps it to XATTR_LIST_MAX and if more xattr names
1303 * are found it will return -E2BIG.
1304 *
1305 * Return: On success the required size or the size of the copied xattrs is
1306 * returned. On error a negative error code is returned.
1307 */
1308ssize_t simple_xattr_list(struct inode *inode, struct simple_xattrs *xattrs,
1309			  char *buffer, size_t size)
1310{
1311	bool trusted = ns_capable_noaudit(&init_user_ns, CAP_SYS_ADMIN);
1312	struct simple_xattr *xattr;
1313	struct rb_node *rbp;
1314	ssize_t remaining_size = size;
1315	int err = 0;
1316
1317	err = posix_acl_listxattr(inode, &buffer, &remaining_size);
1318	if (err)
1319		return err;
1320
1321	read_lock(&xattrs->lock);
1322	for (rbp = rb_first(&xattrs->rb_root); rbp; rbp = rb_next(rbp)) {
1323		xattr = rb_entry(rbp, struct simple_xattr, rb_node);
1324
1325		/* skip "trusted." attributes for unprivileged callers */
1326		if (!trusted && xattr_is_trusted(xattr->name))
1327			continue;
1328
1329		err = xattr_list_one(&buffer, &remaining_size, xattr->name);
1330		if (err)
1331			break;
1332	}
1333	read_unlock(&xattrs->lock);
1334
1335	return err ? err : size - remaining_size;
1336}
1337
1338/**
1339 * rbtree_simple_xattr_less - compare two xattr rbtree nodes
1340 * @new_node: new node
1341 * @node: current node
1342 *
1343 * Compare the xattr attached to @new_node with the xattr attached to @node.
1344 * Note that this function technically tolerates duplicate entries.
1345 *
1346 * Return: True if insertion point in the rbtree is found.
1347 */
1348static bool rbtree_simple_xattr_less(struct rb_node *new_node,
1349				     const struct rb_node *node)
1350{
1351	return rbtree_simple_xattr_node_cmp(new_node, node) < 0;
1352}
1353
1354/**
1355 * simple_xattr_add - add xattr objects
1356 * @xattrs: the header of the xattr object
1357 * @new_xattr: the xattr object to add
1358 *
1359 * Add an xattr object to @xattrs. This assumes no replacement or removal
1360 * of matching xattrs is wanted. Should only be called during inode
1361 * initialization when a few distinct initial xattrs are supposed to be set.
1362 */
1363void simple_xattr_add(struct simple_xattrs *xattrs,
1364		      struct simple_xattr *new_xattr)
1365{
1366	write_lock(&xattrs->lock);
1367	rb_add(&new_xattr->rb_node, &xattrs->rb_root, rbtree_simple_xattr_less);
1368	write_unlock(&xattrs->lock);
1369}
1370
1371/**
1372 * simple_xattrs_init - initialize new xattr header
1373 * @xattrs: header to initialize
1374 *
1375 * Initialize relevant fields of a an xattr header.
1376 */
1377void simple_xattrs_init(struct simple_xattrs *xattrs)
1378{
1379	xattrs->rb_root = RB_ROOT;
1380	rwlock_init(&xattrs->lock);
1381}
1382
1383/**
1384 * simple_xattrs_free - free xattrs
1385 * @xattrs: xattr header whose xattrs to destroy
1386 * @freed_space: approximate number of bytes of memory freed from @xattrs
1387 *
1388 * Destroy all xattrs in @xattr. When this is called no one can hold a
1389 * reference to any of the xattrs anymore.
1390 */
1391void simple_xattrs_free(struct simple_xattrs *xattrs, size_t *freed_space)
1392{
1393	struct rb_node *rbp;
1394
1395	if (freed_space)
1396		*freed_space = 0;
1397	rbp = rb_first(&xattrs->rb_root);
1398	while (rbp) {
1399		struct simple_xattr *xattr;
1400		struct rb_node *rbp_next;
1401
1402		rbp_next = rb_next(rbp);
1403		xattr = rb_entry(rbp, struct simple_xattr, rb_node);
1404		rb_erase(&xattr->rb_node, &xattrs->rb_root);
1405		if (freed_space)
1406			*freed_space += simple_xattr_space(xattr->name,
1407							   xattr->size);
1408		simple_xattr_free(xattr);
1409		rbp = rbp_next;
1410	}
1411}
v6.13.7
   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/filelock.h>
  13#include <linux/slab.h>
  14#include <linux/file.h>
  15#include <linux/xattr.h>
  16#include <linux/mount.h>
  17#include <linux/namei.h>
  18#include <linux/security.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#include "internal.h"
  29
  30static const char *
  31strcmp_prefix(const char *a, const char *a_prefix)
  32{
  33	while (*a_prefix && *a == *a_prefix) {
  34		a++;
  35		a_prefix++;
  36	}
  37	return *a_prefix ? NULL : a;
  38}
  39
  40/*
  41 * In order to implement different sets of xattr operations for each xattr
  42 * prefix, a filesystem should create a null-terminated array of struct
  43 * xattr_handler (one for each prefix) and hang a pointer to it off of the
  44 * s_xattr field of the superblock.
  45 */
  46#define for_each_xattr_handler(handlers, handler)		\
  47	if (handlers)						\
  48		for ((handler) = *(handlers)++;			\
  49			(handler) != NULL;			\
  50			(handler) = *(handlers)++)
  51
  52/*
  53 * Find the xattr_handler with the matching prefix.
  54 */
  55static const struct xattr_handler *
  56xattr_resolve_name(struct inode *inode, const char **name)
  57{
  58	const struct xattr_handler * const *handlers = inode->i_sb->s_xattr;
  59	const struct xattr_handler *handler;
  60
  61	if (!(inode->i_opflags & IOP_XATTR)) {
  62		if (unlikely(is_bad_inode(inode)))
  63			return ERR_PTR(-EIO);
  64		return ERR_PTR(-EOPNOTSUPP);
  65	}
  66	for_each_xattr_handler(handlers, handler) {
  67		const char *n;
  68
  69		n = strcmp_prefix(*name, xattr_prefix(handler));
  70		if (n) {
  71			if (!handler->prefix ^ !*n) {
  72				if (*n)
  73					continue;
  74				return ERR_PTR(-EINVAL);
  75			}
  76			*name = n;
  77			return handler;
  78		}
  79	}
  80	return ERR_PTR(-EOPNOTSUPP);
  81}
  82
  83/**
  84 * may_write_xattr - check whether inode allows writing xattr
  85 * @idmap: idmap of the mount the inode was found from
  86 * @inode: the inode on which to set an xattr
  87 *
  88 * Check whether the inode allows writing xattrs. Specifically, we can never
  89 * set or remove an extended attribute on a read-only filesystem  or on an
  90 * immutable / append-only inode.
  91 *
  92 * We also need to ensure that the inode has a mapping in the mount to
  93 * not risk writing back invalid i_{g,u}id values.
  94 *
  95 * Return: On success zero is returned. On error a negative errno is returned.
  96 */
  97int may_write_xattr(struct mnt_idmap *idmap, struct inode *inode)
  98{
  99	if (IS_IMMUTABLE(inode))
 100		return -EPERM;
 101	if (IS_APPEND(inode))
 102		return -EPERM;
 103	if (HAS_UNMAPPED_ID(idmap, inode))
 104		return -EPERM;
 105	return 0;
 106}
 107
 108/*
 109 * Check permissions for extended attribute access.  This is a bit complicated
 110 * because different namespaces have very different rules.
 111 */
 112static int
 113xattr_permission(struct mnt_idmap *idmap, struct inode *inode,
 114		 const char *name, int mask)
 115{
 116	if (mask & MAY_WRITE) {
 117		int ret;
 118
 119		ret = may_write_xattr(idmap, inode);
 120		if (ret)
 121			return ret;
 122	}
 123
 124	/*
 125	 * No restriction for security.* and system.* from the VFS.  Decision
 126	 * on these is left to the underlying filesystem / security module.
 127	 */
 128	if (!strncmp(name, XATTR_SECURITY_PREFIX, XATTR_SECURITY_PREFIX_LEN) ||
 129	    !strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN))
 130		return 0;
 131
 132	/*
 133	 * The trusted.* namespace can only be accessed by privileged users.
 134	 */
 135	if (!strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN)) {
 136		if (!capable(CAP_SYS_ADMIN))
 137			return (mask & MAY_WRITE) ? -EPERM : -ENODATA;
 138		return 0;
 139	}
 140
 141	/*
 142	 * In the user.* namespace, only regular files and directories can have
 143	 * extended attributes. For sticky directories, only the owner and
 144	 * privileged users can write attributes.
 145	 */
 146	if (!strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN)) {
 147		if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
 148			return (mask & MAY_WRITE) ? -EPERM : -ENODATA;
 149		if (S_ISDIR(inode->i_mode) && (inode->i_mode & S_ISVTX) &&
 150		    (mask & MAY_WRITE) &&
 151		    !inode_owner_or_capable(idmap, inode))
 152			return -EPERM;
 153	}
 154
 155	return inode_permission(idmap, inode, mask);
 156}
 157
 158/*
 159 * Look for any handler that deals with the specified namespace.
 160 */
 161int
 162xattr_supports_user_prefix(struct inode *inode)
 163{
 164	const struct xattr_handler * const *handlers = inode->i_sb->s_xattr;
 165	const struct xattr_handler *handler;
 166
 167	if (!(inode->i_opflags & IOP_XATTR)) {
 168		if (unlikely(is_bad_inode(inode)))
 169			return -EIO;
 170		return -EOPNOTSUPP;
 171	}
 172
 173	for_each_xattr_handler(handlers, handler) {
 174		if (!strncmp(xattr_prefix(handler), XATTR_USER_PREFIX,
 175			     XATTR_USER_PREFIX_LEN))
 176			return 0;
 177	}
 178
 179	return -EOPNOTSUPP;
 180}
 181EXPORT_SYMBOL(xattr_supports_user_prefix);
 182
 183int
 184__vfs_setxattr(struct mnt_idmap *idmap, struct dentry *dentry,
 185	       struct inode *inode, const char *name, const void *value,
 186	       size_t size, int flags)
 187{
 188	const struct xattr_handler *handler;
 189
 190	if (is_posix_acl_xattr(name))
 191		return -EOPNOTSUPP;
 192
 193	handler = xattr_resolve_name(inode, &name);
 194	if (IS_ERR(handler))
 195		return PTR_ERR(handler);
 196	if (!handler->set)
 197		return -EOPNOTSUPP;
 198	if (size == 0)
 199		value = "";  /* empty EA, do not remove */
 200	return handler->set(handler, idmap, dentry, inode, name, value,
 201			    size, flags);
 202}
 203EXPORT_SYMBOL(__vfs_setxattr);
 204
 205/**
 206 *  __vfs_setxattr_noperm - perform setxattr operation without performing
 207 *  permission checks.
 208 *
 209 *  @idmap: idmap of the mount the inode was found from
 210 *  @dentry: object to perform setxattr on
 211 *  @name: xattr name to set
 212 *  @value: value to set @name to
 213 *  @size: size of @value
 214 *  @flags: flags to pass into filesystem operations
 215 *
 216 *  returns the result of the internal setxattr or setsecurity operations.
 217 *
 218 *  This function requires the caller to lock the inode's i_mutex before it
 219 *  is executed. It also assumes that the caller will make the appropriate
 220 *  permission checks.
 221 */
 222int __vfs_setxattr_noperm(struct mnt_idmap *idmap,
 223			  struct dentry *dentry, const char *name,
 224			  const void *value, size_t size, int flags)
 225{
 226	struct inode *inode = dentry->d_inode;
 227	int error = -EAGAIN;
 228	int issec = !strncmp(name, XATTR_SECURITY_PREFIX,
 229				   XATTR_SECURITY_PREFIX_LEN);
 230
 231	if (issec)
 232		inode->i_flags &= ~S_NOSEC;
 233	if (inode->i_opflags & IOP_XATTR) {
 234		error = __vfs_setxattr(idmap, dentry, inode, name, value,
 235				       size, flags);
 236		if (!error) {
 237			fsnotify_xattr(dentry);
 238			security_inode_post_setxattr(dentry, name, value,
 239						     size, flags);
 240		}
 241	} else {
 242		if (unlikely(is_bad_inode(inode)))
 243			return -EIO;
 244	}
 245	if (error == -EAGAIN) {
 246		error = -EOPNOTSUPP;
 247
 248		if (issec) {
 249			const char *suffix = name + XATTR_SECURITY_PREFIX_LEN;
 250
 251			error = security_inode_setsecurity(inode, suffix, value,
 252							   size, flags);
 253			if (!error)
 254				fsnotify_xattr(dentry);
 255		}
 256	}
 257
 258	return error;
 259}
 260
 261/**
 262 * __vfs_setxattr_locked - set an extended attribute while holding the inode
 263 * lock
 264 *
 265 *  @idmap: idmap of the mount of the target inode
 266 *  @dentry: object to perform setxattr on
 267 *  @name: xattr name to set
 268 *  @value: value to set @name to
 269 *  @size: size of @value
 270 *  @flags: flags to pass into filesystem operations
 271 *  @delegated_inode: on return, will contain an inode pointer that
 272 *  a delegation was broken on, NULL if none.
 273 */
 274int
 275__vfs_setxattr_locked(struct mnt_idmap *idmap, struct dentry *dentry,
 276		      const char *name, const void *value, size_t size,
 277		      int flags, struct inode **delegated_inode)
 278{
 279	struct inode *inode = dentry->d_inode;
 280	int error;
 281
 282	error = xattr_permission(idmap, inode, name, MAY_WRITE);
 283	if (error)
 284		return error;
 285
 286	error = security_inode_setxattr(idmap, dentry, name, value, size,
 287					flags);
 288	if (error)
 289		goto out;
 290
 291	error = try_break_deleg(inode, delegated_inode);
 292	if (error)
 293		goto out;
 294
 295	error = __vfs_setxattr_noperm(idmap, dentry, name, value,
 296				      size, flags);
 297
 298out:
 299	return error;
 300}
 301EXPORT_SYMBOL_GPL(__vfs_setxattr_locked);
 302
 303int
 304vfs_setxattr(struct mnt_idmap *idmap, struct dentry *dentry,
 305	     const char *name, const void *value, size_t size, int flags)
 306{
 307	struct inode *inode = dentry->d_inode;
 308	struct inode *delegated_inode = NULL;
 309	const void  *orig_value = value;
 310	int error;
 311
 312	if (size && strcmp(name, XATTR_NAME_CAPS) == 0) {
 313		error = cap_convert_nscap(idmap, dentry, &value, size);
 314		if (error < 0)
 315			return error;
 316		size = error;
 317	}
 318
 319retry_deleg:
 320	inode_lock(inode);
 321	error = __vfs_setxattr_locked(idmap, dentry, name, value, size,
 322				      flags, &delegated_inode);
 323	inode_unlock(inode);
 324
 325	if (delegated_inode) {
 326		error = break_deleg_wait(&delegated_inode);
 327		if (!error)
 328			goto retry_deleg;
 329	}
 330	if (value != orig_value)
 331		kfree(value);
 332
 333	return error;
 334}
 335EXPORT_SYMBOL_GPL(vfs_setxattr);
 336
 337static ssize_t
 338xattr_getsecurity(struct mnt_idmap *idmap, struct inode *inode,
 339		  const char *name, void *value, size_t size)
 340{
 341	void *buffer = NULL;
 342	ssize_t len;
 343
 344	if (!value || !size) {
 345		len = security_inode_getsecurity(idmap, inode, name,
 346						 &buffer, false);
 347		goto out_noalloc;
 348	}
 349
 350	len = security_inode_getsecurity(idmap, inode, name, &buffer,
 351					 true);
 352	if (len < 0)
 353		return len;
 354	if (size < len) {
 355		len = -ERANGE;
 356		goto out;
 357	}
 358	memcpy(value, buffer, len);
 359out:
 360	kfree(buffer);
 361out_noalloc:
 362	return len;
 363}
 364
 365/*
 366 * vfs_getxattr_alloc - allocate memory, if necessary, before calling getxattr
 367 *
 368 * Allocate memory, if not already allocated, or re-allocate correct size,
 369 * before retrieving the extended attribute.  The xattr value buffer should
 370 * always be freed by the caller, even on error.
 371 *
 372 * Returns the result of alloc, if failed, or the getxattr operation.
 373 */
 374int
 375vfs_getxattr_alloc(struct mnt_idmap *idmap, struct dentry *dentry,
 376		   const char *name, char **xattr_value, size_t xattr_size,
 377		   gfp_t flags)
 378{
 379	const struct xattr_handler *handler;
 380	struct inode *inode = dentry->d_inode;
 381	char *value = *xattr_value;
 382	int error;
 383
 384	error = xattr_permission(idmap, inode, name, MAY_READ);
 385	if (error)
 386		return error;
 387
 388	handler = xattr_resolve_name(inode, &name);
 389	if (IS_ERR(handler))
 390		return PTR_ERR(handler);
 391	if (!handler->get)
 392		return -EOPNOTSUPP;
 393	error = handler->get(handler, dentry, inode, name, NULL, 0);
 394	if (error < 0)
 395		return error;
 396
 397	if (!value || (error > xattr_size)) {
 398		value = krealloc(*xattr_value, error + 1, flags);
 399		if (!value)
 400			return -ENOMEM;
 401		memset(value, 0, error + 1);
 402	}
 403
 404	error = handler->get(handler, dentry, inode, name, value, error);
 405	*xattr_value = value;
 406	return error;
 407}
 408
 409ssize_t
 410__vfs_getxattr(struct dentry *dentry, struct inode *inode, const char *name,
 411	       void *value, size_t size)
 412{
 413	const struct xattr_handler *handler;
 414
 415	if (is_posix_acl_xattr(name))
 416		return -EOPNOTSUPP;
 417
 418	handler = xattr_resolve_name(inode, &name);
 419	if (IS_ERR(handler))
 420		return PTR_ERR(handler);
 421	if (!handler->get)
 422		return -EOPNOTSUPP;
 423	return handler->get(handler, dentry, inode, name, value, size);
 424}
 425EXPORT_SYMBOL(__vfs_getxattr);
 426
 427ssize_t
 428vfs_getxattr(struct mnt_idmap *idmap, struct dentry *dentry,
 429	     const char *name, void *value, size_t size)
 430{
 431	struct inode *inode = dentry->d_inode;
 432	int error;
 433
 434	error = xattr_permission(idmap, inode, name, MAY_READ);
 435	if (error)
 436		return error;
 437
 438	error = security_inode_getxattr(dentry, name);
 439	if (error)
 440		return error;
 441
 442	if (!strncmp(name, XATTR_SECURITY_PREFIX,
 443				XATTR_SECURITY_PREFIX_LEN)) {
 444		const char *suffix = name + XATTR_SECURITY_PREFIX_LEN;
 445		int ret = xattr_getsecurity(idmap, inode, suffix, value,
 446					    size);
 447		/*
 448		 * Only overwrite the return value if a security module
 449		 * is actually active.
 450		 */
 451		if (ret == -EOPNOTSUPP)
 452			goto nolsm;
 453		return ret;
 454	}
 455nolsm:
 456	return __vfs_getxattr(dentry, inode, name, value, size);
 457}
 458EXPORT_SYMBOL_GPL(vfs_getxattr);
 459
 460/**
 461 * vfs_listxattr - retrieve \0 separated list of xattr names
 462 * @dentry: the dentry from whose inode the xattr names are retrieved
 463 * @list: buffer to store xattr names into
 464 * @size: size of the buffer
 465 *
 466 * This function returns the names of all xattrs associated with the
 467 * inode of @dentry.
 468 *
 469 * Note, for legacy reasons the vfs_listxattr() function lists POSIX
 470 * ACLs as well. Since POSIX ACLs are decoupled from IOP_XATTR the
 471 * vfs_listxattr() function doesn't check for this flag since a
 472 * filesystem could implement POSIX ACLs without implementing any other
 473 * xattrs.
 474 *
 475 * However, since all codepaths that remove IOP_XATTR also assign of
 476 * inode operations that either don't implement or implement a stub
 477 * ->listxattr() operation.
 478 *
 479 * Return: On success, the size of the buffer that was used. On error a
 480 *         negative error code.
 481 */
 482ssize_t
 483vfs_listxattr(struct dentry *dentry, char *list, size_t size)
 484{
 485	struct inode *inode = d_inode(dentry);
 486	ssize_t error;
 487
 488	error = security_inode_listxattr(dentry);
 489	if (error)
 490		return error;
 491
 492	if (inode->i_op->listxattr) {
 493		error = inode->i_op->listxattr(dentry, list, size);
 494	} else {
 495		error = security_inode_listsecurity(inode, list, size);
 496		if (size && error > size)
 497			error = -ERANGE;
 498	}
 499	return error;
 500}
 501EXPORT_SYMBOL_GPL(vfs_listxattr);
 502
 503int
 504__vfs_removexattr(struct mnt_idmap *idmap, struct dentry *dentry,
 505		  const char *name)
 506{
 507	struct inode *inode = d_inode(dentry);
 508	const struct xattr_handler *handler;
 509
 510	if (is_posix_acl_xattr(name))
 511		return -EOPNOTSUPP;
 512
 513	handler = xattr_resolve_name(inode, &name);
 514	if (IS_ERR(handler))
 515		return PTR_ERR(handler);
 516	if (!handler->set)
 517		return -EOPNOTSUPP;
 518	return handler->set(handler, idmap, dentry, inode, name, NULL, 0,
 519			    XATTR_REPLACE);
 520}
 521EXPORT_SYMBOL(__vfs_removexattr);
 522
 523/**
 524 * __vfs_removexattr_locked - set an extended attribute while holding the inode
 525 * lock
 526 *
 527 *  @idmap: idmap of the mount of the target inode
 528 *  @dentry: object to perform setxattr on
 529 *  @name: name of xattr to remove
 530 *  @delegated_inode: on return, will contain an inode pointer that
 531 *  a delegation was broken on, NULL if none.
 532 */
 533int
 534__vfs_removexattr_locked(struct mnt_idmap *idmap,
 535			 struct dentry *dentry, const char *name,
 536			 struct inode **delegated_inode)
 537{
 538	struct inode *inode = dentry->d_inode;
 539	int error;
 540
 541	error = xattr_permission(idmap, inode, name, MAY_WRITE);
 542	if (error)
 543		return error;
 544
 545	error = security_inode_removexattr(idmap, dentry, name);
 546	if (error)
 547		goto out;
 548
 549	error = try_break_deleg(inode, delegated_inode);
 550	if (error)
 551		goto out;
 552
 553	error = __vfs_removexattr(idmap, dentry, name);
 554	if (error)
 555		return error;
 556
 557	fsnotify_xattr(dentry);
 558	security_inode_post_removexattr(dentry, name);
 
 
 559
 560out:
 561	return error;
 562}
 563EXPORT_SYMBOL_GPL(__vfs_removexattr_locked);
 564
 565int
 566vfs_removexattr(struct mnt_idmap *idmap, struct dentry *dentry,
 567		const char *name)
 568{
 569	struct inode *inode = dentry->d_inode;
 570	struct inode *delegated_inode = NULL;
 571	int error;
 572
 573retry_deleg:
 574	inode_lock(inode);
 575	error = __vfs_removexattr_locked(idmap, dentry,
 576					 name, &delegated_inode);
 577	inode_unlock(inode);
 578
 579	if (delegated_inode) {
 580		error = break_deleg_wait(&delegated_inode);
 581		if (!error)
 582			goto retry_deleg;
 583	}
 584
 585	return error;
 586}
 587EXPORT_SYMBOL_GPL(vfs_removexattr);
 588
 589int import_xattr_name(struct xattr_name *kname, const char __user *name)
 590{
 591	int error = strncpy_from_user(kname->name, name,
 592					sizeof(kname->name));
 593	if (error == 0 || error == sizeof(kname->name))
 594		return -ERANGE;
 595	if (error < 0)
 596		return error;
 597	return 0;
 598}
 599
 600/*
 601 * Extended attribute SET operations
 602 */
 603
 604int setxattr_copy(const char __user *name, struct kernel_xattr_ctx *ctx)
 605{
 606	int error;
 607
 608	if (ctx->flags & ~(XATTR_CREATE|XATTR_REPLACE))
 609		return -EINVAL;
 610
 611	error = import_xattr_name(ctx->kname, name);
 612	if (error)
 
 
 
 613		return error;
 614
 
 615	if (ctx->size) {
 616		if (ctx->size > XATTR_SIZE_MAX)
 617			return -E2BIG;
 618
 619		ctx->kvalue = vmemdup_user(ctx->cvalue, ctx->size);
 620		if (IS_ERR(ctx->kvalue)) {
 621			error = PTR_ERR(ctx->kvalue);
 622			ctx->kvalue = NULL;
 623		}
 624	}
 625
 626	return error;
 627}
 628
 629static int do_setxattr(struct mnt_idmap *idmap, struct dentry *dentry,
 630		struct kernel_xattr_ctx *ctx)
 631{
 632	if (is_posix_acl_xattr(ctx->kname->name))
 633		return do_set_acl(idmap, dentry, ctx->kname->name,
 634				  ctx->kvalue, ctx->size);
 635
 636	return vfs_setxattr(idmap, dentry, ctx->kname->name,
 637			ctx->kvalue, ctx->size, ctx->flags);
 638}
 639
 640int file_setxattr(struct file *f, struct kernel_xattr_ctx *ctx)
 
 
 
 641{
 642	int error = mnt_want_write_file(f);
 
 
 
 
 
 
 
 
 
 
 
 
 643
 644	if (!error) {
 645		audit_file(f);
 646		error = do_setxattr(file_mnt_idmap(f), f->f_path.dentry, ctx);
 647		mnt_drop_write_file(f);
 648	}
 649	return error;
 650}
 651
 652/* unconditionally consumes filename */
 653int filename_setxattr(int dfd, struct filename *filename,
 654		      unsigned int lookup_flags, struct kernel_xattr_ctx *ctx)
 655{
 656	struct path path;
 657	int error;
 658
 659retry:
 660	error = filename_lookup(dfd, filename, lookup_flags, &path, NULL);
 661	if (error)
 662		goto out;
 663	error = mnt_want_write(path.mnt);
 664	if (!error) {
 665		error = do_setxattr(mnt_idmap(path.mnt), path.dentry, ctx);
 
 666		mnt_drop_write(path.mnt);
 667	}
 668	path_put(&path);
 669	if (retry_estale(error, lookup_flags)) {
 670		lookup_flags |= LOOKUP_REVAL;
 671		goto retry;
 672	}
 673
 674out:
 675	putname(filename);
 676	return error;
 677}
 678
 679static int path_setxattrat(int dfd, const char __user *pathname,
 680			   unsigned int at_flags, const char __user *name,
 681			   const void __user *value, size_t size, int flags)
 682{
 683	struct xattr_name kname;
 684	struct kernel_xattr_ctx ctx = {
 685		.cvalue	= value,
 686		.kvalue	= NULL,
 687		.size	= size,
 688		.kname	= &kname,
 689		.flags	= flags,
 690	};
 691	struct filename *filename;
 692	unsigned int lookup_flags = 0;
 693	int error;
 694
 695	if ((at_flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
 696		return -EINVAL;
 697
 698	if (!(at_flags & AT_SYMLINK_NOFOLLOW))
 699		lookup_flags = LOOKUP_FOLLOW;
 700
 701	error = setxattr_copy(name, &ctx);
 702	if (error)
 703		return error;
 704
 705	filename = getname_maybe_null(pathname, at_flags);
 706	if (!filename) {
 707		CLASS(fd, f)(dfd);
 708		if (fd_empty(f))
 709			error = -EBADF;
 710		else
 711			error = file_setxattr(fd_file(f), &ctx);
 712	} else {
 713		error = filename_setxattr(dfd, filename, lookup_flags, &ctx);
 714	}
 715	kvfree(ctx.kvalue);
 716	return error;
 717}
 718
 719SYSCALL_DEFINE6(setxattrat, int, dfd, const char __user *, pathname, unsigned int, at_flags,
 720		const char __user *, name, const struct xattr_args __user *, uargs,
 721		size_t, usize)
 722{
 723	struct xattr_args args = {};
 724	int error;
 725
 726	BUILD_BUG_ON(sizeof(struct xattr_args) < XATTR_ARGS_SIZE_VER0);
 727	BUILD_BUG_ON(sizeof(struct xattr_args) != XATTR_ARGS_SIZE_LATEST);
 728
 729	if (unlikely(usize < XATTR_ARGS_SIZE_VER0))
 730		return -EINVAL;
 731	if (usize > PAGE_SIZE)
 732		return -E2BIG;
 733
 734	error = copy_struct_from_user(&args, sizeof(args), uargs, usize);
 735	if (error)
 736		return error;
 737
 738	return path_setxattrat(dfd, pathname, at_flags, name,
 739			       u64_to_user_ptr(args.value), args.size,
 740			       args.flags);
 741}
 742
 743SYSCALL_DEFINE5(setxattr, const char __user *, pathname,
 744		const char __user *, name, const void __user *, value,
 745		size_t, size, int, flags)
 746{
 747	return path_setxattrat(AT_FDCWD, pathname, 0, name, value, size, flags);
 748}
 749
 750SYSCALL_DEFINE5(lsetxattr, const char __user *, pathname,
 751		const char __user *, name, const void __user *, value,
 752		size_t, size, int, flags)
 753{
 754	return path_setxattrat(AT_FDCWD, pathname, AT_SYMLINK_NOFOLLOW, name,
 755			       value, size, flags);
 756}
 757
 758SYSCALL_DEFINE5(fsetxattr, int, fd, const char __user *, name,
 759		const void __user *,value, size_t, size, int, flags)
 760{
 761	return path_setxattrat(fd, NULL, AT_EMPTY_PATH, name,
 762			       value, size, flags);
 
 
 
 
 
 
 
 
 
 
 
 
 
 763}
 764
 765/*
 766 * Extended attribute GET operations
 767 */
 768static ssize_t
 769do_getxattr(struct mnt_idmap *idmap, struct dentry *d,
 770	struct kernel_xattr_ctx *ctx)
 771{
 772	ssize_t error;
 773	char *kname = ctx->kname->name;
 774	void *kvalue = NULL;
 775
 776	if (ctx->size) {
 777		if (ctx->size > XATTR_SIZE_MAX)
 778			ctx->size = XATTR_SIZE_MAX;
 779		kvalue = kvzalloc(ctx->size, GFP_KERNEL);
 780		if (!kvalue)
 781			return -ENOMEM;
 782	}
 783
 784	if (is_posix_acl_xattr(kname))
 785		error = do_get_acl(idmap, d, kname, kvalue, ctx->size);
 786	else
 787		error = vfs_getxattr(idmap, d, kname, kvalue, ctx->size);
 788	if (error > 0) {
 789		if (ctx->size && copy_to_user(ctx->value, kvalue, error))
 790			error = -EFAULT;
 791	} else if (error == -ERANGE && ctx->size >= XATTR_SIZE_MAX) {
 792		/* The file system tried to returned a value bigger
 793		   than XATTR_SIZE_MAX bytes. Not possible. */
 794		error = -E2BIG;
 795	}
 796
 797	kvfree(kvalue);
 798	return error;
 799}
 800
 801ssize_t file_getxattr(struct file *f, struct kernel_xattr_ctx *ctx)
 
 
 802{
 803	audit_file(f);
 804	return do_getxattr(file_mnt_idmap(f), f->f_path.dentry, ctx);
 805}
 806
 807/* unconditionally consumes filename */
 808ssize_t filename_getxattr(int dfd, struct filename *filename,
 809			  unsigned int lookup_flags, struct kernel_xattr_ctx *ctx)
 810{
 811	struct path path;
 812	ssize_t error;
 813retry:
 814	error = filename_lookup(dfd, filename, lookup_flags, &path, NULL);
 815	if (error)
 816		goto out;
 817	error = do_getxattr(mnt_idmap(path.mnt), path.dentry, ctx);
 818	path_put(&path);
 819	if (retry_estale(error, lookup_flags)) {
 820		lookup_flags |= LOOKUP_REVAL;
 821		goto retry;
 822	}
 823out:
 824	putname(filename);
 825	return error;
 826}
 827
 828static ssize_t path_getxattrat(int dfd, const char __user *pathname,
 829			       unsigned int at_flags, const char __user *name,
 830			       void __user *value, size_t size)
 831{
 832	struct xattr_name kname;
 833	struct kernel_xattr_ctx ctx = {
 834		.value    = value,
 
 835		.size     = size,
 836		.kname    = &kname,
 837		.flags    = 0,
 838	};
 839	struct filename *filename;
 840	ssize_t error;
 841
 842	if ((at_flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
 843		return -EINVAL;
 
 
 
 844
 845	error = import_xattr_name(&kname, name);
 846	if (error)
 847		return error;
 848
 849	filename = getname_maybe_null(pathname, at_flags);
 850	if (!filename) {
 851		CLASS(fd, f)(dfd);
 852		if (fd_empty(f))
 853			return -EBADF;
 854		return file_getxattr(fd_file(f), &ctx);
 855	} else {
 856		int lookup_flags = 0;
 857		if (!(at_flags & AT_SYMLINK_NOFOLLOW))
 858			lookup_flags = LOOKUP_FOLLOW;
 859		return filename_getxattr(dfd, filename, lookup_flags, &ctx);
 860	}
 861}
 862
 863SYSCALL_DEFINE6(getxattrat, int, dfd, const char __user *, pathname, unsigned int, at_flags,
 864		const char __user *, name, struct xattr_args __user *, uargs, size_t, usize)
 
 865{
 866	struct xattr_args args = {};
 867	int error;
 868
 869	BUILD_BUG_ON(sizeof(struct xattr_args) < XATTR_ARGS_SIZE_VER0);
 870	BUILD_BUG_ON(sizeof(struct xattr_args) != XATTR_ARGS_SIZE_LATEST);
 871
 872	if (unlikely(usize < XATTR_ARGS_SIZE_VER0))
 873		return -EINVAL;
 874	if (usize > PAGE_SIZE)
 875		return -E2BIG;
 876
 877	error = copy_struct_from_user(&args, sizeof(args), uargs, usize);
 878	if (error)
 879		return error;
 880
 881	if (args.flags != 0)
 882		return -EINVAL;
 883
 884	return path_getxattrat(dfd, pathname, at_flags, name,
 885			       u64_to_user_ptr(args.value), args.size);
 
 886}
 887
 888SYSCALL_DEFINE4(getxattr, const char __user *, pathname,
 889		const char __user *, name, void __user *, value, size_t, size)
 890{
 891	return path_getxattrat(AT_FDCWD, pathname, 0, name, value, size);
 892}
 893
 894SYSCALL_DEFINE4(lgetxattr, const char __user *, pathname,
 895		const char __user *, name, void __user *, value, size_t, size)
 896{
 897	return path_getxattrat(AT_FDCWD, pathname, AT_SYMLINK_NOFOLLOW, name,
 898			       value, size);
 899}
 900
 901SYSCALL_DEFINE4(fgetxattr, int, fd, const char __user *, name,
 902		void __user *, value, size_t, size)
 903{
 904	return path_getxattrat(fd, NULL, AT_EMPTY_PATH, name, value, size);
 
 
 
 
 
 
 
 
 
 905}
 906
 907/*
 908 * Extended attribute LIST operations
 909 */
 910static ssize_t
 911listxattr(struct dentry *d, char __user *list, size_t size)
 912{
 913	ssize_t error;
 914	char *klist = NULL;
 915
 916	if (size) {
 917		if (size > XATTR_LIST_MAX)
 918			size = XATTR_LIST_MAX;
 919		klist = kvmalloc(size, GFP_KERNEL);
 920		if (!klist)
 921			return -ENOMEM;
 922	}
 923
 924	error = vfs_listxattr(d, klist, size);
 925	if (error > 0) {
 926		if (size && copy_to_user(list, klist, error))
 927			error = -EFAULT;
 928	} else if (error == -ERANGE && size >= XATTR_LIST_MAX) {
 929		/* The file system tried to returned a list bigger
 930		   than XATTR_LIST_MAX bytes. Not possible. */
 931		error = -E2BIG;
 932	}
 933
 934	kvfree(klist);
 935
 936	return error;
 937}
 938
 939static
 940ssize_t file_listxattr(struct file *f, char __user *list, size_t size)
 941{
 942	audit_file(f);
 943	return listxattr(f->f_path.dentry, list, size);
 944}
 945
 946/* unconditionally consumes filename */
 947static
 948ssize_t filename_listxattr(int dfd, struct filename *filename,
 949			   unsigned int lookup_flags,
 950			   char __user *list, size_t size)
 951{
 952	struct path path;
 953	ssize_t error;
 954retry:
 955	error = filename_lookup(dfd, filename, lookup_flags, &path, NULL);
 956	if (error)
 957		goto out;
 958	error = listxattr(path.dentry, list, size);
 959	path_put(&path);
 960	if (retry_estale(error, lookup_flags)) {
 961		lookup_flags |= LOOKUP_REVAL;
 962		goto retry;
 963	}
 964out:
 965	putname(filename);
 966	return error;
 967}
 968
 969static ssize_t path_listxattrat(int dfd, const char __user *pathname,
 970				unsigned int at_flags, char __user *list,
 971				size_t size)
 972{
 973	struct filename *filename;
 974	int lookup_flags;
 975
 976	if ((at_flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
 977		return -EINVAL;
 978
 979	filename = getname_maybe_null(pathname, at_flags);
 980	if (!filename) {
 981		CLASS(fd, f)(dfd);
 982		if (fd_empty(f))
 983			return -EBADF;
 984		return file_listxattr(fd_file(f), list, size);
 985	}
 986
 987	lookup_flags = (at_flags & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
 988	return filename_listxattr(dfd, filename, lookup_flags, list, size);
 989}
 990
 991SYSCALL_DEFINE5(listxattrat, int, dfd, const char __user *, pathname,
 992		unsigned int, at_flags,
 993		char __user *, list, size_t, size)
 994{
 995	return path_listxattrat(dfd, pathname, at_flags, list, size);
 996}
 997
 998SYSCALL_DEFINE3(listxattr, const char __user *, pathname, char __user *, list,
 999		size_t, size)
1000{
1001	return path_listxattrat(AT_FDCWD, pathname, 0, list, size);
1002}
1003
1004SYSCALL_DEFINE3(llistxattr, const char __user *, pathname, char __user *, list,
1005		size_t, size)
1006{
1007	return path_listxattrat(AT_FDCWD, pathname, AT_SYMLINK_NOFOLLOW, list, size);
1008}
1009
1010SYSCALL_DEFINE3(flistxattr, int, fd, char __user *, list, size_t, size)
1011{
1012	return path_listxattrat(fd, NULL, AT_EMPTY_PATH, list, size);
 
 
 
 
 
 
 
 
1013}
1014
1015/*
1016 * Extended attribute REMOVE operations
1017 */
1018static long
1019removexattr(struct mnt_idmap *idmap, struct dentry *d, const char *name)
 
1020{
1021	if (is_posix_acl_xattr(name))
1022		return vfs_remove_acl(idmap, d, name);
1023	return vfs_removexattr(idmap, d, name);
1024}
 
 
 
 
1025
1026static int file_removexattr(struct file *f, struct xattr_name *kname)
1027{
1028	int error = mnt_want_write_file(f);
1029
1030	if (!error) {
1031		audit_file(f);
1032		error = removexattr(file_mnt_idmap(f),
1033				    f->f_path.dentry, kname->name);
1034		mnt_drop_write_file(f);
1035	}
1036	return error;
1037}
1038
1039/* unconditionally consumes filename */
1040static int filename_removexattr(int dfd, struct filename *filename,
1041				unsigned int lookup_flags, struct xattr_name *kname)
1042{
1043	struct path path;
1044	int error;
1045
1046retry:
1047	error = filename_lookup(dfd, filename, lookup_flags, &path, NULL);
1048	if (error)
1049		goto out;
1050	error = mnt_want_write(path.mnt);
1051	if (!error) {
1052		error = removexattr(mnt_idmap(path.mnt), path.dentry, kname->name);
1053		mnt_drop_write(path.mnt);
1054	}
1055	path_put(&path);
1056	if (retry_estale(error, lookup_flags)) {
1057		lookup_flags |= LOOKUP_REVAL;
1058		goto retry;
1059	}
1060out:
1061	putname(filename);
1062	return error;
1063}
1064
1065static int path_removexattrat(int dfd, const char __user *pathname,
1066			      unsigned int at_flags, const char __user *name)
1067{
1068	struct xattr_name kname;
1069	struct filename *filename;
1070	unsigned int lookup_flags;
1071	int error;
1072
1073	if ((at_flags & ~(AT_SYMLINK_NOFOLLOW | AT_EMPTY_PATH)) != 0)
1074		return -EINVAL;
1075
1076	error = import_xattr_name(&kname, name);
1077	if (error)
1078		return error;
1079
1080	filename = getname_maybe_null(pathname, at_flags);
1081	if (!filename) {
1082		CLASS(fd, f)(dfd);
1083		if (fd_empty(f))
1084			return -EBADF;
1085		return file_removexattr(fd_file(f), &kname);
1086	}
1087	lookup_flags = (at_flags & AT_SYMLINK_NOFOLLOW) ? 0 : LOOKUP_FOLLOW;
1088	return filename_removexattr(dfd, filename, lookup_flags, &kname);
1089}
1090
1091SYSCALL_DEFINE4(removexattrat, int, dfd, const char __user *, pathname,
1092		unsigned int, at_flags, const char __user *, name)
1093{
1094	return path_removexattrat(dfd, pathname, at_flags, name);
1095}
1096
1097SYSCALL_DEFINE2(removexattr, const char __user *, pathname,
1098		const char __user *, name)
1099{
1100	return path_removexattrat(AT_FDCWD, pathname, 0, name);
1101}
1102
1103SYSCALL_DEFINE2(lremovexattr, const char __user *, pathname,
1104		const char __user *, name)
1105{
1106	return path_removexattrat(AT_FDCWD, pathname, AT_SYMLINK_NOFOLLOW, name);
1107}
1108
1109SYSCALL_DEFINE2(fremovexattr, int, fd, const char __user *, name)
1110{
1111	return path_removexattrat(fd, NULL, AT_EMPTY_PATH, name);
 
 
 
 
 
 
 
 
 
 
 
 
 
1112}
1113
1114int xattr_list_one(char **buffer, ssize_t *remaining_size, const char *name)
1115{
1116	size_t len;
1117
1118	len = strlen(name) + 1;
1119	if (*buffer) {
1120		if (*remaining_size < len)
1121			return -ERANGE;
1122		memcpy(*buffer, name, len);
1123		*buffer += len;
1124	}
1125	*remaining_size -= len;
1126	return 0;
1127}
1128
1129/**
1130 * generic_listxattr - run through a dentry's xattr list() operations
1131 * @dentry: dentry to list the xattrs
1132 * @buffer: result buffer
1133 * @buffer_size: size of @buffer
1134 *
1135 * Combine the results of the list() operation from every xattr_handler in the
1136 * xattr_handler stack.
1137 *
1138 * Note that this will not include the entries for POSIX ACLs.
1139 */
1140ssize_t
1141generic_listxattr(struct dentry *dentry, char *buffer, size_t buffer_size)
1142{
1143	const struct xattr_handler *handler, * const *handlers = dentry->d_sb->s_xattr;
1144	ssize_t remaining_size = buffer_size;
 
1145
1146	for_each_xattr_handler(handlers, handler) {
1147		int err;
1148
1149		if (!handler->name || (handler->list && !handler->list(dentry)))
1150			continue;
1151		err = xattr_list_one(&buffer, &remaining_size, handler->name);
1152		if (err)
1153			return err;
1154	}
1155
1156	return buffer_size - remaining_size;
1157}
1158EXPORT_SYMBOL(generic_listxattr);
1159
1160/**
1161 * xattr_full_name  -  Compute full attribute name from suffix
1162 *
1163 * @handler:	handler of the xattr_handler operation
1164 * @name:	name passed to the xattr_handler operation
1165 *
1166 * The get and set xattr handler operations are called with the remainder of
1167 * the attribute name after skipping the handler's prefix: for example, "foo"
1168 * is passed to the get operation of a handler with prefix "user." to get
1169 * attribute "user.foo".  The full name is still "there" in the name though.
1170 *
1171 * Note: the list xattr handler operation when called from the vfs is passed a
1172 * NULL name; some file systems use this operation internally, with varying
1173 * semantics.
1174 */
1175const char *xattr_full_name(const struct xattr_handler *handler,
1176			    const char *name)
1177{
1178	size_t prefix_len = strlen(xattr_prefix(handler));
1179
1180	return name - prefix_len;
1181}
1182EXPORT_SYMBOL(xattr_full_name);
1183
1184/**
1185 * simple_xattr_space - estimate the memory used by a simple xattr
1186 * @name: the full name of the xattr
1187 * @size: the size of its value
1188 *
1189 * This takes no account of how much larger the two slab objects actually are:
1190 * that would depend on the slab implementation, when what is required is a
1191 * deterministic number, which grows with name length and size and quantity.
1192 *
1193 * Return: The approximate number of bytes of memory used by such an xattr.
1194 */
1195size_t simple_xattr_space(const char *name, size_t size)
1196{
1197	/*
1198	 * Use "40" instead of sizeof(struct simple_xattr), to return the
1199	 * same result on 32-bit and 64-bit, and even if simple_xattr grows.
1200	 */
1201	return 40 + size + strlen(name);
1202}
1203
1204/**
1205 * simple_xattr_free - free an xattr object
1206 * @xattr: the xattr object
1207 *
1208 * Free the xattr object. Can handle @xattr being NULL.
1209 */
1210void simple_xattr_free(struct simple_xattr *xattr)
1211{
1212	if (xattr)
1213		kfree(xattr->name);
1214	kvfree(xattr);
1215}
1216
1217/**
1218 * simple_xattr_alloc - allocate new xattr object
1219 * @value: value of the xattr object
1220 * @size: size of @value
1221 *
1222 * Allocate a new xattr object and initialize respective members. The caller is
1223 * responsible for handling the name of the xattr.
1224 *
1225 * Return: On success a new xattr object is returned. On failure NULL is
1226 * returned.
1227 */
1228struct simple_xattr *simple_xattr_alloc(const void *value, size_t size)
1229{
1230	struct simple_xattr *new_xattr;
1231	size_t len;
1232
1233	/* wrap around? */
1234	len = sizeof(*new_xattr) + size;
1235	if (len < sizeof(*new_xattr))
1236		return NULL;
1237
1238	new_xattr = kvmalloc(len, GFP_KERNEL_ACCOUNT);
1239	if (!new_xattr)
1240		return NULL;
1241
1242	new_xattr->size = size;
1243	memcpy(new_xattr->value, value, size);
1244	return new_xattr;
1245}
1246
1247/**
1248 * rbtree_simple_xattr_cmp - compare xattr name with current rbtree xattr entry
1249 * @key: xattr name
1250 * @node: current node
1251 *
1252 * Compare the xattr name with the xattr name attached to @node in the rbtree.
1253 *
1254 * Return: Negative value if continuing left, positive if continuing right, 0
1255 * if the xattr attached to @node matches @key.
1256 */
1257static int rbtree_simple_xattr_cmp(const void *key, const struct rb_node *node)
1258{
1259	const char *xattr_name = key;
1260	const struct simple_xattr *xattr;
1261
1262	xattr = rb_entry(node, struct simple_xattr, rb_node);
1263	return strcmp(xattr->name, xattr_name);
1264}
1265
1266/**
1267 * rbtree_simple_xattr_node_cmp - compare two xattr rbtree nodes
1268 * @new_node: new node
1269 * @node: current node
1270 *
1271 * Compare the xattr attached to @new_node with the xattr attached to @node.
1272 *
1273 * Return: Negative value if continuing left, positive if continuing right, 0
1274 * if the xattr attached to @new_node matches the xattr attached to @node.
1275 */
1276static int rbtree_simple_xattr_node_cmp(struct rb_node *new_node,
1277					const struct rb_node *node)
1278{
1279	struct simple_xattr *xattr;
1280	xattr = rb_entry(new_node, struct simple_xattr, rb_node);
1281	return rbtree_simple_xattr_cmp(xattr->name, node);
1282}
1283
1284/**
1285 * simple_xattr_get - get an xattr object
1286 * @xattrs: the header of the xattr object
1287 * @name: the name of the xattr to retrieve
1288 * @buffer: the buffer to store the value into
1289 * @size: the size of @buffer
1290 *
1291 * Try to find and retrieve the xattr object associated with @name.
1292 * If @buffer is provided store the value of @xattr in @buffer
1293 * otherwise just return the length. The size of @buffer is limited
1294 * to XATTR_SIZE_MAX which currently is 65536.
1295 *
1296 * Return: On success the length of the xattr value is returned. On error a
1297 * negative error code is returned.
1298 */
1299int simple_xattr_get(struct simple_xattrs *xattrs, const char *name,
1300		     void *buffer, size_t size)
1301{
1302	struct simple_xattr *xattr = NULL;
1303	struct rb_node *rbp;
1304	int ret = -ENODATA;
1305
1306	read_lock(&xattrs->lock);
1307	rbp = rb_find(name, &xattrs->rb_root, rbtree_simple_xattr_cmp);
1308	if (rbp) {
1309		xattr = rb_entry(rbp, struct simple_xattr, rb_node);
1310		ret = xattr->size;
1311		if (buffer) {
1312			if (size < xattr->size)
1313				ret = -ERANGE;
1314			else
1315				memcpy(buffer, xattr->value, xattr->size);
1316		}
1317	}
1318	read_unlock(&xattrs->lock);
1319	return ret;
1320}
1321
1322/**
1323 * simple_xattr_set - set an xattr object
1324 * @xattrs: the header of the xattr object
1325 * @name: the name of the xattr to retrieve
1326 * @value: the value to store along the xattr
1327 * @size: the size of @value
1328 * @flags: the flags determining how to set the xattr
1329 *
1330 * Set a new xattr object.
1331 * If @value is passed a new xattr object will be allocated. If XATTR_REPLACE
1332 * is specified in @flags a matching xattr object for @name must already exist.
1333 * If it does it will be replaced with the new xattr object. If it doesn't we
1334 * fail. If XATTR_CREATE is specified and a matching xattr does already exist
1335 * we fail. If it doesn't we create a new xattr. If @flags is zero we simply
1336 * insert the new xattr replacing any existing one.
1337 *
1338 * If @value is empty and a matching xattr object is found we delete it if
1339 * XATTR_REPLACE is specified in @flags or @flags is zero.
1340 *
1341 * If @value is empty and no matching xattr object for @name is found we do
1342 * nothing if XATTR_CREATE is specified in @flags or @flags is zero. For
1343 * XATTR_REPLACE we fail as mentioned above.
1344 *
1345 * Return: On success, the removed or replaced xattr is returned, to be freed
1346 * by the caller; or NULL if none. On failure a negative error code is returned.
1347 */
1348struct simple_xattr *simple_xattr_set(struct simple_xattrs *xattrs,
1349				      const char *name, const void *value,
1350				      size_t size, int flags)
1351{
1352	struct simple_xattr *old_xattr = NULL, *new_xattr = NULL;
1353	struct rb_node *parent = NULL, **rbp;
1354	int err = 0, ret;
1355
1356	/* value == NULL means remove */
1357	if (value) {
1358		new_xattr = simple_xattr_alloc(value, size);
1359		if (!new_xattr)
1360			return ERR_PTR(-ENOMEM);
1361
1362		new_xattr->name = kstrdup(name, GFP_KERNEL_ACCOUNT);
1363		if (!new_xattr->name) {
1364			simple_xattr_free(new_xattr);
1365			return ERR_PTR(-ENOMEM);
1366		}
1367	}
1368
1369	write_lock(&xattrs->lock);
1370	rbp = &xattrs->rb_root.rb_node;
1371	while (*rbp) {
1372		parent = *rbp;
1373		ret = rbtree_simple_xattr_cmp(name, *rbp);
1374		if (ret < 0)
1375			rbp = &(*rbp)->rb_left;
1376		else if (ret > 0)
1377			rbp = &(*rbp)->rb_right;
1378		else
1379			old_xattr = rb_entry(*rbp, struct simple_xattr, rb_node);
1380		if (old_xattr)
1381			break;
1382	}
1383
1384	if (old_xattr) {
1385		/* Fail if XATTR_CREATE is requested and the xattr exists. */
1386		if (flags & XATTR_CREATE) {
1387			err = -EEXIST;
1388			goto out_unlock;
1389		}
1390
1391		if (new_xattr)
1392			rb_replace_node(&old_xattr->rb_node,
1393					&new_xattr->rb_node, &xattrs->rb_root);
1394		else
1395			rb_erase(&old_xattr->rb_node, &xattrs->rb_root);
1396	} else {
1397		/* Fail if XATTR_REPLACE is requested but no xattr is found. */
1398		if (flags & XATTR_REPLACE) {
1399			err = -ENODATA;
1400			goto out_unlock;
1401		}
1402
1403		/*
1404		 * If XATTR_CREATE or no flags are specified together with a
1405		 * new value simply insert it.
1406		 */
1407		if (new_xattr) {
1408			rb_link_node(&new_xattr->rb_node, parent, rbp);
1409			rb_insert_color(&new_xattr->rb_node, &xattrs->rb_root);
1410		}
1411
1412		/*
1413		 * If XATTR_CREATE or no flags are specified and neither an
1414		 * old or new xattr exist then we don't need to do anything.
1415		 */
1416	}
1417
1418out_unlock:
1419	write_unlock(&xattrs->lock);
1420	if (!err)
1421		return old_xattr;
1422	simple_xattr_free(new_xattr);
1423	return ERR_PTR(err);
1424}
1425
1426static bool xattr_is_trusted(const char *name)
1427{
1428	return !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN);
1429}
1430
1431/**
1432 * simple_xattr_list - list all xattr objects
1433 * @inode: inode from which to get the xattrs
1434 * @xattrs: the header of the xattr object
1435 * @buffer: the buffer to store all xattrs into
1436 * @size: the size of @buffer
1437 *
1438 * List all xattrs associated with @inode. If @buffer is NULL we returned
1439 * the required size of the buffer. If @buffer is provided we store the
1440 * xattrs value into it provided it is big enough.
1441 *
1442 * Note, the number of xattr names that can be listed with listxattr(2) is
1443 * limited to XATTR_LIST_MAX aka 65536 bytes. If a larger buffer is passed
1444 * then vfs_listxattr() caps it to XATTR_LIST_MAX and if more xattr names
1445 * are found it will return -E2BIG.
1446 *
1447 * Return: On success the required size or the size of the copied xattrs is
1448 * returned. On error a negative error code is returned.
1449 */
1450ssize_t simple_xattr_list(struct inode *inode, struct simple_xattrs *xattrs,
1451			  char *buffer, size_t size)
1452{
1453	bool trusted = ns_capable_noaudit(&init_user_ns, CAP_SYS_ADMIN);
1454	struct simple_xattr *xattr;
1455	struct rb_node *rbp;
1456	ssize_t remaining_size = size;
1457	int err = 0;
1458
1459	err = posix_acl_listxattr(inode, &buffer, &remaining_size);
1460	if (err)
1461		return err;
1462
1463	read_lock(&xattrs->lock);
1464	for (rbp = rb_first(&xattrs->rb_root); rbp; rbp = rb_next(rbp)) {
1465		xattr = rb_entry(rbp, struct simple_xattr, rb_node);
1466
1467		/* skip "trusted." attributes for unprivileged callers */
1468		if (!trusted && xattr_is_trusted(xattr->name))
1469			continue;
1470
1471		err = xattr_list_one(&buffer, &remaining_size, xattr->name);
1472		if (err)
1473			break;
1474	}
1475	read_unlock(&xattrs->lock);
1476
1477	return err ? err : size - remaining_size;
1478}
1479
1480/**
1481 * rbtree_simple_xattr_less - compare two xattr rbtree nodes
1482 * @new_node: new node
1483 * @node: current node
1484 *
1485 * Compare the xattr attached to @new_node with the xattr attached to @node.
1486 * Note that this function technically tolerates duplicate entries.
1487 *
1488 * Return: True if insertion point in the rbtree is found.
1489 */
1490static bool rbtree_simple_xattr_less(struct rb_node *new_node,
1491				     const struct rb_node *node)
1492{
1493	return rbtree_simple_xattr_node_cmp(new_node, node) < 0;
1494}
1495
1496/**
1497 * simple_xattr_add - add xattr objects
1498 * @xattrs: the header of the xattr object
1499 * @new_xattr: the xattr object to add
1500 *
1501 * Add an xattr object to @xattrs. This assumes no replacement or removal
1502 * of matching xattrs is wanted. Should only be called during inode
1503 * initialization when a few distinct initial xattrs are supposed to be set.
1504 */
1505void simple_xattr_add(struct simple_xattrs *xattrs,
1506		      struct simple_xattr *new_xattr)
1507{
1508	write_lock(&xattrs->lock);
1509	rb_add(&new_xattr->rb_node, &xattrs->rb_root, rbtree_simple_xattr_less);
1510	write_unlock(&xattrs->lock);
1511}
1512
1513/**
1514 * simple_xattrs_init - initialize new xattr header
1515 * @xattrs: header to initialize
1516 *
1517 * Initialize relevant fields of a an xattr header.
1518 */
1519void simple_xattrs_init(struct simple_xattrs *xattrs)
1520{
1521	xattrs->rb_root = RB_ROOT;
1522	rwlock_init(&xattrs->lock);
1523}
1524
1525/**
1526 * simple_xattrs_free - free xattrs
1527 * @xattrs: xattr header whose xattrs to destroy
1528 * @freed_space: approximate number of bytes of memory freed from @xattrs
1529 *
1530 * Destroy all xattrs in @xattr. When this is called no one can hold a
1531 * reference to any of the xattrs anymore.
1532 */
1533void simple_xattrs_free(struct simple_xattrs *xattrs, size_t *freed_space)
1534{
1535	struct rb_node *rbp;
1536
1537	if (freed_space)
1538		*freed_space = 0;
1539	rbp = rb_first(&xattrs->rb_root);
1540	while (rbp) {
1541		struct simple_xattr *xattr;
1542		struct rb_node *rbp_next;
1543
1544		rbp_next = rb_next(rbp);
1545		xattr = rb_entry(rbp, struct simple_xattr, rb_node);
1546		rb_erase(&xattr->rb_node, &xattrs->rb_root);
1547		if (freed_space)
1548			*freed_space += simple_xattr_space(xattr->name,
1549							   xattr->size);
1550		simple_xattr_free(xattr);
1551		rbp = rbp_next;
1552	}
1553}