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(error: -EIO);
64 return ERR_PTR(error: -EOPNOTSUPP);
65 }
66 for_each_xattr_handler(handlers, handler) {
67 const char *n;
68
69 n = strcmp_prefix(a: *name, a_prefix: xattr_prefix(handler));
70 if (n) {
71 if (!handler->prefix ^ !*n) {
72 if (*n)
73 continue;
74 return ERR_PTR(error: -EINVAL);
75 }
76 *name = n;
77 return handler;
78 }
79 }
80 return ERR_PTR(error: -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: &name);
194 if (IS_ERR(ptr: handler))
195 return PTR_ERR(ptr: 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, name: 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, ivalue: &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: &delegated_inode);
327 if (!error)
328 goto retry_deleg;
329 }
330 if (value != orig_value)
331 kfree(objp: 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: &buffer, alloc: false);
347 goto out_noalloc;
348 }
349
350 len = security_inode_getsecurity(idmap, inode, name, buffer: &buffer,
351 alloc: 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(objp: 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: &name);
389 if (IS_ERR(ptr: handler))
390 return PTR_ERR(ptr: 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(objp: *xattr_value, new_size: 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: &name);
419 if (IS_ERR(ptr: handler))
420 return PTR_ERR(ptr: 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, name: 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, buffer: list, buffer_size: 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: &name);
514 if (IS_ERR(ptr: handler))
515 return PTR_ERR(ptr: 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: &delegated_inode);
581 if (!error)
582 goto retry_deleg;
583 }
584
585 return error;
586}
587EXPORT_SYMBOL_GPL(vfs_removexattr);
588
589/*
590 * Extended attribute SET operations
591 */
592
593int setxattr_copy(const char __user *name, struct xattr_ctx *ctx)
594{
595 int error;
596
597 if (ctx->flags & ~(XATTR_CREATE|XATTR_REPLACE))
598 return -EINVAL;
599
600 error = strncpy_from_user(dst: ctx->kname->name, src: name,
601 count: sizeof(ctx->kname->name));
602 if (error == 0 || error == sizeof(ctx->kname->name))
603 return -ERANGE;
604 if (error < 0)
605 return error;
606
607 error = 0;
608 if (ctx->size) {
609 if (ctx->size > XATTR_SIZE_MAX)
610 return -E2BIG;
611
612 ctx->kvalue = vmemdup_user(ctx->cvalue, ctx->size);
613 if (IS_ERR(ptr: ctx->kvalue)) {
614 error = PTR_ERR(ptr: ctx->kvalue);
615 ctx->kvalue = NULL;
616 }
617 }
618
619 return error;
620}
621
622int do_setxattr(struct mnt_idmap *idmap, struct dentry *dentry,
623 struct xattr_ctx *ctx)
624{
625 if (is_posix_acl_xattr(name: ctx->kname->name))
626 return do_set_acl(idmap, dentry, acl_name: ctx->kname->name,
627 kvalue: ctx->kvalue, size: ctx->size);
628
629 return vfs_setxattr(idmap, dentry, ctx->kname->name,
630 ctx->kvalue, ctx->size, ctx->flags);
631}
632
633static long
634setxattr(struct mnt_idmap *idmap, struct dentry *d,
635 const char __user *name, const void __user *value, size_t size,
636 int flags)
637{
638 struct xattr_name kname;
639 struct xattr_ctx ctx = {
640 .cvalue = value,
641 .kvalue = NULL,
642 .size = size,
643 .kname = &kname,
644 .flags = flags,
645 };
646 int error;
647
648 error = setxattr_copy(name, ctx: &ctx);
649 if (error)
650 return error;
651
652 error = do_setxattr(idmap, dentry: d, ctx: &ctx);
653
654 kvfree(addr: ctx.kvalue);
655 return error;
656}
657
658static int path_setxattr(const char __user *pathname,
659 const char __user *name, const void __user *value,
660 size_t size, int flags, unsigned int lookup_flags)
661{
662 struct path path;
663 int error;
664
665retry:
666 error = user_path_at(AT_FDCWD, name: pathname, flags: lookup_flags, path: &path);
667 if (error)
668 return error;
669 error = mnt_want_write(mnt: path.mnt);
670 if (!error) {
671 error = setxattr(idmap: mnt_idmap(mnt: path.mnt), d: path.dentry, name,
672 value, size, flags);
673 mnt_drop_write(mnt: path.mnt);
674 }
675 path_put(&path);
676 if (retry_estale(error, flags: lookup_flags)) {
677 lookup_flags |= LOOKUP_REVAL;
678 goto retry;
679 }
680 return error;
681}
682
683SYSCALL_DEFINE5(setxattr, const char __user *, pathname,
684 const char __user *, name, const void __user *, value,
685 size_t, size, int, flags)
686{
687 return path_setxattr(pathname, name, value, size, flags, LOOKUP_FOLLOW);
688}
689
690SYSCALL_DEFINE5(lsetxattr, const char __user *, pathname,
691 const char __user *, name, const void __user *, value,
692 size_t, size, int, flags)
693{
694 return path_setxattr(pathname, name, value, size, flags, lookup_flags: 0);
695}
696
697SYSCALL_DEFINE5(fsetxattr, int, fd, const char __user *, name,
698 const void __user *,value, size_t, size, int, flags)
699{
700 struct fd f = fdget(fd);
701 int error = -EBADF;
702
703 if (!f.file)
704 return error;
705 audit_file(file: f.file);
706 error = mnt_want_write_file(file: f.file);
707 if (!error) {
708 error = setxattr(idmap: file_mnt_idmap(file: f.file),
709 d: f.file->f_path.dentry, name,
710 value, size, flags);
711 mnt_drop_write_file(file: f.file);
712 }
713 fdput(fd: f);
714 return error;
715}
716
717/*
718 * Extended attribute GET operations
719 */
720ssize_t
721do_getxattr(struct mnt_idmap *idmap, struct dentry *d,
722 struct xattr_ctx *ctx)
723{
724 ssize_t error;
725 char *kname = ctx->kname->name;
726
727 if (ctx->size) {
728 if (ctx->size > XATTR_SIZE_MAX)
729 ctx->size = XATTR_SIZE_MAX;
730 ctx->kvalue = kvzalloc(size: ctx->size, GFP_KERNEL);
731 if (!ctx->kvalue)
732 return -ENOMEM;
733 }
734
735 if (is_posix_acl_xattr(name: ctx->kname->name))
736 error = do_get_acl(idmap, dentry: d, acl_name: kname, kvalue: ctx->kvalue, size: ctx->size);
737 else
738 error = vfs_getxattr(idmap, d, kname, ctx->kvalue, ctx->size);
739 if (error > 0) {
740 if (ctx->size && copy_to_user(to: ctx->value, from: ctx->kvalue, n: error))
741 error = -EFAULT;
742 } else if (error == -ERANGE && ctx->size >= XATTR_SIZE_MAX) {
743 /* The file system tried to returned a value bigger
744 than XATTR_SIZE_MAX bytes. Not possible. */
745 error = -E2BIG;
746 }
747
748 return error;
749}
750
751static ssize_t
752getxattr(struct mnt_idmap *idmap, struct dentry *d,
753 const char __user *name, void __user *value, size_t size)
754{
755 ssize_t error;
756 struct xattr_name kname;
757 struct xattr_ctx ctx = {
758 .value = value,
759 .kvalue = NULL,
760 .size = size,
761 .kname = &kname,
762 .flags = 0,
763 };
764
765 error = strncpy_from_user(dst: kname.name, src: name, count: sizeof(kname.name));
766 if (error == 0 || error == sizeof(kname.name))
767 error = -ERANGE;
768 if (error < 0)
769 return error;
770
771 error = do_getxattr(idmap, d, ctx: &ctx);
772
773 kvfree(addr: ctx.kvalue);
774 return error;
775}
776
777static ssize_t path_getxattr(const char __user *pathname,
778 const char __user *name, void __user *value,
779 size_t size, unsigned int lookup_flags)
780{
781 struct path path;
782 ssize_t error;
783retry:
784 error = user_path_at(AT_FDCWD, name: pathname, flags: lookup_flags, path: &path);
785 if (error)
786 return error;
787 error = getxattr(idmap: mnt_idmap(mnt: path.mnt), d: path.dentry, name, value, size);
788 path_put(&path);
789 if (retry_estale(error, flags: lookup_flags)) {
790 lookup_flags |= LOOKUP_REVAL;
791 goto retry;
792 }
793 return error;
794}
795
796SYSCALL_DEFINE4(getxattr, const char __user *, pathname,
797 const char __user *, name, void __user *, value, size_t, size)
798{
799 return path_getxattr(pathname, name, value, size, LOOKUP_FOLLOW);
800}
801
802SYSCALL_DEFINE4(lgetxattr, const char __user *, pathname,
803 const char __user *, name, void __user *, value, size_t, size)
804{
805 return path_getxattr(pathname, name, value, size, lookup_flags: 0);
806}
807
808SYSCALL_DEFINE4(fgetxattr, int, fd, const char __user *, name,
809 void __user *, value, size_t, size)
810{
811 struct fd f = fdget(fd);
812 ssize_t error = -EBADF;
813
814 if (!f.file)
815 return error;
816 audit_file(file: f.file);
817 error = getxattr(idmap: file_mnt_idmap(file: f.file), d: f.file->f_path.dentry,
818 name, value, size);
819 fdput(fd: f);
820 return error;
821}
822
823/*
824 * Extended attribute LIST operations
825 */
826static ssize_t
827listxattr(struct dentry *d, char __user *list, size_t size)
828{
829 ssize_t error;
830 char *klist = NULL;
831
832 if (size) {
833 if (size > XATTR_LIST_MAX)
834 size = XATTR_LIST_MAX;
835 klist = kvmalloc(size, GFP_KERNEL);
836 if (!klist)
837 return -ENOMEM;
838 }
839
840 error = vfs_listxattr(d, klist, size);
841 if (error > 0) {
842 if (size && copy_to_user(to: list, from: klist, n: error))
843 error = -EFAULT;
844 } else if (error == -ERANGE && size >= XATTR_LIST_MAX) {
845 /* The file system tried to returned a list bigger
846 than XATTR_LIST_MAX bytes. Not possible. */
847 error = -E2BIG;
848 }
849
850 kvfree(addr: klist);
851
852 return error;
853}
854
855static ssize_t path_listxattr(const char __user *pathname, char __user *list,
856 size_t size, unsigned int lookup_flags)
857{
858 struct path path;
859 ssize_t error;
860retry:
861 error = user_path_at(AT_FDCWD, name: pathname, flags: lookup_flags, path: &path);
862 if (error)
863 return error;
864 error = listxattr(d: path.dentry, list, size);
865 path_put(&path);
866 if (retry_estale(error, flags: lookup_flags)) {
867 lookup_flags |= LOOKUP_REVAL;
868 goto retry;
869 }
870 return error;
871}
872
873SYSCALL_DEFINE3(listxattr, const char __user *, pathname, char __user *, list,
874 size_t, size)
875{
876 return path_listxattr(pathname, list, size, LOOKUP_FOLLOW);
877}
878
879SYSCALL_DEFINE3(llistxattr, const char __user *, pathname, char __user *, list,
880 size_t, size)
881{
882 return path_listxattr(pathname, list, size, lookup_flags: 0);
883}
884
885SYSCALL_DEFINE3(flistxattr, int, fd, char __user *, list, size_t, size)
886{
887 struct fd f = fdget(fd);
888 ssize_t error = -EBADF;
889
890 if (!f.file)
891 return error;
892 audit_file(file: f.file);
893 error = listxattr(d: f.file->f_path.dentry, list, size);
894 fdput(fd: f);
895 return error;
896}
897
898/*
899 * Extended attribute REMOVE operations
900 */
901static long
902removexattr(struct mnt_idmap *idmap, struct dentry *d,
903 const char __user *name)
904{
905 int error;
906 char kname[XATTR_NAME_MAX + 1];
907
908 error = strncpy_from_user(dst: kname, src: name, count: sizeof(kname));
909 if (error == 0 || error == sizeof(kname))
910 error = -ERANGE;
911 if (error < 0)
912 return error;
913
914 if (is_posix_acl_xattr(name: kname))
915 return vfs_remove_acl(idmap, dentry: d, acl_name: kname);
916
917 return vfs_removexattr(idmap, d, kname);
918}
919
920static int path_removexattr(const char __user *pathname,
921 const char __user *name, unsigned int lookup_flags)
922{
923 struct path path;
924 int error;
925retry:
926 error = user_path_at(AT_FDCWD, name: pathname, flags: lookup_flags, path: &path);
927 if (error)
928 return error;
929 error = mnt_want_write(mnt: path.mnt);
930 if (!error) {
931 error = removexattr(idmap: mnt_idmap(mnt: path.mnt), d: path.dentry, name);
932 mnt_drop_write(mnt: path.mnt);
933 }
934 path_put(&path);
935 if (retry_estale(error, flags: lookup_flags)) {
936 lookup_flags |= LOOKUP_REVAL;
937 goto retry;
938 }
939 return error;
940}
941
942SYSCALL_DEFINE2(removexattr, const char __user *, pathname,
943 const char __user *, name)
944{
945 return path_removexattr(pathname, name, LOOKUP_FOLLOW);
946}
947
948SYSCALL_DEFINE2(lremovexattr, const char __user *, pathname,
949 const char __user *, name)
950{
951 return path_removexattr(pathname, name, lookup_flags: 0);
952}
953
954SYSCALL_DEFINE2(fremovexattr, int, fd, const char __user *, name)
955{
956 struct fd f = fdget(fd);
957 int error = -EBADF;
958
959 if (!f.file)
960 return error;
961 audit_file(file: f.file);
962 error = mnt_want_write_file(file: f.file);
963 if (!error) {
964 error = removexattr(idmap: file_mnt_idmap(file: f.file),
965 d: f.file->f_path.dentry, name);
966 mnt_drop_write_file(file: f.file);
967 }
968 fdput(fd: f);
969 return error;
970}
971
972int xattr_list_one(char **buffer, ssize_t *remaining_size, const char *name)
973{
974 size_t len;
975
976 len = strlen(name) + 1;
977 if (*buffer) {
978 if (*remaining_size < len)
979 return -ERANGE;
980 memcpy(*buffer, name, len);
981 *buffer += len;
982 }
983 *remaining_size -= len;
984 return 0;
985}
986
987/**
988 * generic_listxattr - run through a dentry's xattr list() operations
989 * @dentry: dentry to list the xattrs
990 * @buffer: result buffer
991 * @buffer_size: size of @buffer
992 *
993 * Combine the results of the list() operation from every xattr_handler in the
994 * xattr_handler stack.
995 *
996 * Note that this will not include the entries for POSIX ACLs.
997 */
998ssize_t
999generic_listxattr(struct dentry *dentry, char *buffer, size_t buffer_size)
1000{
1001 const struct xattr_handler *handler, * const *handlers = dentry->d_sb->s_xattr;
1002 ssize_t remaining_size = buffer_size;
1003 int err = 0;
1004
1005 for_each_xattr_handler(handlers, handler) {
1006 if (!handler->name || (handler->list && !handler->list(dentry)))
1007 continue;
1008 err = xattr_list_one(buffer: &buffer, remaining_size: &remaining_size, name: handler->name);
1009 if (err)
1010 return err;
1011 }
1012
1013 return err ? err : buffer_size - remaining_size;
1014}
1015EXPORT_SYMBOL(generic_listxattr);
1016
1017/**
1018 * xattr_full_name - Compute full attribute name from suffix
1019 *
1020 * @handler: handler of the xattr_handler operation
1021 * @name: name passed to the xattr_handler operation
1022 *
1023 * The get and set xattr handler operations are called with the remainder of
1024 * the attribute name after skipping the handler's prefix: for example, "foo"
1025 * is passed to the get operation of a handler with prefix "user." to get
1026 * attribute "user.foo". The full name is still "there" in the name though.
1027 *
1028 * Note: the list xattr handler operation when called from the vfs is passed a
1029 * NULL name; some file systems use this operation internally, with varying
1030 * semantics.
1031 */
1032const char *xattr_full_name(const struct xattr_handler *handler,
1033 const char *name)
1034{
1035 size_t prefix_len = strlen(xattr_prefix(handler));
1036
1037 return name - prefix_len;
1038}
1039EXPORT_SYMBOL(xattr_full_name);
1040
1041/**
1042 * simple_xattr_space - estimate the memory used by a simple xattr
1043 * @name: the full name of the xattr
1044 * @size: the size of its value
1045 *
1046 * This takes no account of how much larger the two slab objects actually are:
1047 * that would depend on the slab implementation, when what is required is a
1048 * deterministic number, which grows with name length and size and quantity.
1049 *
1050 * Return: The approximate number of bytes of memory used by such an xattr.
1051 */
1052size_t simple_xattr_space(const char *name, size_t size)
1053{
1054 /*
1055 * Use "40" instead of sizeof(struct simple_xattr), to return the
1056 * same result on 32-bit and 64-bit, and even if simple_xattr grows.
1057 */
1058 return 40 + size + strlen(name);
1059}
1060
1061/**
1062 * simple_xattr_free - free an xattr object
1063 * @xattr: the xattr object
1064 *
1065 * Free the xattr object. Can handle @xattr being NULL.
1066 */
1067void simple_xattr_free(struct simple_xattr *xattr)
1068{
1069 if (xattr)
1070 kfree(objp: xattr->name);
1071 kvfree(addr: xattr);
1072}
1073
1074/**
1075 * simple_xattr_alloc - allocate new xattr object
1076 * @value: value of the xattr object
1077 * @size: size of @value
1078 *
1079 * Allocate a new xattr object and initialize respective members. The caller is
1080 * responsible for handling the name of the xattr.
1081 *
1082 * Return: On success a new xattr object is returned. On failure NULL is
1083 * returned.
1084 */
1085struct simple_xattr *simple_xattr_alloc(const void *value, size_t size)
1086{
1087 struct simple_xattr *new_xattr;
1088 size_t len;
1089
1090 /* wrap around? */
1091 len = sizeof(*new_xattr) + size;
1092 if (len < sizeof(*new_xattr))
1093 return NULL;
1094
1095 new_xattr = kvmalloc(size: len, GFP_KERNEL_ACCOUNT);
1096 if (!new_xattr)
1097 return NULL;
1098
1099 new_xattr->size = size;
1100 memcpy(new_xattr->value, value, size);
1101 return new_xattr;
1102}
1103
1104/**
1105 * rbtree_simple_xattr_cmp - compare xattr name with current rbtree xattr entry
1106 * @key: xattr name
1107 * @node: current node
1108 *
1109 * Compare the xattr name with the xattr name attached to @node in the rbtree.
1110 *
1111 * Return: Negative value if continuing left, positive if continuing right, 0
1112 * if the xattr attached to @node matches @key.
1113 */
1114static int rbtree_simple_xattr_cmp(const void *key, const struct rb_node *node)
1115{
1116 const char *xattr_name = key;
1117 const struct simple_xattr *xattr;
1118
1119 xattr = rb_entry(node, struct simple_xattr, rb_node);
1120 return strcmp(xattr->name, xattr_name);
1121}
1122
1123/**
1124 * rbtree_simple_xattr_node_cmp - compare two xattr rbtree nodes
1125 * @new_node: new node
1126 * @node: current node
1127 *
1128 * Compare the xattr attached to @new_node with the xattr attached to @node.
1129 *
1130 * Return: Negative value if continuing left, positive if continuing right, 0
1131 * if the xattr attached to @new_node matches the xattr attached to @node.
1132 */
1133static int rbtree_simple_xattr_node_cmp(struct rb_node *new_node,
1134 const struct rb_node *node)
1135{
1136 struct simple_xattr *xattr;
1137 xattr = rb_entry(new_node, struct simple_xattr, rb_node);
1138 return rbtree_simple_xattr_cmp(key: xattr->name, node);
1139}
1140
1141/**
1142 * simple_xattr_get - get an xattr object
1143 * @xattrs: the header of the xattr object
1144 * @name: the name of the xattr to retrieve
1145 * @buffer: the buffer to store the value into
1146 * @size: the size of @buffer
1147 *
1148 * Try to find and retrieve the xattr object associated with @name.
1149 * If @buffer is provided store the value of @xattr in @buffer
1150 * otherwise just return the length. The size of @buffer is limited
1151 * to XATTR_SIZE_MAX which currently is 65536.
1152 *
1153 * Return: On success the length of the xattr value is returned. On error a
1154 * negative error code is returned.
1155 */
1156int simple_xattr_get(struct simple_xattrs *xattrs, const char *name,
1157 void *buffer, size_t size)
1158{
1159 struct simple_xattr *xattr = NULL;
1160 struct rb_node *rbp;
1161 int ret = -ENODATA;
1162
1163 read_lock(&xattrs->lock);
1164 rbp = rb_find(key: name, tree: &xattrs->rb_root, cmp: rbtree_simple_xattr_cmp);
1165 if (rbp) {
1166 xattr = rb_entry(rbp, struct simple_xattr, rb_node);
1167 ret = xattr->size;
1168 if (buffer) {
1169 if (size < xattr->size)
1170 ret = -ERANGE;
1171 else
1172 memcpy(buffer, xattr->value, xattr->size);
1173 }
1174 }
1175 read_unlock(&xattrs->lock);
1176 return ret;
1177}
1178
1179/**
1180 * simple_xattr_set - set an xattr object
1181 * @xattrs: the header of the xattr object
1182 * @name: the name of the xattr to retrieve
1183 * @value: the value to store along the xattr
1184 * @size: the size of @value
1185 * @flags: the flags determining how to set the xattr
1186 *
1187 * Set a new xattr object.
1188 * If @value is passed a new xattr object will be allocated. If XATTR_REPLACE
1189 * is specified in @flags a matching xattr object for @name must already exist.
1190 * If it does it will be replaced with the new xattr object. If it doesn't we
1191 * fail. If XATTR_CREATE is specified and a matching xattr does already exist
1192 * we fail. If it doesn't we create a new xattr. If @flags is zero we simply
1193 * insert the new xattr replacing any existing one.
1194 *
1195 * If @value is empty and a matching xattr object is found we delete it if
1196 * XATTR_REPLACE is specified in @flags or @flags is zero.
1197 *
1198 * If @value is empty and no matching xattr object for @name is found we do
1199 * nothing if XATTR_CREATE is specified in @flags or @flags is zero. For
1200 * XATTR_REPLACE we fail as mentioned above.
1201 *
1202 * Return: On success, the removed or replaced xattr is returned, to be freed
1203 * by the caller; or NULL if none. On failure a negative error code is returned.
1204 */
1205struct simple_xattr *simple_xattr_set(struct simple_xattrs *xattrs,
1206 const char *name, const void *value,
1207 size_t size, int flags)
1208{
1209 struct simple_xattr *old_xattr = NULL, *new_xattr = NULL;
1210 struct rb_node *parent = NULL, **rbp;
1211 int err = 0, ret;
1212
1213 /* value == NULL means remove */
1214 if (value) {
1215 new_xattr = simple_xattr_alloc(value, size);
1216 if (!new_xattr)
1217 return ERR_PTR(error: -ENOMEM);
1218
1219 new_xattr->name = kstrdup(s: name, GFP_KERNEL_ACCOUNT);
1220 if (!new_xattr->name) {
1221 simple_xattr_free(xattr: new_xattr);
1222 return ERR_PTR(error: -ENOMEM);
1223 }
1224 }
1225
1226 write_lock(&xattrs->lock);
1227 rbp = &xattrs->rb_root.rb_node;
1228 while (*rbp) {
1229 parent = *rbp;
1230 ret = rbtree_simple_xattr_cmp(key: name, node: *rbp);
1231 if (ret < 0)
1232 rbp = &(*rbp)->rb_left;
1233 else if (ret > 0)
1234 rbp = &(*rbp)->rb_right;
1235 else
1236 old_xattr = rb_entry(*rbp, struct simple_xattr, rb_node);
1237 if (old_xattr)
1238 break;
1239 }
1240
1241 if (old_xattr) {
1242 /* Fail if XATTR_CREATE is requested and the xattr exists. */
1243 if (flags & XATTR_CREATE) {
1244 err = -EEXIST;
1245 goto out_unlock;
1246 }
1247
1248 if (new_xattr)
1249 rb_replace_node(victim: &old_xattr->rb_node,
1250 new: &new_xattr->rb_node, root: &xattrs->rb_root);
1251 else
1252 rb_erase(&old_xattr->rb_node, &xattrs->rb_root);
1253 } else {
1254 /* Fail if XATTR_REPLACE is requested but no xattr is found. */
1255 if (flags & XATTR_REPLACE) {
1256 err = -ENODATA;
1257 goto out_unlock;
1258 }
1259
1260 /*
1261 * If XATTR_CREATE or no flags are specified together with a
1262 * new value simply insert it.
1263 */
1264 if (new_xattr) {
1265 rb_link_node(node: &new_xattr->rb_node, parent, rb_link: rbp);
1266 rb_insert_color(&new_xattr->rb_node, &xattrs->rb_root);
1267 }
1268
1269 /*
1270 * If XATTR_CREATE or no flags are specified and neither an
1271 * old or new xattr exist then we don't need to do anything.
1272 */
1273 }
1274
1275out_unlock:
1276 write_unlock(&xattrs->lock);
1277 if (!err)
1278 return old_xattr;
1279 simple_xattr_free(xattr: new_xattr);
1280 return ERR_PTR(error: err);
1281}
1282
1283static bool xattr_is_trusted(const char *name)
1284{
1285 return !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN);
1286}
1287
1288/**
1289 * simple_xattr_list - list all xattr objects
1290 * @inode: inode from which to get the xattrs
1291 * @xattrs: the header of the xattr object
1292 * @buffer: the buffer to store all xattrs into
1293 * @size: the size of @buffer
1294 *
1295 * List all xattrs associated with @inode. If @buffer is NULL we returned
1296 * the required size of the buffer. If @buffer is provided we store the
1297 * xattrs value into it provided it is big enough.
1298 *
1299 * Note, the number of xattr names that can be listed with listxattr(2) is
1300 * limited to XATTR_LIST_MAX aka 65536 bytes. If a larger buffer is passed
1301 * then vfs_listxattr() caps it to XATTR_LIST_MAX and if more xattr names
1302 * are found it will return -E2BIG.
1303 *
1304 * Return: On success the required size or the size of the copied xattrs is
1305 * returned. On error a negative error code is returned.
1306 */
1307ssize_t simple_xattr_list(struct inode *inode, struct simple_xattrs *xattrs,
1308 char *buffer, size_t size)
1309{
1310 bool trusted = ns_capable_noaudit(ns: &init_user_ns, CAP_SYS_ADMIN);
1311 struct simple_xattr *xattr;
1312 struct rb_node *rbp;
1313 ssize_t remaining_size = size;
1314 int err = 0;
1315
1316 err = posix_acl_listxattr(inode, buffer: &buffer, remaining_size: &remaining_size);
1317 if (err)
1318 return err;
1319
1320 read_lock(&xattrs->lock);
1321 for (rbp = rb_first(&xattrs->rb_root); rbp; rbp = rb_next(rbp)) {
1322 xattr = rb_entry(rbp, struct simple_xattr, rb_node);
1323
1324 /* skip "trusted." attributes for unprivileged callers */
1325 if (!trusted && xattr_is_trusted(name: xattr->name))
1326 continue;
1327
1328 err = xattr_list_one(buffer: &buffer, remaining_size: &remaining_size, name: xattr->name);
1329 if (err)
1330 break;
1331 }
1332 read_unlock(&xattrs->lock);
1333
1334 return err ? err : size - remaining_size;
1335}
1336
1337/**
1338 * rbtree_simple_xattr_less - compare two xattr rbtree nodes
1339 * @new_node: new node
1340 * @node: current node
1341 *
1342 * Compare the xattr attached to @new_node with the xattr attached to @node.
1343 * Note that this function technically tolerates duplicate entries.
1344 *
1345 * Return: True if insertion point in the rbtree is found.
1346 */
1347static bool rbtree_simple_xattr_less(struct rb_node *new_node,
1348 const struct rb_node *node)
1349{
1350 return rbtree_simple_xattr_node_cmp(new_node, node) < 0;
1351}
1352
1353/**
1354 * simple_xattr_add - add xattr objects
1355 * @xattrs: the header of the xattr object
1356 * @new_xattr: the xattr object to add
1357 *
1358 * Add an xattr object to @xattrs. This assumes no replacement or removal
1359 * of matching xattrs is wanted. Should only be called during inode
1360 * initialization when a few distinct initial xattrs are supposed to be set.
1361 */
1362void simple_xattr_add(struct simple_xattrs *xattrs,
1363 struct simple_xattr *new_xattr)
1364{
1365 write_lock(&xattrs->lock);
1366 rb_add(node: &new_xattr->rb_node, tree: &xattrs->rb_root, less: rbtree_simple_xattr_less);
1367 write_unlock(&xattrs->lock);
1368}
1369
1370/**
1371 * simple_xattrs_init - initialize new xattr header
1372 * @xattrs: header to initialize
1373 *
1374 * Initialize relevant fields of a an xattr header.
1375 */
1376void simple_xattrs_init(struct simple_xattrs *xattrs)
1377{
1378 xattrs->rb_root = RB_ROOT;
1379 rwlock_init(&xattrs->lock);
1380}
1381
1382/**
1383 * simple_xattrs_free - free xattrs
1384 * @xattrs: xattr header whose xattrs to destroy
1385 * @freed_space: approximate number of bytes of memory freed from @xattrs
1386 *
1387 * Destroy all xattrs in @xattr. When this is called no one can hold a
1388 * reference to any of the xattrs anymore.
1389 */
1390void simple_xattrs_free(struct simple_xattrs *xattrs, size_t *freed_space)
1391{
1392 struct rb_node *rbp;
1393
1394 if (freed_space)
1395 *freed_space = 0;
1396 rbp = rb_first(&xattrs->rb_root);
1397 while (rbp) {
1398 struct simple_xattr *xattr;
1399 struct rb_node *rbp_next;
1400
1401 rbp_next = rb_next(rbp);
1402 xattr = rb_entry(rbp, struct simple_xattr, rb_node);
1403 rb_erase(&xattr->rb_node, &xattrs->rb_root);
1404 if (freed_space)
1405 *freed_space += simple_xattr_space(name: xattr->name,
1406 size: xattr->size);
1407 simple_xattr_free(xattr);
1408 rbp = rbp_next;
1409 }
1410}
1411

source code of linux/fs/xattr.c