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(error: -EIO);
65 return ERR_PTR(error: -EOPNOTSUPP);
66 }
67 for_each_xattr_handler(handlers, handler) {
68 const char *n;
69
70 n = strcmp_prefix(a: *name, a_prefix: xattr_prefix(handler));
71 if (n) {
72 if (!handler->prefix ^ !*n) {
73 if (*n)
74 continue;
75 return ERR_PTR(error: -EINVAL);
76 }
77 *name = n;
78 return handler;
79 }
80 }
81 return ERR_PTR(error: -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: &name);
195 if (IS_ERR(ptr: handler))
196 return PTR_ERR(ptr: 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, name: 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, ivalue: &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: &delegated_inode);
328 if (!error)
329 goto retry_deleg;
330 }
331 if (value != orig_value)
332 kfree(objp: 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: &buffer, alloc: false);
348 goto out_noalloc;
349 }
350
351 len = security_inode_getsecurity(idmap, inode, name, buffer: &buffer,
352 alloc: 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(objp: 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: &name);
390 if (IS_ERR(ptr: handler))
391 return PTR_ERR(ptr: 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(objp: *xattr_value, new_size: 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: &name);
420 if (IS_ERR(ptr: handler))
421 return PTR_ERR(ptr: 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, name: 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, buffer: list, buffer_size: 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: &name);
515 if (IS_ERR(ptr: handler))
516 return PTR_ERR(ptr: 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, xattr_name: 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: &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(dst: ctx->kname->name, src: name,
602 count: 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(ptr: ctx->kvalue)) {
615 error = PTR_ERR(ptr: 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(name: ctx->kname->name))
627 return do_set_acl(idmap, dentry, acl_name: ctx->kname->name,
628 kvalue: ctx->kvalue, size: 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: &ctx);
650 if (error)
651 return error;
652
653 error = do_setxattr(idmap, dentry: d, ctx: &ctx);
654
655 kvfree(addr: 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, name: pathname, flags: lookup_flags, path: &path);
668 if (error)
669 return error;
670 error = mnt_want_write(mnt: path.mnt);
671 if (!error) {
672 error = setxattr(idmap: mnt_idmap(mnt: path.mnt), d: path.dentry, name,
673 value, size, flags);
674 mnt_drop_write(mnt: path.mnt);
675 }
676 path_put(&path);
677 if (retry_estale(error, flags: 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, lookup_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(file: f.file);
707 error = mnt_want_write_file(file: f.file);
708 if (!error) {
709 error = setxattr(idmap: file_mnt_idmap(file: f.file),
710 d: f.file->f_path.dentry, name,
711 value, size, flags);
712 mnt_drop_write_file(file: f.file);
713 }
714 fdput(fd: 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(size: ctx->size, GFP_KERNEL);
732 if (!ctx->kvalue)
733 return -ENOMEM;
734 }
735
736 if (is_posix_acl_xattr(name: ctx->kname->name))
737 error = do_get_acl(idmap, dentry: d, acl_name: kname, kvalue: ctx->kvalue, size: 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(to: ctx->value, from: ctx->kvalue, n: 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(dst: kname.name, src: name, count: 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: &ctx);
773
774 kvfree(addr: 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, name: pathname, flags: lookup_flags, path: &path);
786 if (error)
787 return error;
788 error = getxattr(idmap: mnt_idmap(mnt: path.mnt), d: path.dentry, name, value, size);
789 path_put(&path);
790 if (retry_estale(error, flags: 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, lookup_flags: 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(file: f.file);
818 error = getxattr(idmap: file_mnt_idmap(file: f.file), d: f.file->f_path.dentry,
819 name, value, size);
820 fdput(fd: 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(to: list, from: klist, n: 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(addr: 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, name: pathname, flags: lookup_flags, path: &path);
863 if (error)
864 return error;
865 error = listxattr(d: path.dentry, list, size);
866 path_put(&path);
867 if (retry_estale(error, flags: 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, lookup_flags: 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(file: f.file);
894 error = listxattr(d: f.file->f_path.dentry, list, size);
895 fdput(fd: 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(dst: kname, src: name, count: 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(name: kname))
916 return vfs_remove_acl(idmap, dentry: d, acl_name: 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, name: pathname, flags: lookup_flags, path: &path);
928 if (error)
929 return error;
930 error = mnt_want_write(mnt: path.mnt);
931 if (!error) {
932 error = removexattr(idmap: mnt_idmap(mnt: path.mnt), d: path.dentry, name);
933 mnt_drop_write(mnt: path.mnt);
934 }
935 path_put(&path);
936 if (retry_estale(error, flags: 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, lookup_flags: 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(file: f.file);
963 error = mnt_want_write_file(file: f.file);
964 if (!error) {
965 error = removexattr(idmap: file_mnt_idmap(file: f.file),
966 d: f.file->f_path.dentry, name);
967 mnt_drop_write_file(file: f.file);
968 }
969 fdput(fd: 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: &buffer, remaining_size: &remaining_size, name: 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(objp: xattr->name);
1072 kvfree(addr: 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(size: 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(key: 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(key: name, tree: &xattrs->rb_root, cmp: 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(error: -ENOMEM);
1219
1220 new_xattr->name = kstrdup(s: name, GFP_KERNEL_ACCOUNT);
1221 if (!new_xattr->name) {
1222 simple_xattr_free(xattr: new_xattr);
1223 return ERR_PTR(error: -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(key: name, node: *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(victim: &old_xattr->rb_node,
1251 new: &new_xattr->rb_node, root: &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(node: &new_xattr->rb_node, parent, rb_link: 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(xattr: new_xattr);
1281 return ERR_PTR(error: 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(ns: &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: &buffer, remaining_size: &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(name: xattr->name))
1327 continue;
1328
1329 err = xattr_list_one(buffer: &buffer, remaining_size: &remaining_size, name: 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(node: &new_xattr->rb_node, tree: &xattrs->rb_root, less: 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(name: xattr->name,
1407 size: xattr->size);
1408 simple_xattr_free(xattr);
1409 rbp = rbp_next;
1410 }
1411}
1412

source code of linux/fs/xattr.c