Loading...
1/*
2 File: fs/xattr.c
3
4 Extended attribute handling.
5
6 Copyright (C) 2001 by Andreas Gruenbacher <a.gruenbacher@computer.org>
7 Copyright (C) 2001 SGI - Silicon Graphics, Inc <linux-xfs@oss.sgi.com>
8 Copyright (c) 2004 Red Hat, Inc., James Morris <jmorris@redhat.com>
9 */
10#include <linux/fs.h>
11#include <linux/slab.h>
12#include <linux/file.h>
13#include <linux/xattr.h>
14#include <linux/mount.h>
15#include <linux/namei.h>
16#include <linux/security.h>
17#include <linux/evm.h>
18#include <linux/syscalls.h>
19#include <linux/export.h>
20#include <linux/fsnotify.h>
21#include <linux/audit.h>
22#include <linux/vmalloc.h>
23#include <linux/posix_acl_xattr.h>
24
25#include <linux/uaccess.h>
26
27static const char *
28strcmp_prefix(const char *a, const char *a_prefix)
29{
30 while (*a_prefix && *a == *a_prefix) {
31 a++;
32 a_prefix++;
33 }
34 return *a_prefix ? NULL : a;
35}
36
37/*
38 * In order to implement different sets of xattr operations for each xattr
39 * prefix, a filesystem should create a null-terminated array of struct
40 * xattr_handler (one for each prefix) and hang a pointer to it off of the
41 * s_xattr field of the superblock.
42 */
43#define for_each_xattr_handler(handlers, handler) \
44 if (handlers) \
45 for ((handler) = *(handlers)++; \
46 (handler) != NULL; \
47 (handler) = *(handlers)++)
48
49/*
50 * Find the xattr_handler with the matching prefix.
51 */
52static const struct xattr_handler *
53xattr_resolve_name(struct inode *inode, const char **name)
54{
55 const struct xattr_handler **handlers = inode->i_sb->s_xattr;
56 const struct xattr_handler *handler;
57
58 if (!(inode->i_opflags & IOP_XATTR)) {
59 if (unlikely(is_bad_inode(inode)))
60 return ERR_PTR(-EIO);
61 return ERR_PTR(-EOPNOTSUPP);
62 }
63 for_each_xattr_handler(handlers, handler) {
64 const char *n;
65
66 n = strcmp_prefix(*name, xattr_prefix(handler));
67 if (n) {
68 if (!handler->prefix ^ !*n) {
69 if (*n)
70 continue;
71 return ERR_PTR(-EINVAL);
72 }
73 *name = n;
74 return handler;
75 }
76 }
77 return ERR_PTR(-EOPNOTSUPP);
78}
79
80/*
81 * Check permissions for extended attribute access. This is a bit complicated
82 * because different namespaces have very different rules.
83 */
84static int
85xattr_permission(struct inode *inode, const char *name, int mask)
86{
87 /*
88 * We can never set or remove an extended attribute on a read-only
89 * filesystem or on an immutable / append-only inode.
90 */
91 if (mask & MAY_WRITE) {
92 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
93 return -EPERM;
94 /*
95 * Updating an xattr will likely cause i_uid and i_gid
96 * to be writen back improperly if their true value is
97 * unknown to the vfs.
98 */
99 if (HAS_UNMAPPED_ID(inode))
100 return -EPERM;
101 }
102
103 /*
104 * No restriction for security.* and system.* from the VFS. Decision
105 * on these is left to the underlying filesystem / security module.
106 */
107 if (!strncmp(name, XATTR_SECURITY_PREFIX, XATTR_SECURITY_PREFIX_LEN) ||
108 !strncmp(name, XATTR_SYSTEM_PREFIX, XATTR_SYSTEM_PREFIX_LEN))
109 return 0;
110
111 /*
112 * The trusted.* namespace can only be accessed by privileged users.
113 */
114 if (!strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN)) {
115 if (!capable(CAP_SYS_ADMIN))
116 return (mask & MAY_WRITE) ? -EPERM : -ENODATA;
117 return 0;
118 }
119
120 /*
121 * In the user.* namespace, only regular files and directories can have
122 * extended attributes. For sticky directories, only the owner and
123 * privileged users can write attributes.
124 */
125 if (!strncmp(name, XATTR_USER_PREFIX, XATTR_USER_PREFIX_LEN)) {
126 if (!S_ISREG(inode->i_mode) && !S_ISDIR(inode->i_mode))
127 return (mask & MAY_WRITE) ? -EPERM : -ENODATA;
128 if (S_ISDIR(inode->i_mode) && (inode->i_mode & S_ISVTX) &&
129 (mask & MAY_WRITE) && !inode_owner_or_capable(inode))
130 return -EPERM;
131 }
132
133 return inode_permission(inode, mask);
134}
135
136int
137__vfs_setxattr(struct dentry *dentry, struct inode *inode, const char *name,
138 const void *value, size_t size, int flags)
139{
140 const struct xattr_handler *handler;
141
142 handler = xattr_resolve_name(inode, &name);
143 if (IS_ERR(handler))
144 return PTR_ERR(handler);
145 if (!handler->set)
146 return -EOPNOTSUPP;
147 if (size == 0)
148 value = ""; /* empty EA, do not remove */
149 return handler->set(handler, dentry, inode, name, value, size, flags);
150}
151EXPORT_SYMBOL(__vfs_setxattr);
152
153/**
154 * __vfs_setxattr_noperm - perform setxattr operation without performing
155 * permission checks.
156 *
157 * @dentry - object to perform setxattr on
158 * @name - xattr name to set
159 * @value - value to set @name to
160 * @size - size of @value
161 * @flags - flags to pass into filesystem operations
162 *
163 * returns the result of the internal setxattr or setsecurity operations.
164 *
165 * This function requires the caller to lock the inode's i_mutex before it
166 * is executed. It also assumes that the caller will make the appropriate
167 * permission checks.
168 */
169int __vfs_setxattr_noperm(struct dentry *dentry, const char *name,
170 const void *value, size_t size, int flags)
171{
172 struct inode *inode = dentry->d_inode;
173 int error = -EAGAIN;
174 int issec = !strncmp(name, XATTR_SECURITY_PREFIX,
175 XATTR_SECURITY_PREFIX_LEN);
176
177 if (issec)
178 inode->i_flags &= ~S_NOSEC;
179 if (inode->i_opflags & IOP_XATTR) {
180 error = __vfs_setxattr(dentry, inode, name, value, size, flags);
181 if (!error) {
182 fsnotify_xattr(dentry);
183 security_inode_post_setxattr(dentry, name, value,
184 size, flags);
185 }
186 } else {
187 if (unlikely(is_bad_inode(inode)))
188 return -EIO;
189 }
190 if (error == -EAGAIN) {
191 error = -EOPNOTSUPP;
192
193 if (issec) {
194 const char *suffix = name + XATTR_SECURITY_PREFIX_LEN;
195
196 error = security_inode_setsecurity(inode, suffix, value,
197 size, flags);
198 if (!error)
199 fsnotify_xattr(dentry);
200 }
201 }
202
203 return error;
204}
205
206
207int
208vfs_setxattr(struct dentry *dentry, const char *name, const void *value,
209 size_t size, int flags)
210{
211 struct inode *inode = dentry->d_inode;
212 int error;
213
214 error = xattr_permission(inode, name, MAY_WRITE);
215 if (error)
216 return error;
217
218 inode_lock(inode);
219 error = security_inode_setxattr(dentry, name, value, size, flags);
220 if (error)
221 goto out;
222
223 error = __vfs_setxattr_noperm(dentry, name, value, size, flags);
224
225out:
226 inode_unlock(inode);
227 return error;
228}
229EXPORT_SYMBOL_GPL(vfs_setxattr);
230
231ssize_t
232xattr_getsecurity(struct inode *inode, const char *name, void *value,
233 size_t size)
234{
235 void *buffer = NULL;
236 ssize_t len;
237
238 if (!value || !size) {
239 len = security_inode_getsecurity(inode, name, &buffer, false);
240 goto out_noalloc;
241 }
242
243 len = security_inode_getsecurity(inode, name, &buffer, true);
244 if (len < 0)
245 return len;
246 if (size < len) {
247 len = -ERANGE;
248 goto out;
249 }
250 memcpy(value, buffer, len);
251out:
252 security_release_secctx(buffer, len);
253out_noalloc:
254 return len;
255}
256EXPORT_SYMBOL_GPL(xattr_getsecurity);
257
258/*
259 * vfs_getxattr_alloc - allocate memory, if necessary, before calling getxattr
260 *
261 * Allocate memory, if not already allocated, or re-allocate correct size,
262 * before retrieving the extended attribute.
263 *
264 * Returns the result of alloc, if failed, or the getxattr operation.
265 */
266ssize_t
267vfs_getxattr_alloc(struct dentry *dentry, const char *name, char **xattr_value,
268 size_t xattr_size, gfp_t flags)
269{
270 const struct xattr_handler *handler;
271 struct inode *inode = dentry->d_inode;
272 char *value = *xattr_value;
273 int error;
274
275 error = xattr_permission(inode, name, MAY_READ);
276 if (error)
277 return error;
278
279 handler = xattr_resolve_name(inode, &name);
280 if (IS_ERR(handler))
281 return PTR_ERR(handler);
282 if (!handler->get)
283 return -EOPNOTSUPP;
284 error = handler->get(handler, dentry, inode, name, NULL, 0);
285 if (error < 0)
286 return error;
287
288 if (!value || (error > xattr_size)) {
289 value = krealloc(*xattr_value, error + 1, flags);
290 if (!value)
291 return -ENOMEM;
292 memset(value, 0, error + 1);
293 }
294
295 error = handler->get(handler, dentry, inode, name, value, error);
296 *xattr_value = value;
297 return error;
298}
299
300ssize_t
301__vfs_getxattr(struct dentry *dentry, struct inode *inode, const char *name,
302 void *value, size_t size)
303{
304 const struct xattr_handler *handler;
305
306 handler = xattr_resolve_name(inode, &name);
307 if (IS_ERR(handler))
308 return PTR_ERR(handler);
309 if (!handler->get)
310 return -EOPNOTSUPP;
311 return handler->get(handler, dentry, inode, name, value, size);
312}
313EXPORT_SYMBOL(__vfs_getxattr);
314
315ssize_t
316vfs_getxattr(struct dentry *dentry, const char *name, void *value, size_t size)
317{
318 struct inode *inode = dentry->d_inode;
319 int error;
320
321 error = xattr_permission(inode, name, MAY_READ);
322 if (error)
323 return error;
324
325 error = security_inode_getxattr(dentry, name);
326 if (error)
327 return error;
328
329 if (!strncmp(name, XATTR_SECURITY_PREFIX,
330 XATTR_SECURITY_PREFIX_LEN)) {
331 const char *suffix = name + XATTR_SECURITY_PREFIX_LEN;
332 int ret = xattr_getsecurity(inode, suffix, value, size);
333 /*
334 * Only overwrite the return value if a security module
335 * is actually active.
336 */
337 if (ret == -EOPNOTSUPP)
338 goto nolsm;
339 return ret;
340 }
341nolsm:
342 return __vfs_getxattr(dentry, inode, name, value, size);
343}
344EXPORT_SYMBOL_GPL(vfs_getxattr);
345
346ssize_t
347vfs_listxattr(struct dentry *dentry, char *list, size_t size)
348{
349 struct inode *inode = d_inode(dentry);
350 ssize_t error;
351
352 error = security_inode_listxattr(dentry);
353 if (error)
354 return error;
355 if (inode->i_op->listxattr && (inode->i_opflags & IOP_XATTR)) {
356 error = -EOPNOTSUPP;
357 error = inode->i_op->listxattr(dentry, list, size);
358 } else {
359 error = security_inode_listsecurity(inode, list, size);
360 if (size && error > size)
361 error = -ERANGE;
362 }
363 return error;
364}
365EXPORT_SYMBOL_GPL(vfs_listxattr);
366
367int
368__vfs_removexattr(struct dentry *dentry, const char *name)
369{
370 struct inode *inode = d_inode(dentry);
371 const struct xattr_handler *handler;
372
373 handler = xattr_resolve_name(inode, &name);
374 if (IS_ERR(handler))
375 return PTR_ERR(handler);
376 if (!handler->set)
377 return -EOPNOTSUPP;
378 return handler->set(handler, dentry, inode, name, NULL, 0, XATTR_REPLACE);
379}
380EXPORT_SYMBOL(__vfs_removexattr);
381
382int
383vfs_removexattr(struct dentry *dentry, const char *name)
384{
385 struct inode *inode = dentry->d_inode;
386 int error;
387
388 error = xattr_permission(inode, name, MAY_WRITE);
389 if (error)
390 return error;
391
392 inode_lock(inode);
393 error = security_inode_removexattr(dentry, name);
394 if (error)
395 goto out;
396
397 error = __vfs_removexattr(dentry, name);
398
399 if (!error) {
400 fsnotify_xattr(dentry);
401 evm_inode_post_removexattr(dentry, name);
402 }
403
404out:
405 inode_unlock(inode);
406 return error;
407}
408EXPORT_SYMBOL_GPL(vfs_removexattr);
409
410
411/*
412 * Extended attribute SET operations
413 */
414static long
415setxattr(struct dentry *d, const char __user *name, const void __user *value,
416 size_t size, int flags)
417{
418 int error;
419 void *kvalue = NULL;
420 char kname[XATTR_NAME_MAX + 1];
421
422 if (flags & ~(XATTR_CREATE|XATTR_REPLACE))
423 return -EINVAL;
424
425 error = strncpy_from_user(kname, name, sizeof(kname));
426 if (error == 0 || error == sizeof(kname))
427 error = -ERANGE;
428 if (error < 0)
429 return error;
430
431 if (size) {
432 if (size > XATTR_SIZE_MAX)
433 return -E2BIG;
434 kvalue = kmalloc(size, GFP_KERNEL | __GFP_NOWARN);
435 if (!kvalue) {
436 kvalue = vmalloc(size);
437 if (!kvalue)
438 return -ENOMEM;
439 }
440 if (copy_from_user(kvalue, value, size)) {
441 error = -EFAULT;
442 goto out;
443 }
444 if ((strcmp(kname, XATTR_NAME_POSIX_ACL_ACCESS) == 0) ||
445 (strcmp(kname, XATTR_NAME_POSIX_ACL_DEFAULT) == 0))
446 posix_acl_fix_xattr_from_user(kvalue, size);
447 }
448
449 error = vfs_setxattr(d, kname, kvalue, size, flags);
450out:
451 kvfree(kvalue);
452
453 return error;
454}
455
456static int path_setxattr(const char __user *pathname,
457 const char __user *name, const void __user *value,
458 size_t size, int flags, unsigned int lookup_flags)
459{
460 struct path path;
461 int error;
462retry:
463 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
464 if (error)
465 return error;
466 error = mnt_want_write(path.mnt);
467 if (!error) {
468 error = setxattr(path.dentry, name, value, size, flags);
469 mnt_drop_write(path.mnt);
470 }
471 path_put(&path);
472 if (retry_estale(error, lookup_flags)) {
473 lookup_flags |= LOOKUP_REVAL;
474 goto retry;
475 }
476 return error;
477}
478
479SYSCALL_DEFINE5(setxattr, const char __user *, pathname,
480 const char __user *, name, const void __user *, value,
481 size_t, size, int, flags)
482{
483 return path_setxattr(pathname, name, value, size, flags, LOOKUP_FOLLOW);
484}
485
486SYSCALL_DEFINE5(lsetxattr, const char __user *, pathname,
487 const char __user *, name, const void __user *, value,
488 size_t, size, int, flags)
489{
490 return path_setxattr(pathname, name, value, size, flags, 0);
491}
492
493SYSCALL_DEFINE5(fsetxattr, int, fd, const char __user *, name,
494 const void __user *,value, size_t, size, int, flags)
495{
496 struct fd f = fdget(fd);
497 int error = -EBADF;
498
499 if (!f.file)
500 return error;
501 audit_file(f.file);
502 error = mnt_want_write_file(f.file);
503 if (!error) {
504 error = setxattr(f.file->f_path.dentry, name, value, size, flags);
505 mnt_drop_write_file(f.file);
506 }
507 fdput(f);
508 return error;
509}
510
511/*
512 * Extended attribute GET operations
513 */
514static ssize_t
515getxattr(struct dentry *d, const char __user *name, void __user *value,
516 size_t size)
517{
518 ssize_t error;
519 void *kvalue = NULL;
520 char kname[XATTR_NAME_MAX + 1];
521
522 error = strncpy_from_user(kname, name, sizeof(kname));
523 if (error == 0 || error == sizeof(kname))
524 error = -ERANGE;
525 if (error < 0)
526 return error;
527
528 if (size) {
529 if (size > XATTR_SIZE_MAX)
530 size = XATTR_SIZE_MAX;
531 kvalue = kzalloc(size, GFP_KERNEL | __GFP_NOWARN);
532 if (!kvalue) {
533 kvalue = vmalloc(size);
534 if (!kvalue)
535 return -ENOMEM;
536 }
537 }
538
539 error = vfs_getxattr(d, kname, kvalue, size);
540 if (error > 0) {
541 if ((strcmp(kname, XATTR_NAME_POSIX_ACL_ACCESS) == 0) ||
542 (strcmp(kname, XATTR_NAME_POSIX_ACL_DEFAULT) == 0))
543 posix_acl_fix_xattr_to_user(kvalue, size);
544 if (size && copy_to_user(value, kvalue, error))
545 error = -EFAULT;
546 } else if (error == -ERANGE && size >= XATTR_SIZE_MAX) {
547 /* The file system tried to returned a value bigger
548 than XATTR_SIZE_MAX bytes. Not possible. */
549 error = -E2BIG;
550 }
551
552 kvfree(kvalue);
553
554 return error;
555}
556
557static ssize_t path_getxattr(const char __user *pathname,
558 const char __user *name, void __user *value,
559 size_t size, unsigned int lookup_flags)
560{
561 struct path path;
562 ssize_t error;
563retry:
564 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
565 if (error)
566 return error;
567 error = getxattr(path.dentry, name, value, size);
568 path_put(&path);
569 if (retry_estale(error, lookup_flags)) {
570 lookup_flags |= LOOKUP_REVAL;
571 goto retry;
572 }
573 return error;
574}
575
576SYSCALL_DEFINE4(getxattr, const char __user *, pathname,
577 const char __user *, name, void __user *, value, size_t, size)
578{
579 return path_getxattr(pathname, name, value, size, LOOKUP_FOLLOW);
580}
581
582SYSCALL_DEFINE4(lgetxattr, const char __user *, pathname,
583 const char __user *, name, void __user *, value, size_t, size)
584{
585 return path_getxattr(pathname, name, value, size, 0);
586}
587
588SYSCALL_DEFINE4(fgetxattr, int, fd, const char __user *, name,
589 void __user *, value, size_t, size)
590{
591 struct fd f = fdget(fd);
592 ssize_t error = -EBADF;
593
594 if (!f.file)
595 return error;
596 audit_file(f.file);
597 error = getxattr(f.file->f_path.dentry, name, value, size);
598 fdput(f);
599 return error;
600}
601
602/*
603 * Extended attribute LIST operations
604 */
605static ssize_t
606listxattr(struct dentry *d, char __user *list, size_t size)
607{
608 ssize_t error;
609 char *klist = NULL;
610
611 if (size) {
612 if (size > XATTR_LIST_MAX)
613 size = XATTR_LIST_MAX;
614 klist = kmalloc(size, __GFP_NOWARN | GFP_KERNEL);
615 if (!klist) {
616 klist = vmalloc(size);
617 if (!klist)
618 return -ENOMEM;
619 }
620 }
621
622 error = vfs_listxattr(d, klist, size);
623 if (error > 0) {
624 if (size && copy_to_user(list, klist, error))
625 error = -EFAULT;
626 } else if (error == -ERANGE && size >= XATTR_LIST_MAX) {
627 /* The file system tried to returned a list bigger
628 than XATTR_LIST_MAX bytes. Not possible. */
629 error = -E2BIG;
630 }
631
632 kvfree(klist);
633
634 return error;
635}
636
637static ssize_t path_listxattr(const char __user *pathname, char __user *list,
638 size_t size, unsigned int lookup_flags)
639{
640 struct path path;
641 ssize_t error;
642retry:
643 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
644 if (error)
645 return error;
646 error = listxattr(path.dentry, list, size);
647 path_put(&path);
648 if (retry_estale(error, lookup_flags)) {
649 lookup_flags |= LOOKUP_REVAL;
650 goto retry;
651 }
652 return error;
653}
654
655SYSCALL_DEFINE3(listxattr, const char __user *, pathname, char __user *, list,
656 size_t, size)
657{
658 return path_listxattr(pathname, list, size, LOOKUP_FOLLOW);
659}
660
661SYSCALL_DEFINE3(llistxattr, const char __user *, pathname, char __user *, list,
662 size_t, size)
663{
664 return path_listxattr(pathname, list, size, 0);
665}
666
667SYSCALL_DEFINE3(flistxattr, int, fd, char __user *, list, size_t, size)
668{
669 struct fd f = fdget(fd);
670 ssize_t error = -EBADF;
671
672 if (!f.file)
673 return error;
674 audit_file(f.file);
675 error = listxattr(f.file->f_path.dentry, list, size);
676 fdput(f);
677 return error;
678}
679
680/*
681 * Extended attribute REMOVE operations
682 */
683static long
684removexattr(struct dentry *d, const char __user *name)
685{
686 int error;
687 char kname[XATTR_NAME_MAX + 1];
688
689 error = strncpy_from_user(kname, name, sizeof(kname));
690 if (error == 0 || error == sizeof(kname))
691 error = -ERANGE;
692 if (error < 0)
693 return error;
694
695 return vfs_removexattr(d, kname);
696}
697
698static int path_removexattr(const char __user *pathname,
699 const char __user *name, unsigned int lookup_flags)
700{
701 struct path path;
702 int error;
703retry:
704 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
705 if (error)
706 return error;
707 error = mnt_want_write(path.mnt);
708 if (!error) {
709 error = removexattr(path.dentry, name);
710 mnt_drop_write(path.mnt);
711 }
712 path_put(&path);
713 if (retry_estale(error, lookup_flags)) {
714 lookup_flags |= LOOKUP_REVAL;
715 goto retry;
716 }
717 return error;
718}
719
720SYSCALL_DEFINE2(removexattr, const char __user *, pathname,
721 const char __user *, name)
722{
723 return path_removexattr(pathname, name, LOOKUP_FOLLOW);
724}
725
726SYSCALL_DEFINE2(lremovexattr, const char __user *, pathname,
727 const char __user *, name)
728{
729 return path_removexattr(pathname, name, 0);
730}
731
732SYSCALL_DEFINE2(fremovexattr, int, fd, const char __user *, name)
733{
734 struct fd f = fdget(fd);
735 int error = -EBADF;
736
737 if (!f.file)
738 return error;
739 audit_file(f.file);
740 error = mnt_want_write_file(f.file);
741 if (!error) {
742 error = removexattr(f.file->f_path.dentry, name);
743 mnt_drop_write_file(f.file);
744 }
745 fdput(f);
746 return error;
747}
748
749/*
750 * Combine the results of the list() operation from every xattr_handler in the
751 * list.
752 */
753ssize_t
754generic_listxattr(struct dentry *dentry, char *buffer, size_t buffer_size)
755{
756 const struct xattr_handler *handler, **handlers = dentry->d_sb->s_xattr;
757 unsigned int size = 0;
758
759 if (!buffer) {
760 for_each_xattr_handler(handlers, handler) {
761 if (!handler->name ||
762 (handler->list && !handler->list(dentry)))
763 continue;
764 size += strlen(handler->name) + 1;
765 }
766 } else {
767 char *buf = buffer;
768 size_t len;
769
770 for_each_xattr_handler(handlers, handler) {
771 if (!handler->name ||
772 (handler->list && !handler->list(dentry)))
773 continue;
774 len = strlen(handler->name);
775 if (len + 1 > buffer_size)
776 return -ERANGE;
777 memcpy(buf, handler->name, len + 1);
778 buf += len + 1;
779 buffer_size -= len + 1;
780 }
781 size = buf - buffer;
782 }
783 return size;
784}
785EXPORT_SYMBOL(generic_listxattr);
786
787/**
788 * xattr_full_name - Compute full attribute name from suffix
789 *
790 * @handler: handler of the xattr_handler operation
791 * @name: name passed to the xattr_handler operation
792 *
793 * The get and set xattr handler operations are called with the remainder of
794 * the attribute name after skipping the handler's prefix: for example, "foo"
795 * is passed to the get operation of a handler with prefix "user." to get
796 * attribute "user.foo". The full name is still "there" in the name though.
797 *
798 * Note: the list xattr handler operation when called from the vfs is passed a
799 * NULL name; some file systems use this operation internally, with varying
800 * semantics.
801 */
802const char *xattr_full_name(const struct xattr_handler *handler,
803 const char *name)
804{
805 size_t prefix_len = strlen(xattr_prefix(handler));
806
807 return name - prefix_len;
808}
809EXPORT_SYMBOL(xattr_full_name);
810
811/*
812 * Allocate new xattr and copy in the value; but leave the name to callers.
813 */
814struct simple_xattr *simple_xattr_alloc(const void *value, size_t size)
815{
816 struct simple_xattr *new_xattr;
817 size_t len;
818
819 /* wrap around? */
820 len = sizeof(*new_xattr) + size;
821 if (len < sizeof(*new_xattr))
822 return NULL;
823
824 new_xattr = kmalloc(len, GFP_KERNEL);
825 if (!new_xattr)
826 return NULL;
827
828 new_xattr->size = size;
829 memcpy(new_xattr->value, value, size);
830 return new_xattr;
831}
832
833/*
834 * xattr GET operation for in-memory/pseudo filesystems
835 */
836int simple_xattr_get(struct simple_xattrs *xattrs, const char *name,
837 void *buffer, size_t size)
838{
839 struct simple_xattr *xattr;
840 int ret = -ENODATA;
841
842 spin_lock(&xattrs->lock);
843 list_for_each_entry(xattr, &xattrs->head, list) {
844 if (strcmp(name, xattr->name))
845 continue;
846
847 ret = xattr->size;
848 if (buffer) {
849 if (size < xattr->size)
850 ret = -ERANGE;
851 else
852 memcpy(buffer, xattr->value, xattr->size);
853 }
854 break;
855 }
856 spin_unlock(&xattrs->lock);
857 return ret;
858}
859
860/**
861 * simple_xattr_set - xattr SET operation for in-memory/pseudo filesystems
862 * @xattrs: target simple_xattr list
863 * @name: name of the extended attribute
864 * @value: value of the xattr. If %NULL, will remove the attribute.
865 * @size: size of the new xattr
866 * @flags: %XATTR_{CREATE|REPLACE}
867 *
868 * %XATTR_CREATE is set, the xattr shouldn't exist already; otherwise fails
869 * with -EEXIST. If %XATTR_REPLACE is set, the xattr should exist;
870 * otherwise, fails with -ENODATA.
871 *
872 * Returns 0 on success, -errno on failure.
873 */
874int simple_xattr_set(struct simple_xattrs *xattrs, const char *name,
875 const void *value, size_t size, int flags)
876{
877 struct simple_xattr *xattr;
878 struct simple_xattr *new_xattr = NULL;
879 int err = 0;
880
881 /* value == NULL means remove */
882 if (value) {
883 new_xattr = simple_xattr_alloc(value, size);
884 if (!new_xattr)
885 return -ENOMEM;
886
887 new_xattr->name = kstrdup(name, GFP_KERNEL);
888 if (!new_xattr->name) {
889 kfree(new_xattr);
890 return -ENOMEM;
891 }
892 }
893
894 spin_lock(&xattrs->lock);
895 list_for_each_entry(xattr, &xattrs->head, list) {
896 if (!strcmp(name, xattr->name)) {
897 if (flags & XATTR_CREATE) {
898 xattr = new_xattr;
899 err = -EEXIST;
900 } else if (new_xattr) {
901 list_replace(&xattr->list, &new_xattr->list);
902 } else {
903 list_del(&xattr->list);
904 }
905 goto out;
906 }
907 }
908 if (flags & XATTR_REPLACE) {
909 xattr = new_xattr;
910 err = -ENODATA;
911 } else {
912 list_add(&new_xattr->list, &xattrs->head);
913 xattr = NULL;
914 }
915out:
916 spin_unlock(&xattrs->lock);
917 if (xattr) {
918 kfree(xattr->name);
919 kfree(xattr);
920 }
921 return err;
922
923}
924
925static bool xattr_is_trusted(const char *name)
926{
927 return !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN);
928}
929
930static int xattr_list_one(char **buffer, ssize_t *remaining_size,
931 const char *name)
932{
933 size_t len = strlen(name) + 1;
934 if (*buffer) {
935 if (*remaining_size < len)
936 return -ERANGE;
937 memcpy(*buffer, name, len);
938 *buffer += len;
939 }
940 *remaining_size -= len;
941 return 0;
942}
943
944/*
945 * xattr LIST operation for in-memory/pseudo filesystems
946 */
947ssize_t simple_xattr_list(struct inode *inode, struct simple_xattrs *xattrs,
948 char *buffer, size_t size)
949{
950 bool trusted = capable(CAP_SYS_ADMIN);
951 struct simple_xattr *xattr;
952 ssize_t remaining_size = size;
953 int err = 0;
954
955#ifdef CONFIG_FS_POSIX_ACL
956 if (inode->i_acl) {
957 err = xattr_list_one(&buffer, &remaining_size,
958 XATTR_NAME_POSIX_ACL_ACCESS);
959 if (err)
960 return err;
961 }
962 if (inode->i_default_acl) {
963 err = xattr_list_one(&buffer, &remaining_size,
964 XATTR_NAME_POSIX_ACL_DEFAULT);
965 if (err)
966 return err;
967 }
968#endif
969
970 spin_lock(&xattrs->lock);
971 list_for_each_entry(xattr, &xattrs->head, list) {
972 /* skip "trusted." attributes for unprivileged callers */
973 if (!trusted && xattr_is_trusted(xattr->name))
974 continue;
975
976 err = xattr_list_one(&buffer, &remaining_size, xattr->name);
977 if (err)
978 break;
979 }
980 spin_unlock(&xattrs->lock);
981
982 return err ? err : size - remaining_size;
983}
984
985/*
986 * Adds an extended attribute to the list
987 */
988void simple_xattr_list_add(struct simple_xattrs *xattrs,
989 struct simple_xattr *new_xattr)
990{
991 spin_lock(&xattrs->lock);
992 list_add(&new_xattr->list, &xattrs->head);
993 spin_unlock(&xattrs->lock);
994}
1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 File: fs/xattr.c
4
5 Extended attribute handling.
6
7 Copyright (C) 2001 by Andreas Gruenbacher <a.gruenbacher@computer.org>
8 Copyright (C) 2001 SGI - Silicon Graphics, Inc <linux-xfs@oss.sgi.com>
9 Copyright (c) 2004 Red Hat, Inc., James Morris <jmorris@redhat.com>
10 */
11#include <linux/fs.h>
12#include <linux/slab.h>
13#include <linux/file.h>
14#include <linux/xattr.h>
15#include <linux/mount.h>
16#include <linux/namei.h>
17#include <linux/security.h>
18#include <linux/evm.h>
19#include <linux/syscalls.h>
20#include <linux/export.h>
21#include <linux/fsnotify.h>
22#include <linux/audit.h>
23#include <linux/vmalloc.h>
24#include <linux/posix_acl_xattr.h>
25
26#include <linux/uaccess.h>
27
28#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 **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 * @mnt_userns: User namespace 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 user_namespace *mnt_userns, 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(mnt_userns, 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 user_namespace *mnt_userns, struct inode *inode,
114 const char *name, int mask)
115{
116 if (mask & MAY_WRITE) {
117 int ret;
118
119 ret = may_write_xattr(mnt_userns, 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(mnt_userns, inode))
152 return -EPERM;
153 }
154
155 return inode_permission(mnt_userns, inode, mask);
156}
157
158/*
159 * Look for any handler that deals with the specified namespace.
160 */
161int
162xattr_supported_namespace(struct inode *inode, const char *prefix)
163{
164 const struct xattr_handler **handlers = inode->i_sb->s_xattr;
165 const struct xattr_handler *handler;
166 size_t preflen;
167
168 if (!(inode->i_opflags & IOP_XATTR)) {
169 if (unlikely(is_bad_inode(inode)))
170 return -EIO;
171 return -EOPNOTSUPP;
172 }
173
174 preflen = strlen(prefix);
175
176 for_each_xattr_handler(handlers, handler) {
177 if (!strncmp(xattr_prefix(handler), prefix, preflen))
178 return 0;
179 }
180
181 return -EOPNOTSUPP;
182}
183EXPORT_SYMBOL(xattr_supported_namespace);
184
185int
186__vfs_setxattr(struct user_namespace *mnt_userns, struct dentry *dentry,
187 struct inode *inode, const char *name, const void *value,
188 size_t size, int flags)
189{
190 const struct xattr_handler *handler;
191
192 if (is_posix_acl_xattr(name))
193 return -EOPNOTSUPP;
194
195 handler = xattr_resolve_name(inode, &name);
196 if (IS_ERR(handler))
197 return PTR_ERR(handler);
198 if (!handler->set)
199 return -EOPNOTSUPP;
200 if (size == 0)
201 value = ""; /* empty EA, do not remove */
202 return handler->set(handler, mnt_userns, dentry, inode, name, value,
203 size, flags);
204}
205EXPORT_SYMBOL(__vfs_setxattr);
206
207/**
208 * __vfs_setxattr_noperm - perform setxattr operation without performing
209 * permission checks.
210 *
211 * @mnt_userns: user namespace of the mount the inode was found from
212 * @dentry: object to perform setxattr on
213 * @name: xattr name to set
214 * @value: value to set @name to
215 * @size: size of @value
216 * @flags: flags to pass into filesystem operations
217 *
218 * returns the result of the internal setxattr or setsecurity operations.
219 *
220 * This function requires the caller to lock the inode's i_mutex before it
221 * is executed. It also assumes that the caller will make the appropriate
222 * permission checks.
223 */
224int __vfs_setxattr_noperm(struct user_namespace *mnt_userns,
225 struct dentry *dentry, const char *name,
226 const void *value, size_t size, int flags)
227{
228 struct inode *inode = dentry->d_inode;
229 int error = -EAGAIN;
230 int issec = !strncmp(name, XATTR_SECURITY_PREFIX,
231 XATTR_SECURITY_PREFIX_LEN);
232
233 if (issec)
234 inode->i_flags &= ~S_NOSEC;
235 if (inode->i_opflags & IOP_XATTR) {
236 error = __vfs_setxattr(mnt_userns, dentry, inode, name, value,
237 size, flags);
238 if (!error) {
239 fsnotify_xattr(dentry);
240 security_inode_post_setxattr(dentry, name, value,
241 size, flags);
242 }
243 } else {
244 if (unlikely(is_bad_inode(inode)))
245 return -EIO;
246 }
247 if (error == -EAGAIN) {
248 error = -EOPNOTSUPP;
249
250 if (issec) {
251 const char *suffix = name + XATTR_SECURITY_PREFIX_LEN;
252
253 error = security_inode_setsecurity(inode, suffix, value,
254 size, flags);
255 if (!error)
256 fsnotify_xattr(dentry);
257 }
258 }
259
260 return error;
261}
262
263/**
264 * __vfs_setxattr_locked - set an extended attribute while holding the inode
265 * lock
266 *
267 * @mnt_userns: user namespace of the mount of the target inode
268 * @dentry: object to perform setxattr on
269 * @name: xattr name to set
270 * @value: value to set @name to
271 * @size: size of @value
272 * @flags: flags to pass into filesystem operations
273 * @delegated_inode: on return, will contain an inode pointer that
274 * a delegation was broken on, NULL if none.
275 */
276int
277__vfs_setxattr_locked(struct user_namespace *mnt_userns, struct dentry *dentry,
278 const char *name, const void *value, size_t size,
279 int flags, struct inode **delegated_inode)
280{
281 struct inode *inode = dentry->d_inode;
282 int error;
283
284 error = xattr_permission(mnt_userns, inode, name, MAY_WRITE);
285 if (error)
286 return error;
287
288 error = security_inode_setxattr(mnt_userns, dentry, name, value, size,
289 flags);
290 if (error)
291 goto out;
292
293 error = try_break_deleg(inode, delegated_inode);
294 if (error)
295 goto out;
296
297 error = __vfs_setxattr_noperm(mnt_userns, dentry, name, value,
298 size, flags);
299
300out:
301 return error;
302}
303EXPORT_SYMBOL_GPL(__vfs_setxattr_locked);
304
305int
306vfs_setxattr(struct user_namespace *mnt_userns, struct dentry *dentry,
307 const char *name, const void *value, size_t size, int flags)
308{
309 struct inode *inode = dentry->d_inode;
310 struct inode *delegated_inode = NULL;
311 const void *orig_value = value;
312 int error;
313
314 if (size && strcmp(name, XATTR_NAME_CAPS) == 0) {
315 error = cap_convert_nscap(mnt_userns, dentry, &value, size);
316 if (error < 0)
317 return error;
318 size = error;
319 }
320
321retry_deleg:
322 inode_lock(inode);
323 error = __vfs_setxattr_locked(mnt_userns, dentry, name, value, size,
324 flags, &delegated_inode);
325 inode_unlock(inode);
326
327 if (delegated_inode) {
328 error = break_deleg_wait(&delegated_inode);
329 if (!error)
330 goto retry_deleg;
331 }
332 if (value != orig_value)
333 kfree(value);
334
335 return error;
336}
337EXPORT_SYMBOL_GPL(vfs_setxattr);
338
339static ssize_t
340xattr_getsecurity(struct user_namespace *mnt_userns, struct inode *inode,
341 const char *name, void *value, size_t size)
342{
343 void *buffer = NULL;
344 ssize_t len;
345
346 if (!value || !size) {
347 len = security_inode_getsecurity(mnt_userns, inode, name,
348 &buffer, false);
349 goto out_noalloc;
350 }
351
352 len = security_inode_getsecurity(mnt_userns, inode, name, &buffer,
353 true);
354 if (len < 0)
355 return len;
356 if (size < len) {
357 len = -ERANGE;
358 goto out;
359 }
360 memcpy(value, buffer, len);
361out:
362 kfree(buffer);
363out_noalloc:
364 return len;
365}
366
367/*
368 * vfs_getxattr_alloc - allocate memory, if necessary, before calling getxattr
369 *
370 * Allocate memory, if not already allocated, or re-allocate correct size,
371 * before retrieving the extended attribute. The xattr value buffer should
372 * always be freed by the caller, even on error.
373 *
374 * Returns the result of alloc, if failed, or the getxattr operation.
375 */
376int
377vfs_getxattr_alloc(struct user_namespace *mnt_userns, struct dentry *dentry,
378 const char *name, char **xattr_value, size_t xattr_size,
379 gfp_t flags)
380{
381 const struct xattr_handler *handler;
382 struct inode *inode = dentry->d_inode;
383 char *value = *xattr_value;
384 int error;
385
386 error = xattr_permission(mnt_userns, inode, name, MAY_READ);
387 if (error)
388 return error;
389
390 handler = xattr_resolve_name(inode, &name);
391 if (IS_ERR(handler))
392 return PTR_ERR(handler);
393 if (!handler->get)
394 return -EOPNOTSUPP;
395 error = handler->get(handler, dentry, inode, name, NULL, 0);
396 if (error < 0)
397 return error;
398
399 if (!value || (error > xattr_size)) {
400 value = krealloc(*xattr_value, error + 1, flags);
401 if (!value)
402 return -ENOMEM;
403 memset(value, 0, error + 1);
404 }
405
406 error = handler->get(handler, dentry, inode, name, value, error);
407 *xattr_value = value;
408 return error;
409}
410
411ssize_t
412__vfs_getxattr(struct dentry *dentry, struct inode *inode, const char *name,
413 void *value, size_t size)
414{
415 const struct xattr_handler *handler;
416
417 if (is_posix_acl_xattr(name))
418 return -EOPNOTSUPP;
419
420 handler = xattr_resolve_name(inode, &name);
421 if (IS_ERR(handler))
422 return PTR_ERR(handler);
423 if (!handler->get)
424 return -EOPNOTSUPP;
425 return handler->get(handler, dentry, inode, name, value, size);
426}
427EXPORT_SYMBOL(__vfs_getxattr);
428
429ssize_t
430vfs_getxattr(struct user_namespace *mnt_userns, struct dentry *dentry,
431 const char *name, void *value, size_t size)
432{
433 struct inode *inode = dentry->d_inode;
434 int error;
435
436 error = xattr_permission(mnt_userns, inode, name, MAY_READ);
437 if (error)
438 return error;
439
440 error = security_inode_getxattr(dentry, name);
441 if (error)
442 return error;
443
444 if (!strncmp(name, XATTR_SECURITY_PREFIX,
445 XATTR_SECURITY_PREFIX_LEN)) {
446 const char *suffix = name + XATTR_SECURITY_PREFIX_LEN;
447 int ret = xattr_getsecurity(mnt_userns, inode, suffix, value,
448 size);
449 /*
450 * Only overwrite the return value if a security module
451 * is actually active.
452 */
453 if (ret == -EOPNOTSUPP)
454 goto nolsm;
455 return ret;
456 }
457nolsm:
458 return __vfs_getxattr(dentry, inode, name, value, size);
459}
460EXPORT_SYMBOL_GPL(vfs_getxattr);
461
462ssize_t
463vfs_listxattr(struct dentry *dentry, char *list, size_t size)
464{
465 struct inode *inode = d_inode(dentry);
466 ssize_t error;
467
468 error = security_inode_listxattr(dentry);
469 if (error)
470 return error;
471 if (inode->i_op->listxattr && (inode->i_opflags & IOP_XATTR)) {
472 error = inode->i_op->listxattr(dentry, list, size);
473 } else {
474 error = security_inode_listsecurity(inode, list, size);
475 if (size && error > size)
476 error = -ERANGE;
477 }
478 return error;
479}
480EXPORT_SYMBOL_GPL(vfs_listxattr);
481
482int
483__vfs_removexattr(struct user_namespace *mnt_userns, struct dentry *dentry,
484 const char *name)
485{
486 struct inode *inode = d_inode(dentry);
487 const struct xattr_handler *handler;
488
489 if (is_posix_acl_xattr(name))
490 return -EOPNOTSUPP;
491
492 handler = xattr_resolve_name(inode, &name);
493 if (IS_ERR(handler))
494 return PTR_ERR(handler);
495 if (!handler->set)
496 return -EOPNOTSUPP;
497 return handler->set(handler, mnt_userns, dentry, inode, name, NULL, 0,
498 XATTR_REPLACE);
499}
500EXPORT_SYMBOL(__vfs_removexattr);
501
502/**
503 * __vfs_removexattr_locked - set an extended attribute while holding the inode
504 * lock
505 *
506 * @mnt_userns: user namespace of the mount of the target inode
507 * @dentry: object to perform setxattr on
508 * @name: name of xattr to remove
509 * @delegated_inode: on return, will contain an inode pointer that
510 * a delegation was broken on, NULL if none.
511 */
512int
513__vfs_removexattr_locked(struct user_namespace *mnt_userns,
514 struct dentry *dentry, const char *name,
515 struct inode **delegated_inode)
516{
517 struct inode *inode = dentry->d_inode;
518 int error;
519
520 error = xattr_permission(mnt_userns, inode, name, MAY_WRITE);
521 if (error)
522 return error;
523
524 error = security_inode_removexattr(mnt_userns, dentry, name);
525 if (error)
526 goto out;
527
528 error = try_break_deleg(inode, delegated_inode);
529 if (error)
530 goto out;
531
532 error = __vfs_removexattr(mnt_userns, dentry, name);
533
534 if (!error) {
535 fsnotify_xattr(dentry);
536 evm_inode_post_removexattr(dentry, name);
537 }
538
539out:
540 return error;
541}
542EXPORT_SYMBOL_GPL(__vfs_removexattr_locked);
543
544int
545vfs_removexattr(struct user_namespace *mnt_userns, struct dentry *dentry,
546 const char *name)
547{
548 struct inode *inode = dentry->d_inode;
549 struct inode *delegated_inode = NULL;
550 int error;
551
552retry_deleg:
553 inode_lock(inode);
554 error = __vfs_removexattr_locked(mnt_userns, dentry,
555 name, &delegated_inode);
556 inode_unlock(inode);
557
558 if (delegated_inode) {
559 error = break_deleg_wait(&delegated_inode);
560 if (!error)
561 goto retry_deleg;
562 }
563
564 return error;
565}
566EXPORT_SYMBOL_GPL(vfs_removexattr);
567
568/*
569 * Extended attribute SET operations
570 */
571
572int setxattr_copy(const char __user *name, struct xattr_ctx *ctx)
573{
574 int error;
575
576 if (ctx->flags & ~(XATTR_CREATE|XATTR_REPLACE))
577 return -EINVAL;
578
579 error = strncpy_from_user(ctx->kname->name, name,
580 sizeof(ctx->kname->name));
581 if (error == 0 || error == sizeof(ctx->kname->name))
582 return -ERANGE;
583 if (error < 0)
584 return error;
585
586 error = 0;
587 if (ctx->size) {
588 if (ctx->size > XATTR_SIZE_MAX)
589 return -E2BIG;
590
591 ctx->kvalue = vmemdup_user(ctx->cvalue, ctx->size);
592 if (IS_ERR(ctx->kvalue)) {
593 error = PTR_ERR(ctx->kvalue);
594 ctx->kvalue = NULL;
595 }
596 }
597
598 return error;
599}
600
601int do_setxattr(struct mnt_idmap *idmap, struct dentry *dentry,
602 struct xattr_ctx *ctx)
603{
604 if (is_posix_acl_xattr(ctx->kname->name))
605 return do_set_acl(idmap, dentry, ctx->kname->name,
606 ctx->kvalue, ctx->size);
607
608 return vfs_setxattr(mnt_idmap_owner(idmap), dentry, ctx->kname->name,
609 ctx->kvalue, ctx->size, ctx->flags);
610}
611
612static long
613setxattr(struct mnt_idmap *idmap, struct dentry *d,
614 const char __user *name, const void __user *value, size_t size,
615 int flags)
616{
617 struct xattr_name kname;
618 struct xattr_ctx ctx = {
619 .cvalue = value,
620 .kvalue = NULL,
621 .size = size,
622 .kname = &kname,
623 .flags = flags,
624 };
625 int error;
626
627 error = setxattr_copy(name, &ctx);
628 if (error)
629 return error;
630
631 error = do_setxattr(idmap, d, &ctx);
632
633 kvfree(ctx.kvalue);
634 return error;
635}
636
637static int path_setxattr(const char __user *pathname,
638 const char __user *name, const void __user *value,
639 size_t size, int flags, unsigned int lookup_flags)
640{
641 struct path path;
642 int error;
643
644retry:
645 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
646 if (error)
647 return error;
648 error = mnt_want_write(path.mnt);
649 if (!error) {
650 error = setxattr(mnt_idmap(path.mnt), path.dentry, name,
651 value, size, flags);
652 mnt_drop_write(path.mnt);
653 }
654 path_put(&path);
655 if (retry_estale(error, lookup_flags)) {
656 lookup_flags |= LOOKUP_REVAL;
657 goto retry;
658 }
659 return error;
660}
661
662SYSCALL_DEFINE5(setxattr, const char __user *, pathname,
663 const char __user *, name, const void __user *, value,
664 size_t, size, int, flags)
665{
666 return path_setxattr(pathname, name, value, size, flags, LOOKUP_FOLLOW);
667}
668
669SYSCALL_DEFINE5(lsetxattr, const char __user *, pathname,
670 const char __user *, name, const void __user *, value,
671 size_t, size, int, flags)
672{
673 return path_setxattr(pathname, name, value, size, flags, 0);
674}
675
676SYSCALL_DEFINE5(fsetxattr, int, fd, const char __user *, name,
677 const void __user *,value, size_t, size, int, flags)
678{
679 struct fd f = fdget(fd);
680 int error = -EBADF;
681
682 if (!f.file)
683 return error;
684 audit_file(f.file);
685 error = mnt_want_write_file(f.file);
686 if (!error) {
687 error = setxattr(file_mnt_idmap(f.file),
688 f.file->f_path.dentry, name,
689 value, size, flags);
690 mnt_drop_write_file(f.file);
691 }
692 fdput(f);
693 return error;
694}
695
696/*
697 * Extended attribute GET operations
698 */
699ssize_t
700do_getxattr(struct mnt_idmap *idmap, struct dentry *d,
701 struct xattr_ctx *ctx)
702{
703 ssize_t error;
704 char *kname = ctx->kname->name;
705
706 if (ctx->size) {
707 if (ctx->size > XATTR_SIZE_MAX)
708 ctx->size = XATTR_SIZE_MAX;
709 ctx->kvalue = kvzalloc(ctx->size, GFP_KERNEL);
710 if (!ctx->kvalue)
711 return -ENOMEM;
712 }
713
714 if (is_posix_acl_xattr(ctx->kname->name))
715 error = do_get_acl(idmap, d, kname, ctx->kvalue, ctx->size);
716 else
717 error = vfs_getxattr(mnt_idmap_owner(idmap), d, kname,
718 ctx->kvalue, ctx->size);
719 if (error > 0) {
720 if (ctx->size && copy_to_user(ctx->value, ctx->kvalue, error))
721 error = -EFAULT;
722 } else if (error == -ERANGE && ctx->size >= XATTR_SIZE_MAX) {
723 /* The file system tried to returned a value bigger
724 than XATTR_SIZE_MAX bytes. Not possible. */
725 error = -E2BIG;
726 }
727
728 return error;
729}
730
731static ssize_t
732getxattr(struct mnt_idmap *idmap, struct dentry *d,
733 const char __user *name, void __user *value, size_t size)
734{
735 ssize_t error;
736 struct xattr_name kname;
737 struct xattr_ctx ctx = {
738 .value = value,
739 .kvalue = NULL,
740 .size = size,
741 .kname = &kname,
742 .flags = 0,
743 };
744
745 error = strncpy_from_user(kname.name, name, sizeof(kname.name));
746 if (error == 0 || error == sizeof(kname.name))
747 error = -ERANGE;
748 if (error < 0)
749 return error;
750
751 error = do_getxattr(idmap, d, &ctx);
752
753 kvfree(ctx.kvalue);
754 return error;
755}
756
757static ssize_t path_getxattr(const char __user *pathname,
758 const char __user *name, void __user *value,
759 size_t size, unsigned int lookup_flags)
760{
761 struct path path;
762 ssize_t error;
763retry:
764 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
765 if (error)
766 return error;
767 error = getxattr(mnt_idmap(path.mnt), path.dentry, name, value, size);
768 path_put(&path);
769 if (retry_estale(error, lookup_flags)) {
770 lookup_flags |= LOOKUP_REVAL;
771 goto retry;
772 }
773 return error;
774}
775
776SYSCALL_DEFINE4(getxattr, const char __user *, pathname,
777 const char __user *, name, void __user *, value, size_t, size)
778{
779 return path_getxattr(pathname, name, value, size, LOOKUP_FOLLOW);
780}
781
782SYSCALL_DEFINE4(lgetxattr, const char __user *, pathname,
783 const char __user *, name, void __user *, value, size_t, size)
784{
785 return path_getxattr(pathname, name, value, size, 0);
786}
787
788SYSCALL_DEFINE4(fgetxattr, int, fd, const char __user *, name,
789 void __user *, value, size_t, size)
790{
791 struct fd f = fdget(fd);
792 ssize_t error = -EBADF;
793
794 if (!f.file)
795 return error;
796 audit_file(f.file);
797 error = getxattr(file_mnt_idmap(f.file), f.file->f_path.dentry,
798 name, value, size);
799 fdput(f);
800 return error;
801}
802
803/*
804 * Extended attribute LIST operations
805 */
806static ssize_t
807listxattr(struct dentry *d, char __user *list, size_t size)
808{
809 ssize_t error;
810 char *klist = NULL;
811
812 if (size) {
813 if (size > XATTR_LIST_MAX)
814 size = XATTR_LIST_MAX;
815 klist = kvmalloc(size, GFP_KERNEL);
816 if (!klist)
817 return -ENOMEM;
818 }
819
820 error = vfs_listxattr(d, klist, size);
821 if (error > 0) {
822 if (size && copy_to_user(list, klist, error))
823 error = -EFAULT;
824 } else if (error == -ERANGE && size >= XATTR_LIST_MAX) {
825 /* The file system tried to returned a list bigger
826 than XATTR_LIST_MAX bytes. Not possible. */
827 error = -E2BIG;
828 }
829
830 kvfree(klist);
831
832 return error;
833}
834
835static ssize_t path_listxattr(const char __user *pathname, char __user *list,
836 size_t size, unsigned int lookup_flags)
837{
838 struct path path;
839 ssize_t error;
840retry:
841 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
842 if (error)
843 return error;
844 error = listxattr(path.dentry, list, size);
845 path_put(&path);
846 if (retry_estale(error, lookup_flags)) {
847 lookup_flags |= LOOKUP_REVAL;
848 goto retry;
849 }
850 return error;
851}
852
853SYSCALL_DEFINE3(listxattr, const char __user *, pathname, char __user *, list,
854 size_t, size)
855{
856 return path_listxattr(pathname, list, size, LOOKUP_FOLLOW);
857}
858
859SYSCALL_DEFINE3(llistxattr, const char __user *, pathname, char __user *, list,
860 size_t, size)
861{
862 return path_listxattr(pathname, list, size, 0);
863}
864
865SYSCALL_DEFINE3(flistxattr, int, fd, char __user *, list, size_t, size)
866{
867 struct fd f = fdget(fd);
868 ssize_t error = -EBADF;
869
870 if (!f.file)
871 return error;
872 audit_file(f.file);
873 error = listxattr(f.file->f_path.dentry, list, size);
874 fdput(f);
875 return error;
876}
877
878/*
879 * Extended attribute REMOVE operations
880 */
881static long
882removexattr(struct mnt_idmap *idmap, struct dentry *d,
883 const char __user *name)
884{
885 int error;
886 char kname[XATTR_NAME_MAX + 1];
887
888 error = strncpy_from_user(kname, name, sizeof(kname));
889 if (error == 0 || error == sizeof(kname))
890 error = -ERANGE;
891 if (error < 0)
892 return error;
893
894 if (is_posix_acl_xattr(kname))
895 return vfs_remove_acl(mnt_idmap_owner(idmap), d, kname);
896
897 return vfs_removexattr(mnt_idmap_owner(idmap), d, kname);
898}
899
900static int path_removexattr(const char __user *pathname,
901 const char __user *name, unsigned int lookup_flags)
902{
903 struct path path;
904 int error;
905retry:
906 error = user_path_at(AT_FDCWD, pathname, lookup_flags, &path);
907 if (error)
908 return error;
909 error = mnt_want_write(path.mnt);
910 if (!error) {
911 error = removexattr(mnt_idmap(path.mnt), path.dentry, name);
912 mnt_drop_write(path.mnt);
913 }
914 path_put(&path);
915 if (retry_estale(error, lookup_flags)) {
916 lookup_flags |= LOOKUP_REVAL;
917 goto retry;
918 }
919 return error;
920}
921
922SYSCALL_DEFINE2(removexattr, const char __user *, pathname,
923 const char __user *, name)
924{
925 return path_removexattr(pathname, name, LOOKUP_FOLLOW);
926}
927
928SYSCALL_DEFINE2(lremovexattr, const char __user *, pathname,
929 const char __user *, name)
930{
931 return path_removexattr(pathname, name, 0);
932}
933
934SYSCALL_DEFINE2(fremovexattr, int, fd, const char __user *, name)
935{
936 struct fd f = fdget(fd);
937 int error = -EBADF;
938
939 if (!f.file)
940 return error;
941 audit_file(f.file);
942 error = mnt_want_write_file(f.file);
943 if (!error) {
944 error = removexattr(file_mnt_idmap(f.file),
945 f.file->f_path.dentry, name);
946 mnt_drop_write_file(f.file);
947 }
948 fdput(f);
949 return error;
950}
951
952/*
953 * Combine the results of the list() operation from every xattr_handler in the
954 * list.
955 */
956ssize_t
957generic_listxattr(struct dentry *dentry, char *buffer, size_t buffer_size)
958{
959 const struct xattr_handler *handler, **handlers = dentry->d_sb->s_xattr;
960 unsigned int size = 0;
961
962 if (!buffer) {
963 for_each_xattr_handler(handlers, handler) {
964 if (!handler->name ||
965 (handler->list && !handler->list(dentry)))
966 continue;
967 size += strlen(handler->name) + 1;
968 }
969 } else {
970 char *buf = buffer;
971 size_t len;
972
973 for_each_xattr_handler(handlers, handler) {
974 if (!handler->name ||
975 (handler->list && !handler->list(dentry)))
976 continue;
977 len = strlen(handler->name);
978 if (len + 1 > buffer_size)
979 return -ERANGE;
980 memcpy(buf, handler->name, len + 1);
981 buf += len + 1;
982 buffer_size -= len + 1;
983 }
984 size = buf - buffer;
985 }
986 return size;
987}
988EXPORT_SYMBOL(generic_listxattr);
989
990/**
991 * xattr_full_name - Compute full attribute name from suffix
992 *
993 * @handler: handler of the xattr_handler operation
994 * @name: name passed to the xattr_handler operation
995 *
996 * The get and set xattr handler operations are called with the remainder of
997 * the attribute name after skipping the handler's prefix: for example, "foo"
998 * is passed to the get operation of a handler with prefix "user." to get
999 * attribute "user.foo". The full name is still "there" in the name though.
1000 *
1001 * Note: the list xattr handler operation when called from the vfs is passed a
1002 * NULL name; some file systems use this operation internally, with varying
1003 * semantics.
1004 */
1005const char *xattr_full_name(const struct xattr_handler *handler,
1006 const char *name)
1007{
1008 size_t prefix_len = strlen(xattr_prefix(handler));
1009
1010 return name - prefix_len;
1011}
1012EXPORT_SYMBOL(xattr_full_name);
1013
1014/**
1015 * free_simple_xattr - free an xattr object
1016 * @xattr: the xattr object
1017 *
1018 * Free the xattr object. Can handle @xattr being NULL.
1019 */
1020static inline void free_simple_xattr(struct simple_xattr *xattr)
1021{
1022 if (xattr)
1023 kfree(xattr->name);
1024 kvfree(xattr);
1025}
1026
1027/**
1028 * simple_xattr_alloc - allocate new xattr object
1029 * @value: value of the xattr object
1030 * @size: size of @value
1031 *
1032 * Allocate a new xattr object and initialize respective members. The caller is
1033 * responsible for handling the name of the xattr.
1034 *
1035 * Return: On success a new xattr object is returned. On failure NULL is
1036 * returned.
1037 */
1038struct simple_xattr *simple_xattr_alloc(const void *value, size_t size)
1039{
1040 struct simple_xattr *new_xattr;
1041 size_t len;
1042
1043 /* wrap around? */
1044 len = sizeof(*new_xattr) + size;
1045 if (len < sizeof(*new_xattr))
1046 return NULL;
1047
1048 new_xattr = kvmalloc(len, GFP_KERNEL);
1049 if (!new_xattr)
1050 return NULL;
1051
1052 new_xattr->size = size;
1053 memcpy(new_xattr->value, value, size);
1054 return new_xattr;
1055}
1056
1057/**
1058 * rbtree_simple_xattr_cmp - compare xattr name with current rbtree xattr entry
1059 * @key: xattr name
1060 * @node: current node
1061 *
1062 * Compare the xattr name with the xattr name attached to @node in the rbtree.
1063 *
1064 * Return: Negative value if continuing left, positive if continuing right, 0
1065 * if the xattr attached to @node matches @key.
1066 */
1067static int rbtree_simple_xattr_cmp(const void *key, const struct rb_node *node)
1068{
1069 const char *xattr_name = key;
1070 const struct simple_xattr *xattr;
1071
1072 xattr = rb_entry(node, struct simple_xattr, rb_node);
1073 return strcmp(xattr->name, xattr_name);
1074}
1075
1076/**
1077 * rbtree_simple_xattr_node_cmp - compare two xattr rbtree nodes
1078 * @new_node: new node
1079 * @node: current node
1080 *
1081 * Compare the xattr attached to @new_node with the xattr attached to @node.
1082 *
1083 * Return: Negative value if continuing left, positive if continuing right, 0
1084 * if the xattr attached to @new_node matches the xattr attached to @node.
1085 */
1086static int rbtree_simple_xattr_node_cmp(struct rb_node *new_node,
1087 const struct rb_node *node)
1088{
1089 struct simple_xattr *xattr;
1090 xattr = rb_entry(new_node, struct simple_xattr, rb_node);
1091 return rbtree_simple_xattr_cmp(xattr->name, node);
1092}
1093
1094/**
1095 * simple_xattr_get - get an xattr object
1096 * @xattrs: the header of the xattr object
1097 * @name: the name of the xattr to retrieve
1098 * @buffer: the buffer to store the value into
1099 * @size: the size of @buffer
1100 *
1101 * Try to find and retrieve the xattr object associated with @name.
1102 * If @buffer is provided store the value of @xattr in @buffer
1103 * otherwise just return the length. The size of @buffer is limited
1104 * to XATTR_SIZE_MAX which currently is 65536.
1105 *
1106 * Return: On success the length of the xattr value is returned. On error a
1107 * negative error code is returned.
1108 */
1109int simple_xattr_get(struct simple_xattrs *xattrs, const char *name,
1110 void *buffer, size_t size)
1111{
1112 struct simple_xattr *xattr = NULL;
1113 struct rb_node *rbp;
1114 int ret = -ENODATA;
1115
1116 read_lock(&xattrs->lock);
1117 rbp = rb_find(name, &xattrs->rb_root, rbtree_simple_xattr_cmp);
1118 if (rbp) {
1119 xattr = rb_entry(rbp, struct simple_xattr, rb_node);
1120 ret = xattr->size;
1121 if (buffer) {
1122 if (size < xattr->size)
1123 ret = -ERANGE;
1124 else
1125 memcpy(buffer, xattr->value, xattr->size);
1126 }
1127 }
1128 read_unlock(&xattrs->lock);
1129 return ret;
1130}
1131
1132/**
1133 * simple_xattr_set - set an xattr object
1134 * @xattrs: the header of the xattr object
1135 * @name: the name of the xattr to retrieve
1136 * @value: the value to store along the xattr
1137 * @size: the size of @value
1138 * @flags: the flags determining how to set the xattr
1139 * @removed_size: the size of the removed xattr
1140 *
1141 * Set a new xattr object.
1142 * If @value is passed a new xattr object will be allocated. If XATTR_REPLACE
1143 * is specified in @flags a matching xattr object for @name must already exist.
1144 * If it does it will be replaced with the new xattr object. If it doesn't we
1145 * fail. If XATTR_CREATE is specified and a matching xattr does already exist
1146 * we fail. If it doesn't we create a new xattr. If @flags is zero we simply
1147 * insert the new xattr replacing any existing one.
1148 *
1149 * If @value is empty and a matching xattr object is found we delete it if
1150 * XATTR_REPLACE is specified in @flags or @flags is zero.
1151 *
1152 * If @value is empty and no matching xattr object for @name is found we do
1153 * nothing if XATTR_CREATE is specified in @flags or @flags is zero. For
1154 * XATTR_REPLACE we fail as mentioned above.
1155 *
1156 * Return: On success zero and on error a negative error code is returned.
1157 */
1158int simple_xattr_set(struct simple_xattrs *xattrs, const char *name,
1159 const void *value, size_t size, int flags,
1160 ssize_t *removed_size)
1161{
1162 struct simple_xattr *xattr = NULL, *new_xattr = NULL;
1163 struct rb_node *parent = NULL, **rbp;
1164 int err = 0, ret;
1165
1166 if (removed_size)
1167 *removed_size = -1;
1168
1169 /* value == NULL means remove */
1170 if (value) {
1171 new_xattr = simple_xattr_alloc(value, size);
1172 if (!new_xattr)
1173 return -ENOMEM;
1174
1175 new_xattr->name = kstrdup(name, GFP_KERNEL);
1176 if (!new_xattr->name) {
1177 free_simple_xattr(new_xattr);
1178 return -ENOMEM;
1179 }
1180 }
1181
1182 write_lock(&xattrs->lock);
1183 rbp = &xattrs->rb_root.rb_node;
1184 while (*rbp) {
1185 parent = *rbp;
1186 ret = rbtree_simple_xattr_cmp(name, *rbp);
1187 if (ret < 0)
1188 rbp = &(*rbp)->rb_left;
1189 else if (ret > 0)
1190 rbp = &(*rbp)->rb_right;
1191 else
1192 xattr = rb_entry(*rbp, struct simple_xattr, rb_node);
1193 if (xattr)
1194 break;
1195 }
1196
1197 if (xattr) {
1198 /* Fail if XATTR_CREATE is requested and the xattr exists. */
1199 if (flags & XATTR_CREATE) {
1200 err = -EEXIST;
1201 goto out_unlock;
1202 }
1203
1204 if (new_xattr)
1205 rb_replace_node(&xattr->rb_node, &new_xattr->rb_node,
1206 &xattrs->rb_root);
1207 else
1208 rb_erase(&xattr->rb_node, &xattrs->rb_root);
1209 if (!err && removed_size)
1210 *removed_size = xattr->size;
1211 } else {
1212 /* Fail if XATTR_REPLACE is requested but no xattr is found. */
1213 if (flags & XATTR_REPLACE) {
1214 err = -ENODATA;
1215 goto out_unlock;
1216 }
1217
1218 /*
1219 * If XATTR_CREATE or no flags are specified together with a
1220 * new value simply insert it.
1221 */
1222 if (new_xattr) {
1223 rb_link_node(&new_xattr->rb_node, parent, rbp);
1224 rb_insert_color(&new_xattr->rb_node, &xattrs->rb_root);
1225 }
1226
1227 /*
1228 * If XATTR_CREATE or no flags are specified and neither an
1229 * old or new xattr exist then we don't need to do anything.
1230 */
1231 }
1232
1233out_unlock:
1234 write_unlock(&xattrs->lock);
1235 if (err)
1236 free_simple_xattr(new_xattr);
1237 else
1238 free_simple_xattr(xattr);
1239 return err;
1240
1241}
1242
1243static bool xattr_is_trusted(const char *name)
1244{
1245 return !strncmp(name, XATTR_TRUSTED_PREFIX, XATTR_TRUSTED_PREFIX_LEN);
1246}
1247
1248static int xattr_list_one(char **buffer, ssize_t *remaining_size,
1249 const char *name)
1250{
1251 size_t len = strlen(name) + 1;
1252 if (*buffer) {
1253 if (*remaining_size < len)
1254 return -ERANGE;
1255 memcpy(*buffer, name, len);
1256 *buffer += len;
1257 }
1258 *remaining_size -= len;
1259 return 0;
1260}
1261
1262/**
1263 * simple_xattr_list - list all xattr objects
1264 * @inode: inode from which to get the xattrs
1265 * @xattrs: the header of the xattr object
1266 * @buffer: the buffer to store all xattrs into
1267 * @size: the size of @buffer
1268 *
1269 * List all xattrs associated with @inode. If @buffer is NULL we returned
1270 * the required size of the buffer. If @buffer is provided we store the
1271 * xattrs value into it provided it is big enough.
1272 *
1273 * Note, the number of xattr names that can be listed with listxattr(2) is
1274 * limited to XATTR_LIST_MAX aka 65536 bytes. If a larger buffer is passed
1275 * then vfs_listxattr() caps it to XATTR_LIST_MAX and if more xattr names
1276 * are found it will return -E2BIG.
1277 *
1278 * Return: On success the required size or the size of the copied xattrs is
1279 * returned. On error a negative error code is returned.
1280 */
1281ssize_t simple_xattr_list(struct inode *inode, struct simple_xattrs *xattrs,
1282 char *buffer, size_t size)
1283{
1284 bool trusted = ns_capable_noaudit(&init_user_ns, CAP_SYS_ADMIN);
1285 struct simple_xattr *xattr;
1286 struct rb_node *rbp;
1287 ssize_t remaining_size = size;
1288 int err = 0;
1289
1290#ifdef CONFIG_FS_POSIX_ACL
1291 if (IS_POSIXACL(inode)) {
1292 if (inode->i_acl) {
1293 err = xattr_list_one(&buffer, &remaining_size,
1294 XATTR_NAME_POSIX_ACL_ACCESS);
1295 if (err)
1296 return err;
1297 }
1298 if (inode->i_default_acl) {
1299 err = xattr_list_one(&buffer, &remaining_size,
1300 XATTR_NAME_POSIX_ACL_DEFAULT);
1301 if (err)
1302 return err;
1303 }
1304 }
1305#endif
1306
1307 read_lock(&xattrs->lock);
1308 for (rbp = rb_first(&xattrs->rb_root); rbp; rbp = rb_next(rbp)) {
1309 xattr = rb_entry(rbp, struct simple_xattr, rb_node);
1310
1311 /* skip "trusted." attributes for unprivileged callers */
1312 if (!trusted && xattr_is_trusted(xattr->name))
1313 continue;
1314
1315 err = xattr_list_one(&buffer, &remaining_size, xattr->name);
1316 if (err)
1317 break;
1318 }
1319 read_unlock(&xattrs->lock);
1320
1321 return err ? err : size - remaining_size;
1322}
1323
1324/**
1325 * rbtree_simple_xattr_less - compare two xattr rbtree nodes
1326 * @new_node: new node
1327 * @node: current node
1328 *
1329 * Compare the xattr attached to @new_node with the xattr attached to @node.
1330 * Note that this function technically tolerates duplicate entries.
1331 *
1332 * Return: True if insertion point in the rbtree is found.
1333 */
1334static bool rbtree_simple_xattr_less(struct rb_node *new_node,
1335 const struct rb_node *node)
1336{
1337 return rbtree_simple_xattr_node_cmp(new_node, node) < 0;
1338}
1339
1340/**
1341 * simple_xattr_add - add xattr objects
1342 * @xattrs: the header of the xattr object
1343 * @new_xattr: the xattr object to add
1344 *
1345 * Add an xattr object to @xattrs. This assumes no replacement or removal
1346 * of matching xattrs is wanted. Should only be called during inode
1347 * initialization when a few distinct initial xattrs are supposed to be set.
1348 */
1349void simple_xattr_add(struct simple_xattrs *xattrs,
1350 struct simple_xattr *new_xattr)
1351{
1352 write_lock(&xattrs->lock);
1353 rb_add(&new_xattr->rb_node, &xattrs->rb_root, rbtree_simple_xattr_less);
1354 write_unlock(&xattrs->lock);
1355}
1356
1357/**
1358 * simple_xattrs_init - initialize new xattr header
1359 * @xattrs: header to initialize
1360 *
1361 * Initialize relevant fields of a an xattr header.
1362 */
1363void simple_xattrs_init(struct simple_xattrs *xattrs)
1364{
1365 xattrs->rb_root = RB_ROOT;
1366 rwlock_init(&xattrs->lock);
1367}
1368
1369/**
1370 * simple_xattrs_free - free xattrs
1371 * @xattrs: xattr header whose xattrs to destroy
1372 *
1373 * Destroy all xattrs in @xattr. When this is called no one can hold a
1374 * reference to any of the xattrs anymore.
1375 */
1376void simple_xattrs_free(struct simple_xattrs *xattrs)
1377{
1378 struct rb_node *rbp;
1379
1380 rbp = rb_first(&xattrs->rb_root);
1381 while (rbp) {
1382 struct simple_xattr *xattr;
1383 struct rb_node *rbp_next;
1384
1385 rbp_next = rb_next(rbp);
1386 xattr = rb_entry(rbp, struct simple_xattr, rb_node);
1387 rb_erase(&xattr->rb_node, &xattrs->rb_root);
1388 free_simple_xattr(xattr);
1389 rbp = rbp_next;
1390 }
1391}