[AGPGART] Replace kmalloc+memset's with kzalloc's
[linux-2.6.22.y-op.git] / fs / reiserfs / xattr.c
blob87ac9dc8b381e8325ec2e26a1411c2e6d9449cc9
1 /*
2 * linux/fs/reiserfs/xattr.c
4 * Copyright (c) 2002 by Jeff Mahoney, <jeffm@suse.com>
6 */
8 /*
9 * In order to implement EA/ACLs in a clean, backwards compatible manner,
10 * they are implemented as files in a "private" directory.
11 * Each EA is in it's own file, with the directory layout like so (/ is assumed
12 * to be relative to fs root). Inside the /.reiserfs_priv/xattrs directory,
13 * directories named using the capital-hex form of the objectid and
14 * generation number are used. Inside each directory are individual files
15 * named with the name of the extended attribute.
17 * So, for objectid 12648430, we could have:
18 * /.reiserfs_priv/xattrs/C0FFEE.0/system.posix_acl_access
19 * /.reiserfs_priv/xattrs/C0FFEE.0/system.posix_acl_default
20 * /.reiserfs_priv/xattrs/C0FFEE.0/user.Content-Type
21 * .. or similar.
23 * The file contents are the text of the EA. The size is known based on the
24 * stat data describing the file.
26 * In the case of system.posix_acl_access and system.posix_acl_default, since
27 * these are special cases for filesystem ACLs, they are interpreted by the
28 * kernel, in addition, they are negatively and positively cached and attached
29 * to the inode so that unnecessary lookups are avoided.
32 #include <linux/reiserfs_fs.h>
33 #include <linux/dcache.h>
34 #include <linux/namei.h>
35 #include <linux/errno.h>
36 #include <linux/fs.h>
37 #include <linux/file.h>
38 #include <linux/pagemap.h>
39 #include <linux/xattr.h>
40 #include <linux/reiserfs_xattr.h>
41 #include <linux/reiserfs_acl.h>
42 #include <asm/uaccess.h>
43 #include <asm/checksum.h>
44 #include <linux/smp_lock.h>
45 #include <linux/stat.h>
46 #include <asm/semaphore.h>
48 #define FL_READONLY 128
49 #define FL_DIR_SEM_HELD 256
50 #define PRIVROOT_NAME ".reiserfs_priv"
51 #define XAROOT_NAME "xattrs"
53 static struct reiserfs_xattr_handler *find_xattr_handler_prefix(const char
54 *prefix);
56 static struct dentry *create_xa_root(struct super_block *sb)
58 struct dentry *privroot = dget(REISERFS_SB(sb)->priv_root);
59 struct dentry *xaroot;
61 /* This needs to be created at mount-time */
62 if (!privroot)
63 return ERR_PTR(-EOPNOTSUPP);
65 xaroot = lookup_one_len(XAROOT_NAME, privroot, strlen(XAROOT_NAME));
66 if (IS_ERR(xaroot)) {
67 goto out;
68 } else if (!xaroot->d_inode) {
69 int err;
70 down(&privroot->d_inode->i_sem);
71 err =
72 privroot->d_inode->i_op->mkdir(privroot->d_inode, xaroot,
73 0700);
74 up(&privroot->d_inode->i_sem);
76 if (err) {
77 dput(xaroot);
78 dput(privroot);
79 return ERR_PTR(err);
81 REISERFS_SB(sb)->xattr_root = dget(xaroot);
84 out:
85 dput(privroot);
86 return xaroot;
89 /* This will return a dentry, or error, refering to the xa root directory.
90 * If the xa root doesn't exist yet, the dentry will be returned without
91 * an associated inode. This dentry can be used with ->mkdir to create
92 * the xa directory. */
93 static struct dentry *__get_xa_root(struct super_block *s)
95 struct dentry *privroot = dget(REISERFS_SB(s)->priv_root);
96 struct dentry *xaroot = NULL;
98 if (IS_ERR(privroot) || !privroot)
99 return privroot;
101 xaroot = lookup_one_len(XAROOT_NAME, privroot, strlen(XAROOT_NAME));
102 if (IS_ERR(xaroot)) {
103 goto out;
104 } else if (!xaroot->d_inode) {
105 dput(xaroot);
106 xaroot = NULL;
107 goto out;
110 REISERFS_SB(s)->xattr_root = dget(xaroot);
112 out:
113 dput(privroot);
114 return xaroot;
117 /* Returns the dentry (or NULL) referring to the root of the extended
118 * attribute directory tree. If it has already been retreived, it is used.
119 * Otherwise, we attempt to retreive it from disk. It may also return
120 * a pointer-encoded error.
122 static inline struct dentry *get_xa_root(struct super_block *s)
124 struct dentry *dentry = dget(REISERFS_SB(s)->xattr_root);
126 if (!dentry)
127 dentry = __get_xa_root(s);
129 return dentry;
132 /* Opens the directory corresponding to the inode's extended attribute store.
133 * If flags allow, the tree to the directory may be created. If creation is
134 * prohibited, -ENODATA is returned. */
135 static struct dentry *open_xa_dir(const struct inode *inode, int flags)
137 struct dentry *xaroot, *xadir;
138 char namebuf[17];
140 xaroot = get_xa_root(inode->i_sb);
141 if (IS_ERR(xaroot)) {
142 return xaroot;
143 } else if (!xaroot) {
144 if (flags == 0 || flags & XATTR_CREATE) {
145 xaroot = create_xa_root(inode->i_sb);
146 if (IS_ERR(xaroot))
147 return xaroot;
149 if (!xaroot)
150 return ERR_PTR(-ENODATA);
153 /* ok, we have xaroot open */
155 snprintf(namebuf, sizeof(namebuf), "%X.%X",
156 le32_to_cpu(INODE_PKEY(inode)->k_objectid),
157 inode->i_generation);
158 xadir = lookup_one_len(namebuf, xaroot, strlen(namebuf));
159 if (IS_ERR(xadir)) {
160 dput(xaroot);
161 return xadir;
164 if (!xadir->d_inode) {
165 int err;
166 if (flags == 0 || flags & XATTR_CREATE) {
167 /* Although there is nothing else trying to create this directory,
168 * another directory with the same hash may be created, so we need
169 * to protect against that */
170 err =
171 xaroot->d_inode->i_op->mkdir(xaroot->d_inode, xadir,
172 0700);
173 if (err) {
174 dput(xaroot);
175 dput(xadir);
176 return ERR_PTR(err);
179 if (!xadir->d_inode) {
180 dput(xaroot);
181 dput(xadir);
182 return ERR_PTR(-ENODATA);
186 dput(xaroot);
187 return xadir;
190 /* Returns a dentry corresponding to a specific extended attribute file
191 * for the inode. If flags allow, the file is created. Otherwise, a
192 * valid or negative dentry, or an error is returned. */
193 static struct dentry *get_xa_file_dentry(const struct inode *inode,
194 const char *name, int flags)
196 struct dentry *xadir, *xafile;
197 int err = 0;
199 xadir = open_xa_dir(inode, flags);
200 if (IS_ERR(xadir)) {
201 return ERR_PTR(PTR_ERR(xadir));
202 } else if (xadir && !xadir->d_inode) {
203 dput(xadir);
204 return ERR_PTR(-ENODATA);
207 xafile = lookup_one_len(name, xadir, strlen(name));
208 if (IS_ERR(xafile)) {
209 dput(xadir);
210 return ERR_PTR(PTR_ERR(xafile));
213 if (xafile->d_inode) { /* file exists */
214 if (flags & XATTR_CREATE) {
215 err = -EEXIST;
216 dput(xafile);
217 goto out;
219 } else if (flags & XATTR_REPLACE || flags & FL_READONLY) {
220 goto out;
221 } else {
222 /* inode->i_sem is down, so nothing else can try to create
223 * the same xattr */
224 err = xadir->d_inode->i_op->create(xadir->d_inode, xafile,
225 0700 | S_IFREG, NULL);
227 if (err) {
228 dput(xafile);
229 goto out;
233 out:
234 dput(xadir);
235 if (err)
236 xafile = ERR_PTR(err);
237 return xafile;
240 /* Opens a file pointer to the attribute associated with inode */
241 static struct file *open_xa_file(const struct inode *inode, const char *name,
242 int flags)
244 struct dentry *xafile;
245 struct file *fp;
247 xafile = get_xa_file_dentry(inode, name, flags);
248 if (IS_ERR(xafile))
249 return ERR_PTR(PTR_ERR(xafile));
250 else if (!xafile->d_inode) {
251 dput(xafile);
252 return ERR_PTR(-ENODATA);
255 fp = dentry_open(xafile, NULL, O_RDWR);
256 /* dentry_open dputs the dentry if it fails */
258 return fp;
262 * this is very similar to fs/reiserfs/dir.c:reiserfs_readdir, but
263 * we need to drop the path before calling the filldir struct. That
264 * would be a big performance hit to the non-xattr case, so I've copied
265 * the whole thing for now. --clm
267 * the big difference is that I go backwards through the directory,
268 * and don't mess with f->f_pos, but the idea is the same. Do some
269 * action on each and every entry in the directory.
271 * we're called with i_sem held, so there are no worries about the directory
272 * changing underneath us.
274 static int __xattr_readdir(struct file *filp, void *dirent, filldir_t filldir)
276 struct inode *inode = filp->f_dentry->d_inode;
277 struct cpu_key pos_key; /* key of current position in the directory (key of directory entry) */
278 INITIALIZE_PATH(path_to_entry);
279 struct buffer_head *bh;
280 int entry_num;
281 struct item_head *ih, tmp_ih;
282 int search_res;
283 char *local_buf;
284 loff_t next_pos;
285 char small_buf[32]; /* avoid kmalloc if we can */
286 struct reiserfs_de_head *deh;
287 int d_reclen;
288 char *d_name;
289 off_t d_off;
290 ino_t d_ino;
291 struct reiserfs_dir_entry de;
293 /* form key for search the next directory entry using f_pos field of
294 file structure */
295 next_pos = max_reiserfs_offset(inode);
297 while (1) {
298 research:
299 if (next_pos <= DOT_DOT_OFFSET)
300 break;
301 make_cpu_key(&pos_key, inode, next_pos, TYPE_DIRENTRY, 3);
303 search_res =
304 search_by_entry_key(inode->i_sb, &pos_key, &path_to_entry,
305 &de);
306 if (search_res == IO_ERROR) {
307 // FIXME: we could just skip part of directory which could
308 // not be read
309 pathrelse(&path_to_entry);
310 return -EIO;
313 if (search_res == NAME_NOT_FOUND)
314 de.de_entry_num--;
316 set_de_name_and_namelen(&de);
317 entry_num = de.de_entry_num;
318 deh = &(de.de_deh[entry_num]);
320 bh = de.de_bh;
321 ih = de.de_ih;
323 if (!is_direntry_le_ih(ih)) {
324 reiserfs_warning(inode->i_sb, "not direntry %h", ih);
325 break;
327 copy_item_head(&tmp_ih, ih);
329 /* we must have found item, that is item of this directory, */
330 RFALSE(COMP_SHORT_KEYS(&(ih->ih_key), &pos_key),
331 "vs-9000: found item %h does not match to dir we readdir %K",
332 ih, &pos_key);
334 if (deh_offset(deh) <= DOT_DOT_OFFSET) {
335 break;
338 /* look for the previous entry in the directory */
339 next_pos = deh_offset(deh) - 1;
341 if (!de_visible(deh))
342 /* it is hidden entry */
343 continue;
345 d_reclen = entry_length(bh, ih, entry_num);
346 d_name = B_I_DEH_ENTRY_FILE_NAME(bh, ih, deh);
347 d_off = deh_offset(deh);
348 d_ino = deh_objectid(deh);
350 if (!d_name[d_reclen - 1])
351 d_reclen = strlen(d_name);
353 if (d_reclen > REISERFS_MAX_NAME(inode->i_sb->s_blocksize)) {
354 /* too big to send back to VFS */
355 continue;
358 /* Ignore the .reiserfs_priv entry */
359 if (reiserfs_xattrs(inode->i_sb) &&
360 !old_format_only(inode->i_sb) &&
361 deh_objectid(deh) ==
362 le32_to_cpu(INODE_PKEY
363 (REISERFS_SB(inode->i_sb)->priv_root->d_inode)->
364 k_objectid))
365 continue;
367 if (d_reclen <= 32) {
368 local_buf = small_buf;
369 } else {
370 local_buf =
371 reiserfs_kmalloc(d_reclen, GFP_NOFS, inode->i_sb);
372 if (!local_buf) {
373 pathrelse(&path_to_entry);
374 return -ENOMEM;
376 if (item_moved(&tmp_ih, &path_to_entry)) {
377 reiserfs_kfree(local_buf, d_reclen,
378 inode->i_sb);
380 /* sigh, must retry. Do this same offset again */
381 next_pos = d_off;
382 goto research;
386 // Note, that we copy name to user space via temporary
387 // buffer (local_buf) because filldir will block if
388 // user space buffer is swapped out. At that time
389 // entry can move to somewhere else
390 memcpy(local_buf, d_name, d_reclen);
392 /* the filldir function might need to start transactions,
393 * or do who knows what. Release the path now that we've
394 * copied all the important stuff out of the deh
396 pathrelse(&path_to_entry);
398 if (filldir(dirent, local_buf, d_reclen, d_off, d_ino,
399 DT_UNKNOWN) < 0) {
400 if (local_buf != small_buf) {
401 reiserfs_kfree(local_buf, d_reclen,
402 inode->i_sb);
404 goto end;
406 if (local_buf != small_buf) {
407 reiserfs_kfree(local_buf, d_reclen, inode->i_sb);
409 } /* while */
411 end:
412 pathrelse(&path_to_entry);
413 return 0;
417 * this could be done with dedicated readdir ops for the xattr files,
418 * but I want to get something working asap
419 * this is stolen from vfs_readdir
422 static
423 int xattr_readdir(struct file *file, filldir_t filler, void *buf)
425 struct inode *inode = file->f_dentry->d_inode;
426 int res = -ENOTDIR;
427 if (!file->f_op || !file->f_op->readdir)
428 goto out;
429 down(&inode->i_sem);
430 // down(&inode->i_zombie);
431 res = -ENOENT;
432 if (!IS_DEADDIR(inode)) {
433 lock_kernel();
434 res = __xattr_readdir(file, buf, filler);
435 unlock_kernel();
437 // up(&inode->i_zombie);
438 up(&inode->i_sem);
439 out:
440 return res;
443 /* Internal operations on file data */
444 static inline void reiserfs_put_page(struct page *page)
446 kunmap(page);
447 page_cache_release(page);
450 static struct page *reiserfs_get_page(struct inode *dir, unsigned long n)
452 struct address_space *mapping = dir->i_mapping;
453 struct page *page;
454 /* We can deadlock if we try to free dentries,
455 and an unlink/rmdir has just occured - GFP_NOFS avoids this */
456 mapping->flags = (mapping->flags & ~__GFP_BITS_MASK) | GFP_NOFS;
457 page = read_cache_page(mapping, n,
458 (filler_t *) mapping->a_ops->readpage, NULL);
459 if (!IS_ERR(page)) {
460 wait_on_page_locked(page);
461 kmap(page);
462 if (!PageUptodate(page))
463 goto fail;
465 if (PageError(page))
466 goto fail;
468 return page;
470 fail:
471 reiserfs_put_page(page);
472 return ERR_PTR(-EIO);
475 static inline __u32 xattr_hash(const char *msg, int len)
477 return csum_partial(msg, len, 0);
480 /* Generic extended attribute operations that can be used by xa plugins */
483 * inode->i_sem: down
486 reiserfs_xattr_set(struct inode *inode, const char *name, const void *buffer,
487 size_t buffer_size, int flags)
489 int err = 0;
490 struct file *fp;
491 struct page *page;
492 char *data;
493 struct address_space *mapping;
494 size_t file_pos = 0;
495 size_t buffer_pos = 0;
496 struct inode *xinode;
497 struct iattr newattrs;
498 __u32 xahash = 0;
500 if (IS_RDONLY(inode))
501 return -EROFS;
503 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
504 return -EPERM;
506 if (get_inode_sd_version(inode) == STAT_DATA_V1)
507 return -EOPNOTSUPP;
509 /* Empty xattrs are ok, they're just empty files, no hash */
510 if (buffer && buffer_size)
511 xahash = xattr_hash(buffer, buffer_size);
513 open_file:
514 fp = open_xa_file(inode, name, flags);
515 if (IS_ERR(fp)) {
516 err = PTR_ERR(fp);
517 goto out;
520 xinode = fp->f_dentry->d_inode;
521 REISERFS_I(inode)->i_flags |= i_has_xattr_dir;
523 /* we need to copy it off.. */
524 if (xinode->i_nlink > 1) {
525 fput(fp);
526 err = reiserfs_xattr_del(inode, name);
527 if (err < 0)
528 goto out;
529 /* We just killed the old one, we're not replacing anymore */
530 if (flags & XATTR_REPLACE)
531 flags &= ~XATTR_REPLACE;
532 goto open_file;
535 /* Resize it so we're ok to write there */
536 newattrs.ia_size = buffer_size;
537 newattrs.ia_valid = ATTR_SIZE | ATTR_CTIME;
538 down(&xinode->i_sem);
539 err = notify_change(fp->f_dentry, &newattrs);
540 if (err)
541 goto out_filp;
543 mapping = xinode->i_mapping;
544 while (buffer_pos < buffer_size || buffer_pos == 0) {
545 size_t chunk;
546 size_t skip = 0;
547 size_t page_offset = (file_pos & (PAGE_CACHE_SIZE - 1));
548 if (buffer_size - buffer_pos > PAGE_CACHE_SIZE)
549 chunk = PAGE_CACHE_SIZE;
550 else
551 chunk = buffer_size - buffer_pos;
553 page = reiserfs_get_page(xinode, file_pos >> PAGE_CACHE_SHIFT);
554 if (IS_ERR(page)) {
555 err = PTR_ERR(page);
556 goto out_filp;
559 lock_page(page);
560 data = page_address(page);
562 if (file_pos == 0) {
563 struct reiserfs_xattr_header *rxh;
564 skip = file_pos = sizeof(struct reiserfs_xattr_header);
565 if (chunk + skip > PAGE_CACHE_SIZE)
566 chunk = PAGE_CACHE_SIZE - skip;
567 rxh = (struct reiserfs_xattr_header *)data;
568 rxh->h_magic = cpu_to_le32(REISERFS_XATTR_MAGIC);
569 rxh->h_hash = cpu_to_le32(xahash);
572 err = mapping->a_ops->prepare_write(fp, page, page_offset,
573 page_offset + chunk + skip);
574 if (!err) {
575 if (buffer)
576 memcpy(data + skip, buffer + buffer_pos, chunk);
577 err =
578 mapping->a_ops->commit_write(fp, page, page_offset,
579 page_offset + chunk +
580 skip);
582 unlock_page(page);
583 reiserfs_put_page(page);
584 buffer_pos += chunk;
585 file_pos += chunk;
586 skip = 0;
587 if (err || buffer_size == 0 || !buffer)
588 break;
591 /* We can't mark the inode dirty if it's not hashed. This is the case
592 * when we're inheriting the default ACL. If we dirty it, the inode
593 * gets marked dirty, but won't (ever) make it onto the dirty list until
594 * it's synced explicitly to clear I_DIRTY. This is bad. */
595 if (!hlist_unhashed(&inode->i_hash)) {
596 inode->i_ctime = CURRENT_TIME_SEC;
597 mark_inode_dirty(inode);
600 out_filp:
601 up(&xinode->i_sem);
602 fput(fp);
604 out:
605 return err;
609 * inode->i_sem: down
612 reiserfs_xattr_get(const struct inode *inode, const char *name, void *buffer,
613 size_t buffer_size)
615 ssize_t err = 0;
616 struct file *fp;
617 size_t isize;
618 size_t file_pos = 0;
619 size_t buffer_pos = 0;
620 struct page *page;
621 struct inode *xinode;
622 __u32 hash = 0;
624 if (name == NULL)
625 return -EINVAL;
627 /* We can't have xattrs attached to v1 items since they don't have
628 * generation numbers */
629 if (get_inode_sd_version(inode) == STAT_DATA_V1)
630 return -EOPNOTSUPP;
632 fp = open_xa_file(inode, name, FL_READONLY);
633 if (IS_ERR(fp)) {
634 err = PTR_ERR(fp);
635 goto out;
638 xinode = fp->f_dentry->d_inode;
639 isize = xinode->i_size;
640 REISERFS_I(inode)->i_flags |= i_has_xattr_dir;
642 /* Just return the size needed */
643 if (buffer == NULL) {
644 err = isize - sizeof(struct reiserfs_xattr_header);
645 goto out_dput;
648 if (buffer_size < isize - sizeof(struct reiserfs_xattr_header)) {
649 err = -ERANGE;
650 goto out_dput;
653 while (file_pos < isize) {
654 size_t chunk;
655 char *data;
656 size_t skip = 0;
657 if (isize - file_pos > PAGE_CACHE_SIZE)
658 chunk = PAGE_CACHE_SIZE;
659 else
660 chunk = isize - file_pos;
662 page = reiserfs_get_page(xinode, file_pos >> PAGE_CACHE_SHIFT);
663 if (IS_ERR(page)) {
664 err = PTR_ERR(page);
665 goto out_dput;
668 lock_page(page);
669 data = page_address(page);
670 if (file_pos == 0) {
671 struct reiserfs_xattr_header *rxh =
672 (struct reiserfs_xattr_header *)data;
673 skip = file_pos = sizeof(struct reiserfs_xattr_header);
674 chunk -= skip;
675 /* Magic doesn't match up.. */
676 if (rxh->h_magic != cpu_to_le32(REISERFS_XATTR_MAGIC)) {
677 unlock_page(page);
678 reiserfs_put_page(page);
679 reiserfs_warning(inode->i_sb,
680 "Invalid magic for xattr (%s) "
681 "associated with %k", name,
682 INODE_PKEY(inode));
683 err = -EIO;
684 goto out_dput;
686 hash = le32_to_cpu(rxh->h_hash);
688 memcpy(buffer + buffer_pos, data + skip, chunk);
689 unlock_page(page);
690 reiserfs_put_page(page);
691 file_pos += chunk;
692 buffer_pos += chunk;
693 skip = 0;
695 err = isize - sizeof(struct reiserfs_xattr_header);
697 if (xattr_hash(buffer, isize - sizeof(struct reiserfs_xattr_header)) !=
698 hash) {
699 reiserfs_warning(inode->i_sb,
700 "Invalid hash for xattr (%s) associated "
701 "with %k", name, INODE_PKEY(inode));
702 err = -EIO;
705 out_dput:
706 fput(fp);
708 out:
709 return err;
712 static int
713 __reiserfs_xattr_del(struct dentry *xadir, const char *name, int namelen)
715 struct dentry *dentry;
716 struct inode *dir = xadir->d_inode;
717 int err = 0;
719 dentry = lookup_one_len(name, xadir, namelen);
720 if (IS_ERR(dentry)) {
721 err = PTR_ERR(dentry);
722 goto out;
723 } else if (!dentry->d_inode) {
724 err = -ENODATA;
725 goto out_file;
728 /* Skip directories.. */
729 if (S_ISDIR(dentry->d_inode->i_mode))
730 goto out_file;
732 if (!is_reiserfs_priv_object(dentry->d_inode)) {
733 reiserfs_warning(dir->i_sb, "OID %08x [%.*s/%.*s] doesn't have "
734 "priv flag set [parent is %sset].",
735 le32_to_cpu(INODE_PKEY(dentry->d_inode)->
736 k_objectid), xadir->d_name.len,
737 xadir->d_name.name, namelen, name,
738 is_reiserfs_priv_object(xadir->
739 d_inode) ? "" :
740 "not ");
741 dput(dentry);
742 return -EIO;
745 err = dir->i_op->unlink(dir, dentry);
746 if (!err)
747 d_delete(dentry);
749 out_file:
750 dput(dentry);
752 out:
753 return err;
756 int reiserfs_xattr_del(struct inode *inode, const char *name)
758 struct dentry *dir;
759 int err;
761 if (IS_RDONLY(inode))
762 return -EROFS;
764 dir = open_xa_dir(inode, FL_READONLY);
765 if (IS_ERR(dir)) {
766 err = PTR_ERR(dir);
767 goto out;
770 err = __reiserfs_xattr_del(dir, name, strlen(name));
771 dput(dir);
773 if (!err) {
774 inode->i_ctime = CURRENT_TIME_SEC;
775 mark_inode_dirty(inode);
778 out:
779 return err;
782 /* The following are side effects of other operations that aren't explicitly
783 * modifying extended attributes. This includes operations such as permissions
784 * or ownership changes, object deletions, etc. */
786 static int
787 reiserfs_delete_xattrs_filler(void *buf, const char *name, int namelen,
788 loff_t offset, ino_t ino, unsigned int d_type)
790 struct dentry *xadir = (struct dentry *)buf;
792 return __reiserfs_xattr_del(xadir, name, namelen);
796 /* This is called w/ inode->i_sem downed */
797 int reiserfs_delete_xattrs(struct inode *inode)
799 struct file *fp;
800 struct dentry *dir, *root;
801 int err = 0;
803 /* Skip out, an xattr has no xattrs associated with it */
804 if (is_reiserfs_priv_object(inode) ||
805 get_inode_sd_version(inode) == STAT_DATA_V1 ||
806 !reiserfs_xattrs(inode->i_sb)) {
807 return 0;
809 reiserfs_read_lock_xattrs(inode->i_sb);
810 dir = open_xa_dir(inode, FL_READONLY);
811 reiserfs_read_unlock_xattrs(inode->i_sb);
812 if (IS_ERR(dir)) {
813 err = PTR_ERR(dir);
814 goto out;
815 } else if (!dir->d_inode) {
816 dput(dir);
817 return 0;
820 fp = dentry_open(dir, NULL, O_RDWR);
821 if (IS_ERR(fp)) {
822 err = PTR_ERR(fp);
823 /* dentry_open dputs the dentry if it fails */
824 goto out;
827 lock_kernel();
828 err = xattr_readdir(fp, reiserfs_delete_xattrs_filler, dir);
829 if (err) {
830 unlock_kernel();
831 goto out_dir;
834 /* Leftovers besides . and .. -- that's not good. */
835 if (dir->d_inode->i_nlink <= 2) {
836 root = get_xa_root(inode->i_sb);
837 reiserfs_write_lock_xattrs(inode->i_sb);
838 err = vfs_rmdir(root->d_inode, dir);
839 reiserfs_write_unlock_xattrs(inode->i_sb);
840 dput(root);
841 } else {
842 reiserfs_warning(inode->i_sb,
843 "Couldn't remove all entries in directory");
845 unlock_kernel();
847 out_dir:
848 fput(fp);
850 out:
851 if (!err)
852 REISERFS_I(inode)->i_flags =
853 REISERFS_I(inode)->i_flags & ~i_has_xattr_dir;
854 return err;
857 struct reiserfs_chown_buf {
858 struct inode *inode;
859 struct dentry *xadir;
860 struct iattr *attrs;
863 /* XXX: If there is a better way to do this, I'd love to hear about it */
864 static int
865 reiserfs_chown_xattrs_filler(void *buf, const char *name, int namelen,
866 loff_t offset, ino_t ino, unsigned int d_type)
868 struct reiserfs_chown_buf *chown_buf = (struct reiserfs_chown_buf *)buf;
869 struct dentry *xafile, *xadir = chown_buf->xadir;
870 struct iattr *attrs = chown_buf->attrs;
871 int err = 0;
873 xafile = lookup_one_len(name, xadir, namelen);
874 if (IS_ERR(xafile))
875 return PTR_ERR(xafile);
876 else if (!xafile->d_inode) {
877 dput(xafile);
878 return -ENODATA;
881 if (!S_ISDIR(xafile->d_inode->i_mode))
882 err = notify_change(xafile, attrs);
883 dput(xafile);
885 return err;
888 int reiserfs_chown_xattrs(struct inode *inode, struct iattr *attrs)
890 struct file *fp;
891 struct dentry *dir;
892 int err = 0;
893 struct reiserfs_chown_buf buf;
894 unsigned int ia_valid = attrs->ia_valid;
896 /* Skip out, an xattr has no xattrs associated with it */
897 if (is_reiserfs_priv_object(inode) ||
898 get_inode_sd_version(inode) == STAT_DATA_V1 ||
899 !reiserfs_xattrs(inode->i_sb)) {
900 return 0;
902 reiserfs_read_lock_xattrs(inode->i_sb);
903 dir = open_xa_dir(inode, FL_READONLY);
904 reiserfs_read_unlock_xattrs(inode->i_sb);
905 if (IS_ERR(dir)) {
906 if (PTR_ERR(dir) != -ENODATA)
907 err = PTR_ERR(dir);
908 goto out;
909 } else if (!dir->d_inode) {
910 dput(dir);
911 goto out;
914 fp = dentry_open(dir, NULL, O_RDWR);
915 if (IS_ERR(fp)) {
916 err = PTR_ERR(fp);
917 /* dentry_open dputs the dentry if it fails */
918 goto out;
921 lock_kernel();
923 attrs->ia_valid &= (ATTR_UID | ATTR_GID | ATTR_CTIME);
924 buf.xadir = dir;
925 buf.attrs = attrs;
926 buf.inode = inode;
928 err = xattr_readdir(fp, reiserfs_chown_xattrs_filler, &buf);
929 if (err) {
930 unlock_kernel();
931 goto out_dir;
934 err = notify_change(dir, attrs);
935 unlock_kernel();
937 out_dir:
938 fput(fp);
940 out:
941 attrs->ia_valid = ia_valid;
942 return err;
945 /* Actual operations that are exported to VFS-land */
948 * Inode operation getxattr()
949 * Preliminary locking: we down dentry->d_inode->i_sem
951 ssize_t
952 reiserfs_getxattr(struct dentry * dentry, const char *name, void *buffer,
953 size_t size)
955 struct reiserfs_xattr_handler *xah = find_xattr_handler_prefix(name);
956 int err;
958 if (!xah || !reiserfs_xattrs(dentry->d_sb) ||
959 get_inode_sd_version(dentry->d_inode) == STAT_DATA_V1)
960 return -EOPNOTSUPP;
962 reiserfs_read_lock_xattr_i(dentry->d_inode);
963 reiserfs_read_lock_xattrs(dentry->d_sb);
964 err = xah->get(dentry->d_inode, name, buffer, size);
965 reiserfs_read_unlock_xattrs(dentry->d_sb);
966 reiserfs_read_unlock_xattr_i(dentry->d_inode);
967 return err;
971 * Inode operation setxattr()
973 * dentry->d_inode->i_sem down
976 reiserfs_setxattr(struct dentry *dentry, const char *name, const void *value,
977 size_t size, int flags)
979 struct reiserfs_xattr_handler *xah = find_xattr_handler_prefix(name);
980 int err;
981 int lock;
983 if (!xah || !reiserfs_xattrs(dentry->d_sb) ||
984 get_inode_sd_version(dentry->d_inode) == STAT_DATA_V1)
985 return -EOPNOTSUPP;
987 if (IS_RDONLY(dentry->d_inode))
988 return -EROFS;
990 if (IS_IMMUTABLE(dentry->d_inode) || IS_APPEND(dentry->d_inode))
991 return -EROFS;
993 reiserfs_write_lock_xattr_i(dentry->d_inode);
994 lock = !has_xattr_dir(dentry->d_inode);
995 if (lock)
996 reiserfs_write_lock_xattrs(dentry->d_sb);
997 else
998 reiserfs_read_lock_xattrs(dentry->d_sb);
999 err = xah->set(dentry->d_inode, name, value, size, flags);
1000 if (lock)
1001 reiserfs_write_unlock_xattrs(dentry->d_sb);
1002 else
1003 reiserfs_read_unlock_xattrs(dentry->d_sb);
1004 reiserfs_write_unlock_xattr_i(dentry->d_inode);
1005 return err;
1009 * Inode operation removexattr()
1011 * dentry->d_inode->i_sem down
1013 int reiserfs_removexattr(struct dentry *dentry, const char *name)
1015 int err;
1016 struct reiserfs_xattr_handler *xah = find_xattr_handler_prefix(name);
1018 if (!xah || !reiserfs_xattrs(dentry->d_sb) ||
1019 get_inode_sd_version(dentry->d_inode) == STAT_DATA_V1)
1020 return -EOPNOTSUPP;
1022 if (IS_RDONLY(dentry->d_inode))
1023 return -EROFS;
1025 if (IS_IMMUTABLE(dentry->d_inode) || IS_APPEND(dentry->d_inode))
1026 return -EPERM;
1028 reiserfs_write_lock_xattr_i(dentry->d_inode);
1029 reiserfs_read_lock_xattrs(dentry->d_sb);
1031 /* Deletion pre-operation */
1032 if (xah->del) {
1033 err = xah->del(dentry->d_inode, name);
1034 if (err)
1035 goto out;
1038 err = reiserfs_xattr_del(dentry->d_inode, name);
1040 dentry->d_inode->i_ctime = CURRENT_TIME_SEC;
1041 mark_inode_dirty(dentry->d_inode);
1043 out:
1044 reiserfs_read_unlock_xattrs(dentry->d_sb);
1045 reiserfs_write_unlock_xattr_i(dentry->d_inode);
1046 return err;
1049 /* This is what filldir will use:
1050 * r_pos will always contain the amount of space required for the entire
1051 * list. If r_pos becomes larger than r_size, we need more space and we
1052 * return an error indicating this. If r_pos is less than r_size, then we've
1053 * filled the buffer successfully and we return success */
1054 struct reiserfs_listxattr_buf {
1055 int r_pos;
1056 int r_size;
1057 char *r_buf;
1058 struct inode *r_inode;
1061 static int
1062 reiserfs_listxattr_filler(void *buf, const char *name, int namelen,
1063 loff_t offset, ino_t ino, unsigned int d_type)
1065 struct reiserfs_listxattr_buf *b = (struct reiserfs_listxattr_buf *)buf;
1066 int len = 0;
1067 if (name[0] != '.'
1068 || (namelen != 1 && (name[1] != '.' || namelen != 2))) {
1069 struct reiserfs_xattr_handler *xah =
1070 find_xattr_handler_prefix(name);
1071 if (!xah)
1072 return 0; /* Unsupported xattr name, skip it */
1074 /* We call ->list() twice because the operation isn't required to just
1075 * return the name back - we want to make sure we have enough space */
1076 len += xah->list(b->r_inode, name, namelen, NULL);
1078 if (len) {
1079 if (b->r_pos + len + 1 <= b->r_size) {
1080 char *p = b->r_buf + b->r_pos;
1081 p += xah->list(b->r_inode, name, namelen, p);
1082 *p++ = '\0';
1084 b->r_pos += len + 1;
1088 return 0;
1092 * Inode operation listxattr()
1094 * Preliminary locking: we down dentry->d_inode->i_sem
1096 ssize_t reiserfs_listxattr(struct dentry * dentry, char *buffer, size_t size)
1098 struct file *fp;
1099 struct dentry *dir;
1100 int err = 0;
1101 struct reiserfs_listxattr_buf buf;
1103 if (!dentry->d_inode)
1104 return -EINVAL;
1106 if (!reiserfs_xattrs(dentry->d_sb) ||
1107 get_inode_sd_version(dentry->d_inode) == STAT_DATA_V1)
1108 return -EOPNOTSUPP;
1110 reiserfs_read_lock_xattr_i(dentry->d_inode);
1111 reiserfs_read_lock_xattrs(dentry->d_sb);
1112 dir = open_xa_dir(dentry->d_inode, FL_READONLY);
1113 reiserfs_read_unlock_xattrs(dentry->d_sb);
1114 if (IS_ERR(dir)) {
1115 err = PTR_ERR(dir);
1116 if (err == -ENODATA)
1117 err = 0; /* Not an error if there aren't any xattrs */
1118 goto out;
1121 fp = dentry_open(dir, NULL, O_RDWR);
1122 if (IS_ERR(fp)) {
1123 err = PTR_ERR(fp);
1124 /* dentry_open dputs the dentry if it fails */
1125 goto out;
1128 buf.r_buf = buffer;
1129 buf.r_size = buffer ? size : 0;
1130 buf.r_pos = 0;
1131 buf.r_inode = dentry->d_inode;
1133 REISERFS_I(dentry->d_inode)->i_flags |= i_has_xattr_dir;
1135 err = xattr_readdir(fp, reiserfs_listxattr_filler, &buf);
1136 if (err)
1137 goto out_dir;
1139 if (buf.r_pos > buf.r_size && buffer != NULL)
1140 err = -ERANGE;
1141 else
1142 err = buf.r_pos;
1144 out_dir:
1145 fput(fp);
1147 out:
1148 reiserfs_read_unlock_xattr_i(dentry->d_inode);
1149 return err;
1152 /* This is the implementation for the xattr plugin infrastructure */
1153 static struct list_head xattr_handlers = LIST_HEAD_INIT(xattr_handlers);
1154 static DEFINE_RWLOCK(handler_lock);
1156 static struct reiserfs_xattr_handler *find_xattr_handler_prefix(const char
1157 *prefix)
1159 struct reiserfs_xattr_handler *xah = NULL;
1160 struct list_head *p;
1162 read_lock(&handler_lock);
1163 list_for_each(p, &xattr_handlers) {
1164 xah = list_entry(p, struct reiserfs_xattr_handler, handlers);
1165 if (strncmp(xah->prefix, prefix, strlen(xah->prefix)) == 0)
1166 break;
1167 xah = NULL;
1170 read_unlock(&handler_lock);
1171 return xah;
1174 static void __unregister_handlers(void)
1176 struct reiserfs_xattr_handler *xah;
1177 struct list_head *p, *tmp;
1179 list_for_each_safe(p, tmp, &xattr_handlers) {
1180 xah = list_entry(p, struct reiserfs_xattr_handler, handlers);
1181 if (xah->exit)
1182 xah->exit();
1184 list_del_init(p);
1186 INIT_LIST_HEAD(&xattr_handlers);
1189 int __init reiserfs_xattr_register_handlers(void)
1191 int err = 0;
1192 struct reiserfs_xattr_handler *xah;
1193 struct list_head *p;
1195 write_lock(&handler_lock);
1197 /* If we're already initialized, nothing to do */
1198 if (!list_empty(&xattr_handlers)) {
1199 write_unlock(&handler_lock);
1200 return 0;
1203 /* Add the handlers */
1204 list_add_tail(&user_handler.handlers, &xattr_handlers);
1205 list_add_tail(&trusted_handler.handlers, &xattr_handlers);
1206 #ifdef CONFIG_REISERFS_FS_SECURITY
1207 list_add_tail(&security_handler.handlers, &xattr_handlers);
1208 #endif
1209 #ifdef CONFIG_REISERFS_FS_POSIX_ACL
1210 list_add_tail(&posix_acl_access_handler.handlers, &xattr_handlers);
1211 list_add_tail(&posix_acl_default_handler.handlers, &xattr_handlers);
1212 #endif
1214 /* Run initializers, if available */
1215 list_for_each(p, &xattr_handlers) {
1216 xah = list_entry(p, struct reiserfs_xattr_handler, handlers);
1217 if (xah->init) {
1218 err = xah->init();
1219 if (err) {
1220 list_del_init(p);
1221 break;
1226 /* Clean up other handlers, if any failed */
1227 if (err)
1228 __unregister_handlers();
1230 write_unlock(&handler_lock);
1231 return err;
1234 void reiserfs_xattr_unregister_handlers(void)
1236 write_lock(&handler_lock);
1237 __unregister_handlers();
1238 write_unlock(&handler_lock);
1241 /* This will catch lookups from the fs root to .reiserfs_priv */
1242 static int
1243 xattr_lookup_poison(struct dentry *dentry, struct qstr *q1, struct qstr *name)
1245 struct dentry *priv_root = REISERFS_SB(dentry->d_sb)->priv_root;
1246 if (name->len == priv_root->d_name.len &&
1247 name->hash == priv_root->d_name.hash &&
1248 !memcmp(name->name, priv_root->d_name.name, name->len)) {
1249 return -ENOENT;
1250 } else if (q1->len == name->len &&
1251 !memcmp(q1->name, name->name, name->len))
1252 return 0;
1253 return 1;
1256 static struct dentry_operations xattr_lookup_poison_ops = {
1257 .d_compare = xattr_lookup_poison,
1260 /* We need to take a copy of the mount flags since things like
1261 * MS_RDONLY don't get set until *after* we're called.
1262 * mount_flags != mount_options */
1263 int reiserfs_xattr_init(struct super_block *s, int mount_flags)
1265 int err = 0;
1267 /* We need generation numbers to ensure that the oid mapping is correct
1268 * v3.5 filesystems don't have them. */
1269 if (!old_format_only(s)) {
1270 set_bit(REISERFS_XATTRS, &(REISERFS_SB(s)->s_mount_opt));
1271 } else if (reiserfs_xattrs_optional(s)) {
1272 /* Old format filesystem, but optional xattrs have been enabled
1273 * at mount time. Error out. */
1274 reiserfs_warning(s, "xattrs/ACLs not supported on pre v3.6 "
1275 "format filesystem. Failing mount.");
1276 err = -EOPNOTSUPP;
1277 goto error;
1278 } else {
1279 /* Old format filesystem, but no optional xattrs have been enabled. This
1280 * means we silently disable xattrs on the filesystem. */
1281 clear_bit(REISERFS_XATTRS, &(REISERFS_SB(s)->s_mount_opt));
1284 /* If we don't have the privroot located yet - go find it */
1285 if (reiserfs_xattrs(s) && !REISERFS_SB(s)->priv_root) {
1286 struct dentry *dentry;
1287 dentry = lookup_one_len(PRIVROOT_NAME, s->s_root,
1288 strlen(PRIVROOT_NAME));
1289 if (!IS_ERR(dentry)) {
1290 if (!(mount_flags & MS_RDONLY) && !dentry->d_inode) {
1291 struct inode *inode = dentry->d_parent->d_inode;
1292 down(&inode->i_sem);
1293 err = inode->i_op->mkdir(inode, dentry, 0700);
1294 up(&inode->i_sem);
1295 if (err) {
1296 dput(dentry);
1297 dentry = NULL;
1300 if (dentry && dentry->d_inode)
1301 reiserfs_warning(s,
1302 "Created %s on %s - reserved for "
1303 "xattr storage.",
1304 PRIVROOT_NAME,
1305 reiserfs_bdevname
1306 (inode->i_sb));
1307 } else if (!dentry->d_inode) {
1308 dput(dentry);
1309 dentry = NULL;
1311 } else
1312 err = PTR_ERR(dentry);
1314 if (!err && dentry) {
1315 s->s_root->d_op = &xattr_lookup_poison_ops;
1316 reiserfs_mark_inode_private(dentry->d_inode);
1317 REISERFS_SB(s)->priv_root = dentry;
1318 } else if (!(mount_flags & MS_RDONLY)) { /* xattrs are unavailable */
1319 /* If we're read-only it just means that the dir hasn't been
1320 * created. Not an error -- just no xattrs on the fs. We'll
1321 * check again if we go read-write */
1322 reiserfs_warning(s, "xattrs/ACLs enabled and couldn't "
1323 "find/create .reiserfs_priv. Failing mount.");
1324 err = -EOPNOTSUPP;
1328 error:
1329 /* This is only nonzero if there was an error initializing the xattr
1330 * directory or if there is a condition where we don't support them. */
1331 if (err) {
1332 clear_bit(REISERFS_XATTRS, &(REISERFS_SB(s)->s_mount_opt));
1333 clear_bit(REISERFS_XATTRS_USER, &(REISERFS_SB(s)->s_mount_opt));
1334 clear_bit(REISERFS_POSIXACL, &(REISERFS_SB(s)->s_mount_opt));
1337 /* The super_block MS_POSIXACL must mirror the (no)acl mount option. */
1338 s->s_flags = s->s_flags & ~MS_POSIXACL;
1339 if (reiserfs_posixacl(s))
1340 s->s_flags |= MS_POSIXACL;
1342 return err;
1345 static int
1346 __reiserfs_permission(struct inode *inode, int mask, struct nameidata *nd,
1347 int need_lock)
1349 umode_t mode = inode->i_mode;
1351 if (mask & MAY_WRITE) {
1353 * Nobody gets write access to a read-only fs.
1355 if (IS_RDONLY(inode) &&
1356 (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode)))
1357 return -EROFS;
1360 * Nobody gets write access to an immutable file.
1362 if (IS_IMMUTABLE(inode))
1363 return -EACCES;
1366 /* We don't do permission checks on the internal objects.
1367 * Permissions are determined by the "owning" object. */
1368 if (is_reiserfs_priv_object(inode))
1369 return 0;
1371 if (current->fsuid == inode->i_uid) {
1372 mode >>= 6;
1373 #ifdef CONFIG_REISERFS_FS_POSIX_ACL
1374 } else if (reiserfs_posixacl(inode->i_sb) &&
1375 get_inode_sd_version(inode) != STAT_DATA_V1) {
1376 struct posix_acl *acl;
1378 /* ACL can't contain additional permissions if
1379 the ACL_MASK entry is 0 */
1380 if (!(mode & S_IRWXG))
1381 goto check_groups;
1383 if (need_lock) {
1384 reiserfs_read_lock_xattr_i(inode);
1385 reiserfs_read_lock_xattrs(inode->i_sb);
1387 acl = reiserfs_get_acl(inode, ACL_TYPE_ACCESS);
1388 if (need_lock) {
1389 reiserfs_read_unlock_xattrs(inode->i_sb);
1390 reiserfs_read_unlock_xattr_i(inode);
1392 if (IS_ERR(acl)) {
1393 if (PTR_ERR(acl) == -ENODATA)
1394 goto check_groups;
1395 return PTR_ERR(acl);
1398 if (acl) {
1399 int err = posix_acl_permission(inode, acl, mask);
1400 posix_acl_release(acl);
1401 if (err == -EACCES) {
1402 goto check_capabilities;
1404 return err;
1405 } else {
1406 goto check_groups;
1408 #endif
1409 } else {
1410 check_groups:
1411 if (in_group_p(inode->i_gid))
1412 mode >>= 3;
1416 * If the DACs are ok we don't need any capability check.
1418 if (((mode & mask & (MAY_READ | MAY_WRITE | MAY_EXEC)) == mask))
1419 return 0;
1421 check_capabilities:
1423 * Read/write DACs are always overridable.
1424 * Executable DACs are overridable if at least one exec bit is set.
1426 if (!(mask & MAY_EXEC) ||
1427 (inode->i_mode & S_IXUGO) || S_ISDIR(inode->i_mode))
1428 if (capable(CAP_DAC_OVERRIDE))
1429 return 0;
1432 * Searching includes executable on directories, else just read.
1434 if (mask == MAY_READ || (S_ISDIR(inode->i_mode) && !(mask & MAY_WRITE)))
1435 if (capable(CAP_DAC_READ_SEARCH))
1436 return 0;
1438 return -EACCES;
1441 int reiserfs_permission(struct inode *inode, int mask, struct nameidata *nd)
1443 return __reiserfs_permission(inode, mask, nd, 1);
1447 reiserfs_permission_locked(struct inode *inode, int mask, struct nameidata *nd)
1449 return __reiserfs_permission(inode, mask, nd, 0);