2 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
23 #include "xfs_trans.h"
27 #include "xfs_alloc.h"
28 #include "xfs_dmapi.h"
29 #include "xfs_quota.h"
30 #include "xfs_mount.h"
31 #include "xfs_bmap_btree.h"
32 #include "xfs_alloc_btree.h"
33 #include "xfs_ialloc_btree.h"
34 #include "xfs_dir2_sf.h"
35 #include "xfs_attr_sf.h"
36 #include "xfs_dinode.h"
37 #include "xfs_inode.h"
39 #include "xfs_btree.h"
40 #include "xfs_ialloc.h"
41 #include "xfs_rtalloc.h"
42 #include "xfs_error.h"
43 #include "xfs_itable.h"
47 #include "xfs_buf_item.h"
48 #include "xfs_utils.h"
49 #include "xfs_vnodeops.h"
51 #include <linux/capability.h>
52 #include <linux/xattr.h>
53 #include <linux/namei.h>
54 #include <linux/security.h>
55 #include <linux/falloc.h>
58 * Bring the atime in the XFS inode uptodate.
59 * Used before logging the inode to disk or when the Linux inode goes away.
62 xfs_synchronize_atime(
65 struct inode *inode = ip->i_vnode;
68 ip->i_d.di_atime.t_sec = (__int32_t)inode->i_atime.tv_sec;
69 ip->i_d.di_atime.t_nsec = (__int32_t)inode->i_atime.tv_nsec;
74 * If the linux inode exists, mark it dirty.
75 * Used when commiting a dirty inode into a transaction so that
76 * the inode will get written back by the linux code
79 xfs_mark_inode_dirty_sync(
82 struct inode *inode = ip->i_vnode;
85 mark_inode_dirty_sync(inode);
89 * Change the requested timestamp in the given inode.
90 * We don't lock across timestamp updates, and we don't log them but
91 * we do record the fact that there is dirty information in core.
93 * NOTE -- callers MUST combine XFS_ICHGTIME_MOD or XFS_ICHGTIME_CHG
94 * with XFS_ICHGTIME_ACC to be sure that access time
95 * update will take. Calling first with XFS_ICHGTIME_ACC
96 * and then XFS_ICHGTIME_MOD may fail to modify the access
97 * timestamp if the filesystem is mounted noacctm.
104 struct inode *inode = vn_to_inode(XFS_ITOV(ip));
108 if (flags & XFS_ICHGTIME_MOD) {
110 ip->i_d.di_mtime.t_sec = (__int32_t)tv.tv_sec;
111 ip->i_d.di_mtime.t_nsec = (__int32_t)tv.tv_nsec;
113 if (flags & XFS_ICHGTIME_ACC) {
115 ip->i_d.di_atime.t_sec = (__int32_t)tv.tv_sec;
116 ip->i_d.di_atime.t_nsec = (__int32_t)tv.tv_nsec;
118 if (flags & XFS_ICHGTIME_CHG) {
120 ip->i_d.di_ctime.t_sec = (__int32_t)tv.tv_sec;
121 ip->i_d.di_ctime.t_nsec = (__int32_t)tv.tv_nsec;
125 * We update the i_update_core field _after_ changing
126 * the timestamps in order to coordinate properly with
127 * xfs_iflush() so that we don't lose timestamp updates.
128 * This keeps us from having to hold the inode lock
129 * while doing this. We use the SYNCHRONIZE macro to
130 * ensure that the compiler does not reorder the update
131 * of i_update_core above the timestamp updates above.
134 ip->i_update_core = 1;
135 if (!(inode->i_state & I_NEW))
136 mark_inode_dirty_sync(inode);
140 * Variant on the above which avoids querying the system clock
141 * in situations where we know the Linux inode timestamps have
142 * just been updated (and so we can update our inode cheaply).
153 * Atime updates for read() & friends are handled lazily now, and
154 * explicit updates must go through xfs_ichgtime()
156 ASSERT((flags & XFS_ICHGTIME_ACC) == 0);
158 if (flags & XFS_ICHGTIME_MOD) {
159 tvp = &inode->i_mtime;
160 ip->i_d.di_mtime.t_sec = (__int32_t)tvp->tv_sec;
161 ip->i_d.di_mtime.t_nsec = (__int32_t)tvp->tv_nsec;
163 if (flags & XFS_ICHGTIME_CHG) {
164 tvp = &inode->i_ctime;
165 ip->i_d.di_ctime.t_sec = (__int32_t)tvp->tv_sec;
166 ip->i_d.di_ctime.t_nsec = (__int32_t)tvp->tv_nsec;
170 * We update the i_update_core field _after_ changing
171 * the timestamps in order to coordinate properly with
172 * xfs_iflush() so that we don't lose timestamp updates.
173 * This keeps us from having to hold the inode lock
174 * while doing this. We use the SYNCHRONIZE macro to
175 * ensure that the compiler does not reorder the update
176 * of i_update_core above the timestamp updates above.
179 ip->i_update_core = 1;
180 if (!(inode->i_state & I_NEW))
181 mark_inode_dirty_sync(inode);
186 * Pull the link count and size up from the xfs inode to the linux inode
192 struct xfs_inode *ip = XFS_I(inode);
195 /* we're under i_sem so i_size can't change under us */
196 size = XFS_ISIZE(ip);
197 if (i_size_read(inode) != size)
198 i_size_write(inode, size);
202 * Hook in SELinux. This is not quite correct yet, what we really need
203 * here (as we do for default ACLs) is a mechanism by which creation of
204 * these attrs can be journalled at inode creation time (along with the
205 * inode, of course, such that log replay can't cause these to be lost).
212 struct xfs_inode *ip = XFS_I(inode);
218 error = security_inode_init_security(inode, dir, &name,
221 if (error == -EOPNOTSUPP)
226 error = xfs_attr_set(ip, name, value, length, ATTR_SECURE);
228 xfs_iflags_set(ip, XFS_IMODIFIED);
237 struct xfs_name *namep,
238 struct dentry *dentry)
240 namep->name = dentry->d_name.name;
241 namep->len = dentry->d_name.len;
248 struct dentry *dentry)
250 struct xfs_name teardown;
253 * If we can't add the ACL or we fail in
254 * xfs_init_security we must back out.
255 * ENOSPC can hit here, among other things.
257 xfs_dentry_to_name(&teardown, dentry);
259 xfs_remove(XFS_I(dir), &teardown, XFS_I(inode));
266 struct dentry *dentry,
271 struct xfs_inode *ip = NULL;
272 xfs_acl_t *default_acl = NULL;
273 struct xfs_name name;
274 int (*test_default_acl)(struct inode *) = _ACL_DEFAULT_EXISTS;
278 * Irix uses Missed'em'V split, but doesn't want to see
279 * the upper 5 bits of (14bit) major.
281 if (unlikely(!sysv_valid_dev(rdev) || MAJOR(rdev) & ~0x1ff))
284 if (test_default_acl && test_default_acl(dir)) {
285 if (!_ACL_ALLOC(default_acl)) {
288 if (!_ACL_GET_DEFAULT(dir, default_acl)) {
289 _ACL_FREE(default_acl);
294 xfs_dentry_to_name(&name, dentry);
296 if (IS_POSIXACL(dir) && !default_acl)
297 mode &= ~current->fs->umask;
299 switch (mode & S_IFMT) {
304 rdev = sysv_encode_dev(rdev);
306 error = xfs_create(XFS_I(dir), &name, mode, rdev, &ip, NULL);
309 error = xfs_mkdir(XFS_I(dir), &name, mode, &ip, NULL);
321 error = xfs_init_security(inode, dir);
323 goto out_cleanup_inode;
326 error = _ACL_INHERIT(inode, mode, default_acl);
328 goto out_cleanup_inode;
329 xfs_iflags_set(ip, XFS_IMODIFIED);
330 _ACL_FREE(default_acl);
335 xfs_validate_fields(inode);
336 d_instantiate(dentry, inode);
337 xfs_validate_fields(dir);
341 xfs_cleanup_inode(dir, inode, dentry);
344 _ACL_FREE(default_acl);
351 struct dentry *dentry,
353 struct nameidata *nd)
355 return xfs_vn_mknod(dir, dentry, mode, 0);
361 struct dentry *dentry,
364 return xfs_vn_mknod(dir, dentry, mode|S_IFDIR, 0);
367 STATIC struct dentry *
370 struct dentry *dentry,
371 struct nameidata *nd)
373 struct xfs_inode *cip;
374 struct xfs_name name;
377 if (dentry->d_name.len >= MAXNAMELEN)
378 return ERR_PTR(-ENAMETOOLONG);
380 xfs_dentry_to_name(&name, dentry);
381 error = xfs_lookup(XFS_I(dir), &name, &cip, NULL);
382 if (unlikely(error)) {
383 if (unlikely(error != ENOENT))
384 return ERR_PTR(-error);
389 return d_splice_alias(cip->i_vnode, dentry);
392 STATIC struct dentry *
395 struct dentry *dentry,
396 struct nameidata *nd)
398 struct xfs_inode *ip;
399 struct xfs_name xname;
400 struct xfs_name ci_name;
404 if (dentry->d_name.len >= MAXNAMELEN)
405 return ERR_PTR(-ENAMETOOLONG);
407 xfs_dentry_to_name(&xname, dentry);
408 error = xfs_lookup(XFS_I(dir), &xname, &ip, &ci_name);
409 if (unlikely(error)) {
410 if (unlikely(error != ENOENT))
411 return ERR_PTR(-error);
413 * call d_add(dentry, NULL) here when d_drop_negative_children
414 * is called in xfs_vn_mknod (ie. allow negative dentries
415 * with CI filesystems).
420 /* if exact match, just splice and exit */
422 return d_splice_alias(ip->i_vnode, dentry);
424 /* else case-insensitive match... */
425 dname.name = ci_name.name;
426 dname.len = ci_name.len;
427 dentry = d_add_ci(ip->i_vnode, dentry, &dname);
428 kmem_free(ci_name.name);
434 struct dentry *old_dentry,
436 struct dentry *dentry)
438 struct inode *inode; /* inode of guy being linked to */
439 struct xfs_name name;
442 inode = old_dentry->d_inode;
443 xfs_dentry_to_name(&name, dentry);
446 error = xfs_link(XFS_I(dir), XFS_I(inode), &name);
447 if (unlikely(error)) {
452 xfs_iflags_set(XFS_I(dir), XFS_IMODIFIED);
453 xfs_validate_fields(inode);
454 d_instantiate(dentry, inode);
461 struct dentry *dentry)
464 struct xfs_name name;
467 inode = dentry->d_inode;
468 xfs_dentry_to_name(&name, dentry);
470 error = xfs_remove(XFS_I(dir), &name, XFS_I(inode));
471 if (likely(!error)) {
472 xfs_validate_fields(dir); /* size needs update */
473 xfs_validate_fields(inode);
475 * With unlink, the VFS makes the dentry "negative": no inode,
476 * but still hashed. This is incompatible with case-insensitive
477 * mode, so invalidate (unhash) the dentry in CI-mode.
479 if (xfs_sb_version_hasasciici(&XFS_M(dir->i_sb)->m_sb))
480 d_invalidate(dentry);
488 struct dentry *dentry,
492 struct xfs_inode *cip = NULL;
493 struct xfs_name name;
498 (irix_symlink_mode ? 0777 & ~current->fs->umask : S_IRWXUGO);
499 xfs_dentry_to_name(&name, dentry);
501 error = xfs_symlink(XFS_I(dir), &name, symname, mode, &cip, NULL);
505 inode = cip->i_vnode;
507 error = xfs_init_security(inode, dir);
509 goto out_cleanup_inode;
511 d_instantiate(dentry, inode);
512 xfs_validate_fields(dir);
513 xfs_validate_fields(inode);
517 xfs_cleanup_inode(dir, inode, dentry);
525 struct dentry *odentry,
527 struct dentry *ndentry)
529 struct inode *new_inode = ndentry->d_inode;
530 struct xfs_name oname;
531 struct xfs_name nname;
534 xfs_dentry_to_name(&oname, odentry);
535 xfs_dentry_to_name(&nname, ndentry);
537 error = xfs_rename(XFS_I(odir), &oname, XFS_I(odentry->d_inode),
538 XFS_I(ndir), &nname, new_inode ?
539 XFS_I(new_inode) : NULL);
540 if (likely(!error)) {
542 xfs_validate_fields(new_inode);
543 xfs_validate_fields(odir);
545 xfs_validate_fields(ndir);
551 * careful here - this function can get called recursively, so
552 * we need to be very careful about how much stack we use.
553 * uio is kmalloced for this reason...
557 struct dentry *dentry,
558 struct nameidata *nd)
563 link = kmalloc(MAXPATHLEN+1, GFP_KERNEL);
567 error = -xfs_readlink(XFS_I(dentry->d_inode), link);
571 nd_set_link(nd, link);
577 nd_set_link(nd, ERR_PTR(error));
583 struct dentry *dentry,
584 struct nameidata *nd,
587 char *s = nd_get_link(nd);
593 #ifdef CONFIG_XFS_POSIX_ACL
599 struct xfs_inode *ip = XFS_I(inode);
602 xfs_itrace_entry(ip);
604 if (XFS_IFORK_Q(ip)) {
605 error = xfs_acl_iaccess(ip, mask, NULL);
618 return generic_permission(inode, mask, xfs_check_acl);
621 #define xfs_vn_permission NULL
626 struct vfsmount *mnt,
627 struct dentry *dentry,
630 struct inode *inode = dentry->d_inode;
631 struct xfs_inode *ip = XFS_I(inode);
632 struct xfs_mount *mp = ip->i_mount;
634 xfs_itrace_entry(ip);
636 if (XFS_FORCED_SHUTDOWN(mp))
637 return XFS_ERROR(EIO);
639 stat->size = XFS_ISIZE(ip);
640 stat->dev = inode->i_sb->s_dev;
641 stat->mode = ip->i_d.di_mode;
642 stat->nlink = ip->i_d.di_nlink;
643 stat->uid = ip->i_d.di_uid;
644 stat->gid = ip->i_d.di_gid;
645 stat->ino = ip->i_ino;
647 stat->ino += mp->m_inoadd;
649 stat->atime = inode->i_atime;
650 stat->mtime.tv_sec = ip->i_d.di_mtime.t_sec;
651 stat->mtime.tv_nsec = ip->i_d.di_mtime.t_nsec;
652 stat->ctime.tv_sec = ip->i_d.di_ctime.t_sec;
653 stat->ctime.tv_nsec = ip->i_d.di_ctime.t_nsec;
655 XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks);
658 switch (inode->i_mode & S_IFMT) {
661 stat->blksize = BLKDEV_IOSIZE;
662 stat->rdev = MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
663 sysv_minor(ip->i_df.if_u2.if_rdev));
666 if (XFS_IS_REALTIME_INODE(ip)) {
668 * If the file blocks are being allocated from a
669 * realtime volume, then return the inode's realtime
670 * extent size or the realtime volume's extent size.
673 xfs_get_extsz_hint(ip) << mp->m_sb.sb_blocklog;
675 stat->blksize = xfs_preferred_iosize(mp);
685 struct dentry *dentry,
688 struct inode *inode = dentry->d_inode;
689 unsigned int ia_valid = attr->ia_valid;
690 bhv_vattr_t vattr = { 0 };
694 if (ia_valid & ATTR_UID) {
695 vattr.va_mask |= XFS_AT_UID;
696 vattr.va_uid = attr->ia_uid;
698 if (ia_valid & ATTR_GID) {
699 vattr.va_mask |= XFS_AT_GID;
700 vattr.va_gid = attr->ia_gid;
702 if (ia_valid & ATTR_SIZE) {
703 vattr.va_mask |= XFS_AT_SIZE;
704 vattr.va_size = attr->ia_size;
706 if (ia_valid & ATTR_ATIME) {
707 vattr.va_mask |= XFS_AT_ATIME;
708 vattr.va_atime = attr->ia_atime;
709 inode->i_atime = attr->ia_atime;
711 if (ia_valid & ATTR_MTIME) {
712 vattr.va_mask |= XFS_AT_MTIME;
713 vattr.va_mtime = attr->ia_mtime;
715 if (ia_valid & ATTR_CTIME) {
716 vattr.va_mask |= XFS_AT_CTIME;
717 vattr.va_ctime = attr->ia_ctime;
719 if (ia_valid & ATTR_MODE) {
720 vattr.va_mask |= XFS_AT_MODE;
721 vattr.va_mode = attr->ia_mode;
722 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
723 inode->i_mode &= ~S_ISGID;
726 if (ia_valid & (ATTR_MTIME_SET | ATTR_ATIME_SET))
729 if ((ia_valid & ATTR_NO_BLOCK))
730 flags |= ATTR_NONBLOCK;
733 error = xfs_setattr(XFS_I(inode), &vattr, flags, NULL);
735 vn_revalidate(vn_from_inode(inode));
740 * block_truncate_page can return an error, but we can't propagate it
741 * at all here. Leave a complaint + stack trace in the syslog because
742 * this could be bad. If it is bad, we need to propagate the error further.
749 error = block_truncate_page(inode->i_mapping, inode->i_size,
764 xfs_inode_t *ip = XFS_I(inode);
766 /* preallocation on directories not yet supported */
768 if (S_ISDIR(inode->i_mode))
775 xfs_ilock(ip, XFS_IOLOCK_EXCL);
776 error = xfs_change_file_space(ip, XFS_IOC_RESVSP, &bf,
777 0, NULL, ATTR_NOLOCK);
778 if (!error && !(mode & FALLOC_FL_KEEP_SIZE) &&
779 offset + len > i_size_read(inode))
780 new_size = offset + len;
782 /* Change file size if needed */
786 va.va_mask = XFS_AT_SIZE;
787 va.va_size = new_size;
788 error = xfs_setattr(ip, &va, ATTR_NOLOCK, NULL);
791 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
796 const struct inode_operations xfs_inode_operations = {
797 .permission = xfs_vn_permission,
798 .truncate = xfs_vn_truncate,
799 .getattr = xfs_vn_getattr,
800 .setattr = xfs_vn_setattr,
801 .setxattr = generic_setxattr,
802 .getxattr = generic_getxattr,
803 .removexattr = generic_removexattr,
804 .listxattr = xfs_vn_listxattr,
805 .fallocate = xfs_vn_fallocate,
808 const struct inode_operations xfs_dir_inode_operations = {
809 .create = xfs_vn_create,
810 .lookup = xfs_vn_lookup,
812 .unlink = xfs_vn_unlink,
813 .symlink = xfs_vn_symlink,
814 .mkdir = xfs_vn_mkdir,
816 * Yes, XFS uses the same method for rmdir and unlink.
818 * There are some subtile differences deeper in the code,
819 * but we use S_ISDIR to check for those.
821 .rmdir = xfs_vn_unlink,
822 .mknod = xfs_vn_mknod,
823 .rename = xfs_vn_rename,
824 .permission = xfs_vn_permission,
825 .getattr = xfs_vn_getattr,
826 .setattr = xfs_vn_setattr,
827 .setxattr = generic_setxattr,
828 .getxattr = generic_getxattr,
829 .removexattr = generic_removexattr,
830 .listxattr = xfs_vn_listxattr,
833 const struct inode_operations xfs_dir_ci_inode_operations = {
834 .create = xfs_vn_create,
835 .lookup = xfs_vn_ci_lookup,
837 .unlink = xfs_vn_unlink,
838 .symlink = xfs_vn_symlink,
839 .mkdir = xfs_vn_mkdir,
841 * Yes, XFS uses the same method for rmdir and unlink.
843 * There are some subtile differences deeper in the code,
844 * but we use S_ISDIR to check for those.
846 .rmdir = xfs_vn_unlink,
847 .mknod = xfs_vn_mknod,
848 .rename = xfs_vn_rename,
849 .permission = xfs_vn_permission,
850 .getattr = xfs_vn_getattr,
851 .setattr = xfs_vn_setattr,
852 .setxattr = generic_setxattr,
853 .getxattr = generic_getxattr,
854 .removexattr = generic_removexattr,
855 .listxattr = xfs_vn_listxattr,
858 const struct inode_operations xfs_symlink_inode_operations = {
859 .readlink = generic_readlink,
860 .follow_link = xfs_vn_follow_link,
861 .put_link = xfs_vn_put_link,
862 .permission = xfs_vn_permission,
863 .getattr = xfs_vn_getattr,
864 .setattr = xfs_vn_setattr,
865 .setxattr = generic_setxattr,
866 .getxattr = generic_getxattr,
867 .removexattr = generic_removexattr,
868 .listxattr = xfs_vn_listxattr,