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(
67 vp = XFS_ITOV_NULL(ip);
69 ip->i_d.di_atime.t_sec = (__int32_t)vp->i_atime.tv_sec;
70 ip->i_d.di_atime.t_nsec = (__int32_t)vp->i_atime.tv_nsec;
75 * If the linux inode exists, mark it dirty.
76 * Used when commiting a dirty inode into a transaction so that
77 * the inode will get written back by the linux code
80 xfs_mark_inode_dirty_sync(
85 vp = XFS_ITOV_NULL(ip);
87 mark_inode_dirty_sync(vn_to_inode(vp));
91 * Change the requested timestamp in the given inode.
92 * We don't lock across timestamp updates, and we don't log them but
93 * we do record the fact that there is dirty information in core.
95 * NOTE -- callers MUST combine XFS_ICHGTIME_MOD or XFS_ICHGTIME_CHG
96 * with XFS_ICHGTIME_ACC to be sure that access time
97 * update will take. Calling first with XFS_ICHGTIME_ACC
98 * and then XFS_ICHGTIME_MOD may fail to modify the access
99 * timestamp if the filesystem is mounted noacctm.
106 struct inode *inode = vn_to_inode(XFS_ITOV(ip));
110 if (flags & XFS_ICHGTIME_MOD) {
112 ip->i_d.di_mtime.t_sec = (__int32_t)tv.tv_sec;
113 ip->i_d.di_mtime.t_nsec = (__int32_t)tv.tv_nsec;
115 if (flags & XFS_ICHGTIME_ACC) {
117 ip->i_d.di_atime.t_sec = (__int32_t)tv.tv_sec;
118 ip->i_d.di_atime.t_nsec = (__int32_t)tv.tv_nsec;
120 if (flags & XFS_ICHGTIME_CHG) {
122 ip->i_d.di_ctime.t_sec = (__int32_t)tv.tv_sec;
123 ip->i_d.di_ctime.t_nsec = (__int32_t)tv.tv_nsec;
127 * We update the i_update_core field _after_ changing
128 * the timestamps in order to coordinate properly with
129 * xfs_iflush() so that we don't lose timestamp updates.
130 * This keeps us from having to hold the inode lock
131 * while doing this. We use the SYNCHRONIZE macro to
132 * ensure that the compiler does not reorder the update
133 * of i_update_core above the timestamp updates above.
136 ip->i_update_core = 1;
137 if (!(inode->i_state & I_NEW))
138 mark_inode_dirty_sync(inode);
142 * Variant on the above which avoids querying the system clock
143 * in situations where we know the Linux inode timestamps have
144 * just been updated (and so we can update our inode cheaply).
155 * Atime updates for read() & friends are handled lazily now, and
156 * explicit updates must go through xfs_ichgtime()
158 ASSERT((flags & XFS_ICHGTIME_ACC) == 0);
161 * We're not supposed to change timestamps in readonly-mounted
162 * filesystems. Throw it away if anyone asks us.
164 if (unlikely(IS_RDONLY(inode)))
167 if (flags & XFS_ICHGTIME_MOD) {
168 tvp = &inode->i_mtime;
169 ip->i_d.di_mtime.t_sec = (__int32_t)tvp->tv_sec;
170 ip->i_d.di_mtime.t_nsec = (__int32_t)tvp->tv_nsec;
172 if (flags & XFS_ICHGTIME_CHG) {
173 tvp = &inode->i_ctime;
174 ip->i_d.di_ctime.t_sec = (__int32_t)tvp->tv_sec;
175 ip->i_d.di_ctime.t_nsec = (__int32_t)tvp->tv_nsec;
179 * We update the i_update_core field _after_ changing
180 * the timestamps in order to coordinate properly with
181 * xfs_iflush() so that we don't lose timestamp updates.
182 * This keeps us from having to hold the inode lock
183 * while doing this. We use the SYNCHRONIZE macro to
184 * ensure that the compiler does not reorder the update
185 * of i_update_core above the timestamp updates above.
188 ip->i_update_core = 1;
189 if (!(inode->i_state & I_NEW))
190 mark_inode_dirty_sync(inode);
195 * Pull the link count and size up from the xfs inode to the linux inode
201 struct xfs_inode *ip = XFS_I(inode);
204 /* we're under i_sem so i_size can't change under us */
205 size = XFS_ISIZE(ip);
206 if (i_size_read(inode) != size)
207 i_size_write(inode, size);
211 * Hook in SELinux. This is not quite correct yet, what we really need
212 * here (as we do for default ACLs) is a mechanism by which creation of
213 * these attrs can be journalled at inode creation time (along with the
214 * inode, of course, such that log replay can't cause these to be lost).
221 struct inode *ip = vn_to_inode(vp);
227 error = security_inode_init_security(ip, dir, &name, &value, &length);
229 if (error == -EOPNOTSUPP)
234 error = xfs_attr_set(XFS_I(ip), name, value,
235 length, ATTR_SECURE);
237 xfs_iflags_set(XFS_I(ip), XFS_IMODIFIED);
248 struct dentry *dentry,
251 struct dentry teardown = {};
254 * If we can't add the ACL or we fail in
255 * xfs_init_security we must back out.
256 * ENOSPC can hit here, among other things.
258 teardown.d_inode = vn_to_inode(vp);
259 teardown.d_name = dentry->d_name;
262 xfs_rmdir(XFS_I(dir), &teardown);
264 xfs_remove(XFS_I(dir), &teardown);
271 struct dentry *dentry,
276 struct xfs_inode *ip = NULL;
277 xfs_acl_t *default_acl = NULL;
278 attrexists_t test_default_acl = _ACL_DEFAULT_EXISTS;
282 * Irix uses Missed'em'V split, but doesn't want to see
283 * the upper 5 bits of (14bit) major.
285 if (unlikely(!sysv_valid_dev(rdev) || MAJOR(rdev) & ~0x1ff))
288 if (test_default_acl && test_default_acl(dir)) {
289 if (!_ACL_ALLOC(default_acl)) {
292 if (!_ACL_GET_DEFAULT(dir, default_acl)) {
293 _ACL_FREE(default_acl);
298 if (IS_POSIXACL(dir) && !default_acl)
299 mode &= ~current->fs->umask;
301 switch (mode & S_IFMT) {
306 rdev = sysv_encode_dev(rdev);
308 error = xfs_create(XFS_I(dir), dentry, mode, rdev, &ip, NULL);
311 error = xfs_mkdir(XFS_I(dir), dentry, mode, &ip, NULL);
323 error = xfs_init_security(inode, dir);
325 goto out_cleanup_inode;
328 error = _ACL_INHERIT(inode, mode, default_acl);
330 goto out_cleanup_inode;
331 xfs_iflags_set(ip, XFS_IMODIFIED);
332 _ACL_FREE(default_acl);
337 xfs_validate_fields(inode);
338 d_instantiate(dentry, inode);
339 xfs_validate_fields(dir);
343 xfs_cleanup_inode(dir, inode, dentry, mode);
346 _ACL_FREE(default_acl);
353 struct dentry *dentry,
355 struct nameidata *nd)
357 return xfs_vn_mknod(dir, dentry, mode, 0);
363 struct dentry *dentry,
366 return xfs_vn_mknod(dir, dentry, mode|S_IFDIR, 0);
369 STATIC struct dentry *
372 struct dentry *dentry,
373 struct nameidata *nd)
375 struct xfs_inode *cip;
378 if (dentry->d_name.len >= MAXNAMELEN)
379 return ERR_PTR(-ENAMETOOLONG);
381 error = xfs_lookup(XFS_I(dir), dentry, &cip);
382 if (unlikely(error)) {
383 if (unlikely(error != ENOENT))
384 return ERR_PTR(-error);
389 return d_splice_alias(cip->i_vnode, dentry);
394 struct dentry *old_dentry,
396 struct dentry *dentry)
398 struct inode *inode; /* inode of guy being linked to */
401 inode = old_dentry->d_inode;
404 error = xfs_link(XFS_I(dir), XFS_I(inode), dentry);
405 if (unlikely(error)) {
410 xfs_iflags_set(XFS_I(dir), XFS_IMODIFIED);
411 xfs_validate_fields(inode);
412 d_instantiate(dentry, inode);
419 struct dentry *dentry)
424 inode = dentry->d_inode;
426 error = xfs_remove(XFS_I(dir), dentry);
427 if (likely(!error)) {
428 xfs_validate_fields(dir); /* size needs update */
429 xfs_validate_fields(inode);
437 struct dentry *dentry,
441 struct xfs_inode *cip = NULL;
446 (irix_symlink_mode ? 0777 & ~current->fs->umask : S_IRWXUGO);
448 error = xfs_symlink(XFS_I(dir), dentry, (char *)symname, mode,
453 inode = cip->i_vnode;
455 error = xfs_init_security(inode, dir);
457 goto out_cleanup_inode;
459 d_instantiate(dentry, inode);
460 xfs_validate_fields(dir);
461 xfs_validate_fields(inode);
465 xfs_cleanup_inode(dir, inode, dentry, 0);
473 struct dentry *dentry)
475 struct inode *inode = dentry->d_inode;
478 error = xfs_rmdir(XFS_I(dir), dentry);
479 if (likely(!error)) {
480 xfs_validate_fields(inode);
481 xfs_validate_fields(dir);
489 struct dentry *odentry,
491 struct dentry *ndentry)
493 struct inode *new_inode = ndentry->d_inode;
496 error = xfs_rename(XFS_I(odir), odentry, XFS_I(ndir), ndentry);
497 if (likely(!error)) {
499 xfs_validate_fields(new_inode);
500 xfs_validate_fields(odir);
502 xfs_validate_fields(ndir);
508 * careful here - this function can get called recursively, so
509 * we need to be very careful about how much stack we use.
510 * uio is kmalloced for this reason...
514 struct dentry *dentry,
515 struct nameidata *nd)
520 link = kmalloc(MAXPATHLEN+1, GFP_KERNEL);
524 error = -xfs_readlink(XFS_I(dentry->d_inode), link);
528 nd_set_link(nd, link);
534 nd_set_link(nd, ERR_PTR(error));
540 struct dentry *dentry,
541 struct nameidata *nd,
544 char *s = nd_get_link(nd);
550 #ifdef CONFIG_XFS_POSIX_ACL
556 struct xfs_inode *ip = XFS_I(inode);
559 xfs_itrace_entry(ip);
561 if (XFS_IFORK_Q(ip)) {
562 error = xfs_acl_iaccess(ip, mask, NULL);
574 struct nameidata *nd)
576 return generic_permission(inode, mask, xfs_check_acl);
579 #define xfs_vn_permission NULL
584 struct vfsmount *mnt,
585 struct dentry *dentry,
588 struct inode *inode = dentry->d_inode;
589 struct xfs_inode *ip = XFS_I(inode);
590 struct xfs_mount *mp = ip->i_mount;
592 xfs_itrace_entry(ip);
594 if (XFS_FORCED_SHUTDOWN(mp))
595 return XFS_ERROR(EIO);
597 stat->size = XFS_ISIZE(ip);
598 stat->dev = inode->i_sb->s_dev;
599 stat->mode = ip->i_d.di_mode;
600 stat->nlink = ip->i_d.di_nlink;
601 stat->uid = ip->i_d.di_uid;
602 stat->gid = ip->i_d.di_gid;
603 stat->ino = ip->i_ino;
605 stat->ino += mp->m_inoadd;
607 stat->atime = inode->i_atime;
608 stat->mtime.tv_sec = ip->i_d.di_mtime.t_sec;
609 stat->mtime.tv_nsec = ip->i_d.di_mtime.t_nsec;
610 stat->ctime.tv_sec = ip->i_d.di_ctime.t_sec;
611 stat->ctime.tv_nsec = ip->i_d.di_ctime.t_nsec;
613 XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks);
616 switch (inode->i_mode & S_IFMT) {
619 stat->blksize = BLKDEV_IOSIZE;
620 stat->rdev = MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
621 sysv_minor(ip->i_df.if_u2.if_rdev));
624 if (XFS_IS_REALTIME_INODE(ip)) {
626 * If the file blocks are being allocated from a
627 * realtime volume, then return the inode's realtime
628 * extent size or the realtime volume's extent size.
631 xfs_get_extsz_hint(ip) << mp->m_sb.sb_blocklog;
633 stat->blksize = xfs_preferred_iosize(mp);
643 struct dentry *dentry,
646 struct inode *inode = dentry->d_inode;
647 unsigned int ia_valid = attr->ia_valid;
648 bhv_vattr_t vattr = { 0 };
652 if (ia_valid & ATTR_UID) {
653 vattr.va_mask |= XFS_AT_UID;
654 vattr.va_uid = attr->ia_uid;
656 if (ia_valid & ATTR_GID) {
657 vattr.va_mask |= XFS_AT_GID;
658 vattr.va_gid = attr->ia_gid;
660 if (ia_valid & ATTR_SIZE) {
661 vattr.va_mask |= XFS_AT_SIZE;
662 vattr.va_size = attr->ia_size;
664 if (ia_valid & ATTR_ATIME) {
665 vattr.va_mask |= XFS_AT_ATIME;
666 vattr.va_atime = attr->ia_atime;
667 inode->i_atime = attr->ia_atime;
669 if (ia_valid & ATTR_MTIME) {
670 vattr.va_mask |= XFS_AT_MTIME;
671 vattr.va_mtime = attr->ia_mtime;
673 if (ia_valid & ATTR_CTIME) {
674 vattr.va_mask |= XFS_AT_CTIME;
675 vattr.va_ctime = attr->ia_ctime;
677 if (ia_valid & ATTR_MODE) {
678 vattr.va_mask |= XFS_AT_MODE;
679 vattr.va_mode = attr->ia_mode;
680 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
681 inode->i_mode &= ~S_ISGID;
684 if (ia_valid & (ATTR_MTIME_SET | ATTR_ATIME_SET))
687 if ((ia_valid & ATTR_NO_BLOCK))
688 flags |= ATTR_NONBLOCK;
691 error = xfs_setattr(XFS_I(inode), &vattr, flags, NULL);
693 vn_revalidate(vn_from_inode(inode));
701 block_truncate_page(inode->i_mapping, inode->i_size, xfs_get_blocks);
706 struct dentry *dentry,
712 bhv_vnode_t *vp = vn_from_inode(dentry->d_inode);
713 char *attr = (char *)name;
718 namesp = attr_lookup_namespace(attr, attr_namespaces, ATTR_NAMECOUNT);
721 attr += namesp->attr_namelen;
722 error = namesp->attr_capable(vp, NULL);
726 /* Convert Linux syscall to XFS internal ATTR flags */
727 if (flags & XATTR_CREATE)
728 xflags |= ATTR_CREATE;
729 if (flags & XATTR_REPLACE)
730 xflags |= ATTR_REPLACE;
731 xflags |= namesp->attr_flag;
732 return namesp->attr_set(vp, attr, (void *)data, size, xflags);
737 struct dentry *dentry,
742 bhv_vnode_t *vp = vn_from_inode(dentry->d_inode);
743 char *attr = (char *)name;
748 namesp = attr_lookup_namespace(attr, attr_namespaces, ATTR_NAMECOUNT);
751 attr += namesp->attr_namelen;
752 error = namesp->attr_capable(vp, NULL);
756 /* Convert Linux syscall to XFS internal ATTR flags */
758 xflags |= ATTR_KERNOVAL;
761 xflags |= namesp->attr_flag;
762 return namesp->attr_get(vp, attr, (void *)data, size, xflags);
767 struct dentry *dentry,
771 bhv_vnode_t *vp = vn_from_inode(dentry->d_inode);
772 int error, xflags = ATTR_KERNAMELS;
776 xflags |= ATTR_KERNOVAL;
777 xflags |= capable(CAP_SYS_ADMIN) ? ATTR_KERNFULLS : ATTR_KERNORMALS;
779 error = attr_generic_list(vp, data, size, xflags, &result);
787 struct dentry *dentry,
790 bhv_vnode_t *vp = vn_from_inode(dentry->d_inode);
791 char *attr = (char *)name;
796 namesp = attr_lookup_namespace(attr, attr_namespaces, ATTR_NAMECOUNT);
799 attr += namesp->attr_namelen;
800 error = namesp->attr_capable(vp, NULL);
803 xflags |= namesp->attr_flag;
804 return namesp->attr_remove(vp, attr, xflags);
817 xfs_inode_t *ip = XFS_I(inode);
819 /* preallocation on directories not yet supported */
821 if (S_ISDIR(inode->i_mode))
828 xfs_ilock(ip, XFS_IOLOCK_EXCL);
829 error = xfs_change_file_space(ip, XFS_IOC_RESVSP, &bf,
830 0, NULL, ATTR_NOLOCK);
831 if (!error && !(mode & FALLOC_FL_KEEP_SIZE) &&
832 offset + len > i_size_read(inode))
833 new_size = offset + len;
835 /* Change file size if needed */
839 va.va_mask = XFS_AT_SIZE;
840 va.va_size = new_size;
841 error = xfs_setattr(ip, &va, ATTR_NOLOCK, NULL);
844 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
849 const struct inode_operations xfs_inode_operations = {
850 .permission = xfs_vn_permission,
851 .truncate = xfs_vn_truncate,
852 .getattr = xfs_vn_getattr,
853 .setattr = xfs_vn_setattr,
854 .setxattr = xfs_vn_setxattr,
855 .getxattr = xfs_vn_getxattr,
856 .listxattr = xfs_vn_listxattr,
857 .removexattr = xfs_vn_removexattr,
858 .fallocate = xfs_vn_fallocate,
861 const struct inode_operations xfs_dir_inode_operations = {
862 .create = xfs_vn_create,
863 .lookup = xfs_vn_lookup,
865 .unlink = xfs_vn_unlink,
866 .symlink = xfs_vn_symlink,
867 .mkdir = xfs_vn_mkdir,
868 .rmdir = xfs_vn_rmdir,
869 .mknod = xfs_vn_mknod,
870 .rename = xfs_vn_rename,
871 .permission = xfs_vn_permission,
872 .getattr = xfs_vn_getattr,
873 .setattr = xfs_vn_setattr,
874 .setxattr = xfs_vn_setxattr,
875 .getxattr = xfs_vn_getxattr,
876 .listxattr = xfs_vn_listxattr,
877 .removexattr = xfs_vn_removexattr,
880 const struct inode_operations xfs_symlink_inode_operations = {
881 .readlink = generic_readlink,
882 .follow_link = xfs_vn_follow_link,
883 .put_link = xfs_vn_put_link,
884 .permission = xfs_vn_permission,
885 .getattr = xfs_vn_getattr,
886 .setattr = xfs_vn_setattr,
887 .setxattr = xfs_vn_setxattr,
888 .getxattr = xfs_vn_getxattr,
889 .listxattr = xfs_vn_listxattr,
890 .removexattr = xfs_vn_removexattr,