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)
378 if (dentry->d_name.len >= MAXNAMELEN)
379 return ERR_PTR(-ENAMETOOLONG);
381 error = xfs_lookup(XFS_I(dir), dentry, &cvp);
382 if (unlikely(error)) {
383 if (unlikely(error != ENOENT))
384 return ERR_PTR(-error);
389 return d_splice_alias(vn_to_inode(cvp), 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 bhv_vnode_t *cvp; /* used to lookup symlink to put in dentry */
448 (irix_symlink_mode ? 0777 & ~current->fs->umask : S_IRWXUGO);
450 error = xfs_symlink(XFS_I(dir), dentry, (char *)symname, mode,
452 if (likely(!error && cvp)) {
453 error = xfs_init_security(cvp, dir);
454 if (likely(!error)) {
455 ip = vn_to_inode(cvp);
456 d_instantiate(dentry, ip);
457 xfs_validate_fields(dir);
458 xfs_validate_fields(ip);
460 xfs_cleanup_inode(dir, cvp, dentry, 0);
469 struct dentry *dentry)
471 struct inode *inode = dentry->d_inode;
474 error = xfs_rmdir(XFS_I(dir), dentry);
475 if (likely(!error)) {
476 xfs_validate_fields(inode);
477 xfs_validate_fields(dir);
485 struct dentry *odentry,
487 struct dentry *ndentry)
489 struct inode *new_inode = ndentry->d_inode;
490 bhv_vnode_t *tvp; /* target directory */
493 tvp = vn_from_inode(ndir);
495 error = xfs_rename(XFS_I(odir), odentry, tvp, ndentry);
496 if (likely(!error)) {
498 xfs_validate_fields(new_inode);
499 xfs_validate_fields(odir);
501 xfs_validate_fields(ndir);
507 * careful here - this function can get called recursively, so
508 * we need to be very careful about how much stack we use.
509 * uio is kmalloced for this reason...
513 struct dentry *dentry,
514 struct nameidata *nd)
519 link = kmalloc(MAXPATHLEN+1, GFP_KERNEL);
523 error = -xfs_readlink(XFS_I(dentry->d_inode), link);
527 nd_set_link(nd, link);
533 nd_set_link(nd, ERR_PTR(error));
539 struct dentry *dentry,
540 struct nameidata *nd,
543 char *s = nd_get_link(nd);
549 #ifdef CONFIG_XFS_POSIX_ACL
555 struct xfs_inode *ip = XFS_I(inode);
558 xfs_itrace_entry(ip);
560 if (XFS_IFORK_Q(ip)) {
561 error = xfs_acl_iaccess(ip, mask, NULL);
573 struct nameidata *nd)
575 return generic_permission(inode, mask, xfs_check_acl);
578 #define xfs_vn_permission NULL
583 struct vfsmount *mnt,
584 struct dentry *dentry,
587 struct inode *inode = dentry->d_inode;
588 struct xfs_inode *ip = XFS_I(inode);
589 struct xfs_mount *mp = ip->i_mount;
591 xfs_itrace_entry(ip);
593 if (XFS_FORCED_SHUTDOWN(mp))
594 return XFS_ERROR(EIO);
596 stat->size = XFS_ISIZE(ip);
597 stat->dev = inode->i_sb->s_dev;
598 stat->mode = ip->i_d.di_mode;
599 stat->nlink = ip->i_d.di_nlink;
600 stat->uid = ip->i_d.di_uid;
601 stat->gid = ip->i_d.di_gid;
602 stat->ino = ip->i_ino;
604 stat->ino += mp->m_inoadd;
606 stat->atime = inode->i_atime;
607 stat->mtime.tv_sec = ip->i_d.di_mtime.t_sec;
608 stat->mtime.tv_nsec = ip->i_d.di_mtime.t_nsec;
609 stat->ctime.tv_sec = ip->i_d.di_ctime.t_sec;
610 stat->ctime.tv_nsec = ip->i_d.di_ctime.t_nsec;
612 XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks);
615 switch (inode->i_mode & S_IFMT) {
618 stat->blksize = BLKDEV_IOSIZE;
619 stat->rdev = MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
620 sysv_minor(ip->i_df.if_u2.if_rdev));
623 if (XFS_IS_REALTIME_INODE(ip)) {
625 * If the file blocks are being allocated from a
626 * realtime volume, then return the inode's realtime
627 * extent size or the realtime volume's extent size.
630 xfs_get_extsz_hint(ip) << mp->m_sb.sb_blocklog;
632 stat->blksize = xfs_preferred_iosize(mp);
642 struct dentry *dentry,
645 struct inode *inode = dentry->d_inode;
646 unsigned int ia_valid = attr->ia_valid;
647 bhv_vattr_t vattr = { 0 };
651 if (ia_valid & ATTR_UID) {
652 vattr.va_mask |= XFS_AT_UID;
653 vattr.va_uid = attr->ia_uid;
655 if (ia_valid & ATTR_GID) {
656 vattr.va_mask |= XFS_AT_GID;
657 vattr.va_gid = attr->ia_gid;
659 if (ia_valid & ATTR_SIZE) {
660 vattr.va_mask |= XFS_AT_SIZE;
661 vattr.va_size = attr->ia_size;
663 if (ia_valid & ATTR_ATIME) {
664 vattr.va_mask |= XFS_AT_ATIME;
665 vattr.va_atime = attr->ia_atime;
666 inode->i_atime = attr->ia_atime;
668 if (ia_valid & ATTR_MTIME) {
669 vattr.va_mask |= XFS_AT_MTIME;
670 vattr.va_mtime = attr->ia_mtime;
672 if (ia_valid & ATTR_CTIME) {
673 vattr.va_mask |= XFS_AT_CTIME;
674 vattr.va_ctime = attr->ia_ctime;
676 if (ia_valid & ATTR_MODE) {
677 vattr.va_mask |= XFS_AT_MODE;
678 vattr.va_mode = attr->ia_mode;
679 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
680 inode->i_mode &= ~S_ISGID;
683 if (ia_valid & (ATTR_MTIME_SET | ATTR_ATIME_SET))
686 if ((ia_valid & ATTR_NO_BLOCK))
687 flags |= ATTR_NONBLOCK;
690 error = xfs_setattr(XFS_I(inode), &vattr, flags, NULL);
692 vn_revalidate(vn_from_inode(inode));
700 block_truncate_page(inode->i_mapping, inode->i_size, xfs_get_blocks);
705 struct dentry *dentry,
711 bhv_vnode_t *vp = vn_from_inode(dentry->d_inode);
712 char *attr = (char *)name;
717 namesp = attr_lookup_namespace(attr, attr_namespaces, ATTR_NAMECOUNT);
720 attr += namesp->attr_namelen;
721 error = namesp->attr_capable(vp, NULL);
725 /* Convert Linux syscall to XFS internal ATTR flags */
726 if (flags & XATTR_CREATE)
727 xflags |= ATTR_CREATE;
728 if (flags & XATTR_REPLACE)
729 xflags |= ATTR_REPLACE;
730 xflags |= namesp->attr_flag;
731 return namesp->attr_set(vp, attr, (void *)data, size, xflags);
736 struct dentry *dentry,
741 bhv_vnode_t *vp = vn_from_inode(dentry->d_inode);
742 char *attr = (char *)name;
747 namesp = attr_lookup_namespace(attr, attr_namespaces, ATTR_NAMECOUNT);
750 attr += namesp->attr_namelen;
751 error = namesp->attr_capable(vp, NULL);
755 /* Convert Linux syscall to XFS internal ATTR flags */
757 xflags |= ATTR_KERNOVAL;
760 xflags |= namesp->attr_flag;
761 return namesp->attr_get(vp, attr, (void *)data, size, xflags);
766 struct dentry *dentry,
770 bhv_vnode_t *vp = vn_from_inode(dentry->d_inode);
771 int error, xflags = ATTR_KERNAMELS;
775 xflags |= ATTR_KERNOVAL;
776 xflags |= capable(CAP_SYS_ADMIN) ? ATTR_KERNFULLS : ATTR_KERNORMALS;
778 error = attr_generic_list(vp, data, size, xflags, &result);
786 struct dentry *dentry,
789 bhv_vnode_t *vp = vn_from_inode(dentry->d_inode);
790 char *attr = (char *)name;
795 namesp = attr_lookup_namespace(attr, attr_namespaces, ATTR_NAMECOUNT);
798 attr += namesp->attr_namelen;
799 error = namesp->attr_capable(vp, NULL);
802 xflags |= namesp->attr_flag;
803 return namesp->attr_remove(vp, attr, xflags);
816 xfs_inode_t *ip = XFS_I(inode);
818 /* preallocation on directories not yet supported */
820 if (S_ISDIR(inode->i_mode))
827 xfs_ilock(ip, XFS_IOLOCK_EXCL);
828 error = xfs_change_file_space(ip, XFS_IOC_RESVSP, &bf,
829 0, NULL, ATTR_NOLOCK);
830 if (!error && !(mode & FALLOC_FL_KEEP_SIZE) &&
831 offset + len > i_size_read(inode))
832 new_size = offset + len;
834 /* Change file size if needed */
838 va.va_mask = XFS_AT_SIZE;
839 va.va_size = new_size;
840 error = xfs_setattr(ip, &va, ATTR_NOLOCK, NULL);
843 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
848 const struct inode_operations xfs_inode_operations = {
849 .permission = xfs_vn_permission,
850 .truncate = xfs_vn_truncate,
851 .getattr = xfs_vn_getattr,
852 .setattr = xfs_vn_setattr,
853 .setxattr = xfs_vn_setxattr,
854 .getxattr = xfs_vn_getxattr,
855 .listxattr = xfs_vn_listxattr,
856 .removexattr = xfs_vn_removexattr,
857 .fallocate = xfs_vn_fallocate,
860 const struct inode_operations xfs_dir_inode_operations = {
861 .create = xfs_vn_create,
862 .lookup = xfs_vn_lookup,
864 .unlink = xfs_vn_unlink,
865 .symlink = xfs_vn_symlink,
866 .mkdir = xfs_vn_mkdir,
867 .rmdir = xfs_vn_rmdir,
868 .mknod = xfs_vn_mknod,
869 .rename = xfs_vn_rename,
870 .permission = xfs_vn_permission,
871 .getattr = xfs_vn_getattr,
872 .setattr = xfs_vn_setattr,
873 .setxattr = xfs_vn_setxattr,
874 .getxattr = xfs_vn_getxattr,
875 .listxattr = xfs_vn_listxattr,
876 .removexattr = xfs_vn_removexattr,
879 const struct inode_operations xfs_symlink_inode_operations = {
880 .readlink = generic_readlink,
881 .follow_link = xfs_vn_follow_link,
882 .put_link = xfs_vn_put_link,
883 .permission = xfs_vn_permission,
884 .getattr = xfs_vn_getattr,
885 .setattr = xfs_vn_setattr,
886 .setxattr = xfs_vn_setxattr,
887 .getxattr = xfs_vn_getxattr,
888 .listxattr = xfs_vn_listxattr,
889 .removexattr = xfs_vn_removexattr,