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"
49 #include "xfs_buf_item.h"
50 #include "xfs_utils.h"
52 #include <linux/capability.h>
53 #include <linux/xattr.h>
54 #include <linux/namei.h>
55 #include <linux/security.h>
58 * Get a XFS inode from a given vnode.
66 bdp = bhv_lookup_range(VN_BHV_HEAD(vp),
67 VNODE_POSITION_XFS, VNODE_POSITION_XFS);
68 if (unlikely(bdp == NULL))
70 return XFS_BHVTOI(bdp);
74 * Bring the atime in the XFS inode uptodate.
75 * Used before logging the inode to disk or when the Linux inode goes away.
78 xfs_synchronize_atime(
83 vp = XFS_ITOV_NULL(ip);
85 struct inode *inode = &vp->v_inode;
86 ip->i_d.di_atime.t_sec = (__int32_t)inode->i_atime.tv_sec;
87 ip->i_d.di_atime.t_nsec = (__int32_t)inode->i_atime.tv_nsec;
92 * Change the requested timestamp in the given inode.
93 * We don't lock across timestamp updates, and we don't log them but
94 * we do record the fact that there is dirty information in core.
96 * NOTE -- callers MUST combine XFS_ICHGTIME_MOD or XFS_ICHGTIME_CHG
97 * with XFS_ICHGTIME_ACC to be sure that access time
98 * update will take. Calling first with XFS_ICHGTIME_ACC
99 * and then XFS_ICHGTIME_MOD may fail to modify the access
100 * timestamp if the filesystem is mounted noacctm.
107 struct inode *inode = vn_to_inode(XFS_ITOV(ip));
111 if (flags & XFS_ICHGTIME_MOD) {
113 ip->i_d.di_mtime.t_sec = (__int32_t)tv.tv_sec;
114 ip->i_d.di_mtime.t_nsec = (__int32_t)tv.tv_nsec;
116 if (flags & XFS_ICHGTIME_ACC) {
118 ip->i_d.di_atime.t_sec = (__int32_t)tv.tv_sec;
119 ip->i_d.di_atime.t_nsec = (__int32_t)tv.tv_nsec;
121 if (flags & XFS_ICHGTIME_CHG) {
123 ip->i_d.di_ctime.t_sec = (__int32_t)tv.tv_sec;
124 ip->i_d.di_ctime.t_nsec = (__int32_t)tv.tv_nsec;
128 * We update the i_update_core field _after_ changing
129 * the timestamps in order to coordinate properly with
130 * xfs_iflush() so that we don't lose timestamp updates.
131 * This keeps us from having to hold the inode lock
132 * while doing this. We use the SYNCHRONIZE macro to
133 * ensure that the compiler does not reorder the update
134 * of i_update_core above the timestamp updates above.
137 ip->i_update_core = 1;
138 if (!(inode->i_state & I_LOCK))
139 mark_inode_dirty_sync(inode);
143 * Variant on the above which avoids querying the system clock
144 * in situations where we know the Linux inode timestamps have
145 * just been updated (and so we can update our inode cheaply).
156 * Atime updates for read() & friends are handled lazily now, and
157 * explicit updates must go through xfs_ichgtime()
159 ASSERT((flags & XFS_ICHGTIME_ACC) == 0);
162 * We're not supposed to change timestamps in readonly-mounted
163 * filesystems. Throw it away if anyone asks us.
165 if (unlikely(IS_RDONLY(inode)))
168 if (flags & XFS_ICHGTIME_MOD) {
169 tvp = &inode->i_mtime;
170 ip->i_d.di_mtime.t_sec = (__int32_t)tvp->tv_sec;
171 ip->i_d.di_mtime.t_nsec = (__int32_t)tvp->tv_nsec;
173 if (flags & XFS_ICHGTIME_CHG) {
174 tvp = &inode->i_ctime;
175 ip->i_d.di_ctime.t_sec = (__int32_t)tvp->tv_sec;
176 ip->i_d.di_ctime.t_nsec = (__int32_t)tvp->tv_nsec;
180 * We update the i_update_core field _after_ changing
181 * the timestamps in order to coordinate properly with
182 * xfs_iflush() so that we don't lose timestamp updates.
183 * This keeps us from having to hold the inode lock
184 * while doing this. We use the SYNCHRONIZE macro to
185 * ensure that the compiler does not reorder the update
186 * of i_update_core above the timestamp updates above.
189 ip->i_update_core = 1;
190 if (!(inode->i_state & I_LOCK))
191 mark_inode_dirty_sync(inode);
196 * Pull the link count and size up from the xfs inode to the linux inode
203 vattr->va_mask = XFS_AT_NLINK|XFS_AT_SIZE|XFS_AT_NBLOCKS;
204 if (!bhv_vop_getattr(vn_from_inode(ip), vattr, ATTR_LAZY, NULL)) {
205 ip->i_nlink = vattr->va_nlink;
206 ip->i_blocks = vattr->va_nblocks;
208 /* we're under i_sem so i_size can't change under us */
209 if (i_size_read(ip) != vattr->va_size)
210 i_size_write(ip, vattr->va_size);
215 * Hook in SELinux. This is not quite correct yet, what we really need
216 * here (as we do for default ACLs) is a mechanism by which creation of
217 * these attrs can be journalled at inode creation time (along with the
218 * inode, of course, such that log replay can't cause these to be lost).
225 struct inode *ip = vn_to_inode(vp);
231 error = security_inode_init_security(ip, dir, &name, &value, &length);
233 if (error == -EOPNOTSUPP)
238 error = bhv_vop_attr_set(vp, name, value, length, ATTR_SECURE, NULL);
248 * Determine whether a process has a valid fs_struct (kernel daemons
249 * like knfsd don't have an fs_struct).
251 * XXX(hch): nfsd is broken, better fix it instead.
254 xfs_has_fs_struct(struct task_struct *task)
256 return (task->fs != init_task.fs);
263 struct dentry *dentry,
266 struct dentry teardown = {};
269 * If we can't add the ACL or we fail in
270 * xfs_init_security we must back out.
271 * ENOSPC can hit here, among other things.
273 teardown.d_inode = vn_to_inode(vp);
274 teardown.d_name = dentry->d_name;
277 bhv_vop_rmdir(dvp, &teardown, NULL);
279 bhv_vop_remove(dvp, &teardown, NULL);
286 struct dentry *dentry,
291 bhv_vattr_t vattr = { 0 };
292 bhv_vnode_t *vp = NULL, *dvp = vn_from_inode(dir);
293 xfs_acl_t *default_acl = NULL;
294 attrexists_t test_default_acl = _ACL_DEFAULT_EXISTS;
298 * Irix uses Missed'em'V split, but doesn't want to see
299 * the upper 5 bits of (14bit) major.
301 if (unlikely(!sysv_valid_dev(rdev) || MAJOR(rdev) & ~0x1ff))
304 if (unlikely(test_default_acl && test_default_acl(dvp))) {
305 if (!_ACL_ALLOC(default_acl)) {
308 if (!_ACL_GET_DEFAULT(dvp, default_acl)) {
309 _ACL_FREE(default_acl);
314 if (IS_POSIXACL(dir) && !default_acl && xfs_has_fs_struct(current))
315 mode &= ~current->fs->umask;
317 vattr.va_mask = XFS_AT_TYPE|XFS_AT_MODE;
318 vattr.va_mode = mode;
320 switch (mode & S_IFMT) {
321 case S_IFCHR: case S_IFBLK: case S_IFIFO: case S_IFSOCK:
322 vattr.va_rdev = sysv_encode_dev(rdev);
323 vattr.va_mask |= XFS_AT_RDEV;
326 error = bhv_vop_create(dvp, dentry, &vattr, &vp, NULL);
329 error = bhv_vop_mkdir(dvp, dentry, &vattr, &vp, NULL);
336 if (unlikely(!error)) {
337 error = xfs_init_security(vp, dir);
339 xfs_cleanup_inode(dvp, vp, dentry, mode);
342 if (unlikely(default_acl)) {
344 error = _ACL_INHERIT(vp, &vattr, default_acl);
348 xfs_cleanup_inode(dvp, vp, dentry, mode);
350 _ACL_FREE(default_acl);
353 if (likely(!error)) {
355 ip = vn_to_inode(vp);
357 if (S_ISCHR(mode) || S_ISBLK(mode))
359 else if (S_ISDIR(mode))
360 xfs_validate_fields(ip, &vattr);
361 d_instantiate(dentry, ip);
362 xfs_validate_fields(dir, &vattr);
370 struct dentry *dentry,
372 struct nameidata *nd)
374 return xfs_vn_mknod(dir, dentry, mode, 0);
380 struct dentry *dentry,
383 return xfs_vn_mknod(dir, dentry, mode|S_IFDIR, 0);
386 STATIC struct dentry *
389 struct dentry *dentry,
390 struct nameidata *nd)
392 bhv_vnode_t *vp = vn_from_inode(dir), *cvp;
395 if (dentry->d_name.len >= MAXNAMELEN)
396 return ERR_PTR(-ENAMETOOLONG);
398 error = bhv_vop_lookup(vp, dentry, &cvp, 0, NULL, NULL);
399 if (unlikely(error)) {
400 if (unlikely(error != ENOENT))
401 return ERR_PTR(-error);
406 return d_splice_alias(vn_to_inode(cvp), dentry);
411 struct dentry *old_dentry,
413 struct dentry *dentry)
415 struct inode *ip; /* inode of guy being linked to */
416 bhv_vnode_t *tdvp; /* target directory for new name/link */
417 bhv_vnode_t *vp; /* vp of name being linked */
421 ip = old_dentry->d_inode; /* inode being linked to */
422 tdvp = vn_from_inode(dir);
423 vp = vn_from_inode(ip);
426 error = bhv_vop_link(tdvp, vp, dentry, NULL);
427 if (unlikely(error)) {
431 xfs_validate_fields(ip, &vattr);
432 d_instantiate(dentry, ip);
440 struct dentry *dentry)
443 bhv_vnode_t *dvp; /* directory containing name to remove */
447 inode = dentry->d_inode;
448 dvp = vn_from_inode(dir);
450 error = bhv_vop_remove(dvp, dentry, NULL);
451 if (likely(!error)) {
452 xfs_validate_fields(dir, &vattr); /* size needs update */
453 xfs_validate_fields(inode, &vattr);
461 struct dentry *dentry,
465 bhv_vattr_t va = { 0 };
466 bhv_vnode_t *dvp; /* directory containing name of symlink */
467 bhv_vnode_t *cvp; /* used to lookup symlink to put in dentry */
470 dvp = vn_from_inode(dir);
473 va.va_mode = S_IFLNK |
474 (irix_symlink_mode ? 0777 & ~current->fs->umask : S_IRWXUGO);
475 va.va_mask = XFS_AT_TYPE|XFS_AT_MODE;
477 error = bhv_vop_symlink(dvp, dentry, &va, (char *)symname, &cvp, NULL);
478 if (likely(!error && cvp)) {
479 error = xfs_init_security(cvp, dir);
480 if (likely(!error)) {
481 ip = vn_to_inode(cvp);
482 d_instantiate(dentry, ip);
483 xfs_validate_fields(dir, &va);
484 xfs_validate_fields(ip, &va);
486 xfs_cleanup_inode(dvp, cvp, dentry, 0);
495 struct dentry *dentry)
497 struct inode *inode = dentry->d_inode;
498 bhv_vnode_t *dvp = vn_from_inode(dir);
502 error = bhv_vop_rmdir(dvp, dentry, NULL);
503 if (likely(!error)) {
504 xfs_validate_fields(inode, &vattr);
505 xfs_validate_fields(dir, &vattr);
513 struct dentry *odentry,
515 struct dentry *ndentry)
517 struct inode *new_inode = ndentry->d_inode;
518 bhv_vnode_t *fvp; /* from directory */
519 bhv_vnode_t *tvp; /* target directory */
523 fvp = vn_from_inode(odir);
524 tvp = vn_from_inode(ndir);
526 error = bhv_vop_rename(fvp, odentry, tvp, ndentry, NULL);
527 if (likely(!error)) {
529 xfs_validate_fields(new_inode, &vattr);
530 xfs_validate_fields(odir, &vattr);
532 xfs_validate_fields(ndir, &vattr);
538 * careful here - this function can get called recursively, so
539 * we need to be very careful about how much stack we use.
540 * uio is kmalloced for this reason...
544 struct dentry *dentry,
545 struct nameidata *nd)
556 link = kmalloc(MAXPATHLEN+1, GFP_KERNEL);
558 nd_set_link(nd, ERR_PTR(-ENOMEM));
562 uio = kmalloc(sizeof(uio_t), GFP_KERNEL);
565 nd_set_link(nd, ERR_PTR(-ENOMEM));
569 vp = vn_from_inode(dentry->d_inode);
572 iov.iov_len = MAXPATHLEN;
576 uio->uio_segflg = UIO_SYSSPACE;
577 uio->uio_resid = MAXPATHLEN;
580 error = bhv_vop_readlink(vp, uio, 0, NULL);
581 if (unlikely(error)) {
583 link = ERR_PTR(-error);
585 link[MAXPATHLEN - uio->uio_resid] = '\0';
589 nd_set_link(nd, link);
595 struct dentry *dentry,
596 struct nameidata *nd,
599 char *s = nd_get_link(nd);
605 #ifdef CONFIG_XFS_POSIX_ACL
610 struct nameidata *nd)
612 return -bhv_vop_access(vn_from_inode(inode), mode << 6, NULL);
615 #define xfs_vn_permission NULL
620 struct vfsmount *mnt,
621 struct dentry *dentry,
624 struct inode *inode = dentry->d_inode;
625 bhv_vnode_t *vp = vn_from_inode(inode);
626 bhv_vattr_t vattr = { .va_mask = XFS_AT_STAT };
629 error = bhv_vop_getattr(vp, &vattr, ATTR_LAZY, NULL);
630 if (likely(!error)) {
631 stat->size = i_size_read(inode);
632 stat->dev = inode->i_sb->s_dev;
633 stat->rdev = (vattr.va_rdev == 0) ? 0 :
634 MKDEV(sysv_major(vattr.va_rdev) & 0x1ff,
635 sysv_minor(vattr.va_rdev));
636 stat->mode = vattr.va_mode;
637 stat->nlink = vattr.va_nlink;
638 stat->uid = vattr.va_uid;
639 stat->gid = vattr.va_gid;
640 stat->ino = vattr.va_nodeid;
641 stat->atime = vattr.va_atime;
642 stat->mtime = vattr.va_mtime;
643 stat->ctime = vattr.va_ctime;
644 stat->blocks = vattr.va_nblocks;
645 stat->blksize = vattr.va_blocksize;
652 struct dentry *dentry,
655 struct inode *inode = dentry->d_inode;
656 unsigned int ia_valid = attr->ia_valid;
657 bhv_vnode_t *vp = vn_from_inode(inode);
658 bhv_vattr_t vattr = { 0 };
662 if (ia_valid & ATTR_UID) {
663 vattr.va_mask |= XFS_AT_UID;
664 vattr.va_uid = attr->ia_uid;
666 if (ia_valid & ATTR_GID) {
667 vattr.va_mask |= XFS_AT_GID;
668 vattr.va_gid = attr->ia_gid;
670 if (ia_valid & ATTR_SIZE) {
671 vattr.va_mask |= XFS_AT_SIZE;
672 vattr.va_size = attr->ia_size;
674 if (ia_valid & ATTR_ATIME) {
675 vattr.va_mask |= XFS_AT_ATIME;
676 vattr.va_atime = attr->ia_atime;
677 inode->i_atime = attr->ia_atime;
679 if (ia_valid & ATTR_MTIME) {
680 vattr.va_mask |= XFS_AT_MTIME;
681 vattr.va_mtime = attr->ia_mtime;
683 if (ia_valid & ATTR_CTIME) {
684 vattr.va_mask |= XFS_AT_CTIME;
685 vattr.va_ctime = attr->ia_ctime;
687 if (ia_valid & ATTR_MODE) {
688 vattr.va_mask |= XFS_AT_MODE;
689 vattr.va_mode = attr->ia_mode;
690 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
691 inode->i_mode &= ~S_ISGID;
694 if (ia_valid & (ATTR_MTIME_SET | ATTR_ATIME_SET))
697 if ((ia_valid & ATTR_NO_BLOCK))
698 flags |= ATTR_NONBLOCK;
701 error = bhv_vop_setattr(vp, &vattr, flags, NULL);
703 __vn_revalidate(vp, &vattr);
711 block_truncate_page(inode->i_mapping, inode->i_size, xfs_get_blocks);
716 struct dentry *dentry,
722 bhv_vnode_t *vp = vn_from_inode(dentry->d_inode);
723 char *attr = (char *)name;
728 namesp = attr_lookup_namespace(attr, attr_namespaces, ATTR_NAMECOUNT);
731 attr += namesp->attr_namelen;
732 error = namesp->attr_capable(vp, NULL);
736 /* Convert Linux syscall to XFS internal ATTR flags */
737 if (flags & XATTR_CREATE)
738 xflags |= ATTR_CREATE;
739 if (flags & XATTR_REPLACE)
740 xflags |= ATTR_REPLACE;
741 xflags |= namesp->attr_flag;
742 return namesp->attr_set(vp, attr, (void *)data, size, xflags);
747 struct dentry *dentry,
752 bhv_vnode_t *vp = vn_from_inode(dentry->d_inode);
753 char *attr = (char *)name;
758 namesp = attr_lookup_namespace(attr, attr_namespaces, ATTR_NAMECOUNT);
761 attr += namesp->attr_namelen;
762 error = namesp->attr_capable(vp, NULL);
766 /* Convert Linux syscall to XFS internal ATTR flags */
768 xflags |= ATTR_KERNOVAL;
771 xflags |= namesp->attr_flag;
772 return namesp->attr_get(vp, attr, (void *)data, size, xflags);
777 struct dentry *dentry,
781 bhv_vnode_t *vp = vn_from_inode(dentry->d_inode);
782 int error, xflags = ATTR_KERNAMELS;
786 xflags |= ATTR_KERNOVAL;
787 xflags |= capable(CAP_SYS_ADMIN) ? ATTR_KERNFULLS : ATTR_KERNORMALS;
789 error = attr_generic_list(vp, data, size, xflags, &result);
797 struct dentry *dentry,
800 bhv_vnode_t *vp = vn_from_inode(dentry->d_inode);
801 char *attr = (char *)name;
806 namesp = attr_lookup_namespace(attr, attr_namespaces, ATTR_NAMECOUNT);
809 attr += namesp->attr_namelen;
810 error = namesp->attr_capable(vp, NULL);
813 xflags |= namesp->attr_flag;
814 return namesp->attr_remove(vp, attr, xflags);
818 struct inode_operations xfs_inode_operations = {
819 .permission = xfs_vn_permission,
820 .truncate = xfs_vn_truncate,
821 .getattr = xfs_vn_getattr,
822 .setattr = xfs_vn_setattr,
823 .setxattr = xfs_vn_setxattr,
824 .getxattr = xfs_vn_getxattr,
825 .listxattr = xfs_vn_listxattr,
826 .removexattr = xfs_vn_removexattr,
829 struct inode_operations xfs_dir_inode_operations = {
830 .create = xfs_vn_create,
831 .lookup = xfs_vn_lookup,
833 .unlink = xfs_vn_unlink,
834 .symlink = xfs_vn_symlink,
835 .mkdir = xfs_vn_mkdir,
836 .rmdir = xfs_vn_rmdir,
837 .mknod = xfs_vn_mknod,
838 .rename = xfs_vn_rename,
839 .permission = xfs_vn_permission,
840 .getattr = xfs_vn_getattr,
841 .setattr = xfs_vn_setattr,
842 .setxattr = xfs_vn_setxattr,
843 .getxattr = xfs_vn_getxattr,
844 .listxattr = xfs_vn_listxattr,
845 .removexattr = xfs_vn_removexattr,
848 struct inode_operations xfs_symlink_inode_operations = {
849 .readlink = generic_readlink,
850 .follow_link = xfs_vn_follow_link,
851 .put_link = xfs_vn_put_link,
852 .permission = xfs_vn_permission,
853 .getattr = xfs_vn_getattr,
854 .setattr = xfs_vn_setattr,
855 .setxattr = xfs_vn_setxattr,
856 .getxattr = xfs_vn_getxattr,
857 .listxattr = xfs_vn_listxattr,
858 .removexattr = xfs_vn_removexattr,