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
24 #include "xfs_trans.h"
28 #include "xfs_alloc.h"
29 #include "xfs_dmapi.h"
30 #include "xfs_quota.h"
31 #include "xfs_mount.h"
32 #include "xfs_bmap_btree.h"
33 #include "xfs_alloc_btree.h"
34 #include "xfs_ialloc_btree.h"
35 #include "xfs_dir2_sf.h"
36 #include "xfs_attr_sf.h"
37 #include "xfs_dinode.h"
38 #include "xfs_inode.h"
40 #include "xfs_btree.h"
41 #include "xfs_ialloc.h"
42 #include "xfs_rtalloc.h"
43 #include "xfs_error.h"
44 #include "xfs_itable.h"
48 #include "xfs_buf_item.h"
49 #include "xfs_utils.h"
50 #include "xfs_vnodeops.h"
52 #include <linux/capability.h>
53 #include <linux/xattr.h>
54 #include <linux/namei.h>
55 #include <linux/posix_acl.h>
56 #include <linux/security.h>
57 #include <linux/falloc.h>
58 #include <linux/fiemap.h>
61 * Bring the atime in the XFS inode uptodate.
62 * Used before logging the inode to disk or when the Linux inode goes away.
65 xfs_synchronize_atime(
68 struct inode *inode = VFS_I(ip);
70 if (!(inode->i_state & I_CLEAR)) {
71 ip->i_d.di_atime.t_sec = (__int32_t)inode->i_atime.tv_sec;
72 ip->i_d.di_atime.t_nsec = (__int32_t)inode->i_atime.tv_nsec;
77 * If the linux inode is valid, mark it dirty.
78 * Used when commiting a dirty inode into a transaction so that
79 * the inode will get written back by the linux code
82 xfs_mark_inode_dirty_sync(
85 struct inode *inode = VFS_I(ip);
87 if (!(inode->i_state & (I_WILL_FREE|I_FREEING|I_CLEAR)))
88 mark_inode_dirty_sync(inode);
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.
101 struct inode *inode = VFS_I(ip);
105 tv = current_fs_time(inode->i_sb);
107 if ((flags & XFS_ICHGTIME_MOD) &&
108 !timespec_equal(&inode->i_mtime, &tv)) {
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;
114 if ((flags & XFS_ICHGTIME_CHG) &&
115 !timespec_equal(&inode->i_ctime, &tv)) {
117 ip->i_d.di_ctime.t_sec = (__int32_t)tv.tv_sec;
118 ip->i_d.di_ctime.t_nsec = (__int32_t)tv.tv_nsec;
123 * We update the i_update_core field _after_ changing
124 * the timestamps in order to coordinate properly with
125 * xfs_iflush() so that we don't lose timestamp updates.
126 * This keeps us from having to hold the inode lock
127 * while doing this. We use the SYNCHRONIZE macro to
128 * ensure that the compiler does not reorder the update
129 * of i_update_core above the timestamp updates above.
133 ip->i_update_core = 1;
134 xfs_mark_inode_dirty_sync(ip);
139 * Hook in SELinux. This is not quite correct yet, what we really need
140 * here (as we do for default ACLs) is a mechanism by which creation of
141 * these attrs can be journalled at inode creation time (along with the
142 * inode, of course, such that log replay can't cause these to be lost).
149 struct xfs_inode *ip = XFS_I(inode);
155 error = security_inode_init_security(inode, dir, &name,
158 if (error == -EOPNOTSUPP)
163 error = xfs_attr_set(ip, name, value, length, ATTR_SECURE);
172 struct xfs_name *namep,
173 struct dentry *dentry)
175 namep->name = dentry->d_name.name;
176 namep->len = dentry->d_name.len;
183 struct dentry *dentry)
185 struct xfs_name teardown;
188 * If we can't add the ACL or we fail in
189 * xfs_init_security we must back out.
190 * ENOSPC can hit here, among other things.
192 xfs_dentry_to_name(&teardown, dentry);
194 xfs_remove(XFS_I(dir), &teardown, XFS_I(inode));
201 struct dentry *dentry,
206 struct xfs_inode *ip = NULL;
207 struct posix_acl *default_acl = NULL;
208 struct xfs_name name;
212 * Irix uses Missed'em'V split, but doesn't want to see
213 * the upper 5 bits of (14bit) major.
215 if (S_ISCHR(mode) || S_ISBLK(mode)) {
216 if (unlikely(!sysv_valid_dev(rdev) || MAJOR(rdev) & ~0x1ff))
218 rdev = sysv_encode_dev(rdev);
223 if (IS_POSIXACL(dir)) {
224 default_acl = xfs_get_acl(dir, ACL_TYPE_DEFAULT);
225 if (IS_ERR(default_acl))
226 return -PTR_ERR(default_acl);
229 mode &= ~current_umask();
232 xfs_dentry_to_name(&name, dentry);
233 error = xfs_create(XFS_I(dir), &name, mode, rdev, &ip, NULL);
239 error = xfs_init_security(inode, dir);
241 goto out_cleanup_inode;
244 error = -xfs_inherit_acl(inode, default_acl);
246 goto out_cleanup_inode;
247 posix_acl_release(default_acl);
251 d_instantiate(dentry, inode);
255 xfs_cleanup_inode(dir, inode, dentry);
257 posix_acl_release(default_acl);
264 struct dentry *dentry,
266 struct nameidata *nd)
268 return xfs_vn_mknod(dir, dentry, mode, 0);
274 struct dentry *dentry,
277 return xfs_vn_mknod(dir, dentry, mode|S_IFDIR, 0);
280 STATIC struct dentry *
283 struct dentry *dentry,
284 struct nameidata *nd)
286 struct xfs_inode *cip;
287 struct xfs_name name;
290 if (dentry->d_name.len >= MAXNAMELEN)
291 return ERR_PTR(-ENAMETOOLONG);
293 xfs_dentry_to_name(&name, dentry);
294 error = xfs_lookup(XFS_I(dir), &name, &cip, NULL);
295 if (unlikely(error)) {
296 if (unlikely(error != ENOENT))
297 return ERR_PTR(-error);
302 return d_splice_alias(VFS_I(cip), dentry);
305 STATIC struct dentry *
308 struct dentry *dentry,
309 struct nameidata *nd)
311 struct xfs_inode *ip;
312 struct xfs_name xname;
313 struct xfs_name ci_name;
317 if (dentry->d_name.len >= MAXNAMELEN)
318 return ERR_PTR(-ENAMETOOLONG);
320 xfs_dentry_to_name(&xname, dentry);
321 error = xfs_lookup(XFS_I(dir), &xname, &ip, &ci_name);
322 if (unlikely(error)) {
323 if (unlikely(error != ENOENT))
324 return ERR_PTR(-error);
326 * call d_add(dentry, NULL) here when d_drop_negative_children
327 * is called in xfs_vn_mknod (ie. allow negative dentries
328 * with CI filesystems).
333 /* if exact match, just splice and exit */
335 return d_splice_alias(VFS_I(ip), dentry);
337 /* else case-insensitive match... */
338 dname.name = ci_name.name;
339 dname.len = ci_name.len;
340 dentry = d_add_ci(dentry, VFS_I(ip), &dname);
341 kmem_free(ci_name.name);
347 struct dentry *old_dentry,
349 struct dentry *dentry)
351 struct inode *inode = old_dentry->d_inode;
352 struct xfs_name name;
355 xfs_dentry_to_name(&name, dentry);
357 error = xfs_link(XFS_I(dir), XFS_I(inode), &name);
361 atomic_inc(&inode->i_count);
362 d_instantiate(dentry, inode);
369 struct dentry *dentry)
371 struct xfs_name name;
374 xfs_dentry_to_name(&name, dentry);
376 error = -xfs_remove(XFS_I(dir), &name, XFS_I(dentry->d_inode));
381 * With unlink, the VFS makes the dentry "negative": no inode,
382 * but still hashed. This is incompatible with case-insensitive
383 * mode, so invalidate (unhash) the dentry in CI-mode.
385 if (xfs_sb_version_hasasciici(&XFS_M(dir->i_sb)->m_sb))
386 d_invalidate(dentry);
393 struct dentry *dentry,
397 struct xfs_inode *cip = NULL;
398 struct xfs_name name;
403 (irix_symlink_mode ? 0777 & ~current_umask() : S_IRWXUGO);
404 xfs_dentry_to_name(&name, dentry);
406 error = xfs_symlink(XFS_I(dir), &name, symname, mode, &cip, NULL);
412 error = xfs_init_security(inode, dir);
414 goto out_cleanup_inode;
416 d_instantiate(dentry, inode);
420 xfs_cleanup_inode(dir, inode, dentry);
428 struct dentry *odentry,
430 struct dentry *ndentry)
432 struct inode *new_inode = ndentry->d_inode;
433 struct xfs_name oname;
434 struct xfs_name nname;
436 xfs_dentry_to_name(&oname, odentry);
437 xfs_dentry_to_name(&nname, ndentry);
439 return -xfs_rename(XFS_I(odir), &oname, XFS_I(odentry->d_inode),
440 XFS_I(ndir), &nname, new_inode ?
441 XFS_I(new_inode) : NULL);
445 * careful here - this function can get called recursively, so
446 * we need to be very careful about how much stack we use.
447 * uio is kmalloced for this reason...
451 struct dentry *dentry,
452 struct nameidata *nd)
457 link = kmalloc(MAXPATHLEN+1, GFP_KERNEL);
461 error = -xfs_readlink(XFS_I(dentry->d_inode), link);
465 nd_set_link(nd, link);
471 nd_set_link(nd, ERR_PTR(error));
477 struct dentry *dentry,
478 struct nameidata *nd,
481 char *s = nd_get_link(nd);
492 return generic_permission(inode, mask, xfs_check_acl);
497 struct vfsmount *mnt,
498 struct dentry *dentry,
501 struct inode *inode = dentry->d_inode;
502 struct xfs_inode *ip = XFS_I(inode);
503 struct xfs_mount *mp = ip->i_mount;
505 xfs_itrace_entry(ip);
507 if (XFS_FORCED_SHUTDOWN(mp))
508 return XFS_ERROR(EIO);
510 stat->size = XFS_ISIZE(ip);
511 stat->dev = inode->i_sb->s_dev;
512 stat->mode = ip->i_d.di_mode;
513 stat->nlink = ip->i_d.di_nlink;
514 stat->uid = ip->i_d.di_uid;
515 stat->gid = ip->i_d.di_gid;
516 stat->ino = ip->i_ino;
517 stat->atime = inode->i_atime;
518 stat->mtime.tv_sec = ip->i_d.di_mtime.t_sec;
519 stat->mtime.tv_nsec = ip->i_d.di_mtime.t_nsec;
520 stat->ctime.tv_sec = ip->i_d.di_ctime.t_sec;
521 stat->ctime.tv_nsec = ip->i_d.di_ctime.t_nsec;
523 XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks);
526 switch (inode->i_mode & S_IFMT) {
529 stat->blksize = BLKDEV_IOSIZE;
530 stat->rdev = MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
531 sysv_minor(ip->i_df.if_u2.if_rdev));
534 if (XFS_IS_REALTIME_INODE(ip)) {
536 * If the file blocks are being allocated from a
537 * realtime volume, then return the inode's realtime
538 * extent size or the realtime volume's extent size.
541 xfs_get_extsz_hint(ip) << mp->m_sb.sb_blocklog;
543 stat->blksize = xfs_preferred_iosize(mp);
553 struct dentry *dentry,
556 return -xfs_setattr(XFS_I(dentry->d_inode), iattr, 0);
560 * block_truncate_page can return an error, but we can't propagate it
561 * at all here. Leave a complaint + stack trace in the syslog because
562 * this could be bad. If it is bad, we need to propagate the error further.
569 error = block_truncate_page(inode->i_mapping, inode->i_size,
584 xfs_inode_t *ip = XFS_I(inode);
586 /* preallocation on directories not yet supported */
588 if (S_ISDIR(inode->i_mode))
595 xfs_ilock(ip, XFS_IOLOCK_EXCL);
596 error = xfs_change_file_space(ip, XFS_IOC_RESVSP, &bf,
598 if (!error && !(mode & FALLOC_FL_KEEP_SIZE) &&
599 offset + len > i_size_read(inode))
600 new_size = offset + len;
602 /* Change file size if needed */
606 iattr.ia_valid = ATTR_SIZE;
607 iattr.ia_size = new_size;
608 error = xfs_setattr(ip, &iattr, XFS_ATTR_NOLOCK);
611 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
616 #define XFS_FIEMAP_FLAGS (FIEMAP_FLAG_SYNC|FIEMAP_FLAG_XATTR)
619 * Call fiemap helper to fill in user data.
620 * Returns positive errors to xfs_getbmap.
625 struct getbmapx *bmv,
629 struct fiemap_extent_info *fieinfo = *arg;
630 u32 fiemap_flags = 0;
631 u64 logical, physical, length;
633 /* Do nothing for a hole */
634 if (bmv->bmv_block == -1LL)
637 logical = BBTOB(bmv->bmv_offset);
638 physical = BBTOB(bmv->bmv_block);
639 length = BBTOB(bmv->bmv_length);
641 if (bmv->bmv_oflags & BMV_OF_PREALLOC)
642 fiemap_flags |= FIEMAP_EXTENT_UNWRITTEN;
643 else if (bmv->bmv_oflags & BMV_OF_DELALLOC) {
644 fiemap_flags |= FIEMAP_EXTENT_DELALLOC;
645 physical = 0; /* no block yet */
647 if (bmv->bmv_oflags & BMV_OF_LAST)
648 fiemap_flags |= FIEMAP_EXTENT_LAST;
650 error = fiemap_fill_next_extent(fieinfo, logical, physical,
651 length, fiemap_flags);
654 *full = 1; /* user array now full */
663 struct fiemap_extent_info *fieinfo,
667 xfs_inode_t *ip = XFS_I(inode);
671 error = fiemap_check_flags(fieinfo, XFS_FIEMAP_FLAGS);
675 /* Set up bmap header for xfs internal routine */
676 bm.bmv_offset = BTOBB(start);
677 /* Special case for whole file */
678 if (length == FIEMAP_MAX_OFFSET)
679 bm.bmv_length = -1LL;
681 bm.bmv_length = BTOBB(length);
683 /* our formatter will tell xfs_getbmap when to stop. */
684 bm.bmv_count = MAXEXTNUM;
685 bm.bmv_iflags = BMV_IF_PREALLOC;
686 if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR)
687 bm.bmv_iflags |= BMV_IF_ATTRFORK;
688 if (!(fieinfo->fi_flags & FIEMAP_FLAG_SYNC))
689 bm.bmv_iflags |= BMV_IF_DELALLOC;
691 error = xfs_getbmap(ip, &bm, xfs_fiemap_format, fieinfo);
698 static const struct inode_operations xfs_inode_operations = {
699 .permission = xfs_vn_permission,
700 .truncate = xfs_vn_truncate,
701 .getattr = xfs_vn_getattr,
702 .setattr = xfs_vn_setattr,
703 .setxattr = generic_setxattr,
704 .getxattr = generic_getxattr,
705 .removexattr = generic_removexattr,
706 .listxattr = xfs_vn_listxattr,
707 .fallocate = xfs_vn_fallocate,
708 .fiemap = xfs_vn_fiemap,
711 static const struct inode_operations xfs_dir_inode_operations = {
712 .create = xfs_vn_create,
713 .lookup = xfs_vn_lookup,
715 .unlink = xfs_vn_unlink,
716 .symlink = xfs_vn_symlink,
717 .mkdir = xfs_vn_mkdir,
719 * Yes, XFS uses the same method for rmdir and unlink.
721 * There are some subtile differences deeper in the code,
722 * but we use S_ISDIR to check for those.
724 .rmdir = xfs_vn_unlink,
725 .mknod = xfs_vn_mknod,
726 .rename = xfs_vn_rename,
727 .permission = xfs_vn_permission,
728 .getattr = xfs_vn_getattr,
729 .setattr = xfs_vn_setattr,
730 .setxattr = generic_setxattr,
731 .getxattr = generic_getxattr,
732 .removexattr = generic_removexattr,
733 .listxattr = xfs_vn_listxattr,
736 static const struct inode_operations xfs_dir_ci_inode_operations = {
737 .create = xfs_vn_create,
738 .lookup = xfs_vn_ci_lookup,
740 .unlink = xfs_vn_unlink,
741 .symlink = xfs_vn_symlink,
742 .mkdir = xfs_vn_mkdir,
744 * Yes, XFS uses the same method for rmdir and unlink.
746 * There are some subtile differences deeper in the code,
747 * but we use S_ISDIR to check for those.
749 .rmdir = xfs_vn_unlink,
750 .mknod = xfs_vn_mknod,
751 .rename = xfs_vn_rename,
752 .permission = xfs_vn_permission,
753 .getattr = xfs_vn_getattr,
754 .setattr = xfs_vn_setattr,
755 .setxattr = generic_setxattr,
756 .getxattr = generic_getxattr,
757 .removexattr = generic_removexattr,
758 .listxattr = xfs_vn_listxattr,
761 static const struct inode_operations xfs_symlink_inode_operations = {
762 .readlink = generic_readlink,
763 .follow_link = xfs_vn_follow_link,
764 .put_link = xfs_vn_put_link,
765 .permission = xfs_vn_permission,
766 .getattr = xfs_vn_getattr,
767 .setattr = xfs_vn_setattr,
768 .setxattr = generic_setxattr,
769 .getxattr = generic_getxattr,
770 .removexattr = generic_removexattr,
771 .listxattr = xfs_vn_listxattr,
775 xfs_diflags_to_iflags(
777 struct xfs_inode *ip)
779 if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE)
780 inode->i_flags |= S_IMMUTABLE;
782 inode->i_flags &= ~S_IMMUTABLE;
783 if (ip->i_d.di_flags & XFS_DIFLAG_APPEND)
784 inode->i_flags |= S_APPEND;
786 inode->i_flags &= ~S_APPEND;
787 if (ip->i_d.di_flags & XFS_DIFLAG_SYNC)
788 inode->i_flags |= S_SYNC;
790 inode->i_flags &= ~S_SYNC;
791 if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME)
792 inode->i_flags |= S_NOATIME;
794 inode->i_flags &= ~S_NOATIME;
798 * Initialize the Linux inode, set up the operation vectors and
801 * When reading existing inodes from disk this is called directly
802 * from xfs_iget, when creating a new inode it is called from
803 * xfs_ialloc after setting up the inode.
805 * We are always called with an uninitialised linux inode here.
806 * We need to initialise the necessary fields and take a reference
811 struct xfs_inode *ip)
813 struct inode *inode = &ip->i_vnode;
815 inode->i_ino = ip->i_ino;
816 inode->i_state = I_NEW|I_LOCK;
817 inode_add_to_lists(ip->i_mount->m_super, inode);
819 inode->i_mode = ip->i_d.di_mode;
820 inode->i_nlink = ip->i_d.di_nlink;
821 inode->i_uid = ip->i_d.di_uid;
822 inode->i_gid = ip->i_d.di_gid;
824 switch (inode->i_mode & S_IFMT) {
828 MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
829 sysv_minor(ip->i_df.if_u2.if_rdev));
836 inode->i_generation = ip->i_d.di_gen;
837 i_size_write(inode, ip->i_d.di_size);
838 inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec;
839 inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec;
840 inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec;
841 inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec;
842 inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec;
843 inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec;
844 xfs_diflags_to_iflags(inode, ip);
846 switch (inode->i_mode & S_IFMT) {
848 inode->i_op = &xfs_inode_operations;
849 inode->i_fop = &xfs_file_operations;
850 inode->i_mapping->a_ops = &xfs_address_space_operations;
853 if (xfs_sb_version_hasasciici(&XFS_M(inode->i_sb)->m_sb))
854 inode->i_op = &xfs_dir_ci_inode_operations;
856 inode->i_op = &xfs_dir_inode_operations;
857 inode->i_fop = &xfs_dir_file_operations;
860 inode->i_op = &xfs_symlink_inode_operations;
861 if (!(ip->i_df.if_flags & XFS_IFINLINE))
862 inode->i_mapping->a_ops = &xfs_address_space_operations;
865 inode->i_op = &xfs_inode_operations;
866 init_special_inode(inode, inode->i_mode, inode->i_rdev);
870 xfs_iflags_clear(ip, XFS_INEW);
873 unlock_new_inode(inode);