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"
27 #include "xfs_alloc.h"
28 #include "xfs_quota.h"
29 #include "xfs_mount.h"
30 #include "xfs_bmap_btree.h"
31 #include "xfs_dinode.h"
32 #include "xfs_inode.h"
34 #include "xfs_rtalloc.h"
35 #include "xfs_error.h"
36 #include "xfs_itable.h"
39 #include "xfs_buf_item.h"
40 #include "xfs_utils.h"
41 #include "xfs_vnodeops.h"
42 #include "xfs_trace.h"
44 #include <linux/capability.h>
45 #include <linux/xattr.h>
46 #include <linux/namei.h>
47 #include <linux/posix_acl.h>
48 #include <linux/security.h>
49 #include <linux/falloc.h>
50 #include <linux/fiemap.h>
51 #include <linux/slab.h>
54 * Bring the timestamps in the XFS inode uptodate.
56 * Used before writing the inode to disk.
59 xfs_synchronize_times(
62 struct inode *inode = VFS_I(ip);
64 ip->i_d.di_atime.t_sec = (__int32_t)inode->i_atime.tv_sec;
65 ip->i_d.di_atime.t_nsec = (__int32_t)inode->i_atime.tv_nsec;
66 ip->i_d.di_ctime.t_sec = (__int32_t)inode->i_ctime.tv_sec;
67 ip->i_d.di_ctime.t_nsec = (__int32_t)inode->i_ctime.tv_nsec;
68 ip->i_d.di_mtime.t_sec = (__int32_t)inode->i_mtime.tv_sec;
69 ip->i_d.di_mtime.t_nsec = (__int32_t)inode->i_mtime.tv_nsec;
73 * If the linux inode is valid, mark it dirty.
74 * Used when commiting a dirty inode into a transaction so that
75 * the inode will get written back by the linux code
78 xfs_mark_inode_dirty_sync(
81 struct inode *inode = VFS_I(ip);
83 if (!(inode->i_state & (I_WILL_FREE|I_FREEING)))
84 mark_inode_dirty_sync(inode);
91 struct inode *inode = VFS_I(ip);
93 if (!(inode->i_state & (I_WILL_FREE|I_FREEING)))
94 mark_inode_dirty(inode);
98 * Hook in SELinux. This is not quite correct yet, what we really need
99 * here (as we do for default ACLs) is a mechanism by which creation of
100 * these attrs can be journalled at inode creation time (along with the
101 * inode, of course, such that log replay can't cause these to be lost).
108 struct xfs_inode *ip = XFS_I(inode);
114 error = security_inode_init_security(inode, dir, (char **)&name,
117 if (error == -EOPNOTSUPP)
122 error = xfs_attr_set(ip, name, value, length, ATTR_SECURE);
131 struct xfs_name *namep,
132 struct dentry *dentry)
134 namep->name = dentry->d_name.name;
135 namep->len = dentry->d_name.len;
142 struct dentry *dentry)
144 struct xfs_name teardown;
147 * If we can't add the ACL or we fail in
148 * xfs_init_security we must back out.
149 * ENOSPC can hit here, among other things.
151 xfs_dentry_to_name(&teardown, dentry);
153 xfs_remove(XFS_I(dir), &teardown, XFS_I(inode));
160 struct dentry *dentry,
165 struct xfs_inode *ip = NULL;
166 struct posix_acl *default_acl = NULL;
167 struct xfs_name name;
171 * Irix uses Missed'em'V split, but doesn't want to see
172 * the upper 5 bits of (14bit) major.
174 if (S_ISCHR(mode) || S_ISBLK(mode)) {
175 if (unlikely(!sysv_valid_dev(rdev) || MAJOR(rdev) & ~0x1ff))
177 rdev = sysv_encode_dev(rdev);
182 if (IS_POSIXACL(dir)) {
183 default_acl = xfs_get_acl(dir, ACL_TYPE_DEFAULT);
184 if (IS_ERR(default_acl))
185 return -PTR_ERR(default_acl);
188 mode &= ~current_umask();
191 xfs_dentry_to_name(&name, dentry);
192 error = xfs_create(XFS_I(dir), &name, mode, rdev, &ip);
198 error = xfs_init_security(inode, dir);
200 goto out_cleanup_inode;
203 error = -xfs_inherit_acl(inode, default_acl);
205 goto out_cleanup_inode;
206 posix_acl_release(default_acl);
210 d_instantiate(dentry, inode);
214 xfs_cleanup_inode(dir, inode, dentry);
216 posix_acl_release(default_acl);
223 struct dentry *dentry,
225 struct nameidata *nd)
227 return xfs_vn_mknod(dir, dentry, mode, 0);
233 struct dentry *dentry,
236 return xfs_vn_mknod(dir, dentry, mode|S_IFDIR, 0);
239 STATIC struct dentry *
242 struct dentry *dentry,
243 struct nameidata *nd)
245 struct xfs_inode *cip;
246 struct xfs_name name;
249 if (dentry->d_name.len >= MAXNAMELEN)
250 return ERR_PTR(-ENAMETOOLONG);
252 xfs_dentry_to_name(&name, dentry);
253 error = xfs_lookup(XFS_I(dir), &name, &cip, NULL);
254 if (unlikely(error)) {
255 if (unlikely(error != ENOENT))
256 return ERR_PTR(-error);
261 return d_splice_alias(VFS_I(cip), dentry);
264 STATIC struct dentry *
267 struct dentry *dentry,
268 struct nameidata *nd)
270 struct xfs_inode *ip;
271 struct xfs_name xname;
272 struct xfs_name ci_name;
276 if (dentry->d_name.len >= MAXNAMELEN)
277 return ERR_PTR(-ENAMETOOLONG);
279 xfs_dentry_to_name(&xname, dentry);
280 error = xfs_lookup(XFS_I(dir), &xname, &ip, &ci_name);
281 if (unlikely(error)) {
282 if (unlikely(error != ENOENT))
283 return ERR_PTR(-error);
285 * call d_add(dentry, NULL) here when d_drop_negative_children
286 * is called in xfs_vn_mknod (ie. allow negative dentries
287 * with CI filesystems).
292 /* if exact match, just splice and exit */
294 return d_splice_alias(VFS_I(ip), dentry);
296 /* else case-insensitive match... */
297 dname.name = ci_name.name;
298 dname.len = ci_name.len;
299 dentry = d_add_ci(dentry, VFS_I(ip), &dname);
300 kmem_free(ci_name.name);
306 struct dentry *old_dentry,
308 struct dentry *dentry)
310 struct inode *inode = old_dentry->d_inode;
311 struct xfs_name name;
314 xfs_dentry_to_name(&name, dentry);
316 error = xfs_link(XFS_I(dir), XFS_I(inode), &name);
321 d_instantiate(dentry, inode);
328 struct dentry *dentry)
330 struct xfs_name name;
333 xfs_dentry_to_name(&name, dentry);
335 error = -xfs_remove(XFS_I(dir), &name, XFS_I(dentry->d_inode));
340 * With unlink, the VFS makes the dentry "negative": no inode,
341 * but still hashed. This is incompatible with case-insensitive
342 * mode, so invalidate (unhash) the dentry in CI-mode.
344 if (xfs_sb_version_hasasciici(&XFS_M(dir->i_sb)->m_sb))
345 d_invalidate(dentry);
352 struct dentry *dentry,
356 struct xfs_inode *cip = NULL;
357 struct xfs_name name;
362 (irix_symlink_mode ? 0777 & ~current_umask() : S_IRWXUGO);
363 xfs_dentry_to_name(&name, dentry);
365 error = xfs_symlink(XFS_I(dir), &name, symname, mode, &cip);
371 error = xfs_init_security(inode, dir);
373 goto out_cleanup_inode;
375 d_instantiate(dentry, inode);
379 xfs_cleanup_inode(dir, inode, dentry);
387 struct dentry *odentry,
389 struct dentry *ndentry)
391 struct inode *new_inode = ndentry->d_inode;
392 struct xfs_name oname;
393 struct xfs_name nname;
395 xfs_dentry_to_name(&oname, odentry);
396 xfs_dentry_to_name(&nname, ndentry);
398 return -xfs_rename(XFS_I(odir), &oname, XFS_I(odentry->d_inode),
399 XFS_I(ndir), &nname, new_inode ?
400 XFS_I(new_inode) : NULL);
404 * careful here - this function can get called recursively, so
405 * we need to be very careful about how much stack we use.
406 * uio is kmalloced for this reason...
410 struct dentry *dentry,
411 struct nameidata *nd)
416 link = kmalloc(MAXPATHLEN+1, GFP_KERNEL);
420 error = -xfs_readlink(XFS_I(dentry->d_inode), link);
424 nd_set_link(nd, link);
430 nd_set_link(nd, ERR_PTR(error));
436 struct dentry *dentry,
437 struct nameidata *nd,
440 char *s = nd_get_link(nd);
448 struct vfsmount *mnt,
449 struct dentry *dentry,
452 struct inode *inode = dentry->d_inode;
453 struct xfs_inode *ip = XFS_I(inode);
454 struct xfs_mount *mp = ip->i_mount;
456 trace_xfs_getattr(ip);
458 if (XFS_FORCED_SHUTDOWN(mp))
459 return XFS_ERROR(EIO);
461 stat->size = XFS_ISIZE(ip);
462 stat->dev = inode->i_sb->s_dev;
463 stat->mode = ip->i_d.di_mode;
464 stat->nlink = ip->i_d.di_nlink;
465 stat->uid = ip->i_d.di_uid;
466 stat->gid = ip->i_d.di_gid;
467 stat->ino = ip->i_ino;
468 stat->atime = inode->i_atime;
469 stat->mtime = inode->i_mtime;
470 stat->ctime = inode->i_ctime;
472 XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks);
475 switch (inode->i_mode & S_IFMT) {
478 stat->blksize = BLKDEV_IOSIZE;
479 stat->rdev = MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
480 sysv_minor(ip->i_df.if_u2.if_rdev));
483 if (XFS_IS_REALTIME_INODE(ip)) {
485 * If the file blocks are being allocated from a
486 * realtime volume, then return the inode's realtime
487 * extent size or the realtime volume's extent size.
490 xfs_get_extsz_hint(ip) << mp->m_sb.sb_blocklog;
492 stat->blksize = xfs_preferred_iosize(mp);
502 struct dentry *dentry,
505 return -xfs_setattr(XFS_I(dentry->d_inode), iattr, 0);
518 xfs_inode_t *ip = XFS_I(inode);
519 int cmd = XFS_IOC_RESVSP;
521 /* preallocation on directories not yet supported */
523 if (S_ISDIR(inode->i_mode))
530 xfs_ilock(ip, XFS_IOLOCK_EXCL);
532 if (mode & FALLOC_FL_PUNCH_HOLE)
533 cmd = XFS_IOC_UNRESVSP;
535 /* check the new inode size is valid before allocating */
536 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
537 offset + len > i_size_read(inode)) {
538 new_size = offset + len;
539 error = inode_newsize_ok(inode, new_size);
544 error = -xfs_change_file_space(ip, cmd, &bf, 0, XFS_ATTR_NOLOCK);
548 /* Change file size if needed */
552 iattr.ia_valid = ATTR_SIZE;
553 iattr.ia_size = new_size;
554 error = -xfs_setattr(ip, &iattr, XFS_ATTR_NOLOCK);
558 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
563 #define XFS_FIEMAP_FLAGS (FIEMAP_FLAG_SYNC|FIEMAP_FLAG_XATTR)
566 * Call fiemap helper to fill in user data.
567 * Returns positive errors to xfs_getbmap.
572 struct getbmapx *bmv,
576 struct fiemap_extent_info *fieinfo = *arg;
577 u32 fiemap_flags = 0;
578 u64 logical, physical, length;
580 /* Do nothing for a hole */
581 if (bmv->bmv_block == -1LL)
584 logical = BBTOB(bmv->bmv_offset);
585 physical = BBTOB(bmv->bmv_block);
586 length = BBTOB(bmv->bmv_length);
588 if (bmv->bmv_oflags & BMV_OF_PREALLOC)
589 fiemap_flags |= FIEMAP_EXTENT_UNWRITTEN;
590 else if (bmv->bmv_oflags & BMV_OF_DELALLOC) {
591 fiemap_flags |= FIEMAP_EXTENT_DELALLOC;
592 physical = 0; /* no block yet */
594 if (bmv->bmv_oflags & BMV_OF_LAST)
595 fiemap_flags |= FIEMAP_EXTENT_LAST;
597 error = fiemap_fill_next_extent(fieinfo, logical, physical,
598 length, fiemap_flags);
601 *full = 1; /* user array now full */
610 struct fiemap_extent_info *fieinfo,
614 xfs_inode_t *ip = XFS_I(inode);
618 error = fiemap_check_flags(fieinfo, XFS_FIEMAP_FLAGS);
622 /* Set up bmap header for xfs internal routine */
623 bm.bmv_offset = BTOBB(start);
624 /* Special case for whole file */
625 if (length == FIEMAP_MAX_OFFSET)
626 bm.bmv_length = -1LL;
628 bm.bmv_length = BTOBB(length);
630 /* We add one because in getbmap world count includes the header */
631 bm.bmv_count = !fieinfo->fi_extents_max ? MAXEXTNUM :
632 fieinfo->fi_extents_max + 1;
633 bm.bmv_count = min_t(__s32, bm.bmv_count,
634 (PAGE_SIZE * 16 / sizeof(struct getbmapx)));
635 bm.bmv_iflags = BMV_IF_PREALLOC | BMV_IF_NO_HOLES;
636 if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR)
637 bm.bmv_iflags |= BMV_IF_ATTRFORK;
638 if (!(fieinfo->fi_flags & FIEMAP_FLAG_SYNC))
639 bm.bmv_iflags |= BMV_IF_DELALLOC;
641 error = xfs_getbmap(ip, &bm, xfs_fiemap_format, fieinfo);
648 static const struct inode_operations xfs_inode_operations = {
649 .check_acl = xfs_check_acl,
650 .getattr = xfs_vn_getattr,
651 .setattr = xfs_vn_setattr,
652 .setxattr = generic_setxattr,
653 .getxattr = generic_getxattr,
654 .removexattr = generic_removexattr,
655 .listxattr = xfs_vn_listxattr,
656 .fallocate = xfs_vn_fallocate,
657 .fiemap = xfs_vn_fiemap,
660 static const struct inode_operations xfs_dir_inode_operations = {
661 .create = xfs_vn_create,
662 .lookup = xfs_vn_lookup,
664 .unlink = xfs_vn_unlink,
665 .symlink = xfs_vn_symlink,
666 .mkdir = xfs_vn_mkdir,
668 * Yes, XFS uses the same method for rmdir and unlink.
670 * There are some subtile differences deeper in the code,
671 * but we use S_ISDIR to check for those.
673 .rmdir = xfs_vn_unlink,
674 .mknod = xfs_vn_mknod,
675 .rename = xfs_vn_rename,
676 .check_acl = xfs_check_acl,
677 .getattr = xfs_vn_getattr,
678 .setattr = xfs_vn_setattr,
679 .setxattr = generic_setxattr,
680 .getxattr = generic_getxattr,
681 .removexattr = generic_removexattr,
682 .listxattr = xfs_vn_listxattr,
685 static const struct inode_operations xfs_dir_ci_inode_operations = {
686 .create = xfs_vn_create,
687 .lookup = xfs_vn_ci_lookup,
689 .unlink = xfs_vn_unlink,
690 .symlink = xfs_vn_symlink,
691 .mkdir = xfs_vn_mkdir,
693 * Yes, XFS uses the same method for rmdir and unlink.
695 * There are some subtile differences deeper in the code,
696 * but we use S_ISDIR to check for those.
698 .rmdir = xfs_vn_unlink,
699 .mknod = xfs_vn_mknod,
700 .rename = xfs_vn_rename,
701 .check_acl = xfs_check_acl,
702 .getattr = xfs_vn_getattr,
703 .setattr = xfs_vn_setattr,
704 .setxattr = generic_setxattr,
705 .getxattr = generic_getxattr,
706 .removexattr = generic_removexattr,
707 .listxattr = xfs_vn_listxattr,
710 static const struct inode_operations xfs_symlink_inode_operations = {
711 .readlink = generic_readlink,
712 .follow_link = xfs_vn_follow_link,
713 .put_link = xfs_vn_put_link,
714 .check_acl = xfs_check_acl,
715 .getattr = xfs_vn_getattr,
716 .setattr = xfs_vn_setattr,
717 .setxattr = generic_setxattr,
718 .getxattr = generic_getxattr,
719 .removexattr = generic_removexattr,
720 .listxattr = xfs_vn_listxattr,
724 xfs_diflags_to_iflags(
726 struct xfs_inode *ip)
728 if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE)
729 inode->i_flags |= S_IMMUTABLE;
731 inode->i_flags &= ~S_IMMUTABLE;
732 if (ip->i_d.di_flags & XFS_DIFLAG_APPEND)
733 inode->i_flags |= S_APPEND;
735 inode->i_flags &= ~S_APPEND;
736 if (ip->i_d.di_flags & XFS_DIFLAG_SYNC)
737 inode->i_flags |= S_SYNC;
739 inode->i_flags &= ~S_SYNC;
740 if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME)
741 inode->i_flags |= S_NOATIME;
743 inode->i_flags &= ~S_NOATIME;
747 * Initialize the Linux inode, set up the operation vectors and
750 * When reading existing inodes from disk this is called directly
751 * from xfs_iget, when creating a new inode it is called from
752 * xfs_ialloc after setting up the inode.
754 * We are always called with an uninitialised linux inode here.
755 * We need to initialise the necessary fields and take a reference
760 struct xfs_inode *ip)
762 struct inode *inode = &ip->i_vnode;
764 inode->i_ino = ip->i_ino;
765 inode->i_state = I_NEW;
767 inode_sb_list_add(inode);
768 /* make the inode look hashed for the writeback code */
769 hlist_add_fake(&inode->i_hash);
771 inode->i_mode = ip->i_d.di_mode;
772 inode->i_nlink = ip->i_d.di_nlink;
773 inode->i_uid = ip->i_d.di_uid;
774 inode->i_gid = ip->i_d.di_gid;
776 switch (inode->i_mode & S_IFMT) {
780 MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
781 sysv_minor(ip->i_df.if_u2.if_rdev));
788 inode->i_generation = ip->i_d.di_gen;
789 i_size_write(inode, ip->i_d.di_size);
790 inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec;
791 inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec;
792 inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec;
793 inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec;
794 inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec;
795 inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec;
796 xfs_diflags_to_iflags(inode, ip);
798 switch (inode->i_mode & S_IFMT) {
800 inode->i_op = &xfs_inode_operations;
801 inode->i_fop = &xfs_file_operations;
802 inode->i_mapping->a_ops = &xfs_address_space_operations;
805 if (xfs_sb_version_hasasciici(&XFS_M(inode->i_sb)->m_sb))
806 inode->i_op = &xfs_dir_ci_inode_operations;
808 inode->i_op = &xfs_dir_inode_operations;
809 inode->i_fop = &xfs_dir_file_operations;
812 inode->i_op = &xfs_symlink_inode_operations;
813 if (!(ip->i_df.if_flags & XFS_IFINLINE))
814 inode->i_mapping->a_ops = &xfs_address_space_operations;
817 inode->i_op = &xfs_inode_operations;
818 init_special_inode(inode, inode->i_mode, inode->i_rdev);
822 xfs_iflags_clear(ip, XFS_INEW);
825 unlock_new_inode(inode);