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"
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/posix_acl.h>
55 #include <linux/security.h>
56 #include <linux/falloc.h>
57 #include <linux/fiemap.h>
60 * Bring the atime in the XFS inode uptodate.
61 * Used before logging the inode to disk or when the Linux inode goes away.
64 xfs_synchronize_atime(
67 struct inode *inode = VFS_I(ip);
69 if (!(inode->i_state & I_CLEAR)) {
70 ip->i_d.di_atime.t_sec = (__int32_t)inode->i_atime.tv_sec;
71 ip->i_d.di_atime.t_nsec = (__int32_t)inode->i_atime.tv_nsec;
76 * If the linux inode is valid, mark it dirty.
77 * Used when commiting a dirty inode into a transaction so that
78 * the inode will get written back by the linux code
81 xfs_mark_inode_dirty_sync(
84 struct inode *inode = VFS_I(ip);
86 if (!(inode->i_state & (I_WILL_FREE|I_FREEING|I_CLEAR)))
87 mark_inode_dirty_sync(inode);
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.
100 struct inode *inode = VFS_I(ip);
104 tv = current_fs_time(inode->i_sb);
106 if ((flags & XFS_ICHGTIME_MOD) &&
107 !timespec_equal(&inode->i_mtime, &tv)) {
109 ip->i_d.di_mtime.t_sec = (__int32_t)tv.tv_sec;
110 ip->i_d.di_mtime.t_nsec = (__int32_t)tv.tv_nsec;
113 if ((flags & XFS_ICHGTIME_CHG) &&
114 !timespec_equal(&inode->i_ctime, &tv)) {
116 ip->i_d.di_ctime.t_sec = (__int32_t)tv.tv_sec;
117 ip->i_d.di_ctime.t_nsec = (__int32_t)tv.tv_nsec;
122 * We update the i_update_core field _after_ changing
123 * the timestamps in order to coordinate properly with
124 * xfs_iflush() so that we don't lose timestamp updates.
125 * This keeps us from having to hold the inode lock
126 * while doing this. We use the SYNCHRONIZE macro to
127 * ensure that the compiler does not reorder the update
128 * of i_update_core above the timestamp updates above.
132 ip->i_update_core = 1;
133 xfs_mark_inode_dirty_sync(ip);
138 * Hook in SELinux. This is not quite correct yet, what we really need
139 * here (as we do for default ACLs) is a mechanism by which creation of
140 * these attrs can be journalled at inode creation time (along with the
141 * inode, of course, such that log replay can't cause these to be lost).
148 struct xfs_inode *ip = XFS_I(inode);
154 error = security_inode_init_security(inode, dir, &name,
157 if (error == -EOPNOTSUPP)
162 error = xfs_attr_set(ip, name, value, length, ATTR_SECURE);
171 struct xfs_name *namep,
172 struct dentry *dentry)
174 namep->name = dentry->d_name.name;
175 namep->len = dentry->d_name.len;
182 struct dentry *dentry)
184 struct xfs_name teardown;
187 * If we can't add the ACL or we fail in
188 * xfs_init_security we must back out.
189 * ENOSPC can hit here, among other things.
191 xfs_dentry_to_name(&teardown, dentry);
193 xfs_remove(XFS_I(dir), &teardown, XFS_I(inode));
200 struct dentry *dentry,
205 struct xfs_inode *ip = NULL;
206 struct posix_acl *default_acl = NULL;
207 struct xfs_name name;
211 * Irix uses Missed'em'V split, but doesn't want to see
212 * the upper 5 bits of (14bit) major.
214 if (S_ISCHR(mode) || S_ISBLK(mode)) {
215 if (unlikely(!sysv_valid_dev(rdev) || MAJOR(rdev) & ~0x1ff))
217 rdev = sysv_encode_dev(rdev);
222 if (IS_POSIXACL(dir)) {
223 default_acl = xfs_get_acl(dir, ACL_TYPE_DEFAULT);
224 if (IS_ERR(default_acl))
225 return -PTR_ERR(default_acl);
228 mode &= ~current_umask();
231 xfs_dentry_to_name(&name, dentry);
232 error = xfs_create(XFS_I(dir), &name, mode, rdev, &ip, NULL);
238 error = xfs_init_security(inode, dir);
240 goto out_cleanup_inode;
243 error = -xfs_inherit_acl(inode, default_acl);
245 goto out_cleanup_inode;
246 posix_acl_release(default_acl);
250 d_instantiate(dentry, inode);
254 xfs_cleanup_inode(dir, inode, dentry);
256 posix_acl_release(default_acl);
263 struct dentry *dentry,
265 struct nameidata *nd)
267 return xfs_vn_mknod(dir, dentry, mode, 0);
273 struct dentry *dentry,
276 return xfs_vn_mknod(dir, dentry, mode|S_IFDIR, 0);
279 STATIC struct dentry *
282 struct dentry *dentry,
283 struct nameidata *nd)
285 struct xfs_inode *cip;
286 struct xfs_name name;
289 if (dentry->d_name.len >= MAXNAMELEN)
290 return ERR_PTR(-ENAMETOOLONG);
292 xfs_dentry_to_name(&name, dentry);
293 error = xfs_lookup(XFS_I(dir), &name, &cip, NULL);
294 if (unlikely(error)) {
295 if (unlikely(error != ENOENT))
296 return ERR_PTR(-error);
301 return d_splice_alias(VFS_I(cip), dentry);
304 STATIC struct dentry *
307 struct dentry *dentry,
308 struct nameidata *nd)
310 struct xfs_inode *ip;
311 struct xfs_name xname;
312 struct xfs_name ci_name;
316 if (dentry->d_name.len >= MAXNAMELEN)
317 return ERR_PTR(-ENAMETOOLONG);
319 xfs_dentry_to_name(&xname, dentry);
320 error = xfs_lookup(XFS_I(dir), &xname, &ip, &ci_name);
321 if (unlikely(error)) {
322 if (unlikely(error != ENOENT))
323 return ERR_PTR(-error);
325 * call d_add(dentry, NULL) here when d_drop_negative_children
326 * is called in xfs_vn_mknod (ie. allow negative dentries
327 * with CI filesystems).
332 /* if exact match, just splice and exit */
334 return d_splice_alias(VFS_I(ip), dentry);
336 /* else case-insensitive match... */
337 dname.name = ci_name.name;
338 dname.len = ci_name.len;
339 dentry = d_add_ci(dentry, VFS_I(ip), &dname);
340 kmem_free(ci_name.name);
346 struct dentry *old_dentry,
348 struct dentry *dentry)
350 struct inode *inode = old_dentry->d_inode;
351 struct xfs_name name;
354 xfs_dentry_to_name(&name, dentry);
356 error = xfs_link(XFS_I(dir), XFS_I(inode), &name);
360 atomic_inc(&inode->i_count);
361 d_instantiate(dentry, inode);
368 struct dentry *dentry)
370 struct xfs_name name;
373 xfs_dentry_to_name(&name, dentry);
375 error = -xfs_remove(XFS_I(dir), &name, XFS_I(dentry->d_inode));
380 * With unlink, the VFS makes the dentry "negative": no inode,
381 * but still hashed. This is incompatible with case-insensitive
382 * mode, so invalidate (unhash) the dentry in CI-mode.
384 if (xfs_sb_version_hasasciici(&XFS_M(dir->i_sb)->m_sb))
385 d_invalidate(dentry);
392 struct dentry *dentry,
396 struct xfs_inode *cip = NULL;
397 struct xfs_name name;
402 (irix_symlink_mode ? 0777 & ~current_umask() : S_IRWXUGO);
403 xfs_dentry_to_name(&name, dentry);
405 error = xfs_symlink(XFS_I(dir), &name, symname, mode, &cip, NULL);
411 error = xfs_init_security(inode, dir);
413 goto out_cleanup_inode;
415 d_instantiate(dentry, inode);
419 xfs_cleanup_inode(dir, inode, dentry);
427 struct dentry *odentry,
429 struct dentry *ndentry)
431 struct inode *new_inode = ndentry->d_inode;
432 struct xfs_name oname;
433 struct xfs_name nname;
435 xfs_dentry_to_name(&oname, odentry);
436 xfs_dentry_to_name(&nname, ndentry);
438 return -xfs_rename(XFS_I(odir), &oname, XFS_I(odentry->d_inode),
439 XFS_I(ndir), &nname, new_inode ?
440 XFS_I(new_inode) : NULL);
444 * careful here - this function can get called recursively, so
445 * we need to be very careful about how much stack we use.
446 * uio is kmalloced for this reason...
450 struct dentry *dentry,
451 struct nameidata *nd)
456 link = kmalloc(MAXPATHLEN+1, GFP_KERNEL);
460 error = -xfs_readlink(XFS_I(dentry->d_inode), link);
464 nd_set_link(nd, link);
470 nd_set_link(nd, ERR_PTR(error));
476 struct dentry *dentry,
477 struct nameidata *nd,
480 char *s = nd_get_link(nd);
488 struct vfsmount *mnt,
489 struct dentry *dentry,
492 struct inode *inode = dentry->d_inode;
493 struct xfs_inode *ip = XFS_I(inode);
494 struct xfs_mount *mp = ip->i_mount;
496 xfs_itrace_entry(ip);
498 if (XFS_FORCED_SHUTDOWN(mp))
499 return XFS_ERROR(EIO);
501 stat->size = XFS_ISIZE(ip);
502 stat->dev = inode->i_sb->s_dev;
503 stat->mode = ip->i_d.di_mode;
504 stat->nlink = ip->i_d.di_nlink;
505 stat->uid = ip->i_d.di_uid;
506 stat->gid = ip->i_d.di_gid;
507 stat->ino = ip->i_ino;
508 stat->atime = inode->i_atime;
509 stat->mtime.tv_sec = ip->i_d.di_mtime.t_sec;
510 stat->mtime.tv_nsec = ip->i_d.di_mtime.t_nsec;
511 stat->ctime.tv_sec = ip->i_d.di_ctime.t_sec;
512 stat->ctime.tv_nsec = ip->i_d.di_ctime.t_nsec;
514 XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks);
517 switch (inode->i_mode & S_IFMT) {
520 stat->blksize = BLKDEV_IOSIZE;
521 stat->rdev = MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
522 sysv_minor(ip->i_df.if_u2.if_rdev));
525 if (XFS_IS_REALTIME_INODE(ip)) {
527 * If the file blocks are being allocated from a
528 * realtime volume, then return the inode's realtime
529 * extent size or the realtime volume's extent size.
532 xfs_get_extsz_hint(ip) << mp->m_sb.sb_blocklog;
534 stat->blksize = xfs_preferred_iosize(mp);
544 struct dentry *dentry,
547 return -xfs_setattr(XFS_I(dentry->d_inode), iattr, 0);
551 * block_truncate_page can return an error, but we can't propagate it
552 * at all here. Leave a complaint + stack trace in the syslog because
553 * this could be bad. If it is bad, we need to propagate the error further.
560 error = block_truncate_page(inode->i_mapping, inode->i_size,
575 xfs_inode_t *ip = XFS_I(inode);
577 /* preallocation on directories not yet supported */
579 if (S_ISDIR(inode->i_mode))
586 xfs_ilock(ip, XFS_IOLOCK_EXCL);
587 error = xfs_change_file_space(ip, XFS_IOC_RESVSP, &bf,
589 if (!error && !(mode & FALLOC_FL_KEEP_SIZE) &&
590 offset + len > i_size_read(inode))
591 new_size = offset + len;
593 /* Change file size if needed */
597 iattr.ia_valid = ATTR_SIZE;
598 iattr.ia_size = new_size;
599 error = xfs_setattr(ip, &iattr, XFS_ATTR_NOLOCK);
602 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
607 #define XFS_FIEMAP_FLAGS (FIEMAP_FLAG_SYNC|FIEMAP_FLAG_XATTR)
610 * Call fiemap helper to fill in user data.
611 * Returns positive errors to xfs_getbmap.
616 struct getbmapx *bmv,
620 struct fiemap_extent_info *fieinfo = *arg;
621 u32 fiemap_flags = 0;
622 u64 logical, physical, length;
624 /* Do nothing for a hole */
625 if (bmv->bmv_block == -1LL)
628 logical = BBTOB(bmv->bmv_offset);
629 physical = BBTOB(bmv->bmv_block);
630 length = BBTOB(bmv->bmv_length);
632 if (bmv->bmv_oflags & BMV_OF_PREALLOC)
633 fiemap_flags |= FIEMAP_EXTENT_UNWRITTEN;
634 else if (bmv->bmv_oflags & BMV_OF_DELALLOC) {
635 fiemap_flags |= FIEMAP_EXTENT_DELALLOC;
636 physical = 0; /* no block yet */
638 if (bmv->bmv_oflags & BMV_OF_LAST)
639 fiemap_flags |= FIEMAP_EXTENT_LAST;
641 error = fiemap_fill_next_extent(fieinfo, logical, physical,
642 length, fiemap_flags);
645 *full = 1; /* user array now full */
654 struct fiemap_extent_info *fieinfo,
658 xfs_inode_t *ip = XFS_I(inode);
662 error = fiemap_check_flags(fieinfo, XFS_FIEMAP_FLAGS);
666 /* Set up bmap header for xfs internal routine */
667 bm.bmv_offset = BTOBB(start);
668 /* Special case for whole file */
669 if (length == FIEMAP_MAX_OFFSET)
670 bm.bmv_length = -1LL;
672 bm.bmv_length = BTOBB(length);
674 /* We add one because in getbmap world count includes the header */
675 bm.bmv_count = fieinfo->fi_extents_max + 1;
676 bm.bmv_iflags = BMV_IF_PREALLOC;
677 if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR)
678 bm.bmv_iflags |= BMV_IF_ATTRFORK;
679 if (!(fieinfo->fi_flags & FIEMAP_FLAG_SYNC))
680 bm.bmv_iflags |= BMV_IF_DELALLOC;
682 error = xfs_getbmap(ip, &bm, xfs_fiemap_format, fieinfo);
689 static const struct inode_operations xfs_inode_operations = {
690 .check_acl = xfs_check_acl,
691 .truncate = xfs_vn_truncate,
692 .getattr = xfs_vn_getattr,
693 .setattr = xfs_vn_setattr,
694 .setxattr = generic_setxattr,
695 .getxattr = generic_getxattr,
696 .removexattr = generic_removexattr,
697 .listxattr = xfs_vn_listxattr,
698 .fallocate = xfs_vn_fallocate,
699 .fiemap = xfs_vn_fiemap,
702 static const struct inode_operations xfs_dir_inode_operations = {
703 .create = xfs_vn_create,
704 .lookup = xfs_vn_lookup,
706 .unlink = xfs_vn_unlink,
707 .symlink = xfs_vn_symlink,
708 .mkdir = xfs_vn_mkdir,
710 * Yes, XFS uses the same method for rmdir and unlink.
712 * There are some subtile differences deeper in the code,
713 * but we use S_ISDIR to check for those.
715 .rmdir = xfs_vn_unlink,
716 .mknod = xfs_vn_mknod,
717 .rename = xfs_vn_rename,
718 .check_acl = xfs_check_acl,
719 .getattr = xfs_vn_getattr,
720 .setattr = xfs_vn_setattr,
721 .setxattr = generic_setxattr,
722 .getxattr = generic_getxattr,
723 .removexattr = generic_removexattr,
724 .listxattr = xfs_vn_listxattr,
727 static const struct inode_operations xfs_dir_ci_inode_operations = {
728 .create = xfs_vn_create,
729 .lookup = xfs_vn_ci_lookup,
731 .unlink = xfs_vn_unlink,
732 .symlink = xfs_vn_symlink,
733 .mkdir = xfs_vn_mkdir,
735 * Yes, XFS uses the same method for rmdir and unlink.
737 * There are some subtile differences deeper in the code,
738 * but we use S_ISDIR to check for those.
740 .rmdir = xfs_vn_unlink,
741 .mknod = xfs_vn_mknod,
742 .rename = xfs_vn_rename,
743 .check_acl = xfs_check_acl,
744 .getattr = xfs_vn_getattr,
745 .setattr = xfs_vn_setattr,
746 .setxattr = generic_setxattr,
747 .getxattr = generic_getxattr,
748 .removexattr = generic_removexattr,
749 .listxattr = xfs_vn_listxattr,
752 static const struct inode_operations xfs_symlink_inode_operations = {
753 .readlink = generic_readlink,
754 .follow_link = xfs_vn_follow_link,
755 .put_link = xfs_vn_put_link,
756 .check_acl = xfs_check_acl,
757 .getattr = xfs_vn_getattr,
758 .setattr = xfs_vn_setattr,
759 .setxattr = generic_setxattr,
760 .getxattr = generic_getxattr,
761 .removexattr = generic_removexattr,
762 .listxattr = xfs_vn_listxattr,
766 xfs_diflags_to_iflags(
768 struct xfs_inode *ip)
770 if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE)
771 inode->i_flags |= S_IMMUTABLE;
773 inode->i_flags &= ~S_IMMUTABLE;
774 if (ip->i_d.di_flags & XFS_DIFLAG_APPEND)
775 inode->i_flags |= S_APPEND;
777 inode->i_flags &= ~S_APPEND;
778 if (ip->i_d.di_flags & XFS_DIFLAG_SYNC)
779 inode->i_flags |= S_SYNC;
781 inode->i_flags &= ~S_SYNC;
782 if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME)
783 inode->i_flags |= S_NOATIME;
785 inode->i_flags &= ~S_NOATIME;
789 * Initialize the Linux inode, set up the operation vectors and
792 * When reading existing inodes from disk this is called directly
793 * from xfs_iget, when creating a new inode it is called from
794 * xfs_ialloc after setting up the inode.
796 * We are always called with an uninitialised linux inode here.
797 * We need to initialise the necessary fields and take a reference
802 struct xfs_inode *ip)
804 struct inode *inode = &ip->i_vnode;
806 inode->i_ino = ip->i_ino;
807 inode->i_state = I_NEW|I_LOCK;
808 inode_add_to_lists(ip->i_mount->m_super, inode);
810 inode->i_mode = ip->i_d.di_mode;
811 inode->i_nlink = ip->i_d.di_nlink;
812 inode->i_uid = ip->i_d.di_uid;
813 inode->i_gid = ip->i_d.di_gid;
815 switch (inode->i_mode & S_IFMT) {
819 MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
820 sysv_minor(ip->i_df.if_u2.if_rdev));
827 inode->i_generation = ip->i_d.di_gen;
828 i_size_write(inode, ip->i_d.di_size);
829 inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec;
830 inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec;
831 inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec;
832 inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec;
833 inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec;
834 inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec;
835 xfs_diflags_to_iflags(inode, ip);
837 switch (inode->i_mode & S_IFMT) {
839 inode->i_op = &xfs_inode_operations;
840 inode->i_fop = &xfs_file_operations;
841 inode->i_mapping->a_ops = &xfs_address_space_operations;
844 if (xfs_sb_version_hasasciici(&XFS_M(inode->i_sb)->m_sb))
845 inode->i_op = &xfs_dir_ci_inode_operations;
847 inode->i_op = &xfs_dir_inode_operations;
848 inode->i_fop = &xfs_dir_file_operations;
851 inode->i_op = &xfs_symlink_inode_operations;
852 if (!(ip->i_df.if_flags & XFS_IFINLINE))
853 inode->i_mapping->a_ops = &xfs_address_space_operations;
856 inode->i_op = &xfs_inode_operations;
857 init_special_inode(inode, inode->i_mode, inode->i_rdev);
861 xfs_iflags_clear(ip, XFS_INEW);
864 unlock_new_inode(inode);