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/fiemap.h>
50 #include <linux/slab.h>
53 * Bring the timestamps in the XFS inode uptodate.
55 * Used before writing the inode to disk.
58 xfs_synchronize_times(
61 struct inode *inode = VFS_I(ip);
63 ip->i_d.di_atime.t_sec = (__int32_t)inode->i_atime.tv_sec;
64 ip->i_d.di_atime.t_nsec = (__int32_t)inode->i_atime.tv_nsec;
65 ip->i_d.di_ctime.t_sec = (__int32_t)inode->i_ctime.tv_sec;
66 ip->i_d.di_ctime.t_nsec = (__int32_t)inode->i_ctime.tv_nsec;
67 ip->i_d.di_mtime.t_sec = (__int32_t)inode->i_mtime.tv_sec;
68 ip->i_d.di_mtime.t_nsec = (__int32_t)inode->i_mtime.tv_nsec;
72 * If the linux inode is valid, mark it dirty.
73 * Used when committing a dirty inode into a transaction so that
74 * the inode will get written back by the linux code
77 xfs_mark_inode_dirty_sync(
80 struct inode *inode = VFS_I(ip);
82 if (!(inode->i_state & (I_WILL_FREE|I_FREEING)))
83 mark_inode_dirty_sync(inode);
90 struct inode *inode = VFS_I(ip);
92 if (!(inode->i_state & (I_WILL_FREE|I_FREEING)))
93 mark_inode_dirty(inode);
97 int xfs_initxattrs(struct inode *inode, const struct xattr *xattr_array,
100 const struct xattr *xattr;
101 struct xfs_inode *ip = XFS_I(inode);
104 for (xattr = xattr_array; xattr->name != NULL; xattr++) {
105 error = xfs_attr_set(ip, xattr->name, xattr->value,
106 xattr->value_len, ATTR_SECURE);
114 * Hook in SELinux. This is not quite correct yet, what we really need
115 * here (as we do for default ACLs) is a mechanism by which creation of
116 * these attrs can be journalled at inode creation time (along with the
117 * inode, of course, such that log replay can't cause these to be lost).
124 const struct qstr *qstr)
126 return security_inode_init_security(inode, dir, qstr,
127 &xfs_initxattrs, NULL);
132 struct xfs_name *namep,
133 struct dentry *dentry)
135 namep->name = dentry->d_name.name;
136 namep->len = dentry->d_name.len;
143 struct dentry *dentry)
145 struct xfs_name teardown;
148 * If we can't add the ACL or we fail in
149 * xfs_init_security we must back out.
150 * ENOSPC can hit here, among other things.
152 xfs_dentry_to_name(&teardown, dentry);
154 xfs_remove(XFS_I(dir), &teardown, XFS_I(inode));
161 struct dentry *dentry,
166 struct xfs_inode *ip = NULL;
167 struct posix_acl *default_acl = NULL;
168 struct xfs_name name;
172 * Irix uses Missed'em'V split, but doesn't want to see
173 * the upper 5 bits of (14bit) major.
175 if (S_ISCHR(mode) || S_ISBLK(mode)) {
176 if (unlikely(!sysv_valid_dev(rdev) || MAJOR(rdev) & ~0x1ff))
178 rdev = sysv_encode_dev(rdev);
183 if (IS_POSIXACL(dir)) {
184 default_acl = xfs_get_acl(dir, ACL_TYPE_DEFAULT);
185 if (IS_ERR(default_acl))
186 return PTR_ERR(default_acl);
189 mode &= ~current_umask();
192 xfs_dentry_to_name(&name, dentry);
193 error = xfs_create(XFS_I(dir), &name, mode, rdev, &ip);
199 error = xfs_init_security(inode, dir, &dentry->d_name);
201 goto out_cleanup_inode;
204 error = -xfs_inherit_acl(inode, default_acl);
206 goto out_cleanup_inode;
207 posix_acl_release(default_acl);
211 d_instantiate(dentry, inode);
215 xfs_cleanup_inode(dir, inode, dentry);
217 posix_acl_release(default_acl);
224 struct dentry *dentry,
226 struct nameidata *nd)
228 return xfs_vn_mknod(dir, dentry, mode, 0);
234 struct dentry *dentry,
237 return xfs_vn_mknod(dir, dentry, mode|S_IFDIR, 0);
240 STATIC struct dentry *
243 struct dentry *dentry,
244 struct nameidata *nd)
246 struct xfs_inode *cip;
247 struct xfs_name name;
250 if (dentry->d_name.len >= MAXNAMELEN)
251 return ERR_PTR(-ENAMETOOLONG);
253 xfs_dentry_to_name(&name, dentry);
254 error = xfs_lookup(XFS_I(dir), &name, &cip, NULL);
255 if (unlikely(error)) {
256 if (unlikely(error != ENOENT))
257 return ERR_PTR(-error);
262 return d_splice_alias(VFS_I(cip), dentry);
265 STATIC struct dentry *
268 struct dentry *dentry,
269 struct nameidata *nd)
271 struct xfs_inode *ip;
272 struct xfs_name xname;
273 struct xfs_name ci_name;
277 if (dentry->d_name.len >= MAXNAMELEN)
278 return ERR_PTR(-ENAMETOOLONG);
280 xfs_dentry_to_name(&xname, dentry);
281 error = xfs_lookup(XFS_I(dir), &xname, &ip, &ci_name);
282 if (unlikely(error)) {
283 if (unlikely(error != ENOENT))
284 return ERR_PTR(-error);
286 * call d_add(dentry, NULL) here when d_drop_negative_children
287 * is called in xfs_vn_mknod (ie. allow negative dentries
288 * with CI filesystems).
293 /* if exact match, just splice and exit */
295 return d_splice_alias(VFS_I(ip), dentry);
297 /* else case-insensitive match... */
298 dname.name = ci_name.name;
299 dname.len = ci_name.len;
300 dentry = d_add_ci(dentry, VFS_I(ip), &dname);
301 kmem_free(ci_name.name);
307 struct dentry *old_dentry,
309 struct dentry *dentry)
311 struct inode *inode = old_dentry->d_inode;
312 struct xfs_name name;
315 xfs_dentry_to_name(&name, dentry);
317 error = xfs_link(XFS_I(dir), XFS_I(inode), &name);
322 d_instantiate(dentry, inode);
329 struct dentry *dentry)
331 struct xfs_name name;
334 xfs_dentry_to_name(&name, dentry);
336 error = -xfs_remove(XFS_I(dir), &name, XFS_I(dentry->d_inode));
341 * With unlink, the VFS makes the dentry "negative": no inode,
342 * but still hashed. This is incompatible with case-insensitive
343 * mode, so invalidate (unhash) the dentry in CI-mode.
345 if (xfs_sb_version_hasasciici(&XFS_M(dir->i_sb)->m_sb))
346 d_invalidate(dentry);
353 struct dentry *dentry,
357 struct xfs_inode *cip = NULL;
358 struct xfs_name name;
363 (irix_symlink_mode ? 0777 & ~current_umask() : S_IRWXUGO);
364 xfs_dentry_to_name(&name, dentry);
366 error = xfs_symlink(XFS_I(dir), &name, symname, mode, &cip);
372 error = xfs_init_security(inode, dir, &dentry->d_name);
374 goto out_cleanup_inode;
376 d_instantiate(dentry, inode);
380 xfs_cleanup_inode(dir, inode, dentry);
388 struct dentry *odentry,
390 struct dentry *ndentry)
392 struct inode *new_inode = ndentry->d_inode;
393 struct xfs_name oname;
394 struct xfs_name nname;
396 xfs_dentry_to_name(&oname, odentry);
397 xfs_dentry_to_name(&nname, ndentry);
399 return -xfs_rename(XFS_I(odir), &oname, XFS_I(odentry->d_inode),
400 XFS_I(ndir), &nname, new_inode ?
401 XFS_I(new_inode) : NULL);
405 * careful here - this function can get called recursively, so
406 * we need to be very careful about how much stack we use.
407 * uio is kmalloced for this reason...
411 struct dentry *dentry,
412 struct nameidata *nd)
417 link = kmalloc(MAXPATHLEN+1, GFP_KERNEL);
421 error = -xfs_readlink(XFS_I(dentry->d_inode), link);
425 nd_set_link(nd, link);
431 nd_set_link(nd, ERR_PTR(error));
437 struct dentry *dentry,
438 struct nameidata *nd,
441 char *s = nd_get_link(nd);
449 struct vfsmount *mnt,
450 struct dentry *dentry,
453 struct inode *inode = dentry->d_inode;
454 struct xfs_inode *ip = XFS_I(inode);
455 struct xfs_mount *mp = ip->i_mount;
457 trace_xfs_getattr(ip);
459 if (XFS_FORCED_SHUTDOWN(mp))
460 return XFS_ERROR(EIO);
462 stat->size = XFS_ISIZE(ip);
463 stat->dev = inode->i_sb->s_dev;
464 stat->mode = ip->i_d.di_mode;
465 stat->nlink = ip->i_d.di_nlink;
466 stat->uid = ip->i_d.di_uid;
467 stat->gid = ip->i_d.di_gid;
468 stat->ino = ip->i_ino;
469 stat->atime = inode->i_atime;
470 stat->mtime = inode->i_mtime;
471 stat->ctime = inode->i_ctime;
473 XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks);
476 switch (inode->i_mode & S_IFMT) {
479 stat->blksize = BLKDEV_IOSIZE;
480 stat->rdev = MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
481 sysv_minor(ip->i_df.if_u2.if_rdev));
484 if (XFS_IS_REALTIME_INODE(ip)) {
486 * If the file blocks are being allocated from a
487 * realtime volume, then return the inode's realtime
488 * extent size or the realtime volume's extent size.
491 xfs_get_extsz_hint(ip) << mp->m_sb.sb_blocklog;
493 stat->blksize = xfs_preferred_iosize(mp);
503 struct dentry *dentry,
506 return -xfs_setattr(XFS_I(dentry->d_inode), iattr, 0);
509 #define XFS_FIEMAP_FLAGS (FIEMAP_FLAG_SYNC|FIEMAP_FLAG_XATTR)
512 * Call fiemap helper to fill in user data.
513 * Returns positive errors to xfs_getbmap.
518 struct getbmapx *bmv,
522 struct fiemap_extent_info *fieinfo = *arg;
523 u32 fiemap_flags = 0;
524 u64 logical, physical, length;
526 /* Do nothing for a hole */
527 if (bmv->bmv_block == -1LL)
530 logical = BBTOB(bmv->bmv_offset);
531 physical = BBTOB(bmv->bmv_block);
532 length = BBTOB(bmv->bmv_length);
534 if (bmv->bmv_oflags & BMV_OF_PREALLOC)
535 fiemap_flags |= FIEMAP_EXTENT_UNWRITTEN;
536 else if (bmv->bmv_oflags & BMV_OF_DELALLOC) {
537 fiemap_flags |= FIEMAP_EXTENT_DELALLOC;
538 physical = 0; /* no block yet */
540 if (bmv->bmv_oflags & BMV_OF_LAST)
541 fiemap_flags |= FIEMAP_EXTENT_LAST;
543 error = fiemap_fill_next_extent(fieinfo, logical, physical,
544 length, fiemap_flags);
547 *full = 1; /* user array now full */
556 struct fiemap_extent_info *fieinfo,
560 xfs_inode_t *ip = XFS_I(inode);
564 error = fiemap_check_flags(fieinfo, XFS_FIEMAP_FLAGS);
568 /* Set up bmap header for xfs internal routine */
569 bm.bmv_offset = BTOBB(start);
570 /* Special case for whole file */
571 if (length == FIEMAP_MAX_OFFSET)
572 bm.bmv_length = -1LL;
574 bm.bmv_length = BTOBB(length);
576 /* We add one because in getbmap world count includes the header */
577 bm.bmv_count = !fieinfo->fi_extents_max ? MAXEXTNUM :
578 fieinfo->fi_extents_max + 1;
579 bm.bmv_count = min_t(__s32, bm.bmv_count,
580 (PAGE_SIZE * 16 / sizeof(struct getbmapx)));
581 bm.bmv_iflags = BMV_IF_PREALLOC | BMV_IF_NO_HOLES;
582 if (fieinfo->fi_flags & FIEMAP_FLAG_XATTR)
583 bm.bmv_iflags |= BMV_IF_ATTRFORK;
584 if (!(fieinfo->fi_flags & FIEMAP_FLAG_SYNC))
585 bm.bmv_iflags |= BMV_IF_DELALLOC;
587 error = xfs_getbmap(ip, &bm, xfs_fiemap_format, fieinfo);
594 static const struct inode_operations xfs_inode_operations = {
595 .check_acl = xfs_check_acl,
596 .getattr = xfs_vn_getattr,
597 .setattr = xfs_vn_setattr,
598 .setxattr = generic_setxattr,
599 .getxattr = generic_getxattr,
600 .removexattr = generic_removexattr,
601 .listxattr = xfs_vn_listxattr,
602 .fiemap = xfs_vn_fiemap,
605 static const struct inode_operations xfs_dir_inode_operations = {
606 .create = xfs_vn_create,
607 .lookup = xfs_vn_lookup,
609 .unlink = xfs_vn_unlink,
610 .symlink = xfs_vn_symlink,
611 .mkdir = xfs_vn_mkdir,
613 * Yes, XFS uses the same method for rmdir and unlink.
615 * There are some subtile differences deeper in the code,
616 * but we use S_ISDIR to check for those.
618 .rmdir = xfs_vn_unlink,
619 .mknod = xfs_vn_mknod,
620 .rename = xfs_vn_rename,
621 .check_acl = xfs_check_acl,
622 .getattr = xfs_vn_getattr,
623 .setattr = xfs_vn_setattr,
624 .setxattr = generic_setxattr,
625 .getxattr = generic_getxattr,
626 .removexattr = generic_removexattr,
627 .listxattr = xfs_vn_listxattr,
630 static const struct inode_operations xfs_dir_ci_inode_operations = {
631 .create = xfs_vn_create,
632 .lookup = xfs_vn_ci_lookup,
634 .unlink = xfs_vn_unlink,
635 .symlink = xfs_vn_symlink,
636 .mkdir = xfs_vn_mkdir,
638 * Yes, XFS uses the same method for rmdir and unlink.
640 * There are some subtile differences deeper in the code,
641 * but we use S_ISDIR to check for those.
643 .rmdir = xfs_vn_unlink,
644 .mknod = xfs_vn_mknod,
645 .rename = xfs_vn_rename,
646 .check_acl = xfs_check_acl,
647 .getattr = xfs_vn_getattr,
648 .setattr = xfs_vn_setattr,
649 .setxattr = generic_setxattr,
650 .getxattr = generic_getxattr,
651 .removexattr = generic_removexattr,
652 .listxattr = xfs_vn_listxattr,
655 static const struct inode_operations xfs_symlink_inode_operations = {
656 .readlink = generic_readlink,
657 .follow_link = xfs_vn_follow_link,
658 .put_link = xfs_vn_put_link,
659 .check_acl = xfs_check_acl,
660 .getattr = xfs_vn_getattr,
661 .setattr = xfs_vn_setattr,
662 .setxattr = generic_setxattr,
663 .getxattr = generic_getxattr,
664 .removexattr = generic_removexattr,
665 .listxattr = xfs_vn_listxattr,
669 xfs_diflags_to_iflags(
671 struct xfs_inode *ip)
673 if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE)
674 inode->i_flags |= S_IMMUTABLE;
676 inode->i_flags &= ~S_IMMUTABLE;
677 if (ip->i_d.di_flags & XFS_DIFLAG_APPEND)
678 inode->i_flags |= S_APPEND;
680 inode->i_flags &= ~S_APPEND;
681 if (ip->i_d.di_flags & XFS_DIFLAG_SYNC)
682 inode->i_flags |= S_SYNC;
684 inode->i_flags &= ~S_SYNC;
685 if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME)
686 inode->i_flags |= S_NOATIME;
688 inode->i_flags &= ~S_NOATIME;
692 * Initialize the Linux inode, set up the operation vectors and
695 * When reading existing inodes from disk this is called directly
696 * from xfs_iget, when creating a new inode it is called from
697 * xfs_ialloc after setting up the inode.
699 * We are always called with an uninitialised linux inode here.
700 * We need to initialise the necessary fields and take a reference
705 struct xfs_inode *ip)
707 struct inode *inode = &ip->i_vnode;
709 inode->i_ino = ip->i_ino;
710 inode->i_state = I_NEW;
712 inode_sb_list_add(inode);
713 /* make the inode look hashed for the writeback code */
714 hlist_add_fake(&inode->i_hash);
716 inode->i_mode = ip->i_d.di_mode;
717 inode->i_nlink = ip->i_d.di_nlink;
718 inode->i_uid = ip->i_d.di_uid;
719 inode->i_gid = ip->i_d.di_gid;
721 switch (inode->i_mode & S_IFMT) {
725 MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
726 sysv_minor(ip->i_df.if_u2.if_rdev));
733 inode->i_generation = ip->i_d.di_gen;
734 i_size_write(inode, ip->i_d.di_size);
735 inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec;
736 inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec;
737 inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec;
738 inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec;
739 inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec;
740 inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec;
741 xfs_diflags_to_iflags(inode, ip);
743 switch (inode->i_mode & S_IFMT) {
745 inode->i_op = &xfs_inode_operations;
746 inode->i_fop = &xfs_file_operations;
747 inode->i_mapping->a_ops = &xfs_address_space_operations;
750 if (xfs_sb_version_hasasciici(&XFS_M(inode->i_sb)->m_sb))
751 inode->i_op = &xfs_dir_ci_inode_operations;
753 inode->i_op = &xfs_dir_inode_operations;
754 inode->i_fop = &xfs_dir_file_operations;
757 inode->i_op = &xfs_symlink_inode_operations;
758 if (!(ip->i_df.if_flags & XFS_IFINLINE))
759 inode->i_mapping->a_ops = &xfs_address_space_operations;
762 inode->i_op = &xfs_inode_operations;
763 init_special_inode(inode, inode->i_mode, inode->i_rdev);
767 xfs_iflags_clear(ip, XFS_INEW);
770 unlock_new_inode(inode);