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"
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_dir_sf.h"
36 #include "xfs_dir2_sf.h"
37 #include "xfs_attr_sf.h"
38 #include "xfs_dinode.h"
39 #include "xfs_inode.h"
41 #include "xfs_btree.h"
42 #include "xfs_ialloc.h"
43 #include "xfs_rtalloc.h"
44 #include "xfs_error.h"
45 #include "xfs_itable.h"
51 #include "xfs_buf_item.h"
52 #include "xfs_utils.h"
54 #include <linux/xattr.h>
55 #include <linux/namei.h>
58 * Change the requested timestamp in the given inode.
59 * We don't lock across timestamp updates, and we don't log them but
60 * we do record the fact that there is dirty information in core.
62 * NOTE -- callers MUST combine XFS_ICHGTIME_MOD or XFS_ICHGTIME_CHG
63 * with XFS_ICHGTIME_ACC to be sure that access time
64 * update will take. Calling first with XFS_ICHGTIME_ACC
65 * and then XFS_ICHGTIME_MOD may fail to modify the access
66 * timestamp if the filesystem is mounted noacctm.
73 struct inode *inode = LINVFS_GET_IP(XFS_ITOV(ip));
77 * We're not supposed to change timestamps in readonly-mounted
78 * filesystems. Throw it away if anyone asks us.
80 if (unlikely(IS_RDONLY(inode)))
84 * Don't update access timestamps on reads if mounted "noatime".
85 * Throw it away if anyone asks us.
88 (ip->i_mount->m_flags & XFS_MOUNT_NOATIME || IS_NOATIME(inode)) &&
89 (flags & (XFS_ICHGTIME_ACC|XFS_ICHGTIME_MOD|XFS_ICHGTIME_CHG)) ==
94 if (flags & XFS_ICHGTIME_MOD) {
96 ip->i_d.di_mtime.t_sec = (__int32_t)tv.tv_sec;
97 ip->i_d.di_mtime.t_nsec = (__int32_t)tv.tv_nsec;
99 if (flags & XFS_ICHGTIME_ACC) {
101 ip->i_d.di_atime.t_sec = (__int32_t)tv.tv_sec;
102 ip->i_d.di_atime.t_nsec = (__int32_t)tv.tv_nsec;
104 if (flags & XFS_ICHGTIME_CHG) {
106 ip->i_d.di_ctime.t_sec = (__int32_t)tv.tv_sec;
107 ip->i_d.di_ctime.t_nsec = (__int32_t)tv.tv_nsec;
111 * We update the i_update_core field _after_ changing
112 * the timestamps in order to coordinate properly with
113 * xfs_iflush() so that we don't lose timestamp updates.
114 * This keeps us from having to hold the inode lock
115 * while doing this. We use the SYNCHRONIZE macro to
116 * ensure that the compiler does not reorder the update
117 * of i_update_core above the timestamp updates above.
120 ip->i_update_core = 1;
121 if (!(inode->i_state & I_LOCK))
122 mark_inode_dirty_sync(inode);
126 * Variant on the above which avoids querying the system clock
127 * in situations where we know the Linux inode timestamps have
128 * just been updated (and so we can update our inode cheaply).
129 * We also skip the readonly and noatime checks here, they are
130 * also catered for already.
141 * We're not supposed to change timestamps in readonly-mounted
142 * filesystems. Throw it away if anyone asks us.
144 if (unlikely(IS_RDONLY(inode)))
148 * Don't update access timestamps on reads if mounted "noatime".
149 * Throw it away if anyone asks us.
152 (ip->i_mount->m_flags & XFS_MOUNT_NOATIME || IS_NOATIME(inode)) &&
153 ((flags & (XFS_ICHGTIME_ACC|XFS_ICHGTIME_MOD|XFS_ICHGTIME_CHG)) ==
157 if (flags & XFS_ICHGTIME_MOD) {
158 tvp = &inode->i_mtime;
159 ip->i_d.di_mtime.t_sec = (__int32_t)tvp->tv_sec;
160 ip->i_d.di_mtime.t_nsec = (__int32_t)tvp->tv_nsec;
162 if (flags & XFS_ICHGTIME_ACC) {
163 tvp = &inode->i_atime;
164 ip->i_d.di_atime.t_sec = (__int32_t)tvp->tv_sec;
165 ip->i_d.di_atime.t_nsec = (__int32_t)tvp->tv_nsec;
167 if (flags & XFS_ICHGTIME_CHG) {
168 tvp = &inode->i_ctime;
169 ip->i_d.di_ctime.t_sec = (__int32_t)tvp->tv_sec;
170 ip->i_d.di_ctime.t_nsec = (__int32_t)tvp->tv_nsec;
174 * We update the i_update_core field _after_ changing
175 * the timestamps in order to coordinate properly with
176 * xfs_iflush() so that we don't lose timestamp updates.
177 * This keeps us from having to hold the inode lock
178 * while doing this. We use the SYNCHRONIZE macro to
179 * ensure that the compiler does not reorder the update
180 * of i_update_core above the timestamp updates above.
183 ip->i_update_core = 1;
184 if (!(inode->i_state & I_LOCK))
185 mark_inode_dirty_sync(inode);
190 * Pull the link count and size up from the xfs inode to the linux inode
196 vnode_t *vp = LINVFS_GET_VP(ip);
200 va.va_mask = XFS_AT_NLINK|XFS_AT_SIZE|XFS_AT_NBLOCKS;
201 VOP_GETATTR(vp, &va, ATTR_LAZY, NULL, error);
202 if (likely(!error)) {
203 ip->i_nlink = va.va_nlink;
204 ip->i_blocks = va.va_nblocks;
206 /* we're under i_sem so i_size can't change under us */
207 if (i_size_read(ip) != va.va_size)
208 i_size_write(ip, va.va_size);
213 * Determine whether a process has a valid fs_struct (kernel daemons
214 * like knfsd don't have an fs_struct).
216 * XXX(hch): nfsd is broken, better fix it instead.
219 has_fs_struct(struct task_struct *task)
221 return (task->fs != init_task.fs);
227 struct dentry *dentry,
233 vnode_t *vp = NULL, *dvp = LINVFS_GET_VP(dir);
234 xfs_acl_t *default_acl = NULL;
235 attrexists_t test_default_acl = _ACL_DEFAULT_EXISTS;
239 * Irix uses Missed'em'V split, but doesn't want to see
240 * the upper 5 bits of (14bit) major.
242 if (!sysv_valid_dev(rdev) || MAJOR(rdev) & ~0x1ff)
245 if (test_default_acl && test_default_acl(dvp)) {
246 if (!_ACL_ALLOC(default_acl))
248 if (!_ACL_GET_DEFAULT(dvp, default_acl)) {
249 _ACL_FREE(default_acl);
254 if (IS_POSIXACL(dir) && !default_acl && has_fs_struct(current))
255 mode &= ~current->fs->umask;
257 memset(&va, 0, sizeof(va));
258 va.va_mask = XFS_AT_TYPE|XFS_AT_MODE;
261 switch (mode & S_IFMT) {
262 case S_IFCHR: case S_IFBLK: case S_IFIFO: case S_IFSOCK:
263 va.va_rdev = sysv_encode_dev(rdev);
264 va.va_mask |= XFS_AT_RDEV;
267 VOP_CREATE(dvp, dentry, &va, &vp, NULL, error);
270 VOP_MKDIR(dvp, dentry, &va, &vp, NULL, error);
279 error = _ACL_INHERIT(vp, &va, default_acl);
283 struct dentry teardown = {};
287 * If we can't add the ACL we must back out.
288 * ENOSPC can hit here, among other things.
290 teardown.d_inode = ip = LINVFS_GET_IP(vp);
291 teardown.d_name = dentry->d_name;
296 VOP_RMDIR(dvp, &teardown, NULL, err2);
298 VOP_REMOVE(dvp, &teardown, NULL, err2);
302 _ACL_FREE(default_acl);
307 ip = LINVFS_GET_IP(vp);
309 if (S_ISCHR(mode) || S_ISBLK(mode))
311 else if (S_ISDIR(mode))
313 d_instantiate(dentry, ip);
314 validate_fields(dir);
322 struct dentry *dentry,
324 struct nameidata *nd)
326 return linvfs_mknod(dir, dentry, mode, 0);
332 struct dentry *dentry,
335 return linvfs_mknod(dir, dentry, mode|S_IFDIR, 0);
338 STATIC struct dentry *
341 struct dentry *dentry,
342 struct nameidata *nd)
344 struct vnode *vp = LINVFS_GET_VP(dir), *cvp;
347 if (dentry->d_name.len >= MAXNAMELEN)
348 return ERR_PTR(-ENAMETOOLONG);
350 VOP_LOOKUP(vp, dentry, &cvp, 0, NULL, NULL, error);
352 if (unlikely(error != ENOENT))
353 return ERR_PTR(-error);
358 return d_splice_alias(LINVFS_GET_IP(cvp), dentry);
363 struct dentry *old_dentry,
365 struct dentry *dentry)
367 struct inode *ip; /* inode of guy being linked to */
368 vnode_t *tdvp; /* target directory for new name/link */
369 vnode_t *vp; /* vp of name being linked */
372 ip = old_dentry->d_inode; /* inode being linked to */
373 if (S_ISDIR(ip->i_mode))
376 tdvp = LINVFS_GET_VP(dir);
377 vp = LINVFS_GET_VP(ip);
379 VOP_LINK(tdvp, vp, dentry, NULL, error);
384 d_instantiate(dentry, ip);
392 struct dentry *dentry)
395 vnode_t *dvp; /* directory containing name to remove */
398 inode = dentry->d_inode;
399 dvp = LINVFS_GET_VP(dir);
401 VOP_REMOVE(dvp, dentry, NULL, error);
403 validate_fields(dir); /* For size only */
404 validate_fields(inode);
413 struct dentry *dentry,
418 vnode_t *dvp; /* directory containing name of symlink */
419 vnode_t *cvp; /* used to lookup symlink to put in dentry */
422 dvp = LINVFS_GET_VP(dir);
425 memset(&va, 0, sizeof(va));
426 va.va_mode = S_IFLNK |
427 (irix_symlink_mode ? 0777 & ~current->fs->umask : S_IRWXUGO);
428 va.va_mask = XFS_AT_TYPE|XFS_AT_MODE;
431 VOP_SYMLINK(dvp, dentry, &va, (char *)symname, &cvp, NULL, error);
433 ip = LINVFS_GET_IP(cvp);
434 d_instantiate(dentry, ip);
435 validate_fields(dir);
436 validate_fields(ip); /* size needs update */
444 struct dentry *dentry)
446 struct inode *inode = dentry->d_inode;
447 vnode_t *dvp = LINVFS_GET_VP(dir);
450 VOP_RMDIR(dvp, dentry, NULL, error);
452 validate_fields(inode);
453 validate_fields(dir);
461 struct dentry *odentry,
463 struct dentry *ndentry)
465 struct inode *new_inode = ndentry->d_inode;
466 vnode_t *fvp; /* from directory */
467 vnode_t *tvp; /* target directory */
470 fvp = LINVFS_GET_VP(odir);
471 tvp = LINVFS_GET_VP(ndir);
473 VOP_RENAME(fvp, odentry, tvp, ndentry, NULL, error);
478 validate_fields(new_inode);
480 validate_fields(odir);
482 validate_fields(ndir);
487 * careful here - this function can get called recursively, so
488 * we need to be very careful about how much stack we use.
489 * uio is kmalloced for this reason...
493 struct dentry *dentry,
494 struct nameidata *nd)
505 link = (char *)kmalloc(MAXNAMELEN+1, GFP_KERNEL);
507 nd_set_link(nd, ERR_PTR(-ENOMEM));
511 uio = (uio_t *)kmalloc(sizeof(uio_t), GFP_KERNEL);
514 nd_set_link(nd, ERR_PTR(-ENOMEM));
518 vp = LINVFS_GET_VP(dentry->d_inode);
521 iov.iov_len = MAXNAMELEN;
525 uio->uio_segflg = UIO_SYSSPACE;
526 uio->uio_resid = MAXNAMELEN;
529 VOP_READLINK(vp, uio, 0, NULL, error);
532 link = ERR_PTR(-error);
534 link[MAXNAMELEN - uio->uio_resid] = '\0';
538 nd_set_link(nd, link);
544 struct dentry *dentry,
545 struct nameidata *nd,
548 char *s = nd_get_link(nd);
554 #ifdef CONFIG_XFS_POSIX_ACL
559 struct nameidata *nd)
561 vnode_t *vp = LINVFS_GET_VP(inode);
564 mode <<= 6; /* convert from linux to vnode access bits */
565 VOP_ACCESS(vp, mode, NULL, error);
569 #define linvfs_permission NULL
574 struct vfsmount *mnt,
575 struct dentry *dentry,
578 struct inode *inode = dentry->d_inode;
579 vnode_t *vp = LINVFS_GET_VP(inode);
582 if (unlikely(vp->v_flag & VMODIFIED))
583 error = vn_revalidate(vp);
585 generic_fillattr(inode, stat);
591 struct dentry *dentry,
594 struct inode *inode = dentry->d_inode;
595 unsigned int ia_valid = attr->ia_valid;
596 vnode_t *vp = LINVFS_GET_VP(inode);
601 memset(&vattr, 0, sizeof(vattr_t));
602 if (ia_valid & ATTR_UID) {
603 vattr.va_mask |= XFS_AT_UID;
604 vattr.va_uid = attr->ia_uid;
606 if (ia_valid & ATTR_GID) {
607 vattr.va_mask |= XFS_AT_GID;
608 vattr.va_gid = attr->ia_gid;
610 if (ia_valid & ATTR_SIZE) {
611 vattr.va_mask |= XFS_AT_SIZE;
612 vattr.va_size = attr->ia_size;
614 if (ia_valid & ATTR_ATIME) {
615 vattr.va_mask |= XFS_AT_ATIME;
616 vattr.va_atime = attr->ia_atime;
618 if (ia_valid & ATTR_MTIME) {
619 vattr.va_mask |= XFS_AT_MTIME;
620 vattr.va_mtime = attr->ia_mtime;
622 if (ia_valid & ATTR_CTIME) {
623 vattr.va_mask |= XFS_AT_CTIME;
624 vattr.va_ctime = attr->ia_ctime;
626 if (ia_valid & ATTR_MODE) {
627 vattr.va_mask |= XFS_AT_MODE;
628 vattr.va_mode = attr->ia_mode;
629 if (!in_group_p(inode->i_gid) && !capable(CAP_FSETID))
630 inode->i_mode &= ~S_ISGID;
633 if (ia_valid & (ATTR_MTIME_SET | ATTR_ATIME_SET))
636 if ((ia_valid & ATTR_NO_BLOCK))
637 flags |= ATTR_NONBLOCK;
640 VOP_SETATTR(vp, &vattr, flags, NULL, error);
651 block_truncate_page(inode->i_mapping, inode->i_size, linvfs_get_block);
656 struct dentry *dentry,
662 vnode_t *vp = LINVFS_GET_VP(dentry->d_inode);
663 char *attr = (char *)name;
668 namesp = attr_lookup_namespace(attr, attr_namespaces, ATTR_NAMECOUNT);
671 attr += namesp->attr_namelen;
672 error = namesp->attr_capable(vp, NULL);
676 /* Convert Linux syscall to XFS internal ATTR flags */
677 if (flags & XATTR_CREATE)
678 xflags |= ATTR_CREATE;
679 if (flags & XATTR_REPLACE)
680 xflags |= ATTR_REPLACE;
681 xflags |= namesp->attr_flag;
682 return namesp->attr_set(vp, attr, (void *)data, size, xflags);
687 struct dentry *dentry,
692 vnode_t *vp = LINVFS_GET_VP(dentry->d_inode);
693 char *attr = (char *)name;
698 namesp = attr_lookup_namespace(attr, attr_namespaces, ATTR_NAMECOUNT);
701 attr += namesp->attr_namelen;
702 error = namesp->attr_capable(vp, NULL);
706 /* Convert Linux syscall to XFS internal ATTR flags */
708 xflags |= ATTR_KERNOVAL;
711 xflags |= namesp->attr_flag;
712 return namesp->attr_get(vp, attr, (void *)data, size, xflags);
717 struct dentry *dentry,
721 vnode_t *vp = LINVFS_GET_VP(dentry->d_inode);
722 int error, xflags = ATTR_KERNAMELS;
726 xflags |= ATTR_KERNOVAL;
727 xflags |= capable(CAP_SYS_ADMIN) ? ATTR_KERNFULLS : ATTR_KERNORMALS;
729 error = attr_generic_list(vp, data, size, xflags, &result);
737 struct dentry *dentry,
740 vnode_t *vp = LINVFS_GET_VP(dentry->d_inode);
741 char *attr = (char *)name;
746 namesp = attr_lookup_namespace(attr, attr_namespaces, ATTR_NAMECOUNT);
749 attr += namesp->attr_namelen;
750 error = namesp->attr_capable(vp, NULL);
753 xflags |= namesp->attr_flag;
754 return namesp->attr_remove(vp, attr, xflags);
758 struct inode_operations linvfs_file_inode_operations = {
759 .permission = linvfs_permission,
760 .truncate = linvfs_truncate,
761 .getattr = linvfs_getattr,
762 .setattr = linvfs_setattr,
763 .setxattr = linvfs_setxattr,
764 .getxattr = linvfs_getxattr,
765 .listxattr = linvfs_listxattr,
766 .removexattr = linvfs_removexattr,
769 struct inode_operations linvfs_dir_inode_operations = {
770 .create = linvfs_create,
771 .lookup = linvfs_lookup,
773 .unlink = linvfs_unlink,
774 .symlink = linvfs_symlink,
775 .mkdir = linvfs_mkdir,
776 .rmdir = linvfs_rmdir,
777 .mknod = linvfs_mknod,
778 .rename = linvfs_rename,
779 .permission = linvfs_permission,
780 .getattr = linvfs_getattr,
781 .setattr = linvfs_setattr,
782 .setxattr = linvfs_setxattr,
783 .getxattr = linvfs_getxattr,
784 .listxattr = linvfs_listxattr,
785 .removexattr = linvfs_removexattr,
788 struct inode_operations linvfs_symlink_inode_operations = {
789 .readlink = generic_readlink,
790 .follow_link = linvfs_follow_link,
791 .put_link = linvfs_put_link,
792 .permission = linvfs_permission,
793 .getattr = linvfs_getattr,
794 .setattr = linvfs_setattr,
795 .setxattr = linvfs_setxattr,
796 .getxattr = linvfs_getxattr,
797 .listxattr = linvfs_listxattr,
798 .removexattr = linvfs_removexattr,