signal: Ensure generic siginfos the kernel sends have all bits initialized
[cris-mirror.git] / fs / xfs / xfs_ioctl.c
blob20dc65fef6a42fa52ad412b9cd5619aa1434245b
1 /*
2 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
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
18 #include "xfs.h"
19 #include "xfs_fs.h"
20 #include "xfs_shared.h"
21 #include "xfs_format.h"
22 #include "xfs_log_format.h"
23 #include "xfs_trans_resv.h"
24 #include "xfs_mount.h"
25 #include "xfs_inode.h"
26 #include "xfs_ioctl.h"
27 #include "xfs_alloc.h"
28 #include "xfs_rtalloc.h"
29 #include "xfs_itable.h"
30 #include "xfs_error.h"
31 #include "xfs_attr.h"
32 #include "xfs_bmap.h"
33 #include "xfs_bmap_util.h"
34 #include "xfs_fsops.h"
35 #include "xfs_discard.h"
36 #include "xfs_quota.h"
37 #include "xfs_export.h"
38 #include "xfs_trace.h"
39 #include "xfs_icache.h"
40 #include "xfs_symlink.h"
41 #include "xfs_trans.h"
42 #include "xfs_pnfs.h"
43 #include "xfs_acl.h"
44 #include "xfs_btree.h"
45 #include <linux/fsmap.h>
46 #include "xfs_fsmap.h"
47 #include "scrub/xfs_scrub.h"
49 #include <linux/capability.h>
50 #include <linux/cred.h>
51 #include <linux/dcache.h>
52 #include <linux/mount.h>
53 #include <linux/namei.h>
54 #include <linux/pagemap.h>
55 #include <linux/slab.h>
56 #include <linux/exportfs.h>
59 * xfs_find_handle maps from userspace xfs_fsop_handlereq structure to
60 * a file or fs handle.
62 * XFS_IOC_PATH_TO_FSHANDLE
63 * returns fs handle for a mount point or path within that mount point
64 * XFS_IOC_FD_TO_HANDLE
65 * returns full handle for a FD opened in user space
66 * XFS_IOC_PATH_TO_HANDLE
67 * returns full handle for a path
69 int
70 xfs_find_handle(
71 unsigned int cmd,
72 xfs_fsop_handlereq_t *hreq)
74 int hsize;
75 xfs_handle_t handle;
76 struct inode *inode;
77 struct fd f = {NULL};
78 struct path path;
79 int error;
80 struct xfs_inode *ip;
82 if (cmd == XFS_IOC_FD_TO_HANDLE) {
83 f = fdget(hreq->fd);
84 if (!f.file)
85 return -EBADF;
86 inode = file_inode(f.file);
87 } else {
88 error = user_lpath((const char __user *)hreq->path, &path);
89 if (error)
90 return error;
91 inode = d_inode(path.dentry);
93 ip = XFS_I(inode);
96 * We can only generate handles for inodes residing on a XFS filesystem,
97 * and only for regular files, directories or symbolic links.
99 error = -EINVAL;
100 if (inode->i_sb->s_magic != XFS_SB_MAGIC)
101 goto out_put;
103 error = -EBADF;
104 if (!S_ISREG(inode->i_mode) &&
105 !S_ISDIR(inode->i_mode) &&
106 !S_ISLNK(inode->i_mode))
107 goto out_put;
110 memcpy(&handle.ha_fsid, ip->i_mount->m_fixedfsid, sizeof(xfs_fsid_t));
112 if (cmd == XFS_IOC_PATH_TO_FSHANDLE) {
114 * This handle only contains an fsid, zero the rest.
116 memset(&handle.ha_fid, 0, sizeof(handle.ha_fid));
117 hsize = sizeof(xfs_fsid_t);
118 } else {
119 handle.ha_fid.fid_len = sizeof(xfs_fid_t) -
120 sizeof(handle.ha_fid.fid_len);
121 handle.ha_fid.fid_pad = 0;
122 handle.ha_fid.fid_gen = inode->i_generation;
123 handle.ha_fid.fid_ino = ip->i_ino;
124 hsize = sizeof(xfs_handle_t);
127 error = -EFAULT;
128 if (copy_to_user(hreq->ohandle, &handle, hsize) ||
129 copy_to_user(hreq->ohandlen, &hsize, sizeof(__s32)))
130 goto out_put;
132 error = 0;
134 out_put:
135 if (cmd == XFS_IOC_FD_TO_HANDLE)
136 fdput(f);
137 else
138 path_put(&path);
139 return error;
143 * No need to do permission checks on the various pathname components
144 * as the handle operations are privileged.
146 STATIC int
147 xfs_handle_acceptable(
148 void *context,
149 struct dentry *dentry)
151 return 1;
155 * Convert userspace handle data into a dentry.
157 struct dentry *
158 xfs_handle_to_dentry(
159 struct file *parfilp,
160 void __user *uhandle,
161 u32 hlen)
163 xfs_handle_t handle;
164 struct xfs_fid64 fid;
167 * Only allow handle opens under a directory.
169 if (!S_ISDIR(file_inode(parfilp)->i_mode))
170 return ERR_PTR(-ENOTDIR);
172 if (hlen != sizeof(xfs_handle_t))
173 return ERR_PTR(-EINVAL);
174 if (copy_from_user(&handle, uhandle, hlen))
175 return ERR_PTR(-EFAULT);
176 if (handle.ha_fid.fid_len !=
177 sizeof(handle.ha_fid) - sizeof(handle.ha_fid.fid_len))
178 return ERR_PTR(-EINVAL);
180 memset(&fid, 0, sizeof(struct fid));
181 fid.ino = handle.ha_fid.fid_ino;
182 fid.gen = handle.ha_fid.fid_gen;
184 return exportfs_decode_fh(parfilp->f_path.mnt, (struct fid *)&fid, 3,
185 FILEID_INO32_GEN | XFS_FILEID_TYPE_64FLAG,
186 xfs_handle_acceptable, NULL);
189 STATIC struct dentry *
190 xfs_handlereq_to_dentry(
191 struct file *parfilp,
192 xfs_fsop_handlereq_t *hreq)
194 return xfs_handle_to_dentry(parfilp, hreq->ihandle, hreq->ihandlen);
198 xfs_open_by_handle(
199 struct file *parfilp,
200 xfs_fsop_handlereq_t *hreq)
202 const struct cred *cred = current_cred();
203 int error;
204 int fd;
205 int permflag;
206 struct file *filp;
207 struct inode *inode;
208 struct dentry *dentry;
209 fmode_t fmode;
210 struct path path;
212 if (!capable(CAP_SYS_ADMIN))
213 return -EPERM;
215 dentry = xfs_handlereq_to_dentry(parfilp, hreq);
216 if (IS_ERR(dentry))
217 return PTR_ERR(dentry);
218 inode = d_inode(dentry);
220 /* Restrict xfs_open_by_handle to directories & regular files. */
221 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))) {
222 error = -EPERM;
223 goto out_dput;
226 #if BITS_PER_LONG != 32
227 hreq->oflags |= O_LARGEFILE;
228 #endif
230 permflag = hreq->oflags;
231 fmode = OPEN_FMODE(permflag);
232 if ((!(permflag & O_APPEND) || (permflag & O_TRUNC)) &&
233 (fmode & FMODE_WRITE) && IS_APPEND(inode)) {
234 error = -EPERM;
235 goto out_dput;
238 if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
239 error = -EPERM;
240 goto out_dput;
243 /* Can't write directories. */
244 if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
245 error = -EISDIR;
246 goto out_dput;
249 fd = get_unused_fd_flags(0);
250 if (fd < 0) {
251 error = fd;
252 goto out_dput;
255 path.mnt = parfilp->f_path.mnt;
256 path.dentry = dentry;
257 filp = dentry_open(&path, hreq->oflags, cred);
258 dput(dentry);
259 if (IS_ERR(filp)) {
260 put_unused_fd(fd);
261 return PTR_ERR(filp);
264 if (S_ISREG(inode->i_mode)) {
265 filp->f_flags |= O_NOATIME;
266 filp->f_mode |= FMODE_NOCMTIME;
269 fd_install(fd, filp);
270 return fd;
272 out_dput:
273 dput(dentry);
274 return error;
278 xfs_readlink_by_handle(
279 struct file *parfilp,
280 xfs_fsop_handlereq_t *hreq)
282 struct dentry *dentry;
283 __u32 olen;
284 int error;
286 if (!capable(CAP_SYS_ADMIN))
287 return -EPERM;
289 dentry = xfs_handlereq_to_dentry(parfilp, hreq);
290 if (IS_ERR(dentry))
291 return PTR_ERR(dentry);
293 /* Restrict this handle operation to symlinks only. */
294 if (!d_is_symlink(dentry)) {
295 error = -EINVAL;
296 goto out_dput;
299 if (copy_from_user(&olen, hreq->ohandlen, sizeof(__u32))) {
300 error = -EFAULT;
301 goto out_dput;
304 error = vfs_readlink(dentry, hreq->ohandle, olen);
306 out_dput:
307 dput(dentry);
308 return error;
312 xfs_set_dmattrs(
313 xfs_inode_t *ip,
314 uint evmask,
315 uint16_t state)
317 xfs_mount_t *mp = ip->i_mount;
318 xfs_trans_t *tp;
319 int error;
321 if (!capable(CAP_SYS_ADMIN))
322 return -EPERM;
324 if (XFS_FORCED_SHUTDOWN(mp))
325 return -EIO;
327 error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
328 if (error)
329 return error;
331 xfs_ilock(ip, XFS_ILOCK_EXCL);
332 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
334 ip->i_d.di_dmevmask = evmask;
335 ip->i_d.di_dmstate = state;
337 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
338 error = xfs_trans_commit(tp);
340 return error;
343 STATIC int
344 xfs_fssetdm_by_handle(
345 struct file *parfilp,
346 void __user *arg)
348 int error;
349 struct fsdmidata fsd;
350 xfs_fsop_setdm_handlereq_t dmhreq;
351 struct dentry *dentry;
353 if (!capable(CAP_MKNOD))
354 return -EPERM;
355 if (copy_from_user(&dmhreq, arg, sizeof(xfs_fsop_setdm_handlereq_t)))
356 return -EFAULT;
358 error = mnt_want_write_file(parfilp);
359 if (error)
360 return error;
362 dentry = xfs_handlereq_to_dentry(parfilp, &dmhreq.hreq);
363 if (IS_ERR(dentry)) {
364 mnt_drop_write_file(parfilp);
365 return PTR_ERR(dentry);
368 if (IS_IMMUTABLE(d_inode(dentry)) || IS_APPEND(d_inode(dentry))) {
369 error = -EPERM;
370 goto out;
373 if (copy_from_user(&fsd, dmhreq.data, sizeof(fsd))) {
374 error = -EFAULT;
375 goto out;
378 error = xfs_set_dmattrs(XFS_I(d_inode(dentry)), fsd.fsd_dmevmask,
379 fsd.fsd_dmstate);
381 out:
382 mnt_drop_write_file(parfilp);
383 dput(dentry);
384 return error;
387 STATIC int
388 xfs_attrlist_by_handle(
389 struct file *parfilp,
390 void __user *arg)
392 int error = -ENOMEM;
393 attrlist_cursor_kern_t *cursor;
394 struct xfs_fsop_attrlist_handlereq __user *p = arg;
395 xfs_fsop_attrlist_handlereq_t al_hreq;
396 struct dentry *dentry;
397 char *kbuf;
399 if (!capable(CAP_SYS_ADMIN))
400 return -EPERM;
401 if (copy_from_user(&al_hreq, arg, sizeof(xfs_fsop_attrlist_handlereq_t)))
402 return -EFAULT;
403 if (al_hreq.buflen < sizeof(struct attrlist) ||
404 al_hreq.buflen > XFS_XATTR_LIST_MAX)
405 return -EINVAL;
408 * Reject flags, only allow namespaces.
410 if (al_hreq.flags & ~(ATTR_ROOT | ATTR_SECURE))
411 return -EINVAL;
413 dentry = xfs_handlereq_to_dentry(parfilp, &al_hreq.hreq);
414 if (IS_ERR(dentry))
415 return PTR_ERR(dentry);
417 kbuf = kmem_zalloc_large(al_hreq.buflen, KM_SLEEP);
418 if (!kbuf)
419 goto out_dput;
421 cursor = (attrlist_cursor_kern_t *)&al_hreq.pos;
422 error = xfs_attr_list(XFS_I(d_inode(dentry)), kbuf, al_hreq.buflen,
423 al_hreq.flags, cursor);
424 if (error)
425 goto out_kfree;
427 if (copy_to_user(&p->pos, cursor, sizeof(attrlist_cursor_kern_t))) {
428 error = -EFAULT;
429 goto out_kfree;
432 if (copy_to_user(al_hreq.buffer, kbuf, al_hreq.buflen))
433 error = -EFAULT;
435 out_kfree:
436 kmem_free(kbuf);
437 out_dput:
438 dput(dentry);
439 return error;
443 xfs_attrmulti_attr_get(
444 struct inode *inode,
445 unsigned char *name,
446 unsigned char __user *ubuf,
447 uint32_t *len,
448 uint32_t flags)
450 unsigned char *kbuf;
451 int error = -EFAULT;
453 if (*len > XFS_XATTR_SIZE_MAX)
454 return -EINVAL;
455 kbuf = kmem_zalloc_large(*len, KM_SLEEP);
456 if (!kbuf)
457 return -ENOMEM;
459 error = xfs_attr_get(XFS_I(inode), name, kbuf, (int *)len, flags);
460 if (error)
461 goto out_kfree;
463 if (copy_to_user(ubuf, kbuf, *len))
464 error = -EFAULT;
466 out_kfree:
467 kmem_free(kbuf);
468 return error;
472 xfs_attrmulti_attr_set(
473 struct inode *inode,
474 unsigned char *name,
475 const unsigned char __user *ubuf,
476 uint32_t len,
477 uint32_t flags)
479 unsigned char *kbuf;
480 int error;
482 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
483 return -EPERM;
484 if (len > XFS_XATTR_SIZE_MAX)
485 return -EINVAL;
487 kbuf = memdup_user(ubuf, len);
488 if (IS_ERR(kbuf))
489 return PTR_ERR(kbuf);
491 error = xfs_attr_set(XFS_I(inode), name, kbuf, len, flags);
492 if (!error)
493 xfs_forget_acl(inode, name, flags);
494 kfree(kbuf);
495 return error;
499 xfs_attrmulti_attr_remove(
500 struct inode *inode,
501 unsigned char *name,
502 uint32_t flags)
504 int error;
506 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
507 return -EPERM;
508 error = xfs_attr_remove(XFS_I(inode), name, flags);
509 if (!error)
510 xfs_forget_acl(inode, name, flags);
511 return error;
514 STATIC int
515 xfs_attrmulti_by_handle(
516 struct file *parfilp,
517 void __user *arg)
519 int error;
520 xfs_attr_multiop_t *ops;
521 xfs_fsop_attrmulti_handlereq_t am_hreq;
522 struct dentry *dentry;
523 unsigned int i, size;
524 unsigned char *attr_name;
526 if (!capable(CAP_SYS_ADMIN))
527 return -EPERM;
528 if (copy_from_user(&am_hreq, arg, sizeof(xfs_fsop_attrmulti_handlereq_t)))
529 return -EFAULT;
531 /* overflow check */
532 if (am_hreq.opcount >= INT_MAX / sizeof(xfs_attr_multiop_t))
533 return -E2BIG;
535 dentry = xfs_handlereq_to_dentry(parfilp, &am_hreq.hreq);
536 if (IS_ERR(dentry))
537 return PTR_ERR(dentry);
539 error = -E2BIG;
540 size = am_hreq.opcount * sizeof(xfs_attr_multiop_t);
541 if (!size || size > 16 * PAGE_SIZE)
542 goto out_dput;
544 ops = memdup_user(am_hreq.ops, size);
545 if (IS_ERR(ops)) {
546 error = PTR_ERR(ops);
547 goto out_dput;
550 error = -ENOMEM;
551 attr_name = kmalloc(MAXNAMELEN, GFP_KERNEL);
552 if (!attr_name)
553 goto out_kfree_ops;
555 error = 0;
556 for (i = 0; i < am_hreq.opcount; i++) {
557 ops[i].am_error = strncpy_from_user((char *)attr_name,
558 ops[i].am_attrname, MAXNAMELEN);
559 if (ops[i].am_error == 0 || ops[i].am_error == MAXNAMELEN)
560 error = -ERANGE;
561 if (ops[i].am_error < 0)
562 break;
564 switch (ops[i].am_opcode) {
565 case ATTR_OP_GET:
566 ops[i].am_error = xfs_attrmulti_attr_get(
567 d_inode(dentry), attr_name,
568 ops[i].am_attrvalue, &ops[i].am_length,
569 ops[i].am_flags);
570 break;
571 case ATTR_OP_SET:
572 ops[i].am_error = mnt_want_write_file(parfilp);
573 if (ops[i].am_error)
574 break;
575 ops[i].am_error = xfs_attrmulti_attr_set(
576 d_inode(dentry), attr_name,
577 ops[i].am_attrvalue, ops[i].am_length,
578 ops[i].am_flags);
579 mnt_drop_write_file(parfilp);
580 break;
581 case ATTR_OP_REMOVE:
582 ops[i].am_error = mnt_want_write_file(parfilp);
583 if (ops[i].am_error)
584 break;
585 ops[i].am_error = xfs_attrmulti_attr_remove(
586 d_inode(dentry), attr_name,
587 ops[i].am_flags);
588 mnt_drop_write_file(parfilp);
589 break;
590 default:
591 ops[i].am_error = -EINVAL;
595 if (copy_to_user(am_hreq.ops, ops, size))
596 error = -EFAULT;
598 kfree(attr_name);
599 out_kfree_ops:
600 kfree(ops);
601 out_dput:
602 dput(dentry);
603 return error;
607 xfs_ioc_space(
608 struct file *filp,
609 unsigned int cmd,
610 xfs_flock64_t *bf)
612 struct inode *inode = file_inode(filp);
613 struct xfs_inode *ip = XFS_I(inode);
614 struct iattr iattr;
615 enum xfs_prealloc_flags flags = 0;
616 uint iolock = XFS_IOLOCK_EXCL;
617 int error;
620 * Only allow the sys admin to reserve space unless
621 * unwritten extents are enabled.
623 if (!xfs_sb_version_hasextflgbit(&ip->i_mount->m_sb) &&
624 !capable(CAP_SYS_ADMIN))
625 return -EPERM;
627 if (inode->i_flags & (S_IMMUTABLE|S_APPEND))
628 return -EPERM;
630 if (!(filp->f_mode & FMODE_WRITE))
631 return -EBADF;
633 if (!S_ISREG(inode->i_mode))
634 return -EINVAL;
636 if (filp->f_flags & O_DSYNC)
637 flags |= XFS_PREALLOC_SYNC;
638 if (filp->f_mode & FMODE_NOCMTIME)
639 flags |= XFS_PREALLOC_INVISIBLE;
641 error = mnt_want_write_file(filp);
642 if (error)
643 return error;
645 xfs_ilock(ip, iolock);
646 error = xfs_break_layouts(inode, &iolock);
647 if (error)
648 goto out_unlock;
650 xfs_ilock(ip, XFS_MMAPLOCK_EXCL);
651 iolock |= XFS_MMAPLOCK_EXCL;
653 switch (bf->l_whence) {
654 case 0: /*SEEK_SET*/
655 break;
656 case 1: /*SEEK_CUR*/
657 bf->l_start += filp->f_pos;
658 break;
659 case 2: /*SEEK_END*/
660 bf->l_start += XFS_ISIZE(ip);
661 break;
662 default:
663 error = -EINVAL;
664 goto out_unlock;
668 * length of <= 0 for resv/unresv/zero is invalid. length for
669 * alloc/free is ignored completely and we have no idea what userspace
670 * might have set it to, so set it to zero to allow range
671 * checks to pass.
673 switch (cmd) {
674 case XFS_IOC_ZERO_RANGE:
675 case XFS_IOC_RESVSP:
676 case XFS_IOC_RESVSP64:
677 case XFS_IOC_UNRESVSP:
678 case XFS_IOC_UNRESVSP64:
679 if (bf->l_len <= 0) {
680 error = -EINVAL;
681 goto out_unlock;
683 break;
684 default:
685 bf->l_len = 0;
686 break;
689 if (bf->l_start < 0 ||
690 bf->l_start > inode->i_sb->s_maxbytes ||
691 bf->l_start + bf->l_len < 0 ||
692 bf->l_start + bf->l_len >= inode->i_sb->s_maxbytes) {
693 error = -EINVAL;
694 goto out_unlock;
697 switch (cmd) {
698 case XFS_IOC_ZERO_RANGE:
699 flags |= XFS_PREALLOC_SET;
700 error = xfs_zero_file_space(ip, bf->l_start, bf->l_len);
701 break;
702 case XFS_IOC_RESVSP:
703 case XFS_IOC_RESVSP64:
704 flags |= XFS_PREALLOC_SET;
705 error = xfs_alloc_file_space(ip, bf->l_start, bf->l_len,
706 XFS_BMAPI_PREALLOC);
707 break;
708 case XFS_IOC_UNRESVSP:
709 case XFS_IOC_UNRESVSP64:
710 error = xfs_free_file_space(ip, bf->l_start, bf->l_len);
711 break;
712 case XFS_IOC_ALLOCSP:
713 case XFS_IOC_ALLOCSP64:
714 case XFS_IOC_FREESP:
715 case XFS_IOC_FREESP64:
716 flags |= XFS_PREALLOC_CLEAR;
717 if (bf->l_start > XFS_ISIZE(ip)) {
718 error = xfs_alloc_file_space(ip, XFS_ISIZE(ip),
719 bf->l_start - XFS_ISIZE(ip), 0);
720 if (error)
721 goto out_unlock;
724 iattr.ia_valid = ATTR_SIZE;
725 iattr.ia_size = bf->l_start;
727 error = xfs_vn_setattr_size(file_dentry(filp), &iattr);
728 break;
729 default:
730 ASSERT(0);
731 error = -EINVAL;
734 if (error)
735 goto out_unlock;
737 error = xfs_update_prealloc_flags(ip, flags);
739 out_unlock:
740 xfs_iunlock(ip, iolock);
741 mnt_drop_write_file(filp);
742 return error;
745 STATIC int
746 xfs_ioc_bulkstat(
747 xfs_mount_t *mp,
748 unsigned int cmd,
749 void __user *arg)
751 xfs_fsop_bulkreq_t bulkreq;
752 int count; /* # of records returned */
753 xfs_ino_t inlast; /* last inode number */
754 int done;
755 int error;
757 /* done = 1 if there are more stats to get and if bulkstat */
758 /* should be called again (unused here, but used in dmapi) */
760 if (!capable(CAP_SYS_ADMIN))
761 return -EPERM;
763 if (XFS_FORCED_SHUTDOWN(mp))
764 return -EIO;
766 if (copy_from_user(&bulkreq, arg, sizeof(xfs_fsop_bulkreq_t)))
767 return -EFAULT;
769 if (copy_from_user(&inlast, bulkreq.lastip, sizeof(__s64)))
770 return -EFAULT;
772 if ((count = bulkreq.icount) <= 0)
773 return -EINVAL;
775 if (bulkreq.ubuffer == NULL)
776 return -EINVAL;
778 if (cmd == XFS_IOC_FSINUMBERS)
779 error = xfs_inumbers(mp, &inlast, &count,
780 bulkreq.ubuffer, xfs_inumbers_fmt);
781 else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE)
782 error = xfs_bulkstat_one(mp, inlast, bulkreq.ubuffer,
783 sizeof(xfs_bstat_t), NULL, &done);
784 else /* XFS_IOC_FSBULKSTAT */
785 error = xfs_bulkstat(mp, &inlast, &count, xfs_bulkstat_one,
786 sizeof(xfs_bstat_t), bulkreq.ubuffer,
787 &done);
789 if (error)
790 return error;
792 if (bulkreq.ocount != NULL) {
793 if (copy_to_user(bulkreq.lastip, &inlast,
794 sizeof(xfs_ino_t)))
795 return -EFAULT;
797 if (copy_to_user(bulkreq.ocount, &count, sizeof(count)))
798 return -EFAULT;
801 return 0;
804 STATIC int
805 xfs_ioc_fsgeometry_v1(
806 xfs_mount_t *mp,
807 void __user *arg)
809 xfs_fsop_geom_t fsgeo;
810 int error;
812 error = xfs_fs_geometry(mp, &fsgeo, 3);
813 if (error)
814 return error;
817 * Caller should have passed an argument of type
818 * xfs_fsop_geom_v1_t. This is a proper subset of the
819 * xfs_fsop_geom_t that xfs_fs_geometry() fills in.
821 if (copy_to_user(arg, &fsgeo, sizeof(xfs_fsop_geom_v1_t)))
822 return -EFAULT;
823 return 0;
826 STATIC int
827 xfs_ioc_fsgeometry(
828 xfs_mount_t *mp,
829 void __user *arg)
831 xfs_fsop_geom_t fsgeo;
832 int error;
834 error = xfs_fs_geometry(mp, &fsgeo, 4);
835 if (error)
836 return error;
838 if (copy_to_user(arg, &fsgeo, sizeof(fsgeo)))
839 return -EFAULT;
840 return 0;
844 * Linux extended inode flags interface.
847 STATIC unsigned int
848 xfs_merge_ioc_xflags(
849 unsigned int flags,
850 unsigned int start)
852 unsigned int xflags = start;
854 if (flags & FS_IMMUTABLE_FL)
855 xflags |= FS_XFLAG_IMMUTABLE;
856 else
857 xflags &= ~FS_XFLAG_IMMUTABLE;
858 if (flags & FS_APPEND_FL)
859 xflags |= FS_XFLAG_APPEND;
860 else
861 xflags &= ~FS_XFLAG_APPEND;
862 if (flags & FS_SYNC_FL)
863 xflags |= FS_XFLAG_SYNC;
864 else
865 xflags &= ~FS_XFLAG_SYNC;
866 if (flags & FS_NOATIME_FL)
867 xflags |= FS_XFLAG_NOATIME;
868 else
869 xflags &= ~FS_XFLAG_NOATIME;
870 if (flags & FS_NODUMP_FL)
871 xflags |= FS_XFLAG_NODUMP;
872 else
873 xflags &= ~FS_XFLAG_NODUMP;
875 return xflags;
878 STATIC unsigned int
879 xfs_di2lxflags(
880 uint16_t di_flags)
882 unsigned int flags = 0;
884 if (di_flags & XFS_DIFLAG_IMMUTABLE)
885 flags |= FS_IMMUTABLE_FL;
886 if (di_flags & XFS_DIFLAG_APPEND)
887 flags |= FS_APPEND_FL;
888 if (di_flags & XFS_DIFLAG_SYNC)
889 flags |= FS_SYNC_FL;
890 if (di_flags & XFS_DIFLAG_NOATIME)
891 flags |= FS_NOATIME_FL;
892 if (di_flags & XFS_DIFLAG_NODUMP)
893 flags |= FS_NODUMP_FL;
894 return flags;
897 STATIC int
898 xfs_ioc_fsgetxattr(
899 xfs_inode_t *ip,
900 int attr,
901 void __user *arg)
903 struct fsxattr fa;
905 memset(&fa, 0, sizeof(struct fsxattr));
907 xfs_ilock(ip, XFS_ILOCK_SHARED);
908 fa.fsx_xflags = xfs_ip2xflags(ip);
909 fa.fsx_extsize = ip->i_d.di_extsize << ip->i_mount->m_sb.sb_blocklog;
910 fa.fsx_cowextsize = ip->i_d.di_cowextsize <<
911 ip->i_mount->m_sb.sb_blocklog;
912 fa.fsx_projid = xfs_get_projid(ip);
914 if (attr) {
915 if (ip->i_afp) {
916 if (ip->i_afp->if_flags & XFS_IFEXTENTS)
917 fa.fsx_nextents = xfs_iext_count(ip->i_afp);
918 else
919 fa.fsx_nextents = ip->i_d.di_anextents;
920 } else
921 fa.fsx_nextents = 0;
922 } else {
923 if (ip->i_df.if_flags & XFS_IFEXTENTS)
924 fa.fsx_nextents = xfs_iext_count(&ip->i_df);
925 else
926 fa.fsx_nextents = ip->i_d.di_nextents;
928 xfs_iunlock(ip, XFS_ILOCK_SHARED);
930 if (copy_to_user(arg, &fa, sizeof(fa)))
931 return -EFAULT;
932 return 0;
935 STATIC uint16_t
936 xfs_flags2diflags(
937 struct xfs_inode *ip,
938 unsigned int xflags)
940 /* can't set PREALLOC this way, just preserve it */
941 uint16_t di_flags =
942 (ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
944 if (xflags & FS_XFLAG_IMMUTABLE)
945 di_flags |= XFS_DIFLAG_IMMUTABLE;
946 if (xflags & FS_XFLAG_APPEND)
947 di_flags |= XFS_DIFLAG_APPEND;
948 if (xflags & FS_XFLAG_SYNC)
949 di_flags |= XFS_DIFLAG_SYNC;
950 if (xflags & FS_XFLAG_NOATIME)
951 di_flags |= XFS_DIFLAG_NOATIME;
952 if (xflags & FS_XFLAG_NODUMP)
953 di_flags |= XFS_DIFLAG_NODUMP;
954 if (xflags & FS_XFLAG_NODEFRAG)
955 di_flags |= XFS_DIFLAG_NODEFRAG;
956 if (xflags & FS_XFLAG_FILESTREAM)
957 di_flags |= XFS_DIFLAG_FILESTREAM;
958 if (S_ISDIR(VFS_I(ip)->i_mode)) {
959 if (xflags & FS_XFLAG_RTINHERIT)
960 di_flags |= XFS_DIFLAG_RTINHERIT;
961 if (xflags & FS_XFLAG_NOSYMLINKS)
962 di_flags |= XFS_DIFLAG_NOSYMLINKS;
963 if (xflags & FS_XFLAG_EXTSZINHERIT)
964 di_flags |= XFS_DIFLAG_EXTSZINHERIT;
965 if (xflags & FS_XFLAG_PROJINHERIT)
966 di_flags |= XFS_DIFLAG_PROJINHERIT;
967 } else if (S_ISREG(VFS_I(ip)->i_mode)) {
968 if (xflags & FS_XFLAG_REALTIME)
969 di_flags |= XFS_DIFLAG_REALTIME;
970 if (xflags & FS_XFLAG_EXTSIZE)
971 di_flags |= XFS_DIFLAG_EXTSIZE;
974 return di_flags;
977 STATIC uint64_t
978 xfs_flags2diflags2(
979 struct xfs_inode *ip,
980 unsigned int xflags)
982 uint64_t di_flags2 =
983 (ip->i_d.di_flags2 & XFS_DIFLAG2_REFLINK);
985 if (xflags & FS_XFLAG_DAX)
986 di_flags2 |= XFS_DIFLAG2_DAX;
987 if (xflags & FS_XFLAG_COWEXTSIZE)
988 di_flags2 |= XFS_DIFLAG2_COWEXTSIZE;
990 return di_flags2;
993 STATIC void
994 xfs_diflags_to_linux(
995 struct xfs_inode *ip)
997 struct inode *inode = VFS_I(ip);
998 unsigned int xflags = xfs_ip2xflags(ip);
1000 if (xflags & FS_XFLAG_IMMUTABLE)
1001 inode->i_flags |= S_IMMUTABLE;
1002 else
1003 inode->i_flags &= ~S_IMMUTABLE;
1004 if (xflags & FS_XFLAG_APPEND)
1005 inode->i_flags |= S_APPEND;
1006 else
1007 inode->i_flags &= ~S_APPEND;
1008 if (xflags & FS_XFLAG_SYNC)
1009 inode->i_flags |= S_SYNC;
1010 else
1011 inode->i_flags &= ~S_SYNC;
1012 if (xflags & FS_XFLAG_NOATIME)
1013 inode->i_flags |= S_NOATIME;
1014 else
1015 inode->i_flags &= ~S_NOATIME;
1016 #if 0 /* disabled until the flag switching races are sorted out */
1017 if (xflags & FS_XFLAG_DAX)
1018 inode->i_flags |= S_DAX;
1019 else
1020 inode->i_flags &= ~S_DAX;
1021 #endif
1024 static int
1025 xfs_ioctl_setattr_xflags(
1026 struct xfs_trans *tp,
1027 struct xfs_inode *ip,
1028 struct fsxattr *fa)
1030 struct xfs_mount *mp = ip->i_mount;
1031 uint64_t di_flags2;
1033 /* Can't change realtime flag if any extents are allocated. */
1034 if ((ip->i_d.di_nextents || ip->i_delayed_blks) &&
1035 XFS_IS_REALTIME_INODE(ip) != (fa->fsx_xflags & FS_XFLAG_REALTIME))
1036 return -EINVAL;
1038 /* If realtime flag is set then must have realtime device */
1039 if (fa->fsx_xflags & FS_XFLAG_REALTIME) {
1040 if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
1041 (ip->i_d.di_extsize % mp->m_sb.sb_rextsize))
1042 return -EINVAL;
1045 /* Clear reflink if we are actually able to set the rt flag. */
1046 if ((fa->fsx_xflags & FS_XFLAG_REALTIME) && xfs_is_reflink_inode(ip))
1047 ip->i_d.di_flags2 &= ~XFS_DIFLAG2_REFLINK;
1049 /* Don't allow us to set DAX mode for a reflinked file for now. */
1050 if ((fa->fsx_xflags & FS_XFLAG_DAX) && xfs_is_reflink_inode(ip))
1051 return -EINVAL;
1054 * Can't modify an immutable/append-only file unless
1055 * we have appropriate permission.
1057 if (((ip->i_d.di_flags & (XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND)) ||
1058 (fa->fsx_xflags & (FS_XFLAG_IMMUTABLE | FS_XFLAG_APPEND))) &&
1059 !capable(CAP_LINUX_IMMUTABLE))
1060 return -EPERM;
1062 /* diflags2 only valid for v3 inodes. */
1063 di_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1064 if (di_flags2 && ip->i_d.di_version < 3)
1065 return -EINVAL;
1067 ip->i_d.di_flags = xfs_flags2diflags(ip, fa->fsx_xflags);
1068 ip->i_d.di_flags2 = di_flags2;
1070 xfs_diflags_to_linux(ip);
1071 xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1072 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1073 XFS_STATS_INC(mp, xs_ig_attrchg);
1074 return 0;
1078 * If we are changing DAX flags, we have to ensure the file is clean and any
1079 * cached objects in the address space are invalidated and removed. This
1080 * requires us to lock out other IO and page faults similar to a truncate
1081 * operation. The locks need to be held until the transaction has been committed
1082 * so that the cache invalidation is atomic with respect to the DAX flag
1083 * manipulation.
1085 static int
1086 xfs_ioctl_setattr_dax_invalidate(
1087 struct xfs_inode *ip,
1088 struct fsxattr *fa,
1089 int *join_flags)
1091 struct inode *inode = VFS_I(ip);
1092 struct super_block *sb = inode->i_sb;
1093 int error;
1095 *join_flags = 0;
1098 * It is only valid to set the DAX flag on regular files and
1099 * directories on filesystems where the block size is equal to the page
1100 * size. On directories it serves as an inherit hint.
1102 if (fa->fsx_xflags & FS_XFLAG_DAX) {
1103 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)))
1104 return -EINVAL;
1105 if (bdev_dax_supported(sb, sb->s_blocksize) < 0)
1106 return -EINVAL;
1109 /* If the DAX state is not changing, we have nothing to do here. */
1110 if ((fa->fsx_xflags & FS_XFLAG_DAX) && IS_DAX(inode))
1111 return 0;
1112 if (!(fa->fsx_xflags & FS_XFLAG_DAX) && !IS_DAX(inode))
1113 return 0;
1115 /* lock, flush and invalidate mapping in preparation for flag change */
1116 xfs_ilock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1117 error = filemap_write_and_wait(inode->i_mapping);
1118 if (error)
1119 goto out_unlock;
1120 error = invalidate_inode_pages2(inode->i_mapping);
1121 if (error)
1122 goto out_unlock;
1124 *join_flags = XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL;
1125 return 0;
1127 out_unlock:
1128 xfs_iunlock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1129 return error;
1134 * Set up the transaction structure for the setattr operation, checking that we
1135 * have permission to do so. On success, return a clean transaction and the
1136 * inode locked exclusively ready for further operation specific checks. On
1137 * failure, return an error without modifying or locking the inode.
1139 * The inode might already be IO locked on call. If this is the case, it is
1140 * indicated in @join_flags and we take full responsibility for ensuring they
1141 * are unlocked from now on. Hence if we have an error here, we still have to
1142 * unlock them. Otherwise, once they are joined to the transaction, they will
1143 * be unlocked on commit/cancel.
1145 static struct xfs_trans *
1146 xfs_ioctl_setattr_get_trans(
1147 struct xfs_inode *ip,
1148 int join_flags)
1150 struct xfs_mount *mp = ip->i_mount;
1151 struct xfs_trans *tp;
1152 int error = -EROFS;
1154 if (mp->m_flags & XFS_MOUNT_RDONLY)
1155 goto out_unlock;
1156 error = -EIO;
1157 if (XFS_FORCED_SHUTDOWN(mp))
1158 goto out_unlock;
1160 error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
1161 if (error)
1162 return ERR_PTR(error);
1164 xfs_ilock(ip, XFS_ILOCK_EXCL);
1165 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | join_flags);
1166 join_flags = 0;
1169 * CAP_FOWNER overrides the following restrictions:
1171 * The user ID of the calling process must be equal to the file owner
1172 * ID, except in cases where the CAP_FSETID capability is applicable.
1174 if (!inode_owner_or_capable(VFS_I(ip))) {
1175 error = -EPERM;
1176 goto out_cancel;
1179 if (mp->m_flags & XFS_MOUNT_WSYNC)
1180 xfs_trans_set_sync(tp);
1182 return tp;
1184 out_cancel:
1185 xfs_trans_cancel(tp);
1186 out_unlock:
1187 if (join_flags)
1188 xfs_iunlock(ip, join_flags);
1189 return ERR_PTR(error);
1193 * extent size hint validation is somewhat cumbersome. Rules are:
1195 * 1. extent size hint is only valid for directories and regular files
1196 * 2. FS_XFLAG_EXTSIZE is only valid for regular files
1197 * 3. FS_XFLAG_EXTSZINHERIT is only valid for directories.
1198 * 4. can only be changed on regular files if no extents are allocated
1199 * 5. can be changed on directories at any time
1200 * 6. extsize hint of 0 turns off hints, clears inode flags.
1201 * 7. Extent size must be a multiple of the appropriate block size.
1202 * 8. for non-realtime files, the extent size hint must be limited
1203 * to half the AG size to avoid alignment extending the extent beyond the
1204 * limits of the AG.
1206 * Please keep this function in sync with xfs_scrub_inode_extsize.
1208 static int
1209 xfs_ioctl_setattr_check_extsize(
1210 struct xfs_inode *ip,
1211 struct fsxattr *fa)
1213 struct xfs_mount *mp = ip->i_mount;
1215 if ((fa->fsx_xflags & FS_XFLAG_EXTSIZE) && !S_ISREG(VFS_I(ip)->i_mode))
1216 return -EINVAL;
1218 if ((fa->fsx_xflags & FS_XFLAG_EXTSZINHERIT) &&
1219 !S_ISDIR(VFS_I(ip)->i_mode))
1220 return -EINVAL;
1222 if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_d.di_nextents &&
1223 ((ip->i_d.di_extsize << mp->m_sb.sb_blocklog) != fa->fsx_extsize))
1224 return -EINVAL;
1226 if (fa->fsx_extsize != 0) {
1227 xfs_extlen_t size;
1228 xfs_fsblock_t extsize_fsb;
1230 extsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1231 if (extsize_fsb > MAXEXTLEN)
1232 return -EINVAL;
1234 if (XFS_IS_REALTIME_INODE(ip) ||
1235 (fa->fsx_xflags & FS_XFLAG_REALTIME)) {
1236 size = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
1237 } else {
1238 size = mp->m_sb.sb_blocksize;
1239 if (extsize_fsb > mp->m_sb.sb_agblocks / 2)
1240 return -EINVAL;
1243 if (fa->fsx_extsize % size)
1244 return -EINVAL;
1245 } else
1246 fa->fsx_xflags &= ~(FS_XFLAG_EXTSIZE | FS_XFLAG_EXTSZINHERIT);
1248 return 0;
1252 * CoW extent size hint validation rules are:
1254 * 1. CoW extent size hint can only be set if reflink is enabled on the fs.
1255 * The inode does not have to have any shared blocks, but it must be a v3.
1256 * 2. FS_XFLAG_COWEXTSIZE is only valid for directories and regular files;
1257 * for a directory, the hint is propagated to new files.
1258 * 3. Can be changed on files & directories at any time.
1259 * 4. CoW extsize hint of 0 turns off hints, clears inode flags.
1260 * 5. Extent size must be a multiple of the appropriate block size.
1261 * 6. The extent size hint must be limited to half the AG size to avoid
1262 * alignment extending the extent beyond the limits of the AG.
1264 * Please keep this function in sync with xfs_scrub_inode_cowextsize.
1266 static int
1267 xfs_ioctl_setattr_check_cowextsize(
1268 struct xfs_inode *ip,
1269 struct fsxattr *fa)
1271 struct xfs_mount *mp = ip->i_mount;
1273 if (!(fa->fsx_xflags & FS_XFLAG_COWEXTSIZE))
1274 return 0;
1276 if (!xfs_sb_version_hasreflink(&ip->i_mount->m_sb) ||
1277 ip->i_d.di_version != 3)
1278 return -EINVAL;
1280 if (!S_ISREG(VFS_I(ip)->i_mode) && !S_ISDIR(VFS_I(ip)->i_mode))
1281 return -EINVAL;
1283 if (fa->fsx_cowextsize != 0) {
1284 xfs_extlen_t size;
1285 xfs_fsblock_t cowextsize_fsb;
1287 cowextsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
1288 if (cowextsize_fsb > MAXEXTLEN)
1289 return -EINVAL;
1291 size = mp->m_sb.sb_blocksize;
1292 if (cowextsize_fsb > mp->m_sb.sb_agblocks / 2)
1293 return -EINVAL;
1295 if (fa->fsx_cowextsize % size)
1296 return -EINVAL;
1297 } else
1298 fa->fsx_xflags &= ~FS_XFLAG_COWEXTSIZE;
1300 return 0;
1303 static int
1304 xfs_ioctl_setattr_check_projid(
1305 struct xfs_inode *ip,
1306 struct fsxattr *fa)
1308 /* Disallow 32bit project ids if projid32bit feature is not enabled. */
1309 if (fa->fsx_projid > (uint16_t)-1 &&
1310 !xfs_sb_version_hasprojid32bit(&ip->i_mount->m_sb))
1311 return -EINVAL;
1314 * Project Quota ID state is only allowed to change from within the init
1315 * namespace. Enforce that restriction only if we are trying to change
1316 * the quota ID state. Everything else is allowed in user namespaces.
1318 if (current_user_ns() == &init_user_ns)
1319 return 0;
1321 if (xfs_get_projid(ip) != fa->fsx_projid)
1322 return -EINVAL;
1323 if ((fa->fsx_xflags & FS_XFLAG_PROJINHERIT) !=
1324 (ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT))
1325 return -EINVAL;
1327 return 0;
1330 STATIC int
1331 xfs_ioctl_setattr(
1332 xfs_inode_t *ip,
1333 struct fsxattr *fa)
1335 struct xfs_mount *mp = ip->i_mount;
1336 struct xfs_trans *tp;
1337 struct xfs_dquot *udqp = NULL;
1338 struct xfs_dquot *pdqp = NULL;
1339 struct xfs_dquot *olddquot = NULL;
1340 int code;
1341 int join_flags = 0;
1343 trace_xfs_ioctl_setattr(ip);
1345 code = xfs_ioctl_setattr_check_projid(ip, fa);
1346 if (code)
1347 return code;
1350 * If disk quotas is on, we make sure that the dquots do exist on disk,
1351 * before we start any other transactions. Trying to do this later
1352 * is messy. We don't care to take a readlock to look at the ids
1353 * in inode here, because we can't hold it across the trans_reserve.
1354 * If the IDs do change before we take the ilock, we're covered
1355 * because the i_*dquot fields will get updated anyway.
1357 if (XFS_IS_QUOTA_ON(mp)) {
1358 code = xfs_qm_vop_dqalloc(ip, ip->i_d.di_uid,
1359 ip->i_d.di_gid, fa->fsx_projid,
1360 XFS_QMOPT_PQUOTA, &udqp, NULL, &pdqp);
1361 if (code)
1362 return code;
1366 * Changing DAX config may require inode locking for mapping
1367 * invalidation. These need to be held all the way to transaction commit
1368 * or cancel time, so need to be passed through to
1369 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1370 * appropriately.
1372 code = xfs_ioctl_setattr_dax_invalidate(ip, fa, &join_flags);
1373 if (code)
1374 goto error_free_dquots;
1376 tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1377 if (IS_ERR(tp)) {
1378 code = PTR_ERR(tp);
1379 goto error_free_dquots;
1383 if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp) &&
1384 xfs_get_projid(ip) != fa->fsx_projid) {
1385 code = xfs_qm_vop_chown_reserve(tp, ip, udqp, NULL, pdqp,
1386 capable(CAP_FOWNER) ? XFS_QMOPT_FORCE_RES : 0);
1387 if (code) /* out of quota */
1388 goto error_trans_cancel;
1391 code = xfs_ioctl_setattr_check_extsize(ip, fa);
1392 if (code)
1393 goto error_trans_cancel;
1395 code = xfs_ioctl_setattr_check_cowextsize(ip, fa);
1396 if (code)
1397 goto error_trans_cancel;
1399 code = xfs_ioctl_setattr_xflags(tp, ip, fa);
1400 if (code)
1401 goto error_trans_cancel;
1404 * Change file ownership. Must be the owner or privileged. CAP_FSETID
1405 * overrides the following restrictions:
1407 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1408 * successful return from chown()
1411 if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
1412 !capable_wrt_inode_uidgid(VFS_I(ip), CAP_FSETID))
1413 VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
1415 /* Change the ownerships and register project quota modifications */
1416 if (xfs_get_projid(ip) != fa->fsx_projid) {
1417 if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp)) {
1418 olddquot = xfs_qm_vop_chown(tp, ip,
1419 &ip->i_pdquot, pdqp);
1421 ASSERT(ip->i_d.di_version > 1);
1422 xfs_set_projid(ip, fa->fsx_projid);
1426 * Only set the extent size hint if we've already determined that the
1427 * extent size hint should be set on the inode. If no extent size flags
1428 * are set on the inode then unconditionally clear the extent size hint.
1430 if (ip->i_d.di_flags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1431 ip->i_d.di_extsize = fa->fsx_extsize >> mp->m_sb.sb_blocklog;
1432 else
1433 ip->i_d.di_extsize = 0;
1434 if (ip->i_d.di_version == 3 &&
1435 (ip->i_d.di_flags2 & XFS_DIFLAG2_COWEXTSIZE))
1436 ip->i_d.di_cowextsize = fa->fsx_cowextsize >>
1437 mp->m_sb.sb_blocklog;
1438 else
1439 ip->i_d.di_cowextsize = 0;
1441 code = xfs_trans_commit(tp);
1444 * Release any dquot(s) the inode had kept before chown.
1446 xfs_qm_dqrele(olddquot);
1447 xfs_qm_dqrele(udqp);
1448 xfs_qm_dqrele(pdqp);
1450 return code;
1452 error_trans_cancel:
1453 xfs_trans_cancel(tp);
1454 error_free_dquots:
1455 xfs_qm_dqrele(udqp);
1456 xfs_qm_dqrele(pdqp);
1457 return code;
1460 STATIC int
1461 xfs_ioc_fssetxattr(
1462 xfs_inode_t *ip,
1463 struct file *filp,
1464 void __user *arg)
1466 struct fsxattr fa;
1467 int error;
1469 if (copy_from_user(&fa, arg, sizeof(fa)))
1470 return -EFAULT;
1472 error = mnt_want_write_file(filp);
1473 if (error)
1474 return error;
1475 error = xfs_ioctl_setattr(ip, &fa);
1476 mnt_drop_write_file(filp);
1477 return error;
1480 STATIC int
1481 xfs_ioc_getxflags(
1482 xfs_inode_t *ip,
1483 void __user *arg)
1485 unsigned int flags;
1487 flags = xfs_di2lxflags(ip->i_d.di_flags);
1488 if (copy_to_user(arg, &flags, sizeof(flags)))
1489 return -EFAULT;
1490 return 0;
1493 STATIC int
1494 xfs_ioc_setxflags(
1495 struct xfs_inode *ip,
1496 struct file *filp,
1497 void __user *arg)
1499 struct xfs_trans *tp;
1500 struct fsxattr fa;
1501 unsigned int flags;
1502 int join_flags = 0;
1503 int error;
1505 if (copy_from_user(&flags, arg, sizeof(flags)))
1506 return -EFAULT;
1508 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
1509 FS_NOATIME_FL | FS_NODUMP_FL | \
1510 FS_SYNC_FL))
1511 return -EOPNOTSUPP;
1513 fa.fsx_xflags = xfs_merge_ioc_xflags(flags, xfs_ip2xflags(ip));
1515 error = mnt_want_write_file(filp);
1516 if (error)
1517 return error;
1520 * Changing DAX config may require inode locking for mapping
1521 * invalidation. These need to be held all the way to transaction commit
1522 * or cancel time, so need to be passed through to
1523 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1524 * appropriately.
1526 error = xfs_ioctl_setattr_dax_invalidate(ip, &fa, &join_flags);
1527 if (error)
1528 goto out_drop_write;
1530 tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1531 if (IS_ERR(tp)) {
1532 error = PTR_ERR(tp);
1533 goto out_drop_write;
1536 error = xfs_ioctl_setattr_xflags(tp, ip, &fa);
1537 if (error) {
1538 xfs_trans_cancel(tp);
1539 goto out_drop_write;
1542 error = xfs_trans_commit(tp);
1543 out_drop_write:
1544 mnt_drop_write_file(filp);
1545 return error;
1548 static bool
1549 xfs_getbmap_format(
1550 struct kgetbmap *p,
1551 struct getbmapx __user *u,
1552 size_t recsize)
1554 if (put_user(p->bmv_offset, &u->bmv_offset) ||
1555 put_user(p->bmv_block, &u->bmv_block) ||
1556 put_user(p->bmv_length, &u->bmv_length) ||
1557 put_user(0, &u->bmv_count) ||
1558 put_user(0, &u->bmv_entries))
1559 return false;
1560 if (recsize < sizeof(struct getbmapx))
1561 return true;
1562 if (put_user(0, &u->bmv_iflags) ||
1563 put_user(p->bmv_oflags, &u->bmv_oflags) ||
1564 put_user(0, &u->bmv_unused1) ||
1565 put_user(0, &u->bmv_unused2))
1566 return false;
1567 return true;
1570 STATIC int
1571 xfs_ioc_getbmap(
1572 struct file *file,
1573 unsigned int cmd,
1574 void __user *arg)
1576 struct getbmapx bmx = { 0 };
1577 struct kgetbmap *buf;
1578 size_t recsize;
1579 int error, i;
1581 switch (cmd) {
1582 case XFS_IOC_GETBMAPA:
1583 bmx.bmv_iflags = BMV_IF_ATTRFORK;
1584 /*FALLTHRU*/
1585 case XFS_IOC_GETBMAP:
1586 if (file->f_mode & FMODE_NOCMTIME)
1587 bmx.bmv_iflags |= BMV_IF_NO_DMAPI_READ;
1588 /* struct getbmap is a strict subset of struct getbmapx. */
1589 recsize = sizeof(struct getbmap);
1590 break;
1591 case XFS_IOC_GETBMAPX:
1592 recsize = sizeof(struct getbmapx);
1593 break;
1594 default:
1595 return -EINVAL;
1598 if (copy_from_user(&bmx, arg, recsize))
1599 return -EFAULT;
1601 if (bmx.bmv_count < 2)
1602 return -EINVAL;
1603 if (bmx.bmv_count > ULONG_MAX / recsize)
1604 return -ENOMEM;
1606 buf = kmem_zalloc_large(bmx.bmv_count * sizeof(*buf), 0);
1607 if (!buf)
1608 return -ENOMEM;
1610 error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, buf);
1611 if (error)
1612 goto out_free_buf;
1614 error = -EFAULT;
1615 if (copy_to_user(arg, &bmx, recsize))
1616 goto out_free_buf;
1617 arg += recsize;
1619 for (i = 0; i < bmx.bmv_entries; i++) {
1620 if (!xfs_getbmap_format(buf + i, arg, recsize))
1621 goto out_free_buf;
1622 arg += recsize;
1625 error = 0;
1626 out_free_buf:
1627 kmem_free(buf);
1628 return 0;
1631 struct getfsmap_info {
1632 struct xfs_mount *mp;
1633 struct fsmap_head __user *data;
1634 unsigned int idx;
1635 __u32 last_flags;
1638 STATIC int
1639 xfs_getfsmap_format(struct xfs_fsmap *xfm, void *priv)
1641 struct getfsmap_info *info = priv;
1642 struct fsmap fm;
1644 trace_xfs_getfsmap_mapping(info->mp, xfm);
1646 info->last_flags = xfm->fmr_flags;
1647 xfs_fsmap_from_internal(&fm, xfm);
1648 if (copy_to_user(&info->data->fmh_recs[info->idx++], &fm,
1649 sizeof(struct fsmap)))
1650 return -EFAULT;
1652 return 0;
1655 STATIC int
1656 xfs_ioc_getfsmap(
1657 struct xfs_inode *ip,
1658 struct fsmap_head __user *arg)
1660 struct getfsmap_info info = { NULL };
1661 struct xfs_fsmap_head xhead = {0};
1662 struct fsmap_head head;
1663 bool aborted = false;
1664 int error;
1666 if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
1667 return -EFAULT;
1668 if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
1669 memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
1670 sizeof(head.fmh_keys[0].fmr_reserved)) ||
1671 memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
1672 sizeof(head.fmh_keys[1].fmr_reserved)))
1673 return -EINVAL;
1675 xhead.fmh_iflags = head.fmh_iflags;
1676 xhead.fmh_count = head.fmh_count;
1677 xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
1678 xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
1680 trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1681 trace_xfs_getfsmap_high_key(ip->i_mount, &xhead.fmh_keys[1]);
1683 info.mp = ip->i_mount;
1684 info.data = arg;
1685 error = xfs_getfsmap(ip->i_mount, &xhead, xfs_getfsmap_format, &info);
1686 if (error == XFS_BTREE_QUERY_RANGE_ABORT) {
1687 error = 0;
1688 aborted = true;
1689 } else if (error)
1690 return error;
1692 /* If we didn't abort, set the "last" flag in the last fmx */
1693 if (!aborted && info.idx) {
1694 info.last_flags |= FMR_OF_LAST;
1695 if (copy_to_user(&info.data->fmh_recs[info.idx - 1].fmr_flags,
1696 &info.last_flags, sizeof(info.last_flags)))
1697 return -EFAULT;
1700 /* copy back header */
1701 head.fmh_entries = xhead.fmh_entries;
1702 head.fmh_oflags = xhead.fmh_oflags;
1703 if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
1704 return -EFAULT;
1706 return 0;
1709 STATIC int
1710 xfs_ioc_scrub_metadata(
1711 struct xfs_inode *ip,
1712 void __user *arg)
1714 struct xfs_scrub_metadata scrub;
1715 int error;
1717 if (!capable(CAP_SYS_ADMIN))
1718 return -EPERM;
1720 if (copy_from_user(&scrub, arg, sizeof(scrub)))
1721 return -EFAULT;
1723 error = xfs_scrub_metadata(ip, &scrub);
1724 if (error)
1725 return error;
1727 if (copy_to_user(arg, &scrub, sizeof(scrub)))
1728 return -EFAULT;
1730 return 0;
1734 xfs_ioc_swapext(
1735 xfs_swapext_t *sxp)
1737 xfs_inode_t *ip, *tip;
1738 struct fd f, tmp;
1739 int error = 0;
1741 /* Pull information for the target fd */
1742 f = fdget((int)sxp->sx_fdtarget);
1743 if (!f.file) {
1744 error = -EINVAL;
1745 goto out;
1748 if (!(f.file->f_mode & FMODE_WRITE) ||
1749 !(f.file->f_mode & FMODE_READ) ||
1750 (f.file->f_flags & O_APPEND)) {
1751 error = -EBADF;
1752 goto out_put_file;
1755 tmp = fdget((int)sxp->sx_fdtmp);
1756 if (!tmp.file) {
1757 error = -EINVAL;
1758 goto out_put_file;
1761 if (!(tmp.file->f_mode & FMODE_WRITE) ||
1762 !(tmp.file->f_mode & FMODE_READ) ||
1763 (tmp.file->f_flags & O_APPEND)) {
1764 error = -EBADF;
1765 goto out_put_tmp_file;
1768 if (IS_SWAPFILE(file_inode(f.file)) ||
1769 IS_SWAPFILE(file_inode(tmp.file))) {
1770 error = -EINVAL;
1771 goto out_put_tmp_file;
1775 * We need to ensure that the fds passed in point to XFS inodes
1776 * before we cast and access them as XFS structures as we have no
1777 * control over what the user passes us here.
1779 if (f.file->f_op != &xfs_file_operations ||
1780 tmp.file->f_op != &xfs_file_operations) {
1781 error = -EINVAL;
1782 goto out_put_tmp_file;
1785 ip = XFS_I(file_inode(f.file));
1786 tip = XFS_I(file_inode(tmp.file));
1788 if (ip->i_mount != tip->i_mount) {
1789 error = -EINVAL;
1790 goto out_put_tmp_file;
1793 if (ip->i_ino == tip->i_ino) {
1794 error = -EINVAL;
1795 goto out_put_tmp_file;
1798 if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
1799 error = -EIO;
1800 goto out_put_tmp_file;
1803 error = xfs_swap_extents(ip, tip, sxp);
1805 out_put_tmp_file:
1806 fdput(tmp);
1807 out_put_file:
1808 fdput(f);
1809 out:
1810 return error;
1814 * Note: some of the ioctl's return positive numbers as a
1815 * byte count indicating success, such as readlink_by_handle.
1816 * So we don't "sign flip" like most other routines. This means
1817 * true errors need to be returned as a negative value.
1819 long
1820 xfs_file_ioctl(
1821 struct file *filp,
1822 unsigned int cmd,
1823 unsigned long p)
1825 struct inode *inode = file_inode(filp);
1826 struct xfs_inode *ip = XFS_I(inode);
1827 struct xfs_mount *mp = ip->i_mount;
1828 void __user *arg = (void __user *)p;
1829 int error;
1831 trace_xfs_file_ioctl(ip);
1833 switch (cmd) {
1834 case FITRIM:
1835 return xfs_ioc_trim(mp, arg);
1836 case XFS_IOC_ALLOCSP:
1837 case XFS_IOC_FREESP:
1838 case XFS_IOC_RESVSP:
1839 case XFS_IOC_UNRESVSP:
1840 case XFS_IOC_ALLOCSP64:
1841 case XFS_IOC_FREESP64:
1842 case XFS_IOC_RESVSP64:
1843 case XFS_IOC_UNRESVSP64:
1844 case XFS_IOC_ZERO_RANGE: {
1845 xfs_flock64_t bf;
1847 if (copy_from_user(&bf, arg, sizeof(bf)))
1848 return -EFAULT;
1849 return xfs_ioc_space(filp, cmd, &bf);
1851 case XFS_IOC_DIOINFO: {
1852 struct dioattr da;
1853 xfs_buftarg_t *target =
1854 XFS_IS_REALTIME_INODE(ip) ?
1855 mp->m_rtdev_targp : mp->m_ddev_targp;
1857 da.d_mem = da.d_miniosz = target->bt_logical_sectorsize;
1858 da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1860 if (copy_to_user(arg, &da, sizeof(da)))
1861 return -EFAULT;
1862 return 0;
1865 case XFS_IOC_FSBULKSTAT_SINGLE:
1866 case XFS_IOC_FSBULKSTAT:
1867 case XFS_IOC_FSINUMBERS:
1868 return xfs_ioc_bulkstat(mp, cmd, arg);
1870 case XFS_IOC_FSGEOMETRY_V1:
1871 return xfs_ioc_fsgeometry_v1(mp, arg);
1873 case XFS_IOC_FSGEOMETRY:
1874 return xfs_ioc_fsgeometry(mp, arg);
1876 case XFS_IOC_GETVERSION:
1877 return put_user(inode->i_generation, (int __user *)arg);
1879 case XFS_IOC_FSGETXATTR:
1880 return xfs_ioc_fsgetxattr(ip, 0, arg);
1881 case XFS_IOC_FSGETXATTRA:
1882 return xfs_ioc_fsgetxattr(ip, 1, arg);
1883 case XFS_IOC_FSSETXATTR:
1884 return xfs_ioc_fssetxattr(ip, filp, arg);
1885 case XFS_IOC_GETXFLAGS:
1886 return xfs_ioc_getxflags(ip, arg);
1887 case XFS_IOC_SETXFLAGS:
1888 return xfs_ioc_setxflags(ip, filp, arg);
1890 case XFS_IOC_FSSETDM: {
1891 struct fsdmidata dmi;
1893 if (copy_from_user(&dmi, arg, sizeof(dmi)))
1894 return -EFAULT;
1896 error = mnt_want_write_file(filp);
1897 if (error)
1898 return error;
1900 error = xfs_set_dmattrs(ip, dmi.fsd_dmevmask,
1901 dmi.fsd_dmstate);
1902 mnt_drop_write_file(filp);
1903 return error;
1906 case XFS_IOC_GETBMAP:
1907 case XFS_IOC_GETBMAPA:
1908 case XFS_IOC_GETBMAPX:
1909 return xfs_ioc_getbmap(filp, cmd, arg);
1911 case FS_IOC_GETFSMAP:
1912 return xfs_ioc_getfsmap(ip, arg);
1914 case XFS_IOC_SCRUB_METADATA:
1915 return xfs_ioc_scrub_metadata(ip, arg);
1917 case XFS_IOC_FD_TO_HANDLE:
1918 case XFS_IOC_PATH_TO_HANDLE:
1919 case XFS_IOC_PATH_TO_FSHANDLE: {
1920 xfs_fsop_handlereq_t hreq;
1922 if (copy_from_user(&hreq, arg, sizeof(hreq)))
1923 return -EFAULT;
1924 return xfs_find_handle(cmd, &hreq);
1926 case XFS_IOC_OPEN_BY_HANDLE: {
1927 xfs_fsop_handlereq_t hreq;
1929 if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1930 return -EFAULT;
1931 return xfs_open_by_handle(filp, &hreq);
1933 case XFS_IOC_FSSETDM_BY_HANDLE:
1934 return xfs_fssetdm_by_handle(filp, arg);
1936 case XFS_IOC_READLINK_BY_HANDLE: {
1937 xfs_fsop_handlereq_t hreq;
1939 if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1940 return -EFAULT;
1941 return xfs_readlink_by_handle(filp, &hreq);
1943 case XFS_IOC_ATTRLIST_BY_HANDLE:
1944 return xfs_attrlist_by_handle(filp, arg);
1946 case XFS_IOC_ATTRMULTI_BY_HANDLE:
1947 return xfs_attrmulti_by_handle(filp, arg);
1949 case XFS_IOC_SWAPEXT: {
1950 struct xfs_swapext sxp;
1952 if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
1953 return -EFAULT;
1954 error = mnt_want_write_file(filp);
1955 if (error)
1956 return error;
1957 error = xfs_ioc_swapext(&sxp);
1958 mnt_drop_write_file(filp);
1959 return error;
1962 case XFS_IOC_FSCOUNTS: {
1963 xfs_fsop_counts_t out;
1965 error = xfs_fs_counts(mp, &out);
1966 if (error)
1967 return error;
1969 if (copy_to_user(arg, &out, sizeof(out)))
1970 return -EFAULT;
1971 return 0;
1974 case XFS_IOC_SET_RESBLKS: {
1975 xfs_fsop_resblks_t inout;
1976 uint64_t in;
1978 if (!capable(CAP_SYS_ADMIN))
1979 return -EPERM;
1981 if (mp->m_flags & XFS_MOUNT_RDONLY)
1982 return -EROFS;
1984 if (copy_from_user(&inout, arg, sizeof(inout)))
1985 return -EFAULT;
1987 error = mnt_want_write_file(filp);
1988 if (error)
1989 return error;
1991 /* input parameter is passed in resblks field of structure */
1992 in = inout.resblks;
1993 error = xfs_reserve_blocks(mp, &in, &inout);
1994 mnt_drop_write_file(filp);
1995 if (error)
1996 return error;
1998 if (copy_to_user(arg, &inout, sizeof(inout)))
1999 return -EFAULT;
2000 return 0;
2003 case XFS_IOC_GET_RESBLKS: {
2004 xfs_fsop_resblks_t out;
2006 if (!capable(CAP_SYS_ADMIN))
2007 return -EPERM;
2009 error = xfs_reserve_blocks(mp, NULL, &out);
2010 if (error)
2011 return error;
2013 if (copy_to_user(arg, &out, sizeof(out)))
2014 return -EFAULT;
2016 return 0;
2019 case XFS_IOC_FSGROWFSDATA: {
2020 xfs_growfs_data_t in;
2022 if (copy_from_user(&in, arg, sizeof(in)))
2023 return -EFAULT;
2025 error = mnt_want_write_file(filp);
2026 if (error)
2027 return error;
2028 error = xfs_growfs_data(mp, &in);
2029 mnt_drop_write_file(filp);
2030 return error;
2033 case XFS_IOC_FSGROWFSLOG: {
2034 xfs_growfs_log_t in;
2036 if (copy_from_user(&in, arg, sizeof(in)))
2037 return -EFAULT;
2039 error = mnt_want_write_file(filp);
2040 if (error)
2041 return error;
2042 error = xfs_growfs_log(mp, &in);
2043 mnt_drop_write_file(filp);
2044 return error;
2047 case XFS_IOC_FSGROWFSRT: {
2048 xfs_growfs_rt_t in;
2050 if (copy_from_user(&in, arg, sizeof(in)))
2051 return -EFAULT;
2053 error = mnt_want_write_file(filp);
2054 if (error)
2055 return error;
2056 error = xfs_growfs_rt(mp, &in);
2057 mnt_drop_write_file(filp);
2058 return error;
2061 case XFS_IOC_GOINGDOWN: {
2062 uint32_t in;
2064 if (!capable(CAP_SYS_ADMIN))
2065 return -EPERM;
2067 if (get_user(in, (uint32_t __user *)arg))
2068 return -EFAULT;
2070 return xfs_fs_goingdown(mp, in);
2073 case XFS_IOC_ERROR_INJECTION: {
2074 xfs_error_injection_t in;
2076 if (!capable(CAP_SYS_ADMIN))
2077 return -EPERM;
2079 if (copy_from_user(&in, arg, sizeof(in)))
2080 return -EFAULT;
2082 return xfs_errortag_add(mp, in.errtag);
2085 case XFS_IOC_ERROR_CLEARALL:
2086 if (!capable(CAP_SYS_ADMIN))
2087 return -EPERM;
2089 return xfs_errortag_clearall(mp);
2091 case XFS_IOC_FREE_EOFBLOCKS: {
2092 struct xfs_fs_eofblocks eofb;
2093 struct xfs_eofblocks keofb;
2095 if (!capable(CAP_SYS_ADMIN))
2096 return -EPERM;
2098 if (mp->m_flags & XFS_MOUNT_RDONLY)
2099 return -EROFS;
2101 if (copy_from_user(&eofb, arg, sizeof(eofb)))
2102 return -EFAULT;
2104 error = xfs_fs_eofblocks_from_user(&eofb, &keofb);
2105 if (error)
2106 return error;
2108 return xfs_icache_free_eofblocks(mp, &keofb);
2111 default:
2112 return -ENOTTY;