xref: /openbmc/linux/fs/xfs/xfs_ioctl.c (revision 56a0eccd)
1 /*
2  * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3  * All Rights Reserved.
4  *
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.
8  *
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.
13  *
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
17  */
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 
45 #include <linux/capability.h>
46 #include <linux/dcache.h>
47 #include <linux/mount.h>
48 #include <linux/namei.h>
49 #include <linux/pagemap.h>
50 #include <linux/slab.h>
51 #include <linux/exportfs.h>
52 
53 /*
54  * xfs_find_handle maps from userspace xfs_fsop_handlereq structure to
55  * a file or fs handle.
56  *
57  * XFS_IOC_PATH_TO_FSHANDLE
58  *    returns fs handle for a mount point or path within that mount point
59  * XFS_IOC_FD_TO_HANDLE
60  *    returns full handle for a FD opened in user space
61  * XFS_IOC_PATH_TO_HANDLE
62  *    returns full handle for a path
63  */
64 int
65 xfs_find_handle(
66 	unsigned int		cmd,
67 	xfs_fsop_handlereq_t	*hreq)
68 {
69 	int			hsize;
70 	xfs_handle_t		handle;
71 	struct inode		*inode;
72 	struct fd		f = {NULL};
73 	struct path		path;
74 	int			error;
75 	struct xfs_inode	*ip;
76 
77 	if (cmd == XFS_IOC_FD_TO_HANDLE) {
78 		f = fdget(hreq->fd);
79 		if (!f.file)
80 			return -EBADF;
81 		inode = file_inode(f.file);
82 	} else {
83 		error = user_lpath((const char __user *)hreq->path, &path);
84 		if (error)
85 			return error;
86 		inode = d_inode(path.dentry);
87 	}
88 	ip = XFS_I(inode);
89 
90 	/*
91 	 * We can only generate handles for inodes residing on a XFS filesystem,
92 	 * and only for regular files, directories or symbolic links.
93 	 */
94 	error = -EINVAL;
95 	if (inode->i_sb->s_magic != XFS_SB_MAGIC)
96 		goto out_put;
97 
98 	error = -EBADF;
99 	if (!S_ISREG(inode->i_mode) &&
100 	    !S_ISDIR(inode->i_mode) &&
101 	    !S_ISLNK(inode->i_mode))
102 		goto out_put;
103 
104 
105 	memcpy(&handle.ha_fsid, ip->i_mount->m_fixedfsid, sizeof(xfs_fsid_t));
106 
107 	if (cmd == XFS_IOC_PATH_TO_FSHANDLE) {
108 		/*
109 		 * This handle only contains an fsid, zero the rest.
110 		 */
111 		memset(&handle.ha_fid, 0, sizeof(handle.ha_fid));
112 		hsize = sizeof(xfs_fsid_t);
113 	} else {
114 		handle.ha_fid.fid_len = sizeof(xfs_fid_t) -
115 					sizeof(handle.ha_fid.fid_len);
116 		handle.ha_fid.fid_pad = 0;
117 		handle.ha_fid.fid_gen = inode->i_generation;
118 		handle.ha_fid.fid_ino = ip->i_ino;
119 
120 		hsize = XFS_HSIZE(handle);
121 	}
122 
123 	error = -EFAULT;
124 	if (copy_to_user(hreq->ohandle, &handle, hsize) ||
125 	    copy_to_user(hreq->ohandlen, &hsize, sizeof(__s32)))
126 		goto out_put;
127 
128 	error = 0;
129 
130  out_put:
131 	if (cmd == XFS_IOC_FD_TO_HANDLE)
132 		fdput(f);
133 	else
134 		path_put(&path);
135 	return error;
136 }
137 
138 /*
139  * No need to do permission checks on the various pathname components
140  * as the handle operations are privileged.
141  */
142 STATIC int
143 xfs_handle_acceptable(
144 	void			*context,
145 	struct dentry		*dentry)
146 {
147 	return 1;
148 }
149 
150 /*
151  * Convert userspace handle data into a dentry.
152  */
153 struct dentry *
154 xfs_handle_to_dentry(
155 	struct file		*parfilp,
156 	void __user		*uhandle,
157 	u32			hlen)
158 {
159 	xfs_handle_t		handle;
160 	struct xfs_fid64	fid;
161 
162 	/*
163 	 * Only allow handle opens under a directory.
164 	 */
165 	if (!S_ISDIR(file_inode(parfilp)->i_mode))
166 		return ERR_PTR(-ENOTDIR);
167 
168 	if (hlen != sizeof(xfs_handle_t))
169 		return ERR_PTR(-EINVAL);
170 	if (copy_from_user(&handle, uhandle, hlen))
171 		return ERR_PTR(-EFAULT);
172 	if (handle.ha_fid.fid_len !=
173 	    sizeof(handle.ha_fid) - sizeof(handle.ha_fid.fid_len))
174 		return ERR_PTR(-EINVAL);
175 
176 	memset(&fid, 0, sizeof(struct fid));
177 	fid.ino = handle.ha_fid.fid_ino;
178 	fid.gen = handle.ha_fid.fid_gen;
179 
180 	return exportfs_decode_fh(parfilp->f_path.mnt, (struct fid *)&fid, 3,
181 			FILEID_INO32_GEN | XFS_FILEID_TYPE_64FLAG,
182 			xfs_handle_acceptable, NULL);
183 }
184 
185 STATIC struct dentry *
186 xfs_handlereq_to_dentry(
187 	struct file		*parfilp,
188 	xfs_fsop_handlereq_t	*hreq)
189 {
190 	return xfs_handle_to_dentry(parfilp, hreq->ihandle, hreq->ihandlen);
191 }
192 
193 int
194 xfs_open_by_handle(
195 	struct file		*parfilp,
196 	xfs_fsop_handlereq_t	*hreq)
197 {
198 	const struct cred	*cred = current_cred();
199 	int			error;
200 	int			fd;
201 	int			permflag;
202 	struct file		*filp;
203 	struct inode		*inode;
204 	struct dentry		*dentry;
205 	fmode_t			fmode;
206 	struct path		path;
207 
208 	if (!capable(CAP_SYS_ADMIN))
209 		return -EPERM;
210 
211 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
212 	if (IS_ERR(dentry))
213 		return PTR_ERR(dentry);
214 	inode = d_inode(dentry);
215 
216 	/* Restrict xfs_open_by_handle to directories & regular files. */
217 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))) {
218 		error = -EPERM;
219 		goto out_dput;
220 	}
221 
222 #if BITS_PER_LONG != 32
223 	hreq->oflags |= O_LARGEFILE;
224 #endif
225 
226 	permflag = hreq->oflags;
227 	fmode = OPEN_FMODE(permflag);
228 	if ((!(permflag & O_APPEND) || (permflag & O_TRUNC)) &&
229 	    (fmode & FMODE_WRITE) && IS_APPEND(inode)) {
230 		error = -EPERM;
231 		goto out_dput;
232 	}
233 
234 	if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
235 		error = -EACCES;
236 		goto out_dput;
237 	}
238 
239 	/* Can't write directories. */
240 	if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
241 		error = -EISDIR;
242 		goto out_dput;
243 	}
244 
245 	fd = get_unused_fd_flags(0);
246 	if (fd < 0) {
247 		error = fd;
248 		goto out_dput;
249 	}
250 
251 	path.mnt = parfilp->f_path.mnt;
252 	path.dentry = dentry;
253 	filp = dentry_open(&path, hreq->oflags, cred);
254 	dput(dentry);
255 	if (IS_ERR(filp)) {
256 		put_unused_fd(fd);
257 		return PTR_ERR(filp);
258 	}
259 
260 	if (S_ISREG(inode->i_mode)) {
261 		filp->f_flags |= O_NOATIME;
262 		filp->f_mode |= FMODE_NOCMTIME;
263 	}
264 
265 	fd_install(fd, filp);
266 	return fd;
267 
268  out_dput:
269 	dput(dentry);
270 	return error;
271 }
272 
273 int
274 xfs_readlink_by_handle(
275 	struct file		*parfilp,
276 	xfs_fsop_handlereq_t	*hreq)
277 {
278 	struct dentry		*dentry;
279 	__u32			olen;
280 	void			*link;
281 	int			error;
282 
283 	if (!capable(CAP_SYS_ADMIN))
284 		return -EPERM;
285 
286 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
287 	if (IS_ERR(dentry))
288 		return PTR_ERR(dentry);
289 
290 	/* Restrict this handle operation to symlinks only. */
291 	if (!d_is_symlink(dentry)) {
292 		error = -EINVAL;
293 		goto out_dput;
294 	}
295 
296 	if (copy_from_user(&olen, hreq->ohandlen, sizeof(__u32))) {
297 		error = -EFAULT;
298 		goto out_dput;
299 	}
300 
301 	link = kmalloc(MAXPATHLEN+1, GFP_KERNEL);
302 	if (!link) {
303 		error = -ENOMEM;
304 		goto out_dput;
305 	}
306 
307 	error = xfs_readlink(XFS_I(d_inode(dentry)), link);
308 	if (error)
309 		goto out_kfree;
310 	error = readlink_copy(hreq->ohandle, olen, link);
311 	if (error)
312 		goto out_kfree;
313 
314  out_kfree:
315 	kfree(link);
316  out_dput:
317 	dput(dentry);
318 	return error;
319 }
320 
321 int
322 xfs_set_dmattrs(
323 	xfs_inode_t     *ip,
324 	u_int		evmask,
325 	u_int16_t	state)
326 {
327 	xfs_mount_t	*mp = ip->i_mount;
328 	xfs_trans_t	*tp;
329 	int		error;
330 
331 	if (!capable(CAP_SYS_ADMIN))
332 		return -EPERM;
333 
334 	if (XFS_FORCED_SHUTDOWN(mp))
335 		return -EIO;
336 
337 	tp = xfs_trans_alloc(mp, XFS_TRANS_SET_DMATTRS);
338 	error = xfs_trans_reserve(tp, &M_RES(mp)->tr_ichange, 0, 0);
339 	if (error) {
340 		xfs_trans_cancel(tp);
341 		return error;
342 	}
343 	xfs_ilock(ip, XFS_ILOCK_EXCL);
344 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
345 
346 	ip->i_d.di_dmevmask = evmask;
347 	ip->i_d.di_dmstate  = state;
348 
349 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
350 	error = xfs_trans_commit(tp);
351 
352 	return error;
353 }
354 
355 STATIC int
356 xfs_fssetdm_by_handle(
357 	struct file		*parfilp,
358 	void			__user *arg)
359 {
360 	int			error;
361 	struct fsdmidata	fsd;
362 	xfs_fsop_setdm_handlereq_t dmhreq;
363 	struct dentry		*dentry;
364 
365 	if (!capable(CAP_MKNOD))
366 		return -EPERM;
367 	if (copy_from_user(&dmhreq, arg, sizeof(xfs_fsop_setdm_handlereq_t)))
368 		return -EFAULT;
369 
370 	error = mnt_want_write_file(parfilp);
371 	if (error)
372 		return error;
373 
374 	dentry = xfs_handlereq_to_dentry(parfilp, &dmhreq.hreq);
375 	if (IS_ERR(dentry)) {
376 		mnt_drop_write_file(parfilp);
377 		return PTR_ERR(dentry);
378 	}
379 
380 	if (IS_IMMUTABLE(d_inode(dentry)) || IS_APPEND(d_inode(dentry))) {
381 		error = -EPERM;
382 		goto out;
383 	}
384 
385 	if (copy_from_user(&fsd, dmhreq.data, sizeof(fsd))) {
386 		error = -EFAULT;
387 		goto out;
388 	}
389 
390 	error = xfs_set_dmattrs(XFS_I(d_inode(dentry)), fsd.fsd_dmevmask,
391 				 fsd.fsd_dmstate);
392 
393  out:
394 	mnt_drop_write_file(parfilp);
395 	dput(dentry);
396 	return error;
397 }
398 
399 STATIC int
400 xfs_attrlist_by_handle(
401 	struct file		*parfilp,
402 	void			__user *arg)
403 {
404 	int			error = -ENOMEM;
405 	attrlist_cursor_kern_t	*cursor;
406 	xfs_fsop_attrlist_handlereq_t al_hreq;
407 	struct dentry		*dentry;
408 	char			*kbuf;
409 
410 	if (!capable(CAP_SYS_ADMIN))
411 		return -EPERM;
412 	if (copy_from_user(&al_hreq, arg, sizeof(xfs_fsop_attrlist_handlereq_t)))
413 		return -EFAULT;
414 	if (al_hreq.buflen < sizeof(struct attrlist) ||
415 	    al_hreq.buflen > XFS_XATTR_LIST_MAX)
416 		return -EINVAL;
417 
418 	/*
419 	 * Reject flags, only allow namespaces.
420 	 */
421 	if (al_hreq.flags & ~(ATTR_ROOT | ATTR_SECURE))
422 		return -EINVAL;
423 
424 	dentry = xfs_handlereq_to_dentry(parfilp, &al_hreq.hreq);
425 	if (IS_ERR(dentry))
426 		return PTR_ERR(dentry);
427 
428 	kbuf = kmem_zalloc_large(al_hreq.buflen, KM_SLEEP);
429 	if (!kbuf)
430 		goto out_dput;
431 
432 	cursor = (attrlist_cursor_kern_t *)&al_hreq.pos;
433 	error = xfs_attr_list(XFS_I(d_inode(dentry)), kbuf, al_hreq.buflen,
434 					al_hreq.flags, cursor);
435 	if (error)
436 		goto out_kfree;
437 
438 	if (copy_to_user(al_hreq.buffer, kbuf, al_hreq.buflen))
439 		error = -EFAULT;
440 
441 out_kfree:
442 	kmem_free(kbuf);
443 out_dput:
444 	dput(dentry);
445 	return error;
446 }
447 
448 int
449 xfs_attrmulti_attr_get(
450 	struct inode		*inode,
451 	unsigned char		*name,
452 	unsigned char		__user *ubuf,
453 	__uint32_t		*len,
454 	__uint32_t		flags)
455 {
456 	unsigned char		*kbuf;
457 	int			error = -EFAULT;
458 
459 	if (*len > XFS_XATTR_SIZE_MAX)
460 		return -EINVAL;
461 	kbuf = kmem_zalloc_large(*len, KM_SLEEP);
462 	if (!kbuf)
463 		return -ENOMEM;
464 
465 	error = xfs_attr_get(XFS_I(inode), name, kbuf, (int *)len, flags);
466 	if (error)
467 		goto out_kfree;
468 
469 	if (copy_to_user(ubuf, kbuf, *len))
470 		error = -EFAULT;
471 
472 out_kfree:
473 	kmem_free(kbuf);
474 	return error;
475 }
476 
477 int
478 xfs_attrmulti_attr_set(
479 	struct inode		*inode,
480 	unsigned char		*name,
481 	const unsigned char	__user *ubuf,
482 	__uint32_t		len,
483 	__uint32_t		flags)
484 {
485 	unsigned char		*kbuf;
486 	int			error;
487 
488 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
489 		return -EPERM;
490 	if (len > XFS_XATTR_SIZE_MAX)
491 		return -EINVAL;
492 
493 	kbuf = memdup_user(ubuf, len);
494 	if (IS_ERR(kbuf))
495 		return PTR_ERR(kbuf);
496 
497 	error = xfs_attr_set(XFS_I(inode), name, kbuf, len, flags);
498 	if (!error)
499 		xfs_forget_acl(inode, name, flags);
500 	kfree(kbuf);
501 	return error;
502 }
503 
504 int
505 xfs_attrmulti_attr_remove(
506 	struct inode		*inode,
507 	unsigned char		*name,
508 	__uint32_t		flags)
509 {
510 	int			error;
511 
512 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
513 		return -EPERM;
514 	error = xfs_attr_remove(XFS_I(inode), name, flags);
515 	if (!error)
516 		xfs_forget_acl(inode, name, flags);
517 	return error;
518 }
519 
520 STATIC int
521 xfs_attrmulti_by_handle(
522 	struct file		*parfilp,
523 	void			__user *arg)
524 {
525 	int			error;
526 	xfs_attr_multiop_t	*ops;
527 	xfs_fsop_attrmulti_handlereq_t am_hreq;
528 	struct dentry		*dentry;
529 	unsigned int		i, size;
530 	unsigned char		*attr_name;
531 
532 	if (!capable(CAP_SYS_ADMIN))
533 		return -EPERM;
534 	if (copy_from_user(&am_hreq, arg, sizeof(xfs_fsop_attrmulti_handlereq_t)))
535 		return -EFAULT;
536 
537 	/* overflow check */
538 	if (am_hreq.opcount >= INT_MAX / sizeof(xfs_attr_multiop_t))
539 		return -E2BIG;
540 
541 	dentry = xfs_handlereq_to_dentry(parfilp, &am_hreq.hreq);
542 	if (IS_ERR(dentry))
543 		return PTR_ERR(dentry);
544 
545 	error = -E2BIG;
546 	size = am_hreq.opcount * sizeof(xfs_attr_multiop_t);
547 	if (!size || size > 16 * PAGE_SIZE)
548 		goto out_dput;
549 
550 	ops = memdup_user(am_hreq.ops, size);
551 	if (IS_ERR(ops)) {
552 		error = PTR_ERR(ops);
553 		goto out_dput;
554 	}
555 
556 	error = -ENOMEM;
557 	attr_name = kmalloc(MAXNAMELEN, GFP_KERNEL);
558 	if (!attr_name)
559 		goto out_kfree_ops;
560 
561 	error = 0;
562 	for (i = 0; i < am_hreq.opcount; i++) {
563 		ops[i].am_error = strncpy_from_user((char *)attr_name,
564 				ops[i].am_attrname, MAXNAMELEN);
565 		if (ops[i].am_error == 0 || ops[i].am_error == MAXNAMELEN)
566 			error = -ERANGE;
567 		if (ops[i].am_error < 0)
568 			break;
569 
570 		switch (ops[i].am_opcode) {
571 		case ATTR_OP_GET:
572 			ops[i].am_error = xfs_attrmulti_attr_get(
573 					d_inode(dentry), attr_name,
574 					ops[i].am_attrvalue, &ops[i].am_length,
575 					ops[i].am_flags);
576 			break;
577 		case ATTR_OP_SET:
578 			ops[i].am_error = mnt_want_write_file(parfilp);
579 			if (ops[i].am_error)
580 				break;
581 			ops[i].am_error = xfs_attrmulti_attr_set(
582 					d_inode(dentry), attr_name,
583 					ops[i].am_attrvalue, ops[i].am_length,
584 					ops[i].am_flags);
585 			mnt_drop_write_file(parfilp);
586 			break;
587 		case ATTR_OP_REMOVE:
588 			ops[i].am_error = mnt_want_write_file(parfilp);
589 			if (ops[i].am_error)
590 				break;
591 			ops[i].am_error = xfs_attrmulti_attr_remove(
592 					d_inode(dentry), attr_name,
593 					ops[i].am_flags);
594 			mnt_drop_write_file(parfilp);
595 			break;
596 		default:
597 			ops[i].am_error = -EINVAL;
598 		}
599 	}
600 
601 	if (copy_to_user(am_hreq.ops, ops, size))
602 		error = -EFAULT;
603 
604 	kfree(attr_name);
605  out_kfree_ops:
606 	kfree(ops);
607  out_dput:
608 	dput(dentry);
609 	return error;
610 }
611 
612 int
613 xfs_ioc_space(
614 	struct xfs_inode	*ip,
615 	struct inode		*inode,
616 	struct file		*filp,
617 	int			ioflags,
618 	unsigned int		cmd,
619 	xfs_flock64_t		*bf)
620 {
621 	struct iattr		iattr;
622 	enum xfs_prealloc_flags	flags = 0;
623 	uint			iolock = XFS_IOLOCK_EXCL;
624 	int			error;
625 
626 	/*
627 	 * Only allow the sys admin to reserve space unless
628 	 * unwritten extents are enabled.
629 	 */
630 	if (!xfs_sb_version_hasextflgbit(&ip->i_mount->m_sb) &&
631 	    !capable(CAP_SYS_ADMIN))
632 		return -EPERM;
633 
634 	if (inode->i_flags & (S_IMMUTABLE|S_APPEND))
635 		return -EPERM;
636 
637 	if (!(filp->f_mode & FMODE_WRITE))
638 		return -EBADF;
639 
640 	if (!S_ISREG(inode->i_mode))
641 		return -EINVAL;
642 
643 	if (filp->f_flags & O_DSYNC)
644 		flags |= XFS_PREALLOC_SYNC;
645 	if (ioflags & XFS_IO_INVIS)
646 		flags |= XFS_PREALLOC_INVISIBLE;
647 
648 	error = mnt_want_write_file(filp);
649 	if (error)
650 		return error;
651 
652 	xfs_ilock(ip, iolock);
653 	error = xfs_break_layouts(inode, &iolock, false);
654 	if (error)
655 		goto out_unlock;
656 
657 	xfs_ilock(ip, XFS_MMAPLOCK_EXCL);
658 	iolock |= XFS_MMAPLOCK_EXCL;
659 
660 	switch (bf->l_whence) {
661 	case 0: /*SEEK_SET*/
662 		break;
663 	case 1: /*SEEK_CUR*/
664 		bf->l_start += filp->f_pos;
665 		break;
666 	case 2: /*SEEK_END*/
667 		bf->l_start += XFS_ISIZE(ip);
668 		break;
669 	default:
670 		error = -EINVAL;
671 		goto out_unlock;
672 	}
673 
674 	/*
675 	 * length of <= 0 for resv/unresv/zero is invalid.  length for
676 	 * alloc/free is ignored completely and we have no idea what userspace
677 	 * might have set it to, so set it to zero to allow range
678 	 * checks to pass.
679 	 */
680 	switch (cmd) {
681 	case XFS_IOC_ZERO_RANGE:
682 	case XFS_IOC_RESVSP:
683 	case XFS_IOC_RESVSP64:
684 	case XFS_IOC_UNRESVSP:
685 	case XFS_IOC_UNRESVSP64:
686 		if (bf->l_len <= 0) {
687 			error = -EINVAL;
688 			goto out_unlock;
689 		}
690 		break;
691 	default:
692 		bf->l_len = 0;
693 		break;
694 	}
695 
696 	if (bf->l_start < 0 ||
697 	    bf->l_start > inode->i_sb->s_maxbytes ||
698 	    bf->l_start + bf->l_len < 0 ||
699 	    bf->l_start + bf->l_len >= inode->i_sb->s_maxbytes) {
700 		error = -EINVAL;
701 		goto out_unlock;
702 	}
703 
704 	switch (cmd) {
705 	case XFS_IOC_ZERO_RANGE:
706 		flags |= XFS_PREALLOC_SET;
707 		error = xfs_zero_file_space(ip, bf->l_start, bf->l_len);
708 		break;
709 	case XFS_IOC_RESVSP:
710 	case XFS_IOC_RESVSP64:
711 		flags |= XFS_PREALLOC_SET;
712 		error = xfs_alloc_file_space(ip, bf->l_start, bf->l_len,
713 						XFS_BMAPI_PREALLOC);
714 		break;
715 	case XFS_IOC_UNRESVSP:
716 	case XFS_IOC_UNRESVSP64:
717 		error = xfs_free_file_space(ip, bf->l_start, bf->l_len);
718 		break;
719 	case XFS_IOC_ALLOCSP:
720 	case XFS_IOC_ALLOCSP64:
721 	case XFS_IOC_FREESP:
722 	case XFS_IOC_FREESP64:
723 		flags |= XFS_PREALLOC_CLEAR;
724 		if (bf->l_start > XFS_ISIZE(ip)) {
725 			error = xfs_alloc_file_space(ip, XFS_ISIZE(ip),
726 					bf->l_start - XFS_ISIZE(ip), 0);
727 			if (error)
728 				goto out_unlock;
729 		}
730 
731 		iattr.ia_valid = ATTR_SIZE;
732 		iattr.ia_size = bf->l_start;
733 
734 		error = xfs_setattr_size(ip, &iattr);
735 		break;
736 	default:
737 		ASSERT(0);
738 		error = -EINVAL;
739 	}
740 
741 	if (error)
742 		goto out_unlock;
743 
744 	error = xfs_update_prealloc_flags(ip, flags);
745 
746 out_unlock:
747 	xfs_iunlock(ip, iolock);
748 	mnt_drop_write_file(filp);
749 	return error;
750 }
751 
752 STATIC int
753 xfs_ioc_bulkstat(
754 	xfs_mount_t		*mp,
755 	unsigned int		cmd,
756 	void			__user *arg)
757 {
758 	xfs_fsop_bulkreq_t	bulkreq;
759 	int			count;	/* # of records returned */
760 	xfs_ino_t		inlast;	/* last inode number */
761 	int			done;
762 	int			error;
763 
764 	/* done = 1 if there are more stats to get and if bulkstat */
765 	/* should be called again (unused here, but used in dmapi) */
766 
767 	if (!capable(CAP_SYS_ADMIN))
768 		return -EPERM;
769 
770 	if (XFS_FORCED_SHUTDOWN(mp))
771 		return -EIO;
772 
773 	if (copy_from_user(&bulkreq, arg, sizeof(xfs_fsop_bulkreq_t)))
774 		return -EFAULT;
775 
776 	if (copy_from_user(&inlast, bulkreq.lastip, sizeof(__s64)))
777 		return -EFAULT;
778 
779 	if ((count = bulkreq.icount) <= 0)
780 		return -EINVAL;
781 
782 	if (bulkreq.ubuffer == NULL)
783 		return -EINVAL;
784 
785 	if (cmd == XFS_IOC_FSINUMBERS)
786 		error = xfs_inumbers(mp, &inlast, &count,
787 					bulkreq.ubuffer, xfs_inumbers_fmt);
788 	else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE)
789 		error = xfs_bulkstat_one(mp, inlast, bulkreq.ubuffer,
790 					sizeof(xfs_bstat_t), NULL, &done);
791 	else	/* XFS_IOC_FSBULKSTAT */
792 		error = xfs_bulkstat(mp, &inlast, &count, xfs_bulkstat_one,
793 				     sizeof(xfs_bstat_t), bulkreq.ubuffer,
794 				     &done);
795 
796 	if (error)
797 		return error;
798 
799 	if (bulkreq.ocount != NULL) {
800 		if (copy_to_user(bulkreq.lastip, &inlast,
801 						sizeof(xfs_ino_t)))
802 			return -EFAULT;
803 
804 		if (copy_to_user(bulkreq.ocount, &count, sizeof(count)))
805 			return -EFAULT;
806 	}
807 
808 	return 0;
809 }
810 
811 STATIC int
812 xfs_ioc_fsgeometry_v1(
813 	xfs_mount_t		*mp,
814 	void			__user *arg)
815 {
816 	xfs_fsop_geom_t         fsgeo;
817 	int			error;
818 
819 	error = xfs_fs_geometry(mp, &fsgeo, 3);
820 	if (error)
821 		return error;
822 
823 	/*
824 	 * Caller should have passed an argument of type
825 	 * xfs_fsop_geom_v1_t.  This is a proper subset of the
826 	 * xfs_fsop_geom_t that xfs_fs_geometry() fills in.
827 	 */
828 	if (copy_to_user(arg, &fsgeo, sizeof(xfs_fsop_geom_v1_t)))
829 		return -EFAULT;
830 	return 0;
831 }
832 
833 STATIC int
834 xfs_ioc_fsgeometry(
835 	xfs_mount_t		*mp,
836 	void			__user *arg)
837 {
838 	xfs_fsop_geom_t		fsgeo;
839 	int			error;
840 
841 	error = xfs_fs_geometry(mp, &fsgeo, 4);
842 	if (error)
843 		return error;
844 
845 	if (copy_to_user(arg, &fsgeo, sizeof(fsgeo)))
846 		return -EFAULT;
847 	return 0;
848 }
849 
850 /*
851  * Linux extended inode flags interface.
852  */
853 
854 STATIC unsigned int
855 xfs_merge_ioc_xflags(
856 	unsigned int	flags,
857 	unsigned int	start)
858 {
859 	unsigned int	xflags = start;
860 
861 	if (flags & FS_IMMUTABLE_FL)
862 		xflags |= FS_XFLAG_IMMUTABLE;
863 	else
864 		xflags &= ~FS_XFLAG_IMMUTABLE;
865 	if (flags & FS_APPEND_FL)
866 		xflags |= FS_XFLAG_APPEND;
867 	else
868 		xflags &= ~FS_XFLAG_APPEND;
869 	if (flags & FS_SYNC_FL)
870 		xflags |= FS_XFLAG_SYNC;
871 	else
872 		xflags &= ~FS_XFLAG_SYNC;
873 	if (flags & FS_NOATIME_FL)
874 		xflags |= FS_XFLAG_NOATIME;
875 	else
876 		xflags &= ~FS_XFLAG_NOATIME;
877 	if (flags & FS_NODUMP_FL)
878 		xflags |= FS_XFLAG_NODUMP;
879 	else
880 		xflags &= ~FS_XFLAG_NODUMP;
881 
882 	return xflags;
883 }
884 
885 STATIC unsigned int
886 xfs_di2lxflags(
887 	__uint16_t	di_flags)
888 {
889 	unsigned int	flags = 0;
890 
891 	if (di_flags & XFS_DIFLAG_IMMUTABLE)
892 		flags |= FS_IMMUTABLE_FL;
893 	if (di_flags & XFS_DIFLAG_APPEND)
894 		flags |= FS_APPEND_FL;
895 	if (di_flags & XFS_DIFLAG_SYNC)
896 		flags |= FS_SYNC_FL;
897 	if (di_flags & XFS_DIFLAG_NOATIME)
898 		flags |= FS_NOATIME_FL;
899 	if (di_flags & XFS_DIFLAG_NODUMP)
900 		flags |= FS_NODUMP_FL;
901 	return flags;
902 }
903 
904 STATIC int
905 xfs_ioc_fsgetxattr(
906 	xfs_inode_t		*ip,
907 	int			attr,
908 	void			__user *arg)
909 {
910 	struct fsxattr		fa;
911 
912 	memset(&fa, 0, sizeof(struct fsxattr));
913 
914 	xfs_ilock(ip, XFS_ILOCK_SHARED);
915 	fa.fsx_xflags = xfs_ip2xflags(ip);
916 	fa.fsx_extsize = ip->i_d.di_extsize << ip->i_mount->m_sb.sb_blocklog;
917 	fa.fsx_projid = xfs_get_projid(ip);
918 
919 	if (attr) {
920 		if (ip->i_afp) {
921 			if (ip->i_afp->if_flags & XFS_IFEXTENTS)
922 				fa.fsx_nextents = ip->i_afp->if_bytes /
923 							sizeof(xfs_bmbt_rec_t);
924 			else
925 				fa.fsx_nextents = ip->i_d.di_anextents;
926 		} else
927 			fa.fsx_nextents = 0;
928 	} else {
929 		if (ip->i_df.if_flags & XFS_IFEXTENTS)
930 			fa.fsx_nextents = ip->i_df.if_bytes /
931 						sizeof(xfs_bmbt_rec_t);
932 		else
933 			fa.fsx_nextents = ip->i_d.di_nextents;
934 	}
935 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
936 
937 	if (copy_to_user(arg, &fa, sizeof(fa)))
938 		return -EFAULT;
939 	return 0;
940 }
941 
942 STATIC void
943 xfs_set_diflags(
944 	struct xfs_inode	*ip,
945 	unsigned int		xflags)
946 {
947 	unsigned int		di_flags;
948 	uint64_t		di_flags2;
949 
950 	/* can't set PREALLOC this way, just preserve it */
951 	di_flags = (ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
952 	if (xflags & FS_XFLAG_IMMUTABLE)
953 		di_flags |= XFS_DIFLAG_IMMUTABLE;
954 	if (xflags & FS_XFLAG_APPEND)
955 		di_flags |= XFS_DIFLAG_APPEND;
956 	if (xflags & FS_XFLAG_SYNC)
957 		di_flags |= XFS_DIFLAG_SYNC;
958 	if (xflags & FS_XFLAG_NOATIME)
959 		di_flags |= XFS_DIFLAG_NOATIME;
960 	if (xflags & FS_XFLAG_NODUMP)
961 		di_flags |= XFS_DIFLAG_NODUMP;
962 	if (xflags & FS_XFLAG_NODEFRAG)
963 		di_flags |= XFS_DIFLAG_NODEFRAG;
964 	if (xflags & FS_XFLAG_FILESTREAM)
965 		di_flags |= XFS_DIFLAG_FILESTREAM;
966 	if (S_ISDIR(VFS_I(ip)->i_mode)) {
967 		if (xflags & FS_XFLAG_RTINHERIT)
968 			di_flags |= XFS_DIFLAG_RTINHERIT;
969 		if (xflags & FS_XFLAG_NOSYMLINKS)
970 			di_flags |= XFS_DIFLAG_NOSYMLINKS;
971 		if (xflags & FS_XFLAG_EXTSZINHERIT)
972 			di_flags |= XFS_DIFLAG_EXTSZINHERIT;
973 		if (xflags & FS_XFLAG_PROJINHERIT)
974 			di_flags |= XFS_DIFLAG_PROJINHERIT;
975 	} else if (S_ISREG(VFS_I(ip)->i_mode)) {
976 		if (xflags & FS_XFLAG_REALTIME)
977 			di_flags |= XFS_DIFLAG_REALTIME;
978 		if (xflags & FS_XFLAG_EXTSIZE)
979 			di_flags |= XFS_DIFLAG_EXTSIZE;
980 	}
981 	ip->i_d.di_flags = di_flags;
982 
983 	/* diflags2 only valid for v3 inodes. */
984 	if (ip->i_d.di_version < 3)
985 		return;
986 
987 	di_flags2 = 0;
988 	if (xflags & FS_XFLAG_DAX)
989 		di_flags2 |= XFS_DIFLAG2_DAX;
990 
991 	ip->i_d.di_flags2 = di_flags2;
992 
993 }
994 
995 STATIC void
996 xfs_diflags_to_linux(
997 	struct xfs_inode	*ip)
998 {
999 	struct inode		*inode = VFS_I(ip);
1000 	unsigned int		xflags = xfs_ip2xflags(ip);
1001 
1002 	if (xflags & FS_XFLAG_IMMUTABLE)
1003 		inode->i_flags |= S_IMMUTABLE;
1004 	else
1005 		inode->i_flags &= ~S_IMMUTABLE;
1006 	if (xflags & FS_XFLAG_APPEND)
1007 		inode->i_flags |= S_APPEND;
1008 	else
1009 		inode->i_flags &= ~S_APPEND;
1010 	if (xflags & FS_XFLAG_SYNC)
1011 		inode->i_flags |= S_SYNC;
1012 	else
1013 		inode->i_flags &= ~S_SYNC;
1014 	if (xflags & FS_XFLAG_NOATIME)
1015 		inode->i_flags |= S_NOATIME;
1016 	else
1017 		inode->i_flags &= ~S_NOATIME;
1018 	if (xflags & FS_XFLAG_DAX)
1019 		inode->i_flags |= S_DAX;
1020 	else
1021 		inode->i_flags &= ~S_DAX;
1022 
1023 }
1024 
1025 static int
1026 xfs_ioctl_setattr_xflags(
1027 	struct xfs_trans	*tp,
1028 	struct xfs_inode	*ip,
1029 	struct fsxattr		*fa)
1030 {
1031 	struct xfs_mount	*mp = ip->i_mount;
1032 
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;
1037 
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;
1043 	}
1044 
1045 	/*
1046 	 * Can't modify an immutable/append-only file unless
1047 	 * we have appropriate permission.
1048 	 */
1049 	if (((ip->i_d.di_flags & (XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND)) ||
1050 	     (fa->fsx_xflags & (FS_XFLAG_IMMUTABLE | FS_XFLAG_APPEND))) &&
1051 	    !capable(CAP_LINUX_IMMUTABLE))
1052 		return -EPERM;
1053 
1054 	xfs_set_diflags(ip, fa->fsx_xflags);
1055 	xfs_diflags_to_linux(ip);
1056 	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1057 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1058 	XFS_STATS_INC(mp, xs_ig_attrchg);
1059 	return 0;
1060 }
1061 
1062 /*
1063  * If we are changing DAX flags, we have to ensure the file is clean and any
1064  * cached objects in the address space are invalidated and removed. This
1065  * requires us to lock out other IO and page faults similar to a truncate
1066  * operation. The locks need to be held until the transaction has been committed
1067  * so that the cache invalidation is atomic with respect to the DAX flag
1068  * manipulation.
1069  */
1070 static int
1071 xfs_ioctl_setattr_dax_invalidate(
1072 	struct xfs_inode	*ip,
1073 	struct fsxattr		*fa,
1074 	int			*join_flags)
1075 {
1076 	struct inode		*inode = VFS_I(ip);
1077 	int			error;
1078 
1079 	*join_flags = 0;
1080 
1081 	/*
1082 	 * It is only valid to set the DAX flag on regular files and
1083 	 * directories on filesystems where the block size is equal to the page
1084 	 * size. On directories it serves as an inherit hint.
1085 	 */
1086 	if (fa->fsx_xflags & FS_XFLAG_DAX) {
1087 		if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)))
1088 			return -EINVAL;
1089 		if (ip->i_mount->m_sb.sb_blocksize != PAGE_SIZE)
1090 			return -EINVAL;
1091 	}
1092 
1093 	/* If the DAX state is not changing, we have nothing to do here. */
1094 	if ((fa->fsx_xflags & FS_XFLAG_DAX) && IS_DAX(inode))
1095 		return 0;
1096 	if (!(fa->fsx_xflags & FS_XFLAG_DAX) && !IS_DAX(inode))
1097 		return 0;
1098 
1099 	/* lock, flush and invalidate mapping in preparation for flag change */
1100 	xfs_ilock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1101 	error = filemap_write_and_wait(inode->i_mapping);
1102 	if (error)
1103 		goto out_unlock;
1104 	error = invalidate_inode_pages2(inode->i_mapping);
1105 	if (error)
1106 		goto out_unlock;
1107 
1108 	*join_flags = XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL;
1109 	return 0;
1110 
1111 out_unlock:
1112 	xfs_iunlock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1113 	return error;
1114 
1115 }
1116 
1117 /*
1118  * Set up the transaction structure for the setattr operation, checking that we
1119  * have permission to do so. On success, return a clean transaction and the
1120  * inode locked exclusively ready for further operation specific checks. On
1121  * failure, return an error without modifying or locking the inode.
1122  *
1123  * The inode might already be IO locked on call. If this is the case, it is
1124  * indicated in @join_flags and we take full responsibility for ensuring they
1125  * are unlocked from now on. Hence if we have an error here, we still have to
1126  * unlock them. Otherwise, once they are joined to the transaction, they will
1127  * be unlocked on commit/cancel.
1128  */
1129 static struct xfs_trans *
1130 xfs_ioctl_setattr_get_trans(
1131 	struct xfs_inode	*ip,
1132 	int			join_flags)
1133 {
1134 	struct xfs_mount	*mp = ip->i_mount;
1135 	struct xfs_trans	*tp;
1136 	int			error = -EROFS;
1137 
1138 	if (mp->m_flags & XFS_MOUNT_RDONLY)
1139 		goto out_unlock;
1140 	error = -EIO;
1141 	if (XFS_FORCED_SHUTDOWN(mp))
1142 		goto out_unlock;
1143 
1144 	tp = xfs_trans_alloc(mp, XFS_TRANS_SETATTR_NOT_SIZE);
1145 	error = xfs_trans_reserve(tp, &M_RES(mp)->tr_ichange, 0, 0);
1146 	if (error)
1147 		goto out_cancel;
1148 
1149 	xfs_ilock(ip, XFS_ILOCK_EXCL);
1150 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | join_flags);
1151 	join_flags = 0;
1152 
1153 	/*
1154 	 * CAP_FOWNER overrides the following restrictions:
1155 	 *
1156 	 * The user ID of the calling process must be equal to the file owner
1157 	 * ID, except in cases where the CAP_FSETID capability is applicable.
1158 	 */
1159 	if (!inode_owner_or_capable(VFS_I(ip))) {
1160 		error = -EPERM;
1161 		goto out_cancel;
1162 	}
1163 
1164 	if (mp->m_flags & XFS_MOUNT_WSYNC)
1165 		xfs_trans_set_sync(tp);
1166 
1167 	return tp;
1168 
1169 out_cancel:
1170 	xfs_trans_cancel(tp);
1171 out_unlock:
1172 	if (join_flags)
1173 		xfs_iunlock(ip, join_flags);
1174 	return ERR_PTR(error);
1175 }
1176 
1177 /*
1178  * extent size hint validation is somewhat cumbersome. Rules are:
1179  *
1180  * 1. extent size hint is only valid for directories and regular files
1181  * 2. FS_XFLAG_EXTSIZE is only valid for regular files
1182  * 3. FS_XFLAG_EXTSZINHERIT is only valid for directories.
1183  * 4. can only be changed on regular files if no extents are allocated
1184  * 5. can be changed on directories at any time
1185  * 6. extsize hint of 0 turns off hints, clears inode flags.
1186  * 7. Extent size must be a multiple of the appropriate block size.
1187  * 8. for non-realtime files, the extent size hint must be limited
1188  *    to half the AG size to avoid alignment extending the extent beyond the
1189  *    limits of the AG.
1190  */
1191 static int
1192 xfs_ioctl_setattr_check_extsize(
1193 	struct xfs_inode	*ip,
1194 	struct fsxattr		*fa)
1195 {
1196 	struct xfs_mount	*mp = ip->i_mount;
1197 
1198 	if ((fa->fsx_xflags & FS_XFLAG_EXTSIZE) && !S_ISREG(VFS_I(ip)->i_mode))
1199 		return -EINVAL;
1200 
1201 	if ((fa->fsx_xflags & FS_XFLAG_EXTSZINHERIT) &&
1202 	    !S_ISDIR(VFS_I(ip)->i_mode))
1203 		return -EINVAL;
1204 
1205 	if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_d.di_nextents &&
1206 	    ((ip->i_d.di_extsize << mp->m_sb.sb_blocklog) != fa->fsx_extsize))
1207 		return -EINVAL;
1208 
1209 	if (fa->fsx_extsize != 0) {
1210 		xfs_extlen_t    size;
1211 		xfs_fsblock_t   extsize_fsb;
1212 
1213 		extsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1214 		if (extsize_fsb > MAXEXTLEN)
1215 			return -EINVAL;
1216 
1217 		if (XFS_IS_REALTIME_INODE(ip) ||
1218 		    (fa->fsx_xflags & FS_XFLAG_REALTIME)) {
1219 			size = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
1220 		} else {
1221 			size = mp->m_sb.sb_blocksize;
1222 			if (extsize_fsb > mp->m_sb.sb_agblocks / 2)
1223 				return -EINVAL;
1224 		}
1225 
1226 		if (fa->fsx_extsize % size)
1227 			return -EINVAL;
1228 	} else
1229 		fa->fsx_xflags &= ~(FS_XFLAG_EXTSIZE | FS_XFLAG_EXTSZINHERIT);
1230 
1231 	return 0;
1232 }
1233 
1234 static int
1235 xfs_ioctl_setattr_check_projid(
1236 	struct xfs_inode	*ip,
1237 	struct fsxattr		*fa)
1238 {
1239 	/* Disallow 32bit project ids if projid32bit feature is not enabled. */
1240 	if (fa->fsx_projid > (__uint16_t)-1 &&
1241 	    !xfs_sb_version_hasprojid32bit(&ip->i_mount->m_sb))
1242 		return -EINVAL;
1243 
1244 	/*
1245 	 * Project Quota ID state is only allowed to change from within the init
1246 	 * namespace. Enforce that restriction only if we are trying to change
1247 	 * the quota ID state. Everything else is allowed in user namespaces.
1248 	 */
1249 	if (current_user_ns() == &init_user_ns)
1250 		return 0;
1251 
1252 	if (xfs_get_projid(ip) != fa->fsx_projid)
1253 		return -EINVAL;
1254 	if ((fa->fsx_xflags & FS_XFLAG_PROJINHERIT) !=
1255 	    (ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT))
1256 		return -EINVAL;
1257 
1258 	return 0;
1259 }
1260 
1261 STATIC int
1262 xfs_ioctl_setattr(
1263 	xfs_inode_t		*ip,
1264 	struct fsxattr		*fa)
1265 {
1266 	struct xfs_mount	*mp = ip->i_mount;
1267 	struct xfs_trans	*tp;
1268 	struct xfs_dquot	*udqp = NULL;
1269 	struct xfs_dquot	*pdqp = NULL;
1270 	struct xfs_dquot	*olddquot = NULL;
1271 	int			code;
1272 	int			join_flags = 0;
1273 
1274 	trace_xfs_ioctl_setattr(ip);
1275 
1276 	code = xfs_ioctl_setattr_check_projid(ip, fa);
1277 	if (code)
1278 		return code;
1279 
1280 	/*
1281 	 * If disk quotas is on, we make sure that the dquots do exist on disk,
1282 	 * before we start any other transactions. Trying to do this later
1283 	 * is messy. We don't care to take a readlock to look at the ids
1284 	 * in inode here, because we can't hold it across the trans_reserve.
1285 	 * If the IDs do change before we take the ilock, we're covered
1286 	 * because the i_*dquot fields will get updated anyway.
1287 	 */
1288 	if (XFS_IS_QUOTA_ON(mp)) {
1289 		code = xfs_qm_vop_dqalloc(ip, ip->i_d.di_uid,
1290 					 ip->i_d.di_gid, fa->fsx_projid,
1291 					 XFS_QMOPT_PQUOTA, &udqp, NULL, &pdqp);
1292 		if (code)
1293 			return code;
1294 	}
1295 
1296 	/*
1297 	 * Changing DAX config may require inode locking for mapping
1298 	 * invalidation. These need to be held all the way to transaction commit
1299 	 * or cancel time, so need to be passed through to
1300 	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1301 	 * appropriately.
1302 	 */
1303 	code = xfs_ioctl_setattr_dax_invalidate(ip, fa, &join_flags);
1304 	if (code)
1305 		goto error_free_dquots;
1306 
1307 	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1308 	if (IS_ERR(tp)) {
1309 		code = PTR_ERR(tp);
1310 		goto error_free_dquots;
1311 	}
1312 
1313 
1314 	if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp) &&
1315 	    xfs_get_projid(ip) != fa->fsx_projid) {
1316 		code = xfs_qm_vop_chown_reserve(tp, ip, udqp, NULL, pdqp,
1317 				capable(CAP_FOWNER) ?  XFS_QMOPT_FORCE_RES : 0);
1318 		if (code)	/* out of quota */
1319 			goto error_trans_cancel;
1320 	}
1321 
1322 	code = xfs_ioctl_setattr_check_extsize(ip, fa);
1323 	if (code)
1324 		goto error_trans_cancel;
1325 
1326 	code = xfs_ioctl_setattr_xflags(tp, ip, fa);
1327 	if (code)
1328 		goto error_trans_cancel;
1329 
1330 	/*
1331 	 * Change file ownership.  Must be the owner or privileged.  CAP_FSETID
1332 	 * overrides the following restrictions:
1333 	 *
1334 	 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1335 	 * successful return from chown()
1336 	 */
1337 
1338 	if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
1339 	    !capable_wrt_inode_uidgid(VFS_I(ip), CAP_FSETID))
1340 		VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
1341 
1342 	/* Change the ownerships and register project quota modifications */
1343 	if (xfs_get_projid(ip) != fa->fsx_projid) {
1344 		if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp)) {
1345 			olddquot = xfs_qm_vop_chown(tp, ip,
1346 						&ip->i_pdquot, pdqp);
1347 		}
1348 		ASSERT(ip->i_d.di_version > 1);
1349 		xfs_set_projid(ip, fa->fsx_projid);
1350 	}
1351 
1352 	/*
1353 	 * Only set the extent size hint if we've already determined that the
1354 	 * extent size hint should be set on the inode. If no extent size flags
1355 	 * are set on the inode then unconditionally clear the extent size hint.
1356 	 */
1357 	if (ip->i_d.di_flags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1358 		ip->i_d.di_extsize = fa->fsx_extsize >> mp->m_sb.sb_blocklog;
1359 	else
1360 		ip->i_d.di_extsize = 0;
1361 
1362 	code = xfs_trans_commit(tp);
1363 
1364 	/*
1365 	 * Release any dquot(s) the inode had kept before chown.
1366 	 */
1367 	xfs_qm_dqrele(olddquot);
1368 	xfs_qm_dqrele(udqp);
1369 	xfs_qm_dqrele(pdqp);
1370 
1371 	return code;
1372 
1373 error_trans_cancel:
1374 	xfs_trans_cancel(tp);
1375 error_free_dquots:
1376 	xfs_qm_dqrele(udqp);
1377 	xfs_qm_dqrele(pdqp);
1378 	return code;
1379 }
1380 
1381 STATIC int
1382 xfs_ioc_fssetxattr(
1383 	xfs_inode_t		*ip,
1384 	struct file		*filp,
1385 	void			__user *arg)
1386 {
1387 	struct fsxattr		fa;
1388 	int error;
1389 
1390 	if (copy_from_user(&fa, arg, sizeof(fa)))
1391 		return -EFAULT;
1392 
1393 	error = mnt_want_write_file(filp);
1394 	if (error)
1395 		return error;
1396 	error = xfs_ioctl_setattr(ip, &fa);
1397 	mnt_drop_write_file(filp);
1398 	return error;
1399 }
1400 
1401 STATIC int
1402 xfs_ioc_getxflags(
1403 	xfs_inode_t		*ip,
1404 	void			__user *arg)
1405 {
1406 	unsigned int		flags;
1407 
1408 	flags = xfs_di2lxflags(ip->i_d.di_flags);
1409 	if (copy_to_user(arg, &flags, sizeof(flags)))
1410 		return -EFAULT;
1411 	return 0;
1412 }
1413 
1414 STATIC int
1415 xfs_ioc_setxflags(
1416 	struct xfs_inode	*ip,
1417 	struct file		*filp,
1418 	void			__user *arg)
1419 {
1420 	struct xfs_trans	*tp;
1421 	struct fsxattr		fa;
1422 	unsigned int		flags;
1423 	int			join_flags = 0;
1424 	int			error;
1425 
1426 	if (copy_from_user(&flags, arg, sizeof(flags)))
1427 		return -EFAULT;
1428 
1429 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
1430 		      FS_NOATIME_FL | FS_NODUMP_FL | \
1431 		      FS_SYNC_FL))
1432 		return -EOPNOTSUPP;
1433 
1434 	fa.fsx_xflags = xfs_merge_ioc_xflags(flags, xfs_ip2xflags(ip));
1435 
1436 	error = mnt_want_write_file(filp);
1437 	if (error)
1438 		return error;
1439 
1440 	/*
1441 	 * Changing DAX config may require inode locking for mapping
1442 	 * invalidation. These need to be held all the way to transaction commit
1443 	 * or cancel time, so need to be passed through to
1444 	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1445 	 * appropriately.
1446 	 */
1447 	error = xfs_ioctl_setattr_dax_invalidate(ip, &fa, &join_flags);
1448 	if (error)
1449 		goto out_drop_write;
1450 
1451 	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1452 	if (IS_ERR(tp)) {
1453 		error = PTR_ERR(tp);
1454 		goto out_drop_write;
1455 	}
1456 
1457 	error = xfs_ioctl_setattr_xflags(tp, ip, &fa);
1458 	if (error) {
1459 		xfs_trans_cancel(tp);
1460 		goto out_drop_write;
1461 	}
1462 
1463 	error = xfs_trans_commit(tp);
1464 out_drop_write:
1465 	mnt_drop_write_file(filp);
1466 	return error;
1467 }
1468 
1469 STATIC int
1470 xfs_getbmap_format(void **ap, struct getbmapx *bmv, int *full)
1471 {
1472 	struct getbmap __user	*base = (struct getbmap __user *)*ap;
1473 
1474 	/* copy only getbmap portion (not getbmapx) */
1475 	if (copy_to_user(base, bmv, sizeof(struct getbmap)))
1476 		return -EFAULT;
1477 
1478 	*ap += sizeof(struct getbmap);
1479 	return 0;
1480 }
1481 
1482 STATIC int
1483 xfs_ioc_getbmap(
1484 	struct xfs_inode	*ip,
1485 	int			ioflags,
1486 	unsigned int		cmd,
1487 	void			__user *arg)
1488 {
1489 	struct getbmapx		bmx;
1490 	int			error;
1491 
1492 	if (copy_from_user(&bmx, arg, sizeof(struct getbmapx)))
1493 		return -EFAULT;
1494 
1495 	if (bmx.bmv_count < 2)
1496 		return -EINVAL;
1497 
1498 	bmx.bmv_iflags = (cmd == XFS_IOC_GETBMAPA ? BMV_IF_ATTRFORK : 0);
1499 	if (ioflags & XFS_IO_INVIS)
1500 		bmx.bmv_iflags |= BMV_IF_NO_DMAPI_READ;
1501 
1502 	error = xfs_getbmap(ip, &bmx, xfs_getbmap_format,
1503 			    (__force struct getbmap *)arg+1);
1504 	if (error)
1505 		return error;
1506 
1507 	/* copy back header - only size of getbmap */
1508 	if (copy_to_user(arg, &bmx, sizeof(struct getbmap)))
1509 		return -EFAULT;
1510 	return 0;
1511 }
1512 
1513 STATIC int
1514 xfs_getbmapx_format(void **ap, struct getbmapx *bmv, int *full)
1515 {
1516 	struct getbmapx __user	*base = (struct getbmapx __user *)*ap;
1517 
1518 	if (copy_to_user(base, bmv, sizeof(struct getbmapx)))
1519 		return -EFAULT;
1520 
1521 	*ap += sizeof(struct getbmapx);
1522 	return 0;
1523 }
1524 
1525 STATIC int
1526 xfs_ioc_getbmapx(
1527 	struct xfs_inode	*ip,
1528 	void			__user *arg)
1529 {
1530 	struct getbmapx		bmx;
1531 	int			error;
1532 
1533 	if (copy_from_user(&bmx, arg, sizeof(bmx)))
1534 		return -EFAULT;
1535 
1536 	if (bmx.bmv_count < 2)
1537 		return -EINVAL;
1538 
1539 	if (bmx.bmv_iflags & (~BMV_IF_VALID))
1540 		return -EINVAL;
1541 
1542 	error = xfs_getbmap(ip, &bmx, xfs_getbmapx_format,
1543 			    (__force struct getbmapx *)arg+1);
1544 	if (error)
1545 		return error;
1546 
1547 	/* copy back header */
1548 	if (copy_to_user(arg, &bmx, sizeof(struct getbmapx)))
1549 		return -EFAULT;
1550 
1551 	return 0;
1552 }
1553 
1554 int
1555 xfs_ioc_swapext(
1556 	xfs_swapext_t	*sxp)
1557 {
1558 	xfs_inode_t     *ip, *tip;
1559 	struct fd	f, tmp;
1560 	int		error = 0;
1561 
1562 	/* Pull information for the target fd */
1563 	f = fdget((int)sxp->sx_fdtarget);
1564 	if (!f.file) {
1565 		error = -EINVAL;
1566 		goto out;
1567 	}
1568 
1569 	if (!(f.file->f_mode & FMODE_WRITE) ||
1570 	    !(f.file->f_mode & FMODE_READ) ||
1571 	    (f.file->f_flags & O_APPEND)) {
1572 		error = -EBADF;
1573 		goto out_put_file;
1574 	}
1575 
1576 	tmp = fdget((int)sxp->sx_fdtmp);
1577 	if (!tmp.file) {
1578 		error = -EINVAL;
1579 		goto out_put_file;
1580 	}
1581 
1582 	if (!(tmp.file->f_mode & FMODE_WRITE) ||
1583 	    !(tmp.file->f_mode & FMODE_READ) ||
1584 	    (tmp.file->f_flags & O_APPEND)) {
1585 		error = -EBADF;
1586 		goto out_put_tmp_file;
1587 	}
1588 
1589 	if (IS_SWAPFILE(file_inode(f.file)) ||
1590 	    IS_SWAPFILE(file_inode(tmp.file))) {
1591 		error = -EINVAL;
1592 		goto out_put_tmp_file;
1593 	}
1594 
1595 	ip = XFS_I(file_inode(f.file));
1596 	tip = XFS_I(file_inode(tmp.file));
1597 
1598 	if (ip->i_mount != tip->i_mount) {
1599 		error = -EINVAL;
1600 		goto out_put_tmp_file;
1601 	}
1602 
1603 	if (ip->i_ino == tip->i_ino) {
1604 		error = -EINVAL;
1605 		goto out_put_tmp_file;
1606 	}
1607 
1608 	if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
1609 		error = -EIO;
1610 		goto out_put_tmp_file;
1611 	}
1612 
1613 	error = xfs_swap_extents(ip, tip, sxp);
1614 
1615  out_put_tmp_file:
1616 	fdput(tmp);
1617  out_put_file:
1618 	fdput(f);
1619  out:
1620 	return error;
1621 }
1622 
1623 /*
1624  * Note: some of the ioctl's return positive numbers as a
1625  * byte count indicating success, such as readlink_by_handle.
1626  * So we don't "sign flip" like most other routines.  This means
1627  * true errors need to be returned as a negative value.
1628  */
1629 long
1630 xfs_file_ioctl(
1631 	struct file		*filp,
1632 	unsigned int		cmd,
1633 	unsigned long		p)
1634 {
1635 	struct inode		*inode = file_inode(filp);
1636 	struct xfs_inode	*ip = XFS_I(inode);
1637 	struct xfs_mount	*mp = ip->i_mount;
1638 	void			__user *arg = (void __user *)p;
1639 	int			ioflags = 0;
1640 	int			error;
1641 
1642 	if (filp->f_mode & FMODE_NOCMTIME)
1643 		ioflags |= XFS_IO_INVIS;
1644 
1645 	trace_xfs_file_ioctl(ip);
1646 
1647 	switch (cmd) {
1648 	case FITRIM:
1649 		return xfs_ioc_trim(mp, arg);
1650 	case XFS_IOC_ALLOCSP:
1651 	case XFS_IOC_FREESP:
1652 	case XFS_IOC_RESVSP:
1653 	case XFS_IOC_UNRESVSP:
1654 	case XFS_IOC_ALLOCSP64:
1655 	case XFS_IOC_FREESP64:
1656 	case XFS_IOC_RESVSP64:
1657 	case XFS_IOC_UNRESVSP64:
1658 	case XFS_IOC_ZERO_RANGE: {
1659 		xfs_flock64_t		bf;
1660 
1661 		if (copy_from_user(&bf, arg, sizeof(bf)))
1662 			return -EFAULT;
1663 		return xfs_ioc_space(ip, inode, filp, ioflags, cmd, &bf);
1664 	}
1665 	case XFS_IOC_DIOINFO: {
1666 		struct dioattr	da;
1667 		xfs_buftarg_t	*target =
1668 			XFS_IS_REALTIME_INODE(ip) ?
1669 			mp->m_rtdev_targp : mp->m_ddev_targp;
1670 
1671 		da.d_mem =  da.d_miniosz = target->bt_logical_sectorsize;
1672 		da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1673 
1674 		if (copy_to_user(arg, &da, sizeof(da)))
1675 			return -EFAULT;
1676 		return 0;
1677 	}
1678 
1679 	case XFS_IOC_FSBULKSTAT_SINGLE:
1680 	case XFS_IOC_FSBULKSTAT:
1681 	case XFS_IOC_FSINUMBERS:
1682 		return xfs_ioc_bulkstat(mp, cmd, arg);
1683 
1684 	case XFS_IOC_FSGEOMETRY_V1:
1685 		return xfs_ioc_fsgeometry_v1(mp, arg);
1686 
1687 	case XFS_IOC_FSGEOMETRY:
1688 		return xfs_ioc_fsgeometry(mp, arg);
1689 
1690 	case XFS_IOC_GETVERSION:
1691 		return put_user(inode->i_generation, (int __user *)arg);
1692 
1693 	case XFS_IOC_FSGETXATTR:
1694 		return xfs_ioc_fsgetxattr(ip, 0, arg);
1695 	case XFS_IOC_FSGETXATTRA:
1696 		return xfs_ioc_fsgetxattr(ip, 1, arg);
1697 	case XFS_IOC_FSSETXATTR:
1698 		return xfs_ioc_fssetxattr(ip, filp, arg);
1699 	case XFS_IOC_GETXFLAGS:
1700 		return xfs_ioc_getxflags(ip, arg);
1701 	case XFS_IOC_SETXFLAGS:
1702 		return xfs_ioc_setxflags(ip, filp, arg);
1703 
1704 	case XFS_IOC_FSSETDM: {
1705 		struct fsdmidata	dmi;
1706 
1707 		if (copy_from_user(&dmi, arg, sizeof(dmi)))
1708 			return -EFAULT;
1709 
1710 		error = mnt_want_write_file(filp);
1711 		if (error)
1712 			return error;
1713 
1714 		error = xfs_set_dmattrs(ip, dmi.fsd_dmevmask,
1715 				dmi.fsd_dmstate);
1716 		mnt_drop_write_file(filp);
1717 		return error;
1718 	}
1719 
1720 	case XFS_IOC_GETBMAP:
1721 	case XFS_IOC_GETBMAPA:
1722 		return xfs_ioc_getbmap(ip, ioflags, cmd, arg);
1723 
1724 	case XFS_IOC_GETBMAPX:
1725 		return xfs_ioc_getbmapx(ip, arg);
1726 
1727 	case XFS_IOC_FD_TO_HANDLE:
1728 	case XFS_IOC_PATH_TO_HANDLE:
1729 	case XFS_IOC_PATH_TO_FSHANDLE: {
1730 		xfs_fsop_handlereq_t	hreq;
1731 
1732 		if (copy_from_user(&hreq, arg, sizeof(hreq)))
1733 			return -EFAULT;
1734 		return xfs_find_handle(cmd, &hreq);
1735 	}
1736 	case XFS_IOC_OPEN_BY_HANDLE: {
1737 		xfs_fsop_handlereq_t	hreq;
1738 
1739 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1740 			return -EFAULT;
1741 		return xfs_open_by_handle(filp, &hreq);
1742 	}
1743 	case XFS_IOC_FSSETDM_BY_HANDLE:
1744 		return xfs_fssetdm_by_handle(filp, arg);
1745 
1746 	case XFS_IOC_READLINK_BY_HANDLE: {
1747 		xfs_fsop_handlereq_t	hreq;
1748 
1749 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1750 			return -EFAULT;
1751 		return xfs_readlink_by_handle(filp, &hreq);
1752 	}
1753 	case XFS_IOC_ATTRLIST_BY_HANDLE:
1754 		return xfs_attrlist_by_handle(filp, arg);
1755 
1756 	case XFS_IOC_ATTRMULTI_BY_HANDLE:
1757 		return xfs_attrmulti_by_handle(filp, arg);
1758 
1759 	case XFS_IOC_SWAPEXT: {
1760 		struct xfs_swapext	sxp;
1761 
1762 		if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
1763 			return -EFAULT;
1764 		error = mnt_want_write_file(filp);
1765 		if (error)
1766 			return error;
1767 		error = xfs_ioc_swapext(&sxp);
1768 		mnt_drop_write_file(filp);
1769 		return error;
1770 	}
1771 
1772 	case XFS_IOC_FSCOUNTS: {
1773 		xfs_fsop_counts_t out;
1774 
1775 		error = xfs_fs_counts(mp, &out);
1776 		if (error)
1777 			return error;
1778 
1779 		if (copy_to_user(arg, &out, sizeof(out)))
1780 			return -EFAULT;
1781 		return 0;
1782 	}
1783 
1784 	case XFS_IOC_SET_RESBLKS: {
1785 		xfs_fsop_resblks_t inout;
1786 		__uint64_t	   in;
1787 
1788 		if (!capable(CAP_SYS_ADMIN))
1789 			return -EPERM;
1790 
1791 		if (mp->m_flags & XFS_MOUNT_RDONLY)
1792 			return -EROFS;
1793 
1794 		if (copy_from_user(&inout, arg, sizeof(inout)))
1795 			return -EFAULT;
1796 
1797 		error = mnt_want_write_file(filp);
1798 		if (error)
1799 			return error;
1800 
1801 		/* input parameter is passed in resblks field of structure */
1802 		in = inout.resblks;
1803 		error = xfs_reserve_blocks(mp, &in, &inout);
1804 		mnt_drop_write_file(filp);
1805 		if (error)
1806 			return error;
1807 
1808 		if (copy_to_user(arg, &inout, sizeof(inout)))
1809 			return -EFAULT;
1810 		return 0;
1811 	}
1812 
1813 	case XFS_IOC_GET_RESBLKS: {
1814 		xfs_fsop_resblks_t out;
1815 
1816 		if (!capable(CAP_SYS_ADMIN))
1817 			return -EPERM;
1818 
1819 		error = xfs_reserve_blocks(mp, NULL, &out);
1820 		if (error)
1821 			return error;
1822 
1823 		if (copy_to_user(arg, &out, sizeof(out)))
1824 			return -EFAULT;
1825 
1826 		return 0;
1827 	}
1828 
1829 	case XFS_IOC_FSGROWFSDATA: {
1830 		xfs_growfs_data_t in;
1831 
1832 		if (copy_from_user(&in, arg, sizeof(in)))
1833 			return -EFAULT;
1834 
1835 		error = mnt_want_write_file(filp);
1836 		if (error)
1837 			return error;
1838 		error = xfs_growfs_data(mp, &in);
1839 		mnt_drop_write_file(filp);
1840 		return error;
1841 	}
1842 
1843 	case XFS_IOC_FSGROWFSLOG: {
1844 		xfs_growfs_log_t in;
1845 
1846 		if (copy_from_user(&in, arg, sizeof(in)))
1847 			return -EFAULT;
1848 
1849 		error = mnt_want_write_file(filp);
1850 		if (error)
1851 			return error;
1852 		error = xfs_growfs_log(mp, &in);
1853 		mnt_drop_write_file(filp);
1854 		return error;
1855 	}
1856 
1857 	case XFS_IOC_FSGROWFSRT: {
1858 		xfs_growfs_rt_t in;
1859 
1860 		if (copy_from_user(&in, arg, sizeof(in)))
1861 			return -EFAULT;
1862 
1863 		error = mnt_want_write_file(filp);
1864 		if (error)
1865 			return error;
1866 		error = xfs_growfs_rt(mp, &in);
1867 		mnt_drop_write_file(filp);
1868 		return error;
1869 	}
1870 
1871 	case XFS_IOC_GOINGDOWN: {
1872 		__uint32_t in;
1873 
1874 		if (!capable(CAP_SYS_ADMIN))
1875 			return -EPERM;
1876 
1877 		if (get_user(in, (__uint32_t __user *)arg))
1878 			return -EFAULT;
1879 
1880 		return xfs_fs_goingdown(mp, in);
1881 	}
1882 
1883 	case XFS_IOC_ERROR_INJECTION: {
1884 		xfs_error_injection_t in;
1885 
1886 		if (!capable(CAP_SYS_ADMIN))
1887 			return -EPERM;
1888 
1889 		if (copy_from_user(&in, arg, sizeof(in)))
1890 			return -EFAULT;
1891 
1892 		return xfs_errortag_add(in.errtag, mp);
1893 	}
1894 
1895 	case XFS_IOC_ERROR_CLEARALL:
1896 		if (!capable(CAP_SYS_ADMIN))
1897 			return -EPERM;
1898 
1899 		return xfs_errortag_clearall(mp, 1);
1900 
1901 	case XFS_IOC_FREE_EOFBLOCKS: {
1902 		struct xfs_fs_eofblocks eofb;
1903 		struct xfs_eofblocks keofb;
1904 
1905 		if (!capable(CAP_SYS_ADMIN))
1906 			return -EPERM;
1907 
1908 		if (mp->m_flags & XFS_MOUNT_RDONLY)
1909 			return -EROFS;
1910 
1911 		if (copy_from_user(&eofb, arg, sizeof(eofb)))
1912 			return -EFAULT;
1913 
1914 		error = xfs_fs_eofblocks_from_user(&eofb, &keofb);
1915 		if (error)
1916 			return error;
1917 
1918 		return xfs_icache_free_eofblocks(mp, &keofb);
1919 	}
1920 
1921 	default:
1922 		return -ENOTTY;
1923 	}
1924 }
1925