xref: /openbmc/linux/fs/xfs/xfs_ioctl.c (revision c059ee9d)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (c) 2000-2005 Silicon Graphics, Inc.
4  * All Rights Reserved.
5  */
6 #include "xfs.h"
7 #include "xfs_fs.h"
8 #include "xfs_shared.h"
9 #include "xfs_format.h"
10 #include "xfs_log_format.h"
11 #include "xfs_trans_resv.h"
12 #include "xfs_mount.h"
13 #include "xfs_inode.h"
14 #include "xfs_rtalloc.h"
15 #include "xfs_iwalk.h"
16 #include "xfs_itable.h"
17 #include "xfs_error.h"
18 #include "xfs_da_format.h"
19 #include "xfs_da_btree.h"
20 #include "xfs_attr.h"
21 #include "xfs_bmap.h"
22 #include "xfs_bmap_util.h"
23 #include "xfs_fsops.h"
24 #include "xfs_discard.h"
25 #include "xfs_quota.h"
26 #include "xfs_export.h"
27 #include "xfs_trace.h"
28 #include "xfs_icache.h"
29 #include "xfs_trans.h"
30 #include "xfs_acl.h"
31 #include "xfs_btree.h"
32 #include <linux/fsmap.h>
33 #include "xfs_fsmap.h"
34 #include "scrub/xfs_scrub.h"
35 #include "xfs_sb.h"
36 #include "xfs_ag.h"
37 #include "xfs_health.h"
38 #include "xfs_reflink.h"
39 #include "xfs_ioctl.h"
40 #include "xfs_xattr.h"
41 
42 #include <linux/mount.h>
43 #include <linux/namei.h>
44 #include <linux/fileattr.h>
45 
46 /*
47  * xfs_find_handle maps from userspace xfs_fsop_handlereq structure to
48  * a file or fs handle.
49  *
50  * XFS_IOC_PATH_TO_FSHANDLE
51  *    returns fs handle for a mount point or path within that mount point
52  * XFS_IOC_FD_TO_HANDLE
53  *    returns full handle for a FD opened in user space
54  * XFS_IOC_PATH_TO_HANDLE
55  *    returns full handle for a path
56  */
57 int
58 xfs_find_handle(
59 	unsigned int		cmd,
60 	xfs_fsop_handlereq_t	*hreq)
61 {
62 	int			hsize;
63 	xfs_handle_t		handle;
64 	struct inode		*inode;
65 	struct fd		f = {NULL};
66 	struct path		path;
67 	int			error;
68 	struct xfs_inode	*ip;
69 
70 	if (cmd == XFS_IOC_FD_TO_HANDLE) {
71 		f = fdget(hreq->fd);
72 		if (!f.file)
73 			return -EBADF;
74 		inode = file_inode(f.file);
75 	} else {
76 		error = user_path_at(AT_FDCWD, hreq->path, 0, &path);
77 		if (error)
78 			return error;
79 		inode = d_inode(path.dentry);
80 	}
81 	ip = XFS_I(inode);
82 
83 	/*
84 	 * We can only generate handles for inodes residing on a XFS filesystem,
85 	 * and only for regular files, directories or symbolic links.
86 	 */
87 	error = -EINVAL;
88 	if (inode->i_sb->s_magic != XFS_SB_MAGIC)
89 		goto out_put;
90 
91 	error = -EBADF;
92 	if (!S_ISREG(inode->i_mode) &&
93 	    !S_ISDIR(inode->i_mode) &&
94 	    !S_ISLNK(inode->i_mode))
95 		goto out_put;
96 
97 
98 	memcpy(&handle.ha_fsid, ip->i_mount->m_fixedfsid, sizeof(xfs_fsid_t));
99 
100 	if (cmd == XFS_IOC_PATH_TO_FSHANDLE) {
101 		/*
102 		 * This handle only contains an fsid, zero the rest.
103 		 */
104 		memset(&handle.ha_fid, 0, sizeof(handle.ha_fid));
105 		hsize = sizeof(xfs_fsid_t);
106 	} else {
107 		handle.ha_fid.fid_len = sizeof(xfs_fid_t) -
108 					sizeof(handle.ha_fid.fid_len);
109 		handle.ha_fid.fid_pad = 0;
110 		handle.ha_fid.fid_gen = inode->i_generation;
111 		handle.ha_fid.fid_ino = ip->i_ino;
112 		hsize = sizeof(xfs_handle_t);
113 	}
114 
115 	error = -EFAULT;
116 	if (copy_to_user(hreq->ohandle, &handle, hsize) ||
117 	    copy_to_user(hreq->ohandlen, &hsize, sizeof(__s32)))
118 		goto out_put;
119 
120 	error = 0;
121 
122  out_put:
123 	if (cmd == XFS_IOC_FD_TO_HANDLE)
124 		fdput(f);
125 	else
126 		path_put(&path);
127 	return error;
128 }
129 
130 /*
131  * No need to do permission checks on the various pathname components
132  * as the handle operations are privileged.
133  */
134 STATIC int
135 xfs_handle_acceptable(
136 	void			*context,
137 	struct dentry		*dentry)
138 {
139 	return 1;
140 }
141 
142 /*
143  * Convert userspace handle data into a dentry.
144  */
145 struct dentry *
146 xfs_handle_to_dentry(
147 	struct file		*parfilp,
148 	void __user		*uhandle,
149 	u32			hlen)
150 {
151 	xfs_handle_t		handle;
152 	struct xfs_fid64	fid;
153 
154 	/*
155 	 * Only allow handle opens under a directory.
156 	 */
157 	if (!S_ISDIR(file_inode(parfilp)->i_mode))
158 		return ERR_PTR(-ENOTDIR);
159 
160 	if (hlen != sizeof(xfs_handle_t))
161 		return ERR_PTR(-EINVAL);
162 	if (copy_from_user(&handle, uhandle, hlen))
163 		return ERR_PTR(-EFAULT);
164 	if (handle.ha_fid.fid_len !=
165 	    sizeof(handle.ha_fid) - sizeof(handle.ha_fid.fid_len))
166 		return ERR_PTR(-EINVAL);
167 
168 	memset(&fid, 0, sizeof(struct fid));
169 	fid.ino = handle.ha_fid.fid_ino;
170 	fid.gen = handle.ha_fid.fid_gen;
171 
172 	return exportfs_decode_fh(parfilp->f_path.mnt, (struct fid *)&fid, 3,
173 			FILEID_INO32_GEN | XFS_FILEID_TYPE_64FLAG,
174 			xfs_handle_acceptable, NULL);
175 }
176 
177 STATIC struct dentry *
178 xfs_handlereq_to_dentry(
179 	struct file		*parfilp,
180 	xfs_fsop_handlereq_t	*hreq)
181 {
182 	return xfs_handle_to_dentry(parfilp, hreq->ihandle, hreq->ihandlen);
183 }
184 
185 int
186 xfs_open_by_handle(
187 	struct file		*parfilp,
188 	xfs_fsop_handlereq_t	*hreq)
189 {
190 	const struct cred	*cred = current_cred();
191 	int			error;
192 	int			fd;
193 	int			permflag;
194 	struct file		*filp;
195 	struct inode		*inode;
196 	struct dentry		*dentry;
197 	fmode_t			fmode;
198 	struct path		path;
199 
200 	if (!capable(CAP_SYS_ADMIN))
201 		return -EPERM;
202 
203 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
204 	if (IS_ERR(dentry))
205 		return PTR_ERR(dentry);
206 	inode = d_inode(dentry);
207 
208 	/* Restrict xfs_open_by_handle to directories & regular files. */
209 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))) {
210 		error = -EPERM;
211 		goto out_dput;
212 	}
213 
214 #if BITS_PER_LONG != 32
215 	hreq->oflags |= O_LARGEFILE;
216 #endif
217 
218 	permflag = hreq->oflags;
219 	fmode = OPEN_FMODE(permflag);
220 	if ((!(permflag & O_APPEND) || (permflag & O_TRUNC)) &&
221 	    (fmode & FMODE_WRITE) && IS_APPEND(inode)) {
222 		error = -EPERM;
223 		goto out_dput;
224 	}
225 
226 	if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
227 		error = -EPERM;
228 		goto out_dput;
229 	}
230 
231 	/* Can't write directories. */
232 	if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
233 		error = -EISDIR;
234 		goto out_dput;
235 	}
236 
237 	fd = get_unused_fd_flags(0);
238 	if (fd < 0) {
239 		error = fd;
240 		goto out_dput;
241 	}
242 
243 	path.mnt = parfilp->f_path.mnt;
244 	path.dentry = dentry;
245 	filp = dentry_open(&path, hreq->oflags, cred);
246 	dput(dentry);
247 	if (IS_ERR(filp)) {
248 		put_unused_fd(fd);
249 		return PTR_ERR(filp);
250 	}
251 
252 	if (S_ISREG(inode->i_mode)) {
253 		filp->f_flags |= O_NOATIME;
254 		filp->f_mode |= FMODE_NOCMTIME;
255 	}
256 
257 	fd_install(fd, filp);
258 	return fd;
259 
260  out_dput:
261 	dput(dentry);
262 	return error;
263 }
264 
265 int
266 xfs_readlink_by_handle(
267 	struct file		*parfilp,
268 	xfs_fsop_handlereq_t	*hreq)
269 {
270 	struct dentry		*dentry;
271 	__u32			olen;
272 	int			error;
273 
274 	if (!capable(CAP_SYS_ADMIN))
275 		return -EPERM;
276 
277 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
278 	if (IS_ERR(dentry))
279 		return PTR_ERR(dentry);
280 
281 	/* Restrict this handle operation to symlinks only. */
282 	if (!d_is_symlink(dentry)) {
283 		error = -EINVAL;
284 		goto out_dput;
285 	}
286 
287 	if (copy_from_user(&olen, hreq->ohandlen, sizeof(__u32))) {
288 		error = -EFAULT;
289 		goto out_dput;
290 	}
291 
292 	error = vfs_readlink(dentry, hreq->ohandle, olen);
293 
294  out_dput:
295 	dput(dentry);
296 	return error;
297 }
298 
299 /*
300  * Format an attribute and copy it out to the user's buffer.
301  * Take care to check values and protect against them changing later,
302  * we may be reading them directly out of a user buffer.
303  */
304 static void
305 xfs_ioc_attr_put_listent(
306 	struct xfs_attr_list_context *context,
307 	int			flags,
308 	unsigned char		*name,
309 	int			namelen,
310 	int			valuelen)
311 {
312 	struct xfs_attrlist	*alist = context->buffer;
313 	struct xfs_attrlist_ent	*aep;
314 	int			arraytop;
315 
316 	ASSERT(!context->seen_enough);
317 	ASSERT(context->count >= 0);
318 	ASSERT(context->count < (ATTR_MAX_VALUELEN/8));
319 	ASSERT(context->firstu >= sizeof(*alist));
320 	ASSERT(context->firstu <= context->bufsize);
321 
322 	/*
323 	 * Only list entries in the right namespace.
324 	 */
325 	if (context->attr_filter != (flags & XFS_ATTR_NSP_ONDISK_MASK))
326 		return;
327 
328 	arraytop = sizeof(*alist) +
329 			context->count * sizeof(alist->al_offset[0]);
330 
331 	/* decrement by the actual bytes used by the attr */
332 	context->firstu -= round_up(offsetof(struct xfs_attrlist_ent, a_name) +
333 			namelen + 1, sizeof(uint32_t));
334 	if (context->firstu < arraytop) {
335 		trace_xfs_attr_list_full(context);
336 		alist->al_more = 1;
337 		context->seen_enough = 1;
338 		return;
339 	}
340 
341 	aep = context->buffer + context->firstu;
342 	aep->a_valuelen = valuelen;
343 	memcpy(aep->a_name, name, namelen);
344 	aep->a_name[namelen] = 0;
345 	alist->al_offset[context->count++] = context->firstu;
346 	alist->al_count = context->count;
347 	trace_xfs_attr_list_add(context);
348 }
349 
350 static unsigned int
351 xfs_attr_filter(
352 	u32			ioc_flags)
353 {
354 	if (ioc_flags & XFS_IOC_ATTR_ROOT)
355 		return XFS_ATTR_ROOT;
356 	if (ioc_flags & XFS_IOC_ATTR_SECURE)
357 		return XFS_ATTR_SECURE;
358 	return 0;
359 }
360 
361 static unsigned int
362 xfs_attr_flags(
363 	u32			ioc_flags)
364 {
365 	if (ioc_flags & XFS_IOC_ATTR_CREATE)
366 		return XATTR_CREATE;
367 	if (ioc_flags & XFS_IOC_ATTR_REPLACE)
368 		return XATTR_REPLACE;
369 	return 0;
370 }
371 
372 int
373 xfs_ioc_attr_list(
374 	struct xfs_inode		*dp,
375 	void __user			*ubuf,
376 	size_t				bufsize,
377 	int				flags,
378 	struct xfs_attrlist_cursor __user *ucursor)
379 {
380 	struct xfs_attr_list_context	context = { };
381 	struct xfs_attrlist		*alist;
382 	void				*buffer;
383 	int				error;
384 
385 	if (bufsize < sizeof(struct xfs_attrlist) ||
386 	    bufsize > XFS_XATTR_LIST_MAX)
387 		return -EINVAL;
388 
389 	/*
390 	 * Reject flags, only allow namespaces.
391 	 */
392 	if (flags & ~(XFS_IOC_ATTR_ROOT | XFS_IOC_ATTR_SECURE))
393 		return -EINVAL;
394 	if (flags == (XFS_IOC_ATTR_ROOT | XFS_IOC_ATTR_SECURE))
395 		return -EINVAL;
396 
397 	/*
398 	 * Validate the cursor.
399 	 */
400 	if (copy_from_user(&context.cursor, ucursor, sizeof(context.cursor)))
401 		return -EFAULT;
402 	if (context.cursor.pad1 || context.cursor.pad2)
403 		return -EINVAL;
404 	if (!context.cursor.initted &&
405 	    (context.cursor.hashval || context.cursor.blkno ||
406 	     context.cursor.offset))
407 		return -EINVAL;
408 
409 	buffer = kvzalloc(bufsize, GFP_KERNEL);
410 	if (!buffer)
411 		return -ENOMEM;
412 
413 	/*
414 	 * Initialize the output buffer.
415 	 */
416 	context.dp = dp;
417 	context.resynch = 1;
418 	context.attr_filter = xfs_attr_filter(flags);
419 	context.buffer = buffer;
420 	context.bufsize = round_down(bufsize, sizeof(uint32_t));
421 	context.firstu = context.bufsize;
422 	context.put_listent = xfs_ioc_attr_put_listent;
423 
424 	alist = context.buffer;
425 	alist->al_count = 0;
426 	alist->al_more = 0;
427 	alist->al_offset[0] = context.bufsize;
428 
429 	error = xfs_attr_list(&context);
430 	if (error)
431 		goto out_free;
432 
433 	if (copy_to_user(ubuf, buffer, bufsize) ||
434 	    copy_to_user(ucursor, &context.cursor, sizeof(context.cursor)))
435 		error = -EFAULT;
436 out_free:
437 	kmem_free(buffer);
438 	return error;
439 }
440 
441 STATIC int
442 xfs_attrlist_by_handle(
443 	struct file		*parfilp,
444 	struct xfs_fsop_attrlist_handlereq __user *p)
445 {
446 	struct xfs_fsop_attrlist_handlereq al_hreq;
447 	struct dentry		*dentry;
448 	int			error = -ENOMEM;
449 
450 	if (!capable(CAP_SYS_ADMIN))
451 		return -EPERM;
452 	if (copy_from_user(&al_hreq, p, sizeof(al_hreq)))
453 		return -EFAULT;
454 
455 	dentry = xfs_handlereq_to_dentry(parfilp, &al_hreq.hreq);
456 	if (IS_ERR(dentry))
457 		return PTR_ERR(dentry);
458 
459 	error = xfs_ioc_attr_list(XFS_I(d_inode(dentry)), al_hreq.buffer,
460 				  al_hreq.buflen, al_hreq.flags, &p->pos);
461 	dput(dentry);
462 	return error;
463 }
464 
465 static int
466 xfs_attrmulti_attr_get(
467 	struct inode		*inode,
468 	unsigned char		*name,
469 	unsigned char		__user *ubuf,
470 	uint32_t		*len,
471 	uint32_t		flags)
472 {
473 	struct xfs_da_args	args = {
474 		.dp		= XFS_I(inode),
475 		.attr_filter	= xfs_attr_filter(flags),
476 		.attr_flags	= xfs_attr_flags(flags),
477 		.name		= name,
478 		.namelen	= strlen(name),
479 		.valuelen	= *len,
480 	};
481 	int			error;
482 
483 	if (*len > XFS_XATTR_SIZE_MAX)
484 		return -EINVAL;
485 
486 	error = xfs_attr_get(&args);
487 	if (error)
488 		goto out_kfree;
489 
490 	*len = args.valuelen;
491 	if (copy_to_user(ubuf, args.value, args.valuelen))
492 		error = -EFAULT;
493 
494 out_kfree:
495 	kmem_free(args.value);
496 	return error;
497 }
498 
499 static int
500 xfs_attrmulti_attr_set(
501 	struct inode		*inode,
502 	unsigned char		*name,
503 	const unsigned char	__user *ubuf,
504 	uint32_t		len,
505 	uint32_t		flags)
506 {
507 	struct xfs_da_args	args = {
508 		.dp		= XFS_I(inode),
509 		.attr_filter	= xfs_attr_filter(flags),
510 		.attr_flags	= xfs_attr_flags(flags),
511 		.name		= name,
512 		.namelen	= strlen(name),
513 	};
514 	int			error;
515 
516 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
517 		return -EPERM;
518 
519 	if (ubuf) {
520 		if (len > XFS_XATTR_SIZE_MAX)
521 			return -EINVAL;
522 		args.value = memdup_user(ubuf, len);
523 		if (IS_ERR(args.value))
524 			return PTR_ERR(args.value);
525 		args.valuelen = len;
526 	}
527 
528 	error = xfs_attr_change(&args);
529 	if (!error && (flags & XFS_IOC_ATTR_ROOT))
530 		xfs_forget_acl(inode, name);
531 	kfree(args.value);
532 	return error;
533 }
534 
535 int
536 xfs_ioc_attrmulti_one(
537 	struct file		*parfilp,
538 	struct inode		*inode,
539 	uint32_t		opcode,
540 	void __user		*uname,
541 	void __user		*value,
542 	uint32_t		*len,
543 	uint32_t		flags)
544 {
545 	unsigned char		*name;
546 	int			error;
547 
548 	if ((flags & XFS_IOC_ATTR_ROOT) && (flags & XFS_IOC_ATTR_SECURE))
549 		return -EINVAL;
550 
551 	name = strndup_user(uname, MAXNAMELEN);
552 	if (IS_ERR(name))
553 		return PTR_ERR(name);
554 
555 	switch (opcode) {
556 	case ATTR_OP_GET:
557 		error = xfs_attrmulti_attr_get(inode, name, value, len, flags);
558 		break;
559 	case ATTR_OP_REMOVE:
560 		value = NULL;
561 		*len = 0;
562 		fallthrough;
563 	case ATTR_OP_SET:
564 		error = mnt_want_write_file(parfilp);
565 		if (error)
566 			break;
567 		error = xfs_attrmulti_attr_set(inode, name, value, *len, flags);
568 		mnt_drop_write_file(parfilp);
569 		break;
570 	default:
571 		error = -EINVAL;
572 		break;
573 	}
574 
575 	kfree(name);
576 	return error;
577 }
578 
579 STATIC int
580 xfs_attrmulti_by_handle(
581 	struct file		*parfilp,
582 	void			__user *arg)
583 {
584 	int			error;
585 	xfs_attr_multiop_t	*ops;
586 	xfs_fsop_attrmulti_handlereq_t am_hreq;
587 	struct dentry		*dentry;
588 	unsigned int		i, size;
589 
590 	if (!capable(CAP_SYS_ADMIN))
591 		return -EPERM;
592 	if (copy_from_user(&am_hreq, arg, sizeof(xfs_fsop_attrmulti_handlereq_t)))
593 		return -EFAULT;
594 
595 	/* overflow check */
596 	if (am_hreq.opcount >= INT_MAX / sizeof(xfs_attr_multiop_t))
597 		return -E2BIG;
598 
599 	dentry = xfs_handlereq_to_dentry(parfilp, &am_hreq.hreq);
600 	if (IS_ERR(dentry))
601 		return PTR_ERR(dentry);
602 
603 	error = -E2BIG;
604 	size = am_hreq.opcount * sizeof(xfs_attr_multiop_t);
605 	if (!size || size > 16 * PAGE_SIZE)
606 		goto out_dput;
607 
608 	ops = memdup_user(am_hreq.ops, size);
609 	if (IS_ERR(ops)) {
610 		error = PTR_ERR(ops);
611 		goto out_dput;
612 	}
613 
614 	error = 0;
615 	for (i = 0; i < am_hreq.opcount; i++) {
616 		ops[i].am_error = xfs_ioc_attrmulti_one(parfilp,
617 				d_inode(dentry), ops[i].am_opcode,
618 				ops[i].am_attrname, ops[i].am_attrvalue,
619 				&ops[i].am_length, ops[i].am_flags);
620 	}
621 
622 	if (copy_to_user(am_hreq.ops, ops, size))
623 		error = -EFAULT;
624 
625 	kfree(ops);
626  out_dput:
627 	dput(dentry);
628 	return error;
629 }
630 
631 /* Return 0 on success or positive error */
632 int
633 xfs_fsbulkstat_one_fmt(
634 	struct xfs_ibulk		*breq,
635 	const struct xfs_bulkstat	*bstat)
636 {
637 	struct xfs_bstat		bs1;
638 
639 	xfs_bulkstat_to_bstat(breq->mp, &bs1, bstat);
640 	if (copy_to_user(breq->ubuffer, &bs1, sizeof(bs1)))
641 		return -EFAULT;
642 	return xfs_ibulk_advance(breq, sizeof(struct xfs_bstat));
643 }
644 
645 int
646 xfs_fsinumbers_fmt(
647 	struct xfs_ibulk		*breq,
648 	const struct xfs_inumbers	*igrp)
649 {
650 	struct xfs_inogrp		ig1;
651 
652 	xfs_inumbers_to_inogrp(&ig1, igrp);
653 	if (copy_to_user(breq->ubuffer, &ig1, sizeof(struct xfs_inogrp)))
654 		return -EFAULT;
655 	return xfs_ibulk_advance(breq, sizeof(struct xfs_inogrp));
656 }
657 
658 STATIC int
659 xfs_ioc_fsbulkstat(
660 	struct file		*file,
661 	unsigned int		cmd,
662 	void			__user *arg)
663 {
664 	struct xfs_mount	*mp = XFS_I(file_inode(file))->i_mount;
665 	struct xfs_fsop_bulkreq	bulkreq;
666 	struct xfs_ibulk	breq = {
667 		.mp		= mp,
668 		.mnt_userns	= file_mnt_user_ns(file),
669 		.ocount		= 0,
670 	};
671 	xfs_ino_t		lastino;
672 	int			error;
673 
674 	/* done = 1 if there are more stats to get and if bulkstat */
675 	/* should be called again (unused here, but used in dmapi) */
676 
677 	if (!capable(CAP_SYS_ADMIN))
678 		return -EPERM;
679 
680 	if (xfs_is_shutdown(mp))
681 		return -EIO;
682 
683 	if (copy_from_user(&bulkreq, arg, sizeof(struct xfs_fsop_bulkreq)))
684 		return -EFAULT;
685 
686 	if (copy_from_user(&lastino, bulkreq.lastip, sizeof(__s64)))
687 		return -EFAULT;
688 
689 	if (bulkreq.icount <= 0)
690 		return -EINVAL;
691 
692 	if (bulkreq.ubuffer == NULL)
693 		return -EINVAL;
694 
695 	breq.ubuffer = bulkreq.ubuffer;
696 	breq.icount = bulkreq.icount;
697 
698 	/*
699 	 * FSBULKSTAT_SINGLE expects that *lastip contains the inode number
700 	 * that we want to stat.  However, FSINUMBERS and FSBULKSTAT expect
701 	 * that *lastip contains either zero or the number of the last inode to
702 	 * be examined by the previous call and return results starting with
703 	 * the next inode after that.  The new bulk request back end functions
704 	 * take the inode to start with, so we have to compute the startino
705 	 * parameter from lastino to maintain correct function.  lastino == 0
706 	 * is a special case because it has traditionally meant "first inode
707 	 * in filesystem".
708 	 */
709 	if (cmd == XFS_IOC_FSINUMBERS) {
710 		breq.startino = lastino ? lastino + 1 : 0;
711 		error = xfs_inumbers(&breq, xfs_fsinumbers_fmt);
712 		lastino = breq.startino - 1;
713 	} else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE) {
714 		breq.startino = lastino;
715 		breq.icount = 1;
716 		error = xfs_bulkstat_one(&breq, xfs_fsbulkstat_one_fmt);
717 	} else {	/* XFS_IOC_FSBULKSTAT */
718 		breq.startino = lastino ? lastino + 1 : 0;
719 		error = xfs_bulkstat(&breq, xfs_fsbulkstat_one_fmt);
720 		lastino = breq.startino - 1;
721 	}
722 
723 	if (error)
724 		return error;
725 
726 	if (bulkreq.lastip != NULL &&
727 	    copy_to_user(bulkreq.lastip, &lastino, sizeof(xfs_ino_t)))
728 		return -EFAULT;
729 
730 	if (bulkreq.ocount != NULL &&
731 	    copy_to_user(bulkreq.ocount, &breq.ocount, sizeof(__s32)))
732 		return -EFAULT;
733 
734 	return 0;
735 }
736 
737 /* Return 0 on success or positive error */
738 static int
739 xfs_bulkstat_fmt(
740 	struct xfs_ibulk		*breq,
741 	const struct xfs_bulkstat	*bstat)
742 {
743 	if (copy_to_user(breq->ubuffer, bstat, sizeof(struct xfs_bulkstat)))
744 		return -EFAULT;
745 	return xfs_ibulk_advance(breq, sizeof(struct xfs_bulkstat));
746 }
747 
748 /*
749  * Check the incoming bulk request @hdr from userspace and initialize the
750  * internal @breq bulk request appropriately.  Returns 0 if the bulk request
751  * should proceed; -ECANCELED if there's nothing to do; or the usual
752  * negative error code.
753  */
754 static int
755 xfs_bulk_ireq_setup(
756 	struct xfs_mount	*mp,
757 	struct xfs_bulk_ireq	*hdr,
758 	struct xfs_ibulk	*breq,
759 	void __user		*ubuffer)
760 {
761 	if (hdr->icount == 0 ||
762 	    (hdr->flags & ~XFS_BULK_IREQ_FLAGS_ALL) ||
763 	    memchr_inv(hdr->reserved, 0, sizeof(hdr->reserved)))
764 		return -EINVAL;
765 
766 	breq->startino = hdr->ino;
767 	breq->ubuffer = ubuffer;
768 	breq->icount = hdr->icount;
769 	breq->ocount = 0;
770 	breq->flags = 0;
771 
772 	/*
773 	 * The @ino parameter is a special value, so we must look it up here.
774 	 * We're not allowed to have IREQ_AGNO, and we only return one inode
775 	 * worth of data.
776 	 */
777 	if (hdr->flags & XFS_BULK_IREQ_SPECIAL) {
778 		if (hdr->flags & XFS_BULK_IREQ_AGNO)
779 			return -EINVAL;
780 
781 		switch (hdr->ino) {
782 		case XFS_BULK_IREQ_SPECIAL_ROOT:
783 			hdr->ino = mp->m_sb.sb_rootino;
784 			break;
785 		default:
786 			return -EINVAL;
787 		}
788 		breq->icount = 1;
789 	}
790 
791 	/*
792 	 * The IREQ_AGNO flag means that we only want results from a given AG.
793 	 * If @hdr->ino is zero, we start iterating in that AG.  If @hdr->ino is
794 	 * beyond the specified AG then we return no results.
795 	 */
796 	if (hdr->flags & XFS_BULK_IREQ_AGNO) {
797 		if (hdr->agno >= mp->m_sb.sb_agcount)
798 			return -EINVAL;
799 
800 		if (breq->startino == 0)
801 			breq->startino = XFS_AGINO_TO_INO(mp, hdr->agno, 0);
802 		else if (XFS_INO_TO_AGNO(mp, breq->startino) < hdr->agno)
803 			return -EINVAL;
804 
805 		breq->flags |= XFS_IBULK_SAME_AG;
806 
807 		/* Asking for an inode past the end of the AG?  We're done! */
808 		if (XFS_INO_TO_AGNO(mp, breq->startino) > hdr->agno)
809 			return -ECANCELED;
810 	} else if (hdr->agno)
811 		return -EINVAL;
812 
813 	/* Asking for an inode past the end of the FS?  We're done! */
814 	if (XFS_INO_TO_AGNO(mp, breq->startino) >= mp->m_sb.sb_agcount)
815 		return -ECANCELED;
816 
817 	if (hdr->flags & XFS_BULK_IREQ_NREXT64)
818 		breq->flags |= XFS_IBULK_NREXT64;
819 
820 	return 0;
821 }
822 
823 /*
824  * Update the userspace bulk request @hdr to reflect the end state of the
825  * internal bulk request @breq.
826  */
827 static void
828 xfs_bulk_ireq_teardown(
829 	struct xfs_bulk_ireq	*hdr,
830 	struct xfs_ibulk	*breq)
831 {
832 	hdr->ino = breq->startino;
833 	hdr->ocount = breq->ocount;
834 }
835 
836 /* Handle the v5 bulkstat ioctl. */
837 STATIC int
838 xfs_ioc_bulkstat(
839 	struct file			*file,
840 	unsigned int			cmd,
841 	struct xfs_bulkstat_req __user	*arg)
842 {
843 	struct xfs_mount		*mp = XFS_I(file_inode(file))->i_mount;
844 	struct xfs_bulk_ireq		hdr;
845 	struct xfs_ibulk		breq = {
846 		.mp			= mp,
847 		.mnt_userns		= file_mnt_user_ns(file),
848 	};
849 	int				error;
850 
851 	if (!capable(CAP_SYS_ADMIN))
852 		return -EPERM;
853 
854 	if (xfs_is_shutdown(mp))
855 		return -EIO;
856 
857 	if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
858 		return -EFAULT;
859 
860 	error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->bulkstat);
861 	if (error == -ECANCELED)
862 		goto out_teardown;
863 	if (error < 0)
864 		return error;
865 
866 	error = xfs_bulkstat(&breq, xfs_bulkstat_fmt);
867 	if (error)
868 		return error;
869 
870 out_teardown:
871 	xfs_bulk_ireq_teardown(&hdr, &breq);
872 	if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
873 		return -EFAULT;
874 
875 	return 0;
876 }
877 
878 STATIC int
879 xfs_inumbers_fmt(
880 	struct xfs_ibulk		*breq,
881 	const struct xfs_inumbers	*igrp)
882 {
883 	if (copy_to_user(breq->ubuffer, igrp, sizeof(struct xfs_inumbers)))
884 		return -EFAULT;
885 	return xfs_ibulk_advance(breq, sizeof(struct xfs_inumbers));
886 }
887 
888 /* Handle the v5 inumbers ioctl. */
889 STATIC int
890 xfs_ioc_inumbers(
891 	struct xfs_mount		*mp,
892 	unsigned int			cmd,
893 	struct xfs_inumbers_req __user	*arg)
894 {
895 	struct xfs_bulk_ireq		hdr;
896 	struct xfs_ibulk		breq = {
897 		.mp			= mp,
898 	};
899 	int				error;
900 
901 	if (!capable(CAP_SYS_ADMIN))
902 		return -EPERM;
903 
904 	if (xfs_is_shutdown(mp))
905 		return -EIO;
906 
907 	if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
908 		return -EFAULT;
909 
910 	error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->inumbers);
911 	if (error == -ECANCELED)
912 		goto out_teardown;
913 	if (error < 0)
914 		return error;
915 
916 	error = xfs_inumbers(&breq, xfs_inumbers_fmt);
917 	if (error)
918 		return error;
919 
920 out_teardown:
921 	xfs_bulk_ireq_teardown(&hdr, &breq);
922 	if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
923 		return -EFAULT;
924 
925 	return 0;
926 }
927 
928 STATIC int
929 xfs_ioc_fsgeometry(
930 	struct xfs_mount	*mp,
931 	void			__user *arg,
932 	int			struct_version)
933 {
934 	struct xfs_fsop_geom	fsgeo;
935 	size_t			len;
936 
937 	xfs_fs_geometry(mp, &fsgeo, struct_version);
938 
939 	if (struct_version <= 3)
940 		len = sizeof(struct xfs_fsop_geom_v1);
941 	else if (struct_version == 4)
942 		len = sizeof(struct xfs_fsop_geom_v4);
943 	else {
944 		xfs_fsop_geom_health(mp, &fsgeo);
945 		len = sizeof(fsgeo);
946 	}
947 
948 	if (copy_to_user(arg, &fsgeo, len))
949 		return -EFAULT;
950 	return 0;
951 }
952 
953 STATIC int
954 xfs_ioc_ag_geometry(
955 	struct xfs_mount	*mp,
956 	void			__user *arg)
957 {
958 	struct xfs_ag_geometry	ageo;
959 	int			error;
960 
961 	if (copy_from_user(&ageo, arg, sizeof(ageo)))
962 		return -EFAULT;
963 	if (ageo.ag_flags)
964 		return -EINVAL;
965 	if (memchr_inv(&ageo.ag_reserved, 0, sizeof(ageo.ag_reserved)))
966 		return -EINVAL;
967 
968 	error = xfs_ag_get_geometry(mp, ageo.ag_number, &ageo);
969 	if (error)
970 		return error;
971 
972 	if (copy_to_user(arg, &ageo, sizeof(ageo)))
973 		return -EFAULT;
974 	return 0;
975 }
976 
977 /*
978  * Linux extended inode flags interface.
979  */
980 
981 static void
982 xfs_fill_fsxattr(
983 	struct xfs_inode	*ip,
984 	int			whichfork,
985 	struct fileattr		*fa)
986 {
987 	struct xfs_mount	*mp = ip->i_mount;
988 	struct xfs_ifork	*ifp = XFS_IFORK_PTR(ip, whichfork);
989 
990 	fileattr_fill_xflags(fa, xfs_ip2xflags(ip));
991 
992 	if (ip->i_diflags & XFS_DIFLAG_EXTSIZE) {
993 		fa->fsx_extsize = XFS_FSB_TO_B(mp, ip->i_extsize);
994 	} else if (ip->i_diflags & XFS_DIFLAG_EXTSZINHERIT) {
995 		/*
996 		 * Don't let a misaligned extent size hint on a directory
997 		 * escape to userspace if it won't pass the setattr checks
998 		 * later.
999 		 */
1000 		if ((ip->i_diflags & XFS_DIFLAG_RTINHERIT) &&
1001 		    ip->i_extsize % mp->m_sb.sb_rextsize > 0) {
1002 			fa->fsx_xflags &= ~(FS_XFLAG_EXTSIZE |
1003 					    FS_XFLAG_EXTSZINHERIT);
1004 			fa->fsx_extsize = 0;
1005 		} else {
1006 			fa->fsx_extsize = XFS_FSB_TO_B(mp, ip->i_extsize);
1007 		}
1008 	}
1009 
1010 	if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
1011 		fa->fsx_cowextsize = XFS_FSB_TO_B(mp, ip->i_cowextsize);
1012 	fa->fsx_projid = ip->i_projid;
1013 	if (ifp && !xfs_need_iread_extents(ifp))
1014 		fa->fsx_nextents = xfs_iext_count(ifp);
1015 	else
1016 		fa->fsx_nextents = xfs_ifork_nextents(ifp);
1017 }
1018 
1019 STATIC int
1020 xfs_ioc_fsgetxattra(
1021 	xfs_inode_t		*ip,
1022 	void			__user *arg)
1023 {
1024 	struct fileattr		fa;
1025 
1026 	xfs_ilock(ip, XFS_ILOCK_SHARED);
1027 	xfs_fill_fsxattr(ip, XFS_ATTR_FORK, &fa);
1028 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
1029 
1030 	return copy_fsxattr_to_user(&fa, arg);
1031 }
1032 
1033 int
1034 xfs_fileattr_get(
1035 	struct dentry		*dentry,
1036 	struct fileattr		*fa)
1037 {
1038 	struct xfs_inode	*ip = XFS_I(d_inode(dentry));
1039 
1040 	if (d_is_special(dentry))
1041 		return -ENOTTY;
1042 
1043 	xfs_ilock(ip, XFS_ILOCK_SHARED);
1044 	xfs_fill_fsxattr(ip, XFS_DATA_FORK, fa);
1045 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
1046 
1047 	return 0;
1048 }
1049 
1050 STATIC uint16_t
1051 xfs_flags2diflags(
1052 	struct xfs_inode	*ip,
1053 	unsigned int		xflags)
1054 {
1055 	/* can't set PREALLOC this way, just preserve it */
1056 	uint16_t		di_flags =
1057 		(ip->i_diflags & XFS_DIFLAG_PREALLOC);
1058 
1059 	if (xflags & FS_XFLAG_IMMUTABLE)
1060 		di_flags |= XFS_DIFLAG_IMMUTABLE;
1061 	if (xflags & FS_XFLAG_APPEND)
1062 		di_flags |= XFS_DIFLAG_APPEND;
1063 	if (xflags & FS_XFLAG_SYNC)
1064 		di_flags |= XFS_DIFLAG_SYNC;
1065 	if (xflags & FS_XFLAG_NOATIME)
1066 		di_flags |= XFS_DIFLAG_NOATIME;
1067 	if (xflags & FS_XFLAG_NODUMP)
1068 		di_flags |= XFS_DIFLAG_NODUMP;
1069 	if (xflags & FS_XFLAG_NODEFRAG)
1070 		di_flags |= XFS_DIFLAG_NODEFRAG;
1071 	if (xflags & FS_XFLAG_FILESTREAM)
1072 		di_flags |= XFS_DIFLAG_FILESTREAM;
1073 	if (S_ISDIR(VFS_I(ip)->i_mode)) {
1074 		if (xflags & FS_XFLAG_RTINHERIT)
1075 			di_flags |= XFS_DIFLAG_RTINHERIT;
1076 		if (xflags & FS_XFLAG_NOSYMLINKS)
1077 			di_flags |= XFS_DIFLAG_NOSYMLINKS;
1078 		if (xflags & FS_XFLAG_EXTSZINHERIT)
1079 			di_flags |= XFS_DIFLAG_EXTSZINHERIT;
1080 		if (xflags & FS_XFLAG_PROJINHERIT)
1081 			di_flags |= XFS_DIFLAG_PROJINHERIT;
1082 	} else if (S_ISREG(VFS_I(ip)->i_mode)) {
1083 		if (xflags & FS_XFLAG_REALTIME)
1084 			di_flags |= XFS_DIFLAG_REALTIME;
1085 		if (xflags & FS_XFLAG_EXTSIZE)
1086 			di_flags |= XFS_DIFLAG_EXTSIZE;
1087 	}
1088 
1089 	return di_flags;
1090 }
1091 
1092 STATIC uint64_t
1093 xfs_flags2diflags2(
1094 	struct xfs_inode	*ip,
1095 	unsigned int		xflags)
1096 {
1097 	uint64_t		di_flags2 =
1098 		(ip->i_diflags2 & (XFS_DIFLAG2_REFLINK |
1099 				   XFS_DIFLAG2_BIGTIME |
1100 				   XFS_DIFLAG2_NREXT64));
1101 
1102 	if (xflags & FS_XFLAG_DAX)
1103 		di_flags2 |= XFS_DIFLAG2_DAX;
1104 	if (xflags & FS_XFLAG_COWEXTSIZE)
1105 		di_flags2 |= XFS_DIFLAG2_COWEXTSIZE;
1106 
1107 	return di_flags2;
1108 }
1109 
1110 static int
1111 xfs_ioctl_setattr_xflags(
1112 	struct xfs_trans	*tp,
1113 	struct xfs_inode	*ip,
1114 	struct fileattr		*fa)
1115 {
1116 	struct xfs_mount	*mp = ip->i_mount;
1117 	uint64_t		i_flags2;
1118 
1119 	/* Can't change realtime flag if any extents are allocated. */
1120 	if ((ip->i_df.if_nextents || ip->i_delayed_blks) &&
1121 	    XFS_IS_REALTIME_INODE(ip) != (fa->fsx_xflags & FS_XFLAG_REALTIME))
1122 		return -EINVAL;
1123 
1124 	/* If realtime flag is set then must have realtime device */
1125 	if (fa->fsx_xflags & FS_XFLAG_REALTIME) {
1126 		if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
1127 		    (ip->i_extsize % mp->m_sb.sb_rextsize))
1128 			return -EINVAL;
1129 	}
1130 
1131 	/* Clear reflink if we are actually able to set the rt flag. */
1132 	if ((fa->fsx_xflags & FS_XFLAG_REALTIME) && xfs_is_reflink_inode(ip))
1133 		ip->i_diflags2 &= ~XFS_DIFLAG2_REFLINK;
1134 
1135 	/* Don't allow us to set DAX mode for a reflinked file for now. */
1136 	if ((fa->fsx_xflags & FS_XFLAG_DAX) && xfs_is_reflink_inode(ip))
1137 		return -EINVAL;
1138 
1139 	/* diflags2 only valid for v3 inodes. */
1140 	i_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1141 	if (i_flags2 && !xfs_has_v3inodes(mp))
1142 		return -EINVAL;
1143 
1144 	ip->i_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
1145 	ip->i_diflags2 = i_flags2;
1146 
1147 	xfs_diflags_to_iflags(ip, false);
1148 	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1149 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1150 	XFS_STATS_INC(mp, xs_ig_attrchg);
1151 	return 0;
1152 }
1153 
1154 static void
1155 xfs_ioctl_setattr_prepare_dax(
1156 	struct xfs_inode	*ip,
1157 	struct fileattr		*fa)
1158 {
1159 	struct xfs_mount	*mp = ip->i_mount;
1160 	struct inode            *inode = VFS_I(ip);
1161 
1162 	if (S_ISDIR(inode->i_mode))
1163 		return;
1164 
1165 	if (xfs_has_dax_always(mp) || xfs_has_dax_never(mp))
1166 		return;
1167 
1168 	if (((fa->fsx_xflags & FS_XFLAG_DAX) &&
1169 	    !(ip->i_diflags2 & XFS_DIFLAG2_DAX)) ||
1170 	    (!(fa->fsx_xflags & FS_XFLAG_DAX) &&
1171 	     (ip->i_diflags2 & XFS_DIFLAG2_DAX)))
1172 		d_mark_dontcache(inode);
1173 }
1174 
1175 /*
1176  * Set up the transaction structure for the setattr operation, checking that we
1177  * have permission to do so. On success, return a clean transaction and the
1178  * inode locked exclusively ready for further operation specific checks. On
1179  * failure, return an error without modifying or locking the inode.
1180  */
1181 static struct xfs_trans *
1182 xfs_ioctl_setattr_get_trans(
1183 	struct xfs_inode	*ip,
1184 	struct xfs_dquot	*pdqp)
1185 {
1186 	struct xfs_mount	*mp = ip->i_mount;
1187 	struct xfs_trans	*tp;
1188 	int			error = -EROFS;
1189 
1190 	if (xfs_is_readonly(mp))
1191 		goto out_error;
1192 	error = -EIO;
1193 	if (xfs_is_shutdown(mp))
1194 		goto out_error;
1195 
1196 	error = xfs_trans_alloc_ichange(ip, NULL, NULL, pdqp,
1197 			has_capability_noaudit(current, CAP_FOWNER), &tp);
1198 	if (error)
1199 		goto out_error;
1200 
1201 	if (xfs_has_wsync(mp))
1202 		xfs_trans_set_sync(tp);
1203 
1204 	return tp;
1205 
1206 out_error:
1207 	return ERR_PTR(error);
1208 }
1209 
1210 /*
1211  * Validate a proposed extent size hint.  For regular files, the hint can only
1212  * be changed if no extents are allocated.
1213  */
1214 static int
1215 xfs_ioctl_setattr_check_extsize(
1216 	struct xfs_inode	*ip,
1217 	struct fileattr		*fa)
1218 {
1219 	struct xfs_mount	*mp = ip->i_mount;
1220 	xfs_failaddr_t		failaddr;
1221 	uint16_t		new_diflags;
1222 
1223 	if (!fa->fsx_valid)
1224 		return 0;
1225 
1226 	if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_df.if_nextents &&
1227 	    XFS_FSB_TO_B(mp, ip->i_extsize) != fa->fsx_extsize)
1228 		return -EINVAL;
1229 
1230 	if (fa->fsx_extsize & mp->m_blockmask)
1231 		return -EINVAL;
1232 
1233 	new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
1234 
1235 	/*
1236 	 * Inode verifiers do not check that the extent size hint is an integer
1237 	 * multiple of the rt extent size on a directory with both rtinherit
1238 	 * and extszinherit flags set.  Don't let sysadmins misconfigure
1239 	 * directories.
1240 	 */
1241 	if ((new_diflags & XFS_DIFLAG_RTINHERIT) &&
1242 	    (new_diflags & XFS_DIFLAG_EXTSZINHERIT)) {
1243 		unsigned int	rtextsize_bytes;
1244 
1245 		rtextsize_bytes = XFS_FSB_TO_B(mp, mp->m_sb.sb_rextsize);
1246 		if (fa->fsx_extsize % rtextsize_bytes)
1247 			return -EINVAL;
1248 	}
1249 
1250 	failaddr = xfs_inode_validate_extsize(ip->i_mount,
1251 			XFS_B_TO_FSB(mp, fa->fsx_extsize),
1252 			VFS_I(ip)->i_mode, new_diflags);
1253 	return failaddr != NULL ? -EINVAL : 0;
1254 }
1255 
1256 static int
1257 xfs_ioctl_setattr_check_cowextsize(
1258 	struct xfs_inode	*ip,
1259 	struct fileattr		*fa)
1260 {
1261 	struct xfs_mount	*mp = ip->i_mount;
1262 	xfs_failaddr_t		failaddr;
1263 	uint64_t		new_diflags2;
1264 	uint16_t		new_diflags;
1265 
1266 	if (!fa->fsx_valid)
1267 		return 0;
1268 
1269 	if (fa->fsx_cowextsize & mp->m_blockmask)
1270 		return -EINVAL;
1271 
1272 	new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
1273 	new_diflags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1274 
1275 	failaddr = xfs_inode_validate_cowextsize(ip->i_mount,
1276 			XFS_B_TO_FSB(mp, fa->fsx_cowextsize),
1277 			VFS_I(ip)->i_mode, new_diflags, new_diflags2);
1278 	return failaddr != NULL ? -EINVAL : 0;
1279 }
1280 
1281 static int
1282 xfs_ioctl_setattr_check_projid(
1283 	struct xfs_inode	*ip,
1284 	struct fileattr		*fa)
1285 {
1286 	if (!fa->fsx_valid)
1287 		return 0;
1288 
1289 	/* Disallow 32bit project ids if 32bit IDs are not enabled. */
1290 	if (fa->fsx_projid > (uint16_t)-1 &&
1291 	    !xfs_has_projid32(ip->i_mount))
1292 		return -EINVAL;
1293 	return 0;
1294 }
1295 
1296 int
1297 xfs_fileattr_set(
1298 	struct user_namespace	*mnt_userns,
1299 	struct dentry		*dentry,
1300 	struct fileattr		*fa)
1301 {
1302 	struct xfs_inode	*ip = XFS_I(d_inode(dentry));
1303 	struct xfs_mount	*mp = ip->i_mount;
1304 	struct xfs_trans	*tp;
1305 	struct xfs_dquot	*pdqp = NULL;
1306 	struct xfs_dquot	*olddquot = NULL;
1307 	int			error;
1308 
1309 	trace_xfs_ioctl_setattr(ip);
1310 
1311 	if (d_is_special(dentry))
1312 		return -ENOTTY;
1313 
1314 	if (!fa->fsx_valid) {
1315 		if (fa->flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL |
1316 				  FS_NOATIME_FL | FS_NODUMP_FL |
1317 				  FS_SYNC_FL | FS_DAX_FL | FS_PROJINHERIT_FL))
1318 			return -EOPNOTSUPP;
1319 	}
1320 
1321 	error = xfs_ioctl_setattr_check_projid(ip, fa);
1322 	if (error)
1323 		return error;
1324 
1325 	/*
1326 	 * If disk quotas is on, we make sure that the dquots do exist on disk,
1327 	 * before we start any other transactions. Trying to do this later
1328 	 * is messy. We don't care to take a readlock to look at the ids
1329 	 * in inode here, because we can't hold it across the trans_reserve.
1330 	 * If the IDs do change before we take the ilock, we're covered
1331 	 * because the i_*dquot fields will get updated anyway.
1332 	 */
1333 	if (fa->fsx_valid && XFS_IS_QUOTA_ON(mp)) {
1334 		error = xfs_qm_vop_dqalloc(ip, VFS_I(ip)->i_uid,
1335 				VFS_I(ip)->i_gid, fa->fsx_projid,
1336 				XFS_QMOPT_PQUOTA, NULL, NULL, &pdqp);
1337 		if (error)
1338 			return error;
1339 	}
1340 
1341 	xfs_ioctl_setattr_prepare_dax(ip, fa);
1342 
1343 	tp = xfs_ioctl_setattr_get_trans(ip, pdqp);
1344 	if (IS_ERR(tp)) {
1345 		error = PTR_ERR(tp);
1346 		goto error_free_dquots;
1347 	}
1348 
1349 	error = xfs_ioctl_setattr_check_extsize(ip, fa);
1350 	if (error)
1351 		goto error_trans_cancel;
1352 
1353 	error = xfs_ioctl_setattr_check_cowextsize(ip, fa);
1354 	if (error)
1355 		goto error_trans_cancel;
1356 
1357 	error = xfs_ioctl_setattr_xflags(tp, ip, fa);
1358 	if (error)
1359 		goto error_trans_cancel;
1360 
1361 	if (!fa->fsx_valid)
1362 		goto skip_xattr;
1363 	/*
1364 	 * Change file ownership.  Must be the owner or privileged.  CAP_FSETID
1365 	 * overrides the following restrictions:
1366 	 *
1367 	 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1368 	 * successful return from chown()
1369 	 */
1370 
1371 	if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
1372 	    !capable_wrt_inode_uidgid(mnt_userns, VFS_I(ip), CAP_FSETID))
1373 		VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
1374 
1375 	/* Change the ownerships and register project quota modifications */
1376 	if (ip->i_projid != fa->fsx_projid) {
1377 		if (XFS_IS_PQUOTA_ON(mp)) {
1378 			olddquot = xfs_qm_vop_chown(tp, ip,
1379 						&ip->i_pdquot, pdqp);
1380 		}
1381 		ip->i_projid = fa->fsx_projid;
1382 	}
1383 
1384 	/*
1385 	 * Only set the extent size hint if we've already determined that the
1386 	 * extent size hint should be set on the inode. If no extent size flags
1387 	 * are set on the inode then unconditionally clear the extent size hint.
1388 	 */
1389 	if (ip->i_diflags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1390 		ip->i_extsize = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1391 	else
1392 		ip->i_extsize = 0;
1393 
1394 	if (xfs_has_v3inodes(mp)) {
1395 		if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
1396 			ip->i_cowextsize = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
1397 		else
1398 			ip->i_cowextsize = 0;
1399 	}
1400 
1401 skip_xattr:
1402 	error = xfs_trans_commit(tp);
1403 
1404 	/*
1405 	 * Release any dquot(s) the inode had kept before chown.
1406 	 */
1407 	xfs_qm_dqrele(olddquot);
1408 	xfs_qm_dqrele(pdqp);
1409 
1410 	return error;
1411 
1412 error_trans_cancel:
1413 	xfs_trans_cancel(tp);
1414 error_free_dquots:
1415 	xfs_qm_dqrele(pdqp);
1416 	return error;
1417 }
1418 
1419 static bool
1420 xfs_getbmap_format(
1421 	struct kgetbmap		*p,
1422 	struct getbmapx __user	*u,
1423 	size_t			recsize)
1424 {
1425 	if (put_user(p->bmv_offset, &u->bmv_offset) ||
1426 	    put_user(p->bmv_block, &u->bmv_block) ||
1427 	    put_user(p->bmv_length, &u->bmv_length) ||
1428 	    put_user(0, &u->bmv_count) ||
1429 	    put_user(0, &u->bmv_entries))
1430 		return false;
1431 	if (recsize < sizeof(struct getbmapx))
1432 		return true;
1433 	if (put_user(0, &u->bmv_iflags) ||
1434 	    put_user(p->bmv_oflags, &u->bmv_oflags) ||
1435 	    put_user(0, &u->bmv_unused1) ||
1436 	    put_user(0, &u->bmv_unused2))
1437 		return false;
1438 	return true;
1439 }
1440 
1441 STATIC int
1442 xfs_ioc_getbmap(
1443 	struct file		*file,
1444 	unsigned int		cmd,
1445 	void			__user *arg)
1446 {
1447 	struct getbmapx		bmx = { 0 };
1448 	struct kgetbmap		*buf;
1449 	size_t			recsize;
1450 	int			error, i;
1451 
1452 	switch (cmd) {
1453 	case XFS_IOC_GETBMAPA:
1454 		bmx.bmv_iflags = BMV_IF_ATTRFORK;
1455 		fallthrough;
1456 	case XFS_IOC_GETBMAP:
1457 		/* struct getbmap is a strict subset of struct getbmapx. */
1458 		recsize = sizeof(struct getbmap);
1459 		break;
1460 	case XFS_IOC_GETBMAPX:
1461 		recsize = sizeof(struct getbmapx);
1462 		break;
1463 	default:
1464 		return -EINVAL;
1465 	}
1466 
1467 	if (copy_from_user(&bmx, arg, recsize))
1468 		return -EFAULT;
1469 
1470 	if (bmx.bmv_count < 2)
1471 		return -EINVAL;
1472 	if (bmx.bmv_count >= INT_MAX / recsize)
1473 		return -ENOMEM;
1474 
1475 	buf = kvcalloc(bmx.bmv_count, sizeof(*buf), GFP_KERNEL);
1476 	if (!buf)
1477 		return -ENOMEM;
1478 
1479 	error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, buf);
1480 	if (error)
1481 		goto out_free_buf;
1482 
1483 	error = -EFAULT;
1484 	if (copy_to_user(arg, &bmx, recsize))
1485 		goto out_free_buf;
1486 	arg += recsize;
1487 
1488 	for (i = 0; i < bmx.bmv_entries; i++) {
1489 		if (!xfs_getbmap_format(buf + i, arg, recsize))
1490 			goto out_free_buf;
1491 		arg += recsize;
1492 	}
1493 
1494 	error = 0;
1495 out_free_buf:
1496 	kmem_free(buf);
1497 	return error;
1498 }
1499 
1500 STATIC int
1501 xfs_ioc_getfsmap(
1502 	struct xfs_inode	*ip,
1503 	struct fsmap_head	__user *arg)
1504 {
1505 	struct xfs_fsmap_head	xhead = {0};
1506 	struct fsmap_head	head;
1507 	struct fsmap		*recs;
1508 	unsigned int		count;
1509 	__u32			last_flags = 0;
1510 	bool			done = false;
1511 	int			error;
1512 
1513 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
1514 		return -EFAULT;
1515 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
1516 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
1517 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
1518 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
1519 		       sizeof(head.fmh_keys[1].fmr_reserved)))
1520 		return -EINVAL;
1521 
1522 	/*
1523 	 * Use an internal memory buffer so that we don't have to copy fsmap
1524 	 * data to userspace while holding locks.  Start by trying to allocate
1525 	 * up to 128k for the buffer, but fall back to a single page if needed.
1526 	 */
1527 	count = min_t(unsigned int, head.fmh_count,
1528 			131072 / sizeof(struct fsmap));
1529 	recs = kvcalloc(count, sizeof(struct fsmap), GFP_KERNEL);
1530 	if (!recs) {
1531 		count = min_t(unsigned int, head.fmh_count,
1532 				PAGE_SIZE / sizeof(struct fsmap));
1533 		recs = kvcalloc(count, sizeof(struct fsmap), GFP_KERNEL);
1534 		if (!recs)
1535 			return -ENOMEM;
1536 	}
1537 
1538 	xhead.fmh_iflags = head.fmh_iflags;
1539 	xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
1540 	xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
1541 
1542 	trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1543 	trace_xfs_getfsmap_high_key(ip->i_mount, &xhead.fmh_keys[1]);
1544 
1545 	head.fmh_entries = 0;
1546 	do {
1547 		struct fsmap __user	*user_recs;
1548 		struct fsmap		*last_rec;
1549 
1550 		user_recs = &arg->fmh_recs[head.fmh_entries];
1551 		xhead.fmh_entries = 0;
1552 		xhead.fmh_count = min_t(unsigned int, count,
1553 					head.fmh_count - head.fmh_entries);
1554 
1555 		/* Run query, record how many entries we got. */
1556 		error = xfs_getfsmap(ip->i_mount, &xhead, recs);
1557 		switch (error) {
1558 		case 0:
1559 			/*
1560 			 * There are no more records in the result set.  Copy
1561 			 * whatever we got to userspace and break out.
1562 			 */
1563 			done = true;
1564 			break;
1565 		case -ECANCELED:
1566 			/*
1567 			 * The internal memory buffer is full.  Copy whatever
1568 			 * records we got to userspace and go again if we have
1569 			 * not yet filled the userspace buffer.
1570 			 */
1571 			error = 0;
1572 			break;
1573 		default:
1574 			goto out_free;
1575 		}
1576 		head.fmh_entries += xhead.fmh_entries;
1577 		head.fmh_oflags = xhead.fmh_oflags;
1578 
1579 		/*
1580 		 * If the caller wanted a record count or there aren't any
1581 		 * new records to return, we're done.
1582 		 */
1583 		if (head.fmh_count == 0 || xhead.fmh_entries == 0)
1584 			break;
1585 
1586 		/* Copy all the records we got out to userspace. */
1587 		if (copy_to_user(user_recs, recs,
1588 				 xhead.fmh_entries * sizeof(struct fsmap))) {
1589 			error = -EFAULT;
1590 			goto out_free;
1591 		}
1592 
1593 		/* Remember the last record flags we copied to userspace. */
1594 		last_rec = &recs[xhead.fmh_entries - 1];
1595 		last_flags = last_rec->fmr_flags;
1596 
1597 		/* Set up the low key for the next iteration. */
1598 		xfs_fsmap_to_internal(&xhead.fmh_keys[0], last_rec);
1599 		trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1600 	} while (!done && head.fmh_entries < head.fmh_count);
1601 
1602 	/*
1603 	 * If there are no more records in the query result set and we're not
1604 	 * in counting mode, mark the last record returned with the LAST flag.
1605 	 */
1606 	if (done && head.fmh_count > 0 && head.fmh_entries > 0) {
1607 		struct fsmap __user	*user_rec;
1608 
1609 		last_flags |= FMR_OF_LAST;
1610 		user_rec = &arg->fmh_recs[head.fmh_entries - 1];
1611 
1612 		if (copy_to_user(&user_rec->fmr_flags, &last_flags,
1613 					sizeof(last_flags))) {
1614 			error = -EFAULT;
1615 			goto out_free;
1616 		}
1617 	}
1618 
1619 	/* copy back header */
1620 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head))) {
1621 		error = -EFAULT;
1622 		goto out_free;
1623 	}
1624 
1625 out_free:
1626 	kmem_free(recs);
1627 	return error;
1628 }
1629 
1630 STATIC int
1631 xfs_ioc_scrub_metadata(
1632 	struct file			*file,
1633 	void				__user *arg)
1634 {
1635 	struct xfs_scrub_metadata	scrub;
1636 	int				error;
1637 
1638 	if (!capable(CAP_SYS_ADMIN))
1639 		return -EPERM;
1640 
1641 	if (copy_from_user(&scrub, arg, sizeof(scrub)))
1642 		return -EFAULT;
1643 
1644 	error = xfs_scrub_metadata(file, &scrub);
1645 	if (error)
1646 		return error;
1647 
1648 	if (copy_to_user(arg, &scrub, sizeof(scrub)))
1649 		return -EFAULT;
1650 
1651 	return 0;
1652 }
1653 
1654 int
1655 xfs_ioc_swapext(
1656 	xfs_swapext_t	*sxp)
1657 {
1658 	xfs_inode_t     *ip, *tip;
1659 	struct fd	f, tmp;
1660 	int		error = 0;
1661 
1662 	/* Pull information for the target fd */
1663 	f = fdget((int)sxp->sx_fdtarget);
1664 	if (!f.file) {
1665 		error = -EINVAL;
1666 		goto out;
1667 	}
1668 
1669 	if (!(f.file->f_mode & FMODE_WRITE) ||
1670 	    !(f.file->f_mode & FMODE_READ) ||
1671 	    (f.file->f_flags & O_APPEND)) {
1672 		error = -EBADF;
1673 		goto out_put_file;
1674 	}
1675 
1676 	tmp = fdget((int)sxp->sx_fdtmp);
1677 	if (!tmp.file) {
1678 		error = -EINVAL;
1679 		goto out_put_file;
1680 	}
1681 
1682 	if (!(tmp.file->f_mode & FMODE_WRITE) ||
1683 	    !(tmp.file->f_mode & FMODE_READ) ||
1684 	    (tmp.file->f_flags & O_APPEND)) {
1685 		error = -EBADF;
1686 		goto out_put_tmp_file;
1687 	}
1688 
1689 	if (IS_SWAPFILE(file_inode(f.file)) ||
1690 	    IS_SWAPFILE(file_inode(tmp.file))) {
1691 		error = -EINVAL;
1692 		goto out_put_tmp_file;
1693 	}
1694 
1695 	/*
1696 	 * We need to ensure that the fds passed in point to XFS inodes
1697 	 * before we cast and access them as XFS structures as we have no
1698 	 * control over what the user passes us here.
1699 	 */
1700 	if (f.file->f_op != &xfs_file_operations ||
1701 	    tmp.file->f_op != &xfs_file_operations) {
1702 		error = -EINVAL;
1703 		goto out_put_tmp_file;
1704 	}
1705 
1706 	ip = XFS_I(file_inode(f.file));
1707 	tip = XFS_I(file_inode(tmp.file));
1708 
1709 	if (ip->i_mount != tip->i_mount) {
1710 		error = -EINVAL;
1711 		goto out_put_tmp_file;
1712 	}
1713 
1714 	if (ip->i_ino == tip->i_ino) {
1715 		error = -EINVAL;
1716 		goto out_put_tmp_file;
1717 	}
1718 
1719 	if (xfs_is_shutdown(ip->i_mount)) {
1720 		error = -EIO;
1721 		goto out_put_tmp_file;
1722 	}
1723 
1724 	error = xfs_swap_extents(ip, tip, sxp);
1725 
1726  out_put_tmp_file:
1727 	fdput(tmp);
1728  out_put_file:
1729 	fdput(f);
1730  out:
1731 	return error;
1732 }
1733 
1734 static int
1735 xfs_ioc_getlabel(
1736 	struct xfs_mount	*mp,
1737 	char			__user *user_label)
1738 {
1739 	struct xfs_sb		*sbp = &mp->m_sb;
1740 	char			label[XFSLABEL_MAX + 1];
1741 
1742 	/* Paranoia */
1743 	BUILD_BUG_ON(sizeof(sbp->sb_fname) > FSLABEL_MAX);
1744 
1745 	/* 1 larger than sb_fname, so this ensures a trailing NUL char */
1746 	memset(label, 0, sizeof(label));
1747 	spin_lock(&mp->m_sb_lock);
1748 	strncpy(label, sbp->sb_fname, XFSLABEL_MAX);
1749 	spin_unlock(&mp->m_sb_lock);
1750 
1751 	if (copy_to_user(user_label, label, sizeof(label)))
1752 		return -EFAULT;
1753 	return 0;
1754 }
1755 
1756 static int
1757 xfs_ioc_setlabel(
1758 	struct file		*filp,
1759 	struct xfs_mount	*mp,
1760 	char			__user *newlabel)
1761 {
1762 	struct xfs_sb		*sbp = &mp->m_sb;
1763 	char			label[XFSLABEL_MAX + 1];
1764 	size_t			len;
1765 	int			error;
1766 
1767 	if (!capable(CAP_SYS_ADMIN))
1768 		return -EPERM;
1769 	/*
1770 	 * The generic ioctl allows up to FSLABEL_MAX chars, but XFS is much
1771 	 * smaller, at 12 bytes.  We copy one more to be sure we find the
1772 	 * (required) NULL character to test the incoming label length.
1773 	 * NB: The on disk label doesn't need to be null terminated.
1774 	 */
1775 	if (copy_from_user(label, newlabel, XFSLABEL_MAX + 1))
1776 		return -EFAULT;
1777 	len = strnlen(label, XFSLABEL_MAX + 1);
1778 	if (len > sizeof(sbp->sb_fname))
1779 		return -EINVAL;
1780 
1781 	error = mnt_want_write_file(filp);
1782 	if (error)
1783 		return error;
1784 
1785 	spin_lock(&mp->m_sb_lock);
1786 	memset(sbp->sb_fname, 0, sizeof(sbp->sb_fname));
1787 	memcpy(sbp->sb_fname, label, len);
1788 	spin_unlock(&mp->m_sb_lock);
1789 
1790 	/*
1791 	 * Now we do several things to satisfy userspace.
1792 	 * In addition to normal logging of the primary superblock, we also
1793 	 * immediately write these changes to sector zero for the primary, then
1794 	 * update all backup supers (as xfs_db does for a label change), then
1795 	 * invalidate the block device page cache.  This is so that any prior
1796 	 * buffered reads from userspace (i.e. from blkid) are invalidated,
1797 	 * and userspace will see the newly-written label.
1798 	 */
1799 	error = xfs_sync_sb_buf(mp);
1800 	if (error)
1801 		goto out;
1802 	/*
1803 	 * growfs also updates backup supers so lock against that.
1804 	 */
1805 	mutex_lock(&mp->m_growlock);
1806 	error = xfs_update_secondary_sbs(mp);
1807 	mutex_unlock(&mp->m_growlock);
1808 
1809 	invalidate_bdev(mp->m_ddev_targp->bt_bdev);
1810 
1811 out:
1812 	mnt_drop_write_file(filp);
1813 	return error;
1814 }
1815 
1816 static inline int
1817 xfs_fs_eofblocks_from_user(
1818 	struct xfs_fs_eofblocks		*src,
1819 	struct xfs_icwalk		*dst)
1820 {
1821 	if (src->eof_version != XFS_EOFBLOCKS_VERSION)
1822 		return -EINVAL;
1823 
1824 	if (src->eof_flags & ~XFS_EOF_FLAGS_VALID)
1825 		return -EINVAL;
1826 
1827 	if (memchr_inv(&src->pad32, 0, sizeof(src->pad32)) ||
1828 	    memchr_inv(src->pad64, 0, sizeof(src->pad64)))
1829 		return -EINVAL;
1830 
1831 	dst->icw_flags = 0;
1832 	if (src->eof_flags & XFS_EOF_FLAGS_SYNC)
1833 		dst->icw_flags |= XFS_ICWALK_FLAG_SYNC;
1834 	if (src->eof_flags & XFS_EOF_FLAGS_UID)
1835 		dst->icw_flags |= XFS_ICWALK_FLAG_UID;
1836 	if (src->eof_flags & XFS_EOF_FLAGS_GID)
1837 		dst->icw_flags |= XFS_ICWALK_FLAG_GID;
1838 	if (src->eof_flags & XFS_EOF_FLAGS_PRID)
1839 		dst->icw_flags |= XFS_ICWALK_FLAG_PRID;
1840 	if (src->eof_flags & XFS_EOF_FLAGS_MINFILESIZE)
1841 		dst->icw_flags |= XFS_ICWALK_FLAG_MINFILESIZE;
1842 
1843 	dst->icw_prid = src->eof_prid;
1844 	dst->icw_min_file_size = src->eof_min_file_size;
1845 
1846 	dst->icw_uid = INVALID_UID;
1847 	if (src->eof_flags & XFS_EOF_FLAGS_UID) {
1848 		dst->icw_uid = make_kuid(current_user_ns(), src->eof_uid);
1849 		if (!uid_valid(dst->icw_uid))
1850 			return -EINVAL;
1851 	}
1852 
1853 	dst->icw_gid = INVALID_GID;
1854 	if (src->eof_flags & XFS_EOF_FLAGS_GID) {
1855 		dst->icw_gid = make_kgid(current_user_ns(), src->eof_gid);
1856 		if (!gid_valid(dst->icw_gid))
1857 			return -EINVAL;
1858 	}
1859 	return 0;
1860 }
1861 
1862 /*
1863  * These long-unused ioctls were removed from the official ioctl API in 5.17,
1864  * but retain these definitions so that we can log warnings about them.
1865  */
1866 #define XFS_IOC_ALLOCSP		_IOW ('X', 10, struct xfs_flock64)
1867 #define XFS_IOC_FREESP		_IOW ('X', 11, struct xfs_flock64)
1868 #define XFS_IOC_ALLOCSP64	_IOW ('X', 36, struct xfs_flock64)
1869 #define XFS_IOC_FREESP64	_IOW ('X', 37, struct xfs_flock64)
1870 
1871 /*
1872  * Note: some of the ioctl's return positive numbers as a
1873  * byte count indicating success, such as readlink_by_handle.
1874  * So we don't "sign flip" like most other routines.  This means
1875  * true errors need to be returned as a negative value.
1876  */
1877 long
1878 xfs_file_ioctl(
1879 	struct file		*filp,
1880 	unsigned int		cmd,
1881 	unsigned long		p)
1882 {
1883 	struct inode		*inode = file_inode(filp);
1884 	struct xfs_inode	*ip = XFS_I(inode);
1885 	struct xfs_mount	*mp = ip->i_mount;
1886 	void			__user *arg = (void __user *)p;
1887 	int			error;
1888 
1889 	trace_xfs_file_ioctl(ip);
1890 
1891 	switch (cmd) {
1892 	case FITRIM:
1893 		return xfs_ioc_trim(mp, arg);
1894 	case FS_IOC_GETFSLABEL:
1895 		return xfs_ioc_getlabel(mp, arg);
1896 	case FS_IOC_SETFSLABEL:
1897 		return xfs_ioc_setlabel(filp, mp, arg);
1898 	case XFS_IOC_ALLOCSP:
1899 	case XFS_IOC_FREESP:
1900 	case XFS_IOC_ALLOCSP64:
1901 	case XFS_IOC_FREESP64:
1902 		xfs_warn_once(mp,
1903 	"%s should use fallocate; XFS_IOC_{ALLOC,FREE}SP ioctl unsupported",
1904 				current->comm);
1905 		return -ENOTTY;
1906 	case XFS_IOC_DIOINFO: {
1907 		struct xfs_buftarg	*target = xfs_inode_buftarg(ip);
1908 		struct dioattr		da;
1909 
1910 		da.d_mem =  da.d_miniosz = target->bt_logical_sectorsize;
1911 		da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1912 
1913 		if (copy_to_user(arg, &da, sizeof(da)))
1914 			return -EFAULT;
1915 		return 0;
1916 	}
1917 
1918 	case XFS_IOC_FSBULKSTAT_SINGLE:
1919 	case XFS_IOC_FSBULKSTAT:
1920 	case XFS_IOC_FSINUMBERS:
1921 		return xfs_ioc_fsbulkstat(filp, cmd, arg);
1922 
1923 	case XFS_IOC_BULKSTAT:
1924 		return xfs_ioc_bulkstat(filp, cmd, arg);
1925 	case XFS_IOC_INUMBERS:
1926 		return xfs_ioc_inumbers(mp, cmd, arg);
1927 
1928 	case XFS_IOC_FSGEOMETRY_V1:
1929 		return xfs_ioc_fsgeometry(mp, arg, 3);
1930 	case XFS_IOC_FSGEOMETRY_V4:
1931 		return xfs_ioc_fsgeometry(mp, arg, 4);
1932 	case XFS_IOC_FSGEOMETRY:
1933 		return xfs_ioc_fsgeometry(mp, arg, 5);
1934 
1935 	case XFS_IOC_AG_GEOMETRY:
1936 		return xfs_ioc_ag_geometry(mp, arg);
1937 
1938 	case XFS_IOC_GETVERSION:
1939 		return put_user(inode->i_generation, (int __user *)arg);
1940 
1941 	case XFS_IOC_FSGETXATTRA:
1942 		return xfs_ioc_fsgetxattra(ip, arg);
1943 
1944 	case XFS_IOC_GETBMAP:
1945 	case XFS_IOC_GETBMAPA:
1946 	case XFS_IOC_GETBMAPX:
1947 		return xfs_ioc_getbmap(filp, cmd, arg);
1948 
1949 	case FS_IOC_GETFSMAP:
1950 		return xfs_ioc_getfsmap(ip, arg);
1951 
1952 	case XFS_IOC_SCRUB_METADATA:
1953 		return xfs_ioc_scrub_metadata(filp, arg);
1954 
1955 	case XFS_IOC_FD_TO_HANDLE:
1956 	case XFS_IOC_PATH_TO_HANDLE:
1957 	case XFS_IOC_PATH_TO_FSHANDLE: {
1958 		xfs_fsop_handlereq_t	hreq;
1959 
1960 		if (copy_from_user(&hreq, arg, sizeof(hreq)))
1961 			return -EFAULT;
1962 		return xfs_find_handle(cmd, &hreq);
1963 	}
1964 	case XFS_IOC_OPEN_BY_HANDLE: {
1965 		xfs_fsop_handlereq_t	hreq;
1966 
1967 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1968 			return -EFAULT;
1969 		return xfs_open_by_handle(filp, &hreq);
1970 	}
1971 
1972 	case XFS_IOC_READLINK_BY_HANDLE: {
1973 		xfs_fsop_handlereq_t	hreq;
1974 
1975 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1976 			return -EFAULT;
1977 		return xfs_readlink_by_handle(filp, &hreq);
1978 	}
1979 	case XFS_IOC_ATTRLIST_BY_HANDLE:
1980 		return xfs_attrlist_by_handle(filp, arg);
1981 
1982 	case XFS_IOC_ATTRMULTI_BY_HANDLE:
1983 		return xfs_attrmulti_by_handle(filp, arg);
1984 
1985 	case XFS_IOC_SWAPEXT: {
1986 		struct xfs_swapext	sxp;
1987 
1988 		if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
1989 			return -EFAULT;
1990 		error = mnt_want_write_file(filp);
1991 		if (error)
1992 			return error;
1993 		error = xfs_ioc_swapext(&sxp);
1994 		mnt_drop_write_file(filp);
1995 		return error;
1996 	}
1997 
1998 	case XFS_IOC_FSCOUNTS: {
1999 		xfs_fsop_counts_t out;
2000 
2001 		xfs_fs_counts(mp, &out);
2002 
2003 		if (copy_to_user(arg, &out, sizeof(out)))
2004 			return -EFAULT;
2005 		return 0;
2006 	}
2007 
2008 	case XFS_IOC_SET_RESBLKS: {
2009 		xfs_fsop_resblks_t inout;
2010 		uint64_t	   in;
2011 
2012 		if (!capable(CAP_SYS_ADMIN))
2013 			return -EPERM;
2014 
2015 		if (xfs_is_readonly(mp))
2016 			return -EROFS;
2017 
2018 		if (copy_from_user(&inout, arg, sizeof(inout)))
2019 			return -EFAULT;
2020 
2021 		error = mnt_want_write_file(filp);
2022 		if (error)
2023 			return error;
2024 
2025 		/* input parameter is passed in resblks field of structure */
2026 		in = inout.resblks;
2027 		error = xfs_reserve_blocks(mp, &in, &inout);
2028 		mnt_drop_write_file(filp);
2029 		if (error)
2030 			return error;
2031 
2032 		if (copy_to_user(arg, &inout, sizeof(inout)))
2033 			return -EFAULT;
2034 		return 0;
2035 	}
2036 
2037 	case XFS_IOC_GET_RESBLKS: {
2038 		xfs_fsop_resblks_t out;
2039 
2040 		if (!capable(CAP_SYS_ADMIN))
2041 			return -EPERM;
2042 
2043 		error = xfs_reserve_blocks(mp, NULL, &out);
2044 		if (error)
2045 			return error;
2046 
2047 		if (copy_to_user(arg, &out, sizeof(out)))
2048 			return -EFAULT;
2049 
2050 		return 0;
2051 	}
2052 
2053 	case XFS_IOC_FSGROWFSDATA: {
2054 		struct xfs_growfs_data in;
2055 
2056 		if (copy_from_user(&in, arg, sizeof(in)))
2057 			return -EFAULT;
2058 
2059 		error = mnt_want_write_file(filp);
2060 		if (error)
2061 			return error;
2062 		error = xfs_growfs_data(mp, &in);
2063 		mnt_drop_write_file(filp);
2064 		return error;
2065 	}
2066 
2067 	case XFS_IOC_FSGROWFSLOG: {
2068 		struct xfs_growfs_log in;
2069 
2070 		if (copy_from_user(&in, arg, sizeof(in)))
2071 			return -EFAULT;
2072 
2073 		error = mnt_want_write_file(filp);
2074 		if (error)
2075 			return error;
2076 		error = xfs_growfs_log(mp, &in);
2077 		mnt_drop_write_file(filp);
2078 		return error;
2079 	}
2080 
2081 	case XFS_IOC_FSGROWFSRT: {
2082 		xfs_growfs_rt_t in;
2083 
2084 		if (copy_from_user(&in, arg, sizeof(in)))
2085 			return -EFAULT;
2086 
2087 		error = mnt_want_write_file(filp);
2088 		if (error)
2089 			return error;
2090 		error = xfs_growfs_rt(mp, &in);
2091 		mnt_drop_write_file(filp);
2092 		return error;
2093 	}
2094 
2095 	case XFS_IOC_GOINGDOWN: {
2096 		uint32_t in;
2097 
2098 		if (!capable(CAP_SYS_ADMIN))
2099 			return -EPERM;
2100 
2101 		if (get_user(in, (uint32_t __user *)arg))
2102 			return -EFAULT;
2103 
2104 		return xfs_fs_goingdown(mp, in);
2105 	}
2106 
2107 	case XFS_IOC_ERROR_INJECTION: {
2108 		xfs_error_injection_t in;
2109 
2110 		if (!capable(CAP_SYS_ADMIN))
2111 			return -EPERM;
2112 
2113 		if (copy_from_user(&in, arg, sizeof(in)))
2114 			return -EFAULT;
2115 
2116 		return xfs_errortag_add(mp, in.errtag);
2117 	}
2118 
2119 	case XFS_IOC_ERROR_CLEARALL:
2120 		if (!capable(CAP_SYS_ADMIN))
2121 			return -EPERM;
2122 
2123 		return xfs_errortag_clearall(mp);
2124 
2125 	case XFS_IOC_FREE_EOFBLOCKS: {
2126 		struct xfs_fs_eofblocks	eofb;
2127 		struct xfs_icwalk	icw;
2128 
2129 		if (!capable(CAP_SYS_ADMIN))
2130 			return -EPERM;
2131 
2132 		if (xfs_is_readonly(mp))
2133 			return -EROFS;
2134 
2135 		if (copy_from_user(&eofb, arg, sizeof(eofb)))
2136 			return -EFAULT;
2137 
2138 		error = xfs_fs_eofblocks_from_user(&eofb, &icw);
2139 		if (error)
2140 			return error;
2141 
2142 		trace_xfs_ioc_free_eofblocks(mp, &icw, _RET_IP_);
2143 
2144 		sb_start_write(mp->m_super);
2145 		error = xfs_blockgc_free_space(mp, &icw);
2146 		sb_end_write(mp->m_super);
2147 		return error;
2148 	}
2149 
2150 	default:
2151 		return -ENOTTY;
2152 	}
2153 }
2154