xref: /openbmc/linux/fs/xfs/xfs_ioctl.c (revision 7ae5c03a)
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_perag	*pag;
959 	struct xfs_ag_geometry	ageo;
960 	int			error;
961 
962 	if (copy_from_user(&ageo, arg, sizeof(ageo)))
963 		return -EFAULT;
964 	if (ageo.ag_flags)
965 		return -EINVAL;
966 	if (memchr_inv(&ageo.ag_reserved, 0, sizeof(ageo.ag_reserved)))
967 		return -EINVAL;
968 
969 	pag = xfs_perag_get(mp, ageo.ag_number);
970 	if (!pag)
971 		return -EINVAL;
972 
973 	error = xfs_ag_get_geometry(pag, &ageo);
974 	xfs_perag_put(pag);
975 	if (error)
976 		return error;
977 
978 	if (copy_to_user(arg, &ageo, sizeof(ageo)))
979 		return -EFAULT;
980 	return 0;
981 }
982 
983 /*
984  * Linux extended inode flags interface.
985  */
986 
987 static void
988 xfs_fill_fsxattr(
989 	struct xfs_inode	*ip,
990 	int			whichfork,
991 	struct fileattr		*fa)
992 {
993 	struct xfs_mount	*mp = ip->i_mount;
994 	struct xfs_ifork	*ifp = xfs_ifork_ptr(ip, whichfork);
995 
996 	fileattr_fill_xflags(fa, xfs_ip2xflags(ip));
997 
998 	if (ip->i_diflags & XFS_DIFLAG_EXTSIZE) {
999 		fa->fsx_extsize = XFS_FSB_TO_B(mp, ip->i_extsize);
1000 	} else if (ip->i_diflags & XFS_DIFLAG_EXTSZINHERIT) {
1001 		/*
1002 		 * Don't let a misaligned extent size hint on a directory
1003 		 * escape to userspace if it won't pass the setattr checks
1004 		 * later.
1005 		 */
1006 		if ((ip->i_diflags & XFS_DIFLAG_RTINHERIT) &&
1007 		    ip->i_extsize % mp->m_sb.sb_rextsize > 0) {
1008 			fa->fsx_xflags &= ~(FS_XFLAG_EXTSIZE |
1009 					    FS_XFLAG_EXTSZINHERIT);
1010 			fa->fsx_extsize = 0;
1011 		} else {
1012 			fa->fsx_extsize = XFS_FSB_TO_B(mp, ip->i_extsize);
1013 		}
1014 	}
1015 
1016 	if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
1017 		fa->fsx_cowextsize = XFS_FSB_TO_B(mp, ip->i_cowextsize);
1018 	fa->fsx_projid = ip->i_projid;
1019 	if (ifp && !xfs_need_iread_extents(ifp))
1020 		fa->fsx_nextents = xfs_iext_count(ifp);
1021 	else
1022 		fa->fsx_nextents = xfs_ifork_nextents(ifp);
1023 }
1024 
1025 STATIC int
1026 xfs_ioc_fsgetxattra(
1027 	xfs_inode_t		*ip,
1028 	void			__user *arg)
1029 {
1030 	struct fileattr		fa;
1031 
1032 	xfs_ilock(ip, XFS_ILOCK_SHARED);
1033 	xfs_fill_fsxattr(ip, XFS_ATTR_FORK, &fa);
1034 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
1035 
1036 	return copy_fsxattr_to_user(&fa, arg);
1037 }
1038 
1039 int
1040 xfs_fileattr_get(
1041 	struct dentry		*dentry,
1042 	struct fileattr		*fa)
1043 {
1044 	struct xfs_inode	*ip = XFS_I(d_inode(dentry));
1045 
1046 	if (d_is_special(dentry))
1047 		return -ENOTTY;
1048 
1049 	xfs_ilock(ip, XFS_ILOCK_SHARED);
1050 	xfs_fill_fsxattr(ip, XFS_DATA_FORK, fa);
1051 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
1052 
1053 	return 0;
1054 }
1055 
1056 STATIC uint16_t
1057 xfs_flags2diflags(
1058 	struct xfs_inode	*ip,
1059 	unsigned int		xflags)
1060 {
1061 	/* can't set PREALLOC this way, just preserve it */
1062 	uint16_t		di_flags =
1063 		(ip->i_diflags & XFS_DIFLAG_PREALLOC);
1064 
1065 	if (xflags & FS_XFLAG_IMMUTABLE)
1066 		di_flags |= XFS_DIFLAG_IMMUTABLE;
1067 	if (xflags & FS_XFLAG_APPEND)
1068 		di_flags |= XFS_DIFLAG_APPEND;
1069 	if (xflags & FS_XFLAG_SYNC)
1070 		di_flags |= XFS_DIFLAG_SYNC;
1071 	if (xflags & FS_XFLAG_NOATIME)
1072 		di_flags |= XFS_DIFLAG_NOATIME;
1073 	if (xflags & FS_XFLAG_NODUMP)
1074 		di_flags |= XFS_DIFLAG_NODUMP;
1075 	if (xflags & FS_XFLAG_NODEFRAG)
1076 		di_flags |= XFS_DIFLAG_NODEFRAG;
1077 	if (xflags & FS_XFLAG_FILESTREAM)
1078 		di_flags |= XFS_DIFLAG_FILESTREAM;
1079 	if (S_ISDIR(VFS_I(ip)->i_mode)) {
1080 		if (xflags & FS_XFLAG_RTINHERIT)
1081 			di_flags |= XFS_DIFLAG_RTINHERIT;
1082 		if (xflags & FS_XFLAG_NOSYMLINKS)
1083 			di_flags |= XFS_DIFLAG_NOSYMLINKS;
1084 		if (xflags & FS_XFLAG_EXTSZINHERIT)
1085 			di_flags |= XFS_DIFLAG_EXTSZINHERIT;
1086 		if (xflags & FS_XFLAG_PROJINHERIT)
1087 			di_flags |= XFS_DIFLAG_PROJINHERIT;
1088 	} else if (S_ISREG(VFS_I(ip)->i_mode)) {
1089 		if (xflags & FS_XFLAG_REALTIME)
1090 			di_flags |= XFS_DIFLAG_REALTIME;
1091 		if (xflags & FS_XFLAG_EXTSIZE)
1092 			di_flags |= XFS_DIFLAG_EXTSIZE;
1093 	}
1094 
1095 	return di_flags;
1096 }
1097 
1098 STATIC uint64_t
1099 xfs_flags2diflags2(
1100 	struct xfs_inode	*ip,
1101 	unsigned int		xflags)
1102 {
1103 	uint64_t		di_flags2 =
1104 		(ip->i_diflags2 & (XFS_DIFLAG2_REFLINK |
1105 				   XFS_DIFLAG2_BIGTIME |
1106 				   XFS_DIFLAG2_NREXT64));
1107 
1108 	if (xflags & FS_XFLAG_DAX)
1109 		di_flags2 |= XFS_DIFLAG2_DAX;
1110 	if (xflags & FS_XFLAG_COWEXTSIZE)
1111 		di_flags2 |= XFS_DIFLAG2_COWEXTSIZE;
1112 
1113 	return di_flags2;
1114 }
1115 
1116 static int
1117 xfs_ioctl_setattr_xflags(
1118 	struct xfs_trans	*tp,
1119 	struct xfs_inode	*ip,
1120 	struct fileattr		*fa)
1121 {
1122 	struct xfs_mount	*mp = ip->i_mount;
1123 	uint64_t		i_flags2;
1124 
1125 	/* Can't change realtime flag if any extents are allocated. */
1126 	if ((ip->i_df.if_nextents || ip->i_delayed_blks) &&
1127 	    XFS_IS_REALTIME_INODE(ip) != (fa->fsx_xflags & FS_XFLAG_REALTIME))
1128 		return -EINVAL;
1129 
1130 	/* If realtime flag is set then must have realtime device */
1131 	if (fa->fsx_xflags & FS_XFLAG_REALTIME) {
1132 		if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
1133 		    (ip->i_extsize % mp->m_sb.sb_rextsize))
1134 			return -EINVAL;
1135 	}
1136 
1137 	/* Clear reflink if we are actually able to set the rt flag. */
1138 	if ((fa->fsx_xflags & FS_XFLAG_REALTIME) && xfs_is_reflink_inode(ip))
1139 		ip->i_diflags2 &= ~XFS_DIFLAG2_REFLINK;
1140 
1141 	/* Don't allow us to set DAX mode for a reflinked file for now. */
1142 	if ((fa->fsx_xflags & FS_XFLAG_DAX) && xfs_is_reflink_inode(ip))
1143 		return -EINVAL;
1144 
1145 	/* diflags2 only valid for v3 inodes. */
1146 	i_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1147 	if (i_flags2 && !xfs_has_v3inodes(mp))
1148 		return -EINVAL;
1149 
1150 	ip->i_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
1151 	ip->i_diflags2 = i_flags2;
1152 
1153 	xfs_diflags_to_iflags(ip, false);
1154 	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1155 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1156 	XFS_STATS_INC(mp, xs_ig_attrchg);
1157 	return 0;
1158 }
1159 
1160 static void
1161 xfs_ioctl_setattr_prepare_dax(
1162 	struct xfs_inode	*ip,
1163 	struct fileattr		*fa)
1164 {
1165 	struct xfs_mount	*mp = ip->i_mount;
1166 	struct inode            *inode = VFS_I(ip);
1167 
1168 	if (S_ISDIR(inode->i_mode))
1169 		return;
1170 
1171 	if (xfs_has_dax_always(mp) || xfs_has_dax_never(mp))
1172 		return;
1173 
1174 	if (((fa->fsx_xflags & FS_XFLAG_DAX) &&
1175 	    !(ip->i_diflags2 & XFS_DIFLAG2_DAX)) ||
1176 	    (!(fa->fsx_xflags & FS_XFLAG_DAX) &&
1177 	     (ip->i_diflags2 & XFS_DIFLAG2_DAX)))
1178 		d_mark_dontcache(inode);
1179 }
1180 
1181 /*
1182  * Set up the transaction structure for the setattr operation, checking that we
1183  * have permission to do so. On success, return a clean transaction and the
1184  * inode locked exclusively ready for further operation specific checks. On
1185  * failure, return an error without modifying or locking the inode.
1186  */
1187 static struct xfs_trans *
1188 xfs_ioctl_setattr_get_trans(
1189 	struct xfs_inode	*ip,
1190 	struct xfs_dquot	*pdqp)
1191 {
1192 	struct xfs_mount	*mp = ip->i_mount;
1193 	struct xfs_trans	*tp;
1194 	int			error = -EROFS;
1195 
1196 	if (xfs_is_readonly(mp))
1197 		goto out_error;
1198 	error = -EIO;
1199 	if (xfs_is_shutdown(mp))
1200 		goto out_error;
1201 
1202 	error = xfs_trans_alloc_ichange(ip, NULL, NULL, pdqp,
1203 			has_capability_noaudit(current, CAP_FOWNER), &tp);
1204 	if (error)
1205 		goto out_error;
1206 
1207 	if (xfs_has_wsync(mp))
1208 		xfs_trans_set_sync(tp);
1209 
1210 	return tp;
1211 
1212 out_error:
1213 	return ERR_PTR(error);
1214 }
1215 
1216 /*
1217  * Validate a proposed extent size hint.  For regular files, the hint can only
1218  * be changed if no extents are allocated.
1219  */
1220 static int
1221 xfs_ioctl_setattr_check_extsize(
1222 	struct xfs_inode	*ip,
1223 	struct fileattr		*fa)
1224 {
1225 	struct xfs_mount	*mp = ip->i_mount;
1226 	xfs_failaddr_t		failaddr;
1227 	uint16_t		new_diflags;
1228 
1229 	if (!fa->fsx_valid)
1230 		return 0;
1231 
1232 	if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_df.if_nextents &&
1233 	    XFS_FSB_TO_B(mp, ip->i_extsize) != fa->fsx_extsize)
1234 		return -EINVAL;
1235 
1236 	if (fa->fsx_extsize & mp->m_blockmask)
1237 		return -EINVAL;
1238 
1239 	new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
1240 
1241 	/*
1242 	 * Inode verifiers do not check that the extent size hint is an integer
1243 	 * multiple of the rt extent size on a directory with both rtinherit
1244 	 * and extszinherit flags set.  Don't let sysadmins misconfigure
1245 	 * directories.
1246 	 */
1247 	if ((new_diflags & XFS_DIFLAG_RTINHERIT) &&
1248 	    (new_diflags & XFS_DIFLAG_EXTSZINHERIT)) {
1249 		unsigned int	rtextsize_bytes;
1250 
1251 		rtextsize_bytes = XFS_FSB_TO_B(mp, mp->m_sb.sb_rextsize);
1252 		if (fa->fsx_extsize % rtextsize_bytes)
1253 			return -EINVAL;
1254 	}
1255 
1256 	failaddr = xfs_inode_validate_extsize(ip->i_mount,
1257 			XFS_B_TO_FSB(mp, fa->fsx_extsize),
1258 			VFS_I(ip)->i_mode, new_diflags);
1259 	return failaddr != NULL ? -EINVAL : 0;
1260 }
1261 
1262 static int
1263 xfs_ioctl_setattr_check_cowextsize(
1264 	struct xfs_inode	*ip,
1265 	struct fileattr		*fa)
1266 {
1267 	struct xfs_mount	*mp = ip->i_mount;
1268 	xfs_failaddr_t		failaddr;
1269 	uint64_t		new_diflags2;
1270 	uint16_t		new_diflags;
1271 
1272 	if (!fa->fsx_valid)
1273 		return 0;
1274 
1275 	if (fa->fsx_cowextsize & mp->m_blockmask)
1276 		return -EINVAL;
1277 
1278 	new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
1279 	new_diflags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1280 
1281 	failaddr = xfs_inode_validate_cowextsize(ip->i_mount,
1282 			XFS_B_TO_FSB(mp, fa->fsx_cowextsize),
1283 			VFS_I(ip)->i_mode, new_diflags, new_diflags2);
1284 	return failaddr != NULL ? -EINVAL : 0;
1285 }
1286 
1287 static int
1288 xfs_ioctl_setattr_check_projid(
1289 	struct xfs_inode	*ip,
1290 	struct fileattr		*fa)
1291 {
1292 	if (!fa->fsx_valid)
1293 		return 0;
1294 
1295 	/* Disallow 32bit project ids if 32bit IDs are not enabled. */
1296 	if (fa->fsx_projid > (uint16_t)-1 &&
1297 	    !xfs_has_projid32(ip->i_mount))
1298 		return -EINVAL;
1299 	return 0;
1300 }
1301 
1302 int
1303 xfs_fileattr_set(
1304 	struct user_namespace	*mnt_userns,
1305 	struct dentry		*dentry,
1306 	struct fileattr		*fa)
1307 {
1308 	struct xfs_inode	*ip = XFS_I(d_inode(dentry));
1309 	struct xfs_mount	*mp = ip->i_mount;
1310 	struct xfs_trans	*tp;
1311 	struct xfs_dquot	*pdqp = NULL;
1312 	struct xfs_dquot	*olddquot = NULL;
1313 	int			error;
1314 
1315 	trace_xfs_ioctl_setattr(ip);
1316 
1317 	if (d_is_special(dentry))
1318 		return -ENOTTY;
1319 
1320 	if (!fa->fsx_valid) {
1321 		if (fa->flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL |
1322 				  FS_NOATIME_FL | FS_NODUMP_FL |
1323 				  FS_SYNC_FL | FS_DAX_FL | FS_PROJINHERIT_FL))
1324 			return -EOPNOTSUPP;
1325 	}
1326 
1327 	error = xfs_ioctl_setattr_check_projid(ip, fa);
1328 	if (error)
1329 		return error;
1330 
1331 	/*
1332 	 * If disk quotas is on, we make sure that the dquots do exist on disk,
1333 	 * before we start any other transactions. Trying to do this later
1334 	 * is messy. We don't care to take a readlock to look at the ids
1335 	 * in inode here, because we can't hold it across the trans_reserve.
1336 	 * If the IDs do change before we take the ilock, we're covered
1337 	 * because the i_*dquot fields will get updated anyway.
1338 	 */
1339 	if (fa->fsx_valid && XFS_IS_QUOTA_ON(mp)) {
1340 		error = xfs_qm_vop_dqalloc(ip, VFS_I(ip)->i_uid,
1341 				VFS_I(ip)->i_gid, fa->fsx_projid,
1342 				XFS_QMOPT_PQUOTA, NULL, NULL, &pdqp);
1343 		if (error)
1344 			return error;
1345 	}
1346 
1347 	xfs_ioctl_setattr_prepare_dax(ip, fa);
1348 
1349 	tp = xfs_ioctl_setattr_get_trans(ip, pdqp);
1350 	if (IS_ERR(tp)) {
1351 		error = PTR_ERR(tp);
1352 		goto error_free_dquots;
1353 	}
1354 
1355 	error = xfs_ioctl_setattr_check_extsize(ip, fa);
1356 	if (error)
1357 		goto error_trans_cancel;
1358 
1359 	error = xfs_ioctl_setattr_check_cowextsize(ip, fa);
1360 	if (error)
1361 		goto error_trans_cancel;
1362 
1363 	error = xfs_ioctl_setattr_xflags(tp, ip, fa);
1364 	if (error)
1365 		goto error_trans_cancel;
1366 
1367 	if (!fa->fsx_valid)
1368 		goto skip_xattr;
1369 	/*
1370 	 * Change file ownership.  Must be the owner or privileged.  CAP_FSETID
1371 	 * overrides the following restrictions:
1372 	 *
1373 	 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1374 	 * successful return from chown()
1375 	 */
1376 
1377 	if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
1378 	    !capable_wrt_inode_uidgid(mnt_userns, VFS_I(ip), CAP_FSETID))
1379 		VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
1380 
1381 	/* Change the ownerships and register project quota modifications */
1382 	if (ip->i_projid != fa->fsx_projid) {
1383 		if (XFS_IS_PQUOTA_ON(mp)) {
1384 			olddquot = xfs_qm_vop_chown(tp, ip,
1385 						&ip->i_pdquot, pdqp);
1386 		}
1387 		ip->i_projid = fa->fsx_projid;
1388 	}
1389 
1390 	/*
1391 	 * Only set the extent size hint if we've already determined that the
1392 	 * extent size hint should be set on the inode. If no extent size flags
1393 	 * are set on the inode then unconditionally clear the extent size hint.
1394 	 */
1395 	if (ip->i_diflags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1396 		ip->i_extsize = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1397 	else
1398 		ip->i_extsize = 0;
1399 
1400 	if (xfs_has_v3inodes(mp)) {
1401 		if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
1402 			ip->i_cowextsize = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
1403 		else
1404 			ip->i_cowextsize = 0;
1405 	}
1406 
1407 skip_xattr:
1408 	error = xfs_trans_commit(tp);
1409 
1410 	/*
1411 	 * Release any dquot(s) the inode had kept before chown.
1412 	 */
1413 	xfs_qm_dqrele(olddquot);
1414 	xfs_qm_dqrele(pdqp);
1415 
1416 	return error;
1417 
1418 error_trans_cancel:
1419 	xfs_trans_cancel(tp);
1420 error_free_dquots:
1421 	xfs_qm_dqrele(pdqp);
1422 	return error;
1423 }
1424 
1425 static bool
1426 xfs_getbmap_format(
1427 	struct kgetbmap		*p,
1428 	struct getbmapx __user	*u,
1429 	size_t			recsize)
1430 {
1431 	if (put_user(p->bmv_offset, &u->bmv_offset) ||
1432 	    put_user(p->bmv_block, &u->bmv_block) ||
1433 	    put_user(p->bmv_length, &u->bmv_length) ||
1434 	    put_user(0, &u->bmv_count) ||
1435 	    put_user(0, &u->bmv_entries))
1436 		return false;
1437 	if (recsize < sizeof(struct getbmapx))
1438 		return true;
1439 	if (put_user(0, &u->bmv_iflags) ||
1440 	    put_user(p->bmv_oflags, &u->bmv_oflags) ||
1441 	    put_user(0, &u->bmv_unused1) ||
1442 	    put_user(0, &u->bmv_unused2))
1443 		return false;
1444 	return true;
1445 }
1446 
1447 STATIC int
1448 xfs_ioc_getbmap(
1449 	struct file		*file,
1450 	unsigned int		cmd,
1451 	void			__user *arg)
1452 {
1453 	struct getbmapx		bmx = { 0 };
1454 	struct kgetbmap		*buf;
1455 	size_t			recsize;
1456 	int			error, i;
1457 
1458 	switch (cmd) {
1459 	case XFS_IOC_GETBMAPA:
1460 		bmx.bmv_iflags = BMV_IF_ATTRFORK;
1461 		fallthrough;
1462 	case XFS_IOC_GETBMAP:
1463 		/* struct getbmap is a strict subset of struct getbmapx. */
1464 		recsize = sizeof(struct getbmap);
1465 		break;
1466 	case XFS_IOC_GETBMAPX:
1467 		recsize = sizeof(struct getbmapx);
1468 		break;
1469 	default:
1470 		return -EINVAL;
1471 	}
1472 
1473 	if (copy_from_user(&bmx, arg, recsize))
1474 		return -EFAULT;
1475 
1476 	if (bmx.bmv_count < 2)
1477 		return -EINVAL;
1478 	if (bmx.bmv_count >= INT_MAX / recsize)
1479 		return -ENOMEM;
1480 
1481 	buf = kvcalloc(bmx.bmv_count, sizeof(*buf), GFP_KERNEL);
1482 	if (!buf)
1483 		return -ENOMEM;
1484 
1485 	error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, buf);
1486 	if (error)
1487 		goto out_free_buf;
1488 
1489 	error = -EFAULT;
1490 	if (copy_to_user(arg, &bmx, recsize))
1491 		goto out_free_buf;
1492 	arg += recsize;
1493 
1494 	for (i = 0; i < bmx.bmv_entries; i++) {
1495 		if (!xfs_getbmap_format(buf + i, arg, recsize))
1496 			goto out_free_buf;
1497 		arg += recsize;
1498 	}
1499 
1500 	error = 0;
1501 out_free_buf:
1502 	kmem_free(buf);
1503 	return error;
1504 }
1505 
1506 STATIC int
1507 xfs_ioc_getfsmap(
1508 	struct xfs_inode	*ip,
1509 	struct fsmap_head	__user *arg)
1510 {
1511 	struct xfs_fsmap_head	xhead = {0};
1512 	struct fsmap_head	head;
1513 	struct fsmap		*recs;
1514 	unsigned int		count;
1515 	__u32			last_flags = 0;
1516 	bool			done = false;
1517 	int			error;
1518 
1519 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
1520 		return -EFAULT;
1521 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
1522 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
1523 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
1524 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
1525 		       sizeof(head.fmh_keys[1].fmr_reserved)))
1526 		return -EINVAL;
1527 
1528 	/*
1529 	 * Use an internal memory buffer so that we don't have to copy fsmap
1530 	 * data to userspace while holding locks.  Start by trying to allocate
1531 	 * up to 128k for the buffer, but fall back to a single page if needed.
1532 	 */
1533 	count = min_t(unsigned int, head.fmh_count,
1534 			131072 / sizeof(struct fsmap));
1535 	recs = kvcalloc(count, sizeof(struct fsmap), GFP_KERNEL);
1536 	if (!recs) {
1537 		count = min_t(unsigned int, head.fmh_count,
1538 				PAGE_SIZE / sizeof(struct fsmap));
1539 		recs = kvcalloc(count, sizeof(struct fsmap), GFP_KERNEL);
1540 		if (!recs)
1541 			return -ENOMEM;
1542 	}
1543 
1544 	xhead.fmh_iflags = head.fmh_iflags;
1545 	xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
1546 	xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
1547 
1548 	trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1549 	trace_xfs_getfsmap_high_key(ip->i_mount, &xhead.fmh_keys[1]);
1550 
1551 	head.fmh_entries = 0;
1552 	do {
1553 		struct fsmap __user	*user_recs;
1554 		struct fsmap		*last_rec;
1555 
1556 		user_recs = &arg->fmh_recs[head.fmh_entries];
1557 		xhead.fmh_entries = 0;
1558 		xhead.fmh_count = min_t(unsigned int, count,
1559 					head.fmh_count - head.fmh_entries);
1560 
1561 		/* Run query, record how many entries we got. */
1562 		error = xfs_getfsmap(ip->i_mount, &xhead, recs);
1563 		switch (error) {
1564 		case 0:
1565 			/*
1566 			 * There are no more records in the result set.  Copy
1567 			 * whatever we got to userspace and break out.
1568 			 */
1569 			done = true;
1570 			break;
1571 		case -ECANCELED:
1572 			/*
1573 			 * The internal memory buffer is full.  Copy whatever
1574 			 * records we got to userspace and go again if we have
1575 			 * not yet filled the userspace buffer.
1576 			 */
1577 			error = 0;
1578 			break;
1579 		default:
1580 			goto out_free;
1581 		}
1582 		head.fmh_entries += xhead.fmh_entries;
1583 		head.fmh_oflags = xhead.fmh_oflags;
1584 
1585 		/*
1586 		 * If the caller wanted a record count or there aren't any
1587 		 * new records to return, we're done.
1588 		 */
1589 		if (head.fmh_count == 0 || xhead.fmh_entries == 0)
1590 			break;
1591 
1592 		/* Copy all the records we got out to userspace. */
1593 		if (copy_to_user(user_recs, recs,
1594 				 xhead.fmh_entries * sizeof(struct fsmap))) {
1595 			error = -EFAULT;
1596 			goto out_free;
1597 		}
1598 
1599 		/* Remember the last record flags we copied to userspace. */
1600 		last_rec = &recs[xhead.fmh_entries - 1];
1601 		last_flags = last_rec->fmr_flags;
1602 
1603 		/* Set up the low key for the next iteration. */
1604 		xfs_fsmap_to_internal(&xhead.fmh_keys[0], last_rec);
1605 		trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1606 	} while (!done && head.fmh_entries < head.fmh_count);
1607 
1608 	/*
1609 	 * If there are no more records in the query result set and we're not
1610 	 * in counting mode, mark the last record returned with the LAST flag.
1611 	 */
1612 	if (done && head.fmh_count > 0 && head.fmh_entries > 0) {
1613 		struct fsmap __user	*user_rec;
1614 
1615 		last_flags |= FMR_OF_LAST;
1616 		user_rec = &arg->fmh_recs[head.fmh_entries - 1];
1617 
1618 		if (copy_to_user(&user_rec->fmr_flags, &last_flags,
1619 					sizeof(last_flags))) {
1620 			error = -EFAULT;
1621 			goto out_free;
1622 		}
1623 	}
1624 
1625 	/* copy back header */
1626 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head))) {
1627 		error = -EFAULT;
1628 		goto out_free;
1629 	}
1630 
1631 out_free:
1632 	kmem_free(recs);
1633 	return error;
1634 }
1635 
1636 STATIC int
1637 xfs_ioc_scrub_metadata(
1638 	struct file			*file,
1639 	void				__user *arg)
1640 {
1641 	struct xfs_scrub_metadata	scrub;
1642 	int				error;
1643 
1644 	if (!capable(CAP_SYS_ADMIN))
1645 		return -EPERM;
1646 
1647 	if (copy_from_user(&scrub, arg, sizeof(scrub)))
1648 		return -EFAULT;
1649 
1650 	error = xfs_scrub_metadata(file, &scrub);
1651 	if (error)
1652 		return error;
1653 
1654 	if (copy_to_user(arg, &scrub, sizeof(scrub)))
1655 		return -EFAULT;
1656 
1657 	return 0;
1658 }
1659 
1660 int
1661 xfs_ioc_swapext(
1662 	xfs_swapext_t	*sxp)
1663 {
1664 	xfs_inode_t     *ip, *tip;
1665 	struct fd	f, tmp;
1666 	int		error = 0;
1667 
1668 	/* Pull information for the target fd */
1669 	f = fdget((int)sxp->sx_fdtarget);
1670 	if (!f.file) {
1671 		error = -EINVAL;
1672 		goto out;
1673 	}
1674 
1675 	if (!(f.file->f_mode & FMODE_WRITE) ||
1676 	    !(f.file->f_mode & FMODE_READ) ||
1677 	    (f.file->f_flags & O_APPEND)) {
1678 		error = -EBADF;
1679 		goto out_put_file;
1680 	}
1681 
1682 	tmp = fdget((int)sxp->sx_fdtmp);
1683 	if (!tmp.file) {
1684 		error = -EINVAL;
1685 		goto out_put_file;
1686 	}
1687 
1688 	if (!(tmp.file->f_mode & FMODE_WRITE) ||
1689 	    !(tmp.file->f_mode & FMODE_READ) ||
1690 	    (tmp.file->f_flags & O_APPEND)) {
1691 		error = -EBADF;
1692 		goto out_put_tmp_file;
1693 	}
1694 
1695 	if (IS_SWAPFILE(file_inode(f.file)) ||
1696 	    IS_SWAPFILE(file_inode(tmp.file))) {
1697 		error = -EINVAL;
1698 		goto out_put_tmp_file;
1699 	}
1700 
1701 	/*
1702 	 * We need to ensure that the fds passed in point to XFS inodes
1703 	 * before we cast and access them as XFS structures as we have no
1704 	 * control over what the user passes us here.
1705 	 */
1706 	if (f.file->f_op != &xfs_file_operations ||
1707 	    tmp.file->f_op != &xfs_file_operations) {
1708 		error = -EINVAL;
1709 		goto out_put_tmp_file;
1710 	}
1711 
1712 	ip = XFS_I(file_inode(f.file));
1713 	tip = XFS_I(file_inode(tmp.file));
1714 
1715 	if (ip->i_mount != tip->i_mount) {
1716 		error = -EINVAL;
1717 		goto out_put_tmp_file;
1718 	}
1719 
1720 	if (ip->i_ino == tip->i_ino) {
1721 		error = -EINVAL;
1722 		goto out_put_tmp_file;
1723 	}
1724 
1725 	if (xfs_is_shutdown(ip->i_mount)) {
1726 		error = -EIO;
1727 		goto out_put_tmp_file;
1728 	}
1729 
1730 	error = xfs_swap_extents(ip, tip, sxp);
1731 
1732  out_put_tmp_file:
1733 	fdput(tmp);
1734  out_put_file:
1735 	fdput(f);
1736  out:
1737 	return error;
1738 }
1739 
1740 static int
1741 xfs_ioc_getlabel(
1742 	struct xfs_mount	*mp,
1743 	char			__user *user_label)
1744 {
1745 	struct xfs_sb		*sbp = &mp->m_sb;
1746 	char			label[XFSLABEL_MAX + 1];
1747 
1748 	/* Paranoia */
1749 	BUILD_BUG_ON(sizeof(sbp->sb_fname) > FSLABEL_MAX);
1750 
1751 	/* 1 larger than sb_fname, so this ensures a trailing NUL char */
1752 	memset(label, 0, sizeof(label));
1753 	spin_lock(&mp->m_sb_lock);
1754 	strncpy(label, sbp->sb_fname, XFSLABEL_MAX);
1755 	spin_unlock(&mp->m_sb_lock);
1756 
1757 	if (copy_to_user(user_label, label, sizeof(label)))
1758 		return -EFAULT;
1759 	return 0;
1760 }
1761 
1762 static int
1763 xfs_ioc_setlabel(
1764 	struct file		*filp,
1765 	struct xfs_mount	*mp,
1766 	char			__user *newlabel)
1767 {
1768 	struct xfs_sb		*sbp = &mp->m_sb;
1769 	char			label[XFSLABEL_MAX + 1];
1770 	size_t			len;
1771 	int			error;
1772 
1773 	if (!capable(CAP_SYS_ADMIN))
1774 		return -EPERM;
1775 	/*
1776 	 * The generic ioctl allows up to FSLABEL_MAX chars, but XFS is much
1777 	 * smaller, at 12 bytes.  We copy one more to be sure we find the
1778 	 * (required) NULL character to test the incoming label length.
1779 	 * NB: The on disk label doesn't need to be null terminated.
1780 	 */
1781 	if (copy_from_user(label, newlabel, XFSLABEL_MAX + 1))
1782 		return -EFAULT;
1783 	len = strnlen(label, XFSLABEL_MAX + 1);
1784 	if (len > sizeof(sbp->sb_fname))
1785 		return -EINVAL;
1786 
1787 	error = mnt_want_write_file(filp);
1788 	if (error)
1789 		return error;
1790 
1791 	spin_lock(&mp->m_sb_lock);
1792 	memset(sbp->sb_fname, 0, sizeof(sbp->sb_fname));
1793 	memcpy(sbp->sb_fname, label, len);
1794 	spin_unlock(&mp->m_sb_lock);
1795 
1796 	/*
1797 	 * Now we do several things to satisfy userspace.
1798 	 * In addition to normal logging of the primary superblock, we also
1799 	 * immediately write these changes to sector zero for the primary, then
1800 	 * update all backup supers (as xfs_db does for a label change), then
1801 	 * invalidate the block device page cache.  This is so that any prior
1802 	 * buffered reads from userspace (i.e. from blkid) are invalidated,
1803 	 * and userspace will see the newly-written label.
1804 	 */
1805 	error = xfs_sync_sb_buf(mp);
1806 	if (error)
1807 		goto out;
1808 	/*
1809 	 * growfs also updates backup supers so lock against that.
1810 	 */
1811 	mutex_lock(&mp->m_growlock);
1812 	error = xfs_update_secondary_sbs(mp);
1813 	mutex_unlock(&mp->m_growlock);
1814 
1815 	invalidate_bdev(mp->m_ddev_targp->bt_bdev);
1816 
1817 out:
1818 	mnt_drop_write_file(filp);
1819 	return error;
1820 }
1821 
1822 static inline int
1823 xfs_fs_eofblocks_from_user(
1824 	struct xfs_fs_eofblocks		*src,
1825 	struct xfs_icwalk		*dst)
1826 {
1827 	if (src->eof_version != XFS_EOFBLOCKS_VERSION)
1828 		return -EINVAL;
1829 
1830 	if (src->eof_flags & ~XFS_EOF_FLAGS_VALID)
1831 		return -EINVAL;
1832 
1833 	if (memchr_inv(&src->pad32, 0, sizeof(src->pad32)) ||
1834 	    memchr_inv(src->pad64, 0, sizeof(src->pad64)))
1835 		return -EINVAL;
1836 
1837 	dst->icw_flags = 0;
1838 	if (src->eof_flags & XFS_EOF_FLAGS_SYNC)
1839 		dst->icw_flags |= XFS_ICWALK_FLAG_SYNC;
1840 	if (src->eof_flags & XFS_EOF_FLAGS_UID)
1841 		dst->icw_flags |= XFS_ICWALK_FLAG_UID;
1842 	if (src->eof_flags & XFS_EOF_FLAGS_GID)
1843 		dst->icw_flags |= XFS_ICWALK_FLAG_GID;
1844 	if (src->eof_flags & XFS_EOF_FLAGS_PRID)
1845 		dst->icw_flags |= XFS_ICWALK_FLAG_PRID;
1846 	if (src->eof_flags & XFS_EOF_FLAGS_MINFILESIZE)
1847 		dst->icw_flags |= XFS_ICWALK_FLAG_MINFILESIZE;
1848 
1849 	dst->icw_prid = src->eof_prid;
1850 	dst->icw_min_file_size = src->eof_min_file_size;
1851 
1852 	dst->icw_uid = INVALID_UID;
1853 	if (src->eof_flags & XFS_EOF_FLAGS_UID) {
1854 		dst->icw_uid = make_kuid(current_user_ns(), src->eof_uid);
1855 		if (!uid_valid(dst->icw_uid))
1856 			return -EINVAL;
1857 	}
1858 
1859 	dst->icw_gid = INVALID_GID;
1860 	if (src->eof_flags & XFS_EOF_FLAGS_GID) {
1861 		dst->icw_gid = make_kgid(current_user_ns(), src->eof_gid);
1862 		if (!gid_valid(dst->icw_gid))
1863 			return -EINVAL;
1864 	}
1865 	return 0;
1866 }
1867 
1868 /*
1869  * These long-unused ioctls were removed from the official ioctl API in 5.17,
1870  * but retain these definitions so that we can log warnings about them.
1871  */
1872 #define XFS_IOC_ALLOCSP		_IOW ('X', 10, struct xfs_flock64)
1873 #define XFS_IOC_FREESP		_IOW ('X', 11, struct xfs_flock64)
1874 #define XFS_IOC_ALLOCSP64	_IOW ('X', 36, struct xfs_flock64)
1875 #define XFS_IOC_FREESP64	_IOW ('X', 37, struct xfs_flock64)
1876 
1877 /*
1878  * Note: some of the ioctl's return positive numbers as a
1879  * byte count indicating success, such as readlink_by_handle.
1880  * So we don't "sign flip" like most other routines.  This means
1881  * true errors need to be returned as a negative value.
1882  */
1883 long
1884 xfs_file_ioctl(
1885 	struct file		*filp,
1886 	unsigned int		cmd,
1887 	unsigned long		p)
1888 {
1889 	struct inode		*inode = file_inode(filp);
1890 	struct xfs_inode	*ip = XFS_I(inode);
1891 	struct xfs_mount	*mp = ip->i_mount;
1892 	void			__user *arg = (void __user *)p;
1893 	int			error;
1894 
1895 	trace_xfs_file_ioctl(ip);
1896 
1897 	switch (cmd) {
1898 	case FITRIM:
1899 		return xfs_ioc_trim(mp, arg);
1900 	case FS_IOC_GETFSLABEL:
1901 		return xfs_ioc_getlabel(mp, arg);
1902 	case FS_IOC_SETFSLABEL:
1903 		return xfs_ioc_setlabel(filp, mp, arg);
1904 	case XFS_IOC_ALLOCSP:
1905 	case XFS_IOC_FREESP:
1906 	case XFS_IOC_ALLOCSP64:
1907 	case XFS_IOC_FREESP64:
1908 		xfs_warn_once(mp,
1909 	"%s should use fallocate; XFS_IOC_{ALLOC,FREE}SP ioctl unsupported",
1910 				current->comm);
1911 		return -ENOTTY;
1912 	case XFS_IOC_DIOINFO: {
1913 		struct xfs_buftarg	*target = xfs_inode_buftarg(ip);
1914 		struct dioattr		da;
1915 
1916 		da.d_mem =  da.d_miniosz = target->bt_logical_sectorsize;
1917 		da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1918 
1919 		if (copy_to_user(arg, &da, sizeof(da)))
1920 			return -EFAULT;
1921 		return 0;
1922 	}
1923 
1924 	case XFS_IOC_FSBULKSTAT_SINGLE:
1925 	case XFS_IOC_FSBULKSTAT:
1926 	case XFS_IOC_FSINUMBERS:
1927 		return xfs_ioc_fsbulkstat(filp, cmd, arg);
1928 
1929 	case XFS_IOC_BULKSTAT:
1930 		return xfs_ioc_bulkstat(filp, cmd, arg);
1931 	case XFS_IOC_INUMBERS:
1932 		return xfs_ioc_inumbers(mp, cmd, arg);
1933 
1934 	case XFS_IOC_FSGEOMETRY_V1:
1935 		return xfs_ioc_fsgeometry(mp, arg, 3);
1936 	case XFS_IOC_FSGEOMETRY_V4:
1937 		return xfs_ioc_fsgeometry(mp, arg, 4);
1938 	case XFS_IOC_FSGEOMETRY:
1939 		return xfs_ioc_fsgeometry(mp, arg, 5);
1940 
1941 	case XFS_IOC_AG_GEOMETRY:
1942 		return xfs_ioc_ag_geometry(mp, arg);
1943 
1944 	case XFS_IOC_GETVERSION:
1945 		return put_user(inode->i_generation, (int __user *)arg);
1946 
1947 	case XFS_IOC_FSGETXATTRA:
1948 		return xfs_ioc_fsgetxattra(ip, arg);
1949 
1950 	case XFS_IOC_GETBMAP:
1951 	case XFS_IOC_GETBMAPA:
1952 	case XFS_IOC_GETBMAPX:
1953 		return xfs_ioc_getbmap(filp, cmd, arg);
1954 
1955 	case FS_IOC_GETFSMAP:
1956 		return xfs_ioc_getfsmap(ip, arg);
1957 
1958 	case XFS_IOC_SCRUB_METADATA:
1959 		return xfs_ioc_scrub_metadata(filp, arg);
1960 
1961 	case XFS_IOC_FD_TO_HANDLE:
1962 	case XFS_IOC_PATH_TO_HANDLE:
1963 	case XFS_IOC_PATH_TO_FSHANDLE: {
1964 		xfs_fsop_handlereq_t	hreq;
1965 
1966 		if (copy_from_user(&hreq, arg, sizeof(hreq)))
1967 			return -EFAULT;
1968 		return xfs_find_handle(cmd, &hreq);
1969 	}
1970 	case XFS_IOC_OPEN_BY_HANDLE: {
1971 		xfs_fsop_handlereq_t	hreq;
1972 
1973 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1974 			return -EFAULT;
1975 		return xfs_open_by_handle(filp, &hreq);
1976 	}
1977 
1978 	case XFS_IOC_READLINK_BY_HANDLE: {
1979 		xfs_fsop_handlereq_t	hreq;
1980 
1981 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1982 			return -EFAULT;
1983 		return xfs_readlink_by_handle(filp, &hreq);
1984 	}
1985 	case XFS_IOC_ATTRLIST_BY_HANDLE:
1986 		return xfs_attrlist_by_handle(filp, arg);
1987 
1988 	case XFS_IOC_ATTRMULTI_BY_HANDLE:
1989 		return xfs_attrmulti_by_handle(filp, arg);
1990 
1991 	case XFS_IOC_SWAPEXT: {
1992 		struct xfs_swapext	sxp;
1993 
1994 		if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
1995 			return -EFAULT;
1996 		error = mnt_want_write_file(filp);
1997 		if (error)
1998 			return error;
1999 		error = xfs_ioc_swapext(&sxp);
2000 		mnt_drop_write_file(filp);
2001 		return error;
2002 	}
2003 
2004 	case XFS_IOC_FSCOUNTS: {
2005 		xfs_fsop_counts_t out;
2006 
2007 		xfs_fs_counts(mp, &out);
2008 
2009 		if (copy_to_user(arg, &out, sizeof(out)))
2010 			return -EFAULT;
2011 		return 0;
2012 	}
2013 
2014 	case XFS_IOC_SET_RESBLKS: {
2015 		xfs_fsop_resblks_t inout;
2016 		uint64_t	   in;
2017 
2018 		if (!capable(CAP_SYS_ADMIN))
2019 			return -EPERM;
2020 
2021 		if (xfs_is_readonly(mp))
2022 			return -EROFS;
2023 
2024 		if (copy_from_user(&inout, arg, sizeof(inout)))
2025 			return -EFAULT;
2026 
2027 		error = mnt_want_write_file(filp);
2028 		if (error)
2029 			return error;
2030 
2031 		/* input parameter is passed in resblks field of structure */
2032 		in = inout.resblks;
2033 		error = xfs_reserve_blocks(mp, &in, &inout);
2034 		mnt_drop_write_file(filp);
2035 		if (error)
2036 			return error;
2037 
2038 		if (copy_to_user(arg, &inout, sizeof(inout)))
2039 			return -EFAULT;
2040 		return 0;
2041 	}
2042 
2043 	case XFS_IOC_GET_RESBLKS: {
2044 		xfs_fsop_resblks_t out;
2045 
2046 		if (!capable(CAP_SYS_ADMIN))
2047 			return -EPERM;
2048 
2049 		error = xfs_reserve_blocks(mp, NULL, &out);
2050 		if (error)
2051 			return error;
2052 
2053 		if (copy_to_user(arg, &out, sizeof(out)))
2054 			return -EFAULT;
2055 
2056 		return 0;
2057 	}
2058 
2059 	case XFS_IOC_FSGROWFSDATA: {
2060 		struct xfs_growfs_data in;
2061 
2062 		if (copy_from_user(&in, arg, sizeof(in)))
2063 			return -EFAULT;
2064 
2065 		error = mnt_want_write_file(filp);
2066 		if (error)
2067 			return error;
2068 		error = xfs_growfs_data(mp, &in);
2069 		mnt_drop_write_file(filp);
2070 		return error;
2071 	}
2072 
2073 	case XFS_IOC_FSGROWFSLOG: {
2074 		struct xfs_growfs_log in;
2075 
2076 		if (copy_from_user(&in, arg, sizeof(in)))
2077 			return -EFAULT;
2078 
2079 		error = mnt_want_write_file(filp);
2080 		if (error)
2081 			return error;
2082 		error = xfs_growfs_log(mp, &in);
2083 		mnt_drop_write_file(filp);
2084 		return error;
2085 	}
2086 
2087 	case XFS_IOC_FSGROWFSRT: {
2088 		xfs_growfs_rt_t in;
2089 
2090 		if (copy_from_user(&in, arg, sizeof(in)))
2091 			return -EFAULT;
2092 
2093 		error = mnt_want_write_file(filp);
2094 		if (error)
2095 			return error;
2096 		error = xfs_growfs_rt(mp, &in);
2097 		mnt_drop_write_file(filp);
2098 		return error;
2099 	}
2100 
2101 	case XFS_IOC_GOINGDOWN: {
2102 		uint32_t in;
2103 
2104 		if (!capable(CAP_SYS_ADMIN))
2105 			return -EPERM;
2106 
2107 		if (get_user(in, (uint32_t __user *)arg))
2108 			return -EFAULT;
2109 
2110 		return xfs_fs_goingdown(mp, in);
2111 	}
2112 
2113 	case XFS_IOC_ERROR_INJECTION: {
2114 		xfs_error_injection_t in;
2115 
2116 		if (!capable(CAP_SYS_ADMIN))
2117 			return -EPERM;
2118 
2119 		if (copy_from_user(&in, arg, sizeof(in)))
2120 			return -EFAULT;
2121 
2122 		return xfs_errortag_add(mp, in.errtag);
2123 	}
2124 
2125 	case XFS_IOC_ERROR_CLEARALL:
2126 		if (!capable(CAP_SYS_ADMIN))
2127 			return -EPERM;
2128 
2129 		return xfs_errortag_clearall(mp);
2130 
2131 	case XFS_IOC_FREE_EOFBLOCKS: {
2132 		struct xfs_fs_eofblocks	eofb;
2133 		struct xfs_icwalk	icw;
2134 
2135 		if (!capable(CAP_SYS_ADMIN))
2136 			return -EPERM;
2137 
2138 		if (xfs_is_readonly(mp))
2139 			return -EROFS;
2140 
2141 		if (copy_from_user(&eofb, arg, sizeof(eofb)))
2142 			return -EFAULT;
2143 
2144 		error = xfs_fs_eofblocks_from_user(&eofb, &icw);
2145 		if (error)
2146 			return error;
2147 
2148 		trace_xfs_ioc_free_eofblocks(mp, &icw, _RET_IP_);
2149 
2150 		sb_start_write(mp->m_super);
2151 		error = xfs_blockgc_free_space(mp, &icw);
2152 		sb_end_write(mp->m_super);
2153 		return error;
2154 	}
2155 
2156 	default:
2157 		return -ENOTTY;
2158 	}
2159 }
2160