xref: /linux/fs/xfs/xfs_ioctl.c (revision 987b741c52c7c6c68d46fbaeb95b8d1087f10b7f)
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_attr.h"
19 #include "xfs_bmap.h"
20 #include "xfs_bmap_util.h"
21 #include "xfs_fsops.h"
22 #include "xfs_discard.h"
23 #include "xfs_quota.h"
24 #include "xfs_export.h"
25 #include "xfs_trace.h"
26 #include "xfs_icache.h"
27 #include "xfs_trans.h"
28 #include "xfs_acl.h"
29 #include "xfs_btree.h"
30 #include <linux/fsmap.h>
31 #include "xfs_fsmap.h"
32 #include "scrub/xfs_scrub.h"
33 #include "xfs_sb.h"
34 #include "xfs_ag.h"
35 #include "xfs_health.h"
36 #include "xfs_reflink.h"
37 #include "xfs_ioctl.h"
38 #include "xfs_da_format.h"
39 #include "xfs_da_btree.h"
40 
41 #include <linux/mount.h>
42 #include <linux/namei.h>
43 #include <linux/fileattr.h>
44 
45 /*
46  * xfs_find_handle maps from userspace xfs_fsop_handlereq structure to
47  * a file or fs handle.
48  *
49  * XFS_IOC_PATH_TO_FSHANDLE
50  *    returns fs handle for a mount point or path within that mount point
51  * XFS_IOC_FD_TO_HANDLE
52  *    returns full handle for a FD opened in user space
53  * XFS_IOC_PATH_TO_HANDLE
54  *    returns full handle for a path
55  */
56 int
57 xfs_find_handle(
58 	unsigned int		cmd,
59 	xfs_fsop_handlereq_t	*hreq)
60 {
61 	int			hsize;
62 	xfs_handle_t		handle;
63 	struct inode		*inode;
64 	struct fd		f = {NULL};
65 	struct path		path;
66 	int			error;
67 	struct xfs_inode	*ip;
68 
69 	if (cmd == XFS_IOC_FD_TO_HANDLE) {
70 		f = fdget(hreq->fd);
71 		if (!f.file)
72 			return -EBADF;
73 		inode = file_inode(f.file);
74 	} else {
75 		error = user_path_at(AT_FDCWD, hreq->path, 0, &path);
76 		if (error)
77 			return error;
78 		inode = d_inode(path.dentry);
79 	}
80 	ip = XFS_I(inode);
81 
82 	/*
83 	 * We can only generate handles for inodes residing on a XFS filesystem,
84 	 * and only for regular files, directories or symbolic links.
85 	 */
86 	error = -EINVAL;
87 	if (inode->i_sb->s_magic != XFS_SB_MAGIC)
88 		goto out_put;
89 
90 	error = -EBADF;
91 	if (!S_ISREG(inode->i_mode) &&
92 	    !S_ISDIR(inode->i_mode) &&
93 	    !S_ISLNK(inode->i_mode))
94 		goto out_put;
95 
96 
97 	memcpy(&handle.ha_fsid, ip->i_mount->m_fixedfsid, sizeof(xfs_fsid_t));
98 
99 	if (cmd == XFS_IOC_PATH_TO_FSHANDLE) {
100 		/*
101 		 * This handle only contains an fsid, zero the rest.
102 		 */
103 		memset(&handle.ha_fid, 0, sizeof(handle.ha_fid));
104 		hsize = sizeof(xfs_fsid_t);
105 	} else {
106 		handle.ha_fid.fid_len = sizeof(xfs_fid_t) -
107 					sizeof(handle.ha_fid.fid_len);
108 		handle.ha_fid.fid_pad = 0;
109 		handle.ha_fid.fid_gen = inode->i_generation;
110 		handle.ha_fid.fid_ino = ip->i_ino;
111 		hsize = sizeof(xfs_handle_t);
112 	}
113 
114 	error = -EFAULT;
115 	if (copy_to_user(hreq->ohandle, &handle, hsize) ||
116 	    copy_to_user(hreq->ohandlen, &hsize, sizeof(__s32)))
117 		goto out_put;
118 
119 	error = 0;
120 
121  out_put:
122 	if (cmd == XFS_IOC_FD_TO_HANDLE)
123 		fdput(f);
124 	else
125 		path_put(&path);
126 	return error;
127 }
128 
129 /*
130  * No need to do permission checks on the various pathname components
131  * as the handle operations are privileged.
132  */
133 STATIC int
134 xfs_handle_acceptable(
135 	void			*context,
136 	struct dentry		*dentry)
137 {
138 	return 1;
139 }
140 
141 /*
142  * Convert userspace handle data into a dentry.
143  */
144 struct dentry *
145 xfs_handle_to_dentry(
146 	struct file		*parfilp,
147 	void __user		*uhandle,
148 	u32			hlen)
149 {
150 	xfs_handle_t		handle;
151 	struct xfs_fid64	fid;
152 
153 	/*
154 	 * Only allow handle opens under a directory.
155 	 */
156 	if (!S_ISDIR(file_inode(parfilp)->i_mode))
157 		return ERR_PTR(-ENOTDIR);
158 
159 	if (hlen != sizeof(xfs_handle_t))
160 		return ERR_PTR(-EINVAL);
161 	if (copy_from_user(&handle, uhandle, hlen))
162 		return ERR_PTR(-EFAULT);
163 	if (handle.ha_fid.fid_len !=
164 	    sizeof(handle.ha_fid) - sizeof(handle.ha_fid.fid_len))
165 		return ERR_PTR(-EINVAL);
166 
167 	memset(&fid, 0, sizeof(struct fid));
168 	fid.ino = handle.ha_fid.fid_ino;
169 	fid.gen = handle.ha_fid.fid_gen;
170 
171 	return exportfs_decode_fh(parfilp->f_path.mnt, (struct fid *)&fid, 3,
172 			FILEID_INO32_GEN | XFS_FILEID_TYPE_64FLAG,
173 			xfs_handle_acceptable, NULL);
174 }
175 
176 STATIC struct dentry *
177 xfs_handlereq_to_dentry(
178 	struct file		*parfilp,
179 	xfs_fsop_handlereq_t	*hreq)
180 {
181 	return xfs_handle_to_dentry(parfilp, hreq->ihandle, hreq->ihandlen);
182 }
183 
184 int
185 xfs_open_by_handle(
186 	struct file		*parfilp,
187 	xfs_fsop_handlereq_t	*hreq)
188 {
189 	const struct cred	*cred = current_cred();
190 	int			error;
191 	int			fd;
192 	int			permflag;
193 	struct file		*filp;
194 	struct inode		*inode;
195 	struct dentry		*dentry;
196 	fmode_t			fmode;
197 	struct path		path;
198 
199 	if (!capable(CAP_SYS_ADMIN))
200 		return -EPERM;
201 
202 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
203 	if (IS_ERR(dentry))
204 		return PTR_ERR(dentry);
205 	inode = d_inode(dentry);
206 
207 	/* Restrict xfs_open_by_handle to directories & regular files. */
208 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))) {
209 		error = -EPERM;
210 		goto out_dput;
211 	}
212 
213 #if BITS_PER_LONG != 32
214 	hreq->oflags |= O_LARGEFILE;
215 #endif
216 
217 	permflag = hreq->oflags;
218 	fmode = OPEN_FMODE(permflag);
219 	if ((!(permflag & O_APPEND) || (permflag & O_TRUNC)) &&
220 	    (fmode & FMODE_WRITE) && IS_APPEND(inode)) {
221 		error = -EPERM;
222 		goto out_dput;
223 	}
224 
225 	if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
226 		error = -EPERM;
227 		goto out_dput;
228 	}
229 
230 	/* Can't write directories. */
231 	if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
232 		error = -EISDIR;
233 		goto out_dput;
234 	}
235 
236 	fd = get_unused_fd_flags(0);
237 	if (fd < 0) {
238 		error = fd;
239 		goto out_dput;
240 	}
241 
242 	path.mnt = parfilp->f_path.mnt;
243 	path.dentry = dentry;
244 	filp = dentry_open(&path, hreq->oflags, cred);
245 	dput(dentry);
246 	if (IS_ERR(filp)) {
247 		put_unused_fd(fd);
248 		return PTR_ERR(filp);
249 	}
250 
251 	if (S_ISREG(inode->i_mode)) {
252 		filp->f_flags |= O_NOATIME;
253 		filp->f_mode |= FMODE_NOCMTIME;
254 	}
255 
256 	fd_install(fd, filp);
257 	return fd;
258 
259  out_dput:
260 	dput(dentry);
261 	return error;
262 }
263 
264 int
265 xfs_readlink_by_handle(
266 	struct file		*parfilp,
267 	xfs_fsop_handlereq_t	*hreq)
268 {
269 	struct dentry		*dentry;
270 	__u32			olen;
271 	int			error;
272 
273 	if (!capable(CAP_SYS_ADMIN))
274 		return -EPERM;
275 
276 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
277 	if (IS_ERR(dentry))
278 		return PTR_ERR(dentry);
279 
280 	/* Restrict this handle operation to symlinks only. */
281 	if (!d_is_symlink(dentry)) {
282 		error = -EINVAL;
283 		goto out_dput;
284 	}
285 
286 	if (copy_from_user(&olen, hreq->ohandlen, sizeof(__u32))) {
287 		error = -EFAULT;
288 		goto out_dput;
289 	}
290 
291 	error = vfs_readlink(dentry, hreq->ohandle, olen);
292 
293  out_dput:
294 	dput(dentry);
295 	return error;
296 }
297 
298 /*
299  * Format an attribute and copy it out to the user's buffer.
300  * Take care to check values and protect against them changing later,
301  * we may be reading them directly out of a user buffer.
302  */
303 static void
304 xfs_ioc_attr_put_listent(
305 	struct xfs_attr_list_context *context,
306 	int			flags,
307 	unsigned char		*name,
308 	int			namelen,
309 	int			valuelen)
310 {
311 	struct xfs_attrlist	*alist = context->buffer;
312 	struct xfs_attrlist_ent	*aep;
313 	int			arraytop;
314 
315 	ASSERT(!context->seen_enough);
316 	ASSERT(context->count >= 0);
317 	ASSERT(context->count < (ATTR_MAX_VALUELEN/8));
318 	ASSERT(context->firstu >= sizeof(*alist));
319 	ASSERT(context->firstu <= context->bufsize);
320 
321 	/*
322 	 * Only list entries in the right namespace.
323 	 */
324 	if (context->attr_filter != (flags & XFS_ATTR_NSP_ONDISK_MASK))
325 		return;
326 
327 	arraytop = sizeof(*alist) +
328 			context->count * sizeof(alist->al_offset[0]);
329 
330 	/* decrement by the actual bytes used by the attr */
331 	context->firstu -= round_up(offsetof(struct xfs_attrlist_ent, a_name) +
332 			namelen + 1, sizeof(uint32_t));
333 	if (context->firstu < arraytop) {
334 		trace_xfs_attr_list_full(context);
335 		alist->al_more = 1;
336 		context->seen_enough = 1;
337 		return;
338 	}
339 
340 	aep = context->buffer + context->firstu;
341 	aep->a_valuelen = valuelen;
342 	memcpy(aep->a_name, name, namelen);
343 	aep->a_name[namelen] = 0;
344 	alist->al_offset[context->count++] = context->firstu;
345 	alist->al_count = context->count;
346 	trace_xfs_attr_list_add(context);
347 }
348 
349 static unsigned int
350 xfs_attr_filter(
351 	u32			ioc_flags)
352 {
353 	if (ioc_flags & XFS_IOC_ATTR_ROOT)
354 		return XFS_ATTR_ROOT;
355 	if (ioc_flags & XFS_IOC_ATTR_SECURE)
356 		return XFS_ATTR_SECURE;
357 	return 0;
358 }
359 
360 static unsigned int
361 xfs_attr_flags(
362 	u32			ioc_flags)
363 {
364 	if (ioc_flags & XFS_IOC_ATTR_CREATE)
365 		return XATTR_CREATE;
366 	if (ioc_flags & XFS_IOC_ATTR_REPLACE)
367 		return XATTR_REPLACE;
368 	return 0;
369 }
370 
371 int
372 xfs_ioc_attr_list(
373 	struct xfs_inode		*dp,
374 	void __user			*ubuf,
375 	int				bufsize,
376 	int				flags,
377 	struct xfs_attrlist_cursor __user *ucursor)
378 {
379 	struct xfs_attr_list_context	context = { };
380 	struct xfs_attrlist		*alist;
381 	void				*buffer;
382 	int				error;
383 
384 	if (bufsize < sizeof(struct xfs_attrlist) ||
385 	    bufsize > XFS_XATTR_LIST_MAX)
386 		return -EINVAL;
387 
388 	/*
389 	 * Reject flags, only allow namespaces.
390 	 */
391 	if (flags & ~(XFS_IOC_ATTR_ROOT | XFS_IOC_ATTR_SECURE))
392 		return -EINVAL;
393 	if (flags == (XFS_IOC_ATTR_ROOT | XFS_IOC_ATTR_SECURE))
394 		return -EINVAL;
395 
396 	/*
397 	 * Validate the cursor.
398 	 */
399 	if (copy_from_user(&context.cursor, ucursor, sizeof(context.cursor)))
400 		return -EFAULT;
401 	if (context.cursor.pad1 || context.cursor.pad2)
402 		return -EINVAL;
403 	if (!context.cursor.initted &&
404 	    (context.cursor.hashval || context.cursor.blkno ||
405 	     context.cursor.offset))
406 		return -EINVAL;
407 
408 	buffer = kvzalloc(bufsize, GFP_KERNEL);
409 	if (!buffer)
410 		return -ENOMEM;
411 
412 	/*
413 	 * Initialize the output buffer.
414 	 */
415 	context.dp = dp;
416 	context.resynch = 1;
417 	context.attr_filter = xfs_attr_filter(flags);
418 	context.buffer = buffer;
419 	context.bufsize = round_down(bufsize, sizeof(uint32_t));
420 	context.firstu = context.bufsize;
421 	context.put_listent = xfs_ioc_attr_put_listent;
422 
423 	alist = context.buffer;
424 	alist->al_count = 0;
425 	alist->al_more = 0;
426 	alist->al_offset[0] = context.bufsize;
427 
428 	error = xfs_attr_list(&context);
429 	if (error)
430 		goto out_free;
431 
432 	if (copy_to_user(ubuf, buffer, bufsize) ||
433 	    copy_to_user(ucursor, &context.cursor, sizeof(context.cursor)))
434 		error = -EFAULT;
435 out_free:
436 	kmem_free(buffer);
437 	return error;
438 }
439 
440 STATIC int
441 xfs_attrlist_by_handle(
442 	struct file		*parfilp,
443 	struct xfs_fsop_attrlist_handlereq __user *p)
444 {
445 	struct xfs_fsop_attrlist_handlereq al_hreq;
446 	struct dentry		*dentry;
447 	int			error = -ENOMEM;
448 
449 	if (!capable(CAP_SYS_ADMIN))
450 		return -EPERM;
451 	if (copy_from_user(&al_hreq, p, sizeof(al_hreq)))
452 		return -EFAULT;
453 
454 	dentry = xfs_handlereq_to_dentry(parfilp, &al_hreq.hreq);
455 	if (IS_ERR(dentry))
456 		return PTR_ERR(dentry);
457 
458 	error = xfs_ioc_attr_list(XFS_I(d_inode(dentry)), al_hreq.buffer,
459 				  al_hreq.buflen, al_hreq.flags, &p->pos);
460 	dput(dentry);
461 	return error;
462 }
463 
464 static int
465 xfs_attrmulti_attr_get(
466 	struct inode		*inode,
467 	unsigned char		*name,
468 	unsigned char		__user *ubuf,
469 	uint32_t		*len,
470 	uint32_t		flags)
471 {
472 	struct xfs_da_args	args = {
473 		.dp		= XFS_I(inode),
474 		.attr_filter	= xfs_attr_filter(flags),
475 		.attr_flags	= xfs_attr_flags(flags),
476 		.name		= name,
477 		.namelen	= strlen(name),
478 		.valuelen	= *len,
479 	};
480 	int			error;
481 
482 	if (*len > XFS_XATTR_SIZE_MAX)
483 		return -EINVAL;
484 
485 	error = xfs_attr_get(&args);
486 	if (error)
487 		goto out_kfree;
488 
489 	*len = args.valuelen;
490 	if (copy_to_user(ubuf, args.value, args.valuelen))
491 		error = -EFAULT;
492 
493 out_kfree:
494 	kmem_free(args.value);
495 	return error;
496 }
497 
498 static int
499 xfs_attrmulti_attr_set(
500 	struct inode		*inode,
501 	unsigned char		*name,
502 	const unsigned char	__user *ubuf,
503 	uint32_t		len,
504 	uint32_t		flags)
505 {
506 	struct xfs_da_args	args = {
507 		.dp		= XFS_I(inode),
508 		.attr_filter	= xfs_attr_filter(flags),
509 		.attr_flags	= xfs_attr_flags(flags),
510 		.name		= name,
511 		.namelen	= strlen(name),
512 	};
513 	int			error;
514 
515 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
516 		return -EPERM;
517 
518 	if (ubuf) {
519 		if (len > XFS_XATTR_SIZE_MAX)
520 			return -EINVAL;
521 		args.value = memdup_user(ubuf, len);
522 		if (IS_ERR(args.value))
523 			return PTR_ERR(args.value);
524 		args.valuelen = len;
525 	}
526 
527 	error = xfs_attr_set(&args);
528 	if (!error && (flags & XFS_IOC_ATTR_ROOT))
529 		xfs_forget_acl(inode, name);
530 	kfree(args.value);
531 	return error;
532 }
533 
534 int
535 xfs_ioc_attrmulti_one(
536 	struct file		*parfilp,
537 	struct inode		*inode,
538 	uint32_t		opcode,
539 	void __user		*uname,
540 	void __user		*value,
541 	uint32_t		*len,
542 	uint32_t		flags)
543 {
544 	unsigned char		*name;
545 	int			error;
546 
547 	if ((flags & XFS_IOC_ATTR_ROOT) && (flags & XFS_IOC_ATTR_SECURE))
548 		return -EINVAL;
549 
550 	name = strndup_user(uname, MAXNAMELEN);
551 	if (IS_ERR(name))
552 		return PTR_ERR(name);
553 
554 	switch (opcode) {
555 	case ATTR_OP_GET:
556 		error = xfs_attrmulti_attr_get(inode, name, value, len, flags);
557 		break;
558 	case ATTR_OP_REMOVE:
559 		value = NULL;
560 		*len = 0;
561 		/* fall through */
562 	case ATTR_OP_SET:
563 		error = mnt_want_write_file(parfilp);
564 		if (error)
565 			break;
566 		error = xfs_attrmulti_attr_set(inode, name, value, *len, flags);
567 		mnt_drop_write_file(parfilp);
568 		break;
569 	default:
570 		error = -EINVAL;
571 		break;
572 	}
573 
574 	kfree(name);
575 	return error;
576 }
577 
578 STATIC int
579 xfs_attrmulti_by_handle(
580 	struct file		*parfilp,
581 	void			__user *arg)
582 {
583 	int			error;
584 	xfs_attr_multiop_t	*ops;
585 	xfs_fsop_attrmulti_handlereq_t am_hreq;
586 	struct dentry		*dentry;
587 	unsigned int		i, size;
588 
589 	if (!capable(CAP_SYS_ADMIN))
590 		return -EPERM;
591 	if (copy_from_user(&am_hreq, arg, sizeof(xfs_fsop_attrmulti_handlereq_t)))
592 		return -EFAULT;
593 
594 	/* overflow check */
595 	if (am_hreq.opcount >= INT_MAX / sizeof(xfs_attr_multiop_t))
596 		return -E2BIG;
597 
598 	dentry = xfs_handlereq_to_dentry(parfilp, &am_hreq.hreq);
599 	if (IS_ERR(dentry))
600 		return PTR_ERR(dentry);
601 
602 	error = -E2BIG;
603 	size = am_hreq.opcount * sizeof(xfs_attr_multiop_t);
604 	if (!size || size > 16 * PAGE_SIZE)
605 		goto out_dput;
606 
607 	ops = memdup_user(am_hreq.ops, size);
608 	if (IS_ERR(ops)) {
609 		error = PTR_ERR(ops);
610 		goto out_dput;
611 	}
612 
613 	error = 0;
614 	for (i = 0; i < am_hreq.opcount; i++) {
615 		ops[i].am_error = xfs_ioc_attrmulti_one(parfilp,
616 				d_inode(dentry), ops[i].am_opcode,
617 				ops[i].am_attrname, ops[i].am_attrvalue,
618 				&ops[i].am_length, ops[i].am_flags);
619 	}
620 
621 	if (copy_to_user(am_hreq.ops, ops, size))
622 		error = -EFAULT;
623 
624 	kfree(ops);
625  out_dput:
626 	dput(dentry);
627 	return error;
628 }
629 
630 int
631 xfs_ioc_space(
632 	struct file		*filp,
633 	xfs_flock64_t		*bf)
634 {
635 	struct inode		*inode = file_inode(filp);
636 	struct xfs_inode	*ip = XFS_I(inode);
637 	struct iattr		iattr;
638 	enum xfs_prealloc_flags	flags = XFS_PREALLOC_CLEAR;
639 	uint			iolock = XFS_IOLOCK_EXCL | XFS_MMAPLOCK_EXCL;
640 	int			error;
641 
642 	if (inode->i_flags & (S_IMMUTABLE|S_APPEND))
643 		return -EPERM;
644 
645 	if (!(filp->f_mode & FMODE_WRITE))
646 		return -EBADF;
647 
648 	if (!S_ISREG(inode->i_mode))
649 		return -EINVAL;
650 
651 	if (xfs_is_always_cow_inode(ip))
652 		return -EOPNOTSUPP;
653 
654 	if (filp->f_flags & O_DSYNC)
655 		flags |= XFS_PREALLOC_SYNC;
656 	if (filp->f_mode & FMODE_NOCMTIME)
657 		flags |= XFS_PREALLOC_INVISIBLE;
658 
659 	error = mnt_want_write_file(filp);
660 	if (error)
661 		return error;
662 
663 	xfs_ilock(ip, iolock);
664 	error = xfs_break_layouts(inode, &iolock, BREAK_UNMAP);
665 	if (error)
666 		goto out_unlock;
667 	inode_dio_wait(inode);
668 
669 	switch (bf->l_whence) {
670 	case 0: /*SEEK_SET*/
671 		break;
672 	case 1: /*SEEK_CUR*/
673 		bf->l_start += filp->f_pos;
674 		break;
675 	case 2: /*SEEK_END*/
676 		bf->l_start += XFS_ISIZE(ip);
677 		break;
678 	default:
679 		error = -EINVAL;
680 		goto out_unlock;
681 	}
682 
683 	if (bf->l_start < 0 || bf->l_start > inode->i_sb->s_maxbytes) {
684 		error = -EINVAL;
685 		goto out_unlock;
686 	}
687 
688 	if (bf->l_start > XFS_ISIZE(ip)) {
689 		error = xfs_alloc_file_space(ip, XFS_ISIZE(ip),
690 				bf->l_start - XFS_ISIZE(ip), 0);
691 		if (error)
692 			goto out_unlock;
693 	}
694 
695 	iattr.ia_valid = ATTR_SIZE;
696 	iattr.ia_size = bf->l_start;
697 	error = xfs_vn_setattr_size(file_mnt_user_ns(filp), file_dentry(filp),
698 				    &iattr);
699 	if (error)
700 		goto out_unlock;
701 
702 	error = xfs_update_prealloc_flags(ip, flags);
703 
704 out_unlock:
705 	xfs_iunlock(ip, iolock);
706 	mnt_drop_write_file(filp);
707 	return error;
708 }
709 
710 /* Return 0 on success or positive error */
711 int
712 xfs_fsbulkstat_one_fmt(
713 	struct xfs_ibulk		*breq,
714 	const struct xfs_bulkstat	*bstat)
715 {
716 	struct xfs_bstat		bs1;
717 
718 	xfs_bulkstat_to_bstat(breq->mp, &bs1, bstat);
719 	if (copy_to_user(breq->ubuffer, &bs1, sizeof(bs1)))
720 		return -EFAULT;
721 	return xfs_ibulk_advance(breq, sizeof(struct xfs_bstat));
722 }
723 
724 int
725 xfs_fsinumbers_fmt(
726 	struct xfs_ibulk		*breq,
727 	const struct xfs_inumbers	*igrp)
728 {
729 	struct xfs_inogrp		ig1;
730 
731 	xfs_inumbers_to_inogrp(&ig1, igrp);
732 	if (copy_to_user(breq->ubuffer, &ig1, sizeof(struct xfs_inogrp)))
733 		return -EFAULT;
734 	return xfs_ibulk_advance(breq, sizeof(struct xfs_inogrp));
735 }
736 
737 STATIC int
738 xfs_ioc_fsbulkstat(
739 	struct file		*file,
740 	unsigned int		cmd,
741 	void			__user *arg)
742 {
743 	struct xfs_mount	*mp = XFS_I(file_inode(file))->i_mount;
744 	struct xfs_fsop_bulkreq	bulkreq;
745 	struct xfs_ibulk	breq = {
746 		.mp		= mp,
747 		.mnt_userns	= file_mnt_user_ns(file),
748 		.ocount		= 0,
749 	};
750 	xfs_ino_t		lastino;
751 	int			error;
752 
753 	/* done = 1 if there are more stats to get and if bulkstat */
754 	/* should be called again (unused here, but used in dmapi) */
755 
756 	if (!capable(CAP_SYS_ADMIN))
757 		return -EPERM;
758 
759 	if (XFS_FORCED_SHUTDOWN(mp))
760 		return -EIO;
761 
762 	if (copy_from_user(&bulkreq, arg, sizeof(struct xfs_fsop_bulkreq)))
763 		return -EFAULT;
764 
765 	if (copy_from_user(&lastino, bulkreq.lastip, sizeof(__s64)))
766 		return -EFAULT;
767 
768 	if (bulkreq.icount <= 0)
769 		return -EINVAL;
770 
771 	if (bulkreq.ubuffer == NULL)
772 		return -EINVAL;
773 
774 	breq.ubuffer = bulkreq.ubuffer;
775 	breq.icount = bulkreq.icount;
776 
777 	/*
778 	 * FSBULKSTAT_SINGLE expects that *lastip contains the inode number
779 	 * that we want to stat.  However, FSINUMBERS and FSBULKSTAT expect
780 	 * that *lastip contains either zero or the number of the last inode to
781 	 * be examined by the previous call and return results starting with
782 	 * the next inode after that.  The new bulk request back end functions
783 	 * take the inode to start with, so we have to compute the startino
784 	 * parameter from lastino to maintain correct function.  lastino == 0
785 	 * is a special case because it has traditionally meant "first inode
786 	 * in filesystem".
787 	 */
788 	if (cmd == XFS_IOC_FSINUMBERS) {
789 		breq.startino = lastino ? lastino + 1 : 0;
790 		error = xfs_inumbers(&breq, xfs_fsinumbers_fmt);
791 		lastino = breq.startino - 1;
792 	} else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE) {
793 		breq.startino = lastino;
794 		breq.icount = 1;
795 		error = xfs_bulkstat_one(&breq, xfs_fsbulkstat_one_fmt);
796 	} else {	/* XFS_IOC_FSBULKSTAT */
797 		breq.startino = lastino ? lastino + 1 : 0;
798 		error = xfs_bulkstat(&breq, xfs_fsbulkstat_one_fmt);
799 		lastino = breq.startino - 1;
800 	}
801 
802 	if (error)
803 		return error;
804 
805 	if (bulkreq.lastip != NULL &&
806 	    copy_to_user(bulkreq.lastip, &lastino, sizeof(xfs_ino_t)))
807 		return -EFAULT;
808 
809 	if (bulkreq.ocount != NULL &&
810 	    copy_to_user(bulkreq.ocount, &breq.ocount, sizeof(__s32)))
811 		return -EFAULT;
812 
813 	return 0;
814 }
815 
816 /* Return 0 on success or positive error */
817 static int
818 xfs_bulkstat_fmt(
819 	struct xfs_ibulk		*breq,
820 	const struct xfs_bulkstat	*bstat)
821 {
822 	if (copy_to_user(breq->ubuffer, bstat, sizeof(struct xfs_bulkstat)))
823 		return -EFAULT;
824 	return xfs_ibulk_advance(breq, sizeof(struct xfs_bulkstat));
825 }
826 
827 /*
828  * Check the incoming bulk request @hdr from userspace and initialize the
829  * internal @breq bulk request appropriately.  Returns 0 if the bulk request
830  * should proceed; -ECANCELED if there's nothing to do; or the usual
831  * negative error code.
832  */
833 static int
834 xfs_bulk_ireq_setup(
835 	struct xfs_mount	*mp,
836 	struct xfs_bulk_ireq	*hdr,
837 	struct xfs_ibulk	*breq,
838 	void __user		*ubuffer)
839 {
840 	if (hdr->icount == 0 ||
841 	    (hdr->flags & ~XFS_BULK_IREQ_FLAGS_ALL) ||
842 	    memchr_inv(hdr->reserved, 0, sizeof(hdr->reserved)))
843 		return -EINVAL;
844 
845 	breq->startino = hdr->ino;
846 	breq->ubuffer = ubuffer;
847 	breq->icount = hdr->icount;
848 	breq->ocount = 0;
849 	breq->flags = 0;
850 
851 	/*
852 	 * The @ino parameter is a special value, so we must look it up here.
853 	 * We're not allowed to have IREQ_AGNO, and we only return one inode
854 	 * worth of data.
855 	 */
856 	if (hdr->flags & XFS_BULK_IREQ_SPECIAL) {
857 		if (hdr->flags & XFS_BULK_IREQ_AGNO)
858 			return -EINVAL;
859 
860 		switch (hdr->ino) {
861 		case XFS_BULK_IREQ_SPECIAL_ROOT:
862 			hdr->ino = mp->m_sb.sb_rootino;
863 			break;
864 		default:
865 			return -EINVAL;
866 		}
867 		breq->icount = 1;
868 	}
869 
870 	/*
871 	 * The IREQ_AGNO flag means that we only want results from a given AG.
872 	 * If @hdr->ino is zero, we start iterating in that AG.  If @hdr->ino is
873 	 * beyond the specified AG then we return no results.
874 	 */
875 	if (hdr->flags & XFS_BULK_IREQ_AGNO) {
876 		if (hdr->agno >= mp->m_sb.sb_agcount)
877 			return -EINVAL;
878 
879 		if (breq->startino == 0)
880 			breq->startino = XFS_AGINO_TO_INO(mp, hdr->agno, 0);
881 		else if (XFS_INO_TO_AGNO(mp, breq->startino) < hdr->agno)
882 			return -EINVAL;
883 
884 		breq->flags |= XFS_IBULK_SAME_AG;
885 
886 		/* Asking for an inode past the end of the AG?  We're done! */
887 		if (XFS_INO_TO_AGNO(mp, breq->startino) > hdr->agno)
888 			return -ECANCELED;
889 	} else if (hdr->agno)
890 		return -EINVAL;
891 
892 	/* Asking for an inode past the end of the FS?  We're done! */
893 	if (XFS_INO_TO_AGNO(mp, breq->startino) >= mp->m_sb.sb_agcount)
894 		return -ECANCELED;
895 
896 	return 0;
897 }
898 
899 /*
900  * Update the userspace bulk request @hdr to reflect the end state of the
901  * internal bulk request @breq.
902  */
903 static void
904 xfs_bulk_ireq_teardown(
905 	struct xfs_bulk_ireq	*hdr,
906 	struct xfs_ibulk	*breq)
907 {
908 	hdr->ino = breq->startino;
909 	hdr->ocount = breq->ocount;
910 }
911 
912 /* Handle the v5 bulkstat ioctl. */
913 STATIC int
914 xfs_ioc_bulkstat(
915 	struct file			*file,
916 	unsigned int			cmd,
917 	struct xfs_bulkstat_req __user	*arg)
918 {
919 	struct xfs_mount		*mp = XFS_I(file_inode(file))->i_mount;
920 	struct xfs_bulk_ireq		hdr;
921 	struct xfs_ibulk		breq = {
922 		.mp			= mp,
923 		.mnt_userns		= file_mnt_user_ns(file),
924 	};
925 	int				error;
926 
927 	if (!capable(CAP_SYS_ADMIN))
928 		return -EPERM;
929 
930 	if (XFS_FORCED_SHUTDOWN(mp))
931 		return -EIO;
932 
933 	if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
934 		return -EFAULT;
935 
936 	error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->bulkstat);
937 	if (error == -ECANCELED)
938 		goto out_teardown;
939 	if (error < 0)
940 		return error;
941 
942 	error = xfs_bulkstat(&breq, xfs_bulkstat_fmt);
943 	if (error)
944 		return error;
945 
946 out_teardown:
947 	xfs_bulk_ireq_teardown(&hdr, &breq);
948 	if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
949 		return -EFAULT;
950 
951 	return 0;
952 }
953 
954 STATIC int
955 xfs_inumbers_fmt(
956 	struct xfs_ibulk		*breq,
957 	const struct xfs_inumbers	*igrp)
958 {
959 	if (copy_to_user(breq->ubuffer, igrp, sizeof(struct xfs_inumbers)))
960 		return -EFAULT;
961 	return xfs_ibulk_advance(breq, sizeof(struct xfs_inumbers));
962 }
963 
964 /* Handle the v5 inumbers ioctl. */
965 STATIC int
966 xfs_ioc_inumbers(
967 	struct xfs_mount		*mp,
968 	unsigned int			cmd,
969 	struct xfs_inumbers_req __user	*arg)
970 {
971 	struct xfs_bulk_ireq		hdr;
972 	struct xfs_ibulk		breq = {
973 		.mp			= mp,
974 	};
975 	int				error;
976 
977 	if (!capable(CAP_SYS_ADMIN))
978 		return -EPERM;
979 
980 	if (XFS_FORCED_SHUTDOWN(mp))
981 		return -EIO;
982 
983 	if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
984 		return -EFAULT;
985 
986 	error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->inumbers);
987 	if (error == -ECANCELED)
988 		goto out_teardown;
989 	if (error < 0)
990 		return error;
991 
992 	error = xfs_inumbers(&breq, xfs_inumbers_fmt);
993 	if (error)
994 		return error;
995 
996 out_teardown:
997 	xfs_bulk_ireq_teardown(&hdr, &breq);
998 	if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
999 		return -EFAULT;
1000 
1001 	return 0;
1002 }
1003 
1004 STATIC int
1005 xfs_ioc_fsgeometry(
1006 	struct xfs_mount	*mp,
1007 	void			__user *arg,
1008 	int			struct_version)
1009 {
1010 	struct xfs_fsop_geom	fsgeo;
1011 	size_t			len;
1012 
1013 	xfs_fs_geometry(&mp->m_sb, &fsgeo, struct_version);
1014 
1015 	if (struct_version <= 3)
1016 		len = sizeof(struct xfs_fsop_geom_v1);
1017 	else if (struct_version == 4)
1018 		len = sizeof(struct xfs_fsop_geom_v4);
1019 	else {
1020 		xfs_fsop_geom_health(mp, &fsgeo);
1021 		len = sizeof(fsgeo);
1022 	}
1023 
1024 	if (copy_to_user(arg, &fsgeo, len))
1025 		return -EFAULT;
1026 	return 0;
1027 }
1028 
1029 STATIC int
1030 xfs_ioc_ag_geometry(
1031 	struct xfs_mount	*mp,
1032 	void			__user *arg)
1033 {
1034 	struct xfs_ag_geometry	ageo;
1035 	int			error;
1036 
1037 	if (copy_from_user(&ageo, arg, sizeof(ageo)))
1038 		return -EFAULT;
1039 	if (ageo.ag_flags)
1040 		return -EINVAL;
1041 	if (memchr_inv(&ageo.ag_reserved, 0, sizeof(ageo.ag_reserved)))
1042 		return -EINVAL;
1043 
1044 	error = xfs_ag_get_geometry(mp, ageo.ag_number, &ageo);
1045 	if (error)
1046 		return error;
1047 
1048 	if (copy_to_user(arg, &ageo, sizeof(ageo)))
1049 		return -EFAULT;
1050 	return 0;
1051 }
1052 
1053 /*
1054  * Linux extended inode flags interface.
1055  */
1056 
1057 static void
1058 xfs_fill_fsxattr(
1059 	struct xfs_inode	*ip,
1060 	int			whichfork,
1061 	struct fileattr		*fa)
1062 {
1063 	struct xfs_mount	*mp = ip->i_mount;
1064 	struct xfs_ifork	*ifp = XFS_IFORK_PTR(ip, whichfork);
1065 
1066 	fileattr_fill_xflags(fa, xfs_ip2xflags(ip));
1067 
1068 	fa->fsx_extsize = XFS_FSB_TO_B(mp, ip->i_extsize);
1069 	if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
1070 		fa->fsx_cowextsize = XFS_FSB_TO_B(mp, ip->i_cowextsize);
1071 	fa->fsx_projid = ip->i_projid;
1072 	if (ifp && !xfs_need_iread_extents(ifp))
1073 		fa->fsx_nextents = xfs_iext_count(ifp);
1074 	else
1075 		fa->fsx_nextents = xfs_ifork_nextents(ifp);
1076 }
1077 
1078 STATIC int
1079 xfs_ioc_fsgetxattra(
1080 	xfs_inode_t		*ip,
1081 	void			__user *arg)
1082 {
1083 	struct fileattr		fa;
1084 
1085 	xfs_ilock(ip, XFS_ILOCK_SHARED);
1086 	xfs_fill_fsxattr(ip, XFS_ATTR_FORK, &fa);
1087 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
1088 
1089 	return copy_fsxattr_to_user(&fa, arg);
1090 }
1091 
1092 int
1093 xfs_fileattr_get(
1094 	struct dentry		*dentry,
1095 	struct fileattr		*fa)
1096 {
1097 	struct xfs_inode	*ip = XFS_I(d_inode(dentry));
1098 
1099 	if (d_is_special(dentry))
1100 		return -ENOTTY;
1101 
1102 	xfs_ilock(ip, XFS_ILOCK_SHARED);
1103 	xfs_fill_fsxattr(ip, XFS_DATA_FORK, fa);
1104 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
1105 
1106 	return 0;
1107 }
1108 
1109 STATIC uint16_t
1110 xfs_flags2diflags(
1111 	struct xfs_inode	*ip,
1112 	unsigned int		xflags)
1113 {
1114 	/* can't set PREALLOC this way, just preserve it */
1115 	uint16_t		di_flags =
1116 		(ip->i_diflags & XFS_DIFLAG_PREALLOC);
1117 
1118 	if (xflags & FS_XFLAG_IMMUTABLE)
1119 		di_flags |= XFS_DIFLAG_IMMUTABLE;
1120 	if (xflags & FS_XFLAG_APPEND)
1121 		di_flags |= XFS_DIFLAG_APPEND;
1122 	if (xflags & FS_XFLAG_SYNC)
1123 		di_flags |= XFS_DIFLAG_SYNC;
1124 	if (xflags & FS_XFLAG_NOATIME)
1125 		di_flags |= XFS_DIFLAG_NOATIME;
1126 	if (xflags & FS_XFLAG_NODUMP)
1127 		di_flags |= XFS_DIFLAG_NODUMP;
1128 	if (xflags & FS_XFLAG_NODEFRAG)
1129 		di_flags |= XFS_DIFLAG_NODEFRAG;
1130 	if (xflags & FS_XFLAG_FILESTREAM)
1131 		di_flags |= XFS_DIFLAG_FILESTREAM;
1132 	if (S_ISDIR(VFS_I(ip)->i_mode)) {
1133 		if (xflags & FS_XFLAG_RTINHERIT)
1134 			di_flags |= XFS_DIFLAG_RTINHERIT;
1135 		if (xflags & FS_XFLAG_NOSYMLINKS)
1136 			di_flags |= XFS_DIFLAG_NOSYMLINKS;
1137 		if (xflags & FS_XFLAG_EXTSZINHERIT)
1138 			di_flags |= XFS_DIFLAG_EXTSZINHERIT;
1139 		if (xflags & FS_XFLAG_PROJINHERIT)
1140 			di_flags |= XFS_DIFLAG_PROJINHERIT;
1141 	} else if (S_ISREG(VFS_I(ip)->i_mode)) {
1142 		if (xflags & FS_XFLAG_REALTIME)
1143 			di_flags |= XFS_DIFLAG_REALTIME;
1144 		if (xflags & FS_XFLAG_EXTSIZE)
1145 			di_flags |= XFS_DIFLAG_EXTSIZE;
1146 	}
1147 
1148 	return di_flags;
1149 }
1150 
1151 STATIC uint64_t
1152 xfs_flags2diflags2(
1153 	struct xfs_inode	*ip,
1154 	unsigned int		xflags)
1155 {
1156 	uint64_t		di_flags2 =
1157 		(ip->i_diflags2 & (XFS_DIFLAG2_REFLINK |
1158 				   XFS_DIFLAG2_BIGTIME));
1159 
1160 	if (xflags & FS_XFLAG_DAX)
1161 		di_flags2 |= XFS_DIFLAG2_DAX;
1162 	if (xflags & FS_XFLAG_COWEXTSIZE)
1163 		di_flags2 |= XFS_DIFLAG2_COWEXTSIZE;
1164 
1165 	return di_flags2;
1166 }
1167 
1168 static int
1169 xfs_ioctl_setattr_xflags(
1170 	struct xfs_trans	*tp,
1171 	struct xfs_inode	*ip,
1172 	struct fileattr		*fa)
1173 {
1174 	struct xfs_mount	*mp = ip->i_mount;
1175 	uint64_t		i_flags2;
1176 
1177 	/* Can't change realtime flag if any extents are allocated. */
1178 	if ((ip->i_df.if_nextents || ip->i_delayed_blks) &&
1179 	    XFS_IS_REALTIME_INODE(ip) != (fa->fsx_xflags & FS_XFLAG_REALTIME))
1180 		return -EINVAL;
1181 
1182 	/* If realtime flag is set then must have realtime device */
1183 	if (fa->fsx_xflags & FS_XFLAG_REALTIME) {
1184 		if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
1185 		    (ip->i_extsize % mp->m_sb.sb_rextsize))
1186 			return -EINVAL;
1187 	}
1188 
1189 	/* Clear reflink if we are actually able to set the rt flag. */
1190 	if ((fa->fsx_xflags & FS_XFLAG_REALTIME) && xfs_is_reflink_inode(ip))
1191 		ip->i_diflags2 &= ~XFS_DIFLAG2_REFLINK;
1192 
1193 	/* Don't allow us to set DAX mode for a reflinked file for now. */
1194 	if ((fa->fsx_xflags & FS_XFLAG_DAX) && xfs_is_reflink_inode(ip))
1195 		return -EINVAL;
1196 
1197 	/* diflags2 only valid for v3 inodes. */
1198 	i_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1199 	if (i_flags2 && !xfs_sb_version_has_v3inode(&mp->m_sb))
1200 		return -EINVAL;
1201 
1202 	ip->i_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
1203 	ip->i_diflags2 = i_flags2;
1204 
1205 	xfs_diflags_to_iflags(ip, false);
1206 	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1207 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1208 	XFS_STATS_INC(mp, xs_ig_attrchg);
1209 	return 0;
1210 }
1211 
1212 static void
1213 xfs_ioctl_setattr_prepare_dax(
1214 	struct xfs_inode	*ip,
1215 	struct fileattr		*fa)
1216 {
1217 	struct xfs_mount	*mp = ip->i_mount;
1218 	struct inode            *inode = VFS_I(ip);
1219 
1220 	if (S_ISDIR(inode->i_mode))
1221 		return;
1222 
1223 	if ((mp->m_flags & XFS_MOUNT_DAX_ALWAYS) ||
1224 	    (mp->m_flags & XFS_MOUNT_DAX_NEVER))
1225 		return;
1226 
1227 	if (((fa->fsx_xflags & FS_XFLAG_DAX) &&
1228 	    !(ip->i_diflags2 & XFS_DIFLAG2_DAX)) ||
1229 	    (!(fa->fsx_xflags & FS_XFLAG_DAX) &&
1230 	     (ip->i_diflags2 & XFS_DIFLAG2_DAX)))
1231 		d_mark_dontcache(inode);
1232 }
1233 
1234 /*
1235  * Set up the transaction structure for the setattr operation, checking that we
1236  * have permission to do so. On success, return a clean transaction and the
1237  * inode locked exclusively ready for further operation specific checks. On
1238  * failure, return an error without modifying or locking the inode.
1239  */
1240 static struct xfs_trans *
1241 xfs_ioctl_setattr_get_trans(
1242 	struct xfs_inode	*ip,
1243 	struct xfs_dquot	*pdqp)
1244 {
1245 	struct xfs_mount	*mp = ip->i_mount;
1246 	struct xfs_trans	*tp;
1247 	int			error = -EROFS;
1248 
1249 	if (mp->m_flags & XFS_MOUNT_RDONLY)
1250 		goto out_error;
1251 	error = -EIO;
1252 	if (XFS_FORCED_SHUTDOWN(mp))
1253 		goto out_error;
1254 
1255 	error = xfs_trans_alloc_ichange(ip, NULL, NULL, pdqp,
1256 			capable(CAP_FOWNER), &tp);
1257 	if (error)
1258 		goto out_error;
1259 
1260 	if (mp->m_flags & XFS_MOUNT_WSYNC)
1261 		xfs_trans_set_sync(tp);
1262 
1263 	return tp;
1264 
1265 out_error:
1266 	return ERR_PTR(error);
1267 }
1268 
1269 /*
1270  * Validate a proposed extent size hint.  For regular files, the hint can only
1271  * be changed if no extents are allocated.
1272  */
1273 static int
1274 xfs_ioctl_setattr_check_extsize(
1275 	struct xfs_inode	*ip,
1276 	struct fileattr		*fa)
1277 {
1278 	struct xfs_mount	*mp = ip->i_mount;
1279 	xfs_failaddr_t		failaddr;
1280 	uint16_t		new_diflags;
1281 
1282 	if (!fa->fsx_valid)
1283 		return 0;
1284 
1285 	if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_df.if_nextents &&
1286 	    XFS_FSB_TO_B(mp, ip->i_extsize) != fa->fsx_extsize)
1287 		return -EINVAL;
1288 
1289 	if (fa->fsx_extsize & mp->m_blockmask)
1290 		return -EINVAL;
1291 
1292 	new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
1293 
1294 	/*
1295 	 * Inode verifiers on older kernels don't check that the extent size
1296 	 * hint is an integer multiple of the rt extent size on a directory
1297 	 * with both rtinherit and extszinherit flags set.  Don't let sysadmins
1298 	 * misconfigure directories.
1299 	 */
1300 	if ((new_diflags & XFS_DIFLAG_RTINHERIT) &&
1301 	    (new_diflags & XFS_DIFLAG_EXTSZINHERIT)) {
1302 		unsigned int	rtextsize_bytes;
1303 
1304 		rtextsize_bytes = XFS_FSB_TO_B(mp, mp->m_sb.sb_rextsize);
1305 		if (fa->fsx_extsize % rtextsize_bytes)
1306 			return -EINVAL;
1307 	}
1308 
1309 	failaddr = xfs_inode_validate_extsize(ip->i_mount,
1310 			XFS_B_TO_FSB(mp, fa->fsx_extsize),
1311 			VFS_I(ip)->i_mode, new_diflags);
1312 	return failaddr != NULL ? -EINVAL : 0;
1313 }
1314 
1315 static int
1316 xfs_ioctl_setattr_check_cowextsize(
1317 	struct xfs_inode	*ip,
1318 	struct fileattr		*fa)
1319 {
1320 	struct xfs_mount	*mp = ip->i_mount;
1321 	xfs_failaddr_t		failaddr;
1322 	uint64_t		new_diflags2;
1323 	uint16_t		new_diflags;
1324 
1325 	if (!fa->fsx_valid)
1326 		return 0;
1327 
1328 	if (fa->fsx_cowextsize & mp->m_blockmask)
1329 		return -EINVAL;
1330 
1331 	new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
1332 	new_diflags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1333 
1334 	failaddr = xfs_inode_validate_cowextsize(ip->i_mount,
1335 			XFS_B_TO_FSB(mp, fa->fsx_cowextsize),
1336 			VFS_I(ip)->i_mode, new_diflags, new_diflags2);
1337 	return failaddr != NULL ? -EINVAL : 0;
1338 }
1339 
1340 static int
1341 xfs_ioctl_setattr_check_projid(
1342 	struct xfs_inode	*ip,
1343 	struct fileattr		*fa)
1344 {
1345 	if (!fa->fsx_valid)
1346 		return 0;
1347 
1348 	/* Disallow 32bit project ids if projid32bit feature is not enabled. */
1349 	if (fa->fsx_projid > (uint16_t)-1 &&
1350 	    !xfs_sb_version_hasprojid32bit(&ip->i_mount->m_sb))
1351 		return -EINVAL;
1352 	return 0;
1353 }
1354 
1355 int
1356 xfs_fileattr_set(
1357 	struct user_namespace	*mnt_userns,
1358 	struct dentry		*dentry,
1359 	struct fileattr		*fa)
1360 {
1361 	struct xfs_inode	*ip = XFS_I(d_inode(dentry));
1362 	struct xfs_mount	*mp = ip->i_mount;
1363 	struct xfs_trans	*tp;
1364 	struct xfs_dquot	*pdqp = NULL;
1365 	struct xfs_dquot	*olddquot = NULL;
1366 	int			error;
1367 
1368 	trace_xfs_ioctl_setattr(ip);
1369 
1370 	if (d_is_special(dentry))
1371 		return -ENOTTY;
1372 
1373 	if (!fa->fsx_valid) {
1374 		if (fa->flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL |
1375 				  FS_NOATIME_FL | FS_NODUMP_FL |
1376 				  FS_SYNC_FL | FS_DAX_FL | FS_PROJINHERIT_FL))
1377 			return -EOPNOTSUPP;
1378 	}
1379 
1380 	error = xfs_ioctl_setattr_check_projid(ip, fa);
1381 	if (error)
1382 		return error;
1383 
1384 	/*
1385 	 * If disk quotas is on, we make sure that the dquots do exist on disk,
1386 	 * before we start any other transactions. Trying to do this later
1387 	 * is messy. We don't care to take a readlock to look at the ids
1388 	 * in inode here, because we can't hold it across the trans_reserve.
1389 	 * If the IDs do change before we take the ilock, we're covered
1390 	 * because the i_*dquot fields will get updated anyway.
1391 	 */
1392 	if (fa->fsx_valid && XFS_IS_QUOTA_ON(mp)) {
1393 		error = xfs_qm_vop_dqalloc(ip, VFS_I(ip)->i_uid,
1394 				VFS_I(ip)->i_gid, fa->fsx_projid,
1395 				XFS_QMOPT_PQUOTA, NULL, NULL, &pdqp);
1396 		if (error)
1397 			return error;
1398 	}
1399 
1400 	xfs_ioctl_setattr_prepare_dax(ip, fa);
1401 
1402 	tp = xfs_ioctl_setattr_get_trans(ip, pdqp);
1403 	if (IS_ERR(tp)) {
1404 		error = PTR_ERR(tp);
1405 		goto error_free_dquots;
1406 	}
1407 
1408 	error = xfs_ioctl_setattr_check_extsize(ip, fa);
1409 	if (error)
1410 		goto error_trans_cancel;
1411 
1412 	error = xfs_ioctl_setattr_check_cowextsize(ip, fa);
1413 	if (error)
1414 		goto error_trans_cancel;
1415 
1416 	error = xfs_ioctl_setattr_xflags(tp, ip, fa);
1417 	if (error)
1418 		goto error_trans_cancel;
1419 
1420 	if (!fa->fsx_valid)
1421 		goto skip_xattr;
1422 	/*
1423 	 * Change file ownership.  Must be the owner or privileged.  CAP_FSETID
1424 	 * overrides the following restrictions:
1425 	 *
1426 	 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1427 	 * successful return from chown()
1428 	 */
1429 
1430 	if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
1431 	    !capable_wrt_inode_uidgid(mnt_userns, VFS_I(ip), CAP_FSETID))
1432 		VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
1433 
1434 	/* Change the ownerships and register project quota modifications */
1435 	if (ip->i_projid != fa->fsx_projid) {
1436 		if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp)) {
1437 			olddquot = xfs_qm_vop_chown(tp, ip,
1438 						&ip->i_pdquot, pdqp);
1439 		}
1440 		ip->i_projid = fa->fsx_projid;
1441 	}
1442 
1443 	/*
1444 	 * Only set the extent size hint if we've already determined that the
1445 	 * extent size hint should be set on the inode. If no extent size flags
1446 	 * are set on the inode then unconditionally clear the extent size hint.
1447 	 */
1448 	if (ip->i_diflags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1449 		ip->i_extsize = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1450 	else
1451 		ip->i_extsize = 0;
1452 
1453 	if (xfs_sb_version_has_v3inode(&mp->m_sb)) {
1454 		if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
1455 			ip->i_cowextsize = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
1456 		else
1457 			ip->i_cowextsize = 0;
1458 	}
1459 
1460 skip_xattr:
1461 	error = xfs_trans_commit(tp);
1462 
1463 	/*
1464 	 * Release any dquot(s) the inode had kept before chown.
1465 	 */
1466 	xfs_qm_dqrele(olddquot);
1467 	xfs_qm_dqrele(pdqp);
1468 
1469 	return error;
1470 
1471 error_trans_cancel:
1472 	xfs_trans_cancel(tp);
1473 error_free_dquots:
1474 	xfs_qm_dqrele(pdqp);
1475 	return error;
1476 }
1477 
1478 static bool
1479 xfs_getbmap_format(
1480 	struct kgetbmap		*p,
1481 	struct getbmapx __user	*u,
1482 	size_t			recsize)
1483 {
1484 	if (put_user(p->bmv_offset, &u->bmv_offset) ||
1485 	    put_user(p->bmv_block, &u->bmv_block) ||
1486 	    put_user(p->bmv_length, &u->bmv_length) ||
1487 	    put_user(0, &u->bmv_count) ||
1488 	    put_user(0, &u->bmv_entries))
1489 		return false;
1490 	if (recsize < sizeof(struct getbmapx))
1491 		return true;
1492 	if (put_user(0, &u->bmv_iflags) ||
1493 	    put_user(p->bmv_oflags, &u->bmv_oflags) ||
1494 	    put_user(0, &u->bmv_unused1) ||
1495 	    put_user(0, &u->bmv_unused2))
1496 		return false;
1497 	return true;
1498 }
1499 
1500 STATIC int
1501 xfs_ioc_getbmap(
1502 	struct file		*file,
1503 	unsigned int		cmd,
1504 	void			__user *arg)
1505 {
1506 	struct getbmapx		bmx = { 0 };
1507 	struct kgetbmap		*buf;
1508 	size_t			recsize;
1509 	int			error, i;
1510 
1511 	switch (cmd) {
1512 	case XFS_IOC_GETBMAPA:
1513 		bmx.bmv_iflags = BMV_IF_ATTRFORK;
1514 		/*FALLTHRU*/
1515 	case XFS_IOC_GETBMAP:
1516 		/* struct getbmap is a strict subset of struct getbmapx. */
1517 		recsize = sizeof(struct getbmap);
1518 		break;
1519 	case XFS_IOC_GETBMAPX:
1520 		recsize = sizeof(struct getbmapx);
1521 		break;
1522 	default:
1523 		return -EINVAL;
1524 	}
1525 
1526 	if (copy_from_user(&bmx, arg, recsize))
1527 		return -EFAULT;
1528 
1529 	if (bmx.bmv_count < 2)
1530 		return -EINVAL;
1531 	if (bmx.bmv_count > ULONG_MAX / recsize)
1532 		return -ENOMEM;
1533 
1534 	buf = kvzalloc(bmx.bmv_count * sizeof(*buf), GFP_KERNEL);
1535 	if (!buf)
1536 		return -ENOMEM;
1537 
1538 	error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, buf);
1539 	if (error)
1540 		goto out_free_buf;
1541 
1542 	error = -EFAULT;
1543 	if (copy_to_user(arg, &bmx, recsize))
1544 		goto out_free_buf;
1545 	arg += recsize;
1546 
1547 	for (i = 0; i < bmx.bmv_entries; i++) {
1548 		if (!xfs_getbmap_format(buf + i, arg, recsize))
1549 			goto out_free_buf;
1550 		arg += recsize;
1551 	}
1552 
1553 	error = 0;
1554 out_free_buf:
1555 	kmem_free(buf);
1556 	return error;
1557 }
1558 
1559 STATIC int
1560 xfs_ioc_getfsmap(
1561 	struct xfs_inode	*ip,
1562 	struct fsmap_head	__user *arg)
1563 {
1564 	struct xfs_fsmap_head	xhead = {0};
1565 	struct fsmap_head	head;
1566 	struct fsmap		*recs;
1567 	unsigned int		count;
1568 	__u32			last_flags = 0;
1569 	bool			done = false;
1570 	int			error;
1571 
1572 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
1573 		return -EFAULT;
1574 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
1575 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
1576 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
1577 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
1578 		       sizeof(head.fmh_keys[1].fmr_reserved)))
1579 		return -EINVAL;
1580 
1581 	/*
1582 	 * Use an internal memory buffer so that we don't have to copy fsmap
1583 	 * data to userspace while holding locks.  Start by trying to allocate
1584 	 * up to 128k for the buffer, but fall back to a single page if needed.
1585 	 */
1586 	count = min_t(unsigned int, head.fmh_count,
1587 			131072 / sizeof(struct fsmap));
1588 	recs = kvzalloc(count * sizeof(struct fsmap), GFP_KERNEL);
1589 	if (!recs) {
1590 		count = min_t(unsigned int, head.fmh_count,
1591 				PAGE_SIZE / sizeof(struct fsmap));
1592 		recs = kvzalloc(count * sizeof(struct fsmap), GFP_KERNEL);
1593 		if (!recs)
1594 			return -ENOMEM;
1595 	}
1596 
1597 	xhead.fmh_iflags = head.fmh_iflags;
1598 	xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
1599 	xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
1600 
1601 	trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1602 	trace_xfs_getfsmap_high_key(ip->i_mount, &xhead.fmh_keys[1]);
1603 
1604 	head.fmh_entries = 0;
1605 	do {
1606 		struct fsmap __user	*user_recs;
1607 		struct fsmap		*last_rec;
1608 
1609 		user_recs = &arg->fmh_recs[head.fmh_entries];
1610 		xhead.fmh_entries = 0;
1611 		xhead.fmh_count = min_t(unsigned int, count,
1612 					head.fmh_count - head.fmh_entries);
1613 
1614 		/* Run query, record how many entries we got. */
1615 		error = xfs_getfsmap(ip->i_mount, &xhead, recs);
1616 		switch (error) {
1617 		case 0:
1618 			/*
1619 			 * There are no more records in the result set.  Copy
1620 			 * whatever we got to userspace and break out.
1621 			 */
1622 			done = true;
1623 			break;
1624 		case -ECANCELED:
1625 			/*
1626 			 * The internal memory buffer is full.  Copy whatever
1627 			 * records we got to userspace and go again if we have
1628 			 * not yet filled the userspace buffer.
1629 			 */
1630 			error = 0;
1631 			break;
1632 		default:
1633 			goto out_free;
1634 		}
1635 		head.fmh_entries += xhead.fmh_entries;
1636 		head.fmh_oflags = xhead.fmh_oflags;
1637 
1638 		/*
1639 		 * If the caller wanted a record count or there aren't any
1640 		 * new records to return, we're done.
1641 		 */
1642 		if (head.fmh_count == 0 || xhead.fmh_entries == 0)
1643 			break;
1644 
1645 		/* Copy all the records we got out to userspace. */
1646 		if (copy_to_user(user_recs, recs,
1647 				 xhead.fmh_entries * sizeof(struct fsmap))) {
1648 			error = -EFAULT;
1649 			goto out_free;
1650 		}
1651 
1652 		/* Remember the last record flags we copied to userspace. */
1653 		last_rec = &recs[xhead.fmh_entries - 1];
1654 		last_flags = last_rec->fmr_flags;
1655 
1656 		/* Set up the low key for the next iteration. */
1657 		xfs_fsmap_to_internal(&xhead.fmh_keys[0], last_rec);
1658 		trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1659 	} while (!done && head.fmh_entries < head.fmh_count);
1660 
1661 	/*
1662 	 * If there are no more records in the query result set and we're not
1663 	 * in counting mode, mark the last record returned with the LAST flag.
1664 	 */
1665 	if (done && head.fmh_count > 0 && head.fmh_entries > 0) {
1666 		struct fsmap __user	*user_rec;
1667 
1668 		last_flags |= FMR_OF_LAST;
1669 		user_rec = &arg->fmh_recs[head.fmh_entries - 1];
1670 
1671 		if (copy_to_user(&user_rec->fmr_flags, &last_flags,
1672 					sizeof(last_flags))) {
1673 			error = -EFAULT;
1674 			goto out_free;
1675 		}
1676 	}
1677 
1678 	/* copy back header */
1679 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head))) {
1680 		error = -EFAULT;
1681 		goto out_free;
1682 	}
1683 
1684 out_free:
1685 	kmem_free(recs);
1686 	return error;
1687 }
1688 
1689 STATIC int
1690 xfs_ioc_scrub_metadata(
1691 	struct file			*file,
1692 	void				__user *arg)
1693 {
1694 	struct xfs_scrub_metadata	scrub;
1695 	int				error;
1696 
1697 	if (!capable(CAP_SYS_ADMIN))
1698 		return -EPERM;
1699 
1700 	if (copy_from_user(&scrub, arg, sizeof(scrub)))
1701 		return -EFAULT;
1702 
1703 	error = xfs_scrub_metadata(file, &scrub);
1704 	if (error)
1705 		return error;
1706 
1707 	if (copy_to_user(arg, &scrub, sizeof(scrub)))
1708 		return -EFAULT;
1709 
1710 	return 0;
1711 }
1712 
1713 int
1714 xfs_ioc_swapext(
1715 	xfs_swapext_t	*sxp)
1716 {
1717 	xfs_inode_t     *ip, *tip;
1718 	struct fd	f, tmp;
1719 	int		error = 0;
1720 
1721 	/* Pull information for the target fd */
1722 	f = fdget((int)sxp->sx_fdtarget);
1723 	if (!f.file) {
1724 		error = -EINVAL;
1725 		goto out;
1726 	}
1727 
1728 	if (!(f.file->f_mode & FMODE_WRITE) ||
1729 	    !(f.file->f_mode & FMODE_READ) ||
1730 	    (f.file->f_flags & O_APPEND)) {
1731 		error = -EBADF;
1732 		goto out_put_file;
1733 	}
1734 
1735 	tmp = fdget((int)sxp->sx_fdtmp);
1736 	if (!tmp.file) {
1737 		error = -EINVAL;
1738 		goto out_put_file;
1739 	}
1740 
1741 	if (!(tmp.file->f_mode & FMODE_WRITE) ||
1742 	    !(tmp.file->f_mode & FMODE_READ) ||
1743 	    (tmp.file->f_flags & O_APPEND)) {
1744 		error = -EBADF;
1745 		goto out_put_tmp_file;
1746 	}
1747 
1748 	if (IS_SWAPFILE(file_inode(f.file)) ||
1749 	    IS_SWAPFILE(file_inode(tmp.file))) {
1750 		error = -EINVAL;
1751 		goto out_put_tmp_file;
1752 	}
1753 
1754 	/*
1755 	 * We need to ensure that the fds passed in point to XFS inodes
1756 	 * before we cast and access them as XFS structures as we have no
1757 	 * control over what the user passes us here.
1758 	 */
1759 	if (f.file->f_op != &xfs_file_operations ||
1760 	    tmp.file->f_op != &xfs_file_operations) {
1761 		error = -EINVAL;
1762 		goto out_put_tmp_file;
1763 	}
1764 
1765 	ip = XFS_I(file_inode(f.file));
1766 	tip = XFS_I(file_inode(tmp.file));
1767 
1768 	if (ip->i_mount != tip->i_mount) {
1769 		error = -EINVAL;
1770 		goto out_put_tmp_file;
1771 	}
1772 
1773 	if (ip->i_ino == tip->i_ino) {
1774 		error = -EINVAL;
1775 		goto out_put_tmp_file;
1776 	}
1777 
1778 	if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
1779 		error = -EIO;
1780 		goto out_put_tmp_file;
1781 	}
1782 
1783 	error = xfs_swap_extents(ip, tip, sxp);
1784 
1785  out_put_tmp_file:
1786 	fdput(tmp);
1787  out_put_file:
1788 	fdput(f);
1789  out:
1790 	return error;
1791 }
1792 
1793 static int
1794 xfs_ioc_getlabel(
1795 	struct xfs_mount	*mp,
1796 	char			__user *user_label)
1797 {
1798 	struct xfs_sb		*sbp = &mp->m_sb;
1799 	char			label[XFSLABEL_MAX + 1];
1800 
1801 	/* Paranoia */
1802 	BUILD_BUG_ON(sizeof(sbp->sb_fname) > FSLABEL_MAX);
1803 
1804 	/* 1 larger than sb_fname, so this ensures a trailing NUL char */
1805 	memset(label, 0, sizeof(label));
1806 	spin_lock(&mp->m_sb_lock);
1807 	strncpy(label, sbp->sb_fname, XFSLABEL_MAX);
1808 	spin_unlock(&mp->m_sb_lock);
1809 
1810 	if (copy_to_user(user_label, label, sizeof(label)))
1811 		return -EFAULT;
1812 	return 0;
1813 }
1814 
1815 static int
1816 xfs_ioc_setlabel(
1817 	struct file		*filp,
1818 	struct xfs_mount	*mp,
1819 	char			__user *newlabel)
1820 {
1821 	struct xfs_sb		*sbp = &mp->m_sb;
1822 	char			label[XFSLABEL_MAX + 1];
1823 	size_t			len;
1824 	int			error;
1825 
1826 	if (!capable(CAP_SYS_ADMIN))
1827 		return -EPERM;
1828 	/*
1829 	 * The generic ioctl allows up to FSLABEL_MAX chars, but XFS is much
1830 	 * smaller, at 12 bytes.  We copy one more to be sure we find the
1831 	 * (required) NULL character to test the incoming label length.
1832 	 * NB: The on disk label doesn't need to be null terminated.
1833 	 */
1834 	if (copy_from_user(label, newlabel, XFSLABEL_MAX + 1))
1835 		return -EFAULT;
1836 	len = strnlen(label, XFSLABEL_MAX + 1);
1837 	if (len > sizeof(sbp->sb_fname))
1838 		return -EINVAL;
1839 
1840 	error = mnt_want_write_file(filp);
1841 	if (error)
1842 		return error;
1843 
1844 	spin_lock(&mp->m_sb_lock);
1845 	memset(sbp->sb_fname, 0, sizeof(sbp->sb_fname));
1846 	memcpy(sbp->sb_fname, label, len);
1847 	spin_unlock(&mp->m_sb_lock);
1848 
1849 	/*
1850 	 * Now we do several things to satisfy userspace.
1851 	 * In addition to normal logging of the primary superblock, we also
1852 	 * immediately write these changes to sector zero for the primary, then
1853 	 * update all backup supers (as xfs_db does for a label change), then
1854 	 * invalidate the block device page cache.  This is so that any prior
1855 	 * buffered reads from userspace (i.e. from blkid) are invalidated,
1856 	 * and userspace will see the newly-written label.
1857 	 */
1858 	error = xfs_sync_sb_buf(mp);
1859 	if (error)
1860 		goto out;
1861 	/*
1862 	 * growfs also updates backup supers so lock against that.
1863 	 */
1864 	mutex_lock(&mp->m_growlock);
1865 	error = xfs_update_secondary_sbs(mp);
1866 	mutex_unlock(&mp->m_growlock);
1867 
1868 	invalidate_bdev(mp->m_ddev_targp->bt_bdev);
1869 
1870 out:
1871 	mnt_drop_write_file(filp);
1872 	return error;
1873 }
1874 
1875 static inline int
1876 xfs_fs_eofblocks_from_user(
1877 	struct xfs_fs_eofblocks		*src,
1878 	struct xfs_eofblocks		*dst)
1879 {
1880 	if (src->eof_version != XFS_EOFBLOCKS_VERSION)
1881 		return -EINVAL;
1882 
1883 	if (src->eof_flags & ~XFS_EOF_FLAGS_VALID)
1884 		return -EINVAL;
1885 
1886 	if (memchr_inv(&src->pad32, 0, sizeof(src->pad32)) ||
1887 	    memchr_inv(src->pad64, 0, sizeof(src->pad64)))
1888 		return -EINVAL;
1889 
1890 	dst->eof_flags = src->eof_flags;
1891 	dst->eof_prid = src->eof_prid;
1892 	dst->eof_min_file_size = src->eof_min_file_size;
1893 
1894 	dst->eof_uid = INVALID_UID;
1895 	if (src->eof_flags & XFS_EOF_FLAGS_UID) {
1896 		dst->eof_uid = make_kuid(current_user_ns(), src->eof_uid);
1897 		if (!uid_valid(dst->eof_uid))
1898 			return -EINVAL;
1899 	}
1900 
1901 	dst->eof_gid = INVALID_GID;
1902 	if (src->eof_flags & XFS_EOF_FLAGS_GID) {
1903 		dst->eof_gid = make_kgid(current_user_ns(), src->eof_gid);
1904 		if (!gid_valid(dst->eof_gid))
1905 			return -EINVAL;
1906 	}
1907 	return 0;
1908 }
1909 
1910 /*
1911  * Note: some of the ioctl's return positive numbers as a
1912  * byte count indicating success, such as readlink_by_handle.
1913  * So we don't "sign flip" like most other routines.  This means
1914  * true errors need to be returned as a negative value.
1915  */
1916 long
1917 xfs_file_ioctl(
1918 	struct file		*filp,
1919 	unsigned int		cmd,
1920 	unsigned long		p)
1921 {
1922 	struct inode		*inode = file_inode(filp);
1923 	struct xfs_inode	*ip = XFS_I(inode);
1924 	struct xfs_mount	*mp = ip->i_mount;
1925 	void			__user *arg = (void __user *)p;
1926 	int			error;
1927 
1928 	trace_xfs_file_ioctl(ip);
1929 
1930 	switch (cmd) {
1931 	case FITRIM:
1932 		return xfs_ioc_trim(mp, arg);
1933 	case FS_IOC_GETFSLABEL:
1934 		return xfs_ioc_getlabel(mp, arg);
1935 	case FS_IOC_SETFSLABEL:
1936 		return xfs_ioc_setlabel(filp, mp, arg);
1937 	case XFS_IOC_ALLOCSP:
1938 	case XFS_IOC_FREESP:
1939 	case XFS_IOC_ALLOCSP64:
1940 	case XFS_IOC_FREESP64: {
1941 		xfs_flock64_t		bf;
1942 
1943 		if (copy_from_user(&bf, arg, sizeof(bf)))
1944 			return -EFAULT;
1945 		return xfs_ioc_space(filp, &bf);
1946 	}
1947 	case XFS_IOC_DIOINFO: {
1948 		struct xfs_buftarg	*target = xfs_inode_buftarg(ip);
1949 		struct dioattr		da;
1950 
1951 		da.d_mem =  da.d_miniosz = target->bt_logical_sectorsize;
1952 		da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1953 
1954 		if (copy_to_user(arg, &da, sizeof(da)))
1955 			return -EFAULT;
1956 		return 0;
1957 	}
1958 
1959 	case XFS_IOC_FSBULKSTAT_SINGLE:
1960 	case XFS_IOC_FSBULKSTAT:
1961 	case XFS_IOC_FSINUMBERS:
1962 		return xfs_ioc_fsbulkstat(filp, cmd, arg);
1963 
1964 	case XFS_IOC_BULKSTAT:
1965 		return xfs_ioc_bulkstat(filp, cmd, arg);
1966 	case XFS_IOC_INUMBERS:
1967 		return xfs_ioc_inumbers(mp, cmd, arg);
1968 
1969 	case XFS_IOC_FSGEOMETRY_V1:
1970 		return xfs_ioc_fsgeometry(mp, arg, 3);
1971 	case XFS_IOC_FSGEOMETRY_V4:
1972 		return xfs_ioc_fsgeometry(mp, arg, 4);
1973 	case XFS_IOC_FSGEOMETRY:
1974 		return xfs_ioc_fsgeometry(mp, arg, 5);
1975 
1976 	case XFS_IOC_AG_GEOMETRY:
1977 		return xfs_ioc_ag_geometry(mp, arg);
1978 
1979 	case XFS_IOC_GETVERSION:
1980 		return put_user(inode->i_generation, (int __user *)arg);
1981 
1982 	case XFS_IOC_FSGETXATTRA:
1983 		return xfs_ioc_fsgetxattra(ip, arg);
1984 
1985 	case XFS_IOC_GETBMAP:
1986 	case XFS_IOC_GETBMAPA:
1987 	case XFS_IOC_GETBMAPX:
1988 		return xfs_ioc_getbmap(filp, cmd, arg);
1989 
1990 	case FS_IOC_GETFSMAP:
1991 		return xfs_ioc_getfsmap(ip, arg);
1992 
1993 	case XFS_IOC_SCRUB_METADATA:
1994 		return xfs_ioc_scrub_metadata(filp, arg);
1995 
1996 	case XFS_IOC_FD_TO_HANDLE:
1997 	case XFS_IOC_PATH_TO_HANDLE:
1998 	case XFS_IOC_PATH_TO_FSHANDLE: {
1999 		xfs_fsop_handlereq_t	hreq;
2000 
2001 		if (copy_from_user(&hreq, arg, sizeof(hreq)))
2002 			return -EFAULT;
2003 		return xfs_find_handle(cmd, &hreq);
2004 	}
2005 	case XFS_IOC_OPEN_BY_HANDLE: {
2006 		xfs_fsop_handlereq_t	hreq;
2007 
2008 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
2009 			return -EFAULT;
2010 		return xfs_open_by_handle(filp, &hreq);
2011 	}
2012 
2013 	case XFS_IOC_READLINK_BY_HANDLE: {
2014 		xfs_fsop_handlereq_t	hreq;
2015 
2016 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
2017 			return -EFAULT;
2018 		return xfs_readlink_by_handle(filp, &hreq);
2019 	}
2020 	case XFS_IOC_ATTRLIST_BY_HANDLE:
2021 		return xfs_attrlist_by_handle(filp, arg);
2022 
2023 	case XFS_IOC_ATTRMULTI_BY_HANDLE:
2024 		return xfs_attrmulti_by_handle(filp, arg);
2025 
2026 	case XFS_IOC_SWAPEXT: {
2027 		struct xfs_swapext	sxp;
2028 
2029 		if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
2030 			return -EFAULT;
2031 		error = mnt_want_write_file(filp);
2032 		if (error)
2033 			return error;
2034 		error = xfs_ioc_swapext(&sxp);
2035 		mnt_drop_write_file(filp);
2036 		return error;
2037 	}
2038 
2039 	case XFS_IOC_FSCOUNTS: {
2040 		xfs_fsop_counts_t out;
2041 
2042 		xfs_fs_counts(mp, &out);
2043 
2044 		if (copy_to_user(arg, &out, sizeof(out)))
2045 			return -EFAULT;
2046 		return 0;
2047 	}
2048 
2049 	case XFS_IOC_SET_RESBLKS: {
2050 		xfs_fsop_resblks_t inout;
2051 		uint64_t	   in;
2052 
2053 		if (!capable(CAP_SYS_ADMIN))
2054 			return -EPERM;
2055 
2056 		if (mp->m_flags & XFS_MOUNT_RDONLY)
2057 			return -EROFS;
2058 
2059 		if (copy_from_user(&inout, arg, sizeof(inout)))
2060 			return -EFAULT;
2061 
2062 		error = mnt_want_write_file(filp);
2063 		if (error)
2064 			return error;
2065 
2066 		/* input parameter is passed in resblks field of structure */
2067 		in = inout.resblks;
2068 		error = xfs_reserve_blocks(mp, &in, &inout);
2069 		mnt_drop_write_file(filp);
2070 		if (error)
2071 			return error;
2072 
2073 		if (copy_to_user(arg, &inout, sizeof(inout)))
2074 			return -EFAULT;
2075 		return 0;
2076 	}
2077 
2078 	case XFS_IOC_GET_RESBLKS: {
2079 		xfs_fsop_resblks_t out;
2080 
2081 		if (!capable(CAP_SYS_ADMIN))
2082 			return -EPERM;
2083 
2084 		error = xfs_reserve_blocks(mp, NULL, &out);
2085 		if (error)
2086 			return error;
2087 
2088 		if (copy_to_user(arg, &out, sizeof(out)))
2089 			return -EFAULT;
2090 
2091 		return 0;
2092 	}
2093 
2094 	case XFS_IOC_FSGROWFSDATA: {
2095 		struct xfs_growfs_data in;
2096 
2097 		if (copy_from_user(&in, arg, sizeof(in)))
2098 			return -EFAULT;
2099 
2100 		error = mnt_want_write_file(filp);
2101 		if (error)
2102 			return error;
2103 		error = xfs_growfs_data(mp, &in);
2104 		mnt_drop_write_file(filp);
2105 		return error;
2106 	}
2107 
2108 	case XFS_IOC_FSGROWFSLOG: {
2109 		struct xfs_growfs_log in;
2110 
2111 		if (copy_from_user(&in, arg, sizeof(in)))
2112 			return -EFAULT;
2113 
2114 		error = mnt_want_write_file(filp);
2115 		if (error)
2116 			return error;
2117 		error = xfs_growfs_log(mp, &in);
2118 		mnt_drop_write_file(filp);
2119 		return error;
2120 	}
2121 
2122 	case XFS_IOC_FSGROWFSRT: {
2123 		xfs_growfs_rt_t in;
2124 
2125 		if (copy_from_user(&in, arg, sizeof(in)))
2126 			return -EFAULT;
2127 
2128 		error = mnt_want_write_file(filp);
2129 		if (error)
2130 			return error;
2131 		error = xfs_growfs_rt(mp, &in);
2132 		mnt_drop_write_file(filp);
2133 		return error;
2134 	}
2135 
2136 	case XFS_IOC_GOINGDOWN: {
2137 		uint32_t in;
2138 
2139 		if (!capable(CAP_SYS_ADMIN))
2140 			return -EPERM;
2141 
2142 		if (get_user(in, (uint32_t __user *)arg))
2143 			return -EFAULT;
2144 
2145 		return xfs_fs_goingdown(mp, in);
2146 	}
2147 
2148 	case XFS_IOC_ERROR_INJECTION: {
2149 		xfs_error_injection_t in;
2150 
2151 		if (!capable(CAP_SYS_ADMIN))
2152 			return -EPERM;
2153 
2154 		if (copy_from_user(&in, arg, sizeof(in)))
2155 			return -EFAULT;
2156 
2157 		return xfs_errortag_add(mp, in.errtag);
2158 	}
2159 
2160 	case XFS_IOC_ERROR_CLEARALL:
2161 		if (!capable(CAP_SYS_ADMIN))
2162 			return -EPERM;
2163 
2164 		return xfs_errortag_clearall(mp);
2165 
2166 	case XFS_IOC_FREE_EOFBLOCKS: {
2167 		struct xfs_fs_eofblocks eofb;
2168 		struct xfs_eofblocks keofb;
2169 
2170 		if (!capable(CAP_SYS_ADMIN))
2171 			return -EPERM;
2172 
2173 		if (mp->m_flags & XFS_MOUNT_RDONLY)
2174 			return -EROFS;
2175 
2176 		if (copy_from_user(&eofb, arg, sizeof(eofb)))
2177 			return -EFAULT;
2178 
2179 		error = xfs_fs_eofblocks_from_user(&eofb, &keofb);
2180 		if (error)
2181 			return error;
2182 
2183 		trace_xfs_ioc_free_eofblocks(mp, &keofb, _RET_IP_);
2184 
2185 		sb_start_write(mp->m_super);
2186 		error = xfs_blockgc_free_space(mp, &keofb);
2187 		sb_end_write(mp->m_super);
2188 		return error;
2189 	}
2190 
2191 	default:
2192 		return -ENOTTY;
2193 	}
2194 }
2195