xref: /linux/fs/xfs/xfs_ioctl.c (revision cffaefd15a8f423cdee5d8eac15d267bc92de314)
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 #include "xfs_rtbitmap.h"
42 
43 #include <linux/mount.h>
44 #include <linux/namei.h>
45 #include <linux/fileattr.h>
46 
47 /*
48  * xfs_find_handle maps from userspace xfs_fsop_handlereq structure to
49  * a file or fs handle.
50  *
51  * XFS_IOC_PATH_TO_FSHANDLE
52  *    returns fs handle for a mount point or path within that mount point
53  * XFS_IOC_FD_TO_HANDLE
54  *    returns full handle for a FD opened in user space
55  * XFS_IOC_PATH_TO_HANDLE
56  *    returns full handle for a path
57  */
58 int
59 xfs_find_handle(
60 	unsigned int		cmd,
61 	xfs_fsop_handlereq_t	*hreq)
62 {
63 	int			hsize;
64 	xfs_handle_t		handle;
65 	struct inode		*inode;
66 	struct fd		f = {NULL};
67 	struct path		path;
68 	int			error;
69 	struct xfs_inode	*ip;
70 
71 	if (cmd == XFS_IOC_FD_TO_HANDLE) {
72 		f = fdget(hreq->fd);
73 		if (!f.file)
74 			return -EBADF;
75 		inode = file_inode(f.file);
76 	} else {
77 		error = user_path_at(AT_FDCWD, hreq->path, 0, &path);
78 		if (error)
79 			return error;
80 		inode = d_inode(path.dentry);
81 	}
82 	ip = XFS_I(inode);
83 
84 	/*
85 	 * We can only generate handles for inodes residing on a XFS filesystem,
86 	 * and only for regular files, directories or symbolic links.
87 	 */
88 	error = -EINVAL;
89 	if (inode->i_sb->s_magic != XFS_SB_MAGIC)
90 		goto out_put;
91 
92 	error = -EBADF;
93 	if (!S_ISREG(inode->i_mode) &&
94 	    !S_ISDIR(inode->i_mode) &&
95 	    !S_ISLNK(inode->i_mode))
96 		goto out_put;
97 
98 
99 	memcpy(&handle.ha_fsid, ip->i_mount->m_fixedfsid, sizeof(xfs_fsid_t));
100 
101 	if (cmd == XFS_IOC_PATH_TO_FSHANDLE) {
102 		/*
103 		 * This handle only contains an fsid, zero the rest.
104 		 */
105 		memset(&handle.ha_fid, 0, sizeof(handle.ha_fid));
106 		hsize = sizeof(xfs_fsid_t);
107 	} else {
108 		handle.ha_fid.fid_len = sizeof(xfs_fid_t) -
109 					sizeof(handle.ha_fid.fid_len);
110 		handle.ha_fid.fid_pad = 0;
111 		handle.ha_fid.fid_gen = inode->i_generation;
112 		handle.ha_fid.fid_ino = ip->i_ino;
113 		hsize = sizeof(xfs_handle_t);
114 	}
115 
116 	error = -EFAULT;
117 	if (copy_to_user(hreq->ohandle, &handle, hsize) ||
118 	    copy_to_user(hreq->ohandlen, &hsize, sizeof(__s32)))
119 		goto out_put;
120 
121 	error = 0;
122 
123  out_put:
124 	if (cmd == XFS_IOC_FD_TO_HANDLE)
125 		fdput(f);
126 	else
127 		path_put(&path);
128 	return error;
129 }
130 
131 /*
132  * No need to do permission checks on the various pathname components
133  * as the handle operations are privileged.
134  */
135 STATIC int
136 xfs_handle_acceptable(
137 	void			*context,
138 	struct dentry		*dentry)
139 {
140 	return 1;
141 }
142 
143 /*
144  * Convert userspace handle data into a dentry.
145  */
146 struct dentry *
147 xfs_handle_to_dentry(
148 	struct file		*parfilp,
149 	void __user		*uhandle,
150 	u32			hlen)
151 {
152 	xfs_handle_t		handle;
153 	struct xfs_fid64	fid;
154 
155 	/*
156 	 * Only allow handle opens under a directory.
157 	 */
158 	if (!S_ISDIR(file_inode(parfilp)->i_mode))
159 		return ERR_PTR(-ENOTDIR);
160 
161 	if (hlen != sizeof(xfs_handle_t))
162 		return ERR_PTR(-EINVAL);
163 	if (copy_from_user(&handle, uhandle, hlen))
164 		return ERR_PTR(-EFAULT);
165 	if (handle.ha_fid.fid_len !=
166 	    sizeof(handle.ha_fid) - sizeof(handle.ha_fid.fid_len))
167 		return ERR_PTR(-EINVAL);
168 
169 	memset(&fid, 0, sizeof(struct fid));
170 	fid.ino = handle.ha_fid.fid_ino;
171 	fid.gen = handle.ha_fid.fid_gen;
172 
173 	return exportfs_decode_fh(parfilp->f_path.mnt, (struct fid *)&fid, 3,
174 			FILEID_INO32_GEN | XFS_FILEID_TYPE_64FLAG,
175 			xfs_handle_acceptable, NULL);
176 }
177 
178 STATIC struct dentry *
179 xfs_handlereq_to_dentry(
180 	struct file		*parfilp,
181 	xfs_fsop_handlereq_t	*hreq)
182 {
183 	return xfs_handle_to_dentry(parfilp, hreq->ihandle, hreq->ihandlen);
184 }
185 
186 int
187 xfs_open_by_handle(
188 	struct file		*parfilp,
189 	xfs_fsop_handlereq_t	*hreq)
190 {
191 	const struct cred	*cred = current_cred();
192 	int			error;
193 	int			fd;
194 	int			permflag;
195 	struct file		*filp;
196 	struct inode		*inode;
197 	struct dentry		*dentry;
198 	fmode_t			fmode;
199 	struct path		path;
200 
201 	if (!capable(CAP_SYS_ADMIN))
202 		return -EPERM;
203 
204 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
205 	if (IS_ERR(dentry))
206 		return PTR_ERR(dentry);
207 	inode = d_inode(dentry);
208 
209 	/* Restrict xfs_open_by_handle to directories & regular files. */
210 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))) {
211 		error = -EPERM;
212 		goto out_dput;
213 	}
214 
215 #if BITS_PER_LONG != 32
216 	hreq->oflags |= O_LARGEFILE;
217 #endif
218 
219 	permflag = hreq->oflags;
220 	fmode = OPEN_FMODE(permflag);
221 	if ((!(permflag & O_APPEND) || (permflag & O_TRUNC)) &&
222 	    (fmode & FMODE_WRITE) && IS_APPEND(inode)) {
223 		error = -EPERM;
224 		goto out_dput;
225 	}
226 
227 	if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
228 		error = -EPERM;
229 		goto out_dput;
230 	}
231 
232 	/* Can't write directories. */
233 	if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
234 		error = -EISDIR;
235 		goto out_dput;
236 	}
237 
238 	fd = get_unused_fd_flags(0);
239 	if (fd < 0) {
240 		error = fd;
241 		goto out_dput;
242 	}
243 
244 	path.mnt = parfilp->f_path.mnt;
245 	path.dentry = dentry;
246 	filp = dentry_open(&path, hreq->oflags, cred);
247 	dput(dentry);
248 	if (IS_ERR(filp)) {
249 		put_unused_fd(fd);
250 		return PTR_ERR(filp);
251 	}
252 
253 	if (S_ISREG(inode->i_mode)) {
254 		filp->f_flags |= O_NOATIME;
255 		filp->f_mode |= FMODE_NOCMTIME;
256 	}
257 
258 	fd_install(fd, filp);
259 	return fd;
260 
261  out_dput:
262 	dput(dentry);
263 	return error;
264 }
265 
266 int
267 xfs_readlink_by_handle(
268 	struct file		*parfilp,
269 	xfs_fsop_handlereq_t	*hreq)
270 {
271 	struct dentry		*dentry;
272 	__u32			olen;
273 	int			error;
274 
275 	if (!capable(CAP_SYS_ADMIN))
276 		return -EPERM;
277 
278 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
279 	if (IS_ERR(dentry))
280 		return PTR_ERR(dentry);
281 
282 	/* Restrict this handle operation to symlinks only. */
283 	if (!d_is_symlink(dentry)) {
284 		error = -EINVAL;
285 		goto out_dput;
286 	}
287 
288 	if (copy_from_user(&olen, hreq->ohandlen, sizeof(__u32))) {
289 		error = -EFAULT;
290 		goto out_dput;
291 	}
292 
293 	error = vfs_readlink(dentry, hreq->ohandle, olen);
294 
295  out_dput:
296 	dput(dentry);
297 	return error;
298 }
299 
300 /*
301  * Format an attribute and copy it out to the user's buffer.
302  * Take care to check values and protect against them changing later,
303  * we may be reading them directly out of a user buffer.
304  */
305 static void
306 xfs_ioc_attr_put_listent(
307 	struct xfs_attr_list_context *context,
308 	int			flags,
309 	unsigned char		*name,
310 	int			namelen,
311 	int			valuelen)
312 {
313 	struct xfs_attrlist	*alist = context->buffer;
314 	struct xfs_attrlist_ent	*aep;
315 	int			arraytop;
316 
317 	ASSERT(!context->seen_enough);
318 	ASSERT(context->count >= 0);
319 	ASSERT(context->count < (ATTR_MAX_VALUELEN/8));
320 	ASSERT(context->firstu >= sizeof(*alist));
321 	ASSERT(context->firstu <= context->bufsize);
322 
323 	/*
324 	 * Only list entries in the right namespace.
325 	 */
326 	if (context->attr_filter != (flags & XFS_ATTR_NSP_ONDISK_MASK))
327 		return;
328 
329 	arraytop = sizeof(*alist) +
330 			context->count * sizeof(alist->al_offset[0]);
331 
332 	/* decrement by the actual bytes used by the attr */
333 	context->firstu -= round_up(offsetof(struct xfs_attrlist_ent, a_name) +
334 			namelen + 1, sizeof(uint32_t));
335 	if (context->firstu < arraytop) {
336 		trace_xfs_attr_list_full(context);
337 		alist->al_more = 1;
338 		context->seen_enough = 1;
339 		return;
340 	}
341 
342 	aep = context->buffer + context->firstu;
343 	aep->a_valuelen = valuelen;
344 	memcpy(aep->a_name, name, namelen);
345 	aep->a_name[namelen] = 0;
346 	alist->al_offset[context->count++] = context->firstu;
347 	alist->al_count = context->count;
348 	trace_xfs_attr_list_add(context);
349 }
350 
351 static unsigned int
352 xfs_attr_filter(
353 	u32			ioc_flags)
354 {
355 	if (ioc_flags & XFS_IOC_ATTR_ROOT)
356 		return XFS_ATTR_ROOT;
357 	if (ioc_flags & XFS_IOC_ATTR_SECURE)
358 		return XFS_ATTR_SECURE;
359 	return 0;
360 }
361 
362 static unsigned int
363 xfs_attr_flags(
364 	u32			ioc_flags)
365 {
366 	if (ioc_flags & XFS_IOC_ATTR_CREATE)
367 		return XATTR_CREATE;
368 	if (ioc_flags & XFS_IOC_ATTR_REPLACE)
369 		return XATTR_REPLACE;
370 	return 0;
371 }
372 
373 int
374 xfs_ioc_attr_list(
375 	struct xfs_inode		*dp,
376 	void __user			*ubuf,
377 	size_t				bufsize,
378 	int				flags,
379 	struct xfs_attrlist_cursor __user *ucursor)
380 {
381 	struct xfs_attr_list_context	context = { };
382 	struct xfs_attrlist		*alist;
383 	void				*buffer;
384 	int				error;
385 
386 	if (bufsize < sizeof(struct xfs_attrlist) ||
387 	    bufsize > XFS_XATTR_LIST_MAX)
388 		return -EINVAL;
389 
390 	/*
391 	 * Reject flags, only allow namespaces.
392 	 */
393 	if (flags & ~(XFS_IOC_ATTR_ROOT | XFS_IOC_ATTR_SECURE))
394 		return -EINVAL;
395 	if (flags == (XFS_IOC_ATTR_ROOT | XFS_IOC_ATTR_SECURE))
396 		return -EINVAL;
397 
398 	/*
399 	 * Validate the cursor.
400 	 */
401 	if (copy_from_user(&context.cursor, ucursor, sizeof(context.cursor)))
402 		return -EFAULT;
403 	if (context.cursor.pad1 || context.cursor.pad2)
404 		return -EINVAL;
405 	if (!context.cursor.initted &&
406 	    (context.cursor.hashval || context.cursor.blkno ||
407 	     context.cursor.offset))
408 		return -EINVAL;
409 
410 	buffer = kvzalloc(bufsize, GFP_KERNEL);
411 	if (!buffer)
412 		return -ENOMEM;
413 
414 	/*
415 	 * Initialize the output buffer.
416 	 */
417 	context.dp = dp;
418 	context.resynch = 1;
419 	context.attr_filter = xfs_attr_filter(flags);
420 	context.buffer = buffer;
421 	context.bufsize = round_down(bufsize, sizeof(uint32_t));
422 	context.firstu = context.bufsize;
423 	context.put_listent = xfs_ioc_attr_put_listent;
424 
425 	alist = context.buffer;
426 	alist->al_count = 0;
427 	alist->al_more = 0;
428 	alist->al_offset[0] = context.bufsize;
429 
430 	error = xfs_attr_list(&context);
431 	if (error)
432 		goto out_free;
433 
434 	if (copy_to_user(ubuf, buffer, bufsize) ||
435 	    copy_to_user(ucursor, &context.cursor, sizeof(context.cursor)))
436 		error = -EFAULT;
437 out_free:
438 	kvfree(buffer);
439 	return error;
440 }
441 
442 STATIC int
443 xfs_attrlist_by_handle(
444 	struct file		*parfilp,
445 	struct xfs_fsop_attrlist_handlereq __user *p)
446 {
447 	struct xfs_fsop_attrlist_handlereq al_hreq;
448 	struct dentry		*dentry;
449 	int			error = -ENOMEM;
450 
451 	if (!capable(CAP_SYS_ADMIN))
452 		return -EPERM;
453 	if (copy_from_user(&al_hreq, p, sizeof(al_hreq)))
454 		return -EFAULT;
455 
456 	dentry = xfs_handlereq_to_dentry(parfilp, &al_hreq.hreq);
457 	if (IS_ERR(dentry))
458 		return PTR_ERR(dentry);
459 
460 	error = xfs_ioc_attr_list(XFS_I(d_inode(dentry)), al_hreq.buffer,
461 				  al_hreq.buflen, al_hreq.flags, &p->pos);
462 	dput(dentry);
463 	return error;
464 }
465 
466 static int
467 xfs_attrmulti_attr_get(
468 	struct inode		*inode,
469 	unsigned char		*name,
470 	unsigned char		__user *ubuf,
471 	uint32_t		*len,
472 	uint32_t		flags)
473 {
474 	struct xfs_da_args	args = {
475 		.dp		= XFS_I(inode),
476 		.attr_filter	= xfs_attr_filter(flags),
477 		.attr_flags	= xfs_attr_flags(flags),
478 		.name		= name,
479 		.namelen	= strlen(name),
480 		.valuelen	= *len,
481 	};
482 	int			error;
483 
484 	if (*len > XFS_XATTR_SIZE_MAX)
485 		return -EINVAL;
486 
487 	error = xfs_attr_get(&args);
488 	if (error)
489 		goto out_kfree;
490 
491 	*len = args.valuelen;
492 	if (copy_to_user(ubuf, args.value, args.valuelen))
493 		error = -EFAULT;
494 
495 out_kfree:
496 	kvfree(args.value);
497 	return error;
498 }
499 
500 static int
501 xfs_attrmulti_attr_set(
502 	struct inode		*inode,
503 	unsigned char		*name,
504 	const unsigned char	__user *ubuf,
505 	uint32_t		len,
506 	uint32_t		flags)
507 {
508 	struct xfs_da_args	args = {
509 		.dp		= XFS_I(inode),
510 		.attr_filter	= xfs_attr_filter(flags),
511 		.attr_flags	= xfs_attr_flags(flags),
512 		.name		= name,
513 		.namelen	= strlen(name),
514 	};
515 	int			error;
516 
517 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
518 		return -EPERM;
519 
520 	if (ubuf) {
521 		if (len > XFS_XATTR_SIZE_MAX)
522 			return -EINVAL;
523 		args.value = memdup_user(ubuf, len);
524 		if (IS_ERR(args.value))
525 			return PTR_ERR(args.value);
526 		args.valuelen = len;
527 	}
528 
529 	error = xfs_attr_change(&args);
530 	if (!error && (flags & XFS_IOC_ATTR_ROOT))
531 		xfs_forget_acl(inode, name);
532 	kfree(args.value);
533 	return error;
534 }
535 
536 int
537 xfs_ioc_attrmulti_one(
538 	struct file		*parfilp,
539 	struct inode		*inode,
540 	uint32_t		opcode,
541 	void __user		*uname,
542 	void __user		*value,
543 	uint32_t		*len,
544 	uint32_t		flags)
545 {
546 	unsigned char		*name;
547 	int			error;
548 
549 	if ((flags & XFS_IOC_ATTR_ROOT) && (flags & XFS_IOC_ATTR_SECURE))
550 		return -EINVAL;
551 
552 	name = strndup_user(uname, MAXNAMELEN);
553 	if (IS_ERR(name))
554 		return PTR_ERR(name);
555 
556 	switch (opcode) {
557 	case ATTR_OP_GET:
558 		error = xfs_attrmulti_attr_get(inode, name, value, len, flags);
559 		break;
560 	case ATTR_OP_REMOVE:
561 		value = NULL;
562 		*len = 0;
563 		fallthrough;
564 	case ATTR_OP_SET:
565 		error = mnt_want_write_file(parfilp);
566 		if (error)
567 			break;
568 		error = xfs_attrmulti_attr_set(inode, name, value, *len, flags);
569 		mnt_drop_write_file(parfilp);
570 		break;
571 	default:
572 		error = -EINVAL;
573 		break;
574 	}
575 
576 	kfree(name);
577 	return error;
578 }
579 
580 STATIC int
581 xfs_attrmulti_by_handle(
582 	struct file		*parfilp,
583 	void			__user *arg)
584 {
585 	int			error;
586 	xfs_attr_multiop_t	*ops;
587 	xfs_fsop_attrmulti_handlereq_t am_hreq;
588 	struct dentry		*dentry;
589 	unsigned int		i, size;
590 
591 	if (!capable(CAP_SYS_ADMIN))
592 		return -EPERM;
593 	if (copy_from_user(&am_hreq, arg, sizeof(xfs_fsop_attrmulti_handlereq_t)))
594 		return -EFAULT;
595 
596 	/* overflow check */
597 	if (am_hreq.opcount >= INT_MAX / sizeof(xfs_attr_multiop_t))
598 		return -E2BIG;
599 
600 	dentry = xfs_handlereq_to_dentry(parfilp, &am_hreq.hreq);
601 	if (IS_ERR(dentry))
602 		return PTR_ERR(dentry);
603 
604 	error = -E2BIG;
605 	size = am_hreq.opcount * sizeof(xfs_attr_multiop_t);
606 	if (!size || size > 16 * PAGE_SIZE)
607 		goto out_dput;
608 
609 	ops = memdup_user(am_hreq.ops, size);
610 	if (IS_ERR(ops)) {
611 		error = PTR_ERR(ops);
612 		goto out_dput;
613 	}
614 
615 	error = 0;
616 	for (i = 0; i < am_hreq.opcount; i++) {
617 		ops[i].am_error = xfs_ioc_attrmulti_one(parfilp,
618 				d_inode(dentry), ops[i].am_opcode,
619 				ops[i].am_attrname, ops[i].am_attrvalue,
620 				&ops[i].am_length, ops[i].am_flags);
621 	}
622 
623 	if (copy_to_user(am_hreq.ops, ops, size))
624 		error = -EFAULT;
625 
626 	kfree(ops);
627  out_dput:
628 	dput(dentry);
629 	return error;
630 }
631 
632 /* Return 0 on success or positive error */
633 int
634 xfs_fsbulkstat_one_fmt(
635 	struct xfs_ibulk		*breq,
636 	const struct xfs_bulkstat	*bstat)
637 {
638 	struct xfs_bstat		bs1;
639 
640 	xfs_bulkstat_to_bstat(breq->mp, &bs1, bstat);
641 	if (copy_to_user(breq->ubuffer, &bs1, sizeof(bs1)))
642 		return -EFAULT;
643 	return xfs_ibulk_advance(breq, sizeof(struct xfs_bstat));
644 }
645 
646 int
647 xfs_fsinumbers_fmt(
648 	struct xfs_ibulk		*breq,
649 	const struct xfs_inumbers	*igrp)
650 {
651 	struct xfs_inogrp		ig1;
652 
653 	xfs_inumbers_to_inogrp(&ig1, igrp);
654 	if (copy_to_user(breq->ubuffer, &ig1, sizeof(struct xfs_inogrp)))
655 		return -EFAULT;
656 	return xfs_ibulk_advance(breq, sizeof(struct xfs_inogrp));
657 }
658 
659 STATIC int
660 xfs_ioc_fsbulkstat(
661 	struct file		*file,
662 	unsigned int		cmd,
663 	void			__user *arg)
664 {
665 	struct xfs_mount	*mp = XFS_I(file_inode(file))->i_mount;
666 	struct xfs_fsop_bulkreq	bulkreq;
667 	struct xfs_ibulk	breq = {
668 		.mp		= mp,
669 		.idmap		= file_mnt_idmap(file),
670 		.ocount		= 0,
671 	};
672 	xfs_ino_t		lastino;
673 	int			error;
674 
675 	/* done = 1 if there are more stats to get and if bulkstat */
676 	/* should be called again (unused here, but used in dmapi) */
677 
678 	if (!capable(CAP_SYS_ADMIN))
679 		return -EPERM;
680 
681 	if (xfs_is_shutdown(mp))
682 		return -EIO;
683 
684 	if (copy_from_user(&bulkreq, arg, sizeof(struct xfs_fsop_bulkreq)))
685 		return -EFAULT;
686 
687 	if (copy_from_user(&lastino, bulkreq.lastip, sizeof(__s64)))
688 		return -EFAULT;
689 
690 	if (bulkreq.icount <= 0)
691 		return -EINVAL;
692 
693 	if (bulkreq.ubuffer == NULL)
694 		return -EINVAL;
695 
696 	breq.ubuffer = bulkreq.ubuffer;
697 	breq.icount = bulkreq.icount;
698 
699 	/*
700 	 * FSBULKSTAT_SINGLE expects that *lastip contains the inode number
701 	 * that we want to stat.  However, FSINUMBERS and FSBULKSTAT expect
702 	 * that *lastip contains either zero or the number of the last inode to
703 	 * be examined by the previous call and return results starting with
704 	 * the next inode after that.  The new bulk request back end functions
705 	 * take the inode to start with, so we have to compute the startino
706 	 * parameter from lastino to maintain correct function.  lastino == 0
707 	 * is a special case because it has traditionally meant "first inode
708 	 * in filesystem".
709 	 */
710 	if (cmd == XFS_IOC_FSINUMBERS) {
711 		breq.startino = lastino ? lastino + 1 : 0;
712 		error = xfs_inumbers(&breq, xfs_fsinumbers_fmt);
713 		lastino = breq.startino - 1;
714 	} else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE) {
715 		breq.startino = lastino;
716 		breq.icount = 1;
717 		error = xfs_bulkstat_one(&breq, xfs_fsbulkstat_one_fmt);
718 	} else {	/* XFS_IOC_FSBULKSTAT */
719 		breq.startino = lastino ? lastino + 1 : 0;
720 		error = xfs_bulkstat(&breq, xfs_fsbulkstat_one_fmt);
721 		lastino = breq.startino - 1;
722 	}
723 
724 	if (error)
725 		return error;
726 
727 	if (bulkreq.lastip != NULL &&
728 	    copy_to_user(bulkreq.lastip, &lastino, sizeof(xfs_ino_t)))
729 		return -EFAULT;
730 
731 	if (bulkreq.ocount != NULL &&
732 	    copy_to_user(bulkreq.ocount, &breq.ocount, sizeof(__s32)))
733 		return -EFAULT;
734 
735 	return 0;
736 }
737 
738 /* Return 0 on success or positive error */
739 static int
740 xfs_bulkstat_fmt(
741 	struct xfs_ibulk		*breq,
742 	const struct xfs_bulkstat	*bstat)
743 {
744 	if (copy_to_user(breq->ubuffer, bstat, sizeof(struct xfs_bulkstat)))
745 		return -EFAULT;
746 	return xfs_ibulk_advance(breq, sizeof(struct xfs_bulkstat));
747 }
748 
749 /*
750  * Check the incoming bulk request @hdr from userspace and initialize the
751  * internal @breq bulk request appropriately.  Returns 0 if the bulk request
752  * should proceed; -ECANCELED if there's nothing to do; or the usual
753  * negative error code.
754  */
755 static int
756 xfs_bulk_ireq_setup(
757 	struct xfs_mount	*mp,
758 	const struct xfs_bulk_ireq *hdr,
759 	struct xfs_ibulk	*breq,
760 	void __user		*ubuffer)
761 {
762 	if (hdr->icount == 0 ||
763 	    (hdr->flags & ~XFS_BULK_IREQ_FLAGS_ALL) ||
764 	    memchr_inv(hdr->reserved, 0, sizeof(hdr->reserved)))
765 		return -EINVAL;
766 
767 	breq->startino = hdr->ino;
768 	breq->ubuffer = ubuffer;
769 	breq->icount = hdr->icount;
770 	breq->ocount = 0;
771 	breq->flags = 0;
772 
773 	/*
774 	 * The @ino parameter is a special value, so we must look it up here.
775 	 * We're not allowed to have IREQ_AGNO, and we only return one inode
776 	 * worth of data.
777 	 */
778 	if (hdr->flags & XFS_BULK_IREQ_SPECIAL) {
779 		if (hdr->flags & XFS_BULK_IREQ_AGNO)
780 			return -EINVAL;
781 
782 		switch (hdr->ino) {
783 		case XFS_BULK_IREQ_SPECIAL_ROOT:
784 			breq->startino = mp->m_sb.sb_rootino;
785 			break;
786 		default:
787 			return -EINVAL;
788 		}
789 		breq->icount = 1;
790 	}
791 
792 	/*
793 	 * The IREQ_AGNO flag means that we only want results from a given AG.
794 	 * If @hdr->ino is zero, we start iterating in that AG.  If @hdr->ino is
795 	 * beyond the specified AG then we return no results.
796 	 */
797 	if (hdr->flags & XFS_BULK_IREQ_AGNO) {
798 		if (hdr->agno >= mp->m_sb.sb_agcount)
799 			return -EINVAL;
800 
801 		if (breq->startino == 0)
802 			breq->startino = XFS_AGINO_TO_INO(mp, hdr->agno, 0);
803 		else if (XFS_INO_TO_AGNO(mp, breq->startino) < hdr->agno)
804 			return -EINVAL;
805 
806 		breq->flags |= XFS_IBULK_SAME_AG;
807 
808 		/* Asking for an inode past the end of the AG?  We're done! */
809 		if (XFS_INO_TO_AGNO(mp, breq->startino) > hdr->agno)
810 			return -ECANCELED;
811 	} else if (hdr->agno)
812 		return -EINVAL;
813 
814 	/* Asking for an inode past the end of the FS?  We're done! */
815 	if (XFS_INO_TO_AGNO(mp, breq->startino) >= mp->m_sb.sb_agcount)
816 		return -ECANCELED;
817 
818 	if (hdr->flags & XFS_BULK_IREQ_NREXT64)
819 		breq->flags |= XFS_IBULK_NREXT64;
820 
821 	return 0;
822 }
823 
824 /*
825  * Update the userspace bulk request @hdr to reflect the end state of the
826  * internal bulk request @breq.
827  */
828 static void
829 xfs_bulk_ireq_teardown(
830 	struct xfs_bulk_ireq	*hdr,
831 	struct xfs_ibulk	*breq)
832 {
833 	hdr->ino = breq->startino;
834 	hdr->ocount = breq->ocount;
835 }
836 
837 /* Handle the v5 bulkstat ioctl. */
838 STATIC int
839 xfs_ioc_bulkstat(
840 	struct file			*file,
841 	unsigned int			cmd,
842 	struct xfs_bulkstat_req __user	*arg)
843 {
844 	struct xfs_mount		*mp = XFS_I(file_inode(file))->i_mount;
845 	struct xfs_bulk_ireq		hdr;
846 	struct xfs_ibulk		breq = {
847 		.mp			= mp,
848 		.idmap			= file_mnt_idmap(file),
849 	};
850 	int				error;
851 
852 	if (!capable(CAP_SYS_ADMIN))
853 		return -EPERM;
854 
855 	if (xfs_is_shutdown(mp))
856 		return -EIO;
857 
858 	if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
859 		return -EFAULT;
860 
861 	error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->bulkstat);
862 	if (error == -ECANCELED)
863 		goto out_teardown;
864 	if (error < 0)
865 		return error;
866 
867 	error = xfs_bulkstat(&breq, xfs_bulkstat_fmt);
868 	if (error)
869 		return error;
870 
871 out_teardown:
872 	xfs_bulk_ireq_teardown(&hdr, &breq);
873 	if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
874 		return -EFAULT;
875 
876 	return 0;
877 }
878 
879 STATIC int
880 xfs_inumbers_fmt(
881 	struct xfs_ibulk		*breq,
882 	const struct xfs_inumbers	*igrp)
883 {
884 	if (copy_to_user(breq->ubuffer, igrp, sizeof(struct xfs_inumbers)))
885 		return -EFAULT;
886 	return xfs_ibulk_advance(breq, sizeof(struct xfs_inumbers));
887 }
888 
889 /* Handle the v5 inumbers ioctl. */
890 STATIC int
891 xfs_ioc_inumbers(
892 	struct xfs_mount		*mp,
893 	unsigned int			cmd,
894 	struct xfs_inumbers_req __user	*arg)
895 {
896 	struct xfs_bulk_ireq		hdr;
897 	struct xfs_ibulk		breq = {
898 		.mp			= mp,
899 	};
900 	int				error;
901 
902 	if (!capable(CAP_SYS_ADMIN))
903 		return -EPERM;
904 
905 	if (xfs_is_shutdown(mp))
906 		return -EIO;
907 
908 	if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
909 		return -EFAULT;
910 
911 	error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->inumbers);
912 	if (error == -ECANCELED)
913 		goto out_teardown;
914 	if (error < 0)
915 		return error;
916 
917 	error = xfs_inumbers(&breq, xfs_inumbers_fmt);
918 	if (error)
919 		return error;
920 
921 out_teardown:
922 	xfs_bulk_ireq_teardown(&hdr, &breq);
923 	if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
924 		return -EFAULT;
925 
926 	return 0;
927 }
928 
929 STATIC int
930 xfs_ioc_fsgeometry(
931 	struct xfs_mount	*mp,
932 	void			__user *arg,
933 	int			struct_version)
934 {
935 	struct xfs_fsop_geom	fsgeo;
936 	size_t			len;
937 
938 	xfs_fs_geometry(mp, &fsgeo, struct_version);
939 
940 	if (struct_version <= 3)
941 		len = sizeof(struct xfs_fsop_geom_v1);
942 	else if (struct_version == 4)
943 		len = sizeof(struct xfs_fsop_geom_v4);
944 	else {
945 		xfs_fsop_geom_health(mp, &fsgeo);
946 		len = sizeof(fsgeo);
947 	}
948 
949 	if (copy_to_user(arg, &fsgeo, len))
950 		return -EFAULT;
951 	return 0;
952 }
953 
954 STATIC int
955 xfs_ioc_ag_geometry(
956 	struct xfs_mount	*mp,
957 	void			__user *arg)
958 {
959 	struct xfs_perag	*pag;
960 	struct xfs_ag_geometry	ageo;
961 	int			error;
962 
963 	if (copy_from_user(&ageo, arg, sizeof(ageo)))
964 		return -EFAULT;
965 	if (ageo.ag_flags)
966 		return -EINVAL;
967 	if (memchr_inv(&ageo.ag_reserved, 0, sizeof(ageo.ag_reserved)))
968 		return -EINVAL;
969 
970 	pag = xfs_perag_get(mp, ageo.ag_number);
971 	if (!pag)
972 		return -EINVAL;
973 
974 	error = xfs_ag_get_geometry(pag, &ageo);
975 	xfs_perag_put(pag);
976 	if (error)
977 		return error;
978 
979 	if (copy_to_user(arg, &ageo, sizeof(ageo)))
980 		return -EFAULT;
981 	return 0;
982 }
983 
984 /*
985  * Linux extended inode flags interface.
986  */
987 
988 static void
989 xfs_fill_fsxattr(
990 	struct xfs_inode	*ip,
991 	int			whichfork,
992 	struct fileattr		*fa)
993 {
994 	struct xfs_mount	*mp = ip->i_mount;
995 	struct xfs_ifork	*ifp = xfs_ifork_ptr(ip, whichfork);
996 
997 	fileattr_fill_xflags(fa, xfs_ip2xflags(ip));
998 
999 	if (ip->i_diflags & XFS_DIFLAG_EXTSIZE) {
1000 		fa->fsx_extsize = XFS_FSB_TO_B(mp, ip->i_extsize);
1001 	} else if (ip->i_diflags & XFS_DIFLAG_EXTSZINHERIT) {
1002 		/*
1003 		 * Don't let a misaligned extent size hint on a directory
1004 		 * escape to userspace if it won't pass the setattr checks
1005 		 * later.
1006 		 */
1007 		if ((ip->i_diflags & XFS_DIFLAG_RTINHERIT) &&
1008 		    xfs_extlen_to_rtxmod(mp, ip->i_extsize) > 0) {
1009 			fa->fsx_xflags &= ~(FS_XFLAG_EXTSIZE |
1010 					    FS_XFLAG_EXTSZINHERIT);
1011 			fa->fsx_extsize = 0;
1012 		} else {
1013 			fa->fsx_extsize = XFS_FSB_TO_B(mp, ip->i_extsize);
1014 		}
1015 	}
1016 
1017 	if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
1018 		fa->fsx_cowextsize = XFS_FSB_TO_B(mp, ip->i_cowextsize);
1019 	fa->fsx_projid = ip->i_projid;
1020 	if (ifp && !xfs_need_iread_extents(ifp))
1021 		fa->fsx_nextents = xfs_iext_count(ifp);
1022 	else
1023 		fa->fsx_nextents = xfs_ifork_nextents(ifp);
1024 }
1025 
1026 STATIC int
1027 xfs_ioc_fsgetxattra(
1028 	xfs_inode_t		*ip,
1029 	void			__user *arg)
1030 {
1031 	struct fileattr		fa;
1032 
1033 	xfs_ilock(ip, XFS_ILOCK_SHARED);
1034 	xfs_fill_fsxattr(ip, XFS_ATTR_FORK, &fa);
1035 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
1036 
1037 	return copy_fsxattr_to_user(&fa, arg);
1038 }
1039 
1040 int
1041 xfs_fileattr_get(
1042 	struct dentry		*dentry,
1043 	struct fileattr		*fa)
1044 {
1045 	struct xfs_inode	*ip = XFS_I(d_inode(dentry));
1046 
1047 	if (d_is_special(dentry))
1048 		return -ENOTTY;
1049 
1050 	xfs_ilock(ip, XFS_ILOCK_SHARED);
1051 	xfs_fill_fsxattr(ip, XFS_DATA_FORK, fa);
1052 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
1053 
1054 	return 0;
1055 }
1056 
1057 STATIC uint16_t
1058 xfs_flags2diflags(
1059 	struct xfs_inode	*ip,
1060 	unsigned int		xflags)
1061 {
1062 	/* can't set PREALLOC this way, just preserve it */
1063 	uint16_t		di_flags =
1064 		(ip->i_diflags & XFS_DIFLAG_PREALLOC);
1065 
1066 	if (xflags & FS_XFLAG_IMMUTABLE)
1067 		di_flags |= XFS_DIFLAG_IMMUTABLE;
1068 	if (xflags & FS_XFLAG_APPEND)
1069 		di_flags |= XFS_DIFLAG_APPEND;
1070 	if (xflags & FS_XFLAG_SYNC)
1071 		di_flags |= XFS_DIFLAG_SYNC;
1072 	if (xflags & FS_XFLAG_NOATIME)
1073 		di_flags |= XFS_DIFLAG_NOATIME;
1074 	if (xflags & FS_XFLAG_NODUMP)
1075 		di_flags |= XFS_DIFLAG_NODUMP;
1076 	if (xflags & FS_XFLAG_NODEFRAG)
1077 		di_flags |= XFS_DIFLAG_NODEFRAG;
1078 	if (xflags & FS_XFLAG_FILESTREAM)
1079 		di_flags |= XFS_DIFLAG_FILESTREAM;
1080 	if (S_ISDIR(VFS_I(ip)->i_mode)) {
1081 		if (xflags & FS_XFLAG_RTINHERIT)
1082 			di_flags |= XFS_DIFLAG_RTINHERIT;
1083 		if (xflags & FS_XFLAG_NOSYMLINKS)
1084 			di_flags |= XFS_DIFLAG_NOSYMLINKS;
1085 		if (xflags & FS_XFLAG_EXTSZINHERIT)
1086 			di_flags |= XFS_DIFLAG_EXTSZINHERIT;
1087 		if (xflags & FS_XFLAG_PROJINHERIT)
1088 			di_flags |= XFS_DIFLAG_PROJINHERIT;
1089 	} else if (S_ISREG(VFS_I(ip)->i_mode)) {
1090 		if (xflags & FS_XFLAG_REALTIME)
1091 			di_flags |= XFS_DIFLAG_REALTIME;
1092 		if (xflags & FS_XFLAG_EXTSIZE)
1093 			di_flags |= XFS_DIFLAG_EXTSIZE;
1094 	}
1095 
1096 	return di_flags;
1097 }
1098 
1099 STATIC uint64_t
1100 xfs_flags2diflags2(
1101 	struct xfs_inode	*ip,
1102 	unsigned int		xflags)
1103 {
1104 	uint64_t		di_flags2 =
1105 		(ip->i_diflags2 & (XFS_DIFLAG2_REFLINK |
1106 				   XFS_DIFLAG2_BIGTIME |
1107 				   XFS_DIFLAG2_NREXT64));
1108 
1109 	if (xflags & FS_XFLAG_DAX)
1110 		di_flags2 |= XFS_DIFLAG2_DAX;
1111 	if (xflags & FS_XFLAG_COWEXTSIZE)
1112 		di_flags2 |= XFS_DIFLAG2_COWEXTSIZE;
1113 
1114 	return di_flags2;
1115 }
1116 
1117 static int
1118 xfs_ioctl_setattr_xflags(
1119 	struct xfs_trans	*tp,
1120 	struct xfs_inode	*ip,
1121 	struct fileattr		*fa)
1122 {
1123 	struct xfs_mount	*mp = ip->i_mount;
1124 	bool			rtflag = (fa->fsx_xflags & FS_XFLAG_REALTIME);
1125 	uint64_t		i_flags2;
1126 
1127 	if (rtflag != XFS_IS_REALTIME_INODE(ip)) {
1128 		/* Can't change realtime flag if any extents are allocated. */
1129 		if (ip->i_df.if_nextents || ip->i_delayed_blks)
1130 			return -EINVAL;
1131 	}
1132 
1133 	if (rtflag) {
1134 		/* If realtime flag is set then must have realtime device */
1135 		if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
1136 		    xfs_extlen_to_rtxmod(mp, ip->i_extsize))
1137 			return -EINVAL;
1138 
1139 		/* Clear reflink if we are actually able to set the rt flag. */
1140 		if (xfs_is_reflink_inode(ip))
1141 			ip->i_diflags2 &= ~XFS_DIFLAG2_REFLINK;
1142 	}
1143 
1144 	/* diflags2 only valid for v3 inodes. */
1145 	i_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1146 	if (i_flags2 && !xfs_has_v3inodes(mp))
1147 		return -EINVAL;
1148 
1149 	ip->i_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
1150 	ip->i_diflags2 = i_flags2;
1151 
1152 	xfs_diflags_to_iflags(ip, false);
1153 
1154 	/*
1155 	 * Make the stable writes flag match that of the device the inode
1156 	 * resides on when flipping the RT flag.
1157 	 */
1158 	if (rtflag != XFS_IS_REALTIME_INODE(ip) && S_ISREG(VFS_I(ip)->i_mode))
1159 		xfs_update_stable_writes(ip);
1160 
1161 	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1162 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1163 	XFS_STATS_INC(mp, xs_ig_attrchg);
1164 	return 0;
1165 }
1166 
1167 static void
1168 xfs_ioctl_setattr_prepare_dax(
1169 	struct xfs_inode	*ip,
1170 	struct fileattr		*fa)
1171 {
1172 	struct xfs_mount	*mp = ip->i_mount;
1173 	struct inode            *inode = VFS_I(ip);
1174 
1175 	if (S_ISDIR(inode->i_mode))
1176 		return;
1177 
1178 	if (xfs_has_dax_always(mp) || xfs_has_dax_never(mp))
1179 		return;
1180 
1181 	if (((fa->fsx_xflags & FS_XFLAG_DAX) &&
1182 	    !(ip->i_diflags2 & XFS_DIFLAG2_DAX)) ||
1183 	    (!(fa->fsx_xflags & FS_XFLAG_DAX) &&
1184 	     (ip->i_diflags2 & XFS_DIFLAG2_DAX)))
1185 		d_mark_dontcache(inode);
1186 }
1187 
1188 /*
1189  * Set up the transaction structure for the setattr operation, checking that we
1190  * have permission to do so. On success, return a clean transaction and the
1191  * inode locked exclusively ready for further operation specific checks. On
1192  * failure, return an error without modifying or locking the inode.
1193  */
1194 static struct xfs_trans *
1195 xfs_ioctl_setattr_get_trans(
1196 	struct xfs_inode	*ip,
1197 	struct xfs_dquot	*pdqp)
1198 {
1199 	struct xfs_mount	*mp = ip->i_mount;
1200 	struct xfs_trans	*tp;
1201 	int			error = -EROFS;
1202 
1203 	if (xfs_is_readonly(mp))
1204 		goto out_error;
1205 	error = -EIO;
1206 	if (xfs_is_shutdown(mp))
1207 		goto out_error;
1208 
1209 	error = xfs_trans_alloc_ichange(ip, NULL, NULL, pdqp,
1210 			has_capability_noaudit(current, CAP_FOWNER), &tp);
1211 	if (error)
1212 		goto out_error;
1213 
1214 	if (xfs_has_wsync(mp))
1215 		xfs_trans_set_sync(tp);
1216 
1217 	return tp;
1218 
1219 out_error:
1220 	return ERR_PTR(error);
1221 }
1222 
1223 /*
1224  * Validate a proposed extent size hint.  For regular files, the hint can only
1225  * be changed if no extents are allocated.
1226  */
1227 static int
1228 xfs_ioctl_setattr_check_extsize(
1229 	struct xfs_inode	*ip,
1230 	struct fileattr		*fa)
1231 {
1232 	struct xfs_mount	*mp = ip->i_mount;
1233 	xfs_failaddr_t		failaddr;
1234 	uint16_t		new_diflags;
1235 
1236 	if (!fa->fsx_valid)
1237 		return 0;
1238 
1239 	if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_df.if_nextents &&
1240 	    XFS_FSB_TO_B(mp, ip->i_extsize) != fa->fsx_extsize)
1241 		return -EINVAL;
1242 
1243 	if (fa->fsx_extsize & mp->m_blockmask)
1244 		return -EINVAL;
1245 
1246 	new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
1247 
1248 	/*
1249 	 * Inode verifiers do not check that the extent size hint is an integer
1250 	 * multiple of the rt extent size on a directory with both rtinherit
1251 	 * and extszinherit flags set.  Don't let sysadmins misconfigure
1252 	 * directories.
1253 	 */
1254 	if ((new_diflags & XFS_DIFLAG_RTINHERIT) &&
1255 	    (new_diflags & XFS_DIFLAG_EXTSZINHERIT)) {
1256 		unsigned int	rtextsize_bytes;
1257 
1258 		rtextsize_bytes = XFS_FSB_TO_B(mp, mp->m_sb.sb_rextsize);
1259 		if (fa->fsx_extsize % rtextsize_bytes)
1260 			return -EINVAL;
1261 	}
1262 
1263 	failaddr = xfs_inode_validate_extsize(ip->i_mount,
1264 			XFS_B_TO_FSB(mp, fa->fsx_extsize),
1265 			VFS_I(ip)->i_mode, new_diflags);
1266 	return failaddr != NULL ? -EINVAL : 0;
1267 }
1268 
1269 static int
1270 xfs_ioctl_setattr_check_cowextsize(
1271 	struct xfs_inode	*ip,
1272 	struct fileattr		*fa)
1273 {
1274 	struct xfs_mount	*mp = ip->i_mount;
1275 	xfs_failaddr_t		failaddr;
1276 	uint64_t		new_diflags2;
1277 	uint16_t		new_diflags;
1278 
1279 	if (!fa->fsx_valid)
1280 		return 0;
1281 
1282 	if (fa->fsx_cowextsize & mp->m_blockmask)
1283 		return -EINVAL;
1284 
1285 	new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
1286 	new_diflags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1287 
1288 	failaddr = xfs_inode_validate_cowextsize(ip->i_mount,
1289 			XFS_B_TO_FSB(mp, fa->fsx_cowextsize),
1290 			VFS_I(ip)->i_mode, new_diflags, new_diflags2);
1291 	return failaddr != NULL ? -EINVAL : 0;
1292 }
1293 
1294 static int
1295 xfs_ioctl_setattr_check_projid(
1296 	struct xfs_inode	*ip,
1297 	struct fileattr		*fa)
1298 {
1299 	if (!fa->fsx_valid)
1300 		return 0;
1301 
1302 	/* Disallow 32bit project ids if 32bit IDs are not enabled. */
1303 	if (fa->fsx_projid > (uint16_t)-1 &&
1304 	    !xfs_has_projid32(ip->i_mount))
1305 		return -EINVAL;
1306 	return 0;
1307 }
1308 
1309 int
1310 xfs_fileattr_set(
1311 	struct mnt_idmap	*idmap,
1312 	struct dentry		*dentry,
1313 	struct fileattr		*fa)
1314 {
1315 	struct xfs_inode	*ip = XFS_I(d_inode(dentry));
1316 	struct xfs_mount	*mp = ip->i_mount;
1317 	struct xfs_trans	*tp;
1318 	struct xfs_dquot	*pdqp = NULL;
1319 	struct xfs_dquot	*olddquot = NULL;
1320 	int			error;
1321 
1322 	trace_xfs_ioctl_setattr(ip);
1323 
1324 	if (d_is_special(dentry))
1325 		return -ENOTTY;
1326 
1327 	if (!fa->fsx_valid) {
1328 		if (fa->flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL |
1329 				  FS_NOATIME_FL | FS_NODUMP_FL |
1330 				  FS_SYNC_FL | FS_DAX_FL | FS_PROJINHERIT_FL))
1331 			return -EOPNOTSUPP;
1332 	}
1333 
1334 	error = xfs_ioctl_setattr_check_projid(ip, fa);
1335 	if (error)
1336 		return error;
1337 
1338 	/*
1339 	 * If disk quotas is on, we make sure that the dquots do exist on disk,
1340 	 * before we start any other transactions. Trying to do this later
1341 	 * is messy. We don't care to take a readlock to look at the ids
1342 	 * in inode here, because we can't hold it across the trans_reserve.
1343 	 * If the IDs do change before we take the ilock, we're covered
1344 	 * because the i_*dquot fields will get updated anyway.
1345 	 */
1346 	if (fa->fsx_valid && XFS_IS_QUOTA_ON(mp)) {
1347 		error = xfs_qm_vop_dqalloc(ip, VFS_I(ip)->i_uid,
1348 				VFS_I(ip)->i_gid, fa->fsx_projid,
1349 				XFS_QMOPT_PQUOTA, NULL, NULL, &pdqp);
1350 		if (error)
1351 			return error;
1352 	}
1353 
1354 	xfs_ioctl_setattr_prepare_dax(ip, fa);
1355 
1356 	tp = xfs_ioctl_setattr_get_trans(ip, pdqp);
1357 	if (IS_ERR(tp)) {
1358 		error = PTR_ERR(tp);
1359 		goto error_free_dquots;
1360 	}
1361 
1362 	error = xfs_ioctl_setattr_check_extsize(ip, fa);
1363 	if (error)
1364 		goto error_trans_cancel;
1365 
1366 	error = xfs_ioctl_setattr_check_cowextsize(ip, fa);
1367 	if (error)
1368 		goto error_trans_cancel;
1369 
1370 	error = xfs_ioctl_setattr_xflags(tp, ip, fa);
1371 	if (error)
1372 		goto error_trans_cancel;
1373 
1374 	if (!fa->fsx_valid)
1375 		goto skip_xattr;
1376 	/*
1377 	 * Change file ownership.  Must be the owner or privileged.  CAP_FSETID
1378 	 * overrides the following restrictions:
1379 	 *
1380 	 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1381 	 * successful return from chown()
1382 	 */
1383 
1384 	if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
1385 	    !capable_wrt_inode_uidgid(idmap, VFS_I(ip), CAP_FSETID))
1386 		VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
1387 
1388 	/* Change the ownerships and register project quota modifications */
1389 	if (ip->i_projid != fa->fsx_projid) {
1390 		if (XFS_IS_PQUOTA_ON(mp)) {
1391 			olddquot = xfs_qm_vop_chown(tp, ip,
1392 						&ip->i_pdquot, pdqp);
1393 		}
1394 		ip->i_projid = fa->fsx_projid;
1395 	}
1396 
1397 	/*
1398 	 * Only set the extent size hint if we've already determined that the
1399 	 * extent size hint should be set on the inode. If no extent size flags
1400 	 * are set on the inode then unconditionally clear the extent size hint.
1401 	 */
1402 	if (ip->i_diflags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1403 		ip->i_extsize = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1404 	else
1405 		ip->i_extsize = 0;
1406 
1407 	if (xfs_has_v3inodes(mp)) {
1408 		if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
1409 			ip->i_cowextsize = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
1410 		else
1411 			ip->i_cowextsize = 0;
1412 	}
1413 
1414 skip_xattr:
1415 	error = xfs_trans_commit(tp);
1416 
1417 	/*
1418 	 * Release any dquot(s) the inode had kept before chown.
1419 	 */
1420 	xfs_qm_dqrele(olddquot);
1421 	xfs_qm_dqrele(pdqp);
1422 
1423 	return error;
1424 
1425 error_trans_cancel:
1426 	xfs_trans_cancel(tp);
1427 error_free_dquots:
1428 	xfs_qm_dqrele(pdqp);
1429 	return error;
1430 }
1431 
1432 static bool
1433 xfs_getbmap_format(
1434 	struct kgetbmap		*p,
1435 	struct getbmapx __user	*u,
1436 	size_t			recsize)
1437 {
1438 	if (put_user(p->bmv_offset, &u->bmv_offset) ||
1439 	    put_user(p->bmv_block, &u->bmv_block) ||
1440 	    put_user(p->bmv_length, &u->bmv_length) ||
1441 	    put_user(0, &u->bmv_count) ||
1442 	    put_user(0, &u->bmv_entries))
1443 		return false;
1444 	if (recsize < sizeof(struct getbmapx))
1445 		return true;
1446 	if (put_user(0, &u->bmv_iflags) ||
1447 	    put_user(p->bmv_oflags, &u->bmv_oflags) ||
1448 	    put_user(0, &u->bmv_unused1) ||
1449 	    put_user(0, &u->bmv_unused2))
1450 		return false;
1451 	return true;
1452 }
1453 
1454 STATIC int
1455 xfs_ioc_getbmap(
1456 	struct file		*file,
1457 	unsigned int		cmd,
1458 	void			__user *arg)
1459 {
1460 	struct getbmapx		bmx = { 0 };
1461 	struct kgetbmap		*buf;
1462 	size_t			recsize;
1463 	int			error, i;
1464 
1465 	switch (cmd) {
1466 	case XFS_IOC_GETBMAPA:
1467 		bmx.bmv_iflags = BMV_IF_ATTRFORK;
1468 		fallthrough;
1469 	case XFS_IOC_GETBMAP:
1470 		/* struct getbmap is a strict subset of struct getbmapx. */
1471 		recsize = sizeof(struct getbmap);
1472 		break;
1473 	case XFS_IOC_GETBMAPX:
1474 		recsize = sizeof(struct getbmapx);
1475 		break;
1476 	default:
1477 		return -EINVAL;
1478 	}
1479 
1480 	if (copy_from_user(&bmx, arg, recsize))
1481 		return -EFAULT;
1482 
1483 	if (bmx.bmv_count < 2)
1484 		return -EINVAL;
1485 	if (bmx.bmv_count >= INT_MAX / recsize)
1486 		return -ENOMEM;
1487 
1488 	buf = kvcalloc(bmx.bmv_count, sizeof(*buf), GFP_KERNEL);
1489 	if (!buf)
1490 		return -ENOMEM;
1491 
1492 	error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, buf);
1493 	if (error)
1494 		goto out_free_buf;
1495 
1496 	error = -EFAULT;
1497 	if (copy_to_user(arg, &bmx, recsize))
1498 		goto out_free_buf;
1499 	arg += recsize;
1500 
1501 	for (i = 0; i < bmx.bmv_entries; i++) {
1502 		if (!xfs_getbmap_format(buf + i, arg, recsize))
1503 			goto out_free_buf;
1504 		arg += recsize;
1505 	}
1506 
1507 	error = 0;
1508 out_free_buf:
1509 	kvfree(buf);
1510 	return error;
1511 }
1512 
1513 STATIC int
1514 xfs_ioc_getfsmap(
1515 	struct xfs_inode	*ip,
1516 	struct fsmap_head	__user *arg)
1517 {
1518 	struct xfs_fsmap_head	xhead = {0};
1519 	struct fsmap_head	head;
1520 	struct fsmap		*recs;
1521 	unsigned int		count;
1522 	__u32			last_flags = 0;
1523 	bool			done = false;
1524 	int			error;
1525 
1526 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
1527 		return -EFAULT;
1528 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
1529 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
1530 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
1531 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
1532 		       sizeof(head.fmh_keys[1].fmr_reserved)))
1533 		return -EINVAL;
1534 
1535 	/*
1536 	 * Use an internal memory buffer so that we don't have to copy fsmap
1537 	 * data to userspace while holding locks.  Start by trying to allocate
1538 	 * up to 128k for the buffer, but fall back to a single page if needed.
1539 	 */
1540 	count = min_t(unsigned int, head.fmh_count,
1541 			131072 / sizeof(struct fsmap));
1542 	recs = kvcalloc(count, sizeof(struct fsmap), GFP_KERNEL);
1543 	if (!recs) {
1544 		count = min_t(unsigned int, head.fmh_count,
1545 				PAGE_SIZE / sizeof(struct fsmap));
1546 		recs = kvcalloc(count, sizeof(struct fsmap), GFP_KERNEL);
1547 		if (!recs)
1548 			return -ENOMEM;
1549 	}
1550 
1551 	xhead.fmh_iflags = head.fmh_iflags;
1552 	xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
1553 	xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
1554 
1555 	trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1556 	trace_xfs_getfsmap_high_key(ip->i_mount, &xhead.fmh_keys[1]);
1557 
1558 	head.fmh_entries = 0;
1559 	do {
1560 		struct fsmap __user	*user_recs;
1561 		struct fsmap		*last_rec;
1562 
1563 		user_recs = &arg->fmh_recs[head.fmh_entries];
1564 		xhead.fmh_entries = 0;
1565 		xhead.fmh_count = min_t(unsigned int, count,
1566 					head.fmh_count - head.fmh_entries);
1567 
1568 		/* Run query, record how many entries we got. */
1569 		error = xfs_getfsmap(ip->i_mount, &xhead, recs);
1570 		switch (error) {
1571 		case 0:
1572 			/*
1573 			 * There are no more records in the result set.  Copy
1574 			 * whatever we got to userspace and break out.
1575 			 */
1576 			done = true;
1577 			break;
1578 		case -ECANCELED:
1579 			/*
1580 			 * The internal memory buffer is full.  Copy whatever
1581 			 * records we got to userspace and go again if we have
1582 			 * not yet filled the userspace buffer.
1583 			 */
1584 			error = 0;
1585 			break;
1586 		default:
1587 			goto out_free;
1588 		}
1589 		head.fmh_entries += xhead.fmh_entries;
1590 		head.fmh_oflags = xhead.fmh_oflags;
1591 
1592 		/*
1593 		 * If the caller wanted a record count or there aren't any
1594 		 * new records to return, we're done.
1595 		 */
1596 		if (head.fmh_count == 0 || xhead.fmh_entries == 0)
1597 			break;
1598 
1599 		/* Copy all the records we got out to userspace. */
1600 		if (copy_to_user(user_recs, recs,
1601 				 xhead.fmh_entries * sizeof(struct fsmap))) {
1602 			error = -EFAULT;
1603 			goto out_free;
1604 		}
1605 
1606 		/* Remember the last record flags we copied to userspace. */
1607 		last_rec = &recs[xhead.fmh_entries - 1];
1608 		last_flags = last_rec->fmr_flags;
1609 
1610 		/* Set up the low key for the next iteration. */
1611 		xfs_fsmap_to_internal(&xhead.fmh_keys[0], last_rec);
1612 		trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1613 	} while (!done && head.fmh_entries < head.fmh_count);
1614 
1615 	/*
1616 	 * If there are no more records in the query result set and we're not
1617 	 * in counting mode, mark the last record returned with the LAST flag.
1618 	 */
1619 	if (done && head.fmh_count > 0 && head.fmh_entries > 0) {
1620 		struct fsmap __user	*user_rec;
1621 
1622 		last_flags |= FMR_OF_LAST;
1623 		user_rec = &arg->fmh_recs[head.fmh_entries - 1];
1624 
1625 		if (copy_to_user(&user_rec->fmr_flags, &last_flags,
1626 					sizeof(last_flags))) {
1627 			error = -EFAULT;
1628 			goto out_free;
1629 		}
1630 	}
1631 
1632 	/* copy back header */
1633 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head))) {
1634 		error = -EFAULT;
1635 		goto out_free;
1636 	}
1637 
1638 out_free:
1639 	kvfree(recs);
1640 	return error;
1641 }
1642 
1643 STATIC int
1644 xfs_ioc_scrub_metadata(
1645 	struct file			*file,
1646 	void				__user *arg)
1647 {
1648 	struct xfs_scrub_metadata	scrub;
1649 	int				error;
1650 
1651 	if (!capable(CAP_SYS_ADMIN))
1652 		return -EPERM;
1653 
1654 	if (copy_from_user(&scrub, arg, sizeof(scrub)))
1655 		return -EFAULT;
1656 
1657 	error = xfs_scrub_metadata(file, &scrub);
1658 	if (error)
1659 		return error;
1660 
1661 	if (copy_to_user(arg, &scrub, sizeof(scrub)))
1662 		return -EFAULT;
1663 
1664 	return 0;
1665 }
1666 
1667 int
1668 xfs_ioc_swapext(
1669 	xfs_swapext_t	*sxp)
1670 {
1671 	xfs_inode_t     *ip, *tip;
1672 	struct fd	f, tmp;
1673 	int		error = 0;
1674 
1675 	/* Pull information for the target fd */
1676 	f = fdget((int)sxp->sx_fdtarget);
1677 	if (!f.file) {
1678 		error = -EINVAL;
1679 		goto out;
1680 	}
1681 
1682 	if (!(f.file->f_mode & FMODE_WRITE) ||
1683 	    !(f.file->f_mode & FMODE_READ) ||
1684 	    (f.file->f_flags & O_APPEND)) {
1685 		error = -EBADF;
1686 		goto out_put_file;
1687 	}
1688 
1689 	tmp = fdget((int)sxp->sx_fdtmp);
1690 	if (!tmp.file) {
1691 		error = -EINVAL;
1692 		goto out_put_file;
1693 	}
1694 
1695 	if (!(tmp.file->f_mode & FMODE_WRITE) ||
1696 	    !(tmp.file->f_mode & FMODE_READ) ||
1697 	    (tmp.file->f_flags & O_APPEND)) {
1698 		error = -EBADF;
1699 		goto out_put_tmp_file;
1700 	}
1701 
1702 	if (IS_SWAPFILE(file_inode(f.file)) ||
1703 	    IS_SWAPFILE(file_inode(tmp.file))) {
1704 		error = -EINVAL;
1705 		goto out_put_tmp_file;
1706 	}
1707 
1708 	/*
1709 	 * We need to ensure that the fds passed in point to XFS inodes
1710 	 * before we cast and access them as XFS structures as we have no
1711 	 * control over what the user passes us here.
1712 	 */
1713 	if (f.file->f_op != &xfs_file_operations ||
1714 	    tmp.file->f_op != &xfs_file_operations) {
1715 		error = -EINVAL;
1716 		goto out_put_tmp_file;
1717 	}
1718 
1719 	ip = XFS_I(file_inode(f.file));
1720 	tip = XFS_I(file_inode(tmp.file));
1721 
1722 	if (ip->i_mount != tip->i_mount) {
1723 		error = -EINVAL;
1724 		goto out_put_tmp_file;
1725 	}
1726 
1727 	if (ip->i_ino == tip->i_ino) {
1728 		error = -EINVAL;
1729 		goto out_put_tmp_file;
1730 	}
1731 
1732 	if (xfs_is_shutdown(ip->i_mount)) {
1733 		error = -EIO;
1734 		goto out_put_tmp_file;
1735 	}
1736 
1737 	error = xfs_swap_extents(ip, tip, sxp);
1738 
1739  out_put_tmp_file:
1740 	fdput(tmp);
1741  out_put_file:
1742 	fdput(f);
1743  out:
1744 	return error;
1745 }
1746 
1747 static int
1748 xfs_ioc_getlabel(
1749 	struct xfs_mount	*mp,
1750 	char			__user *user_label)
1751 {
1752 	struct xfs_sb		*sbp = &mp->m_sb;
1753 	char			label[XFSLABEL_MAX + 1];
1754 
1755 	/* Paranoia */
1756 	BUILD_BUG_ON(sizeof(sbp->sb_fname) > FSLABEL_MAX);
1757 
1758 	/* 1 larger than sb_fname, so this ensures a trailing NUL char */
1759 	memset(label, 0, sizeof(label));
1760 	spin_lock(&mp->m_sb_lock);
1761 	strncpy(label, sbp->sb_fname, XFSLABEL_MAX);
1762 	spin_unlock(&mp->m_sb_lock);
1763 
1764 	if (copy_to_user(user_label, label, sizeof(label)))
1765 		return -EFAULT;
1766 	return 0;
1767 }
1768 
1769 static int
1770 xfs_ioc_setlabel(
1771 	struct file		*filp,
1772 	struct xfs_mount	*mp,
1773 	char			__user *newlabel)
1774 {
1775 	struct xfs_sb		*sbp = &mp->m_sb;
1776 	char			label[XFSLABEL_MAX + 1];
1777 	size_t			len;
1778 	int			error;
1779 
1780 	if (!capable(CAP_SYS_ADMIN))
1781 		return -EPERM;
1782 	/*
1783 	 * The generic ioctl allows up to FSLABEL_MAX chars, but XFS is much
1784 	 * smaller, at 12 bytes.  We copy one more to be sure we find the
1785 	 * (required) NULL character to test the incoming label length.
1786 	 * NB: The on disk label doesn't need to be null terminated.
1787 	 */
1788 	if (copy_from_user(label, newlabel, XFSLABEL_MAX + 1))
1789 		return -EFAULT;
1790 	len = strnlen(label, XFSLABEL_MAX + 1);
1791 	if (len > sizeof(sbp->sb_fname))
1792 		return -EINVAL;
1793 
1794 	error = mnt_want_write_file(filp);
1795 	if (error)
1796 		return error;
1797 
1798 	spin_lock(&mp->m_sb_lock);
1799 	memset(sbp->sb_fname, 0, sizeof(sbp->sb_fname));
1800 	memcpy(sbp->sb_fname, label, len);
1801 	spin_unlock(&mp->m_sb_lock);
1802 
1803 	/*
1804 	 * Now we do several things to satisfy userspace.
1805 	 * In addition to normal logging of the primary superblock, we also
1806 	 * immediately write these changes to sector zero for the primary, then
1807 	 * update all backup supers (as xfs_db does for a label change), then
1808 	 * invalidate the block device page cache.  This is so that any prior
1809 	 * buffered reads from userspace (i.e. from blkid) are invalidated,
1810 	 * and userspace will see the newly-written label.
1811 	 */
1812 	error = xfs_sync_sb_buf(mp);
1813 	if (error)
1814 		goto out;
1815 	/*
1816 	 * growfs also updates backup supers so lock against that.
1817 	 */
1818 	mutex_lock(&mp->m_growlock);
1819 	error = xfs_update_secondary_sbs(mp);
1820 	mutex_unlock(&mp->m_growlock);
1821 
1822 	invalidate_bdev(mp->m_ddev_targp->bt_bdev);
1823 
1824 out:
1825 	mnt_drop_write_file(filp);
1826 	return error;
1827 }
1828 
1829 static inline int
1830 xfs_fs_eofblocks_from_user(
1831 	struct xfs_fs_eofblocks		*src,
1832 	struct xfs_icwalk		*dst)
1833 {
1834 	if (src->eof_version != XFS_EOFBLOCKS_VERSION)
1835 		return -EINVAL;
1836 
1837 	if (src->eof_flags & ~XFS_EOF_FLAGS_VALID)
1838 		return -EINVAL;
1839 
1840 	if (memchr_inv(&src->pad32, 0, sizeof(src->pad32)) ||
1841 	    memchr_inv(src->pad64, 0, sizeof(src->pad64)))
1842 		return -EINVAL;
1843 
1844 	dst->icw_flags = 0;
1845 	if (src->eof_flags & XFS_EOF_FLAGS_SYNC)
1846 		dst->icw_flags |= XFS_ICWALK_FLAG_SYNC;
1847 	if (src->eof_flags & XFS_EOF_FLAGS_UID)
1848 		dst->icw_flags |= XFS_ICWALK_FLAG_UID;
1849 	if (src->eof_flags & XFS_EOF_FLAGS_GID)
1850 		dst->icw_flags |= XFS_ICWALK_FLAG_GID;
1851 	if (src->eof_flags & XFS_EOF_FLAGS_PRID)
1852 		dst->icw_flags |= XFS_ICWALK_FLAG_PRID;
1853 	if (src->eof_flags & XFS_EOF_FLAGS_MINFILESIZE)
1854 		dst->icw_flags |= XFS_ICWALK_FLAG_MINFILESIZE;
1855 
1856 	dst->icw_prid = src->eof_prid;
1857 	dst->icw_min_file_size = src->eof_min_file_size;
1858 
1859 	dst->icw_uid = INVALID_UID;
1860 	if (src->eof_flags & XFS_EOF_FLAGS_UID) {
1861 		dst->icw_uid = make_kuid(current_user_ns(), src->eof_uid);
1862 		if (!uid_valid(dst->icw_uid))
1863 			return -EINVAL;
1864 	}
1865 
1866 	dst->icw_gid = INVALID_GID;
1867 	if (src->eof_flags & XFS_EOF_FLAGS_GID) {
1868 		dst->icw_gid = make_kgid(current_user_ns(), src->eof_gid);
1869 		if (!gid_valid(dst->icw_gid))
1870 			return -EINVAL;
1871 	}
1872 	return 0;
1873 }
1874 
1875 static int
1876 xfs_ioctl_getset_resblocks(
1877 	struct file		*filp,
1878 	unsigned int		cmd,
1879 	void __user		*arg)
1880 {
1881 	struct xfs_mount	*mp = XFS_I(file_inode(filp))->i_mount;
1882 	struct xfs_fsop_resblks	fsop = { };
1883 	int			error;
1884 
1885 	if (!capable(CAP_SYS_ADMIN))
1886 		return -EPERM;
1887 
1888 	if (cmd == XFS_IOC_SET_RESBLKS) {
1889 		if (xfs_is_readonly(mp))
1890 			return -EROFS;
1891 
1892 		if (copy_from_user(&fsop, arg, sizeof(fsop)))
1893 			return -EFAULT;
1894 
1895 		error = mnt_want_write_file(filp);
1896 		if (error)
1897 			return error;
1898 		error = xfs_reserve_blocks(mp, fsop.resblks);
1899 		mnt_drop_write_file(filp);
1900 		if (error)
1901 			return error;
1902 	}
1903 
1904 	spin_lock(&mp->m_sb_lock);
1905 	fsop.resblks = mp->m_resblks;
1906 	fsop.resblks_avail = mp->m_resblks_avail;
1907 	spin_unlock(&mp->m_sb_lock);
1908 
1909 	if (copy_to_user(arg, &fsop, sizeof(fsop)))
1910 		return -EFAULT;
1911 	return 0;
1912 }
1913 
1914 static int
1915 xfs_ioctl_fs_counts(
1916 	struct xfs_mount	*mp,
1917 	struct xfs_fsop_counts __user	*uarg)
1918 {
1919 	struct xfs_fsop_counts	out = {
1920 		.allocino = percpu_counter_read_positive(&mp->m_icount),
1921 		.freeino  = percpu_counter_read_positive(&mp->m_ifree),
1922 		.freedata = percpu_counter_read_positive(&mp->m_fdblocks) -
1923 				xfs_fdblocks_unavailable(mp),
1924 		.freertx  = percpu_counter_read_positive(&mp->m_frextents),
1925 	};
1926 
1927 	if (copy_to_user(uarg, &out, sizeof(out)))
1928 		return -EFAULT;
1929 	return 0;
1930 }
1931 
1932 /*
1933  * These long-unused ioctls were removed from the official ioctl API in 5.17,
1934  * but retain these definitions so that we can log warnings about them.
1935  */
1936 #define XFS_IOC_ALLOCSP		_IOW ('X', 10, struct xfs_flock64)
1937 #define XFS_IOC_FREESP		_IOW ('X', 11, struct xfs_flock64)
1938 #define XFS_IOC_ALLOCSP64	_IOW ('X', 36, struct xfs_flock64)
1939 #define XFS_IOC_FREESP64	_IOW ('X', 37, struct xfs_flock64)
1940 
1941 /*
1942  * Note: some of the ioctl's return positive numbers as a
1943  * byte count indicating success, such as readlink_by_handle.
1944  * So we don't "sign flip" like most other routines.  This means
1945  * true errors need to be returned as a negative value.
1946  */
1947 long
1948 xfs_file_ioctl(
1949 	struct file		*filp,
1950 	unsigned int		cmd,
1951 	unsigned long		p)
1952 {
1953 	struct inode		*inode = file_inode(filp);
1954 	struct xfs_inode	*ip = XFS_I(inode);
1955 	struct xfs_mount	*mp = ip->i_mount;
1956 	void			__user *arg = (void __user *)p;
1957 	int			error;
1958 
1959 	trace_xfs_file_ioctl(ip);
1960 
1961 	switch (cmd) {
1962 	case FITRIM:
1963 		return xfs_ioc_trim(mp, arg);
1964 	case FS_IOC_GETFSLABEL:
1965 		return xfs_ioc_getlabel(mp, arg);
1966 	case FS_IOC_SETFSLABEL:
1967 		return xfs_ioc_setlabel(filp, mp, arg);
1968 	case XFS_IOC_ALLOCSP:
1969 	case XFS_IOC_FREESP:
1970 	case XFS_IOC_ALLOCSP64:
1971 	case XFS_IOC_FREESP64:
1972 		xfs_warn_once(mp,
1973 	"%s should use fallocate; XFS_IOC_{ALLOC,FREE}SP ioctl unsupported",
1974 				current->comm);
1975 		return -ENOTTY;
1976 	case XFS_IOC_DIOINFO: {
1977 		struct xfs_buftarg	*target = xfs_inode_buftarg(ip);
1978 		struct dioattr		da;
1979 
1980 		da.d_mem =  da.d_miniosz = target->bt_logical_sectorsize;
1981 		da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1982 
1983 		if (copy_to_user(arg, &da, sizeof(da)))
1984 			return -EFAULT;
1985 		return 0;
1986 	}
1987 
1988 	case XFS_IOC_FSBULKSTAT_SINGLE:
1989 	case XFS_IOC_FSBULKSTAT:
1990 	case XFS_IOC_FSINUMBERS:
1991 		return xfs_ioc_fsbulkstat(filp, cmd, arg);
1992 
1993 	case XFS_IOC_BULKSTAT:
1994 		return xfs_ioc_bulkstat(filp, cmd, arg);
1995 	case XFS_IOC_INUMBERS:
1996 		return xfs_ioc_inumbers(mp, cmd, arg);
1997 
1998 	case XFS_IOC_FSGEOMETRY_V1:
1999 		return xfs_ioc_fsgeometry(mp, arg, 3);
2000 	case XFS_IOC_FSGEOMETRY_V4:
2001 		return xfs_ioc_fsgeometry(mp, arg, 4);
2002 	case XFS_IOC_FSGEOMETRY:
2003 		return xfs_ioc_fsgeometry(mp, arg, 5);
2004 
2005 	case XFS_IOC_AG_GEOMETRY:
2006 		return xfs_ioc_ag_geometry(mp, arg);
2007 
2008 	case XFS_IOC_GETVERSION:
2009 		return put_user(inode->i_generation, (int __user *)arg);
2010 
2011 	case XFS_IOC_FSGETXATTRA:
2012 		return xfs_ioc_fsgetxattra(ip, arg);
2013 
2014 	case XFS_IOC_GETBMAP:
2015 	case XFS_IOC_GETBMAPA:
2016 	case XFS_IOC_GETBMAPX:
2017 		return xfs_ioc_getbmap(filp, cmd, arg);
2018 
2019 	case FS_IOC_GETFSMAP:
2020 		return xfs_ioc_getfsmap(ip, arg);
2021 
2022 	case XFS_IOC_SCRUB_METADATA:
2023 		return xfs_ioc_scrub_metadata(filp, arg);
2024 
2025 	case XFS_IOC_FD_TO_HANDLE:
2026 	case XFS_IOC_PATH_TO_HANDLE:
2027 	case XFS_IOC_PATH_TO_FSHANDLE: {
2028 		xfs_fsop_handlereq_t	hreq;
2029 
2030 		if (copy_from_user(&hreq, arg, sizeof(hreq)))
2031 			return -EFAULT;
2032 		return xfs_find_handle(cmd, &hreq);
2033 	}
2034 	case XFS_IOC_OPEN_BY_HANDLE: {
2035 		xfs_fsop_handlereq_t	hreq;
2036 
2037 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
2038 			return -EFAULT;
2039 		return xfs_open_by_handle(filp, &hreq);
2040 	}
2041 
2042 	case XFS_IOC_READLINK_BY_HANDLE: {
2043 		xfs_fsop_handlereq_t	hreq;
2044 
2045 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
2046 			return -EFAULT;
2047 		return xfs_readlink_by_handle(filp, &hreq);
2048 	}
2049 	case XFS_IOC_ATTRLIST_BY_HANDLE:
2050 		return xfs_attrlist_by_handle(filp, arg);
2051 
2052 	case XFS_IOC_ATTRMULTI_BY_HANDLE:
2053 		return xfs_attrmulti_by_handle(filp, arg);
2054 
2055 	case XFS_IOC_SWAPEXT: {
2056 		struct xfs_swapext	sxp;
2057 
2058 		if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
2059 			return -EFAULT;
2060 		error = mnt_want_write_file(filp);
2061 		if (error)
2062 			return error;
2063 		error = xfs_ioc_swapext(&sxp);
2064 		mnt_drop_write_file(filp);
2065 		return error;
2066 	}
2067 
2068 	case XFS_IOC_FSCOUNTS:
2069 		return xfs_ioctl_fs_counts(mp, arg);
2070 
2071 	case XFS_IOC_SET_RESBLKS:
2072 	case XFS_IOC_GET_RESBLKS:
2073 		return xfs_ioctl_getset_resblocks(filp, cmd, arg);
2074 
2075 	case XFS_IOC_FSGROWFSDATA: {
2076 		struct xfs_growfs_data in;
2077 
2078 		if (copy_from_user(&in, arg, sizeof(in)))
2079 			return -EFAULT;
2080 
2081 		error = mnt_want_write_file(filp);
2082 		if (error)
2083 			return error;
2084 		error = xfs_growfs_data(mp, &in);
2085 		mnt_drop_write_file(filp);
2086 		return error;
2087 	}
2088 
2089 	case XFS_IOC_FSGROWFSLOG: {
2090 		struct xfs_growfs_log in;
2091 
2092 		if (copy_from_user(&in, arg, sizeof(in)))
2093 			return -EFAULT;
2094 
2095 		error = mnt_want_write_file(filp);
2096 		if (error)
2097 			return error;
2098 		error = xfs_growfs_log(mp, &in);
2099 		mnt_drop_write_file(filp);
2100 		return error;
2101 	}
2102 
2103 	case XFS_IOC_FSGROWFSRT: {
2104 		xfs_growfs_rt_t in;
2105 
2106 		if (copy_from_user(&in, arg, sizeof(in)))
2107 			return -EFAULT;
2108 
2109 		error = mnt_want_write_file(filp);
2110 		if (error)
2111 			return error;
2112 		error = xfs_growfs_rt(mp, &in);
2113 		mnt_drop_write_file(filp);
2114 		return error;
2115 	}
2116 
2117 	case XFS_IOC_GOINGDOWN: {
2118 		uint32_t in;
2119 
2120 		if (!capable(CAP_SYS_ADMIN))
2121 			return -EPERM;
2122 
2123 		if (get_user(in, (uint32_t __user *)arg))
2124 			return -EFAULT;
2125 
2126 		return xfs_fs_goingdown(mp, in);
2127 	}
2128 
2129 	case XFS_IOC_ERROR_INJECTION: {
2130 		xfs_error_injection_t in;
2131 
2132 		if (!capable(CAP_SYS_ADMIN))
2133 			return -EPERM;
2134 
2135 		if (copy_from_user(&in, arg, sizeof(in)))
2136 			return -EFAULT;
2137 
2138 		return xfs_errortag_add(mp, in.errtag);
2139 	}
2140 
2141 	case XFS_IOC_ERROR_CLEARALL:
2142 		if (!capable(CAP_SYS_ADMIN))
2143 			return -EPERM;
2144 
2145 		return xfs_errortag_clearall(mp);
2146 
2147 	case XFS_IOC_FREE_EOFBLOCKS: {
2148 		struct xfs_fs_eofblocks	eofb;
2149 		struct xfs_icwalk	icw;
2150 
2151 		if (!capable(CAP_SYS_ADMIN))
2152 			return -EPERM;
2153 
2154 		if (xfs_is_readonly(mp))
2155 			return -EROFS;
2156 
2157 		if (copy_from_user(&eofb, arg, sizeof(eofb)))
2158 			return -EFAULT;
2159 
2160 		error = xfs_fs_eofblocks_from_user(&eofb, &icw);
2161 		if (error)
2162 			return error;
2163 
2164 		trace_xfs_ioc_free_eofblocks(mp, &icw, _RET_IP_);
2165 
2166 		sb_start_write(mp->m_super);
2167 		error = xfs_blockgc_free_space(mp, &icw);
2168 		sb_end_write(mp->m_super);
2169 		return error;
2170 	}
2171 
2172 	default:
2173 		return -ENOTTY;
2174 	}
2175 }
2176