xref: /kernel/linux/linux-5.10/fs/xfs/xfs_ioctl.c (revision 8c2ecf20)
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
44/*
45 * xfs_find_handle maps from userspace xfs_fsop_handlereq structure to
46 * a file or fs handle.
47 *
48 * XFS_IOC_PATH_TO_FSHANDLE
49 *    returns fs handle for a mount point or path within that mount point
50 * XFS_IOC_FD_TO_HANDLE
51 *    returns full handle for a FD opened in user space
52 * XFS_IOC_PATH_TO_HANDLE
53 *    returns full handle for a path
54 */
55int
56xfs_find_handle(
57	unsigned int		cmd,
58	xfs_fsop_handlereq_t	*hreq)
59{
60	int			hsize;
61	xfs_handle_t		handle;
62	struct inode		*inode;
63	struct fd		f = {NULL};
64	struct path		path;
65	int			error;
66	struct xfs_inode	*ip;
67
68	if (cmd == XFS_IOC_FD_TO_HANDLE) {
69		f = fdget(hreq->fd);
70		if (!f.file)
71			return -EBADF;
72		inode = file_inode(f.file);
73	} else {
74		error = user_path_at(AT_FDCWD, hreq->path, 0, &path);
75		if (error)
76			return error;
77		inode = d_inode(path.dentry);
78	}
79	ip = XFS_I(inode);
80
81	/*
82	 * We can only generate handles for inodes residing on a XFS filesystem,
83	 * and only for regular files, directories or symbolic links.
84	 */
85	error = -EINVAL;
86	if (inode->i_sb->s_magic != XFS_SB_MAGIC)
87		goto out_put;
88
89	error = -EBADF;
90	if (!S_ISREG(inode->i_mode) &&
91	    !S_ISDIR(inode->i_mode) &&
92	    !S_ISLNK(inode->i_mode))
93		goto out_put;
94
95
96	memcpy(&handle.ha_fsid, ip->i_mount->m_fixedfsid, sizeof(xfs_fsid_t));
97
98	if (cmd == XFS_IOC_PATH_TO_FSHANDLE) {
99		/*
100		 * This handle only contains an fsid, zero the rest.
101		 */
102		memset(&handle.ha_fid, 0, sizeof(handle.ha_fid));
103		hsize = sizeof(xfs_fsid_t);
104	} else {
105		handle.ha_fid.fid_len = sizeof(xfs_fid_t) -
106					sizeof(handle.ha_fid.fid_len);
107		handle.ha_fid.fid_pad = 0;
108		handle.ha_fid.fid_gen = inode->i_generation;
109		handle.ha_fid.fid_ino = ip->i_ino;
110		hsize = sizeof(xfs_handle_t);
111	}
112
113	error = -EFAULT;
114	if (copy_to_user(hreq->ohandle, &handle, hsize) ||
115	    copy_to_user(hreq->ohandlen, &hsize, sizeof(__s32)))
116		goto out_put;
117
118	error = 0;
119
120 out_put:
121	if (cmd == XFS_IOC_FD_TO_HANDLE)
122		fdput(f);
123	else
124		path_put(&path);
125	return error;
126}
127
128/*
129 * No need to do permission checks on the various pathname components
130 * as the handle operations are privileged.
131 */
132STATIC int
133xfs_handle_acceptable(
134	void			*context,
135	struct dentry		*dentry)
136{
137	return 1;
138}
139
140/*
141 * Convert userspace handle data into a dentry.
142 */
143struct dentry *
144xfs_handle_to_dentry(
145	struct file		*parfilp,
146	void __user		*uhandle,
147	u32			hlen)
148{
149	xfs_handle_t		handle;
150	struct xfs_fid64	fid;
151
152	/*
153	 * Only allow handle opens under a directory.
154	 */
155	if (!S_ISDIR(file_inode(parfilp)->i_mode))
156		return ERR_PTR(-ENOTDIR);
157
158	if (hlen != sizeof(xfs_handle_t))
159		return ERR_PTR(-EINVAL);
160	if (copy_from_user(&handle, uhandle, hlen))
161		return ERR_PTR(-EFAULT);
162	if (handle.ha_fid.fid_len !=
163	    sizeof(handle.ha_fid) - sizeof(handle.ha_fid.fid_len))
164		return ERR_PTR(-EINVAL);
165
166	memset(&fid, 0, sizeof(struct fid));
167	fid.ino = handle.ha_fid.fid_ino;
168	fid.gen = handle.ha_fid.fid_gen;
169
170	return exportfs_decode_fh(parfilp->f_path.mnt, (struct fid *)&fid, 3,
171			FILEID_INO32_GEN | XFS_FILEID_TYPE_64FLAG,
172			xfs_handle_acceptable, NULL);
173}
174
175STATIC struct dentry *
176xfs_handlereq_to_dentry(
177	struct file		*parfilp,
178	xfs_fsop_handlereq_t	*hreq)
179{
180	return xfs_handle_to_dentry(parfilp, hreq->ihandle, hreq->ihandlen);
181}
182
183int
184xfs_open_by_handle(
185	struct file		*parfilp,
186	xfs_fsop_handlereq_t	*hreq)
187{
188	const struct cred	*cred = current_cred();
189	int			error;
190	int			fd;
191	int			permflag;
192	struct file		*filp;
193	struct inode		*inode;
194	struct dentry		*dentry;
195	fmode_t			fmode;
196	struct path		path;
197
198	if (!capable(CAP_SYS_ADMIN))
199		return -EPERM;
200
201	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
202	if (IS_ERR(dentry))
203		return PTR_ERR(dentry);
204	inode = d_inode(dentry);
205
206	/* Restrict xfs_open_by_handle to directories & regular files. */
207	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))) {
208		error = -EPERM;
209		goto out_dput;
210	}
211
212#if BITS_PER_LONG != 32
213	hreq->oflags |= O_LARGEFILE;
214#endif
215
216	permflag = hreq->oflags;
217	fmode = OPEN_FMODE(permflag);
218	if ((!(permflag & O_APPEND) || (permflag & O_TRUNC)) &&
219	    (fmode & FMODE_WRITE) && IS_APPEND(inode)) {
220		error = -EPERM;
221		goto out_dput;
222	}
223
224	if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
225		error = -EPERM;
226		goto out_dput;
227	}
228
229	/* Can't write directories. */
230	if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
231		error = -EISDIR;
232		goto out_dput;
233	}
234
235	fd = get_unused_fd_flags(0);
236	if (fd < 0) {
237		error = fd;
238		goto out_dput;
239	}
240
241	path.mnt = parfilp->f_path.mnt;
242	path.dentry = dentry;
243	filp = dentry_open(&path, hreq->oflags, cred);
244	dput(dentry);
245	if (IS_ERR(filp)) {
246		put_unused_fd(fd);
247		return PTR_ERR(filp);
248	}
249
250	if (S_ISREG(inode->i_mode)) {
251		filp->f_flags |= O_NOATIME;
252		filp->f_mode |= FMODE_NOCMTIME;
253	}
254
255	fd_install(fd, filp);
256	return fd;
257
258 out_dput:
259	dput(dentry);
260	return error;
261}
262
263int
264xfs_readlink_by_handle(
265	struct file		*parfilp,
266	xfs_fsop_handlereq_t	*hreq)
267{
268	struct dentry		*dentry;
269	__u32			olen;
270	int			error;
271
272	if (!capable(CAP_SYS_ADMIN))
273		return -EPERM;
274
275	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
276	if (IS_ERR(dentry))
277		return PTR_ERR(dentry);
278
279	/* Restrict this handle operation to symlinks only. */
280	if (!d_is_symlink(dentry)) {
281		error = -EINVAL;
282		goto out_dput;
283	}
284
285	if (copy_from_user(&olen, hreq->ohandlen, sizeof(__u32))) {
286		error = -EFAULT;
287		goto out_dput;
288	}
289
290	error = vfs_readlink(dentry, hreq->ohandle, olen);
291
292 out_dput:
293	dput(dentry);
294	return error;
295}
296
297/*
298 * Format an attribute and copy it out to the user's buffer.
299 * Take care to check values and protect against them changing later,
300 * we may be reading them directly out of a user buffer.
301 */
302static void
303xfs_ioc_attr_put_listent(
304	struct xfs_attr_list_context *context,
305	int			flags,
306	unsigned char		*name,
307	int			namelen,
308	int			valuelen)
309{
310	struct xfs_attrlist	*alist = context->buffer;
311	struct xfs_attrlist_ent	*aep;
312	int			arraytop;
313
314	ASSERT(!context->seen_enough);
315	ASSERT(context->count >= 0);
316	ASSERT(context->count < (ATTR_MAX_VALUELEN/8));
317	ASSERT(context->firstu >= sizeof(*alist));
318	ASSERT(context->firstu <= context->bufsize);
319
320	/*
321	 * Only list entries in the right namespace.
322	 */
323	if (context->attr_filter != (flags & XFS_ATTR_NSP_ONDISK_MASK))
324		return;
325
326	arraytop = sizeof(*alist) +
327			context->count * sizeof(alist->al_offset[0]);
328
329	/* decrement by the actual bytes used by the attr */
330	context->firstu -= round_up(offsetof(struct xfs_attrlist_ent, a_name) +
331			namelen + 1, sizeof(uint32_t));
332	if (context->firstu < arraytop) {
333		trace_xfs_attr_list_full(context);
334		alist->al_more = 1;
335		context->seen_enough = 1;
336		return;
337	}
338
339	aep = context->buffer + context->firstu;
340	aep->a_valuelen = valuelen;
341	memcpy(aep->a_name, name, namelen);
342	aep->a_name[namelen] = 0;
343	alist->al_offset[context->count++] = context->firstu;
344	alist->al_count = context->count;
345	trace_xfs_attr_list_add(context);
346}
347
348static unsigned int
349xfs_attr_filter(
350	u32			ioc_flags)
351{
352	if (ioc_flags & XFS_IOC_ATTR_ROOT)
353		return XFS_ATTR_ROOT;
354	if (ioc_flags & XFS_IOC_ATTR_SECURE)
355		return XFS_ATTR_SECURE;
356	return 0;
357}
358
359static unsigned int
360xfs_attr_flags(
361	u32			ioc_flags)
362{
363	if (ioc_flags & XFS_IOC_ATTR_CREATE)
364		return XATTR_CREATE;
365	if (ioc_flags & XFS_IOC_ATTR_REPLACE)
366		return XATTR_REPLACE;
367	return 0;
368}
369
370int
371xfs_ioc_attr_list(
372	struct xfs_inode		*dp,
373	void __user			*ubuf,
374	size_t				bufsize,
375	int				flags,
376	struct xfs_attrlist_cursor __user *ucursor)
377{
378	struct xfs_attr_list_context	context = { };
379	struct xfs_attrlist		*alist;
380	void				*buffer;
381	int				error;
382
383	if (bufsize < sizeof(struct xfs_attrlist) ||
384	    bufsize > XFS_XATTR_LIST_MAX)
385		return -EINVAL;
386
387	/*
388	 * Reject flags, only allow namespaces.
389	 */
390	if (flags & ~(XFS_IOC_ATTR_ROOT | XFS_IOC_ATTR_SECURE))
391		return -EINVAL;
392	if (flags == (XFS_IOC_ATTR_ROOT | XFS_IOC_ATTR_SECURE))
393		return -EINVAL;
394
395	/*
396	 * Validate the cursor.
397	 */
398	if (copy_from_user(&context.cursor, ucursor, sizeof(context.cursor)))
399		return -EFAULT;
400	if (context.cursor.pad1 || context.cursor.pad2)
401		return -EINVAL;
402	if (!context.cursor.initted &&
403	    (context.cursor.hashval || context.cursor.blkno ||
404	     context.cursor.offset))
405		return -EINVAL;
406
407	buffer = kvzalloc(bufsize, GFP_KERNEL);
408	if (!buffer)
409		return -ENOMEM;
410
411	/*
412	 * Initialize the output buffer.
413	 */
414	context.dp = dp;
415	context.resynch = 1;
416	context.attr_filter = xfs_attr_filter(flags);
417	context.buffer = buffer;
418	context.bufsize = round_down(bufsize, sizeof(uint32_t));
419	context.firstu = context.bufsize;
420	context.put_listent = xfs_ioc_attr_put_listent;
421
422	alist = context.buffer;
423	alist->al_count = 0;
424	alist->al_more = 0;
425	alist->al_offset[0] = context.bufsize;
426
427	error = xfs_attr_list(&context);
428	if (error)
429		goto out_free;
430
431	if (copy_to_user(ubuf, buffer, bufsize) ||
432	    copy_to_user(ucursor, &context.cursor, sizeof(context.cursor)))
433		error = -EFAULT;
434out_free:
435	kmem_free(buffer);
436	return error;
437}
438
439STATIC int
440xfs_attrlist_by_handle(
441	struct file		*parfilp,
442	struct xfs_fsop_attrlist_handlereq __user *p)
443{
444	struct xfs_fsop_attrlist_handlereq al_hreq;
445	struct dentry		*dentry;
446	int			error = -ENOMEM;
447
448	if (!capable(CAP_SYS_ADMIN))
449		return -EPERM;
450	if (copy_from_user(&al_hreq, p, sizeof(al_hreq)))
451		return -EFAULT;
452
453	dentry = xfs_handlereq_to_dentry(parfilp, &al_hreq.hreq);
454	if (IS_ERR(dentry))
455		return PTR_ERR(dentry);
456
457	error = xfs_ioc_attr_list(XFS_I(d_inode(dentry)), al_hreq.buffer,
458				  al_hreq.buflen, al_hreq.flags, &p->pos);
459	dput(dentry);
460	return error;
461}
462
463static int
464xfs_attrmulti_attr_get(
465	struct inode		*inode,
466	unsigned char		*name,
467	unsigned char		__user *ubuf,
468	uint32_t		*len,
469	uint32_t		flags)
470{
471	struct xfs_da_args	args = {
472		.dp		= XFS_I(inode),
473		.attr_filter	= xfs_attr_filter(flags),
474		.attr_flags	= xfs_attr_flags(flags),
475		.name		= name,
476		.namelen	= strlen(name),
477		.valuelen	= *len,
478	};
479	int			error;
480
481	if (*len > XFS_XATTR_SIZE_MAX)
482		return -EINVAL;
483
484	error = xfs_attr_get(&args);
485	if (error)
486		goto out_kfree;
487
488	*len = args.valuelen;
489	if (copy_to_user(ubuf, args.value, args.valuelen))
490		error = -EFAULT;
491
492out_kfree:
493	kmem_free(args.value);
494	return error;
495}
496
497static int
498xfs_attrmulti_attr_set(
499	struct inode		*inode,
500	unsigned char		*name,
501	const unsigned char	__user *ubuf,
502	uint32_t		len,
503	uint32_t		flags)
504{
505	struct xfs_da_args	args = {
506		.dp		= XFS_I(inode),
507		.attr_filter	= xfs_attr_filter(flags),
508		.attr_flags	= xfs_attr_flags(flags),
509		.name		= name,
510		.namelen	= strlen(name),
511	};
512	int			error;
513
514	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
515		return -EPERM;
516
517	if (ubuf) {
518		if (len > XFS_XATTR_SIZE_MAX)
519			return -EINVAL;
520		args.value = memdup_user(ubuf, len);
521		if (IS_ERR(args.value))
522			return PTR_ERR(args.value);
523		args.valuelen = len;
524	}
525
526	error = xfs_attr_set(&args);
527	if (!error && (flags & XFS_IOC_ATTR_ROOT))
528		xfs_forget_acl(inode, name);
529	kfree(args.value);
530	return error;
531}
532
533int
534xfs_ioc_attrmulti_one(
535	struct file		*parfilp,
536	struct inode		*inode,
537	uint32_t		opcode,
538	void __user		*uname,
539	void __user		*value,
540	uint32_t		*len,
541	uint32_t		flags)
542{
543	unsigned char		*name;
544	int			error;
545
546	if ((flags & XFS_IOC_ATTR_ROOT) && (flags & XFS_IOC_ATTR_SECURE))
547		return -EINVAL;
548
549	name = strndup_user(uname, MAXNAMELEN);
550	if (IS_ERR(name))
551		return PTR_ERR(name);
552
553	switch (opcode) {
554	case ATTR_OP_GET:
555		error = xfs_attrmulti_attr_get(inode, name, value, len, flags);
556		break;
557	case ATTR_OP_REMOVE:
558		value = NULL;
559		*len = 0;
560		/* fall through */
561	case ATTR_OP_SET:
562		error = mnt_want_write_file(parfilp);
563		if (error)
564			break;
565		error = xfs_attrmulti_attr_set(inode, name, value, *len, flags);
566		mnt_drop_write_file(parfilp);
567		break;
568	default:
569		error = -EINVAL;
570		break;
571	}
572
573	kfree(name);
574	return error;
575}
576
577STATIC int
578xfs_attrmulti_by_handle(
579	struct file		*parfilp,
580	void			__user *arg)
581{
582	int			error;
583	xfs_attr_multiop_t	*ops;
584	xfs_fsop_attrmulti_handlereq_t am_hreq;
585	struct dentry		*dentry;
586	unsigned int		i, size;
587
588	if (!capable(CAP_SYS_ADMIN))
589		return -EPERM;
590	if (copy_from_user(&am_hreq, arg, sizeof(xfs_fsop_attrmulti_handlereq_t)))
591		return -EFAULT;
592
593	/* overflow check */
594	if (am_hreq.opcount >= INT_MAX / sizeof(xfs_attr_multiop_t))
595		return -E2BIG;
596
597	dentry = xfs_handlereq_to_dentry(parfilp, &am_hreq.hreq);
598	if (IS_ERR(dentry))
599		return PTR_ERR(dentry);
600
601	error = -E2BIG;
602	size = am_hreq.opcount * sizeof(xfs_attr_multiop_t);
603	if (!size || size > 16 * PAGE_SIZE)
604		goto out_dput;
605
606	ops = memdup_user(am_hreq.ops, size);
607	if (IS_ERR(ops)) {
608		error = PTR_ERR(ops);
609		goto out_dput;
610	}
611
612	error = 0;
613	for (i = 0; i < am_hreq.opcount; i++) {
614		ops[i].am_error = xfs_ioc_attrmulti_one(parfilp,
615				d_inode(dentry), ops[i].am_opcode,
616				ops[i].am_attrname, ops[i].am_attrvalue,
617				&ops[i].am_length, ops[i].am_flags);
618	}
619
620	if (copy_to_user(am_hreq.ops, ops, size))
621		error = -EFAULT;
622
623	kfree(ops);
624 out_dput:
625	dput(dentry);
626	return error;
627}
628
629int
630xfs_ioc_space(
631	struct file		*filp,
632	xfs_flock64_t		*bf)
633{
634	struct inode		*inode = file_inode(filp);
635	struct xfs_inode	*ip = XFS_I(inode);
636	struct iattr		iattr;
637	enum xfs_prealloc_flags	flags = XFS_PREALLOC_CLEAR;
638	uint			iolock = XFS_IOLOCK_EXCL | XFS_MMAPLOCK_EXCL;
639	int			error;
640
641	if (inode->i_flags & (S_IMMUTABLE|S_APPEND))
642		return -EPERM;
643
644	if (!(filp->f_mode & FMODE_WRITE))
645		return -EBADF;
646
647	if (!S_ISREG(inode->i_mode))
648		return -EINVAL;
649
650	if (xfs_is_always_cow_inode(ip))
651		return -EOPNOTSUPP;
652
653	if (filp->f_flags & O_DSYNC)
654		flags |= XFS_PREALLOC_SYNC;
655	if (filp->f_mode & FMODE_NOCMTIME)
656		flags |= XFS_PREALLOC_INVISIBLE;
657
658	error = mnt_want_write_file(filp);
659	if (error)
660		return error;
661
662	xfs_ilock(ip, iolock);
663	error = xfs_break_layouts(inode, &iolock, BREAK_UNMAP);
664	if (error)
665		goto out_unlock;
666	inode_dio_wait(inode);
667
668	switch (bf->l_whence) {
669	case 0: /*SEEK_SET*/
670		break;
671	case 1: /*SEEK_CUR*/
672		bf->l_start += filp->f_pos;
673		break;
674	case 2: /*SEEK_END*/
675		bf->l_start += XFS_ISIZE(ip);
676		break;
677	default:
678		error = -EINVAL;
679		goto out_unlock;
680	}
681
682	if (bf->l_start < 0 || bf->l_start > inode->i_sb->s_maxbytes) {
683		error = -EINVAL;
684		goto out_unlock;
685	}
686
687	if (bf->l_start > XFS_ISIZE(ip)) {
688		error = xfs_alloc_file_space(ip, XFS_ISIZE(ip),
689				bf->l_start - XFS_ISIZE(ip),
690				XFS_BMAPI_PREALLOC);
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_dentry(filp), &iattr);
698	if (error)
699		goto out_unlock;
700
701	error = xfs_update_prealloc_flags(ip, flags);
702
703out_unlock:
704	xfs_iunlock(ip, iolock);
705	mnt_drop_write_file(filp);
706	return error;
707}
708
709/* Return 0 on success or positive error */
710int
711xfs_fsbulkstat_one_fmt(
712	struct xfs_ibulk		*breq,
713	const struct xfs_bulkstat	*bstat)
714{
715	struct xfs_bstat		bs1;
716
717	xfs_bulkstat_to_bstat(breq->mp, &bs1, bstat);
718	if (copy_to_user(breq->ubuffer, &bs1, sizeof(bs1)))
719		return -EFAULT;
720	return xfs_ibulk_advance(breq, sizeof(struct xfs_bstat));
721}
722
723int
724xfs_fsinumbers_fmt(
725	struct xfs_ibulk		*breq,
726	const struct xfs_inumbers	*igrp)
727{
728	struct xfs_inogrp		ig1;
729
730	xfs_inumbers_to_inogrp(&ig1, igrp);
731	if (copy_to_user(breq->ubuffer, &ig1, sizeof(struct xfs_inogrp)))
732		return -EFAULT;
733	return xfs_ibulk_advance(breq, sizeof(struct xfs_inogrp));
734}
735
736STATIC int
737xfs_ioc_fsbulkstat(
738	xfs_mount_t		*mp,
739	unsigned int		cmd,
740	void			__user *arg)
741{
742	struct xfs_fsop_bulkreq	bulkreq;
743	struct xfs_ibulk	breq = {
744		.mp		= mp,
745		.ocount		= 0,
746	};
747	xfs_ino_t		lastino;
748	int			error;
749
750	/* done = 1 if there are more stats to get and if bulkstat */
751	/* should be called again (unused here, but used in dmapi) */
752
753	if (!capable(CAP_SYS_ADMIN))
754		return -EPERM;
755
756	if (XFS_FORCED_SHUTDOWN(mp))
757		return -EIO;
758
759	if (copy_from_user(&bulkreq, arg, sizeof(struct xfs_fsop_bulkreq)))
760		return -EFAULT;
761
762	if (copy_from_user(&lastino, bulkreq.lastip, sizeof(__s64)))
763		return -EFAULT;
764
765	if (bulkreq.icount <= 0)
766		return -EINVAL;
767
768	if (bulkreq.ubuffer == NULL)
769		return -EINVAL;
770
771	breq.ubuffer = bulkreq.ubuffer;
772	breq.icount = bulkreq.icount;
773
774	/*
775	 * FSBULKSTAT_SINGLE expects that *lastip contains the inode number
776	 * that we want to stat.  However, FSINUMBERS and FSBULKSTAT expect
777	 * that *lastip contains either zero or the number of the last inode to
778	 * be examined by the previous call and return results starting with
779	 * the next inode after that.  The new bulk request back end functions
780	 * take the inode to start with, so we have to compute the startino
781	 * parameter from lastino to maintain correct function.  lastino == 0
782	 * is a special case because it has traditionally meant "first inode
783	 * in filesystem".
784	 */
785	if (cmd == XFS_IOC_FSINUMBERS) {
786		breq.startino = lastino ? lastino + 1 : 0;
787		error = xfs_inumbers(&breq, xfs_fsinumbers_fmt);
788		lastino = breq.startino - 1;
789	} else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE) {
790		breq.startino = lastino;
791		breq.icount = 1;
792		error = xfs_bulkstat_one(&breq, xfs_fsbulkstat_one_fmt);
793	} else {	/* XFS_IOC_FSBULKSTAT */
794		breq.startino = lastino ? lastino + 1 : 0;
795		error = xfs_bulkstat(&breq, xfs_fsbulkstat_one_fmt);
796		lastino = breq.startino - 1;
797	}
798
799	if (error)
800		return error;
801
802	if (bulkreq.lastip != NULL &&
803	    copy_to_user(bulkreq.lastip, &lastino, sizeof(xfs_ino_t)))
804		return -EFAULT;
805
806	if (bulkreq.ocount != NULL &&
807	    copy_to_user(bulkreq.ocount, &breq.ocount, sizeof(__s32)))
808		return -EFAULT;
809
810	return 0;
811}
812
813/* Return 0 on success or positive error */
814static int
815xfs_bulkstat_fmt(
816	struct xfs_ibulk		*breq,
817	const struct xfs_bulkstat	*bstat)
818{
819	if (copy_to_user(breq->ubuffer, bstat, sizeof(struct xfs_bulkstat)))
820		return -EFAULT;
821	return xfs_ibulk_advance(breq, sizeof(struct xfs_bulkstat));
822}
823
824/*
825 * Check the incoming bulk request @hdr from userspace and initialize the
826 * internal @breq bulk request appropriately.  Returns 0 if the bulk request
827 * should proceed; -ECANCELED if there's nothing to do; or the usual
828 * negative error code.
829 */
830static int
831xfs_bulk_ireq_setup(
832	struct xfs_mount	*mp,
833	struct xfs_bulk_ireq	*hdr,
834	struct xfs_ibulk	*breq,
835	void __user		*ubuffer)
836{
837	if (hdr->icount == 0 ||
838	    (hdr->flags & ~XFS_BULK_IREQ_FLAGS_ALL) ||
839	    memchr_inv(hdr->reserved, 0, sizeof(hdr->reserved)))
840		return -EINVAL;
841
842	breq->startino = hdr->ino;
843	breq->ubuffer = ubuffer;
844	breq->icount = hdr->icount;
845	breq->ocount = 0;
846	breq->flags = 0;
847
848	/*
849	 * The @ino parameter is a special value, so we must look it up here.
850	 * We're not allowed to have IREQ_AGNO, and we only return one inode
851	 * worth of data.
852	 */
853	if (hdr->flags & XFS_BULK_IREQ_SPECIAL) {
854		if (hdr->flags & XFS_BULK_IREQ_AGNO)
855			return -EINVAL;
856
857		switch (hdr->ino) {
858		case XFS_BULK_IREQ_SPECIAL_ROOT:
859			hdr->ino = mp->m_sb.sb_rootino;
860			break;
861		default:
862			return -EINVAL;
863		}
864		breq->icount = 1;
865	}
866
867	/*
868	 * The IREQ_AGNO flag means that we only want results from a given AG.
869	 * If @hdr->ino is zero, we start iterating in that AG.  If @hdr->ino is
870	 * beyond the specified AG then we return no results.
871	 */
872	if (hdr->flags & XFS_BULK_IREQ_AGNO) {
873		if (hdr->agno >= mp->m_sb.sb_agcount)
874			return -EINVAL;
875
876		if (breq->startino == 0)
877			breq->startino = XFS_AGINO_TO_INO(mp, hdr->agno, 0);
878		else if (XFS_INO_TO_AGNO(mp, breq->startino) < hdr->agno)
879			return -EINVAL;
880
881		breq->flags |= XFS_IBULK_SAME_AG;
882
883		/* Asking for an inode past the end of the AG?  We're done! */
884		if (XFS_INO_TO_AGNO(mp, breq->startino) > hdr->agno)
885			return -ECANCELED;
886	} else if (hdr->agno)
887		return -EINVAL;
888
889	/* Asking for an inode past the end of the FS?  We're done! */
890	if (XFS_INO_TO_AGNO(mp, breq->startino) >= mp->m_sb.sb_agcount)
891		return -ECANCELED;
892
893	return 0;
894}
895
896/*
897 * Update the userspace bulk request @hdr to reflect the end state of the
898 * internal bulk request @breq.
899 */
900static void
901xfs_bulk_ireq_teardown(
902	struct xfs_bulk_ireq	*hdr,
903	struct xfs_ibulk	*breq)
904{
905	hdr->ino = breq->startino;
906	hdr->ocount = breq->ocount;
907}
908
909/* Handle the v5 bulkstat ioctl. */
910STATIC int
911xfs_ioc_bulkstat(
912	struct xfs_mount		*mp,
913	unsigned int			cmd,
914	struct xfs_bulkstat_req __user	*arg)
915{
916	struct xfs_bulk_ireq		hdr;
917	struct xfs_ibulk		breq = {
918		.mp			= mp,
919	};
920	int				error;
921
922	if (!capable(CAP_SYS_ADMIN))
923		return -EPERM;
924
925	if (XFS_FORCED_SHUTDOWN(mp))
926		return -EIO;
927
928	if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
929		return -EFAULT;
930
931	error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->bulkstat);
932	if (error == -ECANCELED)
933		goto out_teardown;
934	if (error < 0)
935		return error;
936
937	error = xfs_bulkstat(&breq, xfs_bulkstat_fmt);
938	if (error)
939		return error;
940
941out_teardown:
942	xfs_bulk_ireq_teardown(&hdr, &breq);
943	if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
944		return -EFAULT;
945
946	return 0;
947}
948
949STATIC int
950xfs_inumbers_fmt(
951	struct xfs_ibulk		*breq,
952	const struct xfs_inumbers	*igrp)
953{
954	if (copy_to_user(breq->ubuffer, igrp, sizeof(struct xfs_inumbers)))
955		return -EFAULT;
956	return xfs_ibulk_advance(breq, sizeof(struct xfs_inumbers));
957}
958
959/* Handle the v5 inumbers ioctl. */
960STATIC int
961xfs_ioc_inumbers(
962	struct xfs_mount		*mp,
963	unsigned int			cmd,
964	struct xfs_inumbers_req __user	*arg)
965{
966	struct xfs_bulk_ireq		hdr;
967	struct xfs_ibulk		breq = {
968		.mp			= mp,
969	};
970	int				error;
971
972	if (!capable(CAP_SYS_ADMIN))
973		return -EPERM;
974
975	if (XFS_FORCED_SHUTDOWN(mp))
976		return -EIO;
977
978	if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
979		return -EFAULT;
980
981	error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->inumbers);
982	if (error == -ECANCELED)
983		goto out_teardown;
984	if (error < 0)
985		return error;
986
987	error = xfs_inumbers(&breq, xfs_inumbers_fmt);
988	if (error)
989		return error;
990
991out_teardown:
992	xfs_bulk_ireq_teardown(&hdr, &breq);
993	if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
994		return -EFAULT;
995
996	return 0;
997}
998
999STATIC int
1000xfs_ioc_fsgeometry(
1001	struct xfs_mount	*mp,
1002	void			__user *arg,
1003	int			struct_version)
1004{
1005	struct xfs_fsop_geom	fsgeo;
1006	size_t			len;
1007
1008	xfs_fs_geometry(&mp->m_sb, &fsgeo, struct_version);
1009
1010	if (struct_version <= 3)
1011		len = sizeof(struct xfs_fsop_geom_v1);
1012	else if (struct_version == 4)
1013		len = sizeof(struct xfs_fsop_geom_v4);
1014	else {
1015		xfs_fsop_geom_health(mp, &fsgeo);
1016		len = sizeof(fsgeo);
1017	}
1018
1019	if (copy_to_user(arg, &fsgeo, len))
1020		return -EFAULT;
1021	return 0;
1022}
1023
1024STATIC int
1025xfs_ioc_ag_geometry(
1026	struct xfs_mount	*mp,
1027	void			__user *arg)
1028{
1029	struct xfs_ag_geometry	ageo;
1030	int			error;
1031
1032	if (copy_from_user(&ageo, arg, sizeof(ageo)))
1033		return -EFAULT;
1034	if (ageo.ag_flags)
1035		return -EINVAL;
1036	if (memchr_inv(&ageo.ag_reserved, 0, sizeof(ageo.ag_reserved)))
1037		return -EINVAL;
1038
1039	error = xfs_ag_get_geometry(mp, ageo.ag_number, &ageo);
1040	if (error)
1041		return error;
1042
1043	if (copy_to_user(arg, &ageo, sizeof(ageo)))
1044		return -EFAULT;
1045	return 0;
1046}
1047
1048/*
1049 * Linux extended inode flags interface.
1050 */
1051
1052STATIC unsigned int
1053xfs_merge_ioc_xflags(
1054	unsigned int	flags,
1055	unsigned int	start)
1056{
1057	unsigned int	xflags = start;
1058
1059	if (flags & FS_IMMUTABLE_FL)
1060		xflags |= FS_XFLAG_IMMUTABLE;
1061	else
1062		xflags &= ~FS_XFLAG_IMMUTABLE;
1063	if (flags & FS_APPEND_FL)
1064		xflags |= FS_XFLAG_APPEND;
1065	else
1066		xflags &= ~FS_XFLAG_APPEND;
1067	if (flags & FS_SYNC_FL)
1068		xflags |= FS_XFLAG_SYNC;
1069	else
1070		xflags &= ~FS_XFLAG_SYNC;
1071	if (flags & FS_NOATIME_FL)
1072		xflags |= FS_XFLAG_NOATIME;
1073	else
1074		xflags &= ~FS_XFLAG_NOATIME;
1075	if (flags & FS_NODUMP_FL)
1076		xflags |= FS_XFLAG_NODUMP;
1077	else
1078		xflags &= ~FS_XFLAG_NODUMP;
1079	if (flags & FS_DAX_FL)
1080		xflags |= FS_XFLAG_DAX;
1081	else
1082		xflags &= ~FS_XFLAG_DAX;
1083
1084	return xflags;
1085}
1086
1087STATIC unsigned int
1088xfs_di2lxflags(
1089	uint16_t	di_flags,
1090	uint64_t	di_flags2)
1091{
1092	unsigned int	flags = 0;
1093
1094	if (di_flags & XFS_DIFLAG_IMMUTABLE)
1095		flags |= FS_IMMUTABLE_FL;
1096	if (di_flags & XFS_DIFLAG_APPEND)
1097		flags |= FS_APPEND_FL;
1098	if (di_flags & XFS_DIFLAG_SYNC)
1099		flags |= FS_SYNC_FL;
1100	if (di_flags & XFS_DIFLAG_NOATIME)
1101		flags |= FS_NOATIME_FL;
1102	if (di_flags & XFS_DIFLAG_NODUMP)
1103		flags |= FS_NODUMP_FL;
1104	if (di_flags2 & XFS_DIFLAG2_DAX) {
1105		flags |= FS_DAX_FL;
1106	}
1107	return flags;
1108}
1109
1110static void
1111xfs_fill_fsxattr(
1112	struct xfs_inode	*ip,
1113	bool			attr,
1114	struct fsxattr		*fa)
1115{
1116	struct xfs_ifork	*ifp = attr ? ip->i_afp : &ip->i_df;
1117
1118	simple_fill_fsxattr(fa, xfs_ip2xflags(ip));
1119	fa->fsx_extsize = ip->i_d.di_extsize << ip->i_mount->m_sb.sb_blocklog;
1120	fa->fsx_cowextsize = ip->i_d.di_cowextsize <<
1121			ip->i_mount->m_sb.sb_blocklog;
1122	fa->fsx_projid = ip->i_d.di_projid;
1123	if (ifp && (ifp->if_flags & XFS_IFEXTENTS))
1124		fa->fsx_nextents = xfs_iext_count(ifp);
1125	else
1126		fa->fsx_nextents = xfs_ifork_nextents(ifp);
1127}
1128
1129STATIC int
1130xfs_ioc_fsgetxattr(
1131	xfs_inode_t		*ip,
1132	int			attr,
1133	void			__user *arg)
1134{
1135	struct fsxattr		fa;
1136
1137	xfs_ilock(ip, XFS_ILOCK_SHARED);
1138	xfs_fill_fsxattr(ip, attr, &fa);
1139	xfs_iunlock(ip, XFS_ILOCK_SHARED);
1140
1141	if (copy_to_user(arg, &fa, sizeof(fa)))
1142		return -EFAULT;
1143	return 0;
1144}
1145
1146STATIC uint16_t
1147xfs_flags2diflags(
1148	struct xfs_inode	*ip,
1149	unsigned int		xflags)
1150{
1151	/* can't set PREALLOC this way, just preserve it */
1152	uint16_t		di_flags =
1153		(ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
1154
1155	if (xflags & FS_XFLAG_IMMUTABLE)
1156		di_flags |= XFS_DIFLAG_IMMUTABLE;
1157	if (xflags & FS_XFLAG_APPEND)
1158		di_flags |= XFS_DIFLAG_APPEND;
1159	if (xflags & FS_XFLAG_SYNC)
1160		di_flags |= XFS_DIFLAG_SYNC;
1161	if (xflags & FS_XFLAG_NOATIME)
1162		di_flags |= XFS_DIFLAG_NOATIME;
1163	if (xflags & FS_XFLAG_NODUMP)
1164		di_flags |= XFS_DIFLAG_NODUMP;
1165	if (xflags & FS_XFLAG_NODEFRAG)
1166		di_flags |= XFS_DIFLAG_NODEFRAG;
1167	if (xflags & FS_XFLAG_FILESTREAM)
1168		di_flags |= XFS_DIFLAG_FILESTREAM;
1169	if (S_ISDIR(VFS_I(ip)->i_mode)) {
1170		if (xflags & FS_XFLAG_RTINHERIT)
1171			di_flags |= XFS_DIFLAG_RTINHERIT;
1172		if (xflags & FS_XFLAG_NOSYMLINKS)
1173			di_flags |= XFS_DIFLAG_NOSYMLINKS;
1174		if (xflags & FS_XFLAG_EXTSZINHERIT)
1175			di_flags |= XFS_DIFLAG_EXTSZINHERIT;
1176		if (xflags & FS_XFLAG_PROJINHERIT)
1177			di_flags |= XFS_DIFLAG_PROJINHERIT;
1178	} else if (S_ISREG(VFS_I(ip)->i_mode)) {
1179		if (xflags & FS_XFLAG_REALTIME)
1180			di_flags |= XFS_DIFLAG_REALTIME;
1181		if (xflags & FS_XFLAG_EXTSIZE)
1182			di_flags |= XFS_DIFLAG_EXTSIZE;
1183	}
1184
1185	return di_flags;
1186}
1187
1188STATIC uint64_t
1189xfs_flags2diflags2(
1190	struct xfs_inode	*ip,
1191	unsigned int		xflags)
1192{
1193	uint64_t		di_flags2 =
1194		(ip->i_d.di_flags2 & (XFS_DIFLAG2_REFLINK |
1195				      XFS_DIFLAG2_BIGTIME));
1196
1197	if (xflags & FS_XFLAG_DAX)
1198		di_flags2 |= XFS_DIFLAG2_DAX;
1199	if (xflags & FS_XFLAG_COWEXTSIZE)
1200		di_flags2 |= XFS_DIFLAG2_COWEXTSIZE;
1201
1202	return di_flags2;
1203}
1204
1205static int
1206xfs_ioctl_setattr_xflags(
1207	struct xfs_trans	*tp,
1208	struct xfs_inode	*ip,
1209	struct fsxattr		*fa)
1210{
1211	struct xfs_mount	*mp = ip->i_mount;
1212	uint64_t		di_flags2;
1213
1214	/* Can't change realtime flag if any extents are allocated. */
1215	if ((ip->i_df.if_nextents || ip->i_delayed_blks) &&
1216	    XFS_IS_REALTIME_INODE(ip) != (fa->fsx_xflags & FS_XFLAG_REALTIME))
1217		return -EINVAL;
1218
1219	/* If realtime flag is set then must have realtime device */
1220	if (fa->fsx_xflags & FS_XFLAG_REALTIME) {
1221		if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
1222		    (ip->i_d.di_extsize % mp->m_sb.sb_rextsize))
1223			return -EINVAL;
1224	}
1225
1226	/* Clear reflink if we are actually able to set the rt flag. */
1227	if ((fa->fsx_xflags & FS_XFLAG_REALTIME) && xfs_is_reflink_inode(ip))
1228		ip->i_d.di_flags2 &= ~XFS_DIFLAG2_REFLINK;
1229
1230	/* Don't allow us to set DAX mode for a reflinked file for now. */
1231	if ((fa->fsx_xflags & FS_XFLAG_DAX) && xfs_is_reflink_inode(ip))
1232		return -EINVAL;
1233
1234	/* diflags2 only valid for v3 inodes. */
1235	di_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1236	if (di_flags2 && !xfs_sb_version_has_v3inode(&mp->m_sb))
1237		return -EINVAL;
1238
1239	ip->i_d.di_flags = xfs_flags2diflags(ip, fa->fsx_xflags);
1240	ip->i_d.di_flags2 = di_flags2;
1241
1242	xfs_diflags_to_iflags(ip, false);
1243	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1244	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1245	XFS_STATS_INC(mp, xs_ig_attrchg);
1246	return 0;
1247}
1248
1249static void
1250xfs_ioctl_setattr_prepare_dax(
1251	struct xfs_inode	*ip,
1252	struct fsxattr		*fa)
1253{
1254	struct xfs_mount	*mp = ip->i_mount;
1255	struct inode            *inode = VFS_I(ip);
1256
1257	if (S_ISDIR(inode->i_mode))
1258		return;
1259
1260	if ((mp->m_flags & XFS_MOUNT_DAX_ALWAYS) ||
1261	    (mp->m_flags & XFS_MOUNT_DAX_NEVER))
1262		return;
1263
1264	if (((fa->fsx_xflags & FS_XFLAG_DAX) &&
1265	    !(ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)) ||
1266	    (!(fa->fsx_xflags & FS_XFLAG_DAX) &&
1267	     (ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)))
1268		d_mark_dontcache(inode);
1269}
1270
1271/*
1272 * Set up the transaction structure for the setattr operation, checking that we
1273 * have permission to do so. On success, return a clean transaction and the
1274 * inode locked exclusively ready for further operation specific checks. On
1275 * failure, return an error without modifying or locking the inode.
1276 */
1277static struct xfs_trans *
1278xfs_ioctl_setattr_get_trans(
1279	struct xfs_inode	*ip)
1280{
1281	struct xfs_mount	*mp = ip->i_mount;
1282	struct xfs_trans	*tp;
1283	int			error = -EROFS;
1284
1285	if (mp->m_flags & XFS_MOUNT_RDONLY)
1286		goto out_unlock;
1287	error = -EIO;
1288	if (XFS_FORCED_SHUTDOWN(mp))
1289		goto out_unlock;
1290
1291	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
1292	if (error)
1293		goto out_unlock;
1294
1295	xfs_ilock(ip, XFS_ILOCK_EXCL);
1296	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
1297
1298	/*
1299	 * CAP_FOWNER overrides the following restrictions:
1300	 *
1301	 * The user ID of the calling process must be equal to the file owner
1302	 * ID, except in cases where the CAP_FSETID capability is applicable.
1303	 */
1304	if (!inode_owner_or_capable(VFS_I(ip))) {
1305		error = -EPERM;
1306		goto out_cancel;
1307	}
1308
1309	if (mp->m_flags & XFS_MOUNT_WSYNC)
1310		xfs_trans_set_sync(tp);
1311
1312	return tp;
1313
1314out_cancel:
1315	xfs_trans_cancel(tp);
1316out_unlock:
1317	return ERR_PTR(error);
1318}
1319
1320/*
1321 * extent size hint validation is somewhat cumbersome. Rules are:
1322 *
1323 * 1. extent size hint is only valid for directories and regular files
1324 * 2. FS_XFLAG_EXTSIZE is only valid for regular files
1325 * 3. FS_XFLAG_EXTSZINHERIT is only valid for directories.
1326 * 4. can only be changed on regular files if no extents are allocated
1327 * 5. can be changed on directories at any time
1328 * 6. extsize hint of 0 turns off hints, clears inode flags.
1329 * 7. Extent size must be a multiple of the appropriate block size.
1330 * 8. for non-realtime files, the extent size hint must be limited
1331 *    to half the AG size to avoid alignment extending the extent beyond the
1332 *    limits of the AG.
1333 *
1334 * Please keep this function in sync with xfs_scrub_inode_extsize.
1335 */
1336static int
1337xfs_ioctl_setattr_check_extsize(
1338	struct xfs_inode	*ip,
1339	struct fsxattr		*fa)
1340{
1341	struct xfs_mount	*mp = ip->i_mount;
1342	xfs_extlen_t		size;
1343	xfs_fsblock_t		extsize_fsb;
1344
1345	if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_df.if_nextents &&
1346	    ((ip->i_d.di_extsize << mp->m_sb.sb_blocklog) != fa->fsx_extsize))
1347		return -EINVAL;
1348
1349	if (fa->fsx_extsize == 0)
1350		return 0;
1351
1352	extsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1353	if (extsize_fsb > MAXEXTLEN)
1354		return -EINVAL;
1355
1356	if (XFS_IS_REALTIME_INODE(ip) ||
1357	    (fa->fsx_xflags & FS_XFLAG_REALTIME)) {
1358		size = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
1359	} else {
1360		size = mp->m_sb.sb_blocksize;
1361		if (extsize_fsb > mp->m_sb.sb_agblocks / 2)
1362			return -EINVAL;
1363	}
1364
1365	if (fa->fsx_extsize % size)
1366		return -EINVAL;
1367
1368	return 0;
1369}
1370
1371/*
1372 * CoW extent size hint validation rules are:
1373 *
1374 * 1. CoW extent size hint can only be set if reflink is enabled on the fs.
1375 *    The inode does not have to have any shared blocks, but it must be a v3.
1376 * 2. FS_XFLAG_COWEXTSIZE is only valid for directories and regular files;
1377 *    for a directory, the hint is propagated to new files.
1378 * 3. Can be changed on files & directories at any time.
1379 * 4. CoW extsize hint of 0 turns off hints, clears inode flags.
1380 * 5. Extent size must be a multiple of the appropriate block size.
1381 * 6. The extent size hint must be limited to half the AG size to avoid
1382 *    alignment extending the extent beyond the limits of the AG.
1383 *
1384 * Please keep this function in sync with xfs_scrub_inode_cowextsize.
1385 */
1386static int
1387xfs_ioctl_setattr_check_cowextsize(
1388	struct xfs_inode	*ip,
1389	struct fsxattr		*fa)
1390{
1391	struct xfs_mount	*mp = ip->i_mount;
1392	xfs_extlen_t		size;
1393	xfs_fsblock_t		cowextsize_fsb;
1394
1395	if (!(fa->fsx_xflags & FS_XFLAG_COWEXTSIZE))
1396		return 0;
1397
1398	if (!xfs_sb_version_hasreflink(&ip->i_mount->m_sb))
1399		return -EINVAL;
1400
1401	if (fa->fsx_cowextsize == 0)
1402		return 0;
1403
1404	cowextsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
1405	if (cowextsize_fsb > MAXEXTLEN)
1406		return -EINVAL;
1407
1408	size = mp->m_sb.sb_blocksize;
1409	if (cowextsize_fsb > mp->m_sb.sb_agblocks / 2)
1410		return -EINVAL;
1411
1412	if (fa->fsx_cowextsize % size)
1413		return -EINVAL;
1414
1415	return 0;
1416}
1417
1418static int
1419xfs_ioctl_setattr_check_projid(
1420	struct xfs_inode	*ip,
1421	struct fsxattr		*fa)
1422{
1423	/* Disallow 32bit project ids if projid32bit feature is not enabled. */
1424	if (fa->fsx_projid > (uint16_t)-1 &&
1425	    !xfs_sb_version_hasprojid32bit(&ip->i_mount->m_sb))
1426		return -EINVAL;
1427	return 0;
1428}
1429
1430STATIC int
1431xfs_ioctl_setattr(
1432	xfs_inode_t		*ip,
1433	struct fsxattr		*fa)
1434{
1435	struct fsxattr		old_fa;
1436	struct xfs_mount	*mp = ip->i_mount;
1437	struct xfs_trans	*tp;
1438	struct xfs_dquot	*pdqp = NULL;
1439	struct xfs_dquot	*olddquot = NULL;
1440	int			code;
1441
1442	trace_xfs_ioctl_setattr(ip);
1443
1444	code = xfs_ioctl_setattr_check_projid(ip, fa);
1445	if (code)
1446		return code;
1447
1448	/*
1449	 * If disk quotas is on, we make sure that the dquots do exist on disk,
1450	 * before we start any other transactions. Trying to do this later
1451	 * is messy. We don't care to take a readlock to look at the ids
1452	 * in inode here, because we can't hold it across the trans_reserve.
1453	 * If the IDs do change before we take the ilock, we're covered
1454	 * because the i_*dquot fields will get updated anyway.
1455	 */
1456	if (XFS_IS_QUOTA_ON(mp)) {
1457		code = xfs_qm_vop_dqalloc(ip, VFS_I(ip)->i_uid,
1458				VFS_I(ip)->i_gid, fa->fsx_projid,
1459				XFS_QMOPT_PQUOTA, NULL, NULL, &pdqp);
1460		if (code)
1461			return code;
1462	}
1463
1464	xfs_ioctl_setattr_prepare_dax(ip, fa);
1465
1466	tp = xfs_ioctl_setattr_get_trans(ip);
1467	if (IS_ERR(tp)) {
1468		code = PTR_ERR(tp);
1469		goto error_free_dquots;
1470	}
1471
1472	if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp) &&
1473	    ip->i_d.di_projid != fa->fsx_projid) {
1474		code = xfs_qm_vop_chown_reserve(tp, ip, NULL, NULL, pdqp,
1475				capable(CAP_FOWNER) ?  XFS_QMOPT_FORCE_RES : 0);
1476		if (code)	/* out of quota */
1477			goto error_trans_cancel;
1478	}
1479
1480	xfs_fill_fsxattr(ip, false, &old_fa);
1481	code = vfs_ioc_fssetxattr_check(VFS_I(ip), &old_fa, fa);
1482	if (code)
1483		goto error_trans_cancel;
1484
1485	code = xfs_ioctl_setattr_check_extsize(ip, fa);
1486	if (code)
1487		goto error_trans_cancel;
1488
1489	code = xfs_ioctl_setattr_check_cowextsize(ip, fa);
1490	if (code)
1491		goto error_trans_cancel;
1492
1493	code = xfs_ioctl_setattr_xflags(tp, ip, fa);
1494	if (code)
1495		goto error_trans_cancel;
1496
1497	/*
1498	 * Change file ownership.  Must be the owner or privileged.  CAP_FSETID
1499	 * overrides the following restrictions:
1500	 *
1501	 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1502	 * successful return from chown()
1503	 */
1504
1505	if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
1506	    !capable_wrt_inode_uidgid(VFS_I(ip), CAP_FSETID))
1507		VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
1508
1509	/* Change the ownerships and register project quota modifications */
1510	if (ip->i_d.di_projid != fa->fsx_projid) {
1511		if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp)) {
1512			olddquot = xfs_qm_vop_chown(tp, ip,
1513						&ip->i_pdquot, pdqp);
1514		}
1515		ip->i_d.di_projid = fa->fsx_projid;
1516	}
1517
1518	/*
1519	 * Only set the extent size hint if we've already determined that the
1520	 * extent size hint should be set on the inode. If no extent size flags
1521	 * are set on the inode then unconditionally clear the extent size hint.
1522	 */
1523	if (ip->i_d.di_flags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1524		ip->i_d.di_extsize = fa->fsx_extsize >> mp->m_sb.sb_blocklog;
1525	else
1526		ip->i_d.di_extsize = 0;
1527	if (xfs_sb_version_has_v3inode(&mp->m_sb) &&
1528	    (ip->i_d.di_flags2 & XFS_DIFLAG2_COWEXTSIZE))
1529		ip->i_d.di_cowextsize = fa->fsx_cowextsize >>
1530				mp->m_sb.sb_blocklog;
1531	else
1532		ip->i_d.di_cowextsize = 0;
1533
1534	code = xfs_trans_commit(tp);
1535
1536	/*
1537	 * Release any dquot(s) the inode had kept before chown.
1538	 */
1539	xfs_qm_dqrele(olddquot);
1540	xfs_qm_dqrele(pdqp);
1541
1542	return code;
1543
1544error_trans_cancel:
1545	xfs_trans_cancel(tp);
1546error_free_dquots:
1547	xfs_qm_dqrele(pdqp);
1548	return code;
1549}
1550
1551STATIC int
1552xfs_ioc_fssetxattr(
1553	xfs_inode_t		*ip,
1554	struct file		*filp,
1555	void			__user *arg)
1556{
1557	struct fsxattr		fa;
1558	int error;
1559
1560	if (copy_from_user(&fa, arg, sizeof(fa)))
1561		return -EFAULT;
1562
1563	error = mnt_want_write_file(filp);
1564	if (error)
1565		return error;
1566	error = xfs_ioctl_setattr(ip, &fa);
1567	mnt_drop_write_file(filp);
1568	return error;
1569}
1570
1571STATIC int
1572xfs_ioc_getxflags(
1573	xfs_inode_t		*ip,
1574	void			__user *arg)
1575{
1576	unsigned int		flags;
1577
1578	flags = xfs_di2lxflags(ip->i_d.di_flags, ip->i_d.di_flags2);
1579	if (copy_to_user(arg, &flags, sizeof(flags)))
1580		return -EFAULT;
1581	return 0;
1582}
1583
1584STATIC int
1585xfs_ioc_setxflags(
1586	struct xfs_inode	*ip,
1587	struct file		*filp,
1588	void			__user *arg)
1589{
1590	struct xfs_trans	*tp;
1591	struct fsxattr		fa;
1592	struct fsxattr		old_fa;
1593	unsigned int		flags;
1594	int			error;
1595
1596	if (copy_from_user(&flags, arg, sizeof(flags)))
1597		return -EFAULT;
1598
1599	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
1600		      FS_NOATIME_FL | FS_NODUMP_FL | \
1601		      FS_SYNC_FL | FS_DAX_FL))
1602		return -EOPNOTSUPP;
1603
1604	fa.fsx_xflags = xfs_merge_ioc_xflags(flags, xfs_ip2xflags(ip));
1605
1606	error = mnt_want_write_file(filp);
1607	if (error)
1608		return error;
1609
1610	xfs_ioctl_setattr_prepare_dax(ip, &fa);
1611
1612	tp = xfs_ioctl_setattr_get_trans(ip);
1613	if (IS_ERR(tp)) {
1614		error = PTR_ERR(tp);
1615		goto out_drop_write;
1616	}
1617
1618	xfs_fill_fsxattr(ip, false, &old_fa);
1619	error = vfs_ioc_fssetxattr_check(VFS_I(ip), &old_fa, &fa);
1620	if (error) {
1621		xfs_trans_cancel(tp);
1622		goto out_drop_write;
1623	}
1624
1625	error = xfs_ioctl_setattr_xflags(tp, ip, &fa);
1626	if (error) {
1627		xfs_trans_cancel(tp);
1628		goto out_drop_write;
1629	}
1630
1631	error = xfs_trans_commit(tp);
1632out_drop_write:
1633	mnt_drop_write_file(filp);
1634	return error;
1635}
1636
1637static bool
1638xfs_getbmap_format(
1639	struct kgetbmap		*p,
1640	struct getbmapx __user	*u,
1641	size_t			recsize)
1642{
1643	if (put_user(p->bmv_offset, &u->bmv_offset) ||
1644	    put_user(p->bmv_block, &u->bmv_block) ||
1645	    put_user(p->bmv_length, &u->bmv_length) ||
1646	    put_user(0, &u->bmv_count) ||
1647	    put_user(0, &u->bmv_entries))
1648		return false;
1649	if (recsize < sizeof(struct getbmapx))
1650		return true;
1651	if (put_user(0, &u->bmv_iflags) ||
1652	    put_user(p->bmv_oflags, &u->bmv_oflags) ||
1653	    put_user(0, &u->bmv_unused1) ||
1654	    put_user(0, &u->bmv_unused2))
1655		return false;
1656	return true;
1657}
1658
1659STATIC int
1660xfs_ioc_getbmap(
1661	struct file		*file,
1662	unsigned int		cmd,
1663	void			__user *arg)
1664{
1665	struct getbmapx		bmx = { 0 };
1666	struct kgetbmap		*buf;
1667	size_t			recsize;
1668	int			error, i;
1669
1670	switch (cmd) {
1671	case XFS_IOC_GETBMAPA:
1672		bmx.bmv_iflags = BMV_IF_ATTRFORK;
1673		/*FALLTHRU*/
1674	case XFS_IOC_GETBMAP:
1675		if (file->f_mode & FMODE_NOCMTIME)
1676			bmx.bmv_iflags |= BMV_IF_NO_DMAPI_READ;
1677		/* struct getbmap is a strict subset of struct getbmapx. */
1678		recsize = sizeof(struct getbmap);
1679		break;
1680	case XFS_IOC_GETBMAPX:
1681		recsize = sizeof(struct getbmapx);
1682		break;
1683	default:
1684		return -EINVAL;
1685	}
1686
1687	if (copy_from_user(&bmx, arg, recsize))
1688		return -EFAULT;
1689
1690	if (bmx.bmv_count < 2)
1691		return -EINVAL;
1692	if (bmx.bmv_count >= INT_MAX / recsize)
1693		return -ENOMEM;
1694
1695	buf = kvzalloc(bmx.bmv_count * sizeof(*buf), GFP_KERNEL);
1696	if (!buf)
1697		return -ENOMEM;
1698
1699	error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, buf);
1700	if (error)
1701		goto out_free_buf;
1702
1703	error = -EFAULT;
1704	if (copy_to_user(arg, &bmx, recsize))
1705		goto out_free_buf;
1706	arg += recsize;
1707
1708	for (i = 0; i < bmx.bmv_entries; i++) {
1709		if (!xfs_getbmap_format(buf + i, arg, recsize))
1710			goto out_free_buf;
1711		arg += recsize;
1712	}
1713
1714	error = 0;
1715out_free_buf:
1716	kmem_free(buf);
1717	return error;
1718}
1719
1720STATIC int
1721xfs_ioc_getfsmap(
1722	struct xfs_inode	*ip,
1723	struct fsmap_head	__user *arg)
1724{
1725	struct xfs_fsmap_head	xhead = {0};
1726	struct fsmap_head	head;
1727	struct fsmap		*recs;
1728	unsigned int		count;
1729	__u32			last_flags = 0;
1730	bool			done = false;
1731	int			error;
1732
1733	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
1734		return -EFAULT;
1735	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
1736	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
1737		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
1738	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
1739		       sizeof(head.fmh_keys[1].fmr_reserved)))
1740		return -EINVAL;
1741
1742	/*
1743	 * Use an internal memory buffer so that we don't have to copy fsmap
1744	 * data to userspace while holding locks.  Start by trying to allocate
1745	 * up to 128k for the buffer, but fall back to a single page if needed.
1746	 */
1747	count = min_t(unsigned int, head.fmh_count,
1748			131072 / sizeof(struct fsmap));
1749	recs = kvzalloc(count * sizeof(struct fsmap), GFP_KERNEL);
1750	if (!recs) {
1751		count = min_t(unsigned int, head.fmh_count,
1752				PAGE_SIZE / sizeof(struct fsmap));
1753		recs = kvzalloc(count * sizeof(struct fsmap), GFP_KERNEL);
1754		if (!recs)
1755			return -ENOMEM;
1756	}
1757
1758	xhead.fmh_iflags = head.fmh_iflags;
1759	xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
1760	xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
1761
1762	trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1763	trace_xfs_getfsmap_high_key(ip->i_mount, &xhead.fmh_keys[1]);
1764
1765	head.fmh_entries = 0;
1766	do {
1767		struct fsmap __user	*user_recs;
1768		struct fsmap		*last_rec;
1769
1770		user_recs = &arg->fmh_recs[head.fmh_entries];
1771		xhead.fmh_entries = 0;
1772		xhead.fmh_count = min_t(unsigned int, count,
1773					head.fmh_count - head.fmh_entries);
1774
1775		/* Run query, record how many entries we got. */
1776		error = xfs_getfsmap(ip->i_mount, &xhead, recs);
1777		switch (error) {
1778		case 0:
1779			/*
1780			 * There are no more records in the result set.  Copy
1781			 * whatever we got to userspace and break out.
1782			 */
1783			done = true;
1784			break;
1785		case -ECANCELED:
1786			/*
1787			 * The internal memory buffer is full.  Copy whatever
1788			 * records we got to userspace and go again if we have
1789			 * not yet filled the userspace buffer.
1790			 */
1791			error = 0;
1792			break;
1793		default:
1794			goto out_free;
1795		}
1796		head.fmh_entries += xhead.fmh_entries;
1797		head.fmh_oflags = xhead.fmh_oflags;
1798
1799		/*
1800		 * If the caller wanted a record count or there aren't any
1801		 * new records to return, we're done.
1802		 */
1803		if (head.fmh_count == 0 || xhead.fmh_entries == 0)
1804			break;
1805
1806		/* Copy all the records we got out to userspace. */
1807		if (copy_to_user(user_recs, recs,
1808				 xhead.fmh_entries * sizeof(struct fsmap))) {
1809			error = -EFAULT;
1810			goto out_free;
1811		}
1812
1813		/* Remember the last record flags we copied to userspace. */
1814		last_rec = &recs[xhead.fmh_entries - 1];
1815		last_flags = last_rec->fmr_flags;
1816
1817		/* Set up the low key for the next iteration. */
1818		xfs_fsmap_to_internal(&xhead.fmh_keys[0], last_rec);
1819		trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1820	} while (!done && head.fmh_entries < head.fmh_count);
1821
1822	/*
1823	 * If there are no more records in the query result set and we're not
1824	 * in counting mode, mark the last record returned with the LAST flag.
1825	 */
1826	if (done && head.fmh_count > 0 && head.fmh_entries > 0) {
1827		struct fsmap __user	*user_rec;
1828
1829		last_flags |= FMR_OF_LAST;
1830		user_rec = &arg->fmh_recs[head.fmh_entries - 1];
1831
1832		if (copy_to_user(&user_rec->fmr_flags, &last_flags,
1833					sizeof(last_flags))) {
1834			error = -EFAULT;
1835			goto out_free;
1836		}
1837	}
1838
1839	/* copy back header */
1840	if (copy_to_user(arg, &head, sizeof(struct fsmap_head))) {
1841		error = -EFAULT;
1842		goto out_free;
1843	}
1844
1845out_free:
1846	kmem_free(recs);
1847	return error;
1848}
1849
1850STATIC int
1851xfs_ioc_scrub_metadata(
1852	struct xfs_inode		*ip,
1853	void				__user *arg)
1854{
1855	struct xfs_scrub_metadata	scrub;
1856	int				error;
1857
1858	if (!capable(CAP_SYS_ADMIN))
1859		return -EPERM;
1860
1861	if (copy_from_user(&scrub, arg, sizeof(scrub)))
1862		return -EFAULT;
1863
1864	error = xfs_scrub_metadata(ip, &scrub);
1865	if (error)
1866		return error;
1867
1868	if (copy_to_user(arg, &scrub, sizeof(scrub)))
1869		return -EFAULT;
1870
1871	return 0;
1872}
1873
1874int
1875xfs_ioc_swapext(
1876	xfs_swapext_t	*sxp)
1877{
1878	xfs_inode_t     *ip, *tip;
1879	struct fd	f, tmp;
1880	int		error = 0;
1881
1882	/* Pull information for the target fd */
1883	f = fdget((int)sxp->sx_fdtarget);
1884	if (!f.file) {
1885		error = -EINVAL;
1886		goto out;
1887	}
1888
1889	if (!(f.file->f_mode & FMODE_WRITE) ||
1890	    !(f.file->f_mode & FMODE_READ) ||
1891	    (f.file->f_flags & O_APPEND)) {
1892		error = -EBADF;
1893		goto out_put_file;
1894	}
1895
1896	tmp = fdget((int)sxp->sx_fdtmp);
1897	if (!tmp.file) {
1898		error = -EINVAL;
1899		goto out_put_file;
1900	}
1901
1902	if (!(tmp.file->f_mode & FMODE_WRITE) ||
1903	    !(tmp.file->f_mode & FMODE_READ) ||
1904	    (tmp.file->f_flags & O_APPEND)) {
1905		error = -EBADF;
1906		goto out_put_tmp_file;
1907	}
1908
1909	if (IS_SWAPFILE(file_inode(f.file)) ||
1910	    IS_SWAPFILE(file_inode(tmp.file))) {
1911		error = -EINVAL;
1912		goto out_put_tmp_file;
1913	}
1914
1915	/*
1916	 * We need to ensure that the fds passed in point to XFS inodes
1917	 * before we cast and access them as XFS structures as we have no
1918	 * control over what the user passes us here.
1919	 */
1920	if (f.file->f_op != &xfs_file_operations ||
1921	    tmp.file->f_op != &xfs_file_operations) {
1922		error = -EINVAL;
1923		goto out_put_tmp_file;
1924	}
1925
1926	ip = XFS_I(file_inode(f.file));
1927	tip = XFS_I(file_inode(tmp.file));
1928
1929	if (ip->i_mount != tip->i_mount) {
1930		error = -EINVAL;
1931		goto out_put_tmp_file;
1932	}
1933
1934	if (ip->i_ino == tip->i_ino) {
1935		error = -EINVAL;
1936		goto out_put_tmp_file;
1937	}
1938
1939	if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
1940		error = -EIO;
1941		goto out_put_tmp_file;
1942	}
1943
1944	error = xfs_swap_extents(ip, tip, sxp);
1945
1946 out_put_tmp_file:
1947	fdput(tmp);
1948 out_put_file:
1949	fdput(f);
1950 out:
1951	return error;
1952}
1953
1954static int
1955xfs_ioc_getlabel(
1956	struct xfs_mount	*mp,
1957	char			__user *user_label)
1958{
1959	struct xfs_sb		*sbp = &mp->m_sb;
1960	char			label[XFSLABEL_MAX + 1];
1961
1962	/* Paranoia */
1963	BUILD_BUG_ON(sizeof(sbp->sb_fname) > FSLABEL_MAX);
1964
1965	/* 1 larger than sb_fname, so this ensures a trailing NUL char */
1966	memset(label, 0, sizeof(label));
1967	spin_lock(&mp->m_sb_lock);
1968	strncpy(label, sbp->sb_fname, XFSLABEL_MAX);
1969	spin_unlock(&mp->m_sb_lock);
1970
1971	if (copy_to_user(user_label, label, sizeof(label)))
1972		return -EFAULT;
1973	return 0;
1974}
1975
1976static int
1977xfs_ioc_setlabel(
1978	struct file		*filp,
1979	struct xfs_mount	*mp,
1980	char			__user *newlabel)
1981{
1982	struct xfs_sb		*sbp = &mp->m_sb;
1983	char			label[XFSLABEL_MAX + 1];
1984	size_t			len;
1985	int			error;
1986
1987	if (!capable(CAP_SYS_ADMIN))
1988		return -EPERM;
1989	/*
1990	 * The generic ioctl allows up to FSLABEL_MAX chars, but XFS is much
1991	 * smaller, at 12 bytes.  We copy one more to be sure we find the
1992	 * (required) NULL character to test the incoming label length.
1993	 * NB: The on disk label doesn't need to be null terminated.
1994	 */
1995	if (copy_from_user(label, newlabel, XFSLABEL_MAX + 1))
1996		return -EFAULT;
1997	len = strnlen(label, XFSLABEL_MAX + 1);
1998	if (len > sizeof(sbp->sb_fname))
1999		return -EINVAL;
2000
2001	error = mnt_want_write_file(filp);
2002	if (error)
2003		return error;
2004
2005	spin_lock(&mp->m_sb_lock);
2006	memset(sbp->sb_fname, 0, sizeof(sbp->sb_fname));
2007	memcpy(sbp->sb_fname, label, len);
2008	spin_unlock(&mp->m_sb_lock);
2009
2010	/*
2011	 * Now we do several things to satisfy userspace.
2012	 * In addition to normal logging of the primary superblock, we also
2013	 * immediately write these changes to sector zero for the primary, then
2014	 * update all backup supers (as xfs_db does for a label change), then
2015	 * invalidate the block device page cache.  This is so that any prior
2016	 * buffered reads from userspace (i.e. from blkid) are invalidated,
2017	 * and userspace will see the newly-written label.
2018	 */
2019	error = xfs_sync_sb_buf(mp);
2020	if (error)
2021		goto out;
2022	/*
2023	 * growfs also updates backup supers so lock against that.
2024	 */
2025	mutex_lock(&mp->m_growlock);
2026	error = xfs_update_secondary_sbs(mp);
2027	mutex_unlock(&mp->m_growlock);
2028
2029	invalidate_bdev(mp->m_ddev_targp->bt_bdev);
2030
2031out:
2032	mnt_drop_write_file(filp);
2033	return error;
2034}
2035
2036static inline int
2037xfs_fs_eofblocks_from_user(
2038	struct xfs_fs_eofblocks		*src,
2039	struct xfs_eofblocks		*dst)
2040{
2041	if (src->eof_version != XFS_EOFBLOCKS_VERSION)
2042		return -EINVAL;
2043
2044	if (src->eof_flags & ~XFS_EOF_FLAGS_VALID)
2045		return -EINVAL;
2046
2047	if (memchr_inv(&src->pad32, 0, sizeof(src->pad32)) ||
2048	    memchr_inv(src->pad64, 0, sizeof(src->pad64)))
2049		return -EINVAL;
2050
2051	dst->eof_flags = src->eof_flags;
2052	dst->eof_prid = src->eof_prid;
2053	dst->eof_min_file_size = src->eof_min_file_size;
2054
2055	dst->eof_uid = INVALID_UID;
2056	if (src->eof_flags & XFS_EOF_FLAGS_UID) {
2057		dst->eof_uid = make_kuid(current_user_ns(), src->eof_uid);
2058		if (!uid_valid(dst->eof_uid))
2059			return -EINVAL;
2060	}
2061
2062	dst->eof_gid = INVALID_GID;
2063	if (src->eof_flags & XFS_EOF_FLAGS_GID) {
2064		dst->eof_gid = make_kgid(current_user_ns(), src->eof_gid);
2065		if (!gid_valid(dst->eof_gid))
2066			return -EINVAL;
2067	}
2068	return 0;
2069}
2070
2071/*
2072 * Note: some of the ioctl's return positive numbers as a
2073 * byte count indicating success, such as readlink_by_handle.
2074 * So we don't "sign flip" like most other routines.  This means
2075 * true errors need to be returned as a negative value.
2076 */
2077long
2078xfs_file_ioctl(
2079	struct file		*filp,
2080	unsigned int		cmd,
2081	unsigned long		p)
2082{
2083	struct inode		*inode = file_inode(filp);
2084	struct xfs_inode	*ip = XFS_I(inode);
2085	struct xfs_mount	*mp = ip->i_mount;
2086	void			__user *arg = (void __user *)p;
2087	int			error;
2088
2089	trace_xfs_file_ioctl(ip);
2090
2091	switch (cmd) {
2092	case FITRIM:
2093		return xfs_ioc_trim(mp, arg);
2094	case FS_IOC_GETFSLABEL:
2095		return xfs_ioc_getlabel(mp, arg);
2096	case FS_IOC_SETFSLABEL:
2097		return xfs_ioc_setlabel(filp, mp, arg);
2098	case XFS_IOC_ALLOCSP:
2099	case XFS_IOC_FREESP:
2100	case XFS_IOC_ALLOCSP64:
2101	case XFS_IOC_FREESP64: {
2102		xfs_flock64_t		bf;
2103
2104		if (copy_from_user(&bf, arg, sizeof(bf)))
2105			return -EFAULT;
2106		return xfs_ioc_space(filp, &bf);
2107	}
2108	case XFS_IOC_DIOINFO: {
2109		struct xfs_buftarg	*target = xfs_inode_buftarg(ip);
2110		struct dioattr		da;
2111
2112		da.d_mem =  da.d_miniosz = target->bt_logical_sectorsize;
2113		da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
2114
2115		if (copy_to_user(arg, &da, sizeof(da)))
2116			return -EFAULT;
2117		return 0;
2118	}
2119
2120	case XFS_IOC_FSBULKSTAT_SINGLE:
2121	case XFS_IOC_FSBULKSTAT:
2122	case XFS_IOC_FSINUMBERS:
2123		return xfs_ioc_fsbulkstat(mp, cmd, arg);
2124
2125	case XFS_IOC_BULKSTAT:
2126		return xfs_ioc_bulkstat(mp, cmd, arg);
2127	case XFS_IOC_INUMBERS:
2128		return xfs_ioc_inumbers(mp, cmd, arg);
2129
2130	case XFS_IOC_FSGEOMETRY_V1:
2131		return xfs_ioc_fsgeometry(mp, arg, 3);
2132	case XFS_IOC_FSGEOMETRY_V4:
2133		return xfs_ioc_fsgeometry(mp, arg, 4);
2134	case XFS_IOC_FSGEOMETRY:
2135		return xfs_ioc_fsgeometry(mp, arg, 5);
2136
2137	case XFS_IOC_AG_GEOMETRY:
2138		return xfs_ioc_ag_geometry(mp, arg);
2139
2140	case XFS_IOC_GETVERSION:
2141		return put_user(inode->i_generation, (int __user *)arg);
2142
2143	case XFS_IOC_FSGETXATTR:
2144		return xfs_ioc_fsgetxattr(ip, 0, arg);
2145	case XFS_IOC_FSGETXATTRA:
2146		return xfs_ioc_fsgetxattr(ip, 1, arg);
2147	case XFS_IOC_FSSETXATTR:
2148		return xfs_ioc_fssetxattr(ip, filp, arg);
2149	case XFS_IOC_GETXFLAGS:
2150		return xfs_ioc_getxflags(ip, arg);
2151	case XFS_IOC_SETXFLAGS:
2152		return xfs_ioc_setxflags(ip, filp, arg);
2153
2154	case XFS_IOC_GETBMAP:
2155	case XFS_IOC_GETBMAPA:
2156	case XFS_IOC_GETBMAPX:
2157		return xfs_ioc_getbmap(filp, cmd, arg);
2158
2159	case FS_IOC_GETFSMAP:
2160		return xfs_ioc_getfsmap(ip, arg);
2161
2162	case XFS_IOC_SCRUB_METADATA:
2163		return xfs_ioc_scrub_metadata(ip, arg);
2164
2165	case XFS_IOC_FD_TO_HANDLE:
2166	case XFS_IOC_PATH_TO_HANDLE:
2167	case XFS_IOC_PATH_TO_FSHANDLE: {
2168		xfs_fsop_handlereq_t	hreq;
2169
2170		if (copy_from_user(&hreq, arg, sizeof(hreq)))
2171			return -EFAULT;
2172		return xfs_find_handle(cmd, &hreq);
2173	}
2174	case XFS_IOC_OPEN_BY_HANDLE: {
2175		xfs_fsop_handlereq_t	hreq;
2176
2177		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
2178			return -EFAULT;
2179		return xfs_open_by_handle(filp, &hreq);
2180	}
2181
2182	case XFS_IOC_READLINK_BY_HANDLE: {
2183		xfs_fsop_handlereq_t	hreq;
2184
2185		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
2186			return -EFAULT;
2187		return xfs_readlink_by_handle(filp, &hreq);
2188	}
2189	case XFS_IOC_ATTRLIST_BY_HANDLE:
2190		return xfs_attrlist_by_handle(filp, arg);
2191
2192	case XFS_IOC_ATTRMULTI_BY_HANDLE:
2193		return xfs_attrmulti_by_handle(filp, arg);
2194
2195	case XFS_IOC_SWAPEXT: {
2196		struct xfs_swapext	sxp;
2197
2198		if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
2199			return -EFAULT;
2200		error = mnt_want_write_file(filp);
2201		if (error)
2202			return error;
2203		error = xfs_ioc_swapext(&sxp);
2204		mnt_drop_write_file(filp);
2205		return error;
2206	}
2207
2208	case XFS_IOC_FSCOUNTS: {
2209		xfs_fsop_counts_t out;
2210
2211		xfs_fs_counts(mp, &out);
2212
2213		if (copy_to_user(arg, &out, sizeof(out)))
2214			return -EFAULT;
2215		return 0;
2216	}
2217
2218	case XFS_IOC_SET_RESBLKS: {
2219		xfs_fsop_resblks_t inout;
2220		uint64_t	   in;
2221
2222		if (!capable(CAP_SYS_ADMIN))
2223			return -EPERM;
2224
2225		if (mp->m_flags & XFS_MOUNT_RDONLY)
2226			return -EROFS;
2227
2228		if (copy_from_user(&inout, arg, sizeof(inout)))
2229			return -EFAULT;
2230
2231		error = mnt_want_write_file(filp);
2232		if (error)
2233			return error;
2234
2235		/* input parameter is passed in resblks field of structure */
2236		in = inout.resblks;
2237		error = xfs_reserve_blocks(mp, &in, &inout);
2238		mnt_drop_write_file(filp);
2239		if (error)
2240			return error;
2241
2242		if (copy_to_user(arg, &inout, sizeof(inout)))
2243			return -EFAULT;
2244		return 0;
2245	}
2246
2247	case XFS_IOC_GET_RESBLKS: {
2248		xfs_fsop_resblks_t out;
2249
2250		if (!capable(CAP_SYS_ADMIN))
2251			return -EPERM;
2252
2253		error = xfs_reserve_blocks(mp, NULL, &out);
2254		if (error)
2255			return error;
2256
2257		if (copy_to_user(arg, &out, sizeof(out)))
2258			return -EFAULT;
2259
2260		return 0;
2261	}
2262
2263	case XFS_IOC_FSGROWFSDATA: {
2264		xfs_growfs_data_t in;
2265
2266		if (copy_from_user(&in, arg, sizeof(in)))
2267			return -EFAULT;
2268
2269		error = mnt_want_write_file(filp);
2270		if (error)
2271			return error;
2272		error = xfs_growfs_data(mp, &in);
2273		mnt_drop_write_file(filp);
2274		return error;
2275	}
2276
2277	case XFS_IOC_FSGROWFSLOG: {
2278		xfs_growfs_log_t in;
2279
2280		if (copy_from_user(&in, arg, sizeof(in)))
2281			return -EFAULT;
2282
2283		error = mnt_want_write_file(filp);
2284		if (error)
2285			return error;
2286		error = xfs_growfs_log(mp, &in);
2287		mnt_drop_write_file(filp);
2288		return error;
2289	}
2290
2291	case XFS_IOC_FSGROWFSRT: {
2292		xfs_growfs_rt_t in;
2293
2294		if (copy_from_user(&in, arg, sizeof(in)))
2295			return -EFAULT;
2296
2297		error = mnt_want_write_file(filp);
2298		if (error)
2299			return error;
2300		error = xfs_growfs_rt(mp, &in);
2301		mnt_drop_write_file(filp);
2302		return error;
2303	}
2304
2305	case XFS_IOC_GOINGDOWN: {
2306		uint32_t in;
2307
2308		if (!capable(CAP_SYS_ADMIN))
2309			return -EPERM;
2310
2311		if (get_user(in, (uint32_t __user *)arg))
2312			return -EFAULT;
2313
2314		return xfs_fs_goingdown(mp, in);
2315	}
2316
2317	case XFS_IOC_ERROR_INJECTION: {
2318		xfs_error_injection_t in;
2319
2320		if (!capable(CAP_SYS_ADMIN))
2321			return -EPERM;
2322
2323		if (copy_from_user(&in, arg, sizeof(in)))
2324			return -EFAULT;
2325
2326		return xfs_errortag_add(mp, in.errtag);
2327	}
2328
2329	case XFS_IOC_ERROR_CLEARALL:
2330		if (!capable(CAP_SYS_ADMIN))
2331			return -EPERM;
2332
2333		return xfs_errortag_clearall(mp);
2334
2335	case XFS_IOC_FREE_EOFBLOCKS: {
2336		struct xfs_fs_eofblocks eofb;
2337		struct xfs_eofblocks keofb;
2338
2339		if (!capable(CAP_SYS_ADMIN))
2340			return -EPERM;
2341
2342		if (mp->m_flags & XFS_MOUNT_RDONLY)
2343			return -EROFS;
2344
2345		if (copy_from_user(&eofb, arg, sizeof(eofb)))
2346			return -EFAULT;
2347
2348		error = xfs_fs_eofblocks_from_user(&eofb, &keofb);
2349		if (error)
2350			return error;
2351
2352		sb_start_write(mp->m_super);
2353		error = xfs_icache_free_eofblocks(mp, &keofb);
2354		sb_end_write(mp->m_super);
2355		return error;
2356	}
2357
2358	default:
2359		return -ENOTTY;
2360	}
2361}
2362